diff options
| -rw-r--r-- | Makefile | 4 | ||||
| -rw-r--r-- | compat/regex/regcomp.c | 3884 | ||||
| -rw-r--r-- | compat/regex/regex.c | 5001 | ||||
| -rw-r--r-- | compat/regex/regex.h | 462 | ||||
| -rw-r--r-- | compat/regex/regex_internal.c | 1744 | ||||
| -rw-r--r-- | compat/regex/regex_internal.h | 810 | ||||
| -rw-r--r-- | compat/regex/regexec.c | 4369 | ||||
| -rw-r--r-- | config.mak.in | 1 | ||||
| -rw-r--r-- | configure.ac | 21 | ||||
| -rwxr-xr-x | t/t7008-grep-binary.sh | 2 |
10 files changed, 11193 insertions, 5105 deletions
@@ -1900,6 +1900,10 @@ ifdef NO_EXPAT http-walker.s http-walker.o: EXTRA_CPPFLAGS = -DNO_EXPAT endif +ifdef NO_REGEX +compat/regex/regex.o: EXTRA_CPPFLAGS = -DGAWK -DNO_MBSUPPORT +endif + git-%$X: %.o $(GITLIBS) $(QUIET_LINK)$(CC) $(ALL_CFLAGS) -o $@ $(ALL_LDFLAGS) $(filter %.o,$^) $(LIBS) diff --git a/compat/regex/regcomp.c b/compat/regex/regcomp.c new file mode 100644 index 0000000000..8c96ed942c --- /dev/null +++ b/compat/regex/regcomp.c @@ -0,0 +1,3884 @@ +/* Extended regular expression matching and search library. + Copyright (C) 2002-2007,2009,2010 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, write to the Free + Software Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA + 02110-1301 USA. */ + +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? */ + +const char __re_error_msgid[] attribute_hidden = + { +#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 */ + }; + +const size_t __re_error_msgid_idx[] attribute_hidden = + { + 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. */ + + +#ifdef ZOS_USS + +/* For ZOS USS we must define btowc */ + +wchar_t +btowc (int c) +{ + wchar_t wtmp[2]; + char tmp[2]; + + tmp[0] = c; + tmp[1] = 0; + + mbtowc (wtmp, tmp, 1); + 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, int 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) +{ + volatile 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 = re_malloc (unsigned char, dfa->mb_cur_max), *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]); + re_free (buf); + } +#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)) + { + memcpy (errbuf, msg, errbuf_size - 1); + errbuf[errbuf_size - 1] = 0; + } + 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 (s) + 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, 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 + + /* strcasecmp isn't a standard interface. brute force check */ +#if 0 + if (strcasecmp (codeset_name, "UTF-8") == 0 + || strcasecmp (codeset_name, "UTF8") == 0) + dfa->is_utf8 = 1; +#else + if ( (codeset_name[0] == 'U' || codeset_name[0] == 'u') + && (codeset_name[1] == 'T' || codeset_name[1] == 't') + && (codeset_name[2] == 'F' || codeset_name[2] == 'f') + && (codeset_name[3] == '-' + ? codeset_name[4] == '8' && codeset_name[5] == '\0' + : codeset_name[3] == '8' && codeset_name[4] == '\0')) + dfa->is_utf8 = 1; +#endif + + /* 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; + 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; + + dfa->subexp_map = re_malloc (int, preg->re_nsub); + 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 clsoure + 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 of the node has another constraint, add 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 is occured, 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; + 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)) + { + branch = parse_branch (regexp, preg, token, syntax, nest, err); + if (BE (*err != REG_NOERROR && branch == NULL, 0)) + return NULL; + } + 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)) + { + return NULL; + } + if (tree != NULL && exp != NULL) + { + tree = create_tree (dfa, tree, exp, CONCAT); + if (tree == NULL) + { + *err = REG_ESPACE; + return NULL; + } + } + 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) + { + tree = parse_dup_op (tree, regexp, dfa, token, syntax, err); + if (BE (*err != REG_NOERROR && tree == NULL, 0)) + return NULL; + /* In BRE consecutive duplications are not allowed. */ + if ((syntax & RE_CONTEXT_INVALID_DUP) + && (token->type == OP_DUP_ASTERISK + || token->type == OP_OPEN_DUP_NUM)) + { + *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)) + *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); +#ifndef RE_TOKEN_INIT_BUG + re_token_t start_token = *token; +#else + re_token_t start_token; + + memcpy ((void *) &start_token, (void *) token, sizeof start_token); +#endif + + 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); + 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 sinse we may + update it. */ + +static reg_errcode_t +internal_function +# ifdef RE_ENABLE_I18N +build_range_exp (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 (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; + wchar_t cmp_buf[6] = {L'\0', L'\0', L'\0', L'\0', L'\0', L'\0'}; + + 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; + cmp_buf[0] = start_wc; + cmp_buf[4] = end_wc; + if (wcscoll (cmp_buf, cmp_buf + 4) > 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)) + 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) + { + cmp_buf[2] = wc; + if (wcscoll (cmp_buf, cmp_buf + 2) <= 0 + && wcscoll (cmp_buf + 2, cmp_buf + 4) <= 0) + 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 environement. + Seek the collating symbol entry correspondings to NAME. + Return the index of the symbol in the SYMB_TABLE. */ + + auto inline int32_t + __attribute ((always_inline)) + seek_collating_symbol_entry (name, name_len) + const unsigned char *name; + size_t name_len; + { + int32_t hash = elem_hash ((const char *) name, name_len); + int32_t elem = hash % table_size; + if (symb_table[2 * elem] != 0) + { + int32_t second = hash % (table_size - 2) + 1; + + do + { + /* First compare the hashing value. */ + if (symb_table[2 * elem] == hash + /* Compare the length of the name. */ + && name_len == extra[symb_table[2 * elem + 1]] + /* Compare the name. */ + && memcmp (name, &extra[symb_table[2 * elem + 1] + 1], + name_len) == 0) + { + /* Yep, this is the entry. */ + break; + } + + /* Next entry. */ + elem += second; + } + while (symb_table[2 * elem] != 0); + } + return elem; + } + + /* 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 (br_elem) + 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 (symb_table[2 * elem] != 0) + { + /* 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 (symb_table[2 * elem] == 0 && 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 environement. + 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 sinse we may + update it. */ + + auto inline reg_errcode_t + __attribute ((always_inline)) + build_range_exp (sbcset, mbcset, range_alloc, start_elem, end_elem) + re_charset_t *mbcset; + int *range_alloc; + bitset_t sbcset; + bracket_elem_t *start_elem, *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 environement. + 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 sinse we may update it. */ + + auto inline reg_errcode_t + __attribute ((always_inline)) + build_collating_symbol (sbcset, mbcset, coll_sym_alloc, name) + re_charset_t *mbcset; + int *coll_sym_alloc; + bitset_t sbcset; + 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 (symb_table[2 * elem] != 0) + { + /* We found the entry. */ + idx = symb_table[2 * elem + 1]; + /* Skip the name of collating element name. */ + idx += 1 + extra[idx]; + } + else if (symb_table[2 * elem] == 0 && 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 */ + { + *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; + 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; + 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 (sbcset, mbcset, &range_alloc, + &start_elem, &end_elem); +#else +# ifdef RE_ENABLE_I18N + *err = build_range_exp (sbcset, + dfa->mb_cur_max > 1 ? mbcset : NULL, + &range_alloc, &start_elem, &end_elem); +# else + *err = build_range_exp (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 sinse 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; + /* This #include defines a local function! */ +# include <locale/weight.h> + /* 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 (&cp); + if (BE (idx1 == 0 || cp < name + strlen ((const char *) name), 0)) + /* This isn't a valid character. */ + return REG_ECOLLATE; + + /* Build single byte matcing 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 (&cp); +/* + 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 sinse 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 */ + { + *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. */ + 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; + 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/compat/regex/regex.c b/compat/regex/regex.c index be851fc502..3dd8dfa01f 100644 --- a/compat/regex/regex.c +++ b/compat/regex/regex.c @@ -1,4924 +1,87 @@ -/* Extended regular expression matching and search library, - version 0.12. - (Implements POSIX draft P10003.2/D11.2, except for - internationalization features.) +/* Extended regular expression matching and search library. + Copyright (C) 2002, 2003, 2005 Free Software Foundation, Inc. + This file is part of the GNU C Library. + Contributed by Isamu Hasegawa <isamu@yamato.ibm.com>. - Copyright (C) 1993 Free Software Foundation, Inc. + 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. - 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 2, or (at your option) - any later version. - - This program is distributed in the hope that it will be useful, + 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 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., 675 Mass Ave, Cambridge, MA 02139, USA. */ - -/* AIX requires this to be the first thing in the file. */ -#if defined (_AIX) && !defined (REGEX_MALLOC) - #pragma alloca -#endif - -#define _GNU_SOURCE - -/* We need this for `regex.h', and perhaps for the Emacs include files. */ -#include <sys/types.h> - -/* We used to test for `BSTRING' here, but only GCC and Emacs define - `BSTRING', as far as I know, and neither of them use this code. */ -#include <string.h> -#ifndef bcmp -#define bcmp(s1, s2, n) memcmp ((s1), (s2), (n)) -#endif -#ifndef bcopy -#define bcopy(s, d, n) memcpy ((d), (s), (n)) -#endif -#ifndef bzero -#define bzero(s, n) memset ((s), 0, (n)) -#endif - -#include <stdlib.h> - - -/* Define the syntax stuff for \<, \>, etc. */ - -/* This must be nonzero for the wordchar and notwordchar pattern - commands in re_match_2. */ -#ifndef Sword -#define Sword 1 -#endif - -#ifdef SYNTAX_TABLE - -extern char *re_syntax_table; - -#else /* not SYNTAX_TABLE */ - -/* How many characters in the character set. */ -#define CHAR_SET_SIZE 256 - -static char re_syntax_table[CHAR_SET_SIZE]; - -static void -init_syntax_once () -{ - register int c; - static int done = 0; - - if (done) - return; - - bzero (re_syntax_table, sizeof re_syntax_table); - - for (c = 'a'; c <= 'z'; c++) - re_syntax_table[c] = Sword; - - for (c = 'A'; c <= 'Z'; c++) - re_syntax_table[c] = Sword; - - for (c = '0'; c <= '9'; c++) - re_syntax_table[c] = Sword; - - re_syntax_table['_'] = Sword; - - done = 1; -} - -#endif /* not SYNTAX_TABLE */ - -#define SYNTAX(c) re_syntax_table[c] - - -/* Get the interface, including the syntax bits. */ -#include "regex.h" - -/* isalpha etc. are used for the character classes. */ -#include <ctype.h> - -#ifndef isascii -#define isascii(c) 1 -#endif - -#ifdef isblank -#define ISBLANK(c) (isascii (c) && isblank (c)) -#else -#define ISBLANK(c) ((c) == ' ' || (c) == '\t') -#endif -#ifdef isgraph -#define ISGRAPH(c) (isascii (c) && isgraph (c)) -#else -#define ISGRAPH(c) (isascii (c) && isprint (c) && !isspace (c)) -#endif - -#define ISPRINT(c) (isascii (c) && isprint (c)) -#define ISDIGIT(c) (isascii (c) && isdigit (c)) -#define ISALNUM(c) (isascii (c) && isalnum (c)) -#define ISALPHA(c) (isascii (c) && isalpha (c)) -#define ISCNTRL(c) (isascii (c) && iscntrl (c)) -#define ISLOWER(c) (isascii (c) && islower (c)) -#define ISPUNCT(c) (isascii (c) && ispunct (c)) -#define ISSPACE(c) (isascii (c) && isspace (c)) -#define ISUPPER(c) (isascii (c) && isupper (c)) -#define ISXDIGIT(c) (isascii (c) && isxdigit (c)) - -#ifndef NULL -#define NULL 0 -#endif - -/* We remove any previous definition of `SIGN_EXTEND_CHAR', - since ours (we hope) works properly with all combinations of - machines, compilers, `char' and `unsigned char' argument types. - (Per Bothner suggested the basic approach.) */ -#undef SIGN_EXTEND_CHAR -#if __STDC__ -#define SIGN_EXTEND_CHAR(c) ((signed char) (c)) -#else /* not __STDC__ */ -/* As in Harbison and Steele. */ -#define SIGN_EXTEND_CHAR(c) ((((unsigned char) (c)) ^ 128) - 128) -#endif - -/* Should we use malloc or alloca? If REGEX_MALLOC is not defined, we - use `alloca' instead of `malloc'. This is because using malloc in - re_search* or re_match* could cause memory leaks when C-g is used in - Emacs; also, malloc is slower and causes storage fragmentation. On - the other hand, malloc is more portable, and easier to debug. - - Because we sometimes use alloca, some routines have to be macros, - not functions -- `alloca'-allocated space disappears at the end of the - function it is called in. */ - -#ifdef REGEX_MALLOC - -#define REGEX_ALLOCATE malloc -#define REGEX_REALLOCATE(source, osize, nsize) realloc (source, nsize) - -#else /* not REGEX_MALLOC */ - -/* Emacs already defines alloca, sometimes. */ -#ifndef alloca - -/* Make alloca work the best possible way. */ -#ifdef __GNUC__ -#define alloca __builtin_alloca -#else /* not __GNUC__ */ -#if HAVE_ALLOCA_H -#include <alloca.h> -#else /* not __GNUC__ or HAVE_ALLOCA_H */ -#ifndef _AIX /* Already did AIX, up at the top. */ -char *alloca (); -#endif /* not _AIX */ -#endif /* not HAVE_ALLOCA_H */ -#endif /* not __GNUC__ */ - -#endif /* not alloca */ - -#define REGEX_ALLOCATE alloca - -/* Assumes a `char *destination' variable. */ -#define REGEX_REALLOCATE(source, osize, nsize) \ - (destination = (char *) alloca (nsize), \ - bcopy (source, destination, osize), \ - destination) - -#endif /* not REGEX_MALLOC */ - - -/* True if `size1' is non-NULL and PTR is pointing anywhere inside - `string1' or just past its end. This works if PTR is NULL, which is - a good thing. */ -#define FIRST_STRING_P(ptr) \ - (size1 && string1 <= (ptr) && (ptr) <= string1 + size1) - -/* (Re)Allocate N items of type T using malloc, or fail. */ -#define TALLOC(n, t) ((t *) malloc ((n) * sizeof (t))) -#define RETALLOC(addr, n, t) ((addr) = (t *) realloc (addr, (n) * sizeof (t))) -#define REGEX_TALLOC(n, t) ((t *) REGEX_ALLOCATE ((n) * sizeof (t))) - -#define BYTEWIDTH 8 /* In bits. */ - -#define STREQ(s1, s2) ((strcmp (s1, s2) == 0)) - -#define MAX(a, b) ((a) > (b) ? (a) : (b)) -#define MIN(a, b) ((a) < (b) ? (a) : (b)) - -typedef char boolean; -#define false 0 -#define true 1 - -/* These are the command codes that appear in compiled regular - expressions. Some opcodes are followed by argument bytes. A - command code can specify any interpretation whatsoever for its - arguments. Zero bytes may appear in the compiled regular expression. - - The value of `exactn' is needed in search.c (search_buffer) in Emacs. - So regex.h defines a symbol `RE_EXACTN_VALUE' to be 1; the value of - `exactn' we use here must also be 1. */ - -typedef enum -{ - no_op = 0, - - /* Followed by one byte giving n, then by n literal bytes. */ - exactn = 1, - - /* Matches any (more or less) character. */ - anychar, - - /* Matches any one char belonging to specified set. First - following byte is number of bitmap bytes. Then come bytes - for a bitmap saying which chars are in. Bits in each byte - are ordered low-bit-first. A character is in the set if its - bit is 1. A character too large to have a bit in the map is - automatically not in the set. */ - charset, - - /* Same parameters as charset, but match any character that is - not one of those specified. */ - charset_not, - - /* Start remembering the text that is matched, for storing in a - register. Followed by one byte with the register number, in - the range 0 to one less than the pattern buffer's re_nsub - field. Then followed by one byte with the number of groups - inner to this one. (This last has to be part of the - start_memory only because we need it in the on_failure_jump - of re_match_2.) */ - start_memory, - - /* Stop remembering the text that is matched and store it in a - memory register. Followed by one byte with the register - number, in the range 0 to one less than `re_nsub' in the - pattern buffer, and one byte with the number of inner groups, - just like `start_memory'. (We need the number of inner - groups here because we don't have any easy way of finding the - corresponding start_memory when we're at a stop_memory.) */ - stop_memory, - - /* Match a duplicate of something remembered. Followed by one - byte containing the register number. */ - duplicate, - - /* Fail unless at beginning of line. */ - begline, - - /* Fail unless at end of line. */ - endline, - - /* Succeeds if at beginning of buffer (if emacs) or at beginning - of string to be matched (if not). */ - begbuf, - - /* Analogously, for end of buffer/string. */ - endbuf, - - /* Followed by two byte relative address to which to jump. */ - jump, - - /* Same as jump, but marks the end of an alternative. */ - jump_past_alt, - - /* Followed by two-byte relative address of place to resume at - in case of failure. */ - on_failure_jump, - - /* Like on_failure_jump, but pushes a placeholder instead of the - current string position when executed. */ - on_failure_keep_string_jump, - - /* Throw away latest failure point and then jump to following - two-byte relative address. */ - pop_failure_jump, - - /* Change to pop_failure_jump if know won't have to backtrack to - match; otherwise change to jump. This is used to jump - back to the beginning of a repeat. If what follows this jump - clearly won't match what the repeat does, such that we can be - sure that there is no use backtracking out of repetitions - already matched, then we change it to a pop_failure_jump. - Followed by two-byte address. */ - maybe_pop_jump, - - /* Jump to following two-byte address, and push a dummy failure - point. This failure point will be thrown away if an attempt - is made to use it for a failure. A `+' construct makes this - before the first repeat. Also used as an intermediary kind - of jump when compiling an alternative. */ - dummy_failure_jump, - - /* Push a dummy failure point and continue. Used at the end of - alternatives. */ - push_dummy_failure, - - /* Followed by two-byte relative address and two-byte number n. - After matching N times, jump to the address upon failure. */ - succeed_n, - - /* Followed by two-byte relative address, and two-byte number n. - Jump to the address N times, then fail. */ - jump_n, - - /* Set the following two-byte relative address to the - subsequent two-byte number. The address *includes* the two - bytes of number. */ - set_number_at, - - wordchar, /* Matches any word-constituent character. */ - notwordchar, /* Matches any char that is not a word-constituent. */ - - wordbeg, /* Succeeds if at word beginning. */ - wordend, /* Succeeds if at word end. */ - - wordbound, /* Succeeds if at a word boundary. */ - notwordbound /* Succeeds if not at a word boundary. */ - -#ifdef emacs - ,before_dot, /* Succeeds if before point. */ - at_dot, /* Succeeds if at point. */ - after_dot, /* Succeeds if after point. */ - - /* Matches any character whose syntax is specified. Followed by - a byte which contains a syntax code, e.g., Sword. */ - syntaxspec, - - /* Matches any character whose syntax is not that specified. */ - notsyntaxspec -#endif /* emacs */ -} re_opcode_t; - -/* Common operations on the compiled pattern. */ - -/* Store NUMBER in two contiguous bytes starting at DESTINATION. */ - -#define STORE_NUMBER(destination, number) \ - do { \ - (destination)[0] = (number) & 0377; \ - (destination)[1] = (number) >> 8; \ - } while (0) - -/* Same as STORE_NUMBER, except increment DESTINATION to - the byte after where the number is stored. Therefore, DESTINATION - must be an lvalue. */ - -#define STORE_NUMBER_AND_INCR(destination, number) \ - do { \ - STORE_NUMBER (destination, number); \ - (destination) += 2; \ - } while (0) - -/* Put into DESTINATION a number stored in two contiguous bytes starting - at SOURCE. */ - -#define