aboutsummaryrefslogtreecommitdiff
diff options
context:
space:
mode:
Diffstat
-rw-r--r--ChangeLog34
-rwxr-xr-xconfigure26
-rw-r--r--libintl.h1
-rw-r--r--locale/Makefile34
-rw-r--r--locale/categories.def4
-rw-r--r--locale/charmap.c524
-rw-r--r--locale/collate.c212
-rw-r--r--locale/config.h45
-rw-r--r--locale/ctype.c817
-rw-r--r--locale/ctypedump.c163
-rw-r--r--locale/hash.c254
-rw-r--r--locale/hash.h50
-rw-r--r--locale/iso-4217.def36
-rw-r--r--locale/keyword.gperf77
-rw-r--r--locale/keyword.h180
-rw-r--r--locale/langinfo.h4
-rw-r--r--locale/libintl.h77
-rw-r--r--locale/loadlocale.c9
-rw-r--r--locale/locale.c538
-rw-r--r--locale/localedef.c261
-rw-r--r--locale/localedef.h196
-rw-r--r--locale/locfile-lex.c533
-rw-r--r--locale/locfile-parse.c820
-rw-r--r--locale/messages.c76
-rw-r--r--locale/monetary.c132
-rw-r--r--locale/numeric.c62
-rw-r--r--locale/token.h54
-rw-r--r--posix/Makefile6
-rw-r--r--posix/regex.c5400
-rw-r--r--posix/regex.h495
-rw-r--r--regex.h1
31 files changed, 11096 insertions, 25 deletions
diff --git a/ChangeLog b/ChangeLog
index 4d55f3fa15..e0440179aa 100644
--- a/ChangeLog
+++ b/ChangeLog
@@ -1,3 +1,37 @@
+Wed May 17 16:50:21 1995 Roland McGrath <roland@churchy.gnu.ai.mit.edu>
+
+ Merged 1003.2 locale and localedef programs by Ulrich Drepper.
+ * locale/charmap.c: New file.
+ * locale/collate.c: New file.
+ * locale/config.h: New file.
+ * locale/ctype.c: New file.
+ * locale/ctypedump.c: New file.
+ * locale/hash.c: New file.
+ * locale/hash.h: New file.
+ * locale/iso-4217.def: New file.
+ * locale/keyword.gperf: New file.
+ * locale/keyword.h: New file.
+ * locale/libintl.h: New file.
+ * locale/locale.c: New file.
+ * locale/localedef.c: New file.
+ * locale/localedef.h: New file.
+ * locale/locfile-lex.c: New file.
+ * locale/locfile-parse.c: New file.
+ * locale/messages.c: New file.
+ * locale/monetary.c: New file.
+ * locale/numeric.c: New file.
+ * locale/token.h: New file.
+
+ * posix/regex.c, posix/regex.h: New files, incorporated from GNU regex.
+ * posix/Makefile (headers): Add regex.h.
+ (routines): Add regex.
+ (gpl2lgpl): Add regex.c and regex.h.
+
+Tue May 16 17:35:07 1995 Roland McGrath <roland@churchy.gnu.ai.mit.edu>
+
+ * locale/loadlocale.c: Expect macro LOCALE_PATH to be defined,
+ instead of hard-coding "/share/locale".
+
Sat May 13 02:16:42 1995 Roland McGrath <roland@churchy.gnu.ai.mit.edu>
* configure.in (os=gnu*): Always set elf=yes, not just for os=gnu*elf*.
diff --git a/configure b/configure
index c543fb56fa..a543befe61 100755
--- a/configure
+++ b/configure
@@ -571,6 +571,19 @@ echo "$ac_t""$host" 1>&4
# $machine, $vendor, and $os, and changes them whenever convenient.
config_machine=$host_cpu config_vendor=$host_vendor config_os=$host_os
+# Some configurations imply other options.
+case "$host_os" in
+gnu* | linux* | bsd4.4* | netbsd* | freebsd*)
+ # These systems always use GNU tools.
+ gnu_ld=yes gnu_as=yes
+esac
+case "$host_os" in
+gnu* | linux* | sysv4* | solaris2*)
+ # These systems always use the ELF format.
+ elf=yes
+esac
+
+# Compute the list of sysdep directories for this configuration.
sysdep_dir=$srcdir/sysdeps
echo $ac_n "checking sysdep dirs""... $ac_c" 1>&4
if eval "test \"`echo '$''{'libc_cv_sysdirs'+set}'`\" = set"; then
@@ -621,19 +634,6 @@ none)
base_os='' ;;
esac
-# Some configurations imply other options.
-case "$os" in
-gnu* | linux* | bsd4.4* | netbsd* | freebsd*)
- # These systems always use GNU tools.
- gnu_ld=yes gnu_as=yes
-esac
-case "$os" in
-gnu* | linux* | sysv4* | solaris2*)
- # These systems always use the ELF format.
- elf=yes
-esac
-
-
# For sunos4.1.1, try sunos4.1.1, then sunos4.1, then sunos4, then sunos.
tail=$os
ostry=$os
diff --git a/libintl.h b/libintl.h
new file mode 100644
index 0000000000..38a33d232b
--- /dev/null
+++ b/libintl.h
@@ -0,0 +1 @@
+#include <locale/libintl.h>
diff --git a/locale/Makefile b/locale/Makefile
index 87a6706e05..c3abc2c62e 100644
--- a/locale/Makefile
+++ b/locale/Makefile
@@ -21,10 +21,36 @@
#
subdir := locale
-headers := locale.h
-distribute := localeinfo.h categories.def
-routines := setlocale loadlocale localeconv
-categories := ctype messages monetary numeric time collate
+headers = locale.h
+distribute = localeinfo.h categories.def \
+ $(localedef-modules:=.c) $(locale-modules:=.c) \
+ $(lib-modules:=.c) config.h hash.h iso-4217.def \
+ keyword.gperf keyword.h localedef.h token.h
+routines = setlocale loadlocale localeconv nl_langinfo
+categories = ctype messages monetary numeric time collate
aux = $(categories:%=lc-%) $(categories:%=C-%)
+others = localedef locale
+install-bin = localedef locale
+extra-objs = $(localedef-modules:=.o) $(locale-modules:=.o) \
+ $(lib-modules:=.o)
+
+localedef-modules := charmap locfile-lex locfile-parse ctype \
+ monetary messages collate numeric
+locale-modules := ctypedump
+lib-modules := error hash xmalloc
+
+
+GPERF = gperf
+GPERFFLAGS = -acCgopt -k1,2,5,$$
include ../Rules
+
+keyword.h: keyword.gperf
+ $(GPERF) $(GPERFFLAGS) $< > $@.new
+ mv -f $@.new $@
+
+$(objpfx)localedef: $(localedef-modules:%=$(objpfx)%.o)
+$(objpfx)locale: $(locale-modules:%=$(objpfx)%.o)
+$(objpfx)localedef $(objpfx)locale: $(lib-modules:%=$(objpfx)%.o)
+
+CPPFLAGS += -DLOCALE_PATH='"$(localedir)"' -DCHARMAP_PATH='"$(nlsdir)/charmap"'
diff --git a/locale/categories.def b/locale/categories.def
index 6bdad35e17..be3c6bc549 100644
--- a/locale/categories.def
+++ b/locale/categories.def
@@ -101,6 +101,10 @@ DEFINE_CATEGORY (LC_TIME, "LC_TIME",
{ D_FMT, "d_fmt", std, string },
{ T_FMT, "t_fmt", std, string },
{ T_FMT_AMPM, "t_fmt_ampm", std, string },
+ { ERA, "era", opt, string },
+ { ERA_YEAR, "era_year", opt, string },
+ { ERA_D_FMT, "era_d_fmt", opt, string },
+ { ALT_DIGITS, "alt_digits", opt, stringarray, 0, 100 },
{ 0 }
), NO_POSTLOAD, NULL, NULL, NULL )
diff --git a/locale/charmap.c b/locale/charmap.c
new file mode 100644
index 0000000000..ad1075e5bc
--- /dev/null
+++ b/locale/charmap.c
@@ -0,0 +1,524 @@
+/* Copyright (C) 1995 Free Software Foundation, Inc.
+
+The GNU C Library is free software; you can redistribute it and/or
+modify it under the terms of the GNU Library General Public License as
+published by the Free Software Foundation; either version 2 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
+Library General Public License for more details.
+
+You should have received a copy of the GNU Library General Public
+License along with the GNU C Library; see the file COPYING.LIB. If
+not, write to the Free Software Foundation, Inc., 675 Mass Ave,
+Cambridge, MA 02139, USA. */
+
+#include <ctype.h>
+#include <errno.h>
+#include <libintl.h>
+#include <limits.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <unistd.h>
+
+#include "localedef.h"
+#include "hash.h"
+
+/* Data structure for representing charmap database. */
+struct charmap charmap_data;
+
+/* Line number in charmap file. */
+static unsigned int line_no;
+
+/* Prototypes for local functions. */
+static void read_prolog (FILE *infile);
+static unsigned long read_body (FILE *infile);
+
+
+/* Read complete table of symbolic names for character set from file. If
+ this file does not exist or is not readable a default file is tried.
+ If this also is not readable no character map is defined. */
+void
+charmap_read (const char *filename)
+{
+ unsigned long max_char;
+ long path_max = pathconf (".", _PC_PATH_MAX);
+ char buf[path_max];
+ FILE *infile = NULL;
+
+ /* Initialize charmap data. */
+ charmap_data.codeset_name = NULL;
+ charmap_data.mb_cur_max = -1;
+ charmap_data.mb_cur_min = -1;
+ charmap_data.escape_char = '\\';
+ charmap_data.comment_char = '#';
+
+ if (filename != NULL)
+ {
+ strcpy (buf, filename);
+ infile = fopen (filename, "r");
+ if (infile == NULL && filename[0] != '/')
+ {
+ snprintf (buf, path_max, "%s/%s", CHARMAP_PATH, filename);
+ infile = fopen (buf, "r");
+ }
+ }
+ if (infile == NULL)
+ {
+ if (filename != NULL)
+ error (0, errno, gettext ("input file `%s' not found"), filename);
+
+ snprintf (buf, path_max, "%s/%s", CHARMAP_PATH, DEFAULT_CHARMAP);
+ infile = fopen (buf, "r");
+
+ if (infile == NULL)
+ error (4, errno, gettext ("input file `%s' not found"), filename);
+ }
+
+ charmap_data.filename = buf;
+ init_hash (&charmap_data.table, 500);
+ line_no = 0;
+
+ /* Read the prolog of the charmap file. */
+ read_prolog (infile);
+
+ /* Last works on the charmap tables global data. */
+ if (charmap_data.mb_cur_max == -1)
+ charmap_data.mb_cur_max = 1;
+ if (charmap_data.mb_cur_min == -1)
+ charmap_data.mb_cur_min = charmap_data.mb_cur_max;
+
+ if ((size_t) charmap_data.mb_cur_max > sizeof (long))
+ {
+ error (2, 0, gettext ("program limitation: for now only upto %Zu "
+ "bytes per character are allowed"), sizeof (long));
+ }
+
+ /* Now process all entries. */
+ max_char = read_body (infile);
+
+ /* We don't need the file anymore. */
+ fclose (infile);
+
+
+ /* Determine the optimal table size when using the simple modulo hashing
+ function. */
+ if (max_char >= 256)
+ {
+ int size;
+ /* Current best values, initialized to some never reached high value. */
+ int best_count = 10000;
+ int best_size = 10000;
+ int best_product = best_count * best_size;
+
+ /* Give warning. */
+ error (-1, 0, gettext ("computing character table size: this may take "
+ "a while"));
+
+ for (size = 256; size <= best_product; ++size)
+ {
+ /* Array with slot counters. */
+ int cnt[size];
+ /* Current character. */
+ int ch;
+ /* Maximal number of characters in any slot. */
+ int maxcnt = 0;
+ /* Product of current size and maximal count. */
+ int product = 0;
+ /* Iteration pointer through hashing table. */
+ char *ptr = NULL;
+
+ /* Initializes counters to zero. */
+ memset(cnt, 0, size * sizeof (int));
+
+ /* Iterate through whole hashing table. */
+ while (product < best_product
+ && iterate_table (&charmap_data.table, (void **) &ptr,
+ (void **) &ch))
+ {
+ /* Increment slot counter. */
+ ++cnt[ch % size];
+ /* Test for current maximum. */
+ if (cnt[ch % size] > maxcnt)
+ {
+ maxcnt = cnt[ch % size];
+ product = maxcnt * size;
+ }
+ }
+
+ if (product < best_product)
+ {
+ best_count = maxcnt;
+ best_size = size;
+ best_product = best_count * best_size;
+ }
+ }
+
+ charmap_data.hash_size = best_size;
+ charmap_data.hash_layers = best_count;
+ }
+ else
+ {
+ charmap_data.hash_size = 256;
+ charmap_data.hash_layers = 1;
+ }
+}
+
+
+#define SYNTAX_ERROR \
+ do { error (0, 0, gettext ("%s:%u: syntax error in charmap file"), \
+ charmap_data.filename, line_no); \
+ goto end_of_loop; } while (0)
+
+/* Read the prolog of the charmap file until the line containing `CHARMAP'.
+ All possible entries are processed. */
+static void
+read_prolog (FILE *infile)
+{
+ size_t bufsize = sysconf (_SC_LINE_MAX);
+ char buf[bufsize];
+
+ while (1)
+ {
+ char *cp = buf;
+ char len;
+
+ /* Read the next line. */
+ fgets (buf, bufsize, infile);
+ len = strlen (buf);
+
+ /* On EOF simply return. */
+ if (len == 0 || buf[len - 1] != '\n')
+ error (4, 0, gettext ("%s: unexpected end of file in charmap"),
+ charmap_data.filename);
+
+ /* This is the next line. */
+ ++line_no;
+
+ /* Comments and empty lines are ignored. */
+ if (len == 1 || buf[0] == charmap_data.comment_char)
+ continue;
+
+ buf[len - 1] = '\0';
+
+ /* Throw away leading white spaces. This is not defined in POSIX.2
+ so don't do it if conformance is requested. */
+ if (!posix_conformance)
+ while (isspace (*cp))
+ ++cp;
+
+ /* If `CHARMAP' is read the prolog is over. */
+ if (strncmp (cp, "CHARMAP", 7) == 0
+ && (!posix_conformance || cp[7] == '\0'))
+ return;
+
+ /* Now it can be only one of special symbols defining the charmap
+ parameters. All are beginning with '<'. */
+ if (*cp != '<')
+ SYNTAX_ERROR;
+
+ ++cp;
+ if (strncmp (cp, "code_set_name>", 14) == 0)
+ {
+ char *startp;
+
+#define cp_to_arg(no,pred) \
+ cp += no; \
+ while (isspace (*cp)) \
+ ++cp; \
+ if (*cp == '\0' || !pred (*cp)) \
+ SYNTAX_ERROR;
+
+ cp_to_arg (14,isgraph)
+
+ if (charmap_data.codeset_name != NULL)
+ {
+ error (0, 0, gettext ("%s:%u: duplicate code set name "
+ "specification"),
+ charmap_data.filename, line_no);
+ free (charmap_data.codeset_name);
+ }
+
+ startp = cp;
+ while (*cp != '\0' && isgraph (*cp) && !isspace (*cp))
+ ++cp;
+
+ charmap_data.codeset_name = (char *) xmalloc (cp - startp + 1);
+ strncpy (startp, startp, cp - startp);
+ }
+ else if (strncmp (cp, "mb_cur_max>", 11) == 0)
+ {
+ int new_val;
+ cp_to_arg (11,isdigit)
+
+ if (charmap_data.mb_cur_max != -1)
+ error (0, 0,
+ gettext ("%s:%u: duplicate definition of mb_cur_max"),
+ charmap_data.filename, line_no);
+
+ new_val = (int) strtol (cp, &cp, posix_conformance ? 10 : 0);
+ if (new_val < 1)
+ error (0, 0, gettext ("%s:%u: illegal value for mb_cur_max: %d"),
+ charmap_data.filename, line_no, new_val);
+ else
+ charmap_data.mb_cur_max = new_val;
+ }
+ else if (strncmp (cp, "mb_cur_min>", 11) == 0)
+ {
+ int new_val;
+ cp_to_arg (11,isdigit)
+
+ if (charmap_data.mb_cur_max != -1)
+ error (0, 0,
+ gettext ("%s:%u: duplicate definition of mb_cur_min"),
+ charmap_data.filename, line_no);
+
+ new_val = (int) strtol (cp, &cp, posix_conformance ? 10 : 0);
+ if (new_val < 1)
+ error (0, 0, gettext ("%s:%u: illegal value for mb_cur_min: %d"),
+ charmap_data.filename, line_no, new_val);
+ else
+ charmap_data.mb_cur_min = new_val;
+ }
+ else if (strncmp (cp, "escape_char>", 12) == 0)
+ {
+ cp_to_arg (12, isgraph)
+ charmap_data.escape_char = *cp;
+ }
+ else if (strncmp (cp, "comment_char>", 13) == 0)
+ {
+ cp_to_arg (13, isgraph)
+ charmap_data.comment_char = *cp;
+ }
+ else
+ SYNTAX_ERROR;
+ end_of_loop:
+ }
+}
+#undef cp_to_arg
+
+
+static unsigned long
+read_body (FILE *infile)
+{
+ unsigned long max_char = 0;
+ size_t bufsize = sysconf (_SC_LINE_MAX);
+ char buf[bufsize];
+ char name_str[bufsize / 2];
+ char code_str[bufsize / 2];
+
+ while (1)
+ {
+ char *cp = buf;
+ size_t len;
+
+ /* Read the next line. */
+ fgets (buf, bufsize, infile);
+ len = strlen (buf);
+
+ /* On EOF simply return. */
+ if (len == 0)
+ error (0, 0, gettext ("%s: `END CHARMAP' is missing"),
+ charmap_data.filename);
+
+ /* This is the next line. */
+ ++line_no;
+
+ if (len == bufsize - 1)
+ {
+ error (0, 0, gettext ("%s:%u: line too long; use `getconf "
+ "LINE_MAX' to get the current maximum line"
+ "length"), charmap_data.filename, line_no);
+ do
+ {
+ fgets (buf, bufsize, infile);
+ len = strlen (buf);
+ }
+ while (len == bufsize - 1);
+ continue;
+ }
+
+ /* Comments and empty lines are ignored. */
+ if (len == 1 || buf[0] == charmap_data.comment_char)
+ continue;
+
+ buf[len - 1] = '\0';
+
+ /* Throw away leading white spaces. This is not defined in POSIX.2
+ so don't do it if conformance is requested. */
+ if (!posix_conformance)
+ while (isspace (*cp))
+ ++cp;
+
+ if (*cp == '<')
+ {
+ char *end1p, *end2p, *start2p;
+ size_t cnt = 0;
+ unsigned long char_value = 0;
+
+ if (sscanf (cp + 1, "%s %s", name_str, code_str) != 2)
+ SYNTAX_ERROR;
+
+ end1p = cp = name_str;
+ while (*cp != '\0' && *cp != '>')
+ {
+ if (*cp == charmap_data.escape_char)
+ if (*++cp == '\0')
+ SYNTAX_ERROR;
+ *end1p++ = *cp++;
+ }
+ if (*cp == '\0')
+ /* No final '>'. Make error condition. */
+ end1p = name_str;
+ else
+ ++cp;
+
+ *end1p = '\0';
+
+ if (*cp == '.' && *++cp == '.' && *++cp == '.' && *++cp == '<')
+ {
+ /* This might be the alternate form. */
+ start2p = end2p = ++cp;
+ while (*cp != '\0' && *cp != '>')
+ {
+ if (*cp == charmap_data.escape_char)
+ if (*++cp == '\0')
+ SYNTAX_ERROR;
+ *end2p = *cp++;
+ }
+ if (*cp == '\0')
+ /* NO final '>'. Make error condition. */
+ end2p = start2p;
+ else
+ ++cp;
+ }
+ else
+ start2p = end2p = NULL;
+
+
+ if (end1p == name_str || (start2p != NULL && start2p != end2p)
+ || *cp != '\0'
+ || *code_str != charmap_data.escape_char)
+ SYNTAX_ERROR;
+
+ cp = code_str;
+ do
+ {
+ char *begin;
+ long val;
+
+ switch (*++cp)
+ {
+ case 'd':
+ val = strtol ((begin = cp + 1), &cp, 10);
+ break;
+ case 'x':
+ val = strtol ((begin = cp + 1), &cp, 16);
+ break;
+ default:
+ val = strtol ((begin = cp), &cp, 8);
+ break;
+ }
+ if (begin == cp)
+ SYNTAX_ERROR;
+
+ if (posix_conformance && cp - begin < 2)
+ error (0, 0, gettext ("%s:%u: byte constant has less than "
+ "two digits"),
+ charmap_data.filename, line_no);
+
+ if (val < 0 || val > 255)
+ {
+ error (0, 0, gettext ("%s:%u: character encoding must be "
+ "given in 8-bit bytes"),
+ charmap_data.filename, line_no);
+ goto end_of_loop;
+ }
+
+ if (cnt < (size_t) charmap_data.mb_cur_max)
+ {
+ if (cnt < sizeof (long)) /* FIXME */
+ char_value = (char_value << 8) | val;
+ }
+ else
+ {
+ error (0, 0, gettext ("%s:%u: number of bytes in character "
+ "definition exceeds `mb_cur_max'"),
+ charmap_data.filename, line_no);
+ break;
+ }
+ ++cnt;
+ }
+ while (*cp == charmap_data.escape_char);
+
+ /* Ignore the rest of the line (comment). */
+ if (end2p == NULL)
+ {
+ if (insert_entry (&charmap_data.table, name_str,
+ end1p - name_str, (void *) char_value))
+ error (0, 0, gettext ("%s:%u: duplicate entry"),
+ charmap_data.filename, line_no);
+
+ max_char = MAX (max_char, char_value);
+ }
+ else
+ {
+ char *en1, *en2, *start1p;
+ long n1, n2, n;
+
+ start1p = name_str;
+
+ while (*start1p == *start2p && !isdigit (*start1p)
+ && start1p < end1p)
+ ++start1p, ++start2p;
+
+ n1 = strtol (start1p, &en1, 10);
+ n2 = strtol (start2p, &en2, 10);
+
+ if (en1 - start1p != en2 - start2p || en1 != end1p
+ || en2 != end2p)
+ SYNTAX_ERROR;
+
+ if (n1 > n2)
+ error (0, 0, gettext ("%s:%u: starting character is bigger "
+ "than last"),
+ charmap_data.filename, line_no);
+
+ n = n1;
+ while (n <= n2)
+ {
+ snprintf(start1p, en1 - start1p, "%0*d", en1 - start1p, n);
+
+ if (insert_entry (&charmap_data.table, name_str,
+ en1 - name_str,
+ (void *) (char_value + n - n1)))
+ error (0, 0, gettext ("%s:%u: duplicate entry"),
+ charmap_data.filename, line_no);
+
+ max_char = MAX (max_char, char_value + n - n1);
+ ++n;
+ }
+ }
+ }
+ else
+ {
+ if (strncmp (cp, "END CHARMAP", 11) == 0)
+ return max_char;
+
+ SYNTAX_ERROR;
+ }
+ end_of_loop:
+ }
+
+ return max_char;
+}
+
+/*
+ * Local Variables:
+ * mode:c
+ * c-basic-offset:2
+ * End:
+ */
diff --git a/locale/collate.c b/locale/collate.c
new file mode 100644
index 0000000000..f06964efcb
--- /dev/null
+++ b/locale/collate.c
@@ -0,0 +1,212 @@
+/* Copyright (C) 1995 Free Software Foundation, Inc.
+
+The GNU C Library is free software; you can redistribute it and/or
+modify it under the terms of the GNU Library General Public License as
+published by the Free Software Foundation; either version 2 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
+Library General Public License for more details.
+
+You should have received a copy of the GNU Library General Public
+License along with the GNU C Library; see the file COPYING.LIB. If
+not, write to the Free Software Foundation, Inc., 675 Mass Ave,
+Cambridge, MA 02139, USA. */
+
+#include <langinfo.h>
+#include <libintl.h>
+
+#include "localedef.h"
+#include "token.h"
+
+
+/* defined in locfile-lex.c: flag to indicate that unknown element names
+ are allowed. */
+extern int reject_new_char;
+
+
+#define SYNTAX_ERROR \
+ error (0, 0, gettext ("%s:%Zd: syntax error in locale definition file"), \
+ locfile_data.filename, locfile_data.line_no);
+
+void
+collate_input (int token)
+{
+ int read_order_start = 0;
+
+ while (1)
+ {
+ char *ptr;
+ int len;
+
+ if (token == TOK_END)
+ /* This is the end of the category. */
+ {
+ token = xlocfile_lex (&ptr, &len);
+
+ if (token != _NL_NUM_LC_COLLATE)
+ {
+ error (0, 0, gettext ("%s:%Zd: category `%s' does not end "
+ "with `END %s'"), locfile_data.filename,
+ locfile_data.line_no, "LC_COLLATE", "LC_COLLATE");
+ ignore_to_eol (0, 0);
+ }
+ else
+ ignore_to_eol (0, 1);
+
+ /* Start next category. */
+ break;
+ }
+
+#if 0
+ /* Process line. */
+ if (read_order_start == 0)
+ /* We're still in the preambel. */
+ {
+ switch (token)
+ {
+ case TOK_COLLATING_ELEMENT:
+ reject_new_char = 0;
+ token = xlocfile_lex (&ptr, &len);
+ reject_new_char = 1;
+ if (token == TOK_CHAR)
+ {
+ error (0, 0, gettext ("%s:%Zd: symbolic name must not be "
+ "duplicate name in charmap"),
+ locfile_data.filename, locfile_data.line_no);
+ ignore_to_eol (0, 0);
+ break;
+ }
+ else if (token != TOK_ILL_CHAR)
+ {
+ SYNTAX_ERROR;
+ ignore_to_eol (0, 0);
+ break;
+ }
+ else
+ {
+ char elem_name[len + 1];
+ memcpy (elem_name, ptr, len);
+ elem_name[len] = '\0';
+
+ /* Test whether defined in symbol table. */
+
+ token = xlocfile_lex (&ptr, &len);
+ if (token != TOK_FROM)
+ {
+ SYNTAX_ERROR;
+ ignore_to_eol (0, 0);
+ break;
+ }
+
+ token = xlocfile_lex (&ptr, &len);
+ if (token != TOK_STRING)
+ {
+ SYNTAX_ERROR;
+ ignore_to_eol (0, 0);
+ break;
+ }
+
+ /* Insert collating element into table. */
+
+ /* Rest of the line should be empty. */
+ ignore_to_eol (0, 1);
+ }
+ break;
+ case TOK_COLLATING_SYMBOL:
+ reject_new_char = 0;
+ token = xlocfile_lex (&ptr, &len);
+ reject_new_char = 1;
+ if (token == TOK_CHAR)
+ {
+ error (0, 0, gettext ("%s:%Zd: symbolic name must not "
+ "duplicate name in charmap"),
+ locfile_data.filename, locfile_data.line_no);
+ ignore_to_eol (0, 0);
+ break;
+ }
+ else if (token != TOK_ILL_CHAR)
+ {
+ SYNTAX_ERROR;
+ ignore_to_eol (0, 0);
+ break;
+ }
+ else
+ {
+ /* Test whether defined in element table. */
+
+ /* Insert collating symbol into table. */
+
+ ignore_to_eol (0, 1);
+ }
+ case TOK_ORDER_START:
+ nsort_rules = 0;
+
+ do
+ {
+ token = xlocfile_lex (&ptr, &len);
+
+ if (nsort_rules == 0 && token == ENDOFLINE)
+ break;
+
+ if (token != TOK_BACKWARD && token != TOK_FORWARD
+ && token != TOK_POSITION)
+ {
+ SYNTAX_ERROR;
+ break;
+ }
+
+ switch (token)
+ {
+ case TOK_BACKWARD:
+ if ((sort_rule[nsort_rules] & FORWARD_BIT) != 0)
+ error (0, 0, gettext ("%s:%Zd: directives `forward' "
+ "and `backward' are mutually "
+ "exclusive"),
+ locfile_data.filename, locfile_data.lineno);
+ else
+ sort_rule[nsort_rules] |= BACKWARD_BIT;
+ break;
+ case TOK_FORWARD:
+ if ((sort_rule[nsort_rules] & BACKWARD_BIT) != 0)
+ error (0, 0, gettext ("%s:%Zd: directives `forward' "
+ "and `backward' are mutually "
+ "exclusive"),
+ locfile_data.filename, locfile_data.lineno);
+ else
+ sort_rule[nsort_rules] |= FORWARD_BIT;
+ break;
+ case TOK_POSITION:
+ sort_rule[nsort_rules] |= POSITION_BIT;
+ break;
+ }
+
+ ++nsort_rules;
+
+
+ }
+ break;
+ default:
+ SYNTAX_ERROR;
+ ignore_to_eol (token, 0);
+ }
+ }
+ else
+ {
+ }
+#endif
+
+ ignore_to_eol(token,0);
+ /* Get next token. */
+ token = xlocfile_lex (&ptr, &len);
+ }
+}
+
+/*
+ * Local Variables:
+ * mode:c
+ * c-basic-offset:2
+ * End:
+ */
diff --git a/locale/config.h b/locale/config.h
new file mode 100644
index 0000000000..65dfb33e9a
--- /dev/null
+++ b/locale/config.h
@@ -0,0 +1,45 @@
+/* config.h used by locale and localedef programs in GNU libc.
+Copyright (C) 1995 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 Library General Public License as
+published by the Free Software Foundation; either version 2 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
+Library General Public License for more details.
+
+You should have received a copy of the GNU Library General Public
+License along with the GNU C Library; see the file COPYING.LIB. If
+not, write to the Free Software Foundation, Inc., 675 Mass Ave,
+Cambridge, MA 02139, USA. */
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+
+#define PACKAGE "libc"
+#define VERSION __libc_version
+extern const char __libc_version[];
+
+#define DEFAULT_CHARMAP "POSIX"
+
+
+/* These are tested by xmalloc.c and error.c. */
+#define HAVE_VPRINTF 1
+#define STDC_HEADERS 1
+#define HAVE_STRERROR 1
+
+#define program_name program_invocation_name
+
+typedef unsigned short u16;
+typedef int i32;
+typedef int u32;
+
+
+/* Get the global libc configuration info. */
+#include_next <config.h>
+
+#endif /* config.h */
diff --git a/locale/ctype.c b/locale/ctype.c
new file mode 100644
index 0000000000..1ce09bab2b
--- /dev/null
+++ b/locale/ctype.c
@@ -0,0 +1,817 @@
+/* Copyright (C) 1995 Free Software Foundation, Inc.
+
+The GNU C Library is free software; you can redistribute it and/or
+modify it under the terms of the GNU Library General Public License as
+published by the Free Software Foundation; either version 2 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
+Library General Public License for more details.
+
+You should have received a copy of the GNU Library General Public
+License along with the GNU C Library; see the file COPYING.LIB. If
+not, write to the Free Software Foundation, Inc., 675 Mass Ave,
+Cambridge, MA 02139, USA. */
+
+#include <alloca.h>
+#include <fcntl.h>
+#include <libintl.h>
+#include <locale.h>
+#include <localeinfo.h>
+#include <langinfo.h>
+#include <string.h>
+#include <unistd.h>
+#include <netinet/in.h>
+#include <sys/uio.h>
+
+#include "localedef.h"
+#include "token.h"
+
+/* Arrays representing ctype tables. They must be initialized for the
+ right size to hold the full charmap. */
+static u16 *ctype_b;
+static i32 *names_b, *toupper_b, *tolower_b;
+
+/* For accessing the element of the (possibly sparse) array we use this
+ macro. */
+#define ELEM(arr, idx) \
+ (arr)[({ int h = idx % charmap_data.hash_size; \
+ int n = 0; \
+ while (n < charmap_data.hash_layers \
+ && names_b[n * charmap_data.hash_size + h] != idx) \
+ ++n; \
+ if (n >= charmap_data.hash_layers) \
+ error (6, 0, gettext ("internal error in %s, line %u"), \
+ __FUNCTION__, __LINE__); \
+ n * charmap_data.hash_size + h; })]
+
+/* The bit used for representing a special class. */
+#define BITPOS(class) ((class) - TOK_UPPER)
+#define BIT(class) (1 << BITPOS (class))
+
+/* Remember which class or conversion is already done. */
+static unsigned short class_done = 0;
+static unsigned short toupper_done = 0;
+static unsigned short tolower_done = 0;
+
+#define SYNTAX_ERROR \
+ error (0, 0, gettext ("%s:%Zd: syntax error in locale definition file"), \
+ locfile_data.filename, locfile_data.line_no);
+
+
+/* Prototypes for local functions. */
+static void allocate_arrays (void);
+static void set_class_defaults (void);
+static int valid_char (int ch);
+
+
+/* Read CTYPE category. The initial token is given as a parameter. */
+void
+ctype_input (int token)
+{
+ char *ptr;
+ int len;
+
+ /* If necessary allocate arrays. */
+ allocate_arrays ();
+
+ while (token != TOK_END)
+ {
+ switch (token)
+ {
+ case TOK_UPPER: case TOK_LOWER: case TOK_ALPHA: case TOK_DIGIT:
+ case TOK_XDIGIT: case TOK_SPACE: case TOK_PRINT: case TOK_GRAPH:
+ case TOK_BLANK: case TOK_CNTRL: case TOK_PUNCT:
+ {
+ /* TAKE CARE: the order of the tokens in "token.h" determines
+ the bit used to indicate the membership in the class. This
+ also has to correspond to the values used in <ctype.h>. */
+ int bit = BIT (token);
+ int was_ell = 0;
+ int last = -1;
+
+ if ((class_done & bit) != 0)
+ {
+ char tmp[len + 1];
+ memcpy (tmp, ptr, len);
+ tmp[len] = '\0';
+
+ error (0, 0, gettext ("%s:%Zd: duplicate definiton of item "
+ "`%s' in category `LC_CTYPE'"),
+ locfile_data.filename, locfile_data.line_no, tmp);
+ }
+ class_done |= bit;
+
+ do
+ {
+ token = xlocfile_lex (&ptr, &len);
+
+ if (token == TOK_ENDOFLINE)
+ {
+ SYNTAX_ERROR;
+ break;
+ }
+
+ if (token == TOK_ELLIPSIS)
+ {
+ if (was_ell != 0 || last < 0)
+ {
+ error (0, 0, gettext ("%s:%Zd: illegal use of `...'"),
+ locfile_data.filename, locfile_data.line_no);
+ break;
+ }
+ was_ell = 1;
+ continue;
+ }
+
+ if (token != TOK_CHAR)
+ {
+ if (token != TOK_ILL_CHAR)
+ SYNTAX_ERROR;
+ was_ell = 0;
+ last = -1;
+ continue;
+ }
+
+ if (len < 0 || !valid_char (len))
+ {
+ was_ell = 0;
+ last = -1;
+ continue;
+ }
+
+ /* We have found a valid character. Include it to
+ the class' bit set. */
+ if (was_ell == 0)
+ {
+ ELEM (ctype_b, len) |= bit;
+ last = len;
+ }
+ else
+ {
+ int i;
+
+ if (last > len)
+ {
+ error (0, 0, gettext ("%s:%Zd: lower bound of "
+ "ellipsis not smaller"),
+ locfile_data.filename, locfile_data.line_no);
+ was_ell = 0;
+ last = -1;
+ continue;
+ }
+
+ for (i = last + 1; i <= len; ++i)
+ ELEM (ctype_b, i) |= bit;
+
+ last = -1;
+ }
+ was_ell = 0;
+ }
+ while ((token = locfile_lex (&ptr, &len)) == TOK_CHAR
+ && len == ';');
+
+ /* Rest of the line should be empty. */
+ ignore_to_eol (token, 0);
+ }
+ break;
+ case TOK_TOUPPER: case TOK_TOLOWER:
+ {
+ int from;
+ int to = -1;
+ int is_upper = token == TOK_TOUPPER;
+
+ if (((is_upper ? toupper_done : tolower_done) & BIT (token)) != 0)
+ error (0, 0, gettext ("%s:%Zd: duplicate definition of item "
+ "`%s' in category `LC_CTYPE'"),
+ locfile_data.filename, locfile_data.line_no,
+ is_upper ? "toupper" : "tolower");
+ (is_upper ? toupper_done : tolower_done) |= BIT (token);
+
+ do
+ {
+ int ignore;
+
+ token = xlocfile_lex (&ptr, &len);
+ if (token != TOK_CHAR || len != '(')
+ {
+ SYNTAX_ERROR;
+ break;
+ }
+
+ token = xlocfile_lex (&ptr, &len);
+ if (token != TOK_CHAR && token != TOK_ILL_CHAR)
+ {
+ SYNTAX_ERROR;
+ break;
+ }
+ from = len;
+ ignore = token == TOK_ILL_CHAR;
+
+ token = xlocfile_lex (&ptr, &len);
+ if (token != TOK_CHAR || len != ',')
+ {
+ SYNTAX_ERROR;
+ break;
+ }
+
+ token = xlocfile_lex (&ptr, &len);
+ if (token != TOK_CHAR && token != TOK_ILL_CHAR)
+ {
+ SYNTAX_ERROR;
+ break;
+ }
+ to = len;
+ ignore |= token == TOK_ILL_CHAR;
+
+ token = xlocfile_lex (&ptr, &len);
+ if (token != TOK_CHAR || len != ')')
+ {
+ SYNTAX_ERROR;
+ break;
+ }
+
+ if (!ignore && valid_char (from) && valid_char (to))
+ /* Have a valid pair. */
+ ELEM (is_upper ? toupper_b : tolower_b, from) = to;
+ }
+ while ((token = locfile_lex (&ptr, &len)) == TOK_CHAR
+ && len == ';');
+
+ /* Rest of the line should be empty. */
+ ignore_to_eol (token, 1);
+ }
+ break;
+ default:
+ SYNTAX_ERROR;
+ ignore_to_eol (0, 0);
+ break;
+ }
+
+ /* Read next token. */
+ token = xlocfile_lex (&ptr, &len);
+ }
+
+ token = xlocfile_lex (&ptr, &len);
+
+ if (token != _NL_NUM_LC_CTYPE)
+ {
+ error (0, 0, gettext ("%s:%Zd: category `%s' does not end with "
+ "`END %s'"), locfile_data.filename,
+ locfile_data.line_no, "LC_CTYPE", "LC_CTYPE");
+ ignore_to_eol (0, 0);
+ }
+ else
+ ignore_to_eol (0, posix_conformance);
+}
+
+
+void
+ctype_check(void)
+{
+ /* Here are a lot of things to check. See POSIX.2, table 2-6. */
+ #define NCLASS 11
+ static const struct
+ {
+ const char *name;
+ const char allow[NCLASS];
+ }
+ valid_table[NCLASS] =
+ {
+ /* The order is important. See token.h for more information.
