From: Andreas Schwab <schwab@suse.de>
To: Florian Weimer <fweimer@redhat.com>
Cc: libc-alpha@sourceware.org
Subject: Re: [PATCH] Fix handling of collating elements in fnmatch (bug 17396, bug 16976)
Date: Wed, 09 Jan 2019 13:20:00 -0000 [thread overview]
Message-ID: <mvmo98qezmh.fsf@suse.de> (raw)
In-Reply-To: <87zhscfq0h.fsf@oldenburg2.str.redhat.com> (Florian Weimer's message of "Mon, 07 Jan 2019 16:26:06 +0100")
On Jan 07 2019, Florian Weimer <fweimer@redhat.com> wrote:
>> + /* Compare the wide char sequence. */
>> + && memcmp (startp + 1, &wextra[1],
>> + c1 * sizeof (UCHAR)) == 0)
>
> Could this use wmemcmp?
That required adding an internal alias.
Andreas.
[BZ #16976]
[BZ #17396]
* posix/fnmatch_loop.c (internal_fnmatch, internal_fnwmatch): When
looking up collating elements match against (wide) character
sequence instead of name. Correct alignment adjustment.
* posix/fnmatch.c: Don't include "../locale/elem-hash.h".
(WMEMCMP) [HANDLE_MULTIBYTE]: Define.
* posix/Makefile (tests): Add tst-fnmatch4 and tst-fnmatch5.
(LOCALES): Add cs_CZ.ISO-8859-2.
* posix/tst-fnmatch4.c: New file.
* posix/tst-fnmatch5.c: New file.
* include/wchar.h (__wmemcmp): Declare.
* wcsmbs/wmemcmp.c: Define __wmemcmp and add wmemcmp as weak alias.
* sysdeps/i386/i686/multiarch/wmemcmp.c: Likewise.
* sysdeps/x86_64/multiarch/wmemcmp.c: Likewise.
* sysdeps/s390/wmemcmp.c: Likewise.
---
include/wchar.h | 2 +
posix/Makefile | 4 +-
posix/fnmatch.c | 6 +-
posix/fnmatch_loop.c | 228 +++++++-----------
.../s390/wmemcmp.c => posix/tst-fnmatch4.c | 56 +++--
posix/tst-fnmatch5.c | 52 ++++
sysdeps/i386/i686/multiarch/wmemcmp.c | 3 +-
sysdeps/s390/wmemcmp.c | 7 +-
sysdeps/x86_64/multiarch/wmemcmp.c | 3 +-
wcsmbs/wmemcmp.c | 9 +-
10 files changed, 197 insertions(+), 173 deletions(-)
copy sysdeps/s390/wmemcmp.c => posix/tst-fnmatch4.c (52%)
create mode 100644 posix/tst-fnmatch5.c
diff --git a/include/wchar.h b/include/wchar.h
index 86506d28e9..614073bcb3 100644
--- a/include/wchar.h
+++ b/include/wchar.h
@@ -143,6 +143,8 @@ libc_hidden_proto (wmemchr)
libc_hidden_proto (__wmemchr)
libc_hidden_proto (wmemset)
libc_hidden_proto (__wmemset)
+extern int __wmemcmp (const wchar_t *__s1, const wchar_t *__s2, size_t __n)
+ __THROW __attribute_pure__;
/* Now define the internal interfaces. */
extern int __wcscasecmp (const wchar_t *__s1, const wchar_t *__s2)
diff --git a/posix/Makefile b/posix/Makefile
index cfd914ff21..873947f72e 100644
--- a/posix/Makefile
+++ b/posix/Makefile
@@ -93,6 +93,7 @@ tests := test-errno tstgetopt testfnm runtests runptests \
bug-getopt5 tst-getopt_long1 bug-regex34 bug-regex35 \
tst-pathconf tst-rxspencer-no-utf8 \
tst-fnmatch3 bug-regex36 \
+ tst-fnmatch4 tst-fnmatch5 \
tst-posix_spawn-fd tst-posix_spawn-setsid \
tst-posix_fadvise tst-posix_fadvise64 \
tst-sysconf-empty-chroot tst-glob_symlinks tst-fexecve \
