fnmatch: merge from glibc

Also, merge in Gnulib’s more-recent methods of making it easier
to share between Gnulib and glibc.
* lib/fnmatch.c: Reorder includes to match glibc better.
Include libc-config.h instead of config.h.
Include alloca.h only if _LIBC || HAVE_ALLOCA.
Do not include "../locale/elem-hash.h" if _LIBC.
Define macros for btowc, etc. if _LIBC.  All uses simplified.
Define FALLTHROUGH if _LIBC, instead of including attribute.h.
Include intprops.h, since glibc has it now.
(SIZE_MAX): Remove; use (size_t) -1 instead.
Omit the "Comment out all this code" ifdef, since Gnulib
has never really needed it.
(STREQ): Remove; no longer used.
(__libc_use_alloca, alloca, alloca_account): Define as
needed if !_LIBC.
(ISWCTYPE): Remove; all uses replaced by iswctype.
(HANDLE_MULTIBYTE): Remove.  All uses removed by assuming true.
(internal_function): Remove.  All uses removed.
(STRUCT): New macro.
(WIDE_CHAR_VERSION): Define to 0 instead of leaving undefined.
(WMEMCMP): New macro.
(FINDIDX): Define if _LIBC, and include <locale/weight.h>
and <locale/weightwc.h>.
(fnmatch): Prefer __glibc_likely and __glibc_unlikely to
__builtin_expect.  Check for integer overflow more
systematically.  Account for alloca storage better when
recursive.  Use strnlen instead of strlen for efficiency.
* lib/fnmatch_loop.c: Include stdint.h if _LIBC, for int32_t etc.
(struct STRUCT): New type.
(FCT, EXT): New ENDS and ALLOCA_USED args.
All callers changed.
(FCT): Prefer __glibc_unlikely to __builtin_expect.
Simplify by assuming WIDE_CHAR_SUPPORT.
Copy _LIBC code from glibc without worrying Gnulib compatibility.
Cast cold to UCHAR to avoid signedness warning.
(END): Check for invalid pattern.
(EXT): Improve alloca/malloc checking (taken from glibc),
and improve it some more by using intprops.h and checking
for integer overflow and using bool for booleans.
* lib/libc-config.h (compat_symbol): New macro.
(versioned_symbol): Make it ‘extern int dummy’ so that it’s
acceptable to non-GCC when a trailing semicolon is added.
* modules/fnmatch (Depends-on): Add alloca-opt, intprops,
libc-config, strnlen.  Remove alloca.

1 files changed, 216 insertions, 176 deletions

diff --git a/lib/fnmatch.c b/lib/fnmatch.c
index db4da5e6a6..4d017cfebe 100644
--- a/lib/fnmatch.c
+++ b/lib/fnmatch.c
@@ -1,20 +1,22 @@
-/* Copyright (C) 1991-1993, 1996-2007, 2009-2020 Free Software Foundation, Inc.
+/* Copyright (C) 1991-2020 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.
+   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 General Public License
-   along with this program; if not, see <https://www.gnu.org/licenses/>.  */
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, see
+   <https://www.gnu.org/licenses/>.  */
 
 #ifndef _LIBC
-# include <config.h>
+# include <libc-config.h>
 #endif
 
 /* Enable GNU extensions in fnmatch.h.  */
@@ -24,83 +26,87 @@
 
 #include <fnmatch.h>
 
-#include <alloca.h>
 #include <assert.h>
-#include <ctype.h>
 #include <errno.h>
-#include <stddef.h>
-#include <stdbool.h>
-#include <stdlib.h>
+#include <ctype.h>
 #include <string.h>
+#include <stdlib.h>
+#if defined _LIBC || HAVE_ALLOCA
+# include <alloca.h>
+#endif
 #include <wchar.h>
 #include <wctype.h>
+#include <stddef.h>
+#include <stdbool.h>
 
 /* We need some of the locale data (the collation sequence information)
    but there is no interface to get this information in general.  Therefore
    we support a correct implementation only in glibc.  */
 #ifdef _LIBC
 # include "../locale/localeinfo.h"
-# include "../locale/elem-hash.h"
 # include "../locale/coll-lookup.h"
 # include <shlib-compat.h>
 