EXTRACT_NUMBER(destination, source) \ - do { \ - (destination) = *(source) & 0377; \ - (destination) += SIGN_EXTEND_CHAR (*((source) + 1)) << 8; \ - } while (0) - -#ifdef DEBUG -static void -extract_number (dest, source) - int *dest; - unsigned char *source; -{ - int temp = SIGN_EXTEND_CHAR (*(source + 1)); - *dest = *source & 0377; - *dest += temp << 8; -} - -#ifndef EXTRACT_MACROS /* To debug the macros. */ -#undef EXTRACT_NUMBER -#define EXTRACT_NUMBER(dest, src) extract_number (&dest, src) -#endif /* not EXTRACT_MACROS */ - -#endif /* DEBUG */ - -/* Same as EXTRACT_NUMBER, except increment SOURCE to after the number. - SOURCE must be an lvalue. */ - -#define EXTRACT_NUMBER_AND_INCR(destination, source) \ - do { \ - EXTRACT_NUMBER (destination, source); \ - (source) += 2; \ - } while (0) - -#ifdef DEBUG -static void -extract_number_and_incr (destination, source) - int *destination; - unsigned char **source; -{ - extract_number (destination, *source); - *source += 2; -} - -#ifndef EXTRACT_MACROS -#undef EXTRACT_NUMBER_AND_INCR -#define EXTRACT_NUMBER_AND_INCR(dest, src) \ - extract_number_and_incr (&dest, &src) -#endif /* not EXTRACT_MACROS */ - -#endif /* DEBUG */ - -/* If DEBUG is defined, Regex prints many voluminous messages about what - it is doing (if the variable `debug' is nonzero). If linked with the - main program in `iregex.c', you can enter patterns and strings - interactively. And if linked with the main program in `main.c' and - the other test files, you can run the already-written tests. */ - -#ifdef DEBUG - -/* We use standard I/O for debugging. */ -#include <stdio.h> - -/* It is useful to test things that ``must'' be true when debugging. */ -#include <assert.h> - -static int debug = 0; - -#define DEBUG_STATEMENT(e) e -#define DEBUG_PRINT1(x) if (debug) printf (x) -#define DEBUG_PRINT2(x1, x2) if (debug) printf (x1, x2) -#define DEBUG_PRINT3(x1, x2, x3) if (debug) printf (x1, x2, x3) -#define DEBUG_PRINT4(x1, x2, x3, x4) if (debug) printf (x1, x2, x3, x4) -#define DEBUG_PRINT_COMPILED_PATTERN(p, s, e) \ - if (debug) print_partial_compiled_pattern (s, e) -#define DEBUG_PRINT_DOUBLE_STRING(w, s1, sz1, s2, sz2) \ - if (debug) print_double_string (w, s1, sz1, s2, sz2) - - -extern void printchar (); - -/* Print the fastmap in human-readable form. */ - -void -print_fastmap (fastmap) - char *fastmap; -{ - unsigned was_a_range = 0; - unsigned i = 0; - - while (i < (1 << BYTEWIDTH)) - { - if (fastmap[i++]) - { - was_a_range = 0; - printchar (i - 1); - while (i < (1 << BYTEWIDTH) && fastmap[i]) - { - was_a_range = 1; - i++; - } - if (was_a_range) - { - printf ("-"); - printchar (i - 1); - } - } - } - putchar ('\n'); -} - - -/* Print a compiled pattern string in human-readable form, starting at - the START pointer into it and ending just before the pointer END. */ - -void -print_partial_compiled_pattern (start, end) - unsigned char *start; - unsigned char *end; -{ - int mcnt, mcnt2; - unsigned char *p = start; - unsigned char *pend = end; - - if (start == NULL) - { - printf ("(null)\n"); - return; - } - - /* Loop over pattern commands. */ - while (p < pend) - { - switch ((re_opcode_t) *p++) - { - case no_op: - printf ("/no_op"); - break; - - case exactn: - mcnt = *p++; - printf ("/exactn/%d", mcnt); - do - { - putchar ('/'); - printchar (*p++); - } - while (--mcnt); - break; - - case start_memory: - mcnt = *p++; - printf ("/start_memory/%d/%d", mcnt, *p++); - break; - - case stop_memory: - mcnt = *p++; - printf ("/stop_memory/%d/%d", mcnt, *p++); - break; - - case duplicate: - printf ("/duplicate/%d", *p++); - break; - - case anychar: - printf ("/anychar"); - break; - - case charset: - case charset_not: - { - register int c; - - printf ("/charset%s", - (re_opcode_t) *(p - 1) == charset_not ? "_not" : ""); - - assert (p + *p < pend); - - for (c = 0; c < *p; c++) - { - unsigned bit; - unsigned char map_byte = p[1 + c]; - - putchar ('/'); - - for (bit = 0; bit < BYTEWIDTH; bit++) - if (map_byte & (1 << bit)) - printchar (c * BYTEWIDTH + bit); - } - p += 1 + *p; - break; - } - - case begline: - printf ("/begline"); - break; - - case endline: - printf ("/endline"); - break; - - case on_failure_jump: - extract_number_and_incr (&mcnt, &p); - printf ("/on_failure_jump/0/%d", mcnt); - break; - - case on_failure_keep_string_jump: - extract_number_and_incr (&mcnt, &p); - printf ("/on_failure_keep_string_jump/0/%d", mcnt); - break; - - case dummy_failure_jump: - extract_number_and_incr (&mcnt, &p); - printf ("/dummy_failure_jump/0/%d", mcnt); - break; - - case push_dummy_failure: - printf ("/push_dummy_failure"); - break; - - case maybe_pop_jump: - extract_number_and_incr (&mcnt, &p); - printf ("/maybe_pop_jump/0/%d", mcnt); - break; - - case pop_failure_jump: - extract_number_and_incr (&mcnt, &p); - printf ("/pop_failure_jump/0/%d", mcnt); - break; - - case jump_past_alt: - extract_number_and_incr (&mcnt, &p); - printf ("/jump_past_alt/0/%d", mcnt); - break; - - case jump: - extract_number_and_incr (&mcnt, &p); - printf ("/jump/0/%d", mcnt); - break; - - case succeed_n: - extract_number_and_incr (&mcnt, &p); - extract_number_and_incr (&mcnt2, &p); - printf ("/succeed_n/0/%d/0/%d", mcnt, mcnt2); - break; - - case jump_n: - extract_number_and_incr (&mcnt, &p); - extract_number_and_incr (&mcnt2, &p); - printf ("/jump_n/0/%d/0/%d", mcnt, mcnt2); - break; - - case set_number_at: - extract_number_and_incr (&mcnt, &p); - extract_number_and_incr (&mcnt2, &p); - printf ("/set_number_at/0/%d/0/%d", mcnt, mcnt2); - break; - - case wordbound: - printf ("/wordbound"); - break; - - case notwordbound: - printf ("/notwordbound"); - break; - - case wordbeg: - printf ("/wordbeg"); - break; - - case wordend: - printf ("/wordend"); - -#ifdef emacs - case before_dot: - printf ("/before_dot"); - break; - - case at_dot: - printf ("/at_dot"); - break; - - case after_dot: - printf ("/after_dot"); - break; - - case syntaxspec: - printf ("/syntaxspec"); - mcnt = *p++; - printf ("/%d", mcnt); - break; - - case notsyntaxspec: - printf ("/notsyntaxspec"); - mcnt = *p++; - printf ("/%d", mcnt); - break; -#endif /* emacs */ - - case wordchar: - printf ("/wordchar"); - break; - - case notwordchar: - printf ("/notwordchar"); - break; - - case begbuf: - printf ("/begbuf"); - break; - - case endbuf: - printf ("/endbuf"); - break; - - default: - printf ("?%d", *(p-1)); - } - } - printf ("/\n"); -} - - -void -print_compiled_pattern (bufp) - struct re_pattern_buffer *bufp; -{ - unsigned char *buffer = bufp->buffer; - - print_partial_compiled_pattern (buffer, buffer + bufp->used); - printf ("%d bytes used/%d bytes allocated.\n", bufp->used, bufp->allocated); - - if (bufp->fastmap_accurate && bufp->fastmap) - { - printf ("fastmap: "); - print_fastmap (bufp->fastmap); - } - - printf ("re_nsub: %d\t", bufp->re_nsub); - printf ("regs_alloc: %d\t", bufp->regs_allocated); - printf ("can_be_null: %d\t", bufp->can_be_null); - printf ("newline_anchor: %d\n", bufp->newline_anchor); - printf ("no_sub: %d\t", bufp->no_sub); - printf ("not_bol: %d\t", bufp->not_bol); - printf ("not_eol: %d\t", bufp->not_eol); - printf ("syntax: %d\n", bufp->syntax); - /* Perhaps we should print the translate table? */ -} - - -void -print_double_string (where, string1, size1, string2, size2) - const char *where; - const char *string1; - const char *string2; - int size1; - int size2; -{ - unsigned this_char; - - if (where == NULL) - printf ("(null)"); - else - { - if (FIRST_STRING_P (where)) - { - for (this_char = where - string1; this_char < size1; this_char++) - printchar (string1[this_char]); - - where = string2; - } - - for (this_char = where - string2; this_char < size2; this_char++) - printchar (string2[this_char]); - } -} - -#else /* not DEBUG */ - -#undef assert -#define assert(e) - -#define DEBUG_STATEMENT(e) -#define DEBUG_PRINT1(x) -#define DEBUG_PRINT2(x1, x2) -#define DEBUG_PRINT3(x1, x2, x3) -#define DEBUG_PRINT4(x1, x2, x3, x4) -#define DEBUG_PRINT_COMPILED_PATTERN(p, s, e) -#define DEBUG_PRINT_DOUBLE_STRING(w, s1, sz1, s2, sz2) - -#endif /* not DEBUG */ - -/* 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. */ -reg_syntax_t re_syntax_options = RE_SYNTAX_EMACS; - - -/* 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 (syntax) - reg_syntax_t syntax; -{ - reg_syntax_t ret = re_syntax_options; - - re_syntax_options = syntax; - return ret; -} - -/* 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. */ - -static const char *re_error_msg[] = - { NULL, /* REG_NOERROR */ - "No match", /* REG_NOMATCH */ - "Invalid regular expression", /* REG_BADPAT */ - "Invalid collation character", /* REG_ECOLLATE */ - "Invalid character class name", /* REG_ECTYPE */ - "Trailing backslash", /* REG_EESCAPE */ - "Invalid back reference", /* REG_ESUBREG */ - "Unmatched [ or [^", /* REG_EBRACK */ - "Unmatched ( or \\(", /* REG_EPAREN */ - "Unmatched \\{", /* REG_EBRACE */ - "Invalid content of \\{\\}", /* REG_BADBR */ - "Invalid range end", /* REG_ERANGE */ - "Memory exhausted", /* REG_ESPACE */ - "Invalid preceding regular expression", /* REG_BADRPT */ - "Premature end of regular expression", /* REG_EEND */ - "Regular expression too big", /* REG_ESIZE */ - "Unmatched ) or \\)", /* REG_ERPAREN */ - }; - -/* Subroutine declarations and macros for regex_compile. */ - -static void store_op1 (), store_op2 (); -static void insert_op1 (), insert_op2 (); -static boolean at_begline_loc_p (), at_endline_loc_p (); -static boolean group_in_compile_stack (); -static reg_errcode_t compile_range (); - -/* Fetch the next character in the uncompiled pattern---translating it - if necessary. Also cast from a signed character in the constant - string passed to us by the user to an unsigned char that we can use - as an array index (in, e.g., `translate'). */ -#define PATFETCH(c) \ - do {if (p == pend) return REG_EEND; \ - c = (unsigned char) *p++; \ - if (translate) c = translate[c]; \ - } while (0) - -/* Fetch the next character in the uncompiled pattern, with no - translation. */ -#define PATFETCH_RAW(c) \ - do {if (p == pend) return REG_EEND; \ - c = (unsigned char) *p++; \ - } while (0) - -/* Go backwards one character in the pattern. */ -#define PATUNFETCH p-- - - -/* If `translate' is non-null, return translate[D], else just D. We - cast the subscript to translate because some data is declared as - `char *', to avoid warnings when a string constant is passed. But - when we use a character as a subscript we must make it unsigned. */ -#define TRANSLATE(d) (translate ? translate[(unsigned char) (d)] : (d)) - - -/* Macros for outputting the compiled pattern into `buffer'. */ - -/* If the buffer isn't allocated when it comes in, use this. */ -#define INIT_BUF_SIZE 32 - -/* Make sure we have at least N more bytes of space in buffer. */ -#define GET_BUFFER_SPACE(n) \ - while (b - bufp->buffer + (n) > bufp->allocated) \ - EXTEND_BUFFER () - -/* Make sure we have one more byte of buffer space and then add C to it. */ -#define BUF_PUSH(c) \ - do { \ - GET_BUFFER_SPACE (1); \ - *b++ = (unsigned char) (c); \ - } while (0) - - -/* Ensure we have two more bytes of buffer space and then append C1 and C2. */ -#define BUF_PUSH_2(c1, c2) \ - do { \ - GET_BUFFER_SPACE (2); \ - *b++ = (unsigned char) (c1); \ - *b++ = (unsigned char) (c2); \ - } while (0) - - -/* As with BUF_PUSH_2, except for three bytes. */ -#define BUF_PUSH_3(c1, c2, c3) \ - do { \ - GET_BUFFER_SPACE (3); \ - *b++ = (unsigned char) (c1); \ - *b++ = (unsigned char) (c2); \ - *b++ = (unsigned char) (c3); \ - } while (0) - - -/* Store a jump with opcode OP at LOC to location TO. We store a - relative address offset by the three bytes the jump itself occupies. */ -#define STORE_JUMP(op, loc, to) \ - store_op1 (op, loc, (to) - (loc) - 3) - -/* Likewise, for a two-argument jump. */ -#define STORE_JUMP2(op, loc, to, arg) \ - store_op2 (op, loc, (to) - (loc) - 3, arg) - -/* Like `STORE_JUMP', but for inserting. Assume `b' is the buffer end. */ -#define INSERT_JUMP(op, loc, to) \ - insert_op1 (op, loc, (to) - (loc) - 3, b) - -/* Like `STORE_JUMP2', but for inserting. Assume `b' is the buffer end. */ -#define INSERT_JUMP2(op, loc, to, arg) \ - insert_op2 (op, loc, (to) - (loc) - 3, arg, b) - - -/* This is not an arbitrary limit: the arguments which represent offsets - into the pattern are two bytes long. So if 2^16 bytes turns out to - be too small, many things would have to change. */ -#define MAX_BUF_SIZE (1L << 16) - - -/* Extend the buffer by twice its current size via realloc and - reset the pointers that pointed into the old block to point to the - correct places in the new one. If extending the buffer results in it - being larger than MAX_BUF_SIZE, then flag memory exhausted. */ -#define EXTEND_BUFFER() \ - do { \ - unsigned char *old_buffer = bufp->buffer; \ - if (bufp->allocated == MAX_BUF_SIZE) \ - return REG_ESIZE; \ - bufp->allocated <<= 1; \ - if (bufp->allocated > MAX_BUF_SIZE) \ - bufp->allocated = MAX_BUF_SIZE; \ - bufp->buffer = (unsigned char *) realloc (bufp->buffer, bufp->allocated);\ - if (bufp->buffer == NULL) \ - return REG_ESPACE; \ - /* If the buffer moved, move all the pointers into it. */ \ - if (old_buffer != bufp->buffer) \ - { \ - b = (b - old_buffer) + bufp->buffer; \ - begalt = (begalt - old_buffer) + bufp->buffer; \ - if (fixup_alt_jump) \ - fixup_alt_jump = (fixup_alt_jump - old_buffer) + bufp->buffer;\ - if (laststart) \ - laststart = (laststart - old_buffer) + bufp->buffer; \ - if (pending_exact) \ - pending_exact = (pending_exact - old_buffer) + bufp->buffer; \ - } \ - } while (0) - - -/* Since we have one byte reserved for the register number argument to - {start,stop}_memory, the maximum number of groups we can report - things about is what fits in that byte. */ -#define MAX_REGNUM 255 - -/* But patterns can have more than `MAX_REGNUM' registers. We just - ignore the excess. */ -typedef unsigned regnum_t; - - -/* Macros for the compile stack. */ - -/* Since offsets can go either forwards or backwards, this type needs to - be able to hold values from -(MAX_BUF_SIZE - 1) to MAX_BUF_SIZE - 1. */ -typedef int pattern_offset_t; - -typedef struct -{ - pattern_offset_t begalt_offset; - pattern_offset_t fixup_alt_jump; - pattern_offset_t inner_group_offset; - pattern_offset_t laststart_offset; - regnum_t regnum; -} compile_stack_elt_t; - - -typedef struct -{ - compile_stack_elt_t *stack; - unsigned size; - unsigned avail; /* Offset of next open position. */ -} compile_stack_type; - - -#define INIT_COMPILE_STACK_SIZE 32 - -#define COMPILE_STACK_EMPTY (compile_stack.avail == 0) -#define COMPILE_STACK_FULL (compile_stack.avail == compile_stack.size) - -/* The next available element. */ -#define COMPILE_STACK_TOP (compile_stack.stack[compile_stack.avail]) - - -/* Set the bit for character C in a list. */ -#define SET_LIST_BIT(c) \ - (b[((unsigned char) (c)) / BYTEWIDTH] \ - |= 1 << (((unsigned char) c) % BYTEWIDTH)) - - -/* Get the next unsigned number in the uncompiled pattern. */ -#define GET_UNSIGNED_NUMBER(num) \ - { if (p != pend) \ - { \ - PATFETCH (c); \ - while (ISDIGIT (c)) \ - { \ - if (num < 0) \ - num = 0; \ - num = num * 10 + c - '0'; \ - if (p == pend) \ - break; \ - PATFETCH (c); \ - } \ - } \ - } - -#define CHAR_CLASS_MAX_LENGTH 6 /* Namely, `xdigit'. */ - -#define IS_CHAR_CLASS(string) \ - (STREQ (string, "alpha") || STREQ (string, "upper") \ - || STREQ (string, "lower") || STREQ (string, "digit") \ - || STREQ (string, "alnum") || STREQ (string, "xdigit") \ - || STREQ (string, "space") || STREQ (string, "print") \ - || STREQ (string, "punct") || STREQ (string, "graph") \ - || STREQ (string, "cntrl") || STREQ (string, "blank")) - -/* `regex_compile' compiles PATTERN (of length SIZE) according to SYNTAX. - Returns one of error codes defined in `regex.h', or zero for success. - - Assumes the `allocated' (and perhaps `buffer') and `translate' - fields are set in BUFP on entry. - - If it succeeds, results are put in BUFP (if it returns an error, the - contents of BUFP are undefined): - `buffer' is the compiled pattern; - `syntax' is set to SYNTAX; - `used' is set to the length of the compiled pattern; - `fastmap_accurate' is zero; - `re_nsub' is the number of subexpressions in PATTERN; - `not_bol' and `not_eol' are zero; - - The `fastmap' and `newline_anchor' fields are neither - examined nor set. */ - -static reg_errcode_t -regex_compile (pattern, size, syntax, bufp) - const char *pattern; - int size; - reg_syntax_t syntax; - struct re_pattern_buffer *bufp; -{ - /* We fetch characters from PATTERN here. Even though PATTERN is - `char *' (i.e., signed), we declare these variables as unsigned, so - they can be reliably used as array indices. */ - register unsigned char c, c1; - - /* A random temporary spot in PATTERN. */ - const char *p1; - - /* Points to the end of the buffer, where we should append. */ - register unsigned char *b; - - /* Keeps track of unclosed groups. */ - compile_stack_type compile_stack; - - /* Points to the current (ending) position in the pattern. */ - const char *p = pattern; - const char *pend = pattern + size; - - /* How to translate the characters in the pattern. */ - char *translate = bufp->translate; - - /* Address of the count-byte of the most recently inserted `exactn' - command. This makes it possible to tell if a new exact-match - character can be added to that command or if the character requires - a new `exactn' command. */ - unsigned char *pending_exact = 0; - - /* Address of start of the most recently finished expression. - This tells, e.g., postfix * where to find the start of its - operand. Reset at the beginning of groups and alternatives. */ - unsigned char *laststart = 0; - - /* Address of beginning of regexp, or inside of last group. */ - unsigned char *begalt; - - /* Place in the uncompiled pattern (i.e., the {) to - which to go back if the interval is invalid. */ - const char *beg_interval; - - /* Address of the place where a forward jump should go to the end of - the containing expression. Each alternative of an `or' -- except the - last -- ends with a forward jump of this sort. */ - unsigned char *fixup_alt_jump = 0; - - /* Counts open-groups as they are encountered. Remembered for the - matching close-group on the compile stack, so the same register - number is put in the stop_memory as the start_memory. */ - regnum_t regnum = 0; - -#ifdef DEBUG - DEBUG_PRINT1 ("\nCompiling pattern: "); - if (debug) - { - unsigned debug_count; - - for (debug_count = 0; debug_count < size; debug_count++) - printchar (pattern[debug_count]); - putchar ('\n'); - } -#endif /* DEBUG */ - - /* Initialize the compile stack. */ - compile_stack.stack = TALLOC (INIT_COMPILE_STACK_SIZE, compile_stack_elt_t); - if (compile_stack.stack == NULL) - return REG_ESPACE; - - compile_stack.size = INIT_COMPILE_STACK_SIZE; - compile_stack.avail = 0; - - /* Initialize the pattern buffer. */ - bufp->syntax = syntax; - bufp->fastmap_accurate = 0; - bufp->not_bol = bufp->not_eol = 0; - - /* Set `used' to zero, so that if we return an error, the pattern - printer (for debugging) will think there's no pattern. We reset it - at the end. */ - bufp->used = 0; - - /* Always count groups, whether or not bufp->no_sub is set. */ - bufp->re_nsub = 0; - -#if !defined (emacs) && !defined (SYNTAX_TABLE) - /* Initialize the syntax table. */ - init_syntax_once (); -#endif - - if (bufp->allocated == 0) - { - if (bufp->buffer) - { /* 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. */ - RETALLOC (bufp->buffer, INIT_BUF_SIZE, unsigned char); - } - else - { /* Caller did not allocate a buffer. Do it for them. */ - bufp->buffer = TALLOC (INIT_BUF_SIZE, unsigned char); - } - if (!bufp->buffer) return REG_ESPACE; - - bufp->allocated = INIT_BUF_SIZE; - } - - begalt = b = bufp->buffer; - - /* Loop through the uncompiled pattern until we're at the end. */ - while (p != pend) - { - PATFETCH (c); - - switch (c) - { - case '^': - { - if ( /* If at start of pattern, it's an operator. */ - p == pattern + 1 - /* If context independent, it's an operator. */ - || syntax & RE_CONTEXT_INDEP_ANCHORS - /* Otherwise, depends on what's come before. */ - || at_begline_loc_p (pattern, p, syntax)) - BUF_PUSH (begline); - else - goto normal_char; - } - break; - - - case '$': - { - if ( /* If at end of pattern, it's an operator. */ - p == pend - /* If context independent, it's an operator. */ - || syntax & RE_CONTEXT_INDEP_ANCHORS - /* Otherwise, depends on what's next. */ - || at_endline_loc_p (p, pend, syntax)) - BUF_PUSH (endline); - else - goto normal_char; - } - break; - - - case '+': - case '?': - if ((syntax & RE_BK_PLUS_QM) - || (syntax & RE_LIMITED_OPS)) - goto normal_char; - handle_plus: - case '*': - /* If there is no previous pattern... */ - if (!laststart) - { - if (syntax & RE_CONTEXT_INVALID_OPS) - return REG_BADRPT; - else if (!