+ M = Always, D = Default, - = Permitted, X = Mutually exclusive */
+ [BITPOS (TOK_UPPER)] = { "upper", "--MX-XDDXXX" },
+ [BITPOS (TOK_LOWER)] = { "lower", "--MX-XDDXXX" },
+ [BITPOS (TOK_ALPHA)] = { "alpha", "---X-XDDXXX" },
+ [BITPOS (TOK_DIGIT)] = { "digit", "XXX--XDDXXX" },
+ [BITPOS (TOK_XDIGIT)] = { "xdigit", "-----XDDXXX" },
+ [BITPOS (TOK_SPACE)] = { "space", "XXXXX------" },
+ [BITPOS (TOK_PRINT)] = { "print", "---------X-" },
+ [BITPOS (TOK_GRAPH)] = { "graph", "---------X-" },
+ [BITPOS (TOK_BLANK)] = { "blank", "XXXXXM-----" },
+ [BITPOS (TOK_CNTRL)] = { "cntrl", "XXXXX-XX--X" },
+ [BITPOS (TOK_PUNCT)] = { "punct", "XXXXX-DD-X-" }
+ };
+ int ch, cls1, cls2, eq, space_char;
+ u16 tmp;
+
+ /* Set default value for classes not specified. */
+ set_class_defaults ();
+
+ /* Check according to table. */
+ for (ch = 0; ch < charmap_data.hash_size * charmap_data.hash_layers; ++ch)
+ {
+ if (ch != 0 && names_b[ch] == 0)
+ continue;
+ tmp = ELEM (ctype_b, names_b[ch]);
+ for (cls1 = 0; cls1 < NCLASS; ++cls1)
+ if ((tmp & (1 << cls1)) != 0)
+ for (cls2 = 0; cls2 < NCLASS; ++cls2)
+ if (cls2 != cls1 && valid_table[cls1].allow[cls2] != '-')
+ {
+ eq = (tmp & (1 << cls2)) != 0;
+ switch (valid_table[cls1].allow[cls2])
+ {
+ case 'M':
+ if (!eq)
+ error (0, 0, gettext ("character '\\%o' in class `%s' "
+ "must be in class `%s'"), ch,
+ valid_table[cls1].name, valid_table[cls2].name);
+ break;
+ case 'X':
+ if (eq)
+ error (0, 0, gettext ("character '\\%o' inc class `%s' "
+ "must not be in class `%s'"), ch,
+ valid_table[cls1].name, valid_table[cls2].name);
+ break;
+ case 'D':
+ ELEM (ctype_b, names_b[ch]) |= 1 << cls2;
+ break;
+ default:
+ error (5, 0, gettext ("internal error in %s, line %u"),
+ __FUNCTION__, __LINE__);
+ }
+ }
+ }
+
+ /* ... and now test <SP> as a special case. */
+ if (find_entry (&charmap_data.table, "SP", 2, (void **) &space_char) == 0)
+ error (0, 0, gettext ("character <SP> not defined in character map"));
+ else if ((tmp = BITPOS (TOK_SPACE),
+ (ELEM (ctype_b, space_char) & BIT (TOK_SPACE)) == 0)
+ || (tmp = BITPOS (TOK_BLANK),
+ (ELEM (ctype_b, space_char) & BIT (TOK_BLANK)) == 0))
+ error (0, 0, gettext ("<SP> character not in class `%s'"),
+ valid_table[tmp].name);
+ else if ((tmp = BITPOS (TOK_PUNCT),
+ (ELEM (ctype_b, space_char) & BIT (TOK_PUNCT)) != 0)
+ || (tmp = BITPOS (TOK_GRAPH),
+ (ELEM (ctype_b, space_char) & BIT (TOK_GRAPH)) != 0))
+ error (0, 0, gettext ("<SP> character must not be in class `%s'"),
+ valid_table[tmp].name);
+ else
+ ELEM (ctype_b, space_char) |= BIT (TOK_PRINT);
+}
+
+
+/* These macros can change little to big endian and vice versa. */
+#define SWAP16(v) \
+ ((u16) (((((unsigned short) (v)) & 0x00ff) << 8) \
+ | ((((unsigned short) (v)) & 0xff00) >> 8)))
+#define SWAP32(v) \
+ ((u32) (((((u32) (v)) & 0x000000ff) << 24) \
+ | ((((u32) (v)) & 0x0000ff00) << 8) \
+ | ((((u32) (v)) & 0x00ff0000) >> 8) \
+ | ((((u32) (v)) & 0xff000000) >> 24)))
+
+
+int
+ctype_output (void)
+{
+ char *path, *t;
+ int ch;
+ /* File descriptor for output file. */
+ int fd;
+ /* Magic number. */
+ i32 magic = LIMAGIC (LC_CTYPE);
+ /* Number of table. */
+ int tables = 6;
+ /* Number ints in leading information table. */
+#if 0
+ i32 n = 2 + 2 * tables;
+#else
+ i32 n = 5;
+#endif
+ /* Values describing the character set. */
+ char mb_cur_min = (char) charmap_data.mb_cur_min;
+ char mb_cur_max = (char) charmap_data.mb_cur_max;
+ /* Optimal size of hashing table. */
+ i32 hash_size = charmap_data.hash_size;
+ i32 hash_layers = charmap_data.hash_layers;
+ /* Number of elements in the tables. */
+ int size = hash_size * charmap_data.hash_layers;
+ /* Positions of the tables. */
+ i32 pos[14] =
+ {
+ /* No, no. We don't play towers of Hanoi. This is a more or less
+ readable table of the offsets of the different strings in the
+ produced file. It is seperated in three columns which represent
+ the number of values with 1, 2, and 4 bytes. */
+
+#if 0
+ 4 * (2 + n),
+ 1 + 4 * (2 + n),
+ 2 + 4 * (2 + n),
+ 2 + 4 * (3 + n),
+ 2 + 4 * (4 + n),
+ 2 + 2 * (128 + size) + 4 * (4 + n),
+ 2 + 2 * (128 + size) + 4 * ((4 + n) + (size + 128)),
+ 2 + 2 * (128 + size) + 4 * ((4 + n) + 2 * (size + 128)),
+ 2 + 2 * (128 + size) + 4 * ((4 + n) + 2 * (size + 128) + 1 * size),
+ 2 + 2 * (128 + size) + 4 * ((5 + n) + 2 * (size + 128) + 1 * size),
+ 2 + 2 * (128 + size) + 4 * ((6 + n) + 2 * (size + 128) + 1 * size),
+ 2 + 2 * (2 * (128 + size)) + 4 * ((6 + n) + 2 * (size + 128) + 1 * size),
+ 2 + 2 * (2 * (128 + size)) + 4 * ((6 + n) + 3 * (size + 128) + 1 * size),
+ 2 + 2 * (2 * (128 + size)) + 4 * ((6 + n) + 4 * (size + 128) + 1 * size),
+#else
+ 4 * (2 + n),
+ 2 * (128 + size) + 4 * (2 + n),
+ 2 * (128 + size) + 4 * ((2 + n) + (size + 128)),
+ 2 * (128 + size) + 4 * ((2 + n) + 2 * (size + 128)),
+ 2 * (128 + size) + 4 * ((2 + n) + 3 * (size + 128)),
+#endif
+ };
+ /* Parameter to writev. */
+ struct iovec iov[11] =
+ {
+ { &magic, sizeof (i32) },
+ { &n, sizeof (i32) },
+#if 0
+ { pos, sizeof (pos) },
+ { &mb_cur_min, 1 },
+ { &mb_cur_max, 1 },
+ { &hash_size, sizeof (i32) },
+ { &hash_layers, sizeof (i32) },
+#else
+ { pos, 5 * 4 },
+#endif
+ { ctype_b - 128, (size + 128) * sizeof (u16) },
+ { toupper_b - 128, (size + 128) * sizeof (i32) },
+ { tolower_b - 128, (size + 128) * sizeof (i32) },
+ { names_b, size * sizeof (i32) }
+ };
+ int result = 0;
+
+ /* Now we can bring the representations into the right form. */
+ for (ch = -128; ch < -1; ++ch)
+ {
+ ctype_b[ch] = ctype_b[256 + ch];
+ toupper_b[ch] = toupper_b[256 + ch];
+ tolower_b[ch] = tolower_b[256 + ch];
+ }
+ /* Set value for EOF. */
+ ctype_b[-1] = 0;
+ toupper_b[-1] = -1;
+ tolower_b[-1] = -1;
+
+ for (ch = -128; ch < size; ++ch)
+ ctype_b[ch] = htons (ctype_b[ch]);
+
+ /* Construct the output filename from the argument given to
+ localedef on the command line. */
+ path = (char *) alloca (strlen (output_path) +
+ strlen (category[LC_CTYPE].name) + 1);
+ t = stpcpy (path, output_path);
+ strcpy (t, category[LC_CTYPE].name);
+
+ fd = creat (path, 0666);
+ if (fd == -1)
+ {
+ error (0, 0, gettext ("cannot open output file `%s': %m"), path);
+ result = 1;
+ }
+ else
+ {
+ int idx;
+
+#if 0
+ if (writev (fd, iov, 10) == -1)
+#else
+ if (writev (fd, iov, 6) == -1)
+#endif
+ {
+ error (0, 0, gettext ("cannot write output file `%s': %m"), path);
+ result = 1;
+ goto close_and_return;
+ }
+
+ /* Now we have to write the three tables with different endianess. */
+ hash_size = SWAP32 (hash_size);
+ for (idx = -128; idx < size; ++idx)
+ {
+ ctype_b[idx] = SWAP16 (ctype_b[idx]);
+ toupper_b[idx] = SWAP32 (toupper_b[idx]);
+ tolower_b[idx] = SWAP32 (tolower_b[idx]);
+ if (idx >= 0)
+ names_b[idx] = SWAP32 (names_b[idx]);
+ }
+
+#if 0
+ if (writev (fd, iov + 5, 6) == -1)
+#else
+ if (writev (fd, iov + 3, 2) == -1)
+#endif
+ {
+ error (0, 0, gettext ("cannot write output file `%s': %m"), path);
+ result = 1;
+ }
+
+ close_and_return:
+ close (fd);
+ }
+
+ return result;
+}
+
+
+/* If necessary allocate the memory for the arrays according to the
+ current character map. */
+static void
+allocate_arrays (void)
+{
+ /* Init ctype data structures. */
+ if (ctype_b == NULL)
+ /* All data structures are not initialized yet. */
+ {
+ /* You wonder about this amount of memory? This is only because
+ some users do not manage to address the array with unsigned
+ values or data types with range >= 256. '\200' would result
+ in the array index -128. To help these poor people we
+ duplicate the entries for 128 upto 255 below the entry for \0. */
+ int ch, h, n;
+ char *ptr;
+ int size = charmap_data.hash_size * charmap_data.hash_layers;
+
+ ctype_b = (u16 *) xcalloc (size - (-128), sizeof (u16));
+ ctype_b += 128;
+
+
+ names_b = (i32 *) xcalloc (size, sizeof (i32));
+
+ toupper_b = (i32 *) xcalloc ((size - (-128)), sizeof (i32));
+ toupper_b += 128;
+
+ tolower_b = (i32 *) xcalloc ((size - (-128)), sizeof (i32));
+ tolower_b += 128;
+
+ ptr = NULL;
+ /* Mark the place of the NUL character as occupied. */
+ names_b[0] = 1;
+
+ while (iterate_table (&charmap_data.table, (void **) &ptr,
+ (void **) &ch))
+ {
+ /* We already handled the NUL character. */
+ if (ch == 0)
+ continue;
+
+ h = ch % charmap_data.hash_size;
+ n = 0;
+ while (names_b[h + n * charmap_data.hash_size] != 0)
+ ++n;
+
+ names_b[h + n * charmap_data.hash_size] = ch;
+ toupper_b[h + n * charmap_data.hash_size] = ch;
+ tolower_b[h + n * charmap_data.hash_size] = ch;
+ }
+ /* Correct the value for NUL character. */
+ names_b[0] = 0;
+ }
+}
+
+static void
+set_class_defaults (void)
+{
+ /* These function defines the default values for the classes and conversions
+ according to POSIX.2 2.5.2.1.
+ It may seem that the order of these if-blocks is arbitrary but it is NOT.
+ Don't move them unless you know what you do! */
+
+ void set_default (int bit, int from, int to)
+ {
+ char tmp[4];
+ int ch;
+ /* Define string. */
+ strcpy (tmp, "<?>");
+
+ for (ch = from; ch <= to; ++ch)
+ {
+ int code;
+ tmp[1] = ch;
+
+ code = find_char (tmp + 1, 1);
+ if (code == -1)
+ error (5, 0, gettext ("character `%s' not defined while needed "
+ "as default value"), tmp);
+ ELEM (ctype_b, code) |= bit;
+ }
+ }
+
+ /* If necessary allocate arrays. */
+ allocate_arrays ();
+
+ /* Set default values if keyword was not present. */
+ if ((class_done & BIT (TOK_UPPER)) == 0)
+ /* "If this keyword [lower] is not specified, the lowercase letters
+ `A' through `Z', ..., shall automatically belong to this class,
+ with implementation defined character values." */
+ set_default (BIT (TOK_UPPER), 'A', 'Z');
+
+ if ((class_done & BIT (TOK_LOWER)) == 0)
+ /* "If this keyword [lower] is not specified, the lowercase letters
+ `a' through `z', ..., shall automatically belong to this class,
+ with implementation defined character values." */
+ set_default (BIT (TOK_LOWER), 'a', 'z');
+
+ if ((class_done & BIT (TOK_DIGIT)) == 0)
+ /* "If this keyword [digit] is not specified, the digits `0' through
+ `9', ..., shall automatically belong to this class, with
+ implementation-defined character values." */
+ set_default (BIT (TOK_DIGIT), '0', '9');
+
+ if ((class_done & BIT (TOK_SPACE)) == 0)
+ /* "If this keyword [space] is not specified, the characters <space>,
+ <form-feed>, <newline>, <carriage-return>, <tab>, and
+ <vertical-tab>, ..., shall automatically belong to this class,
+ with implementtation-defined character values." */
+ {
+ int code;
+
+ code = find_char ("space", 5);
+ if (code == -1)
+ error (5, 0, gettext ("character `%s' not defined while needed as "
+ "default value"), "<space>");
+ ELEM (ctype_b, code) |= BIT (TOK_SPACE);
+
+ code = find_char ("form-feed", 9);
+ if (code == -1)
+ error (5, 0, gettext ("character `%s' not defined while needed as "
+ "default value"), "<form-feed>");
+ ELEM (ctype_b, code) |= BIT (TOK_SPACE);
+
+ code = find_char ("newline", 7);
+ if (code == -1)
+ error (5, 0, gettext ("character `%s' not defined while needed as "
+ "default value"), "<newline>");
+ ELEM (ctype_b, code) |= BIT (TOK_SPACE);
+
+ code = find_char ("carriage-return", 15);
+ if (code == -1)
+ error (5, 0, gettext ("character `%s' not defined while needed as "
+ "default value"), "<carriage-return>");
+ ELEM (ctype_b, code) |= BIT (TOK_SPACE);
+
+ code = find_char ("tab", 3);
+ if (code == -1)
+ error (5, 0, gettext ("character `%s' not defined while needed as "
+ "default value"), "<tab>");
+ ELEM (ctype_b, code) |= BIT (TOK_SPACE);
+
+ code = find_char ("vertical-tab", 11);
+ if (code == -1)
+ error (5, 0, gettext ("character `%s' not defined while needed as "
+ "default value"), "<vertical-tab>");
+ ELEM (ctype_b, code) |= BIT (TOK_SPACE);
+ }
+
+ if ((class_done & BIT (TOK_XDIGIT)) == 0)
+ /* "If this keyword is not specified, the digits `0' to `9', the
+ uppercase letters `A' through `F', and the lowercase letters `a'
+ through `f', ..., shell automatically belong to this class, with
+ implementation defined character values." */
+ {
+ if ((class_done & BIT (TOK_XDIGIT)) == 0)
+ set_default (BIT (TOK_XDIGIT), '0', '9');
+
+ if ((class_done & BIT (TOK_XDIGIT)) == 0)
+ set_default (BIT (TOK_XDIGIT), 'A', 'F');
+
+ if ((class_done & BIT (TOK_XDIGIT)) == 0)
+ set_default (BIT (TOK_XDIGIT), 'a', 'f');
+ }
+
+ if ((class_done & BIT (TOK_BLANK)) == 0)
+ /* "If this keyword [blank] is unspecified, the characters <space> and
+ <tab> shall belong to this character class." */
+ {
+ int code;
+
+ code = find_char ("space", 5);
+ if (code == -1)
+ error (5, 0, gettext ("character `%s' not defined while needed as "
+ "default value"), "<space>");
+ ELEM (ctype_b, code) |= BIT (TOK_BLANK);
+
+ code = find_char ("tab", 3);
+ if (code == -1)
+ error (5, 0, gettext ("character `%s' not defined while needed as "
+ "default value"), "<tab>");
+ ELEM (ctype_b, code) |= BIT (TOK_BLANK);
+ }
+
+ if ((class_done & BIT (TOK_GRAPH)) == 0)
+ /* "If this keyword [graph] is not specified, characters specified for
+ the keywords `upper', `lower', `alpha', `digit', `xdigit' and `punct',
+ shall belong to this character class." */
+ {
+ int ch;
+ unsigned short int mask = BIT (TOK_UPPER) | BIT (TOK_LOWER) |
+ BIT (TOK_ALPHA) | BIT (TOK_DIGIT) | BIT (TOK_XDIGIT) | BIT (TOK_PUNCT);
+
+ for (ch = 0; ch < charmap_data.hash_size * charmap_data.hash_layers;
+ ++ch)
+ {
+ if (ch != 0 && names_b[ch] == 0)
+ continue;
+ if ((ELEM (ctype_b, names_b[ch]) & mask) != 0)
+ ELEM (ctype_b, names_b[ch]) |= BIT (TOK_GRAPH);
+ }
+ }
+
+ if ((class_done & BIT (TOK_PRINT)) == 0)
+ /* "If this keyword [print] is not provided, characters specified for
+ the keywords `upper', `lower', `alpha', `digit', `xdigit', `punct',
+ and the <space> character shall belong to this character class." */
+ {
+ int ch;
+ int space = find_char ("space", 5);
+ unsigned short int mask = BIT (TOK_UPPER) | BIT (TOK_LOWER) |
+ BIT (TOK_ALPHA) | BIT (TOK_DIGIT) | BIT (TOK_XDIGIT) | BIT (TOK_PUNCT);
+
+ if (space == -1)
+ error (5, 0, gettext ("character `%s' not defined while needed as "
+ "default value"), "<space>");
+
+ for (ch = 0; ch < charmap_data.hash_size * charmap_data.hash_layers;
+ ++ch)
+ {
+ if (ch != 0 && names_b[ch] == 0)
+ continue;
+ if ((ELEM (ctype_b, names_b[ch]) & mask) != 0)
+ ELEM (ctype_b, names_b[ch]) |= BIT (TOK_PRINT);
+ }
+ ELEM (ctype_b, space) |= BIT (TOK_PRINT);
+ }
+
+ if (toupper_done == 0)
+ /* "If this keyword [toupper] is not spcified, the lowercase letters
+ `a' through `z', and their corresponding uppercase letters `A' to
+ `Z', ..., shall automatically be included, with implementation-
+ defined character values." */
+ {
+ char tmp[4];
+ int ch;
+
+ strcpy (tmp, "<?>");
+
+ for (ch = 'a'; ch <= 'z'; ++ch)
+ {
+ int code_to, code_from;
+
+ tmp[1] = ch;
+ code_from = find_char (tmp + 1, 1);
+ if (code_from == -1)
+ error (5, 0, gettext ("character `%s' not defined while needed "
+ "as default value"), tmp);
+
+ /* This conversion is implementation defined. */
+ tmp[1] = ch + ('A' - 'a');
+ code_to = find_char (tmp + 1, 1);
+ if (code_to == -1)
+ error (5, 0, gettext ("character `%s' not defined while needed "
+ "as default value"), tmp);
+
+ ELEM (toupper_b, code_from) = code_to;
+ }
+ }
+
+ if (tolower_done == 0)
+ /* "If this keyword [tolower] is not specified, the mapping shall be
+ the reverse mapping of the one specified to `toupper'." */
+ {
+ int ch;
+
+ for (ch = 0; ch < charmap_data.hash_size * charmap_data.hash_layers;
+ ++ch)
+ {
+ if (ch != 0 && names_b[ch] == 0)
+ continue;
+
+ if (toupper_b[ch] != names_b[ch])
+ ELEM (tolower_b, toupper_b[ch]) = names_b[ch];
+ }
+ }
+}
+
+
+/* Test whether the given character is valid for the current charmap. */
+static int
+valid_char (int ch)
+{
+ /* FIXME: this assumes 32-bit integers. */
+ int ok = ch >= 0
+ && (charmap_data.mb_cur_max < 4
+ ? ch < 1 << (8 * charmap_data.mb_cur_max) : 1);
+
+ return ok;
+}
+
+
+/*
+ * Local Variables:
+ * mode:c
+ * c-basic-offset:2
+ * End:
+ */
diff --git a/locale/ctypedump.c b/locale/ctypedump.c
new file mode 100644
index 0000000000..e00f3e0f11
--- /dev/null
+++ b/locale/ctypedump.c
@@ -0,0 +1,163 @@
+/* Copyright (C) 1995 Free Software Foundation, Inc.
+
+The GNU C Library is free software; you can redistribute it and/or
+modify it under the terms of the GNU Library General Public License as
+published by the Free Software Foundation; either version 2 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
+Library General Public License for more details.
+
+You should have received a copy of the GNU Library General Public
+License along with the GNU C Library; see the file COPYING.LIB. If
+not, write to the Free Software Foundation, Inc., 675 Mass Ave,
+Cambridge, MA 02139, USA. */
+
+#include <ctype.h>
+#include <endian.h>
+#include <stdlib.h>
+#include <stdio.h>
+#include <netinet/in.h> /* Just for htons() */
+
+#include "localedef.h"
+#include "localeinfo.h"
+
+
+/* FIXME: these values should be part of the LC_CTYPE information. */
+#define mb_cur_max 1
+#define mb_cur_min 1
+
+
+#define SWAP32(v) \
+ ((u32) (((((u32) (v)) & 0x000000ff) << 24) \
+ | ((((u32) (v)) & 0x0000ff00) << 8) \
+ | ((((u32) (v)) & 0x00ff0000) >> 8) \
+ | ((((u32) (v)) & 0xff000000) >> 24)))
+
+
+
+static inline void
+print_short_in_char (unsigned short val)
+{
+ const unsigned char *p = (const unsigned char *) &val;
+ printf ("\"\\%03o\\%03o\"", p[0], p[1]);
+}
+
+
+static inline void
+print_int_in_char (unsigned int val)
+{
+ const unsigned char *p = (const unsigned char *) &val;
+ printf ("\"\\%03o\\%03o\\%03o\\%03o\"", p[0], p[1], p[2], p[3]);
+}
+
+
+int
+ctype_output (void)
+{
+ int ch;
+ int result = 0;
+ const char *locname = (getenv ("LC_ALL") ?: getenv ("LC_CTYPE") ?:
+ getenv ("LANG") ?: "POSIX");
+
+ puts ("#include <endian.h>\n");
+
+ if (mb_cur_max == 1)
+ {
+ printf ("const char _nl_%s_LC_CTYPE_class[] = \n", locname);
+ for (ch = -128; ch < (1 << (8 * MB_CUR_MAX)); ++ch)
+ {
+ if (((ch + 128) % 6) == 0)
+ printf (" /* 0x%02x */ ", ch < 0 ? 256 + ch : ch);
+ print_short_in_char (htons (__ctype_b [ch < 0 ? 256 + ch : ch]));
+ fputc (((ch + 128) % 6) == 5 ? '\n' : ' ', stdout);
+ }
+ puts (";");
+ }
+
+ printf ("#if BYTE_ORDER == %s\n",
+ BYTE_ORDER == LITTLE_ENDIAN ? "LITTLE_ENDIAN" : "BIG_ENDIAN");
+
+ if (mb_cur_max == 1)
+ {
+ printf ("const char _nl_%s_LC_CTYPE_toupper[] = \n", locname);
+ for (ch = -128; ch < (1 << (8 * MB_CUR_MAX)); ++ch)
+ {
+ if (((ch + 128) % 3) == 0)
+ printf (" /* 0x%02x */ ", ch < 0 ? 256 + ch : ch);
+ print_int_in_char (__ctype_toupper[ch < 0 ? 256 + ch : ch]);
+ fputc (((ch + 128) % 3) == 2 ? '\n' : ' ', stdout);
+ }
+ puts (";");
+
+ printf ("const char _nl_%s_LC_CTYPE_tolower[] = \n", locname);
+ for (ch = -128; ch < (1 << (8 * MB_CUR_MAX)); ++ch)
+ {
+ if (((ch + 128) % 3) == 0)
+ printf (" /* 0x%02x */ ", ch < 0 ? 256 + ch : ch);
+ print_int_in_char (__ctype_tolower[ch < 0 ? 256 + ch : ch]);
+ fputc (((ch + 128) % 3) == 2 ? '\n' : ' ', stdout);
+ }
+ puts (";");
+ }
+ else
+ /* not implemented */;
+
+ printf ("#elif BYTE_ORDER == %s\n",
+ BYTE_ORDER == LITTLE_ENDIAN ? "BIG_ENDIAN" : "LITTLE_ENDIAN");
+
+ if (mb_cur_max == 1)
+ {
+ printf ("const char _nl_%s_LC_CTYPE_toupper[] = \n", locname);
+ for (ch = -128; ch < (1 << (8 * MB_CUR_MAX)); ++ch)
+ {
+ if (((ch + 128) % 3) == 0)
+ printf (" /* 0x%02x */ ", ch < 0 ? 256 + ch : ch);
+ print_int_in_char (SWAP32 (__ctype_toupper[ch < 0 ? 256 + ch : ch]));
+ fputc (((ch + 128) % 3) == 2 ? '\n' : ' ', stdout);
+ }
+ puts (";");
+
+ printf ("const char _nl_%s_LC_CTYPE_tolower[] = \n", locname);
+ for (ch = -128; ch < (1 << (8 * MB_CUR_MAX)); ++ch)
+ {
+ if (((ch + 128) % 3) == 0)
+ printf (" /* 0x%02x */ ", ch < 0 ? 256 + ch : ch);
+ print_int_in_char (SWAP32 (__ctype_tolower[ch < 0 ? 256 + ch : ch]));
+ fputc (((ch + 128) % 3) == 2 ? '\n' : ' ', stdout);
+ }
+ puts (";");
+ }
+ else
+ /* not implemented */;
+
+ puts ("#else\n#error \"BYTE_ORDER\" BYTE_ORDER \" not handled.\"\n#endif\n");
+
+ printf("const struct locale_data _nl_%s_LC_CTYPE = \n\
+{\n\
+ NULL, 0, /* no file mapped */\n\
+ 5,\n\
+ {\n\
+ _nl_C_LC_CTYPE_class,\n\
+#ifdef BYTE_ORDER == LITTLE_ENDIAN\n\
+ NULL, NULL,\n\
+#endif\n\
+ _nl_C_LC_CTYPE_toupper,\n\
+ _nl_C_LC_CTYPE_tolower,\n\
+#ifdef BYTE_ORDER == BIG_ENDIAN\n\
+ NULL, NULL,\n\
+#endif\n\
+ }\n\
+};\n", locname);
+
+ return result;
+}
+
+/*
+ * Local Variables:
+ * mode:c
+ * c-basic-offset:2
+ * End:
+ */
diff --git a/locale/hash.c b/locale/hash.c
new file mode 100644
index 0000000000..75cb77f882
--- /dev/null
+++ b/locale/hash.c
@@ -0,0 +1,254 @@
+/* Copyright (C) 1995 Free Software Foundation, Inc.
+
+The GNU C Library is free software; you can redistribute it and/or
+modify it under the terms of the GNU Library General Public License as
+published by the Free Software Foundation; either version 2 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
+Library General Public License for more details.
+
+You should have received a copy of the GNU Library General Public
+License along with the GNU C Library; see the file COPYING.LIB. If
+not, write to the Free Software Foundation, Inc., 675 Mass Ave,
+Cambridge, MA 02139, USA. */
+
+#include <obstack.h>
+#include <stdlib.h>
+#include <string.h>
+#include <values.h>
+
+#include "hash.h"
+
+#define obstack_chunk_alloc xmalloc
+#define obstack_chunk_free free
+
+void *xmalloc (size_t n);
+
+typedef struct hash_entry
+ {
+ int used;
+ char *key;
+ void *data;
+ struct hash_entry *next;
+ }
+hash_entry;
+
+/* Prototypes for local functions. */
+static size_t lookup (hash_table *htab, const char *key, size_t keylen,
+ unsigned long hval);
+static unsigned long compute_hashval(const char *key, size_t keylen);
+static unsigned long next_prime(unsigned long seed);
+static int is_prime(unsigned long candidate);
+
+
+int
+init_hash(hash_table *htab, unsigned long init_size)
+{
+ /* We need the size to be a prime. */
+ init_size = next_prime (init_size);
+
+ /* Initialize the data structure. */
+ htab->size = init_size;
+ htab->filled = 0;
+ htab->first = NULL;
+ htab->table = calloc (init_size + 1, sizeof (hash_entry));
+ obstack_init (&htab->mem_pool);
+
+ return htab->table == NULL;
+}
+
+
+int
+delete_hash(hash_table *htab)
+{
+ free (htab->table);
+ obstack_free (&htab->mem_pool, NULL);
+ return 0;
+}
+
+
+int
+insert_entry (hash_table *htab, const char *key, size_t keylen, void *data)
+{
+ unsigned long hval = compute_hashval (key, keylen);
+ hash_entry *table = (hash_entry *) htab->table;
+ size_t idx = lookup (htab, key, keylen, hval);
+
+ if (table[idx].used)
+ /* We don't want to overwrite the old value. */
+ return 1;
+ else
+ {
+ hash_entry **p;
+
+ /* An empty bucket has been found. */
+ table[idx].used = hval;
+ table[idx].key = obstack_copy0 (&htab->mem_pool, key, keylen);
+ table[idx].data = data;
+
+ /* List the new value in the ordered list. */
+ for (p = (hash_entry **) &htab->first; *p != NULL && (*p)->data < data;
+ p = &(*p)->next);
+ if (*p == NULL || (*p)->data > data)
+ /* Insert new value in the list. */
+ {
+ table[idx].next = *p;
+ *p = &table[idx];
+ }
+
+ ++htab->filled;
+ if (100 * htab->filled > 90 * htab->size)
+ {
+ /* Resize the table. */
+ unsigned long old_size = htab->size;
+
+ htab->size = next_prime (htab->size * 2);
+ htab->filled = 0;
+ htab->first = NULL;
+ htab->table = calloc (htab->size, sizeof (hash_entry));
+
+ for (idx = 1; idx <= old_size; ++idx)
+ if (table[idx].used)
+ insert_entry (htab, table[idx].key, strlen(table[idx].key),
+ table[idx].data);
+
+ free (table);
+ }
+ return 0;
+ }
+ /* NOTREACHED */
+}
+
+
+int
+find_entry (hash_table *htab, const char *key, size_t keylen, void **result)
+{
+ hash_entry *table = (hash_entry *) htab->table;
+ size_t idx = lookup (htab, key, keylen, compute_hashval (key, keylen));
+ int retval;
+
+ retval = table[idx].used;
+ *result = retval ? table[idx].data : NULL;
+
+ return retval;
+}
+
+
+int
+iterate_table (hash_table *htab, void **ptr, void **result)
+{
+ if (*ptr == NULL)
+ *ptr = (void *) htab->first;
+ else
+ {
+ *ptr = (void *) (((hash_entry *) *ptr)->next);
+ if (*ptr == NULL)
+ return 0;
+ }
+
+ *result = ((hash_entry *) *ptr)->data;
+ return 1;
+}
+
+
+static size_t
+lookup (hash_table *htab, const char *key, size_t keylen, unsigned long hval)
+{
+ unsigned long hash;
+ size_t idx;
+ hash_entry *table = (hash_entry *) htab->table;
+
+ /* First hash function: simply take the modul but prevent zero. */
+ hash = 1 + hval % htab->size;
+
+ idx = hash;
+
+ if (table[idx].used)
+ {
+ if (table[idx].used == hval && table[idx].key[keylen] == '\0'
+ && strncmp (key, table[idx].key, keylen) == 0)
+ return idx;
+
+ /* Second hash function as suggested in [Knuth]. */
+ hash = 1 + hash % (htab->size - 2);
+
+ do
+ {
+ if (idx <= hash)
+ idx = htab->size + idx - hash;
+ else
+ idx -= hash;
+
+ /* If entry is found use it. */
+ if (table[idx].used == hval && table[idx].key[keylen] == '\0'
+ && strncmp (key, table[idx].key, keylen) == 0)
+ return idx;
+ }
+ while (table[idx].used);
+ }
+ return idx;
+}
+
+
+static unsigned long
+compute_hashval(const char *key, size_t keylen)
+{
+ size_t cnt;
+ unsigned long hval, g;
+ /* Compute the hash value for the given string. */
+ cnt = 0;
+ hval = keylen;
+ while (cnt < keylen)
+ {
+ hval <<= 4;
+ hval += key[cnt++];
+ g = hval & (0xf << (LONGBITS - 4));
+ if (g != 0)
+ {
+ hval ^= g >> (LONGBITS - 8);
+ hval ^= g;
+ }
+ }
+ return hval;
+}
+
+
+static unsigned long
+next_prime(unsigned long seed)
+{
+ /* Make it definitely odd. */
+ seed |= 1;
+
+ while (!is_prime (seed))
+ seed += 2;
+
+ return seed;
+}
+
+
+static int
+is_prime(unsigned long candidate)
+{
+ /* No even number and none less than 10 will be passwd here. */
+ unsigned long div = 3;
+ unsigned long sq = div * div;
+
+ while (sq < candidate && candidate % div != 0)
+ {
+ ++div;
+ sq += 4 * div;
+ ++div;
+ }
+
+ return candidate % div != 0;
+}
+
+/*
+ * Local Variables:
+ * mode:c
+ * c-basic-offset:2
+ * End:
+ */
diff --git a/locale/hash.h b/locale/hash.h
new file mode 100644
index 0000000000..5f60a44a5b
--- /dev/null
+++ b/locale/hash.h
@@ -0,0 +1,50 @@
+/* Copyright (C) 1995 Free Software Foundation, Inc.
+
+The GNU C Library is free software; you can redistribute it and/or
+modify it under the terms of the GNU Library General Public License as
+published by the Free Software Foundation; either version 2 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
+Library General Public License for more details.
+
+You should have received a copy of the GNU Library General Public
+License along with the GNU C Library; see the file COPYING.LIB. If
+not, write to the Free Software Foundation, Inc., 675 Mass Ave,
+Cambridge, MA 02139, USA. */
+
+#ifndef _HASH_H
+#define _HASH_H
+
+#include <obstack.h>
+
+typedef struct hash_table
+ {
+ unsigned long size;
+ unsigned long filled;
+ void *first;
+ void *table;
+ struct obstack mem_pool;
+ }
+hash_table;
+
+
+int init_hash (hash_table *htab, unsigned long init_size);
+int delete_hash(hash_table *htab);
+int insert_entry (hash_table *htab, const char *key, size_t keylen,
+ void *data);
+int find_entry (hash_table *htab, const char *key, size_t keylen,
+ void **result);
+
+int iterate_table (hash_table *htab, void **ptr, void **result);
+
+#endif /* hash.h */
+/*
+ * Local Variables:
+ * mode:c
+ * c-basic-offset:2
+ * End:
+ */
+
diff --git a/locale/iso-4217.def b/locale/iso-4217.def
new file mode 100644
index 0000000000..849420365e
--- /dev/null
+++ b/locale/iso-4217.def
@@ -0,0 +1,36 @@
+/*
+ * Defines the valid international currency symbols according to ISO-4217.
+ * This is used in monetary.c(monetary_check).
+ *
+ * !!! The list has to be sorted !!!
