This patch syncs the regex implementation with gnulib (commit 0ee5212).
Only two changes in GLIBC regex testing are required:
1. posix/bug-regex28.c: as previously discussed [1] the change of
expected results on the pattern should be safe.
2. posix/PCRE.tests: the ERE (a)|\1 is malformed (in the sense that
the \1 doesn't mean anything) and although current GLIBC accepts
it has undefined behavior. This patch removes the specific test.
This sync contains some patches from thread 'Regex: Make libc regex
more usable outside GLIBC.' [2] which have been pushed upstream in
gnulib. This patches also fixes some regex issues (BZ #23233,
BZ #21163, BZ #18986, BZ #13762) and I did not add testcases for
both #23233 and #13762 because I couldn't think a simple way to
trigger the expected failure path to trigger them.
Checked on x86_64-linux-gnu and i686-linux-gnu.
[BZ #23233]
[BZ #21163]
[BZ #18986]
[BZ #13762]
* posix/Makefile (tests): Add bug-regex37 and bug-regex38.
* posix/PCRE.tests: Remove invalid test.
* posix/bug-regex28.c: Fix expected values for used syntax.
* posix/bug-regex37.c: New file.
* posix/bug-regex38.c: Likewise.
* posix/regcomp.c: Sync with gnulib.
* posix/regex.c: Likewise.
* posix/regex.h: Likewise.
* posix/regex_internal.c: Likewise.
* posix/regex_internal.h: Likewise.
* posix/regexec.c: Likewise.
[1] https://sourceware.org/ml/libc-alpha/2017-12/msg00807.html
[2] https://sourceware.org/ml/libc-alpha/2017-12/msg00237.html
The regex code brings in references to wcrtomb, which isn't in all the
standards that contain regex. This patch makes it call __wcrtomb
instead (in fact some places already called __wcrtomb, so this patch
makes it internally consistent about which name is used).
Tested for x86_64 and x86 that installed stripped shared libraries are
unchanged by the patch.
[BZ #18496]
* posix/regex_internal.c (build_wcs_upper_buffer): Call __wcrtomb
instead of wcrtomb.
regcomp brings in references to various wctype functions that aren't
in all the standards including regcomp. This patch fixes this in the
usual way by using the __* versions of these functions (which already
exist, but some didn't have libc_hidden_proto / libc_hidden_def
before).
Tested for x86_64 and x86 (testsuite, and that installed stripped
shared libraries are unchanged by the patch). (Other wide character
function references from the regex code mean that this patch by itself
doesn't fix any XFAILed linknamespace test failures; further patches
will be needed for that.)
[BZ #18495]
* wctype/wcfuncs.c (__iswalnum): Use libc_hidden_def.
(__iswlower): Likewise.
* include/wctype.h (__iswalnum): Declare. Use libc_hidden_proto.
(__iswlower): Likewise.
* posix/regcomp.c (re_compile_fastmap_iter): Call __towlower
instead of towlower.
* posix/regex_internal.c (build_wcs_upper_buffer): Call __iswlower
instead of iswlower. Call __towupper instead of towupper.
* posix/regex_internal.h (IS_WIDE_WORD_CHAR): Call __iswalnum
instead of iswalnum.
The DEBUG macro is checked for its value in one place and if it is
defined in another. Make this consistent across the two cases and use
the same style that we did in mktime.c, which is to check if the macro
is defined and it is set.
* posix/regex_internal.c (re_string_skip_chars): Fix miscomputation
of remain_len that may cause incomplete multi-byte character and
false match.
* posix/bug-regex33.c: New file.
* posix/Makefile (tests): Add bug-regex33.
The problem is that parse_bracket_symbol is miscompiled, and it turns
out it is because of an incorrect attribute on re_string_fetch_byte_case.
Unlike re_string_peek_byte_case, this one is really not pure, it modifies memory
(increments pstr->cur_idx), and with the pure attribute GCC assumed it doesn't
and it cached the presumed value of regexp->cur_idx in a variable across the
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;
}
2008-03-29 Eric Blake <ebb9@byu.net>
Rewrite string searches to O(n) rather than O(n^2).
* string/str-two-way.h: New file. For linear fixed-allocation
string searching.
* string/memmem.c: New implementation.
* string/strstr.c: New implementation.
* string/strcasestr.c: New implementation.
* sysdeps/posix/getaddrinfo.c (getaddrinfo): Call _res_hconf_init
been converted at all, set *last_wc to WEOF. If mbrtowc failed, set wc
to the byte which couldn't be converted.