(syntax & RE_CONTEXT_INDEP_OPS)) - goto normal_char; - } - - { - /* Are we optimizing this jump? */ - boolean keep_string_p = false; - - /* 1 means zero (many) matches is allowed. */ - char zero_times_ok = 0, many_times_ok = 0; - - /* If there is a sequence of repetition chars, collapse it - down to just one (the right one). We can't combine - interval operators with these because of, e.g., `a{2}*', - which should only match an even number of `a's. */ - - for (;;) - { - zero_times_ok |= c != '+'; - many_times_ok |= c != '?'; - - if (p == pend) - break; - - PATFETCH (c); - - if (c == '*' - || (!(syntax & RE_BK_PLUS_QM) && (c == '+' || c == '?'))) - ; - - else if (syntax & RE_BK_PLUS_QM && c == '\\') - { - if (p == pend) return REG_EESCAPE; - - PATFETCH (c1); - if (!(c1 == '+' || c1 == '?')) - { - PATUNFETCH; - PATUNFETCH; - break; - } - - c = c1; - } - else - { - PATUNFETCH; - break; - } - - /* If we get here, we found another repeat character. */ - } - - /* Star, etc. applied to an empty pattern is equivalent - to an empty pattern. */ - if (!laststart) - break; - - /* Now we know whether or not zero matches is allowed - and also whether or not two or more matches is allowed. */ - if (many_times_ok) - { /* More than one repetition is allowed, so put in at the - end a backward relative jump from `b' to before the next - jump we're going to put in below (which jumps from - laststart to after this jump). - - But if we are at the `*' in the exact sequence `.*\n', - insert an unconditional jump backwards to the ., - instead of the beginning of the loop. This way we only - push a failure point once, instead of every time - through the loop. */ - assert (p - 1 > pattern); - - /* Allocate the space for the jump. */ - GET_BUFFER_SPACE (3); - - /* We know we are not at the first character of the pattern, - because laststart was nonzero. And we've already - incremented `p', by the way, to be the character after - the `*'. Do we have to do something analogous here - for null bytes, because of RE_DOT_NOT_NULL? */ - if (TRANSLATE (*(p - 2)) == TRANSLATE ('.') - && zero_times_ok - && p < pend && TRANSLATE (*p) == TRANSLATE ('\n') - && !(syntax & RE_DOT_NEWLINE)) - { /* We have .*\n. */ - STORE_JUMP (jump, b, laststart); - keep_string_p = true; - } - else - /* Anything else. */ - STORE_JUMP (maybe_pop_jump, b, laststart - 3); - - /* We've added more stuff to the buffer. */ - b += 3; - } - - /* On failure, jump from laststart to b + 3, which will be the - end of the buffer after this jump is inserted. */ - GET_BUFFER_SPACE (3); - INSERT_JUMP (keep_string_p ? on_failure_keep_string_jump - : on_failure_jump, - laststart, b + 3); - pending_exact = 0; - b += 3; - - if (!zero_times_ok) - { - /* At least one repetition is required, so insert a - `dummy_failure_jump' before the initial - `on_failure_jump' instruction of the loop. This - effects a skip over that instruction the first time - we hit that loop. */ - GET_BUFFER_SPACE (3); - INSERT_JUMP (dummy_failure_jump, laststart, laststart + 6); - b += 3; - } - } - break; - - - case '.': - laststart = b; - BUF_PUSH (anychar); - break; - - - case '[': - { - boolean had_char_class = false; - - if (p == pend) return REG_EBRACK; - - /* Ensure that we have enough space to push a charset: the - opcode, the length count, and the bitset; 34 bytes in all. */ - GET_BUFFER_SPACE (34); - - laststart = b; - - /* We test `*p == '^' twice, instead of using an if - statement, so we only need one BUF_PUSH. */ - BUF_PUSH (*p == '^' ? charset_not : charset); - if (*p == '^') - p++; - - /* Remember the first position in the bracket expression. */ - p1 = p; - - /* Push the number of bytes in the bitmap. */ - BUF_PUSH ((1 << BYTEWIDTH) / BYTEWIDTH); - - /* Clear the whole map. */ - bzero (b, (1 << BYTEWIDTH) / BYTEWIDTH); - - /* charset_not matches newline according to a syntax bit. */ - if ((re_opcode_t) b[-2] == charset_not - && (syntax & RE_HAT_LISTS_NOT_NEWLINE)) - SET_LIST_BIT ('\n'); - - /* Read in characters and ranges, setting map bits. */ - for (;;) - { - if (p == pend) return REG_EBRACK; - - PATFETCH (c); - - /* \ might escape characters inside [...] and [^...]. */ - if ((syntax & RE_BACKSLASH_ESCAPE_IN_LISTS) && c == '\\') - { - if (p == pend) return REG_EESCAPE; - - PATFETCH (c1); - SET_LIST_BIT (c1); - continue; - } - - /* Could be the end of the bracket expression. If it's - not (i.e., when the bracket expression is `[]' so - far), the ']' character bit gets set way below. */ - if (c == ']' && p != p1 + 1) - break; - - /* Look ahead to see if it's a range when the last thing - was a character class. */ - if (had_char_class && c == '-' && *p != ']') - return REG_ERANGE; - - /* Look ahead to see if it's a range when the last thing - was a character: if this is a hyphen not at the - beginning or the end of a list, then it's the range - operator. */ - if (c == '-' - && !(p - 2 >= pattern && p[-2] == '[') - && !(p - 3 >= pattern && p[-3] == '[' && p[-2] == '^') - && *p != ']') - { - reg_errcode_t ret - = compile_range (&p, pend, translate, syntax, b); - if (ret != REG_NOERROR) return ret; - } - - else if (p[0] == '-' && p[1] != ']') - { /* This handles ranges made up of characters only. */ - reg_errcode_t ret; - - /* Move past the `-'. */ - PATFETCH (c1); - - ret = compile_range (&p, pend, translate, syntax, b); - if (ret != REG_NOERROR) return ret; - } - - /* See if we're at the beginning of a possible character - class. */ - - else if (syntax & RE_CHAR_CLASSES && c == '[' && *p == ':') - { /* Leave room for the null. */ - char str[CHAR_CLASS_MAX_LENGTH + 1]; - - PATFETCH (c); - c1 = 0; - - /* If pattern is `[[:'. */ - if (p == pend) return REG_EBRACK; - - for (;;) - { - PATFETCH (c); - if (c == ':' || c == ']' || p == pend - || c1 == CHAR_CLASS_MAX_LENGTH) - break; - str[c1++] = c; - } - str[c1] = '\0'; - - /* If isn't a word bracketed by `[:' and:`]': - undo the ending character, the letters, and leave - the leading `:' and `[' (but set bits for them). */ - if (c == ':' && *p == ']') - { - int ch; - boolean is_alnum = STREQ (str, "alnum"); - boolean is_alpha = STREQ (str, "alpha"); - boolean is_blank = STREQ (str, "blank"); - boolean is_cntrl = STREQ (str, "cntrl"); - boolean is_digit = STREQ (str, "digit"); - boolean is_graph = STREQ (str, "graph"); - boolean is_lower = STREQ (str, "lower"); - boolean is_print = STREQ (str, "print"); - boolean is_punct = STREQ (str, "punct"); - boolean is_space = STREQ (str, "space"); - boolean is_upper = STREQ (str, "upper"); - boolean is_xdigit = STREQ (str, "xdigit"); - - if (!IS_CHAR_CLASS (str)) return REG_ECTYPE; - - /* Throw away the ] at the end of the character - class. */ - PATFETCH (c); - - if (p == pend) return REG_EBRACK; - - for (ch = 0; ch < 1 << BYTEWIDTH; ch++) - { - if ( (is_alnum && ISALNUM (ch)) - || (is_alpha && ISALPHA (ch)) - || (is_blank && ISBLANK (ch)) - || (is_cntrl && ISCNTRL (ch)) - || (is_digit && ISDIGIT (ch)) - || (is_graph && ISGRAPH (ch)) - || (is_lower && ISLOWER (ch)) - || (is_print && ISPRINT (ch)) - || (is_punct && ISPUNCT (ch)) - || (is_space && ISSPACE (ch)) - || (is_upper && ISUPPER (ch)) - || (is_xdigit && ISXDIGIT (ch))) - SET_LIST_BIT (ch); - } - had_char_class = true; - } - else - { - c1++; - while (c1--) - PATUNFETCH; - SET_LIST_BIT ('['); - SET_LIST_BIT (':'); - had_char_class = false; - } - } - else - { - had_char_class = false; - SET_LIST_BIT (c); - } - } - - /* Discard any (non)matching list bytes that are all 0 at the - end of the map. Decrease the map-length byte too. */ - while ((int) b[-1] > 0 && b[b[-1] - 1] == 0) - b[-1]--; - b += b[-1]; - } - break; - - - case '(': - if (syntax & RE_NO_BK_PARENS) - goto handle_open; - else - goto normal_char; - - - case ')': - if (syntax & RE_NO_BK_PARENS) - goto handle_close; - else - goto normal_char; - - - case '\n': - if (syntax & RE_NEWLINE_ALT) - goto handle_alt; - else - goto normal_char; - - - case '|': - if (syntax & RE_NO_BK_VBAR) - goto handle_alt; - else - goto normal_char; - - - case '{': - if (syntax & RE_INTERVALS && syntax & RE_NO_BK_BRACES) - goto handle_interval; - else - goto normal_char; - - - case '\\': - if (p == pend) return REG_EESCAPE; - - /* Do not translate the character after the \, so that we can - distinguish, e.g., \B from \b, even if we normally would - translate, e.g., B to b. */ - PATFETCH_RAW (c); - - switch (c) - { - case '(': - if (syntax & RE_NO_BK_PARENS) - goto normal_backslash; - - handle_open: - bufp->re_nsub++; - regnum++; - - if (COMPILE_STACK_FULL) - { - RETALLOC (compile_stack.stack, compile_stack.size << 1, - compile_stack_elt_t); - if (compile_stack.stack == NULL) return REG_ESPACE; - - compile_stack.size <<= 1; - } - - /* These are the values to restore when we hit end of this - group. They are all relative offsets, so that if the - whole pattern moves because of realloc, they will still - be valid. */ - COMPILE_STACK_TOP.begalt_offset = begalt - bufp->buffer; - COMPILE_STACK_TOP.fixup_alt_jump - = fixup_alt_jump ? fixup_alt_jump - bufp->buffer + 1 : 0; - COMPILE_STACK_TOP.laststart_offset = b - bufp->buffer; - COMPILE_STACK_TOP.regnum = regnum; - - /* We will eventually replace the 0 with the number of - groups inner to this one. But do not push a - start_memory for groups beyond the last one we can - represent in the compiled pattern. */ - if (regnum <= MAX_REGNUM) - { - COMPILE_STACK_TOP.inner_group_offset = b - bufp->buffer + 2; - BUF_PUSH_3 (start_memory, regnum, 0); - } - - compile_stack.avail++; - - fixup_alt_jump = 0; - laststart = 0; - begalt = b; - /* If we've reached MAX_REGNUM groups, then this open - won't actually generate any code, so we'll have to - clear pending_exact explicitly. */ - pending_exact = 0; - break; - - - case ')': - if (syntax & RE_NO_BK_PARENS) goto normal_backslash; - - if (COMPILE_STACK_EMPTY) - { - if (syntax & RE_UNMATCHED_RIGHT_PAREN_ORD) - goto normal_backslash; - else - return REG_ERPAREN; - } - - handle_close: - if (fixup_alt_jump) - { /* Push a dummy failure point at the end of the - alternative for a possible future - `pop_failure_jump' to pop. See comments at - `push_dummy_failure' in `re_match_2'. */ - BUF_PUSH (push_dummy_failure); - - /* We allocated space for this jump when we assigned - to `fixup_alt_jump', in the `handle_alt' case below. */ - STORE_JUMP (jump_past_alt, fixup_alt_jump, b - 1); - } - - /* See similar code for backslashed left paren above. */ - if (COMPILE_STACK_EMPTY) - { - if (syntax & RE_UNMATCHED_RIGHT_PAREN_ORD) - goto normal_char; - else - return REG_ERPAREN; - } - - /* Since we just checked for an empty stack above, this - ``can't happen''. */ - assert (compile_stack.avail != 0); - { - /* We don't just want to restore into `regnum', because - later groups should continue to be numbered higher, - as in `(ab)c(de)' -- the second group is #2. */ - regnum_t this_group_regnum; - - compile_stack.avail--; - begalt = bufp->buffer + COMPILE_STACK_TOP.begalt_offset; - fixup_alt_jump - = COMPILE_STACK_TOP.fixup_alt_jump - ? bufp->buffer + COMPILE_STACK_TOP.fixup_alt_jump - 1 - : 0; - laststart = bufp->buffer + COMPILE_STACK_TOP.laststart_offset; - this_group_regnum = COMPILE_STACK_TOP.regnum; - /* If we've reached MAX_REGNUM groups, then this open - won't actually generate any code, so we'll have to - clear pending_exact explicitly. */ - pending_exact = 0; - - /* We're at the end of the group, so now we know how many - groups were inside this one. */ - if (this_group_regnum <= MAX_REGNUM) - { - unsigned char *inner_group_loc - = bufp->buffer + COMPILE_STACK_TOP.inner_group_offset; - - *inner_group_loc = regnum - this_group_regnum; - BUF_PUSH_3 (stop_memory, this_group_regnum, - regnum - this_group_regnum); - } - } - break; - - - case '|': /* `\|'. */ - if (syntax & RE_LIMITED_OPS || syntax & RE_NO_BK_VBAR) - goto normal_backslash; - handle_alt: - if (syntax & RE_LIMITED_OPS) - goto normal_char; - - /* Insert before the previous alternative a jump which - jumps to this alternative if the former fails. */ - GET_BUFFER_SPACE (3); - INSERT_JUMP (on_failure_jump, begalt, b + 6); - pending_exact = 0; - b += 3; - - /* The alternative before this one has a jump after it - which gets executed if it gets matched. Adjust that - jump so it will jump to this alternative's analogous - jump (put in below, which in turn will jump to the next - (if any) alternative's such jump, etc.). The last such - jump jumps to the correct final destination. A picture: - _____ _____ - | | | | - | v | v - a | b | c - - If we are at `b', then fixup_alt_jump right now points to a - three-byte space after `a'. We'll put in the jump, set - fixup_alt_jump to right after `b', and leave behind three - bytes which we'll fill in when we get to after `c'. */ - - if (fixup_alt_jump) - STORE_JUMP (jump_past_alt, fixup_alt_jump, b); - - /* Mark and leave space for a jump after this alternative, - to be filled in later either by next alternative or - when know we're at the end of a series of alternatives. */ - fixup_alt_jump = b; - GET_BUFFER_SPACE (3); - b += 3; - - laststart = 0; - begalt = b; - break; - - - case '{': - /* If \{ is a literal. */ - if (!(syntax & RE_INTERVALS) - /* If we're at `\{' and it's not the open-interval - operator. */ - || ((syntax & RE_INTERVALS) && (syntax & RE_NO_BK_BRACES)) - || (p - 2 == pattern && p == pend)) - goto normal_backslash; - - handle_interval: - { - /* If got here, then the syntax allows intervals. */ - - /* At least (most) this many matches must be made. */ - int lower_bound = -1, upper_bound = -1; - - beg_interval = p - 1; - - if (p == pend) - { - if (syntax & RE_NO_BK_BRACES) - goto unfetch_interval; - else - return REG_EBRACE; - } - - GET_UNSIGNED_NUMBER (lower_bound); - - if (c == ',') - { - GET_UNSIGNED_NUMBER (upper_bound); - if (upper_bound < 0) upper_bound = RE_DUP_MAX; - } - else - /* Interval such as `{1}' => match exactly once. */ - upper_bound = lower_bound; - - if (lower_bound < 0 || upper_bound > RE_DUP_MAX - || lower_bound > upper_bound) - { - if (syntax & RE_NO_BK_BRACES) - goto unfetch_interval; - else - return REG_BADBR; - } - - if (!(syntax & RE_NO_BK_BRACES)) - { - if (c != '\\') return REG_EBRACE; - - PATFETCH (c); - } - - if (c != '}') - { - if (syntax & RE_NO_BK_BRACES) - goto unfetch_interval; - else - return REG_BADBR; - } - - /* We just parsed a valid interval. */ - - /* If it's invalid to have no preceding re. */ - if (!laststart) - { - if (syntax & RE_CONTEXT_INVALID_OPS) - return REG_BADRPT; - else if (syntax & RE_CONTEXT_INDEP_OPS) - laststart = b; - else - goto unfetch_interval; - } - - /* If the upper bound is zero, don't want to succeed at - all; jump from `laststart' to `b + 3', which will be - the end of the buffer after we insert the jump. */ - if (upper_bound == 0) - { - GET_BUFFER_SPACE (3); - INSERT_JUMP (jump, laststart, b + 3); - b += 3; - } - - /* Otherwise, we have a nontrivial interval. When - we're all done, the pattern will look like: - set_number_at <jump count> <upper bound> - set_number_at <succeed_n count> <lower bound> - succeed_n <after jump addr> <succeed_n count> - <body of loop> - jump_n <succeed_n addr> <jump count> - (The upper bound and `jump_n' are omitted if - `upper_bound' is 1, though.) */ - else - { /* If the upper bound is > 1, we need to insert - more at the end of the loop. */ - unsigned nbytes = 10 + (upper_bound > 1) * 10; - - GET_BUFFER_SPACE (nbytes); - - /* Initialize lower bound of the `succeed_n', even - though it will be set during matching by its - attendant `set_number_at' (inserted next), - because `re_compile_fastmap' needs to know. - Jump to the `jump_n' we might insert below. */ - INSERT_JUMP2 (succeed_n, laststart, - b + 5 + (upper_bound > 1) * 5, - lower_bound); - b += 5; - - /* Code to initialize the lower bound. Insert - before the `succeed_n'. The `5' is the last two - bytes of this `set_number_at', plus 3 bytes of - the following `succeed_n'. */ - insert_op2 (set_number_at, laststart, 5, lower_bound, b); - b += 5; - - if (upper_bound > 1) - { /* More than one repetition is allowed, so - append a backward jump to the `succeed_n' - that starts this interval. - - When we've reached this during matching, - we'll have matched the interval once, so - jump back only `upper_bound - 1' times. */ - STORE_JUMP2 (jump_n, b, laststart + 5, - upper_bound - 1); - b += 5; - - /* The location we want to set is the second - parameter of the `jump_n'; that is `b-2' as - an absolute address. `laststart' will be - the `set_number_at' we're about to insert; - `laststart+3' the number to set, the source - for the relative address. But we are - inserting into the middle of the pattern -- - so everything is getting moved up by 5. - Conclusion: (b - 2) - (laststart + 3) + 5, - i.e., b - laststart. - - We insert this at the beginning of the loop - so that if we fail during matching, we'll - reinitialize the bounds. */ - insert_op2 (set_number_at, laststart, b - laststart, - upper_bound - 1, b); - b += 5; - } - } - pending_exact = 0; - beg_interval = NULL; - } - break; - - unfetch_interval: - /* If an invalid interval, match the characters as literals. */ - assert (beg_interval); - p = beg_interval; - beg_interval = NULL; - - /* normal_char and normal_backslash need `c'. */ - PATFETCH (c); - - if (!