[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[Emacs-diffs] master d904cc8 2/6: Use Gnulib regex for lib-src
|
From: |
Paul Eggert |
|
Subject: |
[Emacs-diffs] master d904cc8 2/6: Use Gnulib regex for lib-src |
|
Date: |
Sun, 5 Aug 2018 22:36:18 -0400 (EDT) |
branch: master
commit d904cc83f3036db96107a3976cee1a0112547de6
Author: Paul Eggert <address@hidden>
Commit: Paul Eggert <address@hidden>
Use Gnulib regex for lib-src
Emacs regular expressions forked from everyone else long ago.
This makes it official and should allow simplification later.
etags.c now uses the glibc regex API, falling back on a
Gnulib-supplied substitute lib/regex.c if necessary.
Emacs proper now uses its own regular expression module.
Although this patch may look dauntingly large, most of it
was generated automatically by admin/merge-gnulib
and contains an exact copy of the glibc regex source,
and the by-hand changes do not grow the Emacs source code.
* admin/merge-gnulib (GNULIB_MODULES): Add regex.
(AVOIDED_MODULES): Add btowc, langinfo, lock, mbrtowc, mbsinit,
nl_langinfo, wchar, wcrtomb, wctype-h.
* lib-src/Makefile.in (regex-emacs.o): Remove; Gnulib does it now.
(etags_deps, etags_libs): Remove regex-emacs.o.
* lib-src/etags.c: Go back to including regex.h.
(add_regex): Use unsigned char translation array,
since glibc regex requires that.
* lib/Makefile.in (not_emacs_OBJECTS, for_emacs_OBJECTS): New macros.
(libegnu_a_OBJECTS): Use them, to avoid building e-regex.o.
* lib/gnulib.mk.in, m4/gnulib-comp.m4: Regenerate.
* lib/regcomp.c, lib/regex.c, lib/regex.h, lib/regex_internal.c:
* lib/regex_internal.h, lib/regexec.c, m4/builtin-expect.m4:
* m4/eealloc.m4, m4/glibc21.m4, m4/mbstate_t.m4, m4/regex.m4:
New files, copied from Gnulib.
* src/regex-emacs.h, src/conf_post.h:
(RE_TRANSLATE_TYPE, RE_TRANSLATE, RE_TRANSLATE_P):
Move from src/conf_post.h to src/regex-emacs.h,
so that they don’t interfere with compiling lib/regex.c.
---
admin/merge-gnulib | 7 +-
etc/NEWS | 7 +
lib-src/Makefile.in | 8 +-
lib-src/etags.c | 4 +-
lib/Makefile.in | 8 +-
lib/gnulib.mk.in | 23 +
lib/regcomp.c | 3944 +++++++++++++++++++++++++++++++++++++++++++++
lib/regex.c | 81 +
lib/regex.h | 658 ++++++++
lib/regex_internal.c | 1740 ++++++++++++++++++++
lib/regex_internal.h | 911 +++++++++++
lib/regexec.c | 4324 ++++++++++++++++++++++++++++++++++++++++++++++++++
m4/builtin-expect.m4 | 49 +
m4/eealloc.m4 | 31 +
m4/glibc21.m4 | 34 +
m4/gnulib-comp.m4 | 30 +
m4/mbstate_t.m4 | 41 +
m4/regex.m4 | 300 ++++
src/conf_post.h | 7 -
src/regex-emacs.h | 7 +-
20 files changed, 12194 insertions(+), 20 deletions(-)
diff --git a/admin/merge-gnulib b/admin/merge-gnulib
index 1397ecf..abb1929 100755
--- a/admin/merge-gnulib
+++ b/admin/merge-gnulib
@@ -37,7 +37,7 @@ GNULIB_MODULES='
getloadavg getopt-gnu gettime gettimeofday gitlog-to-changelog
ieee754-h ignore-value intprops largefile lstat
manywarnings memrchr minmax mkostemp mktime nstrftime
- pipe2 pselect pthread_sigmask putenv qcopy-acl readlink readlinkat
+ pipe2 pselect pthread_sigmask putenv qcopy-acl readlink readlinkat regex
sig2str socklen stat-time std-gnu11 stdalign stddef stdio
stpcpy strtoimax symlink sys_stat sys_time
tempname time time_r time_rz timegm timer-time timespec-add timespec-sub
@@ -46,11 +46,12 @@ GNULIB_MODULES='
'
AVOIDED_MODULES='
- close dup fchdir fstat
- malloc-posix msvc-inval msvc-nothrow
+ btowc close dup fchdir fstat langinfo lock
+ malloc-posix mbrtowc mbsinit msvc-inval msvc-nothrow nl_langinfo
openat-die opendir raise
save-cwd select setenv sigprocmask stat stdarg stdbool
threadlib tzset unsetenv utime utime-h
+ wchar wcrtomb wctype-h
'
GNULIB_TOOL_FLAGS='
diff --git a/etc/NEWS b/etc/NEWS
index fa8a7af..21887f5 100644
--- a/etc/NEWS
+++ b/etc/NEWS
@@ -31,6 +31,13 @@ functions 'json-serialize', 'json-insert',
'json-parse-string', and
'json-parse-buffer' are typically much faster than their Lisp
counterparts from json.el.
+** The etags program now uses the C library's regular expression matcher
+when possible, and a compatible regex substitute otherwise. This will
+let developers maintain Emacs's own regex code without having to also
+support other programs. The new configure option '--without-included-regex'
+forces etags to use the C library's regex matcher even if the regex
+substitute ordinarily would be used to work around compatibility problems.
+
** Emacs has been ported to the -fcheck-pointer-bounds option of GCC.
This causes Emacs to check bounds of some arrays addressed by its
internal pointers, which can be helpful when debugging the Emacs
diff --git a/lib-src/Makefile.in b/lib-src/Makefile.in
index e70b23c..b2b9017 100644
--- a/lib-src/Makefile.in
+++ b/lib-src/Makefile.in
@@ -361,13 +361,9 @@ TAGS: etags${EXEEXT} ${tagsfiles}
../lib/libgnu.a: $(config_h)
$(MAKE) -C ../lib all
-regex-emacs.o: $(srcdir)/../src/regex-emacs.c $(srcdir)/../src/regex-emacs.h
$(config_h)
- $(AM_V_CC)$(CC) -c $(CPP_CFLAGS) $<
-
-
-etags_deps = ${srcdir}/etags.c regex-emacs.o $(NTLIB) $(config_h)
+etags_deps = ${srcdir}/etags.c $(NTLIB) $(config_h)
etags_cflags = -DEMACS_NAME="\"GNU Emacs\"" -DVERSION="\"${version}\"" -o $@
-etags_libs = regex-emacs.o $(NTLIB) $(LOADLIBES)
+etags_libs = $(NTLIB) $(LOADLIBES)
etags${EXEEXT}: ${etags_deps}
$(AM_V_CCLD)$(CC) ${ALL_CFLAGS} $(etags_cflags) $< $(etags_libs)
diff --git a/lib-src/etags.c b/lib-src/etags.c
index 47d1311..ee50670 100644
--- a/lib-src/etags.c
+++ b/lib-src/etags.c
@@ -135,7 +135,7 @@ char pot_etags_version[] = "@(#) pot revision number is
17.38.1.4";
#endif
#include <getopt.h>
-#include <regex-emacs.h>
+#include <regex.h>
/* Define CTAGS to make the program "ctags" compatible with the usual one.
Leave it undefined to make the program "etags", which makes emacs-style
@@ -6401,7 +6401,7 @@ add_regex (char *regexp_pattern, language *lang)
*patbuf = zeropattern;
if (ignore_case)
{
- static char lc_trans[UCHAR_MAX + 1];
+ static unsigned char lc_trans[UCHAR_MAX + 1];
int i;
for (i = 0; i < UCHAR_MAX + 1; i++)
lc_trans[i] = c_tolower (i);
diff --git a/lib/Makefile.in b/lib/Makefile.in
index 201f4b5..b26db27 100644
--- a/lib/Makefile.in
+++ b/lib/Makefile.in
@@ -79,9 +79,15 @@ endif
Makefile: ../config.status $(srcdir)/Makefile.in
$(MAKE) -C .. src/$@
+# Object modules that need not be built for Emacs.
+# Emacs does not need e-regex.o (it has its own regex-emacs.c),
+# and building it would just waste time.
+not_emacs_OBJECTS = regex.o
+
libgnu_a_OBJECTS = $(gl_LIBOBJS) \
$(patsubst %.c,%.o,$(filter %.c,$(libgnu_a_SOURCES)))
-libegnu_a_OBJECTS = $(patsubst %.o,e-%.o,$(libgnu_a_OBJECTS))
+for_emacs_OBJECTS = $(filter-out $(not_emacs_OBJECTS),$(libgnu_a_OBJECTS))
+libegnu_a_OBJECTS = $(patsubst %.o,e-%.o,$(for_emacs_OBJECTS))
$(libegnu_a_OBJECTS) $(libgnu_a_OBJECTS): $(BUILT_SOURCES)
diff --git a/lib/gnulib.mk.in b/lib/gnulib.mk.in
index 7d28dcc..7ad3908 100644
--- a/lib/gnulib.mk.in
+++ b/lib/gnulib.mk.in
@@ -34,13 +34,19 @@
# --no-libtool \
# --macro-prefix=gl \
# --no-vc-files \
+# --avoid=btowc \
# --avoid=close \
# --avoid=dup \
# --avoid=fchdir \
# --avoid=fstat \
+# --avoid=langinfo \
+# --avoid=lock \
# --avoid=malloc-posix \
+# --avoid=mbrtowc \
+# --avoid=mbsinit \
# --avoid=msvc-inval \
# --avoid=msvc-nothrow \
+# --avoid=nl_langinfo \
# --avoid=openat-die \
# --avoid=opendir \
# --avoid=raise \
@@ -56,6 +62,9 @@
# --avoid=unsetenv \
# --avoid=utime \
# --avoid=utime-h \
+# --avoid=wchar \
+# --avoid=wcrtomb \
+# --avoid=wctype-h \
# alloca-opt \
# binary-io \
# byteswap \
@@ -113,6 +122,7 @@
# qcopy-acl \
# readlink \
# readlinkat \
+# regex \
# sig2str \
# socklen \
# stat-time \
@@ -216,6 +226,7 @@ GETOPT_CDEFS_H = @GETOPT_CDEFS_H@
GETOPT_H = @GETOPT_H@
GFILENOTIFY_CFLAGS = @GFILENOTIFY_CFLAGS@
GFILENOTIFY_LIBS = @GFILENOTIFY_LIBS@
+GLIBC21 = @GLIBC21@
GL_COND_LIBTOOL = @GL_COND_LIBTOOL@
GL_GENERATE_ALLOCA_H = @GL_GENERATE_ALLOCA_H@
GL_GENERATE_BYTESWAP_H = @GL_GENERATE_BYTESWAP_H@
@@ -1024,6 +1035,7 @@ gameuser = @gameuser@
gl_GNULIB_ENABLED_03e0aaad4cb89ca757653bd367a6ccb7 =
@gl_GNULIB_ENABLED_03e0aaad4cb89ca757653bd367a6ccb7@
gl_GNULIB_ENABLED_2049e887c7e5308faad27b3f894bb8c9 =
@gl_GNULIB_ENABLED_2049e887c7e5308faad27b3f894bb8c9@
gl_GNULIB_ENABLED_260941c0e5dc67ec9e87d1fb321c300b =
@gl_GNULIB_ENABLED_260941c0e5dc67ec9e87d1fb321c300b@
+gl_GNULIB_ENABLED_37f71b604aa9c54446783d80f42fe547 =
@gl_GNULIB_ENABLED_37f71b604aa9c54446783d80f42fe547@
gl_GNULIB_ENABLED_5264294aa0a5557541b53c8c741f7f31 =
@gl_GNULIB_ENABLED_5264294aa0a5557541b53c8c741f7f31@
gl_GNULIB_ENABLED_6099e9737f757db36c47fa9d9f02e88c =
@gl_GNULIB_ENABLED_6099e9737f757db36c47fa9d9f02e88c@
gl_GNULIB_ENABLED_682e609604ccaac6be382e4ee3a4eaec =
@gl_GNULIB_ENABLED_682e609604ccaac6be382e4ee3a4eaec@
@@ -2095,6 +2107,17 @@ EXTRA_libgnu_a_SOURCES += at-func.c readlinkat.c
endif
## end gnulib module readlinkat
+## begin gnulib module regex
+ifeq (,$(OMIT_GNULIB_MODULE_regex))
+
+
+EXTRA_DIST += regcomp.c regex.c regex.h regex_internal.c regex_internal.h
regexec.c
+
+EXTRA_libgnu_a_SOURCES += regcomp.c regex.c regex_internal.c regexec.c
+
+endif
+## end gnulib module regex
+
## begin gnulib module root-uid
ifeq (,$(OMIT_GNULIB_MODULE_root-uid))
diff --git a/lib/regcomp.c b/lib/regcomp.c
new file mode 100644
index 0000000..53eb226
--- /dev/null
+++ b/lib/regcomp.c
@@ -0,0 +1,3944 @@
+/* Extended regular expression matching and search library.
+ Copyright (C) 2002-2018 Free Software Foundation, Inc.
+ This file is part of the GNU C Library.
+ Contributed by Isamu Hasegawa <address@hidden>.
+
+ The GNU C Library 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 3 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
+ General Public License for more details.
+
+ You should have received a copy of the GNU General Public
+ License along with the GNU C Library; if not, see
+ <https://www.gnu.org/licenses/>. */
+
+#ifdef _LIBC
+# include <locale/weight.h>
+#endif
+
+static reg_errcode_t re_compile_internal (regex_t *preg, const char * pattern,
+ size_t length, reg_syntax_t syntax);
+static void re_compile_fastmap_iter (regex_t *bufp,
+ const re_dfastate_t *init_state,
+ char *fastmap);
+static reg_errcode_t init_dfa (re_dfa_t *dfa, size_t pat_len);
+#ifdef RE_ENABLE_I18N
+static void free_charset (re_charset_t *cset);
+#endif /* RE_ENABLE_I18N */
+static void free_workarea_compile (regex_t *preg);
+static reg_errcode_t create_initial_state (re_dfa_t *dfa);
+#ifdef RE_ENABLE_I18N
+static void optimize_utf8 (re_dfa_t *dfa);
+#endif
+static reg_errcode_t analyze (regex_t *preg);
+static reg_errcode_t preorder (bin_tree_t *root,
+ reg_errcode_t (fn (void *, bin_tree_t *)),
+ void *extra);
+static reg_errcode_t postorder (bin_tree_t *root,
+ reg_errcode_t (fn (void *, bin_tree_t *)),
+ void *extra);
+static reg_errcode_t optimize_subexps (void *extra, bin_tree_t *node);
+static reg_errcode_t lower_subexps (void *extra, bin_tree_t *node);
+static bin_tree_t *lower_subexp (reg_errcode_t *err, regex_t *preg,
+ bin_tree_t *node);
+static reg_errcode_t calc_first (void *extra, bin_tree_t *node);
+static reg_errcode_t calc_next (void *extra, bin_tree_t *node);
+static reg_errcode_t link_nfa_nodes (void *extra, bin_tree_t *node);
+static Idx duplicate_node (re_dfa_t *dfa, Idx org_idx, unsigned int
constraint);
+static Idx search_duplicated_node (const re_dfa_t *dfa, Idx org_node,
+ unsigned int constraint);
+static reg_errcode_t calc_eclosure (re_dfa_t *dfa);
+static reg_errcode_t calc_eclosure_iter (re_node_set *new_set, re_dfa_t *dfa,
+ Idx node, bool root);
+static reg_errcode_t calc_inveclosure (re_dfa_t *dfa);
+static Idx fetch_number (re_string_t *input, re_token_t *token,
+ reg_syntax_t syntax);
+static int peek_token (re_token_t *token, re_string_t *input,
+ reg_syntax_t syntax);
+static bin_tree_t *parse (re_string_t *regexp, regex_t *preg,
+ reg_syntax_t syntax, reg_errcode_t *err);
+static bin_tree_t *parse_reg_exp (re_string_t *regexp, regex_t *preg,
+ re_token_t *token, reg_syntax_t syntax,
+ Idx nest, reg_errcode_t *err);
+static bin_tree_t *parse_branch (re_string_t *regexp, regex_t *preg,
+ re_token_t *token, reg_syntax_t syntax,
+ Idx nest, reg_errcode_t *err);
+static bin_tree_t *parse_expression (re_string_t *regexp, regex_t *preg,
+ re_token_t *token, reg_syntax_t syntax,
+ Idx nest, reg_errcode_t *err);
+static bin_tree_t *parse_sub_exp (re_string_t *regexp, regex_t *preg,
+ re_token_t *token, reg_syntax_t syntax,
+ Idx nest, reg_errcode_t *err);
+static bin_tree_t *parse_dup_op (bin_tree_t *dup_elem, re_string_t *regexp,
+ re_dfa_t *dfa, re_token_t *token,
+ reg_syntax_t syntax, reg_errcode_t *err);
+static bin_tree_t *parse_bracket_exp (re_string_t *regexp, re_dfa_t *dfa,
+ re_token_t *token, reg_syntax_t syntax,
+ reg_errcode_t *err);
+static reg_errcode_t parse_bracket_element (bracket_elem_t *elem,
+ re_string_t *regexp,
+ re_token_t *token, int token_len,
+ re_dfa_t *dfa,
+ reg_syntax_t syntax,
+ bool accept_hyphen);
+static reg_errcode_t parse_bracket_symbol (bracket_elem_t *elem,
+ re_string_t *regexp,
+ re_token_t *token);
+#ifdef RE_ENABLE_I18N
+static reg_errcode_t build_equiv_class (bitset_t sbcset,
+ re_charset_t *mbcset,
+ Idx *equiv_class_alloc,
+ const unsigned char *name);
+static reg_errcode_t build_charclass (RE_TRANSLATE_TYPE trans,
+ bitset_t sbcset,
+ re_charset_t *mbcset,
+ Idx *char_class_alloc,
+ const char *class_name,
+ reg_syntax_t syntax);
+#else /* not RE_ENABLE_I18N */
+static reg_errcode_t build_equiv_class (bitset_t sbcset,
+ const unsigned char *name);
+static reg_errcode_t build_charclass (RE_TRANSLATE_TYPE trans,
+ bitset_t sbcset,
+ const char *class_name,
+ reg_syntax_t syntax);
+#endif /* not RE_ENABLE_I18N */
+static bin_tree_t *build_charclass_op (re_dfa_t *dfa,
+ RE_TRANSLATE_TYPE trans,
+ const char *class_name,
+ const char *extra,
+ bool non_match, reg_errcode_t *err);
+static bin_tree_t *create_tree (re_dfa_t *dfa,
+ bin_tree_t *left, bin_tree_t *right,
+ re_token_type_t type);
+static bin_tree_t *create_token_tree (re_dfa_t *dfa,
+ bin_tree_t *left, bin_tree_t *right,
+ const re_token_t *token);
+static bin_tree_t *duplicate_tree (const bin_tree_t *src, re_dfa_t *dfa);
+static void free_token (re_token_t *node);
+static reg_errcode_t free_tree (void *extra, bin_tree_t *node);
+static reg_errcode_t mark_opt_subexp (void *extra, bin_tree_t *node);
+
+/* This table gives an error message for each of the error codes listed
+ in regex.h. Obviously the order here has to be same as there.
+ POSIX doesn't require that we do anything for REG_NOERROR,
+ but why not be nice? */
+
+static const char __re_error_msgid[] =
+ {
+#define REG_NOERROR_IDX 0
+ gettext_noop ("Success") /* REG_NOERROR */
+ "\0"
+#define REG_NOMATCH_IDX (REG_NOERROR_IDX + sizeof "Success")
+ gettext_noop ("No match") /* REG_NOMATCH */
+ "\0"
+#define REG_BADPAT_IDX (REG_NOMATCH_IDX + sizeof "No match")
+ gettext_noop ("Invalid regular expression") /* REG_BADPAT */
+ "\0"
+#define REG_ECOLLATE_IDX (REG_BADPAT_IDX + sizeof "Invalid regular expression")
+ gettext_noop ("Invalid collation character") /* REG_ECOLLATE */
+ "\0"
+#define REG_ECTYPE_IDX (REG_ECOLLATE_IDX + sizeof "Invalid collation
character")
+ gettext_noop ("Invalid character class name") /* REG_ECTYPE */
+ "\0"
+#define REG_EESCAPE_IDX (REG_ECTYPE_IDX + sizeof "Invalid character
class name")
+ gettext_noop ("Trailing backslash") /* REG_EESCAPE */
+ "\0"
+#define REG_ESUBREG_IDX (REG_EESCAPE_IDX + sizeof "Trailing backslash")
+ gettext_noop ("Invalid back reference") /* REG_ESUBREG */
+ "\0"
+#define REG_EBRACK_IDX (REG_ESUBREG_IDX + sizeof "Invalid back reference")
+ gettext_noop ("Unmatched [, [^, [:, [., or [=") /* REG_EBRACK */
+ "\0"
+#define REG_EPAREN_IDX (REG_EBRACK_IDX + sizeof "Unmatched [, [^, [:, [., or
[=")
+ gettext_noop ("Unmatched ( or \\(") /* REG_EPAREN */
+ "\0"
+#define REG_EBRACE_IDX (REG_EPAREN_IDX + sizeof "Unmatched ( or \\(")
+ gettext_noop ("Unmatched \\{") /* REG_EBRACE */
+ "\0"
+#define REG_BADBR_IDX (REG_EBRACE_IDX + sizeof "Unmatched \\{")
+ gettext_noop ("Invalid content of \\{\\}") /* REG_BADBR */
+ "\0"
+#define REG_ERANGE_IDX (REG_BADBR_IDX + sizeof "Invalid content of \\{\\}")
+ gettext_noop ("Invalid range end") /* REG_ERANGE */
+ "\0"
+#define REG_ESPACE_IDX (REG_ERANGE_IDX + sizeof "Invalid range end")
+ gettext_noop ("Memory exhausted") /* REG_ESPACE */
+ "\0"
+#define REG_BADRPT_IDX (REG_ESPACE_IDX + sizeof "Memory exhausted")
+ gettext_noop ("Invalid preceding regular expression") /* REG_BADRPT */
+ "\0"
+#define REG_EEND_IDX (REG_BADRPT_IDX + sizeof "Invalid preceding regular
expression")
+ gettext_noop ("Premature end of regular expression") /* REG_EEND */
+ "\0"
+#define REG_ESIZE_IDX (REG_EEND_IDX + sizeof "Premature end of regular
expression")
+ gettext_noop ("Regular expression too big") /* REG_ESIZE */
+ "\0"
+#define REG_ERPAREN_IDX (REG_ESIZE_IDX + sizeof "Regular expression too
big")
+ gettext_noop ("Unmatched ) or \\)") /* REG_ERPAREN */
+ };
+
+static const size_t __re_error_msgid_idx[] =
+ {
+ REG_NOERROR_IDX,
+ REG_NOMATCH_IDX,
+ REG_BADPAT_IDX,
+ REG_ECOLLATE_IDX,
+ REG_ECTYPE_IDX,
+ REG_EESCAPE_IDX,
+ REG_ESUBREG_IDX,
+ REG_EBRACK_IDX,
+ REG_EPAREN_IDX,
+ REG_EBRACE_IDX,
+ REG_BADBR_IDX,
+ REG_ERANGE_IDX,
+ REG_ESPACE_IDX,
+ REG_BADRPT_IDX,
+ REG_EEND_IDX,
+ REG_ESIZE_IDX,
+ REG_ERPAREN_IDX
+ };
+
+/* Entry points for GNU code. */
+
+/* re_compile_pattern is the GNU regular expression compiler: it
+ compiles PATTERN (of length LENGTH) and puts the result in BUFP.
+ Returns 0 if the pattern was valid, otherwise an error string.
+
+ Assumes the 'allocated' (and perhaps 'buffer') and 'translate' fields
+ are set in BUFP on entry. */
+
+const char *
+re_compile_pattern (const char *pattern, size_t length,
+ struct re_pattern_buffer *bufp)
+{
+ reg_errcode_t ret;
+
+ /* And GNU code determines whether or not to get register information
+ by passing null for the REGS argument to re_match, etc., not by
+ setting no_sub, unless RE_NO_SUB is set. */
+ bufp->no_sub = !!(re_syntax_options & RE_NO_SUB);
+
+ /* Match anchors at newline. */
+ bufp->newline_anchor = 1;
+
+ ret = re_compile_internal (bufp, pattern, length, re_syntax_options);
+
+ if (!ret)
+ return NULL;
+ return gettext (__re_error_msgid + __re_error_msgid_idx[(int) ret]);
+}
+#ifdef _LIBC
+weak_alias (__re_compile_pattern, re_compile_pattern)
+#endif
+
+/* Set by 're_set_syntax' to the current regexp syntax to recognize. Can
+ also be assigned to arbitrarily: each pattern buffer stores its own
+ syntax, so it can be changed between regex compilations. */
+/* This has no initializer because initialized variables in Emacs
+ become read-only after dumping. */
+reg_syntax_t re_syntax_options;
+
+
+/* Specify the precise syntax of regexps for compilation. This provides
+ for compatibility for various utilities which historically have
+ different, incompatible syntaxes.
+
+ The argument SYNTAX is a bit mask comprised of the various bits
+ defined in regex.h. We return the old syntax. */
+
+reg_syntax_t
+re_set_syntax (reg_syntax_t syntax)
+{
+ reg_syntax_t ret = re_syntax_options;
+
+ re_syntax_options = syntax;
+ return ret;
+}
+#ifdef _LIBC
+weak_alias (__re_set_syntax, re_set_syntax)
+#endif
+
+int
+re_compile_fastmap (struct re_pattern_buffer *bufp)
+{
+ re_dfa_t *dfa = bufp->buffer;
+ char *fastmap = bufp->fastmap;
+
+ memset (fastmap, '\0', sizeof (char) * SBC_MAX);
+ re_compile_fastmap_iter (bufp, dfa->init_state, fastmap);
+ if (dfa->init_state != dfa->init_state_word)
+ re_compile_fastmap_iter (bufp, dfa->init_state_word, fastmap);
+ if (dfa->init_state != dfa->init_state_nl)
+ re_compile_fastmap_iter (bufp, dfa->init_state_nl, fastmap);
+ if (dfa->init_state != dfa->init_state_begbuf)
+ re_compile_fastmap_iter (bufp, dfa->init_state_begbuf, fastmap);
+ bufp->fastmap_accurate = 1;
+ return 0;
+}
+#ifdef _LIBC
+weak_alias (__re_compile_fastmap, re_compile_fastmap)
+#endif
+
+static inline void
+__attribute__ ((always_inline))
+re_set_fastmap (char *fastmap, bool icase, int ch)
+{
+ fastmap[ch] = 1;
+ if (icase)
+ fastmap[tolower (ch)] = 1;
+}
+
+/* Helper function for re_compile_fastmap.
+ Compile fastmap for the initial_state INIT_STATE. */
+
+static void
+re_compile_fastmap_iter (regex_t *bufp, const re_dfastate_t *init_state,
+ char *fastmap)
+{
+ re_dfa_t *dfa = bufp->buffer;
+ Idx node_cnt;
+ bool icase = (dfa->mb_cur_max == 1 && (bufp->syntax & RE_ICASE));
+ for (node_cnt = 0; node_cnt < init_state->nodes.nelem; ++node_cnt)
+ {
+ Idx node = init_state->nodes.elems[node_cnt];
+ re_token_type_t type = dfa->nodes[node].type;
+
+ if (type == CHARACTER)
+ {
+ re_set_fastmap (fastmap, icase, dfa->nodes[node].opr.c);
+#ifdef RE_ENABLE_I18N
+ if ((bufp->syntax & RE_ICASE) && dfa->mb_cur_max > 1)
+ {
+ unsigned char buf[MB_LEN_MAX];
+ unsigned char *p;
+ wchar_t wc;
+ mbstate_t state;
+
+ p = buf;
+ *p++ = dfa->nodes[node].opr.c;
+ while (++node < dfa->nodes_len
+ && dfa->nodes[node].type == CHARACTER
+ && dfa->nodes[node].mb_partial)
+ *p++ = dfa->nodes[node].opr.c;
+ memset (&state, '\0', sizeof (state));
+ if (__mbrtowc (&wc, (const char *) buf, p - buf,
+ &state) == p - buf
+ && (__wcrtomb ((char *) buf, __towlower (wc), &state)
+ != (size_t) -1))
+ re_set_fastmap (fastmap, false, buf[0]);
+ }
+#endif
+ }
+ else if (type == SIMPLE_BRACKET)
+ {
+ int i, ch;
+ for (i = 0, ch = 0; i < BITSET_WORDS; ++i)
+ {
+ int j;
+ bitset_word_t w = dfa->nodes[node].opr.sbcset[i];
+ for (j = 0; j < BITSET_WORD_BITS; ++j, ++ch)
+ if (w & ((bitset_word_t) 1 << j))
+ re_set_fastmap (fastmap, icase, ch);
+ }
+ }
+#ifdef RE_ENABLE_I18N
+ else if (type == COMPLEX_BRACKET)
+ {
+ re_charset_t *cset = dfa->nodes[node].opr.mbcset;
+ Idx i;
+
+# ifdef _LIBC
+ /* See if we have to try all bytes which start multiple collation
+ elements.
+ e.g. In da_DK, we want to catch 'a' since "aa" is a valid
+ collation element, and don't catch 'b' since 'b' is
+ the only collation element which starts from 'b' (and
+ it is caught by SIMPLE_BRACKET). */
+ if (_NL_CURRENT_WORD (LC_COLLATE, _NL_COLLATE_NRULES) != 0
+ && (cset->ncoll_syms || cset->nranges))
+ {
+ const int32_t *table = (const int32_t *)
+ _NL_CURRENT (LC_COLLATE, _NL_COLLATE_TABLEMB);
+ for (i = 0; i < SBC_MAX; ++i)
+ if (table[i] < 0)
+ re_set_fastmap (fastmap, icase, i);
+ }
+# endif /* _LIBC */
+
+ /* See if we have to start the match at all multibyte characters,
+ i.e. where we would not find an invalid sequence. This only
+ applies to multibyte character sets; for single byte character
+ sets, the SIMPLE_BRACKET again suffices. */
+ if (dfa->mb_cur_max > 1
+ && (cset->nchar_classes || cset->non_match || cset->nranges
+# ifdef _LIBC
+ || cset->nequiv_classes
+# endif /* _LIBC */
+ ))
+ {
+ unsigned char c = 0;
+ do
+ {
+ mbstate_t mbs;
+ memset (&mbs, 0, sizeof (mbs));
+ if (__mbrtowc (NULL, (char *) &c, 1, &mbs) == (size_t) -2)
+ re_set_fastmap (fastmap, false, (int) c);
+ }
+ while (++c != 0);
+ }
+
+ else
+ {
+ /* ... Else catch all bytes which can start the mbchars. */
+ for (i = 0; i < cset->nmbchars; ++i)
+ {
+ char buf[256];
+ mbstate_t state;
+ memset (&state, '\0', sizeof (state));
+ if (__wcrtomb (buf, cset->mbchars[i], &state) != (size_t) -1)
+ re_set_fastmap (fastmap, icase, *(unsigned char *) buf);
+ if ((bufp->syntax & RE_ICASE) && dfa->mb_cur_max > 1)
+ {
+ if (__wcrtomb (buf, __towlower (cset->mbchars[i]), &state)
+ != (size_t) -1)
+ re_set_fastmap (fastmap, false, *(unsigned char *) buf);
+ }
+ }
+ }
+ }
+#endif /* RE_ENABLE_I18N */
+ else if (type == OP_PERIOD
+#ifdef RE_ENABLE_I18N
+ || type == OP_UTF8_PERIOD
+#endif /* RE_ENABLE_I18N */
+ || type == END_OF_RE)
+ {
+ memset (fastmap, '\1', sizeof (char) * SBC_MAX);
+ if (type == END_OF_RE)
+ bufp->can_be_null = 1;
+ return;
+ }
+ }
+}
+
+/* Entry point for POSIX code. */
+/* regcomp takes a regular expression as a string and compiles it.
+
+ PREG is a regex_t *. We do not expect any fields to be initialized,
+ since POSIX says we shouldn't. Thus, we set
+
+ 'buffer' to the compiled pattern;
+ 'used' to the length of the compiled pattern;
+ 'syntax' to RE_SYNTAX_POSIX_EXTENDED if the
+ REG_EXTENDED bit in CFLAGS is set; otherwise, to
+ RE_SYNTAX_POSIX_BASIC;
+ 'newline_anchor' to REG_NEWLINE being set in CFLAGS;
+ 'fastmap' to an allocated space for the fastmap;
+ 'fastmap_accurate' to zero;
+ 're_nsub' to the number of subexpressions in PATTERN.
+
+ PATTERN is the address of the pattern string.
+
+ CFLAGS is a series of bits which affect compilation.
+
+ If REG_EXTENDED is set, we use POSIX extended syntax; otherwise, we
+ use POSIX basic syntax.
+
+ If REG_NEWLINE is set, then . and [^...] don't match newline.
+ Also, regexec will try a match beginning after every newline.
+
+ If REG_ICASE is set, then we considers upper- and lowercase
+ versions of letters to be equivalent when matching.
+
+ If REG_NOSUB is set, then when PREG is passed to regexec, that
+ routine will report only success or failure, and nothing about the
+ registers.
+
+ It returns 0 if it succeeds, nonzero if it doesn't. (See regex.h for
+ the return codes and their meanings.) */
+
+int
+regcomp (regex_t *_Restrict_ preg, const char *_Restrict_ pattern, int cflags)
+{
+ reg_errcode_t ret;
+ reg_syntax_t syntax = ((cflags & REG_EXTENDED) ? RE_SYNTAX_POSIX_EXTENDED
+ : RE_SYNTAX_POSIX_BASIC);
+
+ preg->buffer = NULL;
+ preg->allocated = 0;
+ preg->used = 0;
+
+ /* Try to allocate space for the fastmap. */
+ preg->fastmap = re_malloc (char, SBC_MAX);
+ if (BE (preg->fastmap == NULL, 0))
+ return REG_ESPACE;
+
+ syntax |= (cflags & REG_ICASE) ? RE_ICASE : 0;
+
+ /* If REG_NEWLINE is set, newlines are treated differently. */
+ if (cflags & REG_NEWLINE)
+ { /* REG_NEWLINE implies neither . nor [^...] match newline. */
+ syntax &= ~RE_DOT_NEWLINE;
+ syntax |= RE_HAT_LISTS_NOT_NEWLINE;
+ /* It also changes the matching behavior. */
+ preg->newline_anchor = 1;
+ }
+ else
+ preg->newline_anchor = 0;
+ preg->no_sub = !!(cflags & REG_NOSUB);
+ preg->translate = NULL;
+
+ ret = re_compile_internal (preg, pattern, strlen (pattern), syntax);
+
+ /* POSIX doesn't distinguish between an unmatched open-group and an
+ unmatched close-group: both are REG_EPAREN. */
+ if (ret == REG_ERPAREN)
+ ret = REG_EPAREN;
+
+ /* We have already checked preg->fastmap != NULL. */
+ if (BE (ret == REG_NOERROR, 1))
+ /* Compute the fastmap now, since regexec cannot modify the pattern
+ buffer. This function never fails in this implementation. */
+ (void) re_compile_fastmap (preg);
+ else
+ {
+ /* Some error occurred while compiling the expression. */
+ re_free (preg->fastmap);
+ preg->fastmap = NULL;
+ }
+
+ return (int) ret;
+}
+#ifdef _LIBC
+libc_hidden_def (__regcomp)
+weak_alias (__regcomp, regcomp)
+#endif
+
+/* Returns a message corresponding to an error code, ERRCODE, returned
+ from either regcomp or regexec. We don't use PREG here. */
+
+size_t
+regerror (int errcode, const regex_t *_Restrict_ preg, char *_Restrict_ errbuf,
+ size_t errbuf_size)
+{
+ const char *msg;
+ size_t msg_size;
+
+ if (BE (errcode < 0
+ || errcode >= (int) (sizeof (__re_error_msgid_idx)
+ / sizeof (__re_error_msgid_idx[0])), 0))
+ /* Only error codes returned by the rest of the code should be passed
+ to this routine. If we are given anything else, or if other regex
+ code generates an invalid error code, then the program has a bug.
+ Dump core so we can fix it. */
+ abort ();
+
+ msg = gettext (__re_error_msgid + __re_error_msgid_idx[errcode]);
+
+ msg_size = strlen (msg) + 1; /* Includes the null. */
+
+ if (BE (errbuf_size != 0, 1))
+ {
+ size_t cpy_size = msg_size;
+ if (BE (msg_size > errbuf_size, 0))
+ {
+ cpy_size = errbuf_size - 1;
+ errbuf[cpy_size] = '\0';
+ }
+ memcpy (errbuf, msg, cpy_size);
+ }
+
+ return msg_size;
+}
+#ifdef _LIBC
+weak_alias (__regerror, regerror)
+#endif
+
+
+#ifdef RE_ENABLE_I18N
+/* This static array is used for the map to single-byte characters when
+ UTF-8 is used. Otherwise we would allocate memory just to initialize
+ it the same all the time. UTF-8 is the preferred encoding so this is
+ a worthwhile optimization. */
+static const bitset_t utf8_sb_map =
+{
+ /* Set the first 128 bits. */
+# if defined __GNUC__ && !defined __STRICT_ANSI__
+ [0 ... 0x80 / BITSET_WORD_BITS - 1] = BITSET_WORD_MAX
+# else
+# if 4 * BITSET_WORD_BITS < ASCII_CHARS
+# error "bitset_word_t is narrower than 32 bits"
+# elif 3 * BITSET_WORD_BITS < ASCII_CHARS
+ BITSET_WORD_MAX, BITSET_WORD_MAX, BITSET_WORD_MAX,
+# elif 2 * BITSET_WORD_BITS < ASCII_CHARS
+ BITSET_WORD_MAX, BITSET_WORD_MAX,
+# elif 1 * BITSET_WORD_BITS < ASCII_CHARS
+ BITSET_WORD_MAX,
+# endif
+ (BITSET_WORD_MAX
+ >> (SBC_MAX % BITSET_WORD_BITS == 0
+ ? 0
+ : BITSET_WORD_BITS - SBC_MAX % BITSET_WORD_BITS))
+# endif
+};
+#endif
+
+
+static void
+free_dfa_content (re_dfa_t *dfa)
+{
+ Idx i, j;
+
+ if (dfa->nodes)
+ for (i = 0; i < dfa->nodes_len; ++i)
+ free_token (dfa->nodes + i);
+ re_free (dfa->nexts);
+ for (i = 0; i < dfa->nodes_len; ++i)
+ {
+ if (dfa->eclosures != NULL)
+ re_node_set_free (dfa->eclosures + i);
+ if (dfa->inveclosures != NULL)
+ re_node_set_free (dfa->inveclosures + i);
+ if (dfa->edests != NULL)
+ re_node_set_free (dfa->edests + i);
+ }
+ re_free (dfa->edests);
+ re_free (dfa->eclosures);
+ re_free (dfa->inveclosures);
+ re_free (dfa->nodes);
+
+ if (dfa->state_table)
+ for (i = 0; i <= dfa->state_hash_mask; ++i)
+ {
+ struct re_state_table_entry *entry = dfa->state_table + i;
+ for (j = 0; j < entry->num; ++j)
+ {
+ re_dfastate_t *state = entry->array[j];
+ free_state (state);
+ }
+ re_free (entry->array);
+ }
+ re_free (dfa->state_table);
+#ifdef RE_ENABLE_I18N
+ if (dfa->sb_char != utf8_sb_map)
+ re_free (dfa->sb_char);
+#endif
+ re_free (dfa->subexp_map);
+#ifdef DEBUG
+ re_free (dfa->re_str);
+#endif
+
+ re_free (dfa);
+}
+
+
+/* Free dynamically allocated space used by PREG. */
+
+void
+regfree (regex_t *preg)
+{
+ re_dfa_t *dfa = preg->buffer;
+ if (BE (dfa != NULL, 1))
+ {
+ lock_fini (dfa->lock);
+ free_dfa_content (dfa);
+ }
+ preg->buffer = NULL;
+ preg->allocated = 0;
+
+ re_free (preg->fastmap);
+ preg->fastmap = NULL;
+
+ re_free (preg->translate);
+ preg->translate = NULL;
+}
+#ifdef _LIBC
+libc_hidden_def (__regfree)
+weak_alias (__regfree, regfree)
+#endif
+
+/* Entry points compatible with 4.2 BSD regex library. We don't define
+ them unless specifically requested. */
+
+#if defined _REGEX_RE_COMP || defined _LIBC
+
+/* BSD has one and only one pattern buffer. */
+static struct re_pattern_buffer re_comp_buf;
+
+char *
+# ifdef _LIBC
+/* Make these definitions weak in libc, so POSIX programs can redefine
+ these names if they don't use our functions, and still use
+ regcomp/regexec above without link errors. */
+weak_function
+# endif
+re_comp (const char *s)
+{
+ reg_errcode_t ret;
+ char *fastmap;
+
+ if (!s)
+ {
+ if (!re_comp_buf.buffer)
+ return gettext ("No previous regular expression");
+ return 0;
+ }
+
+ if (re_comp_buf.buffer)
+ {
+ fastmap = re_comp_buf.fastmap;
+ re_comp_buf.fastmap = NULL;
+ __regfree (&re_comp_buf);
+ memset (&re_comp_buf, '\0', sizeof (re_comp_buf));
+ re_comp_buf.fastmap = fastmap;
+ }
+
+ if (re_comp_buf.fastmap == NULL)
+ {
+ re_comp_buf.fastmap = re_malloc (char, SBC_MAX);
+ if (re_comp_buf.fastmap == NULL)
+ return (char *) gettext (__re_error_msgid
+ + __re_error_msgid_idx[(int) REG_ESPACE]);
+ }
+
+ /* Since 're_exec' always passes NULL for the 'regs' argument, we
+ don't need to initialize the pattern buffer fields which affect it. */
+
+ /* Match anchors at newlines. */
+ re_comp_buf.newline_anchor = 1;
+
+ ret = re_compile_internal (&re_comp_buf, s, strlen (s), re_syntax_options);
+
+ if (!ret)
+ return NULL;
+
+ /* Yes, we're discarding 'const' here if !HAVE_LIBINTL. */
+ return (char *) gettext (__re_error_msgid + __re_error_msgid_idx[(int) ret]);
+}
+
+#ifdef _LIBC
+libc_freeres_fn (free_mem)
+{
+ __regfree (&re_comp_buf);
+}
+#endif
+
+#endif /* _REGEX_RE_COMP */
+
+/* Internal entry point.
+ Compile the regular expression PATTERN, whose length is LENGTH.
+ SYNTAX indicate regular expression's syntax. */
+
+static reg_errcode_t
+re_compile_internal (regex_t *preg, const char * pattern, size_t length,
+ reg_syntax_t syntax)
+{
+ reg_errcode_t err = REG_NOERROR;
+ re_dfa_t *dfa;
+ re_string_t regexp;
+
+ /* Initialize the pattern buffer. */
+ preg->fastmap_accurate = 0;
+ preg->syntax = syntax;
+ preg->not_bol = preg->not_eol = 0;
+ preg->used = 0;
+ preg->re_nsub = 0;
+ preg->can_be_null = 0;
+ preg->regs_allocated = REGS_UNALLOCATED;
+
+ /* Initialize the dfa. */
+ dfa = preg->buffer;
+ if (BE (preg->allocated < sizeof (re_dfa_t), 0))
+ {
+ /* If zero allocated, but buffer is non-null, try to realloc
+ enough space. This loses if buffer's address is bogus, but
+ that is the user's responsibility. If ->buffer is NULL this
+ is a simple allocation. */
+ dfa = re_realloc (preg->buffer, re_dfa_t, 1);
+ if (dfa == NULL)
+ return REG_ESPACE;
+ preg->allocated = sizeof (re_dfa_t);
+ preg->buffer = dfa;
+ }
+ preg->used = sizeof (re_dfa_t);
+
+ err = init_dfa (dfa, length);
+ if (BE (err == REG_NOERROR && lock_init (dfa->lock) != 0, 0))
+ err = REG_ESPACE;
+ if (BE (err != REG_NOERROR, 0))
+ {
+ free_dfa_content (dfa);
+ preg->buffer = NULL;
+ preg->allocated = 0;
+ return err;
+ }
+#ifdef DEBUG
+ /* Note: length+1 will not overflow since it is checked in init_dfa. */
+ dfa->re_str = re_malloc (char, length + 1);
+ strncpy (dfa->re_str, pattern, length + 1);
+#endif
+
+ err = re_string_construct (®exp, pattern, length, preg->translate,
+ (syntax & RE_ICASE) != 0, dfa);
+ if (BE (err != REG_NOERROR, 0))
+ {
+ re_compile_internal_free_return:
+ free_workarea_compile (preg);
+ re_string_destruct (®exp);
+ lock_fini (dfa->lock);
+ free_dfa_content (dfa);
+ preg->buffer = NULL;
+ preg->allocated = 0;
+ return err;
+ }
+
+ /* Parse the regular expression, and build a structure tree. */
+ preg->re_nsub = 0;
+ dfa->str_tree = parse (®exp, preg, syntax, &err);
+ if (BE (dfa->str_tree == NULL, 0))
+ goto re_compile_internal_free_return;
+
+ /* Analyze the tree and create the nfa. */
+ err = analyze (preg);
+ if (BE (err != REG_NOERROR, 0))
+ goto re_compile_internal_free_return;
+
+#ifdef RE_ENABLE_I18N
+ /* If possible, do searching in single byte encoding to speed things up. */
+ if (dfa->is_utf8 && !(syntax & RE_ICASE) && preg->translate == NULL)
+ optimize_utf8 (dfa);
+#endif
+
+ /* Then create the initial state of the dfa. */
+ err = create_initial_state (dfa);
+
+ /* Release work areas. */
+ free_workarea_compile (preg);
+ re_string_destruct (®exp);
+
+ if (BE (err != REG_NOERROR, 0))
+ {
+ lock_fini (dfa->lock);
+ free_dfa_content (dfa);
+ preg->buffer = NULL;
+ preg->allocated = 0;
+ }
+
+ return err;
+}
+
+/* Initialize DFA. We use the length of the regular expression PAT_LEN
+ as the initial length of some arrays. */
+
+static reg_errcode_t
+init_dfa (re_dfa_t *dfa, size_t pat_len)
+{
+ __re_size_t table_size;
+#ifndef _LIBC
+ const char *codeset_name;
+#endif
+#ifdef RE_ENABLE_I18N
+ size_t max_i18n_object_size = MAX (sizeof (wchar_t), sizeof (wctype_t));
+#else
+ size_t max_i18n_object_size = 0;
+#endif
+ size_t max_object_size =
+ MAX (sizeof (struct re_state_table_entry),
+ MAX (sizeof (re_token_t),
+ MAX (sizeof (re_node_set),
+ MAX (sizeof (regmatch_t),
+ max_i18n_object_size))));
+
+ memset (dfa, '\0', sizeof (re_dfa_t));
+
+ /* Force allocation of str_tree_storage the first time. */
+ dfa->str_tree_storage_idx = BIN_TREE_STORAGE_SIZE;
+
+ /* Avoid overflows. The extra "/ 2" is for the table_size doubling
+ calculation below, and for similar doubling calculations
+ elsewhere. And it's <= rather than <, because some of the
+ doubling calculations add 1 afterwards. */
+ if (BE (MIN (IDX_MAX, SIZE_MAX / max_object_size) / 2 <= pat_len, 0))
+ return REG_ESPACE;
+
+ dfa->nodes_alloc = pat_len + 1;
+ dfa->nodes = re_malloc (re_token_t, dfa->nodes_alloc);
+
+ /* table_size = 2 ^ ceil(log pat_len) */
+ for (table_size = 1; ; table_size <<= 1)
+ if (table_size > pat_len)
+ break;
+
+ dfa->state_table = calloc (sizeof (struct re_state_table_entry), table_size);
+ dfa->state_hash_mask = table_size - 1;
+
+ dfa->mb_cur_max = MB_CUR_MAX;
+#ifdef _LIBC
+ if (dfa->mb_cur_max == 6
+ && strcmp (_NL_CURRENT (LC_CTYPE, _NL_CTYPE_CODESET_NAME), "UTF-8") == 0)
+ dfa->is_utf8 = 1;
+ dfa->map_notascii = (_NL_CURRENT_WORD (LC_CTYPE, _NL_CTYPE_MAP_TO_NONASCII)
+ != 0);
+#else
+ codeset_name = nl_langinfo (CODESET);
+ if ((codeset_name[0] == 'U' || codeset_name[0] == 'u')
+ && (codeset_name[1] == 'T' || codeset_name[1] == 't')
+ && (codeset_name[2] == 'F' || codeset_name[2] == 'f')
+ && strcmp (codeset_name + 3 + (codeset_name[3] == '-'), "8") == 0)
+ dfa->is_utf8 = 1;
+
+ /* We check exhaustively in the loop below if this charset is a
+ superset of ASCII. */
+ dfa->map_notascii = 0;
+#endif
+
+#ifdef RE_ENABLE_I18N
+ if (dfa->mb_cur_max > 1)
+ {
+ if (dfa->is_utf8)
+ dfa->sb_char = (re_bitset_ptr_t) utf8_sb_map;
+ else
+ {
+ int i, j, ch;
+
+ dfa->sb_char = (re_bitset_ptr_t) calloc (sizeof (bitset_t), 1);
+ if (BE (dfa->sb_char == NULL, 0))
+ return REG_ESPACE;
+
+ /* Set the bits corresponding to single byte chars. */
+ for (i = 0, ch = 0; i < BITSET_WORDS; ++i)
+ for (j = 0; j < BITSET_WORD_BITS; ++j, ++ch)
+ {
+ wint_t wch = __btowc (ch);
+ if (wch != WEOF)
+ dfa->sb_char[i] |= (bitset_word_t) 1 << j;
+# ifndef _LIBC
+ if (isascii (ch) && wch != ch)
+ dfa->map_notascii = 1;
+# endif
+ }
+ }
+ }
+#endif
+
+ if (BE (dfa->nodes == NULL || dfa->state_table == NULL, 0))
+ return REG_ESPACE;
+ return REG_NOERROR;
+}
+
+/* Initialize WORD_CHAR table, which indicate which character is
+ "word". In this case "word" means that it is the word construction
+ character used by some operators like "\<", "\>", etc. */
+
+static void
+init_word_char (re_dfa_t *dfa)
+{
+ int i = 0;
+ int j;
+ int ch = 0;
+ dfa->word_ops_used = 1;
+ if (BE (dfa->map_notascii == 0, 1))
+ {
+ /* Avoid uint32_t and uint64_t as some non-GCC platforms lack
+ them, an issue when this code is used in Gnulib. */
+ 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)
+ {
+ /* Pacify gcc -Woverflow on 32-bit platformns. */
+ dfa->word_char[0] = bits1 << 31 << 1 | bits0;
+ dfa->word_char[1] = bits3 << 31 << 1 | bits2;
+ i = 2;
+ }
+ else if (BITSET_WORD_BITS == 32)
+ {
+ dfa->word_char[0] = bits0;
+ dfa->word_char[1] = bits1;
+ dfa->word_char[2] = bits2;
+ dfa->word_char[3] = bits3;
+ i = 4;
+ }
+ else
+ goto general_case;
+ ch = 128;
+
+ if (BE (dfa->is_utf8, 1))
+ {
+ memset (&dfa->word_char[i], '\0', (SBC_MAX - ch) / 8);
+ return;
+ }
+ }
+
+ general_case:
+ for (; i < BITSET_WORDS; ++i)
+ for (j = 0; j < BITSET_WORD_BITS; ++j, ++ch)
+ if (isalnum (ch) || ch == '_')
+ dfa->word_char[i] |= (bitset_word_t) 1 << j;
+}
+
+/* Free the work area which are only used while compiling. */
+
+static void
+free_workarea_compile (regex_t *preg)
+{
+ re_dfa_t *dfa = preg->buffer;
+ bin_tree_storage_t *storage, *next;
+ for (storage = dfa->str_tree_storage; storage; storage = next)
+ {
+ next = storage->next;
+ re_free (storage);
+ }
+ dfa->str_tree_storage = NULL;
+ dfa->str_tree_storage_idx = BIN_TREE_STORAGE_SIZE;
+ dfa->str_tree = NULL;
+ re_free (dfa->org_indices);
+ dfa->org_indices = NULL;
+}
+
+/* Create initial states for all contexts. */
+
+static reg_errcode_t
+create_initial_state (re_dfa_t *dfa)
+{
+ Idx first, i;
+ reg_errcode_t err;
+ re_node_set init_nodes;
+
+ /* Initial states have the epsilon closure of the node which is
+ the first node of the regular expression. */
+ first = dfa->str_tree->first->node_idx;
+ dfa->init_node = first;
+ err = re_node_set_init_copy (&init_nodes, dfa->eclosures + first);
+ if (BE (err != REG_NOERROR, 0))
+ return err;
+
+ /* The back-references which are in initial states can epsilon transit,
+ since in this case all of the subexpressions can be null.