+ */
+DEFINE_INT_CURR("ATS ") /* Austria */
+DEFINE_INT_CURR("BEF ") /* Belgium */
+DEFINE_INT_CURR("CAD ") /* Canada */
+DEFINE_INT_CURR("CHF ") /* Switzerland */
+DEFINE_INT_CURR("DEM ") /* Germany */
+DEFINE_INT_CURR("DKK ") /* Denmark */
+DEFINE_INT_CURR("EEK ") /* Estonia */
+DEFINE_INT_CURR("ESP ") /* Spain */
+DEFINE_INT_CURR("FIM ") /* Finland */
+DEFINE_INT_CURR("FRF ") /* France */
+DEFINE_INT_CURR("GBP ") /* Great Britain */
+DEFINE_INT_CURR("GRD ") /* Greece */
+DEFINE_INT_CURR("HRD ") /* Croatia */
+DEFINE_INT_CURR("HUF ") /* Hungary */
+DEFINE_INT_CURR("IEP ") /* Ireland */
+DEFINE_INT_CURR("ILS ") /* Israel */
+DEFINE_INT_CURR("ISK ") /* Iceland */
+DEFINE_INT_CURR("ITL ") /* Italy */
+DEFINE_INT_CURR("LTL ") /* Lithuania */
+DEFINE_INT_CURR("LUF ") /* Luxemburg */
+DEFINE_INT_CURR("LVL ") /* Latvia */
+DEFINE_INT_CURR("NLG ") /* Netherlands */
+DEFINE_INT_CURR("NOK ") /* Norway */
+DEFINE_INT_CURR("PLZ ") /* Poland */
+DEFINE_INT_CURR("PTE ") /* Portugal */
+DEFINE_INT_CURR("ROL ") /* Romania */
+DEFINE_INT_CURR("RUR ") /* Russia */
+DEFINE_INT_CURR("SEK ") /* Sweden */
+DEFINE_INT_CURR("SIT ") /* Slovenia */
+DEFINE_INT_CURR("USD ") /* United States */
diff --git a/locale/keyword.gperf b/locale/keyword.gperf
new file mode 100644
index 0000000000..20941bc03d
--- /dev/null
+++ b/locale/keyword.gperf
@@ -0,0 +1,77 @@
+%{
+/* `strncmp' is used for comparison. */
+#include <string.h>
+
+/* This file defines `enum token'. */
+#include "token.h"
+%}
+struct locale_keyword { char *name; enum token token_id; };
+%%
+END, TOK_END
+IGNORE, TOK_IGNORE
+LC_COLLATE, _NL_NUM_LC_COLLATE
+LC_CTYPE, _NL_NUM_LC_CTYPE
+LC_MESSAGES, _NL_NUM_LC_MESSAGES
+LC_MONETARY, _NL_NUM_LC_MONETARY
+LC_NUMERIC, _NL_NUM_LC_NUMERIC
+LC_TIME, _NL_NUM_LC_TIME
+UNDEFINED, TOK_UNDEFINED
+abday, ABDAY_1
+abmon, ABMON_1
+alpha, TOK_ALPHA
+alt_digits, ALT_DIGITS
+am_pm, AM_STR
+backward, TOK_BACKWARD
+blank, TOK_BLANK
+cntrl, TOK_CNTRL
+collating_element, TOK_COLLATING_ELEMENT
+collating_symbol, TOK_COLLATING_SYMBOL
+comment_char, TOK_COMMENT_CHAR
+copy, TOK_COPY
+currency_symbol, CURRENCY_SYMBOL
+d_fmt, D_FMT
+d_t_fmt, D_T_FMT
+day, DAY_1
+decimal_point, DECIMAL_POINT
+digit, TOK_DIGIT
+era, ERA
+era_d_fmt, ERA_D_FMT
+era_year, ERA_YEAR
+escape_char, TOK_ESCAPE_CHAR
+forward, TOK_FORWARD
+frac_digits, FRAC_DIGITS
+from, TOK_FROM
+graph, TOK_GRAPH
+grouping, GROUPING
+int_curr_symbol, INT_CURR_SYMBOL
+int_frac_digits, INT_FRAC_DIGITS
+lower, TOK_LOWER
+mon, MON_1
+mon_decimal_point, MON_DECIMAL_POINT
+mon_grouping, MON_GROUPING
+mon_thousands_sep, MON_THOUSANDS_SEP
+n_cs_precedes, N_CS_PRECEDES
+n_sep_by_space, N_SEP_BY_SPACE
+n_sign_posn, N_SIGN_POSN
+negative_sign, NEGATIVE_SIGN
+noexpr, NOEXPR
+nostr, NOSTR
+order_end, TOK_ORDER_END
+order_start, TOK_ORDER_START
+p_cs_precedes, P_CS_PRECEDES
+p_sep_by_space, P_SEP_BY_SPACE
+p_sign_posn, P_SIGN_POSN
+position, TOK_POSITION
+positive_sign, POSITIVE_SIGN
+print, TOK_PRINT
+punct, TOK_PUNCT
+space, TOK_SPACE
+t_fmt, T_FMT
+t_fmt_ampm, T_FMT_AMPM
+thousands_sep, THOUSANDS_SEP
+tolower, TOK_TOLOWER
+toupper, TOK_TOUPPER
+upper, TOK_UPPER
+xdigit, TOK_XDIGIT
+yesexpr, YESEXPR
+yesstr, YESSTR
diff --git a/locale/keyword.h b/locale/keyword.h
new file mode 100644
index 0000000000..1dc442aee5
--- /dev/null
+++ b/locale/keyword.h
@@ -0,0 +1,180 @@
+/* C code produced by gperf version 2.5 (GNU C++ version) */
+/* Command-line: gperf -acCgopt -k1,2,5, keyword.gperf */
+/* `strncmp' is used for comparison. */
+#include <string.h>
+
+/* This file defines `enum token'. */
+#include "token.h"
+struct locale_keyword { char *name; enum token token_id; };
+
+#define TOTAL_KEYWORDS 68
+#define MIN_WORD_LENGTH 3
+#define MAX_WORD_LENGTH 17
+#define MIN_HASH_VALUE 4
+#define MAX_HASH_VALUE 140
+/* maximum key range = 137, duplicates = 0 */
+
+#ifdef __GNUC__
+inline
+#endif
+static unsigned int
+hash (register const char *str, register int len)
+{
+ static const unsigned char asso_values[] =
+ {
+ 141, 141, 141, 141, 141, 141, 141, 141, 141, 141,
+ 141, 141, 141, 141, 141, 141, 141, 141, 141, 141,
+ 141, 141, 141, 141, 141, 141, 141, 141, 141, 141,
+ 141, 141, 141, 141, 141, 141, 141, 141, 141, 141,
+ 141, 141, 141, 141, 141, 141, 141, 141, 141, 141,
+ 141, 141, 141, 141, 141, 141, 141, 141, 141, 141,
+ 141, 141, 141, 141, 141, 141, 141, 0, 141, 65,
+ 5, 0, 141, 30, 141, 141, 0, 141, 0, 95,
+ 141, 141, 0, 141, 45, 10, 141, 141, 141, 141,
+ 141, 141, 141, 141, 141, 5, 141, 10, 85, 0,
+ 20, 0, 40, 35, 30, 10, 141, 0, 30, 15,
+ 15, 0, 0, 141, 55, 0, 0, 80, 141, 15,
+ 10, 0, 141, 141, 141, 141, 141, 141,
+ };
+ register int hval = len;
+
+ switch (hval)
+ {
+ default:
+ case 5:
+ hval += asso_values[str[4]];
+ case 4:
+ case 3:
+ case 2:
+ hval += asso_values[str[1]];
+ case 1:
+ hval += asso_values[str[0]];
+ }
+ return hval;
+}
+
+#ifdef __GNUC__
+inline
+#endif
+const struct locale_keyword *
+in_word_set (register const char *str, register int len)
+{
+ static const struct locale_keyword wordlist[] =
+ {
+ {"",}, {"",}, {"",}, {"",},
+ {"copy", TOK_COPY},
+ {"space", TOK_SPACE},
+ {"yesstr", YESSTR},
+ {"toupper", TOK_TOUPPER},
+ {"position", TOK_POSITION},
+ {"",},
+ {"t_fmt", T_FMT},
+ {"escape_char", TOK_ESCAPE_CHAR},
+ {"comment_char", TOK_COMMENT_CHAR},
+ {"positive_sign", POSITIVE_SIGN},
+ {"",},
+ {"t_fmt_ampm", T_FMT_AMPM},
+ {"",},
+ {"yesexpr", YESEXPR},
+ {"mon", MON_1},
+ {"p_sep_by_space", P_SEP_BY_SPACE},
+ {"LC_NUMERIC", _NL_NUM_LC_NUMERIC},
+ {"noexpr", NOEXPR},
+ {"tolower", TOK_TOLOWER},
+ {"p_cs_precedes", P_CS_PRECEDES},
+ {"UNDEFINED", TOK_UNDEFINED},
+ {"",},
+ {"collating_symbol", TOK_COLLATING_SYMBOL},
+ {"collating_element", TOK_COLLATING_ELEMENT},
+ {"negative_sign", NEGATIVE_SIGN},
+ {"",},
+ {"d_fmt", D_FMT},
+ {"",},
+ {"mon_thousands_sep", MON_THOUSANDS_SEP},
+ {"day", DAY_1},
+ {"n_sep_by_space", N_SEP_BY_SPACE},
+ {"digit", TOK_DIGIT},
+ {"IGNORE", TOK_IGNORE},
+ {"LC_TIME", _NL_NUM_LC_TIME},
+ {"n_cs_precedes", N_CS_PRECEDES},
+ {"",},
+ {"int_curr_symbol", INT_CURR_SYMBOL},
+ {"",}, {"",},
+ {"thousands_sep", THOUSANDS_SEP},
+ {"",},
+ {"am_pm", AM_STR},
+ {"xdigit", TOK_XDIGIT},
+ {"",},
+ {"decimal_point", DECIMAL_POINT},
+ {"",},
+ {"cntrl", TOK_CNTRL},
+ {"p_sign_posn", P_SIGN_POSN},
+ {"mon_decimal_point", MON_DECIMAL_POINT},
+ {"LC_CTYPE", _NL_NUM_LC_CTYPE},
+ {"",},
+ {"alpha", TOK_ALPHA},
+ {"",},
+ {"forward", TOK_FORWARD},
+ {"era", ERA},
+ {"",},
+ {"print", TOK_PRINT},
+ {"",},
+ {"mon_grouping", MON_GROUPING},
+ {"era_year", ERA_YEAR},
+ {"",}, {"",},
+ {"n_sign_posn", N_SIGN_POSN},
+ {"",},
+ {"END", TOK_END},
+ {"",},
+ {"alt_digits", ALT_DIGITS},
+ {"",},
+ {"d_t_fmt", D_T_FMT},
+ {"",}, {"",},
+ {"nostr", NOSTR},
+ {"LC_MESSAGES", _NL_NUM_LC_MESSAGES},
+ {"",}, {"",}, {"",},
+ {"int_frac_digits", INT_FRAC_DIGITS},
+ {"",}, {"",}, {"",},
+ {"era_d_fmt", ERA_D_FMT},
+ {"punct", TOK_PUNCT},
+ {"",}, {"",}, {"",}, {"",},
+ {"lower", TOK_LOWER},
+ {"",}, {"",}, {"",}, {"",},
+ {"currency_symbol", CURRENCY_SYMBOL},
+ {"",}, {"",},
+ {"grouping", GROUPING},
+ {"from", TOK_FROM},
+ {"abday", ABDAY_1},
+ {"",}, {"",}, {"",}, {"",},
+ {"LC_COLLATE", _NL_NUM_LC_COLLATE},
+ {"LC_MONETARY", _NL_NUM_LC_MONETARY},
+ {"",}, {"",}, {"",}, {"",},
+ {"frac_digits", FRAC_DIGITS},
+ {"",}, {"",}, {"",},
+ {"abmon", ABMON_1},
+ {"",}, {"",},
+ {"backward", TOK_BACKWARD},
+ {"order_end", TOK_ORDER_END},
+ {"blank", TOK_BLANK},
+ {"order_start", TOK_ORDER_START},
+ {"",}, {"",}, {"",},
+ {"graph", TOK_GRAPH},
+ {"",}, {"",}, {"",}, {"",}, {"",}, {"",}, {"",}, {"",}, {"",},
+ {"",}, {"",}, {"",}, {"",}, {"",},
+ {"upper", TOK_UPPER},
+ };
+
+ if (len <= MAX_WORD_LENGTH && len >= MIN_WORD_LENGTH)
+ {
+ register int key = hash (str, len);
+
+ if (key <= MAX_HASH_VALUE && key >= 0)
+ {
+ register const char *s = wordlist[key].name;
+
+ if (*s == *str && !strncmp (str + 1, s + 1, len - 1))
+ return &wordlist[key];
+ }
+ }
+ return 0;
+}
diff --git a/locale/langinfo.h b/locale/langinfo.h
index e94be68430..245dcf8b92 100644
--- a/locale/langinfo.h
+++ b/locale/langinfo.h
@@ -145,8 +145,10 @@ typedef enum
_NL_NUM_LC_MESSAGES,
/* Stubs for unfinished categories. */
- _NL_NUM_LC_COLLATE = 0,
+ _NL_NUM_LC_COLLATE = _NL_ITEM (LC_COLLATE, 0),
+ /* This marks the highest value used. */
+ _NL_NUM
} nl_item;
diff --git a/locale/libintl.h b/locale/libintl.h
new file mode 100644
index 0000000000..802bd45492
--- /dev/null
+++ b/locale/libintl.h
@@ -0,0 +1,77 @@
+/* libintl.h -- Message catalogs for internationalization.
+Copyright (C) 1995 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 Library General Public License as
+published by the Free Software Foundation; either version 2 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
+Library General Public License for more details.
+
+You should have received a copy of the GNU Library General Public
+License along with the GNU C Library; see the file COPYING.LIB. If
+not, write to the Free Software Foundation, Inc., 675 Mass Ave,
+Cambridge, MA 02139, USA. */
+
+#ifndef _LIBINTL_H
+#define _LIBINTL_H 1
+
+#include <locale.h>
+
+
+/* Look up MSGID in the current default message catalog for the current
+ LC_MESSAGES locale. If not found, returns MSGID itself (the default
+ text). */
+extern char *gettext __P ((const char *__msgid));
+
+/* Look up MSGID in the DOMAINNAME message catalog for the current
+ LC_MESSAGES locale. */
+extern char *dgettext __P ((const char *__domainname, const char *__msgid));
+
+/* Look up MSGID in the DOMAINNAME message catalog for the current CATEGORY
+ locale. */
+extern char *__dcgettext __P ((const char *__domainname, const char *__msgid,
+ int __category));
+extern char *dcgettext __P ((const char *__domainname, const char *__msgid,
+ int __category));
+
+
+/* Set the current default message catalog to DOMAINNAME.
+ If DOMAINNAME is null, return the current default.
+ If DOMAINNAME is "", reset to the default of "messages". */
+extern char *textdomain __P ((const char *__domainname));
+
+/* Specify that the DOMAINNAME message catalog will be found
+ in DIRNAME rather than in the system locale data base. */
+extern char *bindtextdomain __P ((const char *__domainname,
+ const char *__dirname));
+
+
+#if 1 /* XXX stub for the moment */
+#define gettext(msgid) (msgid)
+#define textdomain(domain) (void)(domain)
+#else
+#define gettext(msgid) __gettext (msgid)
+#define __gettext(msgid) __dgettext (NULL, (msgid))
+
+#define dgettext(domainname, msgid) __dgettext (domainname, msgid)
+#define __dgettext(domainname, msgid) \
+ __dcgettext (NULL, (msgid), LC_MESSAGES)
+
+#ifdef __GNUC__
+#define __dcgettext(domainname, msgid, category) \
+ (__extension__ \
+ ({ \
+ static char *__translation__; \
+ if (! __translation__) \
+ __translation__ = (__dcgettext) ((domainname), (msgid), (category)); \
+ __translation__; \
+ }))
+#endif
+#endif
+
+#endif /* libintl.h */
diff --git a/locale/loadlocale.c b/locale/loadlocale.c
index 5073bd4be3..7e29c97205 100644
--- a/locale/loadlocale.c
+++ b/locale/loadlocale.c
@@ -69,13 +69,14 @@ _nl_load_locale (int category, char **name)
}
{
- const char localedir[] = "/share/locale/"; /* XXX */
const char *catname = _nl_category_names[category];
size_t namelen = strlen (*name);
size_t catlen = strlen (catname);
- char file[sizeof localedir + namelen + catlen * 2 + 4];
- sprintf (file, "%s%s/%s",
- strchr (*name, '/') != NULL ? "" : localedir, *name, catname);
+ char file[sizeof LOCALE_PATH + 1 + namelen + catlen * 2 + 4];
+ if (strchr (*name, '/') != NULL)
+ sprintf (file, "%s/%s", *name, catname);
+ else
+ sprintf (file, "%s/%s/%s", LOCALE_PATH, *name, catname);
fd = __open (file, O_RDONLY);
if (fd < 0)
return NULL;
diff --git a/locale/locale.c b/locale/locale.c
new file mode 100644
index 0000000000..20385f07d8
--- /dev/null
+++ b/locale/locale.c
@@ -0,0 +1,538 @@
+/* Copyright (C) 1995 Free Software Foundation, Inc.
+
+The GNU C Library is free software; you can redistribute it and/or
+modify it under the terms of the GNU Library General Public License as
+published by the Free Software Foundation; either version 2 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
+Library General Public License for more details.
+
+You should have received a copy of the GNU Library General Public
+License along with the GNU C Library; see the file COPYING.LIB. If
+not, write to the Free Software Foundation, Inc., 675 Mass Ave,
+Cambridge, MA 02139, USA. */
+
+#include <dirent.h>
+#include <getopt.h>
+#include <langinfo.h>
+#include <libintl.h>
+#include <limits.h>
+#include <locale.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <unistd.h>
+
+#include "localedef.h"
+
+
+/* If set dump C code describing the current locale. */
+static int do_dump;
+
+/* If set print the name of the category. */
+static int show_category_name;
+
+/* If set print the name of the item. */
+static int show_keyword_name;
+
+/* Long options. */
+static const struct option long_options[] =
+ {
+ { "all-locales", no_argument, NULL, 'a' },
+ { "category-name", no_argument, &show_category_name, 1 },
+ { "charmaps", no_argument, NULL, 'm' },
+ { "dump", no_argument, &do_dump, 1 },
+ { "help", no_argument, NULL, 'h' },
+ { "keyword-name", no_argument, &show_keyword_name, 1 },
+ { "version", no_argument, NULL, 'v' },
+ { NULL, 0, NULL, 0 }
+ };
+
+
+/* We don't have these constants defined because we don't use them. Give
+ default values. */
+#define CTYPE_MB_CUR_MIN 0
+#define CTYPE_MB_CUR_MAX 0
+#define CTYPE_HASH_SIZE 0
+#define CTYPE_HASH_LAYERS 0
+#define CTYPE_CLASS 0
+#define CTYPE_TOUPPER_EB 0
+#define CTYPE_TOLOWER_EB 0
+#define CTYPE_TOUPPER_EL 0
+#define CTYPE_TOLOWER_EL 0
+
+
+/* We have all categories defined in `categories.def'. Now construct
+ the description and data structure used for all categories. */
+#define DEFINE_CATEGORY(category, name, items, postload, in, check, out) \
+ static struct cat_item category##_desc[] = \
+ { \
+ NO_PAREN items \
+ };
+
+#include "categories.def"
+#undef DEFINE_CATEGORY
+
+static struct category category[] =
+ {
+#define DEFINE_CATEGORY(category, name, items, postload, in, check, out) \
+ { _NL_NUM_##category, name, NELEMS (category##_desc) - 1, \
+ category##_desc, NULL, NULL, NULL, out },
+#include "categories.def"
+#undef DEFINE_CATEGORY
+ };
+#define NCATEGORIES NELEMS (category)
+
+
+/* Prototypes for local functions. */
+static void usage (int status) __attribute__ ((noreturn));
+static void write_locales (void);
+static void write_charmaps (void);
+static void show_locale_vars (void);
+static void show_info (const char *name);
+static void dump_category (const char *name);
+
+
+int
+main (int argc, char *argv[])
+{
+ int optchar;
+ int do_all = 0;
+ int do_help = 0;
+ int do_version = 0;
+ int do_charmaps = 0;
+
+ /* Set initial values for global varaibles. */
+ do_dump = 0;
+ show_category_name = 0;
+ show_keyword_name = 0;
+
+ /* Set locale. Do not set LC_ALL because the other categories must
+ not be affected (acccording to POSIX.2). */
+ setlocale (LC_CTYPE, "");
+ setlocale (LC_MESSAGES, "");
+
+ /* Initialize the message catalog. */
+ textdomain (PACKAGE);
+
+ while ((optchar = getopt_long (argc, argv, "achkmv", long_options, NULL))
+ != EOF)
+ switch (optchar)
+ {
+ case '\0':
+ break;
+ case 'a':
+ do_all = 1;
+ break;
+ case 'c':
+ show_category_name = 1;
+ break;
+ case 'h':
+ do_help = 1;
+ break;
+ case 'k':
+ show_keyword_name = 1;
+ break;
+ case 'm':
+ do_charmaps = 1;
+ break;
+ case 'v':
+ do_version = 1;
+ break;
+ default:
+ error (1, 0, gettext ("illegal option \"%s\""), optarg);
+ break;
+ }
+
+ /* Version information is requested. */
+ if (do_version)
+ {
+ fprintf (stderr, "GNU %s %s\n", PACKAGE, VERSION);
+ exit (EXIT_SUCCESS);
+ }
+
+ /* Help is requested. */
+ if (do_help)
+ usage (EXIT_SUCCESS);
+
+ /* Dump C code. */
+ if (do_dump)
+ {
+ printf ("\
+/* Generated by GNU %s %s. */\n\
+\n\
+#include \"localeinfo.h\"\n", program_invocation_name, VERSION);
+
+ while (optind < argc)
+ dump_category (argv[optind++]);
+
+ exit (EXIT_SUCCESS);
+ }
+
+ /* `-a' requests the names of all available locales. */
+ if (do_all != 0)
+ {
+ write_locales ();
+ exit (EXIT_SUCCESS);
+ }
+
+ /* `m' requests the names of all available charmaps. The names can be
+ used for the -f argument to localedef(3). */
+ if (do_charmaps != 0)
+ {
+ write_charmaps ();
+ exit (EXIT_SUCCESS);
+ }
+
+ /* If no real argument is given we have to print the contents of the
+ current locale definition variables. These are LANG and the LC_*. */
+ if (optind == argc && show_keyword_name == 0 && show_category_name == 0)
+ {
+ show_locale_vars ();
+ exit (EXIT_SUCCESS);
+ }
+
+ /* Process all given names. */
+ while (optind < argc)
+ show_info (argv[optind++]);
+
+ exit (EXIT_SUCCESS);
+}
+
+
+/* Display usage information and exit. */
+static void
+usage(int status)
+{
+ if (status != EXIT_SUCCESS)
+ fprintf (stderr, gettext ("Try `%s --help' for more information.\n"),
+ program_invocation_name);
+ else
+ printf(gettext ("\
+Usage: %s [OPTION]... name\n\
+Mandatory arguments to long options are mandatory for short options too.\n\
+ -h, --help display this help and exit\n\
+ -v, --version output version information and exit\n\
+\n\
+ -a, --all-locales write names of available locales\n\
+ -m, --charmaps write names of available charmaps\n\
+\n\
+ -c, --category-name write names of selected categories\n\
+ -k, --keyword-name write names of selected keywords\n\
+\n\
+ --dump dump C code describing the current locale\n\
+ (this code can be used in the C library)\n\
+"), program_invocation_name);
+
+ exit (status);
+}
+
+
+/* Write the names of all available locales to stdout. */
+static void
+write_locales (void)
+{
+ DIR *dir;
+ struct dirent *dirent;
+
+ /* `POSIX' locale is always available (POSIX.2 4.34.3). */
+ puts ("POSIX");
+
+ dir = opendir (LOCALE_PATH);
+ if (dir == NULL)
+ {
+ error (1, errno, gettext ("cannot read locale directory `%s'"),
+ LOCALE_PATH);
+ return;
+ }
+
+ /* Now we can look for all files in the directory. */
+ while ((dirent = readdir (dir)) != NULL)
+ if (strcmp (dirent->d_name, ".") != 0
+ && strcmp (dirent->d_name, "..") != 0)
+ puts (dirent->d_name);
+
+ closedir (dir);
+}
+
+
+/* Write the names of all available character maps to stdout. */
+static void
+write_charmaps (void)
+{
+ DIR *dir;
+ struct dirent *dirent;
+
+ dir = opendir (CHARMAP_PATH);
+ if (dir == NULL)
+ {
+ error (1, errno, gettext ("cannot read character map directory `%s'"),
+ CHARMAP_PATH);
+ return;
+ }
+
+ /* Now we can look for all files in the directory. */
+ while ((dirent = readdir (dir)) != NULL)
+ if (strcmp (dirent->d_name, ".") != 0
+ && strcmp (dirent->d_name, "..") != 0)
+ puts (dirent->d_name);
+
+ closedir (dir);
+}
+
+
+/* We have to show the contents of the environments determining the
+ locale. */
+static void
+show_locale_vars (void)
+{
+ size_t cat_no;
+ const char *lcall = getenv ("LC_ALL");
+ const char *lang = getenv ("LANG") ? : "POSIX";
+
+ void get_source (const char *name)
+ {
+ char *val = getenv (name);
+
+ if (lcall != NULL || val == NULL)
+ printf ("%s=\"%s\"\n", name, lcall ? : lang);
+ else
+ printf ("%s=%s\n", name, val);
+ }
+
+ /* LANG has to be the first value. */
+ printf ("LANG=%s\n", lang);
+
+ /* Now all categories in an unspecified order. */
+ for (cat_no = 0; cat_no < NCATEGORIES; ++cat_no)
+ get_source (category[cat_no].name);
+
+ /* The last is the LC_ALL value. */
+ printf ("LC_ALL=%s\n", lcall ? : "");
+}
+
+
+/* Show the information request for NAME. */
+static void
+show_info (const char *name)
+{
+ size_t cat_no;
+
+ void print_item (struct cat_item *item)
+ {
+ if (show_keyword_name != 0)
+ printf ("%s=", item->name);
+
+ switch (item->value_type)
+ {
+ case string:
+ printf ("%s%s%s", show_keyword_name ? "\"" : "",
+ nl_langinfo (item->item_id) ? : "",
+ show_keyword_name ? "\"" : "");
+ break;
+ case stringarray:
+ {
+ int cnt;
+ const char *val;
+
+ if (show_keyword_name)
+ putchar ('"');
+
+ for (cnt = 0; cnt < item->max - 1; ++cnt)
+ {
+ val = nl_langinfo (item->item_id + cnt);
+ printf ("%s;", val ? : "");
+ }
+
+ val = nl_langinfo (item->item_id + cnt);
+ printf ("%s", val ? : "");
+
+ if (show_keyword_name)
+ putchar ('"');
+ }
+ break;
+ case byte:
+ {
+ const char *val = nl_langinfo (item->item_id);
+
+ if (val != NULL)
+ printf ("%d", *val == CHAR_MAX ? -1 : *val);
+ }
+ break;
+ case bytearray:
+ {
+ const char *val = nl_langinfo (item->item_id);
+ int cnt = val ? strlen (val) : 0;
+
+ while (cnt > 1)
+ {
+ printf ("%d;", *val == CHAR_MAX ? -1 : *val);
+ --cnt;
+ ++val;
+ }
+
+ printf ("%d", cnt == 0 || *val == CHAR_MAX ? -1 : *val);
+ }
+ break;
+ default:
+ }
+ putchar ('\n');
+ }
+
+ for (cat_no = 0; cat_no < NCATEGORIES; ++cat_no)
+ {
+ size_t item_no;
+
+ if (category[cat_no].outfct != NULL)
+ /* Categories which need special handling of the output are
+ not written. This is especially for LC_CTYPE and LC_COLLATE.
+ It does not make sense to have this large number of cryptic
+ characters displayed. */
+ continue;
+
+ if (strcmp (name, category[cat_no].name) == 0)
+ /* Print the whole category. */
+ {
+ if (show_category_name != 0)
+ puts (category[cat_no].name);
+
+ for (item_no = 0; item_no < category[cat_no].number; ++item_no)
+ print_item (&category[cat_no].item_desc[item_no]);
+
+ return;
+ }
+
+ for (item_no = 0; item_no < category[cat_no].number; ++item_no)
+ if (strcmp (name, category[cat_no].item_desc[item_no].name) == 0)
+ {
+ if (show_category_name != 0)
+ puts (category[cat_no].name);
+
+ print_item (&category[cat_no].item_desc[item_no]);
+ return;
+ }
+ }
+}
+
+
+static void
+dump_category (const char *name)
+{
+ char *locname;
+ size_t cat_no, item_no, nstrings;
+
+ for (cat_no = 0; cat_no < NCATEGORIES; ++cat_no)
+ if (strcmp (name, category[cat_no].name) == 0)
+ break;
+
+ if (cat_no >= NCATEGORIES)
+ return;
+
+ /* The NAME specifies a correct locale category. */
+ if (category[cat_no].outfct != NULL)
+ {
+ category[cat_no].outfct ();
+ return;
+ }
+
+ locname = (getenv ("LC_ALL") ?: getenv (name) ?:
+ getenv ("LANG") ?: (char *) "POSIX");
+
+ /* Determine the number of strings in advance. */
+ nstrings = 0;
+ for (item_no = 0; item_no < category[cat_no].number; ++item_no)
+ switch (category[cat_no].item_desc[item_no].value_type)
+ {
+ case string:
+ case byte:
+ case bytearray:
+ ++nstrings;
+ break;
+ case stringarray:
+ nstrings += category[cat_no].item_desc[item_no].max;
+ default:
+ }
+
+ printf ("\nconst struct locale_data _nl_%s_%s =\n{\n"
+ " NULL, 0, /* no file mapped */\n %Zu,\n {\n",
+ locname, name, nstrings);
+
+ for (item_no = 0; item_no < category[cat_no].number; ++item_no)
+ switch (category[cat_no].item_desc[item_no].value_type)
+ {
+ case string:
+ {
+ const char *val = nl_langinfo (
+ category[cat_no].item_desc[item_no].item_id);
+
+ if (val != NULL)
+ printf (" \"%s\",\n", val);
+ else
+ puts (" NULL,");
+ }
+ break;
+ case stringarray:
+ {
+ const char *val;
+ int cnt;
+
+ for (cnt = 0; cnt < category[cat_no].item_desc[item_no].max; ++cnt)
+ {
+ val = nl_langinfo (
+ category[cat_no].item_desc[item_no].item_id + cnt);
+
+ if (val != NULL)
+ printf (" \"%s\",\n", val);
+ else
+ puts (" NULL,");
+ }
+ }
+ break;
+ case byte:
+ {
+ const char *val = nl_langinfo (
+ category[cat_no].item_desc[item_no].item_id);
+
+ if (val != NULL)
+ printf (" \"\\%o\",\n",
+ *(unsigned char *) val ? : UCHAR_MAX);
+ else
+ puts (" NULL,");
+ }
+ break;
+ case bytearray:
+ {
+ const char *bytes = nl_langinfo (
+ category[cat_no].item_desc[item_no].item_id);
+
+ if (bytes != NULL)
+ {
+ fputs (" \"", stdout);
+ if (*bytes != '\0')
+ do
+ printf ("\\%o", *(unsigned char *) bytes++);
+ while (*bytes != '\0');
+ else
+ printf ("\\%o", UCHAR_MAX);
+
+ puts ("\",");
+ }
+ else
+ puts (" NULL,");
+ }
+ break;
+ default:
+ break;
+ }
+
+ puts (" }\n};");
+}
+
+/*
+ * Local Variables:
+ * mode:c
+ * c-basic-offset:2
+ * End:
+ */
diff --git a/locale/localedef.c b/locale/localedef.c
new file mode 100644
index 0000000000..c331e11888
--- /dev/null
+++ b/locale/localedef.c
@@ -0,0 +1,261 @@
+/* Copyright (C) 1995 Free Software Foundation, Inc.
+
+The GNU C Library is free software; you can redistribute it and/or
+modify it under the terms of the GNU Library General Public License as
+published by the Free Software Foundation; either version 2 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
+Library General Public License for more details.
+
+You should have received a copy of the GNU Library General Public
+License along with the GNU C Library; see the file COPYING.LIB. If
+not, write to the Free Software Foundation, Inc., 675 Mass Ave,
+Cambridge, MA 02139, USA. */
+
+#include <getopt.h>
+#include <libintl.h>
+#include <locale.h>
+#include <stdarg.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <unistd.h>
+#include <sys/stat.h>
+
+#include "localedef.h"
+
+/* The charmap file used. If none given DEFAULT_CHARMAP is used. */
+static char *charmap_file;
+
+/* If set output is always written, even when warning are given. */
+static int force_output;
+
+/* The input file name. */
+static char *input_file;
+
+/* Path leading to the destination directory for the produced files. */
+char *output_path;
+
+/* If this is defined be POSIX conform. */
+int posix_conformance;
+
+/* If not zero give a lot more messages. */
+int verbose;
+
+/* Long options. */
+static const struct option long_options[] =
+ {
+ { "charmap", required_argument, NULL, 'f' },
+ { "debug", no_argument, NULL, 'd' },
+ { "help", no_argument, NULL, 'h' },
+ { "force", no_argument, NULL, 'c' },
+ { "inputfile", required_argument, NULL, 'i' },
+ { "posix", no_argument, &posix_conformance, 1 },
+ { "verbose", no_argument, &verbose, 1},
+ { "version", no_argument, NULL, 'V' },
+ { NULL, 0, NULL, 0 }
+ };
+
+
+/* This is defined in error-msg.h. */
+extern int warning_cntr;
+
+
+/* Prototypes for local functions. */
+static void usage (int status) __attribute__ ((noreturn));
+static int construct_output_path (const char *path);
+
+int
+main(int argc, char *argv[])
+{
+ int optchar;
+ int cannot_write;
+ int do_help = 0;
+ int do_version = 0;
+
+ /* Set initial values for global varaibles. */
+ charmap_file = NULL;
+ force_output = 0;
+ input_file = 0;
+ posix_conformance = getenv ("POSIXLY_CORRECT") != NULL;
+ verbose = 0;
+
+ /* Set locale. Do not set LC_ALL because the other categories must
+ not be affected (acccording to POSIX.2). */
+ setlocale (LC_MESSAGES, "");
+ setlocale (LC_CTYPE, "");
+
+ /* Initialize the message catalog. */
+ textdomain (PACKAGE);
+
+ while ((optchar = getopt_long (argc, argv, "cdf:hi:vV", long_options, NULL))
+ != EOF)
+ switch (optchar)
+ {
+ case '\0':
+ break;
+ case 'c':
+ force_output = 1;
+ break;
+ case 'f':
+ if (charmap_file != NULL)
+ error (0, 0, gettext ("\"%s %s\" overwrites old option \"%s\""),
+ "-f", optarg, charmap_file);
+ charmap_file = optarg;
+ break;
+ case 'h':
+ do_help = 1;
+ break;
+ case 'i':
+ if (input_file != NULL)
+ error (0, 0, gettext ("\"%s %s\" overwrites old option \"%s\""),
+ "-i", optarg, input_file);
+ input_file = optarg;
+ break;
+ case 'v':
+ verbose = 1;
+ break;
+ case 'V':
+ do_version = 1;
+ break;
+ default:
+ usage (4);
+ break;
+ }
+
+ /* POSIX.2 requires to be verbose about missing characters in the
+ character map. */
+ verbose |= posix_conformance;
+
+ /* Version information is requested. */
+ if (do_version)
+ {
+ fprintf (stderr, "GNU %s %s\n", PACKAGE, VERSION);
+ exit (EXIT_SUCCESS);
+ }
+
+ /* Help is requested. */
+ if (do_help)
+ usage (0);
+
+ if (argc - optind != 1)
+ /* We need exactly one non-option parameter. */
+ usage (4);
+
+ /* The parameter describes the output path of the constructed files.
+ If the files cannot be written return a non-zero value. */
+ cannot_write = construct_output_path (argv[optind]);
+
+ /* Now that the parameters are processed we have to reset the local
+ ctype locale. (POSIX.2 4.35.5.2) */
+ setlocale (LC_CTYPE, "POSIX");
+
+ /* Look whether the system really allows locale definitions. */
+ if (sysconf (_SC_2_LOCALEDEF) < 0)
+ error (3, 0,
+ gettext ("warning: system does not define `_POSIX2_LOCALEDEF'"));
+
+ /* Process charmap file. */
+ charmap_read (charmap_file);
+
+ /* Now read the locale file. */
+ locfile_read (input_file);
+
+ /* Check all categories for consistency. */
+ categories_check ();
+
+ /* We are now able to write the data files. If warning were given we
+ do it only if it is explicitly requested (--force). */
+ if (warning_cntr == 0 || force_output != 0)
+ if (cannot_write != 0)
+ error (0, 0, gettext ("cannot write output file `%s': %s"),
+ output_path, strerror (cannot_write));
+ else
+ categories_write ();
+ else
+ error (0, 0,
+ gettext ("no output file produced because warning were issued"));
+
+ exit (EXIT_SUCCESS);
+}
+
+
+/* Display usage information and exit. */
+static void
+usage(int status)
+{
+ if (status != EXIT_SUCCESS)
+ fprintf (stderr, gettext ("Try `%s --help' for more information.\n"),
+ program_invocation_name);
+ else
+ printf(gettext ("\
+Usage: %s [OPTION]... name\n\
+Mandatory arguments to long options are mandatory for short options too.\n\
+ -c, --force create output even if warning messages have been issued\n\
+ -h, --help display this help and exit\n\
+ -V, --version output version information and exit\n\
+\n\
+ -i, --inputfile=FILE source definitions are found in FILE\n\
+ -f, --charmap=FILE symbolic character names defined in FILE\n\
+\n\
+ -v, --verbose print more messages\n\
+ --posix be strictly POSIX conform\n\
+\n\
+System's directory for character maps: %s\n\
+ locale files : %s\n\
+"), program_invocation_name, CHARMAP_PATH, LOCALE_PATH);
+
+ exit (status);
+}
+
+
+/* The parameter to localedef describes the output path. If it does
+ contain a '/' character it is a relativ path. Otherwise it names the
+ locale this definition is for. */
+static int
+construct_output_path (const char *path)
+{
+ int result = 0;
+
+ if (strchr (path, '/') == NULL)
+ {
+ /* This is a system path. */
+ int path_max_len = pathconf (LOCALE_PATH, _PC_PATH_MAX) + 1;
+ output_path = (char *) xmalloc (path_max_len);
+
+ snprintf (output_path, path_max_len, "%s/%s", LOCALE_PATH, path);
+ }
+ else
+ {
+ char *t;
+ /* This is a user path. */
+ output_path = malloc (strlen (path) + 2);
+ t = stpcpy (output_path, path);
+ *t = '\0';
+ }
+
+ if (euidaccess (output_path, W_OK) == -1)
+ /* Perhaps the directory does not exist now. Try to create it. */
+ if (errno == ENOENT)
+ {
+ if (mkdir (output_path, 0777) == -1)
+ result = errno;
+ }
+ else
+ result = errno;
+
+ if (result == 0)
+ strcat (output_path, "/");
+
+ return result;
+}
+
+/*
+ * Local Variables:
+ * mode:c
+ * c-basic-offset:2
+ * End:
+ */
diff --git a/locale/localedef.h b/locale/localedef.h
new file mode 100644
index 0000000000..5958a9c5d2
--- /dev/null
+++ b/locale/localedef.h
@@ -0,0 +1,196 @@
+/* Copyright (C) 1995 Free Software Foundation, Inc.