(syntax & RE_NO_BK_BRACES)) - { - if (p > pattern && p[-1] == '\\') - goto normal_backslash; - } - goto normal_char; - -#ifdef emacs - /* There is no way to specify the before_dot and after_dot - operators. rms says this is ok. --karl */ - case '=': - BUF_PUSH (at_dot); - break; - - case 's': - laststart = b; - PATFETCH (c); - BUF_PUSH_2 (syntaxspec, syntax_spec_code[c]); - break; - - case 'S': - laststart = b; - PATFETCH (c); - BUF_PUSH_2 (notsyntaxspec, syntax_spec_code[c]); - break; -#endif /* emacs */ - - - case 'w': - laststart = b; - BUF_PUSH (wordchar); - break; - - - case 'W': - laststart = b; - BUF_PUSH (notwordchar); - break; - - - case '<': - BUF_PUSH (wordbeg); - break; - - case '>': - BUF_PUSH (wordend); - break; - - case 'b': - BUF_PUSH (wordbound); - break; - - case 'B': - BUF_PUSH (notwordbound); - break; - - case '`': - BUF_PUSH (begbuf); - break; - - case '\'': - BUF_PUSH (endbuf); - 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) - goto normal_char; - - c1 = c - '0'; - - if (c1 > regnum) - return REG_ESUBREG; - - /* Can't back reference to a subexpression if inside of it. */ - if (group_in_compile_stack (compile_stack, c1)) - goto normal_char; - - laststart = b; - BUF_PUSH_2 (duplicate, c1); - break; - - - case '+': - case '?': - if (syntax & RE_BK_PLUS_QM) - goto handle_plus; - else - goto normal_backslash; - - default: - normal_backslash: - /* You might think it would be useful for \ to mean - not to translate; but if we don't translate it - it will never match anything. */ - c = TRANSLATE (c); - goto normal_char; - } - break; - - - default: - /* Expects the character in `c'. */ - normal_char: - /* If no exactn currently being built. */ - if (!pending_exact - - /* If last exactn not at current position. */ - || pending_exact + *pending_exact + 1 != b - - /* We have only one byte following the exactn for the count. */ - || *pending_exact == (1 << BYTEWIDTH) - 1 - - /* If followed by a repetition operator. */ - || *p == '*' || *p == '^' - || ((syntax & RE_BK_PLUS_QM) - ? *p == '\\' && (p[1] == '+' || p[1] == '?') - : (*p == '+' || *p == '?')) - || ((syntax & RE_INTERVALS) - && ((syntax & RE_NO_BK_BRACES) - ? *p == '{' - : (p[0] == '\\' && p[1] == '{')))) - { - /* Start building a new exactn. */ - - laststart = b; - - BUF_PUSH_2 (exactn, 0); - pending_exact = b - 1; - } - - BUF_PUSH (c); - (*pending_exact)++; - break; - } /* switch (c) */ - } /* while p != pend */ - - - /* Through the pattern now. */ - - if (fixup_alt_jump) - STORE_JUMP (jump_past_alt, fixup_alt_jump, b); - - if (!COMPILE_STACK_EMPTY) - return REG_EPAREN; - - free (compile_stack.stack); - - /* We have succeeded; set the length of the buffer. */ - bufp->used = b - bufp->buffer; - -#ifdef DEBUG - if (debug) - { - DEBUG_PRINT1 ("\nCompiled pattern: "); - print_compiled_pattern (bufp); - } -#endif /* DEBUG */ - - return REG_NOERROR; -} /* regex_compile */ - -/* Subroutines for `regex_compile'. */ - -/* Store OP at LOC followed by two-byte integer parameter ARG. */ - -static void -store_op1 (op, loc, arg) - re_opcode_t op; - unsigned char *loc; - int arg; -{ - *loc = (unsigned char) op; - STORE_NUMBER (loc + 1, arg); -} - - -/* Like `store_op1', but for two two-byte parameters ARG1 and ARG2. */ - -static void -store_op2 (op, loc, arg1, arg2) - re_opcode_t op; - unsigned char *loc; - int arg1, arg2; -{ - *loc = (unsigned char) op; - STORE_NUMBER (loc + 1, arg1); - STORE_NUMBER (loc + 3, arg2); -} - - -/* Copy the bytes from LOC to END to open up three bytes of space at LOC - for OP followed by two-byte integer parameter ARG. */ - -static void -insert_op1 (op, loc, arg, end) - re_opcode_t op; - unsigned char *loc; - int arg; - unsigned char *end; -{ - register unsigned char *pfrom = end; - register unsigned char *pto = end + 3; - - while (pfrom != loc) - *--pto = *--pfrom; - - store_op1 (op, loc, arg); -} - - -/* Like `insert_op1', but for two two-byte parameters ARG1 and ARG2. */ - -static void -insert_op2 (op, loc, arg1, arg2, end) - re_opcode_t op; - unsigned char *loc; - int arg1, arg2; - unsigned char *end; -{ - register unsigned char *pfrom = end; - register unsigned char *pto = end + 5; - - while (pfrom != loc) - *--pto = *--pfrom; - - store_op2 (op, loc, arg1, arg2); -} - - -/* P points to just after a ^ in PATTERN. Return true if that ^ comes - after an alternative or a begin-subexpression. We assume there is at - least one character before the ^. */ - -static boolean -at_begline_loc_p (pattern, p, syntax) - const char *pattern, *p; - reg_syntax_t syntax; -{ - const char *prev = p - 2; - boolean prev_prev_backslash = prev > pattern && prev[-1] == '\\'; - - return - /* After a subexpression? */ - (*prev == '(' && (syntax & RE_NO_BK_PARENS || prev_prev_backslash)) - /* After an alternative? */ - || (*prev == '|' && (syntax & RE_NO_BK_VBAR || prev_prev_backslash)); -} - - -/* The dual of at_begline_loc_p. This one is for $. We assume there is - at least one character after the $, i.e., `P < PEND'. */ - -static boolean -at_endline_loc_p (p, pend, syntax) - const char *p, *pend; - int syntax; -{ - const char *next = p; - boolean next_backslash = *next == '\\'; - const char *next_next = p + 1 < pend ? p + 1 : NULL; - - return - /* Before a subexpression? */ - (syntax & RE_NO_BK_PARENS ? *next == ')' - : next_backslash && next_next && *next_next == ')') - /* Before an alternative? */ - || (syntax & RE_NO_BK_VBAR ? *next == '|' - : next_backslash && next_next && *next_next == '|'); -} - - -/* Returns true if REGNUM is in one of COMPILE_STACK's elements and - false if it's not. */ - -static boolean -group_in_compile_stack (compile_stack, regnum) - compile_stack_type compile_stack; - regnum_t regnum; -{ - int this_element; - - for (this_element = compile_stack.avail - 1; - this_element >= 0; - this_element--) - if (compile_stack.stack[this_element].regnum == regnum) - return true; - - return false; -} - - -/* Read the ending character of a range (in a bracket expression) from the - uncompiled pattern *P_PTR (which ends at PEND). We assume the - starting character is in `P[-2]'. (`P[-1]' is the character `-'.) - Then we set the translation of all bits between the starting and - ending characters (inclusive) in the compiled pattern B. - - Return an error code. - - We use these short variable names so we can use the same macros as - `regex_compile' itself. */ - -static reg_errcode_t -compile_range (p_ptr, pend, translate, syntax, b) - const char **p_ptr, *pend; - char *translate; - reg_syntax_t syntax; - unsigned char *b; -{ - unsigned this_char; - - const char *p = *p_ptr; - int range_start, range_end; - - if (p == pend) - return REG_ERANGE; - - /* Even though the pattern is a signed `char *', we need to fetch - with unsigned char *'s; if the high bit of the pattern character - is set, the range endpoints will be negative if we fetch using a - signed char *. - - We also want to fetch the endpoints without translating them; the - appropriate translation is done in the bit-setting loop below. */ - range_start = ((unsigned char *) p)[-2]; - range_end = ((unsigned char *) p)[0]; - - /* Have to increment the pointer into the pattern string, so the - caller isn't still at the ending character. */ - (*p_ptr)++; - - /* If the start is after the end, the range is empty. */ - if (range_start > range_end) - return syntax & RE_NO_EMPTY_RANGES ? REG_ERANGE : REG_NOERROR; - - /* Here we see why `this_char' has to be larger than an `unsigned - char' -- the range is inclusive, so if `range_end' == 0xff - (assuming 8-bit characters), we would otherwise go into an infinite - loop, since all characters <= 0xff. */ - for (this_char = range_start; this_char <= range_end; this_char++) - { - SET_LIST_BIT (TRANSLATE (this_char)); - } - - return REG_NOERROR; -} - -/* Failure stack declarations and macros; both re_compile_fastmap and - re_match_2 use a failure stack. These have to be macros because of - REGEX_ALLOCATE. */ - - -/* Number of failure points for which to initially allocate space - when matching. If this number is exceeded, we allocate more - space, so it is not a hard limit. */ -#ifndef INIT_FAILURE_ALLOC -#define INIT_FAILURE_ALLOC 5 -#endif - -/* Roughly the maximum number of failure points on the stack. Would be - exactly that if always used MAX_FAILURE_SPACE each time we failed. - This is a variable only so users of regex can assign to it; we never - change it ourselves. */ + 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, write to the Free + Software Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA + 02110-1301 USA. */ + +#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 + +#if defined (_MSC_VER) +#include <stdio.h> /* for size_t */ +#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> + +#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" +#ifdef GAWK +#define bool int +#define true (1) +#define false (0) +#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; - -typedef const unsigned char *fail_stack_elt_t; - -typedef struct -{ - fail_stack_elt_t *stack; - unsigned size; - unsigned avail; /* Offset of next open position. */ -} fail_stack_type; - -#define FAIL_STACK_EMPTY() (fail_stack.avail == 0) -#define FAIL_STACK_PTR_EMPTY() (fail_stack_ptr->avail == 0) -#define FAIL_STACK_FULL() (fail_stack.avail == fail_stack.size) -#define FAIL_STACK_TOP() (fail_stack.stack[fail_stack.avail]) - - -/* Initialize `fail_stack'. Do `return -2' if the alloc fails. */ - -#define INIT_FAIL_STACK() \ - do { \ - fail_stack.stack = (fail_stack_elt_t *) \ - REGEX_ALLOCATE (INIT_FAILURE_ALLOC * sizeof (fail_stack_elt_t)); \ - \ - if (fail_stack.stack == NULL) \ - return -2; \ - \ - fail_stack.size = INIT_FAILURE_ALLOC; \ - fail_stack.avail = 0; \ - } while (0) - - -/* Double the size of FAIL_STACK, up to approximately `re_max_failures' items. - - Return 1 if succeeds, and 0 if either ran out of memory - allocating space for it or it was already too large. - - REGEX_REALLOCATE requires `destination' be declared. */ - -#define DOUBLE_FAIL_STACK(fail_stack) \ - ((fail_stack).size > re_max_failures * MAX_FAILURE_ITEMS \ - ? 0 \ - : ((fail_stack).stack = (fail_stack_elt_t *) \ - REGEX_REALLOCATE ((fail_stack).stack, \ - (fail_stack).size * sizeof (fail_stack_elt_t), \ - ((fail_stack).size << 1) * sizeof (fail_stack_elt_t)), \ - \ - (fail_stack).stack == NULL \ - ? 0 \ - : ((fail_stack).size <<= 1, \ - 1))) - - -/* Push PATTERN_OP on FAIL_STACK. - - Return 1 if was able to do so and 0 if ran out of memory allocating - space to do so. */ -#define PUSH_PATTERN_OP(pattern_op, fail_stack) \ - ((FAIL_STACK_FULL () \ - && !DOUBLE_FAIL_STACK (fail_stack)) \ - ? 0 \ - : ((fail_stack).stack[(fail_stack).avail++] = pattern_op, \ - 1)) - -/* This pushes an item onto the failure stack. Must be a four-byte - value. Assumes the variable `fail_stack'. Probably should only - be called from within `PUSH_FAILURE_POINT'. */ -#define PUSH_FAILURE_ITEM(item) \ - fail_stack.stack[fail_stack.avail++] = (fail_stack_elt_t) item - -/* The complement operation. Assumes `fail_stack' is nonempty. */ -#define POP_FAILURE_ITEM() fail_stack.stack[--fail_stack.avail] - -/* Used to omit pushing failure point id's when we're not debugging. */ -#ifdef DEBUG -#define DEBUG_PUSH PUSH_FAILURE_ITEM -#define DEBUG_POP(item_addr) *(item_addr) = POP_FAILURE_ITEM () -#else -#define DEBUG_PUSH(item) -#define DEBUG_POP(item_addr) -#endif - - -/* Push the information about the state we will need - if we ever fail back to it. - - Requires variables fail_stack, regstart, regend, reg_info, and - num_regs be declared. DOUBLE_FAIL_STACK requires `destination' be - declared. - - Does `return FAILURE_CODE' if runs out of memory. */ - -#define PUSH_FAILURE_POINT(pattern_place, string_place, failure_code) \ - do { \ - char *destination; \ - /* Must be int, so when we don't save any registers, the arithmetic \ - of 0 + -1 isn't done as unsigned. */ \ - int this_reg; \ - \ - DEBUG_STATEMENT (failure_id++); \ - DEBUG_STATEMENT (nfailure_points_pushed++); \ - DEBUG_PRINT2 ("\nPUSH_FAILURE_POINT #%u:\n", failure_id); \ - DEBUG_PRINT2 (" Before push, next avail: %d\n", (fail_stack).avail);\ - DEBUG_PRINT2 (" size: %d\n", (fail_stack).size);\ - \ - DEBUG_PRINT2 (" slots needed: %d\n", NUM_FAILURE_ITEMS); \ - DEBUG_PRINT2 (" available: %d\n", REMAINING_AVAIL_SLOTS); \ - \ - /* Ensure we have enough space allocated for what we will push. */ \ - while (REMAINING_AVAIL_SLOTS < NUM_FAILURE_ITEMS) \ - { \ - if (!DOUBLE_FAIL_STACK (fail_stack)) \ - return failure_code; \ - \ - DEBUG_PRINT2 ("\n Doubled stack; size now: %d\n", \ - (fail_stack).size); \ - DEBUG_PRINT2 (" slots available: %d\n", REMAINING_AVAIL_SLOTS);\ - } \ - \ - /* Push the info, starting with the registers. */ \ - DEBUG_PRINT1 ("\n"); \ - \ - for (this_reg = lowest_active_reg; this_reg <= highest_active_reg; \ - this_reg++) \ - { \ - DEBUG_PRINT2 (" Pushing reg: %d\n", this_reg); \ - DEBUG_STATEMENT (num_regs_pushed++); \ - \ - DEBUG_PRINT2 (" start: 0x%x\n", regstart[this_reg]); \ - PUSH_FAILURE_ITEM (regstart[this_reg]); \ - \ - DEBUG_PRINT2 (" end: 0x%x\n", regend[this_reg]); \ - PUSH_FAILURE_ITEM (regend[this_reg]); \ - \ - DEBUG_PRINT2 (" info: 0x%x\n ", reg_info[this_reg]); \ - DEBUG_PRINT2 (" match_null=%d", \ - REG_MATCH_NULL_STRING_P (reg_info[this_reg])); \ - DEBUG_PRINT2 (" active=%d", IS_ACTIVE (reg_info[this_reg])); \ - DEBUG_PRINT2 (" matched_something=%d", \ - MATCHED_SOMETHING (reg_info[this_reg])); \ - DEBUG_PRINT2 (" ever_matched=%d", \ - EVER_MATCHED_SOMETHING (reg_info[this_reg])); \ - DEBUG_PRINT1 ("\n"); \ - PUSH_FAILURE_ITEM (reg_info[this_reg].word); \ - } \ - \ - DEBUG_PRINT2 (" Pushing low active reg: %d\n", lowest_active_reg);\ - PUSH_FAILURE_ITEM (lowest_active_reg); \ - \ - DEBUG_PRINT2 (" Pushing high active reg: %d\n", highest_active_reg);\ - PUSH_FAILURE_ITEM (highest_active_reg); \ - \ - DEBUG_PRINT2 (" Pushing pattern 0x%x: ", pattern_place); \ - DEBUG_PRINT_COMPILED_PATTERN (bufp, pattern_place, pend); \ - PUSH_FAILURE_ITEM (pattern_place); \ - \ - DEBUG_PRINT2 (" Pushing string 0x%x: `", string_place); \ - DEBUG_PRINT_DOUBLE_STRING (string_place, string1, size1, string2, \ - size2); \ - DEBUG_PRINT1 ("'\n"); \ - PUSH_FAILURE_ITEM (string_place); \ - \ - DEBUG_PRINT2 (" Pushing failure id: %u\n", failure_id); \ - DEBUG_PUSH (failure_id); \ - } while (0) - -/* This is the number of items that are pushed and popped on the stack - for each register. */ -#define NUM_REG_ITEMS 3 - -/* Individual items aside from the registers. */ -#ifdef DEBUG -#define NUM_NONREG_ITEMS 5 /* Includes failure point id. */ -#else -#define NUM_NONREG_ITEMS 4 -#endif - -/* We push at most this many items on the stack. */ -#define MAX_FAILURE_ITEMS ((num_regs - 1) * NUM_REG_ITEMS + NUM_NONREG_ITEMS) - -/* We actually push this many items. */ -#define NUM_FAILURE_ITEMS \ - ((highest_active_reg - lowest_active_reg + 1) * NUM_REG_ITEMS \ - + NUM_NONREG_ITEMS) - -/* How many items can still be added to the stack without overflowing it. */ -#define REMAINING_AVAIL_SLOTS ((fail_stack).size - (fail_stack).avail) - - -/* Pops what PUSH_FAIL_STACK pushes. - - We restore into the parameters, all of which should be lvalues: - STR -- the saved data position. - PAT -- the saved pattern position. - LOW_REG, HIGH_REG -- the highest and lowest active registers. - REGSTART, REGEND -- arrays of string positions. - REG_INFO -- array of information about each subexpression. - - Also assumes the variables `fail_stack' and (if debugging), `bufp', - `pend', `string1', `size1', `string2', and `size2'. */ - -#define POP_FAILURE_POINT(str, pat, low_reg, high_reg, regstart, regend, reg_info)\ -{ \ - DEBUG_STATEMENT (fail_stack_elt_t failure_id;) \ - int this_reg; \ - const unsigned char *string_temp; \ - \ - assert (!FAIL_STACK_EMPTY ()); \ - \ - /* Remove failure points and point to how many regs pushed. */ \ - DEBUG_PRINT1 ("POP_FAILURE_POINT:\n"); \ - DEBUG_PRINT2 (" Before pop, next avail: %d\n", fail_stack.avail); \ - DEBUG_PRINT2 (" size: %d\n", fail_stack.size); \ - \ - assert (fail_stack.avail >= NUM_NONREG_ITEMS); \ - \ - DEBUG_POP (&failure_id); \ - DEBUG_PRINT2 (" Popping failure id: %u\n", failure_id); \ - \ - /* If the saved string location is NULL, it came from an \ - on_failure_keep_string_jump opcode, and we want to throw away the \ - saved NULL, thus retaining our current position in the string. */ \ - string_temp = POP_FAILURE_ITEM (); \ - if (string_temp != NULL) \ - str = (const char *) string_temp; \ - \ - DEBUG_PRINT2 (" Popping string 0x%x: `", str); \ - DEBUG_PRINT_DOUBLE_STRING (str, string1, size1, string2, size2); \ - DEBUG_PRINT1 ("'\n"); \ - \ - pat = (unsigned char *) POP_FAILURE_ITEM (); \ - DEBUG_PRINT2 (" Popping pattern 0x%x: ", pat); \ - DEBUG_PRINT_COMPILED_PATTERN (bufp, pat, pend); \ - \ - /* Restore register info. */ \ - high_reg = (unsigned) POP_FAILURE_ITEM (); \ - DEBUG_PRINT2 (" Popping high active reg: %d\n", high_reg); \ - \ - low_reg = (unsigned) POP_FAILURE_ITEM (); \ - DEBUG_PRINT2 (" Popping low active reg: %d\n", low_reg); \ - \ - for (this_reg = high_reg; this_reg >= low_reg; this_reg--) \ - { \ - DEBUG_PRINT2 (" Popping reg: %d\n", this_reg); \ - \ - reg_info[this_reg].word = POP_FAILURE_ITEM (); \ - DEBUG_PRINT2 (" info: 0x%x\n", reg_info[this_reg]); \ - \ - regend[this_reg] = (const char *) POP_FAILURE_ITEM (); \ - DEBUG_PRINT2 (" end: 0x%x\n", regend[this_reg]); \ - \ - regstart[this_reg] = (const char *) POP_FAILURE_ITEM (); \ - DEBUG_PRINT2 (" start: 0x%x\n", regstart[this_reg]); \ - } \ - \ - DEBUG_STATEMENT (nfailure_points_popped++); \ -} /* POP_FAILURE_POINT */ - -/* re_compile_fastmap computes a ``fastmap'' for the compiled pattern in - BUFP. A fastmap records which of the (1 << BYTEWIDTH) possible - characters can start a string that matches the pattern. This fastmap - is used by re_search to skip quickly over impossible starting points. - - The caller must supply the address of a (1 << BYTEWIDTH)-byte data - area as BUFP->fastmap. - - We set the `fastmap', `fastmap_accurate', and `can_be_null' fields in - the pattern buffer. - - Returns 0 if we succeed, -2 if an internal error. */ - -int -re_compile_fastmap (bufp) - struct re_pattern_buffer *bufp; -{ - int j, k; - fail_stack_type fail_stack; -#ifndef REGEX_MALLOC - char *destination; -#endif - /* We don't push any register information onto the failure stack. */ - unsigned num_regs = 0; - - register char *fastmap = bufp->fastmap; - unsigned char *pattern = bufp->buffer; - unsigned long size = bufp->used; - const unsigned char *p = pattern; - register unsigned char *pend = pattern + size; - - /* Assume that each path through the pattern can be null until - proven otherwise. We set this false at the bottom of switch - statement, to which we get only if a particular path doesn't - match the empty string. */ - boolean path_can_be_null = true; - - /* We aren't doing a `succeed_n' to begin with. */ - boolean succeed_n_p = false; - - assert (fastmap != NULL && p != NULL); - - INIT_FAIL_STACK (); - bzero (fastmap, 1 << BYTEWIDTH); /* Assume nothing's valid. */ - bufp->fastmap_accurate = 1; /* It will be when we're done. */ - bufp->can_be_null = 0; - - while (p != pend || !FAIL_STACK_EMPTY ()) - { - if (p == pend) - { - bufp->can_be_null |= path_can_be_null; - - /* Reset for next path. */ - path_can_be_null = true; - - p = fail_stack.stack[--fail_stack.avail]; - } - - /* We should never be about to go beyond the end of the pattern. */ - assert (p < pend); - -#ifdef SWITCH_ENUM_BUG - switch ((int) ((re_opcode_t) *p++)) -#else - switch ((re_opcode_t) *p++) -#endif - { - - /* I guess the idea here is to simply not bother with a fastmap - if a backreference is used, since it's too hard to figure out - the fastmap for the corresponding group. Setting - `can_be_null' stops `re_search_2' from using the fastmap, so - that is all we do. */ - case duplicate: - bufp->can_be_null = 1; - return 0; - - - /* Following are the cases which match a character. These end - with `break'. */ - - case exactn: - fastmap[p[1]] = 1; - break; - - - case charset: - for (j = *p++ * BYTEWIDTH - 1; j >= 0; j--) - if (p[j / BYTEWIDTH] & (1 << (j % BYTEWIDTH))) - fastmap[j] = 1; - break; - - - case charset_not: - /* Chars beyond end of map must be allowed. */ - for (j = *p * BYTEWIDTH; j < (1 << BYTEWIDTH); j++) - fastmap[j] = 1; - - for (j = *p++ * BYTEWIDTH - 1; j >= 0; j--) - if (!(p[j / BYTEWIDTH] & (1 << (j % BYTEWIDTH)))) - fastmap[j] = 1; - break; - - - case wordchar: - for (j = 0; j < (1 << BYTEWIDTH); j++) - if (SYNTAX (j) == Sword) - fastmap[j] = 1; - break; - - - case notwordchar: - for (j = 0; j < (1 << BYTEWIDTH); j++) - if (SYNTAX (j) != Sword) - fastmap[j] = 1; - break; - - - case anychar: - /* `.' matches anything ... */ - for (j = 0; j < (1 << BYTEWIDTH); j++) - fastmap[j] = 1; - - /* ... except perhaps newline. */ - if (!(bufp->syntax & RE_DOT_NEWLINE)) - fastmap['\n'] = 0; - - /* Return if we have already set `can_be_null'; if we have, - then the fastmap is irrelevant. Something's wrong here. */ - else if (bufp->can_be_null) - return 0; - - /* Otherwise, have to check alternative paths. */ - break; - - -#ifdef emacs - case syntaxspec: - k = *p++; - for (j = 0; j < (1 << BYTEWIDTH); j++) - if (SYNTAX (j) == (enum syntaxcode) k) - fastmap[j] = 1; - break; - - - case notsyntaxspec: - k = *p++; - for (j = 0; j < (1 << BYTEWIDTH); j++) - if (SYNTAX (j) != (enum syntaxcode) k) - fastmap[j] = 1; - break; - - - /* All cases after this match the empty string. These end with - `continue'. */ - - - case before_dot: - case at_dot: - case after_dot: - continue; -#endif /* not emacs */ - - - case no_op: - case begline: - case endline: - case begbuf: - case endbuf: - case wordbound: - case notwordbound: - case wordbeg: - case wordend: - case push_dummy_failure: - continue; - - - case jump_n: - case pop_failure_jump: - case maybe_pop_jump: - case jump: - case jump_past_alt: - case dummy_failure_jump: - EXTRACT_NUMBER_AND_INCR (j, p); - p += j; - if (j > 0) - continue; - - /* Jump backward implies we just went through the body of a - loop and matched nothing. Opcode jumped to should be - `on_failure_jump' or `succeed_n'. Just treat it like an - ordinary jump. For a * loop, it has pushed its failure - point already; if so, discard that as redundant. */ - if ((re_opcode_t) *p != on_failure_jump - && (re_opcode_t) *p != succeed_n) - continue; - - p++; - EXTRACT_NUMBER_AND_INCR (j, p); - p += j; - - /* If what's on the stack is where we are now, pop it. */ - if (!FAIL_STACK_EMPTY () - && fail_stack.stack[fail_stack.avail - 1] == p) - fail_stack.avail--; - - continue; - - - case on_failure_jump: - case on_failure_keep_string_jump: - handle_on_failure_jump: - EXTRACT_NUMBER_AND_INCR (j, p); - - /* For some patterns, e.g., `(a?)