aboutsummaryrefslogtreecommitdiff
diff options
context:
space:
mode:
authorUlrich Drepper <drepper@redhat.com>2006-09-07 13:50:31 +0000
committerUlrich Drepper <drepper@redhat.com>2006-09-07 13:50:31 +0000
commitba40cc1540f5ee1e2082bd31b0e602a62ea3273f (patch)
treea1dc81d0ece8d2107f47112178fb5c68ddb21da8 /posix/regex_internal.c
parent* locale/programs/ld-collate.c (collate_read): Goto sym_equiv_free (diff)
downloadglibc-ba40cc1540f5ee1e2082bd31b0e602a62ea3273f.tar.gz
glibc-ba40cc1540f5ee1e2082bd31b0e602a62ea3273f.tar.bz2
glibc-ba40cc1540f5ee1e2082bd31b0e602a62ea3273f.zip
[BZ #3155]
2006-09-07 Jakub Jelinek <jakub@redhat.com> [BZ #3155] * sysdeps/powerpc/powerpc32/fpu/s_lrint.S (__lrint): Don't access stack below r1. * posix/regex_internal.c (re_string_reconstruct): Handle offset < pstr->valid_raw_len && pstr->offsets_needed case. Ensure no bytes read before raw_mbs array. Pass a saved copy of pstr->valid_len - 1 rather than pstr->valid_raw_len - 1 to re_string_context_at. * posix/Makefile: Add rules to build and run bug-regex26 test. * posix/bug-regex26.c: New test. * dlfcn/Makefile (LDLIBS-bug-atexit3-lib.so): Add ld.so.
Diffstat (limited to 'posix/regex_internal.c')
-rw-r--r--posix/regex_internal.c106
1 files changed, 86 insertions, 20 deletions
diff --git a/posix/regex_internal.c b/posix/regex_internal.c
index ac312db0cd..66154e0cea 100644
--- a/posix/regex_internal.c
+++ b/posix/regex_internal.c
@@ -585,34 +585,98 @@ re_string_reconstruct (re_string_t *pstr, int idx, int eflags)
if (BE (offset != 0, 1))
{
- /* Are the characters which are already checked remain? */
- if (BE (offset < pstr->valid_raw_len, 1)
-#ifdef RE_ENABLE_I18N
- /* Handling this would enlarge the code too much.
- Accept a slowdown in that case. */
- && pstr->offsets_needed == 0
-#endif
- )
+ /* Should the already checked characters be kept? */
+ if (BE (offset < pstr->valid_raw_len, 1))
{
/* Yes, move them to the front of the buffer. */
- 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));
+ 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 (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);
+ assert (pstr->valid_len > 0);
#endif
+ }
}
else
{
/* No, skip all characters until IDX. */
+ int prev_valid_len = pstr->valid_len;
+
#ifdef RE_ENABLE_I18N
if (BE (pstr->offsets_needed, 0))
{
@@ -636,6 +700,8 @@ re_string_reconstruct (re_string_t *pstr, int idx, int eflags)
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
@@ -643,7 +709,7 @@ re_string_reconstruct (re_string_t *pstr, int idx, int eflags)
if (isascii (*p) && BE (pstr->trans == NULL, 1))
{
memset (&pstr->cur_state, '\0', sizeof (mbstate_t));
- pstr->valid_len = 0;
+ /* pstr->valid_len = 0; */
wc = (wchar_t) *p;
}
else
@@ -686,7 +752,7 @@ re_string_reconstruct (re_string_t *pstr, int idx, int eflags)
pstr->valid_len = re_string_skip_chars (pstr, idx, &wc) - idx;
if (wc == WEOF)
pstr->tip_context
- = re_string_context_at (pstr, pstr->valid_raw_len - 1, eflags);
+ = 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))