+
+The GNU C Library is free software; you can redistribute it and/or
+modify it under the terms of the GNU Library General Public License as
+published by the Free Software Foundation; either version 2 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
+Library General Public License for more details.
+
+You should have received a copy of the GNU Library General Public
+License along with the GNU C Library; see the file COPYING.LIB. If
+not, write to the Free Software Foundation, Inc., 675 Mass Ave,
+Cambridge, MA 02139, USA. */
+
+#ifndef _LOCALEDEF_H
+#define _LOCALEDEF_H 1
+
+#define __need_wchar_t
+#include <stddef.h>
+
+#include "config.h"
+
+#include "hash.h"
+
+
+/* Needed always. */
+#define MAX(a, b) ({typeof (a) _a = (a); typeof (b) _b = (b); \
+ _a > _b ? _a : _b; })
+#define MIN(a, b) ({typeof (a) _a = (a); typeof (b) _b = (b); \
+ _a < _b ? _a : _b; })
+
+/* Determine number of elements in ARR. */
+#define NELEMS(arr) ((sizeof (arr)) / (sizeof (arr[0])))
+
+/* I simply love these GCC features ... :) */
+#define NO_PAREN(arg, rest...) arg, ##rest
+
+
+/* The character set used in the locale is defined in a character map file.
+ The information of the file is stored in the following struct. */
+struct charmap
+ {
+ char *filename;
+ char *codeset_name;
+ int mb_cur_min;
+ int mb_cur_max;
+ char escape_char;
+ char comment_char;
+ hash_table table;
+ int hash_size;
+ int hash_layers;
+ };
+
+/* Data structure for representing charmap database. */
+extern struct charmap charmap_data;
+
+
+/* We can map the types of the entries into four categories. */
+enum value_type { none, string, stringarray, byte, bytearray, integer };
+
+/* Definition of the data structure which represents a category and its
+ items. */
+struct category
+{
+ int cat_id;
+ const char *name;
+ size_t number;
+ struct cat_item
+ {
+ int item_id;
+ const char *name;
+ enum { std, opt } status;
+ enum value_type value_type;
+ int min;
+ int max;
+ } *item_desc;
+ char **item_value;
+ void (*infct)(int);
+ void (*checkfct)(void);
+ int (*outfct)(void);
+ int filled;
+ char *copy_locale;
+};
+
+/* This a the structure which contains all information about all
+ categories. */
+extern struct category category[];
+
+
+/* The function used to load the contents of a charmap file into the
+ the global variable `charmap_data'. */
+void charmap_read (const char *filename);
+
+/* Find a character constant given by its name in the hash table. */
+static inline wchar_t find_char (const char *name, size_t len)
+{
+ wchar_t retval;
+ if (find_entry (&charmap_data.table, name, len, (void **) &retval) != 0)
+ return retval;
+ else
+ return -1;
+}
+
+/* Path to the directory the output files are written in. */
+extern char *output_path;
+
+/* If this is defined be POSIX conform. */
+extern int posix_conformance;
+
+/* If not zero give a lot more messages. */
+extern int verbose;
+
+/* This structure contains all informations about the status of of
+ reading the locale definition file. */
+struct locfile_data
+ {
+ const char *filename;
+ char escape_char;
+ char comment_char;
+ size_t bufsize;
+ char *buf;
+ char *strbuf;
+ size_t buf_ptr;
+ int continue_line;
+ size_t returned_tokens;
+ size_t line_no;
+ };
+
+/* The status variable. */
+extern struct locfile_data locfile_data;
+
+/* Open the locale definition file. */
+void locfile_open (const char *fname);
+
+/* Return the next token from the locale definition file. */
+int locfile_lex (char **token, int *token_len);
+/* Dito, but check for EOF. */
+int xlocfile_lex (char **token, int *token_len);
+
+/* Ignore the rest of the line. First TOKEN given if != 0. Warn about
+ anything other than end of line if WARN_FLAG nonzero. */
+void ignore_to_eol (int token, int warn_flag);
+
+/* Code a character with UTF-8 if the character map has multi-byte
+ characters. */
+int char_to_utf (char *buf, int char_val);
+
+
+/* Read the locale defintion file FNAME and fill the appropriate
+ data structures. */
+void locfile_read (const char *fname);
+
+/* Check categories for consistency. */
+void categories_check (void);
+
+/* Write out the binary representation of the category data. */
+void categories_write (void);
+
+
+/* Treat reading the LC_COLLATE definition. */
+void collate_input (int token);
+
+/* Treat reading the LC_CTYPE definition. */
+void ctype_input (int token);
+void ctype_check (void);
+int ctype_output (void);
+
+/* Treat reading the LC_MONETARY definition. */
+void monetary_check (void);
+
+/* Treat reading the LC_MESSAGES definition. */
+void messages_check (void);
+
+/* Treat reading the LC_NUMERIC definition. */
+void numeric_check (void);
+
+
+/* Print an error message, possibly with NLS. */
+void error (int status, int errnum, const char *format, ...)
+ __attribute__ ((format (printf, 3, 4)));
+
+/* Library functions. */
+void *xmalloc (size_t n);
+void *xcalloc (size_t n, size_t s);
+void *xrealloc (void *p, size_t n);
+
+/*
+ * Local Variables:
+ * mode:c
+ * c-basic-offset:2
+ * End:
+ */
+#endif /* _LOCALEDEF_H */
diff --git a/locale/locfile-lex.c b/locale/locfile-lex.c
new file mode 100644
index 0000000000..20e4f0f9cd
--- /dev/null
+++ b/locale/locfile-lex.c
@@ -0,0 +1,533 @@
+/* Copyright (C) 1995 Free Software Foundation, Inc.
+
+The GNU C Library is free software; you can redistribute it and/or
+modify it under the terms of the GNU Library General Public License as
+published by the Free Software Foundation; either version 2 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
+Library General Public License for more details.
+
+You should have received a copy of the GNU Library General Public
+License along with the GNU C Library; see the file COPYING.LIB. If
+not, write to the Free Software Foundation, Inc., 675 Mass Ave,
+Cambridge, MA 02139, USA. */
+
+#include <ctype.h>
+#include <langinfo.h>
+#include <libintl.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <unistd.h>
+
+#include "localedef.h"
+#include "token.h"
+
+
+/* Include the hashing table for the keywords. */
+const struct locale_keyword* in_word_set (register const char *str,
+ register int len);
+#include "keyword.h"
+
+
+/* Contains the status of reading the locale definition file. */
+struct locfile_data locfile_data;
+
+/* This is a flag used while collation input. This is the only place
+ where element names beside the ones defined in the character map are
+ allowed. There we must not give error messages. */
+int reject_new_char = 1;
+
+/* Prototypes for local functions. */
+static int get_char (void);
+
+
+#define LD locfile_data
+
+/* Opens the locale definition file and initializes the status data structure
+ for following calls of `locfile_lex'. */
+void
+locfile_open (const char *fname)
+{
+ if (fname == NULL)
+ /* We read from stdin. */
+ LD.filename = "<stdin>";
+ else
+ {
+ if (freopen (fname, "r", stdin) == NULL)
+ error (4, 0, gettext ("input file `%s' not found"), fname);
+ LD.filename = fname;
+ }
+
+ /* Set default values. */
+ LD.escape_char = '\\';
+ LD.comment_char = '#';
+
+ LD.bufsize = sysconf (_SC_LINE_MAX);
+ LD.buf = (char *) xmalloc (LD.bufsize);
+ LD.strbuf = (char *) xmalloc (LD.bufsize);
+
+ LD.buf_ptr = LD.returned_tokens = LD.line_no = 0;
+
+ /* Now sign that we want immediately read a line. */
+ LD.continue_line = 1;
+ LD.buf[LD.buf_ptr] = '\0';
+}
+
+
+int
+xlocfile_lex (char **token, int *token_len)
+{
+ int retval = locfile_lex (token, token_len);
+
+ if (retval == 0)
+ /* I.e. end of file. */
+ error (4, 0, gettext ("%s: unexpected end of file in locale defintion "
+ "file"), locfile_data.filename);
+
+ return retval;
+}
+
+int
+locfile_lex (char **token, int *token_len)
+{
+ int start_again;
+ int retval = 0;
+
+ do
+ {
+ int start_ptr;
+
+ start_again = 0;
+
+ /* Read the next line. Skip over empty lines and comments. */
+ if ((LD.buf[LD.buf_ptr] == '\0' && LD.continue_line != 0)
+ || LD.buf_ptr >= LD.bufsize
+ || (posix_conformance == 0 && LD.buf[LD.buf_ptr] == LD.comment_char))
+ do
+ {
+ size_t linelen;
+
+ LD.buf_ptr = 0;
+
+ if (fgets (LD.buf, LD.bufsize, stdin) == NULL)
+ {
+ /* This makes subsequent calls also return EOF. */
+ LD.buf[0] = '\0';
+ return 0;
+ }
+
+ /* Increment line number counter. */
+ ++LD.line_no;
+
+ /* We now have to look whether this line is continued and
+ whether it at all fits into our buffer. */
+ linelen = strlen (LD.buf);
+
+ if (linelen == LD.bufsize - 1)
+ /* The did not fit into the buffer. */
+ error (2, 0, gettext ("%s:%Zd: line too long; use "
+ "`getconf LINE_MAX' to get the maximum "
+ "line length"), LD.filename, LD.line_no);
+
+ /* Remove '\n' at end of line. */
+ if (LD.buf[linelen - 1] == '\n')
+ LD.buf[--linelen] = '\0';
+
+ if (linelen > 0 && LD.buf[linelen - 1] == LD.escape_char)
+ {
+ LD.buf[--linelen] = '\0';
+ LD.continue_line = 1;
+ }
+ else
+ LD.continue_line = 0;
+
+ while (isspace (LD.buf[LD.buf_ptr]))
+ ++LD.buf_ptr;
+
+ /* We are not so restrictive and allow white spaces before
+ a comment. */
+ if (posix_conformance == 0
+ && LD.buf[LD.buf_ptr] == LD.comment_char
+ && LD.buf_ptr != 0)
+ error (0, 0, gettext ("%s:%Zd: comment does not start in "
+ "column 1"), LD.filename, LD.line_no);
+ }
+ while (LD.buf[LD.buf_ptr] == '\0'
+ || LD.buf[LD.buf_ptr] == LD.comment_char);
+
+
+ /* Get information for return values. */
+ *token = LD.buf + LD.buf_ptr;
+ start_ptr = LD.buf_ptr;
+
+ /* If no further character is in the line this is the end of a logical
+ line. This information is needed in the parser. */
+ if (LD.buf[LD.buf_ptr] == '\0')
+ {
+ LD.buf_ptr = LD.bufsize;
+ retval = TOK_ENDOFLINE;
+ }
+ else if (isalpha (LD.buf[LD.buf_ptr]))
+ /* The token is an identifier. The POSIX standard does not say
+ what characters might be contained but offical POSIX locale
+ definition files contain beside alnum characters '_', '-' and
+ '+'. */
+ {
+ const struct locale_keyword *kw;
+
+ do
+ ++LD.buf_ptr;
+ while (isalnum (LD.buf[LD.buf_ptr]) || LD.buf[LD.buf_ptr] == '_'
+ || LD.buf[LD.buf_ptr] == '-' || LD.buf[LD.buf_ptr] == '+');
+
+ /* Look in table of keywords. */
+ kw = in_word_set (*token, LD.buf_ptr - start_ptr);
+ if (kw == NULL)
+ retval = TOK_IDENT;
+ else
+ {
+ if (kw->token_id == TOK_ESCAPE_CHAR
+ || kw->token_id == TOK_COMMENT_CHAR)
+ /* `escape_char' and `comment_char' are keywords for the
+ lexer. Do not give them to the parser. */
+ {
+ start_again = 1;
+
+ if (!isspace (LD.buf[LD.buf_ptr])
+ || (posix_conformance && LD.returned_tokens > 0))
+ error (0, 0, gettext ("%s:%Zd: syntax error in locale "
+ "definition file"),
+ LD.filename, LD.line_no);
+
+ do
+ ++LD.buf_ptr;
+ while (isspace (LD.buf[LD.buf_ptr]));
+
+ kw->token_id == TOK_ESCAPE_CHAR
+ ? LD.escape_char
+ : LD.comment_char = LD.buf[LD.buf_ptr++];
+
+ ignore_to_eol (0, posix_conformance);
+ }
+ else
+ /* It is one of the normal keywords. */
+ retval = kw->token_id;
+ }
+
+ *token_len = LD.buf_ptr - start_ptr;
+ }
+ else if (LD.buf[LD.buf_ptr] == '"')
+ /* Read a string. All symbolic character descriptions are expanded.
+ This has to be done in a local buffer because a simple symbolic
+ character like <A> may expand to upto 6 bytes. */
+ {
+ char *last = LD.strbuf;
+
+ ++LD.buf_ptr;
+ while (LD.buf[LD.buf_ptr] != '"')
+ {
+ int pre = LD.buf_ptr;
+ int char_val = get_char (); /* token, token_len); */
+
+ if (char_val == 0)
+ {
+ error (4, 0, gettext ("%s:%Zd: unterminated string at end "
+ "of line"), LD.filename, LD.line_no);
+ /* NOTREACHED */
+ }
+
+ if (char_val > 0)
+ /* Unknown characters are simply not stored. */
+ last += char_to_utf (last, char_val);
+ else
+ {
+ char tmp[LD.buf_ptr - pre + 1];
+ memcpy (tmp, &LD.buf[pre], LD.buf_ptr - pre);
+ tmp[LD.buf_ptr - pre] = '\0';
+ error (0, 0, gettext ("%s:%Zd: character `%s' not defined"),
+ LD.filename, LD.line_no, tmp);
+ }
+ }
+ if (LD.buf[LD.buf_ptr] != '\0')
+ ++LD.buf_ptr;
+
+ *last = '\0';
+ *token = LD.strbuf;
+ *token_len = last - LD.strbuf;
+ retval = TOK_STRING;
+ }
+ else if (LD.buf[LD.buf_ptr] == '.' && LD.buf[LD.buf_ptr + 1] == '.'
+ && LD.buf[LD.buf_ptr + 2] == '.')
+ {
+ LD.buf_ptr += 3;
+ retval = TOK_ELLIPSIS;
+ }
+ else if (LD.buf[LD.buf_ptr] == LD.escape_char)
+ {
+ char *endp;
+
+ ++LD.buf_ptr;
+ switch (LD.buf[LD.buf_ptr])
+ {
+ case 'x':
+ if (isdigit (LD.buf[++LD.buf_ptr]))
+ {
+ retval = strtol (&LD.buf[LD.buf_ptr], &endp, 16);
+ if (endp - (LD.buf + LD.buf_ptr) < 2 || retval > 255)
+ retval = 'x';
+ else
+ LD.buf_ptr = endp - LD.buf;
+ }
+ else
+ retval = 'x';
+ break;
+ case 'd':
+ if (isdigit (LD.buf[++LD.buf_ptr]))
+ {
+ retval = strtol (&LD.buf[LD.buf_ptr], &endp, 10);
+ if (endp - (LD.buf + LD.buf_ptr) < 2 || retval > 255)
+ retval = 'd';
+ else
+ LD.buf_ptr = endp - LD.buf;
+ }
+ else
+ retval = 'd';
+ break;
+ case '0'...'9':
+ retval = strtol (&LD.buf[LD.buf_ptr], &endp, 8);
+ if (endp - (LD.buf + LD.buf_ptr) < 2 || retval > 255)
+ retval = LD.buf[LD.buf_ptr++];
+ else
+ LD.buf_ptr = endp - LD.buf;
+ break;
+ case 'a':
+ retval = '\a';
+ ++LD.buf_ptr;
+ break;
+ case 'b':
+ retval = '\b';
+ ++LD.buf_ptr;
+ break;
+ case 'f':
+ retval = '\f';
+ ++LD.buf_ptr;
+ break;
+ case 'n':
+ retval = '\n';
+ ++LD.buf_ptr;
+ break;
+ case 'r':
+ retval = '\r';
+ ++LD.buf_ptr;
+ break;
+ case 't':
+ retval = '\t';
+ ++LD.buf_ptr;
+ break;
+ case 'v':
+ retval = '\v';
+ ++LD.buf_ptr;
+ break;
+ default:
+ retval = LD.buf[LD.buf_ptr++];
+ break;
+ }
+ }
+ else if (isdigit (LD.buf[LD.buf_ptr]))
+ {
+ char *endp;
+
+ *token_len = strtol (&LD.buf[LD.buf_ptr], &endp, 10);
+ LD.buf_ptr = endp - LD.buf;
+ retval = TOK_NUMBER;
+ }
+ else if (LD.buf[LD.buf_ptr] == '-' && LD.buf[LD.buf_ptr + 1] == '1')
+ {
+ LD.buf_ptr += 2;
+ retval = TOK_MINUS1;
+ }
+ else
+ {
+ int ch = get_char (); /* token, token_len); */
+ if (ch != -1)
+ {
+ *token_len = ch;
+ retval = TOK_CHAR;
+ }
+ else
+ retval = TOK_ILL_CHAR;
+ }
+
+ /* Ignore white space. */
+ while (isspace (LD.buf[LD.buf_ptr]))
+ ++LD.buf_ptr;
+ }
+ while (start_again != 0);
+
+ ++LD.returned_tokens;
+ return retval;
+}
+
+
+/* Code a character with UTF-8 if the character map has multi-byte
+ characters. */
+int
+char_to_utf (char *buf, int char_val)
+{
+ if (charmap_data.mb_cur_max == 1)
+ {
+ *buf++ = char_val;
+ return 1;
+ }
+ else
+ {
+/* The number of bits coded in each character. */
+#define CBPC 6
+ static struct coding_tab
+ {
+ int mask;
+ int val;
+ }
+ tab[] =
+ {
+ { 0x7f, 0x00 },
+ { 0x7ff, 0xc0 },
+ { 0xffff, 0xe0 },
+ { 0x1fffff, 0xf0 },
+ { 0x3ffffff, 0xf8 },
+ { 0x7fffffff, 0xfc },
+ { 0, }
+ };
+ struct coding_tab *t;
+ int c;
+ int cnt = 1;
+
+ for (t = tab; char_val > t->mask; ++t, ++cnt)
+ ;
+
+ c = cnt;
+
+ buf += cnt;
+ while (c > 1)
+ {
+ *--buf = 0x80 | (char_val & ((1 << CBPC) - 1));
+ char_val >>= CBPC;
+ --c;
+ }
+
+ *--buf = t->val | char_val;
+
+ return cnt;
+ }
+}
+
+
+/* Ignore rest of line upto ENDOFLINE token, starting with given token.
+ If WARN_FLAG is set warn about any token but ENDOFLINE. */
+void
+ignore_to_eol (int token, int warn_flag)
+{
+ if (token == TOK_ENDOFLINE)
+ return;
+
+ if (LD.buf[LD.buf_ptr] != '\0' && warn_flag)
+ error (0, 0, gettext ("%s:%Zd: trailing garbage at end of line"),
+ locfile_data.filename, locfile_data.line_no);
+
+ while (LD.continue_line)
+ {
+ LD.continue_line = 0;
+
+ /* Increment line number counter. */
+ ++LD.line_no;
+
+ if (fgets (LD.buf, LD.bufsize, stdin) != NULL)
+ {
+ /* We now have to look whether this line is continued and
+ whether it at all fits into our buffer. */
+ int linelen = strlen (LD.buf);
+
+ if (linelen == LD.bufsize - 1)
+ /* The did not fit into the buffer. */
+ error (2, 0, gettext ("%s:%Zd: line too long; use `getconf "
+ "LINE_MAX' to get the current maximum "
+ "line length"), LD.filename, LD.line_no);
+
+ /* Remove '\n' at end of line. */
+ if (LD.buf[linelen - 1] == '\n')
+ --linelen;
+
+ if (LD.buf[linelen - 1] == LD.escape_char)
+ LD.continue_line = 1;
+ }
+ }
+
+ /* This causes to begin the next line. */
+ LD.buf_ptr = LD.bufsize;
+}
+
+
+/* Return the value of the character at the beginning of the input buffer.
+ Symbolic character constants are expanded. */
+static int
+get_char (void)
+{
+ if (LD.buf[LD.buf_ptr] == '<')
+ /* This is a symbolic character name. */
+ {
+ int char_val;
+ char *startp = LD.buf + (++LD.buf_ptr);
+ char *endp = startp;
+
+ while (LD.buf[LD.buf_ptr] != '>' && isprint (LD.buf[LD.buf_ptr]))
+ {
+ if (LD.buf[LD.buf_ptr] == '\0'
+ || (LD.buf[LD.buf_ptr] == LD.escape_char
+ && LD.buf[++LD.buf_ptr] == '\0'))
+ break;
+
+ *endp++ = LD.buf[LD.buf_ptr++];
+ }
+
+ if (LD.buf[LD.buf_ptr] != '>' && LD.buf[LD.buf_ptr] == '\0')
+ {
+ error (0, 0, gettext ("%s:%Zd: end of line in character symbol"),
+ LD.filename, LD.line_no);
+
+ if (startp == endp)
+ return -1;
+ }
+ else
+ ++LD.buf_ptr;
+
+ char_val = find_char (startp, endp - startp);
+ if (char_val == -1 && verbose != 0 && reject_new_char != 0)
+ {
+ /* Locale defintions are often given very general. Missing
+ characters are only reported when explicitely requested. */
+ char tmp[endp - startp + 3];
+
+ tmp[0] = '<';
+ memcpy (tmp + 1, startp, endp - startp);
+ tmp[endp - startp + 1] = '>';
+ tmp[endp - startp + 2] = '\0';
+
+ error (0, 0, gettext ("%s:%Zd: character `%s' not defined"),
+ LD.filename, LD.line_no, tmp);
+ }
+
+ return char_val;
+ }
+ else
+ return (int) LD.buf[LD.buf_ptr++];
+}
+
+/*
+ * Local Variables:
+ * mode:c
+ * c-basic-offset:2
+ * End:
+ */
diff --git a/locale/locfile-parse.c b/locale/locfile-parse.c
new file mode 100644
index 0000000000..15e25c7afe
--- /dev/null
+++ b/locale/locfile-parse.c
@@ -0,0 +1,820 @@
+/* Copyright (C) 1995 Free Software Foundation, Inc.
+
+The GNU C Library is free software; you can redistribute it and/or
+modify it under the terms of the GNU Library General Public License as
+published by the Free Software Foundation; either version 2 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
+Library General Public License for more details.
+
+You should have received a copy of the GNU Library General Public
+License along with the GNU C Library; see the file COPYING.LIB. If
+not, write to the Free Software Foundation, Inc., 675 Mass Ave,
+Cambridge, MA 02139, USA. */
+
+#include <assert.h>
+#include <dirent.h>
+#include <fcntl.h>
+#include <langinfo.h>
+#include <libintl.h>
+#include <limits.h>
+#include <obstack.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <unistd.h>
+#include <sys/stat.h>
+#include <sys/uio.h>
+
+#include "localedef.h"
+#include "localeinfo.h"
+#include "token.h"
+
+/* We don't have these constants defined because we don't use them. Give
+ default values. */
+#define CTYPE_MB_CUR_MIN 0
+#define CTYPE_MB_CUR_MAX 0
+#define CTYPE_HASH_SIZE 0
+#define CTYPE_HASH_LAYERS 0
+#define CTYPE_CLASS 0
+#define CTYPE_TOUPPER_EB 0
+#define CTYPE_TOLOWER_EB 0
+#define CTYPE_TOUPPER_EL 0
+#define CTYPE_TOLOWER_EL 0
+
+
+/* We have all categories defined in `categories.def'. Now construct
+ the description and data structure used for all categories. */
+#define DEFINE_CATEGORY(category, name, items, postload, in, check, out) \
+ struct cat_item category##_desc[] = \
+ { \
+ NO_PAREN items \
+ }; \
+ \
+ char *category##_values[NELEMS (category##_desc) - 1] = { NULL, };
+#include "categories.def"
+#undef DEFINE_CATEGORY
+
+struct category category[] =
+ {
+#define DEFINE_CATEGORY(category, name, items, postload, in, check, out) \
+ [category] = { _NL_NUM_##category, name, NELEMS (category##_desc) - 1, \
+ category##_desc, category##_values, in, check, out },
+#include "categories.def"
+#undef DEFINE_CATEGORY
+ };
+#define NCATEGORIES NELEMS (category)
+
+
+#define SYNTAX_ERROR \
+ error (0, 0, gettext ("%s:%Zd: syntax error in locale definition file"), \
+ locfile_data.filename, locfile_data.line_no)
+
+
+/* Prototypes for local functions. */
+static int get_byte (char *byte_ptr);
+static char *is_locale_name (int cat_no, const char *str, int len);
+
+
+/* Read a locale definition file FILE. The format is defined in
+ POSIX.2 2.5.3. */
+void
+locfile_read (const char *fname)
+{
+ /* Pointer to text of last token. */
+ char *ptr;
+ /* Length of last token (or if NUMBER the value itself). */
+ int len;
+ /* The last returned token. */
+ int token;
+ /* For error correction we remember whether the last token was correct. */
+ int correct_token = 1;
+
+ /* Open the desired input file on stdin. */
+ locfile_open (fname);
+
+ while ((token = locfile_lex (&ptr, &len)) != 0)
+ {
+ int cat_no;
+
+ for (cat_no = 0; cat_no < NCATEGORIES; ++cat_no)
+ if (token == category[cat_no].cat_id)
+ break;
+
+ if (cat_no >= NCATEGORIES)
+ /* A syntax error occured. No valid category defintion starts. */
+ {
+ if (correct_token != 0)
+ error (0, 0, gettext ("%s:%Zd: locale category start expected"),
+ locfile_data.filename, locfile_data.line_no);
+
+ /* To prevent following errors mark as error case. */
+ correct_token = 0;
+
+ /* Synchronization point is the beginning of a new category.
+ Overread all line upto this silently. */
+ ignore_to_eol (0, 0);
+ continue;
+ }
+
+ /* Rest of the line should be empty. */
+ ignore_to_eol (0, 1);
+
+ /* Perhaps these category is already specified. We simply give a
+ warning and overwrite the values. */
+ if (category[cat_no].filled != 0)
+ error (0, 0, gettext ("%s:%Zd: multiple definition of locale "
+ "category %s"), locfile_data.filename,
+ locfile_data.line_no, category[cat_no].name);
+
+ /* We read the first token because this could be the copy statement. */
+ token = xlocfile_lex (&ptr, &len);
+
+ if (token == TOK_COPY)
+ /* Copying the definitions from an existing locale is requested. */
+ {
+ char *str;
+
+ /* Get the name of the locale to copy from. */
+ token = xlocfile_lex (&ptr, &len);
+ if (token != TOK_IDENT && token != TOK_STRING)
+ /* No name, then mark error and ignore everything upto next
+ start of an category section. */
+ {
+ /* To prevent following errors mark as error case. */
+ correct_token = 0;
+
+ /* Synchronization point is the beginning of a new category.
+ Overread all line upto this silently. */
+ ignore_to_eol (0, 0);
+ }
+ else if ((str = is_locale_name (cat_no, ptr, len)) != NULL)
+ /* Yes the name really names an existing locale file. We are
+ returned the complete file name. Store it so that we can
+ copy it in the output phase. */
+ {
+ category[cat_no].copy_locale = str;
+ category[cat_no].filled = 1;
+
+ ignore_to_eol (0, 1);
+ }
+ else
+ /* No, the name does not address a valid locale file. Mark
+ error case and ignore rest of category. */
+ {
+ char tmp[len + 1];
+ memcpy (tmp, ptr, len);
+ tmp[len] = '\0';
+ error (0, 0, gettext ("%s:%Zd: invalid locale `%s' in copy "
+ "statement"), locfile_data.filename,
+ locfile_data.line_no, tmp);
+ correct_token = 0;
+ ignore_to_eol (0, 0);
+ }
+
+ /* This should END as the next token. */
+ token = xlocfile_lex (&ptr, &len);
+
+ if (token == TOK_END)
+ /* This is the end of the category. */
+ {
+ token = xlocfile_lex (&ptr, &len);
+
+ if (token != category[cat_no].cat_id)
+ /* Wrong category name after END. */
+ {
+ error (0, 0, gettext ("%s:%Zd: category `%s' does not "
+ "end with `END %s'"),
+ locfile_data.filename, locfile_data.line_no,
+ category[cat_no].name, category[cat_no].name);
+ ignore_to_eol (0, 0);
+ }
+ else
+ ignore_to_eol (0, 1);
+
+ correct_token = 1;
+ }
+ else
+ /* No END following copy. Give error while not in error case. */
+ {
+ if (correct_token != 0)
+ error (0, 0, gettext ("%s:%Zd: `copy' must be sole rule"),
+ locfile_data.filename, locfile_data.line_no);
+ correct_token = 0;
+ ignore_to_eol (0, 0);
+ }
+
+ continue;
+ }
+
+ /* Now it's time to mark as mentioned in the locale file. */
+ category[cat_no].filled = 1;
+
+ if (category[cat_no].infct != NULL)
+ /* The category needs a special input handling. */
+ {
+ category[cat_no].infct(token);
+ continue;
+ }
+
+ /* Now process the given items. */
+ while (1)
+ {
+ int item_no;
+
+ if (token == TOK_END)
+ /* This is the end of the category. */
+ {
+ token = xlocfile_lex (&ptr, &len);
+
+ if (token != category[cat_no].cat_id)
+ {
+ error (0, 0, gettext ("%s:%Zd: category `%s' does not end "
+ "with `END %s'"),
+ locfile_data.filename, locfile_data.line_no,
+ category[cat_no].name, category[cat_no].name);
+ ignore_to_eol (0, 0);
+ }
+ else
+ ignore_to_eol (0, 1);
+
+ /* Start next category. */
+ break;
+ }
+
+ /* All other lines should describe valid items of the category. */
+ for (item_no = 0; item_no < category[cat_no].number; ++item_no)
+ if (category[cat_no].item_desc[item_no].item_id == token)
+ break;
+
+ if (item_no >= category[cat_no].number)
+ /* This is not a valid item of the category. */
+ {
+ SYNTAX_ERROR;
+ ignore_to_eol (0, 0);
+
+ token = xlocfile_lex (&ptr, &len);
+
+ /* And process next item. */
+ continue;
+ }
+
+ /* Test whether already a value is defined. */
+ if (category[cat_no].item_value[item_no] != NULL)
+ error (0, 0, gettext ("%s:%Zd: category item `%s' already "
+ "defined"),
+ locfile_data.filename, locfile_data.line_no,
+ category[cat_no].item_desc[item_no].name);
+
+ switch (category[cat_no].item_desc[item_no].value_type)
+ {
+ case string:
+ /* Get next token. This is the argument to the item. */
+ token = xlocfile_lex (&ptr, &len);
+
+ if (token != TOK_STRING)
+ SYNTAX_ERROR;
+ else
+ category[cat_no].item_value[item_no] = strdup (ptr);
+ ignore_to_eol (0, ptr != NULL);
+ break;
+ case stringarray:
+ /* This is a difficult case. The number of strings in
+ the array may vary. But for now its only necessary
+ with ALT_DIGITS from LC_TIME. This item is the last
+ so be can solve it by storing the number of string in
+ the first place and the string indeces following
+ that. */
+ {
+ int cnt;
+ char **buffer;
+ if (category[cat_no].item_value[item_no] != NULL)
+ buffer = (char **) category[cat_no].item_value[item_no];
+ else
+ buffer = (char **) xmalloc (
+ sizeof (char *) * category[cat_no].item_desc[item_no].max);
+
+ category[cat_no].item_value[item_no] = (char *) buffer;
+
+ /* As explained we may need a place to store the real number
+ of strings. */
+ if (category[cat_no].item_desc[item_no].min
+ != category[cat_no].item_desc[item_no].max)
+ ++buffer;
+
+ cnt = 0;
+ do
+ {
+ token = xlocfile_lex (&ptr, &len);
+ if (token != TOK_STRING)
+ {
+ SYNTAX_ERROR;
+ break;
+ }
+
+ if (cnt >= category[cat_no].item_desc[item_no].max)
+ {
+ error (0, 0, gettext ("%s:%Zd: too many elements "
+ "for item `%s`"),
+ locfile_data.filename, locfile_data.line_no,
+ category[cat_no].item_desc[item_no].name);
+ break;
+ }
+
+ buffer[cnt++] = strdup (ptr);
+
+ token = locfile_lex (&ptr, &len);
+ }
+ while (token == TOK_CHAR && len == ';');
+
+ ignore_to_eol (token, ptr != NULL);
+
+ if (cnt < category[cat_no].item_desc[item_no].min)
+ error (0, 0, gettext ("%s:%Zd: too few elements for item "
+ "`%s'"),
+ locfile_data.filename, locfile_data.line_no,
+ category[cat_no].item_desc[item_no].name);
+
+ if (category[cat_no].item_desc[item_no].min
+ != category[cat_no].item_desc[item_no].max)
+ *(int *) category[cat_no].item_value[item_no] = cnt;
+ }
+ break;
+ case byte:
+ {
+ int ok;
+ category[cat_no].item_value[item_no] = (char *) xmalloc (
+ __alignof__ (char));
+ ok = get_byte (category[cat_no].item_value[item_no]);
+ ignore_to_eol (0, ok);
+ }
+ break;
+ case bytearray:
+ {
+ char *buffer;
+ int maxsize;
+ int cnt;
+ char byte;
+ int ok;
+
+ buffer = (char *) xmalloc ((maxsize = 30));
+ cnt = 0;
+
+ while ((ok = get_byte (&byte)))
+ {
+ if (cnt >= maxsize)
+ buffer = (char *) xmalloc ((maxsize *= 2));
+
+ buffer[cnt++] = byte;
+
+ token = locfile_lex (&ptr, &len);
+ if (token != TOK_CHAR || len != ';')
+ break;
+ }
+
+ buffer[cnt] = '\0';
+ category[cat_no].item_value[item_no] = buffer;
+ ignore_to_eol (token, ok);
+ }
+ break;
+ default:
+ error (5, 0, gettext ("internal error in %s, line %u"),
+ __FUNCTION__, __LINE__);
+ /* NOTREACHED */
+ }
+
+ /* Get next token. */
+ token = xlocfile_lex (&ptr, &len);
+ } /* while (1) */
+ }
+}
+
+
+/* Check given values for categories for consistency. */
+void
+categories_check (void)
+{
+ int cat_no;
+
+ for (cat_no = 0; cat_no < NCATEGORIES; ++cat_no)
+ if (category[cat_no].copy_locale == NULL)
+ if (category[cat_no].filled != 0)
+ if (category[cat_no].checkfct)
+ category[cat_no].checkfct();
+ else
+ {
+ int item_no;
+
+ for (item_no = 0; item_no < category[cat_no].number; ++item_no)
+ if (category[cat_no].item_value[item_no] == NULL)
+ {
+ int errcode;
+
+ /* If the item is defined in the standard is it an error to
+ have it not defined. */
+ errcode = category[cat_no].item_desc[item_no].status == std
+ ? 5 : 0;
+
+ error (errcode, 0, gettext ("item `%s' of category `%s' "
+ "undefined"),
+ category[cat_no].item_desc[item_no].name,
+ category[cat_no].name);
+ }
+ }
+ else
+ error (0, 0, gettext ("category `%s' not defined"),
+ category[cat_no].name);
+}
+
+
+/* Write out the binary representation of the category data which can be
+ loaded by setlocale(1). */
+void
+categories_write (void)
+{
+ struct locale_file
+ {
+ int magic;
+ int n;
+ int idx[0];
+ } *data;
+ struct obstack obstk;
+ int cat_no;
+
+#define obstack_chunk_alloc xmalloc
+#define obstack_chunk_free free
+ obstack_init (&obstk);
+
+ for (cat_no = 0; cat_no < NCATEGORIES; ++cat_no)
+ {
+ int result = 0;
+
+ if (category[cat_no].copy_locale != NULL)
+ /* Simply copy the addressed locale file of the specified
+ category. Please note that this is tried before the distinction
+ between categories which need special handling is made. */
+ {
+ int source;
+
+ /* Open source file. */
+ source = open (category[cat_no].copy_locale, O_RDONLY);
+ if (source < 0)
+ error (0, 0, gettext ("cannot copy locale definition file `%s'"),
+ category[cat_no].copy_locale);
+ else
+ {
+ /* Construct file name of output file and open for writing. */
+ char path[strlen (output_path)
+ + strlen(category[cat_no].name) + 1];
+ int dest;
+ char *t;
+
+ t = stpcpy (path, output_path);
+ strcpy (t, category[cat_no].name);
+
+ dest = creat (path, 0666);
+ if (dest == -1)
+ error (0, 0, gettext ("cannot open output file `%s': %m"),
+ path);
+ else
+ {
+ char buffer[BUFSIZ];
+ int size;
+
+ /* Copy the files. */
+ do
+ {
+ size = read (source, buffer, BUFSIZ);
+ write (dest, buffer, size);
+ }
+ while (size > 0);
+
+ close (dest);
+
+ /* Show success. */
+ puts (category[cat_no].name);
+ }
+ close (source);
+ }
+
+ /* Next category. */
+ continue;
+ }
+
+ if (category[cat_no].outfct)
+ result = category[cat_no].outfct();
+ else
+ {
+ char *path, *t;
+ int fd;
+ struct iovec *iov;
+ int item_no, len, slen, cnt;
+ int elems = 0;
+
+ /* Count number of elements. */
+ for (item_no = 0; item_no < category[cat_no].number; ++item_no)
+ {
+ switch (category[cat_no].item_desc[item_no].value_type)
+ {
+ case string:
+ case byte:
+ case bytearray:
+ ++elems;
+ break;
+ case stringarray:
+ elems += category[cat_no].item_desc[item_no].max;
+ break;
+ default:
+ error (5, 0, gettext ("internal error in %s, line %u"),
+ __FUNCTION__, __LINE__);
+ /* NOTREACHED */
+ }
+ }
+
+ /* We now have the number of elements. We build the structure
+ and a helper structure for writing all out. */
+ len = sizeof (struct locale_file) + elems * sizeof (int);
+ data = obstack_alloc (&obstk, len);
+ iov = obstack_alloc (&obstk, (elems + 1) * sizeof (struct iovec));
+
+ data->magic = LIMAGIC (cat_no);
+ data->n = elems;
+ iov[0].iov_base = data;
+ iov[0].iov_len = len;
+
+ cnt = 0;
+ for (item_no = 0; item_no < category[cat_no].number; ++item_no)
+ if (category[cat_no].item_value[item_no] == NULL)
+ {
+ switch (category[cat_no].item_desc[item_no].value_type)
+ {
+ case string:
+ case byte:
+ case bytearray:
+ data->idx[cnt] = len;
+ ++len; /* We reserve one single byte for this entry. */
+ iov[1 + cnt].iov_base = (char *) "";
+ iov[1 + cnt].iov_len = 1;
+ ++cnt;
+ break;
+ case stringarray:
+ {
+ int max;
+ int nstr;
+
+ max = category[cat_no].item_desc[item_no].max;
+
+ for (nstr = 0; nstr < max; ++nstr)
+ {
+ data->idx[cnt] = len;
+ ++len;
+ iov[1 + cnt].iov_base = (char *) "";
+ iov[1 + cnt].iov_len = 1;
+ ++cnt;
+ }
+ }
+ }
+ }
+ else
+ switch (category[cat_no].item_desc[item_no].value_type)
+ {
+ case string:
+ case bytearray:
+ data->idx[cnt] = len;
+ slen = strlen (category[cat_no].item_value[item_no]) + 1;
+ len += slen;
+ iov[1 + cnt].iov_base = category[cat_no].item_value[item_no];
+ iov[1 + cnt].iov_len = slen;
+ ++cnt;
+ break;
+ case byte:
+ data->idx[cnt] = len;
+ slen = 1;
+ len += slen;
+ iov[1 + cnt].iov_base = category[cat_no].item_value[item_no];
+ iov[1 + cnt].iov_len = slen;
+ ++cnt;
+ break;
+ case stringarray:
+ {
+ int nstr, nact;
+ char **first;
+
+ if (category[cat_no].item_desc[item_no].min
+ == category[cat_no].item_desc[item_no].max)
+ {
+ nstr = category[cat_no].item_desc[item_no].min;
+ first = (char **) category[cat_no].item_value[item_no];
+ }
+ else
+ {
+ nstr = *(int *) category[cat_no].item_value[item_no];
+ first =
+ ((char **) category[cat_no].item_value[item_no]) + 1;
+ }
+ nact = nstr;
+ while (nstr > 0)
+ {
+ data->idx[cnt] = len;
+ if (*first != NULL)
+ {
+ slen = strlen (*first) + 1;
+ iov[1 + cnt].iov_base = first;
+ }
+ else
+ {
+ slen = 1;
+ iov[1 + cnt].iov_base = (char *) "";
+ }
+ len += slen;
+ iov[1 + cnt].iov_len = slen;
+ ++cnt;
+ ++first;
+ --nstr;
+ }
+ while (nact < category[cat_no].item_desc[item_no].max)
+ {
+ data->idx[cnt] = len;
+ len += 1;
+ iov[1 + cnt].iov_base = (char *) "";
+ iov[1 + cnt].iov_len = 1;
+ ++cnt;
+ ++nact;
+ }
+ }
+ break;
+ default:
+ /* Cannot happen. */
+ break;
+ }
+ assert (cnt <= elems);
+
+ /* Construct the output filename from the argument given to
+ localedef on the command line. */
+ path = (char *) obstack_alloc (&obstk, strlen (output_path) +
+ strlen (category[cat_no].name) + 1);
+ t = stpcpy (path, output_path);
+ strcpy (t, category[cat_no].name);
+
+ fd = creat (path, 0666);
+ if (fd == -1)
+ {
+ error (0, 0, gettext ("cannot open output file `%s': %m"), path);
+ result = 1;
+ }
+ else
+ {
+ if (writev (fd, iov, cnt + 1) == -1)
+ {
+ error (0, 0, gettext ("cannot write output file `%s': %m"),
+ path);
+ result = 1;
+ }
+
+if (elems==0) write(fd, &elems, 1);
+
+ close (fd);
+ }
+ /* The old data is not needed anymore, but keep the obstack
+ intact. */
+ obstack_free (&obstk, data);
+ }
+
+ if (result == 0)
+ puts (category[cat_no].name);
+ }
+ /* Now the whole obstack can be removed. */
+ obstack_free (&obstk, NULL);
+}
+
+
+/* Get the representation of a number. This is a positive integer or
+ the number -1 which is handled as a special symbol by the scanner. */
+static int
+get_byte (char *byte_ptr)
+{
+ int token;
+ char *ptr;
+ int len;
+
+ token = locfile_lex (&ptr, &len);
+ if (token != TOK_NUMBER && token != TOK_MINUS1)
+ /* None of the valid number format. */
+ {
+ error (0, 0, gettext ("%s:%Zd: number expected"),
+ locfile_data.filename, locfile_data.line_no);
+ *byte_ptr = 0;
+ return 0;
+ }
+
+ if (token == TOK_MINUS1)
+ {
+ *byte_ptr = CHAR_MAX;
+ return 1;
+ }
+
+ if (len > CHAR_MAX)
+ /* The value of the numbers has to be less than CHAR_MAX. This is
+ ok for the information they have to express. */
+ {
+ error (0, 0, gettext ("%s:%Zd: invalid number"),
+ locfile_data.filename, locfile_data.line_no);
+ *byte_ptr = 0;
+ return 0;
+ }
+
+ *byte_ptr = len;
+ return 1;
+}
+
+
+/* Test whether the string STR with length LEN is the name of an existing
+ locale and whether a file for category CAT_NO is found in this directory.