@@ -168,7 +169,8 @@ $(objpfx)wordexp-tst.out: wordexp-tst.sh $(objpfx)wordexp-test
endif
LOCALES := cs_CZ.UTF-8 da_DK.ISO-8859-1 de_DE.ISO-8859-1 de_DE.UTF-8 \
- en_US.UTF-8 es_US.ISO-8859-1 es_US.UTF-8 ja_JP.EUC-JP tr_TR.UTF-8
+ en_US.UTF-8 es_US.ISO-8859-1 es_US.UTF-8 ja_JP.EUC-JP tr_TR.UTF-8 \
+ cs_CZ.ISO-8859-2
include ../gen-locales.mk
$(objpfx)bug-regex1.out: $(gen-locales)
diff --git a/posix/fnmatch.c b/posix/fnmatch.c
index 7b225cf2ba..a58e1743ce 100644
--- a/posix/fnmatch.c
+++ b/posix/fnmatch.c
@@ -53,7 +53,6 @@
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>
@@ -237,6 +236,11 @@ __wcschrnul (const wchar_t *s, wint_t c)
# define MEMPCPY(D, S, N) __wmempcpy (D, S, N)
# define MEMCHR(S, C, N) __wmemchr (S, C, N)
# define STRCOLL(S1, S2) wcscoll (S1, S2)
+# ifdef _LIBC
+# define WMEMCMP(S1, S2, N) __wmemcmp (S1, S2, N)
+# else
+# define WMEMCMP(S1, S2, N) wmemcmp (S1, S2, N)
+# endif
# define WIDE_CHAR_VERSION 1
/* Change the name the header defines so it doesn't conflict with
the <locale/weight.h> version included above. */
diff --git a/posix/fnmatch_loop.c b/posix/fnmatch_loop.c
index f888c66dfb..fa39b21395 100644
--- a/posix/fnmatch_loop.c
+++ b/posix/fnmatch_loop.c
@@ -494,26 +494,12 @@ FCT (const CHAR *pattern, const CHAR *string, const CHAR *string_end,
{
int32_t table_size;
const int32_t *symb_table;
-# if WIDE_CHAR_VERSION
- char str[c1];
- unsigned int strcnt;
-# else
-# define str (startp + 1)
-# endif
const unsigned char *extra;
int32_t idx;
int32_t elem;
- int32_t second;
- int32_t hash;
-
# if WIDE_CHAR_VERSION
- /* We have to convert the name to a single-byte
- string. This is possible since the names
- consist of ASCII characters and the internal
- representation is UCS4. */
- for (strcnt = 0; strcnt < c1; ++strcnt)
- str[strcnt] = startp[1 + strcnt];
-#endif
+ CHAR *wextra;
+# endif
table_size =
_NL_CURRENT_WORD (LC_COLLATE,
@@ -525,71 +511,54 @@ FCT (const CHAR *pattern, const CHAR *string, const CHAR *string_end,
_NL_CURRENT (LC_COLLATE,
_NL_COLLATE_SYMB_EXTRAMB);
- /* Locate the character in the hashing table. */
- hash = elem_hash (str, c1);
-
- idx = 0;
- elem = hash % table_size;
- if (symb_table[2 * elem] != 0)
- {
- second = hash % (table_size - 2) + 1;
-
- do
- {
- /* First compare the hashing value. */
- if (symb_table[2 * elem] == hash
- && (c1
- == extra[symb_table[2 * elem + 1]])
- && memcmp (str,
- &extra[symb_table[2 * elem
- + 1]
- + 1], c1) == 0)
- {
- /* Yep, this is the entry. */
- idx = symb_table[2 * elem + 1];
- idx += 1 + extra[idx];
- break;
- }
-
- /* Next entry. */
- elem += second;
- }
- while (symb_table[2 * elem] != 0);
- }
+ for (elem = 0; elem < table_size; elem++)
+ if (symb_table[2 * elem] != 0)
+ {
+ idx = symb_table[2 * elem + 1];
+ /* Skip the name of collating element. */
+ idx += 1 + extra[idx];
