diff options
| author | Arnold D. Robbins <arnold@skeeve.com> | 2016-12-22 17:21:16 +0200 |
|---|---|---|
| committer | Arnold D. Robbins <arnold@skeeve.com> | 2016-12-22 17:21:16 +0200 |
| commit | 725d2f781b48ba83f4c8784a0fbe62fe2a6c107a (patch) | |
| tree | 396c052bd95aa1190d90bc81f3b983dd60550e36 /support | |
| parent | cd35e365b6f8d356645093bab1c67c1867a63aef (diff) | |
| download | gawk-725d2f781b48ba83f4c8784a0fbe62fe2a6c107a.tar.gz | |
Add small regex fix. Add support directory.
Diffstat (limited to 'support')
| -rw-r--r-- | support/dfa.c | 4160 | ||||
| -rw-r--r-- | support/dfa.h | 132 | ||||
| -rw-r--r-- | support/getopt.c | 1293 | ||||
| -rw-r--r-- | support/getopt.h | 206 | ||||
| -rw-r--r-- | support/getopt1.c | 195 | ||||
| -rw-r--r-- | support/getopt_int.h | 129 | ||||
| -rw-r--r-- | support/intprops.h | 464 | ||||
| -rw-r--r-- | support/localeinfo.c | 113 | ||||
| -rw-r--r-- | support/localeinfo.h | 54 | ||||
| -rw-r--r-- | support/random.c | 542 | ||||
| -rw-r--r-- | support/random.h | 43 | ||||
| -rw-r--r-- | support/regcomp.c | 3939 | ||||
| -rw-r--r-- | support/regex.c | 85 | ||||
| -rw-r--r-- | support/regex.h | 591 | ||||
| -rw-r--r-- | support/regex_internal.c | 1761 | ||||
| -rw-r--r-- | support/regex_internal.h | 831 | ||||
| -rw-r--r-- | support/regexec.c | 4370 | ||||
| -rw-r--r-- | support/verify.h | 279 | ||||
| -rw-r--r-- | support/xalloc.h | 362 |
19 files changed, 19549 insertions, 0 deletions
diff --git a/support/dfa.c b/support/dfa.c new file mode 100644 index 00000000..8f34c4c1 --- /dev/null +++ b/support/dfa.c @@ -0,0 +1,4160 @@ +/* dfa.c - deterministic extended regexp routines for GNU + Copyright (C) 1988, 1998, 2000, 2002, 2004-2005, 2007-2016 Free Software + Foundation, Inc. + + This program is free software; you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation; either version 3, or (at your option) + any later version. + + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program; if not, write to the Free Software + Foundation, Inc., + 51 Franklin Street - Fifth Floor, Boston, MA 02110-1301, USA */ + +/* Written June, 1988 by Mike Haertel + Modified July, 1988 by Arthur David Olson to assist BMG speedups */ + +#include <config.h> + +#include <assert.h> +#include <ctype.h> +#include <stdint.h> +#include <stdio.h> + +#ifndef VMS +#include <sys/types.h> +#else +#include <stddef.h> +#endif +#include <stdlib.h> +#include <limits.h> +#include <string.h> +#if HAVE_SETLOCALE +#include <locale.h> +#endif + +/* Gawk doesn't use Gnulib, so don't assume that setlocale is present. */ +#ifndef LC_ALL +# define setlocale(category, locale) NULL +#endif + +#define STREQ(a, b) (strcmp (a, b) == 0) + +/* ISASCIIDIGIT differs from isdigit, as follows: + - Its arg may be any int or unsigned int; it need not be an unsigned char. + - It's guaranteed to evaluate its argument exactly once. + - It's typically faster. + Posix 1003.2-1992 section 2.5.2.1 page 50 lines 1556-1558 says that + only '0' through '9' are digits. Prefer ISASCIIDIGIT to isdigit unless + it's important to use the locale's definition of "digit" even when the + host does not conform to Posix. */ +#define ISASCIIDIGIT(c) ((unsigned) (c) - '0' <= 9) + +#include "gettext.h" +#define _(str) gettext (str) + +#include <wchar.h> + +#include "intprops.h" +#include "xalloc.h" +#include "localeinfo.h" + +#ifndef MIN +# define MIN(a,b) ((a) < (b) ? (a) : (b)) +#endif + +#if defined(__DJGPP__) +#include "mbsupport.h" +#endif + +#include "dfa.h" + +#ifdef GAWK +static int +is_blank (int c) +{ + return (c == ' ' || c == '\t'); +} +#endif /* GAWK */ + +/* HPUX defines these as macros in sys/param.h. */ +#ifdef setbit +# undef setbit +#endif +#ifdef clrbit +# undef clrbit +#endif + +/* First integer value that is greater than any character code. */ +enum { NOTCHAR = 1 << CHAR_BIT }; + +/* This represents part of a character class. It must be unsigned and + at least CHARCLASS_WORD_BITS wide. Any excess bits are zero. */ +typedef unsigned long int charclass_word; + +/* CHARCLASS_WORD_BITS is the number of bits used in a charclass word. + CHARCLASS_PAIR (LO, HI) is part of a charclass initializer, and + represents 64 bits' worth of a charclass, where LO and HI are the + low and high-order 32 bits of the 64-bit quantity. */ +#if ULONG_MAX >> 31 >> 31 < 3 +enum { CHARCLASS_WORD_BITS = 32 }; +# define CHARCLASS_PAIR(lo, hi) lo, hi +#else +enum { CHARCLASS_WORD_BITS = 64 }; +# define CHARCLASS_PAIR(lo, hi) (((charclass_word) (hi) << 32) + (lo)) +#endif + +/* An initializer for a charclass whose 32-bit words are A through H. */ +#define CHARCLASS_INIT(a, b, c, d, e, f, g, h) \ + { \ + CHARCLASS_PAIR (a, b), CHARCLASS_PAIR (c, d), \ + CHARCLASS_PAIR (e, f), CHARCLASS_PAIR (g, h) \ + } + +/* The maximum useful value of a charclass_word; all used bits are 1. */ +static charclass_word const CHARCLASS_WORD_MASK + = ((charclass_word) 1 << (CHARCLASS_WORD_BITS - 1) << 1) - 1; + +/* Number of words required to hold a bit for every character. */ +enum +{ + CHARCLASS_WORDS = (NOTCHAR + CHARCLASS_WORD_BITS - 1) / CHARCLASS_WORD_BITS +}; + +/* Sets of unsigned characters are stored as bit vectors in arrays of ints. */ +typedef charclass_word charclass[CHARCLASS_WORDS]; + +/* Convert a possibly-signed character to an unsigned character. This is + a bit safer than casting to unsigned char, since it catches some type + errors that the cast doesn't. */ +static unsigned char +to_uchar (char ch) +{ + return ch; +} + +/* Contexts tell us whether a character is a newline or a word constituent. + Word-constituent characters are those that satisfy iswalnum, plus '_'. + Each character has a single CTX_* value; bitmasks of CTX_* values denote + a particular character class. + + A state also stores a context value, which is a bitmask of CTX_* values. + A state's context represents a set of characters that the state's + predecessors must match. For example, a state whose context does not + include CTX_LETTER will never have transitions where the previous + character is a word constituent. A state whose context is CTX_ANY + might have transitions from any character. */ + +#define CTX_NONE 1 +#define CTX_LETTER 2 +#define CTX_NEWLINE 4 +#define CTX_ANY 7 + +/* Sometimes characters can only be matched depending on the surrounding + context. Such context decisions depend on what the previous character + was, and the value of the current (lookahead) character. Context + dependent constraints are encoded as 12-bit integers. Each bit that + is set indicates that the constraint succeeds in the corresponding + context. + + bit 8-11 - valid contexts when next character is CTX_NEWLINE + bit 4-7 - valid contexts when next character is CTX_LETTER + bit 0-3 - valid contexts when next character is CTX_NONE + + The macro SUCCEEDS_IN_CONTEXT determines whether a given constraint + succeeds in a particular context. Prev is a bitmask of possible + context values for the previous character, curr is the (single-bit) + context value for the lookahead character. */ +#define NEWLINE_CONSTRAINT(constraint) (((constraint) >> 8) & 0xf) +#define LETTER_CONSTRAINT(constraint) (((constraint) >> 4) & 0xf) +#define OTHER_CONSTRAINT(constraint) ((constraint) & 0xf) + +#define SUCCEEDS_IN_CONTEXT(constraint, prev, curr) \ + ((((curr) & CTX_NONE ? OTHER_CONSTRAINT (constraint) : 0) \ + | ((curr) & CTX_LETTER ? LETTER_CONSTRAINT (constraint) : 0) \ + | ((curr) & CTX_NEWLINE ? NEWLINE_CONSTRAINT (constraint) : 0)) \ + & (prev)) + +/* The following macros describe what a constraint depends on. */ +#define PREV_NEWLINE_CONSTRAINT(constraint) (((constraint) >> 2) & 0x111) +#define PREV_LETTER_CONSTRAINT(constraint) (((constraint) >> 1) & 0x111) +#define PREV_OTHER_CONSTRAINT(constraint) ((constraint) & 0x111) + +#define PREV_NEWLINE_DEPENDENT(constraint) \ + (PREV_NEWLINE_CONSTRAINT (constraint) != PREV_OTHER_CONSTRAINT (constraint)) +#define PREV_LETTER_DEPENDENT(constraint) \ + (PREV_LETTER_CONSTRAINT (constraint) != PREV_OTHER_CONSTRAINT (constraint)) + +/* Tokens that match the empty string subject to some constraint actually + work by applying that constraint to determine what may follow them, + taking into account what has gone before. The following values are + the constraints corresponding to the special tokens previously defined. */ +#define NO_CONSTRAINT 0x777 +#define BEGLINE_CONSTRAINT 0x444 +#define ENDLINE_CONSTRAINT 0x700 +#define BEGWORD_CONSTRAINT 0x050 +#define ENDWORD_CONSTRAINT 0x202 +#define LIMWORD_CONSTRAINT 0x252 +#define NOTLIMWORD_CONSTRAINT 0x525 + +/* The regexp is parsed into an array of tokens in postfix form. Some tokens + are operators and others are terminal symbols. Most (but not all) of these + codes are returned by the lexical analyzer. */ + +typedef ptrdiff_t token; +#define TOKEN_MAX PTRDIFF_MAX + +/* States are indexed by state_num values. These are normally + nonnegative but -1 is used as a special value. */ +typedef ptrdiff_t state_num; + +/* Predefined token values. */ +enum +{ + END = -1, /* END is a terminal symbol that matches the + end of input; any value of END or less in + the parse tree is such a symbol. Accepting + states of the DFA are those that would have + a transition on END. */ + + /* Ordinary character values are terminal symbols that match themselves. */ + + EMPTY = NOTCHAR, /* EMPTY is a terminal symbol that matches + the empty string. */ + + BACKREF, /* BACKREF is generated by \<digit> + or by any other construct that + is not completely handled. If the scanner + detects a transition on backref, it returns + a kind of "semi-success" indicating that + the match will have to be verified with + a backtracking matcher. */ + + BEGLINE, /* BEGLINE is a terminal symbol that matches + the empty string at the beginning of a + line. */ + + ENDLINE, /* ENDLINE is a terminal symbol that matches + the empty string at the end of a line. */ + + BEGWORD, /* BEGWORD is a terminal symbol that matches + the empty string at the beginning of a + word. */ + + ENDWORD, /* ENDWORD is a terminal symbol that matches + the empty string at the end of a word. */ + + LIMWORD, /* LIMWORD is a terminal symbol that matches + the empty string at the beginning or the + end of a word. */ + + NOTLIMWORD, /* NOTLIMWORD is a terminal symbol that + matches the empty string not at + the beginning or end of a word. */ + + QMARK, /* QMARK is an operator of one argument that + matches zero or one occurrences of its + argument. */ + + STAR, /* STAR is an operator of one argument that + matches the Kleene closure (zero or more + occurrences) of its argument. */ + + PLUS, /* PLUS is an operator of one argument that + matches the positive closure (one or more + occurrences) of its argument. */ + + REPMN, /* REPMN is a lexical token corresponding + to the {m,n} construct. REPMN never + appears in the compiled token vector. */ + + CAT, /* CAT is an operator of two arguments that + matches the concatenation of its + arguments. CAT is never returned by the + lexical analyzer. */ + + OR, /* OR is an operator of two arguments that + matches either of its arguments. */ + + LPAREN, /* LPAREN never appears in the parse tree, + it is only a lexeme. */ + + RPAREN, /* RPAREN never appears in the parse tree. */ + + ANYCHAR, /* ANYCHAR is a terminal symbol that matches + a valid multibyte (or single byte) character. + It is used only if MB_CUR_MAX > 1. */ + + MBCSET, /* MBCSET is similar to CSET, but for + multibyte characters. */ + + WCHAR, /* Only returned by lex. wctok contains + the wide character representation. */ + + CSET /* CSET and (and any value greater) is a + terminal symbol that matches any of a + class of characters. */ +}; + + +/* States of the recognizer correspond to sets of positions in the parse + tree, together with the constraints under which they may be matched. + So a position is encoded as an index into the parse tree together with + a constraint. */ +typedef struct +{ + size_t index; /* Index into the parse array. */ + unsigned int constraint; /* Constraint for matching this position. */ +} position; + +/* Sets of positions are stored as arrays. */ +typedef struct +{ + position *elems; /* Elements of this position set. */ + ptrdiff_t nelem; /* Number of elements in this set. */ + ptrdiff_t alloc; /* Number of elements allocated in ELEMS. */ +} position_set; + +/* Sets of leaves are also stored as arrays. */ +typedef struct +{ + size_t *elems; /* Elements of this position set. */ + size_t nelem; /* Number of elements in this set. */ +} leaf_set; + +/* A state of the dfa consists of a set of positions, some flags, + and the token value of the lowest-numbered position of the state that + contains an END token. */ +typedef struct +{ + size_t hash; /* Hash of the positions of this state. */ + position_set elems; /* Positions this state could match. */ + unsigned char context; /* Context from previous state. */ + unsigned short constraint; /* Constraint for this state to accept. */ + token first_end; /* Token value of the first END in elems. */ + position_set mbps; /* Positions which can match multibyte + characters or the follows, e.g., period. + Used only if MB_CUR_MAX > 1. */ + state_num mb_trindex; /* Index of this state in MB_TRANS, or + negative if the state does not have + ANYCHAR. */ +} dfa_state; + +/* Maximum for any transition table count. This should be at least 3, + for the initial state setup. */ +enum { MAX_TRCOUNT = 1024 }; + +/* A bracket operator. + e.g., [a-c], [[:alpha:]], etc. */ +struct mb_char_classes +{ + ptrdiff_t cset; + bool invert; + wchar_t *chars; /* Normal characters. */ + ptrdiff_t nchars; +}; + +struct regex_syntax +{ + /* Syntax bits controlling the behavior of the lexical analyzer. */ + reg_syntax_t syntax_bits; + bool syntax_bits_set; + + /* Flag for case-folding letters into sets. */ + bool case_fold; + + /* True if ^ and $ match only the start and end of data, and do not match + end-of-line within data. */ + bool anchor; + + /* End-of-line byte in data. */ + unsigned char eolbyte; + + /* Cache of char-context values. */ + int sbit[NOTCHAR]; + + /* If never_trail[B], the byte B cannot be a non-initial byte in a + multibyte character. */ + bool never_trail[NOTCHAR]; + + /* Set of characters considered letters. */ + charclass letters; + + /* Set of characters that are newline. */ + charclass newline; +}; + +/* Lexical analyzer. All the dross that deals with the obnoxious + GNU Regex syntax bits is located here. The poor, suffering + reader is referred to the GNU Regex documentation for the + meaning of the @#%!@#%^!@ syntax bits. */ +struct lexer_state +{ + char const *ptr; /* Pointer to next input character. */ + size_t left; /* Number of characters remaining. */ + token lasttok; /* Previous token returned; initially END. */ + size_t parens; /* Count of outstanding left parens. */ + int minrep, maxrep; /* Repeat counts for {m,n}. */ + + /* Wide character representation of the current multibyte character, + or WEOF if there was an encoding error. Used only if + MB_CUR_MAX > 1. */ + wint_t wctok; + + /* Length of the multibyte representation of wctok. */ + int cur_mb_len; + + /* We're separated from beginning or (, | only by zero-width characters. */ + bool laststart; +}; + +/* Recursive descent parser for regular expressions. */ + +struct parser_state +{ + token tok; /* Lookahead token. */ + size_t depth; /* Current depth of a hypothetical stack + holding deferred productions. This is + used to determine the depth that will be + required of the real stack later on in + dfaanalyze. */ +}; + +/* A compiled regular expression. */ +struct dfa +{ + /* Syntax configuration */ + struct regex_syntax syntax; + + /* Fields filled by the scanner. */ + charclass *charclasses; /* Array of character sets for CSET tokens. */ + ptrdiff_t cindex; /* Index for adding new charclasses. */ + ptrdiff_t calloc; /* Number of charclasses allocated. */ + size_t canychar; /* Index of anychar class, or (size_t) -1. */ + + /* Scanner state */ + struct lexer_state lex; + + /* Parser state */ + struct parser_state parse; + + /* Fields filled by the parser. */ + token *tokens; /* Postfix parse array. */ + size_t tindex; /* Index for adding new tokens. */ + size_t talloc; /* Number of tokens currently allocated. */ + size_t depth; /* Depth required of an evaluation stack + used for depth-first traversal of the + parse tree. */ + size_t nleaves; /* Number of leaves on the parse tree. */ + size_t nregexps; /* Count of parallel regexps being built + with dfaparse. */ + bool fast; /* The DFA is fast. */ + token utf8_anychar_classes[5]; /* To lower ANYCHAR in UTF-8 locales. */ + mbstate_t mbs; /* Multibyte conversion state. */ + + /* The following are valid only if MB_CUR_MAX > 1. */ + + /* The value of multibyte_prop[i] is defined by following rule. + if tokens[i] < NOTCHAR + bit 0 : tokens[i] is the first byte of a character, including + single-byte characters. + bit 1 : tokens[i] is the last byte of a character, including + single-byte characters. + + if tokens[i] = MBCSET + ("the index of mbcsets corresponding to this operator" << 2) + 3 + + e.g. + tokens + = 'single_byte_a', 'multi_byte_A', single_byte_b' + = 'sb_a', 'mb_A(1st byte)', 'mb_A(2nd byte)', 'mb_A(3rd byte)', 'sb_b' + multibyte_prop + = 3 , 1 , 0 , 2 , 3 + */ + int *multibyte_prop; + + /* Array of the bracket expression in the DFA. */ + struct mb_char_classes *mbcsets; + ptrdiff_t nmbcsets; + ptrdiff_t mbcsets_alloc; + + /* Fields filled by the superset. */ + struct dfa *superset; /* Hint of the dfa. */ + + /* Fields filled by the state builder. */ + dfa_state *states; /* States of the dfa. */ + state_num sindex; /* Index for adding new states. */ + ptrdiff_t salloc; /* Number of states currently allocated. */ + + /* Fields filled by the parse tree->NFA conversion. */ + position_set *follows; /* Array of follow sets, indexed by position + index. The follow of a position is the set + of positions containing characters that + could conceivably follow a character + matching the given position in a string + matching the regexp. Allocated to the + maximum possible position index. */ + bool searchflag; /* We are supposed to build a searching + as opposed to an exact matcher. A searching + matcher finds the first and shortest string + matching a regexp anywhere in the buffer, + whereas an exact matcher finds the longest + string matching, but anchored to the + beginning of the buffer. */ + + /* Fields filled by dfaexec. */ + state_num tralloc; /* Number of transition tables that have + slots so far, not counting trans[-1] and + trans[-2]. */ + int trcount; /* Number of transition tables that have + been built, other than for initial + states. */ + int min_trcount; /* Number of initial states. Equivalently, + the minimum state number for which trcount + counts transitions. */ + state_num **trans; /* Transition tables for states that can + never accept. If the transitions for a + state have not yet been computed, or the + state could possibly accept, its entry in + this table is NULL. This points to two + past the start of the allocated array, + and trans[-1] and trans[-2] are always + NULL. */ + state_num **fails; /* Transition tables after failing to accept + on a state that potentially could do so. + If trans[i] is non-null, fails[i] must + be null. */ + int *success; /* Table of acceptance conditions used in + dfaexec and computed in build_state. */ + state_num *newlines; /* Transitions on newlines. The entry for a + newline in any transition table is always + -1 so we can count lines without wasting + too many cycles. The transition for a + newline is stored separately and handled + as a special case. Newline is also used + as a sentinel at the end of the buffer. */ + state_num initstate_notbol; /* Initial state for CTX_LETTER and CTX_NONE + context in multibyte locales, in which we + do not distinguish between their contexts, + as not supported word. */ + position_set mb_follows; /* Follow set added by ANYCHAR on demand. */ + state_num **mb_trans; /* Transition tables for states with + ANYCHAR. */ + state_num mb_trcount; /* Number of transition tables for states with + ANYCHAR that have actually been built. */ + + /* Information derived from the locale. This is at the end so that + a quick memset need not clear it specially. */ + + /* dfaexec implementation. */ + char *(*dfaexec) (struct dfa *, char const *, char *, + bool, size_t *, bool *); + + /* The locale is simple, like the C locale. These locales can be + processed more efficiently, e.g., the relationship between lower- + and upper-case letters is 1-1. */ + bool simple_locale; + + /* Other cached information derived from the locale. */ + struct localeinfo localeinfo; +}; + +/* Some macros for user access to dfa internals. */ + +/* S could possibly be an accepting state of R. */ +#define ACCEPTING(s, r) ((r).states[s].constraint) + +/* STATE accepts in the specified context. */ +#define ACCEPTS_IN_CONTEXT(prev, curr, state, dfa) \ + SUCCEEDS_IN_CONTEXT ((dfa).states[state].constraint, prev, curr) + +static void regexp (struct dfa *dfa); + +/* Store into *PWC the result of converting the leading bytes of the + multibyte buffer S of length N bytes, using D->localeinfo.sbctowc + and updating the conversion state in *D. On conversion error, + convert just a single byte, to WEOF. Return the number of bytes + converted. + + This differs from mbrtowc (PWC, S, N, &D->mbs) as follows: + + * PWC points to wint_t, not to wchar_t. + * The last arg is a dfa *D instead of merely a multibyte conversion + state D->mbs. + * N must be at least 1. + * S[N - 1] must be a sentinel byte. + * Shift encodings are not supported. + * The return value is always in the range 1..N. + * D->mbs is always valid afterwards. + * *PWC is always set to something. */ +static size_t +mbs_to_wchar (wint_t *pwc, char const *s, size_t n, struct dfa *d) +{ + unsigned char uc = s[0]; + wint_t wc = d->localeinfo.sbctowc[uc]; + + if (wc == WEOF) + { + wchar_t wch; + size_t nbytes = mbrtowc (&wch, s, n, &d->mbs); + if (0 < nbytes && nbytes < (size_t) -2) + { + *pwc = wch; + return nbytes; + } + memset (&d->mbs, 0, sizeof d->mbs); + } + + *pwc = wc; + return 1; +} + +#ifdef DEBUG + +static void +prtok (token t) +{ + char const *s; + + if (t < 0) + fprintf (stderr, "END"); + else if (t < NOTCHAR) + { + unsigned int ch = t; + fprintf (stderr, "0x%02x", ch); + } + else + { + switch (t) + { + case EMPTY: + s = "EMPTY"; + break; + case BACKREF: + s = "BACKREF"; + break; + case BEGLINE: + s = "BEGLINE"; + break; + case ENDLINE: + s = "ENDLINE"; + break; + case BEGWORD: + s = "BEGWORD"; + break; + case ENDWORD: + s = "ENDWORD"; + break; + case LIMWORD: + s = "LIMWORD"; + break; + case NOTLIMWORD: + s = "NOTLIMWORD"; + break; + case QMARK: + s = "QMARK"; + break; + case STAR: + s = "STAR"; + break; + case PLUS: + s = "PLUS"; + break; + case CAT: + s = "CAT"; + break; + case OR: + s = "OR"; + break; + case LPAREN: + s = "LPAREN"; + break; + case RPAREN: + s = "RPAREN"; + break; + case ANYCHAR: + s = "ANYCHAR"; + break; + case MBCSET: + s = "MBCSET"; + break; + default: + s = "CSET"; + break; + } + fprintf (stderr, "%s", s); + } +} +#endif /* DEBUG */ + +/* Stuff pertaining to charclasses. */ + +static bool +tstbit (unsigned int b, charclass const c) +{ + return c[b / CHARCLASS_WORD_BITS] >> b % CHARCLASS_WORD_BITS & 1; +} + +static void +setbit (unsigned int b, charclass c) +{ + c[b / CHARCLASS_WORD_BITS] |= (charclass_word) 1 << b % CHARCLASS_WORD_BITS; +} + +static void +clrbit (unsigned int b, charclass c) +{ + c[b / CHARCLASS_WORD_BITS] &= ~((charclass_word) 1 + << b % CHARCLASS_WORD_BITS); +} + +static void +copyset (charclass const src, charclass dst) +{ + memcpy (dst, src, sizeof (charclass)); +} + +static void +zeroset (charclass s) +{ + memset (s, 0, sizeof (charclass)); +} + +static void +fillset (charclass s) +{ + int i; + for (i = 0; i < CHARCLASS_WORDS; i++) + s[i] = CHARCLASS_WORD_MASK; +} + +static void +notset (charclass s) +{ + int i; + for (i = 0; i < CHARCLASS_WORDS; ++i) + s[i] = CHARCLASS_WORD_MASK & ~s[i]; +} + +static bool +equal (charclass const s1, charclass const s2) +{ + charclass_word w = 0; + int i; + for (i = 0; i < CHARCLASS_WORDS; i++) + w |= s1[i] ^ s2[i]; + return w == 0; +} + +static bool +emptyset (charclass const s) +{ + charclass_word w = 0; + int i; + for (i = 0; i < CHARCLASS_WORDS; i++) + w |= s[i]; + return w == 0; +} + +/* Grow PA, which points to an array of *NITEMS items, and return the + location of the reallocated array, updating *NITEMS to reflect its + new size. The new array will contain at least NITEMS_INCR_MIN more + items, but will not contain more than NITEMS_MAX items total. + ITEM_SIZE is the size of each item, in bytes. + + ITEM_SIZE and NITEMS_INCR_MIN must be positive. *NITEMS must be + nonnegative. If NITEMS_MAX is -1, it is treated as if it were + infinity. + + If PA is null, then allocate a new array instead of reallocating + the old one. + + Thus, to grow an array A without saving its old contents, do + { free (A); A = xpalloc (NULL, &AITEMS, ...); }. */ + +static void * +xpalloc (void *pa, ptrdiff_t *nitems, ptrdiff_t nitems_incr_min, + ptrdiff_t nitems_max, ptrdiff_t item_size) +{ + ptrdiff_t n0 = *nitems; + + /* The approximate size to use for initial small allocation + requests. This is the largest "small" request for the GNU C + library malloc. */ + enum { DEFAULT_MXFAST = 64 * sizeof (size_t) / 4 }; + + /* If the array is tiny, grow it to about (but no greater than) + DEFAULT_MXFAST bytes. Otherwise, grow it by about 50%. + Adjust the growth according to three constraints: NITEMS_INCR_MIN, + NITEMS_MAX, and what the C language can represent safely. */ + + ptrdiff_t n, nbytes; + if (INT_ADD_WRAPV (n0, n0 >> 1, &n)) + n = PTRDIFF_MAX; + if (0 <= nitems_max && nitems_max < n) + n = nitems_max; + + ptrdiff_t adjusted_nbytes + = ((INT_MULTIPLY_WRAPV (n, item_size, &nbytes) || SIZE_MAX < nbytes) + ? MIN (PTRDIFF_MAX, SIZE_MAX) + : nbytes < DEFAULT_MXFAST ? DEFAULT_MXFAST : 0); + if (adjusted_nbytes) + { + n = adjusted_nbytes / item_size; + nbytes = adjusted_nbytes - adjusted_nbytes % item_size; + } + + if (! pa) + *nitems = 0; + if (n - n0 < nitems_incr_min + && (INT_ADD_WRAPV (n0, nitems_incr_min, &n) + || (0 <= nitems_max && nitems_max < n) + || INT_MULTIPLY_WRAPV (n, item_size, &nbytes))) + xalloc_die (); + pa = xrealloc (pa, nbytes); + *nitems = n; + return pa; +} + +/* Ensure that the array addressed by PA holds at least I + 1 items. + Either return PA, or reallocate the array and return its new address. + Although PA may be null, the returned value is never null. + + The array holds *NITEMS items, where 0 <= I <= *NITEMS; *NITEMS + is updated on reallocation. If PA is null, *NITEMS must be zero. + Do not allocate more than NITEMS_MAX items total; -1 means no limit. + ITEM_SIZE is the size of one item; it must be positive. + Avoid O(N**2) behavior on arrays growing linearly. */ +static void * +maybe_realloc (void *pa, ptrdiff_t i, ptrdiff_t *nitems, + ptrdiff_t nitems_max, ptrdiff_t item_size) +{ + if (i < *nitems) + return pa; + return xpalloc (pa, nitems, 1, nitems_max, item_size); +} + +/* In DFA D, find the index of charclass S, or allocate a new one. */ +static ptrdiff_t +charclass_index (struct dfa *d, charclass const s) +{ + ptrdiff_t i; + + for (i = 0; i < d->cindex; ++i) + if (equal (s, d->charclasses[i])) + return i; + d->charclasses = maybe_realloc (d->charclasses, d->cindex, &d->calloc, + TOKEN_MAX - CSET, sizeof *d->charclasses); + ++d->cindex; + copyset (s, d->charclasses[i]); + return i; +} + +static bool +unibyte_word_constituent (struct dfa const *dfa, unsigned char c) +{ + return dfa->localeinfo.sbctowc[c] != WEOF && (isalnum (c) || (c) == '_'); +} + +static int +char_context (struct dfa const *dfa, unsigned char c) +{ + if (c == dfa->syntax.eolbyte && !dfa->syntax.anchor) + return CTX_NEWLINE; + if (unibyte_word_constituent (dfa, c)) + return CTX_LETTER; + return CTX_NONE; +} + +/* Copy the syntax settings from one dfa instance to another. + Saves considerable computation time if compiling many regular expressions + based on the same setting. */ +void +dfacopysyntax (struct dfa *to, const struct dfa *from) +{ + to->dfaexec = from->dfaexec; + to->simple_locale = from->simple_locale; + to->localeinfo = from->localeinfo; + + to->fast = from->fast; + + to->canychar = from->canychar; + to->lex.cur_mb_len = from->lex.cur_mb_len; + to->syntax = from->syntax; +} + +/* Set a bit in the charclass for the given wchar_t. Do nothing if WC + is represented by a multi-byte sequence. Even for MB_CUR_MAX == 1, + this may happen when folding case in weird Turkish locales where + dotless i/dotted I are not included in the chosen character set. + Return whether a bit was set in the charclass. */ +static bool +setbit_wc (wint_t wc, charclass c) +{ + int b = wctob (wc); + if (b == EOF) + return false; + + setbit (b, c); + return true; +} + +/* Set a bit for B and its case variants in the charclass C. + MB_CUR_MAX must be 1. */ +static void +setbit_case_fold_c (int b, charclass c) +{ + int ub = toupper (b); + int i; + for (i = 0; i < NOTCHAR; i++) + if (toupper (i) == ub) + setbit (i, c); +} + +/* Return true if the locale compatible with the C locale. */ + +static bool +using_simple_locale (bool multibyte) +{ + /* The native character set is known to be compatible with + the C locale. The following test isn't perfect, but it's good + enough in practice, as only ASCII and EBCDIC are in common use + and this test correctly accepts ASCII and rejects EBCDIC. */ + enum { native_c_charset = + ('\b' == 8 && '\t' == 9 && '\n' == 10 && '\v' == 11 && '\f' == 12 + && '\r' == 13 && ' ' == 32 && '!' == 33 && '"' == 34 && '#' == 35 + && '%' == 37 && '&' == 38 && '\'' == 39 && '(' == 40 && ')' == 41 + && '*' == 42 && '+' == 43 && ',' == 44 && '-' == 45 && '.' == 46 + && '/' == 47 && '0' == 48 && '9' == 57 && ':' == 58 && ';' == 59 + && '<' == 60 && '=' == 61 && '>' == 62 && '?' == 63 && 'A' == 65 + && 'Z' == 90 && '[' == 91 && '\\' == 92 && ']' == 93 && '^' == 94 + && '_' == 95 && 'a' == 97 && 'z' == 122 && '{' == 123 && '|' == 124 + && '}' == 125 && '~' == 126) + }; + + if (!native_c_charset || multibyte) + return false; + else + { + /* Treat C and POSIX locales as being compatible. Also, treat + errors as compatible, as these are invariably from stubs. */ + char const *loc = setlocale (LC_ALL, NULL); + return !loc || STREQ (loc, "C") || STREQ (loc, "POSIX"); + } +} + +/* Fetch the next lexical input character. Set C (of type int) to the + next input byte, except set C to EOF if the input is a multibyte + character of length greater than 1. Set WC (of type wint_t) to the + value of the input if it is a valid multibyte character (possibly + of length 1); otherwise set WC to WEOF. If there is no more input, + report EOFERR if EOFERR is not null, and return lasttok = END + otherwise. */ +# define FETCH_WC(dfa, c, wc, eoferr) \ + do { \ + if (! (dfa)->lex.left) \ + { \ + if ((eoferr) != 0) \ + dfaerror (eoferr); \ + else \ + return (dfa)->lex.lasttok = END; \ + } \ + else \ + { \ + wint_t _wc; \ + size_t nbytes = mbs_to_wchar (&_wc, (dfa)->lex.ptr, \ + (dfa)->lex.left, dfa); \ + (dfa)->lex.cur_mb_len = nbytes; \ + (wc) = _wc; \ + (c) = nbytes == 1 ? to_uchar ((dfa)->lex.ptr[0]) : EOF; \ + (dfa)->lex.ptr += nbytes; \ + (dfa)->lex.left -= nbytes; \ + } \ + } while (false) + +typedef int predicate (int); + +/* The following list maps the names of the Posix named character classes + to predicate functions that determine whether a given character is in + the class. The leading [ has already been eaten by the lexical + analyzer. */ +struct dfa_ctype +{ + const char *name; + predicate *func; + bool single_byte_only; +}; + +static const struct dfa_ctype prednames[] = { + {"alpha", isalpha, false}, + {"upper", isupper, false}, + {"lower", islower, false}, + {"digit", isdigit, true}, + {"xdigit", isxdigit, false}, + {"space", isspace, false}, + {"punct", ispunct, false}, + {"alnum", isalnum, false}, + {"print", isprint, false}, + {"graph", isgraph, false}, + {"cntrl", iscntrl, false}, + {"blank", is_blank, false}, + {NULL, NULL, false} +}; + +static const struct dfa_ctype *_GL_ATTRIBUTE_PURE +find_pred (const char *str) +{ + unsigned int i; + for (i = 0; prednames[i].name; ++i) + if (STREQ (str, prednames[i].name)) + return &prednames[i]; + return NULL; +} + +/* Multibyte character handling sub-routine for lex. + Parse a bracket expression and build a struct mb_char_classes. */ +static token +parse_bracket_exp (struct dfa *dfa) +{ + bool invert; + int c, c1, c2; + charclass ccl; + + /* This is a bracket expression that dfaexec is known to + process correctly. */ + bool known_bracket_exp = true; + + /* Used to warn about [:space:]. + Bit 0 = first character is a colon. + Bit 1 = last character is a colon. + Bit 2 = includes any other character but a colon. + Bit 3 = includes ranges, char/equiv classes or collation elements. */ + int colon_warning_state; + + wint_t wc; + wint_t wc2; + wint_t wc1 = 0; + + /* Work area to build a mb_char_classes. */ + struct mb_char_classes *work_mbc; + ptrdiff_t chars_al; + + chars_al = 0; + if (dfa->localeinfo.multibyte) + { + dfa->mbcsets = maybe_realloc (dfa->mbcsets, dfa->nmbcsets, + &dfa->mbcsets_alloc, -1, + sizeof *dfa->mbcsets); + + /* dfa->multibyte_prop[] hold the index of dfa->mbcsets. + We will update dfa->multibyte_prop[] in addtok, because we can't + decide the index in dfa->tokens[]. */ + + /* Initialize work area. */ + work_mbc = &dfa->mbcsets[dfa->nmbcsets++]; + memset (work_mbc, 0, sizeof *work_mbc); + } + else + work_mbc = NULL; + + memset (ccl, 0, sizeof ccl); + FETCH_WC (dfa, c, wc, _("unbalanced [")); + if (c == '^') + { + FETCH_WC (dfa, c, wc, _("unbalanced [")); + invert = true; + known_bracket_exp = dfa->simple_locale; + } + else + invert = false; + + colon_warning_state = (c == ':'); + do + { + c1 = NOTCHAR; /* Mark c1 as not initialized. */ + colon_warning_state &= ~2; + + /* Note that if we're looking at some other [:...:] construct, + we just treat it as a bunch of ordinary characters. We can do + this because we assume regex has checked for syntax errors before + dfa is ever called. */ + if (c == '[') + { + FETCH_WC (dfa, c1, wc1, _("unbalanced [")); + + if ((c1 == ':' && (dfa->syntax.syntax_bits & RE_CHAR_CLASSES)) + || c1 == '.' || c1 == '=') + { + enum { MAX_BRACKET_STRING_LEN = 32 }; + char str[MAX_BRACKET_STRING_LEN + 1]; + size_t len = 0; + for (;;) + { + FETCH_WC (dfa, c, wc, _("unbalanced [")); + if (dfa->lex.left == 0 + || (c == c1 && dfa->lex.ptr[0] == ']')) + break; + if (len < MAX_BRACKET_STRING_LEN) + str[len++] = c; + else + /* This is in any case an invalid class name. */ + str[0] = '\0'; + } + str[len] = '\0'; + + /* Fetch bracket. */ + FETCH_WC (dfa, c, wc, _("unbalanced [")); + if (c1 == ':') + /* Build character class. POSIX allows character + classes to match multicharacter collating elements, + but the regex code does not support that, so do not + worry about that possibility. */ + { + char const *class + = (dfa->syntax.case_fold && (STREQ (str, "upper") + || STREQ (str, "lower")) + ? "alpha" : str); + const struct dfa_ctype *pred = find_pred (class); + if (!pred) + dfaerror (_("invalid character class")); + + if (dfa->localeinfo.multibyte && !pred->single_byte_only) + known_bracket_exp = false; + else + for (c2 = 0; c2 < NOTCHAR; ++c2) + if (pred->func (c2)) + setbit (c2, ccl); + } + else + known_bracket_exp = false; + + colon_warning_state |= 8; + + /* Fetch new lookahead character. */ + FETCH_WC (dfa, c1, wc1, _("unbalanced [")); + continue; + } + + /* We treat '[' as a normal character here. c/c1/wc/wc1 + are already set up. */ + } + + if (c == '\\' && (dfa->syntax.syntax_bits & RE_BACKSLASH_ESCAPE_IN_LISTS)) + FETCH_WC (dfa, c, wc, _("unbalanced [")); + + if (c1 == NOTCHAR) + FETCH_WC (dfa, c1, wc1, _("unbalanced [")); + + if (c1 == '-') + /* build range characters. */ + { + FETCH_WC (dfa, c2, wc2, _("unbalanced [")); + + /* A bracket expression like [a-[.aa.]] matches an unknown set. + Treat it like [-a[.aa.]] while parsing it, and + remember that the set is unknown. */ + if (c2 == '[' && dfa->lex.ptr[0] == '.') + { + known_bracket_exp = false; + c2 = ']'; + } + + if (c2 == ']') + { + /* In the case [x-], the - is an ordinary hyphen, + which is left in c1, the lookahead character. */ + dfa->lex.ptr -= dfa->lex.cur_mb_len; + dfa->lex.left += dfa->lex.cur_mb_len; + } + else + { + if (c2 == '\\' && (dfa->syntax.syntax_bits + & RE_BACKSLASH_ESCAPE_IN_LISTS)) + FETCH_WC (dfa, c2, wc2, _("unbalanced [")); + + colon_warning_state |= 8; + FETCH_WC (dfa, c1, wc1, _("unbalanced [")); + + /* Treat [x-y] as a range if x != y. */ + if (wc != wc2 || wc == WEOF) + { + if (dfa->localeinfo.multibyte) + known_bracket_exp = false; + else if (dfa->simple_locale) + { + int ci; + for (ci = c; ci <= c2; ci++) + setbit (ci, ccl); + if (dfa->syntax.case_fold) + { + int uc = toupper (c); + int uc2 = toupper (c2); + for (ci = 0; ci < NOTCHAR; ci++) + { + int uci = toupper (ci); + if (uc <= uci && uci <= uc2) + setbit (ci, ccl); + } + } + } + else + known_bracket_exp = false; + + continue; + } + } + } + + colon_warning_state |= (c == ':') ? 2 : 4; + + if (!dfa->localeinfo.multibyte) + { + if (dfa->syntax.case_fold) + setbit_case_fold_c (c, ccl); + else + setbit (c, ccl); + continue; + } + + if (wc == WEOF) + known_bracket_exp = false; + else + { + wchar_t folded[CASE_FOLDED_BUFSIZE + 1]; + unsigned int i; + unsigned int n = (dfa->syntax.case_fold + ? case_folded_counterparts (wc, folded + 1) + 1 + : 1); + folded[0] = wc; + for (i = 0; i < n; i++) + if (!setbit_wc (folded[i], ccl)) + { + work_mbc->chars + = maybe_realloc (work_mbc->chars, work_mbc->nchars, + &chars_al, -1, sizeof *work_mbc->chars); + work_mbc->chars[work_mbc->nchars++] = folded[i]; + } + } + } + while ((wc = wc1, (c = c1) != ']')); + + if (colon_warning_state == 7) + dfawarn (_("character class syntax is [[:space:]], not [:space:]")); + + if (! known_bracket_exp) + return BACKREF; + + if (dfa->localeinfo.multibyte) + { + work_mbc->invert = invert; + work_mbc->cset = emptyset (ccl) ? -1 : charclass_index (dfa, ccl); + return MBCSET; + } + + if (invert) + { + assert (!dfa->localeinfo.multibyte); + notset (ccl); + if (dfa->syntax.syntax_bits & RE_HAT_LISTS_NOT_NEWLINE) + clrbit ('\n', ccl); + } + + return CSET + charclass_index (dfa, ccl); +} + +struct lexptr +{ + char const *ptr; + size_t left; +}; + +static void +push_lex_state (struct dfa *dfa, struct lexptr *ls, char const *s) +{ + ls->ptr = dfa->lex.ptr; + ls->left = dfa->lex.left; + dfa->lex.ptr = s; + dfa->lex.left = strlen (s); +} + +static void +pop_lex_state (struct dfa *dfa, struct lexptr const *ls) +{ + dfa->lex.ptr = ls->ptr; + dfa->lex.left = ls->left; +} + +static token +lex (struct dfa *dfa) +{ + int c, c2; + bool backslash = false; + charclass ccl; + int i; + + /* Basic plan: We fetch a character. If it's a backslash, + we set the backslash flag and go through the loop again. + On the plus side, this avoids having a duplicate of the + main switch inside the backslash case. On the minus side, + it means that just about every case begins with + "if (backslash) ...". */ + for (i = 0; i < 2; ++i) + { + FETCH_WC (dfa, c, dfa->lex.wctok, NULL); + + switch (c) + { + case '\\': + if (backslash) + goto normal_char; + if (dfa->lex.left == 0) + dfaerror (_("unfinished \\ escape")); + backslash = true; + break; + + case '^': + if (backslash) + goto normal_char; + if (dfa->syntax.syntax_bits & RE_CONTEXT_INDEP_ANCHORS + || dfa->lex.lasttok == END || dfa->lex.lasttok == LPAREN + || dfa->lex.lasttok == OR) + return dfa->lex.lasttok = BEGLINE; + goto normal_char; + + case '$': + if (backslash) + goto normal_char; + if (dfa->syntax.syntax_bits & RE_CONTEXT_INDEP_ANCHORS + || dfa->lex.left == 0 + || ((dfa->lex.left + > !(dfa->syntax.syntax_bits & RE_NO_BK_PARENS)) + && (dfa->lex.ptr[!(dfa->syntax.syntax_bits & RE_NO_BK_PARENS) + & (dfa->lex.ptr[0] == '\\')] + == ')')) + || ((dfa->lex.left + > !(dfa->syntax.syntax_bits & RE_NO_BK_VBAR)) + && (dfa->lex.ptr[!(dfa->syntax.syntax_bits & RE_NO_BK_VBAR) + & (dfa->lex.ptr[0] == '\\')] + == '|')) + || ((dfa->syntax.syntax_bits & RE_NEWLINE_ALT) + && dfa->lex.left > 0 && dfa->lex.ptr[0] == '\n')) + return dfa->lex.lasttok = ENDLINE; + goto normal_char; + + case '1': + case '2': + case '3': + case '4': + case '5': + case '6': + case '7': + case '8': + case '9': + if (backslash && !(dfa->syntax.syntax_bits & RE_NO_BK_REFS)) + { + dfa->lex.laststart = false; + return dfa->lex.lasttok = BACKREF; + } + goto normal_char; + + case '`': + if (backslash && !(dfa->syntax.syntax_bits & RE_NO_GNU_OPS)) + { + /* FIXME: should be beginning of string */ + return dfa->lex.lasttok = BEGLINE; + } + goto normal_char; + + case '\'': + if (backslash && !(dfa->syntax.syntax_bits & RE_NO_GNU_OPS)) + { + /* FIXME: should be end of string */ + return dfa->lex.lasttok = ENDLINE; + } + goto normal_char; + + case '<': + if (backslash && !(dfa->syntax.syntax_bits & RE_NO_GNU_OPS)) + return dfa->lex.lasttok = BEGWORD; + goto normal_char; + + case '>': + if (backslash && !(dfa->syntax.syntax_bits & RE_NO_GNU_OPS)) + return dfa->lex.lasttok = ENDWORD; + goto normal_char; + + case 'b': + if (backslash && !(dfa->syntax.syntax_bits & RE_NO_GNU_OPS)) + return dfa->lex.lasttok = LIMWORD; + goto normal_char; + + case 'B': + if (backslash && !(dfa->syntax.syntax_bits & RE_NO_GNU_OPS)) + return dfa->lex.lasttok = NOTLIMWORD; + goto normal_char; + + case '?': + if (dfa->syntax.syntax_bits & RE_LIMITED_OPS) + goto normal_char; + if (backslash != ((dfa->syntax.syntax_bits & RE_BK_PLUS_QM) != 0)) + goto normal_char; + if (!(dfa->syntax.syntax_bits & RE_CONTEXT_INDEP_OPS) + && dfa->lex.laststart) + goto normal_char; + return dfa->lex.lasttok = QMARK; + + case '*': + if (backslash) + goto normal_char; + if (!(dfa->syntax.syntax_bits & RE_CONTEXT_INDEP_OPS) + && dfa->lex.laststart) + goto normal_char; + return dfa->lex.lasttok = STAR; + + case '+': + if (dfa->syntax.syntax_bits & RE_LIMITED_OPS) + goto normal_char; + if (backslash != ((dfa->syntax.syntax_bits & RE_BK_PLUS_QM) != 0)) + goto normal_char; + if (!(dfa->syntax.syntax_bits & RE_CONTEXT_INDEP_OPS) + && dfa->lex.laststart) + goto normal_char; + return dfa->lex.lasttok = PLUS; + + case '{': + if (!(dfa->syntax.syntax_bits & RE_INTERVALS)) + goto normal_char; + if (backslash != ((dfa->syntax.syntax_bits & RE_NO_BK_BRACES) == 0)) + goto normal_char; + if (!(dfa->syntax.syntax_bits & RE_CONTEXT_INDEP_OPS) + && dfa->lex.laststart) + goto normal_char; + + /* Cases: + {M} - exact count + {M,} - minimum count, maximum is infinity + {,N} - 0 through N + {,} - 0 to infinity (same as '*') + {M,N} - M through N */ + { + char const *p = dfa->lex.ptr; + char const *lim = p + dfa->lex.left; + dfa->lex.minrep = dfa->lex.maxrep = -1; + for (; p != lim && ISASCIIDIGIT (*p); p++) + dfa->lex.minrep = (dfa->lex.minrep < 0 + ? *p - '0' + : MIN (RE_DUP_MAX + 1, + dfa->lex.minrep * 10 + *p - '0')); + if (p != lim) + { + if (*p != ',') + dfa->lex.maxrep = dfa->lex.minrep; + else + { + if (dfa->lex.minrep < 0) + dfa->lex.minrep = 0; + while (++p != lim && ISASCIIDIGIT (*p)) + dfa->lex.maxrep + = (dfa->lex.maxrep < 0 + ? *p - '0' + : MIN (RE_DUP_MAX + 1, + dfa->lex.maxrep * 10 + *p - '0')); + } + } + if (! ((! backslash || (p != lim && *p++ == '\\')) + && p != lim && *p++ == '}' + && 0 <= dfa->lex.minrep + && (dfa->lex.maxrep < 0 + || dfa->lex.minrep <= dfa->lex.maxrep))) + { + if (dfa->syntax.syntax_bits & RE_INVALID_INTERVAL_ORD) + goto normal_char; + dfaerror (_("invalid content of \\{\\}")); + } + if (RE_DUP_MAX < dfa->lex.maxrep) + dfaerror (_("regular expression too big")); + dfa->lex.ptr = p; + dfa->lex.left = lim - p; + } + dfa->lex.laststart = false; + return dfa->lex.lasttok = REPMN; + + case '|': + if (dfa->syntax.syntax_bits & RE_LIMITED_OPS) + goto normal_char; + if (backslash != ((dfa->syntax.syntax_bits & RE_NO_BK_VBAR) == 0)) + goto normal_char; + dfa->lex.laststart = true; + return dfa->lex.lasttok = OR; + + case '\n': + if (dfa->syntax.syntax_bits & RE_LIMITED_OPS + || backslash || !(dfa->syntax.syntax_bits & RE_NEWLINE_ALT)) + goto normal_char; + dfa->lex.laststart = true; + return dfa->lex.lasttok = OR; + + case '(': + if (backslash != ((dfa->syntax.syntax_bits & RE_NO_BK_PARENS) == 0)) + goto normal_char; + dfa->lex.parens++; + dfa->lex.laststart = true; + return dfa->lex.lasttok = LPAREN; + + case ')': + if (backslash != ((dfa->syntax.syntax_bits & RE_NO_BK_PARENS) == 0)) + goto normal_char; + if (dfa->lex.parens == 0 + && dfa->syntax.syntax_bits & RE_UNMATCHED_RIGHT_PAREN_ORD) + goto normal_char; + dfa->lex.parens--; + dfa->lex.laststart = false; + return dfa->lex.lasttok = RPAREN; + + case '.': + if (backslash) + goto normal_char; + if (dfa->canychar == (size_t) -1) + { + fillset (ccl); + if (!(dfa->syntax.syntax_bits & RE_DOT_NEWLINE)) + clrbit ('\n', ccl); + if (dfa->syntax.syntax_bits & RE_DOT_NOT_NULL) + clrbit ('\0', ccl); + if (dfa->localeinfo.multibyte) + for (c2 = 0; c2 < NOTCHAR; c2++) + if (dfa->localeinfo.sbctowc[c2] == WEOF) + clrbit (c2, ccl); + dfa->canychar = charclass_index (dfa, ccl); + } + dfa->lex.laststart = false; + return dfa->lex.lasttok = (dfa->localeinfo.multibyte + ? ANYCHAR + : CSET + dfa->canychar); + + case 's': + case 'S': + if (!backslash || (dfa->syntax.syntax_bits & RE_NO_GNU_OPS)) + goto normal_char; + if (!dfa->localeinfo.multibyte) + { + zeroset (ccl); + for (c2 = 0; c2 < NOTCHAR; ++c2) + if (isspace (c2)) + setbit (c2, ccl); + if (c == 'S') + notset (ccl); + dfa->lex.laststart = false; + return dfa->lex.lasttok = CSET + charclass_index (dfa, ccl); + } + + /* FIXME: see if optimizing this, as is done with ANYCHAR and + add_utf8_anychar, makes sense. */ + + /* \s and \S are documented to be equivalent to [[:space:]] and + [^[:space:]] respectively, so tell the lexer to process those + strings, each minus its "already processed" '['. */ + { + struct lexptr ls; + push_lex_state (dfa, &ls, &"^[:space:]]"[c == 's']); + dfa->lex.lasttok = parse_bracket_exp (dfa); + pop_lex_state (dfa, &ls); + } + + dfa->lex.laststart = false; + return dfa->lex.lasttok; + + case 'w': + case 'W': + if (!backslash || (dfa->syntax.syntax_bits & RE_NO_GNU_OPS)) + goto normal_char; + + if (!dfa->localeinfo.multibyte) + { + zeroset (ccl); + for (c2 = 0; c2 < NOTCHAR; ++c2) + if (dfa->syntax.sbit[c2] == CTX_LETTER) + setbit (c2, ccl); + if (c == 'W') + notset (ccl); + dfa->lex.laststart = false; + return dfa->lex.lasttok = CSET + charclass_index (dfa, ccl); + } + + /* FIXME: see if optimizing this, as is done with ANYCHAR and + add_utf8_anychar, makes sense. */ + + /* \w and \W are documented to be equivalent to [_[:alnum:]] and + [^_[:alnum:]] respectively, so tell the lexer to process those + strings, each minus its "already processed" '['. */ + { + struct lexptr ls; + push_lex_state (dfa, &ls, &"^_[:alnum:]]"[c == 'w']); + dfa->lex.lasttok = parse_bracket_exp (dfa); + pop_lex_state (dfa, &ls); + } + + dfa->lex.laststart = false; + return dfa->lex.lasttok; + + case '[': + if (backslash) + goto normal_char; + dfa->lex.laststart = false; + return dfa->lex.lasttok = parse_bracket_exp (dfa); + + default: + normal_char: + dfa->lex.laststart = false; + /* For multibyte character sets, folding is done in atom. Always + return WCHAR. */ + if (dfa->localeinfo.multibyte) + return dfa->lex.lasttok = WCHAR; + + if (dfa->syntax.case_fold && isalpha (c)) + { + zeroset (ccl); + setbit_case_fold_c (c, ccl); + return dfa->lex.lasttok = CSET + charclass_index (dfa, ccl); + } + + return dfa->lex.lasttok = c; + } + } + + /* The above loop should consume at most a backslash + and some other character. */ + abort (); + return END; /* keeps pedantic compilers happy. */ +} + +static void +addtok_mb (struct dfa *dfa, token t, int mbprop) +{ + if (dfa->talloc == dfa->tindex) + { + dfa->tokens = x2nrealloc (dfa->tokens, &dfa->talloc, + sizeof *dfa->tokens); + if (dfa->localeinfo.multibyte) + dfa->multibyte_prop = xnrealloc (dfa->multibyte_prop, dfa->talloc, + sizeof *dfa->multibyte_prop); + } + if (dfa->localeinfo.multibyte) + dfa->multibyte_prop[dfa->tindex] = mbprop; + dfa->tokens[dfa->tindex++] = t; + + switch (t) + { + case QMARK: + case STAR: + case PLUS: + break; + + case CAT: + case OR: + dfa->parse.depth--; + break; + + case BACKREF: + dfa->fast = false; + /* fallthrough */ + default: + dfa->nleaves++; + /* fallthrough */ + case EMPTY: + dfa->parse.depth++; + break; + } + if (dfa->parse.depth > dfa->depth) + dfa->depth = dfa->parse.depth; +} + +static void addtok_wc (struct dfa *dfa, wint_t wc); + +/* Add the given token to the parse tree, maintaining the depth count and + updating the maximum depth if necessary. */ +static void +addtok (struct dfa *dfa, token t) +{ + if (dfa->localeinfo.multibyte && t == MBCSET) + { + bool need_or = false; + struct mb_char_classes *work_mbc = &dfa->mbcsets[dfa->nmbcsets - 1]; + ptrdiff_t i; + + /* Extract wide characters into alternations for better performance. + This does not require UTF-8. */ + for (i = 0; i < work_mbc->nchars; i++) + { + addtok_wc (dfa, work_mbc->chars[i]); + if (need_or) + addtok (dfa, OR); + need_or = true; + } + work_mbc->nchars = 0; + + /* Characters have been handled above, so it is possible + that the mbcset is empty now. Do nothing in that case. */ + if (work_mbc->cset != -1) + { + addtok (dfa, CSET + work_mbc->cset); + if (need_or) + addtok (dfa, OR); + } + } + else + { + addtok_mb (dfa, t, 3); + } +} + +/* We treat a multibyte character as a single atom, so that DFA + can treat a multibyte character as a single expression. + + e.g., we construct the following tree from "<mb1><mb2>". + <mb1(1st-byte)><mb1(2nd-byte)><CAT><mb1(3rd-byte)><CAT> + <mb2(1st-byte)><mb2(2nd-byte)><CAT><mb2(3rd-byte)><CAT><CAT> */ +static void +addtok_wc (struct dfa *dfa, wint_t wc) +{ + unsigned char buf[MB_LEN_MAX]; + mbstate_t s = { 0 }; + int i; + size_t stored_bytes = wcrtomb ((char *) buf, wc, &s); + + if (stored_bytes != (size_t) -1) + dfa->lex.cur_mb_len = stored_bytes; + else + { + /* This is merely stop-gap. buf[0] is undefined, yet skipping + the addtok_mb call altogether can corrupt the heap. */ + dfa->lex.cur_mb_len = 1; + buf[0] = 0; + } + + addtok_mb (dfa, buf[0], dfa->lex.cur_mb_len == 1 ? 3 : 1); + for (i = 1; i < dfa->lex.cur_mb_len; i++) + { + addtok_mb (dfa, buf[i], i == dfa->lex.cur_mb_len - 1 ? 2 : 0); + addtok (dfa, CAT); + } +} + +static void +add_utf8_anychar (struct dfa *dfa) +{ + static charclass const utf8_classes[5] = { + /* 80-bf: non-leading bytes. */ + CHARCLASS_INIT (0, 0, 0, 0, 0xffffffff, 0xffffffff, 0, 0), + + /* 00-7f: 1-byte sequence. */ + CHARCLASS_INIT (0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff, 0, 0, 0, 0), + + /* c2-df: 2-byte sequence. */ + CHARCLASS_INIT (0, 0, 0, 0, 0, 0, 0xfffffffc, 0), + + /* e0-ef: 3-byte sequence. */ + CHARCLASS_INIT (0, 0, 0, 0, 0, 0, 0, 0xffff), + + /* f0-f7: 4-byte sequence. */ + CHARCLASS_INIT (0, 0, 0, 0, 0, 0, 0, 0xff0000) + }; + const unsigned int n = sizeof (utf8_classes) / sizeof (utf8_classes[0]); + unsigned int i; + + /* Define the five character classes that are needed below. */ + if (dfa->utf8_anychar_classes[0] == 0) + for (i = 0; i < n; i++) + { + charclass c; + copyset (utf8_classes[i], c); + if (i == 1) + { + if (!(dfa->syntax.syntax_bits & RE_DOT_NEWLINE)) + clrbit ('\n', c); + if (dfa->syntax.syntax_bits & RE_DOT_NOT_NULL) + clrbit ('\0', c); + } + dfa->utf8_anychar_classes[i] = CSET + charclass_index (dfa, c); + } + + /* A valid UTF-8 character is + + ([0x00-0x7f] + |[0xc2-0xdf][0x80-0xbf] + |[0xe0-0xef[0x80-0xbf][0x80-0xbf] + |[0xf0-f7][0x80-0xbf][0x80-0xbf][0x80-0xbf]) + + which I'll write more concisely "B|CA|DAA|EAAA". Factor the [0x00-0x7f] + and you get "B|(C|(D|EA)A)A". And since the token buffer is in reverse + Polish notation, you get "B C D E A CAT OR A CAT OR A CAT OR". */ + for (i = 1; i < n; i++) + addtok (dfa, dfa->utf8_anychar_classes[i]); + while (--i > 1) + { + addtok (dfa, dfa->utf8_anychar_classes[0]); + addtok (dfa, CAT); + addtok (dfa, OR); + } +} + +/* The grammar understood by the parser is as follows. + + regexp: + regexp OR branch + branch + + branch: + branch closure + closure + + closure: + closure QMARK + closure STAR + closure PLUS + closure REPMN + atom + + atom: + <normal character> + <multibyte character> + ANYCHAR + MBCSET + CSET + BACKREF + BEGLINE + ENDLINE + BEGWORD + ENDWORD + LIMWORD + NOTLIMWORD + LPAREN regexp RPAREN + <empty> + + The parser builds a parse tree in postfix form in an array of tokens. */ + +static void +atom (struct dfa *dfa) +{ + if (dfa->parse.tok == WCHAR) + { + if (dfa->lex.wctok == WEOF) + addtok (dfa, BACKREF); + else + { + addtok_wc (dfa, dfa->lex.wctok); + + if (dfa->syntax.case_fold) + { + wchar_t folded[CASE_FOLDED_BUFSIZE]; + unsigned int i, n = case_folded_counterparts (dfa->lex.wctok, + folded); + for (i = 0; i < n; i++) + { + addtok_wc (dfa, folded[i]); + addtok (dfa, OR); + } + } + } + + dfa->parse.tok = lex (dfa); + } + else if (dfa->parse.tok == ANYCHAR && dfa->localeinfo.using_utf8) + { + /* For UTF-8 expand the period to a series of CSETs that define a valid + UTF-8 character. This avoids using the slow multibyte path. I'm + pretty sure it would be both profitable and correct to do it for + any encoding; however, the optimization must be done manually as + it is done above in add_utf8_anychar. So, let's start with + UTF-8: it is the most used, and the structure of the encoding + makes the correctness more obvious. */ + add_utf8_anychar (dfa); + dfa->parse.tok = lex (dfa); + } + else if ((0 <= dfa->parse.tok && dfa->parse.tok < NOTCHAR) + || dfa->parse.tok >= CSET || dfa->parse.tok == BACKREF + || dfa->parse.tok == BEGLINE || dfa->parse.tok == ENDLINE + || dfa->parse.tok == BEGWORD || dfa->parse.tok == ANYCHAR + || dfa->parse.tok == MBCSET || dfa->parse.tok == ENDWORD + || dfa->parse.tok == LIMWORD || dfa->parse.tok == NOTLIMWORD) + { + addtok (dfa, dfa->parse.tok); + dfa->parse.tok = lex (dfa); + } + else if (dfa->parse.tok == LPAREN) + { + dfa->parse.tok = lex (dfa); + regexp (dfa); + if (dfa->parse.tok != RPAREN) + dfaerror (_("unbalanced (")); + dfa->parse.tok = lex (dfa); + } + else + addtok (dfa, EMPTY); +} + +/* Return the number of tokens in the given subexpression. */ +static size_t _GL_ATTRIBUTE_PURE +nsubtoks (struct dfa const *dfa, size_t tindex) +{ + size_t ntoks1; + + switch (dfa->tokens[tindex - 1]) + { + default: + return 1; + case QMARK: + case STAR: + case PLUS: + return 1 + nsubtoks (dfa, tindex - 1); + case CAT: + case OR: + ntoks1 = nsubtoks (dfa, tindex - 1); + return 1 + ntoks1 + nsubtoks (dfa, tindex - 1 - ntoks1); + } +} + +/* Copy the given subexpression to the top of the tree. */ +static void +copytoks (struct dfa *dfa, size_t tindex, size_t ntokens) +{ + size_t i; + + if (dfa->localeinfo.multibyte) + for (i = 0; i < ntokens; ++i) + addtok_mb (dfa, dfa->tokens[tindex + i], dfa->multibyte_prop[tindex + i]); + else + for (i = 0; i < ntokens; ++i) + addtok_mb (dfa, dfa->tokens[tindex + i], 3); +} + +static void +closure (struct dfa *dfa) +{ + int i; + size_t tindex, ntokens; + + atom (dfa); + while (dfa->parse.tok == QMARK || dfa->parse.tok == STAR + || dfa->parse.tok == PLUS || dfa->parse.tok == REPMN) + if (dfa->parse.tok == REPMN && (dfa->lex.minrep || dfa->lex.maxrep)) + { + ntokens = nsubtoks (dfa, dfa->tindex); + tindex = dfa->tindex - ntokens; + if (dfa->lex.maxrep < 0) + addtok (dfa, PLUS); + if (dfa->lex.minrep == 0) + addtok (dfa, QMARK); + for (i = 1; i < dfa->lex.minrep; i++) + { + copytoks (dfa, tindex, ntokens); + addtok (dfa, CAT); + } + for (; i < dfa->lex.maxrep; i++) + { + copytoks (dfa, tindex, ntokens); + addtok (dfa, QMARK); + addtok (dfa, CAT); + } + dfa->parse.tok = lex (dfa); + } + else if (dfa->parse.tok == REPMN) + { + dfa->tindex -= nsubtoks (dfa, dfa->tindex); + dfa->parse.tok = lex (dfa); + closure (dfa); + } + else + { + addtok (dfa, dfa->parse.tok); + dfa->parse.tok = lex (dfa); + } +} + +static void +branch (struct dfa* dfa) +{ + closure (dfa); + while (dfa->parse.tok != RPAREN && dfa->parse.tok != OR + && dfa->parse.tok >= 0) + { + closure (dfa); + addtok (dfa, CAT); + } +} + +static void +regexp (struct dfa *dfa) +{ + branch (dfa); + while (dfa->parse.tok == OR) + { + dfa->parse.tok = lex (dfa); + branch (dfa); + addtok (dfa, OR); + } +} + +/* Main entry point for the parser. S is a string to be parsed, len is the + length of the string, so s can include NUL characters. D is a pointer to + the struct dfa to parse into. */ +static void +dfaparse (char const *s, size_t len, struct dfa *d) +{ + d->lex.ptr = s; + d->lex.left = len; + d->lex.lasttok = END; + d->lex.laststart = true; + d->lex.parens = 0; + if (d->localeinfo.multibyte) + { + d->lex.cur_mb_len = 0; + memset (&d->mbs, 0, sizeof d->mbs); + } + + if (!d->syntax.syntax_bits_set) + dfaerror (_("no syntax specified")); + + d->parse.tok = lex (d); + d->parse.depth = d->depth; + + regexp (d); + + if (d->parse.tok != END) + dfaerror (_("unbalanced )")); + + addtok (d, END - d->nregexps); + addtok (d, CAT); + + if (d->nregexps) + addtok (d, OR); + + ++d->nregexps; +} + +/* Some primitives for operating on sets of positions. */ + +/* Copy one set to another. */ +static void +copy (position_set const *src, position_set *dst) +{ + if (dst->alloc < src->nelem) + { + free (dst->elems); + dst->elems = xpalloc (NULL, &dst->alloc, src->nelem - dst->alloc, -1, + sizeof *dst->elems); + } + memcpy (dst->elems, src->elems, src->nelem * sizeof *dst->elems); + dst->nelem = src->nelem; +} + +static void +alloc_position_set (position_set *s, size_t size) +{ + s->elems = xnmalloc (size, sizeof *s->elems); + if (PTRDIFF_MAX < SIZE_MAX / sizeof *s->elems && PTRDIFF_MAX < size) + xalloc_die (); + s->alloc = size; + s->nelem = 0; +} + +/* Insert position P in set S. S is maintained in sorted order on + decreasing index. If there is already an entry in S with P.index + then merge (logically-OR) P's constraints into the one in S. + S->elems must point to an array large enough to hold the resulting set. */ +static void +insert (position p, position_set *s) +{ + ptrdiff_t count = s->nelem; + ptrdiff_t lo = 0, hi = count; + ptrdiff_t i; + while (lo < hi) + { + ptrdiff_t mid = (lo + hi) >> 1; + if (s->elems[mid].index > p.index) + lo = mid + 1; + else if (s->elems[mid].index == p.index) + { + s->elems[mid].constraint |= p.constraint; + return; + } + else + hi = mid; + } + + s->elems = maybe_realloc (s->elems, count, &s->alloc, -1, sizeof *s->elems); + for (i = count; i > lo; i--) + s->elems[i] = s->elems[i - 1]; + s->elems[lo] = p; + ++s->nelem; +} + +/* Merge S1 and S2 (with the additional constraint C2) into M. The + result is as if the positions of S1, and of S2 with the additional + constraint C2, were inserted into an initially empty set. */ +static void +merge_constrained (position_set const *s1, position_set const *s2, + unsigned int c2, position_set *m) +{ + ptrdiff_t i = 0, j = 0; + + if (m->alloc - s1->nelem < s2->nelem) + { + free (m->elems); + m->alloc = s1->nelem; + m->elems = xpalloc (NULL, &m->alloc, s2->nelem, -1, sizeof *m->elems); + } + m->nelem = 0; + while (i < s1->nelem || j < s2->nelem) + if (! (j < s2->nelem) + || (i < s1->nelem && s1->elems[i].index >= s2->elems[j].index)) + { + unsigned int c = ((i < s1->nelem && j < s2->nelem + && s1->elems[i].index == s2->elems[j].index) + ? s2->elems[j++].constraint & c2 + : 0); + m->elems[m->nelem].index = s1->elems[i].index; + m->elems[m->nelem++].constraint = s1->elems[i++].constraint | c; + } + else + { + if (s2->elems[j].constraint & c2) + { + m->elems[m->nelem].index = s2->elems[j].index; + m->elems[m->nelem++].constraint = s2->elems[j].constraint & c2; + } + j++; + } +} + +/* Merge two sets of positions into a third. The result is exactly as if + the positions of both sets were inserted into an initially empty set. */ +static void +merge (position_set const *s1, position_set const *s2, position_set *m) +{ + return merge_constrained (s1, s2, -1, m); +} + +/* Delete a position from a set. Return the nonzero constraint of the + deleted position, or zero if there was no such position. */ +static unsigned int +delete (size_t del, position_set *s) +{ + size_t count = s->nelem; + size_t lo = 0, hi = count; + while (lo < hi) + { + size_t mid = (lo + hi) >> 1; + if (s->elems[mid].index > del) + lo = mid + 1; + else if (s->elems[mid].index == del) + { + unsigned int c = s->elems[mid].constraint; + size_t i; + for (i = mid; i + 1 < count; i++) + s->elems[i] = s->elems[i + 1]; + s->nelem = i; + return c; + } + else + hi = mid; + } + return 0; +} + +/* Replace a position with the followed set. */ +static void +replace (position_set *dst, size_t del, position_set *add, + unsigned int constraint, position_set *tmp) +{ + unsigned int c = delete (del, dst) & constraint; + + if (c) + { + copy (dst, tmp); + merge_constrained (tmp, add, c, dst); + } +} + +/* Find the index of the state corresponding to the given position set with + the given preceding context, or create a new state if there is no such + state. Context tells whether we got here on a newline or letter. */ +static state_num +state_index (struct dfa *d, position_set const *s, int context) +{ + size_t hash = 0; + int constraint = 0; + state_num i, j; + token first_end = 0; + + for (i = 0; i < s->nelem; ++i) + hash ^= s->elems[i].index + s->elems[i].constraint; + + /* Try to find a state that exactly matches the proposed one. */ + for (i = 0; i < d->sindex; ++i) + { + if (hash != d->states[i].hash || s->nelem != d->states[i].elems.nelem + || context != d->states[i].context) + continue; + for (j = 0; j < s->nelem; ++j) + if (s->elems[j].constraint != d->states[i].elems.elems[j].constraint + || s->elems[j].index != d->states[i].elems.elems[j].index) + break; + if (j == s->nelem) + return i; + } + +#ifdef DEBUG + fprintf (stderr, "new state %zd\n nextpos:", i); + for (j = 0; j < s->nelem; ++j) + { + fprintf (stderr, " %zu:", s->elems[j].index); + prtok (d->tokens[s->elems[j].index]); + } + fprintf (stderr, "\n context:"); + if (context ^ CTX_ANY) + { + if (context & CTX_NONE) + fprintf (stderr, " CTX_NONE"); + if (context & CTX_LETTER) + fprintf (stderr, " CTX_LETTER"); + if (context & CTX_NEWLINE) + fprintf (stderr, " CTX_NEWLINE"); + } + else + fprintf (stderr, " CTX_ANY"); + fprintf (stderr, "\n"); +#endif + + for (j = 0; j < s->nelem; ++j) + { + int c = s->elems[j].constraint; + if (d->tokens[s->elems[j].index] < 0) + { + if (SUCCEEDS_IN_CONTEXT (c, context, CTX_ANY)) + constraint |= c; + if (!first_end) + first_end = d->tokens[s->elems[j].index]; + } + else if (d->tokens[s->elems[j].index] == BACKREF) + constraint = NO_CONSTRAINT; + } + + + /* Create a new state. */ + d->states = maybe_realloc (d->states, d->sindex, &d->salloc, -1, + sizeof *d->states); + d->states[i].hash = hash; + alloc_position_set (&d->states[i].elems, s->nelem); + copy (s, &d->states[i].elems); + d->states[i].context = context; + d->states[i].constraint = constraint; + d->states[i].first_end = first_end; + d->states[i].mbps.nelem = 0; + d->states[i].mbps.elems = NULL; + d->states[i].mb_trindex = -1; + + ++d->sindex; + + return i; +} + +/* Find the epsilon closure of a set of positions. If any position of the set + contains a symbol that matches the empty string in some context, replace + that position with the elements of its follow labeled with an appropriate + constraint. Repeat exhaustively until no funny positions are left. + S->elems must be large enough to hold the result. */ +static void +epsclosure (position_set *initial, struct dfa const *d) +{ + position_set tmp; + alloc_position_set (&tmp, d->nleaves); + for (size_t i = 0; i < d->tindex; ++i) + if (d->follows[i].nelem > 0 && d->tokens[i] >= NOTCHAR + && d->tokens[i] != BACKREF && d->tokens[i] != ANYCHAR + && d->tokens[i] != MBCSET && d->tokens[i] < CSET) + { + unsigned int constraint; + switch (d->tokens[i]) + { + case BEGLINE: + constraint = BEGLINE_CONSTRAINT; + break; + case ENDLINE: + constraint = ENDLINE_CONSTRAINT; + break; + case BEGWORD: + constraint = BEGWORD_CONSTRAINT; + break; + case ENDWORD: + constraint = ENDWORD_CONSTRAINT; + break; + case LIMWORD: + constraint = LIMWORD_CONSTRAINT; + break; + case NOTLIMWORD: + constraint = NOTLIMWORD_CONSTRAINT; + break; + default: + constraint = NO_CONSTRAINT; + break; + } + + delete (i, &d->follows[i]); + + for (size_t j = 0; j < d->tindex; j++) + if (i != j && d->follows[j].nelem > 0) + replace (&d->follows[j], i, &d->follows[i], constraint, &tmp); + + replace (initial, i, &d->follows[i], constraint, &tmp); + } +} + +/* Returns the set of contexts for which there is at least one + character included in C. */ + +static int +charclass_context (struct dfa const *dfa, charclass c) +{ + int context = 0; + unsigned int j; + + for (j = 0; j < CHARCLASS_WORDS; ++j) + { + if (c[j] & dfa->syntax.newline[j]) + context |= CTX_NEWLINE; + if (c[j] & dfa->syntax.letters[j]) + context |= CTX_LETTER; + if (c[j] & ~(dfa->syntax.letters[j] | dfa->syntax.newline[j])) + context |= CTX_NONE; + } + + return context; +} + +/* Returns the contexts on which the position set S depends. Each context + in the set of returned contexts (let's call it SC) may have a different + follow set than other contexts in SC, and also different from the + follow set of the complement set (sc ^ CTX_ANY). However, all contexts + in the complement set will have the same follow set. */ + +static int _GL_ATTRIBUTE_PURE +state_separate_contexts (position_set const *s) +{ + int separate_contexts = 0; + size_t j; + + for (j = 0; j < s->nelem; ++j) + { + if (PREV_NEWLINE_DEPENDENT (s->elems[j].constraint)) + separate_contexts |= CTX_NEWLINE; + if (PREV_LETTER_DEPENDENT (s->elems[j].constraint)) + separate_contexts |= CTX_LETTER; + } + + return separate_contexts; +} + + +/* Perform bottom-up analysis on the parse tree, computing various functions. + Note that at this point, we're pretending constructs like \< are real + characters rather than constraints on what can follow them. + + Nullable: A node is nullable if it is at the root of a regexp that can + match the empty string. + * EMPTY leaves are nullable. + * No other leaf is nullable. + * A QMARK or STAR node is nullable. + * A PLUS node is nullable if its argument is nullable. + * A CAT node is nullable if both its arguments are nullable. + * An OR node is nullable if either argument is nullable. + + Firstpos: The firstpos of a node is the set of positions (nonempty leaves) + that could correspond to the first character of a string matching the + regexp rooted at the given node. + * EMPTY leaves have empty firstpos. + * The firstpos of a nonempty leaf is that leaf itself. + * The firstpos of a QMARK, STAR, or PLUS node is the firstpos of its + argument. + * The firstpos of a CAT node is the firstpos of the left argument, union + the firstpos of the right if the left argument is nullable. + * The firstpos of an OR node is the union of firstpos of each argument. + + Lastpos: The lastpos of a node is the set of positions that could + correspond to the last character of a string matching the regexp at + the given node. + * EMPTY leaves have empty lastpos. + * The lastpos of a nonempty leaf is that leaf itself. + * The lastpos of a QMARK, STAR, or PLUS node is the lastpos of its + argument. + * The lastpos of a CAT node is the lastpos of its right argument, union + the lastpos of the left if the right argument is nullable. + * The lastpos of an OR node is the union of the lastpos of each argument. + + Follow: The follow of a position is the set of positions that could + correspond to the character following a character matching the node in + a string matching the regexp. At this point we consider special symbols + that match the empty string in some context to be just normal characters. + Later, if we find that a special symbol is in a follow set, we will + replace it with the elements of its follow, labeled with an appropriate + constraint. + * Every node in the firstpos of the argument of a STAR or PLUS node is in + the follow of every node in the lastpos. + * Every node in the firstpos of the second argument of a CAT node is in + the follow of every node in the lastpos of the first argument. + + Because of the postfix representation of the parse tree, the depth-first + analysis is conveniently done by a linear scan with the aid of a stack. + Sets are stored as arrays of the elements, obeying a stack-like allocation + scheme; the number of elements in each set deeper in the stack can be + used to determine the address of a particular set's array. */ +static void +dfaanalyze (struct dfa *d, bool searchflag) +{ + /* Array allocated to hold position sets. */ + position *posalloc = xnmalloc (d->nleaves, 2 * sizeof *posalloc); + /* Firstpos and lastpos elements. */ + position *firstpos = posalloc + d->nleaves; + position *lastpos = firstpos + d->nleaves; + + /* Stack for element counts and nullable flags. */ + struct + { + /* Whether the entry is nullable. */ + bool nullable; + + /* Counts of firstpos and lastpos sets. */ + size_t nfirstpos; + size_t nlastpos; + } *stkalloc = xnmalloc (d->depth, sizeof *stkalloc), *stk = stkalloc; + + position_set tmp; /* Temporary set for merging sets. */ + position_set