+ This categories are looked for in the system locale definition file
+ directory.
+ Return the complete file name for the category file. */
+static char *
+is_locale_name (int cat_no, const char *str, int len)
+{
+ static char **locale_names = NULL;
+ static int max_count = 0;
+ static int locale_count = 0;
+ int cnt, exist, fd;
+ char *fname;
+ struct stat st;
+
+ if (locale_names == NULL)
+ /* Read in the list of all available locales. */
+ {
+ DIR *dir;
+ struct dirent *dirent;
+
+ /* LOCALE_NAMES is not NULL anymore, but LOCALE_COUNT == 0. */
+ ++locale_names;
+
+ dir = opendir (LOCALE_PATH);
+ if (dir == NULL)
+ {
+ error (1, errno, gettext ("cannot read locale directory `%s'"),
+ LOCALE_PATH);
+
+ return NULL;
+ }
+
+ /* Now we can look for all files in the directory. */
+ while ((dirent = readdir (dir)) != NULL)
+ if (strcmp (dirent->d_name, ".") != 0
+ && strcmp (dirent->d_name, "..") != 0)
+ {
+ if (max_count == 0)
+ locale_names = (char **) xmalloc ((max_count = 10)
+ * sizeof (char *));
+ else
+ locale_names = (char **) xrealloc (locale_names,
+ (max_count *= 2)
+ * sizeof (char *));
+ locale_names[locale_count++] = strdup (dirent->d_name);
+ }
+ closedir (dir);
+ }
+
+ for (cnt = 0; cnt < locale_count; ++cnt)
+ if (strncmp (str, locale_names[cnt], len) == 0
+ && locale_names[cnt][len] == '\0')
+ break;
+
+ if (cnt >= locale_count)
+ return NULL;
+
+ /* Now search for this specific locale file. */
+ asprintf (&fname, "%s/%s/%s", LOCALE_PATH, locale_names[cnt],
+ category[cat_no].name);
+
+ fd = open (fname, O_RDONLY);
+ if (fd < 0)
+ {
+ free (fname);
+ return NULL;
+ }
+
+ exist = fstat (fd, &st);
+ close (fd);
+
+ if (exist < 0)
+ {
+ free (fname);
+ return NULL;
+ }
+
+ return fname;
+}
+
+/*
+ * Local Variables:
+ * mode:c
+ * c-basic-offset:2
+ * End:
+ */
diff --git a/locale/messages.c b/locale/messages.c
new file mode 100644
index 0000000000..842e5ef5b2
--- /dev/null
+++ b/locale/messages.c
@@ -0,0 +1,76 @@
+/* Copyright (C) 1995 Free Software Foundation, Inc.
+
+The GNU C Library is free software; you can redistribute it and/or
+modify it under the terms of the GNU Library General Public License as
+published by the Free Software Foundation; either version 2 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
+Library General Public License for more details.
+
+You should have received a copy of the GNU Library General Public
+License along with the GNU C Library; see the file COPYING.LIB. If
+not, write to the Free Software Foundation, Inc., 675 Mass Ave,
+Cambridge, MA 02139, USA. */
+
+#include <langinfo.h>
+#include <libintl.h>
+#include <locale.h>
+#include <stdio.h>
+#include <regex.h>
+
+#include "localedef.h"
+
+/* These are defined in locfile-parse.c. */
+extern struct cat_item LC_MESSAGES_desc[];
+extern char *LC_MESSAGES_values[];
+
+void
+messages_check(void)
+{
+ int item_no;
+
+ /* First general check for existence. */
+ for (item_no = 0; item_no < category[LC_MESSAGES].number; ++item_no)
+ if (LC_MESSAGES_values[item_no] == NULL)
+ {
+ int errcode;
+
+ errcode = LC_MESSAGES_desc[item_no].status == std ? 5 : 0;
+
+ error (errcode, 0, gettext ("item `%s' of category `%s' undefined"),
+ LC_MESSAGES_desc[item_no].name, "LC_MESSAGES");
+ }
+ else
+ {
+ /* Some fields need special tests. */
+ if (LC_MESSAGES_desc[item_no].item_id == YESEXPR
+ || LC_MESSAGES_desc[item_no].item_id == NOEXPR)
+ /* The expression has to be a POSIX extended regular expression. */
+ {
+ regex_t re;
+ int result;
+
+ result = regcomp (&re, LC_MESSAGES_values[item_no], REG_EXTENDED);
+
+ if (result != 0)
+ {
+ char errbuf[BUFSIZ];
+
+ (void) regerror (result, &re, errbuf, BUFSIZ);
+ error (0, 0, gettext ("no correct regular expression for "
+ "item `%s' in category `%s': %s"),
+ LC_MESSAGES_desc[item_no].name, "LC_MESSAGES", errbuf);
+ }
+ }
+ }
+}
+
+/*
+ * Local Variables:
+ * mode:c
+ * c-basic-offset:2
+ * End:
+ */
diff --git a/locale/monetary.c b/locale/monetary.c
new file mode 100644
index 0000000000..2683eb2bf0
--- /dev/null
+++ b/locale/monetary.c
@@ -0,0 +1,132 @@
+/* Copyright (C) 1995 Free Software Foundation, Inc.
+
+The GNU C Library is free software; you can redistribute it and/or
+modify it under the terms of the GNU Library General Public License as
+published by the Free Software Foundation; either version 2 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
+Library General Public License for more details.
+
+You should have received a copy of the GNU Library General Public
+License along with the GNU C Library; see the file COPYING.LIB. If
+not, write to the Free Software Foundation, Inc., 675 Mass Ave,
+Cambridge, MA 02139, USA. */
+
+#include <langinfo.h>
+#include <libintl.h>
+#include <stdlib.h>
+#include <string.h>
+
+#include "localedef.h"
+#include "token.h"
+
+
+/* The content iof the field int_curr_symbol has to be taken from
+ ISO-4217. We test for correct values. */
+#define DEFINE_INT_CURR(str) str,
+static const char *const valid_int_curr[] =
+ {
+# include "iso-4217.def"
+ };
+#define NVALID_INT_CURR ((sizeof (valid_int_curr) \
+ / sizeof (valid_int_curr[0])))
+#undef DEFINE_INT_CURR
+
+
+/* These are defined in locfile-parse.c. */
+extern struct cat_item LC_MONETARY_desc[];
+extern char *LC_MONETARY_values[];
+
+static int _curr_strcmp(const char *s1, const char **s2);
+
+
+
+void
+monetary_check(void)
+{
+ int item_no, val;
+
+ for (item_no = 0; LC_MONETARY_desc[item_no].item_id != 0; ++item_no)
+ /* Test whether the entry has been defined. Byte values are simply
+ stored. */
+ if (LC_MONETARY_values[item_no] == NULL)
+ {
+ int errcode;
+
+ errcode = LC_MONETARY_desc[item_no].status = std ? 5 : 0;
+
+ error (errcode, 0, gettext ("item `%s' of category `%s' undefined"),
+ LC_MONETARY_desc[item_no].name, "LC_MONETARY");
+ }
+ else
+ switch (LC_MONETARY_desc[item_no].item_id)
+ {
+ case INT_CURR_SYMBOL:
+ if (strlen (LC_MONETARY_values[item_no]) != 4)
+ error (0, 0,
+ gettext ("item `%s' of category `%s' has wrong length"),
+ LC_MONETARY_desc[item_no].name, "LC_MONETARY");
+ else if (bsearch (LC_MONETARY_values[item_no], valid_int_curr,
+ NVALID_INT_CURR, sizeof (char *),
+ (comparison_fn_t) _curr_strcmp) == NULL)
+ error (0, 0, gettext ("item `%s' does not correspond to any "
+ "valid name in ISO-4217"),
+ LC_MONETARY_desc[item_no].name);
+ break;
+ case P_CS_PRECEDES:
+ case P_SEP_BY_SPACE:
+ case N_CS_PRECEDES:
+ case N_SEP_BY_SPACE:
+ case P_SIGN_POSN:
+ case N_SIGN_POSN:
+ val = (int) *(char *) LC_MONETARY_values[item_no];
+ if (val < LC_MONETARY_desc[item_no].min
+ || val > LC_MONETARY_desc[item_no].max)
+ error (0, 0, gettext ("value for item `%s' in category `%s' "
+ "must be in range %d...%d"),
+ LC_MONETARY_desc[item_no].name, "LC_MONETARY",
+ LC_MONETARY_desc[item_no].min,
+ LC_MONETARY_desc[item_no].max);
+ break;
+ case MON_DECIMAL_POINT:
+ /* The decimal point must not be empty. This is not said
+ explicitly in POSIX but ANSI C (ISO/IEC 9899) says in
+ 4.4.2.1 it has to be != "". */
+ if (LC_MONETARY_values[item_no][0] == '\0')
+ error (0, 0,
+ gettext ("item `%s' in category `%s' must not be empty"),
+ LC_MONETARY_desc[item_no].name, "LC_MONETARY");
+ break;
+ case CURRENCY_SYMBOL:
+ case MON_THOUSANDS_SEP:
+ case MON_GROUPING:
+ case POSITIVE_SIGN:
+ case NEGATIVE_SIGN:
+ case INT_FRAC_DIGITS:
+ case FRAC_DIGITS:
+ /* Everything is ok for these values. */
+ break;
+ default:
+ error (5, 0, gettext ("Internal error in %s, line %u"),
+ __FUNCTION__, __LINE__);
+ /* NOTREACHED */
+ }
+
+}
+
+
+static int
+_curr_strcmp(const char *s1, const char **s2)
+{
+ return strcmp (s1, *s2);
+}
+
+/*
+ * Local Variables:
+ * mode:c
+ * c-basic-offset:2
+ * End:
+ */
diff --git a/locale/numeric.c b/locale/numeric.c
new file mode 100644
index 0000000000..93f4bc481e
--- /dev/null
+++ b/locale/numeric.c
@@ -0,0 +1,62 @@
+/* Copyright (C) 1995 Free Software Foundation, Inc.
+
+The GNU C Library is free software; you can redistribute it and/or
+modify it under the terms of the GNU Library General Public License as
+published by the Free Software Foundation; either version 2 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
+Library General Public License for more details.
+
+You should have received a copy of the GNU Library General Public
+License along with the GNU C Library; see the file COPYING.LIB. If
+not, write to the Free Software Foundation, Inc., 675 Mass Ave,
+Cambridge, MA 02139, USA. */
+
+#include <langinfo.h>
+#include <libintl.h>
+#include <locale.h>
+
+#include "localedef.h"
+
+/* These are defined in locfile-parse.c. */
+extern struct cat_item LC_NUMERIC_desc[];
+extern char *LC_NUMERIC_values[];
+
+void
+numeric_check(void)
+{
+ int item_no;
+
+ /* First general check for existence. */
+ for (item_no = 0; item_no < category[LC_NUMERIC].number; ++item_no)
+ if (LC_NUMERIC_values[item_no] == NULL)
+ {
+ int errcode;
+
+ errcode = LC_NUMERIC_desc[item_no].status = std ? 5 : 0;
+
+ error (errcode, 0, gettext ("item `%s' of category `%s' undefined"),
+ LC_NUMERIC_desc[item_no].name, "LC_NUMERIC");
+ }
+ else
+ {
+ if (LC_NUMERIC_desc[item_no].item_id == DECIMAL_POINT
+ && LC_NUMERIC_values[item_no][0] == '\0')
+ /* The decimal point must not be empty. This is not said
+ explicitly in POSIX but ANSI C (ISO/IEC 9899) says in
+ 4.4.2.1 it has to be != "". */
+ error (0, 0,
+ gettext ("item `%s' in category `%s' must not be empty"),
+ LC_NUMERIC_desc[item_no].name, "LC_NUMERIC");
+ }
+}
+
+/*
+ * Local Variables:
+ * mode:c
+ * c-basic-offset:2
+ * End:
+ */
diff --git a/locale/token.h b/locale/token.h
new file mode 100644
index 0000000000..1227920083
--- /dev/null
+++ b/locale/token.h
@@ -0,0 +1,54 @@
+/* Copyright (C) 1995 Free Software Foundation, Inc.
+
+The GNU C Library is free software; you can redistribute it and/or
+modify it under the terms of the GNU Library General Public License as
+published by the Free Software Foundation; either version 2 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
+Library General Public License for more details.
+
+You should have received a copy of the GNU Library General Public
+License along with the GNU C Library; see the file COPYING.LIB. If
+not, write to the Free Software Foundation, Inc., 675 Mass Ave,
+Cambridge, MA 02139, USA. */
+
+#ifndef _TOKEN_H
+#define _TOKEN_H 1
+
+/* Define those keywords which does not correspond directly to any
+ item in a category. Those already have values. */
+
+enum token
+ {
+ /* We must make sure that these values do not overlap with the
+ other possible return values of the lexer. */
+ TOK_LAST_USED = _NL_NUM,
+
+ /* General tokens. */
+ TOK_END, TOK_COMMENT_CHAR, TOK_COPY, TOK_ESCAPE_CHAR, TOK_FROM,
+ TOK_ENDOFLINE, TOK_IDENT, TOK_STRING, TOK_ELLIPSIS, TOK_CHAR,
+ TOK_ILL_CHAR, TOK_NUMBER, TOK_MINUS1,
+
+ /* Tokens from the collate category. */
+ TOK_IGNORE, TOK_UNDEFINED, TOK_BACKWARD, TOK_FORWARD, TOK_POSITION,
+ TOK_COLLATING_ELEMENT, TOK_COLLATING_SYMBOL, TOK_ORDER_END,
+ TOK_ORDER_START,
+
+ /* Tokens from the ctype category. */
+ TOK_TOLOWER, TOK_TOUPPER,
+ /* The order here is important. It must correspond to the bits
+ used for indicating the class membership (ctype.h). */
+ TOK_UPPER, TOK_LOWER, TOK_ALPHA, TOK_DIGIT, TOK_XDIGIT, TOK_SPACE,
+ TOK_PRINT, TOK_GRAPH, TOK_BLANK, TOK_CNTRL, TOK_PUNCT,
+ };
+
+/*
+ * Local Variables:
+ * mode:c
+ * c-basic-offset:2
+ * End:
+ */
+#endif /* token.h */
diff --git a/posix/Makefile b/posix/Makefile
index 34fe4bb162..6931727288 100644
--- a/posix/Makefile
+++ b/posix/Makefile
@@ -22,7 +22,7 @@
subdir := posix
headers := sys/utsname.h sys/times.h sys/wait.h sys/types.h unistd.h \
- glob.h wordexp.h fnmatch.h gnu/types.h getopt.h \
+ glob.h regex.h wordexp.h fnmatch.h gnu/types.h getopt.h \
posix1_lim.h posix2_lim.h posix_opt.h local_lim.h tar.h \
utsnamelen.h confname.h waitflags.h waitstatus.h sys/unistd.h
@@ -40,7 +40,7 @@ routines := \
getpgid setpgid getpgrp setsid \
getlogin setlogin \
pathconf sysconf fpathconf \
- glob fnmatch \
+ glob fnmatch regex \
confstr \
getopt getopt1
aux := init-posix environ
@@ -48,7 +48,7 @@ tests := tstgetopt testfnm
others := getconf
install-bin := getconf
install-lib := libposix.a
-gpl2lgpl := getopt.c getopt1.c getopt.h # Frob these guys' copying notices.
+gpl2lgpl := getopt.c getopt1.c getopt.h regex.c regex.h
include ../Rules
diff --git a/posix/regex.c b/posix/regex.c
new file mode 100644
index 0000000000..897295aa27
--- /dev/null
+++ b/posix/regex.c
@@ -0,0 +1,5400 @@
+/* Extended regular expression matching and search library,
+ version 0.12.
+ (Implements POSIX draft P10003.2/D11.2, except for
+ internationalization features.)
+
+ Copyright (C) 1993, 1994, 1995 Free Software Foundation, Inc.
+
+This file is part of the GNU C Library. Its master source is NOT part of
+the C library, however. The master source lives in /gd/gnu/lib.
+
+The GNU C Library is free software; you can redistribute it and/or
+modify it under the terms of the GNU Library General Public License as
+published by the Free Software Foundation; either version 2 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
+Library General Public License for more details.
+
+You should have received a copy of the GNU Library General Public
+License along with the GNU C Library; see the file COPYING.LIB. If
+not, write to the Free Software Foundation, Inc., 675 Mass Ave,
+Cambridge, MA 02139, USA. */
+
+/* AIX requires this to be the first thing in the file. */
+#if defined (_AIX) && !defined (REGEX_MALLOC)
+ #pragma alloca
+#endif
+
+#define _GNU_SOURCE
+
+#ifdef HAVE_CONFIG_H
+#include <config.h>
+#endif
+
+/* We need this for `regex.h', and perhaps for the Emacs include files. */
+#include <sys/types.h>
+
+/* This is for other GNU distributions with internationalized messages. */
+#if HAVE_LIBINTL_H || defined (_LIBC)
+# include <libintl.h>
+#else
+# define gettext(msgid) (msgid)
+#endif
+
+/* The `emacs' switch turns on certain matching commands
+ that make sense only in Emacs. */
+#ifdef emacs
+
+#include "lisp.h"
+#include "buffer.h"
+#include "syntax.h"
+
+#else /* not emacs */
+
+/* If we are not linking with Emacs proper,
+ we can't use the relocating allocator
+ even if config.h says that we can. */
+#undef REL_ALLOC
+
+#if defined (STDC_HEADERS) || defined (_LIBC)
+#include <stdlib.h>
+#else
+char *malloc ();
+char *realloc ();
+#endif
+
+/* We used to test for `BSTRING' here, but only GCC and Emacs define
+ `BSTRING', as far as I know, and neither of them use this code. */
+#ifndef INHIBIT_STRING_HEADER
+#if HAVE_STRING_H || STDC_HEADERS || defined (_LIBC)
+#include <string.h>
+#ifndef bcmp
+#define bcmp(s1, s2, n) memcmp ((s1), (s2), (n))
+#endif
+#ifndef bcopy
+#define bcopy(s, d, n) memcpy ((d), (s), (n))
+#endif
+#ifndef bzero
+#define bzero(s, n) memset ((s), 0, (n))
+#endif
+#else
+#include <strings.h>
+#endif
+#endif
+
+/* Define the syntax stuff for \<, \>, etc. */
+
+/* This must be nonzero for the wordchar and notwordchar pattern
+ commands in re_match_2. */
+#ifndef Sword
+#define Sword 1
+#endif
+
+#ifdef SWITCH_ENUM_BUG
+#define SWITCH_ENUM_CAST(x) ((int)(x))
+#else
+#define SWITCH_ENUM_CAST(x) (x)
+#endif
+
+#ifdef SYNTAX_TABLE
+
+extern char *re_syntax_table;
+
+#else /* not SYNTAX_TABLE */
+
+/* How many characters in the character set. */
+#define CHAR_SET_SIZE 256
+
+static char re_syntax_table[CHAR_SET_SIZE];
+
+static void
+init_syntax_once ()
+{
+ register int c;
+ static int done = 0;
+
+ if (done)
+ return;
+
+ bzero (re_syntax_table, sizeof re_syntax_table);
+
+ for (c = 'a'; c <= 'z'; c++)
+ re_syntax_table[c] = Sword;
+
+ for (c = 'A'; c <= 'Z'; c++)
+ re_syntax_table[c] = Sword;
+
+ for (c = '0'; c <= '9'; c++)
+ re_syntax_table[c] = Sword;
+
+ re_syntax_table['_'] = Sword;
+
+ done = 1;
+}
+
+#endif /* not SYNTAX_TABLE */
+
+#define SYNTAX(c) re_syntax_table[c]
+
+#endif /* not emacs */
+
+/* Get the interface, including the syntax bits. */
+#include "regex.h"
+
+/* isalpha etc. are used for the character classes. */
+#include <ctype.h>
+
+/* Jim Meyering writes:
+
+ "... Some ctype macros are valid only for character codes that
+ isascii says are ASCII (SGI's IRIX-4.0.5 is one such system --when
+ using /bin/cc or gcc but without giving an ansi option). So, all
+ ctype uses should be through macros like ISPRINT... If
+ STDC_HEADERS is defined, then autoconf has verified that the ctype
+ macros don't need to be guarded with references to isascii. ...
+ Defining isascii to 1 should let any compiler worth its salt
+ eliminate the && through constant folding." */
+
+#if defined (STDC_HEADERS) || (!defined (isascii) && !defined (HAVE_ISASCII))
+#define ISASCII(c) 1
+#else
+#define ISASCII(c) isascii(c)
+#endif
+
+#ifdef isblank
+#define ISBLANK(c) (ISASCII (c) && isblank (c))
+#else
+#define ISBLANK(c) ((c) == ' ' || (c) == '\t')
+#endif
+#ifdef isgraph
+#define ISGRAPH(c) (ISASCII (c) && isgraph (c))
+#else
+#define ISGRAPH(c) (ISASCII (c) && isprint (c) && !isspace (c))
+#endif
+
+#define ISPRINT(c) (ISASCII (c) && isprint (c))
+#define ISDIGIT(c) (ISASCII (c) && isdigit (c))
+#define ISALNUM(c) (ISASCII (c) && isalnum (c))
+#define ISALPHA(c) (ISASCII (c) && isalpha (c))
+#define ISCNTRL(c) (ISASCII (c) && iscntrl (c))
+#define ISLOWER(c) (ISASCII (c) && islower (c))
+#define ISPUNCT(c) (ISASCII (c) && ispunct (c))
+#define ISSPACE(c) (ISASCII (c) && isspace (c))
+#define ISUPPER(c) (ISASCII (c) && isupper (c))
+#define ISXDIGIT(c) (ISASCII (c) && isxdigit (c))
+
+#ifndef NULL
+#define NULL 0
+#endif
+
+/* We remove any previous definition of `SIGN_EXTEND_CHAR',
+ since ours (we hope) works properly with all combinations of
+ machines, compilers, `char' and `unsigned char' argument types.
+ (Per Bothner suggested the basic approach.) */
+#undef SIGN_EXTEND_CHAR
+#if __STDC__
+#define SIGN_EXTEND_CHAR(c) ((signed char) (c))
+#else /* not __STDC__ */
+/* As in Harbison and Steele. */
+#define SIGN_EXTEND_CHAR(c) ((((unsigned char) (c)) ^ 128) - 128)
+#endif
+
+/* Should we use malloc or alloca? If REGEX_MALLOC is not defined, we
+ use `alloca' instead of `malloc'. This is because using malloc in
+ re_search* or re_match* could cause memory leaks when C-g is used in
+ Emacs; also, malloc is slower and causes storage fragmentation. On
+ the other hand, malloc is more portable, and easier to debug.
+
+ Because we sometimes use alloca, some routines have to be macros,
+ not functions -- `alloca'-allocated space disappears at the end of the
+ function it is called in. */
+
+#ifdef REGEX_MALLOC
+
+#define REGEX_ALLOCATE malloc
+#define REGEX_REALLOCATE(source, osize, nsize) realloc (source, nsize)
+#define REGEX_FREE free
+
+#else /* not REGEX_MALLOC */
+
+/* Emacs already defines alloca, sometimes. */
+#ifndef alloca
+
+/* Make alloca work the best possible way. */
+#ifdef __GNUC__
+#define alloca __builtin_alloca
+#else /* not __GNUC__ */
+#if HAVE_ALLOCA_H
+#include <alloca.h>
+#else /* not __GNUC__ or HAVE_ALLOCA_H */
+#ifndef _AIX /* Already did AIX, up at the top. */
+char *alloca ();
+#endif /* not _AIX */
+#endif /* not HAVE_ALLOCA_H */
+#endif /* not __GNUC__ */
+
+#endif /* not alloca */
+
+#define REGEX_ALLOCATE alloca
+
+/* Assumes a `char *destination' variable. */
+#define REGEX_REALLOCATE(source, osize, nsize) \
+ (destination = (char *) alloca (nsize), \
+ bcopy (source, destination, osize), \
+ destination)
+
+/* No need to do anything to free, after alloca. */
+#define REGEX_FREE(arg) (0)
+
+#endif /* not REGEX_MALLOC */
+
+/* Define how to allocate the failure stack. */
+
+#ifdef REL_ALLOC
+#define REGEX_ALLOCATE_STACK(size) \
+ r_alloc (&failure_stack_ptr, (size))
+#define REGEX_REALLOCATE_STACK(source, osize, nsize) \
+ r_re_alloc (&failure_stack_ptr, (nsize))
+#define REGEX_FREE_STACK(ptr) \
+ r_alloc_free (&failure_stack_ptr)
+
+#else /* not REL_ALLOC */
+
+#ifdef REGEX_MALLOC
+
+#define REGEX_ALLOCATE_STACK malloc
+#define REGEX_REALLOCATE_STACK(source, osize, nsize) realloc (source, nsize)
+#define REGEX_FREE_STACK free
+
+#else /* not REGEX_MALLOC */
+
+#define REGEX_ALLOCATE_STACK alloca
+
+#define REGEX_REALLOCATE_STACK(source, osize, nsize) \
+ REGEX_REALLOCATE (source, osize, nsize)
+/* No need to explicitly free anything. */
+#define REGEX_FREE_STACK(arg)
+
+#endif /* not REGEX_MALLOC */
+#endif /* not REL_ALLOC */
+
+
+/* True if `size1' is non-NULL and PTR is pointing anywhere inside
+ `string1' or just past its end. This works if PTR is NULL, which is
+ a good thing. */
+#define FIRST_STRING_P(ptr) \
+ (size1 && string1 <= (ptr) && (ptr) <= string1 + size1)
+
+/* (Re)Allocate N items of type T using malloc, or fail. */
+#define TALLOC(n, t) ((t *) malloc ((n) * sizeof (t)))
+#define RETALLOC(addr, n, t) ((addr) = (t *) realloc (addr, (n) * sizeof (t)))
+#define RETALLOC_IF(addr, n, t) \
+ if (addr) RETALLOC((addr), (n), t); else (addr) = TALLOC ((n), t)
+#define REGEX_TALLOC(n, t) ((t *) REGEX_ALLOCATE ((n) * sizeof (t)))
+
+#define BYTEWIDTH 8 /* In bits. */
+
+#define STREQ(s1, s2) ((strcmp (s1, s2) == 0))
+
+#undef MAX
+#undef MIN
+#define MAX(a, b) ((a) > (b) ? (a) : (b))
+#define MIN(a, b) ((a) < (b) ? (a) : (b))
+
+typedef char boolean;
+#define false 0
+#define true 1
+
+static int re_match_2_internal ();
+
+/* These are the command codes that appear in compiled regular
+ expressions. Some opcodes are followed by argument bytes. A
+ command code can specify any interpretation whatsoever for its
+ arguments. Zero bytes may appear in the compiled regular expression. */
+
+typedef enum
+{
+ no_op = 0,
+
+ /* Succeed right away--no more backtracking. */
+ succeed,
+
+ /* Followed by one byte giving n, then by n literal bytes. */
+ exactn,
+
+ /* Matches any (more or less) character. */
+ anychar,
+
+ /* Matches any one char belonging to specified set. First
+ following byte is number of bitmap bytes. Then come bytes
+ for a bitmap saying which chars are in. Bits in each byte
+ are ordered low-bit-first. A character is in the set if its
+ bit is 1. A character too large to have a bit in the map is
+ automatically not in the set. */
+ charset,
+
+ /* Same parameters as charset, but match any character that is
+ not one of those specified. */
+ charset_not,
+
+ /* Start remembering the text that is matched, for storing in a
+ register. Followed by one byte with the register number, in
+ the range 0 to one less than the pattern buffer's re_nsub
+ field. Then followed by one byte with the number of groups
+ inner to this one. (This last has to be part of the
+ start_memory only because we need it in the on_failure_jump
+ of re_match_2.) */
+ start_memory,
+
+ /* Stop remembering the text that is matched and store it in a
+ memory register. Followed by one byte with the register
+ number, in the range 0 to one less than `re_nsub' in the
+ pattern buffer, and one byte with the number of inner groups,
+ just like `start_memory'. (We need the number of inner
+ groups here because we don't have any easy way of finding the
+ corresponding start_memory when we're at a stop_memory.) */
+ stop_memory,
+
+ /* Match a duplicate of something remembered. Followed by one
+ byte containing the register number. */
+ duplicate,
+
+ /* Fail unless at beginning of line. */
+ begline,
+
+ /* Fail unless at end of line. */
+ endline,
+
+ /* Succeeds if at beginning of buffer (if emacs) or at beginning
+ of string to be matched (if not). */
+ begbuf,
+
+ /* Analogously, for end of buffer/string. */
+ endbuf,
+
+ /* Followed by two byte relative address to which to jump. */
+ jump,
+
+ /* Same as jump, but marks the end of an alternative. */
+ jump_past_alt,
+
+ /* Followed by two-byte relative address of place to resume at
+ in case of failure. */
+ on_failure_jump,
+
+ /* Like on_failure_jump, but pushes a placeholder instead of the
+ current string position when executed. */
+ on_failure_keep_string_jump,
+
+ /* Throw away latest failure point and then jump to following
+ two-byte relative address. */
+ pop_failure_jump,
+
+ /* Change to pop_failure_jump if know won't have to backtrack to
+ match; otherwise change to jump. This is used to jump
+ back to the beginning of a repeat. If what follows this jump
+ clearly won't match what the repeat does, such that we can be
+ sure that there is no use backtracking out of repetitions
+ already matched, then we change it to a pop_failure_jump.
+ Followed by two-byte address. */
+ maybe_pop_jump,
+
+ /* Jump to following two-byte address, and push a dummy failure
+ point. This failure point will be thrown away if an attempt
+ is made to use it for a failure. A `+' construct makes this
+ before the first repeat. Also used as an intermediary kind
+ of jump when compiling an alternative. */
+ dummy_failure_jump,
+
+ /* Push a dummy failure point and continue. Used at the end of
+ alternatives. */
+ push_dummy_failure,
+
+ /* Followed by two-byte relative address and two-byte number n.
+ After matching N times, jump to the address upon failure. */
+ succeed_n,
+
+ /* Followed by two-byte relative address, and two-byte number n.
+ Jump to the address N times, then fail. */
+ jump_n,
+
+ /* Set the following two-byte relative address to the
+ subsequent two-byte number. The address *includes* the two
+ bytes of number. */
+ set_number_at,
+
+ wordchar, /* Matches any word-constituent character. */
+ notwordchar, /* Matches any char that is not a word-constituent. */
+
+ wordbeg, /* Succeeds if at word beginning. */
+ wordend, /* Succeeds if at word end. */
+
+ wordbound, /* Succeeds if at a word boundary. */
+ notwordbound /* Succeeds if not at a word boundary. */
+
+#ifdef emacs
+ ,before_dot, /* Succeeds if before point. */
+ at_dot, /* Succeeds if at point. */
+ after_dot, /* Succeeds if after point. */
+
+ /* Matches any character whose syntax is specified. Followed by
+ a byte which contains a syntax code, e.g., Sword. */
+ syntaxspec,
+
+ /* Matches any character whose syntax is not that specified. */
+ notsyntaxspec
+#endif /* emacs */
+} re_opcode_t;
+
+/* Common operations on the compiled pattern. */
+
+/* Store NUMBER in two contiguous bytes starting at DESTINATION. */
+
+#define STORE_NUMBER(destination, number) \
+ do { \
+ (destination)[0] = (number) & 0377; \
+ (destination)[1] = (number) >> 8; \
+ } while (0)
+
+/* Same as STORE_NUMBER, except increment DESTINATION to
+ the byte after where the number is stored. Therefore, DESTINATION
+ must be an lvalue. */
+
+#define STORE_NUMBER_AND_INCR(destination, number) \
+ do { \
+ STORE_NUMBER (destination, number); \
+ (destination) += 2; \
+ } while (0)
+
+/* Put into DESTINATION a number stored in two contiguous bytes starting
+ at SOURCE. */
+
+#define EXTRACT_NUMBER(destination, source) \
+ do { \
+ (destination) = *(source) & 0377; \
+ (destination) += SIGN_EXTEND_CHAR (*((source) + 1)) << 8; \
+ } while (0)
+
+#ifdef DEBUG
+static void
+extract_number (dest, source)
+ int *dest;
+ unsigned char *source;
+{
+ int temp = SIGN_EXTEND_CHAR (*(source + 1));
+ *dest = *source & 0377;
+ *dest += temp << 8;
+}
+
+#ifndef EXTRACT_MACROS /* To debug the macros. */
+#undef EXTRACT_NUMBER
+#define EXTRACT_NUMBER(dest, src) extract_number (&dest, src)
+#endif /* not EXTRACT_MACROS */
+
+#endif /* DEBUG */
+
+/* Same as EXTRACT_NUMBER, except increment SOURCE to after the number.