 # define CONCAT(a,b) __CONCAT(a,b)
+# define btowc __btowc
+# define iswctype __iswctype
 # define mbsrtowcs __mbsrtowcs
+# define mempcpy __mempcpy
+# define strnlen __strnlen
+# define towlower __towlower
+# define wcscat __wcscat
+# define wcslen __wcslen
+# define wctype __wctype
+# define wmemchr __wmemchr
+# define wmempcpy __wmempcpy
 # define fnmatch __fnmatch
 extern int fnmatch (const char *pattern, const char *string, int flags);
 #endif
 
-#ifndef SIZE_MAX
-# define SIZE_MAX ((size_t) -1)
+#ifdef _LIBC
+# if __GNUC__ < 7
+#  define FALLTHROUGH ((void) 0)
+# else
+#  define FALLTHROUGH __attribute__ ((__fallthrough__))
+# endif
+#else
+# include "attribute.h"
 #endif
 
-#include "attribute.h"
-#include "flexmember.h"
+#include <intprops.h>
+#include <flexmember.h>
 
 /* We often have to test for FNM_FILE_NAME and FNM_PERIOD being both set.  */
 #define NO_LEADING_PERIOD(flags) \
   ((flags & (FNM_FILE_NAME | FNM_PERIOD)) == (FNM_FILE_NAME | FNM_PERIOD))
 
-/* Comment out all this code if we are using the GNU C Library, and are not
-   actually compiling the library itself, and have not detected a bug
-   in the library.  This code is part of the GNU C
-   Library, but also included in many other GNU distributions.  Compiling
-   and linking in this code is a waste when using the GNU C library
-   (especially if it is a shared library).  Rather than having every GNU
-   program understand 'configure --with-gnu-libc' and omit the object files,
-   it is simpler to just do this in the source for each such file.  */
-
-#if defined _LIBC || !defined __GNU_LIBRARY__ || !HAVE_FNMATCH_GNU
-
-
-# define STREQ(s1, s2) (strcmp (s1, s2) == 0)
+#ifndef _LIBC
+# if HAVE_ALLOCA
+/* The OS usually guarantees only one guard page at the bottom of the stack,
+   and a page size can be as small as 4096 bytes.  So we cannot safely
+   allocate anything larger than 4096 bytes.  Also care for the possibility
+   of a few compiler-allocated temporary stack slots.  */
+#  define __libc_use_alloca(n) ((n) < 4032)
+# else
+/* Just use malloc.  */
+#  define __libc_use_alloca(n) false
+#  undef alloca
+#  define alloca(n) malloc (n)
+# endif
+# define alloca_account(size, avar) ((avar) += (size), alloca (size))
+#endif
 
 /* Provide support for user-defined character classes, based on the functions
    from ISO C 90 amendment 1.  */
-# ifdef CHARCLASS_NAME_MAX
-#  define CHAR_CLASS_MAX_LENGTH CHARCLASS_NAME_MAX
-# else
+#ifdef CHARCLASS_NAME_MAX
+# define CHAR_CLASS_MAX_LENGTH CHARCLASS_NAME_MAX
+#else
 /* This shouldn't happen but some implementation might still have this
    problem.  Use a reasonable default value.  */
-#  define CHAR_CLASS_MAX_LENGTH 256
-# endif
-
-# ifdef _LIBC
-#  define IS_CHAR_CLASS(string) __wctype (string)
-# else
-#  define IS_CHAR_CLASS(string) wctype (string)
-# endif
-
-# ifdef _LIBC
-#  define ISWCTYPE(WC, WT)     __iswctype (WC, WT)
-# else
-#  define ISWCTYPE(WC, WT)     iswctype (WC, WT)
-# endif
+# define CHAR_CLASS_MAX_LENGTH 256
+#endif
 
-# if (HAVE_MBSTATE_T && HAVE_MBSRTOWCS) || _LIBC
-/* In this case we are implementing the multibyte character handling.  */
-#  define HANDLE_MULTIBYTE     1
-# endif
+#define IS_CHAR_CLASS(string) wctype (string)
 