+# if WIDE_CHAR_VERSION
+ /* Skip the byte sequence of the
+ collating element. */
+ idx += 1 + extra[idx];
+ /* Adjust for the alignment. */
+ idx = (idx + 3) & ~3;
+
+ wextra = (CHAR *) &extra[idx + 4];
+
+ if (/* Compare the length of the sequence. */
+ c1 == wextra[0]
+ /* Compare the wide char sequence. */
+ && WMEMCMP (startp + 1, &wextra[1],
+ c1) == 0)
+ /* Yep, this is the entry. */
+ break;
+# else
+ if (/* Compare the length of the sequence. */
+ c1 == extra[idx]
+ /* Compare the byte sequence. */
+ && memcmp (startp + 1,
+ &extra[idx + 1], c1) == 0)
+ /* Yep, this is the entry. */
+ break;
+# endif
+ }
- if (symb_table[2 * elem] != 0)
+ if (elem < table_size)
{
/* Compare the byte sequence but only if
this is not part of a range. */
-# if WIDE_CHAR_VERSION
- int32_t *wextra;
+ if (! is_range
- idx += 1 + extra[idx];
- /* Adjust for the alignment. */
- idx = (idx + 3) & ~3;
-
- wextra = (int32_t *) &extra[idx + 4];
-# endif
-
- if (! is_range)
- {
# if WIDE_CHAR_VERSION
- for (c1 = 0;
- (int32_t) c1 < wextra[idx];
- ++c1)
- if (n[c1] != wextra[1 + c1])
- break;
-
- if ((int32_t) c1 == wextra[idx])
- goto matched;
+ && WMEMCMP (n, &wextra[1], c1) == 0
# else
- for (c1 = 0; c1 < extra[idx]; ++c1)
- if (n[c1] != extra[1 + c1])
- break;
-
- if (c1 == extra[idx])
- goto matched;
+ && memcmp (n, &extra[idx + 1], c1) == 0
# endif
+ )
+ {
+ n += c1 - 1;
+ goto matched;
}
/* Get the collation sequence value. */
@@ -597,9 +566,9 @@ FCT (const CHAR *pattern, const CHAR *string, const CHAR *string_end,
# if WIDE_CHAR_VERSION
cold = wextra[1 + wextra[idx]];
# else
- /* Adjust for the alignment. */
idx += 1 + extra[idx];
- idx = (idx + 3) & ~4;
+ /* Adjust for the alignment. */
+ idx = (idx + 3) & ~3;
cold = *((int32_t *) &extra[idx]);
# endif
@@ -609,10 +578,10 @@ FCT (const CHAR *pattern, const CHAR *string, const CHAR *string_end,
{
/* No valid character. Match it as a
single byte. */
- if (!is_range && *n == str[0])
+ if (!is_range && *n == startp[1])
goto matched;
- cold = str[0];
+ cold = startp[1];
c = *p++;
}
else
@@ -620,7 +589,6 @@ FCT (const CHAR *pattern, const CHAR *string, const CHAR *string_end,
}
}
else
-# undef str
#endif
{
c = FOLD (c);
@@ -712,25 +680,11 @@ FCT (const CHAR *pattern, const CHAR *string, const CHAR *string_end,
{
int32_t table_size;
const int32_t *symb_table;
-# if WIDE_CHAR_VERSION
- char str[c1];
- unsigned int strcnt;
-# else
-# define str (startp + 1)
-# endif
const unsigned char *extra;
int32_t idx;
int32_t elem;
- int32_t second;
- int32_t hash;
-
# if WIDE_CHAR_VERSION
- /* We have to convert the name to a single-byte
- string. This is possible since the names
- consist of ASCII characters and the internal
- representation is UCS4. */
- for (strcnt = 0; strcnt < c1; ++strcnt)
- str[strcnt] = startp[1 + strcnt];
+ CHAR *wextra;