+ Then we add epsilon closures of the nodes which are the next nodes of
+ the back-references. */
+ if (dfa->nbackref > 0)
+ for (i = 0; i < init_nodes.nelem; ++i)
+ {
+ Idx node_idx = init_nodes.elems[i];
+ re_token_type_t type = dfa->nodes[node_idx].type;
+
+ Idx clexp_idx;
+ if (type != OP_BACK_REF)
+ continue;
+ for (clexp_idx = 0; clexp_idx < init_nodes.nelem; ++clexp_idx)
+ {
+ re_token_t *clexp_node;
+ clexp_node = dfa->nodes + init_nodes.elems[clexp_idx];
+ if (clexp_node->type == OP_CLOSE_SUBEXP
+ && clexp_node->opr.idx == dfa->nodes[node_idx].opr.idx)
+ break;
+ }
+ if (clexp_idx == init_nodes.nelem)
+ continue;
+
+ if (type == OP_BACK_REF)
+ {
+ Idx dest_idx = dfa->edests[node_idx].elems[0];
+ if (!re_node_set_contains (&init_nodes, dest_idx))
+ {
+ reg_errcode_t merge_err
+ = re_node_set_merge (&init_nodes, dfa->eclosures + dest_idx);
+ if (merge_err != REG_NOERROR)
+ return merge_err;
+ i = 0;
+ }
+ }
+ }
+
+ /* It must be the first time to invoke acquire_state. */
+ dfa->init_state = re_acquire_state_context (&err, dfa, &init_nodes, 0);
+ /* We don't check ERR here, since the initial state must not be NULL. */
+ if (BE (dfa->init_state == NULL, 0))
+ return err;
+ if (dfa->init_state->has_constraint)
+ {
+ dfa->init_state_word = re_acquire_state_context (&err, dfa, &init_nodes,
+ CONTEXT_WORD);
+ dfa->init_state_nl = re_acquire_state_context (&err, dfa, &init_nodes,
+ CONTEXT_NEWLINE);
+ dfa->init_state_begbuf = re_acquire_state_context (&err, dfa,
+ &init_nodes,
+ CONTEXT_NEWLINE
+ | CONTEXT_BEGBUF);
+ if (BE (dfa->init_state_word == NULL || dfa->init_state_nl == NULL
+ || dfa->init_state_begbuf == NULL, 0))
+ return err;
+ }
+ else
+ dfa->init_state_word = dfa->init_state_nl
+ = dfa->init_state_begbuf = dfa->init_state;
+
+ re_node_set_free (&init_nodes);
+ return REG_NOERROR;
+}
+
+#ifdef RE_ENABLE_I18N
+/* If it is possible to do searching in single byte encoding instead of UTF-8
+ to speed things up, set dfa->mb_cur_max to 1, clear is_utf8 and change
+ DFA nodes where needed. */
+
+static void
+optimize_utf8 (re_dfa_t *dfa)
+{
+ Idx node;
+ int i;
+ bool mb_chars = false;
+ bool has_period = false;
+
+ for (node = 0; node < dfa->nodes_len; ++node)
+ switch (dfa->nodes[node].type)
+ {
+ case CHARACTER:
+ if (dfa->nodes[node].opr.c >= ASCII_CHARS)
+ mb_chars = true;
+ break;
+ case ANCHOR:
+ switch (dfa->nodes[node].opr.ctx_type)
+ {
+ case LINE_FIRST:
+ case LINE_LAST:
+ case BUF_FIRST:
+ case BUF_LAST:
+ break;
+ default:
+ /* Word anchors etc. cannot be handled. It's okay to test
+ opr.ctx_type since constraints (for all DFA nodes) are
+ created by ORing one or more opr.ctx_type values. */
+ return;
+ }
+ break;
+ case OP_PERIOD:
+ has_period = true;
+ break;
+ case OP_BACK_REF:
+ case OP_ALT:
+ case END_OF_RE:
+ case OP_DUP_ASTERISK:
+ case OP_OPEN_SUBEXP:
+ case OP_CLOSE_SUBEXP:
+ break;
+ case COMPLEX_BRACKET:
+ return;
+ case SIMPLE_BRACKET:
+ /* Just double check. */
+ {
+ int rshift = (ASCII_CHARS % BITSET_WORD_BITS == 0
+ ? 0
+ : BITSET_WORD_BITS - ASCII_CHARS % BITSET_WORD_BITS);
+ for (i = ASCII_CHARS / BITSET_WORD_BITS; i < BITSET_WORDS; ++i)
+ {
+ if (dfa->nodes[node].opr.sbcset[i] >> rshift != 0)
+ return;
+ rshift = 0;
+ }
+ }
+ break;
+ default:
+ abort ();
+ }
+
+ if (mb_chars || has_period)
+ for (node = 0; node < dfa->nodes_len; ++node)
+ {
+ if (dfa->nodes[node].type == CHARACTER
+ && dfa->nodes[node].opr.c >= ASCII_CHARS)
+ dfa->nodes[node].mb_partial = 0;
+ else if (dfa->nodes[node].type == OP_PERIOD)
+ dfa->nodes[node].type = OP_UTF8_PERIOD;
+ }
+
+ /* The search can be in single byte locale. */
+ dfa->mb_cur_max = 1;
+ dfa->is_utf8 = 0;
+ dfa->has_mb_node = dfa->nbackref > 0 || has_period;
+}
+#endif
+
+/* Analyze the structure tree, and calculate "first", "next", "edest",
+ "eclosure", and "inveclosure". */
+
+static reg_errcode_t
+analyze (regex_t *preg)
+{
+ re_dfa_t *dfa = preg->buffer;
+ reg_errcode_t ret;
+
+ /* Allocate arrays. */
+ dfa->nexts = re_malloc (Idx, dfa->nodes_alloc);
+ dfa->org_indices = re_malloc (Idx, dfa->nodes_alloc);
+ dfa->edests = re_malloc (re_node_set, dfa->nodes_alloc);
+ dfa->eclosures = re_malloc (re_node_set, dfa->nodes_alloc);
+ if (BE (dfa->nexts == NULL || dfa->org_indices == NULL || dfa->edests == NULL
+ || dfa->eclosures == NULL, 0))
+ return REG_ESPACE;
+
+ dfa->subexp_map = re_malloc (Idx, preg->re_nsub);
+ if (dfa->subexp_map != NULL)
+ {
+ Idx i;
+ for (i = 0; i < preg->re_nsub; i++)
+ dfa->subexp_map[i] = i;
+ preorder (dfa->str_tree, optimize_subexps, dfa);
+ for (i = 0; i < preg->re_nsub; i++)
+ if (dfa->subexp_map[i] != i)
+ break;
+ if (i == preg->re_nsub)
+ {
+ re_free (dfa->subexp_map);
+ dfa->subexp_map = NULL;
+ }
+ }
+
+ ret = postorder (dfa->str_tree, lower_subexps, preg);
+ if (BE (ret != REG_NOERROR, 0))
+ return ret;
+ ret = postorder (dfa->str_tree, calc_first, dfa);
+ if (BE (ret != REG_NOERROR, 0))
+ return ret;
+ preorder (dfa->str_tree, calc_next, dfa);
+ ret = preorder (dfa->str_tree, link_nfa_nodes, dfa);
+ if (BE (ret != REG_NOERROR, 0))
+ return ret;
+ ret = calc_eclosure (dfa);
+ if (BE (ret != REG_NOERROR, 0))
+ return ret;
+
+ /* We only need this during the prune_impossible_nodes pass in regexec.c;
+ skip it if p_i_n will not run, as calc_inveclosure can be quadratic. */
+ if ((!preg->no_sub && preg->re_nsub > 0 && dfa->has_plural_match)
+ || dfa->nbackref)
+ {
+ dfa->inveclosures = re_malloc (re_node_set, dfa->nodes_len);
+ if (BE (dfa->inveclosures == NULL, 0))
+ return REG_ESPACE;
+ ret = calc_inveclosure (dfa);
+ }
+
+ return ret;
+}
+
+/* Our parse trees are very unbalanced, so we cannot use a stack to
+ implement parse tree visits. Instead, we use parent pointers and
+ some hairy code in these two functions. */
+static reg_errcode_t
+postorder (bin_tree_t *root, reg_errcode_t (fn (void *, bin_tree_t *)),
+ void *extra)
+{
+ bin_tree_t *node, *prev;
+
+ for (node = root; ; )
+ {
+ /* Descend down the tree, preferably to the left (or to the right
+ if that's the only child). */
+ while (node->left || node->right)
+ if (node->left)
+ node = node->left;
+ else
+ node = node->right;
+
+ do
+ {
+ reg_errcode_t err = fn (extra, node);
+ if (BE (err != REG_NOERROR, 0))
+ return err;
+ if (node->parent == NULL)
+ return REG_NOERROR;
+ prev = node;
+ node = node->parent;
+ }
+ /* Go up while we have a node that is reached from the right. */
+ while (node->right == prev || node->right == NULL);
+ node = node->right;
+ }
+}
+
+static reg_errcode_t
+preorder (bin_tree_t *root, reg_errcode_t (fn (void *, bin_tree_t *)),
+ void *extra)
+{
+ bin_tree_t *node;
+
+ for (node = root; ; )
+ {
+ reg_errcode_t err = fn (extra, node);
+ if (BE (err != REG_NOERROR, 0))
+ return err;
+
+ /* Go to the left node, or up and to the right. */
+ if (node->left)
+ node = node->left;
+ else
+ {
+ bin_tree_t *prev = NULL;
+ while (node->right == prev || node->right == NULL)
+ {
+ prev = node;
+ node = node->parent;
+ if (!node)
+ return REG_NOERROR;
+ }
+ node = node->right;
+ }
+ }
+}
+
+/* Optimization pass: if a SUBEXP is entirely contained, strip it and tell
+ re_search_internal to map the inner one's opr.idx to this one's. Adjust
+ backreferences as well. Requires a preorder visit. */
+static reg_errcode_t
+optimize_subexps (void *extra, bin_tree_t *node)
+{
+ re_dfa_t *dfa = (re_dfa_t *) extra;
+
+ if (node->token.type == OP_BACK_REF && dfa->subexp_map)
+ {
+ int idx = node->token.opr.idx;
+ node->token.opr.idx = dfa->subexp_map[idx];
+ dfa->used_bkref_map |= 1 << node->token.opr.idx;
+ }
+
+ else if (node->token.type == SUBEXP
+ && node->left && node->left->token.type == SUBEXP)
+ {
+ Idx other_idx = node->left->token.opr.idx;
+
+ node->left = node->left->left;
+ if (node->left)
+ node->left->parent = node;
+
+ dfa->subexp_map[other_idx] = dfa->subexp_map[node->token.opr.idx];
+ if (other_idx < BITSET_WORD_BITS)
+ dfa->used_bkref_map &= ~((bitset_word_t) 1 << other_idx);
+ }
+
+ return REG_NOERROR;
+}
+
+/* Lowering pass: Turn each SUBEXP node into the appropriate concatenation
+ of OP_OPEN_SUBEXP, the body of the SUBEXP (if any) and OP_CLOSE_SUBEXP. */
+static reg_errcode_t
+lower_subexps (void *extra, bin_tree_t *node)
+{
+ regex_t *preg = (regex_t *) extra;
+ reg_errcode_t err = REG_NOERROR;
+
+ if (node->left && node->left->token.type == SUBEXP)
+ {
+ node->left = lower_subexp (&err, preg, node->left);
+ if (node->left)
+ node->left->parent = node;
+ }
+ if (node->right && node->right->token.type == SUBEXP)
+ {
+ node->right = lower_subexp (&err, preg, node->right);
+ if (node->right)
+ node->right->parent = node;
+ }
+
+ return err;
+}
+
+static bin_tree_t *
+lower_subexp (reg_errcode_t *err, regex_t *preg, bin_tree_t *node)
+{
+ re_dfa_t *dfa = preg->buffer;
+ bin_tree_t *body = node->left;
+ bin_tree_t *op, *cls, *tree1, *tree;
+
+ if (preg->no_sub
+ /* We do not optimize empty subexpressions, because otherwise we may
+ have bad CONCAT nodes with NULL children. This is obviously not
+ very common, so we do not lose much. An example that triggers
+ this case is the sed "script" /\(\)/x. */
+ && node->left != NULL
+ && (node->token.opr.idx >= BITSET_WORD_BITS
+ || !(dfa->used_bkref_map
+ & ((bitset_word_t) 1 << node->token.opr.idx))))
+ return node->left;
+
+ /* Convert the SUBEXP node to the concatenation of an
+ OP_OPEN_SUBEXP, the contents, and an OP_CLOSE_SUBEXP. */
+ op = create_tree (dfa, NULL, NULL, OP_OPEN_SUBEXP);
+ cls = create_tree (dfa, NULL, NULL, OP_CLOSE_SUBEXP);
+ tree1 = body ? create_tree (dfa, body, cls, CONCAT) : cls;
+ tree = create_tree (dfa, op, tree1, CONCAT);
+ if (BE (tree == NULL || tree1 == NULL || op == NULL || cls == NULL, 0))
+ {
+ *err = REG_ESPACE;
+ return NULL;
+ }
+
+ op->token.opr.idx = cls->token.opr.idx = node->token.opr.idx;
+ op->token.opt_subexp = cls->token.opt_subexp = node->token.opt_subexp;
+ return tree;
+}
+
+/* Pass 1 in building the NFA: compute FIRST and create unlinked automaton
+ nodes. Requires a postorder visit. */
+static reg_errcode_t
+calc_first (void *extra, bin_tree_t *node)
+{
+ re_dfa_t *dfa = (re_dfa_t *) extra;
+ if (node->token.type == CONCAT)
+ {
+ node->first = node->left->first;
+ node->node_idx = node->left->node_idx;
+ }
+ else
+ {
+ node->first = node;
+ node->node_idx = re_dfa_add_node (dfa, node->token);
+ if (BE (node->node_idx == -1, 0))
+ return REG_ESPACE;
+ if (node->token.type == ANCHOR)
+ dfa->nodes[node->node_idx].constraint = node->token.opr.ctx_type;
+ }
+ return REG_NOERROR;
+}
+
+/* Pass 2: compute NEXT on the tree. Preorder visit. */
+static reg_errcode_t
+calc_next (void *extra, bin_tree_t *node)
+{
+ switch (node->token.type)
+ {
+ case OP_DUP_ASTERISK:
+ node->left->next = node;
+ break;
+ case CONCAT:
+ node->left->next = node->right->first;
+ node->right->next = node->next;
+ break;
+ default:
+ if (node->left)
+ node->left->next = node->next;
+ if (node->right)
+ node->right->next = node->next;
+ break;
+ }
+ return REG_NOERROR;
+}
+
+/* Pass 3: link all DFA nodes to their NEXT node (any order will do). */
+static reg_errcode_t
+link_nfa_nodes (void *extra, bin_tree_t *node)
+{
+ re_dfa_t *dfa = (re_dfa_t *) extra;
+ Idx idx = node->node_idx;
+ reg_errcode_t err = REG_NOERROR;
+
+ switch (node->token.type)
+ {
+ case CONCAT:
+ break;
+
+ case END_OF_RE:
+ assert (node->next == NULL);
+ break;
+
+ case OP_DUP_ASTERISK:
+ case OP_ALT:
+ {
+ Idx left, right;
+ dfa->has_plural_match = 1;
+ if (node->left != NULL)
+ left = node->left->first->node_idx;
+ else
+ left = node->next->node_idx;
+ if (node->right != NULL)
+ right = node->right->first->node_idx;
+ else
+ right = node->next->node_idx;
+ assert (left > -1);
+ assert (right > -1);
+ err = re_node_set_init_2 (dfa->edests + idx, left, right);
+ }
+ break;
+
+ case ANCHOR:
+ case OP_OPEN_SUBEXP:
+ case OP_CLOSE_SUBEXP:
+ err = re_node_set_init_1 (dfa->edests + idx, node->next->node_idx);
+ break;
+
+ case OP_BACK_REF:
+ dfa->nexts[idx] = node->next->node_idx;
+ if (node->token.type == OP_BACK_REF)
+ err = re_node_set_init_1 (dfa->edests + idx, dfa->nexts[idx]);
+ break;
+
+ default:
+ assert (!IS_EPSILON_NODE (node->token.type));
+ dfa->nexts[idx] = node->next->node_idx;
+ break;
+ }
+
+ return err;
+}
+
+/* Duplicate the epsilon closure of the node ROOT_NODE.
+ Note that duplicated nodes have constraint INIT_CONSTRAINT in addition
+ to their own constraint. */
+
+static reg_errcode_t
+duplicate_node_closure (re_dfa_t *dfa, Idx top_org_node, Idx top_clone_node,
+ Idx root_node, unsigned int init_constraint)
+{
+ Idx org_node, clone_node;
+ bool ok;
+ unsigned int constraint = init_constraint;
+ for (org_node = top_org_node, clone_node = top_clone_node;;)
+ {
+ Idx org_dest, clone_dest;
+ if (dfa->nodes[org_node].type == OP_BACK_REF)
+ {
+ /* If the back reference epsilon-transit, its destination must
+ also have the constraint. Then duplicate the epsilon closure
+ of the destination of the back reference, and store it in
+ edests of the back reference. */
+ org_dest = dfa->nexts[org_node];
+ re_node_set_empty (dfa->edests + clone_node);
+ clone_dest = duplicate_node (dfa, org_dest, constraint);
+ if (BE (clone_dest == -1, 0))
+ return REG_ESPACE;
+ dfa->nexts[clone_node] = dfa->nexts[org_node];
+ ok = re_node_set_insert (dfa->edests + clone_node, clone_dest);
+ if (BE (! ok, 0))
+ return REG_ESPACE;
+ }
+ else if (dfa->edests[org_node].nelem == 0)
+ {
+ /* In case of the node can't epsilon-transit, don't duplicate the
+ destination and store the original destination as the
+ destination of the node. */
+ dfa->nexts[clone_node] = dfa->nexts[org_node];
+ break;
+ }
+ else if (dfa->edests[org_node].nelem == 1)
+ {
+ /* In case of the node can epsilon-transit, and it has only one
+ destination. */
+ org_dest = dfa->edests[org_node].elems[0];
+ re_node_set_empty (dfa->edests + clone_node);
+ /* If the node is root_node itself, it means the epsilon closure
+ has a loop. Then tie it to the destination of the root_node. */
+ if (org_node == root_node && clone_node != org_node)
+ {
+ ok = re_node_set_insert (dfa->edests + clone_node, org_dest);
+ if (BE (! ok, 0))
+ return REG_ESPACE;
+ break;
+ }
+ /* In case the node has another constraint, append it. */
+ constraint |= dfa->nodes[org_node].constraint;
+ clone_dest = duplicate_node (dfa, org_dest, constraint);
+ if (BE (clone_dest == -1, 0))
+ return REG_ESPACE;
+ ok = re_node_set_insert (dfa->edests + clone_node, clone_dest);
+ if (BE (! ok, 0))
+ return REG_ESPACE;
+ }
+ else /* dfa->edests[org_node].nelem == 2 */
+ {
+ /* In case of the node can epsilon-transit, and it has two
+ destinations. In the bin_tree_t and DFA, that's '|' and '*'. */
+ org_dest = dfa->edests[org_node].elems[0];
+ re_node_set_empty (dfa->edests + clone_node);
+ /* Search for a duplicated node which satisfies the constraint. */
+ clone_dest = search_duplicated_node (dfa, org_dest, constraint);
+ if (clone_dest == -1)
+ {
+ /* There is no such duplicated node, create a new one. */
+ reg_errcode_t err;
+ clone_dest = duplicate_node (dfa, org_dest, constraint);
+ if (BE (clone_dest == -1, 0))
+ return REG_ESPACE;
+ ok = re_node_set_insert (dfa->edests + clone_node, clone_dest);
+ if (BE (! ok, 0))
+ return REG_ESPACE;
+ err = duplicate_node_closure (dfa, org_dest, clone_dest,
+ root_node, constraint);
+ if (BE (err != REG_NOERROR, 0))
+ return err;
+ }
+ else
+ {
+ /* There is a duplicated node which satisfies the constraint,
+ use it to avoid infinite loop. */
+ ok = re_node_set_insert (dfa->edests + clone_node, clone_dest);
+ if (BE (! ok, 0))
+ return REG_ESPACE;
+ }
+
+ org_dest = dfa->edests[org_node].elems[1];
+ clone_dest = duplicate_node (dfa, org_dest, constraint);
+ if (BE (clone_dest == -1, 0))
+ return REG_ESPACE;
+ ok = re_node_set_insert (dfa->edests + clone_node, clone_dest);
+ if (BE (! ok, 0))
+ return REG_ESPACE;
+ }
+ org_node = org_dest;
+ clone_node = clone_dest;
+ }
+ return REG_NOERROR;
+}
+
+/* Search for a node which is duplicated from the node ORG_NODE, and
+ satisfies the constraint CONSTRAINT. */
+
+static Idx
+search_duplicated_node (const re_dfa_t *dfa, Idx org_node,
+ unsigned int constraint)
+{
+ Idx idx;
+ for (idx = dfa->nodes_len - 1; dfa->nodes[idx].duplicated && idx > 0; --idx)
+ {
+ if (org_node == dfa->org_indices[idx]
+ && constraint == dfa->nodes[idx].constraint)
+ return idx; /* Found. */
+ }
+ return -1; /* Not found. */
+}
+
+/* Duplicate the node whose index is ORG_IDX and set the constraint CONSTRAINT.
+ Return the index of the new node, or -1 if insufficient storage is
+ available. */
+
+static Idx
+duplicate_node (re_dfa_t *dfa, Idx org_idx, unsigned int constraint)
+{
+ Idx dup_idx = re_dfa_add_node (dfa, dfa->nodes[org_idx]);
+ if (BE (dup_idx != -1, 1))
+ {
+ dfa->nodes[dup_idx].constraint = constraint;
+ dfa->nodes[dup_idx].constraint |= dfa->nodes[org_idx].constraint;
+ dfa->nodes[dup_idx].duplicated = 1;
+
+ /* Store the index of the original node. */
+ dfa->org_indices[dup_idx] = org_idx;
+ }
+ return dup_idx;
+}
+
+static reg_errcode_t
+calc_inveclosure (re_dfa_t *dfa)
+{
+ Idx src, idx;
+ bool ok;
+ for (idx = 0; idx < dfa->nodes_len; ++idx)
+ re_node_set_init_empty (dfa->inveclosures + idx);
+
+ for (src = 0; src < dfa->nodes_len; ++src)
+ {
+ Idx *elems = dfa->eclosures[src].elems;
+ for (idx = 0; idx < dfa->eclosures[src].nelem; ++idx)
+ {
+ ok = re_node_set_insert_last (dfa->inveclosures + elems[idx], src);
+ if (BE (! ok, 0))
+ return REG_ESPACE;
+ }
+ }
+
+ return REG_NOERROR;
+}
+
+/* Calculate "eclosure" for all the node in DFA. */
+
+static reg_errcode_t
+calc_eclosure (re_dfa_t *dfa)
+{
+ Idx node_idx;
+ bool incomplete;
+#ifdef DEBUG
+ assert (dfa->nodes_len > 0);
+#endif
+ incomplete = false;
+ /* For each nodes, calculate epsilon closure. */
+ for (node_idx = 0; ; ++node_idx)
+ {
+ reg_errcode_t err;
+ re_node_set eclosure_elem;
+ if (node_idx == dfa->nodes_len)
+ {
+ if (!incomplete)
+ break;
+ incomplete = false;
+ node_idx = 0;
+ }
+
+#ifdef DEBUG
+ assert (dfa->eclosures[node_idx].nelem != -1);
+#endif
+
+ /* If we have already calculated, skip it. */
+ if (dfa->eclosures[node_idx].nelem != 0)
+ continue;
+ /* Calculate epsilon closure of 'node_idx'. */
+ err = calc_eclosure_iter (&eclosure_elem, dfa, node_idx, true);
+ if (BE (err != REG_NOERROR, 0))
+ return err;
+
+ if (dfa->eclosures[node_idx].nelem == 0)
+ {
+ incomplete = true;
+ re_node_set_free (&eclosure_elem);
+ }
+ }
+ return REG_NOERROR;
+}
+
+/* Calculate epsilon closure of NODE. */
+
+static reg_errcode_t
+calc_eclosure_iter (re_node_set *new_set, re_dfa_t *dfa, Idx node, bool root)
+{
+ reg_errcode_t err;
+ Idx i;
+ re_node_set eclosure;
+ bool ok;
+ bool incomplete = false;
+ err = re_node_set_alloc (&eclosure, dfa->edests[node].nelem + 1);
+ if (BE (err != REG_NOERROR, 0))
+ return err;
+
+ /* This indicates that we are calculating this node now.
+ We reference this value to avoid infinite loop. */
+ dfa->eclosures[node].nelem = -1;
+
+ /* If the current node has constraints, duplicate all nodes
+ since they must inherit the constraints. */
+ if (dfa->nodes[node].constraint
+ && dfa->edests[node].nelem
+ && !dfa->nodes[dfa->edests[node].elems[0]].duplicated)
+ {
+ err = duplicate_node_closure (dfa, node, node, node,
+ dfa->nodes[node].constraint);
+ if (BE (err != REG_NOERROR, 0))
+ return err;
+ }
+
+ /* Expand each epsilon destination nodes. */
+ if (IS_EPSILON_NODE(dfa->nodes[node].type))
+ for (i = 0; i < dfa->edests[node].nelem; ++i)
+ {
+ re_node_set eclosure_elem;
+ Idx edest = dfa->edests[node].elems[i];
+ /* If calculating the epsilon closure of 'edest' is in progress,
+ return intermediate result. */
+ if (dfa->eclosures[edest].nelem == -1)
+ {
+ incomplete = true;
+ continue;
+ }
+ /* If we haven't calculated the epsilon closure of 'edest' yet,
+ calculate now. Otherwise use calculated epsilon closure. */
+ if (dfa->eclosures[edest].nelem == 0)
+ {
+ err = calc_eclosure_iter (&eclosure_elem, dfa, edest, false);
+ if (BE (err != REG_NOERROR, 0))
+ return err;
+ }
+ else
+ eclosure_elem = dfa->eclosures[edest];
+ /* Merge the epsilon closure of 'edest'. */
+ err = re_node_set_merge (&eclosure, &eclosure_elem);
+ if (BE (err != REG_NOERROR, 0))
+ return err;
+ /* If the epsilon closure of 'edest' is incomplete,
+ the epsilon closure of this node is also incomplete. */
+ if (dfa->eclosures[edest].nelem == 0)
+ {
+ incomplete = true;
+ re_node_set_free (&eclosure_elem);
+ }
+ }
+
+ /* An epsilon closure includes itself. */
+ ok = re_node_set_insert (&eclosure, node);
+ if (BE (! ok, 0))
+ return REG_ESPACE;
+ if (incomplete && !root)
+ dfa->eclosures[node].nelem = 0;
+ else
+ dfa->eclosures[node] = eclosure;
+ *new_set = eclosure;
+ return REG_NOERROR;
+}
+
+/* Functions for token which are used in the parser. */
+
+/* Fetch a token from INPUT.
+ We must not use this function inside bracket expressions. */
+
+static void
+fetch_token (re_token_t *result, re_string_t *input, reg_syntax_t syntax)
+{
+ re_string_skip_bytes (input, peek_token (result, input, syntax));
+}
+
+/* Peek a token from INPUT, and return the length of the token.
+ We must not use this function inside bracket expressions. */
+
+static int
+peek_token (re_token_t *token, re_string_t *input, reg_syntax_t syntax)
+{
+ unsigned char c;
+
+ if (re_string_eoi (input))
+ {
+ token->type = END_OF_RE;
+ return 0;
+ }
+
+ c = re_string_peek_byte (input, 0);
+ token->opr.c = c;
+
+ token->word_char = 0;
+#ifdef RE_ENABLE_I18N
+ token->mb_partial = 0;
+ if (input->mb_cur_max > 1 &&
+ !re_string_first_byte (input, re_string_cur_idx (input)))
+ {
+ token->type = CHARACTER;
+ token->mb_partial = 1;
+ return 1;
+ }
+#endif
+ if (c == '\\')
+ {
+ unsigned char c2;
+ if (re_string_cur_idx (input) + 1 >= re_string_length (input))
+ {
+ token->type = BACK_SLASH;
+ return 1;
+ }
+
+ c2 = re_string_peek_byte_case (input, 1);
+ token->opr.c = c2;
+ token->type = CHARACTER;
+#ifdef RE_ENABLE_I18N
+ if (input->mb_cur_max > 1)
+ {
+ wint_t wc = re_string_wchar_at (input,
+ re_string_cur_idx (input) + 1);
+ token->word_char = IS_WIDE_WORD_CHAR (wc) != 0;
+ }
+ else
+#endif
+ token->word_char = IS_WORD_CHAR (c2) != 0;
+
+ switch (c2)
+ {
+ case '|':
+ if (!(syntax & RE_LIMITED_OPS) && !(syntax & RE_NO_BK_VBAR))
+ token->type = OP_ALT;
+ break;
+ case '1': case '2': case '3': case '4': case '5':
+ case '6': case '7': case '8': case '9':
+ if (!(syntax & RE_NO_BK_REFS))
+ {
+ token->type = OP_BACK_REF;
+ token->opr.idx = c2 - '1';
+ }
+ break;
+ case '<':
+ if (!(syntax & RE_NO_GNU_OPS))
+ {
+ token->type = ANCHOR;
+ token->opr.ctx_type = WORD_FIRST;
+ }
+ break;
+ case '>':
+ if (!(syntax & RE_NO_GNU_OPS))
+ {
+ token->type = ANCHOR;
+ token->opr.ctx_type = WORD_LAST;
+ }
+ break;
+ case 'b':
+ if (!(syntax & RE_NO_GNU_OPS))
+ {
+ token->type = ANCHOR;
+ token->opr.ctx_type = WORD_DELIM;
+ }
+ break;
+ case 'B':
+ if (!(syntax & RE_NO_GNU_OPS))
+ {
+ token->type = ANCHOR;
+ token->opr.ctx_type = NOT_WORD_DELIM;
+ }
+ break;
+ case 'w':
+ if (!(syntax & RE_NO_GNU_OPS))
+ token->type = OP_WORD;
+ break;
+ case 'W':
+ if (!(syntax & RE_NO_GNU_OPS))
+ token->type = OP_NOTWORD;
+ break;
+ case 's':
+ if (!(syntax & RE_NO_GNU_OPS))
+ token->type = OP_SPACE;
+ break;
+ case 'S':
+ if (!(syntax & RE_NO_GNU_OPS))
+ token->type = OP_NOTSPACE;
+ break;
+ case '`':
+ if (!(syntax & RE_NO_GNU_OPS))
+ {
+ token->type = ANCHOR;
+ token->opr.ctx_type = BUF_FIRST;
+ }
+ break;
+ case '\'':
+ if (!(syntax & RE_NO_GNU_OPS))
+ {
+ token->type = ANCHOR;
+ token->opr.ctx_type = BUF_LAST;
+ }
+ break;
+ case '(':
+ if (!(syntax & RE_NO_BK_PARENS))
+ token->type = OP_OPEN_SUBEXP;
+ break;
+ case ')':
+ if (!(syntax & RE_NO_BK_PARENS))
+ token->type = OP_CLOSE_SUBEXP;
+ break;
+ case '+':
+ if (!(syntax & RE_LIMITED_OPS) && (syntax & RE_BK_PLUS_QM))
+ token->type = OP_DUP_PLUS;
+ break;
+ case '?':
+ if (!(syntax & RE_LIMITED_OPS) && (syntax & RE_BK_PLUS_QM))
+ token->type = OP_DUP_QUESTION;
+ break;
+ case '{':
+ if ((syntax & RE_INTERVALS) && (!(syntax & RE_NO_BK_BRACES)))
+ token->type = OP_OPEN_DUP_NUM;
+ break;
+ case '}':
+ if ((syntax & RE_INTERVALS) && (!(syntax & RE_NO_BK_BRACES)))
+ token->type = OP_CLOSE_DUP_NUM;
+ break;
+ default:
+ break;
+ }
+ return 2;
+ }
+
+ token->type = CHARACTER;
+#ifdef RE_ENABLE_I18N
+ if (input->mb_cur_max > 1)
+ {
+ wint_t wc = re_string_wchar_at (input, re_string_cur_idx (input));
+ token->word_char = IS_WIDE_WORD_CHAR (wc) != 0;
+ }
+ else
+#endif
+ token->word_char = IS_WORD_CHAR (token->opr.c);
+
+ switch (c)
+ {
+ case '\n':
+ if (syntax & RE_NEWLINE_ALT)
+ token->type = OP_ALT;
+ break;
+ case '|':
+ if (!(syntax & RE_LIMITED_OPS) && (syntax & RE_NO_BK_VBAR))
+ token->type = OP_ALT;
+ break;
+ case '*':
+ token->type = OP_DUP_ASTERISK;
+ break;
+ case '+':
+ if (!(syntax & RE_LIMITED_OPS) && !(syntax & RE_BK_PLUS_QM))
+ token->type = OP_DUP_PLUS;
+ break;
+ case '?':
+ if (!(syntax & RE_LIMITED_OPS) && !(syntax & RE_BK_PLUS_QM))
+ token->type = OP_DUP_QUESTION;
+ break;
+ case '{':
+ if ((syntax & RE_INTERVALS) && (syntax & RE_NO_BK_BRACES))
+ token->type = OP_OPEN_DUP_NUM;
+ break;
+ case '}':
+ if ((syntax & RE_INTERVALS) && (syntax & RE_NO_BK_BRACES))
+ token->type = OP_CLOSE_DUP_NUM;
+ break;
+ case '(':
+ if (syntax & RE_NO_BK_PARENS)
+ token->type = OP_OPEN_SUBEXP;
+ break;
+ case ')':
+ if (syntax & RE_NO_BK_PARENS)
+ token->type = OP_CLOSE_SUBEXP;
+ break;
+ case '[':
+ token->type = OP_OPEN_BRACKET;
+ break;
+ case '.':
+ token->type = OP_PERIOD;
+ break;
+ case '^':
+ if (!(syntax & (RE_CONTEXT_INDEP_ANCHORS | RE_CARET_ANCHORS_HERE)) &&
+ re_string_cur_idx (input) != 0)
+ {
+ char prev = re_string_peek_byte (input, -1);
+ if (!(syntax & RE_NEWLINE_ALT) || prev != '\n')
+ break;
+ }
+ token->type = ANCHOR;
+ token->opr.ctx_type = LINE_FIRST;
+ break;
+ case '$':
+ if (!(syntax & RE_CONTEXT_INDEP_ANCHORS) &&
+ re_string_cur_idx (input) + 1 != re_string_length (input))
+ {
+ re_token_t next;
+ re_string_skip_bytes (input, 1);
+ peek_token (&next, input, syntax);
+ re_string_skip_bytes (input, -1);
+ if (next.type != OP_ALT && next.type != OP_CLOSE_SUBEXP)
+ break;
+ }
+ token->type = ANCHOR;
+ token->opr.ctx_type = LINE_LAST;
+ break;
+ default:
+ break;
+ }
+ return 1;
+}
+
+/* Peek a token from INPUT, and return the length of the token.
+ We must not use this function out of bracket expressions. */
+
+static int
+peek_token_bracket (re_token_t *token, re_string_t *input, reg_syntax_t syntax)
+{
+ unsigned char c;
+ if (re_string_eoi (input))
+ {
+ token->type = END_OF_RE;
+ return 0;
+ }
+ c = re_string_peek_byte (input, 0);
+ token->opr.c = c;
+
+#ifdef RE_ENABLE_I18N
+ if (input->mb_cur_max > 1 &&
+ !re_string_first_byte (input, re_string_cur_idx (input)))
+ {
+ token->type = CHARACTER;
+ return 1;
+ }
+#endif /* RE_ENABLE_I18N */
+
+ if (c == '\\' && (syntax & RE_BACKSLASH_ESCAPE_IN_LISTS)
+ && re_string_cur_idx (input) + 1 < re_string_length (input))
+ {
+ /* In this case, '\' escape a character. */
+ unsigned char c2;
+ re_string_skip_bytes (input, 1);
+ c2 = re_string_peek_byte (input, 0);
+ token->opr.c = c2;
+ token->type = CHARACTER;
+ return 1;
+ }
+ if (c == '[') /* '[' is a special char in a bracket exps. */
+ {
+ unsigned char c2;
+ int token_len;
+ if (re_string_cur_idx (input) + 1 < re_string_length (input))
+ c2 = re_string_peek_byte (input, 1);
+ else
+ c2 = 0;
+ token->opr.c = c2;
+ token_len = 2;
+ switch (c2)
+ {
+ case '.':
+ token->type = OP_OPEN_COLL_ELEM;
+ break;
+
+ case '=':
+ token->type = OP_OPEN_EQUIV_CLASS;
+ break;
+
+ case ':':
+ if (syntax & RE_CHAR_CLASSES)
+ {
+ token->type = OP_OPEN_CHAR_CLASS;
+ break;
+ }
+ FALLTHROUGH;
+ default:
+ token->type = CHARACTER;
+ token->opr.c = c;
+ token_len = 1;
+ break;
+ }
+ return token_len;
+ }
+ switch (c)
+ {
+ case '-':
+ token->type = OP_CHARSET_RANGE;
+ break;
+ case ']':
+ token->type = OP_CLOSE_BRACKET;
+ break;
+ case '^':
+ token->type = OP_NON_MATCH_LIST;
+ break;
+ default:
+ token->type = CHARACTER;
+ }
+ return 1;
+}
+
+/* Functions for parser. */
+
+/* Entry point of the parser.
+ Parse the regular expression REGEXP and return the structure tree.
+ If an error occurs, ERR is set by error code, and return NULL.
+ This function build the following tree, from regular expression <reg_exp>:
+ CAT
+ / \
+ / \
+ <reg_exp> EOR
+
+ CAT means concatenation.
+ EOR means end of regular expression. */
+
+static bin_tree_t *
+parse (re_string_t *regexp, regex_t *preg, reg_syntax_t syntax,
+ reg_errcode_t *err)
+{
+ re_dfa_t *dfa = preg->buffer;
+ bin_tree_t *tree, *eor, *root;
+ re_token_t current_token;
+ dfa->syntax = syntax;
+ fetch_token (¤t_token, regexp, syntax | RE_CARET_ANCHORS_HERE);
+ tree = parse_reg_exp (regexp, preg, ¤t_token, syntax, 0, err);
+ if (BE (*err != REG_NOERROR && tree == NULL, 0))
+ return NULL;
+ eor = create_tree (dfa, NULL, NULL, END_OF_RE);
+ if (tree != NULL)
+ root = create_tree (dfa, tree, eor, CONCAT);
+ else
+ root = eor;
+ if (BE (eor == NULL || root == NULL, 0))
+ {
+ *err = REG_ESPACE;
+ return NULL;
+ }
+ return root;
+}
+
+/* This function build the following tree, from regular expression
+ <branch1>|<branch2>:
+ ALT
+ / \
+ / \
+ <branch1> <branch2>
+
+ ALT means alternative, which represents the operator '|'. */
+
+static bin_tree_t *
+parse_reg_exp (re_string_t *regexp, regex_t *preg, re_token_t *token,
+ reg_syntax_t syntax, Idx nest, reg_errcode_t *err)
+{
+ re_dfa_t *dfa = preg->buffer;
+ bin_tree_t *tree, *branch = NULL;
+ bitset_word_t initial_bkref_map = dfa->completed_bkref_map;
+ tree = parse_branch (regexp, preg, token, syntax, nest, err);
+ if (BE (*err != REG_NOERROR && tree == NULL, 0))
+ return NULL;
+
+ while (token->type == OP_ALT)
+ {
+ fetch_token (token, regexp, syntax | RE_CARET_ANCHORS_HERE);
+ if (token->type != OP_ALT && token->type != END_OF_RE
+ && (nest == 0 || token->type != OP_CLOSE_SUBEXP))
+ {
+ bitset_word_t accumulated_bkref_map = dfa->completed_bkref_map;
+ dfa->completed_bkref_map = initial_bkref_map;
+ branch = parse_branch (regexp, preg, token, syntax, nest, err);
+ if (BE (*err != REG_NOERROR && branch == NULL, 0))
+ {
+ if (tree != NULL)
+ postorder (tree, free_tree, NULL);
+ return NULL;
+ }
+ dfa->completed_bkref_map |= accumulated_bkref_map;
+ }
+ else
+ branch = NULL;
+ tree = create_tree (dfa, tree, branch, OP_ALT);
+ if (BE (tree == NULL, 0))
+ {
+ *err = REG_ESPACE;
+ return NULL;
+ }
+ }
+ return tree;
+}
+
+/* This function build the following tree, from regular expression
+ <exp1><exp2>:
+ CAT
+ / \
+ / \
+ <exp1> <exp2>
+
+ CAT means concatenation. */
+
+static bin_tree_t *
+parse_branch (re_string_t *regexp, regex_t *preg, re_token_t *token,
+ reg_syntax_t syntax, Idx nest, reg_errcode_t *err)
+{
+ bin_tree_t *tree, *expr;
+ re_dfa_t *dfa = preg->buffer;
+ tree = parse_expression (regexp, preg, token, syntax, nest, err);
+ if (BE (*err != REG_NOERROR && tree == NULL, 0))
+ return NULL;
+
+ while (token->type != OP_ALT && token->type != END_OF_RE
+ && (nest == 0 || token->type != OP_CLOSE_SUBEXP))
+ {
+ expr = parse_expression (regexp, preg, token, syntax, nest, err);
+ if (BE (*err != REG_NOERROR && expr == NULL, 0))
+ {
+ if (tree != NULL)
+ postorder (tree, free_tree, NULL);
+ return NULL;
+ }
+ if (tree != NULL && expr != NULL)
+ {
+ bin_tree_t *newtree = create_tree (dfa, tree, expr, CONCAT);
+ if (newtree == NULL)
+ {
+ postorder (expr, free_tree, NULL);
+ postorder (tree, free_tree, NULL);
+ *err = REG_ESPACE;
+ return NULL;
+ }
+ tree = newtree;
+ }
+ else if (tree == NULL)
+ tree = expr;
+ /* Otherwise expr == NULL, we don't need to create new tree. */
+ }
+ return tree;
+}
+
+/* This function build the following tree, from regular expression a*:
+ *
+ |
+ a
+*/
+
+static bin_tree_t *
+parse_expression (re_string_t *regexp, regex_t *preg, re_token_t *token,
+ reg_syntax_t syntax, Idx nest, reg_errcode_t *err)
+{
+ re_dfa_t *dfa = preg->buffer;
+ bin_tree_t *tree;
+ switch (token->type)
+ {
+ case CHARACTER:
+ tree = create_token_tree (dfa, NULL, NULL, token);
+ if (BE (tree == NULL, 0))
+ {
+ *err = REG_ESPACE;
+ return NULL;
+ }
+#ifdef RE_ENABLE_I18N
+ if (dfa->mb_cur_max > 1)
+ {
+ while (!re_string_eoi (regexp)
+ && !re_string_first_byte (regexp, re_string_cur_idx (regexp)))
+ {
+ bin_tree_t *mbc_remain;
+ fetch_token (token, regexp, syntax);
+ mbc_remain = create_token_tree (dfa, NULL, NULL, token);
+ tree = create_tree (dfa, tree, mbc_remain, CONCAT);
+ if (BE (mbc_remain == NULL || tree == NULL, 0))
+ {
+ *err = REG_ESPACE;
+ return NULL;
+ }
+ }
+ }
+#endif
+ break;
+
+ case OP_OPEN_SUBEXP:
+ tree = parse_sub_exp (regexp, preg, token, syntax, nest + 1, err);
+ if (BE (*err != REG_NOERROR && tree == NULL, 0))
+ return NULL;
+ break;
+
+ case OP_OPEN_BRACKET:
+ tree = parse_bracket_exp (regexp, dfa, token, syntax, err);
+ if (BE (*err != REG_NOERROR && tree == NULL, 0))
+ return NULL;
+ break;
+
+ case OP_BACK_REF:
+ if (!BE (dfa->completed_bkref_map & (1 << token->opr.idx), 1))
+ {
+ *err = REG_ESUBREG;
+ return NULL;
+ }
+ dfa->used_bkref_map |= 1 << token->opr.idx;
+ tree = create_token_tree (dfa, NULL, NULL, token);
+ if (BE (tree == NULL, 0))
+ {
+ *err = REG_ESPACE;
+ return NULL;
+ }
+ ++dfa->nbackref;
+ dfa->has_mb_node = 1;
+ break;
+
+ case OP_OPEN_DUP_NUM:
+ if (syntax & RE_CONTEXT_INVALID_DUP)
+ {
+ *err = REG_BADRPT;
+ return NULL;
+ }
+ FALLTHROUGH;
+ case OP_DUP_ASTERISK:
+ case OP_DUP_PLUS:
+ case OP_DUP_QUESTION:
+ if (syntax & RE_CONTEXT_INVALID_OPS)
+ {
+ *err = REG_BADRPT;
+ return NULL;
+ }
+ else if (syntax & RE_CONTEXT_INDEP_OPS)
+ {
+ fetch_token (token, regexp, syntax);
+ return parse_expression (regexp, preg, token, syntax, nest, err);
+ }
+ FALLTHROUGH;
+ case OP_CLOSE_SUBEXP:
+ if ((token->type == OP_CLOSE_SUBEXP) &&
+ !(syntax & RE_UNMATCHED_RIGHT_PAREN_ORD))
+ {
+ *err = REG_ERPAREN;
+ return NULL;
+ }
+ FALLTHROUGH;
+ case OP_CLOSE_DUP_NUM:
+ /* We treat it as a normal character. */
+
+ /* Then we can these characters as normal characters. */
+ token->type = CHARACTER;
+ /* mb_partial and word_char bits should be initialized already
+ by peek_token. */
+ tree = create_token_tree (dfa, NULL, NULL, token);
+ if (BE (tree == NULL, 0))
+ {
+ *err = REG_ESPACE;
+ return NULL;
+ }
+ break;
+
+ case ANCHOR:
+ if ((token->opr.ctx_type
+ & (WORD_DELIM | NOT_WORD_DELIM | WORD_FIRST | WORD_LAST))
+ && dfa->word_ops_used == 0)
+ init_word_char (dfa);
+ if (token->opr.ctx_type == WORD_DELIM
+ || token->opr.ctx_type == NOT_WORD_DELIM)
+ {
+ bin_tree_t *tree_first, *tree_last;
+ if (token->opr.ctx_type == WORD_DELIM)
+ {
+ token->opr.ctx_type = WORD_FIRST;
+ tree_first = create_token_tree (dfa, NULL, NULL, token);
+ token->opr.ctx_type = WORD_LAST;
+ }
+ else
+ {
+ token->opr.ctx_type = INSIDE_WORD;
+ tree_first = create_token_tree (dfa, NULL, NULL, token);
+ token->opr.ctx_type = INSIDE_NOTWORD;
+ }
+ tree_last = create_token_tree (dfa, NULL, NULL, token);
+ tree = create_tree (dfa, tree_first, tree_last, OP_ALT);
+ if (BE (tree_first == NULL || tree_last == NULL || tree == NULL, 0))
+ {
+ *err = REG_ESPACE;
+ return NULL;
+ }
+ }
+ else
+ {
+ tree = create_token_tree (dfa, NULL, NULL, token);
+ if (BE (tree == NULL, 0))
+ {
+ *err = REG_ESPACE;
+ return NULL;
+ }
+ }
+ /* We must return here, since ANCHORs can't be followed
+ by repetition operators.
+ eg. RE"^*" is invalid or "<ANCHOR(^)><CHAR(*)>",
+ it must not be "<ANCHOR(^)><REPEAT(*)>". */
+ fetch_token (token, regexp, syntax);
+ return tree;
+
+ case OP_PERIOD:
+ tree = create_token_tree (dfa, NULL, NULL, token);
+ if (BE (tree == NULL, 0))
+ {
+ *err = REG_ESPACE;
+ return NULL;
+ }
+ if (dfa->mb_cur_max > 1)
+ dfa->has_mb_node = 1;
+ break;
+
+ case OP_WORD:
+ case OP_NOTWORD:
+ tree = build_charclass_op (dfa, regexp->trans,
+ "alnum",
+ "_",
+ token->type == OP_NOTWORD, err);
+ if (BE (*err != REG_NOERROR && tree == NULL, 0))
+ return NULL;
+ break;
+
+ case OP_SPACE:
+ case OP_NOTSPACE:
+ tree = build_charclass_op (dfa, regexp->trans,
+ "space",
+ "",
+ token->type == OP_NOTSPACE, err);
+ if (BE (*err != REG_NOERROR && tree == NULL, 0))
+ return NULL;
+ break;
+
+ case OP_ALT:
+ case END_OF_RE:
+ return NULL;
+
+ case BACK_SLASH:
+ *err = REG_EESCAPE;
+ return NULL;
+
+ default:
+ /* Must not happen? */
+#ifdef DEBUG
+ assert (0);
+#endif
+ return NULL;
+ }
+ fetch_token (token, regexp, syntax);
+
+ while (token->type == OP_DUP_ASTERISK || token->type == OP_DUP_PLUS
+ || token->type == OP_DUP_QUESTION || token->type == OP_OPEN_DUP_NUM)
+ {
+ bin_tree_t *dup_tree = parse_dup_op (tree, regexp, dfa, token,
+ syntax, err);
+ if (BE (*err != REG_NOERROR && dup_tree == NULL, 0))
+ {
+ if (tree != NULL)
+ postorder (tree, free_tree, NULL);
+ return NULL;
+ }
+ tree = dup_tree;
+ /* In BRE consecutive duplications are not allowed. */
+ if ((syntax & RE_CONTEXT_INVALID_DUP)
+ && (token->type == OP_DUP_ASTERISK
+ || token->type == OP_OPEN_DUP_NUM))
+ {
+ if (tree != NULL)
+ postorder (tree, free_tree, NULL);
+ *err = REG_BADRPT;
+ return NULL;
+ }
+ }
+
+ return tree;
+}
+
+/* This function build the following tree, from regular expression
+ (<reg_exp>):
+ SUBEXP
+ |
+ <reg_exp>
+*/
+
+static bin_tree_t *
+parse_sub_exp (re_string_t *regexp, regex_t *preg, re_token_t *token,
+ reg_syntax_t syntax, Idx nest, reg_errcode_t *err)
+{
+ re_dfa_t *dfa = preg->buffer;
+ bin_tree_t *tree;
+ size_t cur_nsub;
+ cur_nsub = preg->re_nsub++;
+
+ fetch_token (token, regexp, syntax | RE_CARET_ANCHORS_HERE);
+
+ /* The subexpression may be a null string. */
+ if (token->type == OP_CLOSE_SUBEXP)
+ tree = NULL;
+ else
+ {
+ tree = parse_reg_exp (regexp, preg, token, syntax, nest, err);
+ if (BE (*err == REG_NOERROR && token->type != OP_CLOSE_SUBEXP, 0))
+ {
+ if (tree != NULL)
+ postorder (tree, free_tree, NULL);
+ *err = REG_EPAREN;
+ }
+ if (BE (*err != REG_NOERROR, 0))
+ return NULL;
+ }
+
+ if (cur_nsub <= '9' - '1')
+ dfa->completed_bkref_map |= 1 << cur_nsub;
+
+ tree = create_tree (dfa, tree, NULL, SUBEXP);
+ if (BE (tree == NULL, 0))
+ {
+ *err = REG_ESPACE;
+ return NULL;
+ }
+ tree->token.opr.idx = cur_nsub;
+ return tree;
+}
+
+/* This function parse repetition operators like "*", "+", "{1,3}" etc. */
+
+static bin_tree_t *
+parse_dup_op (bin_tree_t *elem, re_string_t *regexp, re_dfa_t *dfa,
+ re_token_t *token, reg_syntax_t syntax, reg_errcode_t *err)
+{
+ bin_tree_t *tree = NULL, *old_tree = NULL;
+ Idx i, start, end, start_idx = re_string_cur_idx (regexp);
+ re_token_t start_token = *token;
+
+ if (token->type == OP_OPEN_DUP_NUM)
+ {
+ end = 0;
+ start = fetch_number (regexp, token, syntax);
+ if (start == -1)
+ {
+ if (token->type == CHARACTER && token->opr.c == ',')
+ start = 0; /* We treat "{,m}" as "{0,m}". */
+ else
+ {
+ *err = REG_BADBR; /* <re>{} is invalid. */
+ return NULL;
+ }
+ }
+ if (BE (start != -2, 1))
+ {
+ /* We treat "{n}" as "{n,n}". */
+ end = ((token->type == OP_CLOSE_DUP_NUM) ? start
+ : ((token->type == CHARACTER && token->opr.c == ',')
+ ? fetch_number (regexp, token, syntax) : -2));
+ }
+ if (BE (start == -2 || end == -2, 0))
+ {
+ /* Invalid sequence. */
+ if (BE (!(syntax & RE_INVALID_INTERVAL_ORD), 0))
+ {
+ if (token->type == END_OF_RE)
+ *err = REG_EBRACE;
+ else
+ *err = REG_BADBR;
+
+ return NULL;
+ }
+
+ /* If the syntax bit is set, rollback. */
+ re_string_set_index (regexp, start_idx);
+ *token = start_token;
+ token->type = CHARACTER;
+ /* mb_partial and word_char bits should be already initialized by
+ peek_token. */
+ return elem;
+ }
+
+ if (BE ((end != -1 && start > end)
+ || token->type != OP_CLOSE_DUP_NUM, 0))
+ {
+ /* First number greater than second. */
+ *err = REG_BADBR;
+ return NULL;
+ }
+
+ if (BE (RE_DUP_MAX < (end == -1 ? start : end), 0))
+ {
+ *err = REG_ESIZE;
+ return NULL;
+ }
+ }
+ else
+ {
+ start = (token->type == OP_DUP_PLUS) ? 1 : 0;
+ end = (token->type == OP_DUP_QUESTION) ? 1 : -1;
+ }
+
+ fetch_token (token, regexp, syntax);
+
+ if (BE (elem == NULL, 0))
+ return NULL;
+ if (BE (start == 0 && end == 0, 0))
+ {
+ postorder (elem, free_tree, NULL);
+ return NULL;
+ }
+
+ /* Extract "<re>{n,m}" to "<re><re>...<re><re>{0,<m-n>}". */
+ if (BE (start > 0, 0))
+ {
+ tree = elem;
+ for (i = 2; i <= start; ++i)
+ {
+ elem = duplicate_tree (elem, dfa);
+ tree = create_tree (dfa, tree, elem, CONCAT);
+ if (BE (elem == NULL || tree == NULL, 0))
+ goto parse_dup_op_espace;
+ }
+
+ if (start == end)
+ return tree;
+
+ /* Duplicate ELEM before it is marked optional. */
+ elem = duplicate_tree (elem, dfa);
+ if (BE (elem == NULL, 0))
+ goto parse_dup_op_espace;
+ old_tree = tree;
+ }
+ else
+ old_tree = NULL;
+
+ if (elem->token.type == SUBEXP)
+ {
+ uintptr_t subidx = elem->token.opr.idx;
+ postorder (elem, mark_opt_subexp, (void *) subidx);
+ }
+
+ tree = create_tree (dfa, elem, NULL,
+ (end == -1 ? OP_DUP_ASTERISK : OP_ALT));
+ if (BE (tree == NULL, 0))
+ goto parse_dup_op_espace;
+
+ /* This loop is actually executed only when end != -1,
+ to rewrite <re>{0,n} as (<re>(<re>...<re>?)?)?... We have
+ already created the start+1-th copy. */
+ if (TYPE_SIGNED (Idx) || end != -1)
+ for (i = start + 2; i <= end; ++i)
+ {
+ elem = duplicate_tree (elem, dfa);
+ tree = create_tree (dfa, tree, elem, CONCAT);
+ if (BE (elem == NULL || tree == NULL, 0))
+ goto parse_dup_op_espace;
+
+ tree = create_tree (dfa, tree, NULL, OP_ALT);
+ if (BE (tree == NULL, 0))
+ goto parse_dup_op_espace;
+ }
+
+ if (old_tree)
+ tree = create_tree (dfa, old_tree, tree, CONCAT);
+
+ return tree;
+
+ parse_dup_op_espace:
+ *err = REG_ESPACE;
+ return NULL;
+}
+
+/* Size of the names for collating symbol/equivalence_class/character_class.
+ I'm not sure, but maybe enough. */
+#define BRACKET_NAME_BUF_SIZE 32
+
+#ifndef _LIBC
+
+# ifdef RE_ENABLE_I18N
+/* Convert the byte B to the corresponding wide character. In a
+ unibyte locale, treat B as itself if it is an encoding error.
+ In a multibyte locale, return WEOF if B is an encoding error. */
+static wint_t
+parse_byte (unsigned char b, re_charset_t *mbcset)
+{
+ wint_t wc = __btowc (b);
+ return wc == WEOF && !mbcset ? b : wc;
+}
+#endif
+
+ /* Local function for parse_bracket_exp only used in case of NOT _LIBC.
+ Build the range expression which starts from START_ELEM, and ends
+ at END_ELEM. The result are written to MBCSET and SBCSET.
+ RANGE_ALLOC is the allocated size of mbcset->range_starts, and
+ mbcset->range_ends, is a pointer argument since we may
+ update it. */
+
+static reg_errcode_t
+# ifdef RE_ENABLE_I18N
+build_range_exp (const reg_syntax_t syntax,
+ bitset_t sbcset,
+ re_charset_t *mbcset,
+ Idx *range_alloc,
+ const bracket_elem_t *start_elem,
+ const bracket_elem_t *end_elem)
+# else /* not RE_ENABLE_I18N */
+build_range_exp (const reg_syntax_t syntax,
+ bitset_t sbcset,
+ const bracket_elem_t *start_elem,
+ const bracket_elem_t *end_elem)
+# endif /* not RE_ENABLE_I18N */
+{
+ unsigned int start_ch, end_ch;
+ /* Equivalence Classes and Character Classes can't be a range start/end. */
+ if (BE (start_elem->type == EQUIV_CLASS || start_elem->type == CHAR_CLASS
+ || end_elem->type == EQUIV_CLASS || end_elem->type == CHAR_CLASS,
+ 0))
+ return REG_ERANGE;
+
+ /* We can handle no multi character collating elements without libc
+ support. */
+ if (BE ((start_elem->type == COLL_SYM
+ && strlen ((char *) start_elem->opr.name) > 1)
+ || (end_elem->type == COLL_SYM
+ && strlen ((char *) end_elem->opr.name) > 1), 0))
+ return REG_ECOLLATE;
+
+# ifdef RE_ENABLE_I18N
+ {
+ wchar_t wc;
+ wint_t start_wc;
+ wint_t end_wc;
+
+ start_ch = ((start_elem->type == SB_CHAR) ? start_elem->opr.ch
+ : ((start_elem->type == COLL_SYM) ? start_elem->opr.name[0]
+ : 0));
+ end_ch = ((end_elem->type == SB_CHAR) ? end_elem->opr.ch
+ : ((end_elem->type == COLL_SYM) ? end_elem->opr.name[0]
+ : 0));
+ start_wc = ((start_elem->type == SB_CHAR || start_elem->type == COLL_SYM)
+ ? parse_byte (start_ch, mbcset) : start_elem->opr.wch);
+ end_wc = ((end_elem->type == SB_CHAR || end_elem->type == COLL_SYM)
+ ? parse_byte (end_ch, mbcset) : end_elem->opr.wch);
+ if (start_wc == WEOF || end_wc == WEOF)
+ return REG_ECOLLATE;
+ else if (BE ((syntax & RE_NO_EMPTY_RANGES) && start_wc > end_wc, 0))
+ return REG_ERANGE;
+
+ /* Got valid collation sequence values, add them as a new entry.
+ However, for !_LIBC we have no collation elements: if the
+ character set is single byte, the single byte character set
+ that we build below suffices. parse_bracket_exp passes
+ no MBCSET if dfa->mb_cur_max == 1. */
+ if (mbcset)
+ {
+ /* Check the space of the arrays. */
+ if (BE (*range_alloc == mbcset->nranges, 0))
+ {
+ /* There is not enough space, need realloc. */
+ wchar_t *new_array_start, *new_array_end;
+ Idx new_nranges;
+
+ /* +1 in case of mbcset->nranges is 0. */
+ new_nranges = 2 * mbcset->nranges + 1;
+ /* Use realloc since mbcset->range_starts and mbcset->range_ends
+ are NULL if *range_alloc == 0. */
+ new_array_start = re_realloc (mbcset->range_starts, wchar_t,
+ new_nranges);
+ new_array_end = re_realloc (mbcset->range_ends, wchar_t,
+ new_nranges);
+
+ if (BE (new_array_start == NULL || new_array_end == NULL, 0))
+ {
+ re_free (new_array_start);
+ re_free (new_array_end);
+ return REG_ESPACE;
+ }
+
+ mbcset->range_starts = new_array_start;
+ mbcset->range_ends = new_array_end;
+ *range_alloc = new_nranges;
+ }
+
+ mbcset->range_starts[mbcset->nranges] = start_wc;
+ mbcset->range_ends[mbcset->nranges++] = end_wc;
+ }
+
+ /* Build the table for single byte characters. */
+ for (wc = 0; wc < SBC_MAX; ++wc)
+ {
+ if (start_wc <= wc && wc <= end_wc)
+ bitset_set (sbcset, wc);
+ }
+ }
+# else /* not RE_ENABLE_I18N */
+ {
+ unsigned int ch;
+ start_ch = ((start_elem->type == SB_CHAR ) ? start_elem->opr.ch
+ : ((start_elem->type == COLL_SYM) ? start_elem->opr.name[0]
+ : 0));
+ end_ch = ((end_elem->type == SB_CHAR ) ? end_elem->opr.ch
+ : ((end_elem->type == COLL_SYM) ? end_elem->opr.name[0]
+ : 0));
+ if (start_ch > end_ch)
+ return REG_ERANGE;
+ /* Build the table for single byte characters. */
+ for (ch = 0; ch < SBC_MAX; ++ch)
+ if (start_ch <= ch && ch <= end_ch)
+ bitset_set (sbcset, ch);
+ }
+# endif /* not RE_ENABLE_I18N */
+ return REG_NOERROR;
+}
+#endif /* not _LIBC */
+
+#ifndef _LIBC
+/* Helper function for parse_bracket_exp only used in case of NOT _LIBC..
+ Build the collating element which is represented by NAME.
+ The result are written to MBCSET and SBCSET.
+ COLL_SYM_ALLOC is the allocated size of mbcset->coll_sym, is a
+ pointer argument since we may update it. */
+
+static reg_errcode_t
+# ifdef RE_ENABLE_I18N
+build_collating_symbol (bitset_t sbcset, re_charset_t *mbcset,
+ Idx *coll_sym_alloc, const unsigned char *name)
+# else /* not RE_ENABLE_I18N */
+build_collating_symbol (bitset_t sbcset, const unsigned char *name)
+# endif /* not RE_ENABLE_I18N */
+{
+ size_t name_len = strlen ((const char *) name);
+ if (BE (name_len != 1, 0))
+ return REG_ECOLLATE;
+ else
+ {
+ bitset_set (sbcset, name[0]);
+ return REG_NOERROR;
+ }
+}
+#endif /* not _LIBC */
+
+/* This function parse bracket expression like "[abc]", "[a-c]",
+ "[[.a-a.]]" etc. */
+
+static bin_tree_t *
+parse_bracket_exp (re_string_t *regexp, re_dfa_t *dfa, re_token_t *token,
+ reg_syntax_t syntax, reg_errcode_t *err)
+{
+#ifdef _LIBC
+ const unsigned char *collseqmb;
+ const char *collseqwc;
+ uint32_t nrules;
+ int32_t table_size;
+ const int32_t *symb_table;
+ const unsigned char *extra;
+
+ /* Local function for parse_bracket_exp used in _LIBC environment.
+ Seek the collating symbol entry corresponding to NAME.
+ Return the index of the symbol in the SYMB_TABLE,
+ or -1 if not found. */
+
+ auto inline int32_t
+ __attribute__ ((always_inline))
+ seek_collating_symbol_entry (const unsigned char *name, size_t name_len)
+ {
+ int32_t elem;
+
+ for (elem = 0; elem < table_size; elem++)
+ if (symb_table[2 * elem] != 0)
+ {
+ int32_t idx = symb_table[2 * elem + 1];
+ /* Skip the name of collating element name. */
+ idx += 1 + extra[idx];
+ if (/* Compare the length of the name. */
+ name_len == extra[idx]
+ /* Compare the name. */
+ && memcmp (name, &extra[idx + 1], name_len) == 0)
+ /* Yep, this is the entry. */
+ return elem;
+ }
+ return -1;
+ }
+
+ /* Local function for parse_bracket_exp used in _LIBC environment.