 /* Avoid depending on library functions or files
    whose names are inconsistent.  */
@@ -108,60 +114,53 @@ extern int fnmatch (const char *pattern, const char *string, int flags);
 /* Global variable.  */
 static int posixly_correct;
 
-# ifndef internal_function
-/* Inside GNU libc we mark some function in a special way.  In other
-   environments simply ignore the marking.  */
-#  define internal_function
-# endif
-
 /* Note that this evaluates C many times.  */
-# define FOLD(c) ((flags & FNM_CASEFOLD) ? tolower (c) : (c))
-# define CHAR   char
-# define UCHAR  unsigned char
-# define INT    int
-# define FCT    internal_fnmatch
-# define EXT    ext_match
-# define END    end_pattern
-# define L_(CS) CS
-# ifdef _LIBC
-#  define BTOWC(C)      __btowc (C)
-# else
-#  define BTOWC(C)      btowc (C)
-# endif
-# define STRLEN(S) strlen (S)
-# define STRCAT(D, S) strcat (D, S)
-# ifdef _LIBC
-#  define MEMPCPY(D, S, N) __mempcpy (D, S, N)
-# else
-#  define MEMPCPY(D, S, N) mempcpy (D, S, N)
-# endif
-# define MEMCHR(S, C, N) memchr (S, C, N)
-# include "fnmatch_loop.c"
-
-
-# if HANDLE_MULTIBYTE
-#  define FOLD(c) ((flags & FNM_CASEFOLD) ? towlower (c) : (c))
-#  define CHAR  wchar_t
-#  define UCHAR wint_t
-#  define INT   wint_t
-#  define FCT   internal_fnwmatch
-#  define EXT   ext_wmatch
-#  define END   end_wpattern
-#  define L_(CS)        L##CS
-#  define BTOWC(C)      (C)
-#  ifdef _LIBC
-#   define STRLEN(S) __wcslen (S)
-#   define STRCAT(D, S) __wcscat (D, S)
-#   define MEMPCPY(D, S, N) __wmempcpy (D, S, N)
-#  else
-#   define STRLEN(S) wcslen (S)
-#   define STRCAT(D, S) wcscat (D, S)
-#   define MEMPCPY(D, S, N) wmempcpy (D, S, N)
-#  endif
-#  define MEMCHR(S, C, N) wmemchr (S, C, N)
-#  define WIDE_CHAR_VERSION 1
-
-#  undef IS_CHAR_CLASS
+#define FOLD(c) ((flags & FNM_CASEFOLD) ? tolower (c) : (c))
+#define CHAR    char
+#define UCHAR   unsigned char
+#define INT     int
+#define FCT     internal_fnmatch
+#define EXT     ext_match
+#define END     end_pattern
+#define STRUCT  fnmatch_struct
+#define L_(CS)  CS
+#define BTOWC(C) btowc (C)
+#define STRLEN(S) strlen (S)
+#define STRCAT(D, S) strcat (D, S)
+#define MEMPCPY(D, S, N) mempcpy (D, S, N)
+#define MEMCHR(S, C, N) memchr (S, C, N)
+#define WIDE_CHAR_VERSION 0
+#ifdef _LIBC
+# include <locale/weight.h>
+# define FINDIDX findidx
+#endif
+#include "fnmatch_loop.c"
+
+
+#define FOLD(c) ((flags & FNM_CASEFOLD) ? towlower (c) : (c))
+#define CHAR    wchar_t
+#define UCHAR   wint_t
+#define INT     wint_t
+#define FCT     internal_fnwmatch
+#define EXT     ext_wmatch
+#define END     end_wpattern
+#define L_(CS)  L##CS
+#define BTOWC(C) (C)
+#define STRLEN(S) wcslen (S)
+#define STRCAT(D, S) wcscat (D, S)
+#define MEMPCPY(D, S, N) wmempcpy (D, S, N)
+#define MEMCHR(S, C, N) wmemchr (S, C, N)
+#define WIDE_CHAR_VERSION 1
+#ifdef _LIBC
+/* Change the name the header defines so it doesn't conflict with
+   the <locale/weight.h> version included above.  */
+# define findidx findidxwc
+# include <locale/weightwc.h>
+# undef findidx
+# define FINDIDX findidxwc
+#endif
+
+#undef IS_CHAR_CLASS
 /* We have to convert the wide character string in a multibyte string.  But
    we know that the character class names consist of alphanumeric characters
    from the portable character set, and since the wide character encoding
@@ -177,11 +176,11 @@ is_char_class (const wchar_t *wcs)
   do
     {
       /* Test for a printable character from the portable character set.  */
-#  ifdef _LIBC
+#ifdef _LIBC
       if (*wcs < 0x20 || *wcs > 0x7e
           || *wcs == 0x24 || *wcs == 0x40 || *wcs == 0x60)
         return (wctype_t) 0;
-#  else
+#else
       switch (*wcs)
         {
         case L' ': case L'!': case L'"': case L'#': case L'%':
@@ -208,7 +207,7 @@ is_char_class (const wchar_t *wcs)
         default:
           return (wctype_t) 0;
         }
-#  endif
+#endif
 