merged; /* Result of merging sets. */ + int separate_contexts; /* Context wanted by some position. */ + size_t i, j; + position *pos; + +#ifdef DEBUG + fprintf (stderr, "dfaanalyze:\n"); + for (i = 0; i < d->tindex; ++i) + { + fprintf (stderr, " %zu:", i); + prtok (d->tokens[i]); + } + putc ('\n', stderr); +#endif + + d->searchflag = searchflag; + alloc_position_set (&merged, d->nleaves); + d->follows = xcalloc (d->tindex, sizeof *d->follows); + + for (i = 0; i < d->tindex; ++i) + { + switch (d->tokens[i]) + { + case EMPTY: + /* The empty set is nullable. */ + stk->nullable = true; + + /* The firstpos and lastpos of the empty leaf are both empty. */ + stk->nfirstpos = stk->nlastpos = 0; + stk++; + break; + + case STAR: + case PLUS: + /* Every element in the firstpos of the argument is in the follow + of every element in the lastpos. */ + tmp.nelem = stk[-1].nfirstpos; + tmp.elems = firstpos; + pos = lastpos; + for (j = 0; j < stk[-1].nlastpos; ++j) + { + merge (&tmp, &d->follows[pos[j].index], &merged); + copy (&merged, &d->follows[pos[j].index]); + } + /* fallthrough */ + + case QMARK: + /* A QMARK or STAR node is automatically nullable. */ + if (d->tokens[i] != PLUS) + stk[-1].nullable = true; + break; + + case CAT: + /* Every element in the firstpos of the second argument is in the + follow of every element in the lastpos of the first argument. */ + tmp.nelem = stk[-1].nfirstpos; + tmp.elems = firstpos; + pos = lastpos + stk[-1].nlastpos; + for (j = 0; j < stk[-2].nlastpos; ++j) + { + merge (&tmp, &d->follows[pos[j].index], &merged); + copy (&merged, &d->follows[pos[j].index]); + } + + /* The firstpos of a CAT node is the firstpos of the first argument, + union that of the second argument if the first is nullable. */ + if (stk[-2].nullable) + stk[-2].nfirstpos += stk[-1].nfirstpos; + else + firstpos += stk[-1].nfirstpos; + + /* The lastpos of a CAT node is the lastpos of the second argument, + union that of the first argument if the second is nullable. */ + if (stk[-1].nullable) + stk[-2].nlastpos += stk[-1].nlastpos; + else + { + pos = lastpos + stk[-2].nlastpos; + for (j = stk[-1].nlastpos; j-- > 0;) + pos[j] = lastpos[j]; + lastpos += stk[-2].nlastpos; + stk[-2].nlastpos = stk[-1].nlastpos; + } + + /* A CAT node is nullable if both arguments are nullable. */ + stk[-2].nullable &= stk[-1].nullable; + stk--; + break; + + case OR: + /* The firstpos is the union of the firstpos of each argument. */ + stk[-2].nfirstpos += stk[-1].nfirstpos; + + /* The lastpos is the union of the lastpos of each argument. */ + stk[-2].nlastpos += stk[-1].nlastpos; + + /* An OR node is nullable if either argument is nullable. */ + stk[-2].nullable |= stk[-1].nullable; + stk--; + break; + + default: + /* Anything else is a nonempty position. (Note that special + constructs like \< are treated as nonempty strings here; + an "epsilon closure" effectively makes them nullable later. + Backreferences have to get a real position so we can detect + transitions on them later. But they are nullable. */ + stk->nullable = d->tokens[i] == BACKREF; + + /* This position is in its own firstpos and lastpos. */ + stk->nfirstpos = stk->nlastpos = 1; + stk++; + + --firstpos, --lastpos; + firstpos->index = lastpos->index = i; + firstpos->constraint = lastpos->constraint = NO_CONSTRAINT; + + /* Allocate the follow set for this position. */ + alloc_position_set (&d->follows[i], 1); + break; + } +#ifdef DEBUG + /* ... balance the above nonsyntactic #ifdef goo... */ + fprintf (stderr, "node %zu:", i); + prtok (d->tokens[i]); + putc ('\n', stderr); + fprintf (stderr, + stk[-1].nullable ? " nullable: yes\n" : " nullable: no\n"); + fprintf (stderr, " firstpos:"); + for (j = stk[-1].nfirstpos; j-- > 0;) + { + fprintf (stderr, " %zu:", firstpos[j].index); + prtok (d->tokens[firstpos[j].index]); + } + fprintf (stderr, "\n lastpos:"); + for (j = stk[-1].nlastpos; j-- > 0;) + { + fprintf (stderr, " %zu:", lastpos[j].index); + prtok (d->tokens[lastpos[j].index]); + } + putc ('\n', stderr); +#endif + } + +#ifdef DEBUG + for (i = 0; i < d->tindex; ++i) + if (d->tokens[i] < NOTCHAR || d->tokens[i] == BACKREF + || d->tokens[i] == ANYCHAR || d->tokens[i] == MBCSET + || d->tokens[i] >= CSET) + { + fprintf (stderr, "follows(%zu:", i); + prtok (d->tokens[i]); + fprintf (stderr, "):"); + for (j = d->follows[i].nelem; j-- > 0;) + { + fprintf (stderr, " %zu:", d->follows[i].elems[j].index); + prtok (d->tokens[d->follows[i].elems[j].index]); + } + putc ('\n', stderr); + } +#endif + + /* Get the epsilon closure of the firstpos of the regexp. The result will + be the set of positions of state 0. */ + merged.nelem = 0; + for (i = 0; i < stk[-1].nfirstpos; ++i) + insert (firstpos[i], &merged); + + /* For each follow set that is the follow set of a real position, replace + it with its epsilon closure. */ + epsclosure (&merged, d); + + /* Build the initial state. */ + separate_contexts = state_separate_contexts (&merged); + if (separate_contexts & CTX_NEWLINE) + state_index (d, &merged, CTX_NEWLINE); + d->initstate_notbol = d->min_trcount + = state_index (d, &merged, separate_contexts ^ CTX_ANY); + if (separate_contexts & CTX_LETTER) + d->min_trcount = state_index (d, &merged, CTX_LETTER); + d->min_trcount++; + d->trcount = 0; + + free (posalloc); + free (stkalloc); + free (merged.elems); +} + + +/* Return the transition out of state s of d for the input character uc, + updating the slots in trans accordingly. + + Do not worry about all possible input characters; calculate just the group + of positions that match uc. Label it with the set of characters that + every position in the group matches (taking into account, if necessary, + preceding context information of s). Then find the union + of these positions' follows, i.e., the set of positions of the + new state. For each character in the group's label, set the transition + on this character to be to a state corresponding to the set's positions, + and its associated backward context information, if necessary. + + When building a searching matcher, include the positions of state + 0 in every state. + + The group is constructed by building an equivalence-class + partition of the positions of s. + + For each position, find the set of characters C that it matches. Eliminate + any characters from C that fail on grounds of backward context. + + Check whether the group's label L has nonempty + intersection with C. If L - C is nonempty, create a new group labeled + L - C and having the same positions as the current group, and set L to + the intersection of L and C. Insert the position in the group, set + C = C - L, and resume scanning. + + If after comparing with every group there are characters remaining in C, + create a new group labeled with the characters of C and insert this + position in that group. */ +static state_num +dfastate (state_num s, struct dfa *d, unsigned char uc, state_num trans[]) +{ + leaf_set group; /* Positions that match the input char. */ + charclass label; /* The group's label. */ + position_set follows; /* Union of the follows of the group. */ + position_set tmp; /* Temporary space for merging sets. */ + state_num state; /* New state. */ + state_num state_newline; /* New state on a newline transition. */ + state_num state_letter; /* New state on a letter transition. */ + size_t i, j, k; + +#ifdef DEBUG + fprintf (stderr, "build state %td\n", s); +#endif + + group.elems = xnmalloc (d->nleaves, sizeof *group.elems); + group.nelem = 0; + + fillset (label); + + for (i = 0; i < d->states[s].elems.nelem; ++i) + { + charclass matches; /* Set of matching characters. */ + position pos = d->states[s].elems.elems[i]; + bool matched = false; + if (d->tokens[pos.index] >= 0 && d->tokens[pos.index] < NOTCHAR) + { + zeroset (matches); + setbit (d->tokens[pos.index], matches); + if (d->tokens[pos.index] == uc) + matched = true; + } + else if (d->tokens[pos.index] >= CSET) + { + copyset (d->charclasses[d->tokens[pos.index] - CSET], matches); + if (tstbit (uc, d->charclasses[d->tokens[pos.index] - CSET])) + matched = true; + } + else if (d->tokens[pos.index] == ANYCHAR) + { + copyset (d->charclasses[d->canychar], matches); + if (tstbit (uc, d->charclasses[d->canychar])) + matched = true; + + /* ANYCHAR must match with a single character, so we must put + it to D->states[s].mbps which contains the positions which + can match with a single character not a byte. If all + positions which has ANYCHAR does not depend on context of + next character, we put the follows instead of it to + D->states[s].mbps to optimize. */ + if (SUCCEEDS_IN_CONTEXT (pos.constraint, d->states[s].context, + CTX_NONE)) + { + if (d->states[s].mbps.nelem == 0) + alloc_position_set (&d->states[s].mbps, + d->follows[pos.index].nelem); + for (j = 0; j < d->follows[pos.index].nelem; j++) + insert (d->follows[pos.index].elems[j], &d->states[s].mbps); + } + } + else + continue; + + /* Some characters may need to be eliminated from matches because + they fail in the current context. */ + if (pos.constraint != NO_CONSTRAINT) + { + if (!SUCCEEDS_IN_CONTEXT (pos.constraint, + d->states[s].context, CTX_NEWLINE)) + for (j = 0; j < CHARCLASS_WORDS; ++j) + matches[j] &= ~d->syntax.newline[j]; + if (!SUCCEEDS_IN_CONTEXT (pos.constraint, + d->states[s].context, CTX_LETTER)) + for (j = 0; j < CHARCLASS_WORDS; ++j) + matches[j] &= ~d->syntax.letters[j]; + if (!SUCCEEDS_IN_CONTEXT (pos.constraint, + d->states[s].context, CTX_NONE)) + for (j = 0; j < CHARCLASS_WORDS; ++j) + matches[j] &= d->syntax.letters[j] | d->syntax.newline[j]; + + /* If there are no characters left, there's no point in going on. */ + for (j = 0; j < CHARCLASS_WORDS && !matches[j]; ++j) + continue; + if (j == CHARCLASS_WORDS) + continue; + + /* If we have reset the bit that made us declare "matched", reset + that indicator, too. This is required to avoid an infinite loop + with this command: echo cx | LC_ALL=C grep -E 'c\b[x ]' */ + if (!tstbit (uc, matches)) + matched = false; + } + +#ifdef DEBUG + fprintf (stderr, " nextpos %zu:", pos.index); + prtok (d->tokens[pos.index]); + fprintf (stderr, " of"); + for (j = 0; j < NOTCHAR; j++) + if (tstbit (j, matches)) + fprintf (stderr, " 0x%02zx", j); + fprintf (stderr, "\n"); +#endif + + if (matched) + { + for (k = 0; k < CHARCLASS_WORDS; ++k) + label[k] &= matches[k]; + group.elems[group.nelem++] = pos.index; + } + else + { + for (k = 0; k < CHARCLASS_WORDS; ++k) + label[k] &= ~matches[k]; + } + } + + alloc_position_set (&follows, d->nleaves); + alloc_position_set (&tmp, d->nleaves); + + if (group.nelem > 0) + { + int possible_contexts; /* Contexts that the group can match. */ + int separate_contexts; /* Context that new state wants to know. */ + + follows.nelem = 0; + + /* Find the union of the follows of the positions of the group. + This is a hideously inefficient loop. Fix it someday. */ + for (j = 0; j < group.nelem; ++j) + for (k = 0; k < d->follows[group.elems[j]].nelem; ++k) + insert (d->follows[group.elems[j]].elems[k], &follows); + + /* If we are building a searching matcher, throw in the positions + of state 0 as well, if possible. */ + if (d->searchflag) + { + /* If a token in follows.elems is not 1st byte of a multibyte + character, or the states of follows must accept the bytes + which are not 1st byte of the multibyte character. + Then, if a state of follows encounters a byte, it must not be + a 1st byte of a multibyte character nor a single byte character. + In this case, do not add state[0].follows to next state, because + state[0] must accept 1st-byte. + + For example, suppose <sb a> is a certain single byte character, + <mb A> is a certain multibyte character, and the codepoint of + <sb a> equals the 2nd byte of the codepoint of <mb A>. When + state[0] accepts <sb a>, state[i] transits to state[i+1] by + accepting the 1st byte of <mb A>, and state[i+1] accepts the + 2nd byte of <mb A>, if state[i+1] encounters the codepoint of + <sb a>, it must not be <sb a> but the 2nd byte of <mb A>, so do + not add state[0]. */ + + bool mergeit = !d->localeinfo.multibyte; + if (!mergeit) + for (mergeit = true, j = 0; mergeit && j < follows.nelem; j++) + mergeit &= d->multibyte_prop[follows.elems[j].index]; + if (mergeit) + { + merge (&d->states[0].elems, &follows, &tmp); + copy (&tmp, &follows); + } + } + + /* Find out if the new state will want any context information. */ + possible_contexts = charclass_context (d, label); + separate_contexts = state_separate_contexts (&follows); + + /* Find the state(s) corresponding to the union of the follows. */ + if (possible_contexts & ~separate_contexts) + state = state_index (d, &follows, separate_contexts ^ CTX_ANY); + else + state = -1; + if (separate_contexts & possible_contexts & CTX_NEWLINE) + state_newline = state_index (d, &follows, CTX_NEWLINE); + else + state_newline = state; + if (separate_contexts & possible_contexts & CTX_LETTER) + state_letter = state_index (d, &follows, CTX_LETTER); + else + state_letter = state; + } + + /* If we are a searching matcher, the default transition is to a state + containing the positions of state 0, otherwise the default transition + is to fail miserably. */ + else if (d->searchflag) + { + state_newline = 0; + state_letter = d->min_trcount - 1; + state = d->initstate_notbol; + } + else + { + state_newline = -1; + state_letter = -1; + state = -1; + } + + /* Set the transitions for each character in the label. */ + for (i = 0; i < NOTCHAR; i++) + if (tstbit (i, label)) + switch (d->syntax.sbit[i]) + { + case CTX_NEWLINE: + trans[i] = state_newline; + break; + case CTX_LETTER: + trans[i] = state_letter; + break; + default: + trans[i] = state; + break; + } + +#ifdef DEBUG + fprintf (stderr, "trans table %td", s); + for (i = 0; i < NOTCHAR; ++i) + { + if (!(i & 0xf)) + fprintf (stderr, "\n"); + fprintf (stderr, " %2td", trans[i]); + } + fprintf (stderr, "\n"); +#endif + + free (group.elems); + free (follows.elems); + free (tmp.elems); + + /* Keep the newline transition in a special place so we can use it as + a sentinel. */ + if (tstbit (d->syntax.eolbyte, label)) + { + d->newlines[s] = trans[d->syntax.eolbyte]; + trans[d->syntax.eolbyte] = -1; + } + + return trans[uc]; +} + +/* Make sure D's state arrays are large enough to hold NEW_STATE. */ +static void +realloc_trans_if_necessary (struct dfa *d, state_num new_state) +{ + state_num oldalloc = d->tralloc; + if (oldalloc <= new_state) + { + state_num **realtrans = d->trans ? d->trans - 2 : NULL; + ptrdiff_t newalloc, newalloc1; + newalloc1 = realtrans ? d->tralloc + 2 : 0; + realtrans = xpalloc (realtrans, &newalloc1, new_state - oldalloc + 1, + -1, sizeof *realtrans); + realtrans[0] = realtrans[1] = NULL; + d->trans = realtrans + 2; + d->tralloc = newalloc = newalloc1 - 2; + d->fails = xnrealloc (d->fails, newalloc, sizeof *d->fails); + d->success = xnrealloc (d->success, newalloc, sizeof *d->success); + d->newlines = xnrealloc (d->newlines, newalloc, sizeof *d->newlines); + if (d->localeinfo.multibyte) + { + realtrans = d->mb_trans ? d->mb_trans - 2 : NULL; + realtrans = xnrealloc (realtrans, newalloc1, sizeof *realtrans); + if (oldalloc == 0) + realtrans[0] = realtrans[1] = NULL; + d->mb_trans = realtrans + 2; + } + for (; oldalloc < newalloc; oldalloc++) + { + d->trans[oldalloc] = NULL; + d->fails[oldalloc] = NULL; + if (d->localeinfo.multibyte) + d->mb_trans[oldalloc] = NULL; + } + } +} + +/* Calculate the transition table for a new state derived from state s + for a compiled dfa d after input character uc, and return the new + state number. */ + +static state_num +build_state (state_num s, struct dfa *d, unsigned char uc) +{ + /* A pointer to the new transition table, and the table itself. */ + state_num **ptrans = (ACCEPTING (s, *d) ? d->fails : d->trans) + s; + state_num *trans = *ptrans; + + if (!trans) + { + /* MAX_TRCOUNT is an arbitrary upper limit on the number of + transition tables that can exist at once, other than for + initial states. Often-used transition tables are quickly + rebuilt, whereas rarely-used ones are cleared away. */ + if (MAX_TRCOUNT <= d->trcount) + { + for (state_num i = d->min_trcount; i < d->tralloc; i++) + { + free (d->trans[i]); + free (d->fails[i]); + d->trans[i] = d->fails[i] = NULL; + } + d->trcount = 0; + } + + d->trcount++; + *ptrans = trans = xmalloc (NOTCHAR * sizeof *trans); + + /* Fill transition table with a default value which means that the + transited state has not been calculated yet. */ + for (int i = 0; i < NOTCHAR; i++) + trans[i] = -2; + } + + /* Set up the success bits for this state. */ + d->success[s] = 0; + if (ACCEPTS_IN_CONTEXT (d->states[s].context, CTX_NEWLINE, s, *d)) + d->success[s] |= CTX_NEWLINE; + if (ACCEPTS_IN_CONTEXT (d->states[s].context, CTX_LETTER, s, *d)) + d->success[s] |= CTX_LETTER; + if (ACCEPTS_IN_CONTEXT (d->states[s].context, CTX_NONE, s, *d)) + d->success[s] |= CTX_NONE; + + s = dfastate (s, d, uc, trans); + + /* Now go through the new transition table, and make sure that the trans + and fail arrays are allocated large enough to hold a pointer for the + largest state mentioned in the table. */ + state_num maxstate = -1; + for (int i = 0; i < NOTCHAR; i++) + if (maxstate < trans[i]) + maxstate = trans[i]; + realloc_trans_if_necessary (d, maxstate); + + return s; +} + +/* Multibyte character handling sub-routines for dfaexec. */ + +/* Consume a single byte and transit state from 's' to '*next_state'. + This function is almost same as the state transition routin in dfaexec. + But state transition is done just once, otherwise matching succeed or + reach the end of the buffer. */ +static state_num +transit_state_singlebyte (struct dfa *d, state_num s, unsigned char const **pp) +{ + state_num *t; + + if (d->trans[s]) + t = d->trans[s]; + else if (d->fails[s]) + t = d->fails[s]; + else + { + build_state (s, d, **pp); + if (d->trans[s]) + t = d->trans[s]; + else + { + t = d->fails[s]; + assert (t); + } + } + + if (t[**pp] == -2) + build_state (s, d, **pp); + + return t[*(*pp)++]; +} + +/* Transit state from s, then return new state and update the pointer of + the buffer. This function is for a period operator which can match a + multi-byte character. */ +static state_num +transit_state (struct dfa *d, state_num s, unsigned char const **pp, + unsigned char const *end) +{ + state_num s1, s2; + wint_t wc; + int separate_contexts; + size_t i; + + int mbclen = mbs_to_wchar (&wc, (char const *) *pp, end - *pp, d); + + /* This state has some operators which can match a multibyte character. */ + d->mb_follows.nelem = 0; + + /* Calculate the state which can be reached from the state 's' by + consuming 'mbclen' single bytes from the buffer. */ + s1 = s; + for (i = 0; i < mbclen && (i == 0 || d->min_trcount <= s); i++) + s = transit_state_singlebyte (d, s, pp); + *pp += mbclen - i; + + if (wc == WEOF) + { + /* It is an invalid character, so ANYCHAR is not accepted. */ + return s; + } + + /* If all positions which have ANYCHAR do not depend on the context + of the next character, calculate the next state with + pre-calculated follows and cache the result. */ + if (d->states[s1].mb_trindex < 0) + { + if (MAX_TRCOUNT <= d->mb_trcount) + { + state_num s3; + for (s3 = -1; s3 < d->tralloc; s3++) + { + free (d->mb_trans[s3]); + d->mb_trans[s3] = NULL; + } + + for (i = 0; i < d->sindex; i++) + d->states[i].mb_trindex = -1; + d->mb_trcount = 0; + } + d->states[s1].mb_trindex = d->mb_trcount++; + } + + if (! d->mb_trans[s]) + { + enum { TRANSPTR_SIZE = sizeof *d->mb_trans[s] }; + enum { TRANSALLOC_SIZE = MAX_TRCOUNT * TRANSPTR_SIZE }; + d->mb_trans[s] = xmalloc (TRANSALLOC_SIZE); + for (i = 0; i < MAX_TRCOUNT; i++) + d->mb_trans[s][i] = -1; + } + else if (d->mb_trans[s][d->states[s1].mb_trindex] >= 0) + return d->mb_trans[s][d->states[s1].mb_trindex]; + + if (s == -1) + copy (&d->states[s1].mbps, &d->mb_follows); + else + merge (&d->states[s1].mbps, &d->states[s].elems, &d->mb_follows); + + separate_contexts = state_separate_contexts (&d->mb_follows); + s2 = state_index (d, &d->mb_follows, separate_contexts ^ CTX_ANY); + realloc_trans_if_necessary (d, s2); + + d->mb_trans[s][d->states[s1].mb_trindex] = s2; + + return s2; +} + +/* The initial state may encounter a byte which is not a single byte character + nor the first byte of a multibyte character. But it is incorrect for the + initial state to accept such a byte. For example, in Shift JIS the regular + expression "\\" accepts the codepoint 0x5c, but should not accept the second + byte of the codepoint 0x815c. Then the initial state must skip the bytes + that are not a single byte character nor the first byte of a multibyte + character. + + Given DFA state d, use mbs_to_wchar to advance MBP until it reaches + or exceeds P, and return the advanced MBP. If WCP is non-NULL and + the result is greater than P, set *WCP to the final wide character + processed, or to WEOF if no wide character is processed. Otherwise, + if WCP is non-NULL, *WCP may or may not be updated. + + Both P and MBP must be no larger than END. */ +static unsigned char const * +skip_remains_mb (struct dfa *d, unsigned char const *p, + unsigned char const *mbp, char const *end) +{ + wint_t wc; + if (d->syntax.never_trail[*p]) + return p; + while (mbp < p) + mbp += mbs_to_wchar (&wc, (char const *) mbp, + end - (char const *) mbp, d); + return mbp; +} + +/* Search through a buffer looking for a match to the struct dfa *D. + Find the first occurrence of a string matching the regexp in the + buffer, and the shortest possible version thereof. Return a pointer to + the first character after the match, or NULL if none is found. BEGIN + points to the beginning of the buffer, and END points to the first byte + after its end. Note however that we store a sentinel byte (usually + newline) in *END, so the actual buffer must be one byte longer. + When ALLOW_NL, newlines may appear in the matching string. + If COUNT is non-NULL, increment *COUNT once for each newline processed. + If MULTIBYTE, the input consists of multibyte characters and/or + encoding-error bytes. Otherwise, it consists of single-byte characters. + Here is the list of features that make this DFA matcher punt: + - [M-N] range in non-simple locale: regex is up to 25% faster on [a-z] + - [^...] in non-simple locale + - [[=foo=]] or [[.foo.]] + - [[:alpha:]] etc. in multibyte locale (except [[:digit:]] works OK) + - back-reference: (.)\1 + - word-delimiter in multibyte locale: \<, \>, \b, \B + See using_simple_locale for the definition of "simple locale". */ + +static inline char * +dfaexec_main (struct dfa *d, char const *begin, char *end, bool allow_nl, + size_t *count, bool multibyte) +{ + state_num s, s1; /* Current state. */ + unsigned char const *p, *mbp; /* Current input character. */ + state_num **trans, *t; /* Copy of d->trans so it can be optimized + into a register. */ + unsigned char eol = d->syntax.eolbyte; /* Likewise for eolbyte. */ + unsigned char saved_end; + size_t nlcount = 0; + + if (MAX_TRCOUNT <= d->sindex) + { + for (s = d->min_trcount; s < d->sindex; s++) + { + free (d->states[s].elems.elems); + free (d->states[s].mbps.elems); + } + d->sindex = d->min_trcount; + + if (d->trans) + { + for (s = 0; s < d->tralloc; s++) + { + free (d->trans[s]); + free (d->fails[s]); + d->trans[s] = d->fails[s] = NULL; + } + d->trcount = 0; + } + + if (d->localeinfo.multibyte && d->mb_trans) + { + for (s = -1; s < d->tralloc; s++) + { + free (d->mb_trans[s]); + d->mb_trans[s] = NULL; + } + for (s = 0; s < d->min_trcount; s++) + d->states[s].mb_trindex = -1; + d->mb_trcount = 0; + } + } + + if (!d->tralloc) + realloc_trans_if_necessary (d, 0); + + s = s1 = 0; + p = mbp = (unsigned char const *) begin; + trans = d->trans; + saved_end = *(unsigned char *) end; + *end = eol; + + if (multibyte) + { + memset (&d->mbs, 0, sizeof d->mbs); + if (d->mb_follows.alloc == 0) + alloc_position_set (&d->mb_follows, d->nleaves); + } + + for (;;) + { + while ((t = trans[s]) != NULL) + { + if (s < d->min_trcount) + { + if (!multibyte || d->states[s].mbps.nelem == 0) + { + while (t[*p] == s) + p++; + } + if (multibyte) + p = mbp = skip_remains_mb (d, p, mbp, end); + } + + if (multibyte) + { + s1 = s; + + if (d->states[s].mbps.nelem == 0 + || d->localeinfo.sbctowc[*p] != WEOF || (char *) p >= end) + { + /* If an input character does not match ANYCHAR, do it + like a single-byte character. */ + s = t[*p++]; + } + else + { + s = transit_state (d, s, &p, (unsigned char *) end); + mbp = p; + trans = d->trans; + } + } + else + { + s1 = t[*p++]; + t = trans[s1]; + if (! t) + { + state_num tmp = s; + s = s1; + s1 = tmp; /* swap */ + break; + } + if (s < d->min_trcount) + { + while (t[*p] == s1) + p++; + } + s = t[*p++]; + } + } + + if (s < 0) + { + if (s == -2) + { + s = build_state (s1, d, p[-1]); + trans = d->trans; + } + else if ((char *) p <= end && p[-1] == eol && 0 <= d->newlines[s1]) + { + /* The previous character was a newline. Count it, and skip + checking of multibyte character boundary until here. */ + nlcount++; + mbp = p; + + s = (allow_nl ? d->newlines[s1] + : d->syntax.sbit[eol] == CTX_NEWLINE ? 0 + : d->syntax.sbit[eol] == CTX_LETTER ? d->min_trcount - 1 + : d->initstate_notbol); + } + else + { + p = NULL; + goto done; + } + } + else if (d->fails[s]) + { + if ((d->success[s] & d->syntax.sbit[*p]) + || ((char *) p == end + && ACCEPTS_IN_CONTEXT (d->states[s].context, CTX_NEWLINE, s, + *d))) + goto done; + + if (multibyte && s < d->min_trcount) + p = mbp = skip_remains_mb (d, p, mbp, end); + + s1 = s; + if (!multibyte || d->states[s].mbps.nelem == 0 + || d->localeinfo.sbctowc[*p] != WEOF || (char *) p >= end) + { + /* If a input character does not match ANYCHAR, do it + like a single-byte character. */ + s = d->fails[s][*p++]; + } + else + { + s = transit_state (d, s, &p, (unsigned char *) end); + mbp = p; + trans = d->trans; + } + } + else + { + build_state (s, d, p[0]); + trans = d->trans; + } + } + + done: + if (count) + *count += nlcount; + *end = saved_end; + return (char *) p; +} + +/* Specialized versions of dfaexec for multibyte and single-byte cases. + This is for performance, as dfaexec_main is an inline function. */ + +static char * +dfaexec_mb (struct dfa *d, char const *begin, char *end, + bool allow_nl, size_t *count, bool *backref) +{ + return dfaexec_main (d, begin, end, allow_nl, count, true); +} + +static char * +dfaexec_sb (struct dfa *d, char const *begin, char *end, + bool allow_nl, size_t *count, bool *backref) +{ + return dfaexec_main (d, begin, end, allow_nl, count, false); +} + +/* Always set *BACKREF and return BEGIN. Use this wrapper for + any regexp that uses a construct not supported by this code. */ +static char * +dfaexec_noop (struct dfa *d, char const *begin, char *end, + bool allow_nl, size_t *count, bool *backref) +{ + *backref = true; + return (char *) begin; +} + +/* Like dfaexec_main (D, BEGIN, END, ALLOW_NL, COUNT, D->localeinfo.multibyte), + but faster and set *BACKREF if the DFA code does not support this + regexp usage. */ + +char * +dfaexec (struct dfa *d, char const *begin, char *end, + bool allow_nl, size_t *count, bool *backref) +{ + return d->dfaexec (d, begin, end, allow_nl, count, backref); +} + +struct dfa * +dfasuperset (struct dfa const *d) +{ + return d->superset; +} + +bool +dfaisfast (struct dfa const *d) +{ + return d->fast; +} + +static void +free_mbdata (struct dfa *d) +{ + ptrdiff_t i; + + free (d->multibyte_prop); + + for (i = 0; i < d->nmbcsets; ++i) + free (d->mbcsets[i].chars); + + free (d->mbcsets); + free (d->mb_follows.elems); + + if (d->mb_trans) + { + state_num s; + for (s = -1; s < d->tralloc; s++) + free (d->mb_trans[s]); + free (d->mb_trans - 2); + } +} + +/* Return true if every construct in D is supported by this DFA matcher. */ +static bool _GL_ATTRIBUTE_PURE +dfa_supported (struct dfa const *d) +{ + size_t i; + for (i = 0; i < d->tindex; i++) + { + switch (d->tokens[i]) + { + case BEGWORD: + case ENDWORD: + case LIMWORD: + case NOTLIMWORD: + if (!d->localeinfo.multibyte) + continue; + /* fallthrough */ + + case BACKREF: + case MBCSET: + return false; + } + } + return true; +} + +static void +dfaoptimize (struct dfa *d) +{ + size_t i; + bool have_backref = false; + + if (!d->localeinfo.using_utf8) + return; + + for (i = 0; i < d->tindex; ++i) + { + switch (d->tokens[i]) + { + case ANYCHAR: + /* Lowered. */ + abort (); + case BACKREF: + have_backref = true; + break; + case MBCSET: + /* Requires multi-byte algorithm. */ + return; + default: + break; + } + } + + if (!have_backref && d->superset) + { + /* The superset DFA is not likely to be much faster, so remove it. */ + dfafree (d->superset); + free (d->superset); + d->superset = NULL; + } + + free_mbdata (d); + d->localeinfo.multibyte = false; + d->dfaexec = dfaexec_sb; + d->fast = true; +} + +static void +dfassbuild (struct dfa *d) +{ + size_t i, j; + bool have_achar = false; + bool have_nchar = false; + struct dfa *sup = dfaalloc (); + + *sup = *d; + sup->localeinfo.multibyte = false; + sup->dfaexec = dfaexec_sb; + sup->multibyte_prop = NULL; + sup->mbcsets = NULL; + sup->superset = NULL; + sup->states = NULL; + sup->sindex = 0; + sup->follows = NULL; + sup->tralloc = 0; + sup->trans = NULL; + sup->fails = NULL; + sup->success = NULL; + sup->newlines = NULL; + + sup->charclasses = xnmalloc (sup->calloc, sizeof *sup->charclasses); + if (d->cindex) + { + memcpy (sup->charclasses, d->charclasses, + d->cindex * sizeof *sup->charclasses); + } + + sup->tokens = xnmalloc (d->tindex, 2 * sizeof *sup->tokens); + sup->talloc = d->tindex * 2; + + for (i = j = 0; i < d->tindex; i++) + { + switch (d->tokens[i]) + { + case ANYCHAR: + case MBCSET: + case BACKREF: + { + charclass ccl; + fillset (ccl); + sup->tokens[j++] = CSET + charclass_index (sup, ccl); + sup->tokens[j++] = STAR; + if (d->tokens[i + 1] == QMARK || d->tokens[i + 1] == STAR + || d->tokens[i + 1] == PLUS) + i++; + have_achar = true; + } + break; + case BEGWORD: + case ENDWORD: + case LIMWORD: + case NOTLIMWORD: + if (d->localeinfo.multibyte) + { + /* These constraints aren't supported in a multibyte locale. + Ignore them in the superset DFA. */ + sup->tokens[j++] = EMPTY; + break; + } + /* fallthrough */ + default: + sup->tokens[j++] = d->tokens[i]; + if ((0 <= d->tokens[i] && d->tokens[i] < NOTCHAR) + || d->tokens[i] >= CSET) + have_nchar = true; + break; + } + } + sup->tindex = j; + + if (have_nchar && (have_achar || d->localeinfo.multibyte)) + d->superset = sup; + else + { + dfafree (sup); + free (sup); + } +} + +/* Parse and analyze a single string of the given length. */ +void +dfacomp (char const *s, size_t len, struct dfa *d, bool searchflag) +{ + dfaparse (s, len, d); + dfassbuild (d); + + if (dfa_supported (d)) + { + dfaoptimize (d); + dfaanalyze (d, searchflag); + } + else + { + d->dfaexec = dfaexec_noop; + } + + if (d->superset) + { + d->fast = true; + dfaanalyze (d->superset, searchflag); + } +} + +/* Free the storage held by the components of a dfa. */ +void +dfafree (struct dfa *d) +{ + size_t i; + + free (d->charclasses); + free (d->tokens); + + if (d->localeinfo.multibyte) + free_mbdata (d); + + for (i = 0; i < d->sindex; ++i) + { + free (d->states[i].elems.elems); + free (d->states[i].mbps.elems); + } + free (d->states); + + if (d->follows) + { + for (i = 0; i < d->tindex; ++i) + free (d->follows[i].elems); + free (d->follows); + } + + if (d->trans) + { + for (i = 0; i < d->tralloc; ++i) + { + free (d->trans[i]); + free (d->fails[i]); + } + + free (d->trans - 2); + free (d->fails); + free (d->newlines); + free (d->success); + } + + if (d->superset) + dfafree (d->superset); +} + +/* Having found the postfix representation of the regular expression, + try to find a long sequence of characters that must appear in any line + containing the r.e. + Finding a "longest" sequence is beyond the scope here; + we take an easy way out and hope for the best. + (Take "(ab|a)b"--please.) + + We do a bottom-up calculation of sequences of characters that must appear + in matches of r.e.'s represented by trees rooted at the nodes of the postfix + representation: + sequences that must appear at the left of the match ("left") + sequences that must appear at the right of the match ("right") + lists of sequences that must appear somewhere in the match ("in") + sequences that must constitute the match ("is") + + When we get to the root of the tree, we use one of the longest of its + calculated "in" sequences as our answer. + + The sequences calculated for the various types of node (in pseudo ANSI c) + are shown below. "p" is the operand of unary operators (and the left-hand + operand of binary operators); "q" is the right-hand operand of binary + operators. + + "ZERO" means "a zero-length sequence" below. + + Type left right is in + ---- ---- ----- -- -- + char c # c # c # c # c + + ANYCHAR ZERO ZERO ZERO ZERO + + MBCSET ZERO ZERO ZERO ZERO + + CSET ZERO ZERO ZERO ZERO + + STAR ZERO ZERO ZERO ZERO + + QMARK ZERO ZERO ZERO ZERO + + PLUS p->left p->right ZERO p->in + + CAT (p->is==ZERO)? (q->is==ZERO)? (p->is!=ZERO && p->in plus + p->left : q->right : q->is!=ZERO) ? q->in plus + p->is##q->left p->right##q->is p->is##q->is : p->right##q->left + ZERO + + OR longest common longest common (do p->is and substrings common + leading trailing to q->is have same p->in and + (sub)sequence (sub)sequence q->in length and content) ? + of p->left of p->right + and q->left and q->right p->is : NULL + + If there's anything else we recognize in the tree, all four sequences get set + to zero-length sequences. If there's something we don't recognize in the + tree, we just return a zero-length sequence. + + Break ties in favor of infrequent letters (choosing 'zzz' in preference to + 'aaa')? + + And ... is it here or someplace that we might ponder "optimizations" such as + egrep 'psi|epsilon' -> egrep 'psi' + egrep 'pepsi|epsilon' -> egrep 'epsi' + (Yes, we now find "epsi" as a "string + that must occur", but we might also + simplify the *entire* r.e. being sought) + grep '[c]' -> grep 'c' + grep '(ab|a)b' -> grep 'ab' + grep 'ab*' -> grep 'a' + grep 'a*b' -> grep 'b' + + There are several issues: + + Is optimization easy (enough)? + + Does optimization actually accomplish anything, + or is the automaton you get from "psi|epsilon" (for example) + the same as the one you get from "psi" (for example)? + + Are optimizable r.e.'s likely to be used in real-life situations + (something like 'ab*' is probably unlikely; something like is + 'psi|epsilon' is likelier)? */ + +static char * +icatalloc (char *old, char const *new) +{ + char *result; + size_t oldsize; + size_t newsize = strlen (new); + if (newsize == 0) + return old; + oldsize = strlen (old); + result = xrealloc (old, oldsize + newsize + 1); + memcpy (result + oldsize, new, newsize + 1); + return result; +} + +static void +freelist (char **cpp) +{ + while (*cpp) + free (*cpp++); +} + +static char ** +enlist (char **cpp, char *new, size_t len) +{ + size_t i, j; + new = memcpy (xmalloc (len + 1), new, len); + new[len] = '\0'; + /* Is there already something in the list that's new (or longer)? */ + for (i = 0; cpp[i] != NULL; ++i) + if (strstr (cpp[i], new) != NULL) + { + free (new); + return cpp; + } + /* Eliminate any obsoleted strings. */ + j = 0; + while (cpp[j] != NULL) + if (strstr (new, cpp[j]) == NULL) + ++j; + else + { + free (cpp[j]); + if (--i == j) + break; + cpp[j] = cpp[i]; + cpp[i] = NULL; + } + /* Add the new string. */ + cpp = xnrealloc (cpp, i + 2, sizeof *cpp); + cpp[i] = new; + cpp[i + 1] = NULL; + return cpp; +} + +/* Given pointers to two strings, return a pointer to an allocated + list of their distinct common substrings. */ +static char ** +comsubs (char *left, char const *right) +{ + char **cpp = xzalloc (sizeof *cpp); + char *lcp; + + for (lcp = left; *lcp != '\0'; ++lcp) + { + size_t len = 0; + char *rcp = strchr (right, *lcp); + while (rcp != NULL) + { + size_t i; + for (i = 1; lcp[i] != '\0' && lcp[i] == rcp[i]; ++i) + continue; + if (i > len) + len = i; + rcp = strchr (rcp + 1, *lcp); + } + if (len != 0) + cpp = enlist (cpp, lcp, len); + } + return cpp; +} + +static char ** +addlists (char **old, char **new) +{ + for (; *new; new++) + old = enlist (old, *new, strlen (*new)); + return old; +} + +/* Given two lists of substrings, return a new list giving substrings + common to both. */ +static char ** +inboth (char **left, char **right) +{ + char **both = xzalloc (sizeof *both); + size_t lnum, rnum; + + for (lnum = 0; left[lnum] != NULL; ++lnum) + { + for (rnum = 0; right[rnum] != NULL; ++rnum) + { + char **temp = comsubs (left[lnum], right[rnum]); + both = addlists (both, temp); + freelist (temp); + free (temp); + } + } + return both; +} + +typedef struct must must; + +struct must +{ + char **in; + char *left; + char *right; + char *is; + bool begline; + bool endline; + must *prev; +}; + +static must * +allocmust (must *mp, size_t size) +{ + must *new_mp = xmalloc (sizeof *new_mp); + new_mp->in = xzalloc (sizeof *new_mp->in); + new_mp->left = xzalloc (size); + new_mp->right = xzalloc (size); + new_mp->is = xzalloc (size); + new_mp->begline = false; + new_mp->endline = false; + new_mp->prev = mp; + return new_mp; +} + +static void +resetmust (must *mp) +{ + freelist (mp->in); + mp->in[0] = NULL; + mp->left[0] = mp->right[0] = mp->is[0] = '\0'; + mp->begline = false; + mp->endline = false; +} + +static void +freemust (must *mp) +{ + freelist (mp->in); + free (mp->in); + free (mp->left); + free (mp->right); + free (mp->is); + free (mp); +} + +struct dfamust * +dfamust (struct dfa const *d) +{ + must *mp = NULL; + char const *result = ""; + size_t i, ri; + bool exact = false; + bool begline = false; + bool endline = false; + size_t rj; + bool need_begline = false; + bool need_endline = false; + bool case_fold_unibyte = d->syntax.case_fold && MB_CUR_MAX == 1; + struct dfamust *dm; + + for (ri = 0; ri < d->tindex; ++ri) + { + token t = d->tokens[ri]; + switch (t) + { + case BEGLINE: + mp = allocmust (mp, 2); + mp->begline = true; + need_begline = true; + break; + case ENDLINE: + mp = allocmust (mp, 2); + mp->endline = true; + need_endline = true; + break; + case LPAREN: + case RPAREN: + assert (!"neither LPAREN nor RPAREN may appear here"); + + case EMPTY: + case BEGWORD: + case ENDWORD: + case LIMWORD: + case NOTLIMWORD: + case BACKREF: + case ANYCHAR: + case MBCSET: + mp = allocmust (mp, 2); + break; + + case STAR: + case QMARK: + resetmust (mp); + break; + + case OR: + { + char **new; + must *rmp = mp; + must *lmp = mp = mp->prev; + size_t j, ln, rn, n; + + /* Guaranteed to be. Unlikely, but ... */ + if (STREQ (lmp->is, rmp->is)) + { + lmp->begline &= rmp->begline; + lmp->endline &= rmp->endline; + } + else + { + lmp->is[0] = '\0'; + lmp->begline = false; + lmp->endline = false; + } + /* Left side--easy */ + i = 0; + while (lmp->left[i] != '\0' && lmp->left[i] == rmp->left[i]) + ++i; + lmp->left[i] = '\0'; + /* Right side */ + ln = strlen (lmp->right); + rn = strlen (rmp->right); + n = ln; + if (n > rn) + n = rn; + for (i = 0; i < n; ++i) + if (lmp->right[ln - i - 1] != rmp->right[rn - i - 1]) + break; + for (j = 0; j < i; ++j) + lmp->right[j] = lmp->right[(ln - i) + j]; + lmp->right[j] = '\0'; + new = inboth (lmp->in, rmp->in); + freelist (lmp->in); + free (lmp->in); + lmp->in = new; + freemust (rmp); + } + break; + + case PLUS: + mp->is[0] = '\0'; + break; + + case END: + assert (!mp->prev); + for (i = 0; mp->in[i] != NULL; ++i) + if (strlen (mp->in[i]) > strlen (result)) + result = mp->in[i]; + if (STREQ (result, mp->is)) + { + if ((!need_begline || mp->begline) && (!need_endline + || mp->endline)) + exact = true; + begline = mp->begline; + endline = mp->endline; + } + goto done; + + case CAT: + { + must *rmp = mp; + must *lmp = mp = mp->prev; + + /* In. Everything in left, plus everything in + right, plus concatenation of + left's right and right's left. */ + lmp->in = addlists (lmp->in, rmp->in); + if (lmp->right[0] != '\0' && rmp->left[0] != '\0') + { + size_t lrlen = strlen (lmp->right); + size_t rllen = strlen (rmp->left); + char *tp = xmalloc (lrlen + rllen); + memcpy (tp, lmp->right, lrlen); + memcpy (tp + lrlen, rmp->left, rllen); + lmp->in = enlist (lmp->in, tp, lrlen + rllen); + free (tp); + } + /* Left-hand */ + if (lmp->is[0] != '\0') + lmp->left = icatalloc (lmp->left, rmp->left); + /* Right-hand */ + if (rmp->is[0] == '\0') + lmp->right[0] = '\0'; + lmp->right = icatalloc (lmp->right, rmp->right); + /* Guaranteed to be */ + if ((lmp->is[0] != '\0' || lmp->begline) + && (rmp->is[0] != '\0' || rmp->endline)) + { + lmp->is = icatalloc (lmp->is, rmp->is); + lmp->endline = rmp->endline; + } + else + { + lmp->is[0] = '\0'; + lmp->begline = false; + lmp->endline = false; + } + freemust (rmp); + } + break; + + case '\0': + /* Not on *my* shift. */ + goto done; + + default: + if (CSET <= t) + { + /* If T is a singleton, or if case-folding in a unibyte + locale and T's members all case-fold to the same char, + convert T to one of its members. Otherwise, do + nothing further with T. */ + charclass *ccl = &d->charclasses[t - CSET]; + int j; + for (j = 0; j < NOTCHAR; j++) + if (tstbit (j, *ccl)) + break; + if (! (j < NOTCHAR)) + { + mp = allocmust (mp, 2); + break; + } + t = j; + while (++j < NOTCHAR) + if (tstbit (j, *ccl) + && ! (case_fold_unibyte + && toupper (j) == toupper (t))) + break; + if (j < NOTCHAR) + { + mp = allocmust (mp, 2); + break; + } + } + + rj = ri + 2; + if (d->tokens[ri + 1] == CAT) + { + for (; rj < d->tindex - 1; rj += 2) + { + if ((rj != ri && (d->tokens[rj] <= 0 + || NOTCHAR <= d->tokens[rj])) + || d->tokens[rj + 1] != CAT) + break; + } + } + mp = allocmust (mp, ((rj - ri) >> 1) + 1); + mp->is[0] = mp->left[0] = mp->right[0] + = case_fold_unibyte ? toupper (t) : t; + + for (i = 1; ri + 2 < rj; i++) + { + ri += 2; + t = d->tokens[ri]; + mp->is[i] = mp->left[i] = mp->right[i] + = case_fold_unibyte ? toupper (t) : t; + } + mp->is[i] = mp->left[i] = mp->right[i] = '\0'; + mp->in = enlist (mp->in, mp->is, i); + break; + } + } + done:; + + dm = NULL; + if (*result) + { + dm = xmalloc (sizeof *dm); + dm->exact = exact; + dm->begline = begline; + dm->endline = endline; + dm->must = xstrdup (result); + } + + while (mp) + { + must *prev = mp->prev; + freemust (mp); + mp = prev; + } + + return dm; +} + +void +dfamustfree (struct dfamust *dm) +{ + free (dm->must); + free (dm); +} + +struct dfa * +dfaalloc (void) +{ + void *p = xmalloc (sizeof (struct dfa)); + if (p) + { + memset (p, 0, sizeof (struct dfa)); + } + return p; +} + +/* Initialize DFA. */ +void +dfasyntax (struct dfa *dfa, struct localeinfo const *linfo, + reg_syntax_t bits, int dfaopts) +{ + int i; + memset (dfa, 0, offsetof (struct dfa, dfaexec)); + dfa->dfaexec = linfo->multibyte ? dfaexec_mb : dfaexec_sb; + dfa->simple_locale = using_simple_locale (linfo->multibyte); + dfa->localeinfo = *linfo; + + dfa->fast = !dfa->localeinfo.multibyte; + + dfa->canychar = -1; + dfa->lex.cur_mb_len = 1; + dfa->syntax.syntax_bits_set = true; + dfa->syntax.case_fold = (bits & RE_ICASE) != 0; + dfa->syntax.anchor = (dfaopts & DFA_ANCHOR) != 0; + dfa->syntax.eolbyte = dfaopts & DFA_EOL_NUL ? '\0' : '\n'; + dfa->syntax.syntax_bits = bits; + + for (i = CHAR_MIN; i <= CHAR_MAX; ++i) + { + unsigned char uc = i; + + dfa->syntax.sbit[uc] = char_context (dfa, uc); + switch (dfa->syntax.sbit[uc]) + { + case CTX_LETTER: + setbit (uc, dfa->syntax.letters); + break; + case CTX_NEWLINE: + setbit (uc, dfa->syntax.newline); + break; + } + + /* POSIX requires that the five bytes in "\n\r./" (including the + terminating NUL) cannot occur inside a multibyte character. */ + dfa->syntax.never_trail[uc] = (dfa->localeinfo.using_utf8 + ? (uc & 0xc0) != 0x80 + : strchr ("\n\r./", uc) != NULL); + } +} + +/* vim:set shiftwidth=2: */ diff --git a/support/dfa.h b/support/dfa.h new file mode 100644 index 00000000..c68b4df7 --- /dev/null +++ b/support/dfa.h @@ -0,0 +1,132 @@ +/* dfa.h - declarations for GNU deterministic regexp compiler + Copyright (C) 1988, 1998, 2007, 2009-2016 Free Software Foundation, Inc. + + This program is free software; you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation; either version 3, or (at your option) + any later version. + + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program; if not, write to the Free Software + Foundation, Inc., + 51 Franklin Street - Fifth Floor, Boston, MA 02110-1301, USA */ + +/* Written June, 1988 by Mike Haertel */ + +#include <regex.h> +#ifdef HAVE_STDBOOL_H +#include <stdbool.h> +#else +#include "missing_d/gawkbool.h" +#endif /* HAVE_STDBOOL_H */ +#include <stddef.h> + +#if 3 <= __GNUC__ +# define _GL_ATTRIBUTE_MALLOC __attribute__ ((__malloc__)) +#else +# define _GL_ATTRIBUTE_MALLOC +#endif + +struct localeinfo; /* See localeinfo.h. */ + +/* Element of a list of strings, at least one of which is known to + appear in any R.E. matching the DFA. */ +struct dfamust +{ + bool exact; + bool begline; + bool endline; + char *must; +}; + +/* The dfa structure. It is completely opaque. */ +struct dfa; + +/* Entry points. */ + +/* Allocate a struct dfa. The struct dfa is completely opaque. + The returned pointer should be passed directly to free() after + calling dfafree() on it. */ +extern struct dfa *dfaalloc (void) _GL_ATTRIBUTE_MALLOC; + +/* DFA options that can be ORed together, for dfasyntax's 4th arg. */ +enum + { + /* ^ and $ match only the start and end of data, and do not match + end-of-line within data. This is always false for grep, but + possibly true for other apps. */ + DFA_ANCHOR = 1 << 0, + + /* '\0' in data is end-of-line, instead of the traditional '\n'. */ + DFA_EOL_NUL = 1 << 1 + }; + +/* Initialize or reinitialize a DFA. This must be called before + any of the routines below. The arguments are: + 1. The DFA to operate on. + 2. Information about the current locale. + 3. Syntax bits described in regex.h. + 4. Additional DFA options described above. */ +extern void dfasyntax (struct dfa *, struct localeinfo const *, + reg_syntax_t, int); + +/* Build and return the struct dfamust from the given struct dfa. */ +extern struct dfamust *dfamust (struct dfa const *); + +/* Free the storage held by the components of a struct dfamust. */ +extern void dfamustfree (struct dfamust *); + +/* Compile the given string of the given length into the given struct dfa. + Final argument is a flag specifying whether to build a searching or an + exact matcher. */ +extern void dfacomp (char const *, size_t, struct dfa *, bool); + +/* Search through a buffer looking for a match to the given struct dfa. + Find the first occurrence of a string matching the regexp in the + buffer, and the shortest possible version thereof. Return a pointer to + the first character after the match, or NULL if none is found. BEGIN + points to the beginning of the buffer, and END points to the first byte + after its end. Note however that we store a sentinel byte (usually + newline) in *END, so the actual buffer must be one byte longer. + When ALLOW_NL is true, newlines may appear in the matching string. + If COUNT is non-NULL, increment *COUNT once for each newline processed. + Finally, if BACKREF is non-NULL set *BACKREF to indicate whether we + encountered a back-reference. The caller can use this to decide + whether to fall back on a backtracking matcher. */ +extern char *dfaexec (struct dfa *d, char const *begin, char *end, + bool allow_nl, size_t *count, bool *backref); + +/* Return a superset for D. The superset matches everything that D + matches, along with some other strings (though the latter should be + rare, for efficiency reasons). Return a null pointer if no useful + superset is available. */ +extern struct dfa *dfasuperset (struct dfa const *d) _GL_ATTRIBUTE_PURE; + +/* The DFA is likely to be fast. */ +extern bool dfaisfast (struct dfa const *) _GL_ATTRIBUTE_PURE; + +/* Copy the syntax settings from one dfa instance to another. + Saves considerable computation time if compiling many regular expressions + based on the same setting. */ +extern void dfacopysyntax (struct dfa *to, const struct dfa *from); + +/* Free the storage held by the components of a struct dfa. */ +extern void dfafree (struct dfa *); + +/* Error handling. */ + +/* dfawarn() is called by the regexp routines whenever a regex is compiled + that likely doesn't do what the user wanted. It takes a single + argument, a NUL-terminated string describing the situation. The user + must supply a dfawarn. */ +extern void dfawarn (const char *); + +/* dfaerror() is called by the regexp routines whenever an error occurs. It + takes a single argument, a NUL-terminated string describing the error. + The user must supply a dfaerror. */ +extern _Noreturn void dfaerror (const char *); diff --git a/support/getopt.c b/support/getopt.c new file mode 100644 index 00000000..8bc59610 --- /dev/null +++ b/support/getopt.c @@ -0,0 +1,1293 @@ +/* Getopt for GNU. + NOTE: getopt is part of the C library, so if you don't know what + "Keep this file name-space clean" means, talk to drepper@gnu.org + before changing it! + Copyright (C) 1987-2016 Free Software Foundation, Inc. + This file is part of the GNU C Library. + + The GNU C Library is free software; you can redistribute it and/or + modify it under the terms of the GNU Lesser General Public + License as published by the Free Software Foundation; either + version 2.1 of the License, or (at your option) any later version. + + The GNU C Library is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public + License along with the GNU C Library; if not, see + <http://www.gnu.org/licenses/>. */ + +/* This tells Alpha OSF/1 not to define a getopt prototype in <stdio.h>. + Ditto for AIX 3.2 and <stdlib.h>. */ +#ifndef _NO_PROTO +# define _NO_PROTO +#endif + +#ifdef HAVE_CONFIG_H +# include <config.h> +#endif + +#include <stdio.h> + +/* Comment out all this code if we are using the GNU C Library, and are not + actually compiling the library itself. This code is part of the GNU C + Library, but also included in many other GNU distributions. Compiling + and linking in this code is a waste when using the GNU C library + (especially if it is a shared library). Rather than having every GNU + program understand `configure --with-gnu-libc' and omit the object files, + it is simpler to just do this in the source for each such file. */ + +#define GETOPT_INTERFACE_VERSION 2 +#if !defined _LIBC && defined __GLIBC__ && __GLIBC__ >= 2 +# include <gnu-versions.h> +# if _GNU_GETOPT_INTERFACE_VERSION == GETOPT_INTERFACE_VERSION +# define ELIDE_CODE +# endif +#endif + +/* !@#$%^&*() !!!!!!!! */ +#ifdef GAWK +#undef ELIDE_CODE +#endif + +#ifndef ELIDE_CODE + + +/* This needs to come after some library #include + to get __GNU_LIBRARY__ defined. */ +#if defined (__GNU_LIBRARY__) || defined (__CYGWIN__) || defined(__DJGPP__) || defined(__APPLE__) || defined(__MINGW32__) || defined(__sun) /* Illumos */ +/* Don't include stdlib.h for + * non-GNU C libraries + * non-Cygwin + * non-DJGPP + * non-MinGW + * because some of them contain conflicting prototypes for getopt. */ +# include <stdlib.h> +# include <unistd.h> +#endif /* GNU C library. */ + +#include <string.h> + +#ifdef VMS +# include <unixlib.h> +#endif + +#ifdef _LIBC +# include <libintl.h> +#else +# include "gettext.h" +# define _(msgid) gettext (msgid) +#endif + +#if defined _LIBC +# include <wchar.h> +#endif + +#ifndef attribute_hidden +# define attribute_hidden +#endif + +/* This version of `getopt' appears to the caller like standard Unix `getopt' + but it behaves differently for the user, since it allows the user + to intersperse the options with the other arguments. + + As `getopt' works, it permutes the elements of ARGV so that, + when it is done, all the options precede everything else. Thus + all application programs are extended to handle flexible argument order. + + Setting the environment variable POSIXLY_CORRECT disables permutation. + Then the behavior is completely standard. + + GNU application programs can use a third alternative mode in which + they can distinguish the relative order of options and other arguments. */ + +#include "getopt.h" +#include "getopt_int.h" + +/* For communication from `getopt' to the caller. + When `getopt' finds an option that takes an argument, + the argument value is returned here. + Also, when `ordering' is RETURN_IN_ORDER, + each non-option ARGV-element is returned here. */ + +char *optarg; + +/* Index in ARGV of the next element to be scanned. + This is used for communication to and from the caller + and for communication between successive calls to `getopt'. + + On entry to `getopt', zero means this is the first call; initialize. + + When `getopt' returns -1, this is the index of the first of the + non-option elements that the caller should itself scan. + + Otherwise, `optind' communicates from one call to the next + how much of ARGV has been scanned so far. */ + +/* 1003.2 says this must be 1 before any call. */ +int optind = 1; + +/* Callers store zero here to inhibit the error message + for unrecognized options. */ + +int opterr = 1; + +/* Set to an option character which was unrecognized. + This must be initialized on some systems to avoid linking in the + system's own getopt implementation. */ + +int optopt = '?'; + +/* Keep a global copy of all internal members of getopt_data. */ + +static struct _getopt_data getopt_data; + + +#ifndef __GNU_LIBRARY__ + +/* Avoid depending on library functions or files + whose names are inconsistent. */ + +#ifndef getenv +extern char *getenv (); +#endif + +#endif /* not __GNU_LIBRARY__ */ + +#ifdef _LIBC +/* Stored original parameters. + XXX This is no good solution. We should rather copy the args so + that we can compare them later. But we must not use malloc(3). */ +extern int __libc_argc; +extern char **__libc_argv; + +/* Bash 2.0 gives us an environment variable containing flags + indicating ARGV elements that should not be considered arguments. */ + +# ifdef USE_NONOPTION_FLAGS +/* Defined in getopt_init.c */ +extern char *__getopt_nonoption_flags; +# endif + +# ifdef USE_NONOPTION_FLAGS +# define SWAP_FLAGS(ch1, ch2) \ + if (d->__nonoption_flags_len > 0) \ + { \ + char __tmp = __getopt_nonoption_flags[ch1]; \ + __getopt_nonoption_flags[ch1] = __getopt_nonoption_flags[ch2]; \ + __getopt_nonoption_flags[ch2] = __tmp; \ + } +# else +# define SWAP_FLAGS(ch1, ch2) +# endif +#else /* !_LIBC */ +# define SWAP_FLAGS(ch1, ch2) +#endif /* _LIBC */ + +/* Exchange two adjacent subsequences of ARGV. + One subsequence is elements [first_nonopt,last_nonopt) + which contains all the non-options that have been skipped so far. + The other is elements [last_nonopt,optind), which contains all + the options processed since those non-options were skipped. + + `first_nonopt' and `last_nonopt' are relocated so that they describe + the new indices of the non-options in ARGV after they are moved. */ + +static void +exchange (char **argv, struct _getopt_data *d) +{ + int bottom = d->__first_nonopt; + int middle = d->__last_nonopt; + int top = d->optind; + char *tem; + + /* Exchange the shorter segment with the far end of the longer segment. + That puts the shorter segment into the right place. + It leaves the longer segment in the right place overall, + but it consists of two parts that need to be swapped next. */ + +#if defined _LIBC && defined USE_NONOPTION_FLAGS + /* First make sure the handling of the `__getopt_nonoption_flags' + string can work normally. Our top argument must be in the range + of the string. */ + if (d->__nonoption_flags_len > 0 && top >= d->__nonoption_flags_max_len) + { + /* We must extend the array. The user plays games with us and + presents new arguments. */ + char *new_str = malloc (top + 1); + if (new_str == NULL) + d->__nonoption_flags_len = d->__nonoption_flags_max_len = 0; + else + { + memset (__mempcpy (new_str, __getopt_nonoption_flags, + d->__nonoption_flags_max_len), + '\0', top + 1 - d->__nonoption_flags_max_len); + d->__nonoption_flags_max_len = top + 1; + __getopt_nonoption_flags = new_str; + } + } +#endif + + while (top > middle && middle > bottom) + { + if (top - middle > middle - bottom) + { + /* Bottom segment is the short one. */ + int len = middle - bottom; + int i; + + /* Swap it with the top part of the top segment. */ + for (i = 0; i < len; i++) + { + tem = argv[bottom + i]; + argv[bottom + i] = argv[top - (middle - bottom) + i]; + argv[top - (middle - bottom) + i] = tem; + SWAP_FLAGS (bottom + i, top - (middle - bottom) + i); + } + /* Exclude the moved bottom segment from further swapping. */ + top -= len; + } + else + { + /* Top segment is the short one. */ + int len = top - middle; + int i; + + /* Swap it with the bottom part of the bottom segment. */ + for (i = 0; i < len; i++) + { + tem = argv[bottom + i]; + argv[bottom + i] = argv[middle + i]; + argv[middle + i] = tem; + SWAP_FLAGS (bottom + i, middle + i); + } + /* Exclude the moved top segment from further swapping. */ + bottom += len; + } + } + + /* Update records for the slots the non-options now occupy. */ + + d->__first_nonopt += (d->optind - d->__last_nonopt); + d->__last_nonopt = d->optind; +} + +/* Initialize the internal data when the first call is made. */ + +static const char * +_getopt_initialize (int argc, char *const *argv, const char *optstring, + struct _getopt_data *d, int posixly_correct) +{ + /* Start processing options with ARGV-element 1 (since ARGV-element 0 + is the program name); the sequence of previously skipped + non-option ARGV-elements is empty. */ + + d->__first_nonopt = d->__last_nonopt = d->optind; + + d->__nextchar = NULL; + + d->__posixly_correct = posixly_correct | !!getenv ("POSIXLY_CORRECT"); + + /* Determine how to handle the ordering of options and nonoptions. */ + + if (optstring[0] == '-') + { + d->__ordering = RETURN_IN_ORDER; + ++optstring; + } + else if (optstring[0] == '+') + { + d->__ordering = REQUIRE_ORDER; + ++optstring; + } + else if (d->__posixly_correct) + d->__ordering = REQUIRE_ORDER; + else + d->__ordering = PERMUTE; + +#if defined _LIBC && defined USE_NONOPTION_FLAGS + if (!d->__posixly_correct + && argc == __libc_argc && argv == __libc_argv) + { + if (d->__nonoption_flags_max_len == 0) + { + if (__getopt_nonoption_flags == NULL + || __getopt_nonoption_flags[0] == '\0') + d->__nonoption_flags_max_len = -1; + else + { + const char *orig_str = __getopt_nonoption_flags; + int len = d->__nonoption_flags_max_len = strlen (orig_str); + if (d->__nonoption_flags_max_len < argc) + d->__nonoption_flags_max_len = argc; + __getopt_nonoption_flags = + (char *) malloc (d->__nonoption_flags_max_len); + if (__getopt_nonoption_flags == NULL) + d->__nonoption_flags_max_len = -1; + else + memset (__mempcpy (__getopt_nonoption_flags, orig_str, len), + '\0', d->__nonoption_flags_max_len - len); + } + } + d->__nonoption_flags_len = d->__nonoption_flags_max_len; + } + else + d->__nonoption_flags_len = 0; +#endif + + return optstring; +} + +/* Scan elements of ARGV (whose length is ARGC) for option characters + given in OPTSTRING. + + If an element of ARGV starts with '-', and is not exactly "-" or "--", + then it is an option element. The characters of this element + (aside from the initial '-') are option characters. If `getopt' + is called repeatedly, it returns successively each of the option characters + from each of the option elements. + + If `getopt' finds another option character, it returns that character, + updating `optind' and `nextchar' so that the next call to `getopt' can + resume the scan with the following option character or ARGV-element. + + If there are no more option characters, `getopt' returns -1. + Then `optind' is the index in ARGV of the first ARGV-element + that is not an option. (The ARGV-elements have been permuted + so that those that are not options now come last.) + + OPTSTRING is a string containing the legitimate option characters. + If an option character is seen that is not listed in OPTSTRING, + return '?' after printing an error message. If you set `opterr' to + zero, the error message is suppressed but we still return '?'