+ Look up the collation sequence value of BR_ELEM.
+ Return the value if succeeded, UINT_MAX otherwise. */
+
+ auto inline unsigned int
+ __attribute__ ((always_inline))
+ lookup_collation_sequence_value (bracket_elem_t *br_elem)
+ {
+ if (br_elem->type == SB_CHAR)
+ {
+ /*
+ if (MB_CUR_MAX == 1)
+ */
+ if (nrules == 0)
+ return collseqmb[br_elem->opr.ch];
+ else
+ {
+ wint_t wc = __btowc (br_elem->opr.ch);
+ return __collseq_table_lookup (collseqwc, wc);
+ }
+ }
+ else if (br_elem->type == MB_CHAR)
+ {
+ if (nrules != 0)
+ return __collseq_table_lookup (collseqwc, br_elem->opr.wch);
+ }
+ else if (br_elem->type == COLL_SYM)
+ {
+ size_t sym_name_len = strlen ((char *) br_elem->opr.name);
+ if (nrules != 0)
+ {
+ int32_t elem, idx;
+ elem = seek_collating_symbol_entry (br_elem->opr.name,
+ sym_name_len);
+ if (elem != -1)
+ {
+ /* We found the entry. */
+ idx = symb_table[2 * elem + 1];
+ /* Skip the name of collating element name. */
+ idx += 1 + extra[idx];
+ /* Skip the byte sequence of the collating element. */
+ idx += 1 + extra[idx];
+ /* Adjust for the alignment. */
+ idx = (idx + 3) & ~3;
+ /* Skip the multibyte collation sequence value. */
+ idx += sizeof (unsigned int);
+ /* Skip the wide char sequence of the collating element. */
+ idx += sizeof (unsigned int) *
+ (1 + *(unsigned int *) (extra + idx));
+ /* Return the collation sequence value. */
+ return *(unsigned int *) (extra + idx);
+ }
+ else if (sym_name_len == 1)
+ {
+ /* No valid character. Match it as a single byte
+ character. */
+ return collseqmb[br_elem->opr.name[0]];
+ }
+ }
+ else if (sym_name_len == 1)
+ return collseqmb[br_elem->opr.name[0]];
+ }
+ return UINT_MAX;
+ }
+
+ /* Local function for parse_bracket_exp used in _LIBC environment.
+ Build the range expression which starts from START_ELEM, and ends
+ at END_ELEM. The result are written to MBCSET and SBCSET.
+ RANGE_ALLOC is the allocated size of mbcset->range_starts, and
+ mbcset->range_ends, is a pointer argument since we may
+ update it. */
+
+ auto inline reg_errcode_t
+ __attribute__ ((always_inline))
+ build_range_exp (bitset_t sbcset, re_charset_t *mbcset, int *range_alloc,
+ bracket_elem_t *start_elem, bracket_elem_t *end_elem)
+ {
+ unsigned int ch;
+ uint32_t start_collseq;
+ uint32_t end_collseq;
+
+ /* Equivalence Classes and Character Classes can't be a range
+ start/end. */
+ if (BE (start_elem->type == EQUIV_CLASS || start_elem->type == CHAR_CLASS
+ || end_elem->type == EQUIV_CLASS || end_elem->type == CHAR_CLASS,
+ 0))
+ return REG_ERANGE;
+
+ /* FIXME: Implement rational ranges here, too. */
+ start_collseq = lookup_collation_sequence_value (start_elem);
+ end_collseq = lookup_collation_sequence_value (end_elem);
+ /* Check start/end collation sequence values. */
+ if (BE (start_collseq == UINT_MAX || end_collseq == UINT_MAX, 0))
+ return REG_ECOLLATE;
+ if (BE ((syntax & RE_NO_EMPTY_RANGES) && start_collseq > end_collseq, 0))
+ return REG_ERANGE;
+
+ /* Got valid collation sequence values, add them as a new entry.
+ However, if we have no collation elements, and the character set
+ is single byte, the single byte character set that we
+ build below suffices. */
+ if (nrules > 0 || dfa->mb_cur_max > 1)
+ {
+ /* Check the space of the arrays. */
+ if (BE (*range_alloc == mbcset->nranges, 0))
+ {
+ /* There is not enough space, need realloc. */
+ uint32_t *new_array_start;
+ uint32_t *new_array_end;
+ Idx new_nranges;
+
+ /* +1 in case of mbcset->nranges is 0. */
+ new_nranges = 2 * mbcset->nranges + 1;
+ new_array_start = re_realloc (mbcset->range_starts, uint32_t,
+ new_nranges);
+ new_array_end = re_realloc (mbcset->range_ends, uint32_t,
+ new_nranges);
+
+ if (BE (new_array_start == NULL || new_array_end == NULL, 0))
+ return REG_ESPACE;
+
+ mbcset->range_starts = new_array_start;
+ mbcset->range_ends = new_array_end;
+ *range_alloc = new_nranges;
+ }
+
+ mbcset->range_starts[mbcset->nranges] = start_collseq;
+ mbcset->range_ends[mbcset->nranges++] = end_collseq;
+ }
+
+ /* Build the table for single byte characters. */
+ for (ch = 0; ch < SBC_MAX; ch++)
+ {
+ uint32_t ch_collseq;
+ /*
+ if (MB_CUR_MAX == 1)
+ */
+ if (nrules == 0)
+ ch_collseq = collseqmb[ch];
+ else
+ ch_collseq = __collseq_table_lookup (collseqwc, __btowc (ch));
+ if (start_collseq <= ch_collseq && ch_collseq <= end_collseq)
+ bitset_set (sbcset, ch);
+ }
+ return REG_NOERROR;
+ }
+
+ /* Local function for parse_bracket_exp used in _LIBC environment.
+ Build the collating element which is represented by NAME.
+ The result are written to MBCSET and SBCSET.
+ COLL_SYM_ALLOC is the allocated size of mbcset->coll_sym, is a
+ pointer argument since we may update it. */
+
+ auto inline reg_errcode_t
+ __attribute__ ((always_inline))
+ build_collating_symbol (bitset_t sbcset, re_charset_t *mbcset,
+ Idx *coll_sym_alloc, const unsigned char *name)
+ {
+ int32_t elem, idx;
+ size_t name_len = strlen ((const char *) name);
+ if (nrules != 0)
+ {
+ elem = seek_collating_symbol_entry (name, name_len);
+ if (elem != -1)
+ {
+ /* We found the entry. */
+ idx = symb_table[2 * elem + 1];
+ /* Skip the name of collating element name. */
+ idx += 1 + extra[idx];
+ }
+ else if (name_len == 1)
+ {
+ /* No valid character, treat it as a normal
+ character. */
+ bitset_set (sbcset, name[0]);
+ return REG_NOERROR;
+ }
+ else
+ return REG_ECOLLATE;
+
+ /* Got valid collation sequence, add it as a new entry. */
+ /* Check the space of the arrays. */
+ if (BE (*coll_sym_alloc == mbcset->ncoll_syms, 0))
+ {
+ /* Not enough, realloc it. */
+ /* +1 in case of mbcset->ncoll_syms is 0. */
+ Idx new_coll_sym_alloc = 2 * mbcset->ncoll_syms + 1;
+ /* Use realloc since mbcset->coll_syms is NULL
+ if *alloc == 0. */
+ int32_t *new_coll_syms = re_realloc (mbcset->coll_syms, int32_t,
+ new_coll_sym_alloc);
+ if (BE (new_coll_syms == NULL, 0))
+ return REG_ESPACE;
+ mbcset->coll_syms = new_coll_syms;
+ *coll_sym_alloc = new_coll_sym_alloc;
+ }
+ mbcset->coll_syms[mbcset->ncoll_syms++] = idx;
+ return REG_NOERROR;
+ }
+ else
+ {
+ if (BE (name_len != 1, 0))
+ return REG_ECOLLATE;
+ else
+ {
+ bitset_set (sbcset, name[0]);
+ return REG_NOERROR;
+ }
+ }
+ }
+#endif
+
+ re_token_t br_token;
+ re_bitset_ptr_t sbcset;
+#ifdef RE_ENABLE_I18N
+ re_charset_t *mbcset;
+ Idx coll_sym_alloc = 0, range_alloc = 0, mbchar_alloc = 0;
+ Idx equiv_class_alloc = 0, char_class_alloc = 0;
+#endif /* not RE_ENABLE_I18N */
+ bool non_match = false;
+ bin_tree_t *work_tree;
+ int token_len;
+ bool first_round = true;
+#ifdef _LIBC
+ collseqmb = (const unsigned char *)
+ _NL_CURRENT (LC_COLLATE, _NL_COLLATE_COLLSEQMB);
+ nrules = _NL_CURRENT_WORD (LC_COLLATE, _NL_COLLATE_NRULES);
+ if (nrules)
+ {
+ /*
+ if (MB_CUR_MAX > 1)
+ */
+ collseqwc = _NL_CURRENT (LC_COLLATE, _NL_COLLATE_COLLSEQWC);
+ table_size = _NL_CURRENT_WORD (LC_COLLATE, _NL_COLLATE_SYMB_HASH_SIZEMB);
+ symb_table = (const int32_t *) _NL_CURRENT (LC_COLLATE,
+ _NL_COLLATE_SYMB_TABLEMB);
+ extra = (const unsigned char *) _NL_CURRENT (LC_COLLATE,
+ _NL_COLLATE_SYMB_EXTRAMB);
+ }
+#endif
+ sbcset = (re_bitset_ptr_t) calloc (sizeof (bitset_t), 1);
+#ifdef RE_ENABLE_I18N
+ mbcset = (re_charset_t *) calloc (sizeof (re_charset_t), 1);
+#endif /* RE_ENABLE_I18N */
+#ifdef RE_ENABLE_I18N
+ if (BE (sbcset == NULL || mbcset == NULL, 0))
+#else
+ if (BE (sbcset == NULL, 0))
+#endif /* RE_ENABLE_I18N */
+ {
+ re_free (sbcset);
+#ifdef RE_ENABLE_I18N
+ re_free (mbcset);
+#endif
+ *err = REG_ESPACE;
+ return NULL;
+ }
+
+ token_len = peek_token_bracket (token, regexp, syntax);
+ if (BE (token->type == END_OF_RE, 0))
+ {
+ *err = REG_BADPAT;
+ goto parse_bracket_exp_free_return;
+ }
+ if (token->type == OP_NON_MATCH_LIST)
+ {
+#ifdef RE_ENABLE_I18N
+ mbcset->non_match = 1;
+#endif /* not RE_ENABLE_I18N */
+ non_match = true;
+ if (syntax & RE_HAT_LISTS_NOT_NEWLINE)
+ bitset_set (sbcset, '\n');
+ re_string_skip_bytes (regexp, token_len); /* Skip a token. */
+ token_len = peek_token_bracket (token, regexp, syntax);
+ if (BE (token->type == END_OF_RE, 0))
+ {
+ *err = REG_BADPAT;
+ goto parse_bracket_exp_free_return;
+ }
+ }
+
+ /* We treat the first ']' as a normal character. */
+ if (token->type == OP_CLOSE_BRACKET)
+ token->type = CHARACTER;
+
+ while (1)
+ {
+ bracket_elem_t start_elem, end_elem;
+ unsigned char start_name_buf[BRACKET_NAME_BUF_SIZE];
+ unsigned char end_name_buf[BRACKET_NAME_BUF_SIZE];
+ reg_errcode_t ret;
+ int token_len2 = 0;
+ bool is_range_exp = false;
+ re_token_t token2;
+
+ start_elem.opr.name = start_name_buf;
+ start_elem.type = COLL_SYM;
+ ret = parse_bracket_element (&start_elem, regexp, token, token_len, dfa,
+ syntax, first_round);
+ if (BE (ret != REG_NOERROR, 0))
+ {
+ *err = ret;
+ goto parse_bracket_exp_free_return;
+ }
+ first_round = false;
+
+ /* Get information about the next token. We need it in any case. */
+ token_len = peek_token_bracket (token, regexp, syntax);
+
+ /* Do not check for ranges if we know they are not allowed. */
+ if (start_elem.type != CHAR_CLASS && start_elem.type != EQUIV_CLASS)
+ {
+ if (BE (token->type == END_OF_RE, 0))
+ {
+ *err = REG_EBRACK;
+ goto parse_bracket_exp_free_return;
+ }
+ if (token->type == OP_CHARSET_RANGE)
+ {
+ re_string_skip_bytes (regexp, token_len); /* Skip '-'. */
+ token_len2 = peek_token_bracket (&token2, regexp, syntax);
+ if (BE (token2.type == END_OF_RE, 0))
+ {
+ *err = REG_EBRACK;
+ goto parse_bracket_exp_free_return;
+ }
+ if (token2.type == OP_CLOSE_BRACKET)
+ {
+ /* We treat the last '-' as a normal character. */
+ re_string_skip_bytes (regexp, -token_len);
+ token->type = CHARACTER;
+ }
+ else
+ is_range_exp = true;
+ }
+ }
+
+ if (is_range_exp == true)
+ {
+ end_elem.opr.name = end_name_buf;
+ end_elem.type = COLL_SYM;
+ ret = parse_bracket_element (&end_elem, regexp, &token2, token_len2,
+ dfa, syntax, true);
+ if (BE (ret != REG_NOERROR, 0))
+ {
+ *err = ret;
+ goto parse_bracket_exp_free_return;
+ }
+
+ token_len = peek_token_bracket (token, regexp, syntax);
+
+#ifdef _LIBC
+ *err = build_range_exp (sbcset, mbcset, &range_alloc,
+ &start_elem, &end_elem);
+#else
+# ifdef RE_ENABLE_I18N
+ *err = build_range_exp (syntax, sbcset,
+ dfa->mb_cur_max > 1 ? mbcset : NULL,
+ &range_alloc, &start_elem, &end_elem);
+# else
+ *err = build_range_exp (syntax, sbcset, &start_elem, &end_elem);
+# endif
+#endif /* RE_ENABLE_I18N */
+ if (BE (*err != REG_NOERROR, 0))
+ goto parse_bracket_exp_free_return;
+ }
+ else
+ {
+ switch (start_elem.type)
+ {
+ case SB_CHAR:
+ bitset_set (sbcset, start_elem.opr.ch);
+ break;
+#ifdef RE_ENABLE_I18N
+ case MB_CHAR:
+ /* Check whether the array has enough space. */
+ if (BE (mbchar_alloc == mbcset->nmbchars, 0))
+ {
+ wchar_t *new_mbchars;
+ /* Not enough, realloc it. */
+ /* +1 in case of mbcset->nmbchars is 0. */
+ mbchar_alloc = 2 * mbcset->nmbchars + 1;
+ /* Use realloc since array is NULL if *alloc == 0. */
+ new_mbchars = re_realloc (mbcset->mbchars, wchar_t,
+ mbchar_alloc);
+ if (BE (new_mbchars == NULL, 0))
+ goto parse_bracket_exp_espace;
+ mbcset->mbchars = new_mbchars;
+ }
+ mbcset->mbchars[mbcset->nmbchars++] = start_elem.opr.wch;
+ break;
+#endif /* RE_ENABLE_I18N */
+ case EQUIV_CLASS:
+ *err = build_equiv_class (sbcset,
+#ifdef RE_ENABLE_I18N
+ mbcset, &equiv_class_alloc,
+#endif /* RE_ENABLE_I18N */
+ start_elem.opr.name);
+ if (BE (*err != REG_NOERROR, 0))
+ goto parse_bracket_exp_free_return;
+ break;
+ case COLL_SYM:
+ *err = build_collating_symbol (sbcset,
+#ifdef RE_ENABLE_I18N
+ mbcset, &coll_sym_alloc,
+#endif /* RE_ENABLE_I18N */
+ start_elem.opr.name);
+ if (BE (*err != REG_NOERROR, 0))
+ goto parse_bracket_exp_free_return;
+ break;
+ case CHAR_CLASS:
+ *err = build_charclass (regexp->trans, sbcset,
+#ifdef RE_ENABLE_I18N
+ mbcset, &char_class_alloc,
+#endif /* RE_ENABLE_I18N */
+ (const char *) start_elem.opr.name,
+ syntax);
+ if (BE (*err != REG_NOERROR, 0))
+ goto parse_bracket_exp_free_return;
+ break;
+ default:
+ assert (0);
+ break;
+ }
+ }
+ if (BE (token->type == END_OF_RE, 0))
+ {
+ *err = REG_EBRACK;
+ goto parse_bracket_exp_free_return;
+ }
+ if (token->type == OP_CLOSE_BRACKET)
+ break;
+ }
+
+ re_string_skip_bytes (regexp, token_len); /* Skip a token. */
+
+ /* If it is non-matching list. */
+ if (non_match)
+ bitset_not (sbcset);
+
+#ifdef RE_ENABLE_I18N
+ /* Ensure only single byte characters are set. */
+ if (dfa->mb_cur_max > 1)
+ bitset_mask (sbcset, dfa->sb_char);
+
+ if (mbcset->nmbchars || mbcset->ncoll_syms || mbcset->nequiv_classes
+ || mbcset->nranges || (dfa->mb_cur_max > 1 && (mbcset->nchar_classes
+ || mbcset->non_match)))
+ {
+ bin_tree_t *mbc_tree;
+ int sbc_idx;
+ /* Build a tree for complex bracket. */
+ dfa->has_mb_node = 1;
+ br_token.type = COMPLEX_BRACKET;
+ br_token.opr.mbcset = mbcset;
+ mbc_tree = create_token_tree (dfa, NULL, NULL, &br_token);
+ if (BE (mbc_tree == NULL, 0))
+ goto parse_bracket_exp_espace;
+ for (sbc_idx = 0; sbc_idx < BITSET_WORDS; ++sbc_idx)
+ if (sbcset[sbc_idx])
+ break;
+ /* If there are no bits set in sbcset, there is no point
+ of having both SIMPLE_BRACKET and COMPLEX_BRACKET. */
+ if (sbc_idx < BITSET_WORDS)
+ {
+ /* Build a tree for simple bracket. */
+ br_token.type = SIMPLE_BRACKET;
+ br_token.opr.sbcset = sbcset;
+ work_tree = create_token_tree (dfa, NULL, NULL, &br_token);
+ if (BE (work_tree == NULL, 0))
+ goto parse_bracket_exp_espace;
+
+ /* Then join them by ALT node. */
+ work_tree = create_tree (dfa, work_tree, mbc_tree, OP_ALT);
+ if (BE (work_tree == NULL, 0))
+ goto parse_bracket_exp_espace;
+ }
+ else
+ {
+ re_free (sbcset);
+ work_tree = mbc_tree;
+ }
+ }
+ else
+#endif /* not RE_ENABLE_I18N */
+ {
+#ifdef RE_ENABLE_I18N
+ free_charset (mbcset);
+#endif
+ /* Build a tree for simple bracket. */
+ br_token.type = SIMPLE_BRACKET;
+ br_token.opr.sbcset = sbcset;
+ work_tree = create_token_tree (dfa, NULL, NULL, &br_token);
+ if (BE (work_tree == NULL, 0))
+ goto parse_bracket_exp_espace;
+ }
+ return work_tree;
+
+ parse_bracket_exp_espace:
+ *err = REG_ESPACE;
+ parse_bracket_exp_free_return:
+ re_free (sbcset);
+#ifdef RE_ENABLE_I18N
+ free_charset (mbcset);
+#endif /* RE_ENABLE_I18N */
+ return NULL;
+}
+
+/* Parse an element in the bracket expression. */
+
+static reg_errcode_t
+parse_bracket_element (bracket_elem_t *elem, re_string_t *regexp,
+ re_token_t *token, int token_len, re_dfa_t *dfa,
+ reg_syntax_t syntax, bool accept_hyphen)
+{
+#ifdef RE_ENABLE_I18N
+ int cur_char_size;
+ cur_char_size = re_string_char_size_at (regexp, re_string_cur_idx (regexp));
+ if (cur_char_size > 1)
+ {
+ elem->type = MB_CHAR;
+ elem->opr.wch = re_string_wchar_at (regexp, re_string_cur_idx (regexp));
+ re_string_skip_bytes (regexp, cur_char_size);
+ return REG_NOERROR;
+ }
+#endif /* RE_ENABLE_I18N */
+ re_string_skip_bytes (regexp, token_len); /* Skip a token. */
+ if (token->type == OP_OPEN_COLL_ELEM || token->type == OP_OPEN_CHAR_CLASS
+ || token->type == OP_OPEN_EQUIV_CLASS)
+ return parse_bracket_symbol (elem, regexp, token);
+ if (BE (token->type == OP_CHARSET_RANGE, 0) && !accept_hyphen)
+ {
+ /* A '-' must only appear as anything but a range indicator before
+ the closing bracket. Everything else is an error. */
+ re_token_t token2;
+ (void) peek_token_bracket (&token2, regexp, syntax);
+ if (token2.type != OP_CLOSE_BRACKET)
+ /* The actual error value is not standardized since this whole
+ case is undefined. But ERANGE makes good sense. */
+ return REG_ERANGE;
+ }
+ elem->type = SB_CHAR;
+ elem->opr.ch = token->opr.c;
+ return REG_NOERROR;
+}
+
+/* Parse a bracket symbol in the bracket expression. Bracket symbols are
+ such as [:<character_class>:], [.<collating_element>.], and
+ [=<equivalent_class>=]. */
+
+static reg_errcode_t
+parse_bracket_symbol (bracket_elem_t *elem, re_string_t *regexp,
+ re_token_t *token)
+{
+ unsigned char ch, delim = token->opr.c;
+ int i = 0;
+ if (re_string_eoi(regexp))
+ return REG_EBRACK;
+ 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;
+ }
+ re_string_skip_bytes (regexp, 1);
+ elem->opr.name[i] = '\0';
+ switch (token->type)
+ {
+ case OP_OPEN_COLL_ELEM:
+ elem->type = COLL_SYM;
+ break;
+ case OP_OPEN_EQUIV_CLASS:
+ elem->type = EQUIV_CLASS;
+ break;
+ case OP_OPEN_CHAR_CLASS:
+ elem->type = CHAR_CLASS;
+ break;
+ default:
+ break;
+ }
+ return REG_NOERROR;
+}
+
+ /* Helper function for parse_bracket_exp.
+ Build the equivalence class which is represented by NAME.
+ The result are written to MBCSET and SBCSET.
+ EQUIV_CLASS_ALLOC is the allocated size of mbcset->equiv_classes,
+ is a pointer argument since we may update it. */
+
+static reg_errcode_t
+#ifdef RE_ENABLE_I18N
+build_equiv_class (bitset_t sbcset, re_charset_t *mbcset,
+ Idx *equiv_class_alloc, const unsigned char *name)
+#else /* not RE_ENABLE_I18N */
+build_equiv_class (bitset_t sbcset, const unsigned char *name)
+#endif /* not RE_ENABLE_I18N */
+{
+#ifdef _LIBC
+ uint32_t nrules = _NL_CURRENT_WORD (LC_COLLATE, _NL_COLLATE_NRULES);
+ if (nrules != 0)
+ {
+ const int32_t *table, *indirect;
+ const unsigned char *weights, *extra, *cp;
+ unsigned char char_buf[2];
+ int32_t idx1, idx2;
+ unsigned int ch;
+ size_t len;
+ /* Calculate the index for equivalence class. */
+ cp = name;
+ table = (const int32_t *) _NL_CURRENT (LC_COLLATE, _NL_COLLATE_TABLEMB);
+ weights = (const unsigned char *) _NL_CURRENT (LC_COLLATE,
+ _NL_COLLATE_WEIGHTMB);
+ extra = (const unsigned char *) _NL_CURRENT (LC_COLLATE,
+ _NL_COLLATE_EXTRAMB);
+ indirect = (const int32_t *) _NL_CURRENT (LC_COLLATE,
+ _NL_COLLATE_INDIRECTMB);
+ idx1 = findidx (table, indirect, extra, &cp, -1);
+ if (BE (idx1 == 0 || *cp != '\0', 0))
+ /* This isn't a valid character. */
+ return REG_ECOLLATE;
+
+ /* Build single byte matching table for this equivalence class. */
+ len = weights[idx1 & 0xffffff];
+ for (ch = 0; ch < SBC_MAX; ++ch)
+ {
+ char_buf[0] = ch;
+ cp = char_buf;
+ idx2 = findidx (table, indirect, extra, &cp, 1);
+/*
+ idx2 = table[ch];
+*/
+ if (idx2 == 0)
+ /* This isn't a valid character. */
+ continue;
+ /* Compare only if the length matches and the collation rule
+ index is the same. */
+ if (len == weights[idx2 & 0xffffff] && (idx1 >> 24) == (idx2 >> 24))
+ {
+ int cnt = 0;
+
+ while (cnt <= len &&
+ weights[(idx1 & 0xffffff) + 1 + cnt]
+ == weights[(idx2 & 0xffffff) + 1 + cnt])
+ ++cnt;
+
+ if (cnt > len)
+ bitset_set (sbcset, ch);
+ }
+ }
+ /* Check whether the array has enough space. */
+ if (BE (*equiv_class_alloc == mbcset->nequiv_classes, 0))
+ {
+ /* Not enough, realloc it. */
+ /* +1 in case of mbcset->nequiv_classes is 0. */
+ Idx new_equiv_class_alloc = 2 * mbcset->nequiv_classes + 1;
+ /* Use realloc since the array is NULL if *alloc == 0. */
+ int32_t *new_equiv_classes = re_realloc (mbcset->equiv_classes,
+ int32_t,
+ new_equiv_class_alloc);
+ if (BE (new_equiv_classes == NULL, 0))
+ return REG_ESPACE;
+ mbcset->equiv_classes = new_equiv_classes;
+ *equiv_class_alloc = new_equiv_class_alloc;
+ }
+ mbcset->equiv_classes[mbcset->nequiv_classes++] = idx1;
+ }
+ else
+#endif /* _LIBC */
+ {
+ if (BE (strlen ((const char *) name) != 1, 0))
+ return REG_ECOLLATE;
+ bitset_set (sbcset, *name);
+ }
+ return REG_NOERROR;
+}
+
+ /* Helper function for parse_bracket_exp.
+ Build the character class which is represented by NAME.
+ The result are written to MBCSET and SBCSET.
+ CHAR_CLASS_ALLOC is the allocated size of mbcset->char_classes,
+ is a pointer argument since we may update it. */
+
+static reg_errcode_t
+#ifdef RE_ENABLE_I18N
+build_charclass (RE_TRANSLATE_TYPE trans, bitset_t sbcset,
+ re_charset_t *mbcset, Idx *char_class_alloc,
+ const char *class_name, reg_syntax_t syntax)
+#else /* not RE_ENABLE_I18N */
+build_charclass (RE_TRANSLATE_TYPE trans, bitset_t sbcset,
+ const char *class_name, reg_syntax_t syntax)
+#endif /* not RE_ENABLE_I18N */
+{
+ int i;
+ const char *name = class_name;
+
+ /* In case of REG_ICASE "upper" and "lower" match the both of
+ upper and lower cases. */
+ if ((syntax & RE_ICASE)
+ && (strcmp (name, "upper") == 0 || strcmp (name, "lower") == 0))
+ name = "alpha";
+
+#ifdef RE_ENABLE_I18N
+ /* Check the space of the arrays. */
+ if (BE (*char_class_alloc == mbcset->nchar_classes, 0))
+ {
+ /* Not enough, realloc it. */
+ /* +1 in case of mbcset->nchar_classes is 0. */
+ Idx new_char_class_alloc = 2 * mbcset->nchar_classes + 1;
+ /* Use realloc since array is NULL if *alloc == 0. */
+ wctype_t *new_char_classes = re_realloc (mbcset->char_classes, wctype_t,
+ new_char_class_alloc);
+ if (BE (new_char_classes == NULL, 0))
+ return REG_ESPACE;
+ mbcset->char_classes = new_char_classes;
+ *char_class_alloc = new_char_class_alloc;
+ }
+ mbcset->char_classes[mbcset->nchar_classes++] = __wctype (name);
+#endif /* RE_ENABLE_I18N */
+
+#define BUILD_CHARCLASS_LOOP(ctype_func) \
+ do { \
+ if (BE (trans != NULL, 0)) \
+ { \
+ for (i = 0; i < SBC_MAX; ++i) \
+ if (ctype_func (i)) \
+ bitset_set (sbcset, trans[i]); \
+ } \
+ else \
+ { \
+ for (i = 0; i < SBC_MAX; ++i) \
+ if (ctype_func (i)) \
+ bitset_set (sbcset, i); \
+ } \
+ } while (0)
+
+ if (strcmp (name, "alnum") == 0)
+ BUILD_CHARCLASS_LOOP (isalnum);
+ else if (strcmp (name, "cntrl") == 0)
+ BUILD_CHARCLASS_LOOP (iscntrl);
+ else if (strcmp (name, "lower") == 0)
+ BUILD_CHARCLASS_LOOP (islower);
+ else if (strcmp (name, "space") == 0)
+ BUILD_CHARCLASS_LOOP (isspace);
+ else if (strcmp (name, "alpha") == 0)
+ BUILD_CHARCLASS_LOOP (isalpha);
+ else if (strcmp (name, "digit") == 0)
+ BUILD_CHARCLASS_LOOP (isdigit);
+ else if (strcmp (name, "print") == 0)
+ BUILD_CHARCLASS_LOOP (isprint);
+ else if (strcmp (name, "upper") == 0)
+ BUILD_CHARCLASS_LOOP (isupper);
+ else if (strcmp (name, "blank") == 0)
+ BUILD_CHARCLASS_LOOP (isblank);
+ else if (strcmp (name, "graph") == 0)
+ BUILD_CHARCLASS_LOOP (isgraph);
+ else if (strcmp (name, "punct") == 0)
+ BUILD_CHARCLASS_LOOP (ispunct);
+ else if (strcmp (name, "xdigit") == 0)
+ BUILD_CHARCLASS_LOOP (isxdigit);
+ else
+ return REG_ECTYPE;
+
+ return REG_NOERROR;
+}
+
+static bin_tree_t *
+build_charclass_op (re_dfa_t *dfa, RE_TRANSLATE_TYPE trans,
+ const char *class_name,
+ const char *extra, bool non_match,
+ reg_errcode_t *err)
+{
+ re_bitset_ptr_t sbcset;
+#ifdef RE_ENABLE_I18N
+ re_charset_t *mbcset;
+ Idx alloc = 0;
+#endif /* not RE_ENABLE_I18N */
+ reg_errcode_t ret;
+ re_token_t br_token;
+ bin_tree_t *tree;
+
+ sbcset = (re_bitset_ptr_t) calloc (sizeof (bitset_t), 1);
+ if (BE (sbcset == NULL, 0))
+ {
+ *err = REG_ESPACE;
+ return NULL;
+ }
+#ifdef RE_ENABLE_I18N
+ mbcset = (re_charset_t *) calloc (sizeof (re_charset_t), 1);
+ if (BE (mbcset == NULL, 0))
+ {
+ re_free (sbcset);
+ *err = REG_ESPACE;
+ return NULL;
+ }
+ mbcset->non_match = non_match;
+#endif /* RE_ENABLE_I18N */
+
+ /* We don't care the syntax in this case. */
+ ret = build_charclass (trans, sbcset,
+#ifdef RE_ENABLE_I18N
+ mbcset, &alloc,
+#endif /* RE_ENABLE_I18N */
+ class_name, 0);
+
+ if (BE (ret != REG_NOERROR, 0))
+ {
+ re_free (sbcset);
+#ifdef RE_ENABLE_I18N
+ free_charset (mbcset);
+#endif /* RE_ENABLE_I18N */
+ *err = ret;
+ return NULL;
+ }
+ /* \w match '_' also. */
+ for (; *extra; extra++)
+ bitset_set (sbcset, *extra);
+
+ /* If it is non-matching list. */
+ if (non_match)
+ bitset_not (sbcset);
+
+#ifdef RE_ENABLE_I18N
+ /* Ensure only single byte characters are set. */
+ if (dfa->mb_cur_max > 1)
+ bitset_mask (sbcset, dfa->sb_char);
+#endif
+
+ /* Build a tree for simple bracket. */
+#if defined GCC_LINT || defined lint
+ memset (&br_token, 0, sizeof br_token);
+#endif
+ br_token.type = SIMPLE_BRACKET;
+ br_token.opr.sbcset = sbcset;
+ tree = create_token_tree (dfa, NULL, NULL, &br_token);
+ if (BE (tree == NULL, 0))
+ goto build_word_op_espace;
+
+#ifdef RE_ENABLE_I18N
+ if (dfa->mb_cur_max > 1)
+ {
+ bin_tree_t *mbc_tree;
+ /* Build a tree for complex bracket. */
+ br_token.type = COMPLEX_BRACKET;
+ br_token.opr.mbcset = mbcset;
+ dfa->has_mb_node = 1;
+ mbc_tree = create_token_tree (dfa, NULL, NULL, &br_token);
+ if (BE (mbc_tree == NULL, 0))
+ goto build_word_op_espace;
+ /* Then join them by ALT node. */
+ tree = create_tree (dfa, tree, mbc_tree, OP_ALT);
+ if (BE (mbc_tree != NULL, 1))
+ return tree;
+ }
+ else
+ {
+ free_charset (mbcset);
+ return tree;
+ }
+#else /* not RE_ENABLE_I18N */
+ return tree;
+#endif /* not RE_ENABLE_I18N */
+
+ build_word_op_espace:
+ re_free (sbcset);
+#ifdef RE_ENABLE_I18N
+ free_charset (mbcset);
+#endif /* RE_ENABLE_I18N */
+ *err = REG_ESPACE;
+ return NULL;
+}
+
+/* This is intended for the expressions like "a{1,3}".
+ Fetch a number from 'input', and return the number.
+ Return -1 if the number field is empty like "{,1}".
+ Return RE_DUP_MAX + 1 if the number field is too large.
+ Return -2 if an error occurred. */
+
+static Idx
+fetch_number (re_string_t *input, re_token_t *token, reg_syntax_t syntax)
+{
+ Idx num = -1;
+ unsigned char c;
+ while (1)
+ {
+ fetch_token (token, input, syntax);
+ c = token->opr.c;
+ if (BE (token->type == END_OF_RE, 0))
+ return -2;
+ if (token->type == OP_CLOSE_DUP_NUM || c == ',')
+ break;
+ num = ((token->type != CHARACTER || c < '0' || '9' < c || num == -2)
+ ? -2
+ : num == -1
+ ? c - '0'
+ : MIN (RE_DUP_MAX + 1, num * 10 + c - '0'));
+ }
+ return num;
+}
+
+#ifdef RE_ENABLE_I18N
+static void
+free_charset (re_charset_t *cset)
+{
+ re_free (cset->mbchars);
+# ifdef _LIBC
+ re_free (cset->coll_syms);
+ re_free (cset->equiv_classes);
+ re_free (cset->range_starts);
+ re_free (cset->range_ends);
+# endif
+ re_free (cset->char_classes);
+ re_free (cset);
+}
+#endif /* RE_ENABLE_I18N */
+
+/* Functions for binary tree operation. */
+
+/* Create a tree node. */
+
+static bin_tree_t *
+create_tree (re_dfa_t *dfa, bin_tree_t *left, bin_tree_t *right,
+ re_token_type_t type)
+{
+ re_token_t t;
+#if defined GCC_LINT || defined lint
+ memset (&t, 0, sizeof t);
+#endif
+ t.type = type;
+ return create_token_tree (dfa, left, right, &t);
+}
+
+static bin_tree_t *
+create_token_tree (re_dfa_t *dfa, bin_tree_t *left, bin_tree_t *right,
+ const re_token_t *token)
+{
+ bin_tree_t *tree;
+ if (BE (dfa->str_tree_storage_idx == BIN_TREE_STORAGE_SIZE, 0))
+ {
+ bin_tree_storage_t *storage = re_malloc (bin_tree_storage_t, 1);
+
+ if (storage == NULL)
+ return NULL;
+ storage->next = dfa->str_tree_storage;
+ dfa->str_tree_storage = storage;
+ dfa->str_tree_storage_idx = 0;
+ }
+ tree = &dfa->str_tree_storage->data[dfa->str_tree_storage_idx++];
+
+ tree->parent = NULL;
+ tree->left = left;
+ tree->right = right;
+ tree->token = *token;
+ tree->token.duplicated = 0;
+ tree->token.opt_subexp = 0;
+ tree->first = NULL;
+ tree->next = NULL;
+ tree->node_idx = -1;
+
+ if (left != NULL)
+ left->parent = tree;
+ if (right != NULL)
+ right->parent = tree;
+ return tree;
+}
+
+/* Mark the tree SRC as an optional subexpression.
+ To be called from preorder or postorder. */
+
+static reg_errcode_t
+mark_opt_subexp (void *extra, bin_tree_t *node)
+{
+ Idx idx = (uintptr_t) extra;
+ if (node->token.type == SUBEXP && node->token.opr.idx == idx)
+ node->token.opt_subexp = 1;
+
+ return REG_NOERROR;
+}
+
+/* Free the allocated memory inside NODE. */
+
+static void
+free_token (re_token_t *node)
+{
+#ifdef RE_ENABLE_I18N
+ if (node->type == COMPLEX_BRACKET && node->duplicated == 0)
+ free_charset (node->opr.mbcset);
+ else
+#endif /* RE_ENABLE_I18N */
+ if (node->type == SIMPLE_BRACKET && node->duplicated == 0)
+ re_free (node->opr.sbcset);
+}
+
+/* Worker function for tree walking. Free the allocated memory inside NODE
+ and its children. */
+
+static reg_errcode_t
+free_tree (void *extra, bin_tree_t *node)
+{
+ free_token (&node->token);
+ return REG_NOERROR;
+}
+
+
+/* Duplicate the node SRC, and return new node. This is a preorder
+ visit similar to the one implemented by the generic visitor, but
+ we need more infrastructure to maintain two parallel trees --- so,
+ it's easier to duplicate. */
+
+static bin_tree_t *
+duplicate_tree (const bin_tree_t *root, re_dfa_t *dfa)
+{
+ const bin_tree_t *node;
+ bin_tree_t *dup_root;
+ bin_tree_t **p_new = &dup_root, *dup_node = root->parent;
+
+ for (node = root; ; )
+ {
+ /* Create a new tree and link it back to the current parent. */
+ *p_new = create_token_tree (dfa, NULL, NULL, &node->token);
+ if (*p_new == NULL)
+ return NULL;
+ (*p_new)->parent = dup_node;
+ (*p_new)->token.duplicated = 1;
+ dup_node = *p_new;
+
+ /* Go to the left node, or up and to the right. */
+ if (node->left)
+ {
+ node = node->left;
+ p_new = &dup_node->left;
+ }
+ else
+ {
+ const bin_tree_t *prev = NULL;
+ while (node->right == prev || node->right == NULL)
+ {
+ prev = node;
+ node = node->parent;
+ dup_node = dup_node->parent;
+ if (!node)
+ return dup_root;
+ }
+ node = node->right;
+ p_new = &dup_node->right;
+ }
+ }
+}
diff --git a/lib/regex.c b/lib/regex.c
new file mode 100644
index 0000000..499e1f0
--- /dev/null
+++ b/lib/regex.c
@@ -0,0 +1,81 @@
+/* Extended regular expression matching and search library.
+ Copyright (C) 2002-2018 Free Software Foundation, Inc.
+ This file is part of the GNU C Library.
+ Contributed by Isamu Hasegawa <address@hidden>.
+
+ The GNU C Library 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 3 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
+ General Public License for more details.
+
+ You should have received a copy of the GNU General Public
+ License along with the GNU C Library; if not, see
+ <https://www.gnu.org/licenses/>. */
+
+#ifndef _LIBC
+# include <config.h>
+
+# if (__GNUC__ == 4 && 6 <= __GNUC_MINOR__) || 4 < __GNUC__
+# pragma GCC diagnostic ignored "-Wsuggest-attribute=pure"
+# endif
+# if (__GNUC__ == 4 && 3 <= __GNUC_MINOR__) || 4 < __GNUC__
+# pragma GCC diagnostic ignored "-Wold-style-definition"
+# pragma GCC diagnostic ignored "-Wtype-limits"
+# endif
+#endif
+
+/* Make sure no one compiles this code with a C++ compiler. */
+#if defined __cplusplus && defined _LIBC
+# error "This is C code, use a C compiler"
+#endif
+
+#ifdef _LIBC
+/* We have to keep the namespace clean. */
+# define regfree(preg) __regfree (preg)
+# define regexec(pr, st, nm, pm, ef) __regexec (pr, st, nm, pm, ef)
+# define regcomp(preg, pattern, cflags) __regcomp (preg, pattern, cflags)
+# define regerror(errcode, preg, errbuf, errbuf_size) \
+ __regerror(errcode, preg, errbuf, errbuf_size)
+# define re_set_registers(bu, re, nu, st, en) \
+ __re_set_registers (bu, re, nu, st, en)
+# define re_match_2(bufp, string1, size1, string2, size2, pos, regs, stop) \
+ __re_match_2 (bufp, string1, size1, string2, size2, pos, regs, stop)
+# define re_match(bufp, string, size, pos, regs) \
+ __re_match (bufp, string, size, pos, regs)
+# define re_search(bufp, string, size, startpos, range, regs) \
+ __re_search (bufp, string, size, startpos, range, regs)
+# define re_compile_pattern(pattern, length, bufp) \
+ __re_compile_pattern (pattern, length, bufp)
+# define re_set_syntax(syntax) __re_set_syntax (syntax)
+# define re_search_2(bufp, st1, s1, st2, s2, startpos, range, regs, stop) \
+ __re_search_2 (bufp, st1, s1, st2, s2, startpos, range, regs, stop)
+# define re_compile_fastmap(bufp) __re_compile_fastmap (bufp)
+
+# include "../locale/localeinfo.h"
+#endif
+
+/* On some systems, limits.h sets RE_DUP_MAX to a lower value than
+ GNU regex allows. Include it before <regex.h>, which correctly
+ #undefs RE_DUP_MAX and sets it to the right value. */
+#include <limits.h>
+
+#include <regex.h>
+#include "regex_internal.h"
+
+#include "regex_internal.c"
+#include "regcomp.c"
+#include "regexec.c"
+
+/* Binary backward compatibility. */
+#if _LIBC
+# include <shlib-compat.h>
+# if SHLIB_COMPAT (libc, GLIBC_2_0, GLIBC_2_3)
+link_warning (re_max_failures, "the 're_max_failures' variable is obsolete and
will go away.")
+int re_max_failures = 2000;
+# endif
+#endif
diff --git a/lib/regex.h b/lib/regex.h
new file mode 100644
index 0000000..f2ac950
--- /dev/null
+++ b/lib/regex.h
@@ -0,0 +1,658 @@
+/* Definitions for data structures and routines for the regular
+ expression library.
+ Copyright (C) 1985, 1989-2018 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 General Public
+ License as published by the Free Software Foundation; either
+ version 3 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
+ General Public License for more details.
+
+ You should have received a copy of the GNU General Public
+ License along with the GNU C Library; if not, see
+ <https://www.gnu.org/licenses/>. */
+
+#ifndef _REGEX_H
+#define _REGEX_H 1
+
+#include <sys/types.h>
+
+/* Allow the use in C++ code. */
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/* Define __USE_GNU to declare GNU extensions that violate the
+ POSIX name space rules. */
+#ifdef _GNU_SOURCE
+# define __USE_GNU 1
+#endif
+
+#ifdef _REGEX_LARGE_OFFSETS
+
+/* Use types and values that are wide enough to represent signed and
+ unsigned byte offsets in memory. This currently works only when
+ the regex code is used outside of the GNU C library; it is not yet
+ supported within glibc itself, and glibc users should not define
+ _REGEX_LARGE_OFFSETS. */
+
+/* The type of object sizes. */
+typedef size_t __re_size_t;
+
+/* The type of object sizes, in places where the traditional code
+ uses unsigned long int. */
+typedef size_t __re_long_size_t;
+
+#else
+
+/* The traditional GNU regex implementation mishandles strings longer
+ than INT_MAX. */
+typedef unsigned int __re_size_t;
+typedef unsigned long int __re_long_size_t;
+
+#endif
+
+/* The following two types have to be signed and unsigned integer type
+ wide enough to hold a value of a pointer. For most ANSI compilers
+ ptrdiff_t and size_t should be likely OK. Still size of these two
+ types is 2 for Microsoft C. Ugh... */
+typedef long int s_reg_t;
+typedef unsigned long int active_reg_t;
+
+/* The following bits are used to determine the regexp syntax we
+ recognize. The set/not-set meanings are chosen so that Emacs syntax
+ remains the value 0. The bits are given in alphabetical order, and
+ the definitions shifted by one from the previous bit; thus, when we
+ add or remove a bit, only one other definition need change. */
+typedef unsigned long int reg_syntax_t;
+
+#ifdef __USE_GNU
+/* If this bit is not set, then \ inside a bracket expression is literal.
+ If set, then such a \ quotes the following character. */
+# define RE_BACKSLASH_ESCAPE_IN_LISTS ((unsigned long int) 1)
+
+/* If this bit is not set, then + and ? are operators, and \+ and \? are
+ literals.
+ If set, then \+ and \? are operators and + and ? are literals. */
+# define RE_BK_PLUS_QM (RE_BACKSLASH_ESCAPE_IN_LISTS << 1)
+
+/* If this bit is set, then character classes are supported. They are:
+ [:alpha:], [:upper:], [:lower:], [:digit:], [:alnum:], [:xdigit:],
+ [:space:], [:print:], [:punct:], [:graph:], and [:cntrl:].
+ If not set, then character classes are not supported. */
+# define RE_CHAR_CLASSES (RE_BK_PLUS_QM << 1)
+
+/* If this bit is set, then ^ and $ are always anchors (outside bracket
+ expressions, of course).
+ If this bit is not set, then it depends:
+ ^ is an anchor if it is at the beginning of a regular
+ expression or after an open-group or an alternation operator;
+ $ is an anchor if it is at the end of a regular expression, or
+ before a close-group or an alternation operator.
+
+ This bit could be (re)combined with RE_CONTEXT_INDEP_OPS, because
+ POSIX draft 11.2 says that * etc. in leading positions is undefined.
+ We already implemented a previous draft which made those constructs
+ invalid, though, so we haven't changed the code back. */
+# define RE_CONTEXT_INDEP_ANCHORS (RE_CHAR_CLASSES << 1)
+
+/* If this bit is set, then special characters are always special
+ regardless of where they are in the pattern.
+ If this bit is not set, then special characters are special only in
+ some contexts; otherwise they are ordinary. Specifically,
+ * + ? and intervals are only special when not after the beginning,
+ open-group, or alternation operator. */
+# define RE_CONTEXT_INDEP_OPS (RE_CONTEXT_INDEP_ANCHORS << 1)
+
+/* If this bit is set, then *, +, ?, and { cannot be first in an re or
+ immediately after an alternation or begin-group operator. */
+# define RE_CONTEXT_INVALID_OPS (RE_CONTEXT_INDEP_OPS << 1)
+
+/* If this bit is set, then . matches newline.
+ If not set, then it doesn't. */
+# define RE_DOT_NEWLINE (RE_CONTEXT_INVALID_OPS << 1)
+
+/* If this bit is set, then . doesn't match NUL.
+ If not set, then it does. */
+# define RE_DOT_NOT_NULL (RE_DOT_NEWLINE << 1)
+
+/* If this bit is set, nonmatching lists [^...] do not match newline.
+ If not set, they do. */
+# define RE_HAT_LISTS_NOT_NEWLINE (RE_DOT_NOT_NULL << 1)
+
+/* If this bit is set, either \{...\} or {...} defines an
+ interval, depending on RE_NO_BK_BRACES.
+ If not set, \{, \}, {, and } are literals. */
+# define RE_INTERVALS (RE_HAT_LISTS_NOT_NEWLINE << 1)
+
+/* If this bit is set, +, ? and | aren't recognized as operators.
+ If not set, they are. */
+# define RE_LIMITED_OPS (RE_INTERVALS << 1)
+
+/* If this bit is set, newline is an alternation operator.
+ If not set, newline is literal. */
+# define RE_NEWLINE_ALT (RE_LIMITED_OPS << 1)
+
+/* If this bit is set, then '{...}' defines an interval, and \{ and \}
+ are literals.
+ If not set, then '\{...\}' defines an interval. */
+# define RE_NO_BK_BRACES (RE_NEWLINE_ALT << 1)
+
+/* If this bit is set, (...) defines a group, and \( and \) are literals.
+ If not set, \(...\) defines a group, and ( and ) are literals. */
+# define RE_NO_BK_PARENS (RE_NO_BK_BRACES << 1)
+
+/* If this bit is set, then \<digit> matches <digit>.
+ If not set, then \<digit> is a back-reference. */
+# define RE_NO_BK_REFS (RE_NO_BK_PARENS << 1)
+
+/* If this bit is set, then | is an alternation operator, and \| is literal.
+ If not set, then \| is an alternation operator, and | is literal. */
+# define RE_NO_BK_VBAR (RE_NO_BK_REFS << 1)
+
+/* If this bit is set, then an ending range point collating higher
+ than the starting range point, as in [z-a], is invalid.
+ If not set, then when ending range point collates higher than the
+ starting range point, the range is ignored. */
+# define RE_NO_EMPTY_RANGES (RE_NO_BK_VBAR << 1)
+
+/* If this bit is set, then an unmatched ) is ordinary.
+ If not set, then an unmatched ) is invalid. */
+# define RE_UNMATCHED_RIGHT_PAREN_ORD (RE_NO_EMPTY_RANGES << 1)
+
+/* If this bit is set, succeed as soon as we match the whole pattern,
+ without further backtracking. */
+# define RE_NO_POSIX_BACKTRACKING (RE_UNMATCHED_RIGHT_PAREN_ORD << 1)
+
+/* If this bit is set, do not process the GNU regex operators.
+ If not set, then the GNU regex operators are recognized. */
+# define RE_NO_GNU_OPS (RE_NO_POSIX_BACKTRACKING << 1)
+
+/* If this bit is set, turn on internal regex debugging.
+ If not set, and debugging was on, turn it off.
+ This only works if regex.c is compiled -DDEBUG.
+ We define this bit always, so that all that's needed to turn on
+ debugging is to recompile regex.c; the calling code can always have
+ this bit set, and it won't affect anything in the normal case. */
+# define RE_DEBUG (RE_NO_GNU_OPS << 1)
+
+/* If this bit is set, a syntactically invalid interval is treated as
+ a string of ordinary characters. For example, the ERE 'a{1' is
+ treated as 'a\{1'. */
+# define RE_INVALID_INTERVAL_ORD (RE_DEBUG << 1)
+
+/* If this bit is set, then ignore case when matching.
+ If not set, then case is significant. */
+# define RE_ICASE (RE_INVALID_INTERVAL_ORD << 1)
+
+/* This bit is used internally like RE_CONTEXT_INDEP_ANCHORS but only
+ for ^, because it is difficult to scan the regex backwards to find
+ whether ^ should be special. */
+# define RE_CARET_ANCHORS_HERE (RE_ICASE << 1)
+
+/* If this bit is set, then \{ cannot be first in a regex or
+ immediately after an alternation, open-group or \} operator. */
+# define RE_CONTEXT_INVALID_DUP (RE_CARET_ANCHORS_HERE << 1)
+
+/* If this bit is set, then no_sub will be set to 1 during
+ re_compile_pattern. */
+# define RE_NO_SUB (RE_CONTEXT_INVALID_DUP << 1)
+#endif
+
+/* This global variable defines the particular regexp syntax to use (for
+ some interfaces). When a regexp is compiled, the syntax used is
+ stored in the pattern buffer, so changing this does not affect
+ already-compiled regexps. */
+extern reg_syntax_t re_syntax_options;
+
+#ifdef __USE_GNU
+/* Define combinations of the above bits for the standard possibilities.
+ (The [[[ comments delimit what gets put into the Texinfo file, so
+ don't delete them!) */
+/* [[[begin syntaxes]]] */
+# define RE_SYNTAX_EMACS 0
+
+# define RE_SYNTAX_AWK \
+ (RE_BACKSLASH_ESCAPE_IN_LISTS | RE_DOT_NOT_NULL \
+ | RE_NO_BK_PARENS | RE_NO_BK_REFS \
+ | RE_NO_BK_VBAR | RE_NO_EMPTY_RANGES \
+ | RE_DOT_NEWLINE | RE_CONTEXT_INDEP_ANCHORS \
+ | RE_CHAR_CLASSES \
+ | RE_UNMATCHED_RIGHT_PAREN_ORD | RE_NO_GNU_OPS)
+
+# define RE_SYNTAX_GNU_AWK \
+ ((RE_SYNTAX_POSIX_EXTENDED | RE_BACKSLASH_ESCAPE_IN_LISTS \
+ | RE_INVALID_INTERVAL_ORD) \
+ & ~(RE_DOT_NOT_NULL | RE_CONTEXT_INDEP_OPS \
+ | RE_CONTEXT_INVALID_OPS ))
+
+# define RE_SYNTAX_POSIX_AWK \
+ (RE_SYNTAX_POSIX_EXTENDED | RE_BACKSLASH_ESCAPE_IN_LISTS \
+ | RE_INTERVALS | RE_NO_GNU_OPS \
+ | RE_INVALID_INTERVAL_ORD)
+
+# define RE_SYNTAX_GREP
\
+ ((RE_SYNTAX_POSIX_BASIC | RE_NEWLINE_ALT) \
+ & ~(RE_CONTEXT_INVALID_DUP | RE_DOT_NOT_NULL))
+
+# define RE_SYNTAX_EGREP \
+ ((RE_SYNTAX_POSIX_EXTENDED | RE_INVALID_INTERVAL_ORD | RE_NEWLINE_ALT) \
+ & ~(RE_CONTEXT_INVALID_OPS | RE_DOT_NOT_NULL))
+
+/* POSIX grep -E behavior is no longer incompatible with GNU. */
+# define RE_SYNTAX_POSIX_EGREP \
+ RE_SYNTAX_EGREP
+
+/* P1003.2/D11.2, section 4.20.7.1, lines 5078ff. */
+# define RE_SYNTAX_ED RE_SYNTAX_POSIX_BASIC
+
+# define RE_SYNTAX_SED RE_SYNTAX_POSIX_BASIC
+
+/* Syntax bits common to both basic and extended POSIX regex syntax. */
+# define _RE_SYNTAX_POSIX_COMMON \
+ (RE_CHAR_CLASSES | RE_DOT_NEWLINE | RE_DOT_NOT_NULL \
+ | RE_INTERVALS | RE_NO_EMPTY_RANGES)
+
+# define RE_SYNTAX_POSIX_BASIC \
+ (_RE_SYNTAX_POSIX_COMMON | RE_BK_PLUS_QM | RE_CONTEXT_INVALID_DUP)
+
+/* Differs from ..._POSIX_BASIC only in that RE_BK_PLUS_QM becomes
+ RE_LIMITED_OPS, i.e., \? \+ \| are not recognized. Actually, this
+ isn't minimal, since other operators, such as \`, aren't disabled. */
+# define RE_SYNTAX_POSIX_MINIMAL_BASIC \
+ (_RE_SYNTAX_POSIX_COMMON | RE_LIMITED_OPS)
+
+# define RE_SYNTAX_POSIX_EXTENDED \
+ (_RE_SYNTAX_POSIX_COMMON | RE_CONTEXT_INDEP_ANCHORS \
+ | RE_CONTEXT_INDEP_OPS | RE_NO_BK_BRACES \
+ | RE_NO_BK_PARENS | RE_NO_BK_VBAR \
+ | RE_CONTEXT_INVALID_OPS | RE_UNMATCHED_RIGHT_PAREN_ORD)
+
+/* Differs from ..._POSIX_EXTENDED in that RE_CONTEXT_INDEP_OPS is
+ removed and RE_NO_BK_REFS is added. */
+# define RE_SYNTAX_POSIX_MINIMAL_EXTENDED \
+ (_RE_SYNTAX_POSIX_COMMON | RE_CONTEXT_INDEP_ANCHORS \
+ | RE_CONTEXT_INVALID_OPS | RE_NO_BK_BRACES \
+ | RE_NO_BK_PARENS | RE_NO_BK_REFS \
+ | RE_NO_BK_VBAR | RE_UNMATCHED_RIGHT_PAREN_ORD)
+/* [[[end syntaxes]]] */
+
+/* Maximum number of duplicates an interval can allow. POSIX-conforming
+ systems might define this in <limits.h>, but we want our
+ value, so remove any previous define. */
+# ifdef _REGEX_INCLUDE_LIMITS_H
+# include <limits.h>
+# endif
+# ifdef RE_DUP_MAX
+# undef RE_DUP_MAX
+# endif
+
+/* RE_DUP_MAX is 2**15 - 1 because an earlier implementation stored
+ the counter as a 2-byte signed integer. This is no longer true, so
+ RE_DUP_MAX could be increased to (INT_MAX / 10 - 1), or to
+ ((SIZE_MAX - 9) / 10) if _REGEX_LARGE_OFFSETS is defined.
+ However, there would be a huge performance problem if someone
+ actually used a pattern like a\{214748363\}, so RE_DUP_MAX retains
+ its historical value. */
+# define RE_DUP_MAX (0x7fff)
+#endif
+
+
+/* POSIX 'cflags' bits (i.e., information for 'regcomp'). */
+
+/* If this bit is set, then use extended regular expression syntax.
+ If not set, then use basic regular expression syntax. */
+#define REG_EXTENDED 1
+
+/* If this bit is set, then ignore case when matching.
+ If not set, then case is significant. */
+#define REG_ICASE (1 << 1)
+
+/* If this bit is set, then anchors do not match at newline
+ characters in the string.
+ If not set, then anchors do match at newlines. */
+#define REG_NEWLINE (1 << 2)
+
+/* If this bit is set, then report only success or fail in regexec.
+ If not set, then returns differ between not matching and errors. */
+#define REG_NOSUB (1 << 3)
+
+
+/* POSIX 'eflags' bits (i.e., information for regexec). */
+
+/* If this bit is set, then the beginning-of-line operator doesn't match
+ the beginning of the string (presumably because it's not the
+ beginning of a line).
