regex: don't assume uint64_t or uint32_t

This avoids -Werror=overflow errors for 32-bit systems in
the 64-bit case.  Problem reported by Joseph Myers in:
https://sourceware.org/ml/libc-alpha/2017-11/msg00694.html
Also, when this code is used in Gnulib it ports to platforms
that lack uint64_t and uint32_t.  The C standard doesn't guarantee
them, and on some 32-bit compilers there is no uint64_t.
Problem reported by Gianluigi Tiesi in:
http://lists.gnu.org/archive/html/bug-gnulib/2012-03/msg00154.html
* posix/regcomp.c (init_word_char): Don't assume that the types
uint64_t and uint32_t exist.  Adapted from Gnulib patch
2012-05-27T06:40:00!eggert@cs.ucla.edu.  See:
https://git.savannah.gnu.org/cgit/gnulib.git/commit/?id=252b52457da7887667c036d18cc5169777615bb0
This commit is contained in:
Paul Eggert 2017-11-20 16:25:49 -08:00
parent 5a67c4fa01
commit 567d8c1f68
2 changed files with 30 additions and 13 deletions

View File

@ -1,3 +1,19 @@
2017-11-20 Paul Eggert <eggert@cs.ucla.edu>
regex: don't assume uint64_t or uint32_t
This avoids -Werror=overflow errors for 32-bit systems in
the 64-bit case. Problem reported by Joseph Myers in:
https://sourceware.org/ml/libc-alpha/2017-11/msg00694.html
Also, when this code is used in Gnulib it ports to platforms
that lack uint64_t and uint32_t. The C standard doesn't guarantee
them, and on some 32-bit compilers there is no uint64_t.
Problem reported by Gianluigi Tiesi in:
http://lists.gnu.org/archive/html/bug-gnulib/2012-03/msg00154.html
* posix/regcomp.c (init_word_char): Don't assume that the types
uint64_t and uint32_t exist. Adapted from Gnulib patch
2012-05-27T06:40:00!eggert@cs.ucla.edu. See:
https://git.savannah.gnu.org/cgit/gnulib.git/commit/?id=252b52457da7887667c036d18cc5169777615bb0
2017-11-20 Siddhesh Poyarekar <siddhesh@sourceware.org>
* sysdeps/aarch64/memset-reg.h: New file.

View File

@ -925,26 +925,26 @@ init_word_char (re_dfa_t *dfa)
int ch = 0;
if (BE (dfa->map_notascii == 0, 1))
{
if (sizeof (dfa->word_char[0]) == 8)
bitset_word_t bits0 = 0x00000000;
bitset_word_t bits1 = 0x03ff0000;
bitset_word_t bits2 = 0x87fffffe;
bitset_word_t bits3 = 0x07fffffe;
if (BITSET_WORD_BITS == 64)
{
/* The extra temporaries here avoid "implicitly truncated"
warnings in the case when this is dead code, i.e. 32-bit. */
const uint64_t wc0 = UINT64_C (0x03ff000000000000);
const uint64_t wc1 = UINT64_C (0x07fffffe87fffffe);
dfa->word_char[0] = wc0;
dfa->word_char[1] = wc1;
dfa->word_char[0] = bits1 << 31 << 1 | bits0;
dfa->word_char[1] = bits3 << 31 << 1 | bits2;
i = 2;
}
else if (sizeof (dfa->word_char[0]) == 4)
else if (BITSET_WORD_BITS == 32)
{
dfa->word_char[0] = UINT32_C (0x00000000);
dfa->word_char[1] = UINT32_C (0x03ff0000);
dfa->word_char[2] = UINT32_C (0x87fffffe);
dfa->word_char[3] = UINT32_C (0x07fffffe);
dfa->word_char[0] = bits0;
dfa->word_char[1] = bits1;
dfa->word_char[2] = bits2;
dfa->word_char[3] = bits3;
i = 4;
}
else
abort ();
goto general_case;
ch = 128;
if (BE (dfa->is_utf8, 1))
@ -954,6 +954,7 @@ init_word_char (re_dfa_t *dfa)
}
}
general_case:
for (; i < BITSET_WORDS; ++i)
for (int j = 0; j < BITSET_WORD_BITS; ++j, ++ch)
if (isalnum (ch) || ch == '_')