(re_string_reconstruct): Don't clear valid_raw_len before calling
re_string_skip_chars. If wc is WEOF after re_string_skip_chars, set
tip_context using re_string_context_at.
* posix/Makefile: Add rules to build and run bug-regex25 test.
* posix/bug-regex25.c: New test.
mbrtowc for very simple UTF-8 case.
2005-09-01 Paul Eggert <eggert@cs.ucla.edu>
* posix/regex_internal.c (build_wcs_upper_buffer): Fix portability
bugs in int versus size_t comparisons.
2005-09-06 Ulrich Drepper <drepper@redhat.com>
* posix/regex_internal.c (re_acquire_state): Make DFA pointer arg
a pointer-to-const.
(re_acquire_state_context): Likewise.
* posix/regex_internal.h: Adjust prototypes.
2005-08-31 Jim Meyering <jim@meyering.net>
* posix/regcomp.c (search_duplicated_node): Make first pointer arg
a pointer-to-const.
* posix/regex_internal.c (create_ci_newstate, create_cd_newstate,
register_state): Likewise.
* posix/regexec.c (search_cur_bkref_entry, check_dst_limits):
(check_dst_limits_calc_pos_1, check_dst_limits_calc_pos):
(group_nodes_into_DFAstates): Likewise.
* posix/regexec.c (re_search_internal): Simplify update of
rm_so and rm_eo by replacing "if (A == B) A += C - B;"
with the equivalent of "if (A == B) A = C;".
2005-09-06 Ulrich Drepper <drepper@redhat.com>
* posix/regcomp.c (re_compile_internal): Change third parameter type
to size_t.
(init_dfa): Likewise. Make sure that arithmetic on pat_len doesn't
overflow.
* posix/regex_internal.h (struct re_dfa_t): Change type of nodes_alloc
and nodes_len to size_t.
* posix/regex_internal.c (re_dfa_add_node): Use size_t as type for
new_nodes_alloc. Check for overflow.
2005-08-31 Paul Eggert <eggert@cs.ucla.edu>
* posix/regcomp.c (re_compile_fastmap_iter, init_dfa, init_word_char):
(optimize_subexps, lower_subexp):
Don't assume 1<<31 has defined behavior on hosts with 32-bit int,
since the signed shift might overflow. Use 1u<<31 instead.
* posix/regex_internal.h (bitset_set, bitset_clear, bitset_contain):
Likewise.
* posix/regexec.c (check_dst_limits_calc_pos_1): Likewise.
(check_subexp_matching_top): Likewise.
* posix/regcomp.c (optimize_subexps, lower_subexp):
Use CHAR_BIT rather than 8, for clarity.
* posix/regexec.c (check_dst_limits_calc_pos_1):
(check_subexp_matching_top): Likewise.
* posix/regcomp.c (init_dfa): Make table_size unsigned, so that we
don't have to worry about portability issues when shifting it left.
Remove no-longer-needed test for table_size > 0.
* posix/regcomp.c (parse_sub_exp): Do not shift more bits than there
are in a word, as the resulting behavior is undefined.
* posix/regexec.c (check_dst_limits_calc_pos_1): Likewise;
in one case, a <= should have been an <, and in another case the
whole test was missing.
* posix/regex_internal.h (BYTE_BITS): Remove. All uses changed to
the standard name CHAR_BIT.
next_last_offset.
(struct re_dfa_t): Remove unused member states_alloc.
* posix/regcomp.c (init_dfa): Don't initialize unused members.
2005-08-25 Paul Eggert <eggert@cs.ucla.edu>
* posix/regexec.c (set_regs): Don't alloca with an unbounded size.
alloca modernization/simplification for regex.
* posix/regex.c: Remove portability cruft for alloca. This no longer
needs to be at the start of the file, and can be moved into
regex_internal.h and simplified.
* posix/regex_internal.h: Include <alloca.h>.
(__libc_use_alloca) [!defined _LIBC]: New macro.
* posix/regexec.c (build_trtable): Remove "#ifdef _LIBC",
since the code now works outside glibc.
2005-09-06 Ulrich Drepper <drepper@redhat.com>
* include/regex.h: Remove use of _RE_ARGS.
2005-08-25 Paul Eggert <eggert@cs.ucla.edu>
* posix/regexec.c (find_recover_state): Change "err" to "*err".
2005-08-24 Paul Eggert <eggert@cs.ucla.edu>
* posix/regcomp.c (regerror): Pointer args are 'restrict',
as per POSIX.