# endif
table_size =
@@ -743,71 +697,63 @@ FCT (const CHAR *pattern, const CHAR *string, const CHAR *string_end,
_NL_CURRENT (LC_COLLATE,
_NL_COLLATE_SYMB_EXTRAMB);
- /* Locate the character in the hashing
- table. */
- hash = elem_hash (str, c1);
-
- idx = 0;
- elem = hash % table_size;
- if (symb_table[2 * elem] != 0)
- {
- second = hash % (table_size - 2) + 1;
-
- do
- {
- /* First compare the hashing value. */
- if (symb_table[2 * elem] == hash
- && (c1
- == extra[symb_table[2 * elem + 1]])
- && memcmp (str,
- &extra[symb_table[2 * elem + 1]
- + 1], c1) == 0)
- {
- /* Yep, this is the entry. */
- idx = symb_table[2 * elem + 1];
- idx += 1 + extra[idx];
- break;
- }
-
- /* Next entry. */
- elem += second;
- }
- while (symb_table[2 * elem] != 0);
- }
-
- if (symb_table[2 * elem] != 0)
- {
- /* Compare the byte sequence but only if
- this is not part of a range. */
+ for (elem = 0; elem < table_size; elem++)
+ if (symb_table[2 * elem] != 0)
+ {
+ idx = symb_table[2 * elem + 1];
+ /* Skip the name of collating
+ element. */
+ idx += 1 + extra[idx];
# if WIDE_CHAR_VERSION
- int32_t *wextra;
-
- idx += 1 + extra[idx];
- /* Adjust for the alignment. */
- idx = (idx + 3) & ~4;
-
- wextra = (int32_t *) &extra[idx + 4];
+ /* Skip the byte sequence of the
+ collating element. */
+ idx += 1 + extra[idx];
+ /* Adjust for the alignment. */
+ idx = (idx + 3) & ~3;
+
+ wextra = (CHAR *) &extra[idx + 4];
+
+ if (/* Compare the length of the
+ sequence. */
+ c1 == wextra[0]
+ /* Compare the wide char sequence. */
+ && WMEMCMP (startp + 1, &wextra[1],
+ c1) == 0)
+ /* Yep, this is the entry. */
+ break;
+# else
+ if (/* Compare the length of the
+ sequence. */
+ c1 == extra[idx]
+ /* Compare the byte sequence. */
+ && memcmp (startp + 1,
+ &extra[idx + 1], c1) == 0)
+ /* Yep, this is the entry. */
+ break;
# endif
+ }
+
+ if (elem < table_size)
+ {
/* Get the collation sequence value. */
is_seqval = 1;
# if WIDE_CHAR_VERSION
cend = wextra[1 + wextra[idx]];
# else
- /* Adjust for the alignment. */
idx += 1 + extra[idx];
- idx = (idx + 3) & ~4;
+ /* Adjust for the alignment. */
+ idx = (idx + 3) & ~3;
cend = *((int32_t *) &extra[idx]);
# endif
}
- else if (symb_table[2 * elem] != 0 && c1 == 1)
+ else if (c1 == 1)
{
- cend = str[0];
+ cend = startp[1];
c = *p++;
}
else
return FNM_NOMATCH;
}
-# undef str
}
else
{
diff --git a/sysdeps/s390/wmemcmp.c b/posix/tst-fnmatch4.c
similarity index 52%
copy from sysdeps/s390/wmemcmp.c
copy to posix/tst-fnmatch4.c
index ec0b4027f8..dc13f89288 100644
--- a/sysdeps/s390/wmemcmp.c
+++ b/posix/tst-fnmatch4.c
@@ -1,5 +1,5 @@
-/* Multiple versions of wmemcmp.
- Copyright (C) 2015-2019 Free Software Foundation, Inc.
+/* Test for fnmatch handling of collating elements
+ Copyright (C) 2019 Free Software Foundation, Inc.
This file is part of the GNU C Library.