. + + If a char in OPTSTRING is followed by a colon, that means it wants an arg, + so the following text in the same ARGV-element, or the text of the following + ARGV-element, is returned in `optarg'. Two colons mean an option that + wants an optional arg; if there is text in the current ARGV-element, + it is returned in `optarg', otherwise `optarg' is set to zero. + + If OPTSTRING starts with `-' or `+', it requests different methods of + handling the non-option ARGV-elements. + See the comments about RETURN_IN_ORDER and REQUIRE_ORDER, above. + + Long-named options begin with `--' instead of `-'. + Their names may be abbreviated as long as the abbreviation is unique + or is an exact match for some defined option. If they have an + argument, it follows the option name in the same ARGV-element, separated + from the option name by a `=', or else the in next ARGV-element. + When `getopt' finds a long-named option, it returns 0 if that option's + `flag' field is nonzero, the value of the option's `val' field + if the `flag' field is zero. + + The elements of ARGV aren't really const, because we permute them. + But we pretend they're const in the prototype to be compatible + with other systems. + + LONGOPTS is a vector of `struct option' terminated by an + element containing a name which is zero. + + LONGIND returns the index in LONGOPT of the long-named option found. + It is only valid when a long-named option has been found by the most + recent call. + + If LONG_ONLY is nonzero, '-' as well as '--' can introduce + long-named options. */ + +int +_getopt_internal_r (int argc, char *const *argv, const char *optstring, + const struct option *longopts, int *longind, + int long_only, struct _getopt_data *d, int posixly_correct) +{ + int print_errors = d->opterr; + + if (argc < 1) + return -1; + + d->optarg = NULL; + + if (d->optind == 0 || !d->__initialized) + { + if (d->optind == 0) + d->optind = 1; /* Don't scan ARGV[0], the program name. */ + optstring = _getopt_initialize (argc, argv, optstring, d, + posixly_correct); + d->__initialized = 1; + } + else if (optstring[0] == '-' || optstring[0] == '+') + optstring++; + if (optstring[0] == ':') + print_errors = 0; + + /* Test whether ARGV[optind] points to a non-option argument. + Either it does not have option syntax, or there is an environment flag + from the shell indicating it is not an option. The later information + is only used when the used in the GNU libc. */ +#if defined _LIBC && defined USE_NONOPTION_FLAGS +# define NONOPTION_P (argv[d->optind][0] != '-' || argv[d->optind][1] == '\0' \ + || (d->optind < d->__nonoption_flags_len \ + && __getopt_nonoption_flags[d->optind] == '1')) +#else +# define NONOPTION_P (argv[d->optind][0] != '-' || argv[d->optind][1] == '\0') +#endif + + if (d->__nextchar == NULL || *d->__nextchar == '\0') + { + /* Advance to the next ARGV-element. */ + + /* Give FIRST_NONOPT & LAST_NONOPT rational values if OPTIND has been + moved back by the user (who may also have changed the arguments). */ + if (d->__last_nonopt > d->optind) + d->__last_nonopt = d->optind; + if (d->__first_nonopt > d->optind) + d->__first_nonopt = d->optind; + + if (d->__ordering == PERMUTE) + { + /* If we have just processed some options following some non-options, + exchange them so that the options come first. */ + + if (d->__first_nonopt != d->__last_nonopt + && d->__last_nonopt != d->optind) + exchange ((char **) argv, d); + else if (d->__last_nonopt != d->optind) + d->__first_nonopt = d->optind; + + /* Skip any additional non-options + and extend the range of non-options previously skipped. */ + + while (d->optind < argc && NONOPTION_P) + d->optind++; + d->__last_nonopt = d->optind; + } + + /* The special ARGV-element `--' means premature end of options. + Skip it like a null option, + then exchange with previous non-options as if it were an option, + then skip everything else like a non-option. */ + + if (d->optind != argc && !strcmp (argv[d->optind], "--")) + { + d->optind++; + + if (d->__first_nonopt != d->__last_nonopt + && d->__last_nonopt != d->optind) + exchange ((char **) argv, d); + else if (d->__first_nonopt == d->__last_nonopt) + d->__first_nonopt = d->optind; + d->__last_nonopt = argc; + + d->optind = argc; + } + + /* If we have done all the ARGV-elements, stop the scan + and back over any non-options that we skipped and permuted. */ + + if (d->optind == argc) + { + /* Set the next-arg-index to point at the non-options + that we previously skipped, so the caller will digest them. */ + if (d->__first_nonopt != d->__last_nonopt) + d->optind = d->__first_nonopt; + return -1; + } + + /* If we have come to a non-option and did not permute it, + either stop the scan or describe it to the caller and pass it by. */ + + if (NONOPTION_P) + { + if (d->__ordering == REQUIRE_ORDER) + return -1; + d->optarg = argv[d->optind++]; + return 1; + } + + /* We have found another option-ARGV-element. + Skip the initial punctuation. */ + + d->__nextchar = (argv[d->optind] + 1 + + (longopts != NULL && argv[d->optind][1] == '-')); + } + + /* Decode the current option-ARGV-element. */ + + /* Check whether the ARGV-element is a long option. + + If long_only and the ARGV-element has the form "-f", where f is + a valid short option, don't consider it an abbreviated form of + a long option that starts with f. Otherwise there would be no + way to give the -f short option. + + On the other hand, if there's a long option "fubar" and + the ARGV-element is "-fu", do consider that an abbreviation of + the long option, just like "--fu", and not "-f" with arg "u". + + This distinction seems to be the most useful approach. */ + + if (longopts != NULL + && (argv[d->optind][1] == '-' + || (long_only && (argv[d->optind][2] + || !strchr (optstring, argv[d->optind][1]))))) + { + char *nameend; + unsigned int namelen; + const struct option *p; + const struct option *pfound = NULL; + struct option_list + { + const struct option *p; + struct option_list *next; + int needs_free; + } *ambig_list = NULL; + int exact = 0; + int indfound = -1; + int option_index; + + for (nameend = d->__nextchar; *nameend && *nameend != '='; nameend++) + /* Do nothing. */ ; + namelen = nameend - d->__nextchar; + + /* Test all long options for either exact match + or abbreviated matches. */ + for (p = longopts, option_index = 0; p->name; p++, option_index++) + if (!strncmp (p->name, d->__nextchar, namelen)) + { + if (namelen == (unsigned int) strlen (p->name)) + { + /* Exact match found. */ + pfound = p; + indfound = option_index; + exact = 1; + break; + } + else if (pfound == NULL) + { + /* First nonexact match found. */ + pfound = p; + indfound = option_index; + } + else if (long_only + || pfound->has_arg != p->has_arg + || pfound->flag != p->flag + || pfound->val != p->val) + { + /* Second or later nonexact match found. */ + struct option_list *newp = malloc (sizeof (*newp)); + newp->p = p; + newp->needs_free = 1; + newp->next = ambig_list; + ambig_list = newp; + } + } + + if (ambig_list != NULL && !exact) + { + if (print_errors) + { + struct option_list first; + first.p = pfound; + first.next = ambig_list; + first.needs_free = 0; + ambig_list = &first; + +#if defined _LIBC + char *buf = NULL; + size_t buflen = 0; + + FILE *fp = __open_memstream (&buf, &buflen); + if (fp != NULL) + { + fprintf (fp, + _("%s: option '%s' is ambiguous; possibilities:"), + argv[0], argv[d->optind]); + + do + { + fprintf (fp, " '--%s'", ambig_list->p->name); + ambig_list = ambig_list->next; + } + while (ambig_list != NULL); + + fputc_unlocked ('\n', fp); + + if (__glibc_likely (fclose (fp) != EOF)) + { + _IO_flockfile (stderr); + + int old_flags2 = ((_IO_FILE *) stderr)->_flags2; + ((_IO_FILE *) stderr)->_flags2 |= _IO_FLAGS2_NOTCANCEL; + + __fxprintf (NULL, "%s", buf); + + ((_IO_FILE *) stderr)->_flags2 = old_flags2; + _IO_funlockfile (stderr); + + free (buf); + } + } +#else + fprintf (stderr, + _("%s: option '%s' is ambiguous; possibilities:"), + argv[0], argv[d->optind]); + do + { + struct option_list *tmp_next; + + fprintf (stderr, " '--%s'", ambig_list->p->name); + tmp_next = ambig_list->next; + if (ambig_list->needs_free) + free(ambig_list); + ambig_list = tmp_next; + } + while (ambig_list != NULL); + + fputc ('\n', stderr); +#endif + } + d->__nextchar += strlen (d->__nextchar); + d->optind++; + d->optopt = 0; + return '?'; + } + + if (pfound != NULL) + { + option_index = indfound; + d->optind++; + if (*nameend) + { + /* Don't test has_arg with >, because some C compilers don't + allow it to be used on enums. */ + if (pfound->has_arg) + d->optarg = nameend + 1; + else + { + if (print_errors) + { +#if defined _LIBC + char *buf; + int n; +#endif + + if (argv[d->optind - 1][1] == '-') + { + /* --option */ +#if defined _LIBC + n = __asprintf (&buf, _("\ +%s: option '--%s' doesn't allow an argument\n"), + argv[0], pfound->name); +#else + fprintf (stderr, _("\ +%s: option '--%s' doesn't allow an argument\n"), + argv[0], pfound->name); +#endif + } + else + { + /* +option or -option */ +#if defined _LIBC + n = __asprintf (&buf, _("\ +%s: option '%c%s' doesn't allow an argument\n"), + argv[0], argv[d->optind - 1][0], + pfound->name); +#else + fprintf (stderr, _("\ +%s: option '%c%s' doesn't allow an argument\n"), + argv[0], argv[d->optind - 1][0], + pfound->name); +#endif + } + +#if defined _LIBC + if (n >= 0) + { + _IO_flockfile (stderr); + + int old_flags2 = ((_IO_FILE *) stderr)->_flags2; + ((_IO_FILE *) stderr)->_flags2 + |= _IO_FLAGS2_NOTCANCEL; + + __fxprintf (NULL, "%s", buf); + + ((_IO_FILE *) stderr)->_flags2 = old_flags2; + _IO_funlockfile (stderr); + + free (buf); + } +#endif + } + + d->__nextchar += strlen (d->__nextchar); + + d->optopt = pfound->val; + return '?'; + } + } + else if (pfound->has_arg == 1) + { + if (d->optind < argc) + d->optarg = argv[d->optind++]; + else + { + if (print_errors) + { +#if defined _LIBC + char *buf; + + if (__asprintf (&buf, _("\ +%s: option '--%s' requires an argument\n"), + argv[0], pfound->name) >= 0) + { + _IO_flockfile (stderr); + + int old_flags2 = ((_IO_FILE *) stderr)->_flags2; + ((_IO_FILE *) stderr)->_flags2 + |= _IO_FLAGS2_NOTCANCEL; + + __fxprintf (NULL, "%s", buf); + + ((_IO_FILE *) stderr)->_flags2 = old_flags2; + _IO_funlockfile (stderr); + + free (buf); + } +#else + fprintf (stderr, + _("%s: option '--%s' requires an argument\n"), + argv[0], pfound->name); +#endif + } + d->__nextchar += strlen (d->__nextchar); + d->optopt = pfound->val; + return optstring[0] == ':' ? ':' : '?'; + } + } + d->__nextchar += strlen (d->__nextchar); + if (longind != NULL) + *longind = option_index; + if (pfound->flag) + { + *(pfound->flag) = pfound->val; + return 0; + } + return pfound->val; + } + + /* Can't find it as a long option. If this is not getopt_long_only, + or the option starts with '--' or is not a valid short + option, then it's an error. + Otherwise interpret it as a short option. */ + if (!long_only || argv[d->optind][1] == '-' + || strchr (optstring, *d->__nextchar) == NULL) + { + if (print_errors) + { +#if defined _LIBC + char *buf; + int n; +#endif + + if (argv[d->optind][1] == '-') + { + /* --option */ +#if defined _LIBC + n = __asprintf (&buf, _("%s: unrecognized option '--%s'\n"), + argv[0], d->__nextchar); +#else + fprintf (stderr, _("%s: unrecognized option '--%s'\n"), + argv[0], d->__nextchar); +#endif + } + else + { + /* +option or -option */ +#if defined _LIBC + n = __asprintf (&buf, _("%s: unrecognized option '%c%s'\n"), + argv[0], argv[d->optind][0], d->__nextchar); +#else + fprintf (stderr, _("%s: unrecognized option '%c%s'\n"), + argv[0], argv[d->optind][0], d->__nextchar); +#endif + } + +#if defined _LIBC + if (n >= 0) + { + _IO_flockfile (stderr); + + int old_flags2 = ((_IO_FILE *) stderr)->_flags2; + ((_IO_FILE *) stderr)->_flags2 |= _IO_FLAGS2_NOTCANCEL; + + __fxprintf (NULL, "%s", buf); + + ((_IO_FILE *) stderr)->_flags2 = old_flags2; + _IO_funlockfile (stderr); + + free (buf); + } +#endif + } + d->__nextchar = (char *) ""; + d->optind++; + d->optopt = 0; + return '?'; + } + } + + /* Look at and handle the next short option-character. */ + + { + char c = *d->__nextchar++; + char *temp = strchr (optstring, c); + + /* Increment `optind' when we start to process its last character. */ + if (*d->__nextchar == '\0') + ++d->optind; + + if (temp == NULL || c == ':' || c == ';') + { + if (print_errors) + { +#if defined _LIBC + char *buf; + int n; +#endif + +#if defined _LIBC + n = __asprintf (&buf, _("%s: invalid option -- '%c'\n"), + argv[0], c); +#else + fprintf (stderr, _("%s: invalid option -- '%c'\n"), argv[0], c); +#endif + +#if defined _LIBC + if (n >= 0) + { + _IO_flockfile (stderr); + + int old_flags2 = ((_IO_FILE *) stderr)->_flags2; + ((_IO_FILE *) stderr)->_flags2 |= _IO_FLAGS2_NOTCANCEL; + + __fxprintf (NULL, "%s", buf); + + ((_IO_FILE *) stderr)->_flags2 = old_flags2; + _IO_funlockfile (stderr); + + free (buf); + } +#endif + } + d->optopt = c; + return '?'; + } + /* Convenience. Treat POSIX -W foo same as long option --foo */ + if (temp[0] == 'W' && temp[1] == ';') + { + char *nameend; + const struct option *p; + const struct option *pfound = NULL; + int exact = 0; + int ambig = 0; + int indfound = 0; + int option_index; + + if (longopts == NULL) + goto no_longs; + + /* This is an option that requires an argument. */ + if (*d->__nextchar != '\0') + { + d->optarg = d->__nextchar; + /* If we end this ARGV-element by taking the rest as an arg, + we must advance to the next element now. */ + d->optind++; + } + else if (d->optind == argc) + { + if (print_errors) + { +#if defined _LIBC + char *buf; + + if (__asprintf (&buf, + _("%s: option requires an argument -- '%c'\n"), + argv[0], c) >= 0) + { + _IO_flockfile (stderr); + + int old_flags2 = ((_IO_FILE *) stderr)->_flags2; + ((_IO_FILE *) stderr)->_flags2 |= _IO_FLAGS2_NOTCANCEL; + + __fxprintf (NULL, "%s", buf); + + ((_IO_FILE *) stderr)->_flags2 = old_flags2; + _IO_funlockfile (stderr); + + free (buf); + } +#else + fprintf (stderr, + _("%s: option requires an argument -- '%c'\n"), + argv[0], c); +#endif + } + d->optopt = c; + if (optstring[0] == ':') + c = ':'; + else + c = '?'; + return c; + } + else + /* We already incremented `d->optind' once; + increment it again when taking next ARGV-elt as argument. */ + d->optarg = argv[d->optind++]; + + /* optarg is now the argument, see if it's in the + table of longopts. */ + + for (d->__nextchar = nameend = d->optarg; *nameend && *nameend != '='; + nameend++) + /* Do nothing. */ ; + + /* Test all long options for either exact match + or abbreviated matches. */ + for (p = longopts, option_index = 0; p->name; p++, option_index++) + if (!strncmp (p->name, d->__nextchar, nameend - d->__nextchar)) + { + if ((unsigned int) (nameend - d->__nextchar) == strlen (p->name)) + { + /* Exact match found. */ + pfound = p; + indfound = option_index; + exact = 1; + break; + } + else if (pfound == NULL) + { + /* First nonexact match found. */ + pfound = p; + indfound = option_index; + } + else if (long_only + || pfound->has_arg != p->has_arg + || pfound->flag != p->flag + || pfound->val != p->val) + /* Second or later nonexact match found. */ + ambig = 1; + } + if (ambig && !exact) + { + if (print_errors) + { +#if defined _LIBC + char *buf; + + if (__asprintf (&buf, _("%s: option '-W %s' is ambiguous\n"), + argv[0], d->optarg) >= 0) + { + _IO_flockfile (stderr); + + int old_flags2 = ((_IO_FILE *) stderr)->_flags2; + ((_IO_FILE *) stderr)->_flags2 |= _IO_FLAGS2_NOTCANCEL; + + __fxprintf (NULL, "%s", buf); + + ((_IO_FILE *) stderr)->_flags2 = old_flags2; + _IO_funlockfile (stderr); + + free (buf); + } +#else + fprintf (stderr, _("%s: option '-W %s' is ambiguous\n"), + argv[0], d->optarg); +#endif + } + d->__nextchar += strlen (d->__nextchar); + d->optind++; + return '?'; + } + if (pfound != NULL) + { + option_index = indfound; + if (*nameend) + { + /* Don't test has_arg with >, because some C compilers don't + allow it to be used on enums. */ + if (pfound->has_arg) + d->optarg = nameend + 1; + else + { + if (print_errors) + { +#if defined _LIBC + char *buf; + + if (__asprintf (&buf, _("\ +%s: option '-W %s' doesn't allow an argument\n"), + argv[0], pfound->name) >= 0) + { + _IO_flockfile (stderr); + + int old_flags2 = ((_IO_FILE *) stderr)->_flags2; + ((_IO_FILE *) stderr)->_flags2 + |= _IO_FLAGS2_NOTCANCEL; + + __fxprintf (NULL, "%s", buf); + + ((_IO_FILE *) stderr)->_flags2 = old_flags2; + _IO_funlockfile (stderr); + + free (buf); + } +#else + fprintf (stderr, _("\ +%s: option '-W %s' doesn't allow an argument\n"), + argv[0], pfound->name); +#endif + } + + d->__nextchar += strlen (d->__nextchar); + return '?'; + } + } + else if (pfound->has_arg == 1) + { + if (d->optind < argc) + d->optarg = argv[d->optind++]; + else + { + if (print_errors) + { +#if defined _LIBC + char *buf; + + if (__asprintf (&buf, _("\ +%s: option '-W %s' requires an argument\n"), + argv[0], pfound->name) >= 0) + { + _IO_flockfile (stderr); + + int old_flags2 = ((_IO_FILE *) stderr)->_flags2; + ((_IO_FILE *) stderr)->_flags2 + |= _IO_FLAGS2_NOTCANCEL; + + __fxprintf (NULL, "%s", buf); + + ((_IO_FILE *) stderr)->_flags2 = old_flags2; + _IO_funlockfile (stderr); + + free (buf); + } +#else + fprintf (stderr, _("\ +%s: option '-W %s' requires an argument\n"), + argv[0], pfound->name); +#endif + } + d->__nextchar += strlen (d->__nextchar); + return optstring[0] == ':' ? ':' : '?'; + } + } + else + d->optarg = NULL; + d->__nextchar += strlen (d->__nextchar); + if (longind != NULL) + *longind = option_index; + if (pfound->flag) + { + *(pfound->flag) = pfound->val; + return 0; + } + return pfound->val; + } + + no_longs: + d->__nextchar = NULL; + return 'W'; /* Let the application handle it. */ + } + if (temp[1] == ':') + { + if (temp[2] == ':') + { + /* This is an option that accepts an argument optionally. */ + if (*d->__nextchar != '\0') + { + d->optarg = d->__nextchar; + d->optind++; + } + else + d->optarg = NULL; + d->__nextchar = NULL; + } + else + { + /* This is an option that requires an argument. */ + if (*d->__nextchar != '\0') + { + d->optarg = d->__nextchar; + /* If we end this ARGV-element by taking the rest as an arg, + we must advance to the next element now. */ + d->optind++; + } + else if (d->optind == argc) + { + if (print_errors) + { +#if defined _LIBC + char *buf; + + if (__asprintf (&buf, _("\ +%s: option requires an argument -- '%c'\n"), + argv[0], c) >= 0) + { + _IO_flockfile (stderr); + + int old_flags2 = ((_IO_FILE *) stderr)->_flags2; + ((_IO_FILE *) stderr)->_flags2 |= _IO_FLAGS2_NOTCANCEL; + + __fxprintf (NULL, "%s", buf); + + ((_IO_FILE *) stderr)->_flags2 = old_flags2; + _IO_funlockfile (stderr); + + free (buf); + } +#else + fprintf (stderr, + _("%s: option requires an argument -- '%c'\n"), + argv[0], c); +#endif + } + d->optopt = c; + if (optstring[0] == ':') + c = ':'; + else + c = '?'; + } + else + /* We already incremented `optind' once; + increment it again when taking next ARGV-elt as argument. */ + d->optarg = argv[d->optind++]; + d->__nextchar = NULL; + } + } + return c; + } +} + +int +_getopt_internal (int argc, char *const *argv, const char *optstring, + const struct option *longopts, int *longind, int long_only, + int posixly_correct) +{ + int result; + + getopt_data.optind = optind; + getopt_data.opterr = opterr; + + result = _getopt_internal_r (argc, argv, optstring, longopts, + longind, long_only, &getopt_data, + posixly_correct); + + optind = getopt_data.optind; + optarg = getopt_data.optarg; + optopt = getopt_data.optopt; + + return result; +} + +int +getopt (int argc, char *const *argv, const char *optstring) +{ + return _getopt_internal (argc, argv, optstring, + (const struct option *) 0, + (int *) 0, + 0, 0); +} + +#ifdef _LIBC +int +__posix_getopt (int argc, char *const *argv, const char *optstring) +{ + return _getopt_internal (argc, argv, optstring, + (const struct option *) 0, + (int *) 0, + 0, 1); +} +#endif + +#endif /* Not ELIDE_CODE. */ + +#ifdef TEST + +/* Compile with -DTEST to make an executable for use in testing + the above definition of `getopt'. */ + +int +main (int argc, char **argv) +{ + int c; + int digit_optind = 0; + + while (1) + { + int this_option_optind = optind ? optind : 1; + + c = getopt (argc, argv, "abc:d:0123456789"); + if (c == -1) + break; + + switch (c) + { + case '0': + case '1': + case '2': + case '3': + case '4': + case '5': + case '6': + case '7': + case '8': + case '9': + if (digit_optind != 0 && digit_optind != this_option_optind) + printf ("digits occur in two different argv-elements.\n"); + digit_optind = this_option_optind; + printf ("option %c\n", c); + break; + + case 'a': + printf ("option a\n"); + break; + + case 'b': + printf ("option b\n"); + break; + + case 'c': + printf ("option c with value '%s'\n", optarg); + break; + + case '?': + break; + + default: + printf ("?? getopt returned character code 0%o ??\n", c); + } + } + + if (optind < argc) + { + printf ("non-option ARGV-elements: "); + while (optind < argc) + printf ("%s ", argv[optind++]); + printf ("\n"); + } + + exit (0); +} + +#endif /* TEST */ diff --git a/support/getopt.h b/support/getopt.h new file mode 100644 index 00000000..8393569d --- /dev/null +++ b/support/getopt.h @@ -0,0 +1,206 @@ +/* Declarations for getopt. + Copyright (C) 1989-2016 Free Software Foundation, Inc. + This file is part of the GNU C Library. + + The GNU C Library is free software; you can redistribute it and/or + modify it under the terms of the GNU Lesser General Public + License as published by the Free Software Foundation; either + version 2.1 of the License, or (at your option) any later version. + + The GNU C Library is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public + License along with the GNU C Library; if not, see + <http://www.gnu.org/licenses/>. */ + +#ifndef _GETOPT_H + +#ifndef __need_getopt +# define _GETOPT_H 1 +#endif + +/* If __GNU_LIBRARY__ is not already defined, either we are being used + standalone, or this is the first header included in the source file. + If we are being used with glibc, we need to include <features.h>, but + that does not exist if we are standalone. So: if __GNU_LIBRARY__ is + not defined, include <ctype.h>, which will pull in <features.h> for us + if it's from glibc. (Why ctype.h? It's guaranteed to exist and it + doesn't flood the namespace with stuff the way some other headers do.) */ +#if !defined __GNU_LIBRARY__ +# include <ctype.h> +#endif + +#ifndef __THROW +# ifndef __GNUC_PREREQ +# define __GNUC_PREREQ(maj, min) (0) +# endif +# if defined __cplusplus && __GNUC_PREREQ (2,8) +# define __THROW throw () +# else +# define __THROW +# endif +#endif + +#ifdef __cplusplus +extern "C" { +#endif + +#ifdef __KLIBC__ +/* OS/2 kLIBC has already getopt(). So to avoid name clash, rename + them here. */ + +# define optarg gawk_optarg +# define optind gawk_optind +# define opterr gawk_opterr +# define optopt gawk_optopt + +# define getopt gawk_getopt +# define getopt_long gawk_getopt_long +# define getopt_long_only gawk_getopt_long_only +#endif + + +/* For communication from `getopt' to the caller. + When `getopt' finds an option that takes an argument, + the argument value is returned here. + Also, when `ordering' is RETURN_IN_ORDER, + each non-option ARGV-element is returned here. */ + +extern char *optarg; + +/* Index in ARGV of the next element to be scanned. + This is used for communication to and from the caller + and for communication between successive calls to `getopt'. + + On entry to `getopt', zero means this is the first call; initialize. + + When `getopt' returns -1, this is the index of the first of the + non-option elements that the caller should itself scan. + + Otherwise, `optind' communicates from one call to the next + how much of ARGV has been scanned so far. */ + +extern int optind; + +/* Callers store zero here to inhibit the error message `getopt' prints + for unrecognized options. */ + +extern int opterr; + +/* Set to an option character which was unrecognized. */ + +extern int optopt; + +#ifndef __need_getopt +/* Describe the long-named options requested by the application. + The LONG_OPTIONS argument to getopt_long or getopt_long_only is a vector + of `struct option' terminated by an element containing a name which is + zero. + + The field `has_arg' is: + no_argument (or 0) if the option does not take an argument, + required_argument (or 1) if the option requires an argument, + optional_argument (or 2) if the option takes an optional argument. + + If the field `flag' is not NULL, it points to a variable that is set + to the value given in the field `val' when the option is found, but + left unchanged if the option is not found. + + To have a long-named option do something other than set an `int' to + a compiled-in constant, such as set a value from `optarg', set the + option's `flag' field to zero and its `val' field to a nonzero + value (the equivalent single-letter option character, if there is + one). For long options that have a zero `flag' field, `getopt' + returns the contents of the `val' field. */ + +struct option +{ + const char *name; + /* has_arg can't be an enum because some compilers complain about + type mismatches in all the code that assumes it is an int. */ + int has_arg; + int *flag; + int val; +}; + +/* Names for the values of the `has_arg' field of `struct option'. */ + +# define no_argument 0 +# define required_argument 1 +# define optional_argument 2 +#endif /* need getopt */ + + +/* Get definitions and prototypes for functions to process the + arguments in ARGV (ARGC of them, minus the program name) for + options given in OPTS. + + Return the option character from OPTS just read. Return -1 when + there are no more options. For unrecognized options, or options + missing arguments, `optopt' is set to the option letter, and '?' is + returned. + + The OPTS string is a list of characters which are recognized option + letters, optionally followed by colons, specifying that that letter + takes an argument, to be placed in `optarg'. + + If a letter in OPTS is followed by two colons, its argument is + optional. This behavior is specific to the GNU `getopt'. + + The argument `--' causes premature termination of argument + scanning, explicitly telling `getopt' that there are no more + options. + + If OPTS begins with `--', then non-option arguments are treated as + arguments to the option '\0'. This behavior is specific to the GNU + `getopt'. */ + +#ifdef __GNU_LIBRARY__ +/* Many other libraries have conflicting prototypes for getopt, with + differences in the consts, in stdlib.h. To avoid compilation + errors, only prototype getopt for the GNU C library. */ +extern int getopt (int ___argc, char *const *___argv, const char *__shortopts) + __THROW; + +# if defined __need_getopt && defined __USE_POSIX2 \ + && !defined __USE_POSIX_IMPLICITLY && !defined __USE_GNU +/* The GNU getopt has more functionality than the standard version. The + additional functionality can be disable at runtime. This redirection + helps to also do this at runtime. */ +# ifdef __REDIRECT + extern int __REDIRECT_NTH (getopt, (int ___argc, char *const *___argv, + const char *__shortopts), + __posix_getopt); +# else +extern int __posix_getopt (int ___argc, char *const *___argv, + const char *__shortopts) __THROW; +# define getopt __posix_getopt +# endif +# endif +#else /* not __GNU_LIBRARY__ */ +extern int getopt (); +#endif /* __GNU_LIBRARY__ */ + +#ifndef __need_getopt +extern int getopt_long (int ___argc, char *const *___argv, + const char *__shortopts, + const struct option *__longopts, int *__longind) + __THROW; +extern int getopt_long_only (int ___argc, char *const *___argv, + const char *__shortopts, + const struct option *__longopts, int *__longind) + __THROW; + +#endif + +#ifdef __cplusplus +} +#endif + +/* Make sure we later can get all the definitions and declarations. */ +#undef __need_getopt + +#endif /* getopt.h */ diff --git a/support/getopt1.c b/support/getopt1.c new file mode 100644 index 00000000..438fe52b --- /dev/null +++ b/support/getopt1.c @@ -0,0 +1,195 @@ +/* getopt_long and getopt_long_only entry points for GNU getopt. + Copyright (C) 1987-2016 Free Software Foundation, Inc. + This file is part of the GNU C Library. + + The GNU C Library is free software; you can redistribute it and/or + modify it under the terms of the GNU Lesser General Public + License as published by the Free Software Foundation; either + version 2.1 of the License, or (at your option) any later version. + + The GNU C Library is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public + License along with the GNU C Library; if not, see + <http://www.gnu.org/licenses/>. */ + +#ifdef HAVE_CONFIG_H +#include <config.h> +#endif + +#ifdef _LIBC +# include <getopt.h> +#else +# include "getopt.h" +#endif +#include "getopt_int.h" + +#include <stdio.h> + +/* Comment out all this code if we are using the GNU C Library, and are not + actually compiling the library itself. This code is part of the GNU C + Library, but also included in many other GNU distributions. Compiling + and linking in this code is a waste when using the GNU C library + (especially if it is a shared library). Rather than having every GNU + program understand `configure --with-gnu-libc' and omit the object files, + it is simpler to just do this in the source for each such file. */ + +#define GETOPT_INTERFACE_VERSION 2 +#if !defined _LIBC && defined __GLIBC__ && __GLIBC__ >= 2 +#include <gnu-versions.h> +#if _GNU_GETOPT_INTERFACE_VERSION == GETOPT_INTERFACE_VERSION +#define ELIDE_CODE +#endif +#endif + +/* !@#$%^&*() !!!!!!!! */ +#ifdef GAWK +#undef ELIDE_CODE +#endif + +#ifndef ELIDE_CODE + + +/* This needs to come after some library #include + to get __GNU_LIBRARY__ defined. */ +#ifdef __GNU_LIBRARY__ +#include <stdlib.h> +#endif + +#ifndef NULL +#define NULL 0 +#endif + +int +getopt_long (int argc, char *const *argv, const char *options, + const struct option *long_options, int *opt_index) +{ + return _getopt_internal (argc, argv, options, long_options, opt_index, 0, 0); +} + +int +_getopt_long_r (int argc, char *const *argv, const char *options, + const struct option *long_options, int *opt_index, + struct _getopt_data *d) +{ + return _getopt_internal_r (argc, argv, options, long_options, opt_index, + 0, d, 0); +} + +/* Like getopt_long, but '-' as well as '--' can indicate a long option. + If an option that starts with '-' (not '--') doesn't match a long option, + but does match a short option, it is parsed as a short option + instead. */ + +int +getopt_long_only (int argc, char *const *argv, const char *options, + const struct option *long_options, int *opt_index) +{ + return _getopt_internal (argc, argv, options, long_options, opt_index, 1, 0); +} + +int +_getopt_long_only_r (int argc, char *const *argv, const char *options, + const struct option *long_options, int *opt_index, + struct _getopt_data *d) +{ + return _getopt_internal_r (argc, argv, options, long_options, opt_index, + 1, d, 0); +} + +#endif /* Not ELIDE_CODE. */ + +#ifdef TEST + +#include <stdio.h> + +int +main (int argc, char **argv) +{ + int c; + int digit_optind = 0; + + while (1) + { + int this_option_optind = optind ? optind : 1; + int option_index = 0; + static struct option long_options[] = + { + {"add", 1, 0, 0}, + {"append", 0, 0, 0}, + {"delete", 1, 0, 0}, + {"verbose", 0, 0, 0}, + {"create", 0, 0, 0}, + {"file", 1, 0, 0}, + {0, 0, 0, 0} + }; + + c = getopt_long (argc, argv, "abc:d:0123456789", + long_options, &option_index); + if (c == -1) + break; + + switch (c) + { + case 0: + printf ("option %s", long_options[option_index].name); + if (optarg) + printf (" with arg %s", optarg); + printf ("\n"); + break; + + case '0': + case '1': + case '2': + case '3': + case '4': + case '5': + case '6': + case '7': + case '8': + case '9': + if (digit_optind != 0 && digit_optind != this_option_optind) + printf ("digits occur in two different argv-elements.\n"); + digit_optind = this_option_optind; + printf ("option %c\n", c); + break; + + case 'a': + printf ("option a\n"); + break; + + case 'b': + printf ("option b\n"); + break; + + case 'c': + printf ("option c with value `%s'\n", optarg); + break; + + case 'd': + printf ("option d with value `%s'\n", optarg); + break; + + case '?': + break; + + default: + printf ("?? getopt returned character code 0%o ??\n", c); + } + } + + if (optind < argc) + { + printf ("non-option ARGV-elements: "); + while (optind < argc) + printf ("%s ", argv[optind++]); + printf ("\n"); + } + + exit (0); +} + +#endif /* TEST */ diff --git a/support/getopt_int.h b/support/getopt_int.h new file mode 100644 index 00000000..514a1beb --- /dev/null +++ b/support/getopt_int.h @@ -0,0 +1,129 @@ +/* Internal declarations for getopt. + Copyright (C) 1989-2016 Free Software Foundation, Inc. + This file is part of the GNU C Library. + + The GNU C Library is free software; you can redistribute it and/or + modify it under the terms of the GNU Lesser General Public + License as published by the Free Software Foundation; either + version 2.1 of the License, or (at your option) any later version. + + The GNU C Library is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public + License along with the GNU C Library; if not, see + <http://www.gnu.org/licenses/>. */ + +#ifndef _GETOPT_INT_H +#define _GETOPT_INT_H 1 + +extern int _getopt_internal (int ___argc, char *const *___argv, + const char *__shortopts, + const struct option *__longopts, int *__longind, + int __long_only, int posixly_correct); + + +/* Reentrant versions which can handle parsing multiple argument + vectors at the same time. */ + +/* Data type for reentrant functions. */ +struct _getopt_data +{ + /* These have exactly the same meaning as the corresponding global + variables, except that they are used for the reentrant + versions of getopt. */ + int optind; + int opterr; + int optopt; + char *optarg; + + /* Internal members. */ + + /* True if the internal members have been initialized. */ + int __initialized; + + /* The next char to be scanned in the option-element + in which the last option character we returned was found. + This allows us to pick up the scan where we left off. + + If this is zero, or a null string, it means resume the scan + by advancing to the next ARGV-element. */ + char *__nextchar; + + /* Describe how to deal with options that follow non-option ARGV-elements. + + If the caller did not specify anything, + the default is REQUIRE_ORDER if the environment variable + POSIXLY_CORRECT is defined, PERMUTE otherwise. + + REQUIRE_ORDER means don't recognize them as options; + stop option processing when the first non-option is seen. + This is what Unix does. + This mode of operation is selected by either setting the environment + variable POSIXLY_CORRECT, or using `+' as the first character + of the list of option characters. + + PERMUTE is the default. We permute the contents of ARGV as we + scan, so that eventually all the non-options are at the end. + This allows options to be given in any order, even with programs + that were not written to expect this. + + RETURN_IN_ORDER is an option available to programs that were + written to expect options and other ARGV-elements in any order + and that care about the ordering of the two. We describe each + non-option ARGV-element as if it were the argument of an option + with character code 1. Using `-' as the first character of the + list of option characters selects this mode of operation. + + The special argument `--' forces an end of option-scanning regardless + of the value of `ordering'. In the case of RETURN_IN_ORDER, only + `--' can cause `getopt' to return -1 with `optind' != ARGC. */ + + enum + { + REQUIRE_ORDER, PERMUTE, RETURN_IN_ORDER + } __ordering; + + /* If the POSIXLY_CORRECT environment variable is set. */ + int __posixly_correct; + + + /* Handle permutation of arguments. */ + + /* Describe the part of ARGV that contains non-options that have + been skipped. `first_nonopt' is the index in ARGV of the first + of them; `last_nonopt' is the index after the last of them. */ + + int __first_nonopt; + int __last_nonopt; + +#if defined _LIBC && defined USE_NONOPTION_FLAGS + int __nonoption_flags_max_len; + int __nonoption_flags_len; +# endif +}; + +/* The initializer is necessary to set OPTIND and OPTERR to their + default values and to clear the initialization flag. */ +#define _GETOPT_DATA_INITIALIZER { 1, 1 } + +extern int _getopt_internal_r (int ___argc, char *const *___argv, + const char *__shortopts, + const struct option *__longopts, int *__longind, + int __long_only, struct _getopt_data *__data, + int posixly_correct); + +extern int _getopt_long_r (int ___argc, char *const *___argv, + const char *__shortopts, + const struct option *__longopts, int *__longind, + struct _getopt_data *__data); + +extern int _getopt_long_only_r (int ___argc, char *const *___argv, + const char *__shortopts, + const struct option *__longopts, + int *__longind, + struct _getopt_data *__data); + +#endif /* getopt_int.h */ diff --git a/support/intprops.h b/support/intprops.h new file mode 100644 index 00000000..716741ad --- /dev/null +++ b/support/intprops.h @@ -0,0 +1,464 @@ +/* intprops.h -- properties of integer types + + Copyright (C) 2001-2016 Free Software Foundation, Inc. + + This program is free software: you can redistribute it and/or modify it + under the terms of the GNU Lesser General Public License as published + by the Free Software Foundation; either version 2.1 of the License, or + (at your option) any later version. + + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with this program. If not, see <http://www.gnu.org/licenses/>. */ + +/* Written by Paul Eggert. */ + +#ifndef _GL_INTPROPS_H +#define _GL_INTPROPS_H + +#include <limits.h> +#include <verify.h> + +#ifndef __has_builtin +# define __has_builtin(x) 0 +#endif + +/* Return a value with the common real type of E and V and the value of V. */ +#define _GL_INT_CONVERT(e, v) (0 * (e) + (v)) + +/* Act like _GL_INT_CONVERT (E, -V) but work around a bug in IRIX 6.5 cc; see + <http://lists.gnu.org/archive/html/bug-gnulib/2011-05/msg00406.html>. */ +#define _GL_INT_NEGATE_CONVERT(e, v) (0 * (e) - (v)) + +/* The extra casts in the following macros work around compiler bugs, + e.g., in Cray C 5.0.3.0. */ + +/* True if the arithmetic type T is an integer type. bool counts as + an integer. */ +#define TYPE_IS_INTEGER(t) ((t) 1.5 == 1) + +/* True if the real type T is signed. */ +#define TYPE_SIGNED(t) (! ((t) 0 < (t) -1)) + +/* Return 1 if the real expression E, after promotion, has a + signed or floating type. */ +#define EXPR_SIGNED(e) (_GL_INT_NEGATE_CONVERT (e, 1) < 0) + + +/* Minimum and maximum values for integer types and expressions. */ + +/* The width in bits of the integer type or expression T. + Padding bits are not supported; this is checked at compile-time below. */ +#define TYPE_WIDTH(t) (sizeof (t) * CHAR_BIT) + +/* The maximum and minimum values for the integer type T. */ +#define TYPE_MINIMUM(t) ((t) ~ TYPE_MAXIMUM (t)) +#define TYPE_MAXIMUM(t) \ + ((t) (! TYPE_SIGNED (t) \ + ? (t) -1 \ + : ((((t) 1 << (TYPE_WIDTH (t) - 2)) - 1) * 2 + 1))) + +/* The maximum and minimum values for the type of the expression E, + after integer promotion. E should not have side effects. */ +#define _GL_INT_MINIMUM(e) \ + (EXPR_SIGNED (e) \ + ? ~ _GL_SIGNED_INT_MAXIMUM (e) \ + : _GL_INT_CONVERT (e, 0)) +#define _GL_INT_MAXIMUM(e) \ + (EXPR_SIGNED (e) \ + ? _GL_SIGNED_INT_MAXIMUM (e) \ + : _GL_INT_NEGATE_CONVERT (e, 1)) +#define _GL_SIGNED_INT_MAXIMUM(e) \ + (((_GL_INT_CONVERT (e, 1) << (TYPE_WIDTH ((e) + 0) - 2)) - 1) * 2 + 1) + +/* Work around OpenVMS incompatibility with C99. */ +#if !defined LLONG_MAX && defined __INT64_MAX +# define LLONG_MAX __INT64_MAX +# define LLONG_MIN __INT64_MIN +#endif + +/* This include file assumes that signed types are two's complement without + padding bits; the above macros have undefined behavior otherwise. + If this is a problem for you, please let us know how to fix it for your host. + As a sanity check, test the assumption for some signed types that + <limits.h> bounds. */ +verify (TYPE_MINIMUM (signed char) == SCHAR_MIN); +verify (TYPE_MAXIMUM (signed char) == SCHAR_MAX); +verify (TYPE_MINIMUM (short int) == SHRT_MIN); +verify (TYPE_MAXIMUM (short int) == SHRT_MAX); +verify (TYPE_MINIMUM (int) == INT_MIN); +verify (TYPE_MAXIMUM (int) == INT_MAX); +verify (TYPE_MINIMUM (long int) == LONG_MIN); +verify (TYPE_MAXIMUM (long int) == LONG_MAX); +#ifdef LLONG_MAX +verify (TYPE_MINIMUM (long long int) == LLONG_MIN); +verify (TYPE_MAXIMUM (long long int) == LLONG_MAX); +#endif +/* Similarly, sanity-check one ISO/IEC TS 18661-1:2014 macro if defined. */ +#ifdef UINT_WIDTH +verify (TYPE_WIDTH (unsigned int) == UINT_WIDTH); +#endif + +/* Does the __typeof__ keyword work? This could be done by + 'configure', but for now it's easier to do it by hand. */ +#if (2 <= __GNUC__ \ + || (1210 <= __IBMC__ && defined __IBM__TYPEOF__) \ + || (0x5110 <= __SUNPRO_C && !__STDC__)) +# define _GL_HAVE___TYPEOF__ 1 +#else +# define _GL_HAVE___TYPEOF__ 0 +#endif + +/* Return 1 if the integer type or expression T might be signed. Return 0 + if it is definitely unsigned. This macro does not evaluate its argument, + and expands to an integer constant expression. */ +#if _GL_HAVE___TYPEOF__ +# define _GL_SIGNED_TYPE_OR_EXPR(t) TYPE_SIGNED (__typeof__ (t)) +#else +# define _GL_SIGNED_TYPE_OR_EXPR(t) 1 +#endif + +/* Bound on length of the string representing an unsigned integer + value representable in B bits. log10 (2.0) < 146/485. The + smallest value of B where this bound is not tight is 2621. */ +#define INT_BITS_STRLEN_BOUND(b) (((b) * 146 + 484) / 485) + +/* Bound on length of the string representing an integer type or expression T. + Subtract 1 for the sign bit if T is signed, and then add 1 more for + a minus sign if needed. + + Because _GL_SIGNED_TYPE_OR_EXPR sometimes returns 0 when its argument is + signed, this macro may overestimate the true bound by one byte when + applied to unsigned types of size 2, 4, 16, ... bytes. */ +#define INT_STRLEN_BOUND(t) \ + (INT_BITS_STRLEN_BOUND (TYPE_WIDTH (t) - _GL_SIGNED_TYPE_OR_EXPR (t)) \ + + _GL_SIGNED_TYPE_OR_EXPR (t)) + +/* Bound on buffer size needed to represent an integer type or expression T, + including the terminating null. */ +#define INT_BUFSIZE_BOUND(t) (INT_STRLEN_BOUND (t) + 1) + + +/* Range overflow checks. + + The INT_<op>_RANGE_OVERFLOW macros return 1 if the corresponding C + operators might not yield numerically correct answers due to + arithmetic overflow. They do not rely on undefined or + implementation-defined behavior. Their implementations are simple + and straightforward, but they are a bit harder to use than the + INT_<op>_OVERFLOW macros described below. + + Example usage: + + long int i = ...; + long int j = ...; + if (INT_MULTIPLY_RANGE_OVERFLOW (i, j, LONG_MIN, LONG_MAX)) + printf ("multiply would overflow"); + else + printf ("product is %ld", i * j); + + Restrictions on *_RANGE_OVERFLOW macros: + + These macros do not check for all possible numerical problems or + undefined or unspecified behavior: they do not check for division + by zero, for bad shift counts, or for shifting negative numbers. + + These macros may evaluate their arguments zero or multiple times, + so the arguments should not have side effects. The arithmetic + arguments (including the MIN and MAX arguments) must be of the same + integer type after the usual arithmetic conversions, and the type + must have minimum value MIN and maximum MAX. Unsigned types should + use a zero MIN of the proper type. + + These macros are tuned for constant MIN and MAX. For commutative + operations such as A + B, they are also tuned for constant B. */ + +/* Return 1 if A + B would overflow in [MIN,MAX] arithmetic. + See above for restrictions. */ +#define INT_ADD_RANGE_OVERFLOW(a, b, min, max) \ + ((b) < 0 \ + ? (a) < (min) - (b) \ + : (max) - (b) < (a)) + +/* Return 1 if A - B would overflow in [MIN,MAX] arithmetic. + See above for restrictions. */ +#define INT_SUBTRACT_RANGE_OVERFLOW(a, b, min, max) \ + ((b) < 0 \ + ? (max) + (b) < (a) \ + : (a) < (min) + (b)) + +/* Return 1 if - A would overflow in [MIN,MAX] arithmetic. + See above for restrictions. */ +#define INT_NEGATE_RANGE_OVERFLOW(a, min, max) \ + ((min) < 0 \ + ? (a) < - (max) \ + : 0 < (a)) + +/* Return 1 if A * B would overflow in [MIN,MAX] arithmetic. + See above for restrictions. Avoid && and || as they tickle + bugs in Sun C 5.11 2010/08/13 and other compilers; see + <http://lists.gnu.org/archive/html/bug-gnulib/2011-05/msg00401.html>. */ +#define INT_MULTIPLY_RANGE_OVERFLOW(a, b, min, max) \ + ((b) < 0 \ + ? ((a) < 0 \ + ? (a) < (max) / (b) \ + : (b) == -1 \ + ? 0 \ + : (min) / (b) < (a)) \ + : (b) == 0 \ + ? 0 \ + : ((a) < 0 \ + ? (a) < (min) / (b) \ + : (max) / (b) < (a))) + +/* Return 1 if A / B would overflow in [MIN,MAX] arithmetic. + See above for restrictions. Do not check for division by zero. */ +#define INT_DIVIDE_RANGE_OVERFLOW(a, b, min, max) \ + ((min) < 0 && (b) == -1 && (a) < - (max)) + +/* Return 1 if A % B would overflow in [MIN,MAX] arithmetic. + See above for restrictions. Do not check for division by zero. + Mathematically, % should never overflow, but on x86-like hosts + INT_MIN % -1 traps, and the C standard permits this, so treat this + as an overflow too. */ +#define INT_REMAINDER_RANGE_OVERFLOW(a, b, min, max) \ + INT_DIVIDE_RANGE_OVERFLOW (a, b, min, max) + +/* Return 1 if A << B would overflow in [MIN,MAX] arithmetic. + See above for restrictions. Here, MIN and MAX are for A only, and B need + not be of the same type as the other arguments. The C standard says that + behavior is undefined for shifts unless 0 <= B < wordwidth, and that when + A is negative then A << B has undefined behavior and A >> B has + implementation-defined behavior, but do not check these other + restrictions. */ +#define INT_LEFT_SHIFT_RANGE_OVERFLOW(a, b, min, max) \ + ((a) < 0 \ + ? (a) < (min) >> (b) \ + : (max) >> (b) < (a)) + +/* True if __builtin_add_overflow (A, B, P) works when P is non-null. */ +#define _GL_HAS_BUILTIN_OVERFLOW \ + (5 <= __GNUC__ || __has_builtin (__builtin_add_overflow)) + +/* True if __builtin_add_overflow_p (A, B, C) works. */ +#define _GL_HAS_BUILTIN_OVERFLOW_P \ + (7 <= __GNUC__ || __has_builtin (__builtin_add_overflow_p)) + +/* The _GL*_OVERFLOW macros have the same restrictions as the + *_RANGE_OVERFLOW macros, except that they do not assume that operands + (e.g., A and B) have the same type as MIN and MAX. Instead, they assume + that the result (e.g., A + B) has that type. */ +#if _GL_HAS_BUILTIN_OVERFLOW_P +# define _GL_ADD_OVERFLOW(a, b, min, max) \ + __builtin_add_overflow_p (a, b, (__typeof__ ((a) + (b))) 0) +# define _GL_SUBTRACT_OVERFLOW(a, b, min, max) \ + __builtin_sub_overflow_p (a, b, (__typeof__ ((a) - (b))) 0) +# define _GL_MULTIPLY_OVERFLOW(a, b, min, max) \ + __builtin_mul_overflow_p (a, b, (__typeof__ ((a) * (b))) 0) +#else +# define _GL_ADD_OVERFLOW(a, b, min, max) \ + ((min) < 0 ? INT_ADD_RANGE_OVERFLOW (a, b, min, max) \ + : (a) < 0 ? (b) <= (a) + (b) \ + : (b) < 0 ? (a) <= (a) + (b) \ + : (a) + (b) < (b)) +# define _GL_SUBTRACT_OVERFLOW(a, b, min, max) \ + ((min) < 0 ? INT_SUBTRACT_RANGE_OVERFLOW (a, b, min, max) \ + : (a) < 0 ? 1 \ + : (b) < 0 ? (a) - (b) <= (a) \ + : (a) < (b)) +# define _GL_MULTIPLY_OVERFLOW(a, b, min, max) \ + (((min) == 0 && (((a) < 0 && 0 < (b)) || ((b) < 0 && 0 < (a)))) \ + || INT_MULTIPLY_RANGE_OVERFLOW (a, b, min, max)) +#endif +#define _GL_DIVIDE_OVERFLOW(a, b, min, max) \ + ((min) < 0 ? (b) == _GL_INT_NEGATE_CONVERT (min, 1) && (a) < - (max) \ + : (a) < 0 ? (b) <= (a) + (b) - 1 \ + : (b) < 0 && (a) + (b) <= (a)) +#define _GL_REMAINDER_OVERFLOW(a, b, min, max) \ + ((min) < 0 ? (b) == _GL_INT_NEGATE_CONVERT (min, 1) && (a) < - (max) \ + : (a) < 0 ? (a) % (b) != ((max) - (b) + 1) % (b) \ + : (b) < 0 && ! _GL_UNSIGNED_NEG_MULTIPLE (a, b, max)) + +/* Return a nonzero value if A is a mathematical multiple of B, where + A is unsigned, B is negative, and MAX is the maximum value of A's + type. A's type must be the same as (A % B)'s type. Normally (A % + -B == 0) suffices, but things get tricky if -B would overflow. */ +#define _GL_UNSIGNED_NEG_MULTIPLE(a, b, max) \ + (((b) < -_GL_SIGNED_INT_MAXIMUM (b) \ + ? (_GL_SIGNED_INT_MAXIMUM (b) == (max) \ + ? (a) \ + : (a) % (_GL_INT_CONVERT (a, _GL_SIGNED_INT_MAXIMUM (b)) + 1)) \ + : (a) % - (b)) \ + == 0) + +/* Check for integer overflow, and report low order bits of answer. + + The INT_<op>_OVERFLOW macros return 1 if the corresponding C operators + might not yield numerically correct answers due to arithmetic overflow. + The INT_<op>_WRAPV macros also store the low-order bits of the answer. + These macros work correctly on all known practical hosts, and do not rely + on undefined behavior due to signed arithmetic overflow. + + Example usage, assuming A and B are long int: + + if (INT_MULTIPLY_OVERFLOW (a, b)) + printf ("result would overflow\n"); + else + printf ("result is %ld (no overflow)\n", a * b); + + Example usage with WRAPV flavor: + + long int result; + bool overflow = INT_MULTIPLY_WRAPV (a, b, &result); + printf ("result is %ld (%s)\n", result, + overflow ? "after overflow" : "no overflow"); + + Restrictions on these macros: + + These macros do not check for all possible numerical problems or + undefined or unspecified behavior: they do not check for division + by zero, for bad shift counts, or for shifting negative numbers. + + These macros may evaluate their arguments zero or multiple times, so the + arguments should not have side effects. + + The WRAPV macros are not constant expressions. They support only + +, binary -, and *. The result type must be signed. + + These macros are tuned for their last argument being a constant. + + Return 1 if the integer expressions A * B, A - B, -A, A * B, A / B, + A % B, and A << B would overflow, respectively. */ + +#define INT_ADD_OVERFLOW(a, b) \ + _GL_BINARY_OP_OVERFLOW (a, b, _GL_ADD_OVERFLOW) +#define INT_SUBTRACT_OVERFLOW(a, b) \ + _GL_BINARY_OP_OVERFLOW (a, b, _GL_SUBTRACT_OVERFLOW) +#if _GL_HAS_BUILTIN_OVERFLOW_P +# define INT_NEGATE_OVERFLOW(a) INT_SUBTRACT_OVERFLOW (0, a) +#else +# define INT_NEGATE_OVERFLOW(a) \ + INT_NEGATE_RANGE_OVERFLOW (a, _GL_INT_MINIMUM (a), _GL_INT_MAXIMUM (a)) +#endif +#define INT_MULTIPLY_OVERFLOW(a, b) \ + _GL_BINARY_OP_OVERFLOW (a, b, _GL_MULTIPLY_OVERFLOW) +#define INT_DIVIDE_OVERFLOW(a, b) \ + _GL_BINARY_OP_OVERFLOW (a, b, _GL_DIVIDE_OVERFLOW) +#define INT_REMAINDER_OVERFLOW(a, b) \ + _GL_BINARY_OP_OVERFLOW (a, b, _GL_REMAINDER_OVERFLOW) +#define INT_LEFT_SHIFT_OVERFLOW(a, b) \ + INT_LEFT_SHIFT_RANGE_OVERFLOW (a, b, \ + _GL_INT_MINIMUM (a), _GL_INT_MAXIMUM (a)) + +/* Return 1 if the expression A <op> B would overflow, + where OP_RESULT_OVERFLOW (A, B, MIN, MAX) does the actual test, + assuming MIN and MAX are the minimum and maximum for the result type. + Arguments should be free of side effects. */ +#define _GL_BINARY_OP_OVERFLOW(a, b, op_result_overflow) \ + op_result_overflow (a, b, \ + _GL_INT_MINIMUM (0 * (b) + (a)), \ + _GL_INT_MAXIMUM (0 * (b) + (a))) + +/* Store the low-order bits of A + B, A - B, A * B, respectively, into *R. + Return 1 if the result overflows. See above for restrictions. */ +#define INT_ADD_WRAPV(a, b, r) \ + _GL_INT_OP_WRAPV (a, b, r, +, __builtin_add_overflow, INT_ADD_OVERFLOW) +#define INT_SUBTRACT_WRAPV(a, b, r) \ + _GL_INT_OP_WRAPV (a, b, r, -, __builtin_sub_overflow, INT_SUBTRACT_OVERFLOW) +#define INT_MULTIPLY_WRAPV(a, b, r) \ + _GL_INT_OP_WRAPV (a, b, r, *, __builtin_mul_overflow, INT_MULTIPLY_OVERFLOW) + +/* Nonzero if this compiler has GCC bug 68193 or Clang bug 25390. See: + https://gcc.gnu.org/bugzilla/show_bug.cgi?id=68193 + https://llvm.org/bugs/show_bug.cgi?id=25390 + For now, assume all versions of GCC-like compilers generate bogus + warnings for _Generic. This matters only for older compilers that + lack __builtin_add_overflow. */ +#if __GNUC__ +# define _GL__GENERIC_BOGUS 1 +#else +# define _GL__GENERIC_BOGUS 0 +#endif + +/* Store the low-order bits of A <op> B into *R, where OP specifies + the operation. BUILTIN is the builtin operation, and OVERFLOW the + overflow predicate. Return 1 if the result overflows. See above + for restrictions. */ +#if _GL_HAS_BUILTIN_OVERFLOW +# define _GL_INT_OP_WRAPV(a, b, r, op, builtin, overflow) builtin (a, b, r) +#elif 201112 <= __STDC_VERSION__ && !_GL__GENERIC_BOGUS +# define _GL_INT_OP_WRAPV(a, b, r, op, builtin, overflow) \ + (_Generic \ + (*(r), \ + signed char: \ + _GL_INT_OP_CALC (a, b, r, op, overflow, unsigned char, \ + signed char, SCHAR_MIN, SCHAR_MAX), \ + short int: \ + _GL_INT_OP_CALC (a, b, r, op, overflow, unsigned short int, \ + short int, SHRT_MIN, SHRT_MAX), \ + int: \ + _GL_INT_OP_CALC (a, b, r, op, overflow, unsigned int, \ + int, INT_MIN, INT_MAX), \ + long int: \ + _GL_INT_OP_CALC (a, b, r, op, overflow, unsigned long int, \ + long int, LONG_MIN, LONG_MAX), \ + long long int: \ + _GL_INT_OP_CALC (a, b, r, op, overflow, unsigned long long int, \ + long long int, LLONG_MIN, LLONG_MAX))) +#else +# define _GL_INT_OP_WRAPV(a, b, r, op, builtin, overflow) \ + (sizeof *(r) == sizeof (signed char) \ + ? _GL_INT_OP_CALC (a, b, r, op, overflow, unsigned char, \ + signed char, SCHAR_MIN, SCHAR_MAX) \ + : sizeof *(r) == sizeof (short int) \ + ? _GL_INT_OP_CALC (a, b, r, op, overflow, unsigned short int, \ + short int, SHRT_MIN, SHRT_MAX) \ + : sizeof *(r) == sizeof (int) \ + ? _GL_INT_OP_CALC (a, b, r, op, overflow, unsigned int, \ + int, INT_MIN, INT_MAX) \ + : _GL_INT_OP_WRAPV_LONGISH(a, b, r, op, overflow)) +# ifdef LLONG_MAX +# define _GL_INT_OP_WRAPV_LONGISH(a, b, r, op, overflow) \ + (sizeof *(r) == sizeof (long int) \ + ? _GL_INT_OP_CALC (a, b, r, op, overflow, unsigned long int, \ + long int, LONG_MIN, LONG_MAX) \ + : _GL_INT_OP_CALC (a, b, r, op, overflow, unsigned long long int, \ + long long int, LLONG_MIN, LLONG_MAX)) +# else +# define _GL_INT_OP_WRAPV_LONGISH(a, b, r, op, overflow) \ + _GL_INT_OP_CALC (a, b, r, op, overflow, unsigned long int, \ + long int, LONG_MIN, LONG_MAX) +# endif +#endif + +/* Store the low-order bits of A <op> B into *R, where the operation + is given by OP. Use the unsigned type UT for calculation to avoid + overflow problems. *R's type is T, with extremal values TMIN and + TMAX. T must be a signed integer type. Return 1 if the result + overflows. */ +#define _GL_INT_OP_CALC(a, b, r, op, overflow, ut, t, tmin, tmax) \ + (sizeof ((a) op (b)) < sizeof (t) \ + ? _GL_INT_OP_CALC1 ((t) (a), (t) (b), r, op, overflow, ut, t, tmin, tmax) \ + : _GL_INT_OP_CALC1 (a, b, r, op, overflow, ut, t, tmin, tmax)) +#define _GL_INT_OP_CALC1(a, b, r, op, overflow, ut, t, tmin, tmax) \ + ((overflow (a, b) \ + || (EXPR_SIGNED ((a) op (b)) && ((a) op (b)) < (tmin)) \ + || (tmax) < ((a) op (b))) \ + ? (*(r) = _GL_INT_OP_WRAPV_VIA_UNSIGNED (a, b, op, ut, t, tmin, tmax), 1) \ + : (*(r) = _GL_INT_OP_WRAPV_VIA_UNSIGNED (a, b, op, ut, t, tmin, tmax), 0)) + +/* Return A <op> B, where the operation is given by OP. Use the + unsigned type UT for calculation to avoid overflow problems. + Convert the result to type T without overflow by subtracting TMIN + from large values before converting, and adding it afterwards. + Compilers can optimize all the operations except OP. */ +#define _GL_INT_OP_WRAPV_VIA_UNSIGNED(a, b, op, ut, t, tmin, tmax) \ + (((ut) (a) op (ut) (b)) <= (tmax) \ + ? (t) ((ut) (a) op (ut) (b)) \ + : ((t) (((ut) (a) op (ut) (b)) - (tmin)) + (tmin))) + +#endif /* _GL_INTPROPS_H */ diff --git a/support/localeinfo.c b/support/localeinfo.c new file mode 100644 index 00000000..ca96afc7 --- /dev/null +++ b/support/localeinfo.c @@ -0,0 +1,113 @@ +/* locale information + + Copyright 2016 Free Software Foundation, Inc. + + This program is free software; you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation; either version 3, or (at your option) + any later version. + + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program; if not, write to the Free Software + Foundation, Inc., 51 Franklin Street - Fifth Floor, Boston, MA + 02110-1301, USA. */ + +/* Written by Paul Eggert. */ + +#include <config.h> + +#include <localeinfo.h> + +#include <verify.h> + +#include <limits.h> +#include <locale.h> +#include <stdlib.h> +#include <string.h> +#include <wctype.h> + +/* The sbclen implementation relies on this. */ +verify (MB_LEN_MAX <= SCHAR_MAX); + +/* Return true if the locale uses UTF-8. */ + +static bool +is_using_utf8 (void) +{ + wchar_t wc; + mbstate_t mbs = {0}; + return mbrtowc (&wc, "\xc4\x80", 2, &mbs) == 2 && wc == 0x100; +} + +/* Initialize *LOCALEINFO from the current locale. */ + +void +init_localeinfo (struct localeinfo *localeinfo) +{ + int i; + + localeinfo->multibyte = MB_CUR_MAX > 1; + localeinfo->using_utf8 = is_using_utf8 (); + + for (i = CHAR_MIN; i <= CHAR_MAX; i++) + { + char c = i; + unsigned char uc = i; + mbstate_t s = {0}; + wchar_t wc; + size_t len = mbrtowc (&wc, &c, 1, &s); + localeinfo->sbclen[uc] = len <= 1 ? 1 : - (int) - len; + localeinfo->sbctowc[uc] = len <= 1 ? wc : WEOF; + } +} + +/* The set of wchar_t values C such that there's a useful locale + somewhere where C != towupper (C) && C != towlower (towupper (C)). + For example, 0x00B5 (U+00B5 MICRO SIGN) is in this table, because + towupper (0x00B5) == 0x039C (U+039C GREEK CAPITAL LETTER MU), and + towlower (0x039C) == 0x03BC (U+03BC GREEK SMALL LETTER MU). */ +static short const lonesome_lower[] = + { + 0x00B5, 0x0131, 0x017F, 0x01C5, 0x01C8, 0x01CB, 0x01F2, 0x0345, + 0x03C2, 0x03D0, 0x03D1, 0x03D5, 0x03D6, 0x03F0, 0x03F1, + + /* U+03F2 GREEK LUNATE SIGMA SYMBOL lacks a specific uppercase + counterpart in locales predating Unicode 4.0.0 (April 2003). */ + 0x03F2, + + 0x03F5, 0x1E9B, 0x1FBE, + }; + +/* Verify that the worst case fits. This is 1 for towupper, 1 for + towlower, and 1 for each entry in LONESOME_LOWER. */ +verify (1 + 1 + sizeof lonesome_lower / sizeof *lonesome_lower + <= CASE_FOLDED_BUFSIZE); + +/* Find the characters equal to C after case-folding, other than C + itself, and store them into FOLDED. Return the number of characters + stored. */ + +int +case_folded_counterparts (wchar_t c, wchar_t folded[CASE_FOLDED_BUFSIZE]) +{ + int i; + int n = 0; + wint_t uc = towupper (c); + wint_t lc = towlower (uc); + if (uc != c) + folded[n++] = uc; + if (lc != uc && lc != c && towupper (lc) == uc) + folded[n++] = lc; + for (i = 0; i < sizeof lonesome_lower / sizeof *lonesome_lower; i++) + { + wint_t li = lonesome_lower[i]; + if (li != lc && li != uc && li != c && towupper (li) == uc) + folded[n++] = li; + } + return n; +} diff --git a/support/localeinfo.h b/support/localeinfo.h new file mode 100644 index 00000000..cf2f9a69 --- /dev/null +++ b/support/localeinfo.h @@ -0,0 +1,54 @@ +/* locale information + + Copyright 2016 Free Software Foundation, Inc. + + This program is free software; you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation; either version 3, or (at your option) + any later version. + + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program; if not, write to the Free Software + Foundation, Inc., 51 Franklin Street - Fifth Floor, Boston, MA + 02110-1301, USA. */ + +/* Written by Paul Eggert. */ + +#include <limits.h> +#include <stdbool.h> +#include <wchar.h> + +struct localeinfo +{ + /* MB_CUR_MAX > 1. */ + bool multibyte; + + /* The locale uses UTF-8. */ + bool using_utf8; + + /* An array indexed by byte values B that contains 1 if B is a + single-byte character, -1 if B is an encoding error, and -2 if B + is the leading byte of a multibyte character that contains more + than one byte. */ + signed char sbclen[UCHAR_MAX + 1]; + + /* An array indexed by byte values B that contains the corresponding + wide character (if any) for B if sbclen[B] == 1. WEOF means the + byte is not a valid single-byte character, i.e., sbclen[B] == -1 + or -2. */ + wint_t sbctowc[UCHAR_MAX + 1]; +}; + +extern void init_localeinfo (struct localeinfo *); + +/* Maximum number of characters that can be the case-folded + counterparts of a single character, not counting the character + itself. This is a generous upper bound. */ +enum { CASE_FOLDED_BUFSIZE = 32 }; + +extern int case_folded_counterparts (wchar_t, wchar_t[CASE_FOLDED_BUFSIZE]); diff --git a/support/random.c b/support/random.c new file mode 100644 index 00000000..cba1b6bc --- /dev/null +++ b/support/random.c @@ -0,0 +1,542 @@ +/* + * Copyright (c) 1983, 1993 + * The Regents of the University of California. All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions + * are met: + * 1. Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * 2. Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * 3. Neither the name of the University nor the names of its contributors + * may be used to endorse or promote products derived from this software + * without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND + * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE + * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE + * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL + * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS + * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) + * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT + * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY + * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF + * SUCH DAMAGE. + */ + +/* + * Per the statement at http://opensource.org/licenses/bsd-license.php, + * + * The advertising clause in the license appearing on BSD Unix files was + * officially rescinded by the Director of the Office of Technology + * Licensing of the University of California on July 22 1999. He states + * that clause 3 is "hereby deleted in its entirety." + * + * I removed the advertising clause in the above copyright. + * The above web site points to + * ftp://ftp.cs.berkeley.edu/pub/4bsd/README.Impt.License.Change. + * + * Arnold Robbins + * 15 September 2007 + */ + +#if defined(LIBC_SCCS) && !defined(lint) +static const char sccsid[] = "@(#)random.c 8.2 (Berkeley) 5/19/95"; +#endif /* LIBC_SCCS and not lint */ + +#ifdef HAVE_CONFIG_H /* gawk addition */ +#include <config.h> +#endif + +#ifdef HAVE_FCNTL_H +#include <fcntl.h> +#endif +#include <stdio.h> +#include <stdlib.h> +#ifdef HAVE_UNISTD_H +#include <unistd.h> +#endif + +#include "random.h" /* gawk addition */ + +#ifdef HAVE_SYS_TIME_H /* gawk addition */ +#include <sys/time.h> +#endif + +#if 0 +#include <sys/cdefs.h> +__FBSDID("$FreeBSD: /repoman/r/ncvs/src/lib/libc/stdlib/random.c,v 1.24 2004/01/20 03:02:18 das Exp $"); + +#include "namespace.h" +#include <sys/time.h> /* for srandomdev() */ +#include <fcntl.h> /* for srandomdev() */ +#include <stdint.h> +#include <stdio.h> +#include <stdlib.h> +#include <unistd.h> /* for srandomdev() */ +#include "un-namespace.h" +#endif + +/* + * random.c: + * + * An improved random number generation package. In addition to the standard + * rand()/srand() like interface, this package also has a special state info + * interface. The initstate() routine is called with a seed, an array of + * bytes, and a count of how many bytes are being passed in; this array is + * then initialized to contain information for random number generation with + * that much state information. Good sizes for the amount of state + * information are 32, 64, 128, and 256 bytes. The state can be switched by + * calling the setstate() routine with the same array as was initiallized + * with initstate(). By default, the package runs with 128 bytes of state + * information and generates far better random numbers than a linear + * congruential generator. If the amount of state information is less than + * 32 bytes, a simple linear congruential R.N.G. is used. + * + * Internally, the state information is treated as an array of uint32_t's; the + * zeroeth element of the array is the type of R.N.G. being used (small + * integer); the remainder of the array is the state information for the + * R.N.G. Thus, 32 bytes of state information will give 7 ints worth of + * state information, which will allow a degree seven polynomial. (Note: + * the zeroeth word of state information also has some other information + * stored in it -- see setstate() for details). + * + * The random number generation technique is a linear feedback shift register + * approach, employing trinomials (since there are fewer terms to sum up that + * way). In this approach, the least significant bit of all the numbers in + * the state table will act as a linear feedback shift register, and will + * have period 2^deg - 1 (where deg is the degree of the polynomial being + * used, assuming that the polynomial is irreducible and primitive). The + * higher order bits will have longer periods, since their values are also + * influenced by pseudo-random carries out of the lower bits. The total + * period of the generator is approximately deg*(2**deg - 1); thus doubling + * the amount of state information has a vast influence on the period of the + * generator. Note: the deg*(2**deg - 1) is an approximation only good for + * large deg, when the period of the shift is the dominant factor. + * With deg equal to seven, the period is actually much longer than the + * 7*(2**7 - 1) predicted by this formula. + * + * Modified 28 December 1994 by Jacob S. Rosenberg. + * The following changes have been made: + * All references to the type u_int have been changed to unsigned long. + * All references to type int have been changed to type long. Other + * cleanups have been made as well. A warning for both initstate and + * setstate has been inserted to the effect that on Sparc platforms + * the 'arg_state' variable must be forced to begin on word boundaries. + * This can be easily done by casting a long integer array to char *. + * The overall logic has been left STRICTLY alone. This software was + * tested on both a VAX and Sun SpacsStation with exactly the same + * results. The new version and the original give IDENTICAL results. + * The new version is somewhat faster than the original. As the + * documentation says: "By default, the package runs with 128 bytes of + * state information and generates far better random numbers than a linear + * congruential generator. If the amount of state information is less than + * 32 bytes, a simple linear congruential R.N.G. is used." For a buffer of + * 128 bytes, this new version runs about 19 percent faster and for a 16 + * byte buffer it is about 5 percent faster. + */ + +/* + * For each of the currently supported random number generators, we have a + * break value on the amount of state information (you need at least this + * many bytes of state info to support this random number generator), a degree + * for the polynomial (actually a trinomial) that the R.N.G. is based on, and + * the separation between the two lower order coefficients of the trinomial. + */ +#define TYPE_0 0 /* linear congruential */ +#define BREAK_0 8 +#define DEG_0 0 +#define SEP_0 0 + +#define TYPE_1 1 /* x**7 + x**3 + 1 */ +#define BREAK_1 32 +#define DEG_1 7 +#define SEP_1 3 + +#define TYPE_2 2 /* x**15 + x + 1 */ +#define BREAK_2 64 +#define DEG_2 15 +#define SEP_2 1 + +#define TYPE_3 3 /* x**31 + x**3 + 1 */ +#define BREAK_3 128 +#define DEG_3 31 +#define SEP_3 3 + +#define TYPE_4 4 /* x**63 + x + 1 */ +#define BREAK_4 256 +#define DEG_4 63 +#define SEP_4 1 + +/* + * Array versions of the above information to make code run faster -- + * relies on fact that TYPE_i == i. + */ +#define MAX_TYPES 5 /* max number of types above */ + +#ifdef USE_WEAK_SEEDING +#define NSHUFF 0 +#else /* !USE_WEAK_SEEDING */ +#define NSHUFF 50 /* to drop some "seed -> 1st value" linearity */ +#endif /* !USE_WEAK_SEEDING */ + +static const int degrees[MAX_TYPES] = { DEG_0, DEG_1, DEG_2, DEG_3, DEG_4 }; +static const int seps [MAX_TYPES] = { SEP_0, SEP_1, SEP_2, SEP_3, SEP_4 }; + +/* + * Initially, everything is set up as if from: + * + * initstate(1, randtbl, 128); + * + * Note that this initialization takes advantage of the fact that srandom() + * advances the front and rear pointers 10*rand_deg times, and hence the + * rear pointer which starts at 0 will also end up at zero; thus the zeroeth + * element of the state information, which contains info about the current + * position of the rear pointer is just + * + * MAX_TYPES * (rptr - state) + TYPE_3 == TYPE_3. + */ + +static uint32_t randtbl[DEG_3 + 1] = { + TYPE_3, +#ifdef USE_WEAK_SEEDING +/* Historic implementation compatibility */ +/* The random sequences do not vary much with the seed */ + 0x9a319039, 0x32d9c024, 0x9b663182, 0x5da1f342, 0xde3b81e0, 0xdf0a6fb5, + 0xf103bc02, 0x48f340fb, 0x7449e56b, 0xbeb1dbb0, 0xab5c5918, 0x946554fd, + 0x8c2e680f, 0xeb3d799f, 0xb11ee0b7, 0x2d436b86, 0xda672e2a, 0x1588ca88, + 0xe369735d, 0x904f35f7, 0xd7158fd6, 0x6fa6f051, 0x616e6b96, 0xac94efdc, + 0x36413f93, 0xc622c298, 0xf5a42ab8, 0x8a88d77b, 0xf5ad9d0e, 0x8999220b, + 0x27fb47b9, +#else /* !USE_WEAK_SEEDING */ + 0x991539b1, 0x16a5bce3, 0x6774a4cd, 0x3e01511e, 0x4e508aaa, 0x61048c05, + 0xf5500617, 0x846b7115, 0x6a19892c, 0x896a97af, 0xdb48f936, 0x14898454, + 0x37ffd106, 0xb58bff9c, 0x59e17104, 0xcf918a49, 0x09378c83, 0x52c7a471, + 0x8d293ea9, 0x1f4fc301, 0xc3db71be, 0x39b44e1c, 0xf8a44ef9, 0x4c8b80b1, + 0x19edc328, 0x87bf4bdd, 0xc9b240e5, 0xe9ee4b1b, 0x4382aee7, 0x535b6b41, + 0xf3bec5da +#endif /* !USE_WEAK_SEEDING */ +}; + +/* + * fptr and rptr are two pointers into the state info, a front and a rear + * pointer. These two pointers are always rand_sep places aparts, as they + * cycle cyclically through the state information. (Yes, this does mean we + * could get away with just one pointer, but the code for random() is more + * efficient this way). The pointers are left positioned as they would be + * from the call + * + * initstate(1, randtbl, 128); + * + * (The position of the rear pointer, rptr, is really 0 (as explained above + * in the initialization of randtbl) because the state table pointer is set + * to point to randtbl[1] (as explained below). + */ +static uint32_t *fptr = &randtbl[SEP_3 + 1]; +static uint32_t *rptr = &randtbl[1]; + +/* + * The following things are the pointer to the state information table, the + * type of the current generator, the degree of the current polynomial being + * used, and the separation between the two pointers. Note that for efficiency + * of random(), we remember the first location of the state information, not + * the zeroeth. Hence it is valid to access state[-1], which is used to + * store the type of the R.N.G. Also, we remember the last location, since + * this is more efficient than indexing every time to find the address of + * the last element to see if the front and rear pointers have wrapped. + */ +static uint32_t *state = &randtbl[1]; +static int rand_type = TYPE_3; +static int rand_deg = DEG_3; +static int rand_sep = SEP_3; +static uint32_t *end_ptr = &randtbl[DEG_3 + 1]; + +static inline uint32_t good_rand(int32_t); + +static inline uint32_t good_rand (x) + int32_t x; +{ +#ifdef USE_WEAK_SEEDING +/* + * Historic implementation compatibility. + * The random sequences do not vary much with the seed, + * even with overflowing. + */ + return (1103515245 * x + 12345); +#else /* !USE_WEAK_SEEDING */ +/* + * Compute x = (7^5 * x) mod (2^31 - 1) + * wihout overflowing 31 bits: + * (2^31 - 1) = 127773 * (7^5) + 2836 + * From "Random number generators: good ones are hard to find", + * Park and Miller, Communications of the ACM, vol. 31, no. 10, + * October 1988, p. 1195. + */ + int32_t hi, lo; + + /* Can't be initialized with 0, so use another value. */ + if (x == 0) + x = 123459876; + hi = x / 127773; + lo = x % 127773; + x = 16807 * lo - 2836 * hi; + if (x < 0) + x += 0x7fffffff; + return (x); +#endif /* !USE_WEAK_SEEDING */ +} + +/* + * srandom: + * + * Initialize the random number generator based on the given seed. If the + * type is the trivial no-state-information type, just remember the seed. + * Otherwise, initializes state[] based on the given "seed" via a linear + * congruential generator. Then, the pointers are set to known locations + * that are exactly rand_sep places apart. Lastly, it cycles the state + * information a given number of times to get rid of any initial dependencies + * introduced by the L.C.R.N.G. Note that the initialization of randtbl[] + * for default usage relies on values produced by this routine. + */ +void +srandom(x) + unsigned long x; +{ + int i, lim; + + state[0] = (uint32_t)x; + if (rand_type == TYPE_0) + lim = NSHUFF; + else { + for (i = 1; i < rand_deg; i++) + state[i] = good_rand(state[i - 1]); + fptr = &state[rand_sep]; + rptr = &state[0]; + lim = 10 * rand_deg; + } + for (i = 0; i < lim; i++) + (void)random(); +} + +#if 0 /* gawk doesn't use this */ +/* + * srandomdev: + * + * Many programs choose the seed value in a totally predictable manner. + * This often causes problems. We seed the generator using the much more + * secure random(4) interface. Note that this particular seeding + * procedure can generate states which are impossible to reproduce by + * calling srandom() with any value, since the succeeding terms in the + * state buffer are no longer derived from the LC algorithm applied to + * a fixed seed. + */ +void +srandomdev() +{ + int fd, done; + size_t len; + + if (rand_type == TYPE_0) + len = sizeof state[0]; + else + len = rand_deg * sizeof state[0]; + + done = 0; + fd = open("/dev/random", O_RDONLY, 0); + if (fd >= 0) { + if (read(fd, (void *) state, len) == (ssize_t) len) + done = 1; + close(fd); + } + + if (!done) { + struct timeval tv; + unsigned long junk; + + gettimeofday(&tv, NULL); + srandom((getpid() << 16) ^ tv.tv_sec ^ tv.tv_usec ^ junk); + return; + } + + if (rand_type != TYPE_0) { + fptr = &state[rand_sep]; + rptr = &state[0]; + } +} +#endif + +/* + * initstate: + * + * Initialize the state information in the given array of n bytes for future + * random number generation. Based on the number of bytes we are given, and + * the break values for the different R.N.G.'s, we choose the best (largest) + * one we can and set things up for it. srandom() is then called to + * initialize the state information. + * + * Note that on return from srandom(), we set state[-1] to be the type + * multiplexed with the current value of the rear pointer; this is so + * successive calls to initstate() won't lose this information and will be + * able to restart with setstate(). + * + * Note: the first thing we do is save the current state, if any, just like + * setstate() so that it doesn't matter when initstate is called. + * + * Returns a pointer to the old state. + * + * Note: The Sparc platform requires that arg_state begin on an int + * word boundary; otherwise a bus error will occur. Even so, lint will + * complain about mis-alignment, but you should disregard these messages. + */ +char * +initstate(seed, arg_state, n) + unsigned long seed; /* seed for R.N.G. */ + char *arg_state; /* pointer to state array */ + long n; /* # bytes of state info */ +{ + char *ostate = (char *)(&state[-1]); + uint32_t *int_arg_state = (uint32_t *)arg_state; + + if (rand_type == TYPE_0) + state[-1] = rand_type; + else + state[-1] = MAX_TYPES * (rptr - state) + rand_type; + if (n < BREAK_0) { + (void)fprintf(stderr, + "random: not enough state (%ld bytes); ignored.\n", n); + return(0); + } + if (n < BREAK_1) { + rand_type = TYPE_0; + rand_deg = DEG_0; + rand_sep = SEP_0; + } else if (n < BREAK_2) { + rand_type = TYPE_1; + rand_deg = DEG_1; + rand_sep = SEP_1; + } else if (n < BREAK_3) { + rand_type = TYPE_2; + rand_deg = DEG_2; + rand_sep = SEP_2; + } else if (n < BREAK_4) { + rand_type = TYPE_3; + rand_deg = DEG_3; + rand_sep = SEP_3; + } else { + rand_type = TYPE_4; + rand_deg = DEG_4; + rand_sep = SEP_4; + } + state = int_arg_state + 1; /* first location */ + end_ptr = &state[rand_deg]; /* must set end_ptr before srandom */ + srandom(seed); + if (rand_type == TYPE_0) + int_arg_state[0] = rand_type; + else + int_arg_state[0] = MAX_TYPES * (rptr - state) + rand_type; + return(ostate); +} + +/* + * setstate: + * + * Restore the state from the given state array. + * + * Note: it is important that we also remember the locations of the pointers + * in the current state information, and restore the locations of the pointers + * from the old state information. This is done by multiplexing the pointer + * location into the zeroeth word of the state information. + * + * Note that due to the order in which things are done, it is OK to call + * setstate() with the same state as the current state. + * + * Returns a pointer to the old state information. + * + * Note: The Sparc platform requires that arg_state begin on an int + * word boundary; otherwise a bus error will occur. Even so, lint will + * complain about mis-alignment, but you should disregard these messages. + */ +char * +setstate(arg_state) + char *arg_state; /* pointer to state array */ +{ + uint32_t *new_state = (uint32_t *)arg_state; + uint32_t type = new_state[0] % MAX_TYPES; + uint32_t rear = new_state[0] / MAX_TYPES; + char *ostate = (char *)(&state[-1]); + + if (rand_type == TYPE_0) + state[-1] = rand_type; + else + state[-1] = MAX_TYPES * (rptr - state) + rand_type; + switch(type) { + case TYPE_0: + case TYPE_1: + case TYPE_2: + case TYPE_3: + case TYPE_4: + rand_type = type; + rand_deg = degrees[type]; + rand_sep = seps[type]; + break; + default: + (void)fprintf(stderr, + "random: state info corrupted; not changed.