+ SOURCE must be an lvalue. */
+
+#define EXTRACT_NUMBER_AND_INCR(destination, source) \
+ do { \
+ EXTRACT_NUMBER (destination, source); \
+ (source) += 2; \
+ } while (0)
+
+#ifdef DEBUG
+static void
+extract_number_and_incr (destination, source)
+ int *destination;
+ unsigned char **source;
+{
+ extract_number (destination, *source);
+ *source += 2;
+}
+
+#ifndef EXTRACT_MACROS
+#undef EXTRACT_NUMBER_AND_INCR
+#define EXTRACT_NUMBER_AND_INCR(dest, src) \
+ extract_number_and_incr (&dest, &src)
+#endif /* not EXTRACT_MACROS */
+
+#endif /* DEBUG */
+
+/* If DEBUG is defined, Regex prints many voluminous messages about what
+ it is doing (if the variable `debug' is nonzero). If linked with the
+ main program in `iregex.c', you can enter patterns and strings
+ interactively. And if linked with the main program in `main.c' and
+ the other test files, you can run the already-written tests. */
+
+#ifdef DEBUG
+
+/* We use standard I/O for debugging. */
+#include <stdio.h>
+
+/* It is useful to test things that ``must'' be true when debugging. */
+#include <assert.h>
+
+static int debug = 0;
+
+#define DEBUG_STATEMENT(e) e
+#define DEBUG_PRINT1(x) if (debug) printf (x)
+#define DEBUG_PRINT2(x1, x2) if (debug) printf (x1, x2)
+#define DEBUG_PRINT3(x1, x2, x3) if (debug) printf (x1, x2, x3)
+#define DEBUG_PRINT4(x1, x2, x3, x4) if (debug) printf (x1, x2, x3, x4)
+#define DEBUG_PRINT_COMPILED_PATTERN(p, s, e) \
+ if (debug) print_partial_compiled_pattern (s, e)
+#define DEBUG_PRINT_DOUBLE_STRING(w, s1, sz1, s2, sz2) \
+ if (debug) print_double_string (w, s1, sz1, s2, sz2)
+
+
+/* Print the fastmap in human-readable form. */
+
+void
+print_fastmap (fastmap)
+ char *fastmap;
+{
+ unsigned was_a_range = 0;
+ unsigned i = 0;
+
+ while (i < (1 << BYTEWIDTH))
+ {
+ if (fastmap[i++])
+ {
+ was_a_range = 0;
+ putchar (i - 1);
+ while (i < (1 << BYTEWIDTH) && fastmap[i])
+ {
+ was_a_range = 1;
+ i++;
+ }
+ if (was_a_range)
+ {
+ printf ("-");
+ putchar (i - 1);
+ }
+ }
+ }
+ putchar ('\n');
+}
+
+
+/* Print a compiled pattern string in human-readable form, starting at
+ the START pointer into it and ending just before the pointer END. */
+
+void
+print_partial_compiled_pattern (start, end)
+ unsigned char *start;
+ unsigned char *end;
+{
+ int mcnt, mcnt2;
+ unsigned char *p = start;
+ unsigned char *pend = end;
+
+ if (start == NULL)
+ {
+ printf ("(null)\n");
+ return;
+ }
+
+ /* Loop over pattern commands. */
+ while (p < pend)
+ {
+ printf ("%d:\t", p - start);
+
+ switch ((re_opcode_t) *p++)
+ {
+ case no_op:
+ printf ("/no_op");
+ break;
+
+ case exactn:
+ mcnt = *p++;
+ printf ("/exactn/%d", mcnt);
+ do
+ {
+ putchar ('/');
+ putchar (*p++);
+ }
+ while (--mcnt);
+ break;
+
+ case start_memory:
+ mcnt = *p++;
+ printf ("/start_memory/%d/%d", mcnt, *p++);
+ break;
+
+ case stop_memory:
+ mcnt = *p++;
+ printf ("/stop_memory/%d/%d", mcnt, *p++);
+ break;
+
+ case duplicate:
+ printf ("/duplicate/%d", *p++);
+ break;
+
+ case anychar:
+ printf ("/anychar");
+ break;
+
+ case charset:
+ case charset_not:
+ {
+ register int c, last = -100;
+ register int in_range = 0;
+
+ printf ("/charset [%s",
+ (re_opcode_t) *(p - 1) == charset_not ? "^" : "");
+
+ assert (p + *p < pend);
+
+ for (c = 0; c < 256; c++)
+ if (c / 8 < *p
+ && (p[1 + (c/8)] & (1 << (c % 8))))
+ {
+ /* Are we starting a range? */
+ if (last + 1 == c && ! in_range)
+ {
+ putchar ('-');
+ in_range = 1;
+ }
+ /* Have we broken a range? */
+ else if (last + 1 != c && in_range)
+ {
+ putchar (last);
+ in_range = 0;
+ }
+
+ if (! in_range)
+ putchar (c);
+
+ last = c;
+ }
+
+ if (in_range)
+ putchar (last);
+
+ putchar (']');
+
+ p += 1 + *p;
+ }
+ break;
+
+ case begline:
+ printf ("/begline");
+ break;
+
+ case endline:
+ printf ("/endline");
+ break;
+
+ case on_failure_jump:
+ extract_number_and_incr (&mcnt, &p);
+ printf ("/on_failure_jump to %d", p + mcnt - start);
+ break;
+
+ case on_failure_keep_string_jump:
+ extract_number_and_incr (&mcnt, &p);
+ printf ("/on_failure_keep_string_jump to %d", p + mcnt - start);
+ break;
+
+ case dummy_failure_jump:
+ extract_number_and_incr (&mcnt, &p);
+ printf ("/dummy_failure_jump to %d", p + mcnt - start);
+ break;
+
+ case push_dummy_failure:
+ printf ("/push_dummy_failure");
+ break;
+
+ case maybe_pop_jump:
+ extract_number_and_incr (&mcnt, &p);
+ printf ("/maybe_pop_jump to %d", p + mcnt - start);
+ break;
+
+ case pop_failure_jump:
+ extract_number_and_incr (&mcnt, &p);
+ printf ("/pop_failure_jump to %d", p + mcnt - start);
+ break;
+
+ case jump_past_alt:
+ extract_number_and_incr (&mcnt, &p);
+ printf ("/jump_past_alt to %d", p + mcnt - start);
+ break;
+
+ case jump:
+ extract_number_and_incr (&mcnt, &p);
+ printf ("/jump to %d", p + mcnt - start);
+ break;
+
+ case succeed_n:
+ extract_number_and_incr (&mcnt, &p);
+ extract_number_and_incr (&mcnt2, &p);
+ printf ("/succeed_n to %d, %d times", p + mcnt - start, mcnt2);
+ break;
+
+ case jump_n:
+ extract_number_and_incr (&mcnt, &p);
+ extract_number_and_incr (&mcnt2, &p);
+ printf ("/jump_n to %d, %d times", p + mcnt - start, mcnt2);
+ break;
+
+ case set_number_at:
+ extract_number_and_incr (&mcnt, &p);
+ extract_number_and_incr (&mcnt2, &p);
+ printf ("/set_number_at location %d to %d", p + mcnt - start, mcnt2);
+ break;
+
+ case wordbound:
+ printf ("/wordbound");
+ break;
+
+ case notwordbound:
+ printf ("/notwordbound");
+ break;
+
+ case wordbeg:
+ printf ("/wordbeg");
+ break;
+
+ case wordend:
+ printf ("/wordend");
+
+#ifdef emacs
+ case before_dot:
+ printf ("/before_dot");
+ break;
+
+ case at_dot:
+ printf ("/at_dot");
+ break;
+
+ case after_dot:
+ printf ("/after_dot");
+ break;
+
+ case syntaxspec:
+ printf ("/syntaxspec");
+ mcnt = *p++;
+ printf ("/%d", mcnt);
+ break;
+
+ case notsyntaxspec:
+ printf ("/notsyntaxspec");
+ mcnt = *p++;
+ printf ("/%d", mcnt);
+ break;
+#endif /* emacs */
+
+ case wordchar:
+ printf ("/wordchar");
+ break;
+
+ case notwordchar:
+ printf ("/notwordchar");
+ break;
+
+ case begbuf:
+ printf ("/begbuf");
+ break;
+
+ case endbuf:
+ printf ("/endbuf");
+ break;
+
+ default:
+ printf ("?%d", *(p-1));
+ }
+
+ putchar ('\n');
+ }
+
+ printf ("%d:\tend of pattern.\n", p - start);
+}
+
+
+void
+print_compiled_pattern (bufp)
+ struct re_pattern_buffer *bufp;
+{
+ unsigned char *buffer = bufp->buffer;
+
+ print_partial_compiled_pattern (buffer, buffer + bufp->used);
+ printf ("%d bytes used/%d bytes allocated.\n", bufp->used, bufp->allocated);
+
+ if (bufp->fastmap_accurate && bufp->fastmap)
+ {
+ printf ("fastmap: ");
+ print_fastmap (bufp->fastmap);
+ }
+
+ printf ("re_nsub: %d\t", bufp->re_nsub);
+ printf ("regs_alloc: %d\t", bufp->regs_allocated);
+ printf ("can_be_null: %d\t", bufp->can_be_null);
+ printf ("newline_anchor: %d\n", bufp->newline_anchor);
+ printf ("no_sub: %d\t", bufp->no_sub);
+ printf ("not_bol: %d\t", bufp->not_bol);
+ printf ("not_eol: %d\t", bufp->not_eol);
+ printf ("syntax: %d\n", bufp->syntax);
+ /* Perhaps we should print the translate table? */
+}
+
+
+void
+print_double_string (where, string1, size1, string2, size2)
+ const char *where;
+ const char *string1;
+ const char *string2;
+ int size1;
+ int size2;
+{
+ unsigned this_char;
+
+ if (where == NULL)
+ printf ("(null)");
+ else
+ {
+ if (FIRST_STRING_P (where))
+ {
+ for (this_char = where - string1; this_char < size1; this_char++)
+ putchar (string1[this_char]);
+
+ where = string2;
+ }
+
+ for (this_char = where - string2; this_char < size2; this_char++)
+ putchar (string2[this_char]);
+ }
+}
+
+#else /* not DEBUG */
+
+#undef assert
+#define assert(e)
+
+#define DEBUG_STATEMENT(e)
+#define DEBUG_PRINT1(x)
+#define DEBUG_PRINT2(x1, x2)
+#define DEBUG_PRINT3(x1, x2, x3)
+#define DEBUG_PRINT4(x1, x2, x3, x4)
+#define DEBUG_PRINT_COMPILED_PATTERN(p, s, e)
+#define DEBUG_PRINT_DOUBLE_STRING(w, s1, sz1, s2, sz2)
+
+#endif /* not DEBUG */
+
+/* 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 (syntax)
+ reg_syntax_t syntax;
+{
+ reg_syntax_t ret = re_syntax_options;
+
+ re_syntax_options = syntax;
+ return ret;
+}
+
+/* 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[] =
+ { "Success", /* REG_NOERROR */
+ "No match", /* REG_NOMATCH */
+ "Invalid regular expression", /* REG_BADPAT */
+ "Invalid collation character", /* REG_ECOLLATE */
+ "Invalid character class name", /* REG_ECTYPE */
+ "Trailing backslash", /* REG_EESCAPE */
+ "Invalid back reference", /* REG_ESUBREG */
+ "Unmatched [ or [^", /* REG_EBRACK */
+ "Unmatched ( or \\(", /* REG_EPAREN */
+ "Unmatched \\{", /* REG_EBRACE */
+ "Invalid content of \\{\\}", /* REG_BADBR */
+ "Invalid range end", /* REG_ERANGE */
+ "Memory exhausted", /* REG_ESPACE */
+ "Invalid preceding regular expression", /* REG_BADRPT */
+ "Premature end of regular expression", /* REG_EEND */
+ "Regular expression too big", /* REG_ESIZE */
+ "Unmatched ) or \\)", /* REG_ERPAREN */
+ };
+
+/* Avoiding alloca during matching, to placate r_alloc. */
+
+/* Define MATCH_MAY_ALLOCATE unless we need to make sure that the
+ searching and matching functions should not call alloca. On some
+ systems, alloca is implemented in terms of malloc, and if we're
+ using the relocating allocator routines, then malloc could cause a
+ relocation, which might (if the strings being searched are in the
+ ralloc heap) shift the data out from underneath the regexp
+ routines.
+
+ Here's another reason to avoid allocation: Emacs
+ processes input from X in a signal handler; processing X input may
+ call malloc; if input arrives while a matching routine is calling
+ malloc, then we're scrod. But Emacs can't just block input while
+ calling matching routines; then we don't notice interrupts when
+ they come in. So, Emacs blocks input around all regexp calls
+ except the matching calls, which it leaves unprotected, in the
+ faith that they will not malloc. */
+
+/* Normally, this is fine. */
+#define MATCH_MAY_ALLOCATE
+
+/* When using GNU C, we are not REALLY using the C alloca, no matter
+ what config.h may say. So don't take precautions for it. */
+#ifdef __GNUC__
+#undef C_ALLOCA
+#endif
+
+/* The match routines may not allocate if (1) they would do it with malloc
+ and (2) it's not safe for them to use malloc.
+ Note that if REL_ALLOC is defined, matching would not use malloc for the
+ failure stack, but we would still use it for the register vectors;
+ so REL_ALLOC should not affect this. */
+#if (defined (C_ALLOCA) || defined (REGEX_MALLOC)) && defined (emacs)
+#undef MATCH_MAY_ALLOCATE
+#endif
+
+
+/* Failure stack declarations and macros; both re_compile_fastmap and
+ re_match_2 use a failure stack. These have to be macros because of
+ REGEX_ALLOCATE_STACK. */
+
+
+/* Number of failure points for which to initially allocate space
+ when matching. If this number is exceeded, we allocate more
+ space, so it is not a hard limit. */
+#ifndef INIT_FAILURE_ALLOC
+#define INIT_FAILURE_ALLOC 5
+#endif
+
+/* Roughly the maximum number of failure points on the stack. Would be
+ exactly that if always used MAX_FAILURE_SPACE each time we failed.
+ This is a variable only so users of regex can assign to it; we never
+ change it ourselves. */
+#if defined (MATCH_MAY_ALLOCATE)
+int re_max_failures = 200000;
+#else
+int re_max_failures = 2000;
+#endif
+
+union fail_stack_elt
+{
+ unsigned char *pointer;
+ int integer;
+};
+
+typedef union fail_stack_elt fail_stack_elt_t;
+
+typedef struct
+{
+ fail_stack_elt_t *stack;
+ unsigned size;
+ unsigned avail; /* Offset of next open position. */
+} fail_stack_type;
+
+#define FAIL_STACK_EMPTY() (fail_stack.avail == 0)
+#define FAIL_STACK_PTR_EMPTY() (fail_stack_ptr->avail == 0)
+#define FAIL_STACK_FULL() (fail_stack.avail == fail_stack.size)
+
+
+/* Define macros to initialize and free the failure stack.
+ Do `return -2' if the alloc fails. */
+
+#ifdef MATCH_MAY_ALLOCATE
+#define INIT_FAIL_STACK() \
+ do { \
+ fail_stack.stack = (fail_stack_elt_t *) \
+ REGEX_ALLOCATE_STACK (INIT_FAILURE_ALLOC * sizeof (fail_stack_elt_t)); \
+ \
+ if (fail_stack.stack == NULL) \
+ return -2; \
+ \
+ fail_stack.size = INIT_FAILURE_ALLOC; \
+ fail_stack.avail = 0; \
+ } while (0)
+
+#define RESET_FAIL_STACK() REGEX_FREE_STACK (fail_stack.stack)
+#else
+#define INIT_FAIL_STACK() \
+ do { \
+ fail_stack.avail = 0; \
+ } while (0)
+
+#define RESET_FAIL_STACK()
+#endif
+
+
+/* Double the size of FAIL_STACK, up to approximately `re_max_failures' items.
+
+ Return 1 if succeeds, and 0 if either ran out of memory
+ allocating space for it or it was already too large.
+
+ REGEX_REALLOCATE_STACK requires `destination' be declared. */
+
+#define DOUBLE_FAIL_STACK(fail_stack) \
+ ((fail_stack).size > re_max_failures * MAX_FAILURE_ITEMS \
+ ? 0 \
+ : ((fail_stack).stack = (fail_stack_elt_t *) \
+ REGEX_REALLOCATE_STACK ((fail_stack).stack, \
+ (fail_stack).size * sizeof (fail_stack_elt_t), \
+ ((fail_stack).size << 1) * sizeof (fail_stack_elt_t)), \
+ \
+ (fail_stack).stack == NULL \
+ ? 0 \
+ : ((fail_stack).size <<= 1, \
+ 1)))
+
+
+/* Push pointer POINTER on FAIL_STACK.
+ Return 1 if was able to do so and 0 if ran out of memory allocating
+ space to do so. */
+#define PUSH_PATTERN_OP(POINTER, FAIL_STACK) \
+ ((FAIL_STACK_FULL () \
+ && !DOUBLE_FAIL_STACK (FAIL_STACK)) \
+ ? 0 \
+ : ((FAIL_STACK).stack[(FAIL_STACK).avail++].pointer = POINTER, \
+ 1))
+
+/* Push a pointer value onto the failure stack.
+ Assumes the variable `fail_stack'. Probably should only
+ be called from within `PUSH_FAILURE_POINT'. */
+#define PUSH_FAILURE_POINTER(item) \
+ fail_stack.stack[fail_stack.avail++].pointer = (unsigned char *) (item)
+
+/* This pushes an integer-valued item onto the failure stack.
+ Assumes the variable `fail_stack'. Probably should only
+ be called from within `PUSH_FAILURE_POINT'. */
+#define PUSH_FAILURE_INT(item) \
+ fail_stack.stack[fail_stack.avail++].integer = (item)
+
+/* Push a fail_stack_elt_t value onto the failure stack.
+ Assumes the variable `fail_stack'. Probably should only
+ be called from within `PUSH_FAILURE_POINT'. */
+#define PUSH_FAILURE_ELT(item) \
+ fail_stack.stack[fail_stack.avail++] = (item)
+
+/* These three POP... operations complement the three PUSH... operations.
+ All assume that `fail_stack' is nonempty. */
+#define POP_FAILURE_POINTER() fail_stack.stack[--fail_stack.avail].pointer
+#define POP_FAILURE_INT() fail_stack.stack[--fail_stack.avail].integer
+#define POP_FAILURE_ELT() fail_stack.stack[--fail_stack.avail]
+
+/* Used to omit pushing failure point id's when we're not debugging. */
+#ifdef DEBUG
+#define DEBUG_PUSH PUSH_FAILURE_INT
+#define DEBUG_POP(item_addr) *(item_addr) = POP_FAILURE_INT ()
+#else
+#define DEBUG_PUSH(item)
+#define DEBUG_POP(item_addr)
+#endif
+
+
+/* Push the information about the state we will need
+ if we ever fail back to it.
+
+ Requires variables fail_stack, regstart, regend, reg_info, and
+ num_regs be declared. DOUBLE_FAIL_STACK requires `destination' be
+ declared.
+
+ Does `return FAILURE_CODE' if runs out of memory. */
+
+#define PUSH_FAILURE_POINT(pattern_place, string_place, failure_code) \
+ do { \
+ char *destination; \
+ /* Must be int, so when we don't save any registers, the arithmetic \
+ of 0 + -1 isn't done as unsigned. */ \
+ int this_reg; \
+ \
+ DEBUG_STATEMENT (failure_id++); \
+ DEBUG_STATEMENT (nfailure_points_pushed++); \
+ DEBUG_PRINT2 ("\nPUSH_FAILURE_POINT #%u:\n", failure_id); \
+ DEBUG_PRINT2 (" Before push, next avail: %d\n", (fail_stack).avail);\
+ DEBUG_PRINT2 (" size: %d\n", (fail_stack).size);\
+ \
+ DEBUG_PRINT2 (" slots needed: %d\n", NUM_FAILURE_ITEMS); \
+ DEBUG_PRINT2 (" available: %d\n", REMAINING_AVAIL_SLOTS); \
+ \
+ /* Ensure we have enough space allocated for what we will push. */ \
+ while (REMAINING_AVAIL_SLOTS < NUM_FAILURE_ITEMS) \
+ { \
+ if (!DOUBLE_FAIL_STACK (fail_stack)) \
+ return failure_code; \
+ \
+ DEBUG_PRINT2 ("\n Doubled stack; size now: %d\n", \
+ (fail_stack).size); \
+ DEBUG_PRINT2 (" slots available: %d\n", REMAINING_AVAIL_SLOTS);\
+ } \
+ \
+ /* Push the info, starting with the registers. */ \
+ DEBUG_PRINT1 ("\n"); \
+ \
+ for (this_reg = lowest_active_reg; this_reg <= highest_active_reg; \
+ this_reg++) \
+ { \
+ DEBUG_PRINT2 (" Pushing reg: %d\n", this_reg); \
+ DEBUG_STATEMENT (num_regs_pushed++); \
+ \
+ DEBUG_PRINT2 (" start: 0x%x\n", regstart[this_reg]); \
+ PUSH_FAILURE_POINTER (regstart[this_reg]); \
+ \
+ DEBUG_PRINT2 (" end: 0x%x\n", regend[this_reg]); \
+ PUSH_FAILURE_POINTER (regend[this_reg]); \
+ \
+ DEBUG_PRINT2 (" info: 0x%x\n ", reg_info[this_reg]); \
+ DEBUG_PRINT2 (" match_null=%d", \
+ REG_MATCH_NULL_STRING_P (reg_info[this_reg])); \
+ DEBUG_PRINT2 (" active=%d", IS_ACTIVE (reg_info[this_reg])); \
+ DEBUG_PRINT2 (" matched_something=%d", \
+ MATCHED_SOMETHING (reg_info[this_reg])); \
+ DEBUG_PRINT2 (" ever_matched=%d", \
+ EVER_MATCHED_SOMETHING (reg_info[this_reg])); \
+ DEBUG_PRINT1 ("\n"); \
+ PUSH_FAILURE_ELT (reg_info[this_reg].word); \
+ } \
+ \
+ DEBUG_PRINT2 (" Pushing low active reg: %d\n", lowest_active_reg);\
+ PUSH_FAILURE_INT (lowest_active_reg); \
+ \
+ DEBUG_PRINT2 (" Pushing high active reg: %d\n", highest_active_reg);\
+ PUSH_FAILURE_INT (highest_active_reg); \
+ \
+ DEBUG_PRINT2 (" Pushing pattern 0x%x: ", pattern_place); \
+ DEBUG_PRINT_COMPILED_PATTERN (bufp, pattern_place, pend); \
+ PUSH_FAILURE_POINTER (pattern_place); \
+ \
+ DEBUG_PRINT2 (" Pushing string 0x%x: `", string_place); \
+ DEBUG_PRINT_DOUBLE_STRING (string_place, string1, size1, string2, \
+ size2); \
+ DEBUG_PRINT1 ("'\n"); \
+ PUSH_FAILURE_POINTER (string_place); \
+ \
+ DEBUG_PRINT2 (" Pushing failure id: %u\n", failure_id); \
+ DEBUG_PUSH (failure_id); \
+ } while (0)
+
+/* This is the number of items that are pushed and popped on the stack
+ for each register. */
+#define NUM_REG_ITEMS 3
+
+/* Individual items aside from the registers. */
+#ifdef DEBUG
+#define NUM_NONREG_ITEMS 5 /* Includes failure point id. */
+#else
+#define NUM_NONREG_ITEMS 4
+#endif
+
+/* We push at most this many items on the stack. */
+#define MAX_FAILURE_ITEMS ((num_regs - 1) * NUM_REG_ITEMS + NUM_NONREG_ITEMS)
+
+/* We actually push this many items. */
+#define NUM_FAILURE_ITEMS \
+ ((highest_active_reg - lowest_active_reg + 1) * NUM_REG_ITEMS \
+ + NUM_NONREG_ITEMS)
+
+/* How many items can still be added to the stack without overflowing it. */
+#define REMAINING_AVAIL_SLOTS ((fail_stack).size - (fail_stack).avail)
+
+
+/* Pops what PUSH_FAIL_STACK pushes.
+
+ We restore into the parameters, all of which should be lvalues:
+ STR -- the saved data position.
+ PAT -- the saved pattern position.
+ LOW_REG, HIGH_REG -- the highest and lowest active registers.
+ REGSTART, REGEND -- arrays of string positions.
+ REG_INFO -- array of information about each subexpression.
+
+ Also assumes the variables `fail_stack' and (if debugging), `bufp',
+ `pend', `string1', `size1', `string2', and `size2'. */
+
+#define POP_FAILURE_POINT(str, pat, low_reg, high_reg, regstart, regend, reg_info)\
+{ \
+ DEBUG_STATEMENT (fail_stack_elt_t failure_id;) \
+ int this_reg; \
+ const unsigned char *string_temp; \
+ \
+ assert (!FAIL_STACK_EMPTY ()); \
+ \
+ /* Remove failure points and point to how many regs pushed. */ \
+ DEBUG_PRINT1 ("POP_FAILURE_POINT:\n"); \
+ DEBUG_PRINT2 (" Before pop, next avail: %d\n", fail_stack.avail); \
+ DEBUG_PRINT2 (" size: %d\n", fail_stack.size); \
+ \
+ assert (fail_stack.avail >= NUM_NONREG_ITEMS); \
+ \
+ DEBUG_POP (&failure_id); \
+ DEBUG_PRINT2 (" Popping failure id: %u\n", failure_id); \
+ \
+ /* If the saved string location is NULL, it came from an \
+ on_failure_keep_string_jump opcode, and we want to throw away the \
+ saved NULL, thus retaining our current position in the string. */ \
+ string_temp = POP_FAILURE_POINTER (); \
+ if (string_temp != NULL) \
+ str = (const char *) string_temp; \
+ \
+ DEBUG_PRINT2 (" Popping string 0x%x: `", str); \
+ DEBUG_PRINT_DOUBLE_STRING (str, string1, size1, string2, size2); \
+ DEBUG_PRINT1 ("'\n"); \
+ \
+ pat = (unsigned char *) POP_FAILURE_POINTER (); \
+ DEBUG_PRINT2 (" Popping pattern 0x%x: ", pat); \
+ DEBUG_PRINT_COMPILED_PATTERN (bufp, pat, pend); \
+ \
+ /* Restore register info. */ \
+ high_reg = (unsigned) POP_FAILURE_INT (); \
+ DEBUG_PRINT2 (" Popping high active reg: %d\n", high_reg); \
+ \
+ low_reg = (unsigned) POP_FAILURE_INT (); \
+ DEBUG_PRINT2 (" Popping low active reg: %d\n", low_reg); \
+ \
+ for (this_reg = high_reg; this_reg >= low_reg; this_reg--) \
+ { \
+ DEBUG_PRINT2 (" Popping reg: %d\n", this_reg); \
+ \
+ reg_info[this_reg].word = POP_FAILURE_ELT (); \
+ DEBUG_PRINT2 (" info: 0x%x\n", reg_info[this_reg]); \
+ \
+ regend[this_reg] = (const char *) POP_FAILURE_POINTER (); \
+ DEBUG_PRINT2 (" end: 0x%x\n", regend[this_reg]); \
+ \
+ regstart[this_reg] = (const char *) POP_FAILURE_POINTER (); \
+ DEBUG_PRINT2 (" start: 0x%x\n", regstart[this_reg]); \
+ } \
+ \
+ set_regs_matched_done = 0; \
+ DEBUG_STATEMENT (nfailure_points_popped++); \
+} /* POP_FAILURE_POINT */
+
+
+
+/* Structure for per-register (a.k.a. per-group) information.
+ Other register information, such as the
+ starting and ending positions (which are addresses), and the list of
+ inner groups (which is a bits list) are maintained in separate
+ variables.
+
+ We are making a (strictly speaking) nonportable assumption here: that
+ the compiler will pack our bit fields into something that fits into
+ the type of `word', i.e., is something that fits into one item on the
+ failure stack. */
+
+typedef union
+{
+ fail_stack_elt_t word;
+ struct
+ {
+ /* This field is one if this group can match the empty string,
+ zero if not. If not yet determined, `MATCH_NULL_UNSET_VALUE'. */
+#define MATCH_NULL_UNSET_VALUE 3
+ unsigned match_null_string_p : 2;
+ unsigned is_active : 1;
+ unsigned matched_something : 1;
+ unsigned ever_matched_something : 1;
+ } bits;
+} register_info_type;
+
+#define REG_MATCH_NULL_STRING_P(R) ((R).bits.match_null_string_p)
+#define IS_ACTIVE(R) ((R).bits.is_active)
+#define MATCHED_SOMETHING(R) ((R).bits.matched_something)
+#define EVER_MATCHED_SOMETHING(R) ((R).bits.ever_matched_something)
+
+
+/* Call this when have matched a real character; it sets `matched' flags
+ for the subexpressions which we are currently inside. Also records
+ that those subexprs have matched. */
+#define SET_REGS_MATCHED() \
+ do \
+ { \
+ if (!set_regs_matched_done) \
+ { \
+ unsigned r; \
+ set_regs_matched_done = 1; \
+ for (r = lowest_active_reg; r <= highest_active_reg; r++) \
+ { \
+ MATCHED_SOMETHING (reg_info[r]) \
+ = EVER_MATCHED_SOMETHING (reg_info[r]) \
+ = 1; \
+ } \
+ } \
+ } \
+ while (0)
+
+/* Registers are set to a sentinel when they haven't yet matched. */
+static char reg_unset_dummy;
+#define REG_UNSET_VALUE (&reg_unset_dummy)
+#define REG_UNSET(e) ((e) == REG_UNSET_VALUE)
+
+/* Subroutine declarations and macros for regex_compile. */
+
+static void store_op1 (), store_op2 ();
+static void insert_op1 (), insert_op2 ();
+static boolean at_begline_loc_p (), at_endline_loc_p ();
+static boolean group_in_compile_stack ();
+static reg_errcode_t compile_range ();
+
+/* Fetch the next character in the uncompiled pattern---translating it
+ if necessary. Also cast from a signed character in the constant
+ string passed to us by the user to an unsigned char that we can use
+ as an array index (in, e.g., `translate'). */
+#define PATFETCH(c) \
+ do {if (p == pend) return REG_EEND; \
+ c = (unsigned char) *p++; \
+ if (translate) c = translate[c]; \
+ } while (0)
+
+/* Fetch the next character in the uncompiled pattern, with no
+ translation. */
+#define PATFETCH_RAW(c) \
+ do {if (p == pend) return REG_EEND; \
+ c = (unsigned char) *p++; \
+ } while (0)
+
+/* Go backwards one character in the pattern. */
+#define PATUNFETCH p--
+
+
+/* If `translate' is non-null, return translate[D], else just D. We
+ cast the subscript to translate because some data is declared as
+ `char *', to avoid warnings when a string constant is passed. But
+ when we use a character as a subscript we must make it unsigned. */
+#define TRANSLATE(d) (translate ? translate[(unsigned char) (d)] : (d))
+
+
+/* Macros for outputting the compiled pattern into `buffer'. */
+
+/* If the buffer isn't allocated when it comes in, use this. */
+#define INIT_BUF_SIZE 32
+
+/* Make sure we have at least N more bytes of space in buffer. */
+#define GET_BUFFER_SPACE(n) \
+ while (b - bufp->buffer + (n) > bufp->allocated) \
+ EXTEND_BUFFER ()
+
+/* Make sure we have one more byte of buffer space and then add C to it. */
+#define BUF_PUSH(c) \
+ do { \
+ GET_BUFFER_SPACE (1); \
+ *b++ = (unsigned char) (c); \
+ } while (0)
+
+
+/* Ensure we have two more bytes of buffer space and then append C1 and C2. */
+#define BUF_PUSH_2(c1, c2) \
+ do { \
+ GET_BUFFER_SPACE (2); \
+ *b++ = (unsigned char) (c1); \
+ *b++ = (unsigned char) (c2); \
+ } while (0)
+
+
+/* As with BUF_PUSH_2, except for three bytes. */
+#define BUF_PUSH_3(c1, c2, c3) \
+ do { \
+ GET_BUFFER_SPACE (3); \
+ *b++ = (unsigned char) (c1); \
+ *b++ = (unsigned char) (c2); \
+ *b++ = (unsigned char) (c3); \
+ } while (0)
+
+
+/* Store a jump with opcode OP at LOC to location TO. We store a
+ relative address offset by the three bytes the jump itself occupies. */
+#define STORE_JUMP(op, loc, to) \
+ store_op1 (op, loc, (to) - (loc) - 3)
+
+/* Likewise, for a two-argument jump. */
+#define STORE_JUMP2(op, loc, to, arg) \
+ store_op2 (op, loc, (to) - (loc) - 3, arg)
+
+/* Like `STORE_JUMP', but for inserting. Assume `b' is the buffer end. */
+#define INSERT_JUMP(op, loc, to) \
+ insert_op1 (op, loc, (to) - (loc) - 3, b)
+
+/* Like `STORE_JUMP2', but for inserting. Assume `b' is the buffer end. */
+#define INSERT_JUMP2(op, loc, to, arg) \
+ insert_op2 (op, loc, (to) - (loc) - 3, arg, b)
+
+
+/* This is not an arbitrary limit: the arguments which represent offsets
+ into the pattern are two bytes long. So if 2^16 bytes turns out to
+ be too small, many things would have to change. */
+#define MAX_BUF_SIZE (1L << 16)
+
+
+/* Extend the buffer by twice its current size via realloc and
+ reset the pointers that pointed into the old block to point to the
+ correct places in the new one. If extending the buffer results in it
+ being larger than MAX_BUF_SIZE, then flag memory exhausted. */
+#define EXTEND_BUFFER() \
+ do { \
+ unsigned char *old_buffer = bufp->buffer; \
+ if (bufp->allocated == MAX_BUF_SIZE) \
+ return REG_ESIZE; \
+ bufp->allocated <<= 1; \
+ if (bufp->allocated > MAX_BUF_SIZE) \
+ bufp->allocated = MAX_BUF_SIZE; \
+ bufp->buffer = (unsigned char *) realloc (bufp->buffer, bufp->allocated);\
+ if (bufp->buffer == NULL) \
+ return REG_ESPACE; \
+ /* If the buffer moved, move all the pointers into it. */ \
+ if (old_buffer != bufp->buffer) \
+ { \
+ b = (b - old_buffer) + bufp->buffer; \
+ begalt = (begalt - old_buffer) + bufp->buffer; \
+ if (fixup_alt_jump) \
+ fixup_alt_jump = (fixup_alt_jump - old_buffer) + bufp->buffer;\
+ if (laststart) \
+ laststart = (laststart - old_buffer) + bufp->buffer; \
+ if (pending_exact) \
+ pending_exact = (pending_exact - old_buffer) + bufp->buffer; \
+ } \
+ } while (0)
+
+
+/* Since we have one byte reserved for the register number argument to
+ {start,stop}_memory, the maximum number of groups we can report
+ things about is what fits in that byte. */
+#define MAX_REGNUM 255
+
+/* But patterns can have more than `MAX_REGNUM' registers. We just
+ ignore the excess. */
+typedef unsigned regnum_t;
+
+
+/* Macros for the compile stack. */
+
+/* Since offsets can go either forwards or backwards, this type needs to
+ be able to hold values from -(MAX_BUF_SIZE - 1) to MAX_BUF_SIZE - 1. */
+typedef int pattern_offset_t;
+
+typedef struct
+{
+ pattern_offset_t begalt_offset;
+ pattern_offset_t fixup_alt_jump;
+ pattern_offset_t inner_group_offset;
+ pattern_offset_t laststart_offset;
+ regnum_t regnum;
+} compile_stack_elt_t;
+
+
+typedef struct
+{
+ compile_stack_elt_t *stack;
+ unsigned size;
+ unsigned avail; /* Offset of next open position. */
+} compile_stack_type;
+
+
+#define INIT_COMPILE_STACK_SIZE 32
+
+#define COMPILE_STACK_EMPTY (compile_stack.avail == 0)
+#define COMPILE_STACK_FULL (compile_stack.avail == compile_stack.size)
+
+/* The next available element. */
+#define COMPILE_STACK_TOP (compile_stack.stack[compile_stack.avail])
+
+
+/* Set the bit for character C in a list. */
+#define SET_LIST_BIT(c) \
+ (b[((unsigned char) (c)) / BYTEWIDTH] \
+ |= 1 << (((unsigned char) c) % BYTEWIDTH))
+
+
+/* Get the next unsigned number in the uncompiled pattern. */
+#define GET_UNSIGNED_NUMBER(num) \
+ { if (p != pend) \
+ { \
+ PATFETCH (c); \
+ while (ISDIGIT (c)) \
+ { \
+ if (num < 0) \
+ num = 0; \
+ num = num * 10 + c - '0'; \
+ if (p == pend) \
+ break; \
+ PATFETCH (c); \
+ } \
+ } \
+ }
+
+#define CHAR_CLASS_MAX_LENGTH 6 /* Namely, `xdigit'. */
+
+#define IS_CHAR_CLASS(string) \
+ (STREQ (string, "alpha") || STREQ (string, "upper") \
+ || STREQ (string, "lower") || STREQ (string, "digit") \
+ || STREQ (string, "alnum") || STREQ (string, "xdigit") \
+ || STREQ (string, "space") || STREQ (string, "print") \
+ || STREQ (string, "punct") || STREQ (string, "graph") \
+ || STREQ (string, "cntrl") || STREQ (string, "blank"))
+
+#ifndef MATCH_MAY_ALLOCATE
+
+/* If we cannot allocate large objects within re_match_2_internal,
+ we make the fail stack and register vectors global.
+ The fail stack, we grow to the maximum size when a regexp
+ is compiled.
+ The register vectors, we adjust in size each time we
+ compile a regexp, according to the number of registers it needs. */
+
+static fail_stack_type fail_stack;
+
+/* Size with which the following vectors are currently allocated.