The GNU C Library is free software; you can redistribute it and/or
@@ -16,23 +16,35 @@
License along with the GNU C Library; if not, see
<http://www.gnu.org/licenses/>. */
-#include <ifunc-wmemcmp.h>
-
-#if HAVE_WMEMCMP_IFUNC
-# include <wchar.h>
-# include <ifunc-resolve.h>
-
-# if HAVE_WMEMCMP_C
-extern __typeof (wmemcmp) WMEMCMP_C attribute_hidden;
-# endif
-
-# if HAVE_WMEMCMP_Z13
-extern __typeof (wmemcmp) WMEMCMP_Z13 attribute_hidden;
-# endif
-
-s390_libc_ifunc_expr (wmemcmp, wmemcmp,
- (HAVE_WMEMCMP_Z13 && (hwcap & HWCAP_S390_VX))
- ? WMEMCMP_Z13
- : WMEMCMP_DEFAULT
- )
-#endif
+#include <stdio.h>
+#include <locale.h>
+#include <fnmatch.h>
+
+static int
+do_test_locale (const char *locale)
+{
+ const char *pattern = "[[.ch.]]";
+
+ if (setlocale (LC_ALL, locale) == NULL)
+ {
+ printf ("could not set locale %s\n", locale);
+ return 1;
+ }
+
+ if (fnmatch (pattern, "ch", 0) != 0)
+ {
+ printf ("%s didn't match in locale %s\n", pattern, locale);
+ return 1;
+ }
+
+ return 0;
+}
+
+static int
+do_test (void)
+{
+ return (do_test_locale ("cs_CZ.ISO-8859-2")
+ || do_test_locale ("cs_CZ.UTF-8"));
+}
+
+#include <support/test-driver.c>
diff --git a/posix/tst-fnmatch5.c b/posix/tst-fnmatch5.c
new file mode 100644
index 0000000000..5ebff1f424
--- /dev/null
+++ b/posix/tst-fnmatch5.c
@@ -0,0 +1,52 @@
+/* Test for fnmatch handling of collating elements
+ Copyright (C) 2019 Free Software Foundation, Inc.
+ This file is part of the GNU C Library.
+
+ The GNU C Library is free software; you can redistribute it and/or
+ modify it under the terms of the GNU Lesser General Public
+ License as published by the Free Software Foundation; either
+ version 2.1 of the License, or (at your option) any later version.
+
+ The GNU C Library is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ Lesser General Public License for more details.
+
+ You should have received a copy of the GNU Lesser General Public
+ License along with the GNU C Library; if not, see
+ <http://www.gnu.org/licenses/>. */
+
+#include <fnmatch.h>
+#include <locale.h>
+#include <stdio.h>
+#include <string.h>
+
+#define LENGTH 20000000
+
+static char pattern[LENGTH + 7];
+
+static int
+do_test (void)
+{
+ if (setlocale (LC_ALL, "en_US.UTF-8") == NULL)
+ {
+ puts ("could not set locale");
+ return 1;
+ }
+ pattern[0] = '[';
+ pattern[1] = '[';
+ pattern[2] = '.';
+ memset (pattern + 3, 'a', LENGTH);
+ pattern[LENGTH + 3] = '.';
+ pattern[LENGTH + 4] = ']';
+ pattern[LENGTH + 5] = ']';
+ int ret = fnmatch (pattern, "a", 0);
+ if (ret == 0)
+ {
+ puts ("fnmatch returned 0 for invalid pattern");
+ return 1;
+ }
+ return 0;
+}
+
+#include <support/test-driver.c>
diff --git a/sysdeps/i386/i686/multiarch/wmemcmp.c b/sysdeps/i386/i686/multiarch/wmemcmp.c
index ce25991352..7674530c68 100644
--- a/sysdeps/i386/i686/multiarch/wmemcmp.c
+++ b/sysdeps/i386/i686/multiarch/wmemcmp.c