+ If not set, then the beginning-of-line operator does match the
+ beginning of the string. */
+#define REG_NOTBOL 1
+
+/* Like REG_NOTBOL, except for the end-of-line. */
+#define REG_NOTEOL (1 << 1)
+
+/* Use PMATCH[0] to delimit the start and end of the search in the
+ buffer. */
+#define REG_STARTEND (1 << 2)
+
+
+/* If any error codes are removed, changed, or added, update the
+ '__re_error_msgid' table in regcomp.c. */
+
+typedef enum
+{
+ _REG_ENOSYS = -1, /* This will never happen for this implementation. */
+ _REG_NOERROR = 0, /* Success. */
+ _REG_NOMATCH, /* Didn't find a match (for regexec). */
+
+ /* POSIX regcomp return error codes. (In the order listed in the
+ standard.) */
+ _REG_BADPAT, /* Invalid pattern. */
+ _REG_ECOLLATE, /* Invalid collating element. */
+ _REG_ECTYPE, /* Invalid character class name. */
+ _REG_EESCAPE, /* Trailing backslash. */
+ _REG_ESUBREG, /* Invalid back reference. */
+ _REG_EBRACK, /* Unmatched left bracket. */
+ _REG_EPAREN, /* Parenthesis imbalance. */
+ _REG_EBRACE, /* Unmatched \{. */
+ _REG_BADBR, /* Invalid contents of \{\}. */
+ _REG_ERANGE, /* Invalid range end. */
+ _REG_ESPACE, /* Ran out of memory. */
+ _REG_BADRPT, /* No preceding re for repetition op. */
+
+ /* Error codes we've added. */
+ _REG_EEND, /* Premature end. */
+ _REG_ESIZE, /* Too large (e.g., repeat count too large). */
+ _REG_ERPAREN /* Unmatched ) or \); not returned from regcomp. */
+} reg_errcode_t;
+
+#if defined _XOPEN_SOURCE || defined __USE_XOPEN2K
+# define REG_ENOSYS _REG_ENOSYS
+#endif
+#define REG_NOERROR _REG_NOERROR
+#define REG_NOMATCH _REG_NOMATCH
+#define REG_BADPAT _REG_BADPAT
+#define REG_ECOLLATE _REG_ECOLLATE
+#define REG_ECTYPE _REG_ECTYPE
+#define REG_EESCAPE _REG_EESCAPE
+#define REG_ESUBREG _REG_ESUBREG
+#define REG_EBRACK _REG_EBRACK
+#define REG_EPAREN _REG_EPAREN
+#define REG_EBRACE _REG_EBRACE
+#define REG_BADBR _REG_BADBR
+#define REG_ERANGE _REG_ERANGE
+#define REG_ESPACE _REG_ESPACE
+#define REG_BADRPT _REG_BADRPT
+#define REG_EEND _REG_EEND
+#define REG_ESIZE _REG_ESIZE
+#define REG_ERPAREN _REG_ERPAREN
+
+/* This data structure represents a compiled pattern. Before calling
+ the pattern compiler, the fields 'buffer', 'allocated', 'fastmap',
+ and 'translate' can be set. After the pattern has been compiled,
+ the fields 're_nsub', 'not_bol' and 'not_eol' are available. All
+ other fields are private to the regex routines. */
+
+#ifndef RE_TRANSLATE_TYPE
+# define __RE_TRANSLATE_TYPE unsigned char *
+# ifdef __USE_GNU
+# define RE_TRANSLATE_TYPE __RE_TRANSLATE_TYPE
+# endif
+#endif
+
+#ifdef __USE_GNU
+# define __REPB_PREFIX(name) name
+#else
+# define __REPB_PREFIX(name) __##name
+#endif
+
+struct re_pattern_buffer
+{
+ /* Space that holds the compiled pattern. The type
+ 'struct re_dfa_t' is private and is not declared here. */
+ struct re_dfa_t *__REPB_PREFIX(buffer);
+
+ /* Number of bytes to which 'buffer' points. */
+ __re_long_size_t __REPB_PREFIX(allocated);
+
+ /* Number of bytes actually used in 'buffer'. */
+ __re_long_size_t __REPB_PREFIX(used);
+
+ /* Syntax setting with which the pattern was compiled. */
+ reg_syntax_t __REPB_PREFIX(syntax);
+
+ /* Pointer to a fastmap, if any, otherwise zero. re_search uses the
+ fastmap, if there is one, to skip over impossible starting points
+ for matches. */
+ char *__REPB_PREFIX(fastmap);
+
+ /* Either a translate table to apply to all characters before
+ comparing them, or zero for no translation. The translation is
+ applied to a pattern when it is compiled and to a string when it
+ is matched. */
+ __RE_TRANSLATE_TYPE __REPB_PREFIX(translate);
+
+ /* Number of subexpressions found by the compiler. */
+ size_t re_nsub;
+
+ /* Zero if this pattern cannot match the empty string, one else.
+ Well, in truth it's used only in 're_search_2', to see whether or
+ not we should use the fastmap, so we don't set this absolutely
+ perfectly; see 're_compile_fastmap' (the "duplicate" case). */
+ unsigned __REPB_PREFIX(can_be_null) : 1;
+
+ /* If REGS_UNALLOCATED, allocate space in the 'regs' structure
+ for 'max (RE_NREGS, re_nsub + 1)' groups.
+ If REGS_REALLOCATE, reallocate space if necessary.
+ If REGS_FIXED, use what's there. */
+#ifdef __USE_GNU
+# define REGS_UNALLOCATED 0
+# define REGS_REALLOCATE 1
+# define REGS_FIXED 2
+#endif
+ unsigned __REPB_PREFIX(regs_allocated) : 2;
+
+ /* Set to zero when 're_compile_pattern' compiles a pattern; set to
+ one by 're_compile_fastmap' if it updates the fastmap. */
+ unsigned __REPB_PREFIX(fastmap_accurate) : 1;
+
+ /* If set, 're_match_2' does not return information about
+ subexpressions. */
+ unsigned __REPB_PREFIX(no_sub) : 1;
+
+ /* If set, a beginning-of-line anchor doesn't match at the beginning
+ of the string. */
+ unsigned __REPB_PREFIX(not_bol) : 1;
+
+ /* Similarly for an end-of-line anchor. */
+ unsigned __REPB_PREFIX(not_eol) : 1;
+
+ /* If true, an anchor at a newline matches. */
+ unsigned __REPB_PREFIX(newline_anchor) : 1;
+};
+
+typedef struct re_pattern_buffer regex_t;
+
+/* Type for byte offsets within the string. POSIX mandates this. */
+#ifdef _REGEX_LARGE_OFFSETS
+/* POSIX 1003.1-2008 requires that regoff_t be at least as wide as
+ ptrdiff_t and ssize_t. We don't know of any hosts where ptrdiff_t
+ is wider than ssize_t, so ssize_t is safe. ptrdiff_t is not
+ visible here, so use ssize_t. */
+typedef ssize_t regoff_t;
+#else
+/* The traditional GNU regex implementation mishandles strings longer
+ than INT_MAX. */
+typedef int regoff_t;
+#endif
+
+
+#ifdef __USE_GNU
+/* This is the structure we store register match data in. See
+ regex.texinfo for a full description of what registers match. */
+struct re_registers
+{
+ __re_size_t num_regs;
+ regoff_t *start;
+ regoff_t *end;
+};
+
+
+/* If 'regs_allocated' is REGS_UNALLOCATED in the pattern buffer,
+ 're_match_2' returns information about at least this many registers
+ the first time a 'regs' structure is passed. */
+# ifndef RE_NREGS
+# define RE_NREGS 30
+# endif
+#endif
+
+
+/* POSIX specification for registers. Aside from the different names than
+ 're_registers', POSIX uses an array of structures, instead of a
+ structure of arrays. */
+typedef struct
+{
+ regoff_t rm_so; /* Byte offset from string's start to substring's start. */
+ regoff_t rm_eo; /* Byte offset from string's start to substring's end. */
+} regmatch_t;
+
+/* Declarations for routines. */
+
+#ifdef __USE_GNU
+/* Sets the current default syntax to SYNTAX, and return the old syntax.
+ You can also simply assign to the 're_syntax_options' variable. */
+extern reg_syntax_t re_set_syntax (reg_syntax_t __syntax);
+
+/* Compile the regular expression PATTERN, with length LENGTH
+ and syntax given by the global 're_syntax_options', into the buffer
+ BUFFER. Return NULL if successful, and an error string if not.
+
+ To free the allocated storage, you must call 'regfree' on BUFFER.
+ Note that the translate table must either have been initialized by
+ 'regcomp', with a malloc'ed value, or set to NULL before calling
+ 'regfree'. */
+extern const char *re_compile_pattern (const char *__pattern, size_t __length,
+ struct re_pattern_buffer *__buffer);
+
+
+/* Compile a fastmap for the compiled pattern in BUFFER; used to
+ accelerate searches. Return 0 if successful and -2 if was an
+ internal error. */
+extern int re_compile_fastmap (struct re_pattern_buffer *__buffer);
+
+
+/* Search in the string STRING (with length LENGTH) for the pattern
+ compiled into BUFFER. Start searching at position START, for RANGE
+ characters. Return the starting position of the match, -1 for no
+ match, or -2 for an internal error. Also return register
+ information in REGS (if REGS and BUFFER->no_sub are nonzero). */
+extern regoff_t re_search (struct re_pattern_buffer *__buffer,
+ const char *__String, regoff_t __length,
+ regoff_t __start, regoff_t __range,
+ struct re_registers *__regs);
+
+
+/* Like 're_search', but search in the concatenation of STRING1 and
+ STRING2. Also, stop searching at index START + STOP. */
+extern regoff_t re_search_2 (struct re_pattern_buffer *__buffer,
+ const char *__string1, regoff_t __length1,
+ const char *__string2, regoff_t __length2,
+ regoff_t __start, regoff_t __range,
+ struct re_registers *__regs,
+ regoff_t __stop);
+
+
+/* Like 're_search', but return how many characters in STRING the regexp
+ in BUFFER matched, starting at position START. */
+extern regoff_t re_match (struct re_pattern_buffer *__buffer,
+ const char *__String, regoff_t __length,
+ regoff_t __start, struct re_registers *__regs);
+
+
+/* Relates to 're_match' as 're_search_2' relates to 're_search'. */
+extern regoff_t re_match_2 (struct re_pattern_buffer *__buffer,
+ const char *__string1, regoff_t __length1,
+ const char *__string2, regoff_t __length2,
+ regoff_t __start, struct re_registers *__regs,
+ regoff_t __stop);
+
+
+/* Set REGS to hold NUM_REGS registers, storing them in STARTS and
+ ENDS. Subsequent matches using BUFFER and REGS will use this memory
+ for recording register information. STARTS and ENDS must be
+ allocated with malloc, and must each be at least 'NUM_REGS * sizeof
+ (regoff_t)' bytes long.
+
+ If NUM_REGS == 0, then subsequent matches should allocate their own
+ register data.
+
+ Unless this function is called, the first search or match using
+ BUFFER will allocate its own register data, without
+ freeing the old data. */
+extern void re_set_registers (struct re_pattern_buffer *__buffer,
+ struct re_registers *__regs,
+ __re_size_t __num_regs,
+ regoff_t *__starts, regoff_t *__ends);
+#endif /* Use GNU */
+
+#if defined _REGEX_RE_COMP || (defined _LIBC && defined __USE_MISC)
+# ifndef _CRAY
+/* 4.2 bsd compatibility. */
+extern char *re_comp (const char *);
+extern int re_exec (const char *);
+# endif
+#endif
+
+/* For plain 'restrict', use glibc's __restrict if defined.
+ Otherwise, GCC 2.95 and later have "__restrict"; C99 compilers have
+ "restrict", and "configure" may have defined "restrict".
+ Other compilers use __restrict, __restrict__, and _Restrict, and
+ 'configure' might #define 'restrict' to those words, so pick a
+ different name. */
+#ifndef _Restrict_
+# if defined __restrict || 2 < __GNUC__ + (95 <= __GNUC_MINOR__)
+# define _Restrict_ __restrict
+# elif 199901L <= __STDC_VERSION__ || defined restrict
+# define _Restrict_ restrict
+# else
+# define _Restrict_
+# endif
+#endif
+/* For [restrict], use glibc's __restrict_arr if available.
+ Otherwise, GCC 3.1 (not in C++ mode) and C99 support [restrict]. */
+#ifndef _Restrict_arr_
+# ifdef __restrict_arr
+# define _Restrict_arr_ __restrict_arr
+# elif ((199901L <= __STDC_VERSION__ || 3 < __GNUC__ + (1 <= __GNUC_MINOR__)) \
+ && !defined __GNUG__)
+# define _Restrict_arr_ _Restrict_
+# else
+# define _Restrict_arr_
+# endif
+#endif
+
+/* POSIX compatibility. */
+extern int regcomp (regex_t *_Restrict_ __preg,
+ const char *_Restrict_ __pattern,
+ int __cflags);
+
+extern int regexec (const regex_t *_Restrict_ __preg,
+ const char *_Restrict_ __String, size_t __nmatch,
+ regmatch_t __pmatch[_Restrict_arr_],
+ int __eflags);
+
+extern size_t regerror (int __errcode, const regex_t *_Restrict_ __preg,
+ char *_Restrict_ __errbuf, size_t __errbuf_size);
+
+extern void regfree (regex_t *__preg);
+
+
+#ifdef __cplusplus
+}
+#endif /* C++ */
+
+#endif /* regex.h */
diff --git a/lib/regex_internal.c b/lib/regex_internal.c
new file mode 100644
index 0000000..3237356
--- /dev/null
+++ b/lib/regex_internal.c
@@ -0,0 +1,1740 @@
+/* Extended regular expression matching and search library.
+ Copyright (C) 2002-2018 Free Software Foundation, Inc.
+ This file is part of the GNU C Library.
+ Contributed by Isamu Hasegawa <address@hidden>.
+
+ The GNU C Library 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 3 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
+ General Public License for more details.
+
+ You should have received a copy of the GNU General Public
+ License along with the GNU C Library; if not, see
+ <https://www.gnu.org/licenses/>. */
+
+static void re_string_construct_common (const char *str, Idx len,
+ re_string_t *pstr,
+ RE_TRANSLATE_TYPE trans, bool icase,
+ const re_dfa_t *dfa);
+static re_dfastate_t *create_ci_newstate (const re_dfa_t *dfa,
+ const re_node_set *nodes,
+ re_hashval_t hash);
+static re_dfastate_t *create_cd_newstate (const re_dfa_t *dfa,
+ const re_node_set *nodes,
+ unsigned int context,
+ re_hashval_t hash);
+static reg_errcode_t re_string_realloc_buffers (re_string_t *pstr,
+ Idx new_buf_len);
+#ifdef RE_ENABLE_I18N
+static void build_wcs_buffer (re_string_t *pstr);
+static reg_errcode_t build_wcs_upper_buffer (re_string_t *pstr);
+#endif /* RE_ENABLE_I18N */
+static void build_upper_buffer (re_string_t *pstr);
+static void re_string_translate_buffer (re_string_t *pstr);
+static unsigned int re_string_context_at (const re_string_t *input, Idx idx,
+ int eflags) __attribute__ ((pure));
+
+/* Functions for string operation. */
+
+/* This function allocate the buffers. It is necessary to call
+ re_string_reconstruct before using the object. */
+
+static reg_errcode_t
+__attribute_warn_unused_result__
+re_string_allocate (re_string_t *pstr, const char *str, Idx len, Idx init_len,
+ RE_TRANSLATE_TYPE trans, bool icase, const re_dfa_t *dfa)
+{
+ reg_errcode_t ret;
+ Idx init_buf_len;
+
+ /* Ensure at least one character fits into the buffers. */
+ if (init_len < dfa->mb_cur_max)
+ init_len = dfa->mb_cur_max;
+ init_buf_len = (len + 1 < init_len) ? len + 1: init_len;
+ re_string_construct_common (str, len, pstr, trans, icase, dfa);
+
+ ret = re_string_realloc_buffers (pstr, init_buf_len);
+ if (BE (ret != REG_NOERROR, 0))
+ return ret;
+
+ pstr->word_char = dfa->word_char;
+ pstr->word_ops_used = dfa->word_ops_used;
+ pstr->mbs = pstr->mbs_allocated ? pstr->mbs : (unsigned char *) str;
+ pstr->valid_len = (pstr->mbs_allocated || dfa->mb_cur_max > 1) ? 0 : len;
+ pstr->valid_raw_len = pstr->valid_len;
+ return REG_NOERROR;
+}
+
+/* This function allocate the buffers, and initialize them. */
+
+static reg_errcode_t
+__attribute_warn_unused_result__
+re_string_construct (re_string_t *pstr, const char *str, Idx len,
+ RE_TRANSLATE_TYPE trans, bool icase, const re_dfa_t *dfa)
+{
+ reg_errcode_t ret;
+ memset (pstr, '\0', sizeof (re_string_t));
+ re_string_construct_common (str, len, pstr, trans, icase, dfa);
+
+ if (len > 0)
+ {
+ ret = re_string_realloc_buffers (pstr, len + 1);
+ if (BE (ret != REG_NOERROR, 0))
+ return ret;
+ }
+ pstr->mbs = pstr->mbs_allocated ? pstr->mbs : (unsigned char *) str;
+
+ if (icase)
+ {
+#ifdef RE_ENABLE_I18N
+ if (dfa->mb_cur_max > 1)
+ {
+ while (1)
+ {
+ ret = build_wcs_upper_buffer (pstr);
+ if (BE (ret != REG_NOERROR, 0))
+ return ret;
+ if (pstr->valid_raw_len >= len)
+ break;
+ if (pstr->bufs_len > pstr->valid_len + dfa->mb_cur_max)
+ break;
+ ret = re_string_realloc_buffers (pstr, pstr->bufs_len * 2);
+ if (BE (ret != REG_NOERROR, 0))
+ return ret;
+ }
+ }
+ else
+#endif /* RE_ENABLE_I18N */
+ build_upper_buffer (pstr);
+ }
+ else
+ {
+#ifdef RE_ENABLE_I18N
+ if (dfa->mb_cur_max > 1)
+ build_wcs_buffer (pstr);
+ else
+#endif /* RE_ENABLE_I18N */
+ {
+ if (trans != NULL)
+ re_string_translate_buffer (pstr);
+ else
+ {
+ pstr->valid_len = pstr->bufs_len;
+ pstr->valid_raw_len = pstr->bufs_len;
+ }
+ }
+ }
+
+ return REG_NOERROR;
+}
+
+/* Helper functions for re_string_allocate, and re_string_construct. */
+
+static reg_errcode_t
+__attribute_warn_unused_result__
+re_string_realloc_buffers (re_string_t *pstr, Idx new_buf_len)
+{
+#ifdef RE_ENABLE_I18N
+ if (pstr->mb_cur_max > 1)
+ {
+ wint_t *new_wcs;
+
+ /* Avoid overflow in realloc. */
+ const size_t max_object_size = MAX (sizeof (wint_t), sizeof (Idx));
+ if (BE (MIN (IDX_MAX, SIZE_MAX / max_object_size) < new_buf_len, 0))
+ return REG_ESPACE;
+
+ new_wcs = re_realloc (pstr->wcs, wint_t, new_buf_len);
+ if (BE (new_wcs == NULL, 0))
+ return REG_ESPACE;
+ pstr->wcs = new_wcs;
+ if (pstr->offsets != NULL)
+ {
+ Idx *new_offsets = re_realloc (pstr->offsets, Idx, new_buf_len);
+ if (BE (new_offsets == NULL, 0))
+ return REG_ESPACE;
+ pstr->offsets = new_offsets;
+ }
+ }
+#endif /* RE_ENABLE_I18N */
+ if (pstr->mbs_allocated)
+ {
+ unsigned char *new_mbs = re_realloc (pstr->mbs, unsigned char,
+ new_buf_len);
+ if (BE (new_mbs == NULL, 0))
+ return REG_ESPACE;
+ pstr->mbs = new_mbs;
+ }
+ pstr->bufs_len = new_buf_len;
+ return REG_NOERROR;
+}
+
+
+static void
+re_string_construct_common (const char *str, Idx len, re_string_t *pstr,
+ RE_TRANSLATE_TYPE trans, bool icase,
+ const re_dfa_t *dfa)
+{
+ pstr->raw_mbs = (const unsigned char *) str;
+ pstr->len = len;
+ pstr->raw_len = len;
+ pstr->trans = trans;
+ pstr->icase = icase;
+ pstr->mbs_allocated = (trans != NULL || icase);
+ pstr->mb_cur_max = dfa->mb_cur_max;
+ pstr->is_utf8 = dfa->is_utf8;
+ pstr->map_notascii = dfa->map_notascii;
+ pstr->stop = pstr->len;
+ pstr->raw_stop = pstr->stop;
+}
+
+#ifdef RE_ENABLE_I18N
+
+/* Build wide character buffer PSTR->WCS.
+ If the byte sequence of the string are:
+ <mb1>(0), <mb1>(1), <mb2>(0), <mb2>(1), <sb3>
+ Then wide character buffer will be:
+ <wc1> , WEOF , <wc2> , WEOF , <wc3>
+ We use WEOF for padding, they indicate that the position isn't
+ a first byte of a multibyte character.
+
+ Note that this function assumes PSTR->VALID_LEN elements are already
+ built and starts from PSTR->VALID_LEN. */
+
+static void
+build_wcs_buffer (re_string_t *pstr)
+{
+#ifdef _LIBC
+ unsigned char buf[MB_LEN_MAX];
+ assert (MB_LEN_MAX >= pstr->mb_cur_max);
+#else
+ unsigned char buf[64];
+#endif
+ mbstate_t prev_st;
+ Idx byte_idx, end_idx, remain_len;
+ size_t mbclen;
+
+ /* Build the buffers from pstr->valid_len to either pstr->len or
+ pstr->bufs_len. */
+ end_idx = (pstr->bufs_len > pstr->len) ? pstr->len : pstr->bufs_len;
+ for (byte_idx = pstr->valid_len; byte_idx < end_idx;)
+ {
+ wchar_t wc;
+ const char *p;
+
+ remain_len = end_idx - byte_idx;
+ prev_st = pstr->cur_state;
+ /* Apply the translation if we need. */
+ if (BE (pstr->trans != NULL, 0))
+ {
+ int i, ch;
+
+ for (i = 0; i < pstr->mb_cur_max && i < remain_len; ++i)
+ {
+ ch = pstr->raw_mbs [pstr->raw_mbs_idx + byte_idx + i];
+ buf[i] = pstr->mbs[byte_idx + i] = pstr->trans[ch];
+ }
+ p = (const char *) buf;
+ }
+ else
+ p = (const char *) pstr->raw_mbs + pstr->raw_mbs_idx + byte_idx;
+ mbclen = __mbrtowc (&wc, p, remain_len, &pstr->cur_state);
+ if (BE (mbclen == (size_t) -1 || mbclen == 0
+ || (mbclen == (size_t) -2 && pstr->bufs_len >= pstr->len), 0))
+ {
+ /* We treat these cases as a singlebyte character. */
+ mbclen = 1;
+ wc = (wchar_t) pstr->raw_mbs[pstr->raw_mbs_idx + byte_idx];
+ if (BE (pstr->trans != NULL, 0))
+ wc = pstr->trans[wc];
+ pstr->cur_state = prev_st;
+ }
+ else if (BE (mbclen == (size_t) -2, 0))
+ {
+ /* The buffer doesn't have enough space, finish to build. */
+ pstr->cur_state = prev_st;
+ break;
+ }
+
+ /* Write wide character and padding. */
+ pstr->wcs[byte_idx++] = wc;
+ /* Write paddings. */
+ for (remain_len = byte_idx + mbclen - 1; byte_idx < remain_len ;)
+ pstr->wcs[byte_idx++] = WEOF;
+ }
+ pstr->valid_len = byte_idx;
+ pstr->valid_raw_len = byte_idx;
+}
+
+/* Build wide character buffer PSTR->WCS like build_wcs_buffer,
+ but for REG_ICASE. */
+
+static reg_errcode_t
+__attribute_warn_unused_result__
+build_wcs_upper_buffer (re_string_t *pstr)
+{
+ mbstate_t prev_st;
+ Idx src_idx, byte_idx, end_idx, remain_len;
+ size_t mbclen;
+#ifdef _LIBC
+ char buf[MB_LEN_MAX];
+ assert (MB_LEN_MAX >= pstr->mb_cur_max);
+#else
+ char buf[64];
+#endif
+
+ byte_idx = pstr->valid_len;
+ end_idx = (pstr->bufs_len > pstr->len) ? pstr->len : pstr->bufs_len;
+
+ /* The following optimization assumes that ASCII characters can be
+ mapped to wide characters with a simple cast. */
+ if (! pstr->map_notascii && pstr->trans == NULL && !pstr->offsets_needed)
+ {
+ while (byte_idx < end_idx)
+ {
+ wchar_t wc;
+
+ if (isascii (pstr->raw_mbs[pstr->raw_mbs_idx + byte_idx])
+ && mbsinit (&pstr->cur_state))
+ {
+ /* In case of a singlebyte character. */
+ pstr->mbs[byte_idx]
+ = toupper (pstr->raw_mbs[pstr->raw_mbs_idx + byte_idx]);
+ /* The next step uses the assumption that wchar_t is encoded
+ ASCII-safe: all ASCII values can be converted like this. */
+ pstr->wcs[byte_idx] = (wchar_t) pstr->mbs[byte_idx];
+ ++byte_idx;
+ continue;
+ }
+
+ remain_len = end_idx - byte_idx;
+ prev_st = pstr->cur_state;
+ mbclen = __mbrtowc (&wc,
+ ((const char *) pstr->raw_mbs + pstr->raw_mbs_idx
+ + byte_idx), remain_len, &pstr->cur_state);
+ if (BE (mbclen < (size_t) -2, 1))
+ {
+ wchar_t wcu = __towupper (wc);
+ if (wcu != wc)
+ {
+ size_t mbcdlen;
+
+ mbcdlen = __wcrtomb (buf, wcu, &prev_st);
+ if (BE (mbclen == mbcdlen, 1))
+ memcpy (pstr->mbs + byte_idx, buf, mbclen);
+ else
+ {
+ src_idx = byte_idx;
+ goto offsets_needed;
+ }
+ }
+ else
+ memcpy (pstr->mbs + byte_idx,
+ pstr->raw_mbs + pstr->raw_mbs_idx + byte_idx, mbclen);
+ pstr->wcs[byte_idx++] = wcu;
+ /* Write paddings. */
+ for (remain_len = byte_idx + mbclen - 1; byte_idx < remain_len ;)
+ pstr->wcs[byte_idx++] = WEOF;
+ }
+ else if (mbclen == (size_t) -1 || mbclen == 0
+ || (mbclen == (size_t) -2 && pstr->bufs_len >= pstr->len))
+ {
+ /* It is an invalid character, an incomplete character
+ at the end of the string, or '\0'. Just use the byte. */
+ int ch = pstr->raw_mbs[pstr->raw_mbs_idx + byte_idx];
+ pstr->mbs[byte_idx] = ch;
+ /* And also cast it to wide char. */
+ pstr->wcs[byte_idx++] = (wchar_t) ch;
+ if (BE (mbclen == (size_t) -1, 0))
+ pstr->cur_state = prev_st;
+ }
+ else
+ {
+ /* The buffer doesn't have enough space, finish to build. */
+ pstr->cur_state = prev_st;
+ break;
+ }
+ }
+ pstr->valid_len = byte_idx;
+ pstr->valid_raw_len = byte_idx;
+ return REG_NOERROR;
+ }
+ else
+ for (src_idx = pstr->valid_raw_len; byte_idx < end_idx;)
+ {
+ wchar_t wc;
+ const char *p;
+ offsets_needed:
+ remain_len = end_idx - byte_idx;
+ prev_st = pstr->cur_state;
+ if (BE (pstr->trans != NULL, 0))
+ {
+ int i, ch;
+
+ for (i = 0; i < pstr->mb_cur_max && i < remain_len; ++i)
+ {
+ ch = pstr->raw_mbs [pstr->raw_mbs_idx + src_idx + i];
+ buf[i] = pstr->trans[ch];
+ }
+ p = (const char *) buf;
+ }
+ else
+ p = (const char *) pstr->raw_mbs + pstr->raw_mbs_idx + src_idx;
+ mbclen = __mbrtowc (&wc, p, remain_len, &pstr->cur_state);
+ if (BE (mbclen < (size_t) -2, 1))
+ {
+ wchar_t wcu = __towupper (wc);
+ if (wcu != wc)
+ {
+ size_t mbcdlen;
+
+ mbcdlen = __wcrtomb ((char *) buf, wcu, &prev_st);
+ if (BE (mbclen == mbcdlen, 1))
+ memcpy (pstr->mbs + byte_idx, buf, mbclen);
+ else if (mbcdlen != (size_t) -1)
+ {
+ size_t i;
+
+ if (byte_idx + mbcdlen > pstr->bufs_len)
+ {
+ pstr->cur_state = prev_st;
+ break;
+ }
+
+ if (pstr->offsets == NULL)
+ {
+ pstr->offsets = re_malloc (Idx, pstr->bufs_len);
+
+ if (pstr->offsets == NULL)
+ return REG_ESPACE;
+ }
+ if (!pstr->offsets_needed)
+ {
+ for (i = 0; i < (size_t) byte_idx; ++i)
+ pstr->offsets[i] = i;
+ pstr->offsets_needed = 1;
+ }
+
+ memcpy (pstr->mbs + byte_idx, buf, mbcdlen);
+ pstr->wcs[byte_idx] = wcu;
+ pstr->offsets[byte_idx] = src_idx;
+ for (i = 1; i < mbcdlen; ++i)
+ {
+ pstr->offsets[byte_idx + i]
+ = src_idx + (i < mbclen ? i : mbclen - 1);
+ pstr->wcs[byte_idx + i] = WEOF;
+ }
+ pstr->len += mbcdlen - mbclen;
+ if (pstr->raw_stop > src_idx)
+ pstr->stop += mbcdlen - mbclen;
+ end_idx = (pstr->bufs_len > pstr->len)
+ ? pstr->len : pstr->bufs_len;
+ byte_idx += mbcdlen;
+ src_idx += mbclen;
+ continue;
+ }
+ else
+ memcpy (pstr->mbs + byte_idx, p, mbclen);
+ }
+ else
+ memcpy (pstr->mbs + byte_idx, p, mbclen);
+
+ if (BE (pstr->offsets_needed != 0, 0))
+ {
+ size_t i;
+ for (i = 0; i < mbclen; ++i)
+ pstr->offsets[byte_idx + i] = src_idx + i;
+ }
+ src_idx += mbclen;
+
+ pstr->wcs[byte_idx++] = wcu;
+ /* Write paddings. */
+ for (remain_len = byte_idx + mbclen - 1; byte_idx < remain_len ;)
+ pstr->wcs[byte_idx++] = WEOF;
+ }
+ else if (mbclen == (size_t) -1 || mbclen == 0
+ || (mbclen == (size_t) -2 && pstr->bufs_len >= pstr->len))
+ {
+ /* It is an invalid character or '\0'. Just use the byte. */
+ int ch = pstr->raw_mbs[pstr->raw_mbs_idx + src_idx];
+
+ if (BE (pstr->trans != NULL, 0))
+ ch = pstr->trans [ch];
+ pstr->mbs[byte_idx] = ch;
+
+ if (BE (pstr->offsets_needed != 0, 0))
+ pstr->offsets[byte_idx] = src_idx;
+ ++src_idx;
+
+ /* And also cast it to wide char. */
+ pstr->wcs[byte_idx++] = (wchar_t) ch;
+ if (BE (mbclen == (size_t) -1, 0))
+ pstr->cur_state = prev_st;
+ }
+ else
+ {
+ /* The buffer doesn't have enough space, finish to build. */
+ pstr->cur_state = prev_st;
+ break;
+ }
+ }
+ pstr->valid_len = byte_idx;
+ pstr->valid_raw_len = src_idx;
+ return REG_NOERROR;
+}
+
+/* Skip characters until the index becomes greater than NEW_RAW_IDX.
+ Return the index. */
+
+static Idx
+re_string_skip_chars (re_string_t *pstr, Idx new_raw_idx, wint_t *last_wc)
+{
+ mbstate_t prev_st;
+ Idx rawbuf_idx;
+ size_t mbclen;
+ wint_t wc = WEOF;
+
+ /* Skip the characters which are not necessary to check. */
+ for (rawbuf_idx = pstr->raw_mbs_idx + pstr->valid_raw_len;
+ rawbuf_idx < new_raw_idx;)
+ {
+ wchar_t wc2;
+ Idx remain_len = pstr->raw_len - rawbuf_idx;
+ prev_st = pstr->cur_state;
+ mbclen = __mbrtowc (&wc2, (const char *) pstr->raw_mbs + rawbuf_idx,
+ remain_len, &pstr->cur_state);
+ if (BE (mbclen == (size_t) -2 || mbclen == (size_t) -1 || mbclen == 0,
0))
+ {
+ /* We treat these cases as a single byte character. */
+ if (mbclen == 0 || remain_len == 0)
+ wc = L'\0';
+ else
+ wc = *(unsigned char *) (pstr->raw_mbs + rawbuf_idx);
+ mbclen = 1;
+ pstr->cur_state = prev_st;
+ }
+ else
+ wc = wc2;
+ /* Then proceed the next character. */
+ rawbuf_idx += mbclen;
+ }
+ *last_wc = wc;
+ return rawbuf_idx;
+}
+#endif /* RE_ENABLE_I18N */
+
+/* Build the buffer PSTR->MBS, and apply the translation if we need.
+ This function is used in case of REG_ICASE. */
+
+static void
+build_upper_buffer (re_string_t *pstr)
+{
+ Idx char_idx, end_idx;
+ end_idx = (pstr->bufs_len > pstr->len) ? pstr->len : pstr->bufs_len;
+
+ for (char_idx = pstr->valid_len; char_idx < end_idx; ++char_idx)
+ {
+ int ch = pstr->raw_mbs[pstr->raw_mbs_idx + char_idx];
+ if (BE (pstr->trans != NULL, 0))
+ ch = pstr->trans[ch];
+ pstr->mbs[char_idx] = toupper (ch);
+ }
+ pstr->valid_len = char_idx;
+ pstr->valid_raw_len = char_idx;
+}
+
+/* Apply TRANS to the buffer in PSTR. */
+
+static void
+re_string_translate_buffer (re_string_t *pstr)
+{
+ Idx buf_idx, end_idx;
+ end_idx = (pstr->bufs_len > pstr->len) ? pstr->len : pstr->bufs_len;
+
+ for (buf_idx = pstr->valid_len; buf_idx < end_idx; ++buf_idx)
+ {
+ int ch = pstr->raw_mbs[pstr->raw_mbs_idx + buf_idx];
+ pstr->mbs[buf_idx] = pstr->trans[ch];
+ }
+
+ pstr->valid_len = buf_idx;
+ pstr->valid_raw_len = buf_idx;
+}
+
+/* This function re-construct the buffers.
+ Concretely, convert to wide character in case of pstr->mb_cur_max > 1,
+ convert to upper case in case of REG_ICASE, apply translation. */
+
+static reg_errcode_t
+__attribute_warn_unused_result__
+re_string_reconstruct (re_string_t *pstr, Idx idx, int eflags)
+{
+ Idx offset;
+
+ if (BE (pstr->raw_mbs_idx <= idx, 0))
+ offset = idx - pstr->raw_mbs_idx;
+ else
+ {
+ /* Reset buffer. */
+#ifdef RE_ENABLE_I18N
+ if (pstr->mb_cur_max > 1)
+ memset (&pstr->cur_state, '\0', sizeof (mbstate_t));
+#endif /* RE_ENABLE_I18N */
+ pstr->len = pstr->raw_len;
+ pstr->stop = pstr->raw_stop;
+ pstr->valid_len = 0;
+ pstr->raw_mbs_idx = 0;
+ pstr->valid_raw_len = 0;
+ pstr->offsets_needed = 0;
+ pstr->tip_context = ((eflags & REG_NOTBOL) ? CONTEXT_BEGBUF
+ : CONTEXT_NEWLINE | CONTEXT_BEGBUF);
+ if (!pstr->mbs_allocated)
+ pstr->mbs = (unsigned char *) pstr->raw_mbs;
+ offset = idx;
+ }
+
+ if (BE (offset != 0, 1))
+ {
+ /* Should the already checked characters be kept? */
+ if (BE (offset < pstr->valid_raw_len, 1))
+ {
+ /* Yes, move them to the front of the buffer. */
+#ifdef RE_ENABLE_I18N
+ if (BE (pstr->offsets_needed, 0))
+ {
+ Idx 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 defined DEBUG && DEBUG
+ assert (pstr->valid_len > 0);
+#endif
+ }
+ }
+ else
+ {
+#ifdef RE_ENABLE_I18N
+ /* No, skip all characters until IDX. */
+ Idx prev_valid_len = pstr->valid_len;
+
+ if (BE (pstr->offsets_needed, 0))
+ {
+ pstr->len = pstr->raw_len - idx + offset;
+ pstr->stop = pstr->raw_stop - idx + offset;
+ pstr->offsets_needed = 0;
+ }
+#endif
+ pstr->valid_len = 0;
+#ifdef RE_ENABLE_I18N
+ if (pstr->mb_cur_max > 1)
+ {
+ Idx wcs_idx;
+ wint_t wc = WEOF;
+
+ if (pstr->is_utf8)
+ {
+ const unsigned char *raw, *p, *end;
+
+ /* Special case UTF-8. Multi-byte chars start with any
+ 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
+ case, ASCII characters, skip the conversion step. */
+ if (isascii (*p) && BE (pstr->trans == NULL, 1))
+ {
+ memset (&pstr->cur_state, '\0', sizeof (mbstate_t));
+ /* pstr->valid_len = 0; */
+ wc = (wchar_t) *p;
+ }
+ else
+#endif
+ for (; p >= end; --p)
+ if ((*p & 0xc0) != 0x80)
+ {
+ mbstate_t cur_state;
+ wchar_t wc2;
+ Idx mlen = raw + pstr->len - p;
+ unsigned char buf[6];
+ size_t mbclen;
+
+ const unsigned char *pp = p;
+ if (BE (pstr->trans != NULL, 0))
+ {
+ int i = mlen < 6 ? mlen : 6;
+ while (--i >= 0)
+ buf[i] = pstr->trans[p[i]];
+ pp = buf;
+ }
+ /* XXX Don't use mbrtowc, we know which conversion
+ to use (UTF-8 -> UCS4). */
+ memset (&cur_state, 0, sizeof (cur_state));
+ mbclen = __mbrtowc (&wc2, (const char *) pp, mlen,
+ &cur_state);
+ if (raw + offset - p <= mbclen
+ && mbclen < (size_t) -2)
+ {
+ memset (&pstr->cur_state, '\0',
+ sizeof (mbstate_t));
+ pstr->valid_len = mbclen - (raw + offset - p);
+ wc = wc2;
+ }
+ break;
+ }
+ }
+
+ if (wc == WEOF)
+ pstr->valid_len = re_string_skip_chars (pstr, idx, &wc) - idx;
+ if (wc == WEOF)
+ pstr->tip_context
+ = 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))
+ ? CONTEXT_WORD
+ : ((IS_WIDE_NEWLINE (wc)
+ && pstr->newline_anchor)
+ ? CONTEXT_NEWLINE : 0));
+ if (BE (pstr->valid_len, 0))
+ {
+ for (wcs_idx = 0; wcs_idx < pstr->valid_len; ++wcs_idx)
+ pstr->wcs[wcs_idx] = WEOF;
+ if (pstr->mbs_allocated)
+ memset (pstr->mbs, 255, pstr->valid_len);
+ }
+ pstr->valid_raw_len = pstr->valid_len;
+ }
+ else
+#endif /* RE_ENABLE_I18N */
+ {
+ int c = pstr->raw_mbs[pstr->raw_mbs_idx + offset - 1];
+ pstr->valid_raw_len = 0;
+ if (pstr->trans)
+ c = pstr->trans[c];
+ pstr->tip_context = (bitset_contain (pstr->word_char, c)
+ ? CONTEXT_WORD
+ : ((IS_NEWLINE (c) && pstr->newline_anchor)
+ ? CONTEXT_NEWLINE : 0));
+ }
+ }
+ if (!BE (pstr->mbs_allocated, 0))
+ pstr->mbs += offset;
+ }
+ pstr->raw_mbs_idx = idx;
+ pstr->len -= offset;
+ pstr->stop -= offset;
+
+ /* Then build the buffers. */
+#ifdef RE_ENABLE_I18N
+ if (pstr->mb_cur_max > 1)
+ {
+ if (pstr->icase)
+ {
+ reg_errcode_t ret = build_wcs_upper_buffer (pstr);
+ if (BE (ret != REG_NOERROR, 0))
+ return ret;
+ }
+ else
+ build_wcs_buffer (pstr);
+ }
+ else
+#endif /* RE_ENABLE_I18N */
+ if (BE (pstr->mbs_allocated, 0))
+ {
+ if (pstr->icase)
+ build_upper_buffer (pstr);
+ else if (pstr->trans != NULL)
+ re_string_translate_buffer (pstr);
+ }
+ else
+ pstr->valid_len = pstr->len;
+
+ pstr->cur_idx = 0;
+ return REG_NOERROR;
+}
+
+static unsigned char
+__attribute__ ((pure))
+re_string_peek_byte_case (const re_string_t *pstr, Idx idx)
+{
+ int ch;
+ Idx off;
+
+ /* Handle the common (easiest) cases first. */
+ if (BE (!pstr->mbs_allocated, 1))
+ return re_string_peek_byte (pstr, idx);
+
+#ifdef RE_ENABLE_I18N
+ if (pstr->mb_cur_max > 1
+ && ! re_string_is_single_byte_char (pstr, pstr->cur_idx + idx))
+ return re_string_peek_byte (pstr, idx);
+#endif
+
+ off = pstr->cur_idx + idx;
+#ifdef RE_ENABLE_I18N
+ if (pstr->offsets_needed)
+ off = pstr->offsets[off];
+#endif
+
+ ch = pstr->raw_mbs[pstr->raw_mbs_idx + off];
+
+#ifdef RE_ENABLE_I18N
+ /* Ensure that e.g. for tr_TR.UTF-8 BACKSLASH DOTLESS SMALL LETTER I
+ this function returns CAPITAL LETTER I instead of first byte of
+ DOTLESS SMALL LETTER I. The latter would confuse the parser,
+ since peek_byte_case doesn't advance cur_idx in any way. */
+ if (pstr->offsets_needed && !isascii (ch))
+ return re_string_peek_byte (pstr, idx);
+#endif
+
+ return ch;
+}
+
+static unsigned char
+re_string_fetch_byte_case (re_string_t *pstr)
+{
+ if (BE (!pstr->mbs_allocated, 1))
+ return re_string_fetch_byte (pstr);
+
+#ifdef RE_ENABLE_I18N
+ if (pstr->offsets_needed)
+ {
+ Idx off;
+ int ch;
+
+ /* For tr_TR.UTF-8 [[:islower:]] there is
+ [[: CAPITAL LETTER I WITH DOT lower:]] in mbs. Skip
+ in that case the whole multi-byte character and return
+ the original letter. On the other side, with
+ [[: DOTLESS SMALL LETTER I return [[:I, as doing
+ anything else would complicate things too much. */
+
+ if (!re_string_first_byte (pstr, pstr->cur_idx))
+ return re_string_fetch_byte (pstr);
+
+ off = pstr->offsets[pstr->cur_idx];
+ ch = pstr->raw_mbs[pstr->raw_mbs_idx + off];
+
+ if (! isascii (ch))
+ return re_string_fetch_byte (pstr);
+
+ re_string_skip_bytes (pstr,
+ re_string_char_size_at (pstr, pstr->cur_idx));
+ return ch;
+ }
+#endif
+
+ return pstr->raw_mbs[pstr->raw_mbs_idx + pstr->cur_idx++];
+}
+
+static void
+re_string_destruct (re_string_t *pstr)
+{
+#ifdef RE_ENABLE_I18N
+ re_free (pstr->wcs);
+ re_free (pstr->offsets);
+#endif /* RE_ENABLE_I18N */
+ if (pstr->mbs_allocated)
+ re_free (pstr->mbs);
+}
+
+/* Return the context at IDX in INPUT. */
+
+static unsigned int
+re_string_context_at (const re_string_t *input, Idx idx, int eflags)
+{
+ int c;
+ if (BE (idx < 0, 0))
+ /* In this case, we use the value stored in input->tip_context,
+ since we can't know the character in input->mbs[-1] here. */
+ return input->tip_context;
+ if (BE (idx == input->len, 0))
+ return ((eflags & REG_NOTEOL) ? CONTEXT_ENDBUF
+ : CONTEXT_NEWLINE | CONTEXT_ENDBUF);
+#ifdef RE_ENABLE_I18N
+ if (input->mb_cur_max > 1)
+ {
+ wint_t wc;
+ Idx wc_idx = idx;
+ while(input->wcs[wc_idx] == WEOF)
+ {
+#if defined DEBUG && DEBUG
+ /* It must not happen. */
+ assert (wc_idx >= 0);
+#endif
+ --wc_idx;
+ if (wc_idx < 0)
+ return input->tip_context;
+ }
+ wc = input->wcs[wc_idx];
+ if (BE (input->word_ops_used != 0, 0) && IS_WIDE_WORD_CHAR (wc))
+ return CONTEXT_WORD;
+ return (IS_WIDE_NEWLINE (wc) && input->newline_anchor
+ ? CONTEXT_NEWLINE : 0);
+ }
+ else
+#endif
+ {
+ c = re_string_byte_at (input, idx);
+ if (bitset_contain (input->word_char, c))
+ return CONTEXT_WORD;
+ return IS_NEWLINE (c) && input->newline_anchor ? CONTEXT_NEWLINE : 0;
+ }
+}
+
+/* Functions for set operation. */
+
+static reg_errcode_t
+__attribute_warn_unused_result__
+re_node_set_alloc (re_node_set *set, Idx size)
+{
+ set->alloc = size;
+ set->nelem = 0;
+ set->elems = re_malloc (Idx, size);
+ if (BE (set->elems == NULL, 0) && (MALLOC_0_IS_NONNULL || size != 0))
+ return REG_ESPACE;
+ return REG_NOERROR;
+}
+
+static reg_errcode_t
+__attribute_warn_unused_result__
+re_node_set_init_1 (re_node_set *set, Idx elem)
+{
+ set->alloc = 1;
+ set->nelem = 1;
+ set->elems = re_malloc (Idx, 1);
+ if (BE (set->elems == NULL, 0))
+ {
+ set->alloc = set->nelem = 0;
+ return REG_ESPACE;
+ }
+ set->elems[0] = elem;
+ return REG_NOERROR;
+}
+
+static reg_errcode_t
+__attribute_warn_unused_result__
+re_node_set_init_2 (re_node_set *set, Idx elem1, Idx elem2)
+{
+ set->alloc = 2;
+ set->elems = re_malloc (Idx, 2);
+ if (BE (set->elems == NULL, 0))
+ return REG_ESPACE;
+ if (elem1 == elem2)
+ {
+ set->nelem = 1;
+ set->elems[0] = elem1;
+ }
+ else
+ {
+ set->nelem = 2;
+ if (elem1 < elem2)
+ {
+ set->elems[0] = elem1;
+ set->elems[1] = elem2;
+ }
+ else
+ {
+ set->elems[0] = elem2;
+ set->elems[1] = elem1;
+ }
+ }
+ return REG_NOERROR;
+}
+
+static reg_errcode_t
+__attribute_warn_unused_result__
+re_node_set_init_copy (re_node_set *dest, const re_node_set *src)
+{
+ dest->nelem = src->nelem;
+ if (src->nelem > 0)
+ {
+ dest->alloc = dest->nelem;
+ dest->elems = re_malloc (Idx, dest->alloc);
+ if (BE (dest->elems == NULL, 0))
+ {
+ dest->alloc = dest->nelem = 0;
+ return REG_ESPACE;
+ }
+ memcpy (dest->elems, src->elems, src->nelem * sizeof (Idx));
+ }
+ else
+ re_node_set_init_empty (dest);
+ return REG_NOERROR;
+}
+
+/* Calculate the intersection of the sets SRC1 and SRC2. And merge it to
+ DEST. Return value indicate the error code or REG_NOERROR if succeeded.
+ Note: We assume dest->elems is NULL, when dest->alloc is 0. */
+
+static reg_errcode_t
+__attribute_warn_unused_result__
+re_node_set_add_intersect (re_node_set *dest, const re_node_set *src1,
+ const re_node_set *src2)
+{
+ Idx i1, i2, is, id, delta, sbase;
+ if (src1->nelem == 0 || src2->nelem == 0)
+ return REG_NOERROR;
+
+ /* We need dest->nelem + 2 * elems_in_intersection; this is a
+ conservative estimate. */
+ if (src1->nelem + src2->nelem + dest->nelem > dest->alloc)
+ {
+ Idx new_alloc = src1->nelem + src2->nelem + dest->alloc;
+ Idx *new_elems = re_realloc (dest->elems, Idx, new_alloc);
+ if (BE (new_elems == NULL, 0))
+ return REG_ESPACE;
+ dest->elems = new_elems;
+ dest->alloc = new_alloc;
+ }
+
+ /* Find the items in the intersection of SRC1 and SRC2, and copy
+ into the top of DEST those that are not already in DEST itself. */
+ sbase = dest->nelem + src1->nelem + src2->nelem;
+ i1 = src1->nelem - 1;
+ i2 = src2->nelem - 1;
+ id = dest->nelem - 1;
+ for (;;)
+ {
+ if (src1->elems[i1] == src2->elems[i2])
+ {
+ /* Try to find the item in DEST. Maybe we could binary search? */
+ while (id >= 0 && dest->elems[id] > src1->elems[i1])
+ --id;
+
+ if (id < 0 || dest->elems[id] != src1->elems[i1])
+ dest->elems[--sbase] = src1->elems[i1];
+
+ if (--i1 < 0 || --i2 < 0)
+ break;
+ }
+
+ /* Lower the highest of the two items. */
+ else if (src1->elems[i1] < src2->elems[i2])
+ {
+ if (--i2 < 0)
+ break;
+ }
+ else
+ {
+ if (--i1 < 0)
+ break;
+ }
+ }
+
+ id = dest->nelem - 1;
+ is = dest->nelem + src1->nelem + src2->nelem - 1;
+ delta = is - sbase + 1;
+
+ /* Now copy. When DELTA becomes zero, the remaining
+ DEST elements are already in place; this is more or
+ less the same loop that is in re_node_set_merge. */
+ dest->nelem += delta;
+ if (delta > 0 && id >= 0)
+ for (;;)
+ {
+ if (dest->elems[is] > dest->elems[id])
+ {
+ /* Copy from the top. */
+ dest->elems[id + delta--] = dest->elems[is--];
+ if (delta == 0)
+ break;
+ }
+ else
+ {
+ /* Slide from the bottom. */
+ dest->elems[id + delta] = dest->elems[id];
+ if (--id < 0)
+ break;
+ }
+ }
+
+ /* Copy remaining SRC elements. */
+ memcpy (dest->elems, dest->elems + sbase, delta * sizeof (Idx));
+
+ return REG_NOERROR;
+}
+
+/* Calculate the union set of the sets SRC1 and SRC2. And store it to
+ DEST. Return value indicate the error code or REG_NOERROR if succeeded. */
+
+static reg_errcode_t
+__attribute_warn_unused_result__
+re_node_set_init_union (re_node_set *dest, const re_node_set *src1,
+ const re_node_set *src2)
+{
+ Idx i1, i2, id;
+ if (src1 != NULL && src1->nelem > 0 && src2 != NULL && src2->nelem > 0)
+ {
+ dest->alloc = src1->nelem + src2->nelem;
+ dest->elems = re_malloc (Idx, dest->alloc);
+ if (BE (dest->elems == NULL, 0))
+ return REG_ESPACE;
+ }
+ else
+ {
+ if (src1 != NULL && src1->nelem > 0)
+ return re_node_set_init_copy (dest, src1);
+ else if (src2 != NULL && src2->nelem > 0)
+ return re_node_set_init_copy (dest, src2);
+ else
+ re_node_set_init_empty (dest);
+ return REG_NOERROR;
+ }
+ for (i1 = i2 = id = 0 ; i1 < src1->nelem && i2 < src2->nelem ;)
+ {
+ if (src1->elems[i1] > src2->elems[i2])
+ {
+ dest->elems[id++] = src2->elems[i2++];
+ continue;
+ }
+ if (src1->elems[i1] == src2->elems[i2])
+ ++i2;
+ dest->elems[id++] = src1->elems[i1++];
+ }
+ if (i1 < src1->nelem)
+ {
+ memcpy (dest->elems + id, src1->elems + i1,
+ (src1->nelem - i1) * sizeof (Idx));
+ id += src1->nelem - i1;
+ }
+ else if (i2 < src2->nelem)
+ {
+ memcpy (dest->elems + id, src2->elems + i2,
+ (src2->nelem - i2) * sizeof (Idx));
+ id += src2->nelem - i2;
+ }
+ dest->nelem = id;
+ return REG_NOERROR;
+}
+
+/* Calculate the union set of the sets DEST and SRC. And store it to
+ DEST. Return value indicate the error code or REG_NOERROR if succeeded. */
+
+static reg_errcode_t
+__attribute_warn_unused_result__
+re_node_set_merge (re_node_set *dest, const re_node_set *src)
+{
+ Idx is, id, sbase, delta;
+ if (src == NULL || src->nelem == 0)
+ return REG_NOERROR;
+ if (dest->alloc < 2 * src->nelem + dest->nelem)
+ {
+ Idx new_alloc = 2 * (src->nelem + dest->alloc);
+ Idx *new_buffer = re_realloc (dest->elems, Idx, new_alloc);
+ if (BE (new_buffer == NULL, 0))
+ return REG_ESPACE;
+ dest->elems = new_buffer;
+ dest->alloc = new_alloc;
+ }
+
+ if (BE (dest->nelem == 0, 0))
+ {
+ dest->nelem = src->nelem;
+ memcpy (dest->elems, src->elems, src->nelem * sizeof (Idx));
+ return REG_NOERROR;
+ }
+
+ /* Copy into the top of DEST the items of SRC that are not
+ found in DEST. Maybe we could binary search in DEST? */
+ for (sbase = dest->nelem + 2 * src->nelem,
+ is = src->nelem - 1, id = dest->nelem - 1; is >= 0 && id >= 0; )
+ {
+ if (dest->elems[id] == src->elems[is])
+ is--, id--;
+ else if (dest->elems[id] < src->elems[is])
+ dest->elems[--sbase] = src->elems[is--];
+ else /* if (dest->elems[id] > src->elems[is]) */
+ --id;
+ }
+
+ if (is >= 0)
+ {
+ /* If DEST is exhausted, the remaining items of SRC must be unique. */
+ sbase -= is + 1;
+ memcpy (dest->elems + sbase, src->elems, (is + 1) * sizeof (Idx));
+ }
+
+ id = dest->nelem - 1;
+ is = dest->nelem + 2 * src->nelem - 1;
+ delta = is - sbase + 1;
+ if (delta == 0)
+ return REG_NOERROR;
+
+ /* Now copy. When DELTA becomes zero, the remaining
+ DEST elements are already in place. */
+ dest->nelem += delta;
+ for (;;)
+ {
+ if (dest->elems[is] > dest->elems[id])
+ {
+ /* Copy from the top. */
+ dest->elems[id + delta--] = dest->elems[is--];
+ if (delta == 0)
+ break;
+ }
+ else
+ {
+ /* Slide from the bottom. */
+ dest->elems[id + delta] = dest->elems[id];
+ if (--id < 0)
+ {
+ /* Copy remaining SRC elements. */
+ memcpy (dest->elems, dest->elems + sbase,
+ delta * sizeof (Idx));
+ break;
+ }
+ }
+ }
+
+ return REG_NOERROR;
+}
+
+/* Insert the new element ELEM to the re_node_set* SET.
+ SET should not already have ELEM.
+ Return true if successful. */
+
+static bool
+__attribute_warn_unused_result__
+re_node_set_insert (re_node_set *set, Idx elem)
+{
+ Idx idx;
+ /* In case the set is empty. */
+ if (set->alloc == 0)
+ return BE (re_node_set_init_1 (set, elem) == REG_NOERROR, 1);
+
+ if (BE (set->nelem, 0) == 0)
+ {
+ /* We already guaranteed above that set->alloc != 0. */
+ set->elems[0] = elem;
+ ++set->nelem;
+ return true;
+ }
+
+ /* Realloc if we need. */
+ if (set->alloc == set->nelem)
+ {
+ Idx *new_elems;
+ set->alloc = set->alloc * 2;
+ new_elems = re_realloc (set->elems, Idx, set->alloc);
+ if (BE (new_elems == NULL, 0))
+ return false;
+ set->elems = new_elems;
+ }
+
+ /* Move the elements which follows the new element. Test the
+ first element separately to skip a check in the inner loop. */
+ if (elem < set->elems[0])
+ {
+ idx = 0;
+ for (idx = set->nelem; idx > 0; idx--)
+ set->elems[idx] = set->elems[idx - 1];
+ }
+ else
+ {
+ for (idx = set->nelem; set->elems[idx - 1] > elem; idx--)
+ set->elems[idx] = set->elems[idx - 1];
+ }
+
+ /* Insert the new element. */
+ set->elems[idx] = elem;
+ ++set->nelem;
+ return true;
+}
+
+/* Insert the new element ELEM to the re_node_set* SET.
+ SET should not already have any element greater than or equal to ELEM.
+ Return true if successful. */
+
+static bool
+__attribute_warn_unused_result__
+re_node_set_insert_last (re_node_set *set, Idx elem)
+{
+ /* Realloc if we need. */
+ if (set->alloc == set->nelem)
+ {
+ Idx *new_elems;
+ set->alloc = (set->alloc + 1) * 2;
+ new_elems = re_realloc (set->elems, Idx, set->alloc);
+ if (BE (new_elems == NULL, 0))
+ return false;
+ set->elems = new_elems;
+ }
+
+ /* Insert the new element. */
+ set->elems[set->nelem++] = elem;
+ return true;
+}
+
+/* Compare two node sets SET1 and SET2.
+ Return true if SET1 and SET2 are equivalent. */
+
+static bool
+__attribute__ ((pure))
+re_node_set_compare (const re_node_set *set1, const re_node_set *set2)
+{
+ Idx i;
+ if (set1 == NULL || set2 == NULL || set1->nelem != set2->nelem)
+ return false;
+ for (i = set1->nelem ; --i >= 0 ; )
+ if (set1->elems[i] != set2->elems[i])
+ return false;
+ return true;
+}
+
+/* Return (idx + 1) if SET contains the element ELEM, return 0 otherwise. */
+
+static Idx
+__attribute__ ((pure))
+re_node_set_contains (const re_node_set *set, Idx elem)
+{
+ __re_size_t idx, right, mid;
+ if (set->nelem <= 0)
+ return 0;
+
+ /* Binary search the element. */
+ idx = 0;
+ right = set->nelem - 1;
+ while (idx < right)
+ {
+ mid = (idx + right) / 2;
+ if (set->elems[mid] < elem)
+ idx = mid + 1;
+ else
+ right = mid;
+ }
+ return set->elems[idx] == elem ? idx + 1 : 0;
+}
+
+static void
+re_node_set_remove_at (re_node_set *set, Idx idx)
+{
+ if (idx < 0 || idx >= set->nelem)
+ return;
+ --set->nelem;
+ for (; idx < set->nelem; idx++)
+ set->elems[idx] = set->elems[idx + 1];
+}
+
+
+/* Add the token TOKEN to dfa->nodes, and return the index of the token.