\n"); + } + state = new_state + 1; + if (rand_type != TYPE_0) { + rptr = &state[rear]; + fptr = &state[(rear + rand_sep) % rand_deg]; + } + end_ptr = &state[rand_deg]; /* set end_ptr too */ + return(ostate); +} + +/* + * random: + * + * If we are using the trivial TYPE_0 R.N.G., just do the old linear + * congruential bit. Otherwise, we do our fancy trinomial stuff, which is + * the same in all the other cases due to all the global variables that have + * been set up. The basic operation is to add the number at the rear pointer + * into the one at the front pointer. Then both pointers are advanced to + * the next location cyclically in the table. The value returned is the sum + * generated, reduced to 31 bits by throwing away the "least random" low bit. + * + * Note: the code takes advantage of the fact that both the front and + * rear pointers can't wrap on the same call by not testing the rear + * pointer if the front one has wrapped. + * + * Returns a 31-bit random number. + */ +long +random() +{ + uint32_t i; + uint32_t *f, *r; + + if (rand_type == TYPE_0) { + i = state[0]; + state[0] = i = (good_rand(i)) & 0x7fffffff; + } else { + /* + * Use local variables rather than static variables for speed. + */ + f = fptr; r = rptr; + *f += *r; + i = (*f >> 1) & 0x7fffffff; /* chucking least random bit */ + if (++f >= end_ptr) { + f = state; + ++r; + } + else if (++r >= end_ptr) { + r = state; + } + + fptr = f; rptr = r; + } + return((long)i); +} diff --git a/support/random.h b/support/random.h new file mode 100644 index 00000000..84b31414 --- /dev/null +++ b/support/random.h @@ -0,0 +1,43 @@ +/* + * random.h - redefine name of random lib routines to avoid conflicts + */ + +/* + * Copyright (C) 1996, 2001, 2004, 2005, 2013 the Free Software Foundation, Inc. + * + * This file is part of GAWK, the GNU implementation of the + * AWK Programming Language. + * + * GAWK is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 3 of the License, or + * (at your option) any later version. + * + * GAWK is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA + */ + +#define initstate gawk_initstate +#define setstate gawk_setstate +#define random gawk_random +#define srandom gawk_srandom + +#if SIZEOF_UNSIGNED_INT == 4 +typedef unsigned int gawk_uint32_t; +typedef int gawk_int32_t; +#else +#if SIZEOF_UNSIGNED_LONG == 4 +typedef unsigned long gawk_uint32_t; +typedef long gawk_int32_t; +#endif +#endif +#define uint32_t gawk_uint32_t +#define int32_t gawk_int32_t + +extern long random (void); diff --git a/support/regcomp.c b/support/regcomp.c new file mode 100644 index 00000000..5ac53701 --- /dev/null +++ b/support/regcomp.c @@ -0,0 +1,3939 @@ +/* Extended regular expression matching and search library. + Copyright (C) 2002-2016 Free Software Foundation, Inc. + This file is part of the GNU C Library. + Contributed by Isamu Hasegawa <isamu@yamato.ibm.com>. + + The GNU C Library is free software; you can redistribute it and/or + modify it under the terms of the GNU Lesser General Public + License as published by the Free Software Foundation; either + version 2.1 of the License, or (at your option) any later version. + + The GNU C Library is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public + License along with the GNU C Library; if not, see + <http://www.gnu.org/licenses/>. */ + +#ifdef HAVE_STDINT_H +#include <stdint.h> +#endif + +#ifdef HAVE_STRINGS_H +#include <strings.h> +#endif + +#ifdef _LIBC +# include <locale/weight.h> +#endif + +static reg_errcode_t re_compile_internal (regex_t *preg, const char * pattern, + size_t length, reg_syntax_t syntax); +static void re_compile_fastmap_iter (regex_t *bufp, + const re_dfastate_t *init_state, + char *fastmap); +static reg_errcode_t init_dfa (re_dfa_t *dfa, size_t pat_len); +#ifdef RE_ENABLE_I18N +static void free_charset (re_charset_t *cset); +#endif /* RE_ENABLE_I18N */ +static void free_workarea_compile (regex_t *preg); +static reg_errcode_t create_initial_state (re_dfa_t *dfa); +#ifdef RE_ENABLE_I18N +static void optimize_utf8 (re_dfa_t *dfa); +#endif +static reg_errcode_t analyze (regex_t *preg); +static reg_errcode_t preorder (bin_tree_t *root, + reg_errcode_t (fn (void *, bin_tree_t *)), + void *extra); +static reg_errcode_t postorder (bin_tree_t *root, + reg_errcode_t (fn (void *, bin_tree_t *)), + void *extra); +static reg_errcode_t optimize_subexps (void *extra, bin_tree_t *node); +static reg_errcode_t lower_subexps (void *extra, bin_tree_t *node); +static bin_tree_t *lower_subexp (reg_errcode_t *err, regex_t *preg, + bin_tree_t *node); +static reg_errcode_t calc_first (void *extra, bin_tree_t *node); +static reg_errcode_t calc_next (void *extra, bin_tree_t *node); +static reg_errcode_t link_nfa_nodes (void *extra, bin_tree_t *node); +static int duplicate_node (re_dfa_t *dfa, int org_idx, unsigned int constraint); +static int search_duplicated_node (const re_dfa_t *dfa, int org_node, + unsigned int constraint); +static reg_errcode_t calc_eclosure (re_dfa_t *dfa); +static reg_errcode_t calc_eclosure_iter (re_node_set *new_set, re_dfa_t *dfa, + int node, int root); +static reg_errcode_t calc_inveclosure (re_dfa_t *dfa); +static int fetch_number (re_string_t *input, re_token_t *token, + reg_syntax_t syntax); +static int peek_token (re_token_t *token, re_string_t *input, + reg_syntax_t syntax) internal_function; +static bin_tree_t *parse (re_string_t *regexp, regex_t *preg, + reg_syntax_t syntax, reg_errcode_t *err); +static bin_tree_t *parse_reg_exp (re_string_t *regexp, regex_t *preg, + re_token_t *token, reg_syntax_t syntax, + int nest, reg_errcode_t *err); +static bin_tree_t *parse_branch (re_string_t *regexp, regex_t *preg, + re_token_t *token, reg_syntax_t syntax, + int nest, reg_errcode_t *err); +static bin_tree_t *parse_expression (re_string_t *regexp, regex_t *preg, + re_token_t *token, reg_syntax_t syntax, + int nest, reg_errcode_t *err); +static bin_tree_t *parse_sub_exp (re_string_t *regexp, regex_t *preg, + re_token_t *token, reg_syntax_t syntax, + int nest, reg_errcode_t *err); +static bin_tree_t *parse_dup_op (bin_tree_t *dup_elem, re_string_t *regexp, + re_dfa_t *dfa, re_token_t *token, + reg_syntax_t syntax, reg_errcode_t *err); +static bin_tree_t *parse_bracket_exp (re_string_t *regexp, re_dfa_t *dfa, + re_token_t *token, reg_syntax_t syntax, + reg_errcode_t *err); +static reg_errcode_t parse_bracket_element (bracket_elem_t *elem, + re_string_t *regexp, + re_token_t *token, int token_len, + re_dfa_t *dfa, + reg_syntax_t syntax, + int accept_hyphen); +static reg_errcode_t parse_bracket_symbol (bracket_elem_t *elem, + re_string_t *regexp, + re_token_t *token); +#ifdef RE_ENABLE_I18N +static reg_errcode_t build_equiv_class (bitset_t sbcset, + re_charset_t *mbcset, + int *equiv_class_alloc, + const unsigned char *name); +static reg_errcode_t build_charclass (RE_TRANSLATE_TYPE trans, + bitset_t sbcset, + re_charset_t *mbcset, + int *char_class_alloc, + const char *class_name, + reg_syntax_t syntax); +#else /* not RE_ENABLE_I18N */ +static reg_errcode_t build_equiv_class (bitset_t sbcset, + const unsigned char *name); +static reg_errcode_t build_charclass (RE_TRANSLATE_TYPE trans, + bitset_t sbcset, + const char *class_name, + reg_syntax_t syntax); +#endif /* not RE_ENABLE_I18N */ +static bin_tree_t *build_charclass_op (re_dfa_t *dfa, + RE_TRANSLATE_TYPE trans, + const char *class_name, + const char *extra, + int non_match, reg_errcode_t *err); +static bin_tree_t *create_tree (re_dfa_t *dfa, + bin_tree_t *left, bin_tree_t *right, + re_token_type_t type); +static bin_tree_t *create_token_tree (re_dfa_t *dfa, + bin_tree_t *left, bin_tree_t *right, + const re_token_t *token); +static bin_tree_t *duplicate_tree (const bin_tree_t *src, re_dfa_t *dfa); +static void free_token (re_token_t *node); +static reg_errcode_t free_tree (void *extra, bin_tree_t *node); +static reg_errcode_t mark_opt_subexp (void *extra, bin_tree_t *node); + +/* This table gives an error message for each of the error codes listed + in regex.h. Obviously the order here has to be same as there. + POSIX doesn't require that we do anything for REG_NOERROR, + but why not be nice? */ + +static const char __re_error_msgid[] = + { +#define REG_NOERROR_IDX 0 + gettext_noop ("Success") /* REG_NOERROR */ + "\0" +#define REG_NOMATCH_IDX (REG_NOERROR_IDX + sizeof "Success") + gettext_noop ("No match") /* REG_NOMATCH */ + "\0" +#define REG_BADPAT_IDX (REG_NOMATCH_IDX + sizeof "No match") + gettext_noop ("Invalid regular expression") /* REG_BADPAT */ + "\0" +#define REG_ECOLLATE_IDX (REG_BADPAT_IDX + sizeof "Invalid regular expression") + gettext_noop ("Invalid collation character") /* REG_ECOLLATE */ + "\0" +#define REG_ECTYPE_IDX (REG_ECOLLATE_IDX + sizeof "Invalid collation character") + gettext_noop ("Invalid character class name") /* REG_ECTYPE */ + "\0" +#define REG_EESCAPE_IDX (REG_ECTYPE_IDX + sizeof "Invalid character class name") + gettext_noop ("Trailing backslash") /* REG_EESCAPE */ + "\0" +#define REG_ESUBREG_IDX (REG_EESCAPE_IDX + sizeof "Trailing backslash") + gettext_noop ("Invalid back reference") /* REG_ESUBREG */ + "\0" +#define REG_EBRACK_IDX (REG_ESUBREG_IDX + sizeof "Invalid back reference") + gettext_noop ("Unmatched [, [^, [:, [., or [=") /* REG_EBRACK */ + "\0" +#define REG_EPAREN_IDX (REG_EBRACK_IDX + sizeof "Unmatched [, [^, [:, [., or [=") + gettext_noop ("Unmatched ( or \\(") /* REG_EPAREN */ + "\0" +#define REG_EBRACE_IDX (REG_EPAREN_IDX + sizeof "Unmatched ( or \\(") + gettext_noop ("Unmatched \\{") /* REG_EBRACE */ + "\0" +#define REG_BADBR_IDX (REG_EBRACE_IDX + sizeof "Unmatched \\{") + gettext_noop ("Invalid content of \\{\\}") /* REG_BADBR */ + "\0" +#define REG_ERANGE_IDX (REG_BADBR_IDX + sizeof "Invalid content of \\{\\}") + gettext_noop ("Invalid range end") /* REG_ERANGE */ + "\0" +#define REG_ESPACE_IDX (REG_ERANGE_IDX + sizeof "Invalid range end") + gettext_noop ("Memory exhausted") /* REG_ESPACE */ + "\0" +#define REG_BADRPT_IDX (REG_ESPACE_IDX + sizeof "Memory exhausted") + gettext_noop ("Invalid preceding regular expression") /* REG_BADRPT */ + "\0" +#define REG_EEND_IDX (REG_BADRPT_IDX + sizeof "Invalid preceding regular expression") + gettext_noop ("Premature end of regular expression") /* REG_EEND */ + "\0" +#define REG_ESIZE_IDX (REG_EEND_IDX + sizeof "Premature end of regular expression") + gettext_noop ("Regular expression too big") /* REG_ESIZE */ + "\0" +#define REG_ERPAREN_IDX (REG_ESIZE_IDX + sizeof "Regular expression too big") + gettext_noop ("Unmatched ) or \\)") /* REG_ERPAREN */ + }; + +static const size_t __re_error_msgid_idx[] = + { + REG_NOERROR_IDX, + REG_NOMATCH_IDX, + REG_BADPAT_IDX, + REG_ECOLLATE_IDX, + REG_ECTYPE_IDX, + REG_EESCAPE_IDX, + REG_ESUBREG_IDX, + REG_EBRACK_IDX, + REG_EPAREN_IDX, + REG_EBRACE_IDX, + REG_BADBR_IDX, + REG_ERANGE_IDX, + REG_ESPACE_IDX, + REG_BADRPT_IDX, + REG_EEND_IDX, + REG_ESIZE_IDX, + REG_ERPAREN_IDX + }; + +/* Entry points for GNU code. */ + + +#ifndef HAVE_BTOWC +wchar_t +btowc (int c) +{ + wchar_t wtmp[2]; + char tmp[2]; + mbstate_t mbs; + + memset(& mbs, 0, sizeof(mbs)); + tmp[0] = c; + tmp[1] = 0; + + mbrtowc (wtmp, tmp, 1, & mbs); + return wtmp[0]; +} +#endif + +/* re_compile_pattern is the GNU regular expression compiler: it + compiles PATTERN (of length LENGTH) and puts the result in BUFP. + Returns 0 if the pattern was valid, otherwise an error string. + + Assumes the 'allocated' (and perhaps 'buffer') and 'translate' fields + are set in BUFP on entry. */ + +const char * +re_compile_pattern (const char *pattern, size_t length, + struct re_pattern_buffer *bufp) +{ + reg_errcode_t ret; + + /* And GNU code determines whether or not to get register information + by passing null for the REGS argument to re_match, etc., not by + setting no_sub, unless RE_NO_SUB is set. */ + bufp->no_sub = !!(re_syntax_options & RE_NO_SUB); + + /* Match anchors at newline. */ + bufp->newline_anchor = 1; + + ret = re_compile_internal (bufp, pattern, length, re_syntax_options); + + if (!ret) + return NULL; + return gettext (__re_error_msgid + __re_error_msgid_idx[(int) ret]); +} +#ifdef _LIBC +weak_alias (__re_compile_pattern, re_compile_pattern) +#endif + +/* Set by 're_set_syntax' to the current regexp syntax to recognize. Can + also be assigned to arbitrarily: each pattern buffer stores its own + syntax, so it can be changed between regex compilations. */ +/* This has no initializer because initialized variables in Emacs + become read-only after dumping. */ +reg_syntax_t re_syntax_options; + + +/* Specify the precise syntax of regexps for compilation. This provides + for compatibility for various utilities which historically have + different, incompatible syntaxes. + + The argument SYNTAX is a bit mask comprised of the various bits + defined in regex.h. We return the old syntax. */ + +reg_syntax_t +re_set_syntax (reg_syntax_t syntax) +{ + reg_syntax_t ret = re_syntax_options; + + re_syntax_options = syntax; + return ret; +} +#ifdef _LIBC +weak_alias (__re_set_syntax, re_set_syntax) +#endif + +int +re_compile_fastmap (struct re_pattern_buffer *bufp) +{ + re_dfa_t *dfa = (re_dfa_t *) bufp->buffer; + char *fastmap = bufp->fastmap; + + memset (fastmap, '\0', sizeof (char) * SBC_MAX); + re_compile_fastmap_iter (bufp, dfa->init_state, fastmap); + if (dfa->init_state != dfa->init_state_word) + re_compile_fastmap_iter (bufp, dfa->init_state_word, fastmap); + if (dfa->init_state != dfa->init_state_nl) + re_compile_fastmap_iter (bufp, dfa->init_state_nl, fastmap); + if (dfa->init_state != dfa->init_state_begbuf) + re_compile_fastmap_iter (bufp, dfa->init_state_begbuf, fastmap); + bufp->fastmap_accurate = 1; + return 0; +} +#ifdef _LIBC +weak_alias (__re_compile_fastmap, re_compile_fastmap) +#endif + +static inline void +__attribute__ ((always_inline)) +re_set_fastmap (char *fastmap, bool icase, int ch) +{ + fastmap[ch] = 1; + if (icase) + fastmap[tolower (ch)] = 1; +} + +/* Helper function for re_compile_fastmap. + Compile fastmap for the initial_state INIT_STATE. */ + +static void +re_compile_fastmap_iter (regex_t *bufp, const re_dfastate_t *init_state, + char *fastmap) +{ + re_dfa_t *dfa = (re_dfa_t *) bufp->buffer; + int node_cnt; + int icase = (dfa->mb_cur_max == 1 && (bufp->syntax & RE_ICASE)); + for (node_cnt = 0; node_cnt < init_state->nodes.nelem; ++node_cnt) + { + int node = init_state->nodes.elems[node_cnt]; + re_token_type_t type = dfa->nodes[node].type; + + if (type == CHARACTER) + { + re_set_fastmap (fastmap, icase, dfa->nodes[node].opr.c); +#ifdef RE_ENABLE_I18N + if ((bufp->syntax & RE_ICASE) && dfa->mb_cur_max > 1) + { + unsigned char buf[MB_LEN_MAX]; + unsigned char *p; + wchar_t wc; + mbstate_t state; + + p = buf; + *p++ = dfa->nodes[node].opr.c; + while (++node < dfa->nodes_len + && dfa->nodes[node].type == CHARACTER + && dfa->nodes[node].mb_partial) + *p++ = dfa->nodes[node].opr.c; + memset (&state, '\0', sizeof (state)); + if (__mbrtowc (&wc, (const char *) buf, p - buf, + &state) == p - buf + && (__wcrtomb ((char *) buf, towlower (wc), &state) + != (size_t) -1)) + re_set_fastmap (fastmap, 0, buf[0]); + } +#endif + } + else if (type == SIMPLE_BRACKET) + { + int i, ch; + for (i = 0, ch = 0; i < BITSET_WORDS; ++i) + { + int j; + bitset_word_t w = dfa->nodes[node].opr.sbcset[i]; + for (j = 0; j < BITSET_WORD_BITS; ++j, ++ch) + if (w & ((bitset_word_t) 1 << j)) + re_set_fastmap (fastmap, icase, ch); + } + } +#ifdef RE_ENABLE_I18N + else if (type == COMPLEX_BRACKET) + { + re_charset_t *cset = dfa->nodes[node].opr.mbcset; + int i; + +# ifdef _LIBC + /* See if we have to try all bytes which start multiple collation + elements. + e.g. In da_DK, we want to catch 'a' since "aa" is a valid + collation element, and don't catch 'b' since 'b' is + the only collation element which starts from 'b' (and + it is caught by SIMPLE_BRACKET). */ + if (_NL_CURRENT_WORD (LC_COLLATE, _NL_COLLATE_NRULES) != 0 + && (cset->ncoll_syms || cset->nranges)) + { + const int32_t *table = (const int32_t *) + _NL_CURRENT (LC_COLLATE, _NL_COLLATE_TABLEMB); + for (i = 0; i < SBC_MAX; ++i) + if (table[i] < 0) + re_set_fastmap (fastmap, icase, i); + } +# endif /* _LIBC */ + + /* See if we have to start the match at all multibyte characters, + i.e. where we would not find an invalid sequence. This only + applies to multibyte character sets; for single byte character + sets, the SIMPLE_BRACKET again suffices. */ + if (dfa->mb_cur_max > 1 + && (cset->nchar_classes || cset->non_match || cset->nranges +# ifdef _LIBC + || cset->nequiv_classes +# endif /* _LIBC */ + )) + { + unsigned char c = 0; + do + { + mbstate_t mbs; + memset (&mbs, 0, sizeof (mbs)); + if (__mbrtowc (NULL, (char *) &c, 1, &mbs) == (size_t) -2) + re_set_fastmap (fastmap, false, (int) c); + } + while (++c != 0); + } + + else + { + /* ... Else catch all bytes which can start the mbchars. */ + for (i = 0; i < cset->nmbchars; ++i) + { + char buf[256]; + mbstate_t state; + memset (&state, '\0', sizeof (state)); + if (__wcrtomb (buf, cset->mbchars[i], &state) != (size_t) -1) + re_set_fastmap (fastmap, icase, *(unsigned char *) buf); + if ((bufp->syntax & RE_ICASE) && dfa->mb_cur_max > 1) + { + if (__wcrtomb (buf, towlower (cset->mbchars[i]), &state) + != (size_t) -1) + re_set_fastmap (fastmap, false, *(unsigned char *) buf); + } + } + } + } +#endif /* RE_ENABLE_I18N */ + else if (type == OP_PERIOD +#ifdef RE_ENABLE_I18N + || type == OP_UTF8_PERIOD +#endif /* RE_ENABLE_I18N */ + || type == END_OF_RE) + { + memset (fastmap, '\1', sizeof (char) * SBC_MAX); + if (type == END_OF_RE) + bufp->can_be_null = 1; + return; + } + } +} + +/* Entry point for POSIX code. */ +/* regcomp takes a regular expression as a string and compiles it. + + PREG is a regex_t *. We do not expect any fields to be initialized, + since POSIX says we shouldn't. Thus, we set + + 'buffer' to the compiled pattern; + 'used' to the length of the compiled pattern; + 'syntax' to RE_SYNTAX_POSIX_EXTENDED if the + REG_EXTENDED bit in CFLAGS is set; otherwise, to + RE_SYNTAX_POSIX_BASIC; + 'newline_anchor' to REG_NEWLINE being set in CFLAGS; + 'fastmap' to an allocated space for the fastmap; + 'fastmap_accurate' to zero; + 're_nsub' to the number of subexpressions in PATTERN. + + PATTERN is the address of the pattern string. + + CFLAGS is a series of bits which affect compilation. + + If REG_EXTENDED is set, we use POSIX extended syntax; otherwise, we + use POSIX basic syntax. + + If REG_NEWLINE is set, then . and [^...] don't match newline. + Also, regexec will try a match beginning after every newline. + + If REG_ICASE is set, then we considers upper- and lowercase + versions of letters to be equivalent when matching. + + If REG_NOSUB is set, then when PREG is passed to regexec, that + routine will report only success or failure, and nothing about the + registers. + + It returns 0 if it succeeds, nonzero if it doesn't. (See regex.h for + the return codes and their meanings.) */ + +int +regcomp (regex_t *__restrict preg, const char *__restrict pattern, int cflags) +{ + reg_errcode_t ret; + reg_syntax_t syntax = ((cflags & REG_EXTENDED) ? RE_SYNTAX_POSIX_EXTENDED + : RE_SYNTAX_POSIX_BASIC); + + preg->buffer = NULL; + preg->allocated = 0; + preg->used = 0; + + /* Try to allocate space for the fastmap. */ + preg->fastmap = re_malloc (char, SBC_MAX); + if (BE (preg->fastmap == NULL, 0)) + return REG_ESPACE; + + syntax |= (cflags & REG_ICASE) ? RE_ICASE : 0; + + /* If REG_NEWLINE is set, newlines are treated differently. */ + if (cflags & REG_NEWLINE) + { /* REG_NEWLINE implies neither . nor [^...] match newline. */ + syntax &= ~RE_DOT_NEWLINE; + syntax |= RE_HAT_LISTS_NOT_NEWLINE; + /* It also changes the matching behavior. */ + preg->newline_anchor = 1; + } + else + preg->newline_anchor = 0; + preg->no_sub = !!(cflags & REG_NOSUB); + preg->translate = NULL; + + ret = re_compile_internal (preg, pattern, strlen (pattern), syntax); + + /* POSIX doesn't distinguish between an unmatched open-group and an + unmatched close-group: both are REG_EPAREN. */ + if (ret == REG_ERPAREN) + ret = REG_EPAREN; + + /* We have already checked preg->fastmap != NULL. */ + if (BE (ret == REG_NOERROR, 1)) + /* Compute the fastmap now, since regexec cannot modify the pattern + buffer. This function never fails in this implementation. */ + (void) re_compile_fastmap (preg); + else + { + /* Some error occurred while compiling the expression. */ + re_free (preg->fastmap); + preg->fastmap = NULL; + } + + return (int) ret; +} +#ifdef _LIBC +weak_alias (__regcomp, regcomp) +#endif + +/* Returns a message corresponding to an error code, ERRCODE, returned + from either regcomp or regexec. We don't use PREG here. */ + +size_t +regerror (int errcode, const regex_t *__restrict preg, char *__restrict errbuf, + size_t errbuf_size) +{ + const char *msg; + size_t msg_size; + + if (BE (errcode < 0 + || errcode >= (int) (sizeof (__re_error_msgid_idx) + / sizeof (__re_error_msgid_idx[0])), 0)) + /* Only error codes returned by the rest of the code should be passed + to this routine. If we are given anything else, or if other regex + code generates an invalid error code, then the program has a bug. + Dump core so we can fix it. */ + abort (); + + msg = gettext (__re_error_msgid + __re_error_msgid_idx[errcode]); + + msg_size = strlen (msg) + 1; /* Includes the null. */ + + if (BE (errbuf_size != 0, 1)) + { + if (BE (msg_size > errbuf_size, 0)) + { +#if defined HAVE_MEMPCPY || defined _LIBC + *((char *) __mempcpy (errbuf, msg, errbuf_size - 1)) = '\0'; +#else + memcpy (errbuf, msg, errbuf_size - 1); + errbuf[errbuf_size - 1] = 0; +#endif + } + else + memcpy (errbuf, msg, msg_size); + } + + return msg_size; +} +#ifdef _LIBC +weak_alias (__regerror, regerror) +#endif + + +#ifdef RE_ENABLE_I18N +/* This static array is used for the map to single-byte characters when + UTF-8 is used. Otherwise we would allocate memory just to initialize + it the same all the time. UTF-8 is the preferred encoding so this is + a worthwhile optimization. */ +#if __GNUC__ >= 3 +static const bitset_t utf8_sb_map = { + /* Set the first 128 bits. */ + [0 ... 0x80 / BITSET_WORD_BITS - 1] = BITSET_WORD_MAX +}; +#else /* ! (__GNUC__ >= 3) */ +static bitset_t utf8_sb_map; +#endif /* __GNUC__ >= 3 */ +#endif /* RE_ENABLE_I18N */ + + +static void +free_dfa_content (re_dfa_t *dfa) +{ + int i, j; + + if (dfa->nodes) + for (i = 0; i < dfa->nodes_len; ++i) + free_token (dfa->nodes + i); + re_free (dfa->nexts); + for (i = 0; i < dfa->nodes_len; ++i) + { + if (dfa->eclosures != NULL) + re_node_set_free (dfa->eclosures + i); + if (dfa->inveclosures != NULL) + re_node_set_free (dfa->inveclosures + i); + if (dfa->edests != NULL) + re_node_set_free (dfa->edests + i); + } + re_free (dfa->edests); + re_free (dfa->eclosures); + re_free (dfa->inveclosures); + re_free (dfa->nodes); + + if (dfa->state_table) + for (i = 0; i <= dfa->state_hash_mask; ++i) + { + struct re_state_table_entry *entry = dfa->state_table + i; + for (j = 0; j < entry->num; ++j) + { + re_dfastate_t *state = entry->array[j]; + free_state (state); + } + re_free (entry->array); + } + re_free (dfa->state_table); +#ifdef RE_ENABLE_I18N + if (dfa->sb_char != utf8_sb_map) + re_free (dfa->sb_char); +#endif + re_free (dfa->subexp_map); +#ifdef DEBUG + re_free (dfa->re_str); +#endif + + re_free (dfa); +} + + +/* Free dynamically allocated space used by PREG. */ + +void +regfree (regex_t *preg) +{ + re_dfa_t *dfa = (re_dfa_t *) preg->buffer; + if (BE (dfa != NULL, 1)) + free_dfa_content (dfa); + preg->buffer = NULL; + preg->allocated = 0; + + re_free (preg->fastmap); + preg->fastmap = NULL; + + re_free (preg->translate); + preg->translate = NULL; +} +#ifdef _LIBC +weak_alias (__regfree, regfree) +#endif + +/* Entry points compatible with 4.2 BSD regex library. We don't define + them unless specifically requested. */ + +#if defined _REGEX_RE_COMP || defined _LIBC + +/* BSD has one and only one pattern buffer. */ +static struct re_pattern_buffer re_comp_buf; + +char * +# ifdef _LIBC +/* Make these definitions weak in libc, so POSIX programs can redefine + these names if they don't use our functions, and still use + regcomp/regexec above without link errors. */ +weak_function +# endif +re_comp (const char *s) +{ + reg_errcode_t ret; + char *fastmap; + + if (!s) + { + if (!re_comp_buf.buffer) + return gettext ("No previous regular expression"); + return 0; + } + + if (re_comp_buf.buffer) + { + fastmap = re_comp_buf.fastmap; + re_comp_buf.fastmap = NULL; + __regfree (&re_comp_buf); + memset (&re_comp_buf, '\0', sizeof (re_comp_buf)); + re_comp_buf.fastmap = fastmap; + } + + if (re_comp_buf.fastmap == NULL) + { + re_comp_buf.fastmap = (char *) malloc (SBC_MAX); + if (re_comp_buf.fastmap == NULL) + return (char *) gettext (__re_error_msgid + + __re_error_msgid_idx[(int) REG_ESPACE]); + } + + /* Since 're_exec' always passes NULL for the 'regs' argument, we + don't need to initialize the pattern buffer fields which affect it. */ + + /* Match anchors at newlines. */ + re_comp_buf.newline_anchor = 1; + + ret = re_compile_internal (&re_comp_buf, s, strlen (s), re_syntax_options); + + if (!ret) + return NULL; + + /* Yes, we're discarding `const' here if !HAVE_LIBINTL. */ + return (char *) gettext (__re_error_msgid + __re_error_msgid_idx[(int) ret]); +} + +#ifdef _LIBC +libc_freeres_fn (free_mem) +{ + __regfree (&re_comp_buf); +} +#endif + +#endif /* _REGEX_RE_COMP */ + +/* Internal entry point. + Compile the regular expression PATTERN, whose length is LENGTH. + SYNTAX indicate regular expression's syntax. */ + +static reg_errcode_t +re_compile_internal (regex_t *preg, const char * pattern, size_t length, + reg_syntax_t syntax) +{ + reg_errcode_t err = REG_NOERROR; + re_dfa_t *dfa; + re_string_t regexp; + + /* Initialize the pattern buffer. */ + preg->fastmap_accurate = 0; + preg->syntax = syntax; + preg->not_bol = preg->not_eol = 0; + preg->used = 0; + preg->re_nsub = 0; + preg->can_be_null = 0; + preg->regs_allocated = REGS_UNALLOCATED; + + /* Initialize the dfa. */ + dfa = (re_dfa_t *) preg->buffer; + if (BE (preg->allocated < sizeof (re_dfa_t), 0)) + { + /* If zero allocated, but buffer is non-null, try to realloc + enough space. This loses if buffer's address is bogus, but + that is the user's responsibility. If ->buffer is NULL this + is a simple allocation. */ + dfa = re_realloc (preg->buffer, re_dfa_t, 1); + if (dfa == NULL) + return REG_ESPACE; + preg->allocated = sizeof (re_dfa_t); + preg->buffer = (unsigned char *) dfa; + } + preg->used = sizeof (re_dfa_t); + + err = init_dfa (dfa, length); + if (BE (err != REG_NOERROR, 0)) + { + free_dfa_content (dfa); + preg->buffer = NULL; + preg->allocated = 0; + return err; + } +#ifdef DEBUG + /* Note: length+1 will not overflow since it is checked in init_dfa. */ + dfa->re_str = re_malloc (char, length + 1); + strncpy (dfa->re_str, pattern, length + 1); +#endif + + __libc_lock_init (dfa->lock); + + err = re_string_construct (®exp, pattern, length, preg->translate, + (syntax & RE_ICASE) != 0, dfa); + if (BE (err != REG_NOERROR, 0)) + { + re_compile_internal_free_return: + free_workarea_compile (preg); + re_string_destruct (®exp); + free_dfa_content (dfa); + preg->buffer = NULL; + preg->allocated = 0; + return err; + } + + /* Parse the regular expression, and build a structure tree. */ + preg->re_nsub = 0; + dfa->str_tree = parse (®exp, preg, syntax, &err); + if (BE (dfa->str_tree == NULL, 0)) + goto re_compile_internal_free_return; + + /* Analyze the tree and create the nfa. */ + err = analyze (preg); + if (BE (err != REG_NOERROR, 0)) + goto re_compile_internal_free_return; + +#ifdef RE_ENABLE_I18N + /* If possible, do searching in single byte encoding to speed things up. */ + if (dfa->is_utf8 && !(syntax & RE_ICASE) && preg->translate == NULL) + optimize_utf8 (dfa); +#endif + + /* Then create the initial state of the dfa. */ + err = create_initial_state (dfa); + + /* Release work areas. */ + free_workarea_compile (preg); + re_string_destruct (®exp); + + if (BE (err != REG_NOERROR, 0)) + { + free_dfa_content (dfa); + preg->buffer = NULL; + preg->allocated = 0; + } + + return err; +} + +/* Initialize DFA. We use the length of the regular expression PAT_LEN + as the initial length of some arrays. */ + +static reg_errcode_t +init_dfa (re_dfa_t *dfa, size_t pat_len) +{ + unsigned int table_size; +#ifndef _LIBC + char *codeset_name; +#endif + + memset (dfa, '\0', sizeof (re_dfa_t)); + + /* Force allocation of str_tree_storage the first time. */ + dfa->str_tree_storage_idx = BIN_TREE_STORAGE_SIZE; + + /* Avoid overflows. */ + if (pat_len == SIZE_MAX) + return REG_ESPACE; + + dfa->nodes_alloc = pat_len + 1; + dfa->nodes = re_malloc (re_token_t, dfa->nodes_alloc); + + /* table_size = 2 ^ ceil(log pat_len) */ + for (table_size = 1; ; table_size <<= 1) + if (table_size > pat_len) + break; + + dfa->state_table = calloc (sizeof (struct re_state_table_entry), table_size); + dfa->state_hash_mask = table_size - 1; + + dfa->mb_cur_max = MB_CUR_MAX; +#ifdef _LIBC + if (dfa->mb_cur_max == 6 + && strcmp (_NL_CURRENT (LC_CTYPE, _NL_CTYPE_CODESET_NAME), "UTF-8") == 0) + dfa->is_utf8 = 1; + dfa->map_notascii = (_NL_CURRENT_WORD (LC_CTYPE, _NL_CTYPE_MAP_TO_NONASCII) + != 0); +#else +# ifdef HAVE_LANGINFO_CODESET + codeset_name = nl_langinfo (CODESET); +# else + codeset_name = getenv ("LC_ALL"); + if (codeset_name == NULL || codeset_name[0] == '\0') + codeset_name = getenv ("LC_CTYPE"); + if (codeset_name == NULL || codeset_name[0] == '\0') + codeset_name = getenv ("LANG"); + if (codeset_name == NULL) + codeset_name = ""; + else if (strchr (codeset_name, '.') != NULL) + codeset_name = strchr (codeset_name, '.') + 1; +# endif + + if (strcasecmp (codeset_name, "UTF-8") == 0 + || strcasecmp (codeset_name, "UTF8") == 0) + dfa->is_utf8 = 1; + + /* We check exhaustively in the loop below if this charset is a + superset of ASCII. */ + dfa->map_notascii = 0; +#endif + +#ifdef RE_ENABLE_I18N + if (dfa->mb_cur_max > 1) + { + if (dfa->is_utf8) + { +#if !defined(__GNUC__) || __GNUC__ < 3 + static short utf8_sb_map_inited = 0; + + if (! utf8_sb_map_inited) + { + int i; + + utf8_sb_map_inited = 0; + for (i = 0; i <= 0x80 / BITSET_WORD_BITS - 1; i++) + utf8_sb_map[i] = BITSET_WORD_MAX; + } +#endif + dfa->sb_char = (re_bitset_ptr_t) utf8_sb_map; + } + else + { + int i, j, ch; + + dfa->sb_char = (re_bitset_ptr_t) calloc (sizeof (bitset_t), 1); + if (BE (dfa->sb_char == NULL, 0)) + return REG_ESPACE; + + /* Set the bits corresponding to single byte chars. */ + for (i = 0, ch = 0; i < BITSET_WORDS; ++i) + for (j = 0; j < BITSET_WORD_BITS; ++j, ++ch) + { + wint_t wch = __btowc (ch); + if (wch != WEOF) + dfa->sb_char[i] |= (bitset_word_t) 1 << j; +# ifndef _LIBC + if (isascii (ch) && wch != ch) + dfa->map_notascii = 1; +# endif + } + } + } +#endif + + if (BE (dfa->nodes == NULL || dfa->state_table == NULL, 0)) + return REG_ESPACE; + return REG_NOERROR; +} + +/* Initialize WORD_CHAR table, which indicate which character is + "word". In this case "word" means that it is the word construction + character used by some operators like "\<", "\>", etc. */ + +static void +internal_function +init_word_char (re_dfa_t *dfa) +{ + int i, j, ch; + dfa->word_ops_used = 1; +#ifndef GAWK + if (BE (dfa->map_notascii == 0, 1)) + { + if (sizeof (dfa->word_char[0]) == 8) + { + /* The extra temporaries here avoid "implicitly truncated" + warnings in the case when this is dead code, i.e. 32-bit. */ + const uint64_t wc0 = UINT64_C (0x03ff000000000000); + const uint64_t wc1 = UINT64_C (0x07fffffe87fffffe); + dfa->word_char[0] = wc0; + dfa->word_char[1] = wc1; + i = 2; + } + else if (sizeof (dfa->word_char[0]) == 4) + { + dfa->word_char[0] = UINT32_C (0x00000000); + dfa->word_char[1] = UINT32_C (0x03ff0000); + dfa->word_char[2] = UINT32_C (0x87fffffe); + dfa->word_char[3] = UINT32_C (0x07fffffe); + i = 4; + } + else + abort (); + ch = 128; + + if (BE (dfa->is_utf8, 1)) + { + memset (&dfa->word_char[i], '\0', (SBC_MAX - ch) / 8); + return; + } + } +#endif + + for (i = 0, ch = 0; i < BITSET_WORDS; ++i) + for (j = 0; j < BITSET_WORD_BITS; ++j, ++ch) + if (isalnum (ch) || ch == '_') + dfa->word_char[i] |= (bitset_word_t) 1 << j; +} + +/* Free the work area which are only used while compiling. */ + +static void +free_workarea_compile (regex_t *preg) +{ + re_dfa_t *dfa = (re_dfa_t *) preg->buffer; + bin_tree_storage_t *storage, *next; + for (storage = dfa->str_tree_storage; storage; storage = next) + { + next = storage->next; + re_free (storage); + } + dfa->str_tree_storage = NULL; + dfa->str_tree_storage_idx = BIN_TREE_STORAGE_SIZE; + dfa->str_tree = NULL; + re_free (dfa->org_indices); + dfa->org_indices = NULL; +} + +/* Create initial states for all contexts. */ + +static reg_errcode_t +create_initial_state (re_dfa_t *dfa) +{ + int first, i; + reg_errcode_t err; + re_node_set init_nodes; + + /* Initial states have the epsilon closure of the node which is + the first node of the regular expression. */ + first = dfa->str_tree->first->node_idx; + dfa->init_node = first; + err = re_node_set_init_copy (&init_nodes, dfa->eclosures + first); + if (BE (err != REG_NOERROR, 0)) + return err; + + /* The back-references which are in initial states can epsilon transit, + since in this case all of the subexpressions can be null. + Then we add epsilon closures of the nodes which are the next nodes of + the back-references. */ + if (dfa->nbackref > 0) + for (i = 0; i < init_nodes.nelem; ++i) + { + int node_idx = init_nodes.elems[i]; + re_token_type_t type = dfa->nodes[node_idx].type; + + int clexp_idx; + if (type != OP_BACK_REF) + continue; + for (clexp_idx = 0; clexp_idx < init_nodes.nelem; ++clexp_idx) + { + re_token_t *clexp_node; + clexp_node = dfa->nodes + init_nodes.elems[clexp_idx]; + if (clexp_node->type == OP_CLOSE_SUBEXP + && clexp_node->opr.idx == dfa->nodes[node_idx].opr.idx) + break; + } + if (clexp_idx == init_nodes.nelem) + continue; + + if (type == OP_BACK_REF) + { + int dest_idx = dfa->edests[node_idx].elems[0]; + if (!re_node_set_contains (&init_nodes, dest_idx)) + { + reg_errcode_t err = re_node_set_merge (&init_nodes, + dfa->eclosures + + dest_idx); + if (err != REG_NOERROR) + return err; + i = 0; + } + } + } + + /* It must be the first time to invoke acquire_state. */ + dfa->init_state = re_acquire_state_context (&err, dfa, &init_nodes, 0); + /* We don't check ERR here, since the initial state must not be NULL. */ + if (BE (dfa->init_state == NULL, 0)) + return err; + if (dfa->init_state->has_constraint) + { + dfa->init_state_word = re_acquire_state_context (&err, dfa, &init_nodes, + CONTEXT_WORD); + dfa->init_state_nl = re_acquire_state_context (&err, dfa, &init_nodes, + CONTEXT_NEWLINE); + dfa->init_state_begbuf = re_acquire_state_context (&err, dfa, + &init_nodes, + CONTEXT_NEWLINE + | CONTEXT_BEGBUF); + if (BE (dfa->init_state_word == NULL || dfa->init_state_nl == NULL + || dfa->init_state_begbuf == NULL, 0)) + return err; + } + else + dfa->init_state_word = dfa->init_state_nl + = dfa->init_state_begbuf = dfa->init_state; + + re_node_set_free (&init_nodes); + return REG_NOERROR; +} + +#ifdef RE_ENABLE_I18N +/* If it is possible to do searching in single byte encoding instead of UTF-8 + to speed things up, set dfa->mb_cur_max to 1, clear is_utf8 and change + DFA nodes where needed. */ + +static void +optimize_utf8 (re_dfa_t *dfa) +{ + int node, i, mb_chars = 0, has_period = 0; + + for (node = 0; node < dfa->nodes_len; ++node) + switch (dfa->nodes[node].type) + { + case CHARACTER: + if (dfa->nodes[node].opr.c >= 0x80) + mb_chars = 1; + break; + case ANCHOR: + switch (dfa->nodes[node].opr.ctx_type) + { + case LINE_FIRST: + case LINE_LAST: + case BUF_FIRST: + case BUF_LAST: + break; + default: + /* Word anchors etc. cannot be handled. It's okay to test + opr.ctx_type since constraints (for all DFA nodes) are + created by ORing one or more opr.ctx_type values. */ + return; + } + break; + case OP_PERIOD: + has_period = 1; + break; + case OP_BACK_REF: + case OP_ALT: + case END_OF_RE: + case OP_DUP_ASTERISK: + case OP_OPEN_SUBEXP: + case OP_CLOSE_SUBEXP: + break; + case COMPLEX_BRACKET: + return; + case SIMPLE_BRACKET: + /* Just double check. The non-ASCII range starts at 0x80. */ + assert (0x80 % BITSET_WORD_BITS == 0); + for (i = 0x80 / BITSET_WORD_BITS; i < BITSET_WORDS; ++i) + if (dfa->nodes[node].opr.sbcset[i]) + return; + break; + default: + abort (); + } + + if (mb_chars || has_period) + for (node = 0; node < dfa->nodes_len; ++node) + { + if (dfa->nodes[node].type == CHARACTER + && dfa->nodes[node].opr.c >= 0x80) + dfa->nodes[node].mb_partial = 0; + else if (dfa->nodes[node].type == OP_PERIOD) + dfa->nodes[node].type = OP_UTF8_PERIOD; + } + + /* The search can be in single byte locale. */ + dfa->mb_cur_max = 1; + dfa->is_utf8 = 0; + dfa->has_mb_node = dfa->nbackref > 0 || has_period; +} +#endif + +/* Analyze the structure tree, and calculate "first", "next", "edest", + "eclosure", and "inveclosure". */ + +static reg_errcode_t +analyze (regex_t *preg) +{ + re_dfa_t *dfa = (re_dfa_t *) preg->buffer; + reg_errcode_t ret; + + /* Allocate arrays. */ + dfa->nexts = re_malloc (int, dfa->nodes_alloc); + dfa->org_indices = re_malloc (int, dfa->nodes_alloc); + dfa->edests = re_malloc (re_node_set, dfa->nodes_alloc); + dfa->eclosures = re_malloc (re_node_set, dfa->nodes_alloc); + if (BE (dfa->nexts == NULL || dfa->org_indices == NULL || dfa->edests == NULL + || dfa->eclosures == NULL, 0)) + return REG_ESPACE; + + /* some malloc()-checkers don't like zero allocations */ + if (preg->re_nsub > 0) + dfa->subexp_map = re_malloc (int, preg->re_nsub); + else + dfa->subexp_map = NULL; + + if (dfa->subexp_map != NULL) + { + int i; + for (i = 0; i < preg->re_nsub; i++) + dfa->subexp_map[i] = i; + preorder (dfa->str_tree, optimize_subexps, dfa); + for (i = 0; i < preg->re_nsub; i++) + if (dfa->subexp_map[i] != i) + break; + if (i == preg->re_nsub) + { + free (dfa->subexp_map); + dfa->subexp_map = NULL; + } + } + + ret = postorder (dfa->str_tree, lower_subexps, preg); + if (BE (ret != REG_NOERROR, 0)) + return ret; + ret = postorder (dfa->str_tree, calc_first, dfa); + if (BE (ret != REG_NOERROR, 0)) + return ret; + preorder (dfa->str_tree, calc_next, dfa); + ret = preorder (dfa->str_tree, link_nfa_nodes, dfa); + if (BE (ret != REG_NOERROR, 0)) + return ret; + ret = calc_eclosure (dfa); + if (BE (ret != REG_NOERROR, 0)) + return ret; + + /* We only need this during the prune_impossible_nodes pass in regexec.c; + skip it if p_i_n will not run, as calc_inveclosure can be quadratic. */ + if ((!preg->no_sub && preg->re_nsub > 0 && dfa->has_plural_match) + || dfa->nbackref) + { + dfa->inveclosures = re_malloc (re_node_set, dfa->nodes_len); + if (BE (dfa->inveclosures == NULL, 0)) + return REG_ESPACE; + ret = calc_inveclosure (dfa); + } + + return ret; +} + +/* Our parse trees are very unbalanced, so we cannot use a stack to + implement parse tree visits. Instead, we use parent pointers and + some hairy code in these two functions. */ +static reg_errcode_t +postorder (bin_tree_t *root, reg_errcode_t (fn (void *, bin_tree_t *)), + void *extra) +{ + bin_tree_t *node, *prev; + + for (node = root; ; ) + { + /* Descend down the tree, preferably to the left (or to the right + if that's the only child). */ + while (node->left || node->right) + if (node->left) + node = node->left; + else + node = node->right; + + do + { + reg_errcode_t err = fn (extra, node); + if (BE (err != REG_NOERROR, 0)) + return err; + if (node->parent == NULL) + return REG_NOERROR; + prev = node; + node = node->parent; + } + /* Go up while we have a node that is reached from the right. */ + while (node->right == prev || node->right == NULL); + node = node->right; + } +} + +static reg_errcode_t +preorder (bin_tree_t *root, reg_errcode_t (fn (void *, bin_tree_t *)), + void *extra) +{ + bin_tree_t *node; + + for (node = root; ; ) + { + reg_errcode_t err = fn (extra, node); + if (BE (err != REG_NOERROR, 0)) + return err; + + /* Go to the left node, or up and to the right. */ + if (node->left) + node = node->left; + else + { + bin_tree_t *prev = NULL; + while (node->right == prev || node->right == NULL) + { + prev = node; + node = node->parent; + if (!node) + return REG_NOERROR; + } + node = node->right; + } + } +} + +/* Optimization pass: if a SUBEXP is entirely contained, strip it and tell + re_search_internal to map the inner one's opr.idx to this one's. Adjust + backreferences as well. Requires a preorder visit. */ +static reg_errcode_t +optimize_subexps (void *extra, bin_tree_t *node) +{ + re_dfa_t *dfa = (re_dfa_t *) extra; + + if (node->token.type == OP_BACK_REF && dfa->subexp_map) + { + int idx = node->token.opr.idx; + node->token.opr.idx = dfa->subexp_map[idx]; + dfa->used_bkref_map |= 1 << node->token.opr.idx; + } + + else if (node->token.type == SUBEXP + && node->left && node->left->token.type == SUBEXP) + { + int other_idx = node->left->token.opr.idx; + + node->left = node->left->left; + if (node->left) + node->left->parent = node; + + dfa->subexp_map[other_idx] = dfa->subexp_map[node->token.opr.idx]; + if (other_idx < BITSET_WORD_BITS) + dfa->used_bkref_map &= ~((bitset_word_t) 1 << other_idx); + } + + return REG_NOERROR; +} + +/* Lowering pass: Turn each SUBEXP node into the appropriate concatenation + of OP_OPEN_SUBEXP, the body of the SUBEXP (if any) and OP_CLOSE_SUBEXP. */ +static reg_errcode_t +lower_subexps (void *extra, bin_tree_t *node) +{ + regex_t *preg = (regex_t *) extra; + reg_errcode_t err = REG_NOERROR; + + if (node->left && node->left->token.type == SUBEXP) + { + node->left = lower_subexp (&err, preg, node->left); + if (node->left) + node->left->parent = node; + } + if (node->right && node->right->token.type == SUBEXP) + { + node->right = lower_subexp (&err, preg, node->right); + if (node->right) + node->right->parent = node; + } + + return err; +} + +static bin_tree_t * +lower_subexp (reg_errcode_t *err, regex_t *preg, bin_tree_t *node) +{ + re_dfa_t *dfa = (re_dfa_t *) preg->buffer; + bin_tree_t *body = node->left; + bin_tree_t *op, *cls, *tree1, *tree; + + if (preg->no_sub + /* We do not optimize empty subexpressions, because otherwise we may + have bad CONCAT nodes with NULL children. This is obviously not + very common, so we do not lose much. An example that triggers + this case is the sed "script" /\(\)/x. */ + && node->left != NULL + && (node->token.opr.idx >= BITSET_WORD_BITS + || !(dfa->used_bkref_map + & ((bitset_word_t) 1 << node->token.opr.idx)))) + return node->left; + + /* Convert the SUBEXP node to the concatenation of an + OP_OPEN_SUBEXP, the contents, and an OP_CLOSE_SUBEXP. */ + op = create_tree (dfa, NULL, NULL, OP_OPEN_SUBEXP); + cls = create_tree (dfa, NULL, NULL, OP_CLOSE_SUBEXP); + tree1 = body ? create_tree (dfa, body, cls, CONCAT) : cls; + tree = create_tree (dfa, op, tree1, CONCAT); + if (BE (tree == NULL || tree1 == NULL || op == NULL || cls == NULL, 0)) + { + *err = REG_ESPACE; + return NULL; + } + + op->token.opr.idx = cls->token.opr.idx = node->token.opr.idx; + op->token.opt_subexp = cls->token.opt_subexp = node->token.opt_subexp; + return tree; +} + +/* Pass 1 in building the NFA: compute FIRST and create unlinked automaton + nodes. Requires a postorder visit. */ +static reg_errcode_t +calc_first (void *extra, bin_tree_t *node) +{ + re_dfa_t *dfa = (re_dfa_t *) extra; + if (node->token.type == CONCAT) + { + node->first = node->left->first; + node->node_idx = node->left->node_idx; + } + else + { + node->first = node; + node->node_idx = re_dfa_add_node (dfa, node->token); + if (BE (node->node_idx == -1, 0)) + return REG_ESPACE; + if (node->token.type == ANCHOR) + dfa->nodes[node->node_idx].constraint = node->token.opr.ctx_type; + } + return REG_NOERROR; +} + +/* Pass 2: compute NEXT on the tree. Preorder visit. */ +static reg_errcode_t +calc_next (void *extra, bin_tree_t *node) +{ + switch (node->token.type) + { + case OP_DUP_ASTERISK: + node->left->next = node; + break; + case CONCAT: + node->left->next = node->right->first; + node->right->next = node->next; + break; + default: + if (node->left) + node->left->next = node->next; + if (node->right) + node->right->next = node->next; + break; + } + return REG_NOERROR; +} + +/* Pass 3: link all DFA nodes to their NEXT node (any order will do). */ +static reg_errcode_t +link_nfa_nodes (void *extra, bin_tree_t *node) +{ + re_dfa_t *dfa = (re_dfa_t *) extra; + int idx = node->node_idx; + reg_errcode_t err = REG_NOERROR; + + switch (node->token.type) + { + case CONCAT: + break; + + case END_OF_RE: + assert (node->next == NULL); + break; + + case OP_DUP_ASTERISK: + case OP_ALT: + { + int left, right; + dfa->has_plural_match = 1; + if (node->left != NULL) + left = node->left->first->node_idx; + else + left = node->next->node_idx; + if (node->right != NULL) + right = node->right->first->node_idx; + else + right = node->next->node_idx; + assert (left > -1); + assert (right > -1); + err = re_node_set_init_2 (dfa->edests + idx, left, right); + } + break; + + case ANCHOR: + case OP_OPEN_SUBEXP: + case OP_CLOSE_SUBEXP: + err = re_node_set_init_1 (dfa->edests + idx, node->next->node_idx); + break; + + case OP_BACK_REF: + dfa->nexts[idx] = node->next->node_idx; + if (node->token.type == OP_BACK_REF) + err = re_node_set_init_1 (dfa->edests + idx, dfa->nexts[idx]); + break; + + default: + assert (!IS_EPSILON_NODE (node->token.type)); + dfa->nexts[idx] = node->next->node_idx; + break; + } + + return err; +} + +/* Duplicate the epsilon closure of the node ROOT_NODE. + Note that duplicated nodes have constraint INIT_CONSTRAINT in addition + to their own constraint. */ + +static reg_errcode_t +internal_function +duplicate_node_closure (re_dfa_t *dfa, int top_org_node, int top_clone_node, + int root_node, unsigned int init_constraint) +{ + int org_node, clone_node, ret; + unsigned int constraint = init_constraint; + for (org_node = top_org_node, clone_node = top_clone_node;;) + { + int org_dest, clone_dest; + if (dfa->nodes[org_node].type == OP_BACK_REF) + { + /* If the back reference epsilon-transit, its destination must + also have the constraint. Then duplicate the epsilon closure + of the destination of the back reference, and store it in + edests of the back reference. */ + org_dest = dfa->nexts[org_node]; + re_node_set_empty (dfa->edests + clone_node); + clone_dest = duplicate_node (dfa, org_dest, constraint); + if (BE (clone_dest == -1, 0)) + return REG_ESPACE; + dfa->nexts[clone_node] = dfa->nexts[org_node]; + ret = re_node_set_insert (dfa->edests + clone_node, clone_dest); + if (BE (ret < 0, 0)) + return REG_ESPACE; + } + else if (dfa->edests[org_node].nelem == 0) + { + /* In case of the node can't epsilon-transit, don't duplicate the + destination and store the original destination as the + destination of the node. */ + dfa->nexts[clone_node] = dfa->nexts[org_node]; + break; + } + else if (dfa->edests[org_node].nelem == 1) + { + /* In case of the node can epsilon-transit, and it has only one + destination. */ + org_dest = dfa->edests[org_node].elems[0]; + re_node_set_empty (dfa->edests + clone_node); + /* If the node is root_node itself, it means the epsilon closure + has a loop. Then tie it to the destination of the root_node. */ + if (org_node == root_node && clone_node != org_node) + { + ret = re_node_set_insert (dfa->edests + clone_node, org_dest); + if (BE (ret < 0, 0)) + return REG_ESPACE; + break; + } + /* In case the node has another constraint, append it. */ + constraint |= dfa->nodes[org_node].constraint; + clone_dest = duplicate_node (dfa, org_dest, constraint); + if (BE (clone_dest == -1, 0)) + return REG_ESPACE; + ret = re_node_set_insert (dfa->edests + clone_node, clone_dest); + if (BE (ret < 0, 0)) + return REG_ESPACE; + } + else /* dfa->edests[org_node].nelem == 2 */ + { + /* In case of the node can epsilon-transit, and it has two + destinations. In the bin_tree_t and DFA, that's '|' and '*'. */ + org_dest = dfa->edests[org_node].elems[0]; + re_node_set_empty (dfa->edests + clone_node); + /* Search for a duplicated node which satisfies the constraint. */ + clone_dest = search_duplicated_node (dfa, org_dest, constraint); + if (clone_dest == -1) + { + /* There is no such duplicated node, create a new one. */ + reg_errcode_t err; + clone_dest = duplicate_node (dfa, org_dest, constraint); + if (BE (clone_dest == -1, 0)) + return REG_ESPACE; + ret = re_node_set_insert (dfa->edests + clone_node, clone_dest); + if (BE (ret < 0, 0)) + return REG_ESPACE; + err = duplicate_node_closure (dfa, org_dest, clone_dest, + root_node, constraint); + if (BE (err != REG_NOERROR, 0)) + return err; + } + else + { + /* There is a duplicated node which satisfies the constraint, + use it to avoid infinite loop. */ + ret = re_node_set_insert (dfa->edests + clone_node, clone_dest); + if (BE (ret < 0, 0)) + return REG_ESPACE; + } + + org_dest = dfa->edests[org_node].elems[1]; + clone_dest = duplicate_node (dfa, org_dest, constraint); + if (BE (clone_dest == -1, 0)) + return REG_ESPACE; + ret = re_node_set_insert (dfa->edests + clone_node, clone_dest); + if (BE (ret < 0, 0)) + return REG_ESPACE; + } + org_node = org_dest; + clone_node = clone_dest; + } + return REG_NOERROR; +} + +/* Search for a node which is duplicated from the node ORG_NODE, and + satisfies the constraint CONSTRAINT. */ + +static int +search_duplicated_node (const re_dfa_t *dfa, int org_node, + unsigned int constraint) +{ + int idx; + for (idx = dfa->nodes_len - 1; dfa->nodes[idx].duplicated && idx > 0; --idx) + { + if (org_node == dfa->org_indices[idx] + && constraint == dfa->nodes[idx].constraint) + return idx; /* Found. */ + } + return -1; /* Not found. */ +} + +/* Duplicate the node whose index is ORG_IDX and set the constraint CONSTRAINT. + Return the index of the new node, or -1 if insufficient storage is + available. */ + +static int +duplicate_node (re_dfa_t *dfa, int org_idx, unsigned int constraint) +{ + int dup_idx = re_dfa_add_node (dfa, dfa->nodes[org_idx]); + if (BE (dup_idx != -1, 1)) + { + dfa->nodes[dup_idx].constraint = constraint; + dfa->nodes[dup_idx].constraint |= dfa->nodes[org_idx].constraint; + dfa->nodes[dup_idx].duplicated = 1; + + /* Store the index of the original node. */ + dfa->org_indices[dup_idx] = org_idx; + } + return dup_idx; +} + +static reg_errcode_t +calc_inveclosure (re_dfa_t *dfa) +{ + int src, idx, ret; + for (idx = 0; idx < dfa->nodes_len; ++idx) + re_node_set_init_empty (dfa->inveclosures + idx); + + for (src = 0; src < dfa->nodes_len; ++src) + { + int *elems = dfa->eclosures[src].elems; + for (idx = 0; idx < dfa->eclosures[src].nelem; ++idx) + { + ret = re_node_set_insert_last (dfa->inveclosures + elems[idx], src); + if (BE (ret == -1, 0)) + return REG_ESPACE; + } + } + + return REG_NOERROR; +} + +/* Calculate "eclosure" for all the node in DFA. */ + +static reg_errcode_t +calc_eclosure (re_dfa_t *dfa) +{ + int node_idx, incomplete; +#ifdef DEBUG + assert (dfa->nodes_len > 0); +#endif + incomplete = 0; + /* For each nodes, calculate epsilon closure. */ + for (node_idx = 0; ; ++node_idx) + { + reg_errcode_t err; + re_node_set eclosure_elem; + if (node_idx == dfa->nodes_len) + { + if (!incomplete) + break; + incomplete = 0; + node_idx = 0; + } + +#ifdef DEBUG + assert (dfa->eclosures[node_idx].nelem != -1); +#endif + + /* If we have already calculated, skip it. */ + if (dfa->eclosures[node_idx].nelem != 0) + continue; + /* Calculate epsilon closure of 'node_idx'. */ + err = calc_eclosure_iter (&eclosure_elem, dfa, node_idx, 1); + if (BE (err != REG_NOERROR, 0)) + return err; + + if (dfa->eclosures[node_idx].nelem == 0) + { + incomplete = 1; + re_node_set_free (&eclosure_elem); + } + } + return REG_NOERROR; +} + +/* Calculate epsilon closure of NODE. */ + +static reg_errcode_t +calc_eclosure_iter (re_node_set *new_set, re_dfa_t *dfa, int node, int root) +{ + reg_errcode_t err; + int i; + re_node_set eclosure; + int ret; + int incomplete = 0; + err = re_node_set_alloc (&eclosure, dfa->edests[node].nelem + 1); + if (BE (err != REG_NOERROR, 0)) + return err; + + /* This indicates that we are calculating this node now. + We reference this value to avoid infinite loop. */ + dfa->eclosures[node].nelem = -1; + + /* If the current node has constraints, duplicate all nodes + since they must inherit the constraints. */ + if (dfa->nodes[node].constraint + && dfa->edests[node].nelem + && !dfa->nodes[dfa->edests[node].elems[0]].duplicated) + { + err = duplicate_node_closure (dfa, node, node, node, + dfa->nodes[node].constraint); + if (BE (err != REG_NOERROR, 0)) + return err; + } + + /* Expand each epsilon destination nodes. */ + if (IS_EPSILON_NODE(dfa->nodes[node].type)) + for (i = 0; i < dfa->edests[node].nelem; ++i) + { + re_node_set eclosure_elem; + int edest = dfa->edests[node].elems[i]; + /* If calculating the epsilon closure of `edest' is in progress, + return intermediate result. */ + if (dfa->eclosures[edest].nelem == -1) + { + incomplete = 1; + continue; + } + /* If we haven't calculated the epsilon closure of `edest' yet, + calculate now. Otherwise use calculated epsilon closure. */ + if (dfa->eclosures[edest].nelem == 0) + { + err = calc_eclosure_iter (&eclosure_elem, dfa, edest, 0); + if (BE (err != REG_NOERROR, 0)) + return err; + } + else + eclosure_elem = dfa->eclosures[edest]; + /* Merge the epsilon closure of 'edest'. */ + err = re_node_set_merge (&eclosure, &eclosure_elem); + if (BE (err != REG_NOERROR, 0)) + return err; + /* If the epsilon closure of 'edest' is incomplete, + the epsilon closure of this node is also incomplete. */ + if (dfa->eclosures[edest].nelem == 0) + { + incomplete = 1; + re_node_set_free (&eclosure_elem); + } + } + + /* An epsilon closure includes itself. */ + ret = re_node_set_insert (&eclosure, node); + if (BE (ret < 0, 0)) + return REG_ESPACE; + if (incomplete && !root) + dfa->eclosures[node].nelem = 0; + else + dfa->eclosures[node] = eclosure; + *new_set = eclosure; + return REG_NOERROR; +} + +/* Functions for token which are used in the parser. */ + +/* Fetch a token from INPUT. + We must not use this function inside bracket expressions. */ + +static void +internal_function +fetch_token (re_token_t *result, re_string_t *input, reg_syntax_t syntax) +{ + re_string_skip_bytes (input, peek_token (result, input, syntax)); +} + +/* Peek a token from INPUT, and return the length of the token. + We must not use this function inside bracket expressions. */ + +static int +internal_function +peek_token (re_token_t *token, re_string_t *input, reg_syntax_t syntax) +{ + unsigned char c; + + if (re_string_eoi (input)) + { + token->type = END_OF_RE; + return 0; + } + + c = re_string_peek_byte (input, 0); + token->opr.c = c; + + token->word_char = 0; +#ifdef RE_ENABLE_I18N + token->mb_partial = 0; + if (input->mb_cur_max > 1 && + !re_string_first_byte (input, re_string_cur_idx (input))) + { + token->type = CHARACTER; + token->mb_partial = 1; + return 1; + } +#endif + if (c == '\\') + { + unsigned char c2; + if (re_string_cur_idx (input) + 1 >= re_string_length (input)) + { + token->type = BACK_SLASH; + return 1; + } + + c2 = re_string_peek_byte_case (input, 1); + token->opr.c = c2; + token->type = CHARACTER; +#ifdef RE_ENABLE_I18N + if (input->mb_cur_max > 1) + { + wint_t wc = re_string_wchar_at (input, + re_string_cur_idx (input) + 1); + token->word_char = IS_WIDE_WORD_CHAR (wc) != 0; + } + else +#endif + token->word_char = IS_WORD_CHAR (c2) != 0; + + switch (c2) + { + case '|': + if (!(syntax & RE_LIMITED_OPS) && !(syntax & RE_NO_BK_VBAR)) + token->type = OP_ALT; + break; + case '1': case '2': case '3': case '4': case '5': + case '6': case '7': case '8': case '9': + if (!(syntax & RE_NO_BK_REFS)) + { + token->type = OP_BACK_REF; + token->opr.idx = c2 - '1'; + } + break; + case '<': + if (!(syntax & RE_NO_GNU_OPS)) + { + token->type = ANCHOR; + token->opr.ctx_type = WORD_FIRST; + } + break; + case '>': + if (!(syntax & RE_NO_GNU_OPS)) + { + token->type = ANCHOR; + token->opr.ctx_type = WORD_LAST; + } + break; + case 'b': + if (!(syntax & RE_NO_GNU_OPS)) + { + token->type = ANCHOR; + token->opr.ctx_type = WORD_DELIM; + } + break; + case 'B': + if (!(syntax & RE_NO_GNU_OPS)) + { + token->type = ANCHOR; + token->opr.ctx_type = NOT_WORD_DELIM; + } + break; + case 'w': + if (!(syntax & RE_NO_GNU_OPS)) + token->type = OP_WORD; + break; + case 'W': + if (!(syntax & RE_NO_GNU_OPS)) + token->type = OP_NOTWORD; + break; + case 's': + if (!(syntax & RE_NO_GNU_OPS)) + token->type = OP_SPACE; + break; + case 'S': + if (!(syntax & RE_NO_GNU_OPS)) + token->type = OP_NOTSPACE; + break; + case '`': + if (!(syntax & RE_NO_GNU_OPS)) + { + token->type = ANCHOR; + token->opr.ctx_type = BUF_FIRST; + } + break; + case '\'': + if (!(syntax & RE_NO_GNU_OPS)) + { + token->type = ANCHOR; + token->opr.ctx_type = BUF_LAST; + } + break; + case '(': + if (!(syntax & RE_NO_BK_PARENS)) + token->type = OP_OPEN_SUBEXP; + break; + case ')': + if (!(syntax & RE_NO_BK_PARENS)) + token->type = OP_CLOSE_SUBEXP; + break; + case '+': + if (!(syntax & RE_LIMITED_OPS) && (syntax & RE_BK_PLUS_QM)) + token->type = OP_DUP_PLUS; + break; + case '?': + if (!(syntax & RE_LIMITED_OPS) && (syntax & RE_BK_PLUS_QM)) + token->type = OP_DUP_QUESTION; + break; + case '{': + if ((syntax & RE_INTERVALS) && (!(syntax & RE_NO_BK_BRACES))) + token->type = OP_OPEN_DUP_NUM; + break; + case '}': + if ((syntax & RE_INTERVALS) && (!(syntax & RE_NO_BK_BRACES))) + token->type = OP_CLOSE_DUP_NUM; + break; + default: + break; + } + return 2; + } + + token->type = CHARACTER; +#ifdef RE_ENABLE_I18N + if (input->mb_cur_max > 1) + { + wint_t wc = re_string_wchar_at (input, re_string_cur_idx (input)); + token->word_char = IS_WIDE_WORD_CHAR (wc) != 0; + } + else +#endif + token->word_char = IS_WORD_CHAR (token->opr.c); + + switch (c) + { + case '\n': + if (syntax & RE_NEWLINE_ALT) + token->type = OP_ALT; + break; + case '|': + if (!(syntax & RE_LIMITED_OPS) && (syntax & RE_NO_BK_VBAR)) + token->type = OP_ALT; + break; + case '*': + token->type = OP_DUP_ASTERISK; + break; + case '+': + if (!(syntax & RE_LIMITED_OPS) && !(syntax & RE_BK_PLUS_QM)) + token->type = OP_DUP_PLUS; + break; + case '?': + if (!(syntax & RE_LIMITED_OPS) && !(syntax & RE_BK_PLUS_QM)) + token->type = OP_DUP_QUESTION; + break; + case '{': + if ((syntax & RE_INTERVALS) && (syntax & RE_NO_BK_BRACES)) + token->type = OP_OPEN_DUP_NUM; + break; + case '}': + if ((syntax & RE_INTERVALS) && (syntax & RE_NO_BK_BRACES)) + token->type = OP_CLOSE_DUP_NUM; + break; + case '(': + if (syntax & RE_NO_BK_PARENS) + token->type = OP_OPEN_SUBEXP; + break; + case ')': + if (syntax & RE_NO_BK_PARENS) + token->type = OP_CLOSE_SUBEXP; + break; + case '[': + token->type = OP_OPEN_BRACKET; + break; + case '.': + token->type = OP_PERIOD; + break; + case '^': + if (!(syntax & (RE_CONTEXT_INDEP_ANCHORS | RE_CARET_ANCHORS_HERE)) && + re_string_cur_idx (input) != 0) + { + char prev = re_string_peek_byte (input, -1); + if (!(syntax & RE_NEWLINE_ALT) || prev != '\n') + break; + } + token->type = ANCHOR; + token->opr.ctx_type = LINE_FIRST; + break; + case '$': + if (!(syntax & RE_CONTEXT_INDEP_ANCHORS) && + re_string_cur_idx (input) + 1 != re_string_length (input)) + { + re_token_t next; + re_string_skip_bytes (input, 1); + peek_token (&next, input, syntax); + re_string_skip_bytes (input, -1); + if (next.type != OP_ALT && next.type != OP_CLOSE_SUBEXP) + break; + } + token->type = ANCHOR; + token->opr.ctx_type = LINE_LAST; + break; + default: + break; + } + return 1; +} + +/* Peek a token from INPUT, and return the length of the token. + We must not use this function out of bracket expressions. */ + +static int +internal_function +peek_token_bracket (re_token_t *token, re_string_t *input, reg_syntax_t syntax) +{ + unsigned char c; + if (re_string_eoi (input)) + { + token->type = END_OF_RE; + return 0; + } + c = re_string_peek_byte (input, 0); + token->opr.c = c; + +#ifdef RE_ENABLE_I18N + if (input->mb_cur_max > 1 && + !re_string_first_byte (input, re_string_cur_idx (input))) + { + token->type = CHARACTER; + return 1; + } +#endif /* RE_ENABLE_I18N */ + + if (c == '\\' && (syntax & RE_BACKSLASH_ESCAPE_IN_LISTS) + && re_string_cur_idx (input) + 1 < re_string_length (input)) + { + /* In this case, '\' escape a character. */ + unsigned char c2; + re_string_skip_bytes (input, 1); + c2 = re_string_peek_byte (input, 0); + token->opr.c = c2; + token->type = CHARACTER; + return 1; + } + if (c == '[') /* '[' is a special char in a bracket exps. */ + { + unsigned char c2; + int token_len; + if (re_string_cur_idx (input) + 1 < re_string_length (input)) + c2 = re_string_peek_byte (input, 1); + else + c2 = 0; + token->opr.c = c2; + token_len = 2; + switch (c2) + { + case '.': + token->type = OP_OPEN_COLL_ELEM; + break; + case '=': + token->type = OP_OPEN_EQUIV_CLASS; + break; + case ':': + if (syntax & RE_CHAR_CLASSES) + { + token->type = OP_OPEN_CHAR_CLASS; + break; + } + /* else fall through. */ + default: + token->type = CHARACTER; + token->opr.c = c; + token_len = 1; + break; + } + return token_len; + } + switch (c) + { + case '-': + token->type = OP_CHARSET_RANGE; + break; + case ']': + token->type = OP_CLOSE_BRACKET; + break; + case '^': + token->type = OP_NON_MATCH_LIST; + break; + default: + token->type = CHARACTER; + } + return 1; +} + +/* Functions for parser. */ + +/* Entry point of the parser. + Parse the regular expression REGEXP and return the structure tree. + If an error occurs, ERR is set by error code, and return NULL. + This function build the following tree, from regular expression <reg_exp>: + CAT + / \ + / \ + <reg_exp> EOR + + CAT means concatenation. + EOR means end of regular expression. */ + +static bin_tree_t * +parse (re_string_t *regexp, regex_t *preg, reg_syntax_t syntax, + reg_errcode_t *err) +{ + re_dfa_t *dfa = (re_dfa_t *) preg->buffer; + bin_tree_t *tree, *eor, *root; + re_token_t current_token; + dfa->syntax = syntax; + fetch_token (¤t_token, regexp, syntax | RE_CARET_ANCHORS_HERE); + tree = parse_reg_exp (regexp, preg, ¤t_token, syntax, 0, err); + if (BE (*err != REG_NOERROR && tree == NULL, 0)) + return NULL; + eor = create_tree (dfa, NULL, NULL, END_OF_RE); + if (tree != NULL) + root = create_tree (dfa, tree, eor, CONCAT); + else + root = eor; + if (BE (eor == NULL || root == NULL, 0)) + { + *err = REG_ESPACE; + return NULL; + } + return root; +} + +/* This function build the following tree, from regular expression + <branch1>|<branch2>: + ALT + / \ + / \ + <branch1> <branch2> + + ALT means alternative, which represents the operator '|'. */ + +static bin_tree_t * +parse_reg_exp (re_string_t *regexp, regex_t *preg, re_token_t *token, + reg_syntax_t syntax, int nest, reg_errcode_t *err) +{ + re_dfa_t *dfa = (re_dfa_t *) preg->buffer; + bin_tree_t *tree, *branch = NULL; + bitset_word_t initial_bkref_map = dfa->completed_bkref_map; + tree = parse_branch (regexp, preg, token, syntax, nest, err); + if (BE (*err != REG_NOERROR && tree == NULL, 0)) + return NULL; + + while (token->type == OP_ALT) + { + fetch_token (token, regexp, syntax | RE_CARET_ANCHORS_HERE); + if (token->type != OP_ALT && token->type != END_OF_RE + && (nest == 0 || token->type != OP_CLOSE_SUBEXP)) + { + bitset_word_t accumulated_bkref_map = dfa->completed_bkref_map; + dfa->completed_bkref_map = initial_bkref_map; + branch = parse_branch (regexp, preg, token, syntax, nest, err); + if (BE (*err != REG_NOERROR && branch == NULL, 0)) + { + if (tree != NULL) + postorder (tree, free_tree, NULL); + return NULL; + } + dfa->completed_bkref_map |= accumulated_bkref_map; + } + else + branch = NULL; + tree = create_tree (dfa, tree, branch, OP_ALT); + if (BE (tree == NULL, 0)) + { + *err = REG_ESPACE; + return NULL; + } + } + return tree; +} + +/* This function build the following tree, from regular expression + <exp1><exp2>: + CAT + / \ + / \ + <exp1> <exp2> + + CAT means concatenation. */ + +static bin_tree_t * +parse_branch (re_string_t *regexp, regex_t *preg, re_token_t *token, + reg_syntax_t syntax, int nest, reg_errcode_t *err) +{ + bin_tree_t *tree, *exp; + re_dfa_t *dfa = (re_dfa_t *) preg->buffer; + tree = parse_expression (regexp, preg, token, syntax, nest, err); + if (BE (*err != REG_NOERROR && tree == NULL, 0)) + return NULL; + + while (token->type != OP_ALT && token->type != END_OF_RE + && (nest == 0 || token->type != OP_CLOSE_SUBEXP)) + { + exp = parse_expression (regexp, preg, token, syntax, nest, err); + if (BE (*err != REG_NOERROR && exp == NULL, 0)) + { + if (tree != NULL) + postorder (tree, free_tree, NULL); + return NULL; + } + if (tree != NULL && exp != NULL) + { + bin_tree_t *newtree = create_tree (dfa, tree, exp, CONCAT); + if (newtree == NULL) + { + postorder (exp, free_tree, NULL); + postorder (tree, free_tree, NULL); + *err = REG_ESPACE; + return NULL; + } + tree = newtree; + } + else if (tree == NULL) + tree = exp; + /* Otherwise exp == NULL, we don't need to create new tree. */ + } + return tree; +} + +/* This function build the following tree, from regular expression a*: + * + | + a +*/ + +static bin_tree_t * +parse_expression (re_string_t *regexp, regex_t *preg, re_token_t *token, + reg_syntax_t syntax, int nest, reg_errcode_t *err) +{ + re_dfa_t *dfa = (re_dfa_t *) preg->buffer; + bin_tree_t *tree; + switch (token->type) + { + case CHARACTER: + tree = create_token_tree (dfa, NULL, NULL, token); + if (BE (tree == NULL, 0)) + { + *err = REG_ESPACE; + return NULL; + } +#ifdef RE_ENABLE_I18N + if (dfa->mb_cur_max > 1) + { + while (!re_string_eoi (regexp) + && !re_string_first_byte (regexp, re_string_cur_idx (regexp))) + { + bin_tree_t *mbc_remain; + fetch_token (token, regexp, syntax); + mbc_remain = create_token_tree (dfa, NULL, NULL, token); + tree = create_tree (dfa, tree, mbc_remain, CONCAT); + if (BE (mbc_remain == NULL || tree == NULL, 0)) + { + *err = REG_ESPACE; + return NULL; + } + } + } +#endif + break; + case OP_OPEN_SUBEXP: + tree = parse_sub_exp (regexp, preg, token, syntax, nest + 1, err); + if (BE (*err != REG_NOERROR && tree == NULL, 0)) + return NULL; + break; + case OP_OPEN_BRACKET: + tree = parse_bracket_exp (regexp, dfa, token, syntax, err); + if (BE (*err != REG_NOERROR && tree == NULL, 0)) + return NULL; + break; + case OP_BACK_REF: + if (!BE (dfa->completed_bkref_map & (1 << token->opr.idx), 1)) + { + *err = REG_ESUBREG; + return NULL; + } + dfa->used_bkref_map |= 1 << token->opr.idx; + tree = create_token_tree (dfa, NULL, NULL, token); + if (BE (tree == NULL, 0)) + { + *err = REG_ESPACE; + return NULL; + } + ++dfa->nbackref; + dfa->has_mb_node = 1; + break; + case OP_OPEN_DUP_NUM: + if (syntax & RE_CONTEXT_INVALID_DUP) + { + *err = REG_BADRPT; + return NULL; + } + /* FALLTHROUGH */ + case OP_DUP_ASTERISK: + case OP_DUP_PLUS: + case OP_DUP_QUESTION: + if (syntax & RE_CONTEXT_INVALID_OPS) + { + *err = REG_BADRPT; + return NULL; + } + else if (syntax & RE_CONTEXT_INDEP_OPS) + { + fetch_token (token, regexp, syntax); + return parse_expression (regexp, preg, token, syntax, nest, err); + } + /* else fall through */ + case OP_CLOSE_SUBEXP: + if ((token->type == OP_CLOSE_SUBEXP) && + !(syntax & RE_UNMATCHED_RIGHT_PAREN_ORD)) + { + *err = REG_ERPAREN; + return NULL; + } + /* else fall through */ + case OP_CLOSE_DUP_NUM: + /* We treat it as a normal character. */ + + /* Then we can these characters as normal characters. */ + token->type = CHARACTER; + /* mb_partial and word_char bits should be initialized already + by peek_token. */ + tree = create_token_tree (dfa, NULL, NULL, token); + if (BE (tree == NULL, 0)) + { + *err = REG_ESPACE; + return NULL; + } + break; + case ANCHOR: + if ((token->opr.ctx_type + & (WORD_DELIM | NOT_WORD_DELIM | WORD_FIRST | WORD_LAST)) + && dfa->word_ops_used == 0) + init_word_char (dfa); + if (token->opr.ctx_type == WORD_DELIM + || token->opr.ctx_type == NOT_WORD_DELIM) + { + bin_tree_t *tree_first, *tree_last; + if (token->opr.ctx_type == WORD_DELIM) + { + token->opr.ctx_type = WORD_FIRST; + tree_first = create_token_tree (dfa, NULL, NULL, token); + token->opr.ctx_type = WORD_LAST; + } + else + { + token->opr.ctx_type = INSIDE_WORD; + tree_first = create_token_tree (dfa, NULL, NULL, token); + token->opr.ctx_type = INSIDE_NOTWORD; + } + tree_last = create_token_tree (dfa, NULL, NULL, token); + tree = create_tree (dfa, tree_first, tree_last, OP_ALT); + if (BE (tree_first == NULL || tree_last == NULL || tree == NULL, 0)) + { + *err = REG_ESPACE; + return NULL; + } + } + else + { + tree = create_token_tree (dfa, NULL, NULL, token); + if (BE (tree == NULL, 0)) + { + *err = REG_ESPACE; + return NULL; + } + } + /* We must return here, since ANCHORs can't be followed + by repetition operators. + eg. RE"^*" is invalid or "<ANCHOR(^)><CHAR(*)>", + it must not be "<ANCHOR(^)><REPEAT(*)>". */ + fetch_token (token, regexp, syntax); + return tree; + case OP_PERIOD: + tree = create_token_tree (dfa, NULL, NULL, token); + if (BE (tree == NULL, 0)) + { + *err = REG_ESPACE; + return NULL; + } + if (dfa->mb_cur_max > 1) + dfa->has_mb_node = 1; + break; + case OP_WORD: + case OP_NOTWORD: + tree = build_charclass_op (dfa, regexp->trans, + "alnum", + "_", + token->type == OP_NOTWORD, err); + if (BE (*err != REG_NOERROR && tree == NULL, 0)) + return NULL; + break; + case OP_SPACE: + case OP_NOTSPACE: + tree = build_charclass_op (dfa, regexp->trans, + "space", + "", + token->type == OP_NOTSPACE, err); + if (BE (*err != REG_NOERROR && tree == NULL, 0)) + return NULL; + break; + case OP_ALT: + case END_OF_RE: + return NULL; + case BACK_SLASH: + *err = REG_EESCAPE; + return NULL; + default: + /* Must not happen? */ +#ifdef DEBUG + assert (0); +#endif + return NULL; + } + fetch_token (token, regexp, syntax); + + while (token->type == OP_DUP_ASTERISK || token->type == OP_DUP_PLUS + || token->type == OP_DUP_QUESTION || token->type == OP_OPEN_DUP_NUM) + { + bin_tree_t *dup_tree = parse_dup_op (tree, regexp, dfa, token, syntax, err); + if (BE (*err != REG_NOERROR && dup_tree == NULL, 0)) + { + if (tree != NULL) + postorder (tree, free_tree, NULL); + return NULL; + } + tree = dup_tree; + /* In BRE consecutive duplications are not allowed. */ + if ((syntax & RE_CONTEXT_INVALID_DUP) + && (token->type == OP_DUP_ASTERISK + || token->type == OP_OPEN_DUP_NUM)) + { + if (tree != NULL) + postorder (tree, free_tree, NULL); + *err = REG_BADRPT; + return NULL; + } + } + + return tree; +} + +/* This function build the following tree, from regular expression + (<reg_exp>): + SUBEXP + | + <reg_exp> +*/ + +static bin_tree_t * +parse_sub_exp (re_string_t *regexp, regex_t *preg, re_token_t *token, + reg_syntax_t syntax, int nest, reg_errcode_t *err) +{ + re_dfa_t *dfa = (re_dfa_t *) preg->buffer; + bin_tree_t *tree; + size_t cur_nsub; + cur_nsub = preg->re_nsub++; + + fetch_token (token, regexp, syntax | RE_CARET_ANCHORS_HERE); + + /* The subexpression may be a null string. */ + if (token->type == OP_CLOSE_SUBEXP) + tree = NULL; + else + { + tree = parse_reg_exp (regexp, preg, token, syntax, nest, err); + if (BE (*err == REG_NOERROR && token->type != OP_CLOSE_SUBEXP, 0)) + { + if (tree != NULL) + postorder (tree, free_tree, NULL); + *err = REG_EPAREN; + } + if (BE (*err != REG_NOERROR, 0)) + return NULL; + } + + if (cur_nsub <= '9' - '1') + dfa->completed_bkref_map |= 1 << cur_nsub; + + tree = create_tree (dfa, tree, NULL, SUBEXP); + if (BE (tree == NULL, 0)) + { + *err = REG_ESPACE; + return NULL; + } + tree->token.opr.idx = cur_nsub; + return tree; +} + +/* This function parse repetition operators like "*", "+", "{1,3}" etc. */ + +static bin_tree_t * +parse_dup_op (bin_tree_t *elem, re_string_t *regexp, re_dfa_t *dfa, + re_token_t *token, reg_syntax_t syntax, reg_errcode_t *err) +{ + bin_tree_t *tree = NULL, *old_tree = NULL; + int i, start, end, start_idx = re_string_cur_idx (regexp); + re_token_t start_token = *token; + + if (token->type == OP_OPEN_DUP_NUM) + { + end = 0; + start = fetch_number (regexp, token, syntax); + if (start == -1) + { + if (token->type == CHARACTER && token->opr.c == ',') + start = 0; /* We treat "{,m}" as "{0,m}". */ + else + { + *err = REG_BADBR; /* <re>{} is invalid. */ + return NULL; + } + } + if (BE (start != -2, 1)) + { + /* We treat "{n}" as "{n,n}". */ + end = ((token->type == OP_CLOSE_DUP_NUM) ? start + : ((token->type == CHARACTER && token->opr.c == ',') + ? fetch_number (regexp, token, syntax) : -2)); + } + if (BE (start == -2 || end == -2, 0)) + { + /* Invalid sequence. */ + if (BE (!