@@ -26,5 +26,6 @@
# define SYMBOL_NAME wmemcmp
# include "ifunc-ssse3-sse4_2.h"
-libc_ifunc_redirected (__redirect_wmemcmp, wmemcmp, IFUNC_SELECTOR ());
+libc_ifunc_redirected (__redirect_wmemcmp, __wmemcmp, IFUNC_SELECTOR ());
+weak_alias (__wmemcmp, wmemcmp)
#endif
diff --git a/sysdeps/s390/wmemcmp.c b/sysdeps/s390/wmemcmp.c
index ec0b4027f8..2f619550a6 100644
--- a/sysdeps/s390/wmemcmp.c
+++ b/sysdeps/s390/wmemcmp.c
@@ -23,16 +23,17 @@
# include <ifunc-resolve.h>
# if HAVE_WMEMCMP_C
-extern __typeof (wmemcmp) WMEMCMP_C attribute_hidden;
+extern __typeof (__wmemcmp) WMEMCMP_C attribute_hidden;
# endif
# if HAVE_WMEMCMP_Z13
-extern __typeof (wmemcmp) WMEMCMP_Z13 attribute_hidden;
+extern __typeof (__wmemcmp) WMEMCMP_Z13 attribute_hidden;
# endif
-s390_libc_ifunc_expr (wmemcmp, wmemcmp,
+s390_libc_ifunc_expr (__wmemcmp, __wmemcmp,
(HAVE_WMEMCMP_Z13 && (hwcap & HWCAP_S390_VX))
? WMEMCMP_Z13
: WMEMCMP_DEFAULT
)
+weak_alias (__wmemcmp, wmemcmp)
#endif
diff --git a/sysdeps/x86_64/multiarch/wmemcmp.c b/sysdeps/x86_64/multiarch/wmemcmp.c
index 136a7b05e1..826c90bb77 100644
--- a/sysdeps/x86_64/multiarch/wmemcmp.c
+++ b/sysdeps/x86_64/multiarch/wmemcmp.c
@@ -26,5 +26,6 @@
# define SYMBOL_NAME wmemcmp
# include "ifunc-memcmp.h"
-libc_ifunc_redirected (__redirect_wmemcmp, wmemcmp, IFUNC_SELECTOR ());
+libc_ifunc_redirected (__redirect_wmemcmp, __wmemcmp, IFUNC_SELECTOR ());
+weak_alias (__wmemcmp, wmemcmp)
#endif
diff --git a/wcsmbs/wmemcmp.c b/wcsmbs/wmemcmp.c
index 5b243bab8f..5e137fd3dd 100644
--- a/wcsmbs/wmemcmp.c
+++ b/wcsmbs/wmemcmp.c
@@ -18,12 +18,12 @@
#include <wchar.h>
-#ifndef WMEMCMP
-# define WMEMCMP wmemcmp
+#ifdef WMEMCMP
+# define __wmemcmp WMEMCMP
#endif
int
-WMEMCMP (const wchar_t *s1, const wchar_t *s2, size_t n)
+__wmemcmp (const wchar_t *s1, const wchar_t *s2, size_t n)
{
wchar_t c1;
wchar_t c2;
@@ -81,3 +81,6 @@ WMEMCMP (const wchar_t *s1, const wchar_t *s2, size_t n)
return 0;
}
+#ifndef WMEMCMP
+weak_alias (__wmemcmp, wmemcmp)
+#endif
--
2.20.1
--
Andreas Schwab, SUSE Labs, schwab@suse.de
GPG Key fingerprint = 0196 BAD8 1CE9 1970 F4BE 1748 E4D4 88E3 0EEA B9D7
"And now for something completely different."
next prev parent reply other threads:[~2019-01-09 13:20 UTC|newest]
Thread overview: 10+ messages / expand[flat|nested] mbox.gz Atom feed top
2018-12-11 15:55 Andreas Schwab
2019-01-07 15:26 ` Florian Weimer
2019-01-09 13:20 ` Andreas Schwab [this message]
2019-01-09 13:35 ` Adhemerval Zanella
2019-01-09 14:32 ` Andreas Schwab
2019-01-09 18:33 ` Adhemerval Zanella
2019-01-10 8:43 ` Andreas Schwab
2019-01-10 12:25 ` Adhemerval Zanella
2019-01-10 13:20 ` Andreas Schwab
2019-01-09 14:05 ` Florian Weimer
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=mvmo98qezmh.fsf@suse.de \
--to=schwab@suse.de \
--cc=fweimer@redhat.com \
--cc=libc-alpha@sourceware.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).