+ Or return -1 if an error occurred. */
+
+static Idx
+re_dfa_add_node (re_dfa_t *dfa, re_token_t token)
+{
+ if (BE (dfa->nodes_len >= dfa->nodes_alloc, 0))
+ {
+ size_t new_nodes_alloc = dfa->nodes_alloc * 2;
+ Idx *new_nexts, *new_indices;
+ re_node_set *new_edests, *new_eclosures;
+ re_token_t *new_nodes;
+
+ /* Avoid overflows in realloc. */
+ const size_t max_object_size = MAX (sizeof (re_token_t),
+ MAX (sizeof (re_node_set),
+ sizeof (Idx)));
+ if (BE (MIN (IDX_MAX, SIZE_MAX / max_object_size) < new_nodes_alloc, 0))
+ return -1;
+
+ new_nodes = re_realloc (dfa->nodes, re_token_t, new_nodes_alloc);
+ if (BE (new_nodes == NULL, 0))
+ return -1;
+ dfa->nodes = new_nodes;
+ new_nexts = re_realloc (dfa->nexts, Idx, new_nodes_alloc);
+ new_indices = re_realloc (dfa->org_indices, Idx, new_nodes_alloc);
+ new_edests = re_realloc (dfa->edests, re_node_set, new_nodes_alloc);
+ new_eclosures = re_realloc (dfa->eclosures, re_node_set,
new_nodes_alloc);
+ if (BE (new_nexts == NULL || new_indices == NULL
+ || new_edests == NULL || new_eclosures == NULL, 0))
+ {
+ re_free (new_nexts);
+ re_free (new_indices);
+ re_free (new_edests);
+ re_free (new_eclosures);
+ return -1;
+ }
+ dfa->nexts = new_nexts;
+ dfa->org_indices = new_indices;
+ dfa->edests = new_edests;
+ dfa->eclosures = new_eclosures;
+ dfa->nodes_alloc = new_nodes_alloc;
+ }
+ dfa->nodes[dfa->nodes_len] = token;
+ dfa->nodes[dfa->nodes_len].constraint = 0;
+#ifdef RE_ENABLE_I18N
+ dfa->nodes[dfa->nodes_len].accept_mb =
+ ((token.type == OP_PERIOD && dfa->mb_cur_max > 1)
+ || token.type == COMPLEX_BRACKET);
+#endif
+ dfa->nexts[dfa->nodes_len] = -1;
+ re_node_set_init_empty (dfa->edests + dfa->nodes_len);
+ re_node_set_init_empty (dfa->eclosures + dfa->nodes_len);
+ return dfa->nodes_len++;
+}
+
+static re_hashval_t
+calc_state_hash (const re_node_set *nodes, unsigned int context)
+{
+ re_hashval_t hash = nodes->nelem + context;
+ Idx i;
+ for (i = 0 ; i < nodes->nelem ; i++)
+ hash += nodes->elems[i];
+ return hash;
+}
+
+/* Search for the state whose node_set is equivalent to NODES.
+ Return the pointer to the state, if we found it in the DFA.
+ Otherwise create the new one and return it. In case of an error
+ return NULL and set the error code in ERR.
+ Note: - We assume NULL as the invalid state, then it is possible that
+ return value is NULL and ERR is REG_NOERROR.
+ - We never return non-NULL value in case of any errors, it is for
+ optimization. */
+
+static re_dfastate_t *
+__attribute_warn_unused_result__
+re_acquire_state (reg_errcode_t *err, const re_dfa_t *dfa,
+ const re_node_set *nodes)
+{
+ re_hashval_t hash;
+ re_dfastate_t *new_state;
+ struct re_state_table_entry *spot;
+ Idx i;
+#if defined GCC_LINT || defined lint
+ /* Suppress bogus uninitialized-variable warnings. */
+ *err = REG_NOERROR;
+#endif
+ if (BE (nodes->nelem == 0, 0))
+ {
+ *err = REG_NOERROR;
+ return NULL;
+ }
+ hash = calc_state_hash (nodes, 0);
+ spot = dfa->state_table + (hash & dfa->state_hash_mask);
+
+ for (i = 0 ; i < spot->num ; i++)
+ {
+ re_dfastate_t *state = spot->array[i];
+ if (hash != state->hash)
+ continue;
+ if (re_node_set_compare (&state->nodes, nodes))
+ return state;
+ }
+
+ /* There are no appropriate state in the dfa, create the new one. */
+ new_state = create_ci_newstate (dfa, nodes, hash);
+ if (BE (new_state == NULL, 0))
+ *err = REG_ESPACE;
+
+ return new_state;
+}
+
+/* Search for the state whose node_set is equivalent to NODES and
+ whose context is equivalent to CONTEXT.
+ Return the pointer to the state, if we found it in the DFA.
+ Otherwise create the new one and return it. In case of an error
+ return NULL and set the error code in ERR.
+ Note: - We assume NULL as the invalid state, then it is possible that
+ return value is NULL and ERR is REG_NOERROR.
+ - We never return non-NULL value in case of any errors, it is for
+ optimization. */
+
+static re_dfastate_t *
+__attribute_warn_unused_result__
+re_acquire_state_context (reg_errcode_t *err, const re_dfa_t *dfa,
+ const re_node_set *nodes, unsigned int context)
+{
+ re_hashval_t hash;
+ re_dfastate_t *new_state;
+ struct re_state_table_entry *spot;
+ Idx i;
+#if defined GCC_LINT || defined lint
+ /* Suppress bogus uninitialized-variable warnings. */
+ *err = REG_NOERROR;
+#endif
+ if (nodes->nelem == 0)
+ {
+ *err = REG_NOERROR;
+ return NULL;
+ }
+ hash = calc_state_hash (nodes, context);
+ spot = dfa->state_table + (hash & dfa->state_hash_mask);
+
+ for (i = 0 ; i < spot->num ; i++)
+ {
+ re_dfastate_t *state = spot->array[i];
+ if (state->hash == hash
+ && state->context == context
+ && re_node_set_compare (state->entrance_nodes, nodes))
+ return state;
+ }
+ /* There are no appropriate state in 'dfa', create the new one. */
+ new_state = create_cd_newstate (dfa, nodes, context, hash);
+ if (BE (new_state == NULL, 0))
+ *err = REG_ESPACE;
+
+ return new_state;
+}
+
+/* Finish initialization of the new state NEWSTATE, and using its hash value
+ HASH put in the appropriate bucket of DFA's state table. Return value
+ indicates the error code if failed. */
+
+static reg_errcode_t
+__attribute_warn_unused_result__
+register_state (const re_dfa_t *dfa, re_dfastate_t *newstate,
+ re_hashval_t hash)
+{
+ struct re_state_table_entry *spot;
+ reg_errcode_t err;
+ Idx i;
+
+ newstate->hash = hash;
+ err = re_node_set_alloc (&newstate->non_eps_nodes, newstate->nodes.nelem);
+ if (BE (err != REG_NOERROR, 0))
+ return REG_ESPACE;
+ for (i = 0; i < newstate->nodes.nelem; i++)
+ {
+ Idx elem = newstate->nodes.elems[i];
+ if (!IS_EPSILON_NODE (dfa->nodes[elem].type))
+ if (! re_node_set_insert_last (&newstate->non_eps_nodes, elem))
+ return REG_ESPACE;
+ }
+
+ spot = dfa->state_table + (hash & dfa->state_hash_mask);
+ if (BE (spot->alloc <= spot->num, 0))
+ {
+ Idx new_alloc = 2 * spot->num + 2;
+ re_dfastate_t **new_array = re_realloc (spot->array, re_dfastate_t *,
+ new_alloc);
+ if (BE (new_array == NULL, 0))
+ return REG_ESPACE;
+ spot->array = new_array;
+ spot->alloc = new_alloc;
+ }
+ spot->array[spot->num++] = newstate;
+ return REG_NOERROR;
+}
+
+static void
+free_state (re_dfastate_t *state)
+{
+ re_node_set_free (&state->non_eps_nodes);
+ re_node_set_free (&state->inveclosure);
+ if (state->entrance_nodes != &state->nodes)
+ {
+ re_node_set_free (state->entrance_nodes);
+ re_free (state->entrance_nodes);
+ }
+ re_node_set_free (&state->nodes);
+ re_free (state->word_trtable);
+ re_free (state->trtable);
+ re_free (state);
+}
+
+/* Create the new state which is independent of contexts.
+ Return the new state if succeeded, otherwise return NULL. */
+
+static re_dfastate_t *
+__attribute_warn_unused_result__
+create_ci_newstate (const re_dfa_t *dfa, const re_node_set *nodes,
+ re_hashval_t hash)
+{
+ Idx i;
+ reg_errcode_t err;
+ re_dfastate_t *newstate;
+
+ newstate = (re_dfastate_t *) calloc (sizeof (re_dfastate_t), 1);
+ if (BE (newstate == NULL, 0))
+ return NULL;
+ err = re_node_set_init_copy (&newstate->nodes, nodes);
+ if (BE (err != REG_NOERROR, 0))
+ {
+ re_free (newstate);
+ return NULL;
+ }
+
+ newstate->entrance_nodes = &newstate->nodes;
+ for (i = 0 ; i < nodes->nelem ; i++)
+ {
+ re_token_t *node = dfa->nodes + nodes->elems[i];
+ re_token_type_t type = node->type;
+ if (type == CHARACTER && !node->constraint)
+ continue;
+#ifdef RE_ENABLE_I18N
+ newstate->accept_mb |= node->accept_mb;
+#endif /* RE_ENABLE_I18N */
+
+ /* If the state has the halt node, the state is a halt state. */
+ if (type == END_OF_RE)
+ newstate->halt = 1;
+ else if (type == OP_BACK_REF)
+ newstate->has_backref = 1;
+ else if (type == ANCHOR || node->constraint)
+ newstate->has_constraint = 1;
+ }
+ err = register_state (dfa, newstate, hash);
+ if (BE (err != REG_NOERROR, 0))
+ {
+ free_state (newstate);
+ newstate = NULL;
+ }
+ return newstate;
+}
+
+/* Create the new state which is depend on the context CONTEXT.
+ Return the new state if succeeded, otherwise return NULL. */
+
+static re_dfastate_t *
+__attribute_warn_unused_result__
+create_cd_newstate (const re_dfa_t *dfa, const re_node_set *nodes,
+ unsigned int context, re_hashval_t hash)
+{
+ Idx i, nctx_nodes = 0;
+ reg_errcode_t err;
+ re_dfastate_t *newstate;
+
+ newstate = (re_dfastate_t *) calloc (sizeof (re_dfastate_t), 1);
+ if (BE (newstate == NULL, 0))
+ return NULL;
+ err = re_node_set_init_copy (&newstate->nodes, nodes);
+ if (BE (err != REG_NOERROR, 0))
+ {
+ re_free (newstate);
+ return NULL;
+ }
+
+ newstate->context = context;
+ newstate->entrance_nodes = &newstate->nodes;
+
+ for (i = 0 ; i < nodes->nelem ; i++)
+ {
+ re_token_t *node = dfa->nodes + nodes->elems[i];
+ re_token_type_t type = node->type;
+ unsigned int constraint = node->constraint;
+
+ if (type == CHARACTER && !constraint)
+ continue;
+#ifdef RE_ENABLE_I18N
+ newstate->accept_mb |= node->accept_mb;
+#endif /* RE_ENABLE_I18N */
+
+ /* If the state has the halt node, the state is a halt state. */
+ if (type == END_OF_RE)
+ newstate->halt = 1;
+ else if (type == OP_BACK_REF)
+ newstate->has_backref = 1;
+
+ if (constraint)
+ {
+ if (newstate->entrance_nodes == &newstate->nodes)
+ {
+ newstate->entrance_nodes = re_malloc (re_node_set, 1);
+ if (BE (newstate->entrance_nodes == NULL, 0))
+ {
+ free_state (newstate);
+ return NULL;
+ }
+ if (re_node_set_init_copy (newstate->entrance_nodes, nodes)
+ != REG_NOERROR)
+ return NULL;
+ nctx_nodes = 0;
+ newstate->has_constraint = 1;
+ }
+
+ if (NOT_SATISFY_PREV_CONSTRAINT (constraint,context))
+ {
+ re_node_set_remove_at (&newstate->nodes, i - nctx_nodes);
+ ++nctx_nodes;
+ }
+ }
+ }
+ err = register_state (dfa, newstate, hash);
+ if (BE (err != REG_NOERROR, 0))
+ {
+ free_state (newstate);
+ newstate = NULL;
+ }
+ return newstate;
+}
diff --git a/lib/regex_internal.h b/lib/regex_internal.h
new file mode 100644
index 0000000..7bbe802
--- /dev/null
+++ b/lib/regex_internal.h
@@ -0,0 +1,911 @@
+/* Extended regular expression matching and search library.
+ Copyright (C) 2002-2018 Free Software Foundation, Inc.
+ This file is part of the GNU C Library.
+ Contributed by Isamu Hasegawa <address@hidden>.
+
+ The GNU C Library 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 3 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
+ General Public License for more details.
+
+ You should have received a copy of the GNU General Public
+ License along with the GNU C Library; if not, see
+ <https://www.gnu.org/licenses/>. */
+
+#ifndef _REGEX_INTERNAL_H
+#define _REGEX_INTERNAL_H 1
+
+#include <assert.h>
+#include <ctype.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+
+#include <langinfo.h>
+#include <locale.h>
+#include <wchar.h>
+#include <wctype.h>
+#include <stdbool.h>
+#include <stdint.h>
+
+/* Properties of integers. Although Gnulib has intprops.h, glibc does
+ without for now. */
+#ifndef _LIBC
+# include "intprops.h"
+#else
+/* True if the real type T is signed. */
+# define TYPE_SIGNED(t) (! ((t) 0 < (t) -1))
+
+/* True if adding the nonnegative Idx values A and B would overflow.
+ If false, set *R to A + B. A, B, and R may be evaluated more than
+ once, or zero times. Although this is not a full implementation of
+ Gnulib INT_ADD_WRAPV, it is good enough for glibc regex code.
+ FIXME: This implementation is a fragile stopgap, and this file would
+ be simpler and more robust if intprops.h were migrated into glibc. */
+# define INT_ADD_WRAPV(a, b, r) \
+ (IDX_MAX - (a) < (b) ? true : (*(r) = (a) + (b), false))
+#endif
+
+#ifdef _LIBC
+# include <libc-lock.h>
+# define lock_define(name) __libc_lock_define (, name)
+# define lock_init(lock) (__libc_lock_init (lock), 0)
+# define lock_fini(lock) ((void) 0)
+# define lock_lock(lock) __libc_lock_lock (lock)
+# define lock_unlock(lock) __libc_lock_unlock (lock)
+#elif defined GNULIB_LOCK && !defined USE_UNLOCKED_IO
+# include "glthread/lock.h"
+ /* Use gl_lock_define if empty macro arguments are known to work.
+ Otherwise, fall back on less-portable substitutes. */
+# if ((defined __GNUC__ && !defined __STRICT_ANSI__) \
+ || (defined __STDC_VERSION__ && 199901L <= __STDC_VERSION__))
+# define lock_define(name) gl_lock_define (, name)
+# elif USE_POSIX_THREADS
+# define lock_define(name) pthread_mutex_t name;
+# elif USE_PTH_THREADS
+# define lock_define(name) pth_mutex_t name;
+# elif USE_SOLARIS_THREADS
+# define lock_define(name) mutex_t name;
+# elif USE_WINDOWS_THREADS
+# define lock_define(name) gl_lock_t name;
+# else
+# define lock_define(name)
+# endif
+# define lock_init(lock) glthread_lock_init (&(lock))
+# define lock_fini(lock) glthread_lock_destroy (&(lock))
+# define lock_lock(lock) glthread_lock_lock (&(lock))
+# define lock_unlock(lock) glthread_lock_unlock (&(lock))
+#elif defined GNULIB_PTHREAD && !defined USE_UNLOCKED_IO
+# include <pthread.h>
+# define lock_define(name) pthread_mutex_t name;
+# define lock_init(lock) pthread_mutex_init (&(lock), 0)
+# define lock_fini(lock) pthread_mutex_destroy (&(lock))
+# define lock_lock(lock) pthread_mutex_lock (&(lock))
+# define lock_unlock(lock) pthread_mutex_unlock (&(lock))
+#else
+# define lock_define(name)
+# define lock_init(lock) 0
+# define lock_fini(lock) ((void) 0)
+ /* The 'dfa' avoids an "unused variable 'dfa'" warning from GCC. */
+# define lock_lock(lock) ((void) dfa)
+# define lock_unlock(lock) ((void) 0)
+#endif
+
+/* In case that the system doesn't have isblank(). */
+#if !defined _LIBC && ! (defined isblank || (HAVE_ISBLANK &&
HAVE_DECL_ISBLANK))
+# define isblank(ch) ((ch) == ' ' || (ch) == '\t')
+#endif
+
+#ifdef _LIBC
+# ifndef _RE_DEFINE_LOCALE_FUNCTIONS
+# define _RE_DEFINE_LOCALE_FUNCTIONS 1
+# include <locale/localeinfo.h>
+# include <locale/coll-lookup.h>
+# endif
+#endif
+
+/* This is for other GNU distributions with internationalized messages. */
+#if (HAVE_LIBINTL_H && ENABLE_NLS) || defined _LIBC
+# include <libintl.h>
+# ifdef _LIBC
+# undef gettext
+# define gettext(msgid) \
+ __dcgettext (_libc_intl_domainname, msgid, LC_MESSAGES)
+# endif
+#else
+# undef gettext
+# define gettext(msgid) (msgid)
+#endif
+
+#ifndef gettext_noop
+/* This define is so xgettext can find the internationalizable
+ strings. */
+# define gettext_noop(String) String
+#endif
+
+#if (defined MB_CUR_MAX && HAVE_WCTYPE_H && HAVE_ISWCTYPE) || _LIBC
+# define RE_ENABLE_I18N
+#endif
+
+#define BE(expr, val) __builtin_expect (expr, val)
+
+/* Number of ASCII characters. */
+#define ASCII_CHARS 0x80
+
+/* Number of single byte characters. */
+#define SBC_MAX (UCHAR_MAX + 1)
+
+#define COLL_ELEM_LEN_MAX 8
+
+/* The character which represents newline. */
+#define NEWLINE_CHAR '\n'
+#define WIDE_NEWLINE_CHAR L'\n'
+
+/* Rename to standard API for using out of glibc. */
+#ifndef _LIBC
+# undef __wctype
+# undef __iswctype
+# define __wctype wctype
+# define __iswalnum iswalnum
+# define __iswctype iswctype
+# define __towlower towlower
+# define __towupper towupper
+# define __btowc btowc
+# define __mbrtowc mbrtowc
+# define __wcrtomb wcrtomb
+# define __regfree regfree
+# define attribute_hidden
+#endif /* not _LIBC */
+
+#if __GNUC__ < 3 + (__GNUC_MINOR__ < 1)
+# define __attribute__(arg)
+#endif
+
+#ifndef SSIZE_MAX
+# define SSIZE_MAX ((ssize_t) (SIZE_MAX / 2))
+#endif
+
+/* The type of indexes into strings. This is signed, not size_t,
+ since the API requires indexes to fit in regoff_t anyway, and using
+ signed integers makes the code a bit smaller and presumably faster.
+ The traditional GNU regex implementation uses int for indexes.
+ The POSIX-compatible implementation uses a possibly-wider type.
+ The name 'Idx' is three letters to minimize the hassle of
+ reindenting a lot of regex code that formerly used 'int'. */
+typedef regoff_t Idx;
+#ifdef _REGEX_LARGE_OFFSETS
+# define IDX_MAX SSIZE_MAX
+#else
+# define IDX_MAX INT_MAX
+#endif
+
+/* A hash value, suitable for computing hash tables. */
+typedef __re_size_t re_hashval_t;
+
+/* An integer used to represent a set of bits. It must be unsigned,
+ and must be at least as wide as unsigned int. */
+typedef unsigned long int bitset_word_t;
+/* All bits set in a bitset_word_t. */
+#define BITSET_WORD_MAX ULONG_MAX
+
+/* Number of bits in a bitset_word_t. For portability to hosts with
+ padding bits, do not use '(sizeof (bitset_word_t) * CHAR_BIT)';
+ instead, deduce it directly from BITSET_WORD_MAX. Avoid
+ greater-than-32-bit integers and unconditional shifts by more than
+ 31 bits, as they're not portable. */
+#if BITSET_WORD_MAX == 0xffffffffUL
+# define BITSET_WORD_BITS 32
+#elif BITSET_WORD_MAX >> 31 >> 4 == 1
+# define BITSET_WORD_BITS 36
+#elif BITSET_WORD_MAX >> 31 >> 16 == 1
+# define BITSET_WORD_BITS 48
+#elif BITSET_WORD_MAX >> 31 >> 28 == 1
+# define BITSET_WORD_BITS 60
+#elif BITSET_WORD_MAX >> 31 >> 31 >> 1 == 1
+# define BITSET_WORD_BITS 64
+#elif BITSET_WORD_MAX >> 31 >> 31 >> 9 == 1
+# define BITSET_WORD_BITS 72
+#elif BITSET_WORD_MAX >> 31 >> 31 >> 31 >> 31 >> 3 == 1
+# define BITSET_WORD_BITS 128
+#elif BITSET_WORD_MAX >> 31 >> 31 >> 31 >> 31 >> 31 >> 31 >> 31 >> 31 >> 7 == 1
+# define BITSET_WORD_BITS 256
+#elif BITSET_WORD_MAX >> 31 >> 31 >> 31 >> 31 >> 31 >> 31 >> 31 >> 31 >> 7 > 1
+# define BITSET_WORD_BITS 257 /* any value > SBC_MAX will do here */
+# if BITSET_WORD_BITS <= SBC_MAX
+# error "Invalid SBC_MAX"
+# endif
+#else
+# error "Add case for new bitset_word_t size"
+#endif
+
+/* Number of bitset_word_t values in a bitset_t. */
+#define BITSET_WORDS ((SBC_MAX + BITSET_WORD_BITS - 1) / BITSET_WORD_BITS)
+
+typedef bitset_word_t bitset_t[BITSET_WORDS];
+typedef bitset_word_t *re_bitset_ptr_t;
+typedef const bitset_word_t *re_const_bitset_ptr_t;
+
+#define PREV_WORD_CONSTRAINT 0x0001
+#define PREV_NOTWORD_CONSTRAINT 0x0002
+#define NEXT_WORD_CONSTRAINT 0x0004
+#define NEXT_NOTWORD_CONSTRAINT 0x0008
+#define PREV_NEWLINE_CONSTRAINT 0x0010
+#define NEXT_NEWLINE_CONSTRAINT 0x0020
+#define PREV_BEGBUF_CONSTRAINT 0x0040
+#define NEXT_ENDBUF_CONSTRAINT 0x0080
+#define WORD_DELIM_CONSTRAINT 0x0100
+#define NOT_WORD_DELIM_CONSTRAINT 0x0200
+
+typedef enum
+{
+ INSIDE_WORD = PREV_WORD_CONSTRAINT | NEXT_WORD_CONSTRAINT,
+ WORD_FIRST = PREV_NOTWORD_CONSTRAINT | NEXT_WORD_CONSTRAINT,
+ WORD_LAST = PREV_WORD_CONSTRAINT | NEXT_NOTWORD_CONSTRAINT,
+ INSIDE_NOTWORD = PREV_NOTWORD_CONSTRAINT | NEXT_NOTWORD_CONSTRAINT,
+ LINE_FIRST = PREV_NEWLINE_CONSTRAINT,
+ LINE_LAST = NEXT_NEWLINE_CONSTRAINT,
+ BUF_FIRST = PREV_BEGBUF_CONSTRAINT,
+ BUF_LAST = NEXT_ENDBUF_CONSTRAINT,
+ WORD_DELIM = WORD_DELIM_CONSTRAINT,
+ NOT_WORD_DELIM = NOT_WORD_DELIM_CONSTRAINT
+} re_context_type;
+
+typedef struct
+{
+ Idx alloc;
+ Idx nelem;
+ Idx *elems;
+} re_node_set;
+
+typedef enum
+{
+ NON_TYPE = 0,
+
+ /* Node type, These are used by token, node, tree. */
+ CHARACTER = 1,
+ END_OF_RE = 2,
+ SIMPLE_BRACKET = 3,
+ OP_BACK_REF = 4,
+ OP_PERIOD = 5,
+#ifdef RE_ENABLE_I18N
+ COMPLEX_BRACKET = 6,
+ OP_UTF8_PERIOD = 7,
+#endif /* RE_ENABLE_I18N */
+
+ /* We define EPSILON_BIT as a macro so that OP_OPEN_SUBEXP is used
+ when the debugger shows values of this enum type. */
+#define EPSILON_BIT 8
+ OP_OPEN_SUBEXP = EPSILON_BIT | 0,
+ OP_CLOSE_SUBEXP = EPSILON_BIT | 1,
+ OP_ALT = EPSILON_BIT | 2,
+ OP_DUP_ASTERISK = EPSILON_BIT | 3,
+ ANCHOR = EPSILON_BIT | 4,
+
+ /* Tree type, these are used only by tree. */
+ CONCAT = 16,
+ SUBEXP = 17,
+
+ /* Token type, these are used only by token. */
+ OP_DUP_PLUS = 18,
+ OP_DUP_QUESTION,
+ OP_OPEN_BRACKET,
+ OP_CLOSE_BRACKET,
+ OP_CHARSET_RANGE,
+ OP_OPEN_DUP_NUM,
+ OP_CLOSE_DUP_NUM,
+ OP_NON_MATCH_LIST,
+ OP_OPEN_COLL_ELEM,
+ OP_CLOSE_COLL_ELEM,
+ OP_OPEN_EQUIV_CLASS,
+ OP_CLOSE_EQUIV_CLASS,
+ OP_OPEN_CHAR_CLASS,
+ OP_CLOSE_CHAR_CLASS,
+ OP_WORD,
+ OP_NOTWORD,
+ OP_SPACE,
+ OP_NOTSPACE,
+ BACK_SLASH
+
+} re_token_type_t;
+
+#ifdef RE_ENABLE_I18N
+typedef struct
+{
+ /* Multibyte characters. */
+ wchar_t *mbchars;
+
+ /* Collating symbols. */
+# ifdef _LIBC
+ int32_t *coll_syms;
+# endif
+
+ /* Equivalence classes. */
+# ifdef _LIBC
+ int32_t *equiv_classes;
+# endif
+
+ /* Range expressions. */
+# ifdef _LIBC
+ uint32_t *range_starts;
+ uint32_t *range_ends;
+# else /* not _LIBC */
+ wchar_t *range_starts;
+ wchar_t *range_ends;
+# endif /* not _LIBC */
+
+ /* Character classes. */
+ wctype_t *char_classes;
+
+ /* If this character set is the non-matching list. */
+ unsigned int non_match : 1;
+
+ /* # of multibyte characters. */
+ Idx nmbchars;
+
+ /* # of collating symbols. */
+ Idx ncoll_syms;
+
+ /* # of equivalence classes. */
+ Idx nequiv_classes;
+
+ /* # of range expressions. */
+ Idx nranges;
+
+ /* # of character classes. */
+ Idx nchar_classes;
+} re_charset_t;
+#endif /* RE_ENABLE_I18N */
+
+typedef struct
+{
+ union
+ {
+ unsigned char c; /* for CHARACTER */
+ re_bitset_ptr_t sbcset; /* for SIMPLE_BRACKET */
+#ifdef RE_ENABLE_I18N
+ re_charset_t *mbcset; /* for COMPLEX_BRACKET */
+#endif /* RE_ENABLE_I18N */
+ Idx idx; /* for BACK_REF */
+ re_context_type ctx_type; /* for ANCHOR */
+ } opr;
+#if __GNUC__ >= 2 && !defined __STRICT_ANSI__
+ re_token_type_t type : 8;
+#else
+ re_token_type_t type;
+#endif
+ unsigned int constraint : 10; /* context constraint */
+ unsigned int duplicated : 1;
+ unsigned int opt_subexp : 1;
+#ifdef RE_ENABLE_I18N
+ unsigned int accept_mb : 1;
+ /* These 2 bits can be moved into the union if needed (e.g. if running out
+ of bits; move opr.c to opr.c.c and move the flags to opr.c.flags). */
+ unsigned int mb_partial : 1;
+#endif
+ unsigned int word_char : 1;
+} re_token_t;
+
+#define IS_EPSILON_NODE(type) ((type) & EPSILON_BIT)
+
+struct re_string_t
+{
+ /* Indicate the raw buffer which is the original string passed as an
+ argument of regexec(), re_search(), etc.. */
+ const unsigned char *raw_mbs;
+ /* Store the multibyte string. In case of "case insensitive mode" like
+ REG_ICASE, upper cases of the string are stored, otherwise MBS points
+ the same address that RAW_MBS points. */
+ unsigned char *mbs;
+#ifdef RE_ENABLE_I18N
+ /* Store the wide character string which is corresponding to MBS. */
+ wint_t *wcs;
+ Idx *offsets;
+ mbstate_t cur_state;
+#endif
+ /* Index in RAW_MBS. Each character mbs[i] corresponds to
+ raw_mbs[raw_mbs_idx + i]. */
+ Idx raw_mbs_idx;
+ /* The length of the valid characters in the buffers. */
+ Idx valid_len;
+ /* The corresponding number of bytes in raw_mbs array. */
+ Idx valid_raw_len;
+ /* The length of the buffers MBS and WCS. */
+ Idx bufs_len;
+ /* The index in MBS, which is updated by re_string_fetch_byte. */
+ Idx cur_idx;
+ /* length of RAW_MBS array. */
+ Idx raw_len;
+ /* This is RAW_LEN - RAW_MBS_IDX + VALID_LEN - VALID_RAW_LEN. */
+ Idx len;
+ /* End of the buffer may be shorter than its length in the cases such
+ as re_match_2, re_search_2. Then, we use STOP for end of the buffer
+ instead of LEN. */
+ Idx raw_stop;
+ /* This is RAW_STOP - RAW_MBS_IDX adjusted through OFFSETS. */
+ Idx stop;
+
+ /* The context of mbs[0]. We store the context independently, since
+ the context of mbs[0] may be different from raw_mbs[0], which is
+ the beginning of the input string. */
+ unsigned int tip_context;
+ /* The translation passed as a part of an argument of re_compile_pattern. */
+ RE_TRANSLATE_TYPE trans;
+ /* Copy of re_dfa_t's word_char. */
+ re_const_bitset_ptr_t word_char;
+ /* true if REG_ICASE. */
+ unsigned char icase;
+ unsigned char is_utf8;
+ unsigned char map_notascii;
+ unsigned char mbs_allocated;
+ unsigned char offsets_needed;
+ unsigned char newline_anchor;
+ unsigned char word_ops_used;
+ int mb_cur_max;
+};
+typedef struct re_string_t re_string_t;
+
+
+struct re_dfa_t;
+typedef struct re_dfa_t re_dfa_t;
+
+#ifndef _LIBC
+# define IS_IN(libc) false
+#endif
+
+#define re_string_peek_byte(pstr, offset) \
+ ((pstr)->mbs[(pstr)->cur_idx + offset])
+#define re_string_fetch_byte(pstr) \
+ ((pstr)->mbs[(pstr)->cur_idx++])
+#define re_string_first_byte(pstr, idx) \
+ ((idx) == (pstr)->valid_len || (pstr)->wcs[idx] != WEOF)
+#define re_string_is_single_byte_char(pstr, idx) \
+ ((pstr)->wcs[idx] != WEOF && ((pstr)->valid_len == (idx) + 1 \
+ || (pstr)->wcs[(idx) + 1] != WEOF))
+#define re_string_eoi(pstr) ((pstr)->stop <= (pstr)->cur_idx)
+#define re_string_cur_idx(pstr) ((pstr)->cur_idx)
+#define re_string_get_buffer(pstr) ((pstr)->mbs)
+#define re_string_length(pstr) ((pstr)->len)
+#define re_string_byte_at(pstr,idx) ((pstr)->mbs[idx])
+#define re_string_skip_bytes(pstr,idx) ((pstr)->cur_idx += (idx))
+#define re_string_set_index(pstr,idx) ((pstr)->cur_idx = (idx))
+
+#if defined _LIBC || HAVE_ALLOCA
+# include <alloca.h>
+#endif
+
+#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
+/* alloca is implemented with malloc, so just use malloc. */
+# define __libc_use_alloca(n) 0
+# undef alloca
+# define alloca(n) malloc (n)
+# endif
+#endif
+
+#ifdef _LIBC
+# define MALLOC_0_IS_NONNULL 1
+#elif !defined MALLOC_0_IS_NONNULL
+# define MALLOC_0_IS_NONNULL 0
+#endif
+
+#ifndef MAX
+# define MAX(a,b) ((a) < (b) ? (b) : (a))
+#endif
+#ifndef MIN
+# define MIN(a,b) ((a) < (b) ? (a) : (b))
+#endif
+
+#define re_malloc(t,n) ((t *) malloc ((n) * sizeof (t)))
+#define re_realloc(p,t,n) ((t *) realloc (p, (n) * sizeof (t)))
+#define re_free(p) free (p)
+
+struct bin_tree_t
+{
+ struct bin_tree_t *parent;
+ struct bin_tree_t *left;
+ struct bin_tree_t *right;
+ struct bin_tree_t *first;
+ struct bin_tree_t *next;
+
+ re_token_t token;
+
+ /* 'node_idx' is the index in dfa->nodes, if 'type' == 0.
+ Otherwise 'type' indicate the type of this node. */
+ Idx node_idx;
+};
+typedef struct bin_tree_t bin_tree_t;
+
+#define BIN_TREE_STORAGE_SIZE \
+ ((1024 - sizeof (void *)) / sizeof (bin_tree_t))
+
+struct bin_tree_storage_t
+{
+ struct bin_tree_storage_t *next;
+ bin_tree_t data[BIN_TREE_STORAGE_SIZE];
+};
+typedef struct bin_tree_storage_t bin_tree_storage_t;
+
+#define CONTEXT_WORD 1
+#define CONTEXT_NEWLINE (CONTEXT_WORD << 1)
+#define CONTEXT_BEGBUF (CONTEXT_NEWLINE << 1)
+#define CONTEXT_ENDBUF (CONTEXT_BEGBUF << 1)
+
+#define IS_WORD_CONTEXT(c) ((c) & CONTEXT_WORD)
+#define IS_NEWLINE_CONTEXT(c) ((c) & CONTEXT_NEWLINE)
+#define IS_BEGBUF_CONTEXT(c) ((c) & CONTEXT_BEGBUF)
+#define IS_ENDBUF_CONTEXT(c) ((c) & CONTEXT_ENDBUF)
+#define IS_ORDINARY_CONTEXT(c) ((c) == 0)
+
+#define IS_WORD_CHAR(ch) (isalnum (ch) || (ch) == '_')
+#define IS_NEWLINE(ch) ((ch) == NEWLINE_CHAR)
+#define IS_WIDE_WORD_CHAR(ch) (__iswalnum (ch) || (ch) == L'_')
+#define IS_WIDE_NEWLINE(ch) ((ch) == WIDE_NEWLINE_CHAR)
+
+#define NOT_SATISFY_PREV_CONSTRAINT(constraint,context) \
+ ((((constraint) & PREV_WORD_CONSTRAINT) && !IS_WORD_CONTEXT (context)) \
+ || ((constraint & PREV_NOTWORD_CONSTRAINT) && IS_WORD_CONTEXT (context)) \
+ || ((constraint & PREV_NEWLINE_CONSTRAINT) && !IS_NEWLINE_CONTEXT (context))\
+ || ((constraint & PREV_BEGBUF_CONSTRAINT) && !IS_BEGBUF_CONTEXT (context)))
+
+#define NOT_SATISFY_NEXT_CONSTRAINT(constraint,context) \
+ ((((constraint) & NEXT_WORD_CONSTRAINT) && !IS_WORD_CONTEXT (context)) \
+ || (((constraint) & NEXT_NOTWORD_CONSTRAINT) && IS_WORD_CONTEXT (context)) \
+ || (((constraint) & NEXT_NEWLINE_CONSTRAINT) && !IS_NEWLINE_CONTEXT
(context)) \
+ || (((constraint) & NEXT_ENDBUF_CONSTRAINT) && !IS_ENDBUF_CONTEXT (context)))
+
+struct re_dfastate_t
+{
+ re_hashval_t hash;
+ re_node_set nodes;
+ re_node_set non_eps_nodes;
+ re_node_set inveclosure;
+ re_node_set *entrance_nodes;
+ struct re_dfastate_t **trtable, **word_trtable;
+ unsigned int context : 4;
+ unsigned int halt : 1;
+ /* If this state can accept "multi byte".
+ Note that we refer to multibyte characters, and multi character
+ collating elements as "multi byte". */
+ unsigned int accept_mb : 1;
+ /* If this state has backreference node(s). */
+ unsigned int has_backref : 1;
+ unsigned int has_constraint : 1;
+};
+typedef struct re_dfastate_t re_dfastate_t;
+
+struct re_state_table_entry
+{
+ Idx num;
+ Idx alloc;
+ re_dfastate_t **array;
+};
+
+/* Array type used in re_sub_match_last_t and re_sub_match_top_t. */
+
+typedef struct
+{
+ Idx next_idx;
+ Idx alloc;
+ re_dfastate_t **array;
+} state_array_t;
+
+/* Store information about the node NODE whose type is OP_CLOSE_SUBEXP. */
+
+typedef struct
+{
+ Idx node;
+ Idx str_idx; /* The position NODE match at. */
+ state_array_t path;
+} re_sub_match_last_t;
+
+/* Store information about the node NODE whose type is OP_OPEN_SUBEXP.
+ And information about the node, whose type is OP_CLOSE_SUBEXP,
+ corresponding to NODE is stored in LASTS. */
+
+typedef struct
+{
+ Idx str_idx;
+ Idx node;
+ state_array_t *path;
+ Idx alasts; /* Allocation size of LASTS. */
+ Idx nlasts; /* The number of LASTS. */
+ re_sub_match_last_t **lasts;
+} re_sub_match_top_t;
+
+struct re_backref_cache_entry
+{
+ Idx node;
+ Idx str_idx;
+ Idx subexp_from;
+ Idx subexp_to;
+ char more;
+ char unused;
+ unsigned short int eps_reachable_subexps_map;
+};
+
+typedef struct
+{
+ /* The string object corresponding to the input string. */
+ re_string_t input;
+#if defined _LIBC || (defined __STDC_VERSION__ && __STDC_VERSION__ >= 199901L)
+ const re_dfa_t *const dfa;
+#else
+ const re_dfa_t *dfa;
+#endif
+ /* EFLAGS of the argument of regexec. */
+ int eflags;
+ /* Where the matching ends. */
+ Idx match_last;
+ Idx last_node;
+ /* The state log used by the matcher. */
+ re_dfastate_t **state_log;
+ Idx state_log_top;
+ /* Back reference cache. */
+ Idx nbkref_ents;
+ Idx abkref_ents;
+ struct re_backref_cache_entry *bkref_ents;
+ int max_mb_elem_len;
+ Idx nsub_tops;
+ Idx asub_tops;
+ re_sub_match_top_t **sub_tops;
+} re_match_context_t;
+
+typedef struct
+{
+ re_dfastate_t **sifted_states;
+ re_dfastate_t **limited_states;
+ Idx last_node;
+ Idx last_str_idx;
+ re_node_set limits;
+} re_sift_context_t;
+
+struct re_fail_stack_ent_t
+{
+ Idx idx;
+ Idx node;
+ regmatch_t *regs;
+ re_node_set eps_via_nodes;
+};
+
+struct re_fail_stack_t
+{
+ Idx num;
+ Idx alloc;
+ struct re_fail_stack_ent_t *stack;
+};
+
+struct re_dfa_t
+{
+ re_token_t *nodes;
+ size_t nodes_alloc;
+ size_t nodes_len;
+ Idx *nexts;
+ Idx *org_indices;
+ re_node_set *edests;
+ re_node_set *eclosures;
+ re_node_set *inveclosures;
+ struct re_state_table_entry *state_table;
+ re_dfastate_t *init_state;
+ re_dfastate_t *init_state_word;
+ re_dfastate_t *init_state_nl;
+ re_dfastate_t *init_state_begbuf;
+ bin_tree_t *str_tree;
+ bin_tree_storage_t *str_tree_storage;
+ re_bitset_ptr_t sb_char;
+ int str_tree_storage_idx;
+
+ /* number of subexpressions 're_nsub' is in regex_t. */
+ re_hashval_t state_hash_mask;
+ Idx init_node;
+ Idx nbackref; /* The number of backreference in this dfa. */
+
+ /* Bitmap expressing which backreference is used. */
+ bitset_word_t used_bkref_map;
+ bitset_word_t completed_bkref_map;
+
+ unsigned int has_plural_match : 1;
+ /* If this dfa has "multibyte node", which is a backreference or
+ a node which can accept multibyte character or multi character
+ collating element. */
+ unsigned int has_mb_node : 1;
+ unsigned int is_utf8 : 1;
+ unsigned int map_notascii : 1;
+ unsigned int word_ops_used : 1;
+ int mb_cur_max;
+ bitset_t word_char;
+ reg_syntax_t syntax;
+ Idx *subexp_map;
+#ifdef DEBUG
+ char* re_str;
+#endif
+ lock_define (lock)
+};
+
+#define re_node_set_init_empty(set) memset (set, '\0', sizeof (re_node_set))
+#define re_node_set_remove(set,id) \
+ (re_node_set_remove_at (set, re_node_set_contains (set, id) - 1))
+#define re_node_set_empty(p) ((p)->nelem = 0)
+#define re_node_set_free(set) re_free ((set)->elems)
+
+
+typedef enum
+{
+ SB_CHAR,
+ MB_CHAR,
+ EQUIV_CLASS,
+ COLL_SYM,
+ CHAR_CLASS
+} bracket_elem_type;
+
+typedef struct
+{
+ bracket_elem_type type;
+ union
+ {
+ unsigned char ch;
+ unsigned char *name;
+ wchar_t wch;
+ } opr;
+} bracket_elem_t;
+
+
+/* Functions for bitset_t operation. */
+
+static inline void
+bitset_set (bitset_t set, Idx i)
+{
+ set[i / BITSET_WORD_BITS] |= (bitset_word_t) 1 << i % BITSET_WORD_BITS;
+}
+
+static inline void
+bitset_clear (bitset_t set, Idx i)
+{
+ set[i / BITSET_WORD_BITS] &= ~ ((bitset_word_t) 1 << i % BITSET_WORD_BITS);
+}
+
+static inline bool
+bitset_contain (const bitset_t set, Idx i)
+{
+ return (set[i / BITSET_WORD_BITS] >> i % BITSET_WORD_BITS) & 1;
+}
+
+static inline void
+bitset_empty (bitset_t set)
+{
+ memset (set, '\0', sizeof (bitset_t));
+}
+
+static inline void
+bitset_set_all (bitset_t set)
+{
+ memset (set, -1, sizeof (bitset_word_t) * (SBC_MAX / BITSET_WORD_BITS));
+ if (SBC_MAX % BITSET_WORD_BITS != 0)
+ set[BITSET_WORDS - 1] =
+ ((bitset_word_t) 1 << SBC_MAX % BITSET_WORD_BITS) - 1;
+}
+
+static inline void
+bitset_copy (bitset_t dest, const bitset_t src)
+{
+ memcpy (dest, src, sizeof (bitset_t));
+}
+
+static inline void
+bitset_not (bitset_t set)
+{
+ int bitset_i;
+ for (bitset_i = 0; bitset_i < SBC_MAX / BITSET_WORD_BITS; ++bitset_i)
+ set[bitset_i] = ~set[bitset_i];
+ if (SBC_MAX % BITSET_WORD_BITS != 0)
+ set[BITSET_WORDS - 1] =
+ ((((bitset_word_t) 1 << SBC_MAX % BITSET_WORD_BITS) - 1)
+ & ~set[BITSET_WORDS - 1]);
+}
+
+static inline void
+bitset_merge (bitset_t dest, const bitset_t src)
+{
+ int bitset_i;
+ for (bitset_i = 0; bitset_i < BITSET_WORDS; ++bitset_i)
+ dest[bitset_i] |= src[bitset_i];
+}
+
+static inline void
+bitset_mask (bitset_t dest, const bitset_t src)
+{
+ int bitset_i;
+ for (bitset_i = 0; bitset_i < BITSET_WORDS; ++bitset_i)
+ dest[bitset_i] &= src[bitset_i];
+}
+
+#ifdef RE_ENABLE_I18N
+/* Functions for re_string. */
+static int
+__attribute__ ((pure, unused))
+re_string_char_size_at (const re_string_t *pstr, Idx idx)
+{
+ int byte_idx;
+ if (pstr->mb_cur_max == 1)
+ return 1;
+ for (byte_idx = 1; idx + byte_idx < pstr->valid_len; ++byte_idx)
+ if (pstr->wcs[idx + byte_idx] != WEOF)
+ break;
+ return byte_idx;
+}
+
+static wint_t
+__attribute__ ((pure, unused))
+re_string_wchar_at (const re_string_t *pstr, Idx idx)
+{
+ if (pstr->mb_cur_max == 1)
+ return (wint_t) pstr->mbs[idx];
+ return (wint_t) pstr->wcs[idx];
+}
+
+# ifdef _LIBC
+# include <locale/weight.h>
+# endif
+
+static int
+__attribute__ ((pure, unused))
+re_string_elem_size_at (const re_string_t *pstr, Idx idx)
+{
+# ifdef _LIBC
+ const unsigned char *p, *extra;
+ const int32_t *table, *indirect;
+ uint_fast32_t nrules = _NL_CURRENT_WORD (LC_COLLATE, _NL_COLLATE_NRULES);
+
+ if (nrules != 0)
+ {
+ table = (const int32_t *) _NL_CURRENT (LC_COLLATE, _NL_COLLATE_TABLEMB);
+ extra = (const unsigned char *)
+ _NL_CURRENT (LC_COLLATE, _NL_COLLATE_EXTRAMB);
+ indirect = (const int32_t *) _NL_CURRENT (LC_COLLATE,
+ _NL_COLLATE_INDIRECTMB);
+ p = pstr->mbs + idx;
+ findidx (table, indirect, extra, &p, pstr->len - idx);
+ return p - pstr->mbs - idx;
+ }
+ else
+# endif /* _LIBC */
+ return 1;
+}
+#endif /* RE_ENABLE_I18N */
+
+#ifndef __GNUC_PREREQ
+# if defined __GNUC__ && defined __GNUC_MINOR__
+# define __GNUC_PREREQ(maj, min) \
+ ((__GNUC__ << 16) + __GNUC_MINOR__ >= ((maj) << 16) + (min))
+# else
+# define __GNUC_PREREQ(maj, min) 0
+# endif
+#endif
+
+#if __GNUC_PREREQ (3,4)
+# undef __attribute_warn_unused_result__
+# define __attribute_warn_unused_result__ \
+ __attribute__ ((__warn_unused_result__))
+#else
+# define __attribute_warn_unused_result__ /* empty */
+#endif
+
+#ifndef FALLTHROUGH
+# if __GNUC__ < 7
+# define FALLTHROUGH ((void) 0)
+# else
+# define FALLTHROUGH __attribute__ ((__fallthrough__))
+# endif
+#endif
+
+#endif /* _REGEX_INTERNAL_H */
diff --git a/lib/regexec.c b/lib/regexec.c
new file mode 100644
index 0000000..6591311
--- /dev/null
+++ b/lib/regexec.c
@@ -0,0 +1,4324 @@
+/* Extended regular expression matching and search library.
+ Copyright (C) 2002-2018 Free Software Foundation, Inc.
+ This file is part of the GNU C Library.
+ Contributed by Isamu Hasegawa <address@hidden>.
+
+ The GNU C Library 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 3 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
+ General Public License for more details.
+
+ You should have received a copy of the GNU General Public
+ License along with the GNU C Library; if not, see
+ <https://www.gnu.org/licenses/>. */
+
+static reg_errcode_t match_ctx_init (re_match_context_t *cache, int eflags,
+ Idx n);
+static void match_ctx_clean (re_match_context_t *mctx);
+static void match_ctx_free (re_match_context_t *cache);
+static reg_errcode_t match_ctx_add_entry (re_match_context_t *cache, Idx node,
+ Idx str_idx, Idx from, Idx to);
+static Idx search_cur_bkref_entry (const re_match_context_t *mctx, Idx
str_idx);
+static reg_errcode_t match_ctx_add_subtop (re_match_context_t *mctx, Idx node,
+ Idx str_idx);
+static re_sub_match_last_t * match_ctx_add_sublast (re_sub_match_top_t *subtop,
+ Idx node, Idx str_idx);
+static void sift_ctx_init (re_sift_context_t *sctx, re_dfastate_t **sifted_sts,
+ re_dfastate_t **limited_sts, Idx last_node,
+ Idx last_str_idx);
+static reg_errcode_t re_search_internal (const regex_t *preg,
+ const char *string, Idx length,
+ Idx start, Idx last_start, Idx stop,
+ size_t nmatch, regmatch_t pmatch[],
+ int eflags);
+static regoff_t re_search_2_stub (struct re_pattern_buffer *bufp,
+ const char *string1, Idx length1,
+ const char *string2, Idx length2,
+ Idx start, regoff_t range,
+ struct re_registers *regs,
+ Idx stop, bool ret_len);
+static regoff_t re_search_stub (struct re_pattern_buffer *bufp,
+ const char *string, Idx length, Idx start,
+ regoff_t range, Idx stop,
+ struct re_registers *regs,
+ bool ret_len);
+static unsigned re_copy_regs (struct re_registers *regs, regmatch_t *pmatch,
+ Idx nregs, int regs_allocated);
+static reg_errcode_t prune_impossible_nodes (re_match_context_t *mctx);
+static Idx check_matching (re_match_context_t *mctx, bool fl_longest_match,
+ Idx *p_match_first);
+static Idx check_halt_state_context (const re_match_context_t *mctx,
+ const re_dfastate_t *state, Idx idx);
+static void update_regs (const re_dfa_t *dfa, regmatch_t *pmatch,
+ regmatch_t *prev_idx_match, Idx cur_node,
+ Idx cur_idx, Idx nmatch);
+static reg_errcode_t push_fail_stack (struct re_fail_stack_t *fs,
+ Idx str_idx, Idx dest_node, Idx nregs,
+ regmatch_t *regs,
+ re_node_set *eps_via_nodes);
+static reg_errcode_t set_regs (const regex_t *preg,
+ const re_match_context_t *mctx,
+ size_t nmatch, regmatch_t *pmatch,
+ bool fl_backtrack);
+static reg_errcode_t free_fail_stack_return (struct re_fail_stack_t *fs);
+
+#ifdef RE_ENABLE_I18N
+static int sift_states_iter_mb (const re_match_context_t *mctx,
+ re_sift_context_t *sctx,
+ Idx node_idx, Idx str_idx, Idx max_str_idx);
+#endif /* RE_ENABLE_I18N */
+static reg_errcode_t sift_states_backward (const re_match_context_t *mctx,
+ re_sift_context_t *sctx);
+static reg_errcode_t build_sifted_states (const re_match_context_t *mctx,
+ re_sift_context_t *sctx, Idx str_idx,
+ re_node_set *cur_dest);
+static reg_errcode_t update_cur_sifted_state (const re_match_context_t *mctx,
+ re_sift_context_t *sctx,
+ Idx str_idx,
+ re_node_set *dest_nodes);
+static reg_errcode_t add_epsilon_src_nodes (const re_dfa_t *dfa,
+ re_node_set *dest_nodes,
+ const re_node_set *candidates);
+static bool check_dst_limits (const re_match_context_t *mctx,
+ const re_node_set *limits,
+ Idx dst_node, Idx dst_idx, Idx src_node,
+ Idx src_idx);
+static int check_dst_limits_calc_pos_1 (const re_match_context_t *mctx,
+ int boundaries, Idx subexp_idx,
+ Idx from_node, Idx bkref_idx);
+static int check_dst_limits_calc_pos (const re_match_context_t *mctx,
+ Idx limit, Idx subexp_idx,
+ Idx node, Idx str_idx,
+ Idx bkref_idx);
+static reg_errcode_t check_subexp_limits (const re_dfa_t *dfa,
+ re_node_set *dest_nodes,
+ const re_node_set *candidates,
+ re_node_set *limits,
+ struct re_backref_cache_entry
*bkref_ents,
+ Idx str_idx);
+static reg_errcode_t sift_states_bkref (const re_match_context_t *mctx,
+ re_sift_context_t *sctx,
+ Idx str_idx, const re_node_set
*candidates);
+static reg_errcode_t merge_state_array (const re_dfa_t *dfa,
+ re_dfastate_t **dst,
+ re_dfastate_t **src, Idx num);
+static re_dfastate_t *find_recover_state (reg_errcode_t *err,
+ re_match_context_t *mctx);
+static re_dfastate_t *transit_state (reg_errcode_t *err,
+ re_match_context_t *mctx,
+ re_dfastate_t *state);
+static re_dfastate_t *merge_state_with_log (reg_errcode_t *err,
+ re_match_context_t *mctx,
+ re_dfastate_t *next_state);
+static reg_errcode_t check_subexp_matching_top (re_match_context_t *mctx,
+ re_node_set *cur_nodes,
+ Idx str_idx);
+#if 0
+static re_dfastate_t *transit_state_sb (reg_errcode_t *err,
+ re_match_context_t *mctx,
+ re_dfastate_t *pstate);
+#endif
+#ifdef RE_ENABLE_I18N
+static reg_errcode_t transit_state_mb (re_match_context_t *mctx,
+ re_dfastate_t *pstate);
+#endif /* RE_ENABLE_I18N */
+static reg_errcode_t transit_state_bkref (re_match_context_t *mctx,
+ const re_node_set *nodes);
+static reg_errcode_t get_subexp (re_match_context_t *mctx,
+ Idx bkref_node, Idx bkref_str_idx);
+static reg_errcode_t get_subexp_sub (re_match_context_t *mctx,
+ const re_sub_match_top_t *sub_top,
+ re_sub_match_last_t *sub_last,
+ Idx bkref_node, Idx bkref_str);
+static Idx find_subexp_node (const re_dfa_t *dfa, const re_node_set *nodes,
+ Idx subexp_idx, int type);
+static reg_errcode_t check_arrival (re_match_context_t *mctx,
+ state_array_t *path, Idx top_node,
+ Idx top_str, Idx last_node, Idx last_str,
+ int type);
+static reg_errcode_t check_arrival_add_next_nodes (re_match_context_t *mctx,
+ Idx str_idx,
+ re_node_set *cur_nodes,
+ re_node_set *next_nodes);
+static reg_errcode_t check_arrival_expand_ecl (const re_dfa_t *dfa,
+ re_node_set *cur_nodes,
+ Idx ex_subexp, int type);
+static reg_errcode_t check_arrival_expand_ecl_sub (const re_dfa_t *dfa,
+ re_node_set *dst_nodes,
+ Idx target, Idx ex_subexp,
+ int type);
+static reg_errcode_t expand_bkref_cache (re_match_context_t *mctx,
+ re_node_set *cur_nodes, Idx cur_str,
+ Idx subexp_num, int type);
+static bool build_trtable (const re_dfa_t *dfa, re_dfastate_t *state);
+#ifdef RE_ENABLE_I18N
+static int check_node_accept_bytes (const re_dfa_t *dfa, Idx node_idx,
+ const re_string_t *input, Idx idx);
+# ifdef _LIBC
+static unsigned int find_collation_sequence_value (const unsigned char *mbs,
+ size_t name_len);
+# endif /* _LIBC */
+#endif /* RE_ENABLE_I18N */
+static Idx group_nodes_into_DFAstates (const re_dfa_t *dfa,
+ const re_dfastate_t *state,
+ re_node_set *states_node,
+ bitset_t *states_ch);
+static bool check_node_accept (const re_match_context_t *mctx,
+ const re_token_t *node, Idx idx);
+static reg_errcode_t extend_buffers (re_match_context_t *mctx, int min_len);
+
+/* Entry point for POSIX code. */
+
+/* regexec searches for a given pattern, specified by PREG, in the
+ string STRING.
+
+ If NMATCH is zero or REG_NOSUB was set in the cflags argument to
+ 'regcomp', we ignore PMATCH. Otherwise, we assume PMATCH has at
+ least NMATCH elements, and we set them to the offsets of the
+ corresponding matched substrings.
+
+ EFLAGS specifies "execution flags" which affect matching: if
+ REG_NOTBOL is set, then ^ does not match at the beginning of the
+ string; if REG_NOTEOL is set, then $ does not match at the end.
+
+ We return 0 if we find a match and REG_NOMATCH if not. */
+
+int
+regexec (const regex_t *_Restrict_ preg, const char *_Restrict_ string,
+ size_t nmatch, regmatch_t pmatch[], int eflags)
+{
+ reg_errcode_t err;
+ Idx start, length;
+ re_dfa_t *dfa = preg->buffer;
+
+ if (eflags & ~(REG_NOTBOL | REG_NOTEOL | REG_STARTEND))
+ return REG_BADPAT;
+
+ if (eflags & REG_STARTEND)
+ {
+ start = pmatch[0].rm_so;
+ length = pmatch[0].rm_eo;
+ }
+ else
+ {
+ start = 0;
+ length = strlen (string);
+ }
+
+ lock_lock (dfa->lock);
+ if (preg->no_sub)
+ err = re_search_internal (preg, string, length, start, length,
+ length, 0, NULL, eflags);
+ else
+ err = re_search_internal (preg, string, length, start, length,
+ length, nmatch, pmatch, eflags);
+ lock_unlock (dfa->lock);
+ return err != REG_NOERROR;
+}
+
+#ifdef _LIBC
+libc_hidden_def (__regexec)
+
+# include <shlib-compat.h>
+versioned_symbol (libc, __regexec, regexec, GLIBC_2_3_4);
+
+# if SHLIB_COMPAT (libc, GLIBC_2_0, GLIBC_2_3_4)
+__typeof__ (__regexec) __compat_regexec;
+
+int
+attribute_compat_text_section
+__compat_regexec (const regex_t *_Restrict_ preg,
+ const char *_Restrict_ string, size_t nmatch,
+ regmatch_t pmatch[], int eflags)
+{
+ return regexec (preg, string, nmatch, pmatch,
+ eflags & (REG_NOTBOL | REG_NOTEOL));
+}
+compat_symbol (libc, __compat_regexec, regexec, GLIBC_2_0);
+# endif
+#endif
+
+/* Entry points for GNU code. */
+
+/* re_match, re_search, re_match_2, re_search_2
+
+ The former two functions operate on STRING with length LENGTH,
+ while the later two operate on concatenation of STRING1 and STRING2
+ with lengths LENGTH1 and LENGTH2, respectively.