?', `p+j' here points to the - end of the pattern. We don't want to push such a point, - since when we restore it above, entering the switch will - increment `p' past the end of the pattern. We don't need - to push such a point since we obviously won't find any more - fastmap entries beyond `pend'. Such a pattern can match - the null string, though. */ - if (p + j < pend) - { - if (!PUSH_PATTERN_OP (p + j, fail_stack)) - return -2; - } - else - bufp->can_be_null = 1; - - if (succeed_n_p) - { - EXTRACT_NUMBER_AND_INCR (k, p); /* Skip the n. */ - succeed_n_p = false; - } - - continue; - - - case succeed_n: - /* Get to the number of times to succeed. */ - p += 2; - - /* Increment p past the n for when k != 0. */ - EXTRACT_NUMBER_AND_INCR (k, p); - if (k == 0) - { - p -= 4; - succeed_n_p = true; /* Spaghetti code alert. */ - goto handle_on_failure_jump; - } - continue; - - - case set_number_at: - p += 4; - continue; - - - case start_memory: - case stop_memory: - p += 2; - continue; - - - default: - abort (); /* We have listed all the cases. */ - } /* switch *p++ */ - - /* Getting here means we have found the possible starting - characters for one path of the pattern -- and that the empty - string does not match. We need not follow this path further. - Instead, look at the next alternative (remembered on the - stack), or quit if no more. The test at the top of the loop - does these things. */ - path_can_be_null = false; - p = pend; - } /* while p */ - - /* Set `can_be_null' for the last path (also the first path, if the - pattern is empty). */ - bufp->can_be_null |= path_can_be_null; - return 0; -} /* re_compile_fastmap */ - -/* 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 (bufp, regs, num_regs, starts, ends) - struct re_pattern_buffer *bufp; - struct re_registers *regs; - unsigned num_regs; - regoff_t *starts, *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; - } -} - -/* Searching routines. */ - -/* Like re_search_2, below, but only one string is specified, and - doesn't let you say where to stop matching. */ - -int -re_search (bufp, string, size, startpos, range, regs) - struct re_pattern_buffer *bufp; - const char *string; - int size, startpos, range; - struct re_registers *regs; -{ - return re_search_2 (bufp, NULL, 0, string, size, startpos, range, - regs, size); -} - - -/* Using the compiled pattern in BUFP->buffer, first tries to match the - virtual concatenation of STRING1 and STRING2, starting first at index - STARTPOS, then at STARTPOS + 1, and so on. - - STRING1 and STRING2 have length SIZE1 and SIZE2, respectively. - - RANGE is how far to scan while trying to match. RANGE = 0 means try - only at STARTPOS; in general, the last start tried is STARTPOS + - RANGE. - - In REGS, return the indices of the virtual concatenation of STRING1 - and STRING2 that matched the entire BUFP->buffer and its contained - subexpressions. - - Do not consider matching one past the index STOP in the virtual - concatenation of STRING1 and STRING2. - - We return either the position in the strings at which the match was - found, -1 if no match, or -2 if error (such as failure - stack overflow). */ - -int -re_search_2 (bufp, string1, size1, string2, size2, startpos, range, regs, stop) - struct re_pattern_buffer *bufp; - const char *string1, *string2; - int size1, size2; - int startpos; - int range; - struct re_registers *regs; - int stop; -{ - int val; - register char *fastmap = bufp->fastmap; - register char *translate = bufp->translate; - int total_size = size1 + size2; - int endpos = startpos + range; - - /* Check for out-of-range STARTPOS. */ - if (startpos < 0 || startpos > total_size) - return -1; - - /* Fix up RANGE if it might eventually take us outside - the virtual concatenation of STRING1 and STRING2. */ - if (endpos < -1) - range = -1 - startpos; - else if (endpos > total_size) - range = total_size - startpos; - - /* If the search isn't to be a backwards one, don't waste time in a - search for a pattern that must be anchored. */ - if (bufp->used > 0 && (re_opcode_t) bufp->buffer[0] == begbuf && range > 0) - { - if (startpos > 0) - return -1; - else - range = 1; - } - - /* Update the fastmap now if not correct already. */ - if (fastmap && !bufp->fastmap_accurate) - if (re_compile_fastmap (bufp) == -2) - return -2; - - /* Loop through the string, looking for a place to start matching. */ - for (;;) - { - /* If a fastmap is supplied, skip quickly over characters that - cannot be the start of a match. If the pattern can match the - null string, however, we don't need to skip characters; we want - the first null string. */ - if (fastmap && startpos < total_size && !bufp->can_be_null) - { - if (range > 0) /* Searching forwards. */ - { - register const char *d; - register int lim = 0; - int irange = range; - - if (startpos < size1 && startpos + range >= size1) - lim = range - (size1 - startpos); - - d = (startpos >= size1 ? string2 - size1 : string1) + startpos; - - /* Written out as an if-else to avoid testing `translate' - inside the loop. */ - if (translate) - while (range > lim - && !fastmap[(unsigned char) - translate[(unsigned char) *d++]]) - range--; - else - while (range > lim && !fastmap[(unsigned char) *d++]) - range--; - - startpos += irange - range; - } - else /* Searching backwards. */ - { - register char c = (size1 == 0 || startpos >= size1 - ? string2[startpos - size1] - : string1[startpos]); - - if (!fastmap[(unsigned char) TRANSLATE (c)]) - goto advance; - } - } - - /* If can't match the null string, and that's all we have left, fail. */ - if (range >= 0 && startpos == total_size && fastmap - && !bufp->can_be_null) - return -1; - - val = re_match_2 (bufp, string1, size1, string2, size2, - startpos, regs, stop); - if (val >= 0) - return startpos; - - if (val == -2) - return -2; - - advance: - if (!range) - break; - else if (range > 0) - { - range--; - startpos++; - } - else - { - range++; - startpos--; - } - } - return -1; -} /* re_search_2 */ - -/* Declarations and macros for re_match_2. */ - -static int bcmp_translate (); -static boolean alt_match_null_string_p (), - common_op_match_null_string_p (), - group_match_null_string_p (); - -/* Structure for per-register (a.k.a. per-group) information. - This must not be longer than one word, because we push this value - onto the failure stack. Other register information, such as the - starting and ending positions (which are addresses), and the list of - inner groups (which is a bits list) are maintained in separate - variables. - - We are making a (strictly speaking) nonportable assumption here: that - the compiler will pack our bit fields into something that fits into - the type of `word', i.e., is something that fits into one item on the - failure stack. */ -typedef union -{ - fail_stack_elt_t word; - struct - { - /* This field is one if this group can match the empty string, - zero if not. If not yet determined, `MATCH_NULL_UNSET_VALUE'. */ -#define MATCH_NULL_UNSET_VALUE 3 - unsigned match_null_string_p : 2; - unsigned is_active : 1; - unsigned matched_something : 1; - unsigned ever_matched_something : 1; - } bits; -} register_info_type; - -#define REG_MATCH_NULL_STRING_P(R) ((R).bits.match_null_string_p) -#define IS_ACTIVE(R) ((R).bits.is_active) -#define MATCHED_SOMETHING(R) ((R).bits.matched_something) -#define EVER_MATCHED_SOMETHING(R) ((R).bits.ever_matched_something) - - -/* Call this when have matched a real character; it sets `matched' flags - for the subexpressions which we are currently inside. Also records - that those subexprs have matched. */ -#define SET_REGS_MATCHED() \ - do \ - { \ - unsigned r; \ - for (r = lowest_active_reg; r <= highest_active_reg; r++) \ - { \ - MATCHED_SOMETHING (reg_info[r]) \ - = EVER_MATCHED_SOMETHING (reg_info[r]) \ - = 1; \ - } \ - } \ - while (0) - - -/* This converts PTR, a pointer into one of the search strings `string1' - and `string2' into an offset from the beginning of that string. */ -#define POINTER_TO_OFFSET(ptr) \ - (FIRST_STRING_P (ptr) ? (ptr) - string1 : (ptr) - string2 + size1) - -/* Registers are set to a sentinel when they haven't yet matched. */ -#define REG_UNSET_VALUE ((char *) -1) -#define REG_UNSET(e) ((e) == REG_UNSET_VALUE) - - -/* Macros for dealing with the split strings in re_match_2. */ - -#define MATCHING_IN_FIRST_STRING (dend == end_match_1) - -/* Call before fetching a character with *d. This switches over to - string2 if necessary. */ -#define PREFETCH() \ - while (d == dend) \ - { \ - /* End of string2 => fail. */ \ - if (dend == end_match_2) \ - goto fail; \ - /* End of string1 => advance to string2. */ \ - d = string2; \ - dend = end_match_2; \ - } - - -/* Test if at very beginning or at very end of the virtual concatenation - of `string1' and `string2'. If only one string, it's `string2'. */ -#define AT_STRINGS_BEG(d) ((d) == (size1 ? string1 : string2) || !size2) -#define AT_STRINGS_END(d) ((d) == end2) - - -/* Test if D points to a character which is word-constituent. We have - two special cases to check for: if past the end of string1, look at - the first character in string2; and if before the beginning of - string2, look at the last character in string1. */ -#define WORDCHAR_P(d) \ - (SYNTAX ((d) == end1 ? *string2 \ - : (d) == string2 - 1 ? *(end1 - 1) : *(d)) \ - == Sword) - -/* Test if the character before D and the one at D differ with respect - to being word-constituent. */ -#define AT_WORD_BOUNDARY(d) \ - (AT_STRINGS_BEG (d) || AT_STRINGS_END (d) \ - || WORDCHAR_P (d - 1) != WORDCHAR_P (d)) - - -/* Free everything we malloc. */ -#ifdef REGEX_MALLOC -#define FREE_VAR(var) if (var) free (var); var = NULL -#define FREE_VARIABLES() \ - do { \ - FREE_VAR (fail_stack.stack); \ - FREE_VAR (regstart); \ - FREE_VAR (regend); \ - FREE_VAR (old_regstart); \ - FREE_VAR (old_regend); \ - FREE_VAR (best_regstart); \ - FREE_VAR (best_regend); \ - FREE_VAR (reg_info); \ - FREE_VAR (reg_dummy); \ - FREE_VAR (reg_info_dummy); \ - } while (0) -#else /* not REGEX_MALLOC */ -/* Some MIPS systems (at least) want this to free alloca'd storage. */ -#define FREE_VARIABLES() alloca (0) -#endif /* not REGEX_MALLOC */ - - -/* These values must meet several constraints. They must not be valid - register values; since we have a limit of 255 registers (because - we use only one byte in the pattern for the register number), we can - use numbers larger than 255. They must differ by 1, because of - NUM_FAILURE_ITEMS above. And the value for the lowest register must - be larger than the value for the highest register, so we do not try - to actually save any registers when none are active. */ -#define NO_HIGHEST_ACTIVE_REG (1 << BYTEWIDTH) -#define NO_LOWEST_ACTIVE_REG (NO_HIGHEST_ACTIVE_REG + 1) - -/* Matching routines. */ - -#ifndef emacs /* Emacs never uses this. */ -/* re_match is like re_match_2 except it takes only a single string. */ - -int -re_match (bufp, string, size, pos, regs) - struct re_pattern_buffer *bufp; - const char *string; - int size, pos; - struct re_registers *regs; - { - return re_match_2 (bufp, NULL, 0, string, size, pos, regs, size); -} -#endif /* not emacs */ - - -/* re_match_2 matches the compiled pattern in BUFP against the - (virtual) concatenation of STRING1 and STRING2 (of length SIZE1 - and SIZE2, respectively). We start matching at POS, and stop - matching at STOP. - - If REGS is non-null and the `no_sub' field of BUFP is nonzero, we - store offsets for the substring each group matched in REGS. See the - documentation for exactly how many groups we fill. - - We return -1 if no match, -2 if an internal error (such as the - failure stack overflowing). Otherwise, we return the length of the - matched substring. */ - -int -re_match_2 (bufp, string1, size1, string2, size2, pos, regs, stop) - struct re_pattern_buffer *bufp; - const char *string1, *string2; - int size1, size2; - int pos; - struct re_registers *regs; - int stop; -{ - /* General temporaries. */ - int mcnt; - unsigned char *p1; - - /* Just past the end of the corresponding string. */ - const char *end1, *end2; - - /* Pointers into string1 and string2, just past the last characters in - each to consider matching. */ - const char *end_match_1, *end_match_2; - - /* Where we are in the data, and the end of the current string. */ - const char *d, *dend; - - /* Where we are in the pattern, and the end of the pattern. */ - unsigned char *p = bufp->buffer; - register unsigned char *pend = p + bufp->used; - - /* We use this to map every character in the string. */ - char *translate = bufp->translate; - - /* Failure point stack. Each place that can handle a failure further - down the line pushes a failure point on this stack. It consists of - restart, regend, and reg_info for all registers corresponding to - the subexpressions we're currently inside, plus the number of such - registers, and, finally, two char *'s. The first char * is where - to resume scanning the pattern; the second one is where to resume - scanning the strings. If the latter is zero, the failure point is - a ``dummy''; if a failure happens and the failure point is a dummy, - it gets discarded and the next next one is tried. */ - fail_stack_type fail_stack; -#ifdef DEBUG - static unsigned failure_id = 0; - unsigned nfailure_points_pushed = 0, nfailure_points_popped = 0; -#endif - - /* We fill all the registers internally, independent of what we - return, for use in backreferences. The number here includes - an element for register zero. */ - unsigned num_regs = bufp->re_nsub + 1; - - /* The currently active registers. */ - unsigned lowest_active_reg = NO_LOWEST_ACTIVE_REG; - unsigned highest_active_reg = NO_HIGHEST_ACTIVE_REG; - - /* Information on the contents of registers. These are pointers into - the input strings; they record just what was matched (on this - attempt) by a subexpression part of the pattern, that is, the - regnum-th regstart pointer points to where in the pattern we began - matching and the regnum-th regend points to right after where we - stopped matching the regnum-th subexpression. (The zeroth register - keeps track of what the whole pattern matches.) */ - const char **regstart = NULL, **regend = NULL; - - /* If a group that's operated upon by a repetition operator fails to - match anything, then the register for its start will need to be - restored because it will have been set to wherever in the string we - are when we last see its open-group operator. Similarly for a - register's end. */ - const char **old_regstart = NULL, **old_regend = NULL; - - /* The is_active field of reg_info helps us keep track of which (possibly - nested) subexpressions we are currently in. The matched_something - field of reg_info[reg_num] helps us tell whether or not we have - matched any of the pattern so far this time through the reg_num-th - subexpression. These two fields get reset each time through any - loop their register is in. */ - register_info_type *reg_info = NULL; - - /* The following record the register info as found in the above - variables when we find a match better than any we've seen before. - This happens as we backtrack through the failure points, which in - turn happens only if we have not yet matched the entire string. */ - unsigned best_regs_set = false; - const char **best_regstart = NULL, **best_regend = NULL; - - /* Logically, this is `best_regend[0]'. But we don't want to have to - allocate space for that if we're not allocating space for anything - else (see below). Also, we never need info about register 0 for - any of the other register vectors, and it seems rather a kludge to - treat `best_regend' differently than the rest. So we keep track of - the end of the best match so far in a separate variable. We - initialize this to NULL so that when we backtrack the first time - and need to test it, it's not garbage. */ - const char *match_end = NULL; - - /* Used when we pop values we don't care about. */ - const char **reg_dummy = NULL; - register_info_type *reg_info_dummy = NULL; - -#ifdef DEBUG - /* Counts the total number of registers pushed. */ - unsigned num_regs_pushed = 0; -#endif - - DEBUG_PRINT1 ("\n\nEntering re_match_2.\n"); - - INIT_FAIL_STACK (); - - /* Do not bother to initialize all the register variables if there are - no groups in the pattern, as it takes a fair amount of time. If - there are groups, we include space for register 0 (the whole - pattern), even though we never use it, since it simplifies the - array indexing. We should fix this. */ - if (bufp->re_nsub) - { - regstart = REGEX_TALLOC (num_regs, const char *); - regend = REGEX_TALLOC (num_regs, const char *); - old_regstart = REGEX_TALLOC (num_regs, const char *); - old_regend = REGEX_TALLOC (num_regs, const char *); - best_regstart = REGEX_TALLOC (num_regs, const char *); - best_regend = REGEX_TALLOC (num_regs, const char *); - reg_info = REGEX_TALLOC (num_regs, register_info_type); - reg_dummy = REGEX_TALLOC (num_regs, const char *); - reg_info_dummy = REGEX_TALLOC (num_regs, register_info_type); - - if (!(regstart && regend && old_regstart && old_regend && reg_info - && best_regstart && best_regend && reg_dummy && reg_info_dummy)) - { - FREE_VARIABLES (); - return -2; - } - } -#ifdef REGEX_MALLOC - else - { - /* We must initialize all our variables to NULL, so that - `FREE_VARIABLES' doesn't try to free them. */ - regstart = regend = old_regstart = old_regend = best_regstart - = best_regend = reg_dummy = NULL; - reg_info = reg_info_dummy = (register_info_type *) NULL; - } -#endif /* REGEX_MALLOC */ - - /* The starting position is bogus. */ - if (pos < 0 || pos > size1 + size2) - { - FREE_VARIABLES (); - return -1; - } - - /* Initialize subexpression text positions to -1 to mark ones that no - start_memory/stop_memory has been seen for. Also initialize the - register information struct. */ - for (mcnt = 1; mcnt < num_regs; mcnt++) - { - regstart[mcnt] = regend[mcnt] - = old_regstart[mcnt] = old_regend[mcnt] = REG_UNSET_VALUE; - - REG_MATCH_NULL_STRING_P (reg_info[mcnt]) = MATCH_NULL_UNSET_VALUE; - IS_ACTIVE (reg_info[mcnt]) = 0; - MATCHED_SOMETHING (reg_info[mcnt]) = 0; - EVER_MATCHED_SOMETHING (reg_info[mcnt]) = 0; - } - - /* We move `string1' into `string2' if the latter's empty -- but not if - `string1' is null. */ - if (size2 == 0 && string1 != NULL) - { - string2 = string1; - size2 = size1; - string1 = 0; - size1 = 0; - } - end1 = string1 + size1; - end2 = string2 + size2; - - /* Compute where to stop matching, within the two strings. */ - if (stop <= size1) - { - end_match_1 = string1 + stop; - end_match_2 = string2; - } - else - { - end_match_1 = end1; - end_match_2 = string2 + stop - size1; - } - - /* `p' scans through the pattern as `d' scans through the data. - `dend' is the end of the input string that `d' points within. `d' - is advanced into the following input string whenever necessary, but - this happens before fetching; therefore, at the beginning of the - loop, `d' can be pointing at the end of a string, but it cannot - equal `string2'. */ - if (size1 > 0 && pos <= size1) - { - d = string1 + pos; - dend = end_match_1; - } - else - { - d = string2 + pos - size1; - dend = end_match_2; - } - - DEBUG_PRINT1 ("The compiled pattern is: "); - DEBUG_PRINT_COMPILED_PATTERN (bufp, p, pend); - DEBUG_PRINT1 ("The string to match is: `"); - DEBUG_PRINT_DOUBLE_STRING (d, string1, size1, string2, size2); - DEBUG_PRINT1 ("'\n"); - - /* This loops over pattern commands. It exits by returning from the - function if the match is complete, or it drops through if the match - fails at this starting point in the input data. */ - for (;;) - { - DEBUG_PRINT2 ("\n0x%x: ", p); - - if (p == pend) - { /* End of pattern means we might have succeeded. */ - DEBUG_PRINT1 ("end of pattern ... "); - - /* If we haven't matched the entire string, and we want the - longest match, try backtracking. */ - if (d != end_match_2) - { - DEBUG_PRINT1 ("backtracking.\n"); - - if (!FAIL_STACK_EMPTY ()) - { /* More failure points to try. */ - boolean same_str_p = (FIRST_STRING_P (match_end) - == MATCHING_IN_FIRST_STRING); - - /* If exceeds best match so far, save it. */ - if (!best_regs_set - || (same_str_p && d > match_end) - || (!same_str_p && !MATCHING_IN_FIRST_STRING)) - { - best_regs_set = true; - match_end = d; - - DEBUG_PRINT1 ("\nSAVING match as best so far.\n"); - - for (mcnt = 1; mcnt < num_regs; mcnt++) - { - best_regstart[mcnt] = regstart[mcnt]; - best_regend[mcnt] = regend[mcnt]; - } - } - goto fail; - } - - /* If no failure points, don't restore garbage. */ - else if (best_regs_set) - { - restore_best_regs: - /* Restore best match. It may happen that `dend == - end_match_1' while the restored d is in string2. - For example, the pattern `x.