* posix/regex.h (regerror): Likewise.
* manual/pattern.texi (POSIX Regexp Compilation): Likewise.
Similarly for regcomp and regexec. Also, first 2 args of regexec
and 2nd arg of regerror are const.
* posix/regex.c: Do not include <sys/types.h>, as POSIX no longer
requires this. (The code never needed it.)
2005-08-20 Paul Eggert <eggert@cs.ucla.edu>
* posix/regexec.c (sift_states_bkref): re_node_set_insert returns
int, not reg_errcode_t.
* posix/regex_internal.c (calc_state_hash): Put 'inline' before type,
since some broken compilers warn about it otherwise.
* posix/regcomp.c (create_initial_state): Remove duplicate decl.
2005-08-20 Paul Eggert <eggert@cs.ucla.edu>
* posix/regex.h (_RE_ARGS): Remove. No longer needed, since we assume
C89 or better. All uses removed.
2005-09-06 Ulrich Drepper <drepper@redhat.com>
* posix/regex.c: Prevent using C++ compilers.
2005-08-19 Paul Eggert <eggert@cs.ucla.edu>
* posix/regcomp.c (duplicate_node): Return new index, not an error
code, and let the caller return REG_ESPACE if out of space. This
removes an uninitialied-variable warning with GCC 4.0.1, and also
avoids taking the address of a local variable. All callers
changed.
2005-09-06 Ulrich Drepper <drepper@redhat.com>
* include/time.h (__strptime_internal): Rename parameter to avoid
bogus compiler warning.
2005-08-19 Jim Meyering <jim@meyering.net>
* posix/regexec.c (proceed_next_node): Redo local variables to
avoid GCC shadowing warnings.
2005-09-06 Ulrich Drepper <drepper@redhat.com>
* posix/regex_internal.c (re_acquire_state): Minor code rearrangement.
(re_acquire_state_context): Likewise.
2005-08-19 Paul Eggert <eggert@cs.ucla.edu>
* posix/regex_internal.c (re_string_realloc_buffers):
(re_node_set_insert, re_node_set_insert_last, re_dfa_add_node):
Rename local variables to avoid GCC shadowing warnings.
2005-07-08 Eric Blake <ebb9@byu.net>
Paul Eggert <eggert@cs.ucla.edu>
* posix/regcomp.c (init_dfa): Store __btowc value in wint_t, not
wchar_t. Remove now-unnecessary cast.
(build_range_exp): Likewise.
Update.
2005-01-27 Paolo Bonzini <bonzini@gnu.org>
[BZ #558]
* posix/regcomp.c (calc_inveclosure): Return reg_errcode_t.
Initialize the node sets in dfa->inveclosures.
(analyze): Initialize inveclosures only if it is needed.
Check errors from calc_inveclosure.
* posix/regex_internal.c (re_dfa_add_node): Do not initialize
the inveclosure node set.
* posix/regexec.c (re_search_internal): If nmatch includes unused
subexpressions, reset them to { rm_so: -1, rm_eo: -1 } here.
* posix/regcomp.c (parse_bracket_exp) [!RE_ENABLE_I18N]:
Do build a SIMPLE_BRACKET token.
* posix/regexec.c (transit_state_mb): Do not examine nodes
where ACCEPT_MB is not set.
Update.
2004-12-13 Paolo Bonzini <bonzini@gnu.org>
Separate parsing and creation of the NFA. Avoided recursion on
the (very unbalanced) parse tree.
[BZ #611]
* posix/regcomp.c (struct subexp_optimize, analyze_tree, calc_epsdest,
re_dfa_add_tree_node, mark_opt_subexp_iter): Removed.
(optimize_subexps, duplicate_tree, calc_first, calc_next,
mark_opt_subexp): Rewritten.
(preorder, postorder, lower_subexps, lower_subexp, link_nfa_nodes,
create_token_tree, free_tree, free_token): New.
(analyze): Accept a regex_t *. Invoke the passes via the preorder and
postorder generic visitors. Do not initialize the fields in the
re_dfa_t that represent the transitions.
(free_dfa_content): Use free_token.
(re_compile_internal): Analyze before UTF-8 optimizations. Do not
include optimization of subexpressions.
(create_initial_state): Fetch the DFA node index from the first node's
bin_tree_t *.
(optimize_utf8): Abort on unexpected nodes, including OP_DUP_QUESTION.
Return on COMPLEX_BRACKET.