       /* Avoid overrunning the buffer.  */
       if (cp == s + CHAR_CLASS_MAX_LENGTH)
@@ -220,96 +219,137 @@ is_char_class (const wchar_t *wcs)
 
   *cp = '\0';
 
-#  ifdef _LIBC
-  return __wctype (s);
-#  else
   return wctype (s);
-#  endif
 }
-#  define IS_CHAR_CLASS(string) is_char_class (string)
+#define IS_CHAR_CLASS(string) is_char_class (string)
 
-#  include "fnmatch_loop.c"
-# endif
+#include "fnmatch_loop.c"
 
 
 int
 fnmatch (const char *pattern, const char *string, int flags)
 {
-# if HANDLE_MULTIBYTE
-#  define ALLOCA_LIMIT 2000
-  if (__builtin_expect (MB_CUR_MAX, 1) != 1)
+  if (__glibc_unlikely (MB_CUR_MAX != 1))
     {
       mbstate_t ps;
-      size_t patsize;
-      size_t strsize;
-      size_t totsize;
+      size_t n;
+      const char *p;
+      wchar_t *wpattern_malloc = NULL;
       wchar_t *wpattern;
+      wchar_t *wstring_malloc = NULL;
       wchar_t *wstring;
-      int res;
+      size_t alloca_used = 0;
 