*y.*z' against the - strings `x-' and `y-z-', if the two strings are - not consecutive in memory. */ - DEBUG_PRINT1 ("Restoring best registers.\n"); - - d = match_end; - dend = ((d >= string1 && d <= end1) - ? end_match_1 : end_match_2); - - for (mcnt = 1; mcnt < num_regs; mcnt++) - { - regstart[mcnt] = best_regstart[mcnt]; - regend[mcnt] = best_regend[mcnt]; - } - } - } /* d != end_match_2 */ - - DEBUG_PRINT1 ("Accepting match.\n"); - - /* If caller wants register contents data back, do it. */ - if (regs && !bufp->no_sub) - { - /* Have the register data arrays been allocated? */ - if (bufp->regs_allocated == REGS_UNALLOCATED) - { /* No. So allocate them with malloc. We need one - extra element beyond `num_regs' for the `-1' marker - GNU code uses. */ - regs->num_regs = MAX (RE_NREGS, num_regs + 1); - regs->start = TALLOC (regs->num_regs, regoff_t); - regs->end = TALLOC (regs->num_regs, regoff_t); - if (regs->start == NULL || regs->end == NULL) - return -2; - bufp->regs_allocated = REGS_REALLOCATE; - } - else if (bufp->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 (regs->num_regs < num_regs + 1) - { - regs->num_regs = num_regs + 1; - RETALLOC (regs->start, regs->num_regs, regoff_t); - RETALLOC (regs->end, regs->num_regs, regoff_t); - if (regs->start == NULL || regs->end == NULL) - return -2; - } - } - else - assert (bufp->regs_allocated == REGS_FIXED); - - /* Convert the pointer data in `regstart' and `regend' to - indices. Register zero has to be set differently, - since we haven't kept track of any info for it. */ - if (regs->num_regs > 0) - { - regs->start[0] = pos; - regs->end[0] = (MATCHING_IN_FIRST_STRING ? d - string1 - : d - string2 + size1); - } - - /* Go through the first `min (num_regs, regs->num_regs)' - registers, since that is all we initialized. */ - for (mcnt = 1; mcnt < MIN (num_regs, regs->num_regs); mcnt++) - { - if (REG_UNSET (regstart[mcnt]) || REG_UNSET (regend[mcnt])) - regs->start[mcnt] = regs->end[mcnt] = -1; - else - { - regs->start[mcnt] = POINTER_TO_OFFSET (regstart[mcnt]); - regs->end[mcnt] = POINTER_TO_OFFSET (regend[mcnt]); - } - } - - /* If the regs structure we return has more elements than - were in the pattern, set the extra elements to -1. If - we (re)allocated the registers, this is the case, - because we always allocate enough to have at least one - -1 at the end. */ - for (mcnt = num_regs; mcnt < regs->num_regs; mcnt++) - regs->start[mcnt] = regs->end[mcnt] = -1; - } /* regs && !bufp->no_sub */ - - FREE_VARIABLES (); - DEBUG_PRINT4 ("%u failure points pushed, %u popped (%u remain).\n", - nfailure_points_pushed, nfailure_points_popped, - nfailure_points_pushed - nfailure_points_popped); - DEBUG_PRINT2 ("%u registers pushed.\n", num_regs_pushed); - - mcnt = d - pos - (MATCHING_IN_FIRST_STRING - ? string1 - : string2 - size1); - - DEBUG_PRINT2 ("Returning %d from re_match_2.\n", mcnt); - - return mcnt; - } - - /* Otherwise match next pattern command. */ -#ifdef SWITCH_ENUM_BUG - switch ((int) ((re_opcode_t) *p++)) -#else - switch ((re_opcode_t) *p++) -#endif - { - /* Ignore these. Used to ignore the n of succeed_n's which - currently have n == 0. */ - case no_op: - DEBUG_PRINT1 ("EXECUTING no_op.\n"); - break; - - - /* Match the next n pattern characters exactly. The following - byte in the pattern defines n, and the n bytes after that - are the characters to match. */ - case exactn: - mcnt = *p++; - DEBUG_PRINT2 ("EXECUTING exactn %d.\n", mcnt); - - /* This is written out as an if-else so we don't waste time - testing `translate' inside the loop. */ - if (translate) - { - do - { - PREFETCH (); - if (translate[(unsigned char) *d++] != (char) *p++) - goto fail; - } - while (--mcnt); - } - else - { - do - { - PREFETCH (); - if (*d++ != (char) *p++) goto fail; - } - while (--mcnt); - } - SET_REGS_MATCHED (); - break; - - - /* Match any character except possibly a newline or a null. */ - case anychar: - DEBUG_PRINT1 ("EXECUTING anychar.\n"); - - PREFETCH (); - - if ((!(bufp->syntax & RE_DOT_NEWLINE) && TRANSLATE (*d) == '\n') - || (bufp->syntax & RE_DOT_NOT_NULL && TRANSLATE (*d) == '\000')) - goto fail; - - SET_REGS_MATCHED (); - DEBUG_PRINT2 (" Matched `%d'.\n", *d); - d++; - break; - - - case charset: - case charset_not: - { - register unsigned char c; - boolean not = (re_opcode_t) *(p - 1) == charset_not; - - DEBUG_PRINT2 ("EXECUTING charset%s.\n", not ? "_not" : ""); - - PREFETCH (); - c = TRANSLATE (*d); /* The character to match. */ - - /* Cast to `unsigned' instead of `unsigned char' in case the - bit list is a full 32 bytes long. */ - if (c < (unsigned) (*p * BYTEWIDTH) - && p[1 + c / BYTEWIDTH] & (1 << (c % BYTEWIDTH))) - not = !not; - - p += 1 + *p; - - if (!not) goto fail; - - SET_REGS_MATCHED (); - d++; - break; - } - - - /* The beginning of a group is represented by start_memory. - The arguments are the register number in the next byte, and the - number of groups inner to this one in the next. The text - matched within the group is recorded (in the internal - registers data structure) under the register number. */ - case start_memory: - DEBUG_PRINT3 ("EXECUTING start_memory %d (%d):\n", *p, p[1]); - - /* Find out if this group can match the empty string. */ - p1 = p; /* To send to group_match_null_string_p. */ - - if (REG_MATCH_NULL_STRING_P (reg_info[*p]) == MATCH_NULL_UNSET_VALUE) - REG_MATCH_NULL_STRING_P (reg_info[*p]) - = group_match_null_string_p (&p1, pend, reg_info); - - /* Save the position in the string where we were the last time - we were at this open-group operator in case the group is - operated upon by a repetition operator, e.g., with `(a*)*b' - against `ab'; then we want to ignore where we are now in - the string in case this attempt to match fails. */ - old_regstart[*p] = REG_MATCH_NULL_STRING_P (reg_info[*p]) - ? REG_UNSET (regstart[*p]) ? d : regstart[*p] - : regstart[*p]; - DEBUG_PRINT2 (" old_regstart: %d\n", - POINTER_TO_OFFSET (old_regstart[*p])); - - regstart[*p] = d; - DEBUG_PRINT2 (" regstart: %d\n", POINTER_TO_OFFSET (regstart[*p])); - - IS_ACTIVE (reg_info[*p]) = 1; - MATCHED_SOMETHING (reg_info[*p]) = 0; - - /* This is the new highest active register. */ - highest_active_reg = *p; - - /* If nothing was active before, this is the new lowest active - register. */ - if (lowest_active_reg == NO_LOWEST_ACTIVE_REG) - lowest_active_reg = *p; - - /* Move past the register number and inner group count. */ - p += 2; - break; - - - /* The stop_memory opcode represents the end of a group. Its - arguments are the same as start_memory's: the register - number, and the number of inner groups. */ - case stop_memory: - DEBUG_PRINT3 ("EXECUTING stop_memory %d (%d):\n", *p, p[1]); - - /* We need to save the string position the last time we were at - this close-group operator in case the group is operated - upon by a repetition operator, e.g., with `((a*)*(b*)*)*' - against `aba'; then we want to ignore where we are now in - the string in case this attempt to match fails. */ - old_regend[*p] = REG_MATCH_NULL_STRING_P (reg_info[*p]) - ? REG_UNSET (regend[*p]) ? d : regend[*p] - : regend[*p]; - DEBUG_PRINT2 (" old_regend: %d\n", - POINTER_TO_OFFSET (old_regend[*p])); - - regend[*p] = d; - DEBUG_PRINT2 (" regend: %d\n", POINTER_TO_OFFSET (regend[*p])); - - /* This register isn't active anymore. */ - IS_ACTIVE (reg_info[*p]) = 0; - - /* If this was the only register active, nothing is active - anymore. */ - if (lowest_active_reg == highest_active_reg) - { - lowest_active_reg = NO_LOWEST_ACTIVE_REG; - highest_active_reg = NO_HIGHEST_ACTIVE_REG; - } - else - { /* We must scan for the new highest active register, since - it isn't necessarily one less than now: consider - (a(b)c(d(e)f)g). When group 3 ends, after the f), the - new highest active register is 1. */ - unsigned char r = *p - 1; - while (r > 0 && !IS_ACTIVE (reg_info[r])) - r--; - - /* If we end up at register zero, that means that we saved - the registers as the result of an `on_failure_jump', not - a `start_memory', and we jumped to past the innermost - `stop_memory'. For example, in ((.)*) we save - registers 1 and 2 as a result of the *, but when we pop - back to the second ), we are at the stop_memory 1. - Thus, nothing is active. */ - if (r == 0) - { - lowest_active_reg = NO_LOWEST_ACTIVE_REG; - highest_active_reg = NO_HIGHEST_ACTIVE_REG; - } - else - highest_active_reg = r; - } - - /* If just failed to match something this time around with a - group that's operated on by a repetition operator, try to - force exit from the ``loop'', and restore the register - information for this group that we had before trying this - last match. */ - if ((!MATCHED_SOMETHING (reg_info[*p]) - || (re_opcode_t) p[-3] == start_memory) - && (p + 2) < pend) - { - boolean is_a_jump_n = false; - - p1 = p + 2; - mcnt = 0; - switch ((re_opcode_t) *p1++) - { - case jump_n: - is_a_jump_n = true; - case pop_failure_jump: - case maybe_pop_jump: - case jump: - case dummy_failure_jump: - EXTRACT_NUMBER_AND_INCR (mcnt, p1); - if (is_a_jump_n) - p1 += 2; - break; - - default: - /* do nothing */ ; - } - p1 += mcnt; - - /* If the next operation is a jump backwards in the pattern - to an on_failure_jump right before the start_memory - corresponding to this stop_memory, exit from the loop - by forcing a failure after pushing on the stack the - on_failure_jump's jump in the pattern, and d. */ - if (mcnt < 0 && (re_opcode_t) *p1 == on_failure_jump - && (re_opcode_t) p1[3] == start_memory && p1[4] == *p) - { - /* If this group ever matched anything, then restore - what its registers were before trying this last - failed match, e.g., with `(a*)*b' against `ab' for - regstart[1], and, e.g., with `((a*)*(b*)*)*' - against `aba' for regend[3]. - - Also restore the registers for inner groups for, - e.g., `((a*)(b*))*' against `aba' (register 3 would - otherwise get trashed). */ - - if (EVER_MATCHED_SOMETHING (reg_info[*p])) - { - unsigned r; - - EVER_MATCHED_SOMETHING (reg_info[*p]) = 0; - - /* Restore this and inner groups' (if any) registers. */ - for (r = *p; r < *p + *(p + 1); r++) - { - regstart[r] = old_regstart[r]; - - /* xx why this test? */ - if ((int) old_regend[r] >= (int) regstart[r]) - regend[r] = old_regend[r]; - } - } - p1++; - EXTRACT_NUMBER_AND_INCR (mcnt, p1); - PUSH_FAILURE_POINT (p1 + mcnt, d, -2); - - goto fail; - } - } - - /* Move past the register number and the inner group count. */ - p += 2; - break; - - - /* \<digit> has been turned into a `duplicate' command which is - followed by the numeric value of <digit> as the register number. */ - case duplicate: - { - register const char *d2, *dend2; - int regno = *p++; /* Get which register to match against. */ - DEBUG_PRINT2 ("EXECUTING duplicate %d.\n", regno); - - /* Can't back reference a group which we've never matched. */ - if (REG_UNSET (regstart[regno]) || REG_UNSET (regend[regno])) - goto fail; - - /* Where in input to try to start matching. */ - d2 = regstart[regno]; - - /* Where to stop matching; if both the place to start and - the place to stop matching are in the same string, then - set to the place to stop, otherwise, for now have to use - the end of the first string. */ - - dend2 = ((FIRST_STRING_P (regstart[regno]) - == FIRST_STRING_P (regend[regno])) - ? regend[regno] : end_match_1); - for (;;) - { - /* If necessary, advance to next segment in register - contents. */ - while (d2 == dend2) - { - if (dend2 == end_match_2) break; - if (dend2 == regend[regno]) break; - - /* End of string1 => advance to string2. */ - d2 = string2; - dend2 = regend[regno]; - } - /* At end of register contents => success */ - if (d2 == dend2) break; - - /* If necessary, advance to next segment in data. */ - PREFETCH (); - - /* How many characters left in this segment to match. */ - mcnt = dend - d; - - /* Want how many consecutive characters we can match in - one shot, so, if necessary, adjust the count. */ - if (mcnt > dend2 - d2) - mcnt = dend2 - d2; - - /* Compare that many; failure if mismatch, else move - past them. */ - if (translate - ? bcmp_translate (d, d2, mcnt, translate) - : bcmp (d, d2, mcnt)) - goto fail; - d += mcnt, d2 += mcnt; - } - } - break; - - - /* begline matches the empty string at the beginning of the string - (unless `not_bol' is set in `bufp'), and, if - `newline_anchor' is set, after newlines. */ - case begline: - DEBUG_PRINT1 ("EXECUTING begline.\n"); - - if (AT_STRINGS_BEG (d)) - { - if (!bufp->not_bol) break; - } - else if (d[-1] == '\n' && bufp->newline_anchor) - { - break; - } - /* In all other cases, we fail. */ - goto fail; - - - /* endline is the dual of begline. */ - case endline: - DEBUG_PRINT1 ("EXECUTING endline.\n"); - - if (AT_STRINGS_END (d)) - { - if (!bufp->not_eol) break; - } - - /* We have to ``prefetch'' the next character. */ - else if ((d == end1 ? *string2 : *d) == '\n' - && bufp->newline_anchor) - { - break; - } - goto fail; - - - /* Match at the very beginning of the data. */ - case begbuf: - DEBUG_PRINT1 ("EXECUTING begbuf.\n"); - if (AT_STRINGS_BEG (d)) - break; - goto fail; - - - /* Match at the very end of the data. */ - case endbuf: - DEBUG_PRINT1 ("EXECUTING endbuf.\n"); - if (AT_STRINGS_END (d)) - break; - goto fail; - - - /* on_failure_keep_string_jump is used to optimize `.*\n'. It - pushes NULL as the value for the string on the stack. Then - `pop_failure_point' will keep the current value for the - string, instead of restoring it. To see why, consider - matching `foo\nbar' against `.*\n'. The .* matches the foo; - then the . fails against the \n. But the next thing we want - to do is match the \n against the \n; if we restored the - string value, we would be back at the foo. - - Because this is used only in specific cases, we don't need to - check all the things that `on_failure_jump' does, to make - sure the right things get saved on the stack. Hence we don't - share its code. The only reason to push anything on the - stack at all is that otherwise we would have to change - `anychar's code to do something besides goto fail in this - case; that seems worse than this. */ - case on_failure_keep_string_jump: - DEBUG_PRINT1 ("EXECUTING on_failure_keep_string_jump"); - - EXTRACT_NUMBER_AND_INCR (mcnt, p); - DEBUG_PRINT3 (" %d (to 0x%x):\n", mcnt, p + mcnt); - - PUSH_FAILURE_POINT (p + mcnt, NULL, -2); - break; - - - /* Uses of on_failure_jump: - - Each alternative starts with an on_failure_jump that points - to the beginning of the next alternative. Each alternative - except the last ends with a jump that in effect jumps past - the rest of the alternatives. (They really jump to the - ending jump of the following alternative, because tensioning - these jumps is a hassle.) - - Repeats start with an on_failure_jump that points past both - the repetition text and either the following jump or - pop_failure_jump back to this on_failure_jump. */ - case on_failure_jump: - on_failure: - DEBUG_PRINT1 ("EXECUTING on_failure_jump"); - - EXTRACT_NUMBER_AND_INCR (mcnt, p); - DEBUG_PRINT3 (" %d (to 0x%x)", mcnt, p + mcnt); - - /* If this on_failure_jump comes right before a group (i.e., - the original * applied to a group), save the information - for that group and all inner ones, so that if we fail back - to this point, the group's information will be correct. - For example, in \(a*\)*\1, we need the preceding group, - and in \(\(a*\)b*\)\2, we need the inner group. */ - - /* We can't use `p' to check ahead because we push - a failure point to `p + mcnt' after we do this. */ - p1 = p; - - /* We need to skip no_op's before we look for the - start_memory in case this on_failure_jump is happening as - the result of a completed succeed_n, as in \(a\)\{1,3\}b\1 - against aba. */ - while (p1 < pend && (re_opcode_t) *p1 == no_op) - p1++; - - if (p1 < pend && (re_opcode_t) *p1 == start_memory) - { - /* We have a new highest active register now. This will - get reset at the start_memory we are about to get to, - but we will have saved all the registers relevant to - this repetition op, as described above. */ - highest_active_reg = *(p1 + 1) + *(p1 + 2); - if (lowest_active_reg == NO_LOWEST_ACTIVE_REG) - lowest_active_reg = *(p1 + 1); - } - - DEBUG_PRINT1 (":\n"); - PUSH_FAILURE_POINT (p + mcnt, d, -2); - break; - - - /* A smart repeat ends with `maybe_pop_jump'. - We change it to either `pop_failure_jump' or `jump'. */ - case maybe_pop_jump: - EXTRACT_NUMBER_AND_INCR (mcnt, p); - DEBUG_PRINT2 ("EXECUTING maybe_pop_jump %d.\n", mcnt); - { - register unsigned char *p2 = p; - - /* Compare the beginning of the repeat with what in the - pattern follows its end. If we can establish that there - is nothing that they would both match, i.e., that we - would have to backtrack because of (as in, e.g., `a*a') - then we can change to pop_failure_jump, because we'll - never have to backtrack. - - This is not true in the case of alternatives: in - `(a|ab)*' we do need to backtrack to the `ab' alternative - (e.g., if the string was `ab'). But instead of trying to - detect that here, the alternative has put on a dummy - failure point which is what we will end up popping. */ - - /* Skip over open/close-group commands. */ - while (p2 + 2 < pend - && ((re_opcode_t) *p2 == stop_memory - || (re_opcode_t) *p2 == start_memory)) - p2 += 3; /* Skip over args, too. */ - - /* If we're at the end of the pattern, we can change. */ - if (p2 == pend) - { - /* Consider what happens when matching ":\(.*\)" - against ":/". I don't really understand this code - yet. */ - p[-3] = (unsigned char) pop_failure_jump; - DEBUG_PRINT1 - (" End of pattern: change to `pop_failure_jump'.\n"); - } - - else if ((re_opcode_t) *p2 == exactn - || (bufp->newline_anchor && (re_opcode_t) *p2 == endline)) - { - register unsigned char c - = *p2 == (unsigned char) endline ? '\n' : p2[2]; - p1 = p + mcnt; - - /* p1[0] ... p1[2] are the `on_failure_jump' corresponding - to the `maybe_finalize_jump' of this case. Examine what - follows. */ - if ((re_opcode_t) p1[3] == exactn && p1[5] != c) - { - p[-3] = (unsigned char) pop_failure_jump; - DEBUG_PRINT3 (" %c != %c => pop_failure_jump.\n", - c, p1[5]); - } - - else if ((re_opcode_t) p1[3] == charset - || (re_opcode_t) p1[3] == charset_not) - { - int not = (re_opcode_t) p1[3] == charset_not; - - if (c < (unsigned char) (p1[4] * BYTEWIDTH) - && p1[5 + c / BYTEWIDTH] & (1 << (c % BYTEWIDTH))) - not = !not; - - /* `not' is equal to 1 if c would match, which means - that we can't change to pop_failure_jump. */ - if (!not) - { - p[-3] = (unsigned char) pop_failure_jump; - DEBUG_PRINT1 (" No match => pop_failure_jump.\n"); - } - } - } - } - p -= 2; /* Point at relative address again. */ - if ((re_opcode_t) p[-1] != pop_failure_jump) - { - p[-1] = (unsigned char) jump; - DEBUG_PRINT1 (" Match => jump.\n"); - goto unconditional_jump; - } - /* Note fall through. */ - - - /* The end of a simple repeat has a pop_failure_jump back to - its matching on_failure_jump, where the latter will push a - failure point. The pop_failure_jump takes off failure - points put on by this pop_failure_jump's matching - on_failure_jump; we got through the pattern to here from the - matching on_failure_jump, so didn't fail. */ - case pop_failure_jump: - { - /* We need to pass separate storage for the lowest and - highest registers, even though we don't care about the - actual values. Otherwise, we will restore only one - register from the stack, since lowest will == highest in - `pop_failure_point'. */ - unsigned dummy_low_reg, dummy_high_reg; - unsigned char *pdummy; - const char *sdummy; - - DEBUG_PRINT1 ("EXECUTING pop_failure_jump.\n"); - POP_FAILURE_POINT (sdummy, pdummy, - dummy_low_reg, dummy_high_reg, - reg_dummy, reg_dummy, reg_info_dummy); - } - /* Note fall through. */ - - - /* Unconditionally jump (without popping any failure points). */ - case jump: - unconditional_jump: - EXTRACT_NUMBER_AND_INCR (mcnt, p); /* Get the amount to jump. */ - DEBUG_PRINT2 ("EXECUTING jump %d ", mcnt); - p += mcnt; /* Do the jump. */ - DEBUG_PRINT2 ("(to 0x%x).\n", p); - break; - - - /* We need this opcode so we can detect where alternatives end - in `group_match_null_string_p' et al. */ - case jump_past_alt: - DEBUG_PRINT1 ("EXECUTING jump_past_alt.