(syntax & RE_INVALID_INTERVAL_ORD), 0)) + { + if (token->type == END_OF_RE) + *err = REG_EBRACE; + else + *err = REG_BADBR; + + return NULL; + } + + /* If the syntax bit is set, rollback. */ + re_string_set_index (regexp, start_idx); + *token = start_token; + token->type = CHARACTER; + /* mb_partial and word_char bits should be already initialized by + peek_token. */ + return elem; + } + + if (BE ((end != -1 && start > end) || token->type != OP_CLOSE_DUP_NUM, 0)) + { + /* First number greater than second. */ + *err = REG_BADBR; + return NULL; + } + } + else + { + start = (token->type == OP_DUP_PLUS) ? 1 : 0; + end = (token->type == OP_DUP_QUESTION) ? 1 : -1; + } + + fetch_token (token, regexp, syntax); + + if (BE (elem == NULL, 0)) + return NULL; + if (BE (start == 0 && end == 0, 0)) + { + postorder (elem, free_tree, NULL); + return NULL; + } + + /* Extract "<re>{n,m}" to "<re><re>...<re><re>{0,<m-n>}". */ + if (BE (start > 0, 0)) + { + tree = elem; + for (i = 2; i <= start; ++i) + { + elem = duplicate_tree (elem, dfa); + tree = create_tree (dfa, tree, elem, CONCAT); + if (BE (elem == NULL || tree == NULL, 0)) + goto parse_dup_op_espace; + } + + if (start == end) + return tree; + + /* Duplicate ELEM before it is marked optional. */ + elem = duplicate_tree (elem, dfa); + if (BE (elem == NULL, 0)) + goto parse_dup_op_espace; + old_tree = tree; + } + else + old_tree = NULL; + + if (elem->token.type == SUBEXP) + postorder (elem, mark_opt_subexp, (void *) (long) elem->token.opr.idx); + + tree = create_tree (dfa, elem, NULL, (end == -1 ? OP_DUP_ASTERISK : OP_ALT)); + if (BE (tree == NULL, 0)) + goto parse_dup_op_espace; + + /* This loop is actually executed only when end != -1, + to rewrite <re>{0,n} as (<re>(<re>...<re>?)?)?... We have + already created the start+1-th copy. */ + for (i = start + 2; i <= end; ++i) + { + elem = duplicate_tree (elem, dfa); + tree = create_tree (dfa, tree, elem, CONCAT); + if (BE (elem == NULL || tree == NULL, 0)) + goto parse_dup_op_espace; + + tree = create_tree (dfa, tree, NULL, OP_ALT); + if (BE (tree == NULL, 0)) + goto parse_dup_op_espace; + } + + if (old_tree) + tree = create_tree (dfa, old_tree, tree, CONCAT); + + return tree; + + parse_dup_op_espace: + *err = REG_ESPACE; + return NULL; +} + +/* Size of the names for collating symbol/equivalence_class/character_class. + I'm not sure, but maybe enough. */ +#define BRACKET_NAME_BUF_SIZE 32 + +#ifndef _LIBC + /* Local function for parse_bracket_exp only used in case of NOT _LIBC. + Build the range expression which starts from START_ELEM, and ends + at END_ELEM. The result are written to MBCSET and SBCSET. + RANGE_ALLOC is the allocated size of mbcset->range_starts, and + mbcset->range_ends, is a pointer argument since we may + update it. */ + +static reg_errcode_t +internal_function +# ifdef RE_ENABLE_I18N +build_range_exp (reg_syntax_t syntax, bitset_t sbcset, re_charset_t *mbcset, + int *range_alloc, bracket_elem_t *start_elem, + bracket_elem_t *end_elem) +# else /* not RE_ENABLE_I18N */ +build_range_exp (reg_syntax_t syntax, bitset_t sbcset, + bracket_elem_t *start_elem, bracket_elem_t *end_elem) +# endif /* not RE_ENABLE_I18N */ +{ + unsigned int start_ch, end_ch; + /* Equivalence Classes and Character Classes can't be a range start/end. */ + if (BE (start_elem->type == EQUIV_CLASS || start_elem->type == CHAR_CLASS + || end_elem->type == EQUIV_CLASS || end_elem->type == CHAR_CLASS, + 0)) + return REG_ERANGE; + + /* We can handle no multi character collating elements without libc + support. */ + if (BE ((start_elem->type == COLL_SYM + && strlen ((char *) start_elem->opr.name) > 1) + || (end_elem->type == COLL_SYM + && strlen ((char *) end_elem->opr.name) > 1), 0)) + return REG_ECOLLATE; + +# ifdef RE_ENABLE_I18N + { + wchar_t wc; + wint_t start_wc; + wint_t end_wc; + + start_ch = ((start_elem->type == SB_CHAR) ? start_elem->opr.ch + : ((start_elem->type == COLL_SYM) ? start_elem->opr.name[0] + : 0)); + end_ch = ((end_elem->type == SB_CHAR) ? end_elem->opr.ch + : ((end_elem->type == COLL_SYM) ? end_elem->opr.name[0] + : 0)); +#ifdef GAWK + /* + * Fedora Core 2, maybe others, have broken `btowc' that returns -1 + * for any value > 127. Sigh. Note that `start_ch' and `end_ch' are + * unsigned, so we don't have sign extension problems. + */ + start_wc = ((start_elem->type == SB_CHAR || start_elem->type == COLL_SYM) + ? start_ch : start_elem->opr.wch); + end_wc = ((end_elem->type == SB_CHAR || end_elem->type == COLL_SYM) + ? end_ch : end_elem->opr.wch); +#else + start_wc = ((start_elem->type == SB_CHAR || start_elem->type == COLL_SYM) + ? __btowc (start_ch) : start_elem->opr.wch); + end_wc = ((end_elem->type == SB_CHAR || end_elem->type == COLL_SYM) + ? __btowc (end_ch) : end_elem->opr.wch); +#endif + if (start_wc == WEOF || end_wc == WEOF) + return REG_ECOLLATE; + else if (BE ((syntax & RE_NO_EMPTY_RANGES) && start_wc > end_wc, 0)) + return REG_ERANGE; + + /* Got valid collation sequence values, add them as a new entry. + However, for !_LIBC we have no collation elements: if the + character set is single byte, the single byte character set + that we build below suffices. parse_bracket_exp passes + no MBCSET if dfa->mb_cur_max == 1. */ + if (mbcset) + { + /* Check the space of the arrays. */ + if (BE (*range_alloc == mbcset->nranges, 0)) + { + /* There is not enough space, need realloc. */ + wchar_t *new_array_start, *new_array_end; + int new_nranges; + + /* +1 in case of mbcset->nranges is 0. */ + new_nranges = 2 * mbcset->nranges + 1; + /* Use realloc since mbcset->range_starts and mbcset->range_ends + are NULL if *range_alloc == 0. */ + new_array_start = re_realloc (mbcset->range_starts, wchar_t, + new_nranges); + new_array_end = re_realloc (mbcset->range_ends, wchar_t, + new_nranges); + + if (BE (new_array_start == NULL || new_array_end == NULL, 0)) + { + /* if one is not NULL, free it to avoid leaks */ + if (new_array_start != NULL) + re_free(new_array_start); + if (new_array_end != NULL) + re_free(new_array_end); + return REG_ESPACE; + } + + mbcset->range_starts = new_array_start; + mbcset->range_ends = new_array_end; + *range_alloc = new_nranges; + } + + mbcset->range_starts[mbcset->nranges] = start_wc; + mbcset->range_ends[mbcset->nranges++] = end_wc; + } + + /* Build the table for single byte characters. */ + for (wc = 0; wc < SBC_MAX; ++wc) + { + if (start_wc <= wc && wc <= end_wc) + bitset_set (sbcset, wc); + } + } +# else /* not RE_ENABLE_I18N */ + { + unsigned int ch; + start_ch = ((start_elem->type == SB_CHAR ) ? start_elem->opr.ch + : ((start_elem->type == COLL_SYM) ? start_elem->opr.name[0] + : 0)); + end_ch = ((end_elem->type == SB_CHAR ) ? end_elem->opr.ch + : ((end_elem->type == COLL_SYM) ? end_elem->opr.name[0] + : 0)); + if (start_ch > end_ch) + return REG_ERANGE; + /* Build the table for single byte characters. */ + for (ch = 0; ch < SBC_MAX; ++ch) + if (start_ch <= ch && ch <= end_ch) + bitset_set (sbcset, ch); + } +# endif /* not RE_ENABLE_I18N */ + return REG_NOERROR; +} +#endif /* not _LIBC */ + +#ifndef _LIBC +/* Helper function for parse_bracket_exp only used in case of NOT _LIBC.. + Build the collating element which is represented by NAME. + The result are written to MBCSET and SBCSET. + COLL_SYM_ALLOC is the allocated size of mbcset->coll_sym, is a + pointer argument since we may update it. */ + +static reg_errcode_t +internal_function +# ifdef RE_ENABLE_I18N +build_collating_symbol (bitset_t sbcset, re_charset_t *mbcset, + int *coll_sym_alloc, const unsigned char *name) +# else /* not RE_ENABLE_I18N */ +build_collating_symbol (bitset_t sbcset, const unsigned char *name) +# endif /* not RE_ENABLE_I18N */ +{ + size_t name_len = strlen ((const char *) name); + if (BE (name_len != 1, 0)) + return REG_ECOLLATE; + else + { + bitset_set (sbcset, name[0]); + return REG_NOERROR; + } +} +#endif /* not _LIBC */ + +/* This function parse bracket expression like "[abc]", "[a-c]", + "[[.a-a.]]" etc. */ + +static bin_tree_t * +parse_bracket_exp (re_string_t *regexp, re_dfa_t *dfa, re_token_t *token, + reg_syntax_t syntax, reg_errcode_t *err) +{ +#ifdef _LIBC + const unsigned char *collseqmb; + const char *collseqwc; + uint32_t nrules; + int32_t table_size; + const int32_t *symb_table; + const unsigned char *extra; + + /* Local function for parse_bracket_exp used in _LIBC environment. + Seek the collating symbol entry corresponding to NAME. + Return the index of the symbol in the SYMB_TABLE, + or -1 if not found. */ + + auto inline int32_t + __attribute__ ((always_inline)) + seek_collating_symbol_entry (const unsigned char *name, size_t name_len) + { + int32_t elem; + + for (elem = 0; elem < table_size; elem++) + if (symb_table[2 * elem] != 0) + { + int32_t idx = symb_table[2 * elem + 1]; + /* Skip the name of collating element name. */ + idx += 1 + extra[idx]; + if (/* Compare the length of the name. */ + name_len == extra[idx] + /* Compare the name. */ + && memcmp (name, &extra[idx + 1], name_len) == 0) + /* Yep, this is the entry. */ + return elem; + } + return -1; + } + + /* Local function for parse_bracket_exp used in _LIBC environment. + Look up the collation sequence value of BR_ELEM. + Return the value if succeeded, UINT_MAX otherwise. */ + + auto inline unsigned int + __attribute__ ((always_inline)) + lookup_collation_sequence_value (bracket_elem_t *br_elem) + { + if (br_elem->type == SB_CHAR) + { + /* + if (MB_CUR_MAX == 1) + */ + if (nrules == 0) + return collseqmb[br_elem->opr.ch]; + else + { + wint_t wc = __btowc (br_elem->opr.ch); + return __collseq_table_lookup (collseqwc, wc); + } + } + else if (br_elem->type == MB_CHAR) + { + if (nrules != 0) + return __collseq_table_lookup (collseqwc, br_elem->opr.wch); + } + else if (br_elem->type == COLL_SYM) + { + size_t sym_name_len = strlen ((char *) br_elem->opr.name); + if (nrules != 0) + { + int32_t elem, idx; + elem = seek_collating_symbol_entry (br_elem->opr.name, + sym_name_len); + if (elem != -1) + { + /* We found the entry. */ + idx = symb_table[2 * elem + 1]; + /* Skip the name of collating element name. */ + idx += 1 + extra[idx]; + /* Skip the byte sequence of the collating element. */ + idx += 1 + extra[idx]; + /* Adjust for the alignment. */ + idx = (idx + 3) & ~3; + /* Skip the multibyte collation sequence value. */ + idx += sizeof (unsigned int); + /* Skip the wide char sequence of the collating element. */ + idx += sizeof (unsigned int) * + (1 + *(unsigned int *) (extra + idx)); + /* Return the collation sequence value. */ + return *(unsigned int *) (extra + idx); + } + else if (sym_name_len == 1) + { + /* No valid character. Match it as a single byte + character. */ + return collseqmb[br_elem->opr.name[0]]; + } + } + else if (sym_name_len == 1) + return collseqmb[br_elem->opr.name[0]]; + } + return UINT_MAX; + } + + /* Local function for parse_bracket_exp used in _LIBC environment. + Build the range expression which starts from START_ELEM, and ends + at END_ELEM. The result are written to MBCSET and SBCSET. + RANGE_ALLOC is the allocated size of mbcset->range_starts, and + mbcset->range_ends, is a pointer argument since we may + update it. */ + + auto inline reg_errcode_t + __attribute__ ((always_inline)) + build_range_exp (bitset_t sbcset, re_charset_t *mbcset, int *range_alloc, + bracket_elem_t *start_elem, bracket_elem_t *end_elem) + { + unsigned int ch; + uint32_t start_collseq; + uint32_t end_collseq; + + /* Equivalence Classes and Character Classes can't be a range + start/end. */ + if (BE (start_elem->type == EQUIV_CLASS || start_elem->type == CHAR_CLASS + || end_elem->type == EQUIV_CLASS || end_elem->type == CHAR_CLASS, + 0)) + return REG_ERANGE; + + start_collseq = lookup_collation_sequence_value (start_elem); + end_collseq = lookup_collation_sequence_value (end_elem); + /* Check start/end collation sequence values. */ + if (BE (start_collseq == UINT_MAX || end_collseq == UINT_MAX, 0)) + return REG_ECOLLATE; + if (BE ((syntax & RE_NO_EMPTY_RANGES) && start_collseq > end_collseq, 0)) + return REG_ERANGE; + + /* Got valid collation sequence values, add them as a new entry. + However, if we have no collation elements, and the character set + is single byte, the single byte character set that we + build below suffices. */ + if (nrules > 0 || dfa->mb_cur_max > 1) + { + /* Check the space of the arrays. */ + if (BE (*range_alloc == mbcset->nranges, 0)) + { + /* There is not enough space, need realloc. */ + uint32_t *new_array_start; + uint32_t *new_array_end; + int new_nranges; + + /* +1 in case of mbcset->nranges is 0. */ + new_nranges = 2 * mbcset->nranges + 1; + new_array_start = re_realloc (mbcset->range_starts, uint32_t, + new_nranges); + new_array_end = re_realloc (mbcset->range_ends, uint32_t, + new_nranges); + + if (BE (new_array_start == NULL || new_array_end == NULL, 0)) + return REG_ESPACE; + + mbcset->range_starts = new_array_start; + mbcset->range_ends = new_array_end; + *range_alloc = new_nranges; + } + + mbcset->range_starts[mbcset->nranges] = start_collseq; + mbcset->range_ends[mbcset->nranges++] = end_collseq; + } + + /* Build the table for single byte characters. */ + for (ch = 0; ch < SBC_MAX; ch++) + { + uint32_t ch_collseq; + /* + if (MB_CUR_MAX == 1) + */ + if (nrules == 0) + ch_collseq = collseqmb[ch]; + else + ch_collseq = __collseq_table_lookup (collseqwc, __btowc (ch)); + if (start_collseq <= ch_collseq && ch_collseq <= end_collseq) + bitset_set (sbcset, ch); + } + return REG_NOERROR; + } + + /* Local function for parse_bracket_exp used in _LIBC environment. + Build the collating element which is represented by NAME. + The result are written to MBCSET and SBCSET. + COLL_SYM_ALLOC is the allocated size of mbcset->coll_sym, is a + pointer argument since we may update it. */ + + auto inline reg_errcode_t + __attribute__ ((always_inline)) + build_collating_symbol (bitset_t sbcset, re_charset_t *mbcset, + int *coll_sym_alloc, const unsigned char *name) + { + int32_t elem, idx; + size_t name_len = strlen ((const char *) name); + if (nrules != 0) + { + elem = seek_collating_symbol_entry (name, name_len); + if (elem != -1) + { + /* We found the entry. */ + idx = symb_table[2 * elem + 1]; + /* Skip the name of collating element name. */ + idx += 1 + extra[idx]; + } + else if (name_len == 1) + { + /* No valid character, treat it as a normal + character. */ + bitset_set (sbcset, name[0]); + return REG_NOERROR; + } + else + return REG_ECOLLATE; + + /* Got valid collation sequence, add it as a new entry. */ + /* Check the space of the arrays. */ + if (BE (*coll_sym_alloc == mbcset->ncoll_syms, 0)) + { + /* Not enough, realloc it. */ + /* +1 in case of mbcset->ncoll_syms is 0. */ + int new_coll_sym_alloc = 2 * mbcset->ncoll_syms + 1; + /* Use realloc since mbcset->coll_syms is NULL + if *alloc == 0. */ + int32_t *new_coll_syms = re_realloc (mbcset->coll_syms, int32_t, + new_coll_sym_alloc); + if (BE (new_coll_syms == NULL, 0)) + return REG_ESPACE; + mbcset->coll_syms = new_coll_syms; + *coll_sym_alloc = new_coll_sym_alloc; + } + mbcset->coll_syms[mbcset->ncoll_syms++] = idx; + return REG_NOERROR; + } + else + { + if (BE (name_len != 1, 0)) + return REG_ECOLLATE; + else + { + bitset_set (sbcset, name[0]); + return REG_NOERROR; + } + } + } +#endif + + re_token_t br_token; + re_bitset_ptr_t sbcset; +#ifdef RE_ENABLE_I18N + re_charset_t *mbcset; + int coll_sym_alloc = 0, range_alloc = 0, mbchar_alloc = 0; + int equiv_class_alloc = 0, char_class_alloc = 0; +#endif /* not RE_ENABLE_I18N */ + int non_match = 0; + bin_tree_t *work_tree; + int token_len; + int first_round = 1; +#ifdef _LIBC + collseqmb = (const unsigned char *) + _NL_CURRENT (LC_COLLATE, _NL_COLLATE_COLLSEQMB); + nrules = _NL_CURRENT_WORD (LC_COLLATE, _NL_COLLATE_NRULES); + if (nrules) + { + /* + if (MB_CUR_MAX > 1) + */ + collseqwc = _NL_CURRENT (LC_COLLATE, _NL_COLLATE_COLLSEQWC); + table_size = _NL_CURRENT_WORD (LC_COLLATE, _NL_COLLATE_SYMB_HASH_SIZEMB); + symb_table = (const int32_t *) _NL_CURRENT (LC_COLLATE, + _NL_COLLATE_SYMB_TABLEMB); + extra = (const unsigned char *) _NL_CURRENT (LC_COLLATE, + _NL_COLLATE_SYMB_EXTRAMB); + } +#endif + sbcset = (re_bitset_ptr_t) calloc (sizeof (bitset_t), 1); +#ifdef RE_ENABLE_I18N + mbcset = (re_charset_t *) calloc (sizeof (re_charset_t), 1); +#endif /* RE_ENABLE_I18N */ +#ifdef RE_ENABLE_I18N + if (BE (sbcset == NULL || mbcset == NULL, 0)) +#else + if (BE (sbcset == NULL, 0)) +#endif /* RE_ENABLE_I18N */ + { + re_free (sbcset); +#ifdef RE_ENABLE_I18N + re_free (mbcset); +#endif + *err = REG_ESPACE; + return NULL; + } + + token_len = peek_token_bracket (token, regexp, syntax); + if (BE (token->type == END_OF_RE, 0)) + { + *err = REG_BADPAT; + goto parse_bracket_exp_free_return; + } + if (token->type == OP_NON_MATCH_LIST) + { +#ifdef RE_ENABLE_I18N + mbcset->non_match = 1; +#endif /* not RE_ENABLE_I18N */ + non_match = 1; + if (syntax & RE_HAT_LISTS_NOT_NEWLINE) + bitset_set (sbcset, '\n'); + re_string_skip_bytes (regexp, token_len); /* Skip a token. */ + token_len = peek_token_bracket (token, regexp, syntax); + if (BE (token->type == END_OF_RE, 0)) + { + *err = REG_BADPAT; + goto parse_bracket_exp_free_return; + } + } + + /* We treat the first ']' as a normal character. */ + if (token->type == OP_CLOSE_BRACKET) + token->type = CHARACTER; + + while (1) + { + bracket_elem_t start_elem, end_elem; + unsigned char start_name_buf[BRACKET_NAME_BUF_SIZE]; + unsigned char end_name_buf[BRACKET_NAME_BUF_SIZE]; + reg_errcode_t ret; + int token_len2 = 0, is_range_exp = 0; + re_token_t token2; + + start_elem.opr.name = start_name_buf; + start_elem.type = COLL_SYM; + ret = parse_bracket_element (&start_elem, regexp, token, token_len, dfa, + syntax, first_round); + if (BE (ret != REG_NOERROR, 0)) + { + *err = ret; + goto parse_bracket_exp_free_return; + } + first_round = 0; + + /* Get information about the next token. We need it in any case. */ + token_len = peek_token_bracket (token, regexp, syntax); + + /* Do not check for ranges if we know they are not allowed. */ + if (start_elem.type != CHAR_CLASS && start_elem.type != EQUIV_CLASS) + { + if (BE (token->type == END_OF_RE, 0)) + { + *err = REG_EBRACK; + goto parse_bracket_exp_free_return; + } + if (token->type == OP_CHARSET_RANGE) + { + re_string_skip_bytes (regexp, token_len); /* Skip '-'. */ + token_len2 = peek_token_bracket (&token2, regexp, syntax); + if (BE (token2.type == END_OF_RE, 0)) + { + *err = REG_EBRACK; + goto parse_bracket_exp_free_return; + } + if (token2.type == OP_CLOSE_BRACKET) + { + /* We treat the last '-' as a normal character. */ + re_string_skip_bytes (regexp, -token_len); + token->type = CHARACTER; + } + else + is_range_exp = 1; + } + } + + if (is_range_exp == 1) + { + end_elem.opr.name = end_name_buf; + end_elem.type = COLL_SYM; + ret = parse_bracket_element (&end_elem, regexp, &token2, token_len2, + dfa, syntax, 1); + if (BE (ret != REG_NOERROR, 0)) + { + *err = ret; + goto parse_bracket_exp_free_return; + } + + token_len = peek_token_bracket (token, regexp, syntax); + +#ifdef _LIBC + *err = build_range_exp (syntax, sbcset, mbcset, &range_alloc, + &start_elem, &end_elem); +#else +# ifdef RE_ENABLE_I18N + *err = build_range_exp (syntax, sbcset, + dfa->mb_cur_max > 1 ? mbcset : NULL, + &range_alloc, &start_elem, &end_elem); +# else + *err = build_range_exp (syntax, sbcset, &start_elem, &end_elem); +# endif +#endif /* RE_ENABLE_I18N */ + if (BE (*err != REG_NOERROR, 0)) + goto parse_bracket_exp_free_return; + } + else + { + switch (start_elem.type) + { + case SB_CHAR: + bitset_set (sbcset, start_elem.opr.ch); + break; +#ifdef RE_ENABLE_I18N + case MB_CHAR: + /* Check whether the array has enough space. */ + if (BE (mbchar_alloc == mbcset->nmbchars, 0)) + { + wchar_t *new_mbchars; + /* Not enough, realloc it. */ + /* +1 in case of mbcset->nmbchars is 0. */ + mbchar_alloc = 2 * mbcset->nmbchars + 1; + /* Use realloc since array is NULL if *alloc == 0. */ + new_mbchars = re_realloc (mbcset->mbchars, wchar_t, + mbchar_alloc); + if (BE (new_mbchars == NULL, 0)) + goto parse_bracket_exp_espace; + mbcset->mbchars = new_mbchars; + } + mbcset->mbchars[mbcset->nmbchars++] = start_elem.opr.wch; + break; +#endif /* RE_ENABLE_I18N */ + case EQUIV_CLASS: + *err = build_equiv_class (sbcset, +#ifdef RE_ENABLE_I18N + mbcset, &equiv_class_alloc, +#endif /* RE_ENABLE_I18N */ + start_elem.opr.name); + if (BE (*err != REG_NOERROR, 0)) + goto parse_bracket_exp_free_return; + break; + case COLL_SYM: + *err = build_collating_symbol (sbcset, +#ifdef RE_ENABLE_I18N + mbcset, &coll_sym_alloc, +#endif /* RE_ENABLE_I18N */ + start_elem.opr.name); + if (BE (*err != REG_NOERROR, 0)) + goto parse_bracket_exp_free_return; + break; + case CHAR_CLASS: + *err = build_charclass (regexp->trans, sbcset, +#ifdef RE_ENABLE_I18N + mbcset, &char_class_alloc, +#endif /* RE_ENABLE_I18N */ + (const char *) start_elem.opr.name, syntax); + if (BE (*err != REG_NOERROR, 0)) + goto parse_bracket_exp_free_return; + break; + default: + assert (0); + break; + } + } + if (BE (token->type == END_OF_RE, 0)) + { + *err = REG_EBRACK; + goto parse_bracket_exp_free_return; + } + if (token->type == OP_CLOSE_BRACKET) + break; + } + + re_string_skip_bytes (regexp, token_len); /* Skip a token. */ + + /* If it is non-matching list. */ + if (non_match) + bitset_not (sbcset); + +#ifdef RE_ENABLE_I18N + /* Ensure only single byte characters are set. */ + if (dfa->mb_cur_max > 1) + bitset_mask (sbcset, dfa->sb_char); + + if (mbcset->nmbchars || mbcset->ncoll_syms || mbcset->nequiv_classes + || mbcset->nranges || (dfa->mb_cur_max > 1 && (mbcset->nchar_classes + || mbcset->non_match))) + { + bin_tree_t *mbc_tree; + int sbc_idx; + /* Build a tree for complex bracket. */ + dfa->has_mb_node = 1; + br_token.type = COMPLEX_BRACKET; + br_token.opr.mbcset = mbcset; + mbc_tree = create_token_tree (dfa, NULL, NULL, &br_token); + if (BE (mbc_tree == NULL, 0)) + goto parse_bracket_exp_espace; + for (sbc_idx = 0; sbc_idx < BITSET_WORDS; ++sbc_idx) + if (sbcset[sbc_idx]) + break; + /* If there are no bits set in sbcset, there is no point + of having both SIMPLE_BRACKET and COMPLEX_BRACKET. */ + if (sbc_idx < BITSET_WORDS) + { + /* Build a tree for simple bracket. */ + br_token.type = SIMPLE_BRACKET; + br_token.opr.sbcset = sbcset; + work_tree = create_token_tree (dfa, NULL, NULL, &br_token); + if (BE (work_tree == NULL, 0)) + goto parse_bracket_exp_espace; + + /* Then join them by ALT node. */ + work_tree = create_tree (dfa, work_tree, mbc_tree, OP_ALT); + if (BE (work_tree == NULL, 0)) + goto parse_bracket_exp_espace; + } + else + { + re_free (sbcset); + work_tree = mbc_tree; + } + } + else +#endif /* not RE_ENABLE_I18N */ + { +#ifdef RE_ENABLE_I18N + free_charset (mbcset); +#endif + /* Build a tree for simple bracket. */ + br_token.type = SIMPLE_BRACKET; + br_token.opr.sbcset = sbcset; + work_tree = create_token_tree (dfa, NULL, NULL, &br_token); + if (BE (work_tree == NULL, 0)) + goto parse_bracket_exp_espace; + } + return work_tree; + + parse_bracket_exp_espace: + *err = REG_ESPACE; + parse_bracket_exp_free_return: + re_free (sbcset); +#ifdef RE_ENABLE_I18N + free_charset (mbcset); +#endif /* RE_ENABLE_I18N */ + return NULL; +} + +/* Parse an element in the bracket expression. */ + +static reg_errcode_t +parse_bracket_element (bracket_elem_t *elem, re_string_t *regexp, + re_token_t *token, int token_len, re_dfa_t *dfa, + reg_syntax_t syntax, int accept_hyphen) +{ +#ifdef RE_ENABLE_I18N + int cur_char_size; + cur_char_size = re_string_char_size_at (regexp, re_string_cur_idx (regexp)); + if (cur_char_size > 1) + { + elem->type = MB_CHAR; + elem->opr.wch = re_string_wchar_at (regexp, re_string_cur_idx (regexp)); + re_string_skip_bytes (regexp, cur_char_size); + return REG_NOERROR; + } +#endif /* RE_ENABLE_I18N */ + re_string_skip_bytes (regexp, token_len); /* Skip a token. */ + if (token->type == OP_OPEN_COLL_ELEM || token->type == OP_OPEN_CHAR_CLASS + || token->type == OP_OPEN_EQUIV_CLASS) + return parse_bracket_symbol (elem, regexp, token); + if (BE (token->type == OP_CHARSET_RANGE, 0) && !accept_hyphen) + { + /* A '-' must only appear as anything but a range indicator before + the closing bracket. Everything else is an error. */ + re_token_t token2; + (void) peek_token_bracket (&token2, regexp, syntax); + if (token2.type != OP_CLOSE_BRACKET) + /* The actual error value is not standardized since this whole + case is undefined. But ERANGE makes good sense. */ + return REG_ERANGE; + } + elem->type = SB_CHAR; + elem->opr.ch = token->opr.c; + return REG_NOERROR; +} + +/* Parse a bracket symbol in the bracket expression. Bracket symbols are + such as [:<character_class>:], [.<collating_element>.], and + [=<equivalent_class>=]. */ + +static reg_errcode_t +parse_bracket_symbol (bracket_elem_t *elem, re_string_t *regexp, + re_token_t *token) +{ + unsigned char ch, delim = token->opr.c; + int i = 0; + if (re_string_eoi(regexp)) + return REG_EBRACK; + for (;; ++i) + { + if (i >= BRACKET_NAME_BUF_SIZE) + return REG_EBRACK; + if (token->type == OP_OPEN_CHAR_CLASS) + ch = re_string_fetch_byte_case (regexp); + else + ch = re_string_fetch_byte (regexp); + if (re_string_eoi(regexp)) + return REG_EBRACK; + if (ch == delim && re_string_peek_byte (regexp, 0) == ']') + break; + elem->opr.name[i] = ch; + } + re_string_skip_bytes (regexp, 1); + elem->opr.name[i] = '\0'; + switch (token->type) + { + case OP_OPEN_COLL_ELEM: + elem->type = COLL_SYM; + break; + case OP_OPEN_EQUIV_CLASS: + elem->type = EQUIV_CLASS; + break; + case OP_OPEN_CHAR_CLASS: + elem->type = CHAR_CLASS; + break; + default: + break; + } + return REG_NOERROR; +} + + /* Helper function for parse_bracket_exp. + Build the equivalence class which is represented by NAME. + The result are written to MBCSET and SBCSET. + EQUIV_CLASS_ALLOC is the allocated size of mbcset->equiv_classes, + is a pointer argument since we may update it. */ + +static reg_errcode_t +#ifdef RE_ENABLE_I18N +build_equiv_class (bitset_t sbcset, re_charset_t *mbcset, + int *equiv_class_alloc, const unsigned char *name) +#else /* not RE_ENABLE_I18N */ +build_equiv_class (bitset_t sbcset, const unsigned char *name) +#endif /* not RE_ENABLE_I18N */ +{ +#ifdef _LIBC + uint32_t nrules = _NL_CURRENT_WORD (LC_COLLATE, _NL_COLLATE_NRULES); + if (nrules != 0) + { + const int32_t *table, *indirect; + const unsigned char *weights, *extra, *cp; + unsigned char char_buf[2]; + int32_t idx1, idx2; + unsigned int ch; + size_t len; + /* Calculate the index for equivalence class. */ + cp = name; + table = (const int32_t *) _NL_CURRENT (LC_COLLATE, _NL_COLLATE_TABLEMB); + weights = (const unsigned char *) _NL_CURRENT (LC_COLLATE, + _NL_COLLATE_WEIGHTMB); + extra = (const unsigned char *) _NL_CURRENT (LC_COLLATE, + _NL_COLLATE_EXTRAMB); + indirect = (const int32_t *) _NL_CURRENT (LC_COLLATE, + _NL_COLLATE_INDIRECTMB); + idx1 = findidx (table, indirect, extra, &cp, -1); + if (BE (idx1 == 0 || *cp != '\0', 0)) + /* This isn't a valid character. */ + return REG_ECOLLATE; + + /* Build single byte matching table for this equivalence class. */ + char_buf[1] = (unsigned char) '\0'; + len = weights[idx1 & 0xffffff]; + for (ch = 0; ch < SBC_MAX; ++ch) + { + char_buf[0] = ch; + cp = char_buf; + idx2 = findidx (table, indirect, extra, &cp, 1); +/* + idx2 = table[ch]; +*/ + if (idx2 == 0) + /* This isn't a valid character. */ + continue; + /* Compare only if the length matches and the collation rule + index is the same. */ + if (len == weights[idx2 & 0xffffff] && (idx1 >> 24) == (idx2 >> 24)) + { + int cnt = 0; + + while (cnt <= len && + weights[(idx1 & 0xffffff) + 1 + cnt] + == weights[(idx2 & 0xffffff) + 1 + cnt]) + ++cnt; + + if (cnt > len) + bitset_set (sbcset, ch); + } + } + /* Check whether the array has enough space. */ + if (BE (*equiv_class_alloc == mbcset->nequiv_classes, 0)) + { + /* Not enough, realloc it. */ + /* +1 in case of mbcset->nequiv_classes is 0. */ + int new_equiv_class_alloc = 2 * mbcset->nequiv_classes + 1; + /* Use realloc since the array is NULL if *alloc == 0. */ + int32_t *new_equiv_classes = re_realloc (mbcset->equiv_classes, + int32_t, + new_equiv_class_alloc); + if (BE (new_equiv_classes == NULL, 0)) + return REG_ESPACE; + mbcset->equiv_classes = new_equiv_classes; + *equiv_class_alloc = new_equiv_class_alloc; + } + mbcset->equiv_classes[mbcset->nequiv_classes++] = idx1; + } + else +#endif /* _LIBC */ + { + if (BE (strlen ((const char *) name) != 1, 0)) + return REG_ECOLLATE; + bitset_set (sbcset, *name); + } + return REG_NOERROR; +} + + /* Helper function for parse_bracket_exp. + Build the character class which is represented by NAME. + The result are written to MBCSET and SBCSET. + CHAR_CLASS_ALLOC is the allocated size of mbcset->char_classes, + is a pointer argument since we may update it. */ + +static reg_errcode_t +#ifdef RE_ENABLE_I18N +build_charclass (RE_TRANSLATE_TYPE trans, bitset_t sbcset, + re_charset_t *mbcset, int *char_class_alloc, + const char *class_name, reg_syntax_t syntax) +#else /* not RE_ENABLE_I18N */ +build_charclass (RE_TRANSLATE_TYPE trans, bitset_t sbcset, + const char *class_name, reg_syntax_t syntax) +#endif /* not RE_ENABLE_I18N */ +{ + int i; + + /* In case of REG_ICASE "upper" and "lower" match the both of + upper and lower cases. */ + if ((syntax & RE_ICASE) + && (strcmp (class_name, "upper") == 0 || strcmp (class_name, "lower") == 0)) + class_name = "alpha"; + +#ifdef RE_ENABLE_I18N + /* Check the space of the arrays. */ + if (BE (*char_class_alloc == mbcset->nchar_classes, 0)) + { + /* Not enough, realloc it. */ + /* +1 in case of mbcset->nchar_classes is 0. */ + int new_char_class_alloc = 2 * mbcset->nchar_classes + 1; + /* Use realloc since array is NULL if *alloc == 0. */ + wctype_t *new_char_classes = re_realloc (mbcset->char_classes, wctype_t, + new_char_class_alloc); + if (BE (new_char_classes == NULL, 0)) + return REG_ESPACE; + mbcset->char_classes = new_char_classes; + *char_class_alloc = new_char_class_alloc; + } + mbcset->char_classes[mbcset->nchar_classes++] = __wctype (class_name); +#endif /* RE_ENABLE_I18N */ + +#define BUILD_CHARCLASS_LOOP(ctype_func) \ + do { \ + if (BE (trans != NULL, 0)) \ + { \ + for (i = 0; i < SBC_MAX; ++i) \ + if (ctype_func (i)) \ + bitset_set (sbcset, trans[i]); \ + } \ + else \ + { \ + for (i = 0; i < SBC_MAX; ++i) \ + if (ctype_func (i)) \ + bitset_set (sbcset, i); \ + } \ + } while (0) + + if (strcmp (class_name, "alnum") == 0) + BUILD_CHARCLASS_LOOP (isalnum); + else if (strcmp (class_name, "cntrl") == 0) + BUILD_CHARCLASS_LOOP (iscntrl); + else if (strcmp (class_name, "lower") == 0) + BUILD_CHARCLASS_LOOP (islower); + else if (strcmp (class_name, "space") == 0) + BUILD_CHARCLASS_LOOP (isspace); + else if (strcmp (class_name, "alpha") == 0) + BUILD_CHARCLASS_LOOP (isalpha); + else if (strcmp (class_name, "digit") == 0) + BUILD_CHARCLASS_LOOP (isdigit); + else if (strcmp (class_name, "print") == 0) + BUILD_CHARCLASS_LOOP (isprint); + else if (strcmp (class_name, "upper") == 0) + BUILD_CHARCLASS_LOOP (isupper); + else if (strcmp (class_name, "blank") == 0) +#ifndef GAWK + BUILD_CHARCLASS_LOOP (isblank); +#else + /* see comments above */ + BUILD_CHARCLASS_LOOP (is_blank); +#endif + else if (strcmp (class_name, "graph") == 0) + BUILD_CHARCLASS_LOOP (isgraph); + else if (strcmp (class_name, "punct") == 0) + BUILD_CHARCLASS_LOOP (ispunct); + else if (strcmp (class_name, "xdigit") == 0) + BUILD_CHARCLASS_LOOP (isxdigit); + else + return REG_ECTYPE; + + return REG_NOERROR; +} + +static bin_tree_t * +build_charclass_op (re_dfa_t *dfa, RE_TRANSLATE_TYPE trans, + const char *class_name, + const char *extra, int non_match, + reg_errcode_t *err) +{ + re_bitset_ptr_t sbcset; +#ifdef RE_ENABLE_I18N + re_charset_t *mbcset; + int alloc = 0; +#endif /* not RE_ENABLE_I18N */ + reg_errcode_t ret; + re_token_t br_token; + bin_tree_t *tree; + + sbcset = (re_bitset_ptr_t) calloc (sizeof (bitset_t), 1); +#ifdef RE_ENABLE_I18N + mbcset = (re_charset_t *) calloc (sizeof (re_charset_t), 1); +#endif /* RE_ENABLE_I18N */ + +#ifdef RE_ENABLE_I18N + if (BE (sbcset == NULL || mbcset == NULL, 0)) +#else /* not RE_ENABLE_I18N */ + if (BE (sbcset == NULL, 0)) +#endif /* not RE_ENABLE_I18N */ + { + /* if one is not NULL, free it to avoid leaks */ + if (sbcset != NULL) + free(sbcset); +#ifdef RE_ENABLE_I18N + if (mbcset != NULL) + free(mbcset); +#endif + *err = REG_ESPACE; + return NULL; + } + + if (non_match) + { +#ifdef RE_ENABLE_I18N + mbcset->non_match = 1; +#endif /* not RE_ENABLE_I18N */ + } + + /* We don't care the syntax in this case. */ + ret = build_charclass (trans, sbcset, +#ifdef RE_ENABLE_I18N + mbcset, &alloc, +#endif /* RE_ENABLE_I18N */ + class_name, 0); + + if (BE (ret != REG_NOERROR, 0)) + { + re_free (sbcset); +#ifdef RE_ENABLE_I18N + free_charset (mbcset); +#endif /* RE_ENABLE_I18N */ + *err = ret; + return NULL; + } + /* \w match '_' also. */ + for (; *extra; extra++) + bitset_set (sbcset, *extra); + + /* If it is non-matching list. */ + if (non_match) + bitset_not (sbcset); + +#ifdef RE_ENABLE_I18N + /* Ensure only single byte characters are set. */ + if (dfa->mb_cur_max > 1) + bitset_mask (sbcset, dfa->sb_char); +#endif + + /* Build a tree for simple bracket. */ + memset(& br_token, 0, sizeof(br_token)); /* silence "not initialized" errors froms static checkers */ + br_token.type = SIMPLE_BRACKET; + br_token.opr.sbcset = sbcset; + tree = create_token_tree (dfa, NULL, NULL, &br_token); + if (BE (tree == NULL, 0)) + goto build_word_op_espace; + +#ifdef RE_ENABLE_I18N + if (dfa->mb_cur_max > 1) + { + bin_tree_t *mbc_tree; + /* Build a tree for complex bracket. */ + br_token.type = COMPLEX_BRACKET; + br_token.opr.mbcset = mbcset; + dfa->has_mb_node = 1; + mbc_tree = create_token_tree (dfa, NULL, NULL, &br_token); + if (BE (mbc_tree == NULL, 0)) + goto build_word_op_espace; + /* Then join them by ALT node. */ + tree = create_tree (dfa, tree, mbc_tree, OP_ALT); + if (BE (mbc_tree != NULL, 1)) + return tree; + } + else + { + free_charset (mbcset); + return tree; + } +#else /* not RE_ENABLE_I18N */ + return tree; +#endif /* not RE_ENABLE_I18N */ + + build_word_op_espace: + re_free (sbcset); +#ifdef RE_ENABLE_I18N + free_charset (mbcset); +#endif /* RE_ENABLE_I18N */ + *err = REG_ESPACE; + return NULL; +} + +/* This is intended for the expressions like "a{1,3}". + Fetch a number from `input', and return the number. + Return -1, if the number field is empty like "{,1}". + Return -2, If an error is occured. */ + +static int +fetch_number (re_string_t *input, re_token_t *token, reg_syntax_t syntax) +{ + int num = -1; + unsigned char c; + while (1) + { + fetch_token (token, input, syntax); + c = token->opr.c; + if (BE (token->type == END_OF_RE, 0)) + return -2; + if (token->type == OP_CLOSE_DUP_NUM || c == ',') + break; + num = ((token->type != CHARACTER || c < '0' || '9' < c || num == -2) + ? -2 : ((num == -1) ? c - '0' : num * 10 + c - '0')); + num = (num > RE_DUP_MAX) ? -2 : num; + } + return num; +} + +#ifdef RE_ENABLE_I18N +static void +free_charset (re_charset_t *cset) +{ + re_free (cset->mbchars); +# ifdef _LIBC + re_free (cset->coll_syms); + re_free (cset->equiv_classes); + re_free (cset->range_starts); + re_free (cset->range_ends); +# endif + re_free (cset->char_classes); + re_free (cset); +} +#endif /* RE_ENABLE_I18N */ + +/* Functions for binary tree operation. */ + +/* Create a tree node. */ + +static bin_tree_t * +create_tree (re_dfa_t *dfa, bin_tree_t *left, bin_tree_t *right, + re_token_type_t type) +{ + re_token_t t; + memset(& t, 0, sizeof(t)); /* silence "not initialized" errors froms static checkers */ + t.type = type; + return create_token_tree (dfa, left, right, &t); +} + +static bin_tree_t * +create_token_tree (re_dfa_t *dfa, bin_tree_t *left, bin_tree_t *right, + const re_token_t *token) +{ + bin_tree_t *tree; + if (BE (dfa->str_tree_storage_idx == BIN_TREE_STORAGE_SIZE, 0)) + { + bin_tree_storage_t *storage = re_malloc (bin_tree_storage_t, 1); + + if (storage == NULL) + return NULL; + storage->next = dfa->str_tree_storage; + dfa->str_tree_storage = storage; + dfa->str_tree_storage_idx = 0; + } + tree = &dfa->str_tree_storage->data[dfa->str_tree_storage_idx++]; + + tree->parent = NULL; + tree->left = left; + tree->right = right; + tree->token = *token; + tree->token.duplicated = 0; + tree->token.opt_subexp = 0; + tree->first = NULL; + tree->next = NULL; + tree->node_idx = -1; + + if (left != NULL) + left->parent = tree; + if (right != NULL) + right->parent = tree; + return tree; +} + +/* Mark the tree SRC as an optional subexpression. + To be called from preorder or postorder. */ + +static reg_errcode_t +mark_opt_subexp (void *extra, bin_tree_t *node) +{ + int idx = (int) (long) extra; + if (node->token.type == SUBEXP && node->token.opr.idx == idx) + node->token.opt_subexp = 1; + + return REG_NOERROR; +} + +/* Free the allocated memory inside NODE. */ + +static void +free_token (re_token_t *node) +{ +#ifdef RE_ENABLE_I18N + if (node->type == COMPLEX_BRACKET && node->duplicated == 0) + free_charset (node->opr.mbcset); + else +#endif /* RE_ENABLE_I18N */ + if (node->type == SIMPLE_BRACKET && node->duplicated == 0) + re_free (node->opr.sbcset); +} + +/* Worker function for tree walking. Free the allocated memory inside NODE + and its children. */ + +static reg_errcode_t +free_tree (void *extra, bin_tree_t *node) +{ + free_token (&node->token); + return REG_NOERROR; +} + + +/* Duplicate the node SRC, and return new node. This is a preorder + visit similar to the one implemented by the generic visitor, but + we need more infrastructure to maintain two parallel trees --- so, + it's easier to duplicate. */ + +static bin_tree_t * +duplicate_tree (const bin_tree_t *root, re_dfa_t *dfa) +{ + const bin_tree_t *node; + bin_tree_t *dup_root; + bin_tree_t **p_new = &dup_root, *dup_node = root->parent; + + for (node = root; ; ) + { + /* Create a new tree and link it back to the current parent. */ + *p_new = create_token_tree (dfa, NULL, NULL, &node->token); + if (*p_new == NULL) + return NULL; + (*p_new)->parent = dup_node; + (*p_new)->token.duplicated = 1; + dup_node = *p_new; + + /* Go to the left node, or up and to the right. */ + if (node->left) + { + node = node->left; + p_new = &dup_node->left; + } + else + { + const bin_tree_t *prev = NULL; + while (node->right == prev || node->right == NULL) + { + prev = node; + node = node->parent; + dup_node = dup_node->parent; + if (!node) + return dup_root; + } + node = node->right; + p_new = &dup_node->right; + } + } +} diff --git a/support/regex.c b/support/regex.c new file mode 100644 index 00000000..9f133fab --- /dev/null +++ b/support/regex.c @@ -0,0 +1,85 @@ +/* Extended regular expression matching and search library. + Copyright (C) 2002-2016 Free Software Foundation, Inc. + This file is part of the GNU C Library. + Contributed by Isamu Hasegawa <isamu@yamato.ibm.com>. + + The GNU C Library is free software; you can redistribute it and/or + modify it under the terms of the GNU Lesser General Public + License as published by the Free Software Foundation; either + version 2.1 of the License, or (at your option) any later version. + + The GNU C Library is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public + License along with the GNU C Library; if not, see + <http://www.gnu.org/licenses/>. */ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +/* Make sure noone compiles this code with a C++ compiler. */ +#ifdef __cplusplus +# error "This is C code, use a C compiler" +#endif + +#ifdef _LIBC +/* We have to keep the namespace clean. */ +# define regfree(preg) __regfree (preg) +# define regexec(pr, st, nm, pm, ef) __regexec (pr, st, nm, pm, ef) +# define regcomp(preg, pattern, cflags) __regcomp (preg, pattern, cflags) +# define regerror(errcode, preg, errbuf, errbuf_size) \ + __regerror(errcode, preg, errbuf, errbuf_size) +# define re_set_registers(bu, re, nu, st, en) \ + __re_set_registers (bu, re, nu, st, en) +# define re_match_2(bufp, string1, size1, string2, size2, pos, regs, stop) \ + __re_match_2 (bufp, string1, size1, string2, size2, pos, regs, stop) +# define re_match(bufp, string, size, pos, regs) \ + __re_match (bufp, string, size, pos, regs) +# define re_search(bufp, string, size, startpos, range, regs) \ + __re_search (bufp, string, size, startpos, range, regs) +# define re_compile_pattern(pattern, length, bufp) \ + __re_compile_pattern (pattern, length, bufp) +# define re_set_syntax(syntax) __re_set_syntax (syntax) +# define re_search_2(bufp, st1, s1, st2, s2, startpos, range, regs, stop) \ + __re_search_2 (bufp, st1, s1, st2, s2, startpos, range, regs, stop) +# define re_compile_fastmap(bufp) __re_compile_fastmap (bufp) + +# include "../locale/localeinfo.h" +#endif + +/* On some systems, limits.h sets RE_DUP_MAX to a lower value than + GNU regex allows. Include it before <regex.h>, which correctly + #undefs RE_DUP_MAX and sets it to the right value. */ +#include <limits.h> + +/* This header defines the MIN and MAX macros. */ +#ifdef HAVE_SYS_PARAM_H +#include <sys/param.h> +#endif /* HAVE_SYS_PARAM_H */ + +#ifdef GAWK +#undef alloca +#define alloca alloca_is_bad_you_should_never_use_it +#endif +#include <regex.h> +#include "regex_internal.h" + +#include "regex_internal.c" +#ifndef HAVE_STDBOOL_H +#include "missing_d/gawkbool.h" +#endif +#include "regcomp.c" +#include "regexec.c" + +/* Binary backward compatibility. */ +#if _LIBC +# include <shlib-compat.h> +# if SHLIB_COMPAT (libc, GLIBC_2_0, GLIBC_2_3) +link_warning (re_max_failures, "the 're_max_failures' variable is obsolete and will go away.") +int re_max_failures = 2000; +# endif +#endif diff --git a/support/regex.h b/support/regex.h new file mode 100644 index 00000000..143b3afa --- /dev/null +++ b/support/regex.h @@ -0,0 +1,591 @@ +/* Definitions for data structures and routines for the regular + expression library. + Copyright (C) 1985, 1989-2016 Free Software Foundation, Inc. + This file is part of the GNU C Library. + + The GNU C Library is free software; you can redistribute it and/or + modify it under the terms of the GNU Lesser General Public + License as published by the Free Software Foundation; either + version 2.1 of the License, or (at your option) any later version. + + The GNU C Library is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public + License along with the GNU C Library; if not, see + <http://www.gnu.org/licenses/>. */ + +#ifndef _REGEX_H +#define _REGEX_H 1 + +#ifdef HAVE_STDDEF_H +#include <stddef.h> +#endif + +#ifdef HAVE_SYS_TYPES_H +#include <sys/types.h> +#endif + +#ifndef _LIBC +#define __USE_GNU 1 +#endif + +/* Allow the use in C++ code. */ +#ifdef __cplusplus +extern "C" { +#endif + +/* The following two types have to be signed and unsigned integer type + wide enough to hold a value of a pointer. For most ANSI compilers + ptrdiff_t and size_t should be likely OK. Still size of these two + types is 2 for Microsoft C. Ugh... */ +typedef long int s_reg_t; +typedef unsigned long int active_reg_t; + +/* The following bits are used to determine the regexp syntax we + recognize. The set/not-set meanings are chosen so that Emacs syntax + remains the value 0. The bits are given in alphabetical order, and + the definitions shifted by one from the previous bit; thus, when we + add or remove a bit, only one other definition need change. */ +typedef unsigned long int reg_syntax_t; + +#ifdef __USE_GNU +/* If this bit is not set, then \ inside a bracket expression is literal. + If set, then such a \ quotes the following character. */ +# define RE_BACKSLASH_ESCAPE_IN_LISTS ((unsigned long int) 1) + +/* If this bit is not set, then + and ? are operators, and \+ and \? are + literals. + If set, then \+ and \? are operators and + and ? are literals. */ +# define RE_BK_PLUS_QM (RE_BACKSLASH_ESCAPE_IN_LISTS << 1) + +/* If this bit is set, then character classes are supported. They are: + [:alpha:], [:upper:], [:lower:], [:digit:], [:alnum:], [:xdigit:], + [:space:], [:print:], [:punct:], [:graph:], and [:cntrl:]. + If not set, then character classes are not supported. */ +# define RE_CHAR_CLASSES (RE_BK_PLUS_QM << 1) + +/* If this bit is set, then ^ and $ are always anchors (outside bracket + expressions, of course). + If this bit is not set, then it depends: + ^ is an anchor if it is at the beginning of a regular + expression or after an open-group or an alternation operator; + $ is an anchor if it is at the end of a regular expression, or + before a close-group or an alternation operator. + + This bit could be (re)combined with RE_CONTEXT_INDEP_OPS, because + POSIX draft 11.2 says that * etc. in leading positions is undefined. + We already implemented a previous draft which made those constructs + invalid, though, so we haven't changed the code back. */ +# define RE_CONTEXT_INDEP_ANCHORS (RE_CHAR_CLASSES << 1) + +/* If this bit is set, then special characters are always special + regardless of where they are in the pattern. + If this bit is not set, then special characters are special only in + some contexts; otherwise they are ordinary. Specifically, + * + ? and intervals are only special when not after the beginning, + open-group, or alternation operator. */ +# define RE_CONTEXT_INDEP_OPS (RE_CONTEXT_INDEP_ANCHORS << 1) + +/* If this bit is set, then *, +, ?, and { cannot be first in an re or + immediately after an alternation or begin-group operator. */ +# define RE_CONTEXT_INVALID_OPS (RE_CONTEXT_INDEP_OPS << 1) + +/* If this bit is set, then . matches newline. + If not set, then it doesn't. */ +# define RE_DOT_NEWLINE (RE_CONTEXT_INVALID_OPS << 1) + +/* If this bit is set, then . doesn't match NUL. + If not set, then it does. */ +# define RE_DOT_NOT_NULL (RE_DOT_NEWLINE << 1) + +/* If this bit is set, nonmatching lists [^...] do not match newline. + If not set, they do. */ +# define RE_HAT_LISTS_NOT_NEWLINE (RE_DOT_NOT_NULL << 1) + +/* If this bit is set, either \{...\} or {...} defines an + interval, depending on RE_NO_BK_BRACES. + If not set, \{, \}, {, and } are literals. */ +# define RE_INTERVALS (RE_HAT_LISTS_NOT_NEWLINE << 1) + +/* If this bit is set, +, ? and | aren't recognized as operators. + If not set, they are. */ +# define RE_LIMITED_OPS (RE_INTERVALS << 1) + +/* If this bit is set, newline is an alternation operator. + If not set, newline is literal. */ +# define RE_NEWLINE_ALT (RE_LIMITED_OPS << 1) + +/* If this bit is set, then `{...}' defines an interval, and \{ and \} + are literals. + If not set, then `\{...\}' defines an interval. */ +# define RE_NO_BK_BRACES (RE_NEWLINE_ALT << 1) + +/* If this bit is set, (...) defines a group, and \( and \) are literals. + If not set, \(...\) defines a group, and ( and ) are literals. */ +# define RE_NO_BK_PARENS (RE_NO_BK_BRACES << 1) + +/* If this bit is set, then \<digit> matches <digit>. + If not set, then \<digit> is a back-reference. */ +# define RE_NO_BK_REFS (RE_NO_BK_PARENS << 1) + +/* If this bit is set, then | is an alternation operator, and \| is literal. + If not set, then \| is an alternation operator, and | is literal. */ +# define RE_NO_BK_VBAR (RE_NO_BK_REFS << 1) + +/* If this bit is set, then an ending range point collating higher + than the starting range point, as in [z-a], is invalid. + If not set, then when ending range point collates higher than the + starting range point, the range is ignored. */ +# define RE_NO_EMPTY_RANGES (RE_NO_BK_VBAR << 1) + +/* If this bit is set, then an unmatched ) is ordinary. + If not set, then an unmatched ) is invalid. */ +# define RE_UNMATCHED_RIGHT_PAREN_ORD (RE_NO_EMPTY_RANGES << 1) + +/* If this bit is set, succeed as soon as we match the whole pattern, + without further backtracking. */ +# define RE_NO_POSIX_BACKTRACKING (RE_UNMATCHED_RIGHT_PAREN_ORD << 1) + +/* If this bit is set, do not process the GNU regex operators. + If not set, then the GNU regex operators are recognized. */ +# define RE_NO_GNU_OPS (RE_NO_POSIX_BACKTRACKING << 1) + +/* If this bit is set, turn on internal regex debugging. + If not set, and debugging was on, turn it off. + This only works if regex.c is compiled -DDEBUG. + We define this bit always, so that all that's needed to turn on + debugging is to recompile regex.c; the calling code can always have + this bit set, and it won't affect anything in the normal case. */ +# define RE_DEBUG (RE_NO_GNU_OPS << 1) + +/* If this bit is set, a syntactically invalid interval is treated as + a string of ordinary characters. For example, the ERE 'a{1' is + treated as 'a\{1'. */ +# define RE_INVALID_INTERVAL_ORD (RE_DEBUG << 1) + +/* If this bit is set, then ignore case when matching. + If not set, then case is significant. */ +# define RE_ICASE (RE_INVALID_INTERVAL_ORD << 1) + +/* This bit is used internally like RE_CONTEXT_INDEP_ANCHORS but only + for ^, because it is difficult to scan the regex backwards to find + whether ^ should be special. */ +# define RE_CARET_ANCHORS_HERE (RE_ICASE << 1) + +/* If this bit is set, then \{ cannot be first in an bre or + immediately after an alternation or begin-group operator. */ +# define RE_CONTEXT_INVALID_DUP (RE_CARET_ANCHORS_HERE << 1) + +/* If this bit is set, then no_sub will be set to 1 during + re_compile_pattern. */ +# define RE_NO_SUB (RE_CONTEXT_INVALID_DUP << 1) +#endif + +/* This global variable defines the particular regexp syntax to use (for + some interfaces). When a regexp is compiled, the syntax used is + stored in the pattern buffer, so changing this does not affect + already-compiled regexps. */ +extern reg_syntax_t re_syntax_options; + +#ifdef __USE_GNU +/* Define combinations of the above bits for the standard possibilities. + (The [[[ comments delimit what gets put into the Texinfo file, so + don't delete them!) */ +/* [[[begin syntaxes]]] */ +#define RE_SYNTAX_EMACS 0 + +#define RE_SYNTAX_AWK \ + (RE_BACKSLASH_ESCAPE_IN_LISTS | RE_DOT_NOT_NULL \ + | RE_NO_BK_PARENS | RE_NO_BK_REFS \ + | RE_NO_BK_VBAR | RE_NO_EMPTY_RANGES \ + | RE_DOT_NEWLINE | RE_CONTEXT_INDEP_ANCHORS \ + | RE_CHAR_CLASSES \ + | RE_UNMATCHED_RIGHT_PAREN_ORD | RE_NO_GNU_OPS) + +#define RE_SYNTAX_GNU_AWK \ + ((RE_SYNTAX_POSIX_EXTENDED | RE_BACKSLASH_ESCAPE_IN_LISTS \ + | RE_INVALID_INTERVAL_ORD) \ + & ~(RE_DOT_NOT_NULL | RE_CONTEXT_INDEP_OPS \ + | RE_CONTEXT_INVALID_OPS )) + +#define RE_SYNTAX_POSIX_AWK \ + (RE_SYNTAX_POSIX_EXTENDED | RE_BACKSLASH_ESCAPE_IN_LISTS \ + | RE_INTERVALS | RE_NO_GNU_OPS \ + | RE_INVALID_INTERVAL_ORD) + +#define RE_SYNTAX_GREP \ + (RE_BK_PLUS_QM | RE_CHAR_CLASSES \ + | RE_HAT_LISTS_NOT_NEWLINE | RE_INTERVALS \ + | RE_NEWLINE_ALT) + +#define RE_SYNTAX_EGREP \ + (RE_CHAR_CLASSES | RE_CONTEXT_INDEP_ANCHORS \ + | RE_CONTEXT_INDEP_OPS | RE_HAT_LISTS_NOT_NEWLINE \ + | RE_NEWLINE_ALT | RE_NO_BK_PARENS \ + | RE_NO_BK_VBAR) + +#define RE_SYNTAX_POSIX_EGREP \ + (RE_SYNTAX_EGREP | RE_INTERVALS | RE_NO_BK_BRACES \ + | RE_INVALID_INTERVAL_ORD) + +/* P1003.2/D11.2, section 4.20.7.1, lines 5078ff. */ +#define RE_SYNTAX_ED RE_SYNTAX_POSIX_BASIC + +#define RE_SYNTAX_SED RE_SYNTAX_POSIX_BASIC + +/* Syntax bits common to both basic and extended POSIX regex syntax. */ +#define _RE_SYNTAX_POSIX_COMMON \ + (RE_CHAR_CLASSES | RE_DOT_NEWLINE | RE_DOT_NOT_NULL \ + | RE_INTERVALS | RE_NO_EMPTY_RANGES) + +#define RE_SYNTAX_POSIX_BASIC \ + (_RE_SYNTAX_POSIX_COMMON | RE_BK_PLUS_QM | RE_CONTEXT_INVALID_DUP) + +/* Differs from ..._POSIX_BASIC only in that RE_BK_PLUS_QM becomes + RE_LIMITED_OPS, i.e., \? \+ \| are not recognized. Actually, this + isn't minimal, since other operators, such as \`, aren't disabled. */ +#define RE_SYNTAX_POSIX_MINIMAL_BASIC \ + (_RE_SYNTAX_POSIX_COMMON | RE_LIMITED_OPS) + +#define RE_SYNTAX_POSIX_EXTENDED \ + (_RE_SYNTAX_POSIX_COMMON | RE_CONTEXT_INDEP_ANCHORS \ + | RE_CONTEXT_INDEP_OPS | RE_NO_BK_BRACES \ + | RE_NO_BK_PARENS | RE_NO_BK_VBAR \ + | RE_CONTEXT_INVALID_OPS | RE_UNMATCHED_RIGHT_PAREN_ORD) + +/* Differs from ..._POSIX_EXTENDED in that RE_CONTEXT_INDEP_OPS is + removed and RE_NO_BK_REFS is added. */ +#define RE_SYNTAX_POSIX_MINIMAL_EXTENDED \ + (_RE_SYNTAX_POSIX_COMMON | RE_CONTEXT_INDEP_ANCHORS \ + | RE_CONTEXT_INVALID_OPS | RE_NO_BK_BRACES \ + | RE_NO_BK_PARENS | RE_NO_BK_REFS \ + | RE_NO_BK_VBAR | RE_UNMATCHED_RIGHT_PAREN_ORD) +/* [[[end syntaxes]]] */ + +/* Maximum number of duplicates an interval can allow. Some systems + (erroneously) define this in other header files, but we want our + value, so remove any previous define. */ +# ifdef RE_DUP_MAX +# undef RE_DUP_MAX +# endif +/* If sizeof(int) == 2, then ((1 << 15) - 1) overflows. */ +# define RE_DUP_MAX (0x7fff) +#endif + + +/* POSIX `cflags' bits (i.e., information for `regcomp'). */ + +/* If this bit is set, then use extended regular expression syntax. + If not set, then use basic regular expression syntax. */ +#define REG_EXTENDED 1 + +/* If this bit is set, then ignore case when matching. + If not set, then case is significant. */ +#define REG_ICASE (REG_EXTENDED << 1) + +/* If this bit is set, then anchors do not match at newline + characters in the string. + If not set, then anchors do match at newlines. */ +#define REG_NEWLINE (REG_ICASE << 1) + +/* If this bit is set, then report only success or fail in regexec. + If not set, then returns differ between not matching and errors. */ +#define REG_NOSUB (REG_NEWLINE << 1) + + +/* POSIX `eflags' bits (i.e., information for regexec). */ + +/* If this bit is set, then the beginning-of-line operator doesn't match + the beginning of the string (presumably because it's not the + beginning of a line). + If not set, then the beginning-of-line operator does match the + beginning of the string. */ +#define REG_NOTBOL 1 + +/* Like REG_NOTBOL, except for the end-of-line. */ +#define REG_NOTEOL (1 << 1) + +/* Use PMATCH[0] to delimit the start and end of the search in the + buffer. */ +#define REG_STARTEND (1 << 2) + + +/* If any error codes are removed, changed, or added, update the + `re_error_msg' table in regex.c. */ +typedef enum +{ +#if defined _XOPEN_SOURCE || defined __USE_XOPEN2K + REG_ENOSYS = -1, /* This will never happen for this implementation. */ +#endif + + REG_NOERROR = 0, /* Success. */ + REG_NOMATCH, /* Didn't find a match (for regexec). */ + + /* POSIX regcomp return error codes. (In the order listed in the + standard.) */ + REG_BADPAT, /* Invalid pattern. */ + REG_ECOLLATE, /* Invalid collating element. */ + REG_ECTYPE, /* Invalid character class name. */ + REG_EESCAPE, /* Trailing backslash. */ + REG_ESUBREG, /* Invalid back reference. */ + REG_EBRACK, /* Unmatched left bracket. */ + REG_EPAREN, /* Parenthesis imbalance. */ + REG_EBRACE, /* Unmatched \{. */ + REG_BADBR, /* Invalid contents of \{\}. */ + REG_ERANGE, /* Invalid range end. */ + REG_ESPACE, /* Ran out of memory. */ + REG_BADRPT, /* No preceding re for repetition op. */ + + /* Error codes we've added. */ + REG_EEND, /* Premature end. */ + REG_ESIZE, /* Compiled pattern bigger than 2^16 bytes. */ + REG_ERPAREN /* Unmatched ) or \); not returned from regcomp. */ +} reg_errcode_t; + +/* This data structure represents a compiled pattern. Before calling + the pattern compiler, the fields `buffer', `allocated', `fastmap', + and `translate' can be set. After the pattern has been compiled, + the fields `re_nsub', `not_bol' and `not_eol' are available. All + other fields are private to the regex routines. */ + +#ifndef RE_TRANSLATE_TYPE +# define __RE_TRANSLATE_TYPE unsigned char * +# ifdef __USE_GNU +# define RE_TRANSLATE_TYPE __RE_TRANSLATE_TYPE +# endif +#endif + +#ifdef __USE_GNU +# define __REPB_PREFIX(name) name +#else +# define __REPB_PREFIX(name) __##name +#endif + +struct re_pattern_buffer +{ + /* Space that holds the compiled pattern. It is declared as + `unsigned char *' because its elements are sometimes used as + array indexes. */ + unsigned char *__REPB_PREFIX(buffer); + + /* Number of bytes to which `buffer' points. */ + unsigned long int __REPB_PREFIX(allocated); + + /* Number of bytes actually used in `buffer'. */ + unsigned long int __REPB_PREFIX(used); + + /* Syntax setting with which the pattern was compiled. */ + reg_syntax_t __REPB_PREFIX(syntax); + + /* Pointer to a fastmap, if any, otherwise zero. re_search uses the + fastmap, if there is one, to skip over impossible starting points + for matches. */ + char *__REPB_PREFIX(fastmap); + + /* Either a translate table to apply to all characters before + comparing them, or zero for no translation. The translation is + applied to a pattern when it is compiled and to a string when it + is matched. */ + __RE_TRANSLATE_TYPE __REPB_PREFIX(translate); + + /* Number of subexpressions found by the compiler. */ + size_t re_nsub; + + /* Zero if this pattern cannot match the empty string, one else. + Well, in truth it's used only in `re_search_2', to see whether or + not we should use the fastmap, so we don't set this absolutely + perfectly; see `re_compile_fastmap' (the `duplicate' case). */ + unsigned __REPB_PREFIX(can_be_null) : 1; + + /* If REGS_UNALLOCATED, allocate space in the `regs' structure + for `max (RE_NREGS, re_nsub + 1)' groups. + If REGS_REALLOCATE, reallocate space if necessary. + If REGS_FIXED, use what's there. */ +#ifdef __USE_GNU +# define REGS_UNALLOCATED 0 +# define REGS_REALLOCATE 1 +# define REGS_FIXED 2 +#endif + unsigned __REPB_PREFIX(regs_allocated) : 2; + + /* Set to zero when `regex_compile' compiles a pattern; set to one + by `re_compile_fastmap' if it updates the fastmap. */ + unsigned __REPB_PREFIX(fastmap_accurate) : 1; + + /* If set, `re_match_2' does not return information about + subexpressions. */ + unsigned __REPB_PREFIX(no_sub) : 1; + + /* If set, a beginning-of-line anchor doesn't match at the beginning + of the string. */ + unsigned __REPB_PREFIX(not_bol) : 1; + + /* Similarly for an end-of-line anchor. */ + unsigned __REPB_PREFIX(not_eol) : 1; + + /* If true, an anchor at a newline matches. */ + unsigned __REPB_PREFIX(newline_anchor) : 1; +}; + +typedef struct re_pattern_buffer regex_t; + +/* Type for byte offsets within the string. POSIX mandates this. */ +typedef int regoff_t; + + +#ifdef __USE_GNU +/* This is the structure we store register match data in. See + regex.texinfo for a full description of what registers match. */ +struct re_registers +{ + unsigned num_regs; + regoff_t *start; + regoff_t *end; +}; + + +/* If `regs_allocated' is REGS_UNALLOCATED in the pattern buffer, + `re_match_2' returns information about at least this many registers + the first time a `regs' structure is passed. */ +# ifndef RE_NREGS +# define RE_NREGS 30 +# endif +#endif + + +/* POSIX specification for registers. Aside from the different names than + `re_registers', POSIX uses an array of structures, instead of a + structure of arrays. */ +typedef struct +{ + regoff_t rm_so; /* Byte offset from string's start to substring's start. */ + regoff_t rm_eo; /* Byte offset from string's start to substring's end. */ +} regmatch_t; + +/* Declarations for routines. */ + +#ifdef __USE_GNU +/* Sets the current default syntax to SYNTAX, and return the old syntax. + You can also simply assign to the `re_syntax_options' variable. */ +extern reg_syntax_t re_set_syntax (reg_syntax_t syntax); + +/* Compile the regular expression PATTERN, with length LENGTH + and syntax given by the global `re_syntax_options', into the buffer + BUFFER. Return NULL if successful, and an error string if not. + + To free the allocated storage, you must call `regfree' on BUFFER. + Note that the translate table must either have been initialised by + `regcomp', with a malloc'ed value, or set to NULL before calling + `regfree'. */ +extern const char *re_compile_pattern (const char *pattern, size_t length, + struct re_pattern_buffer *buffer); + + +/* Compile a fastmap for the compiled pattern in BUFFER; used to + accelerate searches. Return 0 if successful and -2 if was an + internal error. */ +extern int re_compile_fastmap (struct re_pattern_buffer *buffer); + + +/* Search in the string STRING (with length LENGTH) for the pattern + compiled into BUFFER. Start searching at position START, for RANGE + characters. Return the starting position of the match, -1 for no + match, or -2 for an internal error. Also return register + information in REGS (if REGS and BUFFER->no_sub are nonzero). */ +extern int re_search (struct re_pattern_buffer *buffer, const char *c_string, + int length, int start, int range, + struct re_registers *regs); + + +/* Like `re_search', but search in the concatenation of STRING1 and + STRING2. Also, stop searching at index START + STOP. */ +extern int re_search_2 (struct re_pattern_buffer *buffer, + const char *string1, int length1, + const char *string2, int length2, int start, + int range, struct re_registers *regs, int stop); + + +/* Like `re_search', but return how many characters in STRING the regexp + in BUFFER matched, starting at position START. */ +extern int re_match (struct re_pattern_buffer *buffer, const char *c_string, + int length, int start, struct re_registers *regs); + + +/* Relates to `re_match' as `re_search_2' relates to `re_search'. */ +extern int re_match_2 (struct re_pattern_buffer *buffer, + const char *string1, int length1, + const char *string2, int length2, int start, + struct re_registers *regs, int stop); + + +/* Set REGS to hold NUM_REGS registers, storing them in STARTS and + ENDS. Subsequent matches using BUFFER and REGS will use this memory + for recording register information. STARTS and ENDS must be + allocated with malloc, and must each be at least `NUM_REGS * sizeof + (regoff_t)' bytes long. + + If NUM_REGS == 0, then subsequent matches should allocate their own + register data. + + Unless this function is called, the first search or match using + PATTERN_BUFFER will allocate its own register data, without + freeing the old data. */ +extern void re_set_registers (struct re_pattern_buffer *buffer, + struct re_registers *regs, + unsigned int num_regs, + regoff_t *starts, regoff_t *ends); +#endif /* Use GNU */ + +#if defined _REGEX_RE_COMP || (defined _LIBC && defined __USE_MISC) +# ifndef _CRAY +/* 4.2 bsd compatibility. */ +extern char *re_comp (const char *); +extern int re_exec (const char *); +# endif +#endif + +/* GCC 2.95 and later have "__restrict"; C99 compilers have + "restrict", and "configure" may have defined "restrict". */ +#ifndef __restrict +# if ! (2 < __GNUC__ || (2 == __GNUC__ && 95 <= __GNUC_MINOR__)) +# if defined restrict || 199901L <= __STDC_VERSION__ +# define __restrict restrict +# else +# define __restrict +# endif +# endif +#endif +/* gcc 3.1 and up support the [restrict] syntax. */ +#ifndef __restrict_arr +# if (__GNUC__ > 3 || (__GNUC__ == 3 && __GNUC_MINOR__ >= 1)) \ + && !defined __GNUG__ +# define __restrict_arr __restrict +# else +# define __restrict_arr +# endif +#endif + +/* POSIX compatibility. */ +extern int regcomp (regex_t *__restrict preg, + const char *__restrict pattern, + int cflags); + +extern int regexec (const regex_t *__restrict preg, + const char *__restrict c_string, size_t nmatch, + regmatch_t pmatch[__restrict_arr], + int eflags); + +extern size_t regerror (int errcode, const regex_t *__restrict preg, + char *__restrict errbuf, size_t errbuf_size); + +extern void regfree (regex_t *preg); + + +#ifdef __cplusplus +} +#endif /* C++ */ + +#endif /* regex.h */ diff --git a/support/regex_internal.c b/support/regex_internal.c new file mode 100644 index 00000000..18641ef1 --- /dev/null +++ b/support/regex_internal.c @@ -0,0 +1,1761 @@ +/* Extended regular expression matching and search library. + Copyright (C) 2002-2016 Free Software Foundation, Inc. + This file is part of the GNU C Library. + Contributed by Isamu Hasegawa <isamu@yamato.ibm.com>. + + The GNU C Library is free software; you can redistribute it and/or + modify it under the terms of the GNU Lesser General Public + License as published by the Free Software Foundation; either + version 2.1 of the License, or (at your option) any later version. + + The GNU C Library is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public + License along with the GNU C Library; if not, see + <http://www.gnu.org/licenses/>. */ + +static void re_string_construct_common (const char *str, int len, + re_string_t *pstr, + RE_TRANSLATE_TYPE trans, int icase, + const re_dfa_t *dfa) internal_function; +static re_dfastate_t *create_ci_newstate (const re_dfa_t *dfa, + const re_node_set *nodes, + unsigned int hash) internal_function; +static re_dfastate_t *create_cd_newstate (const re_dfa_t *dfa, + const re_node_set *nodes, + unsigned int context, + unsigned int hash) internal_function; + +#ifdef GAWK +#undef MAX /* safety */ +static int +MAX(size_t a, size_t b) +{ + return (a > b ? a : b); +} +#endif + +/* Functions for string operation. */ + +/* This function allocate the buffers. It is necessary to call + re_string_reconstruct before using the object. */ + +static reg_errcode_t +internal_function __attribute_warn_unused_result__ +re_string_allocate (re_string_t *pstr, const char *str, int len, int init_len, + RE_TRANSLATE_TYPE trans, int icase, const re_dfa_t *dfa) +{ + reg_errcode_t ret; + int init_buf_len; + + /* Ensure at least one character fits into the buffers. */ + if (init_len < dfa->mb_cur_max) + init_len = dfa->mb_cur_max; + init_buf_len = (len + 1 < init_len) ? len + 1: init_len; + re_string_construct_common (str, len, pstr, trans, icase, dfa); + + ret = re_string_realloc_buffers (pstr, init_buf_len); + if (BE (ret != REG_NOERROR, 0)) + return ret; + + pstr->word_char = dfa->word_char; + pstr->word_ops_used = dfa->word_ops_used; + pstr->mbs = pstr->mbs_allocated ? pstr->mbs : (unsigned char *) str; + pstr->valid_len = (pstr->mbs_allocated || dfa->mb_cur_max > 1) ? 0 : len; + pstr->valid_raw_len = pstr->valid_len; + return REG_NOERROR; +} + +/* This function allocate the buffers, and initialize them. */ + +static reg_errcode_t +internal_function __attribute_warn_unused_result__ +re_string_construct (re_string_t *pstr, const char *str, int len, + RE_TRANSLATE_TYPE trans, int icase, const re_dfa_t *dfa) +{ + reg_errcode_t ret; + memset (pstr, '\0', sizeof (re_string_t)); + re_string_construct_common (str, len, pstr, trans, icase, dfa); + + if (len > 0) + { + ret = re_string_realloc_buffers (pstr, len + 1); + if (BE (ret != REG_NOERROR, 0)) + return ret; + } + pstr->mbs = pstr->mbs_allocated ? pstr->mbs : (unsigned char *) str; + + if (icase) + { +#ifdef RE_ENABLE_I18N + if (dfa->mb_cur_max > 1) + { + while (1) + { + ret = build_wcs_upper_buffer (pstr); + if (BE (ret != REG_NOERROR, 0)) + return ret; + if (pstr->valid_raw_len >= len) + break; + if (pstr->bufs_len > pstr->valid_len + dfa->mb_cur_max) + break; + ret = re_string_realloc_buffers (pstr, pstr->bufs_len * 2); + if (BE (ret != REG_NOERROR, 0)) + return ret; + } + } + else +#endif /* RE_ENABLE_I18N */ + build_upper_buffer (pstr); + } + else + { +#ifdef RE_ENABLE_I18N + if (dfa->mb_cur_max > 1) + build_wcs_buffer (pstr); + else +#endif /* RE_ENABLE_I18N */ + { + if (trans != NULL) + re_string_translate_buffer (pstr); + else + { + pstr->valid_len = pstr->bufs_len; + pstr->valid_raw_len = pstr->bufs_len; + } + } + } + + return REG_NOERROR; +} + +/* Helper functions for re_string_allocate, and re_string_construct. */ + +static reg_errcode_t +internal_function __attribute_warn_unused_result__ +re_string_realloc_buffers (re_string_t *pstr, int new_buf_len) +{ +#ifdef RE_ENABLE_I18N + if (pstr->mb_cur_max > 1) + { + wint_t *new_wcs; + + /* Avoid overflow in realloc. */ + const size_t max_object_size = MAX (sizeof (wint_t), sizeof (int)); + if (BE (SIZE_MAX / max_object_size < new_buf_len, 0)) + return REG_ESPACE; + + new_wcs = re_realloc (pstr->wcs, wint_t, new_buf_len); + if (BE (new_wcs == NULL, 0)) + return REG_ESPACE; + pstr->wcs = new_wcs; + if (pstr->offsets != NULL) + { + int *new_offsets = re_realloc (pstr->offsets, int, new_buf_len); + if (BE (new_offsets == NULL, 0)) + return REG_ESPACE; + pstr->offsets = new_offsets; + } + } +#endif /* RE_ENABLE_I18N */ + if (pstr->mbs_allocated) + { + unsigned char *new_mbs = re_realloc (pstr->mbs, unsigned char, + new_buf_len); + if (BE (new_mbs == NULL, 0)) + return REG_ESPACE; + pstr->mbs = new_mbs; + } + pstr->bufs_len = new_buf_len; + return REG_NOERROR; +} + + +static void +internal_function +re_string_construct_common (const char *str, int len, re_string_t *pstr, + RE_TRANSLATE_TYPE trans, int icase, + const re_dfa_t *dfa) +{ + pstr->raw_mbs = (const unsigned char *) str; + pstr->len = len; + pstr->raw_len = len; + pstr->trans = trans; + pstr->icase = icase ? 