+ That is so we can make them bigger as needed,
+ but never make them smaller. */
+static int regs_allocated_size;
+
+static const char ** regstart, ** regend;
+static const char ** old_regstart, ** old_regend;
+static const char **best_regstart, **best_regend;
+static register_info_type *reg_info;
+static const char **reg_dummy;
+static register_info_type *reg_info_dummy;
+
+/* Make the register vectors big enough for NUM_REGS registers,
+ but don't make them smaller. */
+
+static
+regex_grow_registers (num_regs)
+ int num_regs;
+{
+ if (num_regs > regs_allocated_size)
+ {
+ RETALLOC_IF (regstart, num_regs, const char *);
+ RETALLOC_IF (regend, num_regs, const char *);
+ RETALLOC_IF (old_regstart, num_regs, const char *);
+ RETALLOC_IF (old_regend, num_regs, const char *);
+ RETALLOC_IF (best_regstart, num_regs, const char *);
+ RETALLOC_IF (best_regend, num_regs, const char *);
+ RETALLOC_IF (reg_info, num_regs, register_info_type);
+ RETALLOC_IF (reg_dummy, num_regs, const char *);
+ RETALLOC_IF (reg_info_dummy, num_regs, register_info_type);
+
+ regs_allocated_size = num_regs;
+ }
+}
+
+#endif /* not MATCH_MAY_ALLOCATE */
+
+/* `regex_compile' compiles PATTERN (of length SIZE) according to SYNTAX.
+ Returns one of error codes defined in `regex.h', or zero for success.
+
+ Assumes the `allocated' (and perhaps `buffer') and `translate'
+ fields are set in BUFP on entry.
+
+ If it succeeds, results are put in BUFP (if it returns an error, the
+ contents of BUFP are undefined):
+ `buffer' is the compiled pattern;
+ `syntax' is set to SYNTAX;
+ `used' is set to the length of the compiled pattern;
+ `fastmap_accurate' is zero;
+ `re_nsub' is the number of subexpressions in PATTERN;
+ `not_bol' and `not_eol' are zero;
+
+ The `fastmap' and `newline_anchor' fields are neither
+ examined nor set. */
+
+/* Return, freeing storage we allocated. */
+#define FREE_STACK_RETURN(value) \
+ return (free (compile_stack.stack), value)
+
+static reg_errcode_t
+regex_compile (pattern, size, syntax, bufp)
+ const char *pattern;
+ int size;
+ reg_syntax_t syntax;
+ struct re_pattern_buffer *bufp;
+{
+ /* We fetch characters from PATTERN here. Even though PATTERN is
+ `char *' (i.e., signed), we declare these variables as unsigned, so
+ they can be reliably used as array indices. */
+ register unsigned char c, c1;
+
+ /* A random temporary spot in PATTERN. */
+ const char *p1;
+
+ /* Points to the end of the buffer, where we should append. */
+ register unsigned char *b;
+
+ /* Keeps track of unclosed groups. */
+ compile_stack_type compile_stack;
+
+ /* Points to the current (ending) position in the pattern. */
+ const char *p = pattern;
+ const char *pend = pattern + size;
+
+ /* How to translate the characters in the pattern. */
+ char *translate = bufp->translate;
+
+ /* Address of the count-byte of the most recently inserted `exactn'
+ command. This makes it possible to tell if a new exact-match
+ character can be added to that command or if the character requires
+ a new `exactn' command. */
+ unsigned char *pending_exact = 0;
+
+ /* Address of start of the most recently finished expression.
+ This tells, e.g., postfix * where to find the start of its
+ operand. Reset at the beginning of groups and alternatives. */
+ unsigned char *laststart = 0;
+
+ /* Address of beginning of regexp, or inside of last group. */
+ unsigned char *begalt;
+
+ /* Place in the uncompiled pattern (i.e., the {) to
+ which to go back if the interval is invalid. */
+ const char *beg_interval;
+
+ /* Address of the place where a forward jump should go to the end of
+ the containing expression. Each alternative of an `or' -- except the
+ last -- ends with a forward jump of this sort. */
+ unsigned char *fixup_alt_jump = 0;
+
+ /* Counts open-groups as they are encountered. Remembered for the
+ matching close-group on the compile stack, so the same register
+ number is put in the stop_memory as the start_memory. */
+ regnum_t regnum = 0;
+
+#ifdef DEBUG
+ DEBUG_PRINT1 ("\nCompiling pattern: ");
+ if (debug)
+ {
+ unsigned debug_count;
+
+ for (debug_count = 0; debug_count < size; debug_count++)
+ putchar (pattern[debug_count]);
+ putchar ('\n');
+ }
+#endif /* DEBUG */
+
+ /* Initialize the compile stack. */
+ compile_stack.stack = TALLOC (INIT_COMPILE_STACK_SIZE, compile_stack_elt_t);
+ if (compile_stack.stack == NULL)
+ return REG_ESPACE;
+
+ compile_stack.size = INIT_COMPILE_STACK_SIZE;
+ compile_stack.avail = 0;
+
+ /* Initialize the pattern buffer. */
+ bufp->syntax = syntax;
+ bufp->fastmap_accurate = 0;
+ bufp->not_bol = bufp->not_eol = 0;
+
+ /* Set `used' to zero, so that if we return an error, the pattern
+ printer (for debugging) will think there's no pattern. We reset it
+ at the end. */
+ bufp->used = 0;
+
+ /* Always count groups, whether or not bufp->no_sub is set. */
+ bufp->re_nsub = 0;
+
+#if !defined (emacs) && !defined (SYNTAX_TABLE)
+ /* Initialize the syntax table. */
+ init_syntax_once ();
+#endif
+
+ if (bufp->allocated == 0)
+ {
+ if (bufp->buffer)
+ { /* 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. */
+ RETALLOC (bufp->buffer, INIT_BUF_SIZE, unsigned char);
+ }
+ else
+ { /* Caller did not allocate a buffer. Do it for them. */
+ bufp->buffer = TALLOC (INIT_BUF_SIZE, unsigned char);
+ }
+ if (!bufp->buffer) FREE_STACK_RETURN (REG_ESPACE);
+
+ bufp->allocated = INIT_BUF_SIZE;
+ }
+
+ begalt = b = bufp->buffer;
+
+ /* Loop through the uncompiled pattern until we're at the end. */
+ while (p != pend)
+ {
+ PATFETCH (c);
+
+ switch (c)
+ {
+ case '^':
+ {
+ if ( /* If at start of pattern, it's an operator. */
+ p == pattern + 1
+ /* If context independent, it's an operator. */
+ || syntax & RE_CONTEXT_INDEP_ANCHORS
+ /* Otherwise, depends on what's come before. */
+ || at_begline_loc_p (pattern, p, syntax))
+ BUF_PUSH (begline);
+ else
+ goto normal_char;
+ }
+ break;
+
+
+ case '$':
+ {
+ if ( /* If at end of pattern, it's an operator. */
+ p == pend
+ /* If context independent, it's an operator. */
+ || syntax & RE_CONTEXT_INDEP_ANCHORS
+ /* Otherwise, depends on what's next. */
+ || at_endline_loc_p (p, pend, syntax))
+ BUF_PUSH (endline);
+ else
+ goto normal_char;
+ }
+ break;
+
+
+ case '+':
+ case '?':
+ if ((syntax & RE_BK_PLUS_QM)
+ || (syntax & RE_LIMITED_OPS))
+ goto normal_char;
+ handle_plus:
+ case '*':
+ /* If there is no previous pattern... */
+ if (!laststart)
+ {
+ if (syntax & RE_CONTEXT_INVALID_OPS)
+ FREE_STACK_RETURN (REG_BADRPT);
+ else if (!(syntax & RE_CONTEXT_INDEP_OPS))
+ goto normal_char;
+ }
+
+ {
+ /* Are we optimizing this jump? */
+ boolean keep_string_p = false;
+
+ /* 1 means zero (many) matches is allowed. */
+ char zero_times_ok = 0, many_times_ok = 0;
+
+ /* If there is a sequence of repetition chars, collapse it
+ down to just one (the right one). We can't combine
+ interval operators with these because of, e.g., `a{2}*',
+ which should only match an even number of `a's. */
+
+ for (;;)
+ {
+ zero_times_ok |= c != '+';
+ many_times_ok |= c != '?';
+
+ if (p == pend)
+ break;
+
+ PATFETCH (c);
+
+ if (c == '*'
+ || (!(syntax & RE_BK_PLUS_QM) && (c == '+' || c == '?')))
+ ;
+
+ else if (syntax & RE_BK_PLUS_QM && c == '\\')
+ {
+ if (p == pend) FREE_STACK_RETURN (REG_EESCAPE);
+
+ PATFETCH (c1);
+ if (!(c1 == '+' || c1 == '?'))
+ {
+ PATUNFETCH;
+ PATUNFETCH;
+ break;
+ }
+
+ c = c1;
+ }
+ else
+ {
+ PATUNFETCH;
+ break;
+ }
+
+ /* If we get here, we found another repeat character. */
+ }
+
+ /* Star, etc. applied to an empty pattern is equivalent
+ to an empty pattern. */
+ if (!laststart)
+ break;
+
+ /* Now we know whether or not zero matches is allowed
+ and also whether or not two or more matches is allowed. */
+ if (many_times_ok)
+ { /* More than one repetition is allowed, so put in at the
+ end a backward relative jump from `b' to before the next
+ jump we're going to put in below (which jumps from
+ laststart to after this jump).
+
+ But if we are at the `*' in the exact sequence `.*\n',
+ insert an unconditional jump backwards to the .,
+ instead of the beginning of the loop. This way we only
+ push a failure point once, instead of every time
+ through the loop. */
+ assert (p - 1 > pattern);
+
+ /* Allocate the space for the jump. */
+ GET_BUFFER_SPACE (3);
+
+ /* We know we are not at the first character of the pattern,
+ because laststart was nonzero. And we've already
+ incremented `p', by the way, to be the character after
+ the `*'. Do we have to do something analogous here
+ for null bytes, because of RE_DOT_NOT_NULL? */
+ if (TRANSLATE (*(p - 2)) == TRANSLATE ('.')
+ && zero_times_ok
+ && p < pend && TRANSLATE (*p) == TRANSLATE ('\n')
+ && !(syntax & RE_DOT_NEWLINE))
+ { /* We have .*\n. */
+ STORE_JUMP (jump, b, laststart);
+ keep_string_p = true;
+ }
+ else
+ /* Anything else. */
+ STORE_JUMP (maybe_pop_jump, b, laststart - 3);
+
+ /* We've added more stuff to the buffer. */
+ b += 3;
+ }
+
+ /* On failure, jump from laststart to b + 3, which will be the
+ end of the buffer after this jump is inserted. */
+ GET_BUFFER_SPACE (3);
+ INSERT_JUMP (keep_string_p ? on_failure_keep_string_jump
+ : on_failure_jump,
+ laststart, b + 3);
+ pending_exact = 0;
+ b += 3;
+
+ if (!zero_times_ok)
+ {
+ /* At least one repetition is required, so insert a
+ `dummy_failure_jump' before the initial
+ `on_failure_jump' instruction of the loop. This
+ effects a skip over that instruction the first time
+ we hit that loop. */
+ GET_BUFFER_SPACE (3);
+ INSERT_JUMP (dummy_failure_jump, laststart, laststart + 6);
+ b += 3;
+ }
+ }
+ break;
+
+
+ case '.':
+ laststart = b;
+ BUF_PUSH (anychar);
+ break;
+
+
+ case '[':
+ {
+ boolean had_char_class = false;
+
+ if (p == pend) FREE_STACK_RETURN (REG_EBRACK);
+
+ /* Ensure that we have enough space to push a charset: the
+ opcode, the length count, and the bitset; 34 bytes in all. */
+ GET_BUFFER_SPACE (34);
+
+ laststart = b;
+
+ /* We test `*p == '^' twice, instead of using an if
+ statement, so we only need one BUF_PUSH. */
+ BUF_PUSH (*p == '^' ? charset_not : charset);
+ if (*p == '^')
+ p++;
+
+ /* Remember the first position in the bracket expression. */
+ p1 = p;
+
+ /* Push the number of bytes in the bitmap. */
+ BUF_PUSH ((1 << BYTEWIDTH) / BYTEWIDTH);
+
+ /* Clear the whole map. */
+ bzero (b, (1 << BYTEWIDTH) / BYTEWIDTH);
+
+ /* charset_not matches newline according to a syntax bit. */
+ if ((re_opcode_t) b[-2] == charset_not
+ && (syntax & RE_HAT_LISTS_NOT_NEWLINE))
+ SET_LIST_BIT ('\n');
+
+ /* Read in characters and ranges, setting map bits. */
+ for (;;)
+ {
+ if (p == pend) FREE_STACK_RETURN (REG_EBRACK);
+
+ PATFETCH (c);
+
+ /* \ might escape characters inside [...] and [^...]. */
+ if ((syntax & RE_BACKSLASH_ESCAPE_IN_LISTS) && c == '\\')
+ {
+ if (p == pend) FREE_STACK_RETURN (REG_EESCAPE);
+
+ PATFETCH (c1);
+ SET_LIST_BIT (c1);
+ continue;
+ }
+
+ /* Could be the end of the bracket expression. If it's
+ not (i.e., when the bracket expression is `[]' so
+ far), the ']' character bit gets set way below. */
+ if (c == ']' && p != p1 + 1)
+ break;
+
+ /* Look ahead to see if it's a range when the last thing
+ was a character class. */
+ if (had_char_class && c == '-' && *p != ']')
+ FREE_STACK_RETURN (REG_ERANGE);
+
+ /* Look ahead to see if it's a range when the last thing
+ was a character: if this is a hyphen not at the
+ beginning or the end of a list, then it's the range
+ operator. */
+ if (c == '-'
+ && !(p - 2 >= pattern && p[-2] == '[')
+ && !(p - 3 >= pattern && p[-3] == '[' && p[-2] == '^')
+ && *p != ']')
+ {
+ reg_errcode_t ret
+ = compile_range (&p, pend, translate, syntax, b);
+ if (ret != REG_NOERROR) FREE_STACK_RETURN (ret);
+ }
+
+ else if (p[0] == '-' && p[1] != ']')
+ { /* This handles ranges made up of characters only. */
+ reg_errcode_t ret;
+
+ /* Move past the `-'. */
+ PATFETCH (c1);
+
+ ret = compile_range (&p, pend, translate, syntax, b);
+ if (ret != REG_NOERROR) FREE_STACK_RETURN (ret);
+ }
+
+ /* See if we're at the beginning of a possible character
+ class. */
+
+ else if (syntax & RE_CHAR_CLASSES && c == '[' && *p == ':')
+ { /* Leave room for the null. */
+ char str[CHAR_CLASS_MAX_LENGTH + 1];
+
+ PATFETCH (c);
+ c1 = 0;
+
+ /* If pattern is `[[:'. */
+ if (p == pend) FREE_STACK_RETURN (REG_EBRACK);
+
+ for (;;)
+ {
+ PATFETCH (c);
+ if (c == ':' || c == ']' || p == pend
+ || c1 == CHAR_CLASS_MAX_LENGTH)
+ break;
+ str[c1++] = c;
+ }
+ str[c1] = '\0';
+
+ /* If isn't a word bracketed by `[:' and:`]':
+ undo the ending character, the letters, and leave
+ the leading `:' and `[' (but set bits for them). */
+ if (c == ':' && *p == ']')
+ {
+ int ch;
+ boolean is_alnum = STREQ (str, "alnum");
+ boolean is_alpha = STREQ (str, "alpha");
+ boolean is_blank = STREQ (str, "blank");
+ boolean is_cntrl = STREQ (str, "cntrl");
+ boolean is_digit = STREQ (str, "digit");
+ boolean is_graph = STREQ (str, "graph");
+ boolean is_lower = STREQ (str, "lower");
+ boolean is_print = STREQ (str, "print");
+ boolean is_punct = STREQ (str, "punct");
+ boolean is_space = STREQ (str, "space");
+ boolean is_upper = STREQ (str, "upper");
+ boolean is_xdigit = STREQ (str, "xdigit");
+
+ if (!IS_CHAR_CLASS (str))
+ FREE_STACK_RETURN (REG_ECTYPE);
+
+ /* Throw away the ] at the end of the character
+ class. */
+ PATFETCH (c);
+
+ if (p == pend) FREE_STACK_RETURN (REG_EBRACK);
+
+ for (ch = 0; ch < 1 << BYTEWIDTH; ch++)
+ {
+ /* This was split into 3 if's to
+ avoid an arbitrary limit in some compiler. */
+ if ( (is_alnum && ISALNUM (ch))
+ || (is_alpha && ISALPHA (ch))
+ || (is_blank && ISBLANK (ch))
+ || (is_cntrl && ISCNTRL (ch)))
+ SET_LIST_BIT (ch);
+ if ( (is_digit && ISDIGIT (ch))
+ || (is_graph && ISGRAPH (ch))
+ || (is_lower && ISLOWER (ch))
+ || (is_print && ISPRINT (ch)))
+ SET_LIST_BIT (ch);
+ if ( (is_punct && ISPUNCT (ch))
+ || (is_space && ISSPACE (ch))
+ || (is_upper && ISUPPER (ch))
+ || (is_xdigit && ISXDIGIT (ch)))
+ SET_LIST_BIT (ch);
+ }
+ had_char_class = true;
+ }
+ else
+ {
+ c1++;
+ while (c1--)
+ PATUNFETCH;
+ SET_LIST_BIT ('[');
+ SET_LIST_BIT (':');
+ had_char_class = false;
+ }
+ }
+ else
+ {
+ had_char_class = false;
+ SET_LIST_BIT (c);
+ }
+ }
+
+ /* Discard any (non)matching list bytes that are all 0 at the
+ end of the map. Decrease the map-length byte too. */
+ while ((int) b[-1] > 0 && b[b[-1] - 1] == 0)
+ b[-1]--;
+ b += b[-1];
+ }
+ break;
+
+
+ case '(':
+ if (syntax & RE_NO_BK_PARENS)
+ goto handle_open;
+ else
+ goto normal_char;
+
+
+ case ')':
+ if (syntax & RE_NO_BK_PARENS)
+ goto handle_close;
+ else
+ goto normal_char;
+
+
+ case '\n':
+ if (syntax & RE_NEWLINE_ALT)
+ goto handle_alt;
+ else
+ goto normal_char;
+
+
+ case '|':
+ if (syntax & RE_NO_BK_VBAR)
+ goto handle_alt;
+ else
+ goto normal_char;
+
+
+ case '{':
+ if (syntax & RE_INTERVALS && syntax & RE_NO_BK_BRACES)
+ goto handle_interval;
+ else
+ goto normal_char;
+
+
+ case '\\':
+ if (p == pend) FREE_STACK_RETURN (REG_EESCAPE);
+
+ /* Do not translate the character after the \, so that we can
+ distinguish, e.g., \B from \b, even if we normally would
+ translate, e.g., B to b. */
+ PATFETCH_RAW (c);
+
+ switch (c)
+ {
+ case '(':
+ if (syntax & RE_NO_BK_PARENS)
+ goto normal_backslash;
+
+ handle_open:
+ bufp->re_nsub++;
+ regnum++;
+
+ if (COMPILE_STACK_FULL)
+ {
+ RETALLOC (compile_stack.stack, compile_stack.size << 1,
+ compile_stack_elt_t);
+ if (compile_stack.stack == NULL) return REG_ESPACE;
+
+ compile_stack.size <<= 1;
+ }
+
+ /* These are the values to restore when we hit end of this
+ group. They are all relative offsets, so that if the
+ whole pattern moves because of realloc, they will still
+ be valid. */
+ COMPILE_STACK_TOP.begalt_offset = begalt - bufp->buffer;
+ COMPILE_STACK_TOP.fixup_alt_jump
+ = fixup_alt_jump ? fixup_alt_jump - bufp->buffer + 1 : 0;
+ COMPILE_STACK_TOP.laststart_offset = b - bufp->buffer;
+ COMPILE_STACK_TOP.regnum = regnum;
+
+ /* We will eventually replace the 0 with the number of
+ groups inner to this one. But do not push a
+ start_memory for groups beyond the last one we can
+ represent in the compiled pattern. */
+ if (regnum <= MAX_REGNUM)
+ {
+ COMPILE_STACK_TOP.inner_group_offset = b - bufp->buffer + 2;
+ BUF_PUSH_3 (start_memory, regnum, 0);
+ }
+
+ compile_stack.avail++;
+
+ fixup_alt_jump = 0;
+ laststart = 0;
+ begalt = b;
+ /* If we've reached MAX_REGNUM groups, then this open
+ won't actually generate any code, so we'll have to
+ clear pending_exact explicitly. */
+ pending_exact = 0;
+ break;
+
+
+ case ')':
+ if (syntax & RE_NO_BK_PARENS) goto normal_backslash;
+
+ if (COMPILE_STACK_EMPTY)
+ if (syntax & RE_UNMATCHED_RIGHT_PAREN_ORD)
+ goto normal_backslash;
+ else
+ FREE_STACK_RETURN (REG_ERPAREN);
+
+ handle_close:
+ if (fixup_alt_jump)
+ { /* Push a dummy failure point at the end of the
+ alternative for a possible future
+ `pop_failure_jump' to pop. See comments at
+ `push_dummy_failure' in `re_match_2'. */
+ BUF_PUSH (push_dummy_failure);
+
+ /* We allocated space for this jump when we assigned
+ to `fixup_alt_jump', in the `handle_alt' case below. */
+ STORE_JUMP (jump_past_alt, fixup_alt_jump, b - 1);
+ }
+
+ /* See similar code for backslashed left paren above. */
+ if (COMPILE_STACK_EMPTY)
+ if (syntax & RE_UNMATCHED_RIGHT_PAREN_ORD)
+ goto normal_char;
+ else
+ FREE_STACK_RETURN (REG_ERPAREN);
+
+ /* Since we just checked for an empty stack above, this
+ ``can't happen''. */
+ assert (compile_stack.avail != 0);
+ {
+ /* We don't just want to restore into `regnum', because
+ later groups should continue to be numbered higher,
+ as in `(ab)c(de)' -- the second group is #2. */
+ regnum_t this_group_regnum;
+
+ compile_stack.avail--;
+ begalt = bufp->buffer + COMPILE_STACK_TOP.begalt_offset;
+ fixup_alt_jump
+ = COMPILE_STACK_TOP.fixup_alt_jump
+ ? bufp->buffer + COMPILE_STACK_TOP.fixup_alt_jump - 1
+ : 0;
+ laststart = bufp->buffer + COMPILE_STACK_TOP.laststart_offset;
+ this_group_regnum = COMPILE_STACK_TOP.regnum;
+ /* If we've reached MAX_REGNUM groups, then this open
+ won't actually generate any code, so we'll have to
+ clear pending_exact explicitly. */
+ pending_exact = 0;
+
+ /* We're at the end of the group, so now we know how many
+ groups were inside this one. */
+ if (this_group_regnum <= MAX_REGNUM)
+ {
+ unsigned char *inner_group_loc
+ = bufp->buffer + COMPILE_STACK_TOP.inner_group_offset;
+
+ *inner_group_loc = regnum - this_group_regnum;
+ BUF_PUSH_3 (stop_memory, this_group_regnum,
+ regnum - this_group_regnum);
+ }
+ }
+ break;
+
+
+ case '|': /* `\|'. */
+ if (syntax & RE_LIMITED_OPS || syntax & RE_NO_BK_VBAR)
+ goto normal_backslash;
+ handle_alt:
+ if (syntax & RE_LIMITED_OPS)
+ goto normal_char;
+
+ /* Insert before the previous alternative a jump which
+ jumps to this alternative if the former fails. */
+ GET_BUFFER_SPACE (3);
+ INSERT_JUMP (on_failure_jump, begalt, b + 6);
+ pending_exact = 0;
+ b += 3;
+
+ /* The alternative before this one has a jump after it
+ which gets executed if it gets matched. Adjust that
+ jump so it will jump to this alternative's analogous
+ jump (put in below, which in turn will jump to the next
+ (if any) alternative's such jump, etc.). The last such
+ jump jumps to the correct final destination. A picture:
+ _____ _____
+ | | | |
+ | v | v
+ a | b | c
+
+ If we are at `b', then fixup_alt_jump right now points to a
+ three-byte space after `a'. We'll put in the jump, set
+ fixup_alt_jump to right after `b', and leave behind three
+ bytes which we'll fill in when we get to after `c'. */
+
+ if (fixup_alt_jump)
+ STORE_JUMP (jump_past_alt, fixup_alt_jump, b);
+
+ /* Mark and leave space for a jump after this alternative,
+ to be filled in later either by next alternative or
+ when know we're at the end of a series of alternatives. */
+ fixup_alt_jump = b;
+ GET_BUFFER_SPACE (3);
+ b += 3;
+
+ laststart = 0;
+ begalt = b;
+ break;
+
+
+ case '{':
+ /* If \{ is a literal. */
+ if (!(syntax & RE_INTERVALS)
+ /* If we're at `\{' and it's not the open-interval
+ operator. */
+ || ((syntax & RE_INTERVALS) && (syntax & RE_NO_BK_BRACES))
+ || (p - 2 == pattern && p == pend))
+ goto normal_backslash;
+
+ handle_interval:
+ {
+ /* If got here, then the syntax allows intervals. */
+
+ /* At least (most) this many matches must be made. */
+ int lower_bound = -1, upper_bound = -1;
+
+ beg_interval = p - 1;
+
+ if (p == pend)
+ {
+ if (syntax & RE_NO_BK_BRACES)
+ goto unfetch_interval;
+ else
+ FREE_STACK_RETURN (REG_EBRACE);
+ }
+
+ GET_UNSIGNED_NUMBER (lower_bound);
+
+ if (c == ',')
+ {
+ GET_UNSIGNED_NUMBER (upper_bound);
+ if (upper_bound < 0) upper_bound = RE_DUP_MAX;
+ }
+ else
+ /* Interval such as `{1}' => match exactly once. */
+ upper_bound = lower_bound;
+
+ if (lower_bound < 0 || upper_bound > RE_DUP_MAX
+ || lower_bound > upper_bound)
+ {
+ if (syntax & RE_NO_BK_BRACES)
+ goto unfetch_interval;
+ else
+ FREE_STACK_RETURN (REG_BADBR);
+ }
+
+ if (!(syntax & RE_NO_BK_BRACES))
+ {
+ if (c != '\\') FREE_STACK_RETURN (REG_EBRACE);
+
+ PATFETCH (c);
+ }
+
+ if (c != '}')
+ {
+ if (syntax & RE_NO_BK_BRACES)
+ goto unfetch_interval;
+ else
+ FREE_STACK_RETURN (REG_BADBR);
+ }
+
+ /* We just parsed a valid interval. */
+
+ /* If it's invalid to have no preceding re. */
+ if (!laststart)
+ {
+ if (syntax & RE_CONTEXT_INVALID_OPS)
+ FREE_STACK_RETURN (REG_BADRPT);
+ else if (syntax & RE_CONTEXT_INDEP_OPS)
+ laststart = b;
+ else
+ goto unfetch_interval;
+ }
+
+ /* If the upper bound is zero, don't want to succeed at
+ all; jump from `laststart' to `b + 3', which will be
+ the end of the buffer after we insert the jump. */
+ if (upper_bound == 0)
+ {
+ GET_BUFFER_SPACE (3);
+ INSERT_JUMP (jump, laststart, b + 3);
+ b += 3;
+ }
+
+ /* Otherwise, we have a nontrivial interval. When
+ we're all done, the pattern will look like:
+ set_number_at <jump count> <upper bound>
+ set_number_at <succeed_n count> <lower bound>
+ succeed_n <after jump addr> <succeed_n count>
+ <body of loop>
+ jump_n <succeed_n addr> <jump count>
+ (The upper bound and `jump_n' are omitted if
+ `upper_bound' is 1, though.) */
+ else
+ { /* If the upper bound is > 1, we need to insert
+ more at the end of the loop. */
+ unsigned nbytes = 10 + (upper_bound > 1) * 10;
+
+ GET_BUFFER_SPACE (nbytes);
+
+ /* Initialize lower bound of the `succeed_n', even
+ though it will be set during matching by its
+ attendant `set_number_at' (inserted next),
+ because `re_compile_fastmap' needs to know.
+ Jump to the `jump_n' we might insert below. */
+ INSERT_JUMP2 (succeed_n, laststart,
+ b + 5 + (upper_bound > 1) * 5,
+ lower_bound);
+ b += 5;
+
+ /* Code to initialize the lower bound. Insert
+ before the `succeed_n'. The `5' is the last two
+ bytes of this `set_number_at', plus 3 bytes of
+ the following `succeed_n'. */
+ insert_op2 (set_number_at, laststart, 5, lower_bound, b);
+ b += 5;
+
+ if (upper_bound > 1)
+ { /* More than one repetition is allowed, so
+ append a backward jump to the `succeed_n'
+ that starts this interval.
+
+ When we've reached this during matching,
+ we'll have matched the interval once, so
+ jump back only `upper_bound - 1' times. */
+ STORE_JUMP2 (jump_n, b, laststart + 5,
+ upper_bound - 1);
+ b += 5;
+
+ /* The location we want to set is the second
+ parameter of the `jump_n'; that is `b-2' as
+ an absolute address. `laststart' will be
+ the `set_number_at' we're about to insert;
+ `laststart+3' the number to set, the source
+ for the relative address. But we are
+ inserting into the middle of the pattern --
+ so everything is getting moved up by 5.
+ Conclusion: (b - 2) - (laststart + 3) + 5,
+ i.e., b - laststart.
+
+ We insert this at the beginning of the loop
+ so that if we fail during matching, we'll
+ reinitialize the bounds. */
+ insert_op2 (set_number_at, laststart, b - laststart,
+ upper_bound - 1, b);
+ b += 5;
+ }
+ }
+ pending_exact = 0;
+ beg_interval = NULL;
+ }
+ break;
+
+ unfetch_interval:
+ /* If an invalid interval, match the characters as literals. */
+ assert (beg_interval);
+ p = beg_interval;
+ beg_interval = NULL;
+
+ /* normal_char and normal_backslash need `c'. */
+ PATFETCH (c);
+
+ if (!(syntax & RE_NO_BK_BRACES))
+ {
+ if (p > pattern && p[-1] == '\\')
+ goto normal_backslash;
+ }
+ goto normal_char;
+
+#ifdef emacs
+ /* There is no way to specify the before_dot and after_dot
+ operators. rms says this is ok. --karl */
+ case '=':
+ BUF_PUSH (at_dot);
+ break;
+
+ case 's':
+ laststart = b;
+ PATFETCH (c);
+ BUF_PUSH_2 (syntaxspec, syntax_spec_code[c]);
+ break;
+
+ case 'S':
+ laststart = b;
+ PATFETCH (c);
+ BUF_PUSH_2 (notsyntaxspec, syntax_spec_code[c]);
+ break;
+#endif /* emacs */
+
+
+ case 'w':
+ laststart = b;
+ BUF_PUSH (wordchar);
+ break;
+
+
+ case 'W':
+ laststart = b;
+ BUF_PUSH (notwordchar);
+ break;
+
+
+ case '<':
+ BUF_PUSH (wordbeg);
+ break;
+
+ case '>':
+ BUF_PUSH (wordend);
+ break;
+
+ case 'b':
+ BUF_PUSH (wordbound);
+ break;
+
+ case 'B':
+ BUF_PUSH (notwordbound);
+ break;
+
+ case '`':
+ BUF_PUSH (begbuf);
+ break;
+
+ case '\'':
+ BUF_PUSH (endbuf);
+ 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)
+ goto normal_char;
+
+ c1 = c - '0';
+
+ if (c1 > regnum)
+ FREE_STACK_RETURN (REG_ESUBREG);
+
+ /* Can't back reference to a subexpression if inside of it. */
+ if (group_in_compile_stack (compile_stack, c1))
+ goto normal_char;
+
+ laststart = b;
+ BUF_PUSH_2 (duplicate, c1);
+ break;
+
+
+ case '+':
+ case '?':
+ if (syntax & RE_BK_PLUS_QM)
+ goto handle_plus;
+ else
+ goto normal_backslash;
+
+ default:
+ normal_backslash:
+ /* You might think it would be useful for \ to mean
+ not to translate; but if we don't translate it
+ it will never match anything. */
+ c = TRANSLATE (c);
+ goto normal_char;
+ }
+ break;
+
+
+ default:
+ /* Expects the character in `c'. */
+ normal_char:
+ /* If no exactn currently being built. */
+ if (!pending_exact
+
+ /* If last exactn not at current position. */
+ || pending_exact + *pending_exact + 1 != b
+
+ /* We have only one byte following the exactn for the count. */
+ || *pending_exact == (1 << BYTEWIDTH) - 1
+
+ /* If followed by a repetition operator. */
+ || *p == '*' || *p == '^'
+ || ((syntax & RE_BK_PLUS_QM)
+ ? *p == '\\' && (p[1] == '+' || p[1] == '?')
+ : (*p == '+' || *p == '?'))
+ || ((syntax & RE_INTERVALS)
+ && ((syntax & RE_NO_BK_BRACES)
+ ? *p == '{'
+ : (p[0] == '\\' && p[1] == '{'))))
+ {
+ /* Start building a new exactn. */
+
+ laststart = b;
+
+ BUF_PUSH_2 (exactn, 0);
+ pending_exact = b - 1;
+ }
+
+ BUF_PUSH (c);
+ (*pending_exact)++;
+ break;
+ } /* switch (c) */
+ } /* while p != pend */
+
+
+ /* Through the pattern now. */
+
+ if (fixup_alt_jump)
+ STORE_JUMP (jump_past_alt, fixup_alt_jump, b);
+
+ if (!COMPILE_STACK_EMPTY)
+ FREE_STACK_RETURN (REG_EPAREN);
+
+ /* If we don't want backtracking, force success
+ the first time we reach the end of the compiled pattern. */
+ if (syntax & RE_NO_POSIX_BACKTRACKING)
+ BUF_PUSH (succeed);
+
+ free (compile_stack.stack);
+
+ /* We have succeeded; set the length of the buffer. */
+ bufp->used = b - bufp->buffer;
+
+#ifdef DEBUG
+ if (debug)
+ {
+ DEBUG_PRINT1 ("\nCompiled pattern: \n");
+ print_compiled_pattern (bufp);
+ }
+#endif /* DEBUG */
+
+#ifndef MATCH_MAY_ALLOCATE
+ /* Initialize the failure stack to the largest possible stack. This
+ isn't necessary unless we're trying to avoid calling alloca in
+ the search and match routines. */
+ {
+ int num_regs = bufp->re_nsub + 1;
+
+ /* Since DOUBLE_FAIL_STACK refuses to double only if the current size
+ is strictly greater than re_max_failures, the largest possible stack
+ is 2 * re_max_failures failure points. */
+ if (fail_stack.size < (2 * re_max_failures * MAX_FAILURE_ITEMS))
+ {
+ fail_stack.size = (2 * re_max_failures * MAX_FAILURE_ITEMS);
+
+#ifdef emacs
+ if (! fail_stack.stack)
+ fail_stack.stack
+ = (fail_stack_elt_t *) xmalloc (fail_stack.size
+ * sizeof (fail_stack_elt_t));
+ else
+ fail_stack.stack
+ = (fail_stack_elt_t *) xrealloc (fail_stack.stack,
+ (fail_stack.size
+ * sizeof (fail_stack_elt_t)));
+#else /* not emacs */
+ if (! fail_stack.stack)
+ fail_stack.stack
+ = (fail_stack_elt_t *) malloc (fail_stack.size
+ * sizeof (fail_stack_elt_t));
+ else
+ fail_stack.stack
+ = (fail_stack_elt_t *) realloc (fail_stack.stack,
+ (fail_stack.size
+ * sizeof (fail_stack_elt_t)));
+#endif /* not emacs */
+ }
+
+ regex_grow_registers (num_regs);
+ }
+#endif /* not MATCH_MAY_ALLOCATE */
+
+ return REG_NOERROR;
+} /* regex_compile */
+
+/* Subroutines for `regex_compile'. */
+
+/* Store OP at LOC followed by two-byte integer parameter ARG. */
+
+static void
+store_op1 (op, loc, arg)
+ re_opcode_t op;
+ unsigned char *loc;
+ int arg;
+{
+ *loc = (unsigned char) op;
+ STORE_NUMBER (loc + 1, arg);
+}
+
+
+/* Like `store_op1', but for two two-byte parameters ARG1 and ARG2. */
+
+static void
+store_op2 (op, loc, arg1, arg2)
+ re_opcode_t op;
+ unsigned char *loc;
+ int arg1, arg2;
+{
+ *loc = (unsigned char) op;
+ STORE_NUMBER (loc + 1, arg1);
+ STORE_NUMBER (loc + 3, arg2);
+}
+
+
+/* Copy the bytes from LOC to END to open up three bytes of space at LOC
+ for OP followed by two-byte integer parameter ARG. */
+
+static void
+insert_op1 (op, loc, arg, end)
+ re_opcode_t op;
+ unsigned char *loc;
+ int arg;
+ unsigned char *end;
+{
+ register unsigned char *pfrom = end;
+ register unsigned char *pto = end + 3;
+
+ while (pfrom != loc)
+ *--pto = *--pfrom;
+
+ store_op1 (op, loc, arg);
+}
+
+
+/* Like `insert_op1', but for two two-byte parameters ARG1 and ARG2. */
+
+static void
+insert_op2 (op, loc, arg1, arg2, end)
+ re_opcode_t op;
+ unsigned char *loc;
+ int arg1, arg2;
+ unsigned char *end;
+{
+ register unsigned char *pfrom = end;
+ register unsigned char *pto = end + 5;
+
+ while (pfrom != loc)
+ *--pto = *--pfrom;
+
+ store_op2 (op, loc, arg1, arg2);
+}
+
+
+/* P points to just after a ^ in PATTERN. Return true if that ^ comes
+ after an alternative or a begin-subexpression. We assume there is at
+ least one character before the ^. */
+
+static boolean
+at_begline_loc_p (pattern, p, syntax)
+ const char *pattern, *p;
+ reg_syntax_t syntax;
+{
+ const char *prev = p - 2;
+ boolean prev_prev_backslash = prev > pattern && prev[-1] == '\\';
+
+ return
+ /* After a subexpression? */
+ (*prev == '(' && (syntax & RE_NO_BK_PARENS || prev_prev_backslash))
+ /* After an alternative? */
+ || (*prev == '|' && (syntax & RE_NO_BK_VBAR || prev_prev_backslash));
+}
+
+
+/* The dual of at_begline_loc_p. This one is for $. We assume there is
+ at least one character after the $, i.e., `P < PEND'. */
+
+static boolean
+at_endline_loc_p (p, pend, syntax)
+ const char *p, *pend;
+ int syntax;
+{
+ const char *next = p;
+ boolean next_backslash = *next == '\\';
+ const char *next_next = p + 1 < pend ? p + 1 : NULL;
+
+ return
+ /* Before a subexpression? */
+ (syntax & RE_NO_BK_PARENS ? *next == ')'
+ : next_backslash && next_next && *next_next == ')')
+ /* Before an alternative? */
+ || (syntax & RE_NO_BK_VBAR ? *next == '|'
+ : next_backslash && next_next && *next_next == '|');
+}
+
+
+/* Returns true if REGNUM is in one of COMPILE_STACK's elements and
+ false if it's not. */
+
+static boolean
+group_in_compile_stack (compile_stack, regnum)
+ compile_stack_type compile_stack;
+ regnum_t regnum;
+{
+ int this_element;
+
+ for (this_element = compile_stack.avail - 1;
+ this_element >= 0;
+ this_element--)
+ if (compile_stack.stack[this_element].regnum == regnum)
+ return true;
+
+ return false;
+}
+
+
+/* Read the ending character of a range (in a bracket expression) from the
+ uncompiled pattern *P_PTR (which ends at PEND). We assume the
+ starting character is in `P[-2]'. (`P[-1]' is the character `-'.)