\n"); - goto unconditional_jump; - - - /* Normally, the on_failure_jump pushes a failure point, which - then gets popped at pop_failure_jump. We will end up at - pop_failure_jump, also, and with a pattern of, say, `a+', we - are skipping over the on_failure_jump, so we have to push - something meaningless for pop_failure_jump to pop. */ - case dummy_failure_jump: - DEBUG_PRINT1 ("EXECUTING dummy_failure_jump.\n"); - /* It doesn't matter what we push for the string here. What - the code at `fail' tests is the value for the pattern. */ - PUSH_FAILURE_POINT (0, 0, -2); - goto unconditional_jump; - - - /* At the end of an alternative, we need to push a dummy failure - point in case we are followed by a `pop_failure_jump', because - we don't want the failure point for the alternative to be - popped. For example, matching `(a|ab)*' against `aab' - requires that we match the `ab' alternative. */ - case push_dummy_failure: - DEBUG_PRINT1 ("EXECUTING push_dummy_failure.\n"); - /* See comments just above at `dummy_failure_jump' about the - two zeroes. */ - PUSH_FAILURE_POINT (0, 0, -2); - break; - - /* Have to succeed matching what follows at least n times. - After that, handle like `on_failure_jump'. */ - case succeed_n: - EXTRACT_NUMBER (mcnt, p + 2); - DEBUG_PRINT2 ("EXECUTING succeed_n %d.\n", mcnt); - - assert (mcnt >= 0); - /* Originally, this is how many times we HAVE to succeed. */ - if (mcnt > 0) - { - mcnt--; - p += 2; - STORE_NUMBER_AND_INCR (p, mcnt); - DEBUG_PRINT3 (" Setting 0x%x to %d.\n", p, mcnt); - } - else if (mcnt == 0) - { - DEBUG_PRINT2 (" Setting two bytes from 0x%x to no_op.\n", p+2); - p[2] = (unsigned char) no_op; - p[3] = (unsigned char) no_op; - goto on_failure; - } - break; - - case jump_n: - EXTRACT_NUMBER (mcnt, p + 2); - DEBUG_PRINT2 ("EXECUTING jump_n %d.\n", mcnt); - - /* Originally, this is how many times we CAN jump. */ - if (mcnt) - { - mcnt--; - STORE_NUMBER (p + 2, mcnt); - goto unconditional_jump; - } - /* If don't have to jump any more, skip over the rest of command. */ - else - p += 4; - break; - - case set_number_at: - { - DEBUG_PRINT1 ("EXECUTING set_number_at.\n"); - - EXTRACT_NUMBER_AND_INCR (mcnt, p); - p1 = p + mcnt; - EXTRACT_NUMBER_AND_INCR (mcnt, p); - DEBUG_PRINT3 (" Setting 0x%x to %d.\n", p1, mcnt); - STORE_NUMBER (p1, mcnt); - break; - } - - case wordbound: - DEBUG_PRINT1 ("EXECUTING wordbound.\n"); - if (AT_WORD_BOUNDARY (d)) - break; - goto fail; - - case notwordbound: - DEBUG_PRINT1 ("EXECUTING notwordbound.\n"); - if (AT_WORD_BOUNDARY (d)) - goto fail; - break; - - case wordbeg: - DEBUG_PRINT1 ("EXECUTING wordbeg.\n"); - if (WORDCHAR_P (d) && (AT_STRINGS_BEG (d) || !WORDCHAR_P (d - 1))) - break; - goto fail; - - case wordend: - DEBUG_PRINT1 ("EXECUTING wordend.\n"); - if (!AT_STRINGS_BEG (d) && WORDCHAR_P (d - 1) - && (!WORDCHAR_P (d) || AT_STRINGS_END (d))) - break; - goto fail; - -#ifdef emacs -#ifdef emacs19 - case before_dot: - DEBUG_PRINT1 ("EXECUTING before_dot.\n"); - if (PTR_CHAR_POS ((unsigned char *) d) >= point) - goto fail; - break; - - case at_dot: - DEBUG_PRINT1 ("EXECUTING at_dot.\n"); - if (PTR_CHAR_POS ((unsigned char *) d) != point) - goto fail; - break; - - case after_dot: - DEBUG_PRINT1 ("EXECUTING after_dot.\n"); - if (PTR_CHAR_POS ((unsigned char *) d) <= point) - goto fail; - break; -#else /* not emacs19 */ - case at_dot: - DEBUG_PRINT1 ("EXECUTING at_dot.\n"); - if (PTR_CHAR_POS ((unsigned char *) d) + 1 != point) - goto fail; - break; -#endif /* not emacs19 */ - - case syntaxspec: - DEBUG_PRINT2 ("EXECUTING syntaxspec %d.\n", mcnt); - mcnt = *p++; - goto matchsyntax; - - case wordchar: - DEBUG_PRINT1 ("EXECUTING Emacs wordchar.\n"); - mcnt = (int) Sword; - matchsyntax: - PREFETCH (); - if (SYNTAX (*d++) != (enum syntaxcode) mcnt) - goto fail; - SET_REGS_MATCHED (); - break; - - case notsyntaxspec: - DEBUG_PRINT2 ("EXECUTING notsyntaxspec %d.\n", mcnt); - mcnt = *p++; - goto matchnotsyntax; - - case notwordchar: - DEBUG_PRINT1 ("EXECUTING Emacs notwordchar.\n"); - mcnt = (int) Sword; - matchnotsyntax: - PREFETCH (); - if (SYNTAX (*d++) == (enum syntaxcode) mcnt) - goto fail; - SET_REGS_MATCHED (); - break; - -#else /* not emacs */ - case wordchar: - DEBUG_PRINT1 ("EXECUTING non-Emacs wordchar.\n"); - PREFETCH (); - if (!WORDCHAR_P (d)) - goto fail; - SET_REGS_MATCHED (); - d++; - break; - - case notwordchar: - DEBUG_PRINT1 ("EXECUTING non-Emacs notwordchar.\n"); - PREFETCH (); - if (WORDCHAR_P (d)) - goto fail; - SET_REGS_MATCHED (); - d++; - break; -#endif /* not emacs */ - - default: - abort (); - } - continue; /* Successfully executed one pattern command; keep going. */ - - - /* We goto here if a matching operation fails. */ - fail: - if (!FAIL_STACK_EMPTY ()) - { /* A restart point is known. Restore to that state. */ - DEBUG_PRINT1 ("\nFAIL:\n"); - POP_FAILURE_POINT (d, p, - lowest_active_reg, highest_active_reg, - regstart, regend, reg_info); - - /* If this failure point is a dummy, try the next one. */ - if (!p) - goto fail; - - /* If we failed to the end of the pattern, don't examine *p. */ - assert (p <= pend); - if (p < pend) - { - boolean is_a_jump_n = false; - - /* If failed to a backwards jump that's part of a repetition - loop, need to pop this failure point and use the next one. */ - switch ((re_opcode_t) *p) - { - case jump_n: - is_a_jump_n = true; - case maybe_pop_jump: - case pop_failure_jump: - case jump: - p1 = p + 1; - EXTRACT_NUMBER_AND_INCR (mcnt, p1); - p1 += mcnt; - - if ((is_a_jump_n && (re_opcode_t) *p1 == succeed_n) - || (!is_a_jump_n - && (re_opcode_t) *p1 == on_failure_jump)) - goto fail; - break; - default: - /* do nothing */ ; - } - } - - if (d >= string1 && d <= end1) - dend = end_match_1; - } - else - break; /* Matching at this starting point really fails. */ - } /* for (;;) */ - - if (best_regs_set) - goto restore_best_regs; - - FREE_VARIABLES (); - - return -1; /* Failure to match. */ -} /* re_match_2 */ - -/* Subroutine definitions for re_match_2. */ - - -/* We are passed P pointing to a register number after a start_memory. - - Return true if the pattern up to the corresponding stop_memory can - match the empty string, and false otherwise. - - If we find the matching stop_memory, sets P to point to one past its number. - Otherwise, sets P to an undefined byte less than or equal to END. - - We don't handle duplicates properly (yet). */ - -static boolean -group_match_null_string_p (p, end, reg_info) - unsigned char **p, *end; - register_info_type *reg_info; -{ - int mcnt; - /* Point to after the args to the start_memory. */ - unsigned char *p1 = *p + 2; - - while (p1 < end) - { - /* Skip over opcodes that can match nothing, and return true or - false, as appropriate, when we get to one that can't, or to the - matching stop_memory. */ - - switch ((re_opcode_t) *p1) - { - /* Could be either a loop or a series of alternatives. */ - case on_failure_jump: - p1++; - EXTRACT_NUMBER_AND_INCR (mcnt, p1); - - /* If the next operation is not a jump backwards in the - pattern. */ - - if (mcnt >= 0) - { - /* Go through the on_failure_jumps of the alternatives, - seeing if any of the alternatives cannot match nothing. - The last alternative starts with only a jump, - whereas the rest start with on_failure_jump and end - with a jump, e.g., here is the pattern for `a|b|c': - - /on_failure_jump/0/6/exactn/1/a/jump_past_alt/0/6 - /on_failure_jump/0/6/exactn/1/b/jump_past_alt/0/3 - /exactn/1/c - - So, we have to first go through the first (n-1) - alternatives and then deal with the last one separately. */ - - - /* Deal with the first (n-1) alternatives, which start - with an on_failure_jump (see above) that jumps to right - past a jump_past_alt. */ - - while ((re_opcode_t) p1[mcnt-3] == jump_past_alt) - { - /* `mcnt' holds how many bytes long the alternative - is, including the ending `jump_past_alt' and - its number. */ - - if (!alt_match_null_string_p (p1, p1 + mcnt - 3, - reg_info)) - return false; - - /* Move to right after this alternative, including the - jump_past_alt. */ - p1 += mcnt; - - /* Break if it's the beginning of an n-th alternative - that doesn't begin with an on_failure_jump. */ - if ((re_opcode_t) *p1 != on_failure_jump) - break; - - /* Still have to check that it's not an n-th - alternative that starts with an on_failure_jump. */ - p1++; - EXTRACT_NUMBER_AND_INCR (mcnt, p1); - if ((re_opcode_t) p1[mcnt-3] != jump_past_alt) - { - /* Get to the beginning of the n-th alternative. */ - p1 -= 3; - break; - } - } - - /* Deal with the last alternative: go back and get number - of the `jump_past_alt' just before it. `mcnt' contains - the length of the alternative. */ - EXTRACT_NUMBER (mcnt, p1 - 2); - - if (!alt_match_null_string_p (p1, p1 + mcnt, reg_info)) - return false; - - p1 += mcnt; /* Get past the n-th alternative. */ - } /* if mcnt > 0 */ - break; - - - case stop_memory: - assert (p1[1] == **p); - *p = p1 + 2; - return true; - - - default: - if (!common_op_match_null_string_p (&p1, end, reg_info)) - return false; - } - } /* while p1 < end */ - - return false; -} /* group_match_null_string_p */ - - -/* Similar to group_match_null_string_p, but doesn't deal with alternatives: - It expects P to be the first byte of a single alternative and END one - byte past the last. The alternative can contain groups. */ - -static boolean -alt_match_null_string_p (p, end, reg_info) - unsigned char *p, *end; - register_info_type *reg_info; -{ - int mcnt; - unsigned char *p1 = p; - - while (p1 < end) - { - /* Skip over opcodes that can match nothing, and break when we get - to one that can't. */ - - switch ((re_opcode_t) *p1) - { - /* It's a loop. */ - case on_failure_jump: - p1++; - EXTRACT_NUMBER_AND_INCR (mcnt, p1); - p1 += mcnt; - break; - - default: - if (!common_op_match_null_string_p (&p1, end, reg_info)) - return false; - } - } /* while p1 < end */ - - return true; -} /* alt_match_null_string_p */ - - -/* Deals with the ops common to group_match_null_string_p and - alt_match_null_string_p. - - Sets P to one after the op and its arguments, if any. */ - -static boolean -common_op_match_null_string_p (p, end, reg_info) - unsigned char **p, *end; - register_info_type *reg_info; -{ - int mcnt; - boolean ret; - int reg_no; - unsigned char *p1 = *p; - - switch ((re_opcode_t) *p1++) - { - case no_op: - case begline: - case endline: - case begbuf: - case endbuf: - case wordbeg: - case wordend: - case wordbound: - case notwordbound: -#ifdef emacs - case before_dot: - case at_dot: - case after_dot: +# endif #endif - break; - - case start_memory: - reg_no = *p1; - assert (reg_no > 0 && reg_no <= MAX_REGNUM); - ret = group_match_null_string_p (&p1, end, reg_info); - - /* Have to set this here in case we're checking a group which - contains a group and a back reference to it. */ - - if (REG_MATCH_NULL_STRING_P (reg_info[reg_no]) == MATCH_NULL_UNSET_VALUE) - REG_MATCH_NULL_STRING_P (reg_info[reg_no]) = ret; - - if (!ret) - return false; - break; - - /* If this is an optimized succeed_n for zero times, make the jump. */ - case jump: - EXTRACT_NUMBER_AND_INCR (mcnt, p1); - if (mcnt >= 0) - p1 += mcnt; - else - return false; - break; - - case succeed_n: - /* Get to the number of times to succeed. */ - p1 += 2; - EXTRACT_NUMBER_AND_INCR (mcnt, p1); - - if (mcnt == 0) - { - p1 -= 4; - EXTRACT_NUMBER_AND_INCR (mcnt, p1); - p1 += mcnt; - } - else - return false; - break; - - case duplicate: - if (!REG_MATCH_NULL_STRING_P (reg_info[*p1])) - return false; - break; - - case set_number_at: - p1 += 4; - - default: - /* All other opcodes mean we cannot match the empty string. */ - return false; - } - - *p = p1; - return true; -} /* common_op_match_null_string_p */ - - -/* Return zero if TRANSLATE[S1] and TRANSLATE[S2] are identical for LEN - bytes; nonzero otherwise. */ - -static int -bcmp_translate( - unsigned char *s1, - unsigned char *s2, - int len, - char *translate -) -{ - register unsigned char *p1 = s1, *p2 = s2; - while (len) - { - if (translate[*p1++] != translate[*p2++]) return 1; - len--; - } - return 0; -} - -/* Entry points for GNU code. */ - -/* re_compile_pattern is the GNU regular expression compiler: it - compiles PATTERN (of length SIZE) 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. - - We call regex_compile to do the actual compilation. */ - -const char * -re_compile_pattern (pattern, length, bufp) - const char *pattern; - int length; - struct re_pattern_buffer *bufp; -{ - reg_errcode_t ret; - - /* GNU code is written to assume at least RE_NREGS registers will be set - (and at least one extra will be -1). */ - bufp->regs_allocated = REGS_UNALLOCATED; - - /* 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. */ - bufp->no_sub = 0; - - /* Match anchors at newline. */ - bufp->newline_anchor = 1; - - ret = regex_compile (pattern, length, re_syntax_options, bufp); - - return re_error_msg[(int) ret]; -} - -/* Entry points compatible with 4.2 BSD regex library. We don't define - them if this is an Emacs or POSIX compilation. */ - -#if !defined (emacs) && !defined (_POSIX_SOURCE) - -/* BSD has one and only one pattern buffer. */ -static struct re_pattern_buffer re_comp_buf; - -char * -re_comp (s) - const char *s; -{ - reg_errcode_t ret; - - if (!s) - { - if (!re_comp_buf.buffer) - return "No previous regular expression"; - return 0; - } - - if (!re_comp_buf.buffer) - { - re_comp_buf.buffer = (unsigned char *) malloc (200); - if (re_comp_buf.buffer == NULL) - return "Memory exhausted"; - re_comp_buf.allocated = 200; - - re_comp_buf.fastmap = (char *) malloc (1 << BYTEWIDTH); - if (re_comp_buf.fastmap == NULL) - return "Memory exhausted"; - } - - /* 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 = regex_compile (s, strlen (s), re_syntax_options, &re_comp_buf); - - /* Yes, we're discarding `const' here. */ - return (char *) re_error_msg[(int) ret]; -} - - -int -re_exec (s) - const char *s; -{ - const int len = strlen (s); - return - 0 <= re_search (&re_comp_buf, s, len, 0, len, (struct re_registers *) 0); -} -#endif /* not emacs and not _POSIX_SOURCE */ - -/* POSIX.2 functions. Don't define these for Emacs. */ - -#ifndef emacs - -/* 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' and `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 (preg, pattern, cflags) - regex_t *preg; - const char *pattern; - int cflags; -{ - reg_errcode_t ret; - unsigned syntax - = (cflags & REG_EXTENDED) ? - RE_SYNTAX_POSIX_EXTENDED : RE_SYNTAX_POSIX_BASIC; - - /* regex_compile will allocate the space for the compiled pattern. */ - preg->buffer = 0; - preg->allocated = 0; - - /* Don't bother to use a fastmap when searching. This simplifies the - REG_NEWLINE case: if we used a fastmap, we'd have to put all the - characters after newlines into the fastmap. This way, we just try - every character. */ - preg->fastmap = 0; - - if (cflags & REG_ICASE) - { - unsigned i; - - preg->translate = (char *) malloc (CHAR_SET_SIZE); - if (preg->translate == NULL) - return (int) REG_ESPACE; - - /* Map uppercase characters to corresponding lowercase ones. */ - for (i = 0; i < CHAR_SET_SIZE; i++) - preg->translate[i] = ISUPPER (i) ? tolower (i) : i; - } - else - preg->translate = NULL; - - /* 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); - - /* POSIX says a null character in the pattern terminates it, so we - can use strlen here in compiling the pattern. */ - ret = regex_compile (pattern, strlen (pattern), syntax, preg); - - /* 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; - - return (int) ret; -} - - -/* 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 (preg, string, nmatch, pmatch, eflags) - const regex_t *preg; - const char *string; - size_t nmatch; - regmatch_t pmatch[]; - int eflags; -{ - int ret; - struct re_registers regs; - regex_t private_preg; - int len = strlen (string); - boolean want_reg_info = !preg->no_sub && nmatch > 0; - - private_preg = *preg; - - private_preg.not_bol = !!(eflags & REG_NOTBOL); - private_preg.not_eol = !!(eflags & REG_NOTEOL); - - /* The user has told us exactly how many registers to return - information about, via `nmatch'. We have to pass that on to the - matching routines. */ - private_preg.regs_allocated = REGS_FIXED; - - if (want_reg_info) - { - regs.num_regs = nmatch; - regs.start = TALLOC (nmatch, regoff_t); - regs.end = TALLOC (nmatch, regoff_t); - if (regs.start == NULL || regs.end == NULL) - return (int) REG_NOMATCH; - } - - /* Perform the searching operation. */ - ret = re_search (&private_preg, string, len, - /* start: */ 0, /* range: */ len, - want_reg_info ? ®s : (struct re_registers *) 0); - - /* Copy the register information to the POSIX structure. */ - if (want_reg_info) - { - if (ret >= 0) - { - unsigned r; - - for (r = 0; r < nmatch; r++) - { - pmatch[r].rm_so = regs.start[r]; - pmatch[r].rm_eo = regs.end[r]; - } - } - - /* If we needed the temporary register info, free the space now. */ - free (regs.start); - free (regs.end); - } - - /* We want zero return to mean success, unlike `re_search'. */ - return ret >= 0 ? (int) REG_NOERROR : (int) REG_NOMATCH; -} - - -/* 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 *preg, - char *errbuf, size_t errbuf_size) -{ - const char *msg; - size_t msg_size; - - if (errcode < 0 - || errcode >= (sizeof (re_error_msg) / sizeof (re_error_msg[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 = re_error_msg[errcode]; - - /* POSIX doesn't require that we do anything in this case, but why - not be nice. */ - if (! msg) - msg = "Success"; - - msg_size = strlen (msg) + 1; /* Includes the null. */ - - if (errbuf_size != 0) - { - if (msg_size > errbuf_size) - { - strncpy (errbuf, msg, errbuf_size - 1); - errbuf[errbuf_size - 1] = 0; - } - else - strcpy (errbuf, msg); - } - - return msg_size; -} - - -/* Free dynamically allocated space used by PREG. */ - -void -regfree (preg) - regex_t *preg; -{ - if (preg->buffer != NULL) - free (preg->buffer); - preg->buffer = NULL; - - preg->allocated = 0; - preg->used = 0; - - if (preg->fastmap != NULL) - free (preg->fastmap); - preg->fastmap = NULL; - preg->fastmap_accurate = 0; - - if (preg->translate != NULL) - free (preg->translate); - preg->translate = NULL; -} - -#endif /* not emacs */ - -/* -Local variables: -make-backup-files: t -version-control: t -trim-versions-without-asking: nil -End: -*/ diff --git a/compat/regex/regex.h b/compat/regex/regex.h index 6eb64f1402..61c9683872 100644 --- a/compat/regex/regex.h +++ b/compat/regex/regex.h @@ -1,70 +1,90 @@ -/* Definitions for data structures and routines for the regular - expression library, version 0.12. +#include <stdio.h> +#include <stddef.h> - Copyright (C) 1985, 1989, 1990, 1991, 1992, 1993 Free Software Foundation, Inc. +/* Definitions for data structures and routines for the regular + expression library. + Copyright (C) 1985,1989-93,1995-98,2000,2001,2002,2003,2005,2006,2008 + Free Software Foundation, Inc. + This file is part of the GNU C Library. - 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 2, or (at your option) - any later version. + 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. - This program is distributed in the hope that it will be useful, + 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 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., 675 Mass Ave, Cambridge, MA 02139, USA. */ + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. -#ifndef __REGEXP_LIBRARY_H__ -#define __REGEXP_LIBRARY_H__ + You should have received a copy of the GNU Lesser General Public + License along with the GNU C Library; if not, write to the Free + Software Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA + 02110-1301 USA. */ -/* POSIX says that <sys/types.h> must be included (by the caller) before - <regex.h>. */ +#ifndef _REGEX_H +#define _REGEX_H 1 -#ifdef VMS -/* VMS doesn't have `size_t' in <sys/types.h>, even though POSIX says it - should be there. */ +#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 reg_syntax_t; +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 (1) +# 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) +# 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) +# 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. + ^ 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) +# 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. @@ -72,63 +92,94 @@ typedef unsigned reg_syntax_t; 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) +# 