(duplicate_node_closure): Fix comment.
(duplicate_node): Do not initialize the fields in the
re_dfa_t that represent the transitions.
(calc_eclosure, calc_inveclosure): Do not handle OP_DELETED_SUBEXP.
(create_tree): Remove final argument. All callers adjusted. Rewritten
to use create_token_tree.
(parse_reg_exp, parse_branch, parse_expression, parse_bracket_exp,
build_charclass_op): Use create_tree or create_token_tree instead
of re_dfa_add_tree_node.
(parse_dup_op): Likewise. Also free the tree using free_tree for
"<re>{0}", and lower OP_DUP_QUESTION to OP_ALT: "a?" is equivalent
to "a|". Adjust invocation of mark_opt_subexp.
(parse_sub_exp): Create a single SUBEXP node.
* posix/regex_internal.c (re_dfa_add_node): Remove last parameter,
always perform as if it was 1. Do not initialize OPT_SUBEXP and
DUPLICATED, and initialize the DFA fields representing the transitions.
* posix/regex_internal.h (re_dfa_add_node): Adjust prototype.
(re_token_type_t): Move OP_DUP_PLUS and OP_DUP_QUESTION to the tokens
section. Add a tree-only code SUBEXP. Remove OP_DELETED_SUBEXP.
(bin_tree_t): Include a full re_token_t for TOKEN. Turn FIRST and
NEXT into pointers to trees. Remove ECLOSURE.
2004-12-28 Paolo Bonzini <bonzini@gnu.org >
[BZ #605]
* posix/regcomp.c (parse_bracket_exp): Do not modify DFA nodes
that were already created.
* posix/regex_internal.c (re_dfa_add_node): Set accept_mb field
in the token if needed.
(create_ci_newstate, create_cd_newstate): Set accept_mb field
from the tokens' field.
* posix/regex_internal.h (re_token_t): Add accept_mb field.
(ACCEPT_MB_NODE): Removed.
* posix/regexec.c (proceed_next_node, transit_states_mb,
build_sifted_states, check_arrival_add_next_nodes): Use
accept_mb instead of ACCEPT_MB_NODE.
2005-01-06 Ulrich Drepper <drepper@redhat.com>
* misc/sys/cdefs.h: Define __wur.
* libio/stdio.h: Use __wur for a number of interfaces.
* posix/unistd.h: Likewise.
* posix/regex_internal.c (free_state): Free word_trtable.
2004-12-07 Paolo Bonzini <bonzini@gnu.org>
* posix/regexec.c (proceed_next_node): Simplify treatment of epsilon
nodes. Pass the pushed node to push_fail_stack.
(push_fail_stack): Accept a single node rather than an array
of two epsilon destinations.
(build_sifted_states): Only walk non-epsilon nodes.
(check_arrival): Don't pass epsilon nodes to
check_arrival_add_next_nodes.
(check_arrival_add_next_nodes) [DEBUG]: Abort if an epsilon node is
found.
(check_node_accept): Do expensive checks later.
(add_epsilon_src_nodes): Cache result of merging the inveclosures.
* posix/regex_internal.h (re_dfastate_t): Add non_eps_nodes and
inveclosure.
(re_string_elem_size_at, re_string_char_size_at, re_string_wchar_at,
re_string_context_at, re_string_peek_byte_case,
re_string_fetch_byte_case, re_node_set_compare, re_node_set_contains):
Declare as pure.
* posix/regex_internal.c (create_newstate_common): Remove.
(register_state): Move part of it here. Initialize non_eps_nodes.
(free_state): Free inveclosure and non_eps_nodes.
(create_cd_newstate, create_ci_newstate): Allocate the new
re_dfastate_t here.
2004-11-23 Paolo Bonzini <bonzini@gnu.org>
* posix/regcomp.c (analyze_tree): Always call calc_epsdest.
(calc_inveclosure): Use re_node_set_insert_last.
(parse_dup_op): Lower X{1,5} to (X(X(X(XX?)?)?)?)?
rather than X?X?X?X?X?.
* posix/regex_internal.h (re_node_set_insert_last): New declaration.
* posix/regex_internal.c (re_node_set_insert_last): New function.
* posix/PCRE.tests: Add testcases.
2004-11-09 Paolo Bonzini <bonzini@gnu.org>
* posix/regexec.c (transit_state): Remove the check for
out-of-bounds buffers.
(check_matching): Check here for out-of-bounds buffers.