1 : 0; + pstr->mbs_allocated = (trans != NULL || icase); + pstr->mb_cur_max = dfa->mb_cur_max; + pstr->is_utf8 = dfa->is_utf8; + pstr->map_notascii = dfa->map_notascii; + pstr->stop = pstr->len; + pstr->raw_stop = pstr->stop; +} + +#ifdef RE_ENABLE_I18N + +/* Build wide character buffer PSTR->WCS. + If the byte sequence of the string are: + <mb1>(0), <mb1>(1), <mb2>(0), <mb2>(1), <sb3> + Then wide character buffer will be: + <wc1> , WEOF , <wc2> , WEOF , <wc3> + We use WEOF for padding, they indicate that the position isn't + a first byte of a multibyte character. + + Note that this function assumes PSTR->VALID_LEN elements are already + built and starts from PSTR->VALID_LEN. */ + +static void +internal_function +build_wcs_buffer (re_string_t *pstr) +{ +#ifdef _LIBC + unsigned char buf[MB_LEN_MAX]; + assert (MB_LEN_MAX >= pstr->mb_cur_max); +#else + unsigned char buf[64]; +#endif + mbstate_t prev_st; + int byte_idx, end_idx, remain_len; + size_t mbclen; + + /* Build the buffers from pstr->valid_len to either pstr->len or + pstr->bufs_len. */ + end_idx = (pstr->bufs_len > pstr->len) ? pstr->len : pstr->bufs_len; + for (byte_idx = pstr->valid_len; byte_idx < end_idx;) + { + wchar_t wc; + const char *p; + + remain_len = end_idx - byte_idx; + prev_st = pstr->cur_state; + /* Apply the translation if we need. */ + if (BE (pstr->trans != NULL, 0)) + { + int i, ch; + + for (i = 0; i < pstr->mb_cur_max && i < remain_len; ++i) + { + ch = pstr->raw_mbs [pstr->raw_mbs_idx + byte_idx + i]; + buf[i] = pstr->mbs[byte_idx + i] = pstr->trans[ch]; + } + p = (const char *) buf; + } + else + p = (const char *) pstr->raw_mbs + pstr->raw_mbs_idx + byte_idx; + mbclen = __mbrtowc (&wc, p, remain_len, &pstr->cur_state); + if (BE (mbclen == (size_t) -1 || mbclen == 0 + || (mbclen == (size_t) -2 && pstr->bufs_len >= pstr->len), 0)) + { + /* We treat these cases as a singlebyte character. */ + mbclen = 1; + wc = (wchar_t) pstr->raw_mbs[pstr->raw_mbs_idx + byte_idx]; + if (BE (pstr->trans != NULL, 0)) + wc = pstr->trans[wc]; + pstr->cur_state = prev_st; + } + else if (BE (mbclen == (size_t) -2, 0)) + { + /* The buffer doesn't have enough space, finish to build. */ + pstr->cur_state = prev_st; + break; + } + + /* Write wide character and padding. */ + pstr->wcs[byte_idx++] = wc; + /* Write paddings. */ + for (remain_len = byte_idx + mbclen - 1; byte_idx < remain_len ;) + pstr->wcs[byte_idx++] = WEOF; + } + pstr->valid_len = byte_idx; + pstr->valid_raw_len = byte_idx; +} + +/* Build wide character buffer PSTR->WCS like build_wcs_buffer, + but for REG_ICASE. */ + +static reg_errcode_t +internal_function __attribute_warn_unused_result__ +build_wcs_upper_buffer (re_string_t *pstr) +{ + mbstate_t prev_st; + int src_idx, byte_idx, end_idx, remain_len; + size_t mbclen; +#ifdef _LIBC + char buf[MB_LEN_MAX]; + assert (MB_LEN_MAX >= pstr->mb_cur_max); +#else + char buf[64]; +#endif + + byte_idx = pstr->valid_len; + end_idx = (pstr->bufs_len > pstr->len) ? pstr->len : pstr->bufs_len; + + /* The following optimization assumes that ASCII characters can be + mapped to wide characters with a simple cast. */ + if (! pstr->map_notascii && pstr->trans == NULL && !pstr->offsets_needed) + { + while (byte_idx < end_idx) + { + wchar_t wc; + + if (isascii (pstr->raw_mbs[pstr->raw_mbs_idx + byte_idx]) + && mbsinit (&pstr->cur_state)) + { + /* In case of a singlebyte character. */ + pstr->mbs[byte_idx] + = toupper (pstr->raw_mbs[pstr->raw_mbs_idx + byte_idx]); + /* The next step uses the assumption that wchar_t is encoded + ASCII-safe: all ASCII values can be converted like this. */ + pstr->wcs[byte_idx] = (wchar_t) pstr->mbs[byte_idx]; + ++byte_idx; + continue; + } + + remain_len = end_idx - byte_idx; + prev_st = pstr->cur_state; + mbclen = __mbrtowc (&wc, + ((const char *) pstr->raw_mbs + pstr->raw_mbs_idx + + byte_idx), remain_len, &pstr->cur_state); + if (BE (mbclen + 2 > 2, 1)) + { + wchar_t wcu = wc; + if (iswlower (wc)) + { + size_t mbcdlen; + + wcu = towupper (wc); + mbcdlen = __wcrtomb (buf, wcu, &prev_st); + if (BE (mbclen == mbcdlen, 1)) + memcpy (pstr->mbs + byte_idx, buf, mbclen); + else + { + src_idx = byte_idx; + goto offsets_needed; + } + } + else + memcpy (pstr->mbs + byte_idx, + pstr->raw_mbs + pstr->raw_mbs_idx + byte_idx, mbclen); + pstr->wcs[byte_idx++] = wcu; + /* Write paddings. */ + for (remain_len = byte_idx + mbclen - 1; byte_idx < remain_len ;) + pstr->wcs[byte_idx++] = WEOF; + } + else if (mbclen == (size_t) -1 || mbclen == 0 + || (mbclen == (size_t) -2 && pstr->bufs_len >= pstr->len)) + { + /* It is an invalid character, an incomplete character + at the end of the string, or '\0'. Just use the byte. */ + int ch = pstr->raw_mbs[pstr->raw_mbs_idx + byte_idx]; + pstr->mbs[byte_idx] = ch; + /* And also cast it to wide char. */ + pstr->wcs[byte_idx++] = (wchar_t) ch; + if (BE (mbclen == (size_t) -1, 0)) + pstr->cur_state = prev_st; + } + else + { + /* The buffer doesn't have enough space, finish to build. */ + pstr->cur_state = prev_st; + break; + } + } + pstr->valid_len = byte_idx; + pstr->valid_raw_len = byte_idx; + return REG_NOERROR; + } + else + for (src_idx = pstr->valid_raw_len; byte_idx < end_idx;) + { + wchar_t wc; + const char *p; + offsets_needed: + remain_len = end_idx - byte_idx; + prev_st = pstr->cur_state; + if (BE (pstr->trans != NULL, 0)) + { + int i, ch; + + for (i = 0; i < pstr->mb_cur_max && i < remain_len; ++i) + { + ch = pstr->raw_mbs [pstr->raw_mbs_idx + src_idx + i]; + buf[i] = pstr->trans[ch]; + } + p = (const char *) buf; + } + else + p = (const char *) pstr->raw_mbs + pstr->raw_mbs_idx + src_idx; + mbclen = __mbrtowc (&wc, p, remain_len, &pstr->cur_state); + if (BE (mbclen + 2 > 2, 1)) + { + wchar_t wcu = wc; + if (iswlower (wc)) + { + size_t mbcdlen; + + wcu = towupper (wc); + mbcdlen = wcrtomb ((char *) buf, wcu, &prev_st); + if (BE (mbclen == mbcdlen, 1)) + memcpy (pstr->mbs + byte_idx, buf, mbclen); + else if (mbcdlen != (size_t) -1) + { + size_t i; + + if (byte_idx + mbcdlen > pstr->bufs_len) + { + pstr->cur_state = prev_st; + break; + } + + if (pstr->offsets == NULL) + { + pstr->offsets = re_malloc (int, pstr->bufs_len); + + if (pstr->offsets == NULL) + return REG_ESPACE; + } + if (!pstr->offsets_needed) + { + for (i = 0; i < (size_t) byte_idx; ++i) + pstr->offsets[i] = i; + pstr->offsets_needed = 1; + } + + memcpy (pstr->mbs + byte_idx, buf, mbcdlen); + pstr->wcs[byte_idx] = wcu; + pstr->offsets[byte_idx] = src_idx; + for (i = 1; i < mbcdlen; ++i) + { + pstr->offsets[byte_idx + i] + = src_idx + (i < mbclen ? i : mbclen - 1); + pstr->wcs[byte_idx + i] = WEOF; + } + pstr->len += mbcdlen - mbclen; + if (pstr->raw_stop > src_idx) + pstr->stop += mbcdlen - mbclen; + end_idx = (pstr->bufs_len > pstr->len) + ? pstr->len : pstr->bufs_len; + byte_idx += mbcdlen; + src_idx += mbclen; + continue; + } + else + memcpy (pstr->mbs + byte_idx, p, mbclen); + } + else + memcpy (pstr->mbs + byte_idx, p, mbclen); + + if (BE (pstr->offsets_needed != 0, 0)) + { + size_t i; + for (i = 0; i < mbclen; ++i) + pstr->offsets[byte_idx + i] = src_idx + i; + } + src_idx += mbclen; + + pstr->wcs[byte_idx++] = wcu; + /* Write paddings. */ + for (remain_len = byte_idx + mbclen - 1; byte_idx < remain_len ;) + pstr->wcs[byte_idx++] = WEOF; + } + else if (mbclen == (size_t) -1 || mbclen == 0 + || (mbclen == (size_t) -2 && pstr->bufs_len >= pstr->len)) + { + /* It is an invalid character or '\0'. Just use the byte. */ + int ch = pstr->raw_mbs[pstr->raw_mbs_idx + src_idx]; + + if (BE (pstr->trans != NULL, 0)) + ch = pstr->trans [ch]; + pstr->mbs[byte_idx] = ch; + + if (BE (pstr->offsets_needed != 0, 0)) + pstr->offsets[byte_idx] = src_idx; + ++src_idx; + + /* And also cast it to wide char. */ + pstr->wcs[byte_idx++] = (wchar_t) ch; + if (BE (mbclen == (size_t) -1, 0)) + pstr->cur_state = prev_st; + } + else + { + /* The buffer doesn't have enough space, finish to build. */ + pstr->cur_state = prev_st; + break; + } + } + pstr->valid_len = byte_idx; + pstr->valid_raw_len = src_idx; + return REG_NOERROR; +} + +/* Skip characters until the index becomes greater than NEW_RAW_IDX. + Return the index. */ + +static int +internal_function +re_string_skip_chars (re_string_t *pstr, int new_raw_idx, wint_t *last_wc) +{ + mbstate_t prev_st; + int rawbuf_idx; + size_t mbclen; + wint_t wc = WEOF; + + /* Skip the characters which are not necessary to check. */ + for (rawbuf_idx = pstr->raw_mbs_idx + pstr->valid_raw_len; + rawbuf_idx < new_raw_idx;) + { + wchar_t wc2; + int remain_len = pstr->raw_len - rawbuf_idx; + prev_st = pstr->cur_state; + mbclen = __mbrtowc (&wc2, (const char *) pstr->raw_mbs + rawbuf_idx, + remain_len, &pstr->cur_state); + if (BE ((ssize_t) mbclen <= 0, 0)) + { + /* We treat these cases as a single byte character. */ + if (mbclen == 0 || remain_len == 0) + wc = L'\0'; + else + wc = *(unsigned char *) (pstr->raw_mbs + rawbuf_idx); + mbclen = 1; + pstr->cur_state = prev_st; + } + else + wc = (wint_t) wc2; + /* Then proceed the next character. */ + rawbuf_idx += mbclen; + } + *last_wc = (wint_t) wc; + return rawbuf_idx; +} +#endif /* RE_ENABLE_I18N */ + +/* Build the buffer PSTR->MBS, and apply the translation if we need. + This function is used in case of REG_ICASE. */ + +static void +internal_function +build_upper_buffer (re_string_t *pstr) +{ + int char_idx, end_idx; + end_idx = (pstr->bufs_len > pstr->len) ? pstr->len : pstr->bufs_len; + + for (char_idx = pstr->valid_len; char_idx < end_idx; ++char_idx) + { + int ch = pstr->raw_mbs[pstr->raw_mbs_idx + char_idx]; + if (BE (pstr->trans != NULL, 0)) + ch = pstr->trans[ch]; + if (islower (ch)) + pstr->mbs[char_idx] = toupper (ch); + else + pstr->mbs[char_idx] = ch; + } + pstr->valid_len = char_idx; + pstr->valid_raw_len = char_idx; +} + +/* Apply TRANS to the buffer in PSTR. */ + +static void +internal_function +re_string_translate_buffer (re_string_t *pstr) +{ + int buf_idx, end_idx; + end_idx = (pstr->bufs_len > pstr->len) ? pstr->len : pstr->bufs_len; + + for (buf_idx = pstr->valid_len; buf_idx < end_idx; ++buf_idx) + { + int ch = pstr->raw_mbs[pstr->raw_mbs_idx + buf_idx]; + pstr->mbs[buf_idx] = pstr->trans[ch]; + } + + pstr->valid_len = buf_idx; + pstr->valid_raw_len = buf_idx; +} + +/* This function re-construct the buffers. + Concretely, convert to wide character in case of pstr->mb_cur_max > 1, + convert to upper case in case of REG_ICASE, apply translation. */ + +static reg_errcode_t +internal_function __attribute_warn_unused_result__ +re_string_reconstruct (re_string_t *pstr, int idx, int eflags) +{ + int offset = idx - pstr->raw_mbs_idx; + if (BE (offset < 0, 0)) + { + /* Reset buffer. */ +#ifdef RE_ENABLE_I18N + if (pstr->mb_cur_max > 1) + memset (&pstr->cur_state, '\0', sizeof (mbstate_t)); +#endif /* RE_ENABLE_I18N */ + pstr->len = pstr->raw_len; + pstr->stop = pstr->raw_stop; + pstr->valid_len = 0; + pstr->raw_mbs_idx = 0; + pstr->valid_raw_len = 0; + pstr->offsets_needed = 0; + pstr->tip_context = ((eflags & REG_NOTBOL) ? CONTEXT_BEGBUF + : CONTEXT_NEWLINE | CONTEXT_BEGBUF); + if (!pstr->mbs_allocated) + pstr->mbs = (unsigned char *) pstr->raw_mbs; + offset = idx; + } + + if (BE (offset != 0, 1)) + { + /* Should the already checked characters be kept? */ + if (BE (offset < pstr->valid_raw_len, 1)) + { + /* Yes, move them to the front of the buffer. */ +#ifdef RE_ENABLE_I18N + if (BE (pstr->offsets_needed, 0)) + { + int low = 0, high = pstr->valid_len, mid; + do + { + mid = (high + low) / 2; + if (pstr->offsets[mid] > offset) + high = mid; + else if (pstr->offsets[mid] < offset) + low = mid + 1; + else + break; + } + while (low < high); + if (pstr->offsets[mid] < offset) + ++mid; + pstr->tip_context = re_string_context_at (pstr, mid - 1, + eflags); + /* This can be quite complicated, so handle specially + only the common and easy case where the character with + different length representation of lower and upper + case is present at or after offset. */ + if (pstr->valid_len > offset + && mid == offset && pstr->offsets[mid] == offset) + { + memmove (pstr->wcs, pstr->wcs + offset, + (pstr->valid_len - offset) * sizeof (wint_t)); + memmove (pstr->mbs, pstr->mbs + offset, pstr->valid_len - offset); + pstr->valid_len -= offset; + pstr->valid_raw_len -= offset; + for (low = 0; low < pstr->valid_len; low++) + pstr->offsets[low] = pstr->offsets[low + offset] - offset; + } + else + { + /* Otherwise, just find out how long the partial multibyte + character at offset is and fill it with WEOF/255. */ + pstr->len = pstr->raw_len - idx + offset; + pstr->stop = pstr->raw_stop - idx + offset; + pstr->offsets_needed = 0; + while (mid > 0 && pstr->offsets[mid - 1] == offset) + --mid; + while (mid < pstr->valid_len) + if (pstr->wcs[mid] != WEOF) + break; + else + ++mid; + if (mid == pstr->valid_len) + pstr->valid_len = 0; + else + { + pstr->valid_len = pstr->offsets[mid] - offset; + if (pstr->valid_len) + { + for (low = 0; low < pstr->valid_len; ++low) + pstr->wcs[low] = WEOF; + memset (pstr->mbs, 255, pstr->valid_len); + } + } + pstr->valid_raw_len = pstr->valid_len; + } + } + else +#endif + { + pstr->tip_context = re_string_context_at (pstr, offset - 1, + eflags); +#ifdef RE_ENABLE_I18N + if (pstr->mb_cur_max > 1) + memmove (pstr->wcs, pstr->wcs + offset, + (pstr->valid_len - offset) * sizeof (wint_t)); +#endif /* RE_ENABLE_I18N */ + if (BE (pstr->mbs_allocated, 0)) + memmove (pstr->mbs, pstr->mbs + offset, + pstr->valid_len - offset); + pstr->valid_len -= offset; + pstr->valid_raw_len -= offset; +#if defined DEBUG && DEBUG + assert (pstr->valid_len > 0); +#endif + } + } + else + { +#ifdef RE_ENABLE_I18N + /* No, skip all characters until IDX. */ + int prev_valid_len = pstr->valid_len; + + if (BE (pstr->offsets_needed, 0)) + { + pstr->len = pstr->raw_len - idx + offset; + pstr->stop = pstr->raw_stop - idx + offset; + pstr->offsets_needed = 0; + } +#endif + pstr->valid_len = 0; +#ifdef RE_ENABLE_I18N + if (pstr->mb_cur_max > 1) + { + int wcs_idx; + wint_t wc = WEOF; + + if (pstr->is_utf8) + { + const unsigned char *raw, *p, *end; + + /* Special case UTF-8. Multi-byte chars start with any + byte other than 0x80 - 0xbf. */ + raw = pstr->raw_mbs + pstr->raw_mbs_idx; + end = raw + (offset - pstr->mb_cur_max); + if (end < pstr->raw_mbs) + end = pstr->raw_mbs; + p = raw + offset - 1; +#ifdef _LIBC + /* We know the wchar_t encoding is UCS4, so for the simple + case, ASCII characters, skip the conversion step. */ + if (isascii (*p) && BE (pstr->trans == NULL, 1)) + { + memset (&pstr->cur_state, '\0', sizeof (mbstate_t)); + /* pstr->valid_len = 0; */ + wc = (wchar_t) *p; + } + else +#endif + for (; p >= end; --p) + if ((*p & 0xc0) != 0x80) + { + mbstate_t cur_state; + wchar_t wc2; + int mlen = raw + pstr->len - p; + unsigned char buf[6]; + size_t mbclen; + + const unsigned char *pp = p; + if (BE (pstr->trans != NULL, 0)) + { + int i = mlen < 6 ? mlen : 6; + while (--i >= 0) + buf[i] = pstr->trans[p[i]]; + pp = buf; + } + /* XXX Don't use mbrtowc, we know which conversion + to use (UTF-8 -> UCS4). */ + memset (&cur_state, 0, sizeof (cur_state)); + mbclen = __mbrtowc (&wc2, (const char *) pp, mlen, + &cur_state); + if (raw + offset - p <= mbclen + && mbclen < (size_t) -2) + { + memset (&pstr->cur_state, '\0', + sizeof (mbstate_t)); + pstr->valid_len = mbclen - (raw + offset - p); + wc = wc2; + } + break; + } + } + + if (wc == WEOF) + pstr->valid_len = re_string_skip_chars (pstr, idx, &wc) - idx; + if (wc == WEOF) + pstr->tip_context + = re_string_context_at (pstr, prev_valid_len - 1, eflags); + else + pstr->tip_context = ((BE (pstr->word_ops_used != 0, 0) + && IS_WIDE_WORD_CHAR (wc)) + ? CONTEXT_WORD + : ((IS_WIDE_NEWLINE (wc) + && pstr->newline_anchor) + ? CONTEXT_NEWLINE : 0)); + if (BE (pstr->valid_len, 0)) + { + for (wcs_idx = 0; wcs_idx < pstr->valid_len; ++wcs_idx) + pstr->wcs[wcs_idx] = WEOF; + if (pstr->mbs_allocated) + memset (pstr->mbs, 255, pstr->valid_len); + } + pstr->valid_raw_len = pstr->valid_len; + } + else +#endif /* RE_ENABLE_I18N */ + { + int c = pstr->raw_mbs[pstr->raw_mbs_idx + offset - 1]; + pstr->valid_raw_len = 0; + if (pstr->trans) + c = pstr->trans[c]; + pstr->tip_context = (bitset_contain (pstr->word_char, c) + ? CONTEXT_WORD + : ((IS_NEWLINE (c) && pstr->newline_anchor) + ? CONTEXT_NEWLINE : 0)); + } + } + if (!BE (pstr->mbs_allocated, 0)) + pstr->mbs += offset; + } + pstr->raw_mbs_idx = idx; + pstr->len -= offset; + pstr->stop -= offset; + + /* Then build the buffers. */ +#ifdef RE_ENABLE_I18N + if (pstr->mb_cur_max > 1) + { + if (pstr->icase) + { + reg_errcode_t ret = build_wcs_upper_buffer (pstr); + if (BE (ret != REG_NOERROR, 0)) + return ret; + } + else + build_wcs_buffer (pstr); + } + else +#endif /* RE_ENABLE_I18N */ + if (BE (pstr->mbs_allocated, 0)) + { + if (pstr->icase) + build_upper_buffer (pstr); + else if (pstr->trans != NULL) + re_string_translate_buffer (pstr); + } + else + pstr->valid_len = pstr->len; + + pstr->cur_idx = 0; + return REG_NOERROR; +} + +static unsigned char +internal_function __attribute__ ((__pure__)) +re_string_peek_byte_case (const re_string_t *pstr, int idx) +{ + int ch, off; + + /* Handle the common (easiest) cases first. */ + if (BE (!pstr->mbs_allocated, 1)) + return re_string_peek_byte (pstr, idx); + +#ifdef RE_ENABLE_I18N + if (pstr->mb_cur_max > 1 + && ! re_string_is_single_byte_char (pstr, pstr->cur_idx + idx)) + return re_string_peek_byte (pstr, idx); +#endif + + off = pstr->cur_idx + idx; +#ifdef RE_ENABLE_I18N + if (pstr->offsets_needed) + off = pstr->offsets[off]; +#endif + + ch = pstr->raw_mbs[pstr->raw_mbs_idx + off]; + +#ifdef RE_ENABLE_I18N + /* Ensure that e.g. for tr_TR.UTF-8 BACKSLASH DOTLESS SMALL LETTER I + this function returns CAPITAL LETTER I instead of first byte of + DOTLESS SMALL LETTER I. The latter would confuse the parser, + since peek_byte_case doesn't advance cur_idx in any way. */ + if (pstr->offsets_needed && !isascii (ch)) + return re_string_peek_byte (pstr, idx); +#endif + + return ch; +} + +static unsigned char +internal_function +re_string_fetch_byte_case (re_string_t *pstr) +{ + if (BE (!pstr->mbs_allocated, 1)) + return re_string_fetch_byte (pstr); + +#ifdef RE_ENABLE_I18N + if (pstr->offsets_needed) + { + int off, ch; + + /* For tr_TR.UTF-8 [[:islower:]] there is + [[: CAPITAL LETTER I WITH DOT lower:]] in mbs. Skip + in that case the whole multi-byte character and return + the original letter. On the other side, with + [[: DOTLESS SMALL LETTER I return [[:I, as doing + anything else would complicate things too much. */ + + if (!re_string_first_byte (pstr, pstr->cur_idx)) + return re_string_fetch_byte (pstr); + + off = pstr->offsets[pstr->cur_idx]; + ch = pstr->raw_mbs[pstr->raw_mbs_idx + off]; + + if (! isascii (ch)) + return re_string_fetch_byte (pstr); + + re_string_skip_bytes (pstr, + re_string_char_size_at (pstr, pstr->cur_idx)); + return ch; + } +#endif + + return pstr->raw_mbs[pstr->raw_mbs_idx + pstr->cur_idx++]; +} + +static void +internal_function +re_string_destruct (re_string_t *pstr) +{ +#ifdef RE_ENABLE_I18N + re_free (pstr->wcs); + re_free (pstr->offsets); +#endif /* RE_ENABLE_I18N */ + if (pstr->mbs_allocated) + re_free (pstr->mbs); +} + +/* Return the context at IDX in INPUT. */ + +static unsigned int +internal_function +re_string_context_at (const re_string_t *input, int idx, int eflags) +{ + int c; + if (BE (idx < 0, 0)) + /* In this case, we use the value stored in input->tip_context, + since we can't know the character in input->mbs[-1] here. */ + return input->tip_context; + if (BE (idx == input->len, 0)) + return ((eflags & REG_NOTEOL) ? CONTEXT_ENDBUF + : CONTEXT_NEWLINE | CONTEXT_ENDBUF); +#ifdef RE_ENABLE_I18N + if (input->mb_cur_max > 1) + { + wint_t wc; + int wc_idx = idx; + while(input->wcs[wc_idx] == WEOF) + { +#if defined DEBUG && DEBUG + /* It must not happen. */ + assert (wc_idx >= 0); +#endif + --wc_idx; + if (wc_idx < 0) + return input->tip_context; + } + wc = input->wcs[wc_idx]; + if (BE (input->word_ops_used != 0, 0) && IS_WIDE_WORD_CHAR (wc)) + return CONTEXT_WORD; + return (IS_WIDE_NEWLINE (wc) && input->newline_anchor + ? CONTEXT_NEWLINE : 0); + } + else +#endif + { + c = re_string_byte_at (input, idx); + if (bitset_contain (input->word_char, c)) + return CONTEXT_WORD; + return IS_NEWLINE (c) && input->newline_anchor ? CONTEXT_NEWLINE : 0; + } +} + +/* Functions for set operation. */ + +static reg_errcode_t +internal_function __attribute_warn_unused_result__ +re_node_set_alloc (re_node_set *set, int size) +{ + /* + * ADR: valgrind says size can be 0, which then doesn't + * free the block of size 0. Harumph. This seems + * to work ok, though. + */ + if (size == 0) + { + memset(set, 0, sizeof(*set)); + return REG_NOERROR; + } + set->alloc = size; + set->nelem = 0; + set->elems = re_malloc (int, size); + if (BE (set->elems == NULL, 0)) + return REG_ESPACE; + return REG_NOERROR; +} + +static reg_errcode_t +internal_function __attribute_warn_unused_result__ +re_node_set_init_1 (re_node_set *set, int elem) +{ + set->alloc = 1; + set->nelem = 1; + set->elems = re_malloc (int, 1); + if (BE (set->elems == NULL, 0)) + { + set->alloc = set->nelem = 0; + return REG_ESPACE; + } + set->elems[0] = elem; + return REG_NOERROR; +} + +static reg_errcode_t +internal_function __attribute_warn_unused_result__ +re_node_set_init_2 (re_node_set *set, int elem1, int elem2) +{ + set->alloc = 2; + set->elems = re_malloc (int, 2); + if (BE (set->elems == NULL, 0)) + return REG_ESPACE; + if (elem1 == elem2) + { + set->nelem = 1; + set->elems[0] = elem1; + } + else + { + set->nelem = 2; + if (elem1 < elem2) + { + set->elems[0] = elem1; + set->elems[1] = elem2; + } + else + { + set->elems[0] = elem2; + set->elems[1] = elem1; + } + } + return REG_NOERROR; +} + +static reg_errcode_t +internal_function __attribute_warn_unused_result__ +re_node_set_init_copy (re_node_set *dest, const re_node_set *src) +{ + dest->nelem = src->nelem; + if (src->nelem > 0) + { + dest->alloc = dest->nelem; + dest->elems = re_malloc (int, dest->alloc); + if (BE (dest->elems == NULL, 0)) + { + dest->alloc = dest->nelem = 0; + return REG_ESPACE; + } + memcpy (dest->elems, src->elems, src->nelem * sizeof (int)); + } + else + re_node_set_init_empty (dest); + return REG_NOERROR; +} + +/* Calculate the intersection of the sets SRC1 and SRC2. And merge it to + DEST. Return value indicate the error code or REG_NOERROR if succeeded. + Note: We assume dest->elems is NULL, when dest->alloc is 0. */ + +static reg_errcode_t +internal_function __attribute_warn_unused_result__ +re_node_set_add_intersect (re_node_set *dest, const re_node_set *src1, + const re_node_set *src2) +{ + int i1, i2, is, id, delta, sbase; + if (src1->nelem == 0 || src2->nelem == 0) + return REG_NOERROR; + + /* We need dest->nelem + 2 * elems_in_intersection; this is a + conservative estimate. */ + if (src1->nelem + src2->nelem + dest->nelem > dest->alloc) + { + int new_alloc = src1->nelem + src2->nelem + dest->alloc; + int *new_elems = re_realloc (dest->elems, int, new_alloc); + if (BE (new_elems == NULL, 0)) + return REG_ESPACE; + dest->elems = new_elems; + dest->alloc = new_alloc; + } + + /* Find the items in the intersection of SRC1 and SRC2, and copy + into the top of DEST those that are not already in DEST itself. */ + sbase = dest->nelem + src1->nelem + src2->nelem; + i1 = src1->nelem - 1; + i2 = src2->nelem - 1; + id = dest->nelem - 1; + for (;;) + { + if (src1->elems[i1] == src2->elems[i2]) + { + /* Try to find the item in DEST. Maybe we could binary search? */ + while (id >= 0 && dest->elems[id] > src1->elems[i1]) + --id; + + if (id < 0 || dest->elems[id] != src1->elems[i1]) + dest->elems[--sbase] = src1->elems[i1]; + + if (--i1 < 0 || --i2 < 0) + break; + } + + /* Lower the highest of the two items. */ + else if (src1->elems[i1] < src2->elems[i2]) + { + if (--i2 < 0) + break; + } + else + { + if (--i1 < 0) + break; + } + } + + id = dest->nelem - 1; + is = dest->nelem + src1->nelem + src2->nelem - 1; + delta = is - sbase + 1; + + /* Now copy. When DELTA becomes zero, the remaining + DEST elements are already in place; this is more or + less the same loop that is in re_node_set_merge. */ + dest->nelem += delta; + if (delta > 0 && id >= 0) + for (;;) + { + if (dest->elems[is] > dest->elems[id]) + { + /* Copy from the top. */ + dest->elems[id + delta--] = dest->elems[is--]; + if (delta == 0) + break; + } + else + { + /* Slide from the bottom. */ + dest->elems[id + delta] = dest->elems[id]; + if (--id < 0) + break; + } + } + + /* Copy remaining SRC elements. */ + memcpy (dest->elems, dest->elems + sbase, delta * sizeof (int)); + + return REG_NOERROR; +} + +/* Calculate the union set of the sets SRC1 and SRC2. And store it to + DEST. Return value indicate the error code or REG_NOERROR if succeeded. */ + +static reg_errcode_t +internal_function __attribute_warn_unused_result__ +re_node_set_init_union (re_node_set *dest, const re_node_set *src1, + const re_node_set *src2) +{ + int i1, i2, id; + if (src1 != NULL && src1->nelem > 0 && src2 != NULL && src2->nelem > 0) + { + dest->alloc = src1->nelem + src2->nelem; + dest->elems = re_malloc (int, dest->alloc); + if (BE (dest->elems == NULL, 0)) + return REG_ESPACE; + } + else + { + if (src1 != NULL && src1->nelem > 0) + return re_node_set_init_copy (dest, src1); + else if (src2 != NULL && src2->nelem > 0) + return re_node_set_init_copy (dest, src2); + else + re_node_set_init_empty (dest); + return REG_NOERROR; + } + for (i1 = i2 = id = 0 ; i1 < src1->nelem && i2 < src2->nelem ;) + { + if (src1->elems[i1] > src2->elems[i2]) + { + dest->elems[id++] = src2->elems[i2++]; + continue; + } + if (src1->elems[i1] == src2->elems[i2]) + ++i2; + dest->elems[id++] = src1->elems[i1++]; + } + if (i1 < src1->nelem) + { + memcpy (dest->elems + id, src1->elems + i1, + (src1->nelem - i1) * sizeof (int)); + id += src1->nelem - i1; + } + else if (i2 < src2->nelem) + { + memcpy (dest->elems + id, src2->elems + i2, + (src2->nelem - i2) * sizeof (int)); + id += src2->nelem - i2; + } + dest->nelem = id; + return REG_NOERROR; +} + +/* Calculate the union set of the sets DEST and SRC. And store it to + DEST. Return value indicate the error code or REG_NOERROR if succeeded. */ + +static reg_errcode_t +internal_function __attribute_warn_unused_result__ +re_node_set_merge (re_node_set *dest, const re_node_set *src) +{ + int is, id, sbase, delta; + if (src == NULL || src->nelem == 0) + return REG_NOERROR; + if (dest->alloc < 2 * src->nelem + dest->nelem) + { + int new_alloc = 2 * (src->nelem + dest->alloc); + int *new_buffer = re_realloc (dest->elems, int, new_alloc); + if (BE (new_buffer == NULL, 0)) + return REG_ESPACE; + dest->elems = new_buffer; + dest->alloc = new_alloc; + } + + if (BE (dest->nelem == 0, 0)) + { + dest->nelem = src->nelem; + memcpy (dest->elems, src->elems, src->nelem * sizeof (int)); + return REG_NOERROR; + } + + /* Copy into the top of DEST the items of SRC that are not + found in DEST. Maybe we could binary search in DEST? */ + for (sbase = dest->nelem + 2 * src->nelem, + is = src->nelem - 1, id = dest->nelem - 1; is >= 0 && id >= 0; ) + { + if (dest->elems[id] == src->elems[is]) + is--, id--; + else if (dest->elems[id] < src->elems[is]) + dest->elems[--sbase] = src->elems[is--]; + else /* if (dest->elems[id] > src->elems[is]) */ + --id; + } + + if (is >= 0) + { + /* If DEST is exhausted, the remaining items of SRC must be unique. */ + sbase -= is + 1; + memcpy (dest->elems + sbase, src->elems, (is + 1) * sizeof (int)); + } + + id = dest->nelem - 1; + is = dest->nelem + 2 * src->nelem - 1; + delta = is - sbase + 1; + if (delta == 0) + return REG_NOERROR; + + /* Now copy. When DELTA becomes zero, the remaining + DEST elements are already in place. */ + dest->nelem += delta; + for (;;) + { + if (dest->elems[is] > dest->elems[id]) + { + /* Copy from the top. */ + dest->elems[id + delta--] = dest->elems[is--]; + if (delta == 0) + break; + } + else + { + /* Slide from the bottom. */ + dest->elems[id + delta] = dest->elems[id]; + if (--id < 0) + { + /* Copy remaining SRC elements. */ + memcpy (dest->elems, dest->elems + sbase, + delta * sizeof (int)); + break; + } + } + } + + return REG_NOERROR; +} + +/* Insert the new element ELEM to the re_node_set* SET. + SET should not already have ELEM. + return -1 if an error is occured, return 1 otherwise. */ + +static int +internal_function __attribute_warn_unused_result__ +re_node_set_insert (re_node_set *set, int elem) +{ + int idx; + /* In case the set is empty. */ + if (set->alloc == 0) + { + if (BE (re_node_set_init_1 (set, elem) == REG_NOERROR, 1)) + return 1; + else + return -1; + } + + if (BE (set->nelem, 0) == 0) + { + /* We already guaranteed above that set->alloc != 0. */ + set->elems[0] = elem; + ++set->nelem; + return 1; + } + + /* Realloc if we need. */ + if (set->alloc == set->nelem) + { + int *new_elems; + set->alloc = set->alloc * 2; + new_elems = re_realloc (set->elems, int, set->alloc); + if (BE (new_elems == NULL, 0)) + return -1; + set->elems = new_elems; + } + + /* Move the elements which follows the new element. Test the + first element separately to skip a check in the inner loop. */ + if (elem < set->elems[0]) + { + idx = 0; + for (idx = set->nelem; idx > 0; idx--) + set->elems[idx] = set->elems[idx - 1]; + } + else + { + for (idx = set->nelem; set->elems[idx - 1] > elem; idx--) + set->elems[idx] = set->elems[idx - 1]; + } + + /* Insert the new element. */ + set->elems[idx] = elem; + ++set->nelem; + return 1; +} + +/* Insert the new element ELEM to the re_node_set* SET. + SET should not already have any element greater than or equal to ELEM. + Return -1 if an error is occured, return 1 otherwise. */ + +static int +internal_function __attribute_warn_unused_result__ +re_node_set_insert_last (re_node_set *set, int elem) +{ + /* Realloc if we need. */ + if (set->alloc == set->nelem) + { + int *new_elems; + set->alloc = (set->alloc + 1) * 2; + new_elems = re_realloc (set->elems, int, set->alloc); + if (BE (new_elems == NULL, 0)) + return -1; + set->elems = new_elems; + } + + /* Insert the new element. */ + set->elems[set->nelem++] = elem; + return 1; +} + +/* Compare two node sets SET1 and SET2. + return 1 if SET1 and SET2 are equivalent, return 0 otherwise. */ + +static int +internal_function __attribute__ ((__pure__)) +re_node_set_compare (const re_node_set *set1, const re_node_set *set2) +{ + int i; + if (set1 == NULL || set2 == NULL || set1->nelem != set2->nelem) + return 0; + for (i = set1->nelem ; --i >= 0 ; ) + if (set1->elems[i] != set2->elems[i]) + return 0; + return 1; +} + +/* Return (idx + 1) if SET contains the element ELEM, return 0 otherwise. */ + +static int +internal_function __attribute__ ((__pure__)) +re_node_set_contains (const re_node_set *set, int elem) +{ + unsigned int idx, right, mid; + if (set->nelem <= 0) + return 0; + + /* Binary search the element. */ + idx = 0; + right = set->nelem - 1; + while (idx < right) + { + mid = (idx + right) / 2; + if (set->elems[mid] < elem) + idx = mid + 1; + else + right = mid; + } + return set->elems[idx] == elem ? idx + 1 : 0; +} + +static void +internal_function +re_node_set_remove_at (re_node_set *set, int idx) +{ + if (idx < 0 || idx >= set->nelem) + return; + --set->nelem; + for (; idx < set->nelem; idx++) + set->elems[idx] = set->elems[idx + 1]; +} + + +/* Add the token TOKEN to dfa->nodes, and return the index of the token. + Or return -1, if an error will be occured. */ + +static int +internal_function +re_dfa_add_node (re_dfa_t *dfa, re_token_t token) +{ + if (BE (dfa->nodes_len >= dfa->nodes_alloc, 0)) + { + size_t new_nodes_alloc = dfa->nodes_alloc * 2; + int *new_nexts, *new_indices; + re_node_set *new_edests, *new_eclosures; + re_token_t *new_nodes; + + /* Avoid overflows in realloc. */ + const size_t max_object_size = MAX (sizeof (re_token_t), + MAX (sizeof (re_node_set), + sizeof (int))); + if (BE (SIZE_MAX / max_object_size < new_nodes_alloc, 0)) + return -1; + + new_nodes = re_realloc (dfa->nodes, re_token_t, new_nodes_alloc); + if (BE (new_nodes == NULL, 0)) + return -1; + dfa->nodes = new_nodes; + new_nexts = re_realloc (dfa->nexts, int, new_nodes_alloc); + new_indices = re_realloc (dfa->org_indices, int, new_nodes_alloc); + new_edests = re_realloc (dfa->edests, re_node_set, new_nodes_alloc); + new_eclosures = re_realloc (dfa->eclosures, re_node_set, new_nodes_alloc); + if (BE (new_nexts == NULL || new_indices == NULL + || new_edests == NULL || new_eclosures == NULL, 0)) + { + /* if any are not NULL, free them, avoid leaks */ + if (new_nexts != NULL) + re_free(new_nexts); + if (new_indices != NULL) + re_free(new_indices); + if (new_edests != NULL) + re_free(new_edests); + if (new_eclosures != NULL) + re_free(new_eclosures); + return -1; + } + dfa->nexts = new_nexts; + dfa->org_indices = new_indices; + dfa->edests = new_edests; + dfa->eclosures = new_eclosures; + dfa->nodes_alloc = new_nodes_alloc; + } + dfa->nodes[dfa->nodes_len] = token; + dfa->nodes[dfa->nodes_len].constraint = 0; +#ifdef RE_ENABLE_I18N + dfa->nodes[dfa->nodes_len].accept_mb = + (token.type == OP_PERIOD && dfa->mb_cur_max > 1) || token.type == COMPLEX_BRACKET; +#endif + dfa->nexts[dfa->nodes_len] = -1; + re_node_set_init_empty (dfa->edests + dfa->nodes_len); + re_node_set_init_empty (dfa->eclosures + dfa->nodes_len); + return dfa->nodes_len++; +} + +static inline unsigned int +internal_function +calc_state_hash (const re_node_set *nodes, unsigned int context) +{ + unsigned int hash = nodes->nelem + context; + int i; + for (i = 0 ; i < nodes->nelem ; i++) + hash += nodes->elems[i]; + return hash; +} + +/* Search for the state whose node_set is equivalent to NODES. + Return the pointer to the state, if we found it in the DFA. + Otherwise create the new one and return it. In case of an error + return NULL and set the error code in ERR. + Note: - We assume NULL as the invalid state, then it is possible that + return value is NULL and ERR is REG_NOERROR. + - We never return non-NULL value in case of any errors, it is for + optimization. */ + +static re_dfastate_t * +internal_function __attribute_warn_unused_result__ +re_acquire_state (reg_errcode_t *err, const re_dfa_t *dfa, + const re_node_set *nodes) +{ + unsigned int hash; + re_dfastate_t *new_state; + struct re_state_table_entry *spot; + int i; + if (BE (nodes->nelem == 0, 0)) + { + *err = REG_NOERROR; + return NULL; + } + hash = calc_state_hash (nodes, 0); + spot = dfa->state_table + (hash & dfa->state_hash_mask); + + for (i = 0 ; i < spot->num ; i++) + { + re_dfastate_t *state = spot->array[i]; + if (hash != state->hash) + continue; + if (re_node_set_compare (&state->nodes, nodes)) + return state; + } + + /* There are no appropriate state in the dfa, create the new one. */ + new_state = create_ci_newstate (dfa, nodes, hash); + if (BE (new_state == NULL, 0)) + *err = REG_ESPACE; + + return new_state; +} + +/* Search for the state whose node_set is equivalent to NODES and + whose context is equivalent to CONTEXT. + Return the pointer to the state, if we found it in the DFA. + Otherwise create the new one and return it. In case of an error + return NULL and set the error code in ERR. + Note: - We assume NULL as the invalid state, then it is possible that + return value is NULL and ERR is REG_NOERROR. + - We never return non-NULL value in case of any errors, it is for + optimization. */ + +static re_dfastate_t * +internal_function __attribute_warn_unused_result__ +re_acquire_state_context (reg_errcode_t *err, const re_dfa_t *dfa, + const re_node_set *nodes, unsigned int context) +{ + unsigned int hash; + re_dfastate_t *new_state; + struct re_state_table_entry *spot; + int i; + if (nodes->nelem == 0) + { + *err = REG_NOERROR; + return NULL; + } + hash = calc_state_hash (nodes, context); + spot = dfa->state_table + (hash & dfa->state_hash_mask); + + for (i = 0 ; i < spot->num ; i++) + { + re_dfastate_t *state = spot->array[i]; + if (state->hash == hash + && state->context == context + && re_node_set_compare (state->entrance_nodes, nodes)) + return state; + } + /* There are no appropriate state in `dfa', create the new one. */ + new_state = create_cd_newstate (dfa, nodes, context, hash); + if (BE (new_state == NULL, 0)) + *err = REG_ESPACE; + + return new_state; +} + +/* Finish initialization of the new state NEWSTATE, and using its hash value + HASH put in the appropriate bucket of DFA's state table. Return value + indicates the error code if failed. */ + +static reg_errcode_t +__attribute_warn_unused_result__ +register_state (const re_dfa_t *dfa, re_dfastate_t *newstate, + unsigned int hash) +{ + struct re_state_table_entry *spot; + reg_errcode_t err; + int i; + + newstate->hash = hash; + err = re_node_set_alloc (&newstate->non_eps_nodes, newstate->nodes.nelem); + if (BE (err != REG_NOERROR, 0)) + return REG_ESPACE; + for (i = 0; i < newstate->nodes.nelem; i++) + { + int elem = newstate->nodes.elems[i]; + if (!IS_EPSILON_NODE (dfa->nodes[elem].type)) + if (re_node_set_insert_last (&newstate->non_eps_nodes, elem) < 0) + return REG_ESPACE; + } + + spot = dfa->state_table + (hash & dfa->state_hash_mask); + if (BE (spot->alloc <= spot->num, 0)) + { + int new_alloc = 2 * spot->num + 2; + re_dfastate_t **new_array = re_realloc (spot->array, re_dfastate_t *, + new_alloc); + if (BE (new_array == NULL, 0)) + return REG_ESPACE; + spot->array = new_array; + spot->alloc = new_alloc; + } + spot->array[spot->num++] = newstate; + return REG_NOERROR; +} + +static void +free_state (re_dfastate_t *state) +{ + re_node_set_free (&state->non_eps_nodes); + re_node_set_free (&state->inveclosure); + if (state->entrance_nodes != &state->nodes) + { + re_node_set_free (state->entrance_nodes); + re_free (state->entrance_nodes); + } + re_node_set_free (&state->nodes); + re_free (state->word_trtable); + re_free (state->trtable); + re_free (state); +} + +/* Create the new state which is independent of contexts. + Return the new state if succeeded, otherwise return NULL. */ + +static re_dfastate_t * +internal_function __attribute_warn_unused_result__ +create_ci_newstate (const re_dfa_t *dfa, const re_node_set *nodes, + unsigned int hash) +{ + int i; + reg_errcode_t err; + re_dfastate_t *newstate; + + newstate = (re_dfastate_t *) calloc (sizeof (re_dfastate_t), 1); + if (BE (newstate == NULL, 0)) + return NULL; + err = re_node_set_init_copy (&newstate->nodes, nodes); + if (BE (err != REG_NOERROR, 0)) + { + re_free (newstate); + return NULL; + } + + newstate->entrance_nodes = &newstate->nodes; + for (i = 0 ; i < nodes->nelem ; i++) + { + re_token_t *node = dfa->nodes + nodes->elems[i]; + re_token_type_t type = node->type; + if (type == CHARACTER && !node->constraint) + continue; +#ifdef RE_ENABLE_I18N + newstate->accept_mb |= node->accept_mb; +#endif /* RE_ENABLE_I18N */ + + /* If the state has the halt node, the state is a halt state. */ + if (type == END_OF_RE) + newstate->halt = 1; + else if (type == OP_BACK_REF) + newstate->has_backref = 1; + else if (type == ANCHOR || node->constraint) + newstate->has_constraint = 1; + } + err = register_state (dfa, newstate, hash); + if (BE (err != REG_NOERROR, 0)) + { + free_state (newstate); + newstate = NULL; + } + return newstate; +} + +/* Create the new state which is depend on the context CONTEXT. + Return the new state if succeeded, otherwise return NULL. */ + +static re_dfastate_t * +internal_function __attribute_warn_unused_result__ +create_cd_newstate (const re_dfa_t *dfa, const re_node_set *nodes, + unsigned int context, unsigned int hash) +{ + int i, nctx_nodes = 0; + reg_errcode_t err; + re_dfastate_t *newstate; + + newstate = (re_dfastate_t *) calloc (sizeof (re_dfastate_t), 1); + if (BE (newstate == NULL, 0)) + return NULL; + err = re_node_set_init_copy (&newstate->nodes, nodes); + if (BE (err != REG_NOERROR, 0)) + { + re_free (newstate); + return NULL; + } + + newstate->context = context; + newstate->entrance_nodes = &newstate->nodes; + + for (i = 0 ; i < nodes->nelem ; i++) + { + re_token_t *node = dfa->nodes + nodes->elems[i]; + re_token_type_t type = node->type; + unsigned int constraint = node->constraint; + + if (type == CHARACTER && !constraint) + continue; +#ifdef RE_ENABLE_I18N + newstate->accept_mb |= node->accept_mb; +#endif /* RE_ENABLE_I18N */ + + /* If the state has the halt node, the state is a halt state. */ + if (type == END_OF_RE) + newstate->halt = 1; + else if (type == OP_BACK_REF) + newstate->has_backref = 1; + + if (constraint) + { + if (newstate->entrance_nodes == &newstate->nodes) + { + newstate->entrance_nodes = re_malloc (re_node_set, 1); + if (BE (newstate->entrance_nodes == NULL, 0)) + { + free_state (newstate); + return NULL; + } + if (re_node_set_init_copy (newstate->entrance_nodes, nodes) + != REG_NOERROR) + return NULL; + nctx_nodes = 0; + newstate->has_constraint = 1; + } + + if (NOT_SATISFY_PREV_CONSTRAINT (constraint,context)) + { + re_node_set_remove_at (&newstate->nodes, i - nctx_nodes); + ++nctx_nodes; + } + } + } + err = register_state (dfa, newstate, hash); + if (BE (err != REG_NOERROR, 0)) + { + free_state (newstate); + newstate = NULL; + } + return newstate; +} diff --git a/support/regex_internal.h b/support/regex_internal.h new file mode 100644 index 00000000..01465e76 --- /dev/null +++ b/support/regex_internal.h @@ -0,0 +1,831 @@ +/* Extended regular expression matching and search library. + Copyright (C) 2002-2016 Free Software Foundation, Inc. + This file is part of the GNU C Library. + Contributed by Isamu Hasegawa <isamu@yamato.ibm.com>. + + The GNU C Library is free software; you can redistribute it and/or + modify it under the terms of the GNU Lesser General Public + License as published by the Free Software Foundation; either + version 2.1 of the License, or (at your option) any later version. + + The GNU C Library is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public + License along with the GNU C Library; if not, see + <http://www.gnu.org/licenses/>. */ + +#ifndef _REGEX_INTERNAL_H +#define _REGEX_INTERNAL_H 1 + +#include <assert.h> +#include <ctype.h> +#include <stdio.h> +#include <stdlib.h> +#include <string.h> + +#if defined HAVE_LANGINFO_H || defined HAVE_LANGINFO_CODESET || defined _LIBC +# include <langinfo.h> +#endif +#if defined HAVE_LOCALE_H || defined _LIBC +# include <locale.h> +#endif +#if defined HAVE_WCHAR_H || defined _LIBC +# include <wchar.h> +#endif /* HAVE_WCHAR_H || _LIBC */ +#if defined HAVE_WCTYPE_H || defined _LIBC +# include <wctype.h> +#endif /* HAVE_WCTYPE_H || _LIBC */ +#if defined HAVE_STDBOOL_H || defined _LIBC +# include <stdbool.h> +#endif /* HAVE_STDBOOL_H || _LIBC */ +#if defined HAVE_STDINT_H || defined _LIBC +# include <stdint.h> +#endif /* HAVE_STDINT_H || _LIBC */ + +#include "intprops.h" + +#if defined _LIBC +# include <libc-lock.h> +#else +# define __libc_lock_define(CLASS,NAME) +# define __libc_lock_init(NAME) do { } while (0) +# define __libc_lock_lock(NAME) do { } while (0) +# define __libc_lock_unlock(NAME) do { } while (0) +#endif + +#ifndef GAWK +/* In case that the system doesn't have isblank(). */ +#if !defined _LIBC && !defined HAVE_ISBLANK && !defined isblank +# define isblank(ch) ((ch) == ' ' || (ch) == '\t') +#endif +#else /* GAWK */ +/* + * This is a freaking mess. On glibc systems you have to define + * a magic constant to get isblank() out of <ctype.h>, since it's + * a C99 function. To heck with all that and borrow a page from + * dfa.c's book. + */ + +static int +is_blank (int c) +{ + return (c == ' ' || c == '\t'); +} +#endif /* GAWK */ + +#ifdef _LIBC +# ifndef _RE_DEFINE_LOCALE_FUNCTIONS +# define _RE_DEFINE_LOCALE_FUNCTIONS 1 +# include <locale/localeinfo.h> +# include <locale/coll-lookup.h> +# endif +#endif + +/* This is for other GNU distributions with internationalized messages. */ +#if (HAVE_LIBINTL_H && ENABLE_NLS) || defined _LIBC +# include <libintl.h> +# ifdef _LIBC +# undef gettext +# define gettext(msgid) \ + __dcgettext (_libc_intl_domainname, msgid, LC_MESSAGES) +# endif +#else +# define gettext(msgid) (msgid) +#endif + +#ifndef gettext_noop +/* This define is so xgettext can find the internationalizable + strings. */ +# define gettext_noop(String) String +#endif + +/* For loser systems without the definition. */ +#ifndef SIZE_MAX +# define SIZE_MAX ((size_t) -1) +#endif + +#if ! defined(__DJGPP__) && (defined(GAWK) || _LIBC) +# define RE_ENABLE_I18N +#endif + +#if __GNUC__ >= 3 +# define BE(expr, val) __builtin_expect (expr, val) +#else +# define BE(expr, val) (expr) +# ifdef inline +# undef inline +# endif +# define inline +#endif + +/* Number of single byte character. */ +#define SBC_MAX 256 + +#define COLL_ELEM_LEN_MAX 8 + +/* The character which represents newline. */ +#define NEWLINE_CHAR '\n' +#define WIDE_NEWLINE_CHAR L'\n' + +/* Rename to standard API for using out of glibc. */ +#ifndef _LIBC +# ifdef __wctype +# undef __wctype +# endif +# define __wctype wctype +# ifdef __iswctype +# undef __iswctype +# endif +# define __iswctype iswctype +# define __btowc btowc +# define __mbrtowc mbrtowc +#undef __mempcpy /* GAWK */ +# define __mempcpy mempcpy +# define __wcrtomb wcrtomb +# define __regfree regfree +#endif /* not _LIBC */ + +#if __GNUC__ < 3 + (__GNUC_MINOR__ < 1) +# define __attribute__(arg) +#endif + +#ifdef GAWK +/* + * Instead of trying to figure out which GCC version introduced + * this symbol, just define it out and be done. + */ +# undef __attribute_warn_unused_result__ +# define __attribute_warn_unused_result__ +#endif + +/* An integer used to represent a set of bits. It must be unsigned, + and must be at least as wide as unsigned int. */ +typedef unsigned long int bitset_word_t; +/* All bits set in a bitset_word_t. */ +#define BITSET_WORD_MAX ULONG_MAX +/* Number of bits in a bitset_word_t. */ +#define BITSET_WORD_BITS (sizeof (bitset_word_t) * CHAR_BIT) +/* Number of bitset_word_t in a bit_set. */ +#define BITSET_WORDS (SBC_MAX / BITSET_WORD_BITS) +typedef bitset_word_t bitset_t[BITSET_WORDS]; +typedef bitset_word_t *re_bitset_ptr_t; +typedef const bitset_word_t *re_const_bitset_ptr_t; + +#define bitset_set(set,i) \ + (set[i / BITSET_WORD_BITS] |= (bitset_word_t) 1 << i % BITSET_WORD_BITS) +#define bitset_clear(set,i) \ + (set[i / BITSET_WORD_BITS] &= ~((bitset_word_t) 1 << i % BITSET_WORD_BITS)) +#define bitset_contain(set,i) \ + (set[i / BITSET_WORD_BITS] & ((bitset_word_t) 1 << i % BITSET_WORD_BITS)) +#define bitset_empty(set) memset (set, '\0', sizeof (bitset_t)) +#define bitset_set_all(set) memset (set, '\xff', sizeof (bitset_t)) +#define bitset_copy(dest,src) memcpy (dest, src, sizeof (bitset_t)) + +#define PREV_WORD_CONSTRAINT 0x0001 +#define PREV_NOTWORD_CONSTRAINT 0x0002 +#define NEXT_WORD_CONSTRAINT 0x0004 +#define NEXT_NOTWORD_CONSTRAINT 0x0008 +#define PREV_NEWLINE_CONSTRAINT 0x0010 +#define NEXT_NEWLINE_CONSTRAINT 0x0020 +#define PREV_BEGBUF_CONSTRAINT 0x0040 +#define NEXT_ENDBUF_CONSTRAINT 0x0080 +#define WORD_DELIM_CONSTRAINT 0x0100 +#define NOT_WORD_DELIM_CONSTRAINT 0x0200 + +typedef enum +{ + INSIDE_WORD = PREV_WORD_CONSTRAINT | NEXT_WORD_CONSTRAINT, + WORD_FIRST = PREV_NOTWORD_CONSTRAINT | NEXT_WORD_CONSTRAINT, + WORD_LAST = PREV_WORD_CONSTRAINT | NEXT_NOTWORD_CONSTRAINT, + INSIDE_NOTWORD = PREV_NOTWORD_CONSTRAINT | NEXT_NOTWORD_CONSTRAINT, + LINE_FIRST = PREV_NEWLINE_CONSTRAINT, + LINE_LAST = NEXT_NEWLINE_CONSTRAINT, + BUF_FIRST = PREV_BEGBUF_CONSTRAINT, + BUF_LAST = NEXT_ENDBUF_CONSTRAINT, + WORD_DELIM = WORD_DELIM_CONSTRAINT, + NOT_WORD_DELIM = NOT_WORD_DELIM_CONSTRAINT +} re_context_type; + +typedef struct +{ + int alloc; + int nelem; + int *elems; +} re_node_set; + +typedef enum +{ + NON_TYPE = 0, + + /* Node type, These are used by token, node, tree. */ + CHARACTER = 1, + END_OF_RE = 2, + SIMPLE_BRACKET = 3, + OP_BACK_REF = 4, + OP_PERIOD = 5, +#ifdef RE_ENABLE_I18N + COMPLEX_BRACKET = 6, + OP_UTF8_PERIOD = 7, +#endif /* RE_ENABLE_I18N */ + + /* We define EPSILON_BIT as a macro so that OP_OPEN_SUBEXP is used + when the debugger shows values of this enum type. */ +#define EPSILON_BIT 8 + OP_OPEN_SUBEXP = EPSILON_BIT | 0, + OP_CLOSE_SUBEXP = EPSILON_BIT | 1, + OP_ALT = EPSILON_BIT | 2, + OP_DUP_ASTERISK = EPSILON_BIT | 3, + ANCHOR = EPSILON_BIT | 4, + + /* Tree type, these are used only by tree. */ + CONCAT = 16, + SUBEXP = 17, + + /* Token type, these are used only by token. */ + OP_DUP_PLUS = 18, + OP_DUP_QUESTION, + OP_OPEN_BRACKET, + OP_CLOSE_BRACKET, + OP_CHARSET_RANGE, + OP_OPEN_DUP_NUM, + OP_CLOSE_DUP_NUM, + OP_NON_MATCH_LIST, + OP_OPEN_COLL_ELEM, + OP_CLOSE_COLL_ELEM, + OP_OPEN_EQUIV_CLASS, + OP_CLOSE_EQUIV_CLASS, + OP_OPEN_CHAR_CLASS, + OP_CLOSE_CHAR_CLASS, + OP_WORD, + OP_NOTWORD, + OP_SPACE, + OP_NOTSPACE, + BACK_SLASH + +} re_token_type_t; + +#ifdef RE_ENABLE_I18N +typedef struct +{ + /* Multibyte characters. */ + wchar_t *mbchars; + + /* Collating symbols. */ +# ifdef _LIBC + int32_t *coll_syms; +# endif + + /* Equivalence classes. */ +# ifdef _LIBC + int32_t *equiv_classes; +# endif + + /* Range expressions. */ +# ifdef _LIBC + uint32_t *range_starts; + uint32_t *range_ends; +# else /* not _LIBC */ + wchar_t *range_starts; + wchar_t *range_ends; +# endif /* not _LIBC */ + + /* Character classes. */ + wctype_t *char_classes; + + /* If this character set is the non-matching list. */ + unsigned int non_match : 1; + + /* # of multibyte characters. */ + int nmbchars; + + /* # of collating symbols. */ + int ncoll_syms; + + /* # of equivalence classes. */ + int nequiv_classes; + + /* # of range expressions. */ + int nranges; + + /* # of character classes. */ + int nchar_classes; +} re_charset_t; +#endif /* RE_ENABLE_I18N */ + +typedef struct +{ + union + { + unsigned char c; /* for CHARACTER */ + re_bitset_ptr_t sbcset; /* for SIMPLE_BRACKET */ +#ifdef RE_ENABLE_I18N + re_charset_t *mbcset; /* for COMPLEX_BRACKET */ +#endif /* RE_ENABLE_I18N */ + int idx; /* for BACK_REF */ + re_context_type ctx_type; /* for ANCHOR */ + } opr; +#if __GNUC__ >= 2 + re_token_type_t type : 8; +#else + re_token_type_t type; +#endif + unsigned int constraint : 10; /* context constraint */ + unsigned int duplicated : 1; + unsigned int opt_subexp : 1; +#ifdef RE_ENABLE_I18N + unsigned int accept_mb : 1; + /* These 2 bits can be moved into the union if needed (e.g. if running out + of bits; move opr.c to opr.c.c and move the flags to opr.c.flags). */ + unsigned int mb_partial : 1; +#endif + unsigned int word_char : 1; +} re_token_t; + +#define IS_EPSILON_NODE(type) ((type) & EPSILON_BIT) + +struct re_string_t +{ + /* Indicate the raw buffer which is the original string passed as an + argument of regexec(), re_search(), etc.. */ + const unsigned char *raw_mbs; + /* Store the multibyte string. In case of "case insensitive mode" like + REG_ICASE, upper cases of the string are stored, otherwise MBS points + the same address that RAW_MBS points. */ + unsigned char *mbs; +#ifdef RE_ENABLE_I18N + /* Store the wide character string which is corresponding to MBS. */ + wint_t *wcs; + int *offsets; + mbstate_t cur_state; +#endif + /* Index in RAW_MBS. Each character mbs[i] corresponds to + raw_mbs[raw_mbs_idx + i]. */ + int raw_mbs_idx; + /* The length of the valid characters in the buffers. */ + int valid_len; + /* The corresponding number of bytes in raw_mbs array. */ + int valid_raw_len; + /* The length of the buffers MBS and WCS. */ + int bufs_len; + /* The index in MBS, which is updated by re_string_fetch_byte. */ + int cur_idx; + /* length of RAW_MBS array. */ + int raw_len; + /* This is RAW_LEN - RAW_MBS_IDX + VALID_LEN - VALID_RAW_LEN. */ + int len; + /* End of the buffer may be shorter than its length in the cases such + as re_match_2, re_search_2. Then, we use STOP for end of the buffer + instead of LEN. */ + int raw_stop; + /* This is RAW_STOP - RAW_MBS_IDX adjusted through OFFSETS. */ + int stop; + + /* The context of mbs[0]. We store the context independently, since + the context of mbs[0] may be different from raw_mbs[0], which is + the beginning of the input string. */ + unsigned int tip_context; + /* The translation passed as a part of an argument of re_compile_pattern. */ + RE_TRANSLATE_TYPE trans; + /* Copy of re_dfa_t's word_char. */ + re_const_bitset_ptr_t word_char; + /* 1 if REG_ICASE. */ + unsigned char icase; + unsigned char is_utf8; + unsigned char map_notascii; + unsigned char mbs_allocated; + unsigned char offsets_needed; + unsigned char newline_anchor; + unsigned char word_ops_used; + int mb_cur_max; +}; +typedef struct re_string_t re_string_t; + + +struct re_dfa_t; +typedef struct re_dfa_t re_dfa_t; + +#ifndef _LIBC +# ifdef __i386__ +# define internal_function __attribute__ ((regparm (3), stdcall)) +# else +# define internal_function +# endif +#endif + +#ifndef NOT_IN_libc +static reg_errcode_t re_string_realloc_buffers (re_string_t *pstr, + int new_buf_len) + internal_function; +# ifdef RE_ENABLE_I18N +static void build_wcs_buffer (re_string_t *pstr) internal_function; +static reg_errcode_t build_wcs_upper_buffer (re_string_t *pstr) + internal_function; +# endif /* RE_ENABLE_I18N */ +static void build_upper_buffer (re_string_t *pstr) internal_function; +static void re_string_translate_buffer (re_string_t *pstr) internal_function; +static unsigned int re_string_context_at (const re_string_t *input, int idx, + int eflags) + internal_function __attribute__ ((pure)); +#endif +#define re_string_peek_byte(pstr, offset) \ + ((pstr)->mbs[(pstr)->cur_idx + offset]) +#define re_string_fetch_byte(pstr) \ + ((pstr)->mbs[(pstr)->cur_idx++]) +#define re_string_first_byte(pstr, idx) \ + ((idx) == (pstr)->valid_len || (pstr)->wcs[idx] != WEOF) +#define re_string_is_single_byte_char(pstr, idx) \ + ((pstr)->wcs[idx] != WEOF && ((pstr)->valid_len == (idx) + 1 \ + || (pstr)->wcs[(idx) + 1] != WEOF)) +#define re_string_eoi(pstr) ((pstr)->stop <= (pstr)->cur_idx) +#define re_string_cur_idx(pstr) ((pstr)->cur_idx) +#define re_string_get_buffer(pstr) ((pstr)->mbs) +#define re_string_length(pstr) ((pstr)->len) +#define re_string_byte_at(pstr,idx) ((pstr)->mbs[idx]) +#define re_string_skip_bytes(pstr,idx) ((pstr)->cur_idx += (idx)) +#define re_string_set_index(pstr,idx) ((pstr)->cur_idx = (idx)) + +#ifndef _LIBC +# if HAVE_ALLOCA +# include <alloca.h> +/* The OS usually guarantees only one guard page at the bottom of the stack, + and a page size can be as small as 4096 bytes. So we cannot safely + allocate anything larger than 4096 bytes. Also care for the possibility + of a few compiler-allocated temporary stack slots. */ +# define __libc_use_alloca(n) ((n) < 4032) +# else +/* alloca is implemented with malloc, so just use malloc. */ +# define __libc_use_alloca(n) 0 +# endif +#endif + +/* + * GAWK checks for zero-size allocations everywhere else, + * do it here too. + */ +#ifndef GAWK +#define re_malloc(t,n) ((t *) malloc ((n) * sizeof (t))) +#define re_realloc(p,t,n) ((t *) realloc (p, (n) * sizeof (t))) +#else +static void * +test_malloc(size_t count, const char *file, size_t line) +{ + if (count == 0) { + fprintf(stderr, "%s:%lu: allocation of zero bytes\n", + file, (unsigned long) line); + exit(1); + } + return malloc(count); +} + +static void * +test_realloc(void *p, size_t count, const char *file, size_t line) +{ + if (count == 0) { + fprintf(stderr, "%s:%lu: reallocation of zero bytes\n", + file, (unsigned long) line); + exit(1); + } + return realloc(p, count); +} +#define re_malloc(t,n) ((t *) test_malloc (((n) * sizeof (t)), __FILE__, __LINE__)) +#define re_realloc(p,t,n) ((t *) test_realloc (p, (n) * sizeof (t), __FILE__, __LINE__)) +#endif +#define re_free(p) free (p) + +struct bin_tree_t +{ + struct bin_tree_t *parent; + struct bin_tree_t *left; + struct bin_tree_t *right; + struct bin_tree_t *first; + struct bin_tree_t *next; + + re_token_t token; + + /* `node_idx' is the index in dfa->nodes, if `type' == 0. + Otherwise `type' indicate the type of this node. */ + int node_idx; +}; +typedef struct bin_tree_t bin_tree_t; + +#define BIN_TREE_STORAGE_SIZE \ + ((1024 - sizeof (void *)) / sizeof (bin_tree_t)) + +struct bin_tree_storage_t +{ + struct bin_tree_storage_t *next; + bin_tree_t data[BIN_TREE_STORAGE_SIZE]; +}; +typedef struct bin_tree_storage_t bin_tree_storage_t; + +#define CONTEXT_WORD 1 +#define CONTEXT_NEWLINE (CONTEXT_WORD << 1) +#define CONTEXT_BEGBUF (CONTEXT_NEWLINE << 1) +#define CONTEXT_ENDBUF (CONTEXT_BEGBUF << 1) + +#define IS_WORD_CONTEXT(c) ((c) & CONTEXT_WORD) +#define IS_NEWLINE_CONTEXT(c) ((c) & CONTEXT_NEWLINE) +#define IS_BEGBUF_CONTEXT(c) ((c) & CONTEXT_BEGBUF) +#define IS_ENDBUF_CONTEXT(c) ((c) & CONTEXT_ENDBUF) +#define IS_ORDINARY_CONTEXT(c) ((c) == 0) + +#define IS_WORD_CHAR(ch) (isalnum (ch) || (ch) == '_') +#define IS_NEWLINE(ch) ((ch) == NEWLINE_CHAR) +#define IS_WIDE_WORD_CHAR(ch) (iswalnum (ch) || (ch) == L'_') +#define IS_WIDE_NEWLINE(ch) ((ch) == WIDE_NEWLINE_CHAR) + +#define NOT_SATISFY_PREV_CONSTRAINT(constraint,context) \ + ((((constraint) & PREV_WORD_CONSTRAINT) && !IS_WORD_CONTEXT (context)) \ + || ((constraint & PREV_NOTWORD_CONSTRAINT) && IS_WORD_CONTEXT (context)) \ + || ((constraint & PREV_NEWLINE_CONSTRAINT) && !IS_NEWLINE_CONTEXT (context))\ + || ((constraint & PREV_BEGBUF_CONSTRAINT) && !IS_BEGBUF_CONTEXT (context))) + +#define NOT_SATISFY_NEXT_CONSTRAINT(constraint,context) \ + ((((constraint) & NEXT_WORD_CONSTRAINT) && !IS_WORD_CONTEXT (context)) \ + || (((constraint) & NEXT_NOTWORD_CONSTRAINT) && IS_WORD_CONTEXT (context)) \ + || (((constraint) & NEXT_NEWLINE_CONSTRAINT) && !IS_NEWLINE_CONTEXT (context)) \ + || (((constraint) & NEXT_ENDBUF_CONSTRAINT) && !IS_ENDBUF_CONTEXT (context))) + +struct re_dfastate_t +{ + unsigned int hash; + re_node_set nodes; + re_node_set non_eps_nodes; + re_node_set inveclosure; + re_node_set *entrance_nodes; + struct re_dfastate_t **trtable, **word_trtable; + unsigned int context : 4; + unsigned int halt : 1; + /* If this state can accept `multi byte'. + Note that we refer to multibyte characters, and multi character + collating elements as `multi byte'. */ + unsigned int accept_mb : 1; + /* If this state has backreference node(s). */ + unsigned int has_backref : 1; + unsigned int has_constraint : 1; +}; +typedef struct re_dfastate_t re_dfastate_t; + +struct re_state_table_entry +{ + int num; + int alloc; + re_dfastate_t **array; +}; + +/* Array type used in re_sub_match_last_t and re_sub_match_top_t. */ + +typedef struct +{ + int next_idx; + int alloc; + re_dfastate_t **array; +} state_array_t; + +/* Store information about the node NODE whose type is OP_CLOSE_SUBEXP. */ + +typedef struct +{ + int node; + int str_idx; /* The position NODE match at. */ + state_array_t path; +} re_sub_match_last_t; + +/* Store information about the node NODE whose type is OP_OPEN_SUBEXP. + And information about the node, whose type is OP_CLOSE_SUBEXP, + corresponding to NODE is stored in LASTS. */ + +typedef struct +{ + int str_idx; + int node; + state_array_t *path; + int alasts; /* Allocation size of LASTS. */ + int nlasts; /* The number of LASTS. */ + re_sub_match_last_t **lasts; +} re_sub_match_top_t; + +struct re_backref_cache_entry +{ + int node; + int str_idx; + int subexp_from; + int subexp_to; + char more; + char unused; + unsigned short int eps_reachable_subexps_map; +}; + +typedef struct +{ + /* The string object corresponding to the input string. */ + re_string_t input; +#if defined _LIBC || (defined __STDC_VERSION__ && __STDC_VERSION__ >= 199901L) + const re_dfa_t *const dfa; +#else + const re_dfa_t *dfa; +#endif + /* EFLAGS of the argument of regexec. */ + int eflags; + /* Where the matching ends. */ + int match_last; + int last_node; + /* The state log used by the matcher. */ + re_dfastate_t **state_log; + int state_log_top; + /* Back reference cache. */ + int nbkref_ents; + int abkref_ents; + struct re_backref_cache_entry *bkref_ents; + int max_mb_elem_len; + int nsub_tops; + int asub_tops; + re_sub_match_top_t **sub_tops; +} re_match_context_t; + +typedef struct +{ + re_dfastate_t **sifted_states; + re_dfastate_t **limited_states; + int last_node; + int last_str_idx; + re_node_set limits; +} re_sift_context_t; + +struct re_fail_stack_ent_t +{ + int idx; + int node; + regmatch_t *regs; + re_node_set eps_via_nodes; +}; + +struct re_fail_stack_t +{ + int num; + int alloc; + struct re_fail_stack_ent_t *stack; +}; + +struct re_dfa_t +{ + re_token_t *nodes; + size_t nodes_alloc; + size_t nodes_len; + int *nexts; + int *org_indices; + re_node_set *edests; + re_node_set *eclosures; + re_node_set *inveclosures; + struct re_state_table_entry *state_table; + re_dfastate_t *init_state; + re_dfastate_t *init_state_word; + re_dfastate_t *init_state_nl; + re_dfastate_t *init_state_begbuf; + bin_tree_t *str_tree; + bin_tree_storage_t *str_tree_storage; + re_bitset_ptr_t sb_char; + int str_tree_storage_idx; + + /* number of subexpressions `re_nsub' is in regex_t. */ + unsigned int state_hash_mask; + int init_node; + int nbackref; /* The number of backreference in this dfa. */ + + /* Bitmap expressing which backreference is used. */ + bitset_word_t used_bkref_map; + bitset_word_t completed_bkref_map; + + unsigned int has_plural_match : 1; + /* If this dfa has "multibyte node", which is a backreference or + a node which can accept multibyte character or multi character + collating element. */ + unsigned int has_mb_node : 1; + unsigned int is_utf8 : 1; + unsigned int map_notascii : 1; + unsigned int word_ops_used : 1; + int mb_cur_max; + bitset_t word_char; + reg_syntax_t syntax; + int *subexp_map; +#ifdef DEBUG + char* re_str; +#endif +#if defined _LIBC + __libc_lock_define (, lock) +#endif +}; + +#define re_node_set_init_empty(set) memset (set, '\0', sizeof (re_node_set)) +#define re_node_set_remove(set,id) \ + (re_node_set_remove_at (set, re_node_set_contains (set, id) - 1)) +#define re_node_set_empty(p) ((p)->nelem = 0) +#define re_node_set_free(set) re_free ((set)->elems) + + +typedef enum +{ + SB_CHAR, + MB_CHAR, + EQUIV_CLASS, + COLL_SYM, + CHAR_CLASS +} bracket_elem_type; + +typedef struct +{ + bracket_elem_type type; + union + { + unsigned char ch; + unsigned char *name; + wchar_t wch; + } opr; +} bracket_elem_t; + + +/* Inline functions for bitset operation. */ +static void __attribute__ ((unused)) +bitset_not (bitset_t set) +{ + int bitset_i; + for (bitset_i = 0; bitset_i < BITSET_WORDS; ++bitset_i) + set[bitset_i] = ~set[bitset_i]; +} + +static void __attribute__ ((unused)) +bitset_merge (bitset_t dest, const bitset_t src) +{ + int bitset_i; + for (bitset_i = 0; bitset_i < BITSET_WORDS; ++bitset_i) + dest[bitset_i] |= src[bitset_i]; +} + +static void __attribute__ ((unused)) +bitset_mask (bitset_t dest, const bitset_t src) +{ + int bitset_i; + for (bitset_i = 0; bitset_i < BITSET_WORDS; ++bitset_i) + dest[bitset_i] &= src[bitset_i]; +} + +#ifdef RE_ENABLE_I18N +/* Inline functions for re_string. */ +static int +internal_function __attribute__ ((pure, unused)) +re_string_char_size_at (const re_string_t *pstr, int idx) +{ + int byte_idx; + if (pstr->mb_cur_max == 1) + return 1; + for (byte_idx = 1; idx + byte_idx < pstr->valid_len; ++byte_idx) + if (pstr->wcs[idx + byte_idx] != WEOF) + break; + return byte_idx; +} + +static wint_t +internal_function __attribute__ ((pure, unused)) +re_string_wchar_at (const re_string_t *pstr, int idx) +{ + if (pstr->mb_cur_max == 1) + return (wint_t) pstr->mbs[idx]; + return (wint_t) pstr->wcs[idx]; +} + +# ifndef NOT_IN_libc +# ifdef _LIBC +# include <locale/weight.h> +# endif + +static int +internal_function __attribute__ ((pure, unused)) +re_string_elem_size_at (const re_string_t *pstr, int idx) +{ +# ifdef _LIBC + const unsigned char *p, *extra; + const int32_t *table, *indirect; + uint_fast32_t nrules = _NL_CURRENT_WORD (LC_COLLATE, _NL_COLLATE_NRULES); + + if (nrules != 0) + { + table = (const int32_t *) _NL_CURRENT (LC_COLLATE, _NL_COLLATE_TABLEMB); + extra = (const unsigned char *) + _NL_CURRENT (LC_COLLATE, _NL_COLLATE_EXTRAMB); + indirect = (const int32_t *) _NL_CURRENT (LC_COLLATE, + _NL_COLLATE_INDIRECTMB); + p = pstr->mbs + idx; + findidx (table, indirect, extra, &p, pstr->len - idx); + return p - pstr->mbs - idx; + } + else +# endif /* _LIBC */ + return 1; +} +# endif +#endif /* RE_ENABLE_I18N */ + +#endif /* _REGEX_INTERNAL_H */ diff --git a/support/regexec.c b/support/regexec.c new file mode 100644 index 00000000..c8f11e52 --- /dev/null +++ b/support/regexec.c @@ -0,0 +1,4370 @@ +/* Extended regular expression matching and search library. + Copyright (C) 2002-2016 Free Software Foundation, Inc. + This file is part of the GNU C Library. + Contributed by Isamu Hasegawa <isamu@yamato.ibm.com>. + + The GNU C Library is free software; you can redistribute it and/or + modify it under the terms of the GNU Lesser General Public + License as published by the Free Software Foundation; either + version 2.1 of the License, or (at your option) any later version. + + The GNU C Library is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public + License along with the GNU C Library; if not, see + <http://www.gnu.org/licenses/>. */ + +#ifdef HAVE_STDINT_H +#include <stdint.h> +#endif /* HAVE_STDINT_H */ + +static reg_errcode_t match_ctx_init (re_match_context_t *cache, int eflags, + int n) internal_function; +static void match_ctx_clean (re_match_context_t *mctx) internal_function; +static void match_ctx_free (re_match_context_t *cache) internal_function; +static reg_errcode_t match_ctx_add_entry (re_match_context_t *cache, int node, + int str_idx, int from, int to) + internal_function; +static int search_cur_bkref_entry (const re_match_context_t *mctx, int str_idx) + internal_function; +static reg_errcode_t match_ctx_add_subtop (re_match_context_t *mctx, int node, + int str_idx) internal_function; +static re_sub_match_last_t * match_ctx_add_sublast (re_sub_match_top_t *subtop, + int node, int str_idx) + internal_function; +static void sift_ctx_init (re_sift_context_t *sctx, re_dfastate_t **sifted_sts, + re_dfastate_t **limited_sts, int last_node, + int last_str_idx) + internal_function; +static reg_errcode_t re_search_internal (const regex_t *preg, + const char *string, int length, + int start, int range, int stop, + size_t nmatch, regmatch_t pmatch[], + int eflags) internal_function; +static int re_search_2_stub (struct re_pattern_buffer *bufp, + const char *string1, int length1, + const char *string2, int length2, + int start, int range, struct re_registers *regs, + int stop, int ret_len) internal_function; +static int re_search_stub (struct re_pattern_buffer *bufp, + const char *string, int length, int start, + int range, int stop, struct re_registers *regs, + int ret_len) internal_function; +static unsigned re_copy_regs (struct re_registers *regs, regmatch_t *pmatch, + int nregs, int regs_allocated) internal_function; +static reg_errcode_t prune_impossible_nodes (re_match_context_t *mctx); +static int check_matching (re_match_context_t *mctx, int fl_longest_match, + int *p_match_first) internal_function; +static int check_halt_state_context (const re_match_context_t *mctx, + const re_dfastate_t *state, int idx) + internal_function; +static void update_regs (const re_dfa_t *dfa, regmatch_t *pmatch, + regmatch_t *prev_idx_match, int cur_node, + int cur_idx, int nmatch) internal_function; +static reg_errcode_t push_fail_stack (struct re_fail_stack_t *fs, + int str_idx, int dest_node, int nregs, + regmatch_t *regs, + re_node_set *eps_via_nodes) + internal_function; +static reg_errcode_t set_regs (const regex_t *preg, + const re_match_context_t *mctx, + size_t nmatch, regmatch_t *pmatch, + int fl_backtrack) internal_function; +static reg_errcode_t free_fail_stack_return (struct re_fail_stack_t *fs) + internal_function; + +#ifdef RE_ENABLE_I18N +static int sift_states_iter_mb (const re_match_context_t *mctx, + re_sift_context_t *sctx, + int node_idx, int str_idx, int max_str_idx) + internal_function; +#endif /* RE_ENABLE_I18N */ +static reg_errcode_t sift_states_backward (const re_match_context_t *mctx, + re_sift_context_t *sctx) + internal_function; +static reg_errcode_t build_sifted_states (const re_match_context_t *mctx, + re_sift_context_t *sctx, int str_idx, + re_node_set *cur_dest) + internal_function; +static reg_errcode_t update_cur_sifted_state (const re_match_context_t *mctx, + re_sift_context_t *sctx, + int str_idx, + re_node_set *dest_nodes) + internal_function; +static reg_errcode_t add_epsilon_src_nodes (const re_dfa_t *dfa, + re_node_set *dest_nodes, + const re_node_set *candidates) + internal_function; +static int check_dst_limits (const re_match_context_t *mctx, + re_node_set *limits, + int dst_node, int dst_idx, int src_node, + int src_idx) internal_function; +static int check_dst_limits_calc_pos_1 (const re_match_context_t *mctx, + int boundaries, int subexp_idx, + int from_node, int bkref_idx) + internal_function; +static int check_dst_limits_calc_pos (const re_match_context_t *mctx, + int limit, int subexp_idx, + int node, int str_idx, + int bkref_idx) internal_function; +static reg_errcode_t check_subexp_limits (const re_dfa_t *dfa, + re_node_set *dest_nodes, + const re_node_set *candidates, + re_node_set *limits, + struct re_backref_cache_entry *bkref_ents, + int str_idx) internal_function; +static reg_errcode_t sift_states_bkref (const re_match_context_t *mctx, + re_sift_context_t *sctx, + int str_idx, const re_node_set *candidates) + internal_function; +static reg_errcode_t merge_state_array (const re_dfa_t *dfa, + re_dfastate_t **dst, + re_dfastate_t **src, int num) + internal_function; +static re_dfastate_t *find_recover_state (reg_errcode_t *err, + re_match_context_t *mctx) internal_function; +static re_dfastate_t *transit_state (reg_errcode_t *err, + re_match_context_t *mctx, + re_dfastate_t *state) internal_function; +static re_dfastate_t *merge_state_with_log (reg_errcode_t *err, + re_match_context_t *mctx, + re_dfastate_t *next_state) + internal_function; +static reg_errcode_t check_subexp_matching_top (re_match_context_t *mctx, + re_node_set *cur_nodes, + int str_idx) internal_function; +#if 0 +static re_dfastate_t *transit_state_sb (reg_errcode_t *err, + re_match_context_t *mctx, + re_dfastate_t *pstate) + internal_function; +#endif +#ifdef RE_ENABLE_I18N +static reg_errcode_t transit_state_mb (re_match_context_t *mctx, + re_dfastate_t *pstate) + internal_function; +#endif /* RE_ENABLE_I18N */ +static reg_errcode_t transit_state_bkref (re_match_context_t *mctx, + const re_node_set *nodes) + internal_function; +static reg_errcode_t get_subexp (re_match_context_t *mctx, + int bkref_node, int bkref_str_idx) + internal_function; +static reg_errcode_t get_subexp_sub (re_match_context_t *mctx, + const re_sub_match_top_t *sub_top, + re_sub_match_last_t *sub_last, + int bkref_node, int bkref_str) + internal_function; +static int find_subexp_node (const re_dfa_t *dfa, const re_node_set *nodes, + int subexp_idx, int type) internal_function; +static reg_errcode_t check_arrival (re_match_context_t *mctx, + state_array_t *path, int top_node, + int top_str, int last_node, int last_str, + int type) internal_function; +static reg_errcode_t check_arrival_add_next_nodes (re_match_context_t *mctx, + int str_idx, + re_node_set *cur_nodes, + re_node_set *next_nodes) + internal_function; +static reg_errcode_t check_arrival_expand_ecl (const re_dfa_t *dfa, + re_node_set *cur_nodes, + int ex_subexp, int type) + internal_function; +static reg_errcode_t check_arrival_expand_ecl_sub (const re_dfa_t *dfa, + re_node_set *dst_nodes, + int target, int ex_subexp, + int type) internal_function; +static reg_errcode_t expand_bkref_cache (re_match_context_t *mctx, + re_node_set *cur_nodes, int cur_str, + int subexp_num, int type) + internal_function; +static int build_trtable (const re_dfa_t *dfa, + re_dfastate_t *state) internal_function; +#ifdef RE_ENABLE_I18N +static int check_node_accept_bytes (const re_dfa_t *dfa, int node_idx, + const re_string_t *input, int idx) + internal_function; +# ifdef _LIBC +static unsigned int find_collation_sequence_value (const unsigned char *mbs, + size_t name_len) + internal_function; +# endif /* _LIBC */ +#endif /* RE_ENABLE_I18N */ +static int group_nodes_into_DFAstates (const re_dfa_t *dfa, + const re_dfastate_t *state, + re_node_set *states_node, + bitset_t *states_ch) internal_function; +static int check_node_accept (const re_match_context_t *mctx, + const re_token_t *node, int idx) + internal_function; +static reg_errcode_t extend_buffers (re_match_context_t *mctx, int min_len) + internal_function; + +#ifdef GAWK +#undef MIN /* safety */ +static int +MIN(size_t a, size_t b) +{ + return (a < b ? a : b); +} +#endif + +/* Entry point for POSIX code. */ + +/* regexec searches for a given pattern, specified by PREG, in the + string STRING. + + If NMATCH is zero or REG_NOSUB was set in the cflags argument to + `regcomp', we ignore PMATCH. Otherwise, we assume PMATCH has at + least NMATCH elements, and we set them to the offsets of the + corresponding matched substrings. + + EFLAGS specifies `execution flags' which affect matching: if + REG_NOTBOL is set, then ^ does not match at the beginning of the + string; if REG_NOTEOL is set, then $ does not match at the end. + + We return 0 if we find a match and REG_NOMATCH if not. */ + +int +regexec (const regex_t *__restrict preg, const char *__restrict string, + size_t nmatch, regmatch_t pmatch[], int eflags) +{ + reg_errcode_t err; + int start, length; + + if (eflags & ~(REG_NOTBOL | REG_NOTEOL | REG_STARTEND)) + return REG_BADPAT; + + if (eflags & REG_STARTEND) + { + start = pmatch[0].rm_so; + length = pmatch[0].rm_eo; + } + else + { + start = 0; + length = strlen (string); + } + + __libc_lock_lock (dfa->lock); + if (preg->no_sub) + err = re_search_internal (preg, string, length, start, length - start, + length, 0, NULL, eflags); + else + err = re_search_internal (preg, string, length, start, length - start, + length, nmatch, pmatch, eflags); + __libc_lock_unlock (dfa->lock); + return err != REG_NOERROR; +} + +#ifdef _LIBC +# include <shlib-compat.h> +versioned_symbol (libc, __regexec, regexec, GLIBC_2_3_4); + +# if SHLIB_COMPAT (libc, GLIBC_2_0, GLIBC_2_3_4) +__typeof__ (__regexec) __compat_regexec; + +int +attribute_compat_text_section +__compat_regexec (const regex_t *__restrict preg, + const char *__restrict string, size_t nmatch, + regmatch_t pmatch[], int eflags) +{ + return regexec (preg, string, nmatch, pmatch, + eflags & (REG_NOTBOL | REG_NOTEOL)); +} +compat_symbol (libc, __compat_regexec, regexec, GLIBC_2_0); +# endif +#endif + +/* Entry points for GNU code. */ + +/* re_match, re_search, re_match_2, re_search_2 + + The former two functions operate on STRING with length LENGTH, + while the later two operate on concatenation of STRING1 and STRING2 + with lengths LENGTH1 and LENGTH2, respectively. + + re_match() matches the compiled pattern in BUFP against the string, + starting at index START. + + re_search() first tries matching at index START, then it tries to match + starting from index START + 1, and so on. The last start position tried + is START + RANGE. (Thus RANGE = 0 forces re_search to operate the same + way as re_match().) + + The parameter STOP of re_{match,search}_2 specifies that no match exceeding + the first STOP characters of the concatenation of the strings should be + concerned. + + If REGS is not NULL, and BUFP->no_sub is not set, the offsets of the match + and all groups is stored in REGS. (For the "_2" variants, the offsets are + computed relative to the concatenation, not relative to the individual + strings.) + + On success, re_match* functions return the length of the match, re_search* + return the position of the start of the match. Return value -1 means no + match was found and -2 indicates an internal error. */ + +int +re_match (struct re_pattern_buffer *bufp, const char *string, int length, + int start, struct re_registers *regs) +{ + return re_search_stub (bufp, string, length, start, 0, length, regs, 1); +} +#ifdef _LIBC +weak_alias (__re_match, re_match) +#endif + +int +re_search (struct re_pattern_buffer *bufp, const char *string, int length, + int start, int range, struct re_registers *regs) +{ + return re_search_stub (bufp, string, length, start, range, length, regs, 0); +} +#ifdef _LIBC +weak_alias (__re_search, re_search) +#endif + +int +re_match_2 (struct re_pattern_buffer *bufp, const char *string1, int length1, + const char *string2, int length2, int start, + struct re_registers *regs, int stop) +{ + return re_search_2_stub (bufp, string1, length1, string2, length2, + start, 0, regs, stop, 1); +} +#ifdef _LIBC +weak_alias (__re_match_2, re_match_2) +#endif + +int +re_search_2 (struct re_pattern_buffer *bufp, const char *string1, int length1, + const char *string2, int length2, int start, int range, + struct re_registers *regs, int stop) +{ + return re_search_2_stub (bufp, string1, length1, string2, length2, + start, range, regs, stop, 0); +} +#ifdef _LIBC +weak_alias (__re_search_2, re_search_2) +#endif + +static int +internal_function +re_search_2_stub (struct re_pattern_buffer *bufp, const char *string1, + int length1, const char *string2, int length2, int start, + int range, struct re_registers *regs, + int stop, int ret_len) +{ + const char *str; + int rval; + int len; + char *s = NULL; + + if (BE ((length1 < 0 || length2 < 0 || stop < 0 + || INT_ADD_WRAPV (length1, length2, &len)), + 0)) + return -2; + + /* Concatenate the strings. */ + if (length2 > 0) + if (length1 > 0) + { + s = re_malloc (char, len); + + if (BE (s == NULL, 0)) + return -2; +#ifdef _LIBC + memcpy (__mempcpy (s, string1, length1), string2, length2); +#else + memcpy (s, string1, length1); + memcpy (s + length1, string2, length2); +#endif + str = s; + } + else + str = string2; + else + str = string1; + + rval = re_search_stub (bufp, str, len, start, range, stop, regs, ret_len); + re_free (s); + return rval; +} + +/* The parameters have the same meaning as those of re_search. + Additional parameters: + If RET_LEN is nonzero the length of the match is returned (re_match style); + otherwise the position of the match is returned. */ + +static int +internal_function +re_search_stub (struct re_pattern_buffer *bufp, const char *string, int length, + int start, int range, int stop, struct re_registers *regs, + int ret_len) +{ + reg_errcode_t result; + regmatch_t *pmatch; + int nregs, rval; + int eflags = 0; + + /* Check for out-of-range. */ + if (BE (start < 0 || start > length, 0)) + return -1; + if (BE (start + range > length, 0)) + range = length - start; + else if (BE (start + range < 0, 0)) + range = -start; + + __libc_lock_lock (dfa->lock); + + eflags |= (bufp->not_bol) ? REG_NOTBOL : 0; + eflags |= (bufp->not_eol) ? REG_NOTEOL : 0; + + /* Compile fastmap if we haven't yet. */ + if (range > 0 && bufp->fastmap != NULL && !bufp->fastmap_accurate) + re_compile_fastmap (bufp); + + if (BE (bufp->no_sub, 0)) + regs = NULL; + + /* We need at least 1 register. */ + if (regs == NULL) + nregs = 1; + else if (BE (bufp->regs_allocated == REGS_FIXED && + regs->num_regs < bufp->re_nsub + 1, 0)) + { + nregs = regs->num_regs; + if (BE (nregs < 1, 0)) + { + /* Nothing can be copied to regs. */ + regs = NULL; + nregs = 1; + } + } + else + nregs = bufp->re_nsub + 1; + pmatch = re_malloc (regmatch_t, nregs); + if (BE (pmatch == NULL, 0)) + { + rval = -2; + goto out; + } + + result = re_search_internal (bufp, string, length, start, range, stop, + nregs, pmatch, eflags); + + rval = 0; + + /* I hope we needn't fill ther regs with -1's when no match was found. */ + if (result != REG_NOERROR) + rval = -1; + else if (regs != NULL) + { + /* If caller wants register contents data back, copy them. */ + bufp->regs_allocated = re_copy_regs (regs, pmatch, nregs, + bufp->regs_allocated); + if (BE (bufp->regs_allocated == REGS_UNALLOCATED, 0)) + rval = -2; + } + + if (BE (rval == 0, 1)) + { + if (ret_len) + { + assert (pmatch[0].rm_so == start); + rval = pmatch[0].rm_eo - start; + } + else + rval = pmatch[0].rm_so; + } + re_free (pmatch); + out: + __libc_lock_unlock (dfa->lock); + return rval; +} + +static unsigned +internal_function +re_copy_regs (struct re_registers *regs, regmatch_t *pmatch, int nregs, + int regs_allocated) +{ + int rval = REGS_REALLOCATE; + int i; + int need_regs = nregs + 1; + /* We need one extra element beyond `num_regs' for the `-1' marker GNU code + uses. */ + + /* Have the register data arrays been allocated? */ + if (regs_allocated == REGS_UNALLOCATED) + { /* No. So allocate them with malloc. */ + regs->start = re_malloc (regoff_t, need_regs); + if (BE (regs->start == NULL, 0)) + return REGS_UNALLOCATED; + regs->end = re_malloc (regoff_t, need_regs); + if (BE (regs->end == NULL, 0)) + { + re_free (regs->start); + return REGS_UNALLOCATED; + } + regs->num_regs = need_regs; + } + else if (regs_allocated == REGS_REALLOCATE) + { /* Yes. If we need more elements than were already + allocated, reallocate them. If we need fewer, just + leave it alone. */ + if (BE (need_regs > regs->num_regs, 0)) + { + regoff_t *new_start = re_realloc (regs->start, regoff_t, need_regs); + regoff_t *new_end; + if (BE (new_start == NULL, 0)) + return REGS_UNALLOCATED; + new_end = re_realloc (regs->end, regoff_t, need_regs); + if (BE (new_end == NULL, 0)) + { + re_free (new_start); + return REGS_UNALLOCATED; + } + regs->start = new_start; + regs->end = new_end; + regs->num_regs = need_regs; + } + } + else + { + assert (regs_allocated == REGS_FIXED); + /* This function may not be called with REGS_FIXED and nregs too big. */ + assert (regs->num_regs >= nregs); + rval = REGS_FIXED; + } + + /* Copy the regs. */ + for (i = 0; i < nregs; ++i) + { + regs->start[i] = pmatch[i].rm_so; + regs->end[i] = pmatch[i].rm_eo; + } + for ( ; i < regs->num_regs; ++i) + regs->start[i] = regs->end[i] = -1; + + return rval; +} + +/* Set REGS to hold NUM_REGS registers, storing them in STARTS and + ENDS. Subsequent matches using PATTERN_BUFFER and REGS will use + this memory for recording register information. STARTS and ENDS + must be allocated using the malloc library routine, and must each + be at least NUM_REGS * sizeof (regoff_t) bytes long. + + If NUM_REGS == 0, then subsequent matches should allocate their own + register data. + + Unless this function is called, the first search or match using + PATTERN_BUFFER will allocate its own register data, without + freeing the old data. */ + +void +re_set_registers (struct re_pattern_buffer *bufp, struct re_registers *regs, + unsigned num_regs, regoff_t *starts, regoff_t *ends) +{ + if (num_regs) + { + bufp->regs_allocated = REGS_REALLOCATE; + regs->num_regs = num_regs; + regs->start = starts; + regs->end = ends; + } + else + { + bufp->regs_allocated = REGS_UNALLOCATED; + regs->num_regs = 0; + regs->start = regs->end = (regoff_t *) 0; + } +} +#ifdef _LIBC +weak_alias (__re_set_registers, re_set_registers) +#endif + +/* Entry points compatible with 4.2 BSD regex library. We don't define + them unless specifically requested. */ + +#if defined _REGEX_RE_COMP || defined _LIBC +int +# ifdef _LIBC +weak_function +# endif +re_exec (const char *s) +{ + return 0 == regexec (&re_comp_buf, s, 0, NULL, 0); +} +#endif /* _REGEX_RE_COMP */ + +/* Internal entry point. */ + +/* Searches for a compiled pattern PREG in the string STRING, whose + length is LENGTH. NMATCH, PMATCH, and EFLAGS have the same + meaning as with regexec. START, and RANGE have the same meanings + with re_search. + Return REG_NOERROR if we find a match, and REG_NOMATCH if not, + otherwise return the error code. + Note: We assume front end functions already check ranges. + (START + RANGE >= 0 && START + RANGE <= LENGTH) */ + +static reg_errcode_t +__attribute_warn_unused_result__ internal_function +re_search_internal (const regex_t *preg, const char *string, int length, + int start, int range, int stop, size_t nmatch, + regmatch_t pmatch[], int eflags) +{ + reg_errcode_t err; + const re_dfa_t *dfa = (const re_dfa_t *) preg->buffer; + int left_lim, right_lim, incr; + int fl_longest_match, match_first, match_kind, match_last = -1; + int extra_nmatch; + int sb, ch; +#if defined _LIBC || (defined __STDC_VERSION__ && __STDC_VERSION__ >= 199901L) + re_match_context_t mctx = { .dfa = dfa }; +#else + re_match_context_t mctx; +#endif + char *fastmap = (preg->fastmap != NULL && preg->fastmap_accurate + && range && !preg->can_be_null) ? preg->fastmap : NULL; + RE_TRANSLATE_TYPE t = preg->translate; + +#if !(defined _LIBC || (defined __STDC_VERSION__ && __STDC_VERSION__ >= 199901L)) + memset (&mctx, '\0', sizeof (re_match_context_t)); + mctx.dfa = dfa; +#endif + + extra_nmatch = (nmatch > preg->re_nsub) ? nmatch - (preg->re_nsub + 1) : 0; + nmatch -= extra_nmatch; + + /* Check if the DFA haven't been compiled. */ + if (BE (preg->used == 0 || dfa == NULL || dfa->init_state == NULL + || dfa->init_state_word == NULL || dfa->init_state_nl == NULL + || dfa->init_state_begbuf == NULL, 0)) + return REG_NOMATCH; + +#ifdef DEBUG + /* We assume front-end functions already check them. */ + assert (start + range >= 0 && start + range <= length); +#endif + + /* If initial states with non-begbuf contexts have no elements, + the regex must be anchored. If preg->newline_anchor is set, + we'll never use init_state_nl, so do not check it. */ + if (dfa->init_state->nodes.nelem == 0 + && dfa->init_state_word->nodes.nelem == 0 + && (dfa->init_state_nl->nodes.nelem == 0 + || !preg->newline_anchor)) + { + if (start != 0 && start + range != 0) + return REG_NOMATCH; + start = range = 0; + } + + /* We must check the longest matching, if nmatch > 0. */ + fl_longest_match = (nmatch != 0 || dfa->nbackref); + + err = re_string_allocate (&mctx.input, string, length, dfa->nodes_len + 1, + preg->translate, (preg->syntax & RE_ICASE) != 0, + dfa); + if (BE (err != REG_NOERROR, 0)) + goto free_return; + mctx.input.stop = stop; + mctx.input.raw_stop = stop; + mctx.input.newline_anchor = preg->newline_anchor; + + err = match_ctx_init (&mctx, eflags, dfa->nbackref * 2); + if (BE (err != REG_NOERROR, 0)) + goto free_return; + + /* We will log all the DFA states through which the dfa pass, + if nmatch > 1, or this dfa has "multibyte node", which is a + back-reference or a node which can accept multibyte character or + multi character collating element. */ + if (nmatch > 1 || dfa->has_mb_node) + { + /* Avoid overflow. */ + if (BE (SIZE_MAX / sizeof (re_dfastate_t *) <= mctx.input.bufs_len, 0)) + { + err = REG_ESPACE; + goto free_return; + } + + mctx.state_log = re_malloc (re_dfastate_t *, mctx.input.bufs_len + 1); + if (BE (mctx.state_log == NULL, 0)) + { + err = REG_ESPACE; + goto free_return; + } + } + else + mctx.state_log = NULL; + + match_first = start; + mctx.input.tip_context = (eflags & REG_NOTBOL) ? CONTEXT_BEGBUF + : CONTEXT_NEWLINE | CONTEXT_BEGBUF; + + /* Check incrementally whether of not the input string match. */ + incr = (range < 0) ? -1 : 1; + left_lim = (range < 0) ? start + range : start; + right_lim = (range < 0) ? start : start + range; + sb = dfa->mb_cur_max == 1; + match_kind = + (fastmap + ? ((sb || !(preg->syntax & RE_ICASE || t) ? 4 : 0) + | (range >= 0 ? 2 : 0) + | (t != NULL ? 1 : 0)) + : 8); + + for (;; match_first += incr) + { + err = REG_NOMATCH; + if (match_first < left_lim || right_lim < match_first) + goto free_return; + + /* Advance as rapidly as possible through the string, until we + find a plausible place to start matching. This may be done + with varying efficiency, so there are various possibilities: + only the most common of them are specialized, in order to + save on code size. We use a switch statement for speed. */ + switch (match_kind) + { + case 8: + /* No fastmap. */ + break; + + case 7: + /* Fastmap with single-byte translation, match forward. */ + while (BE (match_first < right_lim, 1) + && !fastmap[t[(unsigned char) string[match_first]]]) + ++match_first; + goto forward_match_found_start_or_reached_end; + + case 6: + /* Fastmap without translation, match forward. */ + while (BE (match_first < right_lim, 1) + && !fastmap[(unsigned char) string[match_first]]) + ++match_first; + + forward_match_found_start_or_reached_end: + if (BE (match_first == right_lim, 0)) + { + ch = match_first >= length + ? 0 : (unsigned char) string[match_first]; + if (!fastmap[t ? t[ch] : ch]) + goto free_return; + } + break; + + case 4: + case 5: + /* Fastmap without multi-byte translation, match backwards. */ + while (match_first >= left_lim) + { + ch = match_first >= length + ? 0 : (unsigned char) string[match_first]; + if (fastmap[t ? t[ch] : ch]) + break; + --match_first; + } + if (match_first < left_lim) + goto free_return; + break; + + default: + /* In this case, we can't determine easily the current byte, + since it might be a component byte of a multibyte + character. Then we use the constructed buffer instead. */ + for (;;) + { + /* If MATCH_FIRST is out of the valid range, reconstruct the + buffers. */ + unsigned int offset = match_first - mctx.input.raw_mbs_idx; + if (BE (offset >= (unsigned int) mctx.input.valid_raw_len, 0)) + { + err = re_string_reconstruct (&mctx.input, match_first, + eflags); + if (BE (err != REG_NOERROR, 0)) + goto free_return; + + offset = match_first - mctx.input.raw_mbs_idx; + } + /* If MATCH_FIRST is out of the buffer, leave it as '\0'. + Note that MATCH_FIRST must not be smaller than 0. */ + ch = (match_first >= length + ? 0 : re_string_byte_at (&mctx.input, offset)); + if (fastmap[ch]) + break; + match_first += incr; + if (match_first < left_lim || match_first > right_lim) + { + err = REG_NOMATCH; + goto free_return; + } + } + break; + } + + /* Reconstruct the buffers so that the matcher can assume that + the matching starts from the beginning of the buffer. */ + err = re_string_reconstruct (&mctx.input, match_first, eflags); + if (BE (err != REG_NOERROR, 0)) + goto free_return; + +#ifdef RE_ENABLE_I18N + /* Don't consider this char as a possible match start if it part, + yet isn't the head, of a multibyte character. */ + if (!sb && !re_string_first_byte (&mctx.input, 0)) + continue; +#endif + + /* It seems to be appropriate one, then use the matcher. */ + /* We assume that the matching starts from 0. */ + mctx.state_log_top = mctx.nbkref_ents = mctx.max_mb_elem_len = 0; + match_last = check_matching (&mctx, fl_longest_match, + range >= 0 ? &match_first : NULL); + if (match_last != -1) + { + if (BE (match_last == -2, 0)) + { + err = REG_ESPACE; + goto free_return; + } + else + { + mctx.match_last = match_last; + if ((!preg->no_sub && nmatch > 1) || dfa->nbackref) + { + re_dfastate_t *pstate = mctx.state_log[match_last]; + mctx.last_node = check_halt_state_context (&mctx, pstate, + match_last); + } + if ((!preg->no_sub && nmatch > 1 && dfa->has_plural_match) + || dfa->nbackref) + { + err = prune_impossible_nodes (&mctx); + if (err == REG_NOERROR) + break; + if (BE (err != REG_NOMATCH, 0)) + goto free_return; + match_last = -1; + } + else + break; /* We found a match. */ + } + } + + match_ctx_clean (&mctx); + } + +#ifdef DEBUG + assert (match_last != -1); + assert (err == REG_NOERROR); +#endif + + /* Set pmatch[] if we need. */ + if (nmatch > 0) + { + int reg_idx; + + /* Initialize registers. */ + for (reg_idx = 1; reg_idx < nmatch; ++reg_idx) + pmatch[reg_idx].rm_so = pmatch[reg_idx].rm_eo = -1; + + /* Set the points where matching start/end. */ + pmatch[0].rm_so = 0; + pmatch[0].rm_eo = mctx.match_last; + + if (!preg->no_sub && nmatch > 1) + { + err = set_regs (preg, &mctx, nmatch, pmatch, + dfa->has_plural_match && dfa->nbackref > 0); + if (BE (err != REG_NOERROR, 0)) + goto free_return; + } + + /* At last, add the offset to each register, since we slid + the buffers so that we could assume that the matching starts + from 0. */ + for (reg_idx = 0; reg_idx < nmatch; ++reg_idx) + if (pmatch[reg_idx].rm_so != -1) + { +#ifdef RE_ENABLE_I18N + if (BE (mctx.input.offsets_needed != 0, 0)) + { + pmatch[reg_idx].rm_so = + (pmatch[reg_idx].rm_so == mctx.input.valid_len + ? mctx.input.valid_raw_len + : mctx.input.offsets[pmatch[reg_idx].rm_so]); + pmatch[reg_idx].rm_eo = + (pmatch[reg_idx].rm_eo == mctx.input.valid_len + ? mctx.input.valid_raw_len + : mctx.input.offsets[pmatch[reg_idx].rm_eo]); + } +#else + assert (mctx.input.offsets_needed == 0); +#endif + pmatch[reg_idx].rm_so += match_first; + pmatch[reg_idx].rm_eo += match_first; + } + for (reg_idx = 0; reg_idx < extra_nmatch; ++reg_idx) + { + pmatch[nmatch + reg_idx].rm_so = -1; + pmatch[nmatch + reg_idx].rm_eo = -1; + } + + if (dfa->subexp_map) + for (reg_idx = 0; reg_idx + 1 < nmatch; reg_idx++) + if (dfa->subexp_map[reg_idx] != reg_idx) + { + pmatch[reg_idx + 1].rm_so + = pmatch[dfa->subexp_map[reg_idx] + 1].rm_so; + pmatch[reg_idx + 1].rm_eo + = pmatch[dfa->subexp_map[reg_idx] + 1].rm_eo; + } + } + + free_return: + re_free (mctx.state_log); + if (dfa->nbackref) + match_ctx_free (&mctx); + re_string_destruct (&mctx.input); + return err; +} + +static reg_errcode_t +prune_impossible_nodes (re_match_context_t *mctx) +{ + const re_dfa_t *const dfa = mctx->dfa; + int halt_node, match_last; + reg_errcode_t ret; + re_dfastate_t **sifted_states; + re_dfastate_t **lim_states = NULL; + re_sift_context_t sctx; +#ifdef DEBUG + assert (mctx->state_log != NULL); +#endif + match_last = mctx->match_last; + halt_node = mctx->last_node; + + /* Avoid overflow. */ + if (BE (SIZE_MAX / sizeof (re_dfastate_t *) <= match_last, 0)) + return REG_ESPACE; + + sifted_states = re_malloc (re_dfastate_t *, match_last + 1); + if (BE (sifted_states == NULL, 0)) + { + ret = REG_ESPACE; + goto free_return; + } + if (dfa->nbackref) + { + lim_states = re_malloc (re_dfastate_t *, match_last + 1); + if (BE (lim_states == NULL, 0)) + { + ret = REG_ESPACE; + goto free_return; + } + while (1) + { + memset (lim_states, '\0', + sizeof (re_dfastate_t *) * (match_last + 1)); + sift_ctx_init (&sctx, sifted_states, lim_states, halt_node, + match_last); + ret = sift_states_backward (mctx, &sctx); + re_node_set_free (&sctx.limits); + if (BE (ret != REG_NOERROR, 0)) + goto free_return; + if (sifted_states[0] != NULL || lim_states[0] != NULL) + break; + do + { + --match_last; + if (match_last < 0) + { + ret = REG_NOMATCH; + goto free_return; + } + } while (mctx->state_log[match_last] == NULL + || !mctx->state_log[match_last]->halt); + halt_node = check_halt_state_context (mctx, + mctx->state_log[match_last], + match_last); + } + ret = merge_state_array (dfa, sifted_states, lim_states, + match_last + 1); + re_free (lim_states); + lim_states = NULL; + if (BE (ret != REG_NOERROR, 0)) + goto free_return; + } + else + { + sift_ctx_init (&sctx, sifted_states, lim_states, halt_node, match_last); + ret = sift_states_backward (mctx, &sctx); + re_node_set_free (&sctx.limits); + if (BE (ret != REG_NOERROR, 0)) + goto free_return; + if (sifted_states[0] == NULL) + { + ret = REG_NOMATCH; + goto free_return; + } + } + re_free (mctx->state_log); + mctx->state_log = sifted_states; + sifted_states = NULL; + mctx->last_node = halt_node; + mctx->match_last = match_last; + ret = REG_NOERROR; + free_return: + re_free (sifted_states); + re_free (lim_states); + return ret; +} + +/* Acquire an initial state and return it. + We must select appropriate initial state depending on the context, + since initial states may have constraints like "\<", "^", etc.. */ + +static inline re_dfastate_t * +__attribute__ ((always_inline)) internal_function +acquire_init_state_context (reg_errcode_t *err, const re_match_context_t *mctx, + int idx) +{ + const re_dfa_t *const dfa = mctx->dfa; + if (dfa->init_state->has_constraint) + { + unsigned int context; + context = re_string_context_at (&mctx->input, idx - 1, mctx->eflags); + if (IS_WORD_CONTEXT (context)) + return dfa->init_state_word; + else if (IS_ORDINARY_CONTEXT (context)) + return dfa->init_state; + else if (IS_BEGBUF_CONTEXT (context) && IS_NEWLINE_CONTEXT (context)) + return dfa->init_state_begbuf; + else if (IS_NEWLINE_CONTEXT (context)) + return dfa->init_state_nl; + else if (IS_BEGBUF_CONTEXT (context)) + { + /* It is relatively rare case, then calculate on demand. */ + return re_acquire_state_context (err, dfa, + dfa->init_state->entrance_nodes, + context); + } + else + /* Must not happen? */ + return dfa->init_state; + } + else + return dfa->init_state; +} + +/* Check whether the regular expression match input string INPUT or not, + and return the index where the matching end, return -1 if not match, + or return -2 in case of an error. + FL_LONGEST_MATCH means we want the POSIX longest matching. + If P_MATCH_FIRST is not NULL, and the match fails, it is set to the + next place where we may want to try matching. + Note that the matcher assume that the maching starts from the current + index of the buffer. */ + +static int +internal_function __attribute_warn_unused_result__ +check_matching (re_match_context_t *mctx, int fl_longest_match, + int *p_match_first) +{ + const re_dfa_t *const dfa = mctx->dfa; + reg_errcode_t err; + int match = 0; + int match_last = -1; + int cur_str_idx = re_string_cur_idx (&mctx->input); + re_dfastate_t *cur_state; + int at_init_state = p_match_first != NULL; + int next_start_idx = cur_str_idx; + + err = REG_NOERROR; + cur_state = acquire_init_state_context (&err, mctx, cur_str_idx); + /* An initial state must not be NULL (invalid). */ + if (BE (cur_state == NULL, 0)) + { + assert (err == REG_ESPACE); + return -2; + } + + if (mctx->state_log != NULL) + { + mctx->state_log[cur_str_idx] = cur_state; + + /* Check OP_OPEN_SUBEXP in the initial state in case that we use them + later. E.g. Processing back references. */ + if (BE (dfa->nbackref, 0)) + { + at_init_state = 0; + err = check_subexp_matching_top (mctx, &cur_state->nodes, 0); + if (BE (err != REG_NOERROR, 0)) + return err; + + if (cur_state->has_backref) + { + err = transit_state_bkref (mctx, &cur_state->nodes); + if (BE (err != REG_NOERROR, 0)) + return err; + } + } + } + + /* If the RE accepts NULL string. */ + if (BE (cur_state->halt, 0)) + { + if (!cur_state->has_constraint + || check_halt_state_context (mctx, cur_state, cur_str_idx)) + { + if (!fl_longest_match) + return cur_str_idx; + else + { + match_last = cur_str_idx; + match = 1; + } + } + } + + while (!re_string_eoi (&mctx->input)) + { + re_dfastate_t *old_state = cur_state; + int next_char_idx = re_string_cur_idx (&mctx->input) + 1; + + if ((BE (next_char_idx >= mctx->input.bufs_len, 0) + && mctx->input.bufs_len < mctx->input.len) + || (BE (next_char_idx >= mctx->input.valid_len, 0) + && mctx->input.valid_len < mctx->input.len)) + { + err = extend_buffers (mctx, next_char_idx + 1); + if (BE (err != REG_NOERROR, 0)) + { + assert (err == REG_ESPACE); + return -2; + } + } + + cur_state = transit_state (&err, mctx, cur_state); + if (mctx->state_log != NULL) + cur_state = merge_state_with_log (&err, mctx, cur_state); + + if (cur_state == NULL) + { + /* Reached the invalid state or an error. Try to recover a valid + state using the state log, if available and if we have not + already found a valid (even if not the longest) match. */ + if (BE (err != REG_NOERROR, 0)) + return -2; + + if (mctx->state_log == NULL + || (match && !fl_longest_match) + || (cur_state = find_recover_state (&err, mctx)) == NULL) + break; + } + + if (BE (at_init_state, 0)) + { + if (old_state == cur_state) + next_start_idx = next_char_idx; + else + at_init_state = 0; + } + + if (cur_state->halt) + { + /* Reached a halt state. + Check the halt state can satisfy the current context. */ + if (!cur_state->has_constraint + || check_halt_state_context (mctx, cur_state, + re_string_cur_idx (&mctx->input))) + { + /* We found an appropriate halt state. */ + match_last = re_string_cur_idx (&mctx->input); + match = 1; + + /* We found a match, do not modify match_first below. */ + p_match_first = NULL; + if (!fl_longest_match) + break; + } + } + } + + if (p_match_first) + *p_match_first += next_start_idx; + + return match_last; +} + +/* Check NODE match the current context. */ + +static int +internal_function +check_halt_node_context (const re_dfa_t *dfa, int node, unsigned int context) +{ + re_token_type_t type = dfa->nodes[node].type; + unsigned int constraint = dfa->nodes[node].constraint; + if (type != END_OF_RE) + return 0; + if (!constraint) + return 1; + if (NOT_SATISFY_NEXT_CONSTRAINT (constraint, context)) + return 0; + return 1; +} + +/* Check the halt state STATE match the current context. + Return 0 if not match, if the node, STATE has, is a halt node and + match the context, return the node. */ + +static int +internal_function +check_halt_state_context (const re_match_context_t *mctx, + const re_dfastate_t *state, int idx) +{ + int i; + unsigned int context; +#ifdef DEBUG + assert (state->halt); +#endif + context = re_string_context_at (&mctx->input, idx, mctx->eflags); + for (i = 0; i < state->nodes.nelem; ++i) + if (check_halt_node_context (mctx->dfa, state->nodes.elems[i], context)) + return state->nodes.elems[i]; + return 0; +} + +/* Compute the next node to which "NFA" transit from NODE("NFA" is a NFA + corresponding to the DFA). + Return the destination node, and update EPS_VIA_NODES, return -1 in case + of errors. */ + +static int +internal_function +proceed_next_node (const re_match_context_t *mctx, int nregs, regmatch_t *regs, + int *pidx, int node, re_node_set *eps_via_nodes, + struct re_fail_stack_t *fs) +{ + const re_dfa_t *const dfa = mctx->dfa; + int i, err; + if (IS_EPSILON_NODE (dfa->nodes[node].type)) + { + re_node_set *cur_nodes = &mctx->state_log[*pidx]->nodes; + re_node_set *edests = &dfa->edests[node]; + int dest_node; + err = re_node_set_insert (eps_via_nodes, node); + if (BE (err < 0, 0)) + return -2; + /* Pick up a valid destination, or return -1 if none is found. */ + for (dest_node = -1, i = 0; i < edests->nelem; ++i) + { + int candidate = edests->elems[i]; + if (!re_node_set_contains (cur_nodes, candidate)) + continue; + if (dest_node == -1) + dest_node = candidate; + + else + { + /* In order to avoid infinite loop like "(a*)*", return the second + epsilon-transition if the first was already considered. */ + if (re_node_set_contains (eps_via_nodes, dest_node)) + return candidate; + + /* Otherwise, push the second epsilon-transition on the fail stack. */ + else if (fs != NULL + && push_fail_stack (fs, *pidx, candidate, nregs, regs, + eps_via_nodes)) + return -2; + + /* We know we are going to exit. */ + break; + } + } + return dest_node; + } + else + { + int naccepted = 0; + re_token_type_t type = dfa->nodes[node].type; + +#ifdef RE_ENABLE_I18N + if (dfa->nodes[node].accept_mb) + naccepted = check_node_accept_bytes (dfa, node, &mctx->input, *pidx); + else +#endif /* RE_ENABLE_I18N */ + if (type == OP_BACK_REF) + { + int subexp_idx = dfa->nodes[node].opr.idx + 1; + naccepted = regs[subexp_idx].rm_eo - regs[subexp_idx].rm_so; + if (fs != NULL) + { + if (regs[subexp_idx].rm_so == -1 || regs[subexp_idx].rm_eo == -1) + return -1; + else if (naccepted) + { + char *buf = (char *) re_string_get_buffer (&mctx->input); + if (memcmp (buf + regs[subexp_idx].rm_so, buf + *pidx, + naccepted) != 0) + return -1; + } + } + + if (naccepted == 0) + { + int dest_node; + err = re_node_set_insert (eps_via_nodes, node); + if (BE (err < 0, 0)) + return -2; + dest_node = dfa->edests[node].elems[0]; + if (re_node_set_contains (&mctx->state_log[*pidx]->nodes, + dest_node)) + return dest_node; + } + } + + if (naccepted != 0 + || check_node_accept (mctx, dfa->nodes + node, *pidx)) + { + int dest_node = dfa->nexts[node]; + *pidx = (naccepted == 0) ? *pidx + 1 : *pidx + naccepted; + if (fs && (*pidx > mctx->match_last || mctx->state_log[*pidx] == NULL + || !re_node_set_contains (&mctx->state_log[*pidx]->nodes, + dest_node))) + return -1; + re_node_set_empty (eps_via_nodes); + return dest_node; + } + } + return -1; +} + +static reg_errcode_t +internal_function __attribute_warn_unused_result__ +push_fail_stack (struct re_fail_stack_t *fs, int str_idx, int dest_node, + int nregs, regmatch_t *regs, re_node_set *eps_via_nodes) +{ + reg_errcode_t err; + int num = fs->num++; + if (fs->num == fs->alloc) + { + struct re_fail_stack_ent_t *new_array; + new_array = realloc (fs->stack, (sizeof (struct re_fail_stack_ent_t) + * fs->alloc * 2)); + if (new_array == NULL) + return REG_ESPACE; + fs->alloc *= 2; + fs->stack = new_array; + } + fs->stack[num].idx = str_idx; + fs->stack[num].node = dest_node; + fs->stack[num].regs = re_malloc (regmatch_t, nregs); + if (fs->stack[num].regs == NULL) + return REG_ESPACE; + memcpy (fs->stack[num].regs, regs, sizeof (regmatch_t) * nregs); + err = re_node_set_init_copy (&fs->stack[num].eps_via_nodes, eps_via_nodes); + return err; +} + +static int +internal_function +pop_fail_stack (struct re_fail_stack_t *fs, int *pidx, int nregs, + regmatch_t *regs, re_node_set *eps_via_nodes) +{ + int num = --fs->num; + assert (num >= 0); + *pidx = fs->stack[num].idx; + memcpy (regs, fs->stack[num].regs, sizeof (regmatch_t) * nregs); + re_node_set_free (eps_via_nodes); + re_free (fs->stack[num].regs); + *eps_via_nodes = fs->stack[num].eps_via_nodes; + return fs->stack[num].node; +} + +/* Set the positions where the subexpressions are starts/ends to registers + PMATCH. + Note: We assume that pmatch[0] is already set, and + pmatch[i].rm_so == pmatch[i].rm_eo == -1 for 0 < i < nmatch. */ + +static reg_errcode_t +internal_function __attribute_warn_unused_result__ +set_regs (const regex_t *preg, const re_match_context_t *mctx, size_t nmatch, + regmatch_t *pmatch, int fl_backtrack) +{ + const re_dfa_t *dfa = (const re_dfa_t *) preg->buffer; + int idx, cur_node; + re_node_set eps_via_nodes; + struct re_fail_stack_t *fs; + struct re_fail_stack_t fs_body = { 0, 2, NULL }; + regmatch_t *prev_idx_match; + int prev_idx_match_malloced = 0; + +#ifdef DEBUG + assert (nmatch > 1); + assert (mctx->state_log != NULL); +#endif + if (fl_backtrack) + { + fs = &fs_body; + fs->stack = re_malloc (struct re_fail_stack_ent_t, fs->alloc); + if (fs->stack == NULL) + return REG_ESPACE; + } + else + fs = NULL; + + cur_node = dfa->init_node; + re_node_set_init_empty (&eps_via_nodes); + +#ifdef HAVE_ALLOCA + if (__libc_use_alloca (nmatch * sizeof (regmatch_t))) + prev_idx_match = (regmatch_t *) alloca (nmatch * sizeof (regmatch_t)); + else +#endif + { + prev_idx_match = re_malloc (regmatch_t, nmatch); + if (prev_idx_match == NULL) + { + free_fail_stack_return (fs); + return REG_ESPACE; + } + prev_idx_match_malloced = 1; + } + memcpy (prev_idx_match, pmatch, sizeof (regmatch_t) * nmatch); + + for (idx = pmatch[0].rm_so; idx <= pmatch[0].rm_eo ;) + { + update_regs (dfa, pmatch, prev_idx_match, cur_node, idx, nmatch); + + if (idx == pmatch[0].rm_eo && cur_node == mctx->last_node) + { + int reg_idx; + if (fs) + { + for (reg_idx = 0; reg_idx < nmatch; ++reg_idx) + if (pmatch[reg_idx].rm_so > -1 && pmatch[reg_idx].rm_eo == -1) + break; + if (reg_idx == nmatch) + { + re_node_set_free (&eps_via_nodes); + if (prev_idx_match_malloced) + re_free (prev_idx_match); + return free_fail_stack_return (fs); + } + cur_node = pop_fail_stack (fs, &idx, nmatch, pmatch, + &eps_via_nodes); + } + else + { + re_node_set_free (&eps_via_nodes); + if (prev_idx_match_malloced) + re_free (prev_idx_match); + return REG_NOERROR; + } + } + + /* Proceed to next node. */ + cur_node = proceed_next_node (mctx, nmatch, pmatch, &idx, cur_node, + &eps_via_nodes, fs); + + if (BE (cur_node < 0, 0)) + { + if (BE (cur_node == -2, 0)) + { + re_node_set_free (&eps_via_nodes); + if (prev_idx_match_malloced) + re_free (prev_idx_match); + free_fail_stack_return (fs); + return REG_ESPACE; + } + if (fs) + cur_node = pop_fail_stack (fs, &idx, nmatch, pmatch, + &eps_via_nodes); + else + { + re_node_set_free (&eps_via_nodes); + if (prev_idx_match_malloced) + re_free (prev_idx_match); + return REG_NOMATCH; + } + } + } + re_node_set_free (&eps_via_nodes); + if (prev_idx_match_malloced) + re_free (prev_idx_match); + return free_fail_stack_return (fs); +} + +static reg_errcode_t +internal_function +free_fail_stack_return (struct re_fail_stack_t *fs) +{ + if (fs) + { + int fs_idx; + for (fs_idx = 0; fs_idx < fs->num; ++fs_idx) + { + re_node_set_free (&fs->stack[fs_idx].eps_via_nodes); + re_free (fs->stack[fs_idx].regs); + } + re_free (fs->stack); + } + return REG_NOERROR; +} + +static void +internal_function +update_regs (const re_dfa_t *dfa, regmatch_t *pmatch, + regmatch_t *prev_idx_match, int cur_node, int cur_idx, int nmatch) +{ + int type = dfa->nodes[cur_node].type; + if (type == OP_OPEN_SUBEXP) + { + int reg_num = dfa->nodes[cur_node].opr.idx + 1; + + /* We are at the first node of this sub expression. */ + if (reg_num < nmatch) + { + pmatch[reg_num].rm_so = cur_idx; + pmatch[reg_num].rm_eo = -1; + } + } + else if (type == OP_CLOSE_SUBEXP) + { + int reg_num = dfa->nodes[cur_node].opr.idx + 1; + if (reg_num < nmatch) + { + /* We are at the last node of this sub expression. */ + if (pmatch[reg_num].rm_so < cur_idx) + { + pmatch[reg_num].rm_eo = cur_idx; + /* This is a non-empty match or we are not inside an optional + subexpression. Accept this right away. */ + memcpy (prev_idx_match, pmatch, sizeof (regmatch_t) * nmatch); + } + else + { + if (dfa->nodes[cur_node].opt_subexp + && prev_idx_match[reg_num].rm_so != -1) + /* We transited through an empty match for an optional + subexpression, like (a?)*, and this is not the subexp's + first match. Copy back the old content of the registers + so that matches of an inner subexpression are undone as + well, like in ((a?))*. */ + memcpy (pmatch, prev_idx_match, sizeof (regmatch_t) * nmatch); + else + /* We completed a subexpression, but it may be part of + an optional one, so do not update PREV_IDX_MATCH. */ + pmatch[reg_num].rm_eo = cur_idx; + } + } + } +} + +/* This function checks the STATE_LOG from the SCTX->last_str_idx to 0 + and sift the nodes in each states according to the following rules. + Updated state_log will be wrote to STATE_LOG. + + Rules: We throw away the Node `a' in the STATE_LOG[STR_IDX] if... + 1. When STR_IDX == MATCH_LAST(the last index in the state_log): + If `a' isn't the LAST_NODE and `a' can't epsilon transit to + the LAST_NODE, we throw away the node `a'. + 2. When 0 <= STR_IDX < MATCH_LAST and `a' accepts + string `s' and transit to `b': + i. If 'b' isn't in the STATE_LOG[STR_IDX+strlen('s')], we throw + away the node `a'. + ii. If 'b' is in the STATE_LOG[STR_IDX+strlen('s')] but 'b' is + thrown away, we throw away the node `a'. + 3. When 0 <= STR_IDX < MATCH_LAST and 'a' epsilon transit to 'b': + i. If 'b' isn't in the STATE_LOG[STR_IDX], we throw away the + node `a'. + ii. If 'b' is in the STATE_LOG[STR_IDX] but 'b' is thrown away, + we throw away the node `a'. */ + +#define STATE_NODE_CONTAINS(state,node) \ + ((state) != NULL && re_node_set_contains (&(state)->nodes, node)) + +static reg_errcode_t +internal_function +sift_states_backward (const re_match_context_t *mctx, re_sift_context_t *sctx) +{ + reg_errcode_t err; + int null_cnt = 0; + int str_idx = sctx->last_str_idx; + re_node_set cur_dest; + +#ifdef DEBUG + assert (mctx->state_log != NULL && mctx->state_log[str_idx] != NULL); +#endif + + /* Build sifted state_log[str_idx]. It has the nodes which can epsilon + transit to the last_node and the last_node itself. */ + err = re_node_set_init_1 (&cur_dest, sctx->last_node); + if (BE (err != REG_NOERROR, 0)) + return err; + err = update_cur_sifted_state (mctx, sctx, str_idx, &cur_dest); + if (BE (err != REG_NOERROR, 0)) + goto free_return; + + /* Then check each states in the state_log. */ + while (str_idx > 0) + { + /* Update counters. */ + null_cnt = (sctx->sifted_states[str_idx] == NULL) ? null_cnt + 1 : 0; + if (null_cnt > mctx->max_mb_elem_len) + { + memset (sctx->sifted_states, '\0', + sizeof (re_dfastate_t *) * str_idx); + re_node_set_free (&cur_dest); + return REG_NOERROR; + } + re_node_set_empty (&cur_dest); + --str_idx; + + if (mctx->state_log[str_idx]) + { + err = build_sifted_states (mctx, sctx, str_idx, &cur_dest); + if (BE (err != REG_NOERROR, 0)) + goto free_return; + } + + /* Add all the nodes which satisfy the following conditions: + - It can epsilon transit to a node in CUR_DEST. + - It is in CUR_SRC. + And update state_log. */ + err = update_cur_sifted_state (mctx, sctx, str_idx, &cur_dest); + if (BE (err != REG_NOERROR, 0)) + goto free_return; + } + err = REG_NOERROR; + free_return: + re_node_set_free (&cur_dest); + return err; +} + +static reg_errcode_t +internal_function __attribute_warn_unused_result__ +build_sifted_states (const re_match_context_t *mctx, re_sift_context_t *sctx, + int str_idx, re_node_set *cur_dest) +{ + const re_dfa_t *const dfa = mctx->dfa; + const re_node_set *cur_src = &mctx->state_log[str_idx]->non_eps_nodes; + int i; + + /* Then build the next sifted state. + We build the next sifted state on `cur_dest', and update + `sifted_states[str_idx]' with `cur_dest'. + Note: + `cur_dest' is the sifted state from `state_log[str_idx + 1]'. + `cur_src' points the node_set of the old `state_log[str_idx]' + (with the epsilon nodes pre-filtered out). */ + for (i = 0; i < cur_src->nelem; i++) + { + int prev_node = cur_src->elems[i]; + int naccepted = 0; + int ret; + +#ifdef DEBUG + re_token_type_t type = dfa->nodes[prev_node].type; + assert (!IS_EPSILON_NODE (type)); +#endif +#ifdef RE_ENABLE_I18N + /* If the node may accept `multi byte'. */ + if (dfa->nodes[prev_node].accept_mb) + naccepted = sift_states_iter_mb (mctx, sctx, prev_node, + str_idx, sctx->last_str_idx); +#endif /* RE_ENABLE_I18N */ + + /* We don't check backreferences here. + See update_cur_sifted_state(). */ + if (!naccepted + && check_node_accept (mctx, dfa->nodes + prev_node, str_idx) + && STATE_NODE_CONTAINS (sctx->sifted_states[str_idx + 1], + dfa->nexts[prev_node])) + naccepted = 1; + + if (naccepted == 0) + continue; + + if (sctx->limits.nelem) + { + int to_idx = str_idx + naccepted; + if (check_dst_limits (mctx, &sctx->limits, + dfa->nexts[prev_node], to_idx, + prev_node, str_idx)) + continue; + } + ret = re_node_set_insert (cur_dest, prev_node); + if (BE (ret == -1, 0)) + return REG_ESPACE; + } + + return REG_NOERROR; +} + +/* Helper functions. */ + +static reg_errcode_t +internal_function +clean_state_log_if_needed (re_match_context_t *mctx, int next_state_log_idx) +{ + int top = mctx->state_log_top; + + if ((next_state_log_idx >= mctx->input.bufs_len + && mctx->input.bufs_len < mctx->input.len) + || (next_state_log_idx >= mctx->input.valid_len + && mctx->input.valid_len < mctx->input.len)) + { + reg_errcode_t err; + err = extend_buffers (mctx, next_state_log_idx + 1); + if (BE (err != REG_NOERROR, 0)) + return err; + } + + if (top < next_state_log_idx) + { + memset (mctx->state_log + top + 1, '\0', + sizeof (re_dfastate_t *) * (next_state_log_idx - top)); + mctx->state_log_top = next_state_log_idx; + } + return REG_NOERROR; +} + +static reg_errcode_t +internal_function +merge_state_array (const re_dfa_t *dfa, re_dfastate_t **dst, + re_dfastate_t **src, int num) +{ + int st_idx; + reg_errcode_t err; + for (st_idx = 0; st_idx < num; ++st_idx) + { + if (dst[st_idx] == NULL) + dst[st_idx] = src[st_idx]; + else if (src[st_idx] != NULL) + { + re_node_set merged_set; + err = re_node_set_init_union (&merged_set, &dst[st_idx]->nodes, + &src[st_idx]->nodes); + if (BE (err != REG_NOERROR, 0)) + return err; + dst[st_idx] = re_acquire_state (&err, dfa, &merged_set); + re_node_set_free (&merged_set); + if (BE (err != REG_NOERROR, 0)) + return err; + } + } + return REG_NOERROR; +} + +static reg_errcode_t +internal_function +update_cur_sifted_state (const re_match_context_t *mctx, + re_sift_context_t *sctx, int str_idx, + re_node_set *dest_nodes) +{ + const re_dfa_t *const dfa = mctx->dfa; + reg_errcode_t err = REG_NOERROR; + const re_node_set *candidates; + candidates = ((mctx->state_log[str_idx] == NULL) ? NULL + : &mctx->state_log[str_idx]->nodes); + + if (dest_nodes->nelem == 0) + sctx->sifted_states[str_idx] = NULL; + else + { + if (candidates) + { + /* At first, add the nodes which can epsilon transit to a node in + DEST_NODE. */ + err = add_epsilon_src_nodes (dfa, dest_nodes, candidates); + if (BE (err != REG_NOERROR, 0)) + return err; + + /* Then, check the limitations in the current sift_context. */ + if (sctx->limits.nelem) + { + err = check_subexp_limits (dfa, dest_nodes, candidates, &sctx->limits, + mctx->bkref_ents, str_idx); + if (BE (err != REG_NOERROR, 0)) + return err; + } + } + + sctx->sifted_states[str_idx] = re_acquire_state (&err, dfa, dest_nodes); + if (BE (err != REG_NOERROR, 0)) + return err; + } + + if (candidates && mctx->state_log[str_idx]->has_backref) + { + err = sift_states_bkref (mctx, sctx, str_idx, candidates); + if (BE (err != REG_NOERROR, 0)) + return err; + } + return REG_NOERROR; +} + +static reg_errcode_t +internal_function __attribute_warn_unused_result__ +add_epsilon_src_nodes (const re_dfa_t *dfa, re_node_set *dest_nodes, + const re_node_set *candidates) +{ + reg_errcode_t err = REG_NOERROR; + int i; + + re_dfastate_t *state = re_acquire_state (&err, dfa, dest_nodes); + if (BE (err != REG_NOERROR, 0)) + return err; + + if (!state->inveclosure.alloc) + { + err = re_node_set_alloc (&state->inveclosure, dest_nodes->nelem); + if (BE (err != REG_NOERROR, 0)) + return REG_ESPACE; + for (i = 0; i < dest_nodes->nelem; i++) + { + err = re_node_set_merge (&state->inveclosure, + dfa->inveclosures + dest_nodes->elems[i]); + if (BE (err != REG_NOERROR, 0)) + return REG_ESPACE; + } + } + return re_node_set_add_intersect (dest_nodes, candidates, + &state->inveclosure); +} + +static reg_errcode_t +internal_function +sub_epsilon_src_nodes (const re_dfa_t *dfa, int node, re_node_set *dest_nodes, + const re_node_set *candidates) +{ + int ecl_idx; + reg_errcode_t err; + re_node_set *inv_eclosure = dfa->inveclosures + node; + re_node_set except_nodes; + re_node_set_init_empty (&except_nodes); + for (ecl_idx = 0; ecl_idx < inv_eclosure->nelem; ++ecl_idx) + { + int cur_node = inv_eclosure->elems[ecl_idx]; + if (cur_node == node) + continue; + if (IS_EPSILON_NODE (dfa->nodes[cur_node].type)) + { + int edst1 = dfa->edests[cur_node].elems[0]; + int edst2 = ((dfa->edests[cur_node].nelem > 1) + ? dfa->edests[cur_node].elems[1] : -1); + if ((!re_node_set_contains (inv_eclosure, edst1) + && re_node_set_contains (dest_nodes, edst1)) + || (edst2 > 0 + && !re_node_set_contains (inv_eclosure, edst2) + && re_node_set_contains (dest_nodes, edst2))) + { + err = re_node_set_add_intersect (&except_nodes, candidates, + dfa->inveclosures + cur_node); + if (BE (err != REG_NOERROR, 0)) + { + re_node_set_free (&except_nodes); + return err; + } + } + } + } + for (ecl_idx = 0; ecl_idx < inv_eclosure->nelem; ++ecl_idx) + { + int cur_node = inv_eclosure->elems[ecl_idx]; + if (!re_node_set_contains (&except_nodes, cur_node)) + { + int idx = re_node_set_contains (dest_nodes, cur_node) - 1; + re_node_set_remove_at (dest_nodes, idx); + } + } + re_node_set_free (&except_nodes); + return REG_NOERROR; +} + +static int +internal_function +check_dst_limits (const re_match_context_t *mctx, re_node_set *limits, + int dst_node, int dst_idx, int src_node, int src_idx) +{ + const re_dfa_t *const dfa = mctx->dfa; + int lim_idx, src_pos, dst_pos; + + int dst_bkref_idx = search_cur_bkref_entry (mctx, dst_idx); + int src_bkref_idx = search_cur_bkref_entry (mctx, src_idx); + for (lim_idx = 0; lim_idx < limits->nelem; ++lim_idx) + { + int subexp_idx; + struct re_backref_cache_entry *ent; + ent = mctx->bkref_ents + limits->elems[lim_idx]; + subexp_idx = dfa->nodes[ent->node].opr.idx; + + dst_pos = check_dst_limits_calc_pos (mctx, limits->elems[lim_idx], + subexp_idx, dst_node, dst_idx, + dst_bkref_idx); + src_pos = check_dst_limits_calc_pos (mctx, limits->elems[lim_idx], + subexp_idx, src_node, src_idx, + src_bkref_idx); + + /* In case of: + <src> <dst> ( <subexp> ) + ( <subexp> ) <src> <dst> + ( <subexp1> <src> <subexp2> <dst> <subexp3> ) */ + if (src_pos == dst_pos) + continue; /* This is unrelated limitation. */ + else + return 1; + } + return 0; +} + +static int +internal_function +check_dst_limits_calc_pos_1 (const re_match_context_t *mctx, int boundaries, + int subexp_idx, int from_node, int bkref_idx) +{ + const re_dfa_t *const dfa = mctx->dfa; + const re_node_set *eclosures = dfa->eclosures + from_node; + int node_idx; + + /* Else, we are on the boundary: examine the nodes on the epsilon + closure. */ + for (node_idx = 0; node_idx < eclosures->nelem; ++node_idx) + { + int node = eclosures->elems[node_idx]; + switch (dfa->nodes[node].type) + { + case OP_BACK_REF: + if (bkref_idx != -1) + { + struct re_backref_cache_entry *ent = mctx->bkref_ents + bkref_idx; + do + { + int dst, cpos; + + if (ent->node != node) + continue; + + if (subexp_idx < BITSET_WORD_BITS + && !(ent->eps_reachable_subexps_map + & ((bitset_word_t) 1 << subexp_idx))) + continue; + + /* Recurse trying to reach the OP_OPEN_SUBEXP and + OP_CLOSE_SUBEXP cases below. But, if the + destination node is the same node as the source + node, don't recurse because it would cause an + infinite loop: a regex that exhibits this behavior + is ()\1*\1* */ + dst = dfa->edests[node].elems[0]; + if (dst == from_node) + { + if (boundaries & 1) + return -1; + else /* if (boundaries & 2) */ + return 0; + } + + cpos = + check_dst_limits_calc_pos_1 (mctx, boundaries, subexp_idx, + dst, bkref_idx); + if (cpos == -1 /* && (boundaries & 1) */) + return -1; + if (cpos == 0 && (boundaries & 2)) + return 0; + + if (subexp_idx < BITSET_WORD_BITS) + ent->eps_reachable_subexps_map + &= ~((bitset_word_t) 1 << subexp_idx); + } + while (ent++->more); + } + break; + + case OP_OPEN_SUBEXP: + if ((boundaries & 1) && subexp_idx == dfa->nodes[node].opr.idx) + return -1; + break; + + case OP_CLOSE_SUBEXP: + if ((boundaries & 2) && subexp_idx == dfa->nodes[node].opr.idx) + return 0; + break; + + default: + break; + } + } + + return (boundaries & 2) ? 1 : 0; +} + +static int +internal_function +check_dst_limits_calc_pos (const re_match_context_t *mctx, int limit, + int subexp_idx, int from_node, int str_idx, + int bkref_idx) +{ + struct re_backref_cache_entry *lim = mctx->bkref_ents + limit; + int boundaries; + + /* If we are outside the range of the subexpression, return -1 or 1. */ + if (str_idx < lim->subexp_from) + return -1; + + if (lim->subexp_to < str_idx) + return 1; + + /* If we are within the subexpression, return 0. */ + boundaries = (str_idx == lim->subexp_from); + boundaries |= (str_idx == lim->subexp_to) << 1; + if (boundaries == 0) + return 0; + + /* Else, examine epsilon closure. */ + return check_dst_limits_calc_pos_1 (mctx, boundaries, subexp_idx, + from_node, bkref_idx); +} + +/* Check the limitations of sub expressions LIMITS, and remove the nodes + which are against limitations from DEST_NODES. */ + +static reg_errcode_t +internal_function +check_subexp_limits (const re_dfa_t *dfa, re_node_set *dest_nodes, + const re_node_set *candidates, re_node_set *limits, + struct re_backref_cache_entry *bkref_ents, int str_idx) +{ + reg_errcode_t err; + int node_idx, lim_idx; + + for (lim_idx = 0; lim_idx < limits->nelem; ++lim_idx) + { + int subexp_idx; + struct re_backref_cache_entry *ent; + ent = bkref_ents + limits->elems[lim_idx]; + + if (str_idx <= ent->subexp_from || ent->str_idx < str_idx) + continue; /* This is unrelated limitation. */ + + subexp_idx = dfa->nodes[ent->node].opr.idx; + if (ent->subexp_to == str_idx) + { + int ops_node = -1; + int cls_node = -1; + for (node_idx = 0; node_idx < dest_nodes->nelem; ++node_idx) + { + int node = dest_nodes->elems[node_idx]; + re_token_type_t type = dfa->nodes[node].type; + if (type == OP_OPEN_SUBEXP + && subexp_idx == dfa->nodes[node].opr.idx) + ops_node = node; + else if (type == OP_CLOSE_SUBEXP + && subexp_idx == dfa->nodes[node].opr.idx) + cls_node = node; + } + + /* Check the limitation of the open subexpression. */ + /* Note that (ent->subexp_to = str_idx != ent->subexp_from). */ + if (ops_node >= 0) + { + err = sub_epsilon_src_nodes (dfa, ops_node, dest_nodes, + candidates); + if (BE (err != REG_NOERROR, 0)) + return err; + } + + /* Check the limitation of the close subexpression. */ + if (cls_node >= 0) + for (node_idx = 0; node_idx < dest_nodes->nelem; ++node_idx) + { + int node = dest_nodes->elems[node_idx]; + if (!re_node_set_contains (dfa->inveclosures + node, + cls_node) + && !re_node_set_contains (dfa->eclosures + node, + cls_node)) + { + /* It is against this limitation. + Remove it form the current sifted state. */ + err = sub_epsilon_src_nodes (dfa, node, dest_nodes, + candidates); + if (BE (err != REG_NOERROR, 0)) + return err; + --node_idx; + } + } + } + else /* (ent->subexp_to != str_idx) */ + { + for (node_idx = 0; node_idx < dest_nodes->nelem; ++node_idx) + { + int node = dest_nodes->elems[node_idx]; + re_token_type_t type = dfa->nodes[node].type; + if (type == OP_CLOSE_SUBEXP || type == OP_OPEN_SUBEXP) + { + if (subexp_idx != dfa->nodes[node].opr.idx) + continue; + /* It is against this limitation. + Remove it form the current sifted state. */ + err = sub_epsilon_src_nodes (dfa, node, dest_nodes, + candidates); + if (BE (err != REG_NOERROR, 0)) + return err; + } + } + } + } + return REG_NOERROR; +} + +static reg_errcode_t +internal_function __attribute_warn_unused_result__ +sift_states_bkref (const re_match_context_t *mctx, re_sift_context_t *sctx, + int str_idx, const re_node_set *candidates) +{ + const re_dfa_t *const dfa = mctx->dfa; + reg_errcode_t err; + int node_idx, node; + re_sift_context_t local_sctx; + int first_idx = search_cur_bkref_entry (mctx, str_idx); + + if (first_idx == -1) + return REG_NOERROR; + + local_sctx.sifted_states = NULL; /* Mark that it hasn't been initialized. */ + + for (node_idx = 0; node_idx < candidates->nelem; ++node_idx) + { + int enabled_idx; + re_token_type_t type; + struct re_backref_cache_entry *entry; + node = candidates->elems[node_idx]; + type = dfa->nodes[node].type; + /* Avoid infinite loop for the REs like "()\1+". */ + if (node == sctx->last_node && str_idx == sctx->last_str_idx) + continue; + if (type != OP_BACK_REF) + continue; + + entry = mctx->bkref_ents + first_idx; + enabled_idx = first_idx; + do + { + int subexp_len; + int to_idx; + int dst_node; + int ret; + re_dfastate_t *cur_state; + + if (entry->node != node) + continue; + subexp_len = entry->subexp_to - entry->subexp_from; + to_idx = str_idx + subexp_len; + dst_node = (subexp_len ? dfa->nexts[node] + : dfa->edests[node].elems[0]); + + if (to_idx > sctx->last_str_idx + || sctx->sifted_states[to_idx] == NULL + || !STATE_NODE_CONTAINS (sctx->sifted_states[to_idx], dst_node) + || check_dst_limits (mctx, &sctx->limits, node, + str_idx, dst_node, to_idx)) + continue; + + if (local_sctx.sifted_states == NULL) + { + local_sctx = *sctx; + err = re_node_set_init_copy (&local_sctx.limits, &sctx->limits); + if (BE (err != REG_NOERROR, 0)) + goto free_return; + } + local_sctx.last_node = node; + local_sctx.last_str_idx = str_idx; + ret = re_node_set_insert (&local_sctx.limits, enabled_idx); + if (BE (ret < 0, 0)) + { + err = REG_ESPACE; + goto free_return; + } + cur_state = local_sctx.sifted_states[str_idx]; + err = sift_states_backward (mctx, &local_sctx); + if (BE (err != REG_NOERROR, 0)) + goto free_return; + if (sctx->limited_states != NULL) + { + err = merge_state_array (dfa, sctx->limited_states, + local_sctx.sifted_states, + str_idx + 1); + if (BE (err != REG_NOERROR, 0)) + goto free_return; + } + local_sctx.sifted_states[str_idx] = cur_state; + re_node_set_remove (&local_sctx.limits, enabled_idx); + + /* mctx->bkref_ents may have changed, reload the pointer. */ + entry = mctx->bkref_ents + enabled_idx; + } + while (enabled_idx++, entry++->more); + } + err = REG_NOERROR; + free_return: + if (local_sctx.sifted_states != NULL) + { + re_node_set_free (&local_sctx.limits); + } + + return err; +} + + +#ifdef RE_ENABLE_I18N +static int +internal_function +sift_states_iter_mb (const re_match_context_t *mctx, re_sift_context_t *sctx, + int node_idx, int str_idx, int max_str_idx) +{ + const re_dfa_t *const dfa = mctx->dfa; + int naccepted; + /* Check the node can accept `multi byte'. */ + naccepted = check_node_accept_bytes (dfa, node_idx, &mctx->input, str_idx); + if (naccepted > 0 && str_idx + naccepted <= max_str_idx && + !STATE_NODE_CONTAINS (sctx->sifted_states[str_idx + naccepted], + dfa->nexts[node_idx])) + /* The node can't accept the `multi byte', or the + destination was already thrown away, then the node + could't accept the current input `multi byte'. */ + naccepted = 0; + /* Otherwise, it is sure that the node could accept + `naccepted' bytes input. */ + return naccepted; +} +#endif /* RE_ENABLE_I18N */ + + +/* Functions for state transition. */ + +/* Return the next state to which the current state STATE will transit by + accepting the current input byte, and update STATE_LOG if necessary. + If STATE can accept a multibyte char/collating element/back reference + update the destination of STATE_LOG. */ + +static re_dfastate_t * +internal_function __attribute_warn_unused_result__ +transit_state (reg_errcode_t *err, re_match_context_t *mctx, + re_dfastate_t *state) +{ + re_dfastate_t **trtable; + unsigned char ch; + +#ifdef RE_ENABLE_I18N + /* If the current state can accept multibyte. */ + if (BE (state->accept_mb, 0)) + { + *err = transit_state_mb (mctx, state); + if (BE (*err != REG_NOERROR, 0)) + return NULL; + } +#endif /* RE_ENABLE_I18N */ + + /* Then decide the next state with the single byte. */ +#if 0 + if (0) + /* don't use transition table */ + return transit_state_sb (err, mctx, state); +#endif + + /* Use transition table */ + ch = re_string_fetch_byte (&mctx->input); + for (;;) + { + trtable = state->trtable; + if (BE (trtable != NULL, 1)) + return trtable[ch]; + + trtable = state->word_trtable; + if (BE (trtable != NULL, 1)) + { + unsigned int context; + context + = re_string_context_at (&mctx->input, + re_string_cur_idx (&mctx->input) - 1, + mctx->eflags); + if (IS_WORD_CONTEXT (context)) + return trtable[ch + SBC_MAX]; + else + return trtable[ch]; + } + + if (!build_trtable (mctx->dfa, state)) + { + *err = REG_ESPACE; + return NULL; + } + + /* Retry, we now have a transition table. */ + } +} + +/* Update the state_log if we need */ +re_dfastate_t * +internal_function +merge_state_with_log (reg_errcode_t *err, re_match_context_t *mctx, + re_dfastate_t *next_state) +{ + const re_dfa_t *const dfa = mctx->dfa; + int cur_idx = re_string_cur_idx (&mctx->input); + + if (cur_idx > mctx->state_log_top) + { + mctx->state_log[cur_idx] = next_state; + mctx->state_log_top = cur_idx; + } + else if (mctx->state_log[cur_idx] == 0) + { + mctx->state_log[cur_idx] = next_state; + } + else + { + re_dfastate_t *pstate; + unsigned int context; + re_node_set next_nodes, *log_nodes, *table_nodes = NULL; + /* If (state_log[cur_idx] != 0), it implies that cur_idx is + the destination of a multibyte char/collating element/ + back reference. Then the next state is the union set of + these destinations and the results of the transition table. */ + pstate = mctx->state_log[cur_idx]; + log_nodes = pstate->entrance_nodes; + if (next_state != NULL) + { + table_nodes = next_state->entrance_nodes; + *err = re_node_set_init_union (&next_nodes, table_nodes, + log_nodes); + if (BE (*err != REG_NOERROR, 0)) + return NULL; + } + else + next_nodes = *log_nodes; + /* Note: We already add the nodes of the initial state, + then we don't need to add them here. */ + + context = re_string_context_at (&mctx->input, + re_string_cur_idx (&mctx->input) - 1, + mctx->eflags); + next_state = mctx->state_log[cur_idx] + = re_acquire_state_context (err, dfa, &next_nodes, context); + /* We don't need to check errors here, since the return value of + this function is next_state and ERR is already set. */ + + if (table_nodes != NULL) + re_node_set_free (&next_nodes); + } + + if (BE (dfa->nbackref, 0) && next_state != NULL) + { + /* Check OP_OPEN_SUBEXP in the current state in case that we use them + later. We must check them here, since the back references in the + next state might use them. */ + *err = check_subexp_matching_top (mctx, &next_state->nodes, + cur_idx); + if (BE (*err != REG_NOERROR, 0)) + return NULL; + + /* If the next state has back references. */ + if (next_state->has_backref) + { + *err = transit_state_bkref (mctx, &next_state->nodes); + if (BE (*err != REG_NOERROR, 0)) + return NULL; + next_state = mctx->state_log[cur_idx]; + } + } + + return next_state; +} + +/* Skip bytes in the input that correspond to part of a + multi-byte match, then look in the log for a state + from which to restart matching. */ +static re_dfastate_t * +internal_function +find_recover_state (reg_errcode_t *err, re_match_context_t *mctx) +{ + re_dfastate_t *cur_state; + do + { + int max = mctx->state_log_top; + int cur_str_idx = re_string_cur_idx (&mctx->input); + + do + { + if (++cur_str_idx > max) + return NULL; + re_string_skip_bytes (&mctx->input, 1); + } + while (mctx->state_log[cur_str_idx] == NULL); + + cur_state = merge_state_with_log (err, mctx, NULL); + } + while (*err == REG_NOERROR && cur_state == NULL); + return cur_state; +} + +/* Helper functions for transit_state. */ + +/* From the node set CUR_NODES, pick up the nodes whose types are + OP_OPEN_SUBEXP and which have corresponding back references in the regular + expression. And register them to use them later for evaluating the + corresponding back references. */ + +static reg_errcode_t +internal_function +check_subexp_matching_top (re_match_context_t *mctx, re_node_set *cur_nodes, + int str_idx) +{ + const re_dfa_t *const dfa = mctx->dfa; + int node_idx; + reg_errcode_t err; + + /* TODO: This isn't efficient. + Because there might be more than one nodes whose types are + OP_OPEN_SUBEXP and whose index is SUBEXP_IDX, we must check all + nodes. + E.g. RE: (a){2} */ + for (node_idx = 0; node_idx < cur_nodes->nelem; ++node_idx) + { + int node = cur_nodes->elems[node_idx]; + if (dfa->nodes[node].type == OP_OPEN_SUBEXP + && dfa->nodes[node].opr.idx < BITSET_WORD_BITS + && (dfa->used_bkref_map + & ((bitset_word_t) 1 << dfa->nodes[node].opr.idx))) + { + err = match_ctx_add_subtop (mctx, node, str_idx); + if (BE (err != REG_NOERROR, 0)) + return err; + } + } + return REG_NOERROR; +} + +#if 0 +/* Return the next state to which the current state STATE will transit by + accepting the current input byte. */ + +static re_dfastate_t * +transit_state_sb (reg_errcode_t *err, re_match_context_t *mctx, + re_dfastate_t *state) +{ + const re_dfa_t *const dfa = mctx->dfa; + re_node_set next_nodes; + re_dfastate_t *next_state; + int node_cnt, cur_str_idx = re_string_cur_idx (&mctx->input); + unsigned int context; + + *err = re_node_set_alloc (&next_nodes, state->nodes.nelem + 1); + if (BE (*err != REG_NOERROR, 0)) + return NULL; + for (node_cnt = 0; node_cnt < state->nodes.nelem; ++node_cnt) + { + int cur_node = state->nodes.elems[node_cnt]; + if (check_node_accept (mctx, dfa->nodes + cur_node, cur_str_idx)) + { + *err = re_node_set_merge (&next_nodes, + dfa->eclosures + dfa->nexts[cur_node]); + if (BE (*err != REG_NOERROR, 0)) + { + re_node_set_free (&next_nodes); + return NULL; + } + } + } + context = re_string_context_at (&mctx->input, cur_str_idx, mctx->eflags); + next_state = re_acquire_state_context (err, dfa, &next_nodes, context); + /* We don't need to check errors here, since the return value of + this function is next_state and ERR is already set. */ + + re_node_set_free (&next_nodes); + re_string_skip_bytes (&mctx->input, 1); + return next_state; +} +#endif + +#ifdef RE_ENABLE_I18N +static reg_errcode_t +internal_function +transit_state_mb (re_match_context_t *mctx, re_dfastate_t *pstate) +{ + const re_dfa_t *const dfa = mctx->dfa; + reg_errcode_t err; + int i; + + for (i = 0; i < pstate->nodes.nelem; ++i) + { + re_node_set dest_nodes, *new_nodes; + int cur_node_idx = pstate->nodes.elems[i]; + int naccepted, dest_idx; + unsigned int context; + re_dfastate_t *dest_state; + + if (!dfa->nodes[cur_node_idx].accept_mb) + continue; + + if (dfa->nodes[cur_node_idx].constraint) + { + context = re_string_context_at (&mctx->input, + re_string_cur_idx (&mctx->input), + mctx->eflags); + if (NOT_SATISFY_NEXT_CONSTRAINT (dfa->nodes[cur_node_idx].constraint, + context)) + continue; + } + + /* How many bytes the node can accept? */ + naccepted = check_node_accept_bytes (dfa, cur_node_idx, &mctx->input, + re_string_cur_idx (&mctx->input)); + if (naccepted == 0) + continue; + + /* The node can accepts `naccepted' bytes. */ + dest_idx = re_string_cur_idx (&mctx->input) + naccepted; + mctx->max_mb_elem_len = ((mctx->max_mb_elem_len < naccepted) ? naccepted + : mctx->max_mb_elem_len); + err = clean_state_log_if_needed (mctx, dest_idx); + if (BE (err != REG_NOERROR, 0)) + return err; +#ifdef DEBUG + assert (dfa->nexts[cur_node_idx] != -1); +#endif + new_nodes = dfa->eclosures + dfa->nexts[cur_node_idx]; + + dest_state = mctx->state_log[dest_idx]; + if (dest_state == NULL) + dest_nodes = *new_nodes; + else + { + err = re_node_set_init_union (&dest_nodes, + dest_state->entrance_nodes, new_nodes); + if (BE (err != REG_NOERROR, 0)) + return err; + } + context = re_string_context_at (&mctx->input, dest_idx - 1, + mctx->eflags); + mctx->state_log[dest_idx] + = re_acquire_state_context (&err, dfa, &dest_nodes, context); + if (dest_state != NULL) + re_node_set_free (&dest_nodes); + if (BE (mctx->state_log[dest_idx] == NULL && err != REG_NOERROR, 0)) + return err; + } + return REG_NOERROR; +} +#endif /* RE_ENABLE_I18N */ + +static reg_errcode_t +internal_function +transit_state_bkref (re_match_context_t *mctx, const re_node_set *nodes) +{ + const re_dfa_t *const dfa = mctx->dfa; + reg_errcode_t err; + int i; + int cur_str_idx = re_string_cur_idx (&mctx->input); + + for (i = 0; i < nodes->nelem; ++i) + { + int dest_str_idx, prev_nelem, bkc_idx; + int node_idx = nodes->elems[i]; + unsigned int context; + const re_token_t *node = dfa->nodes + node_idx; + re_node_set *new_dest_nodes; + + /* Check whether `node' is a backreference or not. */ + if (node->type != OP_BACK_REF) + continue; + + if (node->constraint) + { + context = re_string_context_at (&mctx->input, cur_str_idx, + mctx->eflags); + if (NOT_SATISFY_NEXT_CONSTRAINT (node->constraint, context)) + continue; + } + + /* `node' is a backreference. + Check the substring which the substring matched. */ + bkc_idx = mctx->nbkref_ents; + err = get_subexp (mctx, node_idx, cur_str_idx); + if (BE (err != REG_NOERROR, 0)) + goto free_return; + + /* And add the epsilon closures (which is `new_dest_nodes') of + the backreference to appropriate state_log. */ +#ifdef DEBUG + assert (dfa->nexts[node_idx] != -1); +#endif + for (; bkc_idx < mctx->nbkref_ents; ++bkc_idx) + { + int subexp_len; + re_dfastate_t *dest_state; + struct re_backref_cache_entry *bkref_ent; + bkref_ent = mctx->bkref_ents + bkc_idx; + if (bkref_ent->node != node_idx || bkref_ent->str_idx != cur_str_idx) + continue; + subexp_len = bkref_ent->subexp_to - bkref_ent->subexp_from; + new_dest_nodes = (subexp_len == 0 + ? dfa->eclosures + dfa->edests[node_idx].elems[0] + : dfa->eclosures + dfa->nexts[node_idx]); + dest_str_idx = (cur_str_idx + bkref_ent->subexp_to + - bkref_ent->subexp_from); + context = re_string_context_at (&mctx->input, dest_str_idx - 1, + mctx->eflags); + dest_state = mctx->state_log[dest_str_idx]; + prev_nelem = ((mctx->state_log[cur_str_idx] == NULL) ? 