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) +# 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) +# 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) +# 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) +# 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) +# 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) +# 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) +# 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) +# 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) +# 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) +# 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) +# 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) +# 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) +# 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, 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_NO_GNU_OPS << 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 @@ -136,6 +187,7 @@ typedef unsigned reg_syntax_t; 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!) */ @@ -143,13 +195,22 @@ extern reg_syntax_t re_syntax_options; #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_UNMATCHED_RIGHT_PAREN_ORD) - -#define RE_SYNTAX_POSIX_AWK \ - (RE_SYNTAX_POSIX_EXTENDED | RE_BACKSLASH_ESCAPE_IN_LISTS) + (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_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 \ @@ -163,7 +224,8 @@ extern reg_syntax_t re_syntax_options; | RE_NO_BK_VBAR) #define RE_SYNTAX_POSIX_EGREP \ - (RE_SYNTAX_EGREP | RE_INTERVALS | RE_NO_BK_BRACES) + (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 @@ -176,7 +238,7 @@ extern reg_syntax_t re_syntax_options; | RE_INTERVALS | RE_NO_EMPTY_RANGES) #define RE_SYNTAX_POSIX_BASIC \ - (_RE_SYNTAX_POSIX_COMMON | RE_BK_PLUS_QM) + (_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 @@ -185,13 +247,13 @@ extern reg_syntax_t re_syntax_options; (_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_UNMATCHED_RIGHT_PAREN_ORD) + (_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_INVALID_OPS - replaces RE_CONTEXT_INDEP_OPS and RE_NO_BK_REFS is added. */ +/* 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 \ @@ -202,10 +264,12 @@ extern reg_syntax_t re_syntax_options; /* 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 +# ifdef RE_DUP_MAX +# undef RE_DUP_MAX +# endif +/* If sizeof(int) == 2, then ((1 << 15) - 1) overflows. */ +# define RE_DUP_MAX (0x7fff) #endif -#define RE_DUP_MAX ((1 << 15) - 1) /* POSIX `cflags' bits (i.e., information for `regcomp'). */ @@ -240,18 +304,26 @@ extern reg_syntax_t re_syntax_options; /* 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, /* Not implemented. */ + REG_ECOLLATE, /* Inalid collating element. */ REG_ECTYPE, /* Invalid character class name. */ REG_EESCAPE, /* Trailing backslash. */ REG_ESUBREG, /* Invalid back reference. */ @@ -275,85 +347,92 @@ typedef enum compiled, the `re_nsub' field is 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 { -/* [[[begin 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 *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 allocated; + /* Number of bytes to which `buffer' points. */ + unsigned long int __REPB_PREFIX(allocated); - /* Number of bytes actually used in `buffer'. */ - unsigned long used; + /* Number of bytes actually used in `buffer'. */ + unsigned long int __REPB_PREFIX(used); - /* Syntax setting with which the pattern was compiled. */ - reg_syntax_t syntax; + /* 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 *fastmap; + /* 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. */ - char *translate; + /* 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. */ + /* 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 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. */ -#define REGS_UNALLOCATED 0 -#define REGS_REALLOCATE 1 -#define REGS_FIXED 2 - unsigned 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 fastmap_accurate : 1; - - /* If set, `re_match_2' does not return information about - subexpressions. */ - unsigned no_sub : 1; - - /* If set, a beginning-of-line anchor doesn't match at the - beginning of the string. */ - unsigned not_bol : 1; - - /* Similarly for an end-of-line anchor. */ - unsigned not_eol : 1; - - /* If true, an anchor at a newline matches. */ - unsigned newline_anchor : 1; - -/* [[[end pattern_buffer]]] */ -}; + /* 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; -typedef struct re_pattern_buffer regex_t; + /* 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; +}; -/* search.c (search_buffer) in Emacs needs this one opcode value. It is - defined both in `regex.c' and here. */ -#define RE_EXACTN_VALUE 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 @@ -367,8 +446,9 @@ struct re_registers /* 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 +# ifndef RE_NREGS +# define RE_NREGS 30 +# endif #endif @@ -383,38 +463,22 @@ typedef struct /* Declarations for routines. */ -/* To avoid duplicating every routine declaration -- once with a - prototype (if we are ANSI), and once without (if we aren't) -- we - use the following macro to declare argument types. This - unfortunately clutters up the declarations a bit, but I think it's - worth it. */ - -#if __STDC__ - -#define _RE_ARGS(args) args - -#else /* not __STDC__ */ - -#define _RE_ARGS(args) () - -#endif /* not __STDC__ */ - +#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 _RE_ARGS ((reg_syntax_t syntax)); +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. */ -extern const char *re_compile_pattern - _RE_ARGS ((const char *pattern, int length, - struct re_pattern_buffer *buffer)); +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 _RE_ARGS ((struct re_pattern_buffer *buffer)); +extern int re_compile_fastmap (struct re_pattern_buffer *__buffer); /* Search in the string STRING (with length LENGTH) for the pattern @@ -422,31 +486,30 @@ extern int re_compile_fastmap _RE_ARGS ((struct re_pattern_buffer *buffer)); 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 - _RE_ARGS ((struct re_pattern_buffer *buffer, const char *string, - int length, int start, int range, struct re_registers *regs)); +extern int re_search (struct re_pattern_buffer *__buffer, const char *__cstring, + 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 - _RE_ARGS ((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)); +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 - _RE_ARGS ((struct re_pattern_buffer *buffer, const char *string, - int length, int start, struct re_registers *regs)); +extern int re_match (struct re_pattern_buffer *__buffer, const char *__cstring, + 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 - _RE_ARGS ((struct re_pattern_buffer *buffer, const char *string1, - int length1, const char *string2, int length2, - int start, struct re_registers *regs, int stop)); +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 @@ -461,30 +524,59 @@ extern int re_match_2 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 - _RE_ARGS ((struct re_pattern_buffer *buffer, struct re_registers *regs, - unsigned num_regs, regoff_t *starts, regoff_t *ends)); - +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_BSD) +# ifndef _CRAY /* 4.2 bsd compatibility. */ -extern char *re_comp _RE_ARGS ((const char *)); -extern int re_exec _RE_ARGS ((const char *)); +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 _RE_ARGS ((regex_t *preg, const char *pattern, int cflags)); -extern int regexec - _RE_ARGS ((const regex_t *preg, const char *string, size_t nmatch, - regmatch_t pmatch[], int eflags)); -extern size_t regerror - _RE_ARGS ((int errcode, const regex_t *preg, char *errbuf, - size_t errbuf_size)); -extern void regfree _RE_ARGS ((regex_t *preg)); - -#endif /* not __REGEXP_LIBRARY_H__ */ - -/* -Local variables: -make-backup-files: t -version-control: t -trim-versions-without-asking: nil -End: -*/ +extern int regcomp (regex_t *__restrict __preg, + const char *__restrict __pattern, + int __cflags); + +extern int regexec (const regex_t *__restrict __preg, + const char *__restrict __cstring, 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/compat/regex/regex_internal.c b/compat/regex/regex_internal.c new file mode 100644 index 0000000000..193854cf5b --- /dev/null +++ b/compat/regex/regex_internal.c @@ -0,0 +1,1744 @@ +/* Extended regular expression matching and search library. + Copyright (C) 2002-2006, 2010 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, write to the Free + Software Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA + 02110-1301 USA. */ + +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 +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 +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 +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) -2, 0)) + { + /* The buffer doesn't have enough space, finish to build. */ + pstr->cur_state = prev_st; + break; + } + else if (BE (mbclen == (size_t) -1 || mbclen == 0, 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; + } + + /* 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 +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) + { + /* It is an invalid character 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) + { + /* 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->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 (mbclen == (size_t) -2 || mbclen == (size_t) -1 || 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 +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 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; + + if (BE (pstr->trans != NULL, 0)) + { + int i = mlen < 6 ? mlen : 6; + while (--i >= 0) + buf[i] = pstr->trans[p[i]]; + } + /* 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 *) p, 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 __attribute ((pure)) +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) + { +#ifdef 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 +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 +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 +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 +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 +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 +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 +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 +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 +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)) + 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 +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 +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 +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 independ of contexts. + Return the new state if succeeded, otherwise return NULL. */ + +static re_dfastate_t * +internal_function +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 +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/compat/regex/regex_internal.h b/compat/regex/regex_internal.h new file mode 100644 index 0000000000..4184d7f5a6 --- /dev/null +++ b/compat/regex/regex_internal.h @@ -0,0 +1,810 @@ +/* Extended regular expression matching and search library. + Copyright (C) 2002-2005, 2007, 2008, 2010 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, write to the Free + Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA + 02111-1307 USA. */ + +#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(ZOS_USS) +#if defined HAVE_STDINT_H || defined _LIBC +# include <stdint.h> +#endif /* HAVE_STDINT_H || _LIBC */ +#endif /* !ZOS_USS */ +#if defined _LIBC +# include <bits/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/elem-hash.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) \ + INTUSE(__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 + +#ifndef NO_MBSUPPORT +#include "mbsupport.h" /* gawk */ +#endif +#ifndef MB_CUR_MAX +#define MB_CUR_MAX 1 +#endif + +#if (defined MBS_SUPPORT) || _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 +# define attribute_hidden +#endif /* not _LIBC */ + +#ifdef __GNUC__ +# define __attribute(arg) __attribute__ (arg) +#else +# define __attribute(arg) +#endif + +extern const char __re_error_msgid[] attribute_hidden; +extern const size_t __re_error_msgid_idx[] attribute_hidden; + +/* 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 +# if (_MSC_VER) +# include <malloc.h> +# define __libc_use_alloca(n) 0 +# else +# 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) +# endif +# else +/* alloca is implemented with malloc, so just use malloc. */ +# define __libc_use_alloca(n) 0 +# endif +#endif + +#define re_malloc(t,n) ((t *) malloc ((n) * sizeof (t))) +/* SunOS 4.1.x realloc doesn't accept null pointers: pre-Standard C. Sigh. */ +#define re_realloc(p,t,n) ((p != NULL) ? (t *) realloc (p,(n)*sizeof(t)) : (t *) calloc(n,sizeof(t))) +#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 inline void +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 inline void +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 inline void +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 inline int +internal_function __attribute ((pure)) +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 inline wint_t +internal_function __attribute ((pure)) +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 +static int +internal_function __attribute ((pure)) +re_string_elem_size_at (const re_string_t *pstr, int idx) +{ +# ifdef _LIBC + const unsigned char *p, *extra; + const int32_t *table, *indirect; + int32_t tmp; +# include <locale/weight.h> + 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; + tmp = findidx (&p); + return p - pstr->mbs - idx; + } + else +# endif /* _LIBC */ + return 1; +} +# endif +#endif /* RE_ENABLE_I18N */ + +#endif /* _REGEX_INTERNAL_H */ diff --git a/compat/regex/regexec.c b/compat/regex/regexec.c new file mode 100644 index 0000000000..0194965c5d --- /dev/null +++ b/compat/regex/regexec.c @@ -0,0 +1,4369 @@ +/* Extended regular expression matching and search library. + Copyright (C) 2002-2005, 2007, 2009, 2010 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, write to the Free + Software Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA + 02110-1301 USA. */ + +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); +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); +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); +static unsigned re_copy_regs (struct re_registers *regs, regmatch_t *pmatch, + int nregs, int regs_allocated); +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) + internal_function; + +/* 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 stroed 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 +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 = length1 + length2; + int free_str = 0; + + if (BE (length1 < 0 || length2 < 0 || stop < 0, 0)) + return -2; + + /* Concatenate the strings. */ + if (length2 > 0) + if (length1 > 0) + { + char *s = re_malloc (char, len); + + if (BE (s == NULL, 0)) + return -2; + memcpy (s, string1, length1); + memcpy (s + length1, string2, length2); + str = s; + free_str = 1; + } + else + str = string2; + else + str = string1; + + rval = re_search_stub (bufp, str, len, start, range, stop, regs, ret_len); + if (free_str) + re_free ((char *) str); + 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 +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 +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 (s) + 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 + mingings 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 +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->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, 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 the each registers, since we slided + 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 +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) + || (BE (next_char_idx >= mctx->input.valid_len, 0) + && mctx->input.valid_len < mctx->input.len)) + { + err = extend_buffers (mctx); + 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 +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 +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 +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 + || (next_state_log_idx >= mctx->input.valid_len + && mctx->input.valid_len < mctx->input.len)) + { + reg_errcode_t err; + err = extend_buffers (mctx); + 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 +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 +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 +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. */ +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 + correspoding 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 +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); + 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 +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 +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 +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 +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) + 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); + 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_dfastate_t **) + malloc (ndests * 3 * sizeof (re_dfastate_t *)); + if (BE (dest_states == NULL, 0)) + { +out_free: + if (dest_states_malloced) + 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) + 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) + 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) + 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. */ + +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; + /* This #include defines a local function! */ +# include <locale/weight.h> + + /* 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 (&cp); + 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? */ +#if __GNUC__ >= 2 + wchar_t cmp_buf[] = {L'\0', L'\0', wc, L'\0', L'\0', L'\0'}; +#else + wchar_t cmp_buf[] = {L'\0', L'\0', L'\0', L'\0', L'\0', L'\0'}; + cmp_buf[2] = wc; +#endif + for (i = 0; i < cset->nranges; ++i) + { + cmp_buf[0] = cset->range_starts[i]; + cmp_buf[4] = cset->range_ends[i]; + if (wcscoll (cmp_buf, cmp_buf + 2) <= 0 + && wcscoll (cmp_buf + 2, cmp_buf + 4) <= 0) + { + 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) * (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 +extend_buffers (re_match_context_t *mctx) +{ + 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. */ + ret = re_string_realloc_buffers (pstr, 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 +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); + } + 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 +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 +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/config.mak.in b/config.mak.in index 4ffd77420f..a0c34eec15 100644 --- a/config.mak.in +++ b/config.mak.in @@ -59,6 +59,7 @@ NO_INET_NTOP=@NO_INET_NTOP@ NO_INET_PTON=@NO_INET_PTON@ NO_ICONV=@NO_ICONV@ OLD_ICONV=@OLD_ICONV@ +NO_REGEX=@NO_REGEX@ NO_DEFLATE_BOUND=@NO_DEFLATE_BOUND@ INLINE=@INLINE@ SOCKLEN_T=@SOCKLEN_T@ diff --git a/configure.ac b/configure.ac index 708e7b86ce..56731c35c9 100644 --- a/configure.ac +++ b/configure.ac @@ -706,6 +706,27 @@ else fi AC_SUBST(NO_C99_FORMAT) # +# Define NO_REGEX if you have no or inferior regex support in your C library. +AC_CACHE_CHECK([whether the platform regex can handle null bytes], + [ac_cv_c_excellent_regex], [ +AC_EGREP_CPP(yippeeyeswehaveit, + AC_LANG_PROGRAM([AC_INCLUDES_DEFAULT +#include <regex.h> +], +[#ifdef REG_STARTEND +yippeeyeswehaveit +#endif +]), + [ac_cv_c_excellent_regex=yes], + [ac_cv_c_excellent_regex=no]) +]) +if test $ac_cv_c_excellent_regex = yes; then + NO_REGEX= +else + NO_REGEX=YesPlease +fi +AC_SUBST(NO_REGEX) +# # Define FREAD_READS_DIRECTORIES if your are on a system which succeeds # when attempting to read from an fopen'ed directory. AC_CACHE_CHECK([whether system succeeds to read fopen'ed directory], diff --git a/t/t7008-grep-binary.sh b/t/t7008-grep-binary.sh index eb8ca88cce..c0f9f3f705 100755 --- a/t/t7008-grep-binary.sh +++ b/t/t7008-grep-binary.sh @@ -61,7 +61,7 @@ test_expect_success 'git grep -Fi iLE a' ' # This test actually passes on platforms where regexec() supports the # flag REG_STARTEND. -test_expect_failure 'git grep ile a' ' +test_expect_success 'git grep ile a' ' git grep ile a ' |