(re_search_internal): Store into match_kind a set of bits
indicating which incantation of fastmap scanning must be
used. Use a switch statement instead of multiple ifs.
Exit the final "for (;;)" with goto free_return unless
the match succeeded, thus simplifying some conditionals.
* posix/regex_internal.c (re_string_reconstruct,
re_string_context_at): Add several branch predictions for
case-sensitive matching and no transition table being used.
2004-11-10 Ulrich Drepper <drepper@redhat.com>
* posix/tst-waitid.c: Don't use error to print error message, they
won't end up in the .out file.
* nscd/nscd_getgr_r.c: Likewise. Make map externally visible.
* nscd/nscd_gethst_r.c: Likewise.
2004-11-08 Ulrich Drepper <drepper@redhat.com>
* posix/regcomp.c (utf8_sb_map): Define.
(free_dfa_content): Don't free dfa->sb_char if it's a pointer to
utf8_sb_map.
(init_dfa): Use utf8_sb_map instead of initializing memory when the
encoding is UTF-8.
* posix/regcomp.c (init_dfa): Get the codeset name outside glibc as
well. Check if it is spelled UTF8 as well as UTF-8, and check
case-insensitively. Set dfa->map_notascii manually when outside
glibc.
* posix/regex_internal.c (build_wcs_upper_buffer) [!_LIBC]: Enable
optimizations based on map_notascii.
* posix/regex_internal.h [HAVE_LANGINFO_H || HAVE_LANGINFO_CODESET
|| _LIBC]: Include langinfo.h.
* posix/regex_internal.h (struct re_backref_cache_entry): Add "more"
field.
* posix/regexec.c (check_dst_limits): Hoist computation of the source
and destination bkref_idx out of the loop. Pass it to
check_dst_limits_calc_pos.
(check_dst_limits_calc_pos_1): New function, containing the recursive
loop of check_dst_limits_calc_pos; uses the "more" field of
struct re_backref_cache to control the loop.
(check_dst_limits_calc_pos): Store into "boundaries" the position
relative to lim's start and end positions. Do not accept eclosures,
accept bkref_idx instead. Call check_dst_limits_calc_pos_1 to do the
work.
(sift_states_bkref): Use the "more" field of struct re_backref_cache
to control the loop. A big "if" was turned into a continue and the
function was reindented.
(get_subexp): Use the "more" field of struct re_backref_cache
to control the loop.
(match_ctx_add_entry): Initialize the bkref_ents' "more" field.
(search_cur_bkref_entry): Return -1 if out of bounds.
* posix/regexec.c (empty_set): Remove.
(sift_states_backward): Remove cur_src variable. Move inner loop
to build_sifted_states.
(build_sifted_states): Extract from sift_states_backward. Do not
use empty_set.
(update_cur_sifted_state): Do not use empty_set. Special case
dest_nodes->nelem == 0.
2004-02-24 Arnold D. Robbins <arnold@skeeve.com>
* posix/regex_internal.c (build_wcs_upper_buffer): Enclose
`offsets_needed' label in `#ifdef _LIBC' to silence `unused label'
compiler warning.
2004-02-24 Nelson H.F. Beebe <beebe@math.utah.edu>
* posix/regex_internal.c (build_wcs_buffer): Add cast to char* in call
to `wcrtomb'.
* posix/regex_internal.h (bitset_not, bitset_merge, bitset_not_merge,
bitset_mask, re_string_char_size_a, re_string_wchar_at,
re_string_elem_size_at): Change to use prototypes.
(re_string_char_size_at, re_string_wchar_at,
re_string_elem_size_at): Declare as `internal_function'.
2004-01-05 Jakub Jelinek <jakub@redhat.com>
* posix/regcomp.c (regcomp): Fix comment typo.
(regfree): Free preg->translate, clear buffer, allocated, fastmap
and translate fields.
* posix/regcomp.c (build_charclass, buld_charclass_op): Change first
argument to unsigned RE_TRANSLATE_TYPE.
* posix/regex_internal.h (re_string_t): Change trans type to
unsigned RE_TRANSLATE_TYPE.
* posix/regex_internal.c (re_string_construct_common): Cast
trans to unsigned RE_TRANSLATE_TYPE.
(re_string_peek_byte_case, re_string_fetch_byte_case): Avoid fast
path if pstr->trans. Never translate the character through
pstr->trans.
* posix/Makefile (tests): Add bug-regex22.
(bug-regex22-ENV): Set.
* posix/bug-regex22.c: New test.