+ Then we set the translation of all bits between the starting and
+ ending characters (inclusive) in the compiled pattern B.
+
+ Return an error code.
+
+ We use these short variable names so we can use the same macros as
+ `regex_compile' itself. */
+
+static reg_errcode_t
+compile_range (p_ptr, pend, translate, syntax, b)
+ const char **p_ptr, *pend;
+ char *translate;
+ reg_syntax_t syntax;
+ unsigned char *b;
+{
+ unsigned this_char;
+
+ const char *p = *p_ptr;
+ int range_start, range_end;
+
+ if (p == pend)
+ return REG_ERANGE;
+
+ /* Even though the pattern is a signed `char *', we need to fetch
+ with unsigned char *'s; if the high bit of the pattern character
+ is set, the range endpoints will be negative if we fetch using a
+ signed char *.
+
+ We also want to fetch the endpoints without translating them; the
+ appropriate translation is done in the bit-setting loop below. */
+ /* The SVR4 compiler on the 3B2 had trouble with unsigned const char *. */
+ range_start = ((const unsigned char *) p)[-2];
+ range_end = ((const unsigned char *) p)[0];
+
+ /* Have to increment the pointer into the pattern string, so the
+ caller isn't still at the ending character. */
+ (*p_ptr)++;
+
+ /* If the start is after the end, the range is empty. */
+ if (range_start > range_end)
+ return syntax & RE_NO_EMPTY_RANGES ? REG_ERANGE : REG_NOERROR;
+
+ /* Here we see why `this_char' has to be larger than an `unsigned
+ char' -- the range is inclusive, so if `range_end' == 0xff
+ (assuming 8-bit characters), we would otherwise go into an infinite
+ loop, since all characters <= 0xff. */
+ for (this_char = range_start; this_char <= range_end; this_char++)
+ {
+ SET_LIST_BIT (TRANSLATE (this_char));
+ }
+
+ return REG_NOERROR;
+}
+
+/* re_compile_fastmap computes a ``fastmap'' for the compiled pattern in
+ BUFP. A fastmap records which of the (1 << BYTEWIDTH) possible
+ characters can start a string that matches the pattern. This fastmap
+ is used by re_search to skip quickly over impossible starting points.
+
+ The caller must supply the address of a (1 << BYTEWIDTH)-byte data
+ area as BUFP->fastmap.
+
+ We set the `fastmap', `fastmap_accurate', and `can_be_null' fields in
+ the pattern buffer.
+
+ Returns 0 if we succeed, -2 if an internal error. */
+
+int
+re_compile_fastmap (bufp)
+ struct re_pattern_buffer *bufp;
+{
+ int j, k;
+#ifdef MATCH_MAY_ALLOCATE
+ fail_stack_type fail_stack;
+#endif
+#ifndef REGEX_MALLOC
+ char *destination;
+#endif
+ /* We don't push any register information onto the failure stack. */
+ unsigned num_regs = 0;
+
+ register char *fastmap = bufp->fastmap;
+ unsigned char *pattern = bufp->buffer;
+ unsigned long size = bufp->used;
+ unsigned char *p = pattern;
+ register unsigned char *pend = pattern + size;
+
+ /* This holds the pointer to the failure stack, when
+ it is allocated relocatably. */
+ fail_stack_elt_t *failure_stack_ptr;
+
+ /* Assume that each path through the pattern can be null until
+ proven otherwise. We set this false at the bottom of switch
+ statement, to which we get only if a particular path doesn't
+ match the empty string. */
+ boolean path_can_be_null = true;
+
+ /* We aren't doing a `succeed_n' to begin with. */
+ boolean succeed_n_p = false;
+
+ assert (fastmap != NULL && p != NULL);
+
+ INIT_FAIL_STACK ();
+ bzero (fastmap, 1 << BYTEWIDTH); /* Assume nothing's valid. */
+ bufp->fastmap_accurate = 1; /* It will be when we're done. */
+ bufp->can_be_null = 0;
+
+ while (1)
+ {
+ if (p == pend || *p == succeed)
+ {
+ /* We have reached the (effective) end of pattern. */
+ if (!FAIL_STACK_EMPTY ())
+ {
+ bufp->can_be_null |= path_can_be_null;
+
+ /* Reset for next path. */
+ path_can_be_null = true;
+
+ p = fail_stack.stack[--fail_stack.avail].pointer;
+
+ continue;
+ }
+ else
+ break;
+ }
+
+ /* We should never be about to go beyond the end of the pattern. */
+ assert (p < pend);
+
+ switch (SWITCH_ENUM_CAST ((re_opcode_t) *p++))
+ {
+
+ /* I guess the idea here is to simply not bother with a fastmap
+ if a backreference is used, since it's too hard to figure out
+ the fastmap for the corresponding group. Setting
+ `can_be_null' stops `re_search_2' from using the fastmap, so
+ that is all we do. */
+ case duplicate:
+ bufp->can_be_null = 1;
+ goto done;
+
+
+ /* Following are the cases which match a character. These end
+ with `break'. */
+
+ case exactn:
+ fastmap[p[1]] = 1;
+ break;
+
+
+ case charset:
+ for (j = *p++ * BYTEWIDTH - 1; j >= 0; j--)
+ if (p[j / BYTEWIDTH] & (1 << (j % BYTEWIDTH)))
+ fastmap[j] = 1;
+ break;
+
+
+ case charset_not:
+ /* Chars beyond end of map must be allowed. */
+ for (j = *p * BYTEWIDTH; j < (1 << BYTEWIDTH); j++)
+ fastmap[j] = 1;
+
+ for (j = *p++ * BYTEWIDTH - 1; j >= 0; j--)
+ if (!(p[j / BYTEWIDTH] & (1 << (j % BYTEWIDTH))))
+ fastmap[j] = 1;
+ break;
+
+
+ case wordchar:
+ for (j = 0; j < (1 << BYTEWIDTH); j++)
+ if (SYNTAX (j) == Sword)
+ fastmap[j] = 1;
+ break;
+
+
+ case notwordchar:
+ for (j = 0; j < (1 << BYTEWIDTH); j++)
+ if (SYNTAX (j) != Sword)
+ fastmap[j] = 1;
+ break;
+
+
+ case anychar:
+ {
+ int fastmap_newline = fastmap['\n'];
+
+ /* `.' matches anything ... */
+ for (j = 0; j < (1 << BYTEWIDTH); j++)
+ fastmap[j] = 1;
+
+ /* ... except perhaps newline. */
+ if (!(bufp->syntax & RE_DOT_NEWLINE))
+ fastmap['\n'] = fastmap_newline;
+
+ /* Return if we have already set `can_be_null'; if we have,
+ then the fastmap is irrelevant. Something's wrong here. */
+ else if (bufp->can_be_null)
+ goto done;
+
+ /* Otherwise, have to check alternative paths. */
+ break;
+ }
+
+#ifdef emacs
+ case syntaxspec:
+ k = *p++;
+ for (j = 0; j < (1 << BYTEWIDTH); j++)
+ if (SYNTAX (j) == (enum syntaxcode) k)
+ fastmap[j] = 1;
+ break;
+
+
+ case notsyntaxspec:
+ k = *p++;
+ for (j = 0; j < (1 << BYTEWIDTH); j++)
+ if (SYNTAX (j) != (enum syntaxcode) k)
+ fastmap[j] = 1;
+ break;
+
+
+ /* All cases after this match the empty string. These end with
+ `continue'. */
+
+
+ case before_dot:
+ case at_dot:
+ case after_dot:
+ continue;
+#endif /* not emacs */
+
+
+ case no_op:
+ case begline:
+ case endline:
+ case begbuf:
+ case endbuf:
+ case wordbound:
+ case notwordbound:
+ case wordbeg:
+ case wordend:
+ case push_dummy_failure:
+ continue;
+
+
+ case jump_n:
+ case pop_failure_jump:
+ case maybe_pop_jump:
+ case jump:
+ case jump_past_alt:
+ case dummy_failure_jump:
+ EXTRACT_NUMBER_AND_INCR (j, p);
+ p += j;
+ if (j > 0)
+ continue;
+
+ /* Jump backward implies we just went through the body of a
+ loop and matched nothing. Opcode jumped to should be
+ `on_failure_jump' or `succeed_n'. Just treat it like an
+ ordinary jump. For a * loop, it has pushed its failure
+ point already; if so, discard that as redundant. */
+ if ((re_opcode_t) *p != on_failure_jump
+ && (re_opcode_t) *p != succeed_n)
+ continue;
+
+ p++;
+ EXTRACT_NUMBER_AND_INCR (j, p);
+ p += j;
+
+ /* If what's on the stack is where we are now, pop it. */
+ if (!FAIL_STACK_EMPTY ()
+ && fail_stack.stack[fail_stack.avail - 1].pointer == p)
+ fail_stack.avail--;
+
+ continue;
+
+
+ case on_failure_jump:
+ case on_failure_keep_string_jump:
+ handle_on_failure_jump:
+ EXTRACT_NUMBER_AND_INCR (j, p);
+
+ /* For some patterns, e.g., `(a?)?', `p+j' here points to the
+ end of the pattern. We don't want to push such a point,
+ since when we restore it above, entering the switch will
+ increment `p' past the end of the pattern. We don't need
+ to push such a point since we obviously won't find any more
+ fastmap entries beyond `pend'. Such a pattern can match
+ the null string, though. */
+ if (p + j < pend)
+ {
+ if (!PUSH_PATTERN_OP (p + j, fail_stack))
+ {
+ RESET_FAIL_STACK ();
+ return -2;
+ }
+ }
+ else
+ bufp->can_be_null = 1;
+
+ if (succeed_n_p)
+ {
+ EXTRACT_NUMBER_AND_INCR (k, p); /* Skip the n. */
+ succeed_n_p = false;
+ }
+
+ continue;
+
+
+ case succeed_n:
+ /* Get to the number of times to succeed. */
+ p += 2;
+
+ /* Increment p past the n for when k != 0. */
+ EXTRACT_NUMBER_AND_INCR (k, p);
+ if (k == 0)
+ {
+ p -= 4;
+ succeed_n_p = true; /* Spaghetti code alert. */
+ goto handle_on_failure_jump;
+ }
+ continue;
+
+
+ case set_number_at:
+ p += 4;
+ continue;
+
+
+ case start_memory:
+ case stop_memory:
+ p += 2;
+ continue;
+
+
+ default:
+ abort (); /* We have listed all the cases. */
+ } /* switch *p++ */
+
+ /* Getting here means we have found the possible starting
+ characters for one path of the pattern -- and that the empty
+ string does not match. We need not follow this path further.
+ Instead, look at the next alternative (remembered on the
+ stack), or quit if no more. The test at the top of the loop
+ does these things. */
+ path_can_be_null = false;
+ p = pend;
+ } /* while p */
+
+ /* Set `can_be_null' for the last path (also the first path, if the
+ pattern is empty). */
+ bufp->can_be_null |= path_can_be_null;
+
+ done:
+ RESET_FAIL_STACK ();
+ return 0;
+} /* re_compile_fastmap */
+
+/* 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 (bufp, regs, num_regs, starts, ends)
+ struct re_pattern_buffer *bufp;
+ struct re_registers *regs;
+ unsigned num_regs;
+ regoff_t *starts, *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 = (regoff_t *) 0;
+ }
+}
+
+/* Searching routines. */
+
+/* Like re_search_2, below, but only one string is specified, and
+ doesn't let you say where to stop matching. */
+
+int
+re_search (bufp, string, size, startpos, range, regs)
+ struct re_pattern_buffer *bufp;
+ const char *string;
+ int size, startpos, range;
+ struct re_registers *regs;
+{
+ return re_search_2 (bufp, NULL, 0, string, size, startpos, range,
+ regs, size);
+}
+
+
+/* Using the compiled pattern in BUFP->buffer, first tries to match the
+ virtual concatenation of STRING1 and STRING2, starting first at index
+ STARTPOS, then at STARTPOS + 1, and so on.
+
+ STRING1 and STRING2 have length SIZE1 and SIZE2, respectively.
+
+ RANGE is how far to scan while trying to match. RANGE = 0 means try
+ only at STARTPOS; in general, the last start tried is STARTPOS +
+ RANGE.
+
+ In REGS, return the indices of the virtual concatenation of STRING1
+ and STRING2 that matched the entire BUFP->buffer and its contained
+ subexpressions.
+
+ Do not consider matching one past the index STOP in the virtual
+ concatenation of STRING1 and STRING2.
+
+ We return either the position in the strings at which the match was
+ found, -1 if no match, or -2 if error (such as failure
+ stack overflow). */
+
+int
+re_search_2 (bufp, string1, size1, string2, size2, startpos, range, regs, stop)
+ struct re_pattern_buffer *bufp;
+ const char *string1, *string2;
+ int size1, size2;
+ int startpos;
+ int range;
+ struct re_registers *regs;
+ int stop;
+{
+ int val;
+ register char *fastmap = bufp->fastmap;
+ register char *translate = bufp->translate;
+ int total_size = size1 + size2;
+ int endpos = startpos + range;
+
+ /* Check for out-of-range STARTPOS. */
+ if (startpos < 0 || startpos > total_size)
+ return -1;
+
+ /* Fix up RANGE if it might eventually take us outside
+ the virtual concatenation of STRING1 and STRING2. */
+ if (endpos < -1)
+ range = -1 - startpos;
+ else if (endpos > total_size)
+ range = total_size - startpos;
+
+ /* If the search isn't to be a backwards one, don't waste time in a
+ search for a pattern that must be anchored. */
+ if (bufp->used > 0 && (re_opcode_t) bufp->buffer[0] == begbuf && range > 0)
+ {
+ if (startpos > 0)
+ return -1;
+ else
+ range = 1;
+ }
+
+ /* Update the fastmap now if not correct already. */
+ if (fastmap && !bufp->fastmap_accurate)
+ if (re_compile_fastmap (bufp) == -2)
+ return -2;
+
+ /* Loop through the string, looking for a place to start matching. */
+ for (;;)
+ {
+ /* If a fastmap is supplied, skip quickly over characters that
+ cannot be the start of a match. If the pattern can match the
+ null string, however, we don't need to skip characters; we want
+ the first null string. */
+ if (fastmap && startpos < total_size && !bufp->can_be_null)
+ {
+ if (range > 0) /* Searching forwards. */
+ {
+ register const char *d;
+ register int lim = 0;
+ int irange = range;
+
+ if (startpos < size1 && startpos + range >= size1)
+ lim = range - (size1 - startpos);
+
+ d = (startpos >= size1 ? string2 - size1 : string1) + startpos;
+
+ /* Written out as an if-else to avoid testing `translate'
+ inside the loop. */
+ if (translate)
+ while (range > lim
+ && !fastmap[(unsigned char)
+ translate[(unsigned char) *d++]])
+ range--;
+ else
+ while (range > lim && !fastmap[(unsigned char) *d++])
+ range--;
+
+ startpos += irange - range;
+ }
+ else /* Searching backwards. */
+ {
+ register char c = (size1 == 0 || startpos >= size1
+ ? string2[startpos - size1]
+ : string1[startpos]);
+
+ if (!fastmap[(unsigned char) TRANSLATE (c)])
+ goto advance;
+ }
+ }
+
+ /* If can't match the null string, and that's all we have left, fail. */
+ if (range >= 0 && startpos == total_size && fastmap
+ && !bufp->can_be_null)
+ return -1;
+
+ val = re_match_2_internal (bufp, string1, size1, string2, size2,
+ startpos, regs, stop);
+#ifndef REGEX_MALLOC
+#ifdef C_ALLOCA
+ alloca (0);
+#endif
+#endif
+
+ if (val >= 0)
+ return startpos;
+
+ if (val == -2)
+ return -2;
+
+ advance:
+ if (!range)
+ break;
+ else if (range > 0)
+ {
+ range--;
+ startpos++;
+ }
+ else
+ {
+ range++;
+ startpos--;
+ }
+ }
+ return -1;
+} /* re_search_2 */
+
+/* Declarations and macros for re_match_2. */
+
+static int bcmp_translate ();
+static boolean alt_match_null_string_p (),
+ common_op_match_null_string_p (),
+ group_match_null_string_p ();
+
+/* This converts PTR, a pointer into one of the search strings `string1'
+ and `string2' into an offset from the beginning of that string. */
+#define POINTER_TO_OFFSET(ptr) \
+ (FIRST_STRING_P (ptr) \
+ ? ((regoff_t) ((ptr) - string1)) \
+ : ((regoff_t) ((ptr) - string2 + size1)))
+
+/* Macros for dealing with the split strings in re_match_2. */
+
+#define MATCHING_IN_FIRST_STRING (dend == end_match_1)
+
+/* Call before fetching a character with *d. This switches over to
+ string2 if necessary. */
+#define PREFETCH() \
+ while (d == dend) \
+ { \
+ /* End of string2 => fail. */ \
+ if (dend == end_match_2) \
+ goto fail; \
+ /* End of string1 => advance to string2. */ \
+ d = string2; \
+ dend = end_match_2; \
+ }
+
+
+/* Test if at very beginning or at very end of the virtual concatenation
+ of `string1' and `string2'. If only one string, it's `string2'. */
+#define AT_STRINGS_BEG(d) ((d) == (size1 ? string1 : string2) || !size2)
+#define AT_STRINGS_END(d) ((d) == end2)
+
+
+/* Test if D points to a character which is word-constituent. We have
+ two special cases to check for: if past the end of string1, look at
+ the first character in string2; and if before the beginning of
+ string2, look at the last character in string1. */
+#define WORDCHAR_P(d) \
+ (SYNTAX ((d) == end1 ? *string2 \
+ : (d) == string2 - 1 ? *(end1 - 1) : *(d)) \
+ == Sword)
+
+/* Test if the character before D and the one at D differ with respect
+ to being word-constituent. */
+#define AT_WORD_BOUNDARY(d) \
+ (AT_STRINGS_BEG (d) || AT_STRINGS_END (d) \
+ || WORDCHAR_P (d - 1) != WORDCHAR_P (d))
+
+
+/* Free everything we malloc. */
+#ifdef MATCH_MAY_ALLOCATE
+#define FREE_VAR(var) if (var) REGEX_FREE (var); var = NULL
+#define FREE_VARIABLES() \
+ do { \
+ REGEX_FREE_STACK (fail_stack.stack); \
+ FREE_VAR (regstart); \
+ FREE_VAR (regend); \
+ FREE_VAR (old_regstart); \
+ FREE_VAR (old_regend); \
+ FREE_VAR (best_regstart); \
+ FREE_VAR (best_regend); \
+ FREE_VAR (reg_info); \
+ FREE_VAR (reg_dummy); \
+ FREE_VAR (reg_info_dummy); \
+ } while (0)
+#else
+#define FREE_VARIABLES() ((void)0) /* Do nothing! But inhibit gcc warning. */
+#endif /* not MATCH_MAY_ALLOCATE */
+
+/* These values must meet several constraints. They must not be valid
+ register values; since we have a limit of 255 registers (because
+ we use only one byte in the pattern for the register number), we can
+ use numbers larger than 255. They must differ by 1, because of
+ NUM_FAILURE_ITEMS above. And the value for the lowest register must
+ be larger than the value for the highest register, so we do not try
+ to actually save any registers when none are active. */
+#define NO_HIGHEST_ACTIVE_REG (1 << BYTEWIDTH)
+#define NO_LOWEST_ACTIVE_REG (NO_HIGHEST_ACTIVE_REG + 1)
+
+/* Matching routines. */
+
+#ifndef emacs /* Emacs never uses this. */
+/* re_match is like re_match_2 except it takes only a single string. */
+
+int
+re_match (bufp, string, size, pos, regs)
+ struct re_pattern_buffer *bufp;
+ const char *string;
+ int size, pos;
+ struct re_registers *regs;
+{
+ int result = re_match_2_internal (bufp, NULL, 0, string, size,
+ pos, regs, size);
+ alloca (0);
+ return result;
+}
+#endif /* not emacs */
+
+
+/* re_match_2 matches the compiled pattern in BUFP against the
+ the (virtual) concatenation of STRING1 and STRING2 (of length SIZE1
+ and SIZE2, respectively). We start matching at POS, and stop
+ matching at STOP.
+
+ If REGS is non-null and the `no_sub' field of BUFP is nonzero, we
+ store offsets for the substring each group matched in REGS. See the
+ documentation for exactly how many groups we fill.
+
+ We return -1 if no match, -2 if an internal error (such as the
+ failure stack overflowing). Otherwise, we return the length of the
+ matched substring. */
+
+int
+re_match_2 (bufp, string1, size1, string2, size2, pos, regs, stop)
+ struct re_pattern_buffer *bufp;
+ const char *string1, *string2;
+ int size1, size2;
+ int pos;
+ struct re_registers *regs;
+ int stop;
+{
+ int result = re_match_2_internal (bufp, string1, size1, string2, size2,
+ pos, regs, stop);
+ alloca (0);
+ return result;
+}
+
+/* This is a separate function so that we can force an alloca cleanup
+ afterwards. */
+static int
+re_match_2_internal (bufp, string1, size1, string2, size2, pos, regs, stop)
+ struct re_pattern_buffer *bufp;
+ const char *string1, *string2;
+ int size1, size2;
+ int pos;
+ struct re_registers *regs;
+ int stop;
+{
+ /* General temporaries. */
+ int mcnt;
+ unsigned char *p1;
+
+ /* Just past the end of the corresponding string. */
+ const char *end1, *end2;
+
+ /* Pointers into string1 and string2, just past the last characters in
+ each to consider matching. */
+ const char *end_match_1, *end_match_2;
+
+ /* Where we are in the data, and the end of the current string. */
+ const char *d, *dend;
+
+ /* Where we are in the pattern, and the end of the pattern. */
+ unsigned char *p = bufp->buffer;
+ register unsigned char *pend = p + bufp->used;
+
+ /* Mark the opcode just after a start_memory, so we can test for an
+ empty subpattern when we get to the stop_memory. */
+ unsigned char *just_past_start_mem = 0;
+
+ /* We use this to map every character in the string. */
+ char *translate = bufp->translate;
+
+ /* Failure point stack. Each place that can handle a failure further
+ down the line pushes a failure point on this stack. It consists of
+ restart, regend, and reg_info for all registers corresponding to
+ the subexpressions we're currently inside, plus the number of such
+ registers, and, finally, two char *'s. The first char * is where
+ to resume scanning the pattern; the second one is where to resume
+ scanning the strings. If the latter is zero, the failure point is
+ a ``dummy''; if a failure happens and the failure point is a dummy,
+ it gets discarded and the next next one is tried. */
+#ifdef MATCH_MAY_ALLOCATE /* otherwise, this is global. */
+ fail_stack_type fail_stack;
+#endif
+#ifdef DEBUG
+ static unsigned failure_id = 0;
+ unsigned nfailure_points_pushed = 0, nfailure_points_popped = 0;
+#endif
+
+ /* This holds the pointer to the failure stack, when
+ it is allocated relocatably. */
+ fail_stack_elt_t *failure_stack_ptr;
+
+ /* We fill all the registers internally, independent of what we
+ return, for use in backreferences. The number here includes
+ an element for register zero. */
+ unsigned num_regs = bufp->re_nsub + 1;
+
+ /* The currently active registers. */
+ unsigned lowest_active_reg = NO_LOWEST_ACTIVE_REG;
+ unsigned highest_active_reg = NO_HIGHEST_ACTIVE_REG;
+
+ /* Information on the contents of registers. These are pointers into
+ the input strings; they record just what was matched (on this
+ attempt) by a subexpression part of the pattern, that is, the
+ regnum-th regstart pointer points to where in the pattern we began
+ matching and the regnum-th regend points to right after where we
+ stopped matching the regnum-th subexpression. (The zeroth register
+ keeps track of what the whole pattern matches.) */
+#ifdef MATCH_MAY_ALLOCATE /* otherwise, these are global. */
+ const char **regstart, **regend;
+#endif
+
+ /* If a group that's operated upon by a repetition operator fails to
+ match anything, then the register for its start will need to be
+ restored because it will have been set to wherever in the string we
+ are when we last see its open-group operator. Similarly for a
+ register's end. */
+#ifdef MATCH_MAY_ALLOCATE /* otherwise, these are global. */
+ const char **old_regstart, **old_regend;
+#endif
+
+ /* The is_active field of reg_info helps us keep track of which (possibly
+ nested) subexpressions we are currently in. The matched_something
+ field of reg_info[reg_num] helps us tell whether or not we have
+ matched any of the pattern so far this time through the reg_num-th
+ subexpression. These two fields get reset each time through any
+ loop their register is in. */
+#ifdef MATCH_MAY_ALLOCATE /* otherwise, this is global. */
+ register_info_type *reg_info;
+#endif
+
+ /* The following record the register info as found in the above
+ variables when we find a match better than any we've seen before.
+ This happens as we backtrack through the failure points, which in
+ turn happens only if we have not yet matched the entire string. */
+ unsigned best_regs_set = false;
+#ifdef MATCH_MAY_ALLOCATE /* otherwise, these are global. */
+ const char **best_regstart, **best_regend;
+#endif
+
+ /* Logically, this is `best_regend[0]'. But we don't want to have to
+ allocate space for that if we're not allocating space for anything
+ else (see below). Also, we never need info about register 0 for
+ any of the other register vectors, and it seems rather a kludge to
+ treat `best_regend' differently than the rest. So we keep track of
+ the end of the best match so far in a separate variable. We
+ initialize this to NULL so that when we backtrack the first time
+ and need to test it, it's not garbage. */
+ const char *match_end = NULL;
+
+ /* This helps SET_REGS_MATCHED avoid doing redundant work. */
+ int set_regs_matched_done = 0;
+
+ /* Used when we pop values we don't care about. */
+#ifdef MATCH_MAY_ALLOCATE /* otherwise, these are global. */
+ const char **reg_dummy;
+ register_info_type *reg_info_dummy;
+#endif
+
+#ifdef DEBUG
+ /* Counts the total number of registers pushed. */
+ unsigned num_regs_pushed = 0;
+#endif
+
+ DEBUG_PRINT1 ("\n\nEntering re_match_2.\n");
+
+ INIT_FAIL_STACK ();
+
+#ifdef MATCH_MAY_ALLOCATE
+ /* Do not bother to initialize all the register variables if there are
+ no groups in the pattern, as it takes a fair amount of time. If
+ there are groups, we include space for register 0 (the whole
+ pattern), even though we never use it, since it simplifies the
+ array indexing. We should fix this. */
+ if (bufp->re_nsub)
+ {
+ regstart = REGEX_TALLOC (num_regs, const char *);
+ regend = REGEX_TALLOC (num_regs, const char *);
+ old_regstart = REGEX_TALLOC (num_regs, const char *);
+ old_regend = REGEX_TALLOC (num_regs, const char *);
+ best_regstart = REGEX_TALLOC (num_regs, const char *);
+ best_regend = REGEX_TALLOC (num_regs, const char *);
+ reg_info = REGEX_TALLOC (num_regs, register_info_type);
+ reg_dummy = REGEX_TALLOC (num_regs, const char *);
+ reg_info_dummy = REGEX_TALLOC (num_regs, register_info_type);
+
+ if (!(regstart && regend && old_regstart && old_regend && reg_info
+ && best_regstart && best_regend && reg_dummy && reg_info_dummy))
+ {
+ FREE_VARIABLES ();
+ return -2;
+ }
+ }
+ else
+ {
+ /* We must initialize all our variables to NULL, so that
+ `FREE_VARIABLES' doesn't try to free them. */
+ regstart = regend = old_regstart = old_regend = best_regstart
+ = best_regend = reg_dummy = NULL;
+ reg_info = reg_info_dummy = (register_info_type *) NULL;
+ }
+#endif /* MATCH_MAY_ALLOCATE */
+
+ /* The starting position is bogus. */
+ if (pos < 0 || pos > size1 + size2)
+ {
+ FREE_VARIABLES ();
+ return -1;
+ }
+
+ /* Initialize subexpression text positions to -1 to mark ones that no
+ start_memory/stop_memory has been seen for. Also initialize the
+ register information struct. */
+ for (mcnt = 1; mcnt < num_regs; mcnt++)
+ {
+ regstart[mcnt] = regend[mcnt]
+ = old_regstart[mcnt] = old_regend[mcnt] = REG_UNSET_VALUE;
+
+ REG_MATCH_NULL_STRING_P (reg_info[mcnt]) = MATCH_NULL_UNSET_VALUE;
+ IS_ACTIVE (reg_info[mcnt]) = 0;
+ MATCHED_SOMETHING (reg_info[mcnt]) = 0;
+ EVER_MATCHED_SOMETHING (reg_info[mcnt]) = 0;
+ }
+
+ /* We move `string1' into `string2' if the latter's empty -- but not if
+ `string1' is null. */
+ if (size2 == 0 && string1 != NULL)
+ {
+ string2 = string1;
+ size2 = size1;
+ string1 = 0;
+ size1 = 0;
+ }
+ end1 = string1 + size1;
+ end2 = string2 + size2;
+
+ /* Compute where to stop matching, within the two strings. */
+ if (stop <= size1)
+ {
+ end_match_1 = string1 + stop;
+ end_match_2 = string2;
+ }
+ else
+ {
+ end_match_1 = end1;
+ end_match_2 = string2 + stop - size1;
+ }
+
+ /* `p' scans through the pattern as `d' scans through the data.
+ `dend' is the end of the input string that `d' points within. `d'
+ is advanced into the following input string whenever necessary, but
+ this happens before fetching; therefore, at the beginning of the
+ loop, `d' can be pointing at the end of a string, but it cannot
+ equal `string2'. */
+ if (size1 > 0 && pos <= size1)
+ {
+ d = string1 + pos;
+ dend = end_match_1;
+ }
+ else
+ {
+ d = string2 + pos - size1;
+ dend = end_match_2;
+ }
+
+ DEBUG_PRINT1 ("The compiled pattern is: ");
+ DEBUG_PRINT_COMPILED_PATTERN (bufp, p, pend);
+ DEBUG_PRINT1 ("The string to match is: `");
+ DEBUG_PRINT_DOUBLE_STRING (d, string1, size1, string2, size2);
+ DEBUG_PRINT1 ("'\n");
+
+ /* This loops over pattern commands. It exits by returning from the
+ function if the match is complete, or it drops through if the match
+ fails at this starting point in the input data. */
+ for (;;)
+ {
+ DEBUG_PRINT2 ("\n0x%x: ", p);
+
+ if (p == pend)
+ { /* End of pattern means we might have succeeded. */
+ DEBUG_PRINT1 ("end of pattern ... ");
+
+ /* If we haven't matched the entire string, and we want the
+ longest match, try backtracking. */
+ if (d != end_match_2)
+ {
+ /* 1 if this match ends in the same string (string1 or string2)
+ as the best previous match. */
+ boolean same_str_p = (FIRST_STRING_P (match_end)
+ == MATCHING_IN_FIRST_STRING);
+ /* 1 if this match is the best seen so far. */
+ boolean best_match_p;
+
+ /* AIX compiler got confused when this was combined
+ with the previous declaration. */
+ if (same_str_p)
+ best_match_p = d > match_end;
+ else
+ best_match_p = !MATCHING_IN_FIRST_STRING;
+
+ DEBUG_PRINT1 ("backtracking.\n");
+
+ if (!FAIL_STACK_EMPTY ())
+ { /* More failure points to try. */
+
+ /* If exceeds best match so far, save it. */
+ if (!best_regs_set || best_match_p)
+ {
+ best_regs_set = true;
+ match_end = d;
+
+ DEBUG_PRINT1 ("\nSAVING match as best so far.\n");
+
+ for (mcnt = 1; mcnt < num_regs; mcnt++)
+ {
+ best_regstart[mcnt] = regstart[mcnt];
+ best_regend[mcnt] = regend[mcnt];
+ }
+ }
+ goto fail;
+ }
+
+ /* If no failure points, don't restore garbage. And if
+ last match is real best match, don't restore second
+ best one. */
+ else if (best_regs_set && !best_match_p)
+ {
+ restore_best_regs:
+ /* Restore best match. It may happen that `dend ==
+ end_match_1' while the restored d is in string2.
+ For example, the pattern `x.*y.*z' against the
+ strings `x-' and `y-z-', if the two strings are
+ not consecutive in memory. */
+ DEBUG_PRINT1 ("Restoring best registers.\n");
+
+ d = match_end;
+ dend = ((d >= string1 && d <= end1)
+ ? end_match_1 : end_match_2);
+
+ for (mcnt = 1; mcnt < num_regs; mcnt++)
+ {
+ regstart[mcnt] = best_regstart[mcnt];
+ regend[mcnt] = best_regend[mcnt];
+ }
+ }
+ } /* d != end_match_2 */
+
+ succeed_label:
+ DEBUG_PRINT1 ("Accepting match.\n");
+
+ /* If caller wants register contents data back, do it. */
+ if (regs && !bufp->no_sub)
+ {
+ /* Have the register data arrays been allocated? */
+ if (bufp->regs_allocated == REGS_UNALLOCATED)
+ { /* No. So allocate them with malloc. We need one
+ extra element beyond `num_regs' for the `-1' marker
+ GNU code uses. */
+ regs->num_regs = MAX (RE_NREGS, num_regs + 1);
+ regs->start = TALLOC (regs->num_regs, regoff_t);
+ regs->end = TALLOC (regs->num_regs, regoff_t);
+ if (regs->start == NULL || regs->end == NULL)
+ {
+ FREE_VARIABLES ();
+ return -2;
+ }
+ bufp->regs_allocated = REGS_REALLOCATE;
+ }
+ else if (bufp->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 (regs->num_regs < num_regs + 1)
+ {
+ regs->num_regs = num_regs + 1;
+ RETALLOC (regs->start, regs->num_regs, regoff_t);
+ RETALLOC (regs->end, regs->num_regs, regoff_t);
+ if (regs->start == NULL || regs->end == NULL)
+ {
+ FREE_VARIABLES ();
+ return -2;
+ }
+ }
+ }
+ else
+ {
+ /* These braces fend off a "empty body in an else-statement"
+ warning under GCC when assert expands to nothing. */
+ assert (bufp->regs_allocated == REGS_FIXED);
+ }
+
+ /* Convert the pointer data in `regstart' and `regend' to
+ indices. Register zero has to be set differently,
+ since we haven't kept track of any info for it. */
+ if (regs->num_regs > 0)
+ {
+ regs->start[0] = pos;
+ regs->end[0] = (MATCHING_IN_FIRST_STRING
+ ? ((regoff_t) (d - string1))
+ : ((regoff_t) (d - string2 + size1)));
+ }
+
+ /* Go through the first `min (num_regs, regs->num_regs)'
+ registers, since that is all we initialized. */
+ for (mcnt = 1; mcnt < MIN (num_regs, regs->num_regs); mcnt++)
+ {
+ if (REG_UNSET (regstart[mcnt]) || REG_UNSET (regend[mcnt]))
+ regs->start[mcnt] = regs->end[mcnt] = -1;
+ else
+ {
+ regs->start[mcnt]
+ = (regoff_t) POINTER_TO_OFFSET (regstart[mcnt]);
+ regs->end[mcnt]
+ = (regoff_t) POINTER_TO_OFFSET (regend[mcnt]);
+ }
+ }
+
+ /* If the regs structure we return has more elements than
+ were in the pattern, set the extra elements to -1. If
+ we (re)allocated the registers, this is the case,
+ because we always allocate enough to have at least one
+ -1 at the end. */
+ for (mcnt = num_regs; mcnt < regs->num_regs; mcnt++)
+ regs->start[mcnt] = regs->end[mcnt] = -1;
+ } /* regs && !bufp->no_sub */
+
+ DEBUG_PRINT4 ("%u failure points pushed, %u popped (%u remain).\n",
+ nfailure_points_pushed, nfailure_points_popped,
+ nfailure_points_pushed - nfailure_points_popped);
+ DEBUG_PRINT2 ("%u registers pushed.\n", num_regs_pushed);
+
+ mcnt = d - pos - (MATCHING_IN_FIRST_STRING
+ ? string1
+ : string2 - size1);
+
+ DEBUG_PRINT2 ("Returning %d from re_match_2.\n", mcnt);
+
+ FREE_VARIABLES ();
+ return mcnt;
+ }
+
+ /* Otherwise match next pattern command. */
+ switch (SWITCH_ENUM_CAST ((re_opcode_t) *p++))
+ {
+ /* Ignore these. Used to ignore the n of succeed_n's which
+ currently have n == 0. */
+ case no_op:
+ DEBUG_PRINT1 ("EXECUTING no_op.\n");
+ break;
+
+ case succeed:
+ DEBUG_PRINT1 ("EXECUTING succeed.\n");
+ goto succeed_label;
+
+ /* Match the next n pattern characters exactly. The following
+ byte in the pattern defines n, and the n bytes after that
+ are the characters to match. */
+