-      /* Calculate the size needed to convert the strings to
-         wide characters.  */
+      /* Convert the strings into wide characters.  */
       memset (&ps, '\0', sizeof (ps));
-      patsize = mbsrtowcs (NULL, &pattern, 0, &ps) + 1;
-      if (__builtin_expect (patsize != 0, 1))
+      p = pattern;
+      n = strnlen (pattern, 1024);
+      if (__glibc_likely (n < 1024))
         {
+          wpattern = (wchar_t *) alloca_account ((n + 1) * sizeof (wchar_t),
+                                                 alloca_used);
+          n = mbsrtowcs (wpattern, &p, n + 1, &ps);
+          if (__glibc_unlikely (n == (size_t) -1))
+            /* Something wrong.
+               XXX Do we have to set 'errno' to something which mbsrtows hasn't
+               already done?  */
+            return -1;
+          if (p)
+            {
+              memset (&ps, '\0', sizeof (ps));
+              goto prepare_wpattern;
+            }
+        }
+      else
+        {
+        prepare_wpattern:
+          n = mbsrtowcs (NULL, &pattern, 0, &ps);
+          if (__glibc_unlikely (n == (size_t) -1))
+            /* Something wrong.
+               XXX Do we have to set 'errno' to something which mbsrtows hasn't
+               already done?  */
+            return -1;
+          if (__glibc_unlikely (n >= (size_t) -1 / sizeof (wchar_t)))
+            {
+              __set_errno (ENOMEM);
+              return -2;
+            }
+          wpattern_malloc = wpattern
+            = (wchar_t *) malloc ((n + 1) * sizeof (wchar_t));
           assert (mbsinit (&ps));
-          strsize = mbsrtowcs (NULL, &string, 0, &ps) + 1;
-          if (__builtin_expect (strsize != 0, 1))
+          if (wpattern == NULL)
+            return -2;
+          (void) mbsrtowcs (wpattern, &pattern, n + 1, &ps);
+        }
+
+      assert (mbsinit (&ps));
+      n = strnlen (string, 1024);
+      p = string;
+      if (__glibc_likely (n < 1024))
+        {
+          wstring = (wchar_t *) alloca_account ((n + 1) * sizeof (wchar_t),
+                                                alloca_used);
+          n = mbsrtowcs (wstring, &p, n + 1, &ps);
+          if (__glibc_unlikely (n == (size_t) -1))
             {
-              assert (mbsinit (&ps));
-              totsize = patsize + strsize;
-              if (__builtin_expect (! (patsize <= totsize
-                                       && totsize <= SIZE_MAX / sizeof (wchar_t)),
-                                    0))
-                {
-                  errno = ENOMEM;
-                  return -1;
-                }
-
-              /* Allocate room for the wide characters.  */
-              if (__builtin_expect (totsize < ALLOCA_LIMIT, 1))
-                wpattern = (wchar_t *) alloca (totsize * sizeof (wchar_t));
-              else
-                {
-                  wpattern = malloc (totsize * sizeof (wchar_t));
-                  if (__builtin_expect (! wpattern, 0))
-                    {
-                      errno = ENOMEM;
-                      return -1;
-                    }
-                }
-              wstring = wpattern + patsize;
-
-              /* Convert the strings into wide characters.  */
-              mbsrtowcs (wpattern, &pattern, patsize, &ps);
-              assert (mbsinit (&ps));
-              mbsrtowcs (wstring, &string, strsize, &ps);
-
-              res = internal_fnwmatch (wpattern, wstring, wstring + strsize - 1,
-                                       flags & FNM_PERIOD, flags);
-
-              if (__builtin_expect (! (totsize < ALLOCA_LIMIT), 0))
-                free (wpattern);
-              return res;
+              /* Something wrong.
+                 XXX Do we have to set 'errno' to something which
+                 mbsrtows hasn't already done?  */
+            free_return:
+              free (wpattern_malloc);
+              return -1;
+            }
+          if (p)
+            {
+              memset (&ps, '\0', sizeof (ps));
+              goto prepare_wstring;
             }
         }
-    }
+      else
+        {
+        prepare_wstring:
+          n = mbsrtowcs (NULL, &string, 0, &ps);
+          if (__glibc_unlikely (n == (size_t) -1))
+            /* Something wrong.
+               XXX Do we have to set 'errno' to something which mbsrtows hasn't
+               already done?  */
+            goto free_return;
+          if (__glibc_unlikely (n >= (size_t) -1 / sizeof (wchar_t)))
+            {
+              free (wpattern_malloc);
+              __set_errno (ENOMEM);
+              return -2;
+            }
 
-# endif /* HANDLE_MULTIBYTE */
+          wstring_malloc = wstring
+            = (wchar_t *) malloc ((n + 1) * sizeof (wchar_t));
+          if (wstring == NULL)
+            {
+              free (wpattern_malloc);
+              return -2;
+            }
+          assert (mbsinit (&ps));
+          (void) mbsrtowcs (wstring, &string, n + 1, &ps);
+        }
+
+      int res = internal_fnwmatch (wpattern, wstring, wstring + n,
+                                   flags & FNM_PERIOD, flags, NULL,
+                                   alloca_used);
+
+      free (wstring_malloc);
+      free (wpattern_malloc);
+
+      return res;
+    }
 
   return internal_fnmatch (pattern, string, string + strlen (string),
-                           flags & FNM_PERIOD, flags);
+                           flags & FNM_PERIOD, flags, NULL, 0);
 }
 
-# ifdef _LIBC
-#  undef fnmatch
+#undef fnmatch
 versioned_symbol (libc, __fnmatch, fnmatch, GLIBC_2_2_3);
-#  if SHLIB_COMPAT(libc, GLIBC_2_0, GLIBC_2_2_3)
+#if SHLIB_COMPAT(libc, GLIBC_2_0, GLIBC_2_2_3)
 strong_alias (__fnmatch, __fnmatch_old)
 compat_symbol (libc, __fnmatch_old, fnmatch, GLIBC_2_0);
-#  endif
+#endif
 libc_hidden_ver (__fnmatch, fnmatch)
-# endif
-
-#endif  /* _LIBC or not __GNU_LIBRARY__.  */