0 + : mctx->state_log[cur_str_idx]->nodes.nelem); + /* Add `new_dest_node' to state_log. */ + if (dest_state == NULL) + { + mctx->state_log[dest_str_idx] + = re_acquire_state_context (&err, dfa, new_dest_nodes, + context); + if (BE (mctx->state_log[dest_str_idx] == NULL + && err != REG_NOERROR, 0)) + goto free_return; + } + else + { + re_node_set dest_nodes; + err = re_node_set_init_union (&dest_nodes, + dest_state->entrance_nodes, + new_dest_nodes); + if (BE (err != REG_NOERROR, 0)) + { + re_node_set_free (&dest_nodes); + goto free_return; + } + mctx->state_log[dest_str_idx] + = re_acquire_state_context (&err, dfa, &dest_nodes, context); + re_node_set_free (&dest_nodes); + if (BE (mctx->state_log[dest_str_idx] == NULL + && err != REG_NOERROR, 0)) + goto free_return; + } + /* We need to check recursively if the backreference can epsilon + transit. */ + if (subexp_len == 0 + && mctx->state_log[cur_str_idx]->nodes.nelem > prev_nelem) + { + err = check_subexp_matching_top (mctx, new_dest_nodes, + cur_str_idx); + if (BE (err != REG_NOERROR, 0)) + goto free_return; + err = transit_state_bkref (mctx, new_dest_nodes); + if (BE (err != REG_NOERROR, 0)) + goto free_return; + } + } + } + err = REG_NOERROR; + free_return: + return err; +} + +/* Enumerate all the candidates which the backreference BKREF_NODE can match + at BKREF_STR_IDX, and register them by match_ctx_add_entry(). + Note that we might collect inappropriate candidates here. + However, the cost of checking them strictly here is too high, then we + delay these checking for prune_impossible_nodes(). */ + +static reg_errcode_t +internal_function __attribute_warn_unused_result__ +get_subexp (re_match_context_t *mctx, int bkref_node, int bkref_str_idx) +{ + const re_dfa_t *const dfa = mctx->dfa; + int subexp_num, sub_top_idx; + const char *buf = (const char *) re_string_get_buffer (&mctx->input); + /* Return if we have already checked BKREF_NODE at BKREF_STR_IDX. */ + int cache_idx = search_cur_bkref_entry (mctx, bkref_str_idx); + if (cache_idx != -1) + { + const struct re_backref_cache_entry *entry + = mctx->bkref_ents + cache_idx; + do + if (entry->node == bkref_node) + return REG_NOERROR; /* We already checked it. */ + while (entry++->more); + } + + subexp_num = dfa->nodes[bkref_node].opr.idx; + + /* For each sub expression */ + for (sub_top_idx = 0; sub_top_idx < mctx->nsub_tops; ++sub_top_idx) + { + reg_errcode_t err; + re_sub_match_top_t *sub_top = mctx->sub_tops[sub_top_idx]; + re_sub_match_last_t *sub_last; + int sub_last_idx, sl_str, bkref_str_off; + + if (dfa->nodes[sub_top->node].opr.idx != subexp_num) + continue; /* It isn't related. */ + + sl_str = sub_top->str_idx; + bkref_str_off = bkref_str_idx; + /* At first, check the last node of sub expressions we already + evaluated. */ + for (sub_last_idx = 0; sub_last_idx < sub_top->nlasts; ++sub_last_idx) + { + int sl_str_diff; + sub_last = sub_top->lasts[sub_last_idx]; + sl_str_diff = sub_last->str_idx - sl_str; + /* The matched string by the sub expression match with the substring + at the back reference? */ + if (sl_str_diff > 0) + { + if (BE (bkref_str_off + sl_str_diff > mctx->input.valid_len, 0)) + { + /* Not enough chars for a successful match. */ + if (bkref_str_off + sl_str_diff > mctx->input.len) + break; + + err = clean_state_log_if_needed (mctx, + bkref_str_off + + sl_str_diff); + if (BE (err != REG_NOERROR, 0)) + return err; + buf = (const char *) re_string_get_buffer (&mctx->input); + } + if (memcmp (buf + bkref_str_off, buf + sl_str, sl_str_diff) != 0) + /* We don't need to search this sub expression any more. */ + break; + } + bkref_str_off += sl_str_diff; + sl_str += sl_str_diff; + err = get_subexp_sub (mctx, sub_top, sub_last, bkref_node, + bkref_str_idx); + + /* Reload buf, since the preceding call might have reallocated + the buffer. */ + buf = (const char *) re_string_get_buffer (&mctx->input); + + if (err == REG_NOMATCH) + continue; + if (BE (err != REG_NOERROR, 0)) + return err; + } + + if (sub_last_idx < sub_top->nlasts) + continue; + if (sub_last_idx > 0) + ++sl_str; + /* Then, search for the other last nodes of the sub expression. */ + for (; sl_str <= bkref_str_idx; ++sl_str) + { + int cls_node, sl_str_off; + const re_node_set *nodes; + sl_str_off = sl_str - sub_top->str_idx; + /* The matched string by the sub expression match with the substring + at the back reference? */ + if (sl_str_off > 0) + { + if (BE (bkref_str_off >= mctx->input.valid_len, 0)) + { + /* If we are at the end of the input, we cannot match. */ + if (bkref_str_off >= mctx->input.len) + break; + + err = extend_buffers (mctx, bkref_str_off + 1); + if (BE (err != REG_NOERROR, 0)) + return err; + + buf = (const char *) re_string_get_buffer (&mctx->input); + } + if (buf [bkref_str_off++] != buf[sl_str - 1]) + break; /* We don't need to search this sub expression + any more. */ + } + if (mctx->state_log[sl_str] == NULL) + continue; + /* Does this state have a ')' of the sub expression? */ + nodes = &mctx->state_log[sl_str]->nodes; + cls_node = find_subexp_node (dfa, nodes, subexp_num, + OP_CLOSE_SUBEXP); + if (cls_node == -1) + continue; /* No. */ + if (sub_top->path == NULL) + { + sub_top->path = calloc (sizeof (state_array_t), + sl_str - sub_top->str_idx + 1); + if (sub_top->path == NULL) + return REG_ESPACE; + } + /* Can the OP_OPEN_SUBEXP node arrive the OP_CLOSE_SUBEXP node + in the current context? */ + err = check_arrival (mctx, sub_top->path, sub_top->node, + sub_top->str_idx, cls_node, sl_str, + OP_CLOSE_SUBEXP); + if (err == REG_NOMATCH) + continue; + if (BE (err != REG_NOERROR, 0)) + return err; + sub_last = match_ctx_add_sublast (sub_top, cls_node, sl_str); + if (BE (sub_last == NULL, 0)) + return REG_ESPACE; + err = get_subexp_sub (mctx, sub_top, sub_last, bkref_node, + bkref_str_idx); + if (err == REG_NOMATCH) + continue; + } + } + return REG_NOERROR; +} + +/* Helper functions for get_subexp(). */ + +/* Check SUB_LAST can arrive to the back reference BKREF_NODE at BKREF_STR. + If it can arrive, register the sub expression expressed with SUB_TOP + and SUB_LAST. */ + +static reg_errcode_t +internal_function +get_subexp_sub (re_match_context_t *mctx, const re_sub_match_top_t *sub_top, + re_sub_match_last_t *sub_last, int bkref_node, int bkref_str) +{ + reg_errcode_t err; + int to_idx; + /* Can the subexpression arrive the back reference? */ + err = check_arrival (mctx, &sub_last->path, sub_last->node, + sub_last->str_idx, bkref_node, bkref_str, + OP_OPEN_SUBEXP); + if (err != REG_NOERROR) + return err; + err = match_ctx_add_entry (mctx, bkref_node, bkref_str, sub_top->str_idx, + sub_last->str_idx); + if (BE (err != REG_NOERROR, 0)) + return err; + to_idx = bkref_str + sub_last->str_idx - sub_top->str_idx; + return clean_state_log_if_needed (mctx, to_idx); +} + +/* Find the first node which is '(' or ')' and whose index is SUBEXP_IDX. + Search '(' if FL_OPEN, or search ')' otherwise. + TODO: This function isn't efficient... + Because there might be more than one nodes whose types are + OP_OPEN_SUBEXP and whose index is SUBEXP_IDX, we must check all + nodes. + E.g. RE: (a){2} */ + +static int +internal_function +find_subexp_node (const re_dfa_t *dfa, const re_node_set *nodes, + int subexp_idx, int type) +{ + int cls_idx; + for (cls_idx = 0; cls_idx < nodes->nelem; ++cls_idx) + { + int cls_node = nodes->elems[cls_idx]; + const re_token_t *node = dfa->nodes + cls_node; + if (node->type == type + && node->opr.idx == subexp_idx) + return cls_node; + } + return -1; +} + +/* Check whether the node TOP_NODE at TOP_STR can arrive to the node + LAST_NODE at LAST_STR. We record the path onto PATH since it will be + heavily reused. + Return REG_NOERROR if it can arrive, or REG_NOMATCH otherwise. */ + +static reg_errcode_t +internal_function __attribute_warn_unused_result__ +check_arrival (re_match_context_t *mctx, state_array_t *path, int top_node, + int top_str, int last_node, int last_str, int type) +{ + const re_dfa_t *const dfa = mctx->dfa; + reg_errcode_t err = REG_NOERROR; + int subexp_num, backup_cur_idx, str_idx, null_cnt; + re_dfastate_t *cur_state = NULL; + re_node_set *cur_nodes, next_nodes; + re_dfastate_t **backup_state_log; + unsigned int context; + + subexp_num = dfa->nodes[top_node].opr.idx; + /* Extend the buffer if we need. */ + if (BE (path->alloc < last_str + mctx->max_mb_elem_len + 1, 0)) + { + re_dfastate_t **new_array; + int old_alloc = path->alloc; + path->alloc += last_str + mctx->max_mb_elem_len + 1; + new_array = re_realloc (path->array, re_dfastate_t *, path->alloc); + if (BE (new_array == NULL, 0)) + { + path->alloc = old_alloc; + return REG_ESPACE; + } + path->array = new_array; + memset (new_array + old_alloc, '\0', + sizeof (re_dfastate_t *) * (path->alloc - old_alloc)); + } + + str_idx = path->next_idx ? path->next_idx : top_str; + + /* Temporary modify MCTX. */ + backup_state_log = mctx->state_log; + backup_cur_idx = mctx->input.cur_idx; + mctx->state_log = path->array; + mctx->input.cur_idx = str_idx; + + /* Setup initial node set. */ + context = re_string_context_at (&mctx->input, str_idx - 1, mctx->eflags); + if (str_idx == top_str) + { + err = re_node_set_init_1 (&next_nodes, top_node); + if (BE (err != REG_NOERROR, 0)) + return err; + err = check_arrival_expand_ecl (dfa, &next_nodes, subexp_num, type); + if (BE (err != REG_NOERROR, 0)) + { + re_node_set_free (&next_nodes); + return err; + } + } + else + { + cur_state = mctx->state_log[str_idx]; + if (cur_state && cur_state->has_backref) + { + err = re_node_set_init_copy (&next_nodes, &cur_state->nodes); + if (BE (err != REG_NOERROR, 0)) + return err; + } + else + re_node_set_init_empty (&next_nodes); + } + if (str_idx == top_str || (cur_state && cur_state->has_backref)) + { + if (next_nodes.nelem) + { + err = expand_bkref_cache (mctx, &next_nodes, str_idx, + subexp_num, type); + if (BE (err != REG_NOERROR, 0)) + { + re_node_set_free (&next_nodes); + return err; + } + } + cur_state = re_acquire_state_context (&err, dfa, &next_nodes, context); + if (BE (cur_state == NULL && err != REG_NOERROR, 0)) + { + re_node_set_free (&next_nodes); + return err; + } + mctx->state_log[str_idx] = cur_state; + } + + for (null_cnt = 0; str_idx < last_str && null_cnt <= mctx->max_mb_elem_len;) + { + re_node_set_empty (&next_nodes); + if (mctx->state_log[str_idx + 1]) + { + err = re_node_set_merge (&next_nodes, + &mctx->state_log[str_idx + 1]->nodes); + if (BE (err != REG_NOERROR, 0)) + { + re_node_set_free (&next_nodes); + return err; + } + } + if (cur_state) + { + err = check_arrival_add_next_nodes (mctx, str_idx, + &cur_state->non_eps_nodes, + &next_nodes); + if (BE (err != REG_NOERROR, 0)) + { + re_node_set_free (&next_nodes); + return err; + } + } + ++str_idx; + if (next_nodes.nelem) + { + err = check_arrival_expand_ecl (dfa, &next_nodes, subexp_num, type); + if (BE (err != REG_NOERROR, 0)) + { + re_node_set_free (&next_nodes); + return err; + } + err = expand_bkref_cache (mctx, &next_nodes, str_idx, + subexp_num, type); + if (BE (err != REG_NOERROR, 0)) + { + re_node_set_free (&next_nodes); + return err; + } + } + context = re_string_context_at (&mctx->input, str_idx - 1, mctx->eflags); + cur_state = re_acquire_state_context (&err, dfa, &next_nodes, context); + if (BE (cur_state == NULL && err != REG_NOERROR, 0)) + { + re_node_set_free (&next_nodes); + return err; + } + mctx->state_log[str_idx] = cur_state; + null_cnt = cur_state == NULL ? null_cnt + 1 : 0; + } + re_node_set_free (&next_nodes); + cur_nodes = (mctx->state_log[last_str] == NULL ? NULL + : &mctx->state_log[last_str]->nodes); + path->next_idx = str_idx; + + /* Fix MCTX. */ + mctx->state_log = backup_state_log; + mctx->input.cur_idx = backup_cur_idx; + + /* Then check the current node set has the node LAST_NODE. */ + if (cur_nodes != NULL && re_node_set_contains (cur_nodes, last_node)) + return REG_NOERROR; + + return REG_NOMATCH; +} + +/* Helper functions for check_arrival. */ + +/* Calculate the destination nodes of CUR_NODES at STR_IDX, and append them + to NEXT_NODES. + TODO: This function is similar to the functions transit_state*(), + however this function has many additional works. + Can't we unify them? */ + +static reg_errcode_t +internal_function __attribute_warn_unused_result__ +check_arrival_add_next_nodes (re_match_context_t *mctx, int str_idx, + re_node_set *cur_nodes, re_node_set *next_nodes) +{ + const re_dfa_t *const dfa = mctx->dfa; + int result; + int cur_idx; +#ifdef RE_ENABLE_I18N + reg_errcode_t err = REG_NOERROR; +#endif + re_node_set union_set; + re_node_set_init_empty (&union_set); + for (cur_idx = 0; cur_idx < cur_nodes->nelem; ++cur_idx) + { + int naccepted = 0; + int cur_node = cur_nodes->elems[cur_idx]; +#ifdef DEBUG + re_token_type_t type = dfa->nodes[cur_node].type; + assert (!IS_EPSILON_NODE (type)); +#endif +#ifdef RE_ENABLE_I18N + /* If the node may accept `multi byte'. */ + if (dfa->nodes[cur_node].accept_mb) + { + naccepted = check_node_accept_bytes (dfa, cur_node, &mctx->input, + str_idx); + if (naccepted > 1) + { + re_dfastate_t *dest_state; + int next_node = dfa->nexts[cur_node]; + int next_idx = str_idx + naccepted; + dest_state = mctx->state_log[next_idx]; + re_node_set_empty (&union_set); + if (dest_state) + { + err = re_node_set_merge (&union_set, &dest_state->nodes); + if (BE (err != REG_NOERROR, 0)) + { + re_node_set_free (&union_set); + return err; + } + } + result = re_node_set_insert (&union_set, next_node); + if (BE (result < 0, 0)) + { + re_node_set_free (&union_set); + return REG_ESPACE; + } + mctx->state_log[next_idx] = re_acquire_state (&err, dfa, + &union_set); + if (BE (mctx->state_log[next_idx] == NULL + && err != REG_NOERROR, 0)) + { + re_node_set_free (&union_set); + return err; + } + } + } +#endif /* RE_ENABLE_I18N */ + if (naccepted + || check_node_accept (mctx, dfa->nodes + cur_node, str_idx)) + { + result = re_node_set_insert (next_nodes, dfa->nexts[cur_node]); + if (BE (result < 0, 0)) + { + re_node_set_free (&union_set); + return REG_ESPACE; + } + } + } + re_node_set_free (&union_set); + return REG_NOERROR; +} + +/* For all the nodes in CUR_NODES, add the epsilon closures of them to + CUR_NODES, however exclude the nodes which are: + - inside the sub expression whose number is EX_SUBEXP, if FL_OPEN. + - out of the sub expression whose number is EX_SUBEXP, if !FL_OPEN. +*/ + +static reg_errcode_t +internal_function +check_arrival_expand_ecl (const re_dfa_t *dfa, re_node_set *cur_nodes, + int ex_subexp, int type) +{ + reg_errcode_t err; + int idx, outside_node; + re_node_set new_nodes; +#ifdef DEBUG + assert (cur_nodes->nelem); +#endif + err = re_node_set_alloc (&new_nodes, cur_nodes->nelem); + if (BE (err != REG_NOERROR, 0)) + return err; + /* Create a new node set NEW_NODES with the nodes which are epsilon + closures of the node in CUR_NODES. */ + + for (idx = 0; idx < cur_nodes->nelem; ++idx) + { + int cur_node = cur_nodes->elems[idx]; + const re_node_set *eclosure = dfa->eclosures + cur_node; + outside_node = find_subexp_node (dfa, eclosure, ex_subexp, type); + if (outside_node == -1) + { + /* There are no problematic nodes, just merge them. */ + err = re_node_set_merge (&new_nodes, eclosure); + if (BE (err != REG_NOERROR, 0)) + { + re_node_set_free (&new_nodes); + return err; + } + } + else + { + /* There are problematic nodes, re-calculate incrementally. */ + err = check_arrival_expand_ecl_sub (dfa, &new_nodes, cur_node, + ex_subexp, type); + if (BE (err != REG_NOERROR, 0)) + { + re_node_set_free (&new_nodes); + return err; + } + } + } + re_node_set_free (cur_nodes); + *cur_nodes = new_nodes; + return REG_NOERROR; +} + +/* Helper function for check_arrival_expand_ecl. + Check incrementally the epsilon closure of TARGET, and if it isn't + problematic append it to DST_NODES. */ + +static reg_errcode_t +internal_function __attribute_warn_unused_result__ +check_arrival_expand_ecl_sub (const re_dfa_t *dfa, re_node_set *dst_nodes, + int target, int ex_subexp, int type) +{ + int cur_node; + for (cur_node = target; !re_node_set_contains (dst_nodes, cur_node);) + { + int err; + + if (dfa->nodes[cur_node].type == type + && dfa->nodes[cur_node].opr.idx == ex_subexp) + { + if (type == OP_CLOSE_SUBEXP) + { + err = re_node_set_insert (dst_nodes, cur_node); + if (BE (err == -1, 0)) + return REG_ESPACE; + } + break; + } + err = re_node_set_insert (dst_nodes, cur_node); + if (BE (err == -1, 0)) + return REG_ESPACE; + if (dfa->edests[cur_node].nelem == 0) + break; + if (dfa->edests[cur_node].nelem == 2) + { + err = check_arrival_expand_ecl_sub (dfa, dst_nodes, + dfa->edests[cur_node].elems[1], + ex_subexp, type); + if (BE (err != REG_NOERROR, 0)) + return err; + } + cur_node = dfa->edests[cur_node].elems[0]; + } + return REG_NOERROR; +} + + +/* For all the back references in the current state, calculate the + destination of the back references by the appropriate entry + in MCTX->BKREF_ENTS. */ + +static reg_errcode_t +internal_function __attribute_warn_unused_result__ +expand_bkref_cache (re_match_context_t *mctx, re_node_set *cur_nodes, + int cur_str, int subexp_num, int type) +{ + const re_dfa_t *const dfa = mctx->dfa; + reg_errcode_t err; + int cache_idx_start = search_cur_bkref_entry (mctx, cur_str); + struct re_backref_cache_entry *ent; + + if (cache_idx_start == -1) + return REG_NOERROR; + + restart: + ent = mctx->bkref_ents + cache_idx_start; + do + { + int to_idx, next_node; + + /* Is this entry ENT is appropriate? */ + if (!re_node_set_contains (cur_nodes, ent->node)) + continue; /* No. */ + + to_idx = cur_str + ent->subexp_to - ent->subexp_from; + /* Calculate the destination of the back reference, and append it + to MCTX->STATE_LOG. */ + if (to_idx == cur_str) + { + /* The backreference did epsilon transit, we must re-check all the + node in the current state. */ + re_node_set new_dests; + reg_errcode_t err2, err3; + next_node = dfa->edests[ent->node].elems[0]; + if (re_node_set_contains (cur_nodes, next_node)) + continue; + err = re_node_set_init_1 (&new_dests, next_node); + err2 = check_arrival_expand_ecl (dfa, &new_dests, subexp_num, type); + err3 = re_node_set_merge (cur_nodes, &new_dests); + re_node_set_free (&new_dests); + if (BE (err != REG_NOERROR || err2 != REG_NOERROR + || err3 != REG_NOERROR, 0)) + { + err = (err != REG_NOERROR ? err + : (err2 != REG_NOERROR ? err2 : err3)); + return err; + } + /* TODO: It is still inefficient... */ + goto restart; + } + else + { + re_node_set union_set; + next_node = dfa->nexts[ent->node]; + if (mctx->state_log[to_idx]) + { + int ret; + if (re_node_set_contains (&mctx->state_log[to_idx]->nodes, + next_node)) + continue; + err = re_node_set_init_copy (&union_set, + &mctx->state_log[to_idx]->nodes); + ret = re_node_set_insert (&union_set, next_node); + if (BE (err != REG_NOERROR || ret < 0, 0)) + { + re_node_set_free (&union_set); + err = err != REG_NOERROR ? err : REG_ESPACE; + return err; + } + } + else + { + err = re_node_set_init_1 (&union_set, next_node); + if (BE (err != REG_NOERROR, 0)) + return err; + } + mctx->state_log[to_idx] = re_acquire_state (&err, dfa, &union_set); + re_node_set_free (&union_set); + if (BE (mctx->state_log[to_idx] == NULL + && err != REG_NOERROR, 0)) + return err; + } + } + while (ent++->more); + return REG_NOERROR; +} + +/* Build transition table for the state. + Return 1 if succeeded, otherwise return NULL. */ + +static int +internal_function +build_trtable (const re_dfa_t *dfa, re_dfastate_t *state) +{ + reg_errcode_t err; + int i, j, ch, need_word_trtable = 0; + bitset_word_t elem, mask; + bool dests_node_malloced = false; + bool dest_states_malloced = false; + int ndests; /* Number of the destination states from `state'. */ + re_dfastate_t **trtable; + re_dfastate_t **dest_states = NULL, **dest_states_word, **dest_states_nl; + re_node_set follows, *dests_node; + bitset_t *dests_ch; + bitset_t acceptable; + + struct dests_alloc + { + re_node_set dests_node[SBC_MAX]; + bitset_t dests_ch[SBC_MAX]; + } *dests_alloc; + + /* We build DFA states which corresponds to the destination nodes + from `state'. `dests_node[i]' represents the nodes which i-th + destination state contains, and `dests_ch[i]' represents the + characters which i-th destination state accepts. */ +#ifdef HAVE_ALLOCA + if (__libc_use_alloca (sizeof (struct dests_alloc))) + dests_alloc = (struct dests_alloc *) alloca (sizeof (struct dests_alloc)); + else +#endif + { + dests_alloc = re_malloc (struct dests_alloc, 1); + if (BE (dests_alloc == NULL, 0)) + return 0; + dests_node_malloced = true; + } + dests_node = dests_alloc->dests_node; + dests_ch = dests_alloc->dests_ch; + + /* Initialize transiton table. */ + state->word_trtable = state->trtable = NULL; + + /* At first, group all nodes belonging to `state' into several + destinations. */ + ndests = group_nodes_into_DFAstates (dfa, state, dests_node, dests_ch); + if (BE (ndests <= 0, 0)) + { + if (dests_node_malloced) + re_free (dests_alloc); + /* Return 0 in case of an error, 1 otherwise. */ + if (ndests == 0) + { + state->trtable = (re_dfastate_t **) + calloc (sizeof (re_dfastate_t *), SBC_MAX); + if (BE (state->trtable == NULL, 0)) + return 0; + return 1; + } + return 0; + } + + err = re_node_set_alloc (&follows, ndests + 1); + if (BE (err != REG_NOERROR, 0)) + goto out_free; + + /* Avoid arithmetic overflow in size calculation. */ + if (BE ((((SIZE_MAX - (sizeof (re_node_set) + sizeof (bitset_t)) * SBC_MAX) + / (3 * sizeof (re_dfastate_t *))) + < ndests), + 0)) + goto out_free; + +#ifdef HAVE_ALLOCA + if (__libc_use_alloca ((sizeof (re_node_set) + sizeof (bitset_t)) * SBC_MAX + + ndests * 3 * sizeof (re_dfastate_t *))) + dest_states = (re_dfastate_t **) + alloca (ndests * 3 * sizeof (re_dfastate_t *)); + else +#endif + { + dest_states = + re_malloc (re_dfastate_t *, ndests * 3); + if (BE (dest_states == NULL, 0)) + { +out_free: + if (dest_states_malloced) + re_free (dest_states); + re_node_set_free (&follows); + for (i = 0; i < ndests; ++i) + re_node_set_free (dests_node + i); + if (dests_node_malloced) + re_free (dests_alloc); + return 0; + } + dest_states_malloced = true; + } + dest_states_word = dest_states + ndests; + dest_states_nl = dest_states_word + ndests; + bitset_empty (acceptable); + + /* Then build the states for all destinations. */ + for (i = 0; i < ndests; ++i) + { + int next_node; + re_node_set_empty (&follows); + /* Merge the follows of this destination states. */ + for (j = 0; j < dests_node[i].nelem; ++j) + { + next_node = dfa->nexts[dests_node[i].elems[j]]; + if (next_node != -1) + { + err = re_node_set_merge (&follows, dfa->eclosures + next_node); + if (BE (err != REG_NOERROR, 0)) + goto out_free; + } + } + dest_states[i] = re_acquire_state_context (&err, dfa, &follows, 0); + if (BE (dest_states[i] == NULL && err != REG_NOERROR, 0)) + goto out_free; + /* If the new state has context constraint, + build appropriate states for these contexts. */ + if (dest_states[i]->has_constraint) + { + dest_states_word[i] = re_acquire_state_context (&err, dfa, &follows, + CONTEXT_WORD); + if (BE (dest_states_word[i] == NULL && err != REG_NOERROR, 0)) + goto out_free; + + if (dest_states[i] != dest_states_word[i] && dfa->mb_cur_max > 1) + need_word_trtable = 1; + + dest_states_nl[i] = re_acquire_state_context (&err, dfa, &follows, + CONTEXT_NEWLINE); + if (BE (dest_states_nl[i] == NULL && err != REG_NOERROR, 0)) + goto out_free; + } + else + { + dest_states_word[i] = dest_states[i]; + dest_states_nl[i] = dest_states[i]; + } + bitset_merge (acceptable, dests_ch[i]); + } + + if (!BE (need_word_trtable, 0)) + { + /* We don't care about whether the following character is a word + character, or we are in a single-byte character set so we can + discern by looking at the character code: allocate a + 256-entry transition table. */ + trtable = state->trtable = + (re_dfastate_t **) calloc (sizeof (re_dfastate_t *), SBC_MAX); + if (BE (trtable == NULL, 0)) + goto out_free; + + /* For all characters ch...: */ + for (i = 0; i < BITSET_WORDS; ++i) + for (ch = i * BITSET_WORD_BITS, elem = acceptable[i], mask = 1; + elem; + mask <<= 1, elem >>= 1, ++ch) + if (BE (elem & 1, 0)) + { + /* There must be exactly one destination which accepts + character ch. See group_nodes_into_DFAstates. */ + for (j = 0; (dests_ch[j][i] & mask) == 0; ++j) + ; + + /* j-th destination accepts the word character ch. */ + if (dfa->word_char[i] & mask) + trtable[ch] = dest_states_word[j]; + else + trtable[ch] = dest_states[j]; + } + } + else + { + /* We care about whether the following character is a word + character, and we are in a multi-byte character set: discern + by looking at the character code: build two 256-entry + transition tables, one starting at trtable[0] and one + starting at trtable[SBC_MAX]. */ + trtable = state->word_trtable = + (re_dfastate_t **) calloc (sizeof (re_dfastate_t *), 2 * SBC_MAX); + if (BE (trtable == NULL, 0)) + goto out_free; + + /* For all characters ch...: */ + for (i = 0; i < BITSET_WORDS; ++i) + for (ch = i * BITSET_WORD_BITS, elem = acceptable[i], mask = 1; + elem; + mask <<= 1, elem >>= 1, ++ch) + if (BE (elem & 1, 0)) + { + /* There must be exactly one destination which accepts + character ch. See group_nodes_into_DFAstates. */ + for (j = 0; (dests_ch[j][i] & mask) == 0; ++j) + ; + + /* j-th destination accepts the word character ch. */ + trtable[ch] = dest_states[j]; + trtable[ch + SBC_MAX] = dest_states_word[j]; + } + } + + /* new line */ + if (bitset_contain (acceptable, NEWLINE_CHAR)) + { + /* The current state accepts newline character. */ + for (j = 0; j < ndests; ++j) + if (bitset_contain (dests_ch[j], NEWLINE_CHAR)) + { + /* k-th destination accepts newline character. */ + trtable[NEWLINE_CHAR] = dest_states_nl[j]; + if (need_word_trtable) + trtable[NEWLINE_CHAR + SBC_MAX] = dest_states_nl[j]; + /* There must be only one destination which accepts + newline. See group_nodes_into_DFAstates. */ + break; + } + } + + if (dest_states_malloced) + re_free (dest_states); + + re_node_set_free (&follows); + for (i = 0; i < ndests; ++i) + re_node_set_free (dests_node + i); + + if (dests_node_malloced) + re_free (dests_alloc); + + return 1; +} + +/* Group all nodes belonging to STATE into several destinations. + Then for all destinations, set the nodes belonging to the destination + to DESTS_NODE[i] and set the characters accepted by the destination + to DEST_CH[i]. This function return the number of destinations. */ + +static int +internal_function +group_nodes_into_DFAstates (const re_dfa_t *dfa, const re_dfastate_t *state, + re_node_set *dests_node, bitset_t *dests_ch) +{ + reg_errcode_t err; + int result; + int i, j, k; + int ndests; /* Number of the destinations from `state'. */ + bitset_t accepts; /* Characters a node can accept. */ + const re_node_set *cur_nodes = &state->nodes; + bitset_empty (accepts); + ndests = 0; + + /* For all the nodes belonging to `state', */ + for (i = 0; i < cur_nodes->nelem; ++i) + { + re_token_t *node = &dfa->nodes[cur_nodes->elems[i]]; + re_token_type_t type = node->type; + unsigned int constraint = node->constraint; + + /* Enumerate all single byte character this node can accept. */ + if (type == CHARACTER) + bitset_set (accepts, node->opr.c); + else if (type == SIMPLE_BRACKET) + { + bitset_merge (accepts, node->opr.sbcset); + } + else if (type == OP_PERIOD) + { +#ifdef RE_ENABLE_I18N + if (dfa->mb_cur_max > 1) + bitset_merge (accepts, dfa->sb_char); + else +#endif + bitset_set_all (accepts); + if (!(dfa->syntax & RE_DOT_NEWLINE)) + bitset_clear (accepts, '\n'); + if (dfa->syntax & RE_DOT_NOT_NULL) + bitset_clear (accepts, '\0'); + } +#ifdef RE_ENABLE_I18N + else if (type == OP_UTF8_PERIOD) + { + memset (accepts, '\xff', sizeof (bitset_t) / 2); + if (!(dfa->syntax & RE_DOT_NEWLINE)) + bitset_clear (accepts, '\n'); + if (dfa->syntax & RE_DOT_NOT_NULL) + bitset_clear (accepts, '\0'); + } +#endif + else + continue; + + /* Check the `accepts' and sift the characters which are not + match it the context. */ + if (constraint) + { + if (constraint & NEXT_NEWLINE_CONSTRAINT) + { + bool accepts_newline = bitset_contain (accepts, NEWLINE_CHAR); + bitset_empty (accepts); + if (accepts_newline) + bitset_set (accepts, NEWLINE_CHAR); + else + continue; + } + if (constraint & NEXT_ENDBUF_CONSTRAINT) + { + bitset_empty (accepts); + continue; + } + + if (constraint & NEXT_WORD_CONSTRAINT) + { + bitset_word_t any_set = 0; + if (type == CHARACTER && !node->word_char) + { + bitset_empty (accepts); + continue; + } +#ifdef RE_ENABLE_I18N + if (dfa->mb_cur_max > 1) + for (j = 0; j < BITSET_WORDS; ++j) + any_set |= (accepts[j] &= (dfa->word_char[j] | ~dfa->sb_char[j])); + else +#endif + for (j = 0; j < BITSET_WORDS; ++j) + any_set |= (accepts[j] &= dfa->word_char[j]); + if (!any_set) + continue; + } + if (constraint & NEXT_NOTWORD_CONSTRAINT) + { + bitset_word_t any_set = 0; + if (type == CHARACTER && node->word_char) + { + bitset_empty (accepts); + continue; + } +#ifdef RE_ENABLE_I18N + if (dfa->mb_cur_max > 1) + for (j = 0; j < BITSET_WORDS; ++j) + any_set |= (accepts[j] &= ~(dfa->word_char[j] & dfa->sb_char[j])); + else +#endif + for (j = 0; j < BITSET_WORDS; ++j) + any_set |= (accepts[j] &= ~dfa->word_char[j]); + if (!any_set) + continue; + } + } + + /* Then divide `accepts' into DFA states, or create a new + state. Above, we make sure that accepts is not empty. */ + for (j = 0; j < ndests; ++j) + { + bitset_t intersec; /* Intersection sets, see below. */ + bitset_t remains; + /* Flags, see below. */ + bitset_word_t has_intersec, not_subset, not_consumed; + + /* Optimization, skip if this state doesn't accept the character. */ + if (type == CHARACTER && !bitset_contain (dests_ch[j], node->opr.c)) + continue; + + /* Enumerate the intersection set of this state and `accepts'. */ + has_intersec = 0; + for (k = 0; k < BITSET_WORDS; ++k) + has_intersec |= intersec[k] = accepts[k] & dests_ch[j][k]; + /* And skip if the intersection set is empty. */ + if (!has_intersec) + continue; + + /* Then check if this state is a subset of `accepts'. */ + not_subset = not_consumed = 0; + for (k = 0; k < BITSET_WORDS; ++k) + { + not_subset |= remains[k] = ~accepts[k] & dests_ch[j][k]; + not_consumed |= accepts[k] = accepts[k] & ~dests_ch[j][k]; + } + + /* If this state isn't a subset of `accepts', create a + new group state, which has the `remains'. */ + if (not_subset) + { + bitset_copy (dests_ch[ndests], remains); + bitset_copy (dests_ch[j], intersec); + err = re_node_set_init_copy (dests_node + ndests, &dests_node[j]); + if (BE (err != REG_NOERROR, 0)) + goto error_return; + ++ndests; + } + + /* Put the position in the current group. */ + result = re_node_set_insert (&dests_node[j], cur_nodes->elems[i]); + if (BE (result < 0, 0)) + goto error_return; + + /* If all characters are consumed, go to next node. */ + if (!not_consumed) + break; + } + /* Some characters remain, create a new group. */ + if (j == ndests) + { + bitset_copy (dests_ch[ndests], accepts); + err = re_node_set_init_1 (dests_node + ndests, cur_nodes->elems[i]); + if (BE (err != REG_NOERROR, 0)) + goto error_return; + ++ndests; + bitset_empty (accepts); + } + } + return ndests; + error_return: + for (j = 0; j < ndests; ++j) + re_node_set_free (dests_node + j); + return -1; +} + +#ifdef RE_ENABLE_I18N +/* Check how many bytes the node `dfa->nodes[node_idx]' accepts. + Return the number of the bytes the node accepts. + STR_IDX is the current index of the input string. + + This function handles the nodes which can accept one character, or + one collating element like '.', '[a-z]', opposite to the other nodes + can only accept one byte. */ + +# ifdef _LIBC +# include <locale/weight.h> +# endif + +static int +internal_function +check_node_accept_bytes (const re_dfa_t *dfa, int node_idx, + const re_string_t *input, int str_idx) +{ + const re_token_t *node = dfa->nodes + node_idx; + int char_len, elem_len; + int i; + wint_t wc; + + if (BE (node->type == OP_UTF8_PERIOD, 0)) + { + unsigned char c = re_string_byte_at (input, str_idx), d; + if (BE (c < 0xc2, 1)) + return 0; + + if (str_idx + 2 > input->len) + return 0; + + d = re_string_byte_at (input, str_idx + 1); + if (c < 0xe0) + return (d < 0x80 || d > 0xbf) ? 0 : 2; + else if (c < 0xf0) + { + char_len = 3; + if (c == 0xe0 && d < 0xa0) + return 0; + } + else if (c < 0xf8) + { + char_len = 4; + if (c == 0xf0 && d < 0x90) + return 0; + } + else if (c < 0xfc) + { + char_len = 5; + if (c == 0xf8 && d < 0x88) + return 0; + } + else if (c < 0xfe) + { + char_len = 6; + if (c == 0xfc && d < 0x84) + return 0; + } + else + return 0; + + if (str_idx + char_len > input->len) + return 0; + + for (i = 1; i < char_len; ++i) + { + d = re_string_byte_at (input, str_idx + i); + if (d < 0x80 || d > 0xbf) + return 0; + } + return char_len; + } + + char_len = re_string_char_size_at (input, str_idx); + if (node->type == OP_PERIOD) + { + if (char_len <= 1) + return 0; + /* FIXME: I don't think this if is needed, as both '\n' + and '\0' are char_len == 1. */ + /* '.' accepts any one character except the following two cases. */ + if ((!(dfa->syntax & RE_DOT_NEWLINE) && + re_string_byte_at (input, str_idx) == '\n') || + ((dfa->syntax & RE_DOT_NOT_NULL) && + re_string_byte_at (input, str_idx) == '\0')) + return 0; + return char_len; + } + + elem_len = re_string_elem_size_at (input, str_idx); + wc = __btowc(*(input->mbs+str_idx)); + if (((elem_len <= 1 && char_len <= 1) || char_len == 0) && (wc != WEOF && wc < SBC_MAX)) + return 0; + + if (node->type == COMPLEX_BRACKET) + { + const re_charset_t *cset = node->opr.mbcset; +# ifdef _LIBC + const unsigned char *pin + = ((const unsigned char *) re_string_get_buffer (input) + str_idx); + int j; + uint32_t nrules; +# endif /* _LIBC */ + int match_len = 0; + wchar_t wc = ((cset->nranges || cset->nchar_classes || cset->nmbchars) + ? re_string_wchar_at (input, str_idx) : 0); + + /* match with multibyte character? */ + for (i = 0; i < cset->nmbchars; ++i) + if (wc == cset->mbchars[i]) + { + match_len = char_len; + goto check_node_accept_bytes_match; + } + /* match with character_class? */ + for (i = 0; i < cset->nchar_classes; ++i) + { + wctype_t wt = cset->char_classes[i]; + if (__iswctype (wc, wt)) + { + match_len = char_len; + goto check_node_accept_bytes_match; + } + } + +# ifdef _LIBC + nrules = _NL_CURRENT_WORD (LC_COLLATE, _NL_COLLATE_NRULES); + if (nrules != 0) + { + unsigned int in_collseq = 0; + const int32_t *table, *indirect; + const unsigned char *weights, *extra; + const char *collseqwc; + + /* match with collating_symbol? */ + if (cset->ncoll_syms) + extra = (const unsigned char *) + _NL_CURRENT (LC_COLLATE, _NL_COLLATE_SYMB_EXTRAMB); + for (i = 0; i < cset->ncoll_syms; ++i) + { + const unsigned char *coll_sym = extra + cset->coll_syms[i]; + /* Compare the length of input collating element and + the length of current collating element. */ + if (*coll_sym != elem_len) + continue; + /* Compare each bytes. */ + for (j = 0; j < *coll_sym; j++) + if (pin[j] != coll_sym[1 + j]) + break; + if (j == *coll_sym) + { + /* Match if every bytes is equal. */ + match_len = j; + goto check_node_accept_bytes_match; + } + } + + if (cset->nranges) + { + if (elem_len <= char_len) + { + collseqwc = _NL_CURRENT (LC_COLLATE, _NL_COLLATE_COLLSEQWC); + in_collseq = __collseq_table_lookup (collseqwc, wc); + } + else + in_collseq = find_collation_sequence_value (pin, elem_len); + } + /* match with range expression? */ + for (i = 0; i < cset->nranges; ++i) + if (cset->range_starts[i] <= in_collseq + && in_collseq <= cset->range_ends[i]) + { + match_len = elem_len; + goto check_node_accept_bytes_match; + } + + /* match with equivalence_class? */ + if (cset->nequiv_classes) + { + const unsigned char *cp = pin; + table = (const int32_t *) + _NL_CURRENT (LC_COLLATE, _NL_COLLATE_TABLEMB); + weights = (const unsigned char *) + _NL_CURRENT (LC_COLLATE, _NL_COLLATE_WEIGHTMB); + extra = (const unsigned char *) + _NL_CURRENT (LC_COLLATE, _NL_COLLATE_EXTRAMB); + indirect = (const int32_t *) + _NL_CURRENT (LC_COLLATE, _NL_COLLATE_INDIRECTMB); + int32_t idx = findidx (table, indirect, extra, &cp, elem_len); + if (idx > 0) + for (i = 0; i < cset->nequiv_classes; ++i) + { + int32_t equiv_class_idx = cset->equiv_classes[i]; + size_t weight_len = weights[idx & 0xffffff]; + if (weight_len == weights[equiv_class_idx & 0xffffff] + && (idx >> 24) == (equiv_class_idx >> 24)) + { + int cnt = 0; + + idx &= 0xffffff; + equiv_class_idx &= 0xffffff; + + while (cnt <= weight_len + && (weights[equiv_class_idx + 1 + cnt] + == weights[idx + 1 + cnt])) + ++cnt; + if (cnt > weight_len) + { + match_len = elem_len; + goto check_node_accept_bytes_match; + } + } + } + } + } + else +# endif /* _LIBC */ + { + /* match with range expression? */ + for (i = 0; i < cset->nranges; ++i) + { + if (cset->range_starts[i] <= wc + && wc <= cset->range_ends[i]) + { + match_len = char_len; + goto check_node_accept_bytes_match; + } + } + } + check_node_accept_bytes_match: + if (!cset->non_match) + return match_len; + else + { + if (match_len > 0) + return 0; + else + return (elem_len > char_len) ? elem_len : char_len; + } + } + return 0; +} + +# ifdef _LIBC +static unsigned int +internal_function +find_collation_sequence_value (const unsigned char *mbs, size_t mbs_len) +{ + uint32_t nrules = _NL_CURRENT_WORD (LC_COLLATE, _NL_COLLATE_NRULES); + if (nrules == 0) + { + if (mbs_len == 1) + { + /* No valid character. Match it as a single byte character. */ + const unsigned char *collseq = (const unsigned char *) + _NL_CURRENT (LC_COLLATE, _NL_COLLATE_COLLSEQMB); + return collseq[mbs[0]]; + } + return UINT_MAX; + } + else + { + int32_t idx; + const unsigned char *extra = (const unsigned char *) + _NL_CURRENT (LC_COLLATE, _NL_COLLATE_SYMB_EXTRAMB); + int32_t extrasize = (const unsigned char *) + _NL_CURRENT (LC_COLLATE, _NL_COLLATE_SYMB_EXTRAMB + 1) - extra; + + for (idx = 0; idx < extrasize;) + { + int mbs_cnt, found = 0; + int32_t elem_mbs_len; + /* Skip the name of collating element name. */ + idx = idx + extra[idx] + 1; + elem_mbs_len = extra[idx++]; + if (mbs_len == elem_mbs_len) + { + for (mbs_cnt = 0; mbs_cnt < elem_mbs_len; ++mbs_cnt) + if (extra[idx + mbs_cnt] != mbs[mbs_cnt]) + break; + if (mbs_cnt == elem_mbs_len) + /* Found the entry. */ + found = 1; + } + /* Skip the byte sequence of the collating element. */ + idx += elem_mbs_len; + /* Adjust for the alignment. */ + idx = (idx + 3) & ~3; + /* Skip the collation sequence value. */ + idx += sizeof (uint32_t); + /* Skip the wide char sequence of the collating element. */ + idx = idx + sizeof (uint32_t) * (*(int32_t *) (extra + idx) + 1); + /* If we found the entry, return the sequence value. */ + if (found) + return *(uint32_t *) (extra + idx); + /* Skip the collation sequence value. */ + idx += sizeof (uint32_t); + } + return UINT_MAX; + } +} +# endif /* _LIBC */ +#endif /* RE_ENABLE_I18N */ + +/* Check whether the node accepts the byte which is IDX-th + byte of the INPUT. */ + +static int +internal_function +check_node_accept (const re_match_context_t *mctx, const re_token_t *node, + int idx) +{ + unsigned char ch; + ch = re_string_byte_at (&mctx->input, idx); + switch (node->type) + { + case CHARACTER: + if (node->opr.c != ch) + return 0; + break; + + case SIMPLE_BRACKET: + if (!bitset_contain (node->opr.sbcset, ch)) + return 0; + break; + +#ifdef RE_ENABLE_I18N + case OP_UTF8_PERIOD: + if (ch >= 0x80) + return 0; + /* FALLTHROUGH */ +#endif + case OP_PERIOD: + if ((ch == '\n' && !(mctx->dfa->syntax & RE_DOT_NEWLINE)) + || (ch == '\0' && (mctx->dfa->syntax & RE_DOT_NOT_NULL))) + return 0; + break; + + default: + return 0; + } + + if (node->constraint) + { + /* The node has constraints. Check whether the current context + satisfies the constraints. */ + unsigned int context = re_string_context_at (&mctx->input, idx, + mctx->eflags); + if (NOT_SATISFY_NEXT_CONSTRAINT (node->constraint, context)) + return 0; + } + + return 1; +} + +/* Extend the buffers, if the buffers have run out. */ + +static reg_errcode_t +internal_function __attribute_warn_unused_result__ +extend_buffers (re_match_context_t *mctx, int min_len) +{ + reg_errcode_t ret; + re_string_t *pstr = &mctx->input; + + /* Avoid overflow. */ + if (BE (INT_MAX / 2 / sizeof (re_dfastate_t *) <= pstr->bufs_len, 0)) + return REG_ESPACE; + + /* Double the lengthes of the buffers, but allocate at least MIN_LEN. */ + ret = re_string_realloc_buffers (pstr, + MAX (min_len, + MIN (pstr->len, pstr->bufs_len * 2))); + if (BE (ret != REG_NOERROR, 0)) + return ret; + + if (mctx->state_log != NULL) + { + /* And double the length of state_log. */ + /* XXX We have no indication of the size of this buffer. If this + allocation fail we have no indication that the state_log array + does not have the right size. */ + re_dfastate_t **new_array = re_realloc (mctx->state_log, re_dfastate_t *, + pstr->bufs_len + 1); + if (BE (new_array == NULL, 0)) + return REG_ESPACE; + mctx->state_log = new_array; + } + + /* Then reconstruct the buffers. */ + if (pstr->icase) + { +#ifdef RE_ENABLE_I18N + if (pstr->mb_cur_max > 1) + { + ret = build_wcs_upper_buffer (pstr); + if (BE (ret != REG_NOERROR, 0)) + return ret; + } + else +#endif /* RE_ENABLE_I18N */ + build_upper_buffer (pstr); + } + else + { +#ifdef RE_ENABLE_I18N + if (pstr->mb_cur_max > 1) + build_wcs_buffer (pstr); + else +#endif /* RE_ENABLE_I18N */ + { + if (pstr->trans != NULL) + re_string_translate_buffer (pstr); + } + } + return REG_NOERROR; +} + + +/* Functions for matching context. */ + +/* Initialize MCTX. */ + +static reg_errcode_t +internal_function __attribute_warn_unused_result__ +match_ctx_init (re_match_context_t *mctx, int eflags, int n) +{ + mctx->eflags = eflags; + mctx->match_last = -1; + if (n > 0) + { + mctx->bkref_ents = re_malloc (struct re_backref_cache_entry, n); + mctx->sub_tops = re_malloc (re_sub_match_top_t *, n); + if (BE (mctx->bkref_ents == NULL || mctx->sub_tops == NULL, 0)) + return REG_ESPACE; + } + /* Already zero-ed by the caller. + else + mctx->bkref_ents = NULL; + mctx->nbkref_ents = 0; + mctx->nsub_tops = 0; */ + mctx->abkref_ents = n; + mctx->max_mb_elem_len = 1; + mctx->asub_tops = n; + return REG_NOERROR; +} + +/* Clean the entries which depend on the current input in MCTX. + This function must be invoked when the matcher changes the start index + of the input, or changes the input string. */ + +static void +internal_function +match_ctx_clean (re_match_context_t *mctx) +{ + int st_idx; + for (st_idx = 0; st_idx < mctx->nsub_tops; ++st_idx) + { + int sl_idx; + re_sub_match_top_t *top = mctx->sub_tops[st_idx]; + for (sl_idx = 0; sl_idx < top->nlasts; ++sl_idx) + { + re_sub_match_last_t *last = top->lasts[sl_idx]; + re_free (last->path.array); + re_free (last); + } + re_free (top->lasts); + if (top->path) + { + re_free (top->path->array); + re_free (top->path); + } + re_free (top); + } + + mctx->nsub_tops = 0; + mctx->nbkref_ents = 0; +} + +/* Free all the memory associated with MCTX. */ + +static void +internal_function +match_ctx_free (re_match_context_t *mctx) +{ + /* First, free all the memory associated with MCTX->SUB_TOPS. */ + match_ctx_clean (mctx); + re_free (mctx->sub_tops); + re_free (mctx->bkref_ents); +} + +/* Add a new backreference entry to MCTX. + Note that we assume that caller never call this function with duplicate + entry, and call with STR_IDX which isn't smaller than any existing entry. +*/ + +static reg_errcode_t +internal_function __attribute_warn_unused_result__ +match_ctx_add_entry (re_match_context_t *mctx, int node, int str_idx, int from, + int to) +{ + if (mctx->nbkref_ents >= mctx->abkref_ents) + { + struct re_backref_cache_entry* new_entry; + new_entry = re_realloc (mctx->bkref_ents, struct re_backref_cache_entry, + mctx->abkref_ents * 2); + if (BE (new_entry == NULL, 0)) + { + re_free (mctx->bkref_ents); + return REG_ESPACE; + } + mctx->bkref_ents = new_entry; + memset (mctx->bkref_ents + mctx->nbkref_ents, '\0', + sizeof (struct re_backref_cache_entry) * mctx->abkref_ents); + mctx->abkref_ents *= 2; + } + if (mctx->nbkref_ents > 0 + && mctx->bkref_ents[mctx->nbkref_ents - 1].str_idx == str_idx) + mctx->bkref_ents[mctx->nbkref_ents - 1].more = 1; + + mctx->bkref_ents[mctx->nbkref_ents].node = node; + mctx->bkref_ents[mctx->nbkref_ents].str_idx = str_idx; + mctx->bkref_ents[mctx->nbkref_ents].subexp_from = from; + mctx->bkref_ents[mctx->nbkref_ents].subexp_to = to; + + /* This is a cache that saves negative results of check_dst_limits_calc_pos. + If bit N is clear, means that this entry won't epsilon-transition to + an OP_OPEN_SUBEXP or OP_CLOSE_SUBEXP for the N+1-th subexpression. If + it is set, check_dst_limits_calc_pos_1 will recurse and try to find one + such node. + + A backreference does not epsilon-transition unless it is empty, so set + to all zeros if FROM != TO. */ + mctx->bkref_ents[mctx->nbkref_ents].eps_reachable_subexps_map + = (from == to ? ~0 : 0); + + mctx->bkref_ents[mctx->nbkref_ents++].more = 0; + if (mctx->max_mb_elem_len < to - from) + mctx->max_mb_elem_len = to - from; + return REG_NOERROR; +} + +/* Search for the first entry which has the same str_idx, or -1 if none is + found. Note that MCTX->BKREF_ENTS is already sorted by MCTX->STR_IDX. */ + +static int +internal_function +search_cur_bkref_entry (const re_match_context_t *mctx, int str_idx) +{ + int left, right, mid, last; + last = right = mctx->nbkref_ents; + for (left = 0; left < right;) + { + mid = (left + right) / 2; + if (mctx->bkref_ents[mid].str_idx < str_idx) + left = mid + 1; + else + right = mid; + } + if (left < last && mctx->bkref_ents[left].str_idx == str_idx) + return left; + else + return -1; +} + +/* Register the node NODE, whose type is OP_OPEN_SUBEXP, and which matches + at STR_IDX. */ + +static reg_errcode_t +internal_function __attribute_warn_unused_result__ +match_ctx_add_subtop (re_match_context_t *mctx, int node, int str_idx) +{ +#ifdef DEBUG + assert (mctx->sub_tops != NULL); + assert (mctx->asub_tops > 0); +#endif + if (BE (mctx->nsub_tops == mctx->asub_tops, 0)) + { + int new_asub_tops = mctx->asub_tops * 2; + re_sub_match_top_t **new_array = re_realloc (mctx->sub_tops, + re_sub_match_top_t *, + new_asub_tops); + if (BE (new_array == NULL, 0)) + return REG_ESPACE; + mctx->sub_tops = new_array; + mctx->asub_tops = new_asub_tops; + } + mctx->sub_tops[mctx->nsub_tops] = calloc (1, sizeof (re_sub_match_top_t)); + if (BE (mctx->sub_tops[mctx->nsub_tops] == NULL, 0)) + return REG_ESPACE; + mctx->sub_tops[mctx->nsub_tops]->node = node; + mctx->sub_tops[mctx->nsub_tops++]->str_idx = str_idx; + return REG_NOERROR; +} + +/* Register the node NODE, whose type is OP_CLOSE_SUBEXP, and which matches + at STR_IDX, whose corresponding OP_OPEN_SUBEXP is SUB_TOP. */ + +static re_sub_match_last_t * +internal_function +match_ctx_add_sublast (re_sub_match_top_t *subtop, int node, int str_idx) +{ + re_sub_match_last_t *new_entry; + if (BE (subtop->nlasts == subtop->alasts, 0)) + { + int new_alasts = 2 * subtop->alasts + 1; + re_sub_match_last_t **new_array = re_realloc (subtop->lasts, + re_sub_match_last_t *, + new_alasts); + if (BE (new_array == NULL, 0)) + return NULL; + subtop->lasts = new_array; + subtop->alasts = new_alasts; + } + new_entry = calloc (1, sizeof (re_sub_match_last_t)); + if (BE (new_entry != NULL, 1)) + { + subtop->lasts[subtop->nlasts] = new_entry; + new_entry->node = node; + new_entry->str_idx = str_idx; + ++subtop->nlasts; + } + return new_entry; +} + +static void +internal_function +sift_ctx_init (re_sift_context_t *sctx, re_dfastate_t **sifted_sts, + re_dfastate_t **limited_sts, int last_node, int last_str_idx) +{ + sctx->sifted_states = sifted_sts; + sctx->limited_states = limited_sts; + sctx->last_node = last_node; + sctx->last_str_idx = last_str_idx; + re_node_set_init_empty (&sctx->limits); +} diff --git a/support/verify.h b/support/verify.h new file mode 100644 index 00000000..5c8381d2 --- /dev/null +++ b/support/verify.h @@ -0,0 +1,279 @@ +/* Compile-time assert-like macros. + + Copyright (C) 2005-2006, 2009-2016 Free Software Foundation, Inc. + + This program is free software: you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation; either version 3 of the License, or + (at your option) any later version. + + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program. If not, see <http://www.gnu.org/licenses/>. */ + +/* Written by Paul Eggert, Bruno Haible, and Jim Meyering. */ + +#ifndef _GL_VERIFY_H +#define _GL_VERIFY_H + + +/* Define _GL_HAVE__STATIC_ASSERT to 1 if _Static_assert works as per C11. + This is supported by GCC 4.6.0 and later, in C mode, and its use + here generates easier-to-read diagnostics when verify (R) fails. + + Define _GL_HAVE_STATIC_ASSERT to 1 if static_assert works as per C++11. + This will likely be supported by future GCC versions, in C++ mode. + + Use this only with GCC. If we were willing to slow 'configure' + down we could also use it with other compilers, but since this + affects only the quality of diagnostics, why bother? */ +#if (4 < __GNUC__ + (6 <= __GNUC_MINOR__) \ + && (201112L <= __STDC_VERSION__ || !defined __STRICT_ANSI__) \ + && !defined __cplusplus) +# define _GL_HAVE__STATIC_ASSERT 1 +#endif +/* The condition (99 < __GNUC__) is temporary, until we know about the + first G++ release that supports static_assert. */ +#if (99 < __GNUC__) && defined __cplusplus +# define _GL_HAVE_STATIC_ASSERT 1 +#endif + +/* FreeBSD 9.1 <sys/cdefs.h>, included by <stddef.h> and lots of other + system headers, defines a conflicting _Static_assert that is no + better than ours; override it. */ +#ifndef _GL_HAVE_STATIC_ASSERT +# include <stddef.h> +# undef _Static_assert +#endif + +/* Each of these macros verifies that its argument R is nonzero. To + be portable, R should be an integer constant expression. Unlike + assert (R), there is no run-time overhead. + + If _Static_assert works, verify (R) uses it directly. Similarly, + _GL_VERIFY_TRUE works by packaging a _Static_assert inside a struct + that is an operand of sizeof. + + The code below uses several ideas for C++ compilers, and for C + compilers that do not support _Static_assert: + + * The first step is ((R) ? 1 : -1). Given an expression R, of + integral or boolean or floating-point type, this yields an + expression of integral type, whose value is later verified to be + constant and nonnegative. + + * Next this expression W is wrapped in a type + struct _gl_verify_type { + unsigned int _gl_verify_error_if_negative: W; + }. + If W is negative, this yields a compile-time error. No compiler can + deal with a bit-field of negative size. + + One might think that an array size check would have the same + effect, that is, that the type struct { unsigned int dummy[W]; } + would work as well. However, inside a function, some compilers + (such as C++ compilers and GNU C) allow local parameters and + variables inside array size expressions. With these compilers, + an array size check would not properly diagnose this misuse of + the verify macro: + + void function (int n) { verify (n < 0); } + + * For the verify macro, the struct _gl_verify_type will need to + somehow be embedded into a declaration. To be portable, this + declaration must declare an object, a constant, a function, or a + typedef name. If the declared entity uses the type directly, + such as in + + struct dummy {...}; + typedef struct {...} dummy; + extern struct {...} *dummy; + extern void dummy (struct {...} *); + extern struct {...} *dummy (void); + + two uses of the verify macro would yield colliding declarations + if the entity names are not disambiguated. A workaround is to + attach the current line number to the entity name: + + #define _GL_CONCAT0(x, y) x##y + #define _GL_CONCAT(x, y) _GL_CONCAT0 (x, y) + extern struct {...} * _GL_CONCAT (dummy, __LINE__); + + But this has the problem that two invocations of verify from + within the same macro would collide, since the __LINE__ value + would be the same for both invocations. (The GCC __COUNTER__ + macro solves this problem, but is not portable.) + + A solution is to use the sizeof operator. It yields a number, + getting rid of the identity of the type. Declarations like + + extern int dummy [sizeof (struct {...})]; + extern void dummy (int [sizeof (struct {...})]); + extern int (*dummy (void)) [sizeof (struct {...})]; + + can be repeated. + + * Should the implementation use a named struct or an unnamed struct? + Which of the following alternatives can be used? + + extern int dummy [sizeof (struct {...})]; + extern int dummy [sizeof (struct _gl_verify_type {...})]; + extern void dummy (int [sizeof (struct {...})]); + extern void dummy (int [sizeof (struct _gl_verify_type {...})]); + extern int (*dummy (void)) [sizeof (struct {...})]; + extern int (*dummy (void)) [sizeof (struct _gl_verify_type {...})]; + + In the second and sixth case, the struct type is exported to the + outer scope; two such declarations therefore collide. GCC warns + about the first, third, and fourth cases. So the only remaining + possibility is the fifth case: + + extern int (*dummy (void)) [sizeof (struct {...})]; + + * GCC warns about duplicate declarations of the dummy function if + -Wredundant-decls is used. GCC 4.3 and later have a builtin + __COUNTER__ macro that can let us generate unique identifiers for + each dummy function, to suppress this warning. + + * This implementation exploits the fact that older versions of GCC, + which do not support _Static_assert, also do not warn about the + last declaration mentioned above. + + * GCC warns if -Wnested-externs is enabled and verify() is used + within a function body; but inside a function, you can always + arrange to use verify_expr() instead. + + * In C++, any struct definition inside sizeof is invalid. + Use a template type to work around the problem. */ + +/* Concatenate two preprocessor tokens. */ +#define _GL_CONCAT(x, y) _GL_CONCAT0 (x, y) +#define _GL_CONCAT0(x, y) x##y + +/* _GL_COUNTER is an integer, preferably one that changes each time we + use it. Use __COUNTER__ if it works, falling back on __LINE__ + otherwise. __LINE__ isn't perfect, but it's better than a + constant. */ +#if defined __COUNTER__ && __COUNTER__ != __COUNTER__ +# define _GL_COUNTER __COUNTER__ +#else +# define _GL_COUNTER __LINE__ +#endif + +/* Generate a symbol with the given prefix, making it unique if + possible. */ +#define _GL_GENSYM(prefix) _GL_CONCAT (prefix, _GL_COUNTER) + +/* Verify requirement R at compile-time, as an integer constant expression + that returns 1. If R is false, fail at compile-time, preferably + with a diagnostic that includes the string-literal DIAGNOSTIC. */ + +#define _GL_VERIFY_TRUE(R, DIAGNOSTIC) \ + (!!sizeof (_GL_VERIFY_TYPE (R, DIAGNOSTIC))) + +#ifdef __cplusplus +# if !GNULIB_defined_struct__gl_verify_type +template <int w> + struct _gl_verify_type { + unsigned int _gl_verify_error_if_negative: w; + }; +# define GNULIB_defined_struct__gl_verify_type 1 +# endif +# define _GL_VERIFY_TYPE(R, DIAGNOSTIC) \ + _gl_verify_type<(R) ? 1 : -1> +#elif defined _GL_HAVE__STATIC_ASSERT +# define _GL_VERIFY_TYPE(R, DIAGNOSTIC) \ + struct { \ + _Static_assert (R, DIAGNOSTIC); \ + int _gl_dummy; \ + } +#else +# define _GL_VERIFY_TYPE(R, DIAGNOSTIC) \ + struct { unsigned int _gl_verify_error_if_negative: (R) ? 1 : -1; } +#endif + +/* Verify requirement R at compile-time, as a declaration without a + trailing ';'. If R is false, fail at compile-time, preferably + with a diagnostic that includes the string-literal DIAGNOSTIC. + + Unfortunately, unlike C11, this implementation must appear as an + ordinary declaration, and cannot appear inside struct { ... }. */ + +#ifdef _GL_HAVE__STATIC_ASSERT +# define _GL_VERIFY _Static_assert +#else +# define _GL_VERIFY(R, DIAGNOSTIC) \ + extern int (*_GL_GENSYM (_gl_verify_function) (void)) \ + [_GL_VERIFY_TRUE (R, DIAGNOSTIC)] +#endif + +/* _GL_STATIC_ASSERT_H is defined if this code is copied into assert.h. */ +#ifdef _GL_STATIC_ASSERT_H +# if !defined _GL_HAVE__STATIC_ASSERT && !defined _Static_assert +# define _Static_assert(R, DIAGNOSTIC) _GL_VERIFY (R, DIAGNOSTIC) +# endif +# if !defined _GL_HAVE_STATIC_ASSERT && !defined static_assert +# define static_assert _Static_assert /* C11 requires this #define. */ +# endif +#endif + +/* @assert.h omit start@ */ + +/* Each of these macros verifies that its argument R is nonzero. To + be portable, R should be an integer constant expression. Unlike + assert (R), there is no run-time overhead. + + There are two macros, since no single macro can be used in all + contexts in C. verify_true (R) is for scalar contexts, including + integer constant expression contexts. verify (R) is for declaration + contexts, e.g., the top level. */ + +/* Verify requirement R at compile-time, as an integer constant expression. + Return 1. This is equivalent to verify_expr (R, 1). + + verify_true is obsolescent; please use verify_expr instead. */ + +#define verify_true(R) _GL_VERIFY_TRUE (R, "verify_true (" #R ")") + +/* Verify requirement R at compile-time. Return the value of the + expression E. */ + +#define verify_expr(R, E) \ + (_GL_VERIFY_TRUE (R, "verify_expr (" #R ", " #E ")") ? (E) : (E)) + +/* Verify requirement R at compile-time, as a declaration without a + trailing ';'. */ + +#define verify(R) _GL_VERIFY (R, "verify (" #R ")") + +#ifndef __has_builtin +# define __has_builtin(x) 0 +#endif + +/* Assume that R always holds. This lets the compiler optimize + accordingly. R should not have side-effects; it may or may not be + evaluated. Behavior is undefined if R is false. */ + +#if (__has_builtin (__builtin_unreachable) \ + || 4 < __GNUC__ + (5 <= __GNUC_MINOR__)) +# define assume(R) ((R) ? (void) 0 : __builtin_unreachable ()) +#elif 1200 <= _MSC_VER +# define assume(R) __assume (R) +#elif ((defined GCC_LINT || defined lint) \ + && (__has_builtin (__builtin_trap) \ + || 3 < __GNUC__ + (3 < __GNUC_MINOR__ + (4 <= __GNUC_PATCHLEVEL__)))) + /* Doing it this way helps various packages when configured with + --enable-gcc-warnings, which compiles with -Dlint. It's nicer + when 'assume' silences warnings even with older GCCs. */ +# define assume(R) ((R) ? (void) 0 : __builtin_trap ()) +#else +# define assume(R) ((void) (0 && (R))) +#endif + +/* @assert.h omit end@ */ + +#endif diff --git a/support/xalloc.h b/support/xalloc.h new file mode 100644 index 00000000..0d169cf9 --- /dev/null +++ b/support/xalloc.h @@ -0,0 +1,362 @@ +/* xalloc.h -- malloc with out-of-memory checking + + Copyright (C) 1990, 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999, + 2000, 2003, 2004, 2006, 2007, 2008, 2009, 2010 Free Software Foundation, + Inc. + + This program is free software: you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation; either version 3 of the License, or + (at your option) any later version. + + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program. If not, see <http://www.gnu.org/licenses/>. */ + +#ifndef XALLOC_H_ +# define XALLOC_H_ + +# include <stddef.h> + + +# ifdef __cplusplus +extern "C" { +# endif + + +# ifndef __attribute__ +# if __GNUC__ < 2 || (__GNUC__ == 2 && __GNUC_MINOR__ < 8) +# define __attribute__(x) +# endif +# endif + +# ifndef ATTRIBUTE_NORETURN +# define ATTRIBUTE_NORETURN __attribute__ ((__noreturn__)) +# endif + +# ifndef ATTRIBUTE_MALLOC +# if __GNUC__ >= 3 +# define ATTRIBUTE_MALLOC __attribute__ ((__malloc__)) +# else +# define ATTRIBUTE_MALLOC +# endif +# endif + +/* This function is always triggered when memory is exhausted. + It must be defined by the application, either explicitly + or by using gnulib's xalloc-die module. This is the + function to call when one wants the program to die because of a + memory allocation failure. */ +extern void xalloc_die (void) ATTRIBUTE_NORETURN; + +void *xmalloc (size_t s) ATTRIBUTE_MALLOC; +void *xzalloc (size_t s) ATTRIBUTE_MALLOC; +void *xcalloc (size_t n, size_t s) ATTRIBUTE_MALLOC; +void *xrealloc (void *p, size_t s); +void *x2realloc (void *p, size_t *pn); +void *xmemdup (void const *p, size_t s) ATTRIBUTE_MALLOC; +char *xstrdup (char const *str) ATTRIBUTE_MALLOC; + +/* Return 1 if an array of N objects, each of size S, cannot exist due + to size arithmetic overflow. S must be positive and N must be + nonnegative. This is a macro, not an inline function, so that it + works correctly even when SIZE_MAX < N. + + By gnulib convention, SIZE_MAX represents overflow in size + calculations, so the conservative dividend to use here is + SIZE_MAX - 1, since SIZE_MAX might represent an overflowed value. + However, malloc (SIZE_MAX) fails on all known hosts where + sizeof (ptrdiff_t) <= sizeof (size_t), so do not bother to test for + exactly-SIZE_MAX allocations on such hosts; this avoids a test and + branch when S is known to be 1. */ +# define xalloc_oversized(n, s) \ + ((size_t) (sizeof (ptrdiff_t) <= sizeof (size_t) ? -1 : -2) / (s) < (n)) + + +/* In the following macros, T must be an elementary or structure/union or + typedef'ed type, or a pointer to such a type. To apply one of the + following macros to a function pointer or array type, you need to typedef + it first and use the typedef name. */ + +/* Allocate an object of type T dynamically, with error checking. */ +/* extern t *XMALLOC (typename t); */ +# define XMALLOC(t) ((t *) xmalloc (sizeof (t))) + +/* Allocate memory for N elements of type T, with error checking. */ +/* extern t *XNMALLOC (size_t n, typename t); */ +# define XNMALLOC(n, t) \ + ((t *) (sizeof (t) == 1 ? xmalloc (n) : xnmalloc (n, sizeof (t)))) + +/* Allocate an object of type T dynamically, with error checking, + and zero it. */ +/* extern t *XZALLOC (typename t); */ +# define XZALLOC(t) ((t *) xzalloc (sizeof (t))) + +/* Allocate memory for N elements of type T, with error checking, + and zero it. */ +/* extern t *XCALLOC (size_t n, typename t); */ +# define XCALLOC(n, t) \ + ((t *) (sizeof (t) == 1 ? xzalloc (n) : xcalloc (n, sizeof (t)))) + +/* + * Gawk uses this file only to keep dfa.c happy. + * We're therefore safe in manually defining HAVE_INLINE to + * make the !@#$%^&*() thing just work. + */ +#ifdef GAWK +#define HAVE_INLINE 1 /* so there. nyah, nyah, nyah. */ +#endif + +# if HAVE_INLINE +# define static_inline static inline +# else +void *xnmalloc (size_t n, size_t s) ATTRIBUTE_MALLOC; +void *xnrealloc (void *p, size_t n, size_t s); +void *x2nrealloc (void *p, size_t *pn, size_t s); +char *xcharalloc (size_t n) ATTRIBUTE_MALLOC; +# endif + +# ifdef static_inline + +/* Allocate an array of N objects, each with S bytes of memory, + dynamically, with error checking. S must be nonzero. */ + +static_inline void *xnmalloc (size_t n, size_t s) ATTRIBUTE_MALLOC; +static_inline void * +xnmalloc (size_t n, size_t s) +{ + if (xalloc_oversized (n, s)) + xalloc_die (); + return xmalloc (n * s); +} + +#ifdef GAWK +#include <errno.h> +extern void r_fatal(const char *msg, ...) ATTRIBUTE_NORETURN ; + +/* Allocate an array of N objects, each with S bytes of memory, + dynamically, with error checking. S must be nonzero. + Clear the contents afterwards. */ + +void * +xcalloc(size_t nmemb, size_t size) +{ + void *p = xmalloc (nmemb * size); + memset(p, '\0', nmemb * size); + return p; +} + +/* Reallocate a pointer to a new size, with error checking. */ + +void * +xrealloc(void *p, size_t size) +{ + void *new_p = realloc(p, size); + if (new_p == 0) + xalloc_die (); + + return new_p; +} + +/* xalloc_die --- fatal error message when malloc fails, needed by dfa.c */ + +void +xalloc_die (void) +{ + r_fatal(_("xalloc: malloc failed: %s"), strerror(errno)); +} + +/* Clone an object P of size S, with error checking. There's no need + for xnmemdup (P, N, S), since xmemdup (P, N * S) works without any + need for an arithmetic overflow check. */ + +void * +xmemdup (void const *p, size_t s) +{ + return memcpy (xmalloc (s), p, s); +} + +/* xstrdup --- strdup and die if fails */ +char *xstrdup(const char *s) +{ + char *p; + int l; + + if (s == NULL) + r_fatal(_("xstrdup: null parameter")); + + l = strlen(s); + p = xmemdup(s, l + 1); + p[l] = '\0'; + + return p; +} +#endif + +/* Change the size of an allocated block of memory P to an array of N + objects each of S bytes, with error checking. S must be nonzero. */ + +static_inline void * +xnrealloc (void *p, size_t n, size_t s) +{ + if (xalloc_oversized (n, s)) + xalloc_die (); + return xrealloc (p, n * s); +} + +/* If P is null, allocate a block of at least *PN such objects; + otherwise, reallocate P so that it contains more than *PN objects + each of S bytes. *PN must be nonzero unless P is null, and S must + be nonzero. Set *PN to the new number of objects, and return the + pointer to the new block. *PN is never set to zero, and the + returned pointer is never null. + + Repeated reallocations are guaranteed to make progress, either by + allocating an initial block with a nonzero size, or by allocating a + larger block. + + In the following implementation, nonzero sizes are increased by a + factor of approximately 1.5 so that repeated reallocations have + O(N) overall cost rather than O(N**2) cost, but the + specification for this function does not guarantee that rate. + + Here is an example of use: + + int *p = NULL; + size_t used = 0; + size_t allocated = 0; + + void + append_int (int value) + { + if (used == allocated) + p = x2nrealloc (p, &allocated, sizeof *p); + p[used++] = value; + } + + This causes x2nrealloc to allocate a block of some nonzero size the + first time it is called. + + To have finer-grained control over the initial size, set *PN to a + nonzero value before calling this function with P == NULL. For + example: + + int *p = NULL; + size_t used = 0; + size_t allocated = 0; + size_t allocated1 = 1000; + + void + append_int (int value) + { + if (used == allocated) + { + p = x2nrealloc (p, &allocated1, sizeof *p); + allocated = allocated1; + } + p[used++] = value; + } + + */ + +static_inline void * +x2nrealloc (void *p, size_t *pn, size_t s) +{ + size_t n = *pn; + + if (! p) + { + if (! n) + { + /* The approximate size to use for initial small allocation + requests, when the invoking code specifies an old size of + zero. 64 bytes is the largest "small" request for the + GNU C library malloc. */ + enum { DEFAULT_MXFAST = 64 * sizeof (size_t) / 4 }; + + n = DEFAULT_MXFAST / s; + n += !n; + } + } + else + { + /* Set N = ceil (1.5 * N) so that progress is made if N == 1. + Check for overflow, so that N * S stays in size_t range. + The check is slightly conservative, but an exact check isn't + worth the trouble. */ + if ((size_t) -1 / 3 * 2 / s <= n) + xalloc_die (); + n += n / 2 + 1; + } + + *pn = n; + return xrealloc (p, n * s); +} + +/* Return a pointer to a new buffer of N bytes. This is like xmalloc, + except it returns char *. */ + +static_inline char *xcharalloc (size_t n) ATTRIBUTE_MALLOC; +static_inline char * +xcharalloc (size_t n) +{ + return XNMALLOC (n, char); +} + +/* Allocate S bytes of zeroed memory dynamically, with error checking. + There's no need for xnzalloc (N, S), since it would be equivalent + to xcalloc (N, S). */ + +inline void * +xzalloc (size_t s) +{ + return memset (xmalloc (s), 0, s); +} + +# endif + +# ifdef __cplusplus +} + +/* C++ does not allow conversions from void * to other pointer types + without a cast. Use templates to work around the problem when + possible. */ + +template <typename T> inline T * +xrealloc (T *p, size_t s) +{ + return (T *) xrealloc ((void *) p, s); +} + +template <typename T> inline T * +xnrealloc (T *p, size_t n, size_t s) +{ + return (T *) xnrealloc ((void *) p, n, s); +} + +template <typename T> inline T * +x2realloc (T *p, size_t *pn) +{ + return (T *) x2realloc ((void *) p, pn); +} + +template <typename T> inline T * +x2nrealloc (T *p, size_t *pn, size_t s) +{ + return (T *) x2nrealloc ((void *) p, pn, s); +} + +template <typename T> inline T * +xmemdup (T const *p, size_t s) +{ + return (T *) xmemdup ((void const *) p, s); +} + +# endif + + +#endif /* !XALLOC_H_ */ |