+
+ re_match() matches the compiled pattern in BUFP against the string,
+ starting at index START.
+
+ re_search() first tries matching at index START, then it tries to match
+ starting from index START + 1, and so on. The last start position tried
+ is START + RANGE. (Thus RANGE = 0 forces re_search to operate the same
+ way as re_match().)
+
+ The parameter STOP of re_{match,search}_2 specifies that no match exceeding
+ the first STOP characters of the concatenation of the strings should be
+ concerned.
+
+ If REGS is not NULL, and BUFP->no_sub is not set, the offsets of the match
+ and all groups is stored in REGS. (For the "_2" variants, the offsets are
+ computed relative to the concatenation, not relative to the individual
+ strings.)
+
+ On success, re_match* functions return the length of the match, re_search*
+ return the position of the start of the match. Return value -1 means no
+ match was found and -2 indicates an internal error. */
+
+regoff_t
+re_match (struct re_pattern_buffer *bufp, const char *string, Idx length,
+ Idx start, struct re_registers *regs)
+{
+ return re_search_stub (bufp, string, length, start, 0, length, regs, true);
+}
+#ifdef _LIBC
+weak_alias (__re_match, re_match)
+#endif
+
+regoff_t
+re_search (struct re_pattern_buffer *bufp, const char *string, Idx length,
+ Idx start, regoff_t range, struct re_registers *regs)
+{
+ return re_search_stub (bufp, string, length, start, range, length, regs,
+ false);
+}
+#ifdef _LIBC
+weak_alias (__re_search, re_search)
+#endif
+
+regoff_t
+re_match_2 (struct re_pattern_buffer *bufp, const char *string1, Idx length1,
+ const char *string2, Idx length2, Idx start,
+ struct re_registers *regs, Idx stop)
+{
+ return re_search_2_stub (bufp, string1, length1, string2, length2,
+ start, 0, regs, stop, true);
+}
+#ifdef _LIBC
+weak_alias (__re_match_2, re_match_2)
+#endif
+
+regoff_t
+re_search_2 (struct re_pattern_buffer *bufp, const char *string1, Idx length1,
+ const char *string2, Idx length2, Idx start, regoff_t range,
+ struct re_registers *regs, Idx stop)
+{
+ return re_search_2_stub (bufp, string1, length1, string2, length2,
+ start, range, regs, stop, false);
+}
+#ifdef _LIBC
+weak_alias (__re_search_2, re_search_2)
+#endif
+
+static regoff_t
+re_search_2_stub (struct re_pattern_buffer *bufp, const char *string1,
+ Idx length1, const char *string2, Idx length2, Idx start,
+ regoff_t range, struct re_registers *regs,
+ Idx stop, bool ret_len)
+{
+ const char *str;
+ regoff_t rval;
+ Idx len;
+ char *s = NULL;
+
+ if (BE ((length1 < 0 || length2 < 0 || stop < 0
+ || INT_ADD_WRAPV (length1, length2, &len)),
+ 0))
+ return -2;
+
+ /* Concatenate the strings. */
+ if (length2 > 0)
+ if (length1 > 0)
+ {
+ s = re_malloc (char, len);
+
+ if (BE (s == NULL, 0))
+ return -2;
+#ifdef _LIBC
+ memcpy (__mempcpy (s, string1, length1), string2, length2);
+#else
+ memcpy (s, string1, length1);
+ memcpy (s + length1, string2, length2);
+#endif
+ str = s;
+ }
+ else
+ str = string2;
+ else
+ str = string1;
+
+ rval = re_search_stub (bufp, str, len, start, range, stop, regs,
+ ret_len);
+ re_free (s);
+ return rval;
+}
+
+/* The parameters have the same meaning as those of re_search.
+ Additional parameters:
+ If RET_LEN is true the length of the match is returned (re_match style);
+ otherwise the position of the match is returned. */
+
+static regoff_t
+re_search_stub (struct re_pattern_buffer *bufp, const char *string, Idx length,
+ Idx start, regoff_t range, Idx stop, struct re_registers *regs,
+ bool ret_len)
+{
+ reg_errcode_t result;
+ regmatch_t *pmatch;
+ Idx nregs;
+ regoff_t rval;
+ int eflags = 0;
+ re_dfa_t *dfa = bufp->buffer;
+ Idx last_start = start + range;
+
+ /* Check for out-of-range. */
+ if (BE (start < 0 || start > length, 0))
+ return -1;
+ if (BE (length < last_start || (0 <= range && last_start < start), 0))
+ last_start = length;
+ else if (BE (last_start < 0 || (range < 0 && start <= last_start), 0))
+ last_start = 0;
+
+ lock_lock (dfa->lock);
+
+ eflags |= (bufp->not_bol) ? REG_NOTBOL : 0;
+ eflags |= (bufp->not_eol) ? REG_NOTEOL : 0;
+
+ /* Compile fastmap if we haven't yet. */
+ if (start < last_start && bufp->fastmap != NULL && !bufp->fastmap_accurate)
+ re_compile_fastmap (bufp);
+
+ if (BE (bufp->no_sub, 0))
+ regs = NULL;
+
+ /* We need at least 1 register. */
+ if (regs == NULL)
+ nregs = 1;
+ else if (BE (bufp->regs_allocated == REGS_FIXED
+ && regs->num_regs <= bufp->re_nsub, 0))
+ {
+ nregs = regs->num_regs;
+ if (BE (nregs < 1, 0))
+ {
+ /* Nothing can be copied to regs. */
+ regs = NULL;
+ nregs = 1;
+ }
+ }
+ else
+ nregs = bufp->re_nsub + 1;
+ pmatch = re_malloc (regmatch_t, nregs);
+ if (BE (pmatch == NULL, 0))
+ {
+ rval = -2;
+ goto out;
+ }
+
+ result = re_search_internal (bufp, string, length, start, last_start, stop,
+ nregs, pmatch, eflags);
+
+ rval = 0;
+
+ /* I hope we needn't fill their regs with -1's when no match was found. */
+ if (result != REG_NOERROR)
+ rval = result == REG_NOMATCH ? -1 : -2;
+ else if (regs != NULL)
+ {
+ /* If caller wants register contents data back, copy them. */
+ bufp->regs_allocated = re_copy_regs (regs, pmatch, nregs,
+ bufp->regs_allocated);
+ if (BE (bufp->regs_allocated == REGS_UNALLOCATED, 0))
+ rval = -2;
+ }
+
+ if (BE (rval == 0, 1))
+ {
+ if (ret_len)
+ {
+ assert (pmatch[0].rm_so == start);
+ rval = pmatch[0].rm_eo - start;
+ }
+ else
+ rval = pmatch[0].rm_so;
+ }
+ re_free (pmatch);
+ out:
+ lock_unlock (dfa->lock);
+ return rval;
+}
+
+static unsigned
+re_copy_regs (struct re_registers *regs, regmatch_t *pmatch, Idx nregs,
+ int regs_allocated)
+{
+ int rval = REGS_REALLOCATE;
+ Idx i;
+ Idx need_regs = nregs + 1;
+ /* We need one extra element beyond 'num_regs' for the '-1' marker GNU code
+ uses. */
+
+ /* Have the register data arrays been allocated? */
+ if (regs_allocated == REGS_UNALLOCATED)
+ { /* No. So allocate them with malloc. */
+ regs->start = re_malloc (regoff_t, need_regs);
+ if (BE (regs->start == NULL, 0))
+ return REGS_UNALLOCATED;
+ regs->end = re_malloc (regoff_t, need_regs);
+ if (BE (regs->end == NULL, 0))
+ {
+ re_free (regs->start);
+ return REGS_UNALLOCATED;
+ }
+ regs->num_regs = need_regs;
+ }
+ else if (regs_allocated == REGS_REALLOCATE)
+ { /* Yes. If we need more elements than were already
+ allocated, reallocate them. If we need fewer, just
+ leave it alone. */
+ if (BE (need_regs > regs->num_regs, 0))
+ {
+ regoff_t *new_start = re_realloc (regs->start, regoff_t, need_regs);
+ regoff_t *new_end;
+ if (BE (new_start == NULL, 0))
+ return REGS_UNALLOCATED;
+ new_end = re_realloc (regs->end, regoff_t, need_regs);
+ if (BE (new_end == NULL, 0))
+ {
+ re_free (new_start);
+ return REGS_UNALLOCATED;
+ }
+ regs->start = new_start;
+ regs->end = new_end;
+ regs->num_regs = need_regs;
+ }
+ }
+ else
+ {
+ assert (regs_allocated == REGS_FIXED);
+ /* This function may not be called with REGS_FIXED and nregs too big. */
+ assert (regs->num_regs >= nregs);
+ rval = REGS_FIXED;
+ }
+
+ /* Copy the regs. */
+ for (i = 0; i < nregs; ++i)
+ {
+ regs->start[i] = pmatch[i].rm_so;
+ regs->end[i] = pmatch[i].rm_eo;
+ }
+ for ( ; i < regs->num_regs; ++i)
+ regs->start[i] = regs->end[i] = -1;
+
+ return rval;
+}
+
+/* Set REGS to hold NUM_REGS registers, storing them in STARTS and
+ ENDS. Subsequent matches using PATTERN_BUFFER and REGS will use
+ this memory for recording register information. STARTS and ENDS
+ must be allocated using the malloc library routine, and must each
+ be at least NUM_REGS * sizeof (regoff_t) bytes long.
+
+ If NUM_REGS == 0, then subsequent matches should allocate their own
+ register data.
+
+ Unless this function is called, the first search or match using
+ PATTERN_BUFFER will allocate its own register data, without
+ freeing the old data. */
+
+void
+re_set_registers (struct re_pattern_buffer *bufp, struct re_registers *regs,
+ __re_size_t num_regs, regoff_t *starts, regoff_t *ends)
+{
+ if (num_regs)
+ {
+ bufp->regs_allocated = REGS_REALLOCATE;
+ regs->num_regs = num_regs;
+ regs->start = starts;
+ regs->end = ends;
+ }
+ else
+ {
+ bufp->regs_allocated = REGS_UNALLOCATED;
+ regs->num_regs = 0;
+ regs->start = regs->end = NULL;
+ }
+}
+#ifdef _LIBC
+weak_alias (__re_set_registers, re_set_registers)
+#endif
+
+/* Entry points compatible with 4.2 BSD regex library. We don't define
+ them unless specifically requested. */
+
+#if defined _REGEX_RE_COMP || defined _LIBC
+int
+# ifdef _LIBC
+weak_function
+# endif
+re_exec (const char *s)
+{
+ return 0 == regexec (&re_comp_buf, s, 0, NULL, 0);
+}
+#endif /* _REGEX_RE_COMP */
+
+/* Internal entry point. */
+
+/* Searches for a compiled pattern PREG in the string STRING, whose
+ length is LENGTH. NMATCH, PMATCH, and EFLAGS have the same
+ meaning as with regexec. LAST_START is START + RANGE, where
+ START and RANGE have the same meaning as with re_search.
+ Return REG_NOERROR if we find a match, and REG_NOMATCH if not,
+ otherwise return the error code.
+ Note: We assume front end functions already check ranges.
+ (0 <= LAST_START && LAST_START <= LENGTH) */
+
+static reg_errcode_t
+__attribute_warn_unused_result__
+re_search_internal (const regex_t *preg, const char *string, Idx length,
+ Idx start, Idx last_start, Idx stop, size_t nmatch,
+ regmatch_t pmatch[], int eflags)
+{
+ reg_errcode_t err;
+ const re_dfa_t *dfa = preg->buffer;
+ Idx left_lim, right_lim;
+ int incr;
+ bool fl_longest_match;
+ int match_kind;
+ Idx match_first;
+ Idx match_last = -1;
+ Idx extra_nmatch;
+ bool sb;
+ int ch;
+#if defined _LIBC || (defined __STDC_VERSION__ && __STDC_VERSION__ >= 199901L)
+ re_match_context_t mctx = { .dfa = dfa };
+#else
+ re_match_context_t mctx;
+#endif
+ char *fastmap = ((preg->fastmap != NULL && preg->fastmap_accurate
+ && start != last_start && !preg->can_be_null)
+ ? preg->fastmap : NULL);
+ RE_TRANSLATE_TYPE t = preg->translate;
+
+#if !(defined _LIBC || (defined __STDC_VERSION__ && __STDC_VERSION__ >=
199901L))
+ memset (&mctx, '\0', sizeof (re_match_context_t));
+ mctx.dfa = dfa;
+#endif
+
+ extra_nmatch = (nmatch > preg->re_nsub) ? nmatch - (preg->re_nsub + 1) : 0;
+ nmatch -= extra_nmatch;
+
+ /* Check if the DFA haven't been compiled. */
+ if (BE (preg->used == 0 || dfa->init_state == NULL
+ || dfa->init_state_word == NULL || dfa->init_state_nl == NULL
+ || dfa->init_state_begbuf == NULL, 0))
+ return REG_NOMATCH;
+
+#ifdef DEBUG
+ /* We assume front-end functions already check them. */
+ assert (0 <= last_start && last_start <= length);
+#endif
+
+ /* If initial states with non-begbuf contexts have no elements,
+ the regex must be anchored. If preg->newline_anchor is set,
+ we'll never use init_state_nl, so do not check it. */
+ if (dfa->init_state->nodes.nelem == 0
+ && dfa->init_state_word->nodes.nelem == 0
+ && (dfa->init_state_nl->nodes.nelem == 0
+ || !preg->newline_anchor))
+ {
+ if (start != 0 && last_start != 0)
+ return REG_NOMATCH;
+ start = last_start = 0;
+ }
+
+ /* We must check the longest matching, if nmatch > 0. */
+ fl_longest_match = (nmatch != 0 || dfa->nbackref);
+
+ err = re_string_allocate (&mctx.input, string, length, dfa->nodes_len + 1,
+ preg->translate, (preg->syntax & RE_ICASE) != 0,
+ dfa);
+ if (BE (err != REG_NOERROR, 0))
+ goto free_return;
+ mctx.input.stop = stop;
+ mctx.input.raw_stop = stop;
+ mctx.input.newline_anchor = preg->newline_anchor;
+
+ err = match_ctx_init (&mctx, eflags, dfa->nbackref * 2);
+ if (BE (err != REG_NOERROR, 0))
+ goto free_return;
+
+ /* We will log all the DFA states through which the dfa pass,
+ if nmatch > 1, or this dfa has "multibyte node", which is a
+ back-reference or a node which can accept multibyte character or
+ multi character collating element. */
+ if (nmatch > 1 || dfa->has_mb_node)
+ {
+ /* Avoid overflow. */
+ if (BE ((MIN (IDX_MAX, SIZE_MAX / sizeof (re_dfastate_t *))
+ <= mctx.input.bufs_len), 0))
+ {
+ err = REG_ESPACE;
+ goto free_return;
+ }
+
+ mctx.state_log = re_malloc (re_dfastate_t *, mctx.input.bufs_len + 1);
+ if (BE (mctx.state_log == NULL, 0))
+ {
+ err = REG_ESPACE;
+ goto free_return;
+ }
+ }
+ else
+ mctx.state_log = NULL;
+
+ match_first = start;
+ mctx.input.tip_context = (eflags & REG_NOTBOL) ? CONTEXT_BEGBUF
+ : CONTEXT_NEWLINE | CONTEXT_BEGBUF;
+
+ /* Check incrementally whether the input string matches. */
+ incr = (last_start < start) ? -1 : 1;
+ left_lim = (last_start < start) ? last_start : start;
+ right_lim = (last_start < start) ? start : last_start;
+ sb = dfa->mb_cur_max == 1;
+ match_kind =
+ (fastmap
+ ? ((sb || !(preg->syntax & RE_ICASE || t) ? 4 : 0)
+ | (start <= last_start ? 2 : 0)
+ | (t != NULL ? 1 : 0))
+ : 8);
+
+ for (;; match_first += incr)
+ {
+ err = REG_NOMATCH;
+ if (match_first < left_lim || right_lim < match_first)
+ goto free_return;
+
+ /* Advance as rapidly as possible through the string, until we
+ find a plausible place to start matching. This may be done
+ with varying efficiency, so there are various possibilities:
+ only the most common of them are specialized, in order to
+ save on code size. We use a switch statement for speed. */
+ switch (match_kind)
+ {
+ case 8:
+ /* No fastmap. */
+ break;
+
+ case 7:
+ /* Fastmap with single-byte translation, match forward. */
+ while (BE (match_first < right_lim, 1)
+ && !fastmap[t[(unsigned char) string[match_first]]])
+ ++match_first;
+ goto forward_match_found_start_or_reached_end;
+
+ case 6:
+ /* Fastmap without translation, match forward. */
+ while (BE (match_first < right_lim, 1)
+ && !fastmap[(unsigned char) string[match_first]])
+ ++match_first;
+
+ forward_match_found_start_or_reached_end:
+ if (BE (match_first == right_lim, 0))
+ {
+ ch = match_first >= length
+ ? 0 : (unsigned char) string[match_first];
+ if (!fastmap[t ? t[ch] : ch])
+ goto free_return;
+ }
+ break;
+
+ case 4:
+ case 5:
+ /* Fastmap without multi-byte translation, match backwards. */
+ while (match_first >= left_lim)
+ {
+ ch = match_first >= length
+ ? 0 : (unsigned char) string[match_first];
+ if (fastmap[t ? t[ch] : ch])
+ break;
+ --match_first;
+ }
+ if (match_first < left_lim)
+ goto free_return;
+ break;
+
+ default:
+ /* In this case, we can't determine easily the current byte,
+ since it might be a component byte of a multibyte
+ character. Then we use the constructed buffer instead. */
+ for (;;)
+ {
+ /* If MATCH_FIRST is out of the valid range, reconstruct the
+ buffers. */
+ __re_size_t offset = match_first - mctx.input.raw_mbs_idx;
+ if (BE (offset >= (__re_size_t) mctx.input.valid_raw_len, 0))
+ {
+ err = re_string_reconstruct (&mctx.input, match_first,
+ eflags);
+ if (BE (err != REG_NOERROR, 0))
+ goto free_return;
+
+ offset = match_first - mctx.input.raw_mbs_idx;
+ }
+ /* If MATCH_FIRST is out of the buffer, leave it as '\0'.
+ Note that MATCH_FIRST must not be smaller than 0. */
+ ch = (match_first >= length
+ ? 0 : re_string_byte_at (&mctx.input, offset));
+ if (fastmap[ch])
+ break;
+ match_first += incr;
+ if (match_first < left_lim || match_first > right_lim)
+ {
+ err = REG_NOMATCH;
+ goto free_return;
+ }
+ }
+ break;
+ }
+
+ /* Reconstruct the buffers so that the matcher can assume that
+ the matching starts from the beginning of the buffer. */
+ err = re_string_reconstruct (&mctx.input, match_first, eflags);
+ if (BE (err != REG_NOERROR, 0))
+ goto free_return;
+
+#ifdef RE_ENABLE_I18N
+ /* Don't consider this char as a possible match start if it part,
+ yet isn't the head, of a multibyte character. */
+ if (!sb && !re_string_first_byte (&mctx.input, 0))
+ continue;
+#endif
+
+ /* It seems to be appropriate one, then use the matcher. */
+ /* We assume that the matching starts from 0. */
+ mctx.state_log_top = mctx.nbkref_ents = mctx.max_mb_elem_len = 0;
+ match_last = check_matching (&mctx, fl_longest_match,
+ start <= last_start ? &match_first : NULL);
+ if (match_last != -1)
+ {
+ if (BE (match_last == -2, 0))
+ {
+ err = REG_ESPACE;
+ goto free_return;
+ }
+ else
+ {
+ mctx.match_last = match_last;
+ if ((!preg->no_sub && nmatch > 1) || dfa->nbackref)
+ {
+ re_dfastate_t *pstate = mctx.state_log[match_last];
+ mctx.last_node = check_halt_state_context (&mctx, pstate,
+ match_last);
+ }
+ if ((!preg->no_sub && nmatch > 1 && dfa->has_plural_match)
+ || dfa->nbackref)
+ {
+ err = prune_impossible_nodes (&mctx);
+ if (err == REG_NOERROR)
+ break;
+ if (BE (err != REG_NOMATCH, 0))
+ goto free_return;
+ match_last = -1;
+ }
+ else
+ break; /* We found a match. */
+ }
+ }
+
+ match_ctx_clean (&mctx);
+ }
+
+#ifdef DEBUG
+ assert (match_last != -1);
+ assert (err == REG_NOERROR);
+#endif
+
+ /* Set pmatch[] if we need. */
+ if (nmatch > 0)
+ {
+ Idx reg_idx;
+
+ /* Initialize registers. */
+ for (reg_idx = 1; reg_idx < nmatch; ++reg_idx)
+ pmatch[reg_idx].rm_so = pmatch[reg_idx].rm_eo = -1;
+
+ /* Set the points where matching start/end. */
+ pmatch[0].rm_so = 0;
+ pmatch[0].rm_eo = mctx.match_last;
+ /* FIXME: This function should fail if mctx.match_last exceeds
+ the maximum possible regoff_t value. We need a new error
+ code REG_OVERFLOW. */
+
+ if (!preg->no_sub && nmatch > 1)
+ {
+ err = set_regs (preg, &mctx, nmatch, pmatch,
+ dfa->has_plural_match && dfa->nbackref > 0);
+ if (BE (err != REG_NOERROR, 0))
+ goto free_return;
+ }
+
+ /* At last, add the offset to each register, since we slid
+ the buffers so that we could assume that the matching starts
+ from 0. */
+ for (reg_idx = 0; reg_idx < nmatch; ++reg_idx)
+ if (pmatch[reg_idx].rm_so != -1)
+ {
+#ifdef RE_ENABLE_I18N
+ if (BE (mctx.input.offsets_needed != 0, 0))
+ {
+ pmatch[reg_idx].rm_so =
+ (pmatch[reg_idx].rm_so == mctx.input.valid_len
+ ? mctx.input.valid_raw_len
+ : mctx.input.offsets[pmatch[reg_idx].rm_so]);
+ pmatch[reg_idx].rm_eo =
+ (pmatch[reg_idx].rm_eo == mctx.input.valid_len
+ ? mctx.input.valid_raw_len
+ : mctx.input.offsets[pmatch[reg_idx].rm_eo]);
+ }
+#else
+ assert (mctx.input.offsets_needed == 0);
+#endif
+ pmatch[reg_idx].rm_so += match_first;
+ pmatch[reg_idx].rm_eo += match_first;
+ }
+ for (reg_idx = 0; reg_idx < extra_nmatch; ++reg_idx)
+ {
+ pmatch[nmatch + reg_idx].rm_so = -1;
+ pmatch[nmatch + reg_idx].rm_eo = -1;
+ }
+
+ if (dfa->subexp_map)
+ for (reg_idx = 0; reg_idx + 1 < nmatch; reg_idx++)
+ if (dfa->subexp_map[reg_idx] != reg_idx)
+ {
+ pmatch[reg_idx + 1].rm_so
+ = pmatch[dfa->subexp_map[reg_idx] + 1].rm_so;
+ pmatch[reg_idx + 1].rm_eo
+ = pmatch[dfa->subexp_map[reg_idx] + 1].rm_eo;
+ }
+ }
+
+ free_return:
+ re_free (mctx.state_log);
+ if (dfa->nbackref)
+ match_ctx_free (&mctx);
+ re_string_destruct (&mctx.input);
+ return err;
+}
+
+static reg_errcode_t
+__attribute_warn_unused_result__
+prune_impossible_nodes (re_match_context_t *mctx)
+{
+ const re_dfa_t *const dfa = mctx->dfa;
+ Idx halt_node, match_last;
+ reg_errcode_t ret;
+ re_dfastate_t **sifted_states;
+ re_dfastate_t **lim_states = NULL;
+ re_sift_context_t sctx;
+#ifdef DEBUG
+ assert (mctx->state_log != NULL);
+#endif
+ match_last = mctx->match_last;
+ halt_node = mctx->last_node;
+
+ /* Avoid overflow. */
+ if (BE (MIN (IDX_MAX, SIZE_MAX / sizeof (re_dfastate_t *)) <= match_last, 0))
+ return REG_ESPACE;
+
+ sifted_states = re_malloc (re_dfastate_t *, match_last + 1);
+ if (BE (sifted_states == NULL, 0))
+ {
+ ret = REG_ESPACE;
+ goto free_return;
+ }
+ if (dfa->nbackref)
+ {
+ lim_states = re_malloc (re_dfastate_t *, match_last + 1);
+ if (BE (lim_states == NULL, 0))
+ {
+ ret = REG_ESPACE;
+ goto free_return;
+ }
+ while (1)
+ {
+ memset (lim_states, '\0',
+ sizeof (re_dfastate_t *) * (match_last + 1));
+ sift_ctx_init (&sctx, sifted_states, lim_states, halt_node,
+ match_last);
+ ret = sift_states_backward (mctx, &sctx);
+ re_node_set_free (&sctx.limits);
+ if (BE (ret != REG_NOERROR, 0))
+ goto free_return;
+ if (sifted_states[0] != NULL || lim_states[0] != NULL)
+ break;
+ do
+ {
+ --match_last;
+ if (match_last < 0)
+ {
+ ret = REG_NOMATCH;
+ goto free_return;
+ }
+ } while (mctx->state_log[match_last] == NULL
+ || !mctx->state_log[match_last]->halt);
+ halt_node = check_halt_state_context (mctx,
+ mctx->state_log[match_last],
+ match_last);
+ }
+ ret = merge_state_array (dfa, sifted_states, lim_states,
+ match_last + 1);
+ re_free (lim_states);
+ lim_states = NULL;
+ if (BE (ret != REG_NOERROR, 0))
+ goto free_return;
+ }
+ else
+ {
+ sift_ctx_init (&sctx, sifted_states, lim_states, halt_node, match_last);
+ ret = sift_states_backward (mctx, &sctx);
+ re_node_set_free (&sctx.limits);
+ if (BE (ret != REG_NOERROR, 0))
+ goto free_return;
+ if (sifted_states[0] == NULL)
+ {
+ ret = REG_NOMATCH;
+ goto free_return;
+ }
+ }
+ re_free (mctx->state_log);
+ mctx->state_log = sifted_states;
+ sifted_states = NULL;
+ mctx->last_node = halt_node;
+ mctx->match_last = match_last;
+ ret = REG_NOERROR;
+ free_return:
+ re_free (sifted_states);
+ re_free (lim_states);
+ return ret;
+}
+
+/* Acquire an initial state and return it.
+ We must select appropriate initial state depending on the context,
+ since initial states may have constraints like "\<", "^", etc.. */
+
+static inline re_dfastate_t *
+__attribute__ ((always_inline))
+acquire_init_state_context (reg_errcode_t *err, const re_match_context_t *mctx,
+ Idx idx)
+{
+ const re_dfa_t *const dfa = mctx->dfa;
+ if (dfa->init_state->has_constraint)
+ {
+ unsigned int context;
+ context = re_string_context_at (&mctx->input, idx - 1, mctx->eflags);
+ if (IS_WORD_CONTEXT (context))
+ return dfa->init_state_word;
+ else if (IS_ORDINARY_CONTEXT (context))
+ return dfa->init_state;
+ else if (IS_BEGBUF_CONTEXT (context) && IS_NEWLINE_CONTEXT (context))
+ return dfa->init_state_begbuf;
+ else if (IS_NEWLINE_CONTEXT (context))
+ return dfa->init_state_nl;
+ else if (IS_BEGBUF_CONTEXT (context))
+ {
+ /* It is relatively rare case, then calculate on demand. */
+ return re_acquire_state_context (err, dfa,
+ dfa->init_state->entrance_nodes,
+ context);
+ }
+ else
+ /* Must not happen? */
+ return dfa->init_state;
+ }
+ else
+ return dfa->init_state;
+}
+
+/* Check whether the regular expression match input string INPUT or not,
+ and return the index where the matching end. Return -1 if
+ there is no match, and return -2 in case of an error.
+ FL_LONGEST_MATCH means we want the POSIX longest matching.
+ If P_MATCH_FIRST is not NULL, and the match fails, it is set to the
+ next place where we may want to try matching.
+ Note that the matcher assumes that the matching starts from the current
+ index of the buffer. */
+
+static Idx
+__attribute_warn_unused_result__
+check_matching (re_match_context_t *mctx, bool fl_longest_match,
+ Idx *p_match_first)
+{
+ const re_dfa_t *const dfa = mctx->dfa;
+ reg_errcode_t err;
+ Idx match = 0;
+ Idx match_last = -1;
+ Idx cur_str_idx = re_string_cur_idx (&mctx->input);
+ re_dfastate_t *cur_state;
+ bool at_init_state = p_match_first != NULL;
+ Idx next_start_idx = cur_str_idx;
+
+ err = REG_NOERROR;
+ cur_state = acquire_init_state_context (&err, mctx, cur_str_idx);
+ /* An initial state must not be NULL (invalid). */
+ if (BE (cur_state == NULL, 0))
+ {
+ assert (err == REG_ESPACE);
+ return -2;
+ }
+
+ if (mctx->state_log != NULL)
+ {
+ mctx->state_log[cur_str_idx] = cur_state;
+
+ /* Check OP_OPEN_SUBEXP in the initial state in case that we use them
+ later. E.g. Processing back references. */
+ if (BE (dfa->nbackref, 0))
+ {
+ at_init_state = false;
+ err = check_subexp_matching_top (mctx, &cur_state->nodes, 0);
+ if (BE (err != REG_NOERROR, 0))
+ return err;
+
+ if (cur_state->has_backref)
+ {
+ err = transit_state_bkref (mctx, &cur_state->nodes);
+ if (BE (err != REG_NOERROR, 0))
+ return err;
+ }
+ }
+ }
+
+ /* If the RE accepts NULL string. */
+ if (BE (cur_state->halt, 0))
+ {
+ if (!cur_state->has_constraint
+ || check_halt_state_context (mctx, cur_state, cur_str_idx))
+ {
+ if (!fl_longest_match)
+ return cur_str_idx;
+ else
+ {
+ match_last = cur_str_idx;
+ match = 1;
+ }
+ }
+ }
+
+ while (!re_string_eoi (&mctx->input))
+ {
+ re_dfastate_t *old_state = cur_state;
+ Idx next_char_idx = re_string_cur_idx (&mctx->input) + 1;
+
+ if ((BE (next_char_idx >= mctx->input.bufs_len, 0)
+ && mctx->input.bufs_len < mctx->input.len)
+ || (BE (next_char_idx >= mctx->input.valid_len, 0)
+ && mctx->input.valid_len < mctx->input.len))
+ {
+ err = extend_buffers (mctx, next_char_idx + 1);
+ if (BE (err != REG_NOERROR, 0))
+ {
+ assert (err == REG_ESPACE);
+ return -2;
+ }
+ }
+
+ cur_state = transit_state (&err, mctx, cur_state);
+ if (mctx->state_log != NULL)
+ cur_state = merge_state_with_log (&err, mctx, cur_state);
+
+ if (cur_state == NULL)
+ {
+ /* Reached the invalid state or an error. Try to recover a valid
+ state using the state log, if available and if we have not
+ already found a valid (even if not the longest) match. */
+ if (BE (err != REG_NOERROR, 0))
+ return -2;
+
+ if (mctx->state_log == NULL
+ || (match && !fl_longest_match)
+ || (cur_state = find_recover_state (&err, mctx)) == NULL)
+ break;
+ }
+
+ if (BE (at_init_state, 0))
+ {
+ if (old_state == cur_state)
+ next_start_idx = next_char_idx;
+ else
+ at_init_state = false;
+ }
+
+ if (cur_state->halt)
+ {
+ /* Reached a halt state.
+ Check the halt state can satisfy the current context. */
+ if (!cur_state->has_constraint
+ || check_halt_state_context (mctx, cur_state,
+ re_string_cur_idx (&mctx->input)))
+ {
+ /* We found an appropriate halt state. */
+ match_last = re_string_cur_idx (&mctx->input);
+ match = 1;
+
+ /* We found a match, do not modify match_first below. */
+ p_match_first = NULL;
+ if (!fl_longest_match)
+ break;
+ }
+ }
+ }
+
+ if (p_match_first)
+ *p_match_first += next_start_idx;
+
+ return match_last;
+}
+
+/* Check NODE match the current context. */
+
+static bool
+check_halt_node_context (const re_dfa_t *dfa, Idx node, unsigned int context)
+{
+ re_token_type_t type = dfa->nodes[node].type;
+ unsigned int constraint = dfa->nodes[node].constraint;
+ if (type != END_OF_RE)
+ return false;
+ if (!constraint)
+ return true;
+ if (NOT_SATISFY_NEXT_CONSTRAINT (constraint, context))
+ return false;
+ return true;
+}
+
+/* Check the halt state STATE match the current context.
+ Return 0 if not match, if the node, STATE has, is a halt node and
+ match the context, return the node. */
+
+static Idx
+check_halt_state_context (const re_match_context_t *mctx,
+ const re_dfastate_t *state, Idx idx)
+{
+ Idx i;
+ unsigned int context;
+#ifdef DEBUG
+ assert (state->halt);
+#endif
+ context = re_string_context_at (&mctx->input, idx, mctx->eflags);
+ for (i = 0; i < state->nodes.nelem; ++i)
+ if (check_halt_node_context (mctx->dfa, state->nodes.elems[i], context))
+ return state->nodes.elems[i];
+ return 0;
+}
+
+/* Compute the next node to which "NFA" transit from NODE("NFA" is a NFA
+ corresponding to the DFA).
+ Return the destination node, and update EPS_VIA_NODES;
+ return -1 in case of errors. */
+
+static Idx
+proceed_next_node (const re_match_context_t *mctx, Idx nregs, regmatch_t *regs,
+ Idx *pidx, Idx node, re_node_set *eps_via_nodes,
+ struct re_fail_stack_t *fs)
+{
+ const re_dfa_t *const dfa = mctx->dfa;
+ Idx i;
+ bool ok;
+ if (IS_EPSILON_NODE (dfa->nodes[node].type))
+ {
+ re_node_set *cur_nodes = &mctx->state_log[*pidx]->nodes;
+ re_node_set *edests = &dfa->edests[node];
+ Idx dest_node;
+ ok = re_node_set_insert (eps_via_nodes, node);
+ if (BE (! ok, 0))
+ return -2;
+ /* Pick up a valid destination, or return -1 if none
+ is found. */
+ for (dest_node = -1, i = 0; i < edests->nelem; ++i)
+ {
+ Idx candidate = edests->elems[i];
+ if (!re_node_set_contains (cur_nodes, candidate))
+ continue;
+ if (dest_node == -1)
+ dest_node = candidate;
+
+ else
+ {
+ /* In order to avoid infinite loop like "(a*)*", return the second
+ epsilon-transition if the first was already considered. */
+ if (re_node_set_contains (eps_via_nodes, dest_node))
+ return candidate;
+
+ /* Otherwise, push the second epsilon-transition on the fail
stack. */
+ else if (fs != NULL
+ && push_fail_stack (fs, *pidx, candidate, nregs, regs,
+ eps_via_nodes))
+ return -2;
+
+ /* We know we are going to exit. */
+ break;
+ }
+ }
+ return dest_node;
+ }
+ else
+ {
+ Idx naccepted = 0;
+ re_token_type_t type = dfa->nodes[node].type;
+
+#ifdef RE_ENABLE_I18N
+ if (dfa->nodes[node].accept_mb)
+ naccepted = check_node_accept_bytes (dfa, node, &mctx->input, *pidx);
+ else
+#endif /* RE_ENABLE_I18N */
+ if (type == OP_BACK_REF)
+ {
+ Idx subexp_idx = dfa->nodes[node].opr.idx + 1;
+ naccepted = regs[subexp_idx].rm_eo - regs[subexp_idx].rm_so;
+ if (fs != NULL)
+ {
+ if (regs[subexp_idx].rm_so == -1 || regs[subexp_idx].rm_eo == -1)
+ return -1;
+ else if (naccepted)
+ {
+ char *buf = (char *) re_string_get_buffer (&mctx->input);
+ if (memcmp (buf + regs[subexp_idx].rm_so, buf + *pidx,
+ naccepted) != 0)
+ return -1;
+ }
+ }
+
+ if (naccepted == 0)
+ {
+ Idx dest_node;
+ ok = re_node_set_insert (eps_via_nodes, node);
+ if (BE (! ok, 0))
+ return -2;
+ dest_node = dfa->edests[node].elems[0];
+ if (re_node_set_contains (&mctx->state_log[*pidx]->nodes,
+ dest_node))
+ return dest_node;
+ }
+ }
+
+ if (naccepted != 0
+ || check_node_accept (mctx, dfa->nodes + node, *pidx))
+ {
+ Idx dest_node = dfa->nexts[node];
+ *pidx = (naccepted == 0) ? *pidx + 1 : *pidx + naccepted;
+ if (fs && (*pidx > mctx->match_last || mctx->state_log[*pidx] == NULL
+ || !re_node_set_contains (&mctx->state_log[*pidx]->nodes,
+ dest_node)))
+ return -1;
+ re_node_set_empty (eps_via_nodes);
+ return dest_node;
+ }
+ }
+ return -1;
+}
+
+static reg_errcode_t
+__attribute_warn_unused_result__
+push_fail_stack (struct re_fail_stack_t *fs, Idx str_idx, Idx dest_node,
+ Idx nregs, regmatch_t *regs, re_node_set *eps_via_nodes)
+{
+ reg_errcode_t err;
+ Idx num = fs->num++;
+ if (fs->num == fs->alloc)
+ {
+ struct re_fail_stack_ent_t *new_array;
+ new_array = re_realloc (fs->stack, struct re_fail_stack_ent_t,
+ fs->alloc * 2);
+ if (new_array == NULL)
+ return REG_ESPACE;
+ fs->alloc *= 2;
+ fs->stack = new_array;
+ }
+ fs->stack[num].idx = str_idx;
+ fs->stack[num].node = dest_node;
+ fs->stack[num].regs = re_malloc (regmatch_t, nregs);
+ if (fs->stack[num].regs == NULL)
+ return REG_ESPACE;
+ memcpy (fs->stack[num].regs, regs, sizeof (regmatch_t) * nregs);
+ err = re_node_set_init_copy (&fs->stack[num].eps_via_nodes, eps_via_nodes);
+ return err;
+}
+
+static Idx
+pop_fail_stack (struct re_fail_stack_t *fs, Idx *pidx, Idx nregs,
+ regmatch_t *regs, re_node_set *eps_via_nodes)
+{
+ Idx num = --fs->num;
+ assert (num >= 0);
+ *pidx = fs->stack[num].idx;
+ memcpy (regs, fs->stack[num].regs, sizeof (regmatch_t) * nregs);
+ re_node_set_free (eps_via_nodes);
+ re_free (fs->stack[num].regs);
+ *eps_via_nodes = fs->stack[num].eps_via_nodes;
+ return fs->stack[num].node;
+}
+
+/* Set the positions where the subexpressions are starts/ends to registers
+ PMATCH.
+ Note: We assume that pmatch[0] is already set, and
+ pmatch[i].rm_so == pmatch[i].rm_eo == -1 for 0 < i < nmatch. */
+
+static reg_errcode_t
+__attribute_warn_unused_result__
+set_regs (const regex_t *preg, const re_match_context_t *mctx, size_t nmatch,
+ regmatch_t *pmatch, bool fl_backtrack)
+{
+ const re_dfa_t *dfa = preg->buffer;
+ Idx idx, cur_node;
+ re_node_set eps_via_nodes;
+ struct re_fail_stack_t *fs;
+ struct re_fail_stack_t fs_body = { 0, 2, NULL };
+ regmatch_t *prev_idx_match;
+ bool prev_idx_match_malloced = false;
+
+#ifdef DEBUG
+ assert (nmatch > 1);
+ assert (mctx->state_log != NULL);
+#endif
+ if (fl_backtrack)
+ {
+ fs = &fs_body;
+ fs->stack = re_malloc (struct re_fail_stack_ent_t, fs->alloc);
+ if (fs->stack == NULL)
+ return REG_ESPACE;
+ }
+ else
+ fs = NULL;
+
+ cur_node = dfa->init_node;
+ re_node_set_init_empty (&eps_via_nodes);
+
+ if (__libc_use_alloca (nmatch * sizeof (regmatch_t)))
+ prev_idx_match = (regmatch_t *) alloca (nmatch * sizeof (regmatch_t));
+ else
+ {
+ prev_idx_match = re_malloc (regmatch_t, nmatch);
+ if (prev_idx_match == NULL)
+ {
+ free_fail_stack_return (fs);
+ return REG_ESPACE;
+ }
+ prev_idx_match_malloced = true;
+ }
+ memcpy (prev_idx_match, pmatch, sizeof (regmatch_t) * nmatch);
+
+ for (idx = pmatch[0].rm_so; idx <= pmatch[0].rm_eo ;)
+ {
+ update_regs (dfa, pmatch, prev_idx_match, cur_node, idx, nmatch);
+
+ if (idx == pmatch[0].rm_eo && cur_node == mctx->last_node)
+ {
+ Idx reg_idx;
+ if (fs)
+ {
+ for (reg_idx = 0; reg_idx < nmatch; ++reg_idx)
+ if (pmatch[reg_idx].rm_so > -1 && pmatch[reg_idx].rm_eo == -1)
+ break;
+ if (reg_idx == nmatch)
+ {
+ re_node_set_free (&eps_via_nodes);
+ if (prev_idx_match_malloced)
+ re_free (prev_idx_match);
+ return free_fail_stack_return (fs);
+ }
+ cur_node = pop_fail_stack (fs, &idx, nmatch, pmatch,
+ &eps_via_nodes);
+ }
+ else
+ {
+ re_node_set_free (&eps_via_nodes);
+ if (prev_idx_match_malloced)
+ re_free (prev_idx_match);
+ return REG_NOERROR;
+ }
+ }
+
+ /* Proceed to next node. */
+ cur_node = proceed_next_node (mctx, nmatch, pmatch, &idx, cur_node,
+ &eps_via_nodes, fs);
+
+ if (BE (cur_node < 0, 0))
+ {
+ if (BE (cur_node == -2, 0))
+ {
+ re_node_set_free (&eps_via_nodes);
+ if (prev_idx_match_malloced)
+ re_free (prev_idx_match);
+ free_fail_stack_return (fs);
+ return REG_ESPACE;
+ }
+ if (fs)
+ cur_node = pop_fail_stack (fs, &idx, nmatch, pmatch,
+ &eps_via_nodes);
+ else
+ {
+ re_node_set_free (&eps_via_nodes);
+ if (prev_idx_match_malloced)
+ re_free (prev_idx_match);
+ return REG_NOMATCH;
+ }
+ }
+ }
+ re_node_set_free (&eps_via_nodes);
+ if (prev_idx_match_malloced)
+ re_free (prev_idx_match);
+ return free_fail_stack_return (fs);
+}
+
+static reg_errcode_t
+free_fail_stack_return (struct re_fail_stack_t *fs)
+{
+ if (fs)
+ {
+ Idx fs_idx;
+ for (fs_idx = 0; fs_idx < fs->num; ++fs_idx)
+ {
+ re_node_set_free (&fs->stack[fs_idx].eps_via_nodes);
+ re_free (fs->stack[fs_idx].regs);
+ }
+ re_free (fs->stack);
+ }
+ return REG_NOERROR;
+}
+
+static void
+update_regs (const re_dfa_t *dfa, regmatch_t *pmatch,
+ regmatch_t *prev_idx_match, Idx cur_node, Idx cur_idx, Idx nmatch)
+{
+ int type = dfa->nodes[cur_node].type;
+ if (type == OP_OPEN_SUBEXP)
+ {
+ Idx reg_num = dfa->nodes[cur_node].opr.idx + 1;
+
+ /* We are at the first node of this sub expression. */
+ if (reg_num < nmatch)
+ {
+ pmatch[reg_num].rm_so = cur_idx;
+ pmatch[reg_num].rm_eo = -1;
+ }
+ }
+ else if (type == OP_CLOSE_SUBEXP)
+ {
+ Idx reg_num = dfa->nodes[cur_node].opr.idx + 1;
+ if (reg_num < nmatch)
+ {
+ /* We are at the last node of this sub expression. */
+ if (pmatch[reg_num].rm_so < cur_idx)
+ {
+ pmatch[reg_num].rm_eo = cur_idx;
+ /* This is a non-empty match or we are not inside an optional
+ subexpression. Accept this right away. */
+ memcpy (prev_idx_match, pmatch, sizeof (regmatch_t) * nmatch);
+ }
+ else
+ {
+ if (dfa->nodes[cur_node].opt_subexp
+ && prev_idx_match[reg_num].rm_so != -1)
+ /* We transited through an empty match for an optional
+ subexpression, like (a?)*, and this is not the subexp's
+ first match. Copy back the old content of the registers
+ so that matches of an inner subexpression are undone as
+ well, like in ((a?))*. */
+ memcpy (pmatch, prev_idx_match, sizeof (regmatch_t) * nmatch);
+ else
+ /* We completed a subexpression, but it may be part of
+ an optional one, so do not update PREV_IDX_MATCH. */
+ pmatch[reg_num].rm_eo = cur_idx;
+ }
+ }
+ }
+}
+
+/* This function checks the STATE_LOG from the SCTX->last_str_idx to 0
+ and sift the nodes in each states according to the following rules.
+ Updated state_log will be wrote to STATE_LOG.
+
+ Rules: We throw away the Node 'a' in the STATE_LOG[STR_IDX] if...
+ 1. When STR_IDX == MATCH_LAST(the last index in the state_log):
+ If 'a' isn't the LAST_NODE and 'a' can't epsilon transit to
+ the LAST_NODE, we throw away the node 'a'.
+ 2. When 0 <= STR_IDX < MATCH_LAST and 'a' accepts
+ string 's' and transit to 'b':
+ i. If 'b' isn't in the STATE_LOG[STR_IDX+strlen('s')], we throw
+ away the node 'a'.
+ ii. If 'b' is in the STATE_LOG[STR_IDX+strlen('s')] but 'b' is
+ thrown away, we throw away the node 'a'.
+ 3. When 0 <= STR_IDX < MATCH_LAST and 'a' epsilon transit to 'b':
+ i. If 'b' isn't in the STATE_LOG[STR_IDX], we throw away the
+ node 'a'.
+ ii. If 'b' is in the STATE_LOG[STR_IDX] but 'b' is thrown away,
+ we throw away the node 'a'. */
+
+#define STATE_NODE_CONTAINS(state,node) \
+ ((state) != NULL && re_node_set_contains (&(state)->nodes, node))
+
+static reg_errcode_t
+sift_states_backward (const re_match_context_t *mctx, re_sift_context_t *sctx)
+{
+ reg_errcode_t err;
+ int null_cnt = 0;
+ Idx str_idx = sctx->last_str_idx;
+ re_node_set cur_dest;
+
+#ifdef DEBUG
+ assert (mctx->state_log != NULL && mctx->state_log[str_idx] != NULL);
+#endif
+
+ /* Build sifted state_log[str_idx]. It has the nodes which can epsilon
+ transit to the last_node and the last_node itself. */
+ err = re_node_set_init_1 (&cur_dest, sctx->last_node);
+ if (BE (err != REG_NOERROR, 0))
+ return err;
+ err = update_cur_sifted_state (mctx, sctx, str_idx, &cur_dest);
+ if (BE (err != REG_NOERROR, 0))
+ goto free_return;
+
+ /* Then check each states in the state_log. */
+ while (str_idx > 0)
+ {
+ /* Update counters. */
+ null_cnt = (sctx->sifted_states[str_idx] == NULL) ? null_cnt + 1 : 0;
+ if (null_cnt > mctx->max_mb_elem_len)
+ {
+ memset (sctx->sifted_states, '\0',
+ sizeof (re_dfastate_t *) * str_idx);
+ re_node_set_free (&cur_dest);
+ return REG_NOERROR;
+ }
+ re_node_set_empty (&cur_dest);
+ --str_idx;
+
+ if (mctx->state_log[str_idx])
+ {
+ err = build_sifted_states (mctx, sctx, str_idx, &cur_dest);
+ if (BE (err != REG_NOERROR, 0))
+ goto free_return;
+ }
+
+ /* Add all the nodes which satisfy the following conditions:
+ - It can epsilon transit to a node in CUR_DEST.
+ - It is in CUR_SRC.
+ And update state_log. */
+ err = update_cur_sifted_state (mctx, sctx, str_idx, &cur_dest);
+ if (BE (err != REG_NOERROR, 0))
+ goto free_return;
+ }
+ err = REG_NOERROR;
+ free_return:
+ re_node_set_free (&cur_dest);
+ return err;
+}
+
+static reg_errcode_t
+__attribute_warn_unused_result__
+build_sifted_states (const re_match_context_t *mctx, re_sift_context_t *sctx,
+ Idx str_idx, re_node_set *cur_dest)
+{
+ const re_dfa_t *const dfa = mctx->dfa;
+ const re_node_set *cur_src = &mctx->state_log[str_idx]->non_eps_nodes;
+ Idx i;
+
+ /* Then build the next sifted state.
+ We build the next sifted state on 'cur_dest', and update
+ 'sifted_states[str_idx]' with 'cur_dest'.
+ Note:
+ 'cur_dest' is the sifted state from 'state_log[str_idx + 1]'.
+ 'cur_src' points the node_set of the old 'state_log[str_idx]'
+ (with the epsilon nodes pre-filtered out). */
+ for (i = 0; i < cur_src->nelem; i++)
+ {
+ Idx prev_node = cur_src->elems[i];
+ int naccepted = 0;
+ bool ok;
+
+#ifdef DEBUG
+ re_token_type_t type = dfa->nodes[prev_node].type;
+ assert (!IS_EPSILON_NODE (type));
+#endif
+#ifdef RE_ENABLE_I18N
+ /* If the node may accept "multi byte". */
+ if (dfa->nodes[prev_node].accept_mb)
+ naccepted = sift_states_iter_mb (mctx, sctx, prev_node,
+ str_idx, sctx->last_str_idx);
+#endif /* RE_ENABLE_I18N */
+
+ /* We don't check backreferences here.
+ See update_cur_sifted_state(). */
+ if (!naccepted
+ && check_node_accept (mctx, dfa->nodes + prev_node, str_idx)
+ && STATE_NODE_CONTAINS (sctx->sifted_states[str_idx + 1],
+ dfa->nexts[prev_node]))
+ naccepted = 1;
+
+ if (naccepted == 0)
+ continue;
+
+ if (sctx->limits.nelem)
+ {
+ Idx to_idx = str_idx + naccepted;
+ if (check_dst_limits (mctx, &sctx->limits,
+ dfa->nexts[prev_node], to_idx,
+ prev_node, str_idx))
+ continue;
+ }
+ ok = re_node_set_insert (cur_dest, prev_node);
+ if (BE (! ok, 0))
+ return REG_ESPACE;
+ }
+
+ return REG_NOERROR;
+}
+
+/* Helper functions. */
+
+static reg_errcode_t
+clean_state_log_if_needed (re_match_context_t *mctx, Idx next_state_log_idx)
+{
+ Idx top = mctx->state_log_top;
+
+ if ((next_state_log_idx >= mctx->input.bufs_len
+ && mctx->input.bufs_len < mctx->input.len)
+ || (next_state_log_idx >= mctx->input.valid_len
+ && mctx->input.valid_len < mctx->input.len))
+ {
+ reg_errcode_t err;
+ err = extend_buffers (mctx, next_state_log_idx + 1);
+ if (BE (err != REG_NOERROR, 0))
+ return err;
+ }
+
+ if (top < next_state_log_idx)
+ {
+ memset (mctx->state_log + top + 1, '\0',
+ sizeof (re_dfastate_t *) * (next_state_log_idx - top));
+ mctx->state_log_top = next_state_log_idx;
+ }
+ return REG_NOERROR;
+}
+
+static reg_errcode_t
+merge_state_array (const re_dfa_t *dfa, re_dfastate_t **dst,
+ re_dfastate_t **src, Idx num)
+{
+ Idx st_idx;
+ reg_errcode_t err;
+ for (st_idx = 0; st_idx < num; ++st_idx)
+ {
+ if (dst[st_idx] == NULL)
+ dst[st_idx] = src[st_idx];
+ else if (src[st_idx] != NULL)
+ {
+ re_node_set merged_set;
+ err = re_node_set_init_union (&merged_set, &dst[st_idx]->nodes,
+ &src[st_idx]->nodes);
+ if (BE (err != REG_NOERROR, 0))
+ return err;
+ dst[st_idx] = re_acquire_state (&err, dfa, &merged_set);
+ re_node_set_free (&merged_set);
+ if (BE (err != REG_NOERROR, 0))
+ return err;
+ }
+ }
+ return REG_NOERROR;
+}
+
+static reg_errcode_t
+update_cur_sifted_state (const re_match_context_t *mctx,
+ re_sift_context_t *sctx, Idx str_idx,
+ re_node_set *dest_nodes)
+{
+ const re_dfa_t *const dfa = mctx->dfa;
+ reg_errcode_t err = REG_NOERROR;
+ const re_node_set *candidates;
+ candidates = ((mctx->state_log[str_idx] == NULL) ? NULL
+ : &mctx->state_log[str_idx]->nodes);
+
+ if (dest_nodes->nelem == 0)
+ sctx->sifted_states[str_idx] = NULL;
+ else
+ {
+ if (candidates)
+ {
+ /* At first, add the nodes which can epsilon transit to a node in
+ DEST_NODE. */
+ err = add_epsilon_src_nodes (dfa, dest_nodes, candidates);
+ if (BE (err != REG_NOERROR, 0))
+ return err;
+
+ /* Then, check the limitations in the current sift_context. */
+ if (sctx->limits.nelem)
+ {
+ err = check_subexp_limits (dfa, dest_nodes, candidates,
&sctx->limits,
+ mctx->bkref_ents, str_idx);
+ if (BE (err != REG_NOERROR, 0))
+ return err;
+ }
+ }
+
+ sctx->sifted_states[str_idx] = re_acquire_state (&err, dfa, dest_nodes);
+ if (BE (err != REG_NOERROR, 0))
+ return err;
+ }
+
+ if (candidates && mctx->state_log[str_idx]->has_backref)
+ {
+ err = sift_states_bkref (mctx, sctx, str_idx, candidates);
+ if (BE (err != REG_NOERROR, 0))
+ return err;
+ }
+ return REG_NOERROR;
+}
+
+static reg_errcode_t
+__attribute_warn_unused_result__
+add_epsilon_src_nodes (const re_dfa_t *dfa, re_node_set *dest_nodes,
+ const re_node_set *candidates)
+{
+ reg_errcode_t err = REG_NOERROR;
+ Idx i;
+
+ re_dfastate_t *state = re_acquire_state (&err, dfa, dest_nodes);
+ if (BE (err != REG_NOERROR, 0))
+ return err;
+
+ if (!state->inveclosure.alloc)
+ {
+ err = re_node_set_alloc (&state->inveclosure, dest_nodes->nelem);
+ if (BE (err != REG_NOERROR, 0))
+ return REG_ESPACE;
+ for (i = 0; i < dest_nodes->nelem; i++)
+ {
+ err = re_node_set_merge (&state->inveclosure,
+ dfa->inveclosures + dest_nodes->elems[i]);
+ if (BE (err != REG_NOERROR, 0))
+ return REG_ESPACE;
+ }
+ }
+ return re_node_set_add_intersect (dest_nodes, candidates,
+ &state->inveclosure);
+}
+
+static reg_errcode_t
+sub_epsilon_src_nodes (const re_dfa_t *dfa, Idx node, re_node_set *dest_nodes,
+ const re_node_set *candidates)
+{
+ Idx ecl_idx;
+ reg_errcode_t err;
+ re_node_set *inv_eclosure = dfa->inveclosures + node;
+ re_node_set except_nodes;
+ re_node_set_init_empty (&except_nodes);
+ for (ecl_idx = 0; ecl_idx < inv_eclosure->nelem; ++ecl_idx)
+ {
+ Idx cur_node = inv_eclosure->elems[ecl_idx];
+ if (cur_node == node)
+ continue;
+ if (IS_EPSILON_NODE (dfa->nodes[cur_node].type))
+ {
+ Idx edst1 = dfa->edests[cur_node].elems[0];
+ Idx edst2 = ((dfa->edests[cur_node].nelem > 1)
+ ? dfa->edests[cur_node].elems[1] : -1);
+ if ((!re_node_set_contains (inv_eclosure, edst1)
+ && re_node_set_contains (dest_nodes, edst1))
+ || (edst2 > 0
+ && !re_node_set_contains (inv_eclosure, edst2)
+ && re_node_set_contains (dest_nodes, edst2)))
+ {
+ err = re_node_set_add_intersect (&except_nodes, candidates,
+ dfa->inveclosures + cur_node);
+ if (BE (err != REG_NOERROR, 0))
+ {
+ re_node_set_free (&except_nodes);
+ return err;
+ }
+ }
+ }
+ }
+ for (ecl_idx = 0; ecl_idx < inv_eclosure->nelem; ++ecl_idx)
+ {
+ Idx cur_node = inv_eclosure->elems[ecl_idx];
+ if (!re_node_set_contains (&except_nodes, cur_node))
+ {
+ Idx idx = re_node_set_contains (dest_nodes, cur_node) - 1;
+ re_node_set_remove_at (dest_nodes, idx);
+ }
+ }
+ re_node_set_free (&except_nodes);
+ return REG_NOERROR;
+}
+
+static bool
+check_dst_limits (const re_match_context_t *mctx, const re_node_set *limits,
+ Idx dst_node, Idx dst_idx, Idx src_node, Idx src_idx)
+{
+ const re_dfa_t *const dfa = mctx->dfa;
+ Idx lim_idx, src_pos, dst_pos;
+
+ Idx dst_bkref_idx = search_cur_bkref_entry (mctx, dst_idx);
+ Idx src_bkref_idx = search_cur_bkref_entry (mctx, src_idx);
+ for (lim_idx = 0; lim_idx < limits->nelem; ++lim_idx)
+ {
+ Idx subexp_idx;
+ struct re_backref_cache_entry *ent;
+ ent = mctx->bkref_ents + limits->elems[lim_idx];
+ subexp_idx = dfa->nodes[ent->node].opr.idx;
+
+ dst_pos = check_dst_limits_calc_pos (mctx, limits->elems[lim_idx],
+ subexp_idx, dst_node, dst_idx,
+ dst_bkref_idx);
+ src_pos = check_dst_limits_calc_pos (mctx, limits->elems[lim_idx],
+ subexp_idx, src_node, src_idx,
+ src_bkref_idx);
+
+ /* In case of:
+ <src> <dst> ( <subexp> )
+ ( <subexp> ) <src> <dst>
+ ( <subexp1> <src> <subexp2> <dst> <subexp3> ) */
+ if (src_pos == dst_pos)
+ continue; /* This is unrelated limitation. */
+ else
+ return true;
+ }
+ return false;
+}
+
+static int
+check_dst_limits_calc_pos_1 (const re_match_context_t *mctx, int boundaries,
+ Idx subexp_idx, Idx from_node, Idx bkref_idx)
+{
+ const re_dfa_t *const dfa = mctx->dfa;
+ const re_node_set *eclosures = dfa->eclosures + from_node;
+ Idx node_idx;
+
+ /* Else, we are on the boundary: examine the nodes on the epsilon
+ closure. */
+ for (node_idx = 0; node_idx < eclosures->nelem; ++node_idx)
+ {
+ Idx node = eclosures->elems[node_idx];
+ switch (dfa->nodes[node].type)
+ {
+ case OP_BACK_REF:
+ if (bkref_idx != -1)
+ {
+ struct re_backref_cache_entry *ent = mctx->bkref_ents + bkref_idx;
+ do
+ {
+ Idx dst;
+ int cpos;
+
+ if (ent->node != node)
+ continue;
+
+ if (subexp_idx < BITSET_WORD_BITS
+ && !(ent->eps_reachable_subexps_map
+ & ((bitset_word_t) 1 << subexp_idx)))
+ continue;
+
+ /* Recurse trying to reach the OP_OPEN_SUBEXP and
+ OP_CLOSE_SUBEXP cases below. But, if the
+ destination node is the same node as the source
+ node, don't recurse because it would cause an
+ infinite loop: a regex that exhibits this behavior
+ is ()\1*\1* */
+ dst = dfa->edests[node].elems[0];
+ if (dst == from_node)
+ {
+ if (boundaries & 1)
+ return -1;
+ else /* if (boundaries & 2) */
+ return 0;
+ }
+
+ cpos =
+ check_dst_limits_calc_pos_1 (mctx, boundaries, subexp_idx,
+ dst, bkref_idx);
+ if (cpos == -1 /* && (boundaries & 1) */)
+ return -1;
+ if (cpos == 0 && (boundaries & 2))
+ return 0;
+
+ if (subexp_idx < BITSET_WORD_BITS)
+ ent->eps_reachable_subexps_map
+ &= ~((bitset_word_t) 1 << subexp_idx);
+ }
+ while (ent++->more);
+ }
+ break;
+
+ case OP_OPEN_SUBEXP:
+ if ((boundaries & 1) && subexp_idx == dfa->nodes[node].opr.idx)
+ return -1;
+ break;
+
+ case OP_CLOSE_SUBEXP:
+ if ((boundaries & 2) && subexp_idx == dfa->nodes[node].opr.idx)
+ return 0;
+ break;
+
+ default:
+ break;
+ }
+ }
+
+ return (boundaries & 2) ? 1 : 0;
+}
+
+static int
+check_dst_limits_calc_pos (const re_match_context_t *mctx, Idx limit,
+ Idx subexp_idx, Idx from_node, Idx str_idx,
+ Idx bkref_idx)
+{
+ struct re_backref_cache_entry *lim = mctx->bkref_ents + limit;
+ int boundaries;
+
+ /* If we are outside the range of the subexpression, return -1 or 1. */
+ if (str_idx < lim->subexp_from)
+ return -1;
+
+ if (lim->subexp_to < str_idx)
+ return 1;
+
+ /* If we are within the subexpression, return 0. */
+ boundaries = (str_idx == lim->subexp_from);
+ boundaries |= (str_idx == lim->subexp_to) << 1;
+ if (boundaries == 0)
+ return 0;
+
+ /* Else, examine epsilon closure. */
+ return check_dst_limits_calc_pos_1 (mctx, boundaries, subexp_idx,
+ from_node, bkref_idx);
+}
+
+/* Check the limitations of sub expressions LIMITS, and remove the nodes
+ which are against limitations from DEST_NODES. */
+
+static reg_errcode_t
+check_subexp_limits (const re_dfa_t *dfa, re_node_set *dest_nodes,
+ const re_node_set *candidates, re_node_set *limits,
+ struct re_backref_cache_entry *bkref_ents, Idx str_idx)
+{
+ reg_errcode_t err;
+ Idx node_idx, lim_idx;
+
+ for (lim_idx = 0; lim_idx < limits->nelem; ++lim_idx)
+ {
+ Idx subexp_idx;
+ struct re_backref_cache_entry *ent;
+ ent = bkref_ents + limits->elems[lim_idx];
+
+ if (str_idx <= ent->subexp_from || ent->str_idx < str_idx)
+ continue; /* This is unrelated limitation. */
+
+ subexp_idx = dfa->nodes[ent->node].opr.idx;
+ if (ent->subexp_to == str_idx)
+ {
+ Idx ops_node = -1;
+ Idx cls_node = -1;
+ for (node_idx = 0; node_idx < dest_nodes->nelem; ++node_idx)
+ {
+ Idx node = dest_nodes->elems[node_idx];
+ re_token_type_t type = dfa->nodes[node].type;
+ if (type == OP_OPEN_SUBEXP
+ && subexp_idx == dfa->nodes[node].opr.idx)
+ ops_node = node;
+ else if (type == OP_CLOSE_SUBEXP
+ && subexp_idx == dfa->nodes[node].opr.idx)
+ cls_node = node;
+ }
+
+ /* Check the limitation of the open subexpression. */
+ /* Note that (ent->subexp_to = str_idx != ent->subexp_from). */
+ if (ops_node >= 0)
+ {
+ err = sub_epsilon_src_nodes (dfa, ops_node, dest_nodes,
+ candidates);
+ if (BE (err != REG_NOERROR, 0))
+ return err;
+ }
+
+ /* Check the limitation of the close subexpression. */
+ if (cls_node >= 0)
+ for (node_idx = 0; node_idx < dest_nodes->nelem; ++node_idx)
+ {
+ Idx node = dest_nodes->elems[node_idx];
+ if (!re_node_set_contains (dfa->inveclosures + node,
+ cls_node)
+ && !re_node_set_contains (dfa->eclosures + node,
+ cls_node))
+ {
+ /* It is against this limitation.
+ Remove it form the current sifted state. */
+ err = sub_epsilon_src_nodes (dfa, node, dest_nodes,
+ candidates);
+ if (BE (err != REG_NOERROR, 0))
+ return err;
+ --node_idx;
+ }
+ }
+ }
+ else /* (ent->subexp_to != str_idx) */
+ {
+ for (node_idx = 0; node_idx < dest_nodes->nelem; ++node_idx)
+ {
+ Idx node = dest_nodes->elems[node_idx];
+ re_token_type_t type = dfa->nodes[node].type;
+ if (type == OP_CLOSE_SUBEXP || type == OP_OPEN_SUBEXP)
+ {
+ if (subexp_idx != dfa->nodes[node].opr.idx)
+ continue;
+ /* It is against this limitation.
+ Remove it form the current sifted state. */
+ err = sub_epsilon_src_nodes (dfa, node, dest_nodes,
+ candidates);
+ if (BE (err != REG_NOERROR, 0))
+ return err;
+ }
+ }
+ }
+ }
+ return REG_NOERROR;
+}
+
+static reg_errcode_t
+__attribute_warn_unused_result__
+sift_states_bkref (const re_match_context_t *mctx, re_sift_context_t *sctx,
+ Idx str_idx, const re_node_set *candidates)
+{
+ const re_dfa_t *const dfa = mctx->dfa;
+ reg_errcode_t err;
+ Idx node_idx, node;
+ re_sift_context_t local_sctx;
+ Idx first_idx = search_cur_bkref_entry (mctx, str_idx);
+
+ if (first_idx == -1)
+ return REG_NOERROR;
+
+ local_sctx.sifted_states = NULL; /* Mark that it hasn't been initialized. */
+
+ for (node_idx = 0; node_idx < candidates->nelem; ++node_idx)
+ {
+ Idx enabled_idx;
+ re_token_type_t type;
+ struct re_backref_cache_entry *entry;
+ node = candidates->elems[node_idx];
+ type = dfa->nodes[node].type;
+ /* Avoid infinite loop for the REs like "()\1+". */
+ if (node == sctx->last_node && str_idx == sctx->last_str_idx)
+ continue;
+ if (type != OP_BACK_REF)
+ continue;
+
+ entry = mctx->bkref_ents + first_idx;
+ enabled_idx = first_idx;
+ do
+ {
+ Idx subexp_len;
+ Idx to_idx;
+ Idx dst_node;
+ bool ok;
+ re_dfastate_t *cur_state;
+
+ if (entry->node != node)
+ continue;
+ subexp_len = entry->subexp_to - entry->subexp_from;
+ to_idx = str_idx + subexp_len;
+ dst_node = (subexp_len ? dfa->nexts[node]
+ : dfa->edests[node].elems[0]);
+
+ if (to_idx > sctx->last_str_idx
+ || sctx->sifted_states[to_idx] == NULL
+ || !STATE_NODE_CONTAINS (sctx->sifted_states[to_idx], dst_node)
+ || check_dst_limits (mctx, &sctx->limits, node,
+ str_idx, dst_node, to_idx))
+ continue;
+
+ if (local_sctx.sifted_states == NULL)
+ {
+ local_sctx = *sctx;
+ err = re_node_set_init_copy (&local_sctx.limits, &sctx->limits);
+ if (BE (err != REG_NOERROR, 0))
+ goto free_return;
+ }
+ local_sctx.last_node = node;
+ local_sctx.last_str_idx = str_idx;
+ ok = re_node_set_insert (&local_sctx.limits, enabled_idx);
+ if (BE (! ok, 0))
+ {
+ err = REG_ESPACE;
+ goto free_return;
+ }
+ cur_state = local_sctx.sifted_states[str_idx];
+ err = sift_states_backward (mctx, &local_sctx);
+ if (BE (err != REG_NOERROR, 0))
+ goto free_return;
+ if (sctx->limited_states != NULL)
+ {
+ err = merge_state_array (dfa, sctx->limited_states,
+ local_sctx.sifted_states,
+ str_idx + 1);
+ if (BE (err != REG_NOERROR, 0))
+ goto free_return;
+ }
+ local_sctx.sifted_states[str_idx] = cur_state;
+ re_node_set_remove (&local_sctx.limits, enabled_idx);
+
+ /* mctx->bkref_ents may have changed, reload the pointer. */
+ entry = mctx->bkref_ents + enabled_idx;
+ }
+ while (enabled_idx++, entry++->more);
+ }
+ err = REG_NOERROR;
+ free_return:
+ if (local_sctx.sifted_states != NULL)
+ {
+ re_node_set_free (&local_sctx.limits);
+ }
+
+ return err;
+}
+
+
+#ifdef RE_ENABLE_I18N
+static int
+sift_states_iter_mb (const re_match_context_t *mctx, re_sift_context_t *sctx,
+ Idx node_idx, Idx str_idx, Idx max_str_idx)
+{
+ const re_dfa_t *const dfa = mctx->dfa;
+ int naccepted;
+ /* Check the node can accept "multi byte". */
+ naccepted = check_node_accept_bytes (dfa, node_idx, &mctx->input, str_idx);
+ if (naccepted > 0 && str_idx + naccepted <= max_str_idx &&
+ !STATE_NODE_CONTAINS (sctx->sifted_states[str_idx + naccepted],
+ dfa->nexts[node_idx]))
+ /* The node can't accept the "multi byte", or the
+ destination was already thrown away, then the node
+ could't accept the current input "multi byte". */
+ naccepted = 0;
+ /* Otherwise, it is sure that the node could accept
+ 'naccepted' bytes input. */
+ return naccepted;
+}
+#endif /* RE_ENABLE_I18N */
+
+
+/* Functions for state transition. */
+
+/* Return the next state to which the current state STATE will transit by
+ accepting the current input byte, and update STATE_LOG if necessary.
+ If STATE can accept a multibyte char/collating element/back reference
+ update the destination of STATE_LOG. */
+
+static re_dfastate_t *
+__attribute_warn_unused_result__
+transit_state (reg_errcode_t *err, re_match_context_t *mctx,
+ re_dfastate_t *state)
+{
+ re_dfastate_t **trtable;
+ unsigned char ch;
+
+#ifdef RE_ENABLE_I18N
+ /* If the current state can accept multibyte. */
+ if (BE (state->accept_mb, 0))
+ {
+ *err = transit_state_mb (mctx, state);
+ if (BE (*err != REG_NOERROR, 0))
+ return NULL;
+ }
+#endif /* RE_ENABLE_I18N */
+
+ /* Then decide the next state with the single byte. */
+#if 0
+ if (0)
+ /* don't use transition table */
+ return transit_state_sb (err, mctx, state);
+#endif
+
+ /* Use transition table */
+ ch = re_string_fetch_byte (&mctx->input);
+ for (;;)
+ {
+ trtable = state->trtable;
+ if (BE (trtable != NULL, 1))
+ return trtable[ch];
+
+ trtable = state->word_trtable;
+ if (BE (trtable != NULL, 1))
+ {
+ unsigned int context;
+ context
+ = re_string_context_at (&mctx->input,
+ re_string_cur_idx (&mctx->input) - 1,
+ mctx->eflags);
+ if (IS_WORD_CONTEXT (context))
+ return trtable[ch + SBC_MAX];
+ else
+ return trtable[ch];
+ }
+
+ if (!build_trtable (mctx->dfa, state))
+ {
+ *err = REG_ESPACE;
+ return NULL;
+ }
+
+ /* Retry, we now have a transition table. */
+ }
+}
+
+/* Update the state_log if we need */
+static re_dfastate_t *
+merge_state_with_log (reg_errcode_t *err, re_match_context_t *mctx,
+ re_dfastate_t *next_state)
+{
+ const re_dfa_t *const dfa = mctx->dfa;
+ Idx cur_idx = re_string_cur_idx (&mctx->input);
+
+ if (cur_idx > mctx->state_log_top)
+ {
+ mctx->state_log[cur_idx] = next_state;
+ mctx->state_log_top = cur_idx;
+ }
+ else if (mctx->state_log[cur_idx] == 0)
+ {
+ mctx->state_log[cur_idx] = next_state;
+ }
+ else
+ {
+ re_dfastate_t *pstate;
+ unsigned int context;
+ re_node_set next_nodes, *log_nodes, *table_nodes = NULL;
+ /* If (state_log[cur_idx] != 0), it implies that cur_idx is
+ the destination of a multibyte char/collating element/
+ back reference. Then the next state is the union set of
+ these destinations and the results of the transition table. */
+ pstate = mctx->state_log[cur_idx];
+ log_nodes = pstate->entrance_nodes;
+ if (next_state != NULL)
+ {
+ table_nodes = next_state->entrance_nodes;
+ *err = re_node_set_init_union (&next_nodes, table_nodes,
+ log_nodes);
+ if (BE (*err != REG_NOERROR, 0))
+ return NULL;
+ }
+ else
+ next_nodes = *log_nodes;
+ /* Note: We already add the nodes of the initial state,
+ then we don't need to add them here. */
+
+ context = re_string_context_at (&mctx->input,
+ re_string_cur_idx (&mctx->input) - 1,
+ mctx->eflags);
+ next_state = mctx->state_log[cur_idx]
+ = re_acquire_state_context (err, dfa, &next_nodes, context);
+ /* We don't need to check errors here, since the return value of
+ this function is next_state and ERR is already set. */
+
+ if (table_nodes != NULL)
+ re_node_set_free (&next_nodes);
+ }
+
+ if (BE (dfa->nbackref, 0) && next_state != NULL)
+ {
+ /* Check OP_OPEN_SUBEXP in the current state in case that we use them
+ later. We must check them here, since the back references in the
+ next state might use them. */
+ *err = check_subexp_matching_top (mctx, &next_state->nodes,
+ cur_idx);
+ if (BE (*err != REG_NOERROR, 0))
+ return NULL;
+
+ /* If the next state has back references. */
+ if (next_state->has_backref)
+ {
+ *err = transit_state_bkref (mctx, &next_state->nodes);
+ if (BE (*err != REG_NOERROR, 0))
+ return NULL;
+ next_state = mctx->state_log[cur_idx];
+ }
+ }
+
+ return next_state;
+}
+
+/* Skip bytes in the input that correspond to part of a
+ multi-byte match, then look in the log for a state
+ from which to restart matching. */
+static re_dfastate_t *
+find_recover_state (reg_errcode_t *err, re_match_context_t *mctx)
+{
+ re_dfastate_t *cur_state;
+ do
+ {
+ Idx max = mctx->state_log_top;
+ Idx cur_str_idx = re_string_cur_idx (&mctx->input);
+
+ do
+ {
+ if (++cur_str_idx > max)
+ return NULL;
+ re_string_skip_bytes (&mctx->input, 1);
+ }
+ while (mctx->state_log[cur_str_idx] == NULL);
+
+ cur_state = merge_state_with_log (err, mctx, NULL);
+ }
+ while (*err == REG_NOERROR && cur_state == NULL);
+ return cur_state;
+}
+
+/* Helper functions for transit_state. */
+
+/* From the node set CUR_NODES, pick up the nodes whose types are
+ OP_OPEN_SUBEXP and which have corresponding back references in the regular
+ expression. And register them to use them later for evaluating the
+ corresponding back references. */
+
+static reg_errcode_t
+check_subexp_matching_top (re_match_context_t *mctx, re_node_set *cur_nodes,
+ Idx str_idx)
+{
+ const re_dfa_t *const dfa = mctx->dfa;
+ Idx node_idx;
+ reg_errcode_t err;
+
+ /* TODO: This isn't efficient.
+ Because there might be more than one nodes whose types are
+ OP_OPEN_SUBEXP and whose index is SUBEXP_IDX, we must check all
+ nodes.
+ E.g. RE: (a){2} */
+ for (node_idx = 0; node_idx < cur_nodes->nelem; ++node_idx)
+ {
+ Idx node = cur_nodes->elems[node_idx];
+ if (dfa->nodes[node].type == OP_OPEN_SUBEXP
+ && dfa->nodes[node].opr.idx < BITSET_WORD_BITS
+ && (dfa->used_bkref_map
+ & ((bitset_word_t) 1 << dfa->nodes[node].opr.idx)))
+ {
+ err = match_ctx_add_subtop (mctx, node, str_idx);
+ if (BE (err != REG_NOERROR, 0))
+ return err;
+ }
+ }
+ return REG_NOERROR;
+}
+
+#if 0
+/* Return the next state to which the current state STATE will transit by
+ accepting the current input byte. */
+
+static re_dfastate_t *
+transit_state_sb (reg_errcode_t *err, re_match_context_t *mctx,
+ re_dfastate_t *state)
+{
+ const re_dfa_t *const dfa = mctx->dfa;
+ re_node_set next_nodes;
+ re_dfastate_t *next_state;
+ Idx node_cnt, cur_str_idx = re_string_cur_idx (&mctx->input);
+ unsigned int context;
+
+ *err = re_node_set_alloc (&next_nodes, state->nodes.nelem + 1);
+ if (BE (*err != REG_NOERROR, 0))
+ return NULL;
+ for (node_cnt = 0; node_cnt < state->nodes.nelem; ++node_cnt)
+ {
+ Idx cur_node = state->nodes.elems[node_cnt];
+ if (check_node_accept (mctx, dfa->nodes + cur_node, cur_str_idx))
+ {
+ *err = re_node_set_merge (&next_nodes,
+ dfa->eclosures + dfa->nexts[cur_node]);
+ if (BE (*err != REG_NOERROR, 0))
+ {
+ re_node_set_free (&next_nodes);
+ return NULL;
+ }
+ }
+ }
+ context = re_string_context_at (&mctx->input, cur_str_idx, mctx->eflags);
+ next_state = re_acquire_state_context (err, dfa, &next_nodes, context);
+ /* We don't need to check errors here, since the return value of
+ this function is next_state and ERR is already set. */
+
+ re_node_set_free (&next_nodes);
+ re_string_skip_bytes (&mctx->input, 1);
+ return next_state;
+}
+#endif
+
+#ifdef RE_ENABLE_I18N
+static reg_errcode_t
+transit_state_mb (re_match_context_t *mctx, re_dfastate_t *pstate)
+{
+ const re_dfa_t *const dfa = mctx->dfa;
+ reg_errcode_t err;
+ Idx i;
+
+ for (i = 0; i < pstate->nodes.nelem; ++i)
+ {
+ re_node_set dest_nodes, *new_nodes;
+ Idx cur_node_idx = pstate->nodes.elems[i];
+ int naccepted;
+ Idx dest_idx;
+ unsigned int context;
+ re_dfastate_t *dest_state;
+
+ if (!dfa->nodes[cur_node_idx].accept_mb)
+ continue;
+
+ if (dfa->nodes[cur_node_idx].constraint)
+ {
+ context = re_string_context_at (&mctx->input,
+ re_string_cur_idx (&mctx->input),
+ mctx->eflags);
+ if (NOT_SATISFY_NEXT_CONSTRAINT (dfa->nodes[cur_node_idx].constraint,
+ context))
+ continue;
+ }
+
+ /* How many bytes the node can accept? */
+ naccepted = check_node_accept_bytes (dfa, cur_node_idx, &mctx->input,
+ re_string_cur_idx (&mctx->input));
+ if (naccepted == 0)
+ continue;
+
+ /* The node can accepts 'naccepted' bytes. */
+ dest_idx = re_string_cur_idx (&mctx->input) + naccepted;
+ mctx->max_mb_elem_len = ((mctx->max_mb_elem_len < naccepted) ? naccepted
+ : mctx->max_mb_elem_len);
+ err = clean_state_log_if_needed (mctx, dest_idx);
+ if (BE (err != REG_NOERROR, 0))
+ return err;
+#ifdef DEBUG
+ assert (dfa->nexts[cur_node_idx] != -1);
+#endif
+ new_nodes = dfa->eclosures + dfa->nexts[cur_node_idx];
+
+ dest_state = mctx->state_log[dest_idx];
+ if (dest_state == NULL)
+ dest_nodes = *new_nodes;
+ else
+ {
+ err = re_node_set_init_union (&dest_nodes,
+ dest_state->entrance_nodes, new_nodes);
+ if (BE (err != REG_NOERROR, 0))
+ return err;
+ }
+ context = re_string_context_at (&mctx->input, dest_idx - 1,
+ mctx->eflags);
+ mctx->state_log[dest_idx]
+ = re_acquire_state_context (&err, dfa, &dest_nodes, context);
+ if (dest_state != NULL)
+ re_node_set_free (&dest_nodes);
+ if (BE (mctx->state_log[dest_idx] == NULL && err != REG_NOERROR, 0))
+ return err;
+ }
+ return REG_NOERROR;
+}
+#endif /* RE_ENABLE_I18N */
+
+static reg_errcode_t
+transit_state_bkref (re_match_context_t *mctx, const re_node_set *nodes)
+{
+ const re_dfa_t *const dfa = mctx->dfa;
+ reg_errcode_t err;
+ Idx i;
+ Idx cur_str_idx = re_string_cur_idx (&mctx->input);
+
+ for (i = 0; i < nodes->nelem; ++i)
+ {
+ Idx dest_str_idx, prev_nelem, bkc_idx;
+ Idx node_idx = nodes->elems[i];
+ unsigned int context;
+ const re_token_t *node = dfa->nodes + node_idx;
+ re_node_set *new_dest_nodes;
+
+ /* Check whether 'node' is a backreference or not. */
+ if (node->type != OP_BACK_REF)
+ continue;
+
+ if (node->constraint)
+ {
+ context = re_string_context_at (&mctx->input, cur_str_idx,
+ mctx->eflags);
+ if (NOT_SATISFY_NEXT_CONSTRAINT (node->constraint, context))
+ continue;
+ }
+
+ /* 'node' is a backreference.
+ Check the substring which the substring matched. */
+ bkc_idx = mctx->nbkref_ents;
+ err = get_subexp (mctx, node_idx, cur_str_idx);
+ if (BE (err != REG_NOERROR, 0))
+ goto free_return;
+
+ /* And add the epsilon closures (which is 'new_dest_nodes') of
+ the backreference to appropriate state_log. */
+#ifdef DEBUG
+ assert (dfa->nexts[node_idx] != -1);
+#endif
+ for (; bkc_idx < mctx->nbkref_ents; ++bkc_idx)
+ {
+ Idx subexp_len;
+ re_dfastate_t *dest_state;
+ struct re_backref_cache_entry *bkref_ent;
+ bkref_ent = mctx->bkref_ents + bkc_idx;
+ if (bkref_ent->node != node_idx || bkref_ent->str_idx != cur_str_idx)
+ continue;
+ subexp_len = bkref_ent->subexp_to - bkref_ent->subexp_from;
+ new_dest_nodes = (subexp_len == 0
+ ? dfa->eclosures + dfa->edests[node_idx].elems[0]
+ : dfa->eclosures + dfa->nexts[node_idx]);
+ dest_str_idx = (cur_str_idx + bkref_ent->subexp_to
+ - bkref_ent->subexp_from);
+ context = re_string_context_at (&mctx->input, dest_str_idx - 1,
+ mctx->eflags);
+ dest_state = mctx->state_log[dest_str_idx];
+ prev_nelem = ((mctx->state_log[cur_str_idx] == NULL) ? 0
+ : mctx->state_log[cur_str_idx]->nodes.nelem);
+ /* Add 'new_dest_node' to state_log. */
+ if (dest_state == NULL)
+ {
+ mctx->state_log[dest_str_idx]
+ = re_acquire_state_context (&err, dfa, new_dest_nodes,
+ context);
+ if (BE (mctx->state_log[dest_str_idx] == NULL
+ && err != REG_NOERROR, 0))
+ goto free_return;
+ }
+ else
+ {
+ re_node_set dest_nodes;
+ err = re_node_set_init_union (&dest_nodes,
+ dest_state->entrance_nodes,
+ new_dest_nodes);
+ if (BE (err != REG_NOERROR, 0))
+ {
+ re_node_set_free (&dest_nodes);
+ goto free_return;
+ }
+ mctx->state_log[dest_str_idx]
+ = re_acquire_state_context (&err, dfa, &dest_nodes, context);
+ re_node_set_free (&dest_nodes);
+ if (BE (mctx->state_log[dest_str_idx] == NULL
+ && err != REG_NOERROR, 0))
+ goto free_return;
+ }
+ /* We need to check recursively if the backreference can epsilon
+ transit. */
+ if (subexp_len == 0
+ && mctx->state_log[cur_str_idx]->nodes.nelem > prev_nelem)
+ {
+ err = check_subexp_matching_top (mctx, new_dest_nodes,
+ cur_str_idx);
+ if (BE (err != REG_NOERROR, 0))
+ goto free_return;
+ err = transit_state_bkref (mctx, new_dest_nodes);
+ if (BE (err != REG_NOERROR, 0))
+ goto free_return;
+ }
+ }
+ }
+ err = REG_NOERROR;
+ free_return:
+ return err;
+}
+
+/* Enumerate all the candidates which the backreference BKREF_NODE can match
+ at BKREF_STR_IDX, and register them by match_ctx_add_entry().
+ Note that we might collect inappropriate candidates here.
+ However, the cost of checking them strictly here is too high, then we
+ delay these checking for prune_impossible_nodes(). */
+
+static reg_errcode_t
+__attribute_warn_unused_result__
+get_subexp (re_match_context_t *mctx, Idx bkref_node, Idx bkref_str_idx)
+{
+ const re_dfa_t *const dfa = mctx->dfa;
+ Idx subexp_num, sub_top_idx;
+ const char *buf = (const char *) re_string_get_buffer (&mctx->input);
+ /* Return if we have already checked BKREF_NODE at BKREF_STR_IDX. */
+ Idx cache_idx = search_cur_bkref_entry (mctx, bkref_str_idx);
+ if (cache_idx != -1)
+ {
+ const struct re_backref_cache_entry *entry
+ = mctx->bkref_ents + cache_idx;
+ do
+ if (entry->node == bkref_node)
+ return REG_NOERROR; /* We already checked it. */
+ while (entry++->more);
+ }
+
+ subexp_num = dfa->nodes[bkref_node].opr.idx;
+
+ /* For each sub expression */
+ for (sub_top_idx = 0; sub_top_idx < mctx->nsub_tops; ++sub_top_idx)
+ {
+ reg_errcode_t err;
+ re_sub_match_top_t *sub_top = mctx->sub_tops[sub_top_idx];
+ re_sub_match_last_t *sub_last;
+ Idx sub_last_idx, sl_str, bkref_str_off;
+
+ if (dfa->nodes[sub_top->node].opr.idx != subexp_num)
+ continue; /* It isn't related. */
+
+ sl_str = sub_top->str_idx;
+ bkref_str_off = bkref_str_idx;
+ /* At first, check the last node of sub expressions we already
+ evaluated. */
+ for (sub_last_idx = 0; sub_last_idx < sub_top->nlasts; ++sub_last_idx)
+ {
+ regoff_t sl_str_diff;
+ sub_last = sub_top->lasts[sub_last_idx];
+ sl_str_diff = sub_last->str_idx - sl_str;
+ /* The matched string by the sub expression match with the substring
+ at the back reference? */
+ if (sl_str_diff > 0)
+ {
+ if (BE (bkref_str_off + sl_str_diff > mctx->input.valid_len, 0))
+ {
+ /* Not enough chars for a successful match. */
+ if (bkref_str_off + sl_str_diff > mctx->input.len)
+ break;
+
+ err = clean_state_log_if_needed (mctx,
+ bkref_str_off
+ + sl_str_diff);
+ if (BE (err != REG_NOERROR, 0))
+ return err;
+ buf = (const char *) re_string_get_buffer (&mctx->input);
+ }
+ if (memcmp (buf + bkref_str_off, buf + sl_str, sl_str_diff) != 0)
+ /* We don't need to search this sub expression any more. */
+ break;
+ }
+ bkref_str_off += sl_str_diff;
+ sl_str += sl_str_diff;
+ err = get_subexp_sub (mctx, sub_top, sub_last, bkref_node,
+ bkref_str_idx);
+
+ /* Reload buf, since the preceding call might have reallocated
+ the buffer. */
+ buf = (const char *) re_string_get_buffer (&mctx->input);
+
+ if (err == REG_NOMATCH)
+ continue;
+ if (BE (err != REG_NOERROR, 0))
+ return err;
+ }
+
+ if (sub_last_idx < sub_top->nlasts)
+ continue;
+ if (sub_last_idx > 0)
+ ++sl_str;
+ /* Then, search for the other last nodes of the sub expression. */
+ for (; sl_str <= bkref_str_idx; ++sl_str)
+ {
+ Idx cls_node;
+ regoff_t sl_str_off;
+ const re_node_set *nodes;
+ sl_str_off = sl_str - sub_top->str_idx;
+ /* The matched string by the sub expression match with the substring
+ at the back reference? */
+ if (sl_str_off > 0)
+ {
+ if (BE (bkref_str_off >= mctx->input.valid_len, 0))
+ {
+ /* If we are at the end of the input, we cannot match. */
+ if (bkref_str_off >= mctx->input.len)
+ break;
+
+ err = extend_buffers (mctx, bkref_str_off + 1);
+ if (BE (err != REG_NOERROR, 0))
+ return err;
+
+ buf = (const char *) re_string_get_buffer (&mctx->input);
+ }
+ if (buf [bkref_str_off++] != buf[sl_str - 1])
+ break; /* We don't need to search this sub expression
+ any more. */
+ }
+ if (mctx->state_log[sl_str] == NULL)
+ continue;
+ /* Does this state have a ')' of the sub expression? */
+ nodes = &mctx->state_log[sl_str]->nodes;
+ cls_node = find_subexp_node (dfa, nodes, subexp_num,
+ OP_CLOSE_SUBEXP);
+ if (cls_node == -1)
+ continue; /* No. */
+ if (sub_top->path == NULL)
+ {
+ sub_top->path = calloc (sizeof (state_array_t),
+ sl_str - sub_top->str_idx + 1);
+ if (sub_top->path == NULL)
+ return REG_ESPACE;
+ }
+ /* Can the OP_OPEN_SUBEXP node arrive the OP_CLOSE_SUBEXP node
+ in the current context? */
+ err = check_arrival (mctx, sub_top->path, sub_top->node,
+ sub_top->str_idx, cls_node, sl_str,
+ OP_CLOSE_SUBEXP);
+ if (err == REG_NOMATCH)
+ continue;
+ if (BE (err != REG_NOERROR, 0))
+ return err;
+ sub_last = match_ctx_add_sublast (sub_top, cls_node, sl_str);
+ if (BE (sub_last == NULL, 0))
+ return REG_ESPACE;
+ err = get_subexp_sub (mctx, sub_top, sub_last, bkref_node,
+ bkref_str_idx);
+ if (err == REG_NOMATCH)
+ continue;
+ }
+ }
+ return REG_NOERROR;
+}
+
+/* Helper functions for get_subexp(). */
+
+/* Check SUB_LAST can arrive to the back reference BKREF_NODE at BKREF_STR.
+ If it can arrive, register the sub expression expressed with SUB_TOP
+ and SUB_LAST. */
+
+static reg_errcode_t
+get_subexp_sub (re_match_context_t *mctx, const re_sub_match_top_t *sub_top,
+ re_sub_match_last_t *sub_last, Idx bkref_node, Idx bkref_str)
+{
+ reg_errcode_t err;
+ Idx to_idx;
+ /* Can the subexpression arrive the back reference? */
+ err = check_arrival (mctx, &sub_last->path, sub_last->node,
+ sub_last->str_idx, bkref_node, bkref_str,
+ OP_OPEN_SUBEXP);
+ if (err != REG_NOERROR)
+ return err;
+ err = match_ctx_add_entry (mctx, bkref_node, bkref_str, sub_top->str_idx,
+ sub_last->str_idx);
+ if (BE (err != REG_NOERROR, 0))
+ return err;
+ to_idx = bkref_str + sub_last->str_idx - sub_top->str_idx;
+ return clean_state_log_if_needed (mctx, to_idx);
+}
+
+/* Find the first node which is '(' or ')' and whose index is SUBEXP_IDX.
+ Search '(' if FL_OPEN, or search ')' otherwise.
+ TODO: This function isn't efficient...
+ Because there might be more than one nodes whose types are
+ OP_OPEN_SUBEXP and whose index is SUBEXP_IDX, we must check all
+ nodes.
+ E.g. RE: (a){2} */
+
+static Idx
+find_subexp_node (const re_dfa_t *dfa, const re_node_set *nodes,
+ Idx subexp_idx, int type)
+{
+ Idx cls_idx;
+ for (cls_idx = 0; cls_idx < nodes->nelem; ++cls_idx)
+ {
+ Idx cls_node = nodes->elems[cls_idx];
+ const re_token_t *node = dfa->nodes + cls_node;
+ if (node->type == type
+ && node->opr.idx == subexp_idx)
+ return cls_node;
+ }
+ return -1;
+}
+
+/* Check whether the node TOP_NODE at TOP_STR can arrive to the node
+ LAST_NODE at LAST_STR. We record the path onto PATH since it will be
+ heavily reused.
+ Return REG_NOERROR if it can arrive, or REG_NOMATCH otherwise. */
+
+static reg_errcode_t
+__attribute_warn_unused_result__
+check_arrival (re_match_context_t *mctx, state_array_t *path, Idx top_node,
+ Idx top_str, Idx last_node, Idx last_str, int type)
+{
+ const re_dfa_t *const dfa = mctx->dfa;
+ reg_errcode_t err = REG_NOERROR;
+ Idx subexp_num, backup_cur_idx, str_idx, null_cnt;
+ re_dfastate_t *cur_state = NULL;
+ re_node_set *cur_nodes, next_nodes;
+ re_dfastate_t **backup_state_log;
+ unsigned int context;
+
+ subexp_num = dfa->nodes[top_node].opr.idx;
+ /* Extend the buffer if we need. */
+ if (BE (path->alloc < last_str + mctx->max_mb_elem_len + 1, 0))
+ {
+ re_dfastate_t **new_array;
+ Idx old_alloc = path->alloc;
+ Idx incr_alloc = last_str + mctx->max_mb_elem_len + 1;
+ Idx new_alloc;
+ if (BE (IDX_MAX - old_alloc < incr_alloc, 0))
+ return REG_ESPACE;
+ new_alloc = old_alloc + incr_alloc;
+ if (BE (SIZE_MAX / sizeof (re_dfastate_t *) < new_alloc, 0))
+ return REG_ESPACE;
+ new_array = re_realloc (path->array, re_dfastate_t *, new_alloc);
+ if (BE (new_array == NULL, 0))
+ return REG_ESPACE;
+ path->array = new_array;
+ path->alloc = new_alloc;
+ memset (new_array + old_alloc, '\0',
+ sizeof (re_dfastate_t *) * (path->alloc - old_alloc));
+ }
+
+ str_idx = path->next_idx ? path->next_idx : top_str;
+
+ /* Temporary modify MCTX. */
+ backup_state_log = mctx->state_log;
+ backup_cur_idx = mctx->input.cur_idx;
+ mctx->state_log = path->array;
+ mctx->input.cur_idx = str_idx;
+
+ /* Setup initial node set. */
+ context = re_string_context_at (&mctx->input, str_idx - 1, mctx->eflags);
+ if (str_idx == top_str)
+ {
+ err = re_node_set_init_1 (&next_nodes, top_node);
+ if (BE (err != REG_NOERROR, 0))
+ return err;
+ err = check_arrival_expand_ecl (dfa, &next_nodes, subexp_num, type);
+ if (BE (err != REG_NOERROR, 0))
+ {
+ re_node_set_free (&next_nodes);
+ return err;
+ }
+ }
+ else
+ {
+ cur_state = mctx->state_log[str_idx];
+ if (cur_state && cur_state->has_backref)
+ {
+ err = re_node_set_init_copy (&next_nodes, &cur_state->nodes);
+ if (BE (err != REG_NOERROR, 0))
+ return err;
+ }
+ else
+ re_node_set_init_empty (&next_nodes);
+ }
+ if (str_idx == top_str || (cur_state && cur_state->has_backref))
+ {
+ if (next_nodes.nelem)
+ {
+ err = expand_bkref_cache (mctx, &next_nodes, str_idx,
+ subexp_num, type);
+ if (BE (err != REG_NOERROR, 0))
+ {
+ re_node_set_free (&next_nodes);
+ return err;
+ }
+ }
+ cur_state = re_acquire_state_context (&err, dfa, &next_nodes, context);
+ if (BE (cur_state == NULL && err != REG_NOERROR, 0))
+ {
+ re_node_set_free (&next_nodes);
+ return err;
+ }
+ mctx->state_log[str_idx] = cur_state;
+ }
+
+ for (null_cnt = 0; str_idx < last_str && null_cnt <= mctx->max_mb_elem_len;)
+ {
+ re_node_set_empty (&next_nodes);
+ if (mctx->state_log[str_idx + 1])
+ {
+ err = re_node_set_merge (&next_nodes,
+ &mctx->state_log[str_idx + 1]->nodes);
+ if (BE (err != REG_NOERROR, 0))
+ {
+ re_node_set_free (&next_nodes);
+ return err;
+ }
+ }
+ if (cur_state)
+ {
+ err = check_arrival_add_next_nodes (mctx, str_idx,
+ &cur_state->non_eps_nodes,
+ &next_nodes);
+ if (BE (err != REG_NOERROR, 0))
+ {
+ re_node_set_free (&next_nodes);
+ return err;
+ }
+ }
+ ++str_idx;
+ if (next_nodes.nelem)
+ {
+ err = check_arrival_expand_ecl (dfa, &next_nodes, subexp_num, type);
+ if (BE (err != REG_NOERROR, 0))
+ {
+ re_node_set_free (&next_nodes);
+ return err;
+ }
+ err = expand_bkref_cache (mctx, &next_nodes, str_idx,
+ subexp_num, type);
+ if (BE (err != REG_NOERROR, 0))
+ {
+ re_node_set_free (&next_nodes);
+ return err;
+ }
+ }
+ context = re_string_context_at (&mctx->input, str_idx - 1, mctx->eflags);
+ cur_state = re_acquire_state_context (&err, dfa, &next_nodes, context);
+ if (BE (cur_state == NULL && err != REG_NOERROR, 0))
+ {
+ re_node_set_free (&next_nodes);
+ return err;
+ }
+ mctx->state_log[str_idx] = cur_state;
+ null_cnt = cur_state == NULL ? null_cnt + 1 : 0;
+ }
+ re_node_set_free (&next_nodes);
+ cur_nodes = (mctx->state_log[last_str] == NULL ? NULL
+ : &mctx->state_log[last_str]->nodes);
+ path->next_idx = str_idx;
+
+ /* Fix MCTX. */
+ mctx->state_log = backup_state_log;
+ mctx->input.cur_idx = backup_cur_idx;
+
+ /* Then check the current node set has the node LAST_NODE. */
+ if (cur_nodes != NULL && re_node_set_contains (cur_nodes, last_node))
+ return REG_NOERROR;
+
+ return REG_NOMATCH;
+}
+
+/* Helper functions for check_arrival. */
+
+/* Calculate the destination nodes of CUR_NODES at STR_IDX, and append them
+ to NEXT_NODES.
+ TODO: This function is similar to the functions transit_state*(),
+ however this function has many additional works.
+ Can't we unify them? */
+
+static reg_errcode_t
+__attribute_warn_unused_result__
+check_arrival_add_next_nodes (re_match_context_t *mctx, Idx str_idx,
+ re_node_set *cur_nodes, re_node_set *next_nodes)
+{
+ const re_dfa_t *const dfa = mctx->dfa;
+ bool ok;
+ Idx cur_idx;
+#ifdef RE_ENABLE_I18N
+ reg_errcode_t err = REG_NOERROR;
+#endif
+ re_node_set union_set;
+ re_node_set_init_empty (&union_set);
+ for (cur_idx = 0; cur_idx < cur_nodes->nelem; ++cur_idx)
+ {
+ int naccepted = 0;
+ Idx cur_node = cur_nodes->elems[cur_idx];
+#ifdef DEBUG
+ re_token_type_t type = dfa->nodes[cur_node].type;
+ assert (!IS_EPSILON_NODE (type));
+#endif
+#ifdef RE_ENABLE_I18N
+ /* If the node may accept "multi byte". */
+ if (dfa->nodes[cur_node].accept_mb)
+ {
+ naccepted = check_node_accept_bytes (dfa, cur_node, &mctx->input,
+ str_idx);
+ if (naccepted > 1)
+ {
+ re_dfastate_t *dest_state;
+ Idx next_node = dfa->nexts[cur_node];
+ Idx next_idx = str_idx + naccepted;
+ dest_state = mctx->state_log[next_idx];
+ re_node_set_empty (&union_set);
+ if (dest_state)
+ {
+ err = re_node_set_merge (&union_set, &dest_state->nodes);
+ if (BE (err != REG_NOERROR, 0))
+ {
+ re_node_set_free (&union_set);
+ return err;
+ }
+ }
+ ok = re_node_set_insert (&union_set, next_node);
+ if (BE (! ok, 0))
+ {
+ re_node_set_free (&union_set);
+ return REG_ESPACE;
+ }
+ mctx->state_log[next_idx] = re_acquire_state (&err, dfa,
+ &union_set);
+ if (BE (mctx->state_log[next_idx] == NULL
+ && err != REG_NOERROR, 0))
+ {
+ re_node_set_free (&union_set);
+ return err;
+ }
+ }
+ }
+#endif /* RE_ENABLE_I18N */
+ if (naccepted
+ || check_node_accept (mctx, dfa->nodes + cur_node, str_idx))
+ {
+ ok = re_node_set_insert (next_nodes, dfa->nexts[cur_node]);
+ if (BE (! ok, 0))
+ {
+ re_node_set_free (&union_set);
+ return REG_ESPACE;
+ }
+ }
+ }
+ re_node_set_free (&union_set);
+ return REG_NOERROR;
+}
+
+/* For all the nodes in CUR_NODES, add the epsilon closures of them to
+ CUR_NODES, however exclude the nodes which are:
+ - inside the sub expression whose number is EX_SUBEXP, if FL_OPEN.
+ - out of the sub expression whose number is EX_SUBEXP, if !FL_OPEN.
+*/
+
+static reg_errcode_t
+check_arrival_expand_ecl (const re_dfa_t *dfa, re_node_set *cur_nodes,
+ Idx ex_subexp, int type)
+{
+ reg_errcode_t err;
+ Idx idx, outside_node;
+ re_node_set new_nodes;
+#ifdef DEBUG
+ assert (cur_nodes->nelem);
+#endif
+ err = re_node_set_alloc (&new_nodes, cur_nodes->nelem);
+ if (BE (err != REG_NOERROR, 0))
+ return err;
+ /* Create a new node set NEW_NODES with the nodes which are epsilon
+ closures of the node in CUR_NODES. */
+
+ for (idx = 0; idx < cur_nodes->nelem; ++idx)
+ {
+ Idx cur_node = cur_nodes->elems[idx];
+ const re_node_set *eclosure = dfa->eclosures + cur_node;
+ outside_node = find_subexp_node (dfa, eclosure, ex_subexp, type);
+ if (outside_node == -1)
+ {
+ /* There are no problematic nodes, just merge them. */
+ err = re_node_set_merge (&new_nodes, eclosure);
+ if (BE (err != REG_NOERROR, 0))
+ {
+ re_node_set_free (&new_nodes);
+ return err;
+ }
+ }
+ else
+ {
+ /* There are problematic nodes, re-calculate incrementally. */
+ err = check_arrival_expand_ecl_sub (dfa, &new_nodes, cur_node,
+ ex_subexp, type);
+ if (BE (err != REG_NOERROR, 0))
+ {
+ re_node_set_free (&new_nodes);
+ return err;
+ }
+ }
+ }
+ re_node_set_free (cur_nodes);
+ *cur_nodes = new_nodes;
+ return REG_NOERROR;
+}
+
+/* Helper function for check_arrival_expand_ecl.
+ Check incrementally the epsilon closure of TARGET, and if it isn't
+ problematic append it to DST_NODES. */
+
+static reg_errcode_t
+__attribute_warn_unused_result__
+check_arrival_expand_ecl_sub (const re_dfa_t *dfa, re_node_set *dst_nodes,
+ Idx target, Idx ex_subexp, int type)
+{
+ Idx cur_node;
+ for (cur_node = target; !re_node_set_contains (dst_nodes, cur_node);)
+ {
+ bool ok;
+
+ if (dfa->nodes[cur_node].type == type
+ && dfa->nodes[cur_node].opr.idx == ex_subexp)
+ {
+ if (type == OP_CLOSE_SUBEXP)
+ {
+ ok = re_node_set_insert (dst_nodes, cur_node);
+ if (BE (! ok, 0))
+ return REG_ESPACE;
+ }
+ break;
+ }
+ ok = re_node_set_insert (dst_nodes, cur_node);
+ if (BE (! ok, 0))
+ return REG_ESPACE;
+ if (dfa->edests[cur_node].nelem == 0)
+ break;
+ if (dfa->edests[cur_node].nelem == 2)
+ {
+ reg_errcode_t err;
+ err = check_arrival_expand_ecl_sub (dfa, dst_nodes,
+ dfa->edests[cur_node].elems[1],
+ ex_subexp, type);
+ if (BE (err != REG_NOERROR, 0))
+ return err;
+ }
+ cur_node = dfa->edests[cur_node].elems[0];
+ }
+ return REG_NOERROR;
+}
+
+
+/* For all the back references in the current state, calculate the
+ destination of the back references by the appropriate entry
+ in MCTX->BKREF_ENTS. */
+
+static reg_errcode_t
+__attribute_warn_unused_result__
+expand_bkref_cache (re_match_context_t *mctx, re_node_set *cur_nodes,
+ Idx cur_str, Idx subexp_num, int type)
+{
+ const re_dfa_t *const dfa = mctx->dfa;
+ reg_errcode_t err;
+ Idx cache_idx_start = search_cur_bkref_entry (mctx, cur_str);
+ struct re_backref_cache_entry *ent;
+
+ if (cache_idx_start == -1)
+ return REG_NOERROR;
+
+ restart:
+ ent = mctx->bkref_ents + cache_idx_start;
+ do
+ {
+ Idx to_idx, next_node;
+
+ /* Is this entry ENT is appropriate? */
+ if (!re_node_set_contains (cur_nodes, ent->node))
+ continue; /* No. */
+
+ to_idx = cur_str + ent->subexp_to - ent->subexp_from;
+ /* Calculate the destination of the back reference, and append it
+ to MCTX->STATE_LOG. */
+ if (to_idx == cur_str)
+ {
+ /* The backreference did epsilon transit, we must re-check all the
+ node in the current state. */
+ re_node_set new_dests;
+ reg_errcode_t err2, err3;
+ next_node = dfa->edests[ent->node].elems[0];
+ if (re_node_set_contains (cur_nodes, next_node))
+ continue;
+ err = re_node_set_init_1 (&new_dests, next_node);
+ err2 = check_arrival_expand_ecl (dfa, &new_dests, subexp_num, type);
+ err3 = re_node_set_merge (cur_nodes, &new_dests);
+ re_node_set_free (&new_dests);
+ if (BE (err != REG_NOERROR || err2 != REG_NOERROR
+ || err3 != REG_NOERROR, 0))
+ {
+ err = (err != REG_NOERROR ? err
+ : (err2 != REG_NOERROR ? err2 : err3));
+ return err;
+ }
+ /* TODO: It is still inefficient... */
+ goto restart;
+ }
+ else
+ {
+ re_node_set union_set;
+ next_node = dfa->nexts[ent->node];
+ if (mctx->state_log[to_idx])
+ {
+ bool ok;
+ if (re_node_set_contains (&mctx->state_log[to_idx]->nodes,
+ next_node))
+ continue;
+ err = re_node_set_init_copy (&union_set,
+ &mctx->state_log[to_idx]->nodes);
+ ok = re_node_set_insert (&union_set, next_node);
+ if (BE (err != REG_NOERROR || ! ok, 0))
+ {
+ re_node_set_free (&union_set);
+ err = err != REG_NOERROR ? err : REG_ESPACE;
+ return err;
+ }
+ }
+ else
+ {
+ err = re_node_set_init_1 (&union_set, next_node);
+ if (BE (err != REG_NOERROR, 0))
+ return err;
+ }
+ mctx->state_log[to_idx] = re_acquire_state (&err, dfa, &union_set);
+ re_node_set_free (&union_set);
+ if (BE (mctx->state_log[to_idx] == NULL
+ && err != REG_NOERROR, 0))
+ return err;
+ }
+ }
+ while (ent++->more);
+ return REG_NOERROR;
+}
+
+/* Build transition table for the state.
+ Return true if successful. */
+
+static bool
+build_trtable (const re_dfa_t *dfa, re_dfastate_t *state)
+{
+ reg_errcode_t err;
+ Idx i, j;
+ int ch;
+ bool need_word_trtable = false;
+ bitset_word_t elem, mask;
+ bool dests_node_malloced = false;
+ bool dest_states_malloced = false;
+ Idx ndests; /* Number of the destination states from 'state'. */
+ re_dfastate_t **trtable;
+ re_dfastate_t **dest_states = NULL, **dest_states_word, **dest_states_nl;
+ re_node_set follows, *dests_node;
+ bitset_t *dests_ch;
+ bitset_t acceptable;
+
+ struct dests_alloc
+ {
+ re_node_set dests_node[SBC_MAX];
+ bitset_t dests_ch[SBC_MAX];
+ } *dests_alloc;
+
+ /* We build DFA states which corresponds to the destination nodes
+ from 'state'. 'dests_node[i]' represents the nodes which i-th
+ destination state contains, and 'dests_ch[i]' represents the
+ characters which i-th destination state accepts. */
+ if (__libc_use_alloca (sizeof (struct dests_alloc)))
+ dests_alloc = (struct dests_alloc *) alloca (sizeof (struct dests_alloc));
+ else
+ {
+ dests_alloc = re_malloc (struct dests_alloc, 1);
+ if (BE (dests_alloc == NULL, 0))
+ return false;
+ dests_node_malloced = true;
+ }
+ dests_node = dests_alloc->dests_node;
+ dests_ch = dests_alloc->dests_ch;
+
+ /* Initialize transition table. */
+ state->word_trtable = state->trtable = NULL;
+
+ /* At first, group all nodes belonging to 'state' into several
+ destinations. */
+ ndests = group_nodes_into_DFAstates (dfa, state, dests_node, dests_ch);
+ if (BE (ndests <= 0, 0))
+ {
+ if (dests_node_malloced)
+ re_free (dests_alloc);
+ /* Return false in case of an error, true otherwise. */
+ if (ndests == 0)
+ {
+ state->trtable = (re_dfastate_t **)
+ calloc (sizeof (re_dfastate_t *), SBC_MAX);
+ if (BE (state->trtable == NULL, 0))
+ return false;
+ return true;
+ }
+ return false;
+ }
+
+ err = re_node_set_alloc (&follows, ndests + 1);
+ if (BE (err != REG_NOERROR, 0))
+ goto out_free;
+
+ /* Avoid arithmetic overflow in size calculation. */
+ if (BE ((((SIZE_MAX - (sizeof (re_node_set) + sizeof (bitset_t)) * SBC_MAX)
+ / (3 * sizeof (re_dfastate_t *)))
+ < ndests),
+ 0))
+ goto out_free;
+
+ if (__libc_use_alloca ((sizeof (re_node_set) + sizeof (bitset_t)) * SBC_MAX
+ + ndests * 3 * sizeof (re_dfastate_t *)))
+ dest_states = (re_dfastate_t **)
+ alloca (ndests * 3 * sizeof (re_dfastate_t *));
+ else
+ {
+ dest_states = re_malloc (re_dfastate_t *, ndests * 3);
+ if (BE (dest_states == NULL, 0))
+ {
+out_free:
+ if (dest_states_malloced)
+ re_free (dest_states);
+ re_node_set_free (&follows);
+ for (i = 0; i < ndests; ++i)
+ re_node_set_free (dests_node + i);
+ if (dests_node_malloced)
+ re_free (dests_alloc);
+ return false;
+ }
+ dest_states_malloced = true;
+ }
+ dest_states_word = dest_states + ndests;
+ dest_states_nl = dest_states_word + ndests;
+ bitset_empty (acceptable);
+
+ /* Then build the states for all destinations. */
+ for (i = 0; i < ndests; ++i)
+ {
+ Idx next_node;
+ re_node_set_empty (&follows);
+ /* Merge the follows of this destination states. */
+ for (j = 0; j < dests_node[i].nelem; ++j)
+ {
+ next_node = dfa->nexts[dests_node[i].elems[j]];
+ if (next_node != -1)
+ {
+ err = re_node_set_merge (&follows, dfa->eclosures + next_node);
+ if (BE (err != REG_NOERROR, 0))
+ goto out_free;
+ }
+ }
+ dest_states[i] = re_acquire_state_context (&err, dfa, &follows, 0);
+ if (BE (dest_states[i] == NULL && err != REG_NOERROR, 0))
+ goto out_free;
+ /* If the new state has context constraint,
+ build appropriate states for these contexts. */
+ if (dest_states[i]->has_constraint)
+ {
+ dest_states_word[i] = re_acquire_state_context (&err, dfa, &follows,
+ CONTEXT_WORD);
+ if (BE (dest_states_word[i] == NULL && err != REG_NOERROR, 0))
+ goto out_free;
+
+ if (dest_states[i] != dest_states_word[i] && dfa->mb_cur_max > 1)
+ need_word_trtable = true;
+
+ dest_states_nl[i] = re_acquire_state_context (&err, dfa, &follows,
+ CONTEXT_NEWLINE);
+ if (BE (dest_states_nl[i] == NULL && err != REG_NOERROR, 0))
+ goto out_free;
+ }
+ else
+ {
+ dest_states_word[i] = dest_states[i];
+ dest_states_nl[i] = dest_states[i];
+ }
+ bitset_merge (acceptable, dests_ch[i]);
+ }
+
+ if (!BE (need_word_trtable, 0))
+ {
+ /* We don't care about whether the following character is a word
+ character, or we are in a single-byte character set so we can
+ discern by looking at the character code: allocate a
+ 256-entry transition table. */
+ trtable = state->trtable =
+ (re_dfastate_t **) calloc (sizeof (re_dfastate_t *), SBC_MAX);
+ if (BE (trtable == NULL, 0))
+ goto out_free;
+
+ /* For all characters ch...: */
+ for (i = 0; i < BITSET_WORDS; ++i)
+ for (ch = i * BITSET_WORD_BITS, elem = acceptable[i], mask = 1;
+ elem;
+ mask <<= 1, elem >>= 1, ++ch)
+ if (BE (elem & 1, 0))
+ {
+ /* There must be exactly one destination which accepts
+ character ch. See group_nodes_into_DFAstates. */
+ for (j = 0; (dests_ch[j][i] & mask) == 0; ++j)
+ ;
+
+ /* j-th destination accepts the word character ch. */
+ if (dfa->word_char[i] & mask)
+ trtable[ch] = dest_states_word[j];
+ else
+ trtable[ch] = dest_states[j];
+ }
+ }
+ else
+ {
+ /* We care about whether the following character is a word
+ character, and we are in a multi-byte character set: discern
+ by looking at the character code: build two 256-entry
+ transition tables, one starting at trtable[0] and one
+ starting at trtable[SBC_MAX]. */
+ trtable = state->word_trtable =
+ (re_dfastate_t **) calloc (sizeof (re_dfastate_t *), 2 * SBC_MAX);
+ if (BE (trtable == NULL, 0))
+ goto out_free;
+
+ /* For all characters ch...: */
+ for (i = 0; i < BITSET_WORDS; ++i)
+ for (ch = i * BITSET_WORD_BITS, elem = acceptable[i], mask = 1;
+ elem;
+ mask <<= 1, elem >>= 1, ++ch)
+ if (BE (elem & 1, 0))
+ {
+ /* There must be exactly one destination which accepts
+ character ch. See group_nodes_into_DFAstates. */
+ for (j = 0; (dests_ch[j][i] & mask) == 0; ++j)
+ ;
+
+ /* j-th destination accepts the word character ch. */
+ trtable[ch] = dest_states[j];
+ trtable[ch + SBC_MAX] = dest_states_word[j];
+ }
+ }
+
+ /* new line */
+ if (bitset_contain (acceptable, NEWLINE_CHAR))
+ {
+ /* The current state accepts newline character. */
+ for (j = 0; j < ndests; ++j)
+ if (bitset_contain (dests_ch[j], NEWLINE_CHAR))
+ {
+ /* k-th destination accepts newline character. */
+ trtable[NEWLINE_CHAR] = dest_states_nl[j];
+ if (need_word_trtable)
+ trtable[NEWLINE_CHAR + SBC_MAX] = dest_states_nl[j];
+ /* There must be only one destination which accepts
+ newline. See group_nodes_into_DFAstates. */
+ break;
+ }
+ }
+
+ if (dest_states_malloced)
+ re_free (dest_states);
+
+ re_node_set_free (&follows);
+ for (i = 0; i < ndests; ++i)
+ re_node_set_free (dests_node + i);
+
+ if (dests_node_malloced)
+ re_free (dests_alloc);
+
+ return true;
+}
+
+/* Group all nodes belonging to STATE into several destinations.
+ Then for all destinations, set the nodes belonging to the destination
+ to DESTS_NODE[i] and set the characters accepted by the destination
+ to DEST_CH[i]. This function return the number of destinations. */
+
+static Idx
+group_nodes_into_DFAstates (const re_dfa_t *dfa, const re_dfastate_t *state,
+ re_node_set *dests_node, bitset_t *dests_ch)
+{
+ reg_errcode_t err;
+ bool ok;
+ Idx i, j, k;
+ Idx ndests; /* Number of the destinations from 'state'. */
+ bitset_t accepts; /* Characters a node can accept. */
+ const re_node_set *cur_nodes = &state->nodes;
+ bitset_empty (accepts);
+ ndests = 0;
+
+ /* For all the nodes belonging to 'state', */
+ for (i = 0; i < cur_nodes->nelem; ++i)
+ {
+ re_token_t *node = &dfa->nodes[cur_nodes->elems[i]];
+ re_token_type_t type = node->type;
+ unsigned int constraint = node->constraint;
+
+ /* Enumerate all single byte character this node can accept. */
+ if (type == CHARACTER)
+ bitset_set (accepts, node->opr.c);
+ else if (type == SIMPLE_BRACKET)
+ {
+ bitset_merge (accepts, node->opr.sbcset);
+ }
+ else if (type == OP_PERIOD)
+ {
+#ifdef RE_ENABLE_I18N
+ if (dfa->mb_cur_max > 1)
+ bitset_merge (accepts, dfa->sb_char);
+ else
+#endif
+ bitset_set_all (accepts);
+ if (!(dfa->syntax & RE_DOT_NEWLINE))
+ bitset_clear (accepts, '\n');
+ if (dfa->syntax & RE_DOT_NOT_NULL)
+ bitset_clear (accepts, '\0');
+ }
+#ifdef RE_ENABLE_I18N
+ else if (type == OP_UTF8_PERIOD)
+ {
+ if (ASCII_CHARS % BITSET_WORD_BITS == 0)
+ memset (accepts, -1, ASCII_CHARS / CHAR_BIT);
+ else
+ bitset_merge (accepts, utf8_sb_map);
+ if (!(dfa->syntax & RE_DOT_NEWLINE))
+ bitset_clear (accepts, '\n');
+ if (dfa->syntax & RE_DOT_NOT_NULL)
+ bitset_clear (accepts, '\0');
+ }
+#endif
+ else
+ continue;
+
+ /* Check the 'accepts' and sift the characters which are not
+ match it the context. */
+ if (constraint)
+ {
+ if (constraint & NEXT_NEWLINE_CONSTRAINT)
+ {
+ bool accepts_newline = bitset_contain (accepts, NEWLINE_CHAR);
+ bitset_empty (accepts);
+ if (accepts_newline)
+ bitset_set (accepts, NEWLINE_CHAR);
+ else
+ continue;
+ }
+ if (constraint & NEXT_ENDBUF_CONSTRAINT)
+ {
+ bitset_empty (accepts);
+ continue;
+ }
+
+ if (constraint & NEXT_WORD_CONSTRAINT)
+ {
+ bitset_word_t any_set = 0;
+ if (type == CHARACTER && !node->word_char)
+ {
+ bitset_empty (accepts);
+ continue;
+ }
+#ifdef RE_ENABLE_I18N
+ if (dfa->mb_cur_max > 1)
+ for (j = 0; j < BITSET_WORDS; ++j)
+ any_set |= (accepts[j] &= (dfa->word_char[j] |
~dfa->sb_char[j]));
+ else
+#endif
+ for (j = 0; j < BITSET_WORDS; ++j)
+ any_set |= (accepts[j] &= dfa->word_char[j]);
+ if (!any_set)
+ continue;
+ }
+ if (constraint & NEXT_NOTWORD_CONSTRAINT)
+ {
+ bitset_word_t any_set = 0;
+ if (type == CHARACTER && node->word_char)
+ {
+ bitset_empty (accepts);
+ continue;
+ }
+#ifdef RE_ENABLE_I18N
+ if (dfa->mb_cur_max > 1)
+ for (j = 0; j < BITSET_WORDS; ++j)
+ any_set |= (accepts[j] &= ~(dfa->word_char[j] &
dfa->sb_char[j]));
+ else
+#endif
+ for (j = 0; j < BITSET_WORDS; ++j)
+ any_set |= (accepts[j] &= ~dfa->word_char[j]);
+ if (!any_set)
+ continue;
+ }
+ }
+
+ /* Then divide 'accepts' into DFA states, or create a new
+ state. Above, we make sure that accepts is not empty. */
+ for (j = 0; j < ndests; ++j)
+ {
+ bitset_t intersec; /* Intersection sets, see below. */
+ bitset_t remains;
+ /* Flags, see below. */
+ bitset_word_t has_intersec, not_subset, not_consumed;
+
+ /* Optimization, skip if this state doesn't accept the character. */
+ if (type == CHARACTER && !bitset_contain (dests_ch[j], node->opr.c))
+ continue;
+
+ /* Enumerate the intersection set of this state and 'accepts'. */
+ has_intersec = 0;
+ for (k = 0; k < BITSET_WORDS; ++k)
+ has_intersec |= intersec[k] = accepts[k] & dests_ch[j][k];
+ /* And skip if the intersection set is empty. */
+ if (!has_intersec)
+ continue;
+
+ /* Then check if this state is a subset of 'accepts'. */
+ not_subset = not_consumed = 0;
+ for (k = 0; k < BITSET_WORDS; ++k)
+ {
+ not_subset |= remains[k] = ~accepts[k] & dests_ch[j][k];
+ not_consumed |= accepts[k] = accepts[k] & ~dests_ch[j][k];
+ }
+
+ /* If this state isn't a subset of 'accepts', create a
+ new group state, which has the 'remains'. */
+ if (not_subset)
+ {
+ bitset_copy (dests_ch[ndests], remains);
+ bitset_copy (dests_ch[j], intersec);
+ err = re_node_set_init_copy (dests_node + ndests, &dests_node[j]);
+ if (BE (err != REG_NOERROR, 0))
+ goto error_return;
+ ++ndests;
+ }
+
+ /* Put the position in the current group. */
+ ok = re_node_set_insert (&dests_node[j], cur_nodes->elems[i]);
+ if (BE (! ok, 0))
+ goto error_return;
+
+ /* If all characters are consumed, go to next node. */
+ if (!not_consumed)
+ break;
+ }
+ /* Some characters remain, create a new group. */
+ if (j == ndests)
+ {
+ bitset_copy (dests_ch[ndests], accepts);
+ err = re_node_set_init_1 (dests_node + ndests, cur_nodes->elems[i]);
+ if (BE (err != REG_NOERROR, 0))
+ goto error_return;
+ ++ndests;
+ bitset_empty (accepts);
+ }
+ }
+ return ndests;
+ error_return:
+ for (j = 0; j < ndests; ++j)
+ re_node_set_free (dests_node + j);
+ return -1;
+}
+
+#ifdef RE_ENABLE_I18N
+/* Check how many bytes the node 'dfa->nodes[node_idx]' accepts.
+ Return the number of the bytes the node accepts.
+ STR_IDX is the current index of the input string.
+
+ This function handles the nodes which can accept one character, or
+ one collating element like '.', '[a-z]', opposite to the other nodes
+ can only accept one byte. */
+
+# ifdef _LIBC
+# include <locale/weight.h>
+# endif
+
+static int
+check_node_accept_bytes (const re_dfa_t *dfa, Idx node_idx,
+ const re_string_t *input, Idx str_idx)
+{
+ const re_token_t *node = dfa->nodes + node_idx;
+ int char_len, elem_len;
+ Idx i;
+
+ if (BE (node->type == OP_UTF8_PERIOD, 0))
+ {
+ unsigned char c = re_string_byte_at (input, str_idx), d;
+ if (BE (c < 0xc2, 1))
+ return 0;
+
+ if (str_idx + 2 > input->len)
+ return 0;
+
+ d = re_string_byte_at (input, str_idx + 1);
+ if (c < 0xe0)
+ return (d < 0x80 || d > 0xbf) ? 0 : 2;
+ else if (c < 0xf0)
+ {
+ char_len = 3;
+ if (c == 0xe0 && d < 0xa0)
+ return 0;
+ }
+ else if (c < 0xf8)
+ {
+ char_len = 4;
+ if (c == 0xf0 && d < 0x90)
+ return 0;
+ }
+ else if (c < 0xfc)
+ {
+ char_len = 5;
+ if (c == 0xf8 && d < 0x88)
+ return 0;
+ }
+ else if (c < 0xfe)
+ {
+ char_len = 6;
+ if (c == 0xfc && d < 0x84)
+ return 0;
+ }
+ else
+ return 0;
+
+ if (str_idx + char_len > input->len)
+ return 0;
+
+ for (i = 1; i < char_len; ++i)
+ {
+ d = re_string_byte_at (input, str_idx + i);
+ if (d < 0x80 || d > 0xbf)
+ return 0;
+ }
+ return char_len;
+ }
+
+ char_len = re_string_char_size_at (input, str_idx);
+ if (node->type == OP_PERIOD)
+ {
+ if (char_len <= 1)
+ return 0;
+ /* FIXME: I don't think this if is needed, as both '\n'
+ and '\0' are char_len == 1. */
+ /* '.' accepts any one character except the following two cases. */
+ if ((!(dfa->syntax & RE_DOT_NEWLINE) &&
+ re_string_byte_at (input, str_idx) == '\n') ||
+ ((dfa->syntax & RE_DOT_NOT_NULL) &&
+ re_string_byte_at (input, str_idx) == '\0'))
+ return 0;
+ return char_len;
+ }
+
+ elem_len = re_string_elem_size_at (input, str_idx);
+ if ((elem_len <= 1 && char_len <= 1) || char_len == 0)
+ return 0;
+
+ if (node->type == COMPLEX_BRACKET)
+ {
+ const re_charset_t *cset = node->opr.mbcset;
+# ifdef _LIBC
+ const unsigned char *pin
+ = ((const unsigned char *) re_string_get_buffer (input) + str_idx);
+ Idx j;
+ uint32_t nrules;
+# endif /* _LIBC */
+ int match_len = 0;
+ wchar_t wc = ((cset->nranges || cset->nchar_classes || cset->nmbchars)
+ ? re_string_wchar_at (input, str_idx) : 0);
+
+ /* match with multibyte character? */
+ for (i = 0; i < cset->nmbchars; ++i)
+ if (wc == cset->mbchars[i])
+ {
+ match_len = char_len;
+ goto check_node_accept_bytes_match;
+ }
+ /* match with character_class? */
+ for (i = 0; i < cset->nchar_classes; ++i)
+ {
+ wctype_t wt = cset->char_classes[i];
+ if (__iswctype (wc, wt))
+ {
+ match_len = char_len;
+ goto check_node_accept_bytes_match;
+ }
+ }
+
+# ifdef _LIBC
+ nrules = _NL_CURRENT_WORD (LC_COLLATE, _NL_COLLATE_NRULES);
+ if (nrules != 0)
+ {
+ unsigned int in_collseq = 0;
+ const int32_t *table, *indirect;
+ const unsigned char *weights, *extra;
+ const char *collseqwc;
+
+ /* match with collating_symbol? */
+ if (cset->ncoll_syms)
+ extra = (const unsigned char *)
+ _NL_CURRENT (LC_COLLATE, _NL_COLLATE_SYMB_EXTRAMB);
+ for (i = 0; i < cset->ncoll_syms; ++i)
+ {
+ const unsigned char *coll_sym = extra + cset->coll_syms[i];
+ /* Compare the length of input collating element and
+ the length of current collating element. */
+ if (*coll_sym != elem_len)
+ continue;
+ /* Compare each bytes. */
+ for (j = 0; j < *coll_sym; j++)
+ if (pin[j] != coll_sym[1 + j])
+ break;
+ if (j == *coll_sym)
+ {
+ /* Match if every bytes is equal. */
+ match_len = j;
+ goto check_node_accept_bytes_match;
+ }
+ }
+
+ if (cset->nranges)
+ {
+ if (elem_len <= char_len)
+ {
+ collseqwc = _NL_CURRENT (LC_COLLATE, _NL_COLLATE_COLLSEQWC);
+ in_collseq = __collseq_table_lookup (collseqwc, wc);
+ }
+ else
+ in_collseq = find_collation_sequence_value (pin, elem_len);
+ }
+ /* match with range expression? */
+ /* FIXME: Implement rational ranges here, too. */
+ for (i = 0; i < cset->nranges; ++i)
+ if (cset->range_starts[i] <= in_collseq
+ && in_collseq <= cset->range_ends[i])
+ {
+ match_len = elem_len;
+ goto check_node_accept_bytes_match;
+ }
+
+ /* match with equivalence_class? */
+ if (cset->nequiv_classes)
+ {
+ const unsigned char *cp = pin;
+ table = (const int32_t *)
+ _NL_CURRENT (LC_COLLATE, _NL_COLLATE_TABLEMB);
+ weights = (const unsigned char *)
+ _NL_CURRENT (LC_COLLATE, _NL_COLLATE_WEIGHTMB);
+ extra = (const unsigned char *)
+ _NL_CURRENT (LC_COLLATE, _NL_COLLATE_EXTRAMB);
+ indirect = (const int32_t *)
+ _NL_CURRENT (LC_COLLATE, _NL_COLLATE_INDIRECTMB);
+ int32_t idx = findidx (table, indirect, extra, &cp, elem_len);
+ int32_t rule = idx >> 24;
+ idx &= 0xffffff;
+ if (idx > 0)
+ {
+ size_t weight_len = weights[idx];
+ for (i = 0; i < cset->nequiv_classes; ++i)
+ {
+ int32_t equiv_class_idx = cset->equiv_classes[i];
+ int32_t equiv_class_rule = equiv_class_idx >> 24;
+ equiv_class_idx &= 0xffffff;
+ if (weights[equiv_class_idx] == weight_len
+ && equiv_class_rule == rule
+ && memcmp (weights + idx + 1,
+ weights + equiv_class_idx + 1,
+ weight_len) == 0)
+ {
+ match_len = elem_len;
+ goto check_node_accept_bytes_match;
+ }
+ }
+ }
+ }
+ }
+ else
+# endif /* _LIBC */
+ {
+ /* match with range expression? */
+ for (i = 0; i < cset->nranges; ++i)
+ {
+ if (cset->range_starts[i] <= wc && wc <= cset->range_ends[i])
+ {
+ match_len = char_len;
+ goto check_node_accept_bytes_match;
+ }
+ }
+ }
+ check_node_accept_bytes_match:
+ if (!cset->non_match)
+ return match_len;
+ else
+ {
+ if (match_len > 0)
+ return 0;
+ else
+ return (elem_len > char_len) ? elem_len : char_len;
+ }
+ }
+ return 0;
+}
+
+# ifdef _LIBC
+static unsigned int
+find_collation_sequence_value (const unsigned char *mbs, size_t mbs_len)
+{
+ uint32_t nrules = _NL_CURRENT_WORD (LC_COLLATE, _NL_COLLATE_NRULES);
+ if (nrules == 0)
+ {
+ if (mbs_len == 1)
+ {
+ /* No valid character. Match it as a single byte character. */
+ const unsigned char *collseq = (const unsigned char *)
+ _NL_CURRENT (LC_COLLATE, _NL_COLLATE_COLLSEQMB);
+ return collseq[mbs[0]];
+ }
+ return UINT_MAX;
+ }
+ else
+ {
+ int32_t idx;
+ const unsigned char *extra = (const unsigned char *)
+ _NL_CURRENT (LC_COLLATE, _NL_COLLATE_SYMB_EXTRAMB);
+ int32_t extrasize = (const unsigned char *)
+ _NL_CURRENT (LC_COLLATE, _NL_COLLATE_SYMB_EXTRAMB + 1) - extra;
+
+ for (idx = 0; idx < extrasize;)
+ {
+ int mbs_cnt;
+ bool found = false;
+ int32_t elem_mbs_len;
+ /* Skip the name of collating element name. */
+ idx = idx + extra[idx] + 1;
+ elem_mbs_len = extra[idx++];
+ if (mbs_len == elem_mbs_len)
+ {
+ for (mbs_cnt = 0; mbs_cnt < elem_mbs_len; ++mbs_cnt)
+ if (extra[idx + mbs_cnt] != mbs[mbs_cnt])
+ break;
+ if (mbs_cnt == elem_mbs_len)
+ /* Found the entry. */
+ found = true;
+ }
+ /* Skip the byte sequence of the collating element. */
+ idx += elem_mbs_len;
+ /* Adjust for the alignment. */
+ idx = (idx + 3) & ~3;
+ /* Skip the collation sequence value. */
+ idx += sizeof (uint32_t);
+ /* Skip the wide char sequence of the collating element. */
+ idx = idx + sizeof (uint32_t) * (*(int32_t *) (extra + idx) + 1);
+ /* If we found the entry, return the sequence value. */
+ if (found)
+ return *(uint32_t *) (extra + idx);
+ /* Skip the collation sequence value. */
+ idx += sizeof (uint32_t);
+ }
+ return UINT_MAX;
+ }
+}
+# endif /* _LIBC */
+#endif /* RE_ENABLE_I18N */
+
+/* Check whether the node accepts the byte which is IDX-th
+ byte of the INPUT. */
+
+static bool
+check_node_accept (const re_match_context_t *mctx, const re_token_t *node,
+ Idx idx)
+{
+ unsigned char ch;
+ ch = re_string_byte_at (&mctx->input, idx);
+ switch (node->type)
+ {
+ case CHARACTER:
+ if (node->opr.c != ch)
+ return false;
+ break;
+
+ case SIMPLE_BRACKET:
+ if (!bitset_contain (node->opr.sbcset, ch))
+ return false;
+ break;
+
+#ifdef RE_ENABLE_I18N
+ case OP_UTF8_PERIOD:
+ if (ch >= ASCII_CHARS)
+ return false;
+ FALLTHROUGH;
+#endif
+ case OP_PERIOD:
+ if ((ch == '\n' && !(mctx->dfa->syntax & RE_DOT_NEWLINE))
+ || (ch == '\0' && (mctx->dfa->syntax & RE_DOT_NOT_NULL)))
+ return false;
+ break;
+
+ default:
+ return false;
+ }
+
+ if (node->constraint)
+ {
+ /* The node has constraints. Check whether the current context
+ satisfies the constraints. */
+ unsigned int context = re_string_context_at (&mctx->input, idx,
+ mctx->eflags);
+ if (NOT_SATISFY_NEXT_CONSTRAINT (node->constraint, context))
+ return false;
+ }
+
+ return true;
+}
+
+/* Extend the buffers, if the buffers have run out. */
+
+static reg_errcode_t
+__attribute_warn_unused_result__
+extend_buffers (re_match_context_t *mctx, int min_len)
+{
+ reg_errcode_t ret;
+ re_string_t *pstr = &mctx->input;
+
+ /* Avoid overflow. */
+ if (BE (MIN (IDX_MAX, SIZE_MAX / sizeof (re_dfastate_t *)) / 2
+ <= pstr->bufs_len, 0))
+ return REG_ESPACE;
+
+ /* Double the lengths of the buffers, but allocate at least MIN_LEN. */
+ ret = re_string_realloc_buffers (pstr,
+ MAX (min_len,
+ MIN (pstr->len, pstr->bufs_len * 2)));
+ if (BE (ret != REG_NOERROR, 0))
+ return ret;
+
+ if (mctx->state_log != NULL)
+ {
+ /* And double the length of state_log. */
+ /* XXX We have no indication of the size of this buffer. If this
+ allocation fail we have no indication that the state_log array
+ does not have the right size. */
+ re_dfastate_t **new_array = re_realloc (mctx->state_log, re_dfastate_t *,
+ pstr->bufs_len + 1);
+ if (BE (new_array == NULL, 0))
+ return REG_ESPACE;
+ mctx->state_log = new_array;
+ }
+
+ /* Then reconstruct the buffers. */
+ if (pstr->icase)
+ {
+#ifdef RE_ENABLE_I18N
+ if (pstr->mb_cur_max > 1)
+ {
+ ret = build_wcs_upper_buffer (pstr);
+ if (BE (ret != REG_NOERROR, 0))
+ return ret;
+ }
+ else
+#endif /* RE_ENABLE_I18N */
+ build_upper_buffer (pstr);
+ }
+ else
+ {
+#ifdef RE_ENABLE_I18N
+ if (pstr->mb_cur_max > 1)
+ build_wcs_buffer (pstr);
+ else
+#endif /* RE_ENABLE_I18N */
+ {
+ if (pstr->trans != NULL)
+ re_string_translate_buffer (pstr);
+ }
+ }
+ return REG_NOERROR;
+}
+
+
+/* Functions for matching context. */
+
+/* Initialize MCTX. */
+
+static reg_errcode_t
+__attribute_warn_unused_result__
+match_ctx_init (re_match_context_t *mctx, int eflags, Idx n)
+{
+ mctx->eflags = eflags;
+ mctx->match_last = -1;
+ if (n > 0)
+ {
+ /* Avoid overflow. */
+ size_t max_object_size =
+ MAX (sizeof (struct re_backref_cache_entry),
+ sizeof (re_sub_match_top_t *));
+ if (BE (MIN (IDX_MAX, SIZE_MAX / max_object_size) < n, 0))
+ return REG_ESPACE;
+
+ mctx->bkref_ents = re_malloc (struct re_backref_cache_entry, n);
+ mctx->sub_tops = re_malloc (re_sub_match_top_t *, n);
+ if (BE (mctx->bkref_ents == NULL || mctx->sub_tops == NULL, 0))
+ return REG_ESPACE;
+ }
+ /* Already zero-ed by the caller.
+ else
+ mctx->bkref_ents = NULL;
+ mctx->nbkref_ents = 0;
+ mctx->nsub_tops = 0; */
+ mctx->abkref_ents = n;
+ mctx->max_mb_elem_len = 1;
+ mctx->asub_tops = n;
+ return REG_NOERROR;
+}
+
+/* Clean the entries which depend on the current input in MCTX.
+ This function must be invoked when the matcher changes the start index
+ of the input, or changes the input string. */
+
+static void
+match_ctx_clean (re_match_context_t *mctx)
+{
+ Idx st_idx;
+ for (st_idx = 0; st_idx < mctx->nsub_tops; ++st_idx)
+ {
+ Idx sl_idx;
+ re_sub_match_top_t *top = mctx->sub_tops[st_idx];
+ for (sl_idx = 0; sl_idx < top->nlasts; ++sl_idx)
+ {
+ re_sub_match_last_t *last = top->lasts[sl_idx];
+ re_free (last->path.array);
+ re_free (last);
+ }
+ re_free (top->lasts);
+ if (top->path)
+ {
+ re_free (top->path->array);
+ re_free (top->path);
+ }
+ re_free (top);
+ }
+
+ mctx->nsub_tops = 0;
+ mctx->nbkref_ents = 0;
+}
+
+/* Free all the memory associated with MCTX. */
+
+static void
+match_ctx_free (re_match_context_t *mctx)
+{
+ /* First, free all the memory associated with MCTX->SUB_TOPS. */
+ match_ctx_clean (mctx);
+ re_free (mctx->sub_tops);
+ re_free (mctx->bkref_ents);
+}
+
+/* Add a new backreference entry to MCTX.
+ Note that we assume that caller never call this function with duplicate
+ entry, and call with STR_IDX which isn't smaller than any existing entry.
+*/
+
+static reg_errcode_t
+__attribute_warn_unused_result__
+match_ctx_add_entry (re_match_context_t *mctx, Idx node, Idx str_idx, Idx from,
+ Idx to)
+{
+ if (mctx->nbkref_ents >= mctx->abkref_ents)
+ {
+ struct re_backref_cache_entry* new_entry;
+ new_entry = re_realloc (mctx->bkref_ents, struct re_backref_cache_entry,
+ mctx->abkref_ents * 2);
+ if (BE (new_entry == NULL, 0))
+ {
+ re_free (mctx->bkref_ents);
+ return REG_ESPACE;
+ }
+ mctx->bkref_ents = new_entry;
+ memset (mctx->bkref_ents + mctx->nbkref_ents, '\0',
+ sizeof (struct re_backref_cache_entry) * mctx->abkref_ents);
+ mctx->abkref_ents *= 2;
+ }
+ if (mctx->nbkref_ents > 0
+ && mctx->bkref_ents[mctx->nbkref_ents - 1].str_idx == str_idx)
+ mctx->bkref_ents[mctx->nbkref_ents - 1].more = 1;
+
+ mctx->bkref_ents[mctx->nbkref_ents].node = node;
+ mctx->bkref_ents[mctx->nbkref_ents].str_idx = str_idx;
+ mctx->bkref_ents[mctx->nbkref_ents].subexp_from = from;
+ mctx->bkref_ents[mctx->nbkref_ents].subexp_to = to;
+
+ /* This is a cache that saves negative results of check_dst_limits_calc_pos.
+ If bit N is clear, means that this entry won't epsilon-transition to
+ an OP_OPEN_SUBEXP or OP_CLOSE_SUBEXP for the N+1-th subexpression. If
+ it is set, check_dst_limits_calc_pos_1 will recurse and try to find one
+ such node.
+
+ A backreference does not epsilon-transition unless it is empty, so set
+ to all zeros if FROM != TO. */
+ mctx->bkref_ents[mctx->nbkref_ents].eps_reachable_subexps_map
+ = (from == to ? -1 : 0);
+
+ mctx->bkref_ents[mctx->nbkref_ents++].more = 0;
+ if (mctx->max_mb_elem_len < to - from)
+ mctx->max_mb_elem_len = to - from;
+ return REG_NOERROR;
+}
+
+/* Return the first entry with the same str_idx, or -1 if none is
+ found. Note that MCTX->BKREF_ENTS is already sorted by MCTX->STR_IDX. */
+
+static Idx
+search_cur_bkref_entry (const re_match_context_t *mctx, Idx str_idx)
+{
+ Idx left, right, mid, last;
+ last = right = mctx->nbkref_ents;
+ for (left = 0; left < right;)
+ {
+ mid = (left + right) / 2;
+ if (mctx->bkref_ents[mid].str_idx < str_idx)
+ left = mid + 1;
+ else
+ right = mid;
+ }
+ if (left < last && mctx->bkref_ents[left].str_idx == str_idx)
+ return left;
+ else
+ return -1;
+}
+
+/* Register the node NODE, whose type is OP_OPEN_SUBEXP, and which matches
+ at STR_IDX. */
+
+static reg_errcode_t
+__attribute_warn_unused_result__
+match_ctx_add_subtop (re_match_context_t *mctx, Idx node, Idx str_idx)
+{
+#ifdef DEBUG
+ assert (mctx->sub_tops != NULL);
+ assert (mctx->asub_tops > 0);
+#endif
+ if (BE (mctx->nsub_tops == mctx->asub_tops, 0))
+ {
+ Idx new_asub_tops = mctx->asub_tops * 2;
+ re_sub_match_top_t **new_array = re_realloc (mctx->sub_tops,
+ re_sub_match_top_t *,
+ new_asub_tops);
+ if (BE (new_array == NULL, 0))
+ return REG_ESPACE;
+ mctx->sub_tops = new_array;
+ mctx->asub_tops = new_asub_tops;
+ }
+ mctx->sub_tops[mctx->nsub_tops] = calloc (1, sizeof (re_sub_match_top_t));
+ if (BE (mctx->sub_tops[mctx->nsub_tops] == NULL, 0))
+ return REG_ESPACE;
+ mctx->sub_tops[mctx->nsub_tops]->node = node;
+ mctx->sub_tops[mctx->nsub_tops++]->str_idx = str_idx;
+ return REG_NOERROR;
+}
+
+/* Register the node NODE, whose type is OP_CLOSE_SUBEXP, and which matches
+ at STR_IDX, whose corresponding OP_OPEN_SUBEXP is SUB_TOP. */
+
+static re_sub_match_last_t *
+match_ctx_add_sublast (re_sub_match_top_t *subtop, Idx node, Idx str_idx)
+{
+ re_sub_match_last_t *new_entry;
+ if (BE (subtop->nlasts == subtop->alasts, 0))
+ {
+ Idx new_alasts = 2 * subtop->alasts + 1;
+ re_sub_match_last_t **new_array = re_realloc (subtop->lasts,
+ re_sub_match_last_t *,
+ new_alasts);
+ if (BE (new_array == NULL, 0))
+ return NULL;
+ subtop->lasts = new_array;
+ subtop->alasts = new_alasts;
+ }
+ new_entry = calloc (1, sizeof (re_sub_match_last_t));
+ if (BE (new_entry != NULL, 1))
+ {
+ subtop->lasts[subtop->nlasts] = new_entry;
+ new_entry->node = node;
+ new_entry->str_idx = str_idx;
+ ++subtop->nlasts;
+ }
+ return new_entry;
+}
+
+static void
+sift_ctx_init (re_sift_context_t *sctx, re_dfastate_t **sifted_sts,
+ re_dfastate_t **limited_sts, Idx last_node, Idx last_str_idx)
+{
+ sctx->sifted_states = sifted_sts;
+ sctx->limited_states = limited_sts;
+ sctx->last_node = last_node;
+ sctx->last_str_idx = last_str_idx;
+ re_node_set_init_empty (&sctx->limits);
+}
diff --git a/m4/builtin-expect.m4 b/m4/builtin-expect.m4
new file mode 100644
index 0000000..a1eaf96
--- /dev/null
+++ b/m4/builtin-expect.m4
@@ -0,0 +1,49 @@
+dnl Check for __builtin_expect.
+
+dnl Copyright 2016-2018 Free Software Foundation, Inc.
+dnl This file is free software; the Free Software Foundation
+dnl gives unlimited permission to copy and/or distribute it,
+dnl with or without modifications, as long as this notice is preserved.
+
+dnl Written by Paul Eggert.
+
+AC_DEFUN([gl___BUILTIN_EXPECT],
+[
+ AC_CACHE_CHECK([for __builtin_expect],
+ [gl_cv___builtin_expect],
+ [AC_LINK_IFELSE(
+ [AC_LANG_SOURCE([[
+ int
+ main (int argc, char **argv)
+ {
+ argc = __builtin_expect (argc, 100);
+ return argv[argc != 100][0];
+ }]])],
+ [gl_cv___builtin_expect=yes],
+ [AC_LINK_IFELSE(
+ [AC_LANG_SOURCE([[
+ #include <builtins.h>
+ int
+ main (int argc, char **argv)
+ {
+ argc = __builtin_expect (argc, 100);
+ return argv[argc != 100][0];
+ }]])],
+ [gl_cv___builtin_expect="in <builtins.h>"],
+ [gl_cv___builtin_expect=no])])])
+ if test "$gl_cv___builtin_expect" = yes; then
+ AC_DEFINE([HAVE___BUILTIN_EXPECT], [1])
+ elif test "$gl_cv___builtin_expect" = "in <builtins.h>"; then
+ AC_DEFINE([HAVE___BUILTIN_EXPECT], [2])
+ fi
+ AH_VERBATIM([HAVE___BUILTIN_EXPECT],
+ [/* Define to 1 if the compiler supports __builtin_expect,
+ and to 2 if <builtins.h> does. */
+#undef HAVE___BUILTIN_EXPECT
+#ifndef HAVE___BUILTIN_EXPECT
+# define __builtin_expect(e, c) (e)
+#elif HAVE___BUILTIN_EXPECT == 2
+# include <builtins.h>
+#endif
+ ])
+])
diff --git a/m4/eealloc.m4 b/m4/eealloc.m4
new file mode 100644
index 0000000..a5a4e26
--- /dev/null
+++ b/m4/eealloc.m4
@@ -0,0 +1,31 @@
+# eealloc.m4 serial 3
+dnl Copyright (C) 2003, 2009-2018 Free Software Foundation, Inc.
+dnl This file is free software; the Free Software Foundation
+dnl gives unlimited permission to copy and/or distribute it,
+dnl with or without modifications, as long as this notice is preserved.
+
+AC_DEFUN([gl_EEALLOC],
+[
+ AC_REQUIRE([gl_EEMALLOC])
+ AC_REQUIRE([gl_EEREALLOC])
+])
+
+AC_DEFUN([gl_EEMALLOC],
+[
+ _AC_FUNC_MALLOC_IF(
+ [gl_cv_func_malloc_0_nonnull=1],
+ [gl_cv_func_malloc_0_nonnull=0])
+ AC_DEFINE_UNQUOTED([MALLOC_0_IS_NONNULL], [$gl_cv_func_malloc_0_nonnull],
+ [If malloc(0) is != NULL, define this to 1. Otherwise define this
+ to 0.])
+])
+
+AC_DEFUN([gl_EEREALLOC],
+[
+ _AC_FUNC_REALLOC_IF(
+ [gl_cv_func_realloc_0_nonnull=1],
+ [gl_cv_func_realloc_0_nonnull=0])
+ AC_DEFINE_UNQUOTED([REALLOC_0_IS_NONNULL], [$gl_cv_func_realloc_0_nonnull],
+ [If realloc(NULL,0) is != NULL, define this to 1. Otherwise define this
+ to 0.])
+])
diff --git a/m4/glibc21.m4 b/m4/glibc21.m4
new file mode 100644
index 0000000..126aa1a
--- /dev/null
+++ b/m4/glibc21.m4
@@ -0,0 +1,34 @@
+# glibc21.m4 serial 5
+dnl Copyright (C) 2000-2002, 2004, 2008, 2010-2018 Free Software Foundation,
+dnl Inc.
+dnl This file is free software; the Free Software Foundation
+dnl gives unlimited permission to copy and/or distribute it,
+dnl with or without modifications, as long as this notice is preserved.
+
+# Test for the GNU C Library, version 2.1 or newer, or uClibc.
+# From Bruno Haible.
+
+AC_DEFUN([gl_GLIBC21],
+ [
+ AC_CACHE_CHECK([whether we are using the GNU C Library >= 2.1 or uClibc],
+ [ac_cv_gnu_library_2_1],
+ [AC_EGREP_CPP([Lucky],
+ [
+#include <features.h>
+#ifdef __GNU_LIBRARY__
+ #if (__GLIBC__ == 2 && __GLIBC_MINOR__ >= 1) || (__GLIBC__ > 2)
+ Lucky GNU user
+ #endif
+#endif
+#ifdef __UCLIBC__
+ Lucky user
+#endif
+ ],
+ [ac_cv_gnu_library_2_1=yes],
+ [ac_cv_gnu_library_2_1=no])
+ ]
+ )
+ AC_SUBST([GLIBC21])
+ GLIBC21="$ac_cv_gnu_library_2_1"
+ ]
+)
diff --git a/m4/gnulib-comp.m4 b/m4/gnulib-comp.m4
index 494c77c..61aabaa 100644
--- a/m4/gnulib-comp.m4
+++ b/m4/gnulib-comp.m4
@@ -48,6 +48,7 @@ AC_DEFUN([gl_EARLY],
# Code from module allocator:
# Code from module at-internal:
# Code from module binary-io:
+ # Code from module builtin-expect:
# Code from module byteswap:
# Code from module c-ctype:
# Code from module c-strcase:
@@ -129,6 +130,7 @@ AC_DEFUN([gl_EARLY],
# Code from module qcopy-acl:
# Code from module readlink:
# Code from module readlinkat:
+ # Code from module regex:
# Code from module root-uid:
# Code from module sig2str:
# Code from module signal-h:
@@ -358,6 +360,11 @@ AC_DEFUN([gl_INIT],
AC_LIBOBJ([readlinkat])
fi
gl_UNISTD_MODULE_INDICATOR([readlinkat])
+ gl_REGEX
+ if test $ac_use_included_regex = yes; then
+ AC_LIBOBJ([regex])
+ gl_PREREQ_REGEX
+ fi
gl_FUNC_SIG2STR
if test $ac_cv_func_sig2str = no; then
AC_LIBOBJ([sig2str])
@@ -425,6 +432,7 @@ AC_DEFUN([gl_INIT],
gl_UTIMENS
AC_C_VARARRAYS
gl_gnulib_enabled_260941c0e5dc67ec9e87d1fb321c300b=false
+ gl_gnulib_enabled_37f71b604aa9c54446783d80f42fe547=false
gl_gnulib_enabled_cloexec=false
gl_gnulib_enabled_dirfd=false
gl_gnulib_enabled_dosname=false
@@ -448,6 +456,13 @@ AC_DEFUN([gl_INIT],
func_gl_gnulib_m4code_open
fi
}
+ func_gl_gnulib_m4code_37f71b604aa9c54446783d80f42fe547 ()
+ {
+ if ! $gl_gnulib_enabled_37f71b604aa9c54446783d80f42fe547; then
+ gl___BUILTIN_EXPECT
+ gl_gnulib_enabled_37f71b604aa9c54446783d80f42fe547=true
+ fi
+ }
func_gl_gnulib_m4code_cloexec ()
{
if ! $gl_gnulib_enabled_cloexec; then
@@ -651,6 +666,9 @@ AC_DEFUN([gl_INIT],
if test $HAVE_READLINKAT = 0; then
func_gl_gnulib_m4code_03e0aaad4cb89ca757653bd367a6ccb7
fi
+ if test $ac_use_included_regex = yes; then
+ func_gl_gnulib_m4code_37f71b604aa9c54446783d80f42fe547
+ fi
if { test $HAVE_DECL_STRTOIMAX = 0 || test $REPLACE_STRTOIMAX = 1; } && test
$ac_cv_type_long_long_int = yes; then
func_gl_gnulib_m4code_strtoll
fi
@@ -659,6 +677,7 @@ AC_DEFUN([gl_INIT],
fi
m4_pattern_allow([^gl_GNULIB_ENABLED_])
AM_CONDITIONAL([gl_GNULIB_ENABLED_260941c0e5dc67ec9e87d1fb321c300b],
[$gl_gnulib_enabled_260941c0e5dc67ec9e87d1fb321c300b])
+ AM_CONDITIONAL([gl_GNULIB_ENABLED_37f71b604aa9c54446783d80f42fe547],
[$gl_gnulib_enabled_37f71b604aa9c54446783d80f42fe547])
AM_CONDITIONAL([gl_GNULIB_ENABLED_cloexec], [$gl_gnulib_enabled_cloexec])
AM_CONDITIONAL([gl_GNULIB_ENABLED_dirfd], [$gl_gnulib_enabled_dirfd])
AM_CONDITIONAL([gl_GNULIB_ENABLED_dosname], [$gl_gnulib_enabled_dosname])
@@ -924,6 +943,12 @@ AC_DEFUN([gl_FILE_LIST], [
lib/qcopy-acl.c
lib/readlink.c
lib/readlinkat.c
+ lib/regcomp.c
+ lib/regex.c
+ lib/regex.h
+ lib/regex_internal.c
+ lib/regex_internal.h
+ lib/regexec.c
lib/root-uid.h
lib/set-permissions.c
lib/sha1.c
@@ -980,6 +1005,7 @@ AC_DEFUN([gl_FILE_LIST], [
m4/absolute-header.m4
m4/acl.m4
m4/alloca.m4
+ m4/builtin-expect.m4
m4/byteswap.m4
m4/c-strtod.m4
m4/clock_time.m4
@@ -991,6 +1017,7 @@ AC_DEFUN([gl_FILE_LIST], [
m4/dirent_h.m4
m4/dirfd.m4
m4/dup2.m4
+ m4/eealloc.m4
m4/environ.m4
m4/errno_h.m4
m4/euidaccess.m4
@@ -1018,6 +1045,7 @@ AC_DEFUN([gl_FILE_LIST], [
m4/gettime.m4
m4/gettimeofday.m4
m4/gl-openssl.m4
+ m4/glibc21.m4
m4/gnulib-common.m4
m4/group-member.m4
m4/ieee754-h.m4
@@ -1030,6 +1058,7 @@ AC_DEFUN([gl_FILE_LIST], [
m4/lstat.m4
m4/manywarnings-c++.m4
m4/manywarnings.m4
+ m4/mbstate_t.m4
m4/md5.m4
m4/memrchr.m4
m4/minmax.m4
@@ -1048,6 +1077,7 @@ AC_DEFUN([gl_FILE_LIST], [
m4/putenv.m4
m4/readlink.m4
m4/readlinkat.m4
+ m4/regex.m4
m4/sha1.m4
m4/sha256.m4
m4/sha512.m4
diff --git a/m4/mbstate_t.m4 b/m4/mbstate_t.m4
new file mode 100644
index 0000000..004aa0d
--- /dev/null
+++ b/m4/mbstate_t.m4
@@ -0,0 +1,41 @@
+# mbstate_t.m4 serial 13
+dnl Copyright (C) 2000-2002, 2008-2018 Free Software Foundation, Inc.
+dnl This file is free software; the Free Software Foundation
+dnl gives unlimited permission to copy and/or distribute it,
+dnl with or without modifications, as long as this notice is preserved.
+
+# From Paul Eggert.
+
+# BeOS 5 has <wchar.h> but does not define mbstate_t,
+# so you can't declare an object of that type.
+# Check for this incompatibility with Standard C.
+
+# AC_TYPE_MBSTATE_T
+# -----------------
+AC_DEFUN([AC_TYPE_MBSTATE_T],
+[
+ AC_REQUIRE([AC_USE_SYSTEM_EXTENSIONS]) dnl for HP-UX 11.11
+
+ AC_CACHE_CHECK([for mbstate_t], [ac_cv_type_mbstate_t],
+ [AC_COMPILE_IFELSE(
+ [AC_LANG_PROGRAM(
+ [AC_INCLUDES_DEFAULT[
+/* Tru64 with Desktop Toolkit C has a bug: <stdio.h> must be included before
+ <wchar.h>.
+ BSD/OS 4.0.1 has a bug: <stddef.h>, <stdio.h> and <time.h> must be
+ included before <wchar.h>. */
+#include <stddef.h>
+#include <stdio.h>
+#include <time.h>
+#include <wchar.h>]],
+ [[mbstate_t x; return sizeof x;]])],
+ [ac_cv_type_mbstate_t=yes],
+ [ac_cv_type_mbstate_t=no])])
+ if test $ac_cv_type_mbstate_t = yes; then
+ AC_DEFINE([HAVE_MBSTATE_T], [1],
+ [Define to 1 if <wchar.h> declares mbstate_t.])
+ else
+ AC_DEFINE([mbstate_t], [int],
+ [Define to a type if <wchar.h> does not define.])
+ fi
+])
diff --git a/m4/regex.m4 b/m4/regex.m4
new file mode 100644
index 0000000..055d71b
--- /dev/null
+++ b/m4/regex.m4
@@ -0,0 +1,300 @@
+# serial 67
+
+# Copyright (C) 1996-2001, 2003-2018 Free Software Foundation, Inc.
+#
+# This file is free software; the Free Software Foundation
+# gives unlimited permission to copy and/or distribute it,
+# with or without modifications, as long as this notice is preserved.
+
+dnl Initially derived from code in GNU grep.
+dnl Mostly written by Jim Meyering.
+
+AC_PREREQ([2.50])
+
+AC_DEFUN([gl_REGEX],
+[
+ AC_REQUIRE([AC_CANONICAL_HOST]) dnl for cross-compiles
+ AC_ARG_WITH([included-regex],
+ [AS_HELP_STRING([--without-included-regex],
+ [don't compile regex; this is the default on systems
+ with recent-enough versions of the GNU C Library
+ (use with caution on other systems).])])
+
+ case $with_included_regex in #(
+ yes|no) ac_use_included_regex=$with_included_regex
+ ;;
+ '')
+ # If the system regex support is good enough that it passes the
+ # following run test, then default to *not* using the included regex.c.
+ # If cross compiling, assume the test would fail and use the included
+ # regex.c.
+ AC_CHECK_DECLS_ONCE([alarm])
+ AC_CHECK_HEADERS_ONCE([malloc.h])
+ AC_CACHE_CHECK([for working re_compile_pattern],
+ [gl_cv_func_re_compile_pattern_working],
+ [AC_RUN_IFELSE(
+ [AC_LANG_PROGRAM(
+ [[#include <regex.h>
+
+ #include <locale.h>
+ #include <limits.h>
+ #include <string.h>
+
+ #if defined M_CHECK_ACTION || HAVE_DECL_ALARM
+ # include <signal.h>
+ # include <unistd.h>
+ #endif
+
+ #if HAVE_MALLOC_H
+ # include <malloc.h>
+ #endif
+
+ #ifdef M_CHECK_ACTION
+ /* Exit with distinguishable exit code. */
+ static void sigabrt_no_core (int sig) { raise (SIGTERM); }
+ #endif
+ ]],
+ [[int result = 0;
+ static struct re_pattern_buffer regex;
+ unsigned char folded_chars[UCHAR_MAX + 1];
+ int i;
+ const char *s;
+ struct re_registers regs;
+
+ /* Some builds of glibc go into an infinite loop on this
+ test. Use alarm to force death, and mallopt to avoid
+ malloc recursion in diagnosing the corrupted heap. */
+#if HAVE_DECL_ALARM
+ signal (SIGALRM, SIG_DFL);
+ alarm (2);
+#endif
+#ifdef M_CHECK_ACTION
+ signal (SIGABRT, sigabrt_no_core);
+ mallopt (M_CHECK_ACTION, 2);
+#endif
+
+ if (setlocale (LC_ALL, "en_US.UTF-8"))
+ {
+ {
+ /*
https://sourceware.org/ml/libc-hacker/2006-09/msg00008.html
+ This test needs valgrind to catch the bug on Debian
+ GNU/Linux 3.1 x86, but it might catch the bug better
+ on other platforms and it shouldn't hurt to try the
+ test here. */
+ static char const pat[] = "insert into";
+ static char const data[] =
+ "\xFF\0\x12\xA2\xAA\xC4\xB1,K\x12\xC4\xB1*\xACK";
+ re_set_syntax (RE_SYNTAX_GREP | RE_HAT_LISTS_NOT_NEWLINE
+ | RE_ICASE);
+ memset (®ex, 0, sizeof regex);
+ s = re_compile_pattern (pat, sizeof pat - 1, ®ex);
+ if (s)
+ result |= 1;
+ else if (re_search (®ex, data, sizeof data - 1,
+ 0, sizeof data - 1, ®s)
+ != -1)
+ result |= 1;
+ regfree (®ex);
+ }
+
+ {
+ /* This test is from glibc bug 15078.
+ The test case is from Andreas Schwab in
+
<https://sourceware.org/ml/libc-alpha/2013-01/msg00967.html>.
+ */
+ static char const pat[] = "[^x]x";
+ static char const data[] =
+ /*
<U1000><U103B><U103D><U1014><U103A><U102F><U1015><U103A> */
+ "\xe1\x80\x80"
+ "\xe1\x80\xbb"
+ "\xe1\x80\xbd"
+ "\xe1\x80\x94"
+ "\xe1\x80\xba"
+ "\xe1\x80\xaf"
+ "\xe1\x80\x95"
+ "\xe1\x80\xba"
+ "x";
+ re_set_syntax (0);
+ memset (®ex, 0, sizeof regex);
+ s = re_compile_pattern (pat, sizeof pat - 1, ®ex);
+ if (s)
+ result |= 1;
+ else
+ {
+ i = re_search (®ex, data, sizeof data - 1,
+ 0, sizeof data - 1, 0);
+ if (i != 0 && i != 21)
+ result |= 1;
+ }
+ regfree (®ex);
+ }
+
+ if (! setlocale (LC_ALL, "C"))
+ return 1;
+ }
+
+ /* This test is from glibc bug 3957, reported by Andrew Mackey. */
+ re_set_syntax (RE_SYNTAX_EGREP | RE_HAT_LISTS_NOT_NEWLINE);
+ memset (®ex, 0, sizeof regex);
+ s = re_compile_pattern ("a[^x]b", 6, ®ex);
+ if (s)
+ result |= 2;
+ /* This should fail, but succeeds for glibc-2.5. */
+ else if (re_search (®ex, "a\nb", 3, 0, 3, ®s) != -1)
+ result |= 2;
+
+ /* This regular expression is from Spencer ere test number 75
+ in grep-2.3. */
+ re_set_syntax (RE_SYNTAX_POSIX_EGREP);
+ memset (®ex, 0, sizeof regex);
+ for (i = 0; i <= UCHAR_MAX; i++)
+ folded_chars[i] = i;
+ regex.translate = folded_chars;
+ s = re_compile_pattern ("a[[:@:>@:]]b\n", 11, ®ex);
+ /* This should fail with _Invalid character class name_ error. */
+ if (!s)
+ result |= 4;
+
+ /* Ensure that [b-a] is diagnosed as invalid, when
+ using RE_NO_EMPTY_RANGES. */
+ re_set_syntax (RE_SYNTAX_POSIX_EGREP | RE_NO_EMPTY_RANGES);
+ memset (®ex, 0, sizeof regex);
+ s = re_compile_pattern ("a[b-a]", 6, ®ex);
+ if (s == 0)
+ result |= 8;
+
+ /* This should succeed, but does not for glibc-2.1.3. */
+ memset (®ex, 0, sizeof regex);
+ s = re_compile_pattern ("{1", 2, ®ex);
+ if (s)
+ result |= 8;
+
+ /* The following example is derived from a problem report
+ against gawk from Jorge Stolfi <address@hidden>. */
+ memset (®ex, 0, sizeof regex);
+ s = re_compile_pattern ("[an\371]*n", 7, ®ex);
+ if (s)
+ result |= 8;
+ /* This should match, but does not for glibc-2.2.1. */
+ else if (re_match (®ex, "an", 2, 0, ®s) != 2)
+ result |= 8;
+
+ memset (®ex, 0, sizeof regex);
+ s = re_compile_pattern ("x", 1, ®ex);
+ if (s)
+ result |= 8;
+ /* glibc-2.2.93 does not work with a negative RANGE argument. */
+ else if (re_search (®ex, "wxy", 3, 2, -2, ®s) != 1)
+ result |= 8;
+
+ /* The version of regex.c in older versions of gnulib
+ ignored RE_ICASE. Detect that problem too. */
+ re_set_syntax (RE_SYNTAX_EMACS | RE_ICASE);
+ memset (®ex, 0, sizeof regex);
+ s = re_compile_pattern ("x", 1, ®ex);
+ if (s)
+ result |= 16;
+ else if (re_search (®ex, "WXY", 3, 0, 3, ®s) < 0)
+ result |= 16;
+
+ /* Catch a bug reported by Vin Shelton in
+ https://lists.gnu.org/r/bug-coreutils/2007-06/msg00089.html
+ */
+ re_set_syntax (RE_SYNTAX_POSIX_BASIC
+ & ~RE_CONTEXT_INVALID_DUP
+ & ~RE_NO_EMPTY_RANGES);
+ memset (®ex, 0, sizeof regex);
+ s = re_compile_pattern ("[[:alnum:]_-]\\\\+$", 16, ®ex);
+ if (s)
+ result |= 32;
+
+ /* REG_STARTEND was added to glibc on 2004-01-15.
+ Reject older versions. */
+ if (! REG_STARTEND)
+ result |= 64;
+
+#if 0
+ /* It would be nice to reject hosts whose regoff_t values are too
+ narrow (including glibc on hosts with 64-bit ptrdiff_t and
+ 32-bit int), but we should wait until glibc implements this
+ feature. Otherwise, support for equivalence classes and
+ multibyte collation symbols would always be broken except
+ when compiling --without-included-regex. */
+ if (sizeof (regoff_t) < sizeof (ptrdiff_t)
+ || sizeof (regoff_t) < sizeof (ssize_t))
+ result |= 64;
+#endif
+
+ return result;
+ ]])],
+ [gl_cv_func_re_compile_pattern_working=yes],
+ [gl_cv_func_re_compile_pattern_working=no],
+ [case "$host_os" in
+ # Guess no on native Windows.
+ mingw*) gl_cv_func_re_compile_pattern_working="guessing no" ;;
+ # Otherwise, assume it is not working.
+ *) gl_cv_func_re_compile_pattern_working="guessing no" ;;
+ esac
+ ])
+ ])
+ case "$gl_cv_func_re_compile_pattern_working" in #(
+ *yes) ac_use_included_regex=no;; #(
+ *no) ac_use_included_regex=yes;;
+ esac
+ ;;
+ *) AC_MSG_ERROR([Invalid value for --with-included-regex:
$with_included_regex])
+ ;;
+ esac
+
+ if test $ac_use_included_regex = yes; then
+ AC_DEFINE([_REGEX_INCLUDE_LIMITS_H], [1],
+ [Define if you want <regex.h> to include <limits.h>, so that it
+ consistently overrides <limits.h>'s RE_DUP_MAX.])
+ AC_DEFINE([_REGEX_LARGE_OFFSETS], [1],
+ [Define if you want regoff_t to be at least as wide POSIX requires.])
+ AC_DEFINE([re_syntax_options], [rpl_re_syntax_options],
+ [Define to rpl_re_syntax_options if the replacement should be used.])
+ AC_DEFINE([re_set_syntax], [rpl_re_set_syntax],
+ [Define to rpl_re_set_syntax if the replacement should be used.])
+ AC_DEFINE([re_compile_pattern], [rpl_re_compile_pattern],
+ [Define to rpl_re_compile_pattern if the replacement should be used.])
+ AC_DEFINE([re_compile_fastmap], [rpl_re_compile_fastmap],
+ [Define to rpl_re_compile_fastmap if the replacement should be used.])
+ AC_DEFINE([re_search], [rpl_re_search],
+ [Define to rpl_re_search if the replacement should be used.])
+ AC_DEFINE([re_search_2], [rpl_re_search_2],
+ [Define to rpl_re_search_2 if the replacement should be used.])
+ AC_DEFINE([re_match], [rpl_re_match],
+ [Define to rpl_re_match if the replacement should be used.])
+ AC_DEFINE([re_match_2], [rpl_re_match_2],
+ [Define to rpl_re_match_2 if the replacement should be used.])
+ AC_DEFINE([re_set_registers], [rpl_re_set_registers],
+ [Define to rpl_re_set_registers if the replacement should be used.])
+ AC_DEFINE([re_comp], [rpl_re_comp],
+ [Define to rpl_re_comp if the replacement should be used.])
+ AC_DEFINE([re_exec], [rpl_re_exec],
+ [Define to rpl_re_exec if the replacement should be used.])
+ AC_DEFINE([regcomp], [rpl_regcomp],
+ [Define to rpl_regcomp if the replacement should be used.])
+ AC_DEFINE([regexec], [rpl_regexec],
+ [Define to rpl_regexec if the replacement should be used.])
+ AC_DEFINE([regerror], [rpl_regerror],
+ [Define to rpl_regerror if the replacement should be used.])
+ AC_DEFINE([regfree], [rpl_regfree],
+ [Define to rpl_regfree if the replacement should be used.])
+ fi
+])
+
+# Prerequisites of lib/regex.c and lib/regex_internal.c.
+AC_DEFUN([gl_PREREQ_REGEX],
+[
+ AC_REQUIRE([AC_USE_SYSTEM_EXTENSIONS])
+ AC_REQUIRE([AC_C_INLINE])
+ AC_REQUIRE([AC_C_RESTRICT])
+ AC_REQUIRE([AC_TYPE_MBSTATE_T])
+ AC_REQUIRE([gl_EEMALLOC])
+ AC_REQUIRE([gl_GLIBC21])
+ AC_CHECK_HEADERS([libintl.h])
+ AC_CHECK_FUNCS_ONCE([isblank iswctype])
+ AC_CHECK_DECLS([isblank], [], [], [[#include <ctype.h>]])
+])
diff --git a/src/conf_post.h b/src/conf_post.h
index 8d56f0b..9758298 100644
--- a/src/conf_post.h
+++ b/src/conf_post.h
@@ -202,13 +202,6 @@ extern void _DebPrint (const char *fmt, ...);
#endif
#endif
-#ifdef emacs /* Don't do this for lib-src. */
-/* Tell regex-emacs.c to use a type compatible with Emacs. */
-#define RE_TRANSLATE_TYPE Lisp_Object
-#define RE_TRANSLATE(TBL, C) char_table_translate (TBL, C)
-#define RE_TRANSLATE_P(TBL) (!EQ (TBL, make_number (0)))
-#endif
-
/* Tell time_rz.c to use Emacs's getter and setter for TZ.
Only Emacs uses time_rz so this is OK. */
#define getenv_TZ emacs_getenv_TZ
diff --git a/src/regex-emacs.h b/src/regex-emacs.h
index cb6dd76..9a6214a 100644
--- a/src/regex-emacs.h
+++ b/src/regex-emacs.h
@@ -219,7 +219,7 @@ extern ptrdiff_t emacs_re_safe_alloca;
((RE_SYNTAX_POSIX_EXTENDED | RE_BACKSLASH_ESCAPE_IN_LISTS | RE_DEBUG)
\
& ~(RE_DOT_NOT_NULL | RE_INTERVALS | RE_CONTEXT_INDEP_OPS))
-#define RE_SYNTAX_POSIX_AWK \
+#define RE_SYNTAX_POSIX_AWK \
(RE_SYNTAX_POSIX_EXTENDED | RE_BACKSLASH_ESCAPE_IN_LISTS \
| RE_INTERVALS | RE_NO_GNU_OPS)
@@ -350,6 +350,11 @@ typedef enum
REG_ESIZEBR /* n or m too big in \{n,m\} */
} reg_errcode_t;
+/* Use a type compatible with Emacs. */
+#define RE_TRANSLATE_TYPE Lisp_Object
+#define RE_TRANSLATE(TBL, C) char_table_translate (TBL, C)
+#define RE_TRANSLATE_P(TBL) (!EQ (TBL, make_number (0)))
+
/* This data structure represents a compiled pattern. Before calling
the pattern compiler, the fields `buffer', `allocated', `fastmap',
`translate', and `no_sub' can be set. After the pattern has been
- [Emacs-diffs] master updated (ba8eb99 -> 9c022a4), Paul Eggert, 2018/08/05
- [Emacs-diffs] master e097826 5/6: Remove always-0 struct re_pattern_buffer members, Paul Eggert, 2018/08/05
- [Emacs-diffs] master e565226 1/6: Rename src/regex.c to src/regex-emacs.c., Paul Eggert, 2018/08/05
- [Emacs-diffs] master 03dfb60 4/6: Simplify regex-emacs by assuming Emacs syntax, Paul Eggert, 2018/08/05
- [Emacs-diffs] master 9c022a4 6/6: Spruce up some regex-emacs comments, Paul Eggert, 2018/08/05
- [Emacs-diffs] master 3a6abe6 3/6: Simplify regex-emacs code by assuming Emacs, Paul Eggert, 2018/08/05
- [Emacs-diffs] master d904cc8 2/6: Use Gnulib regex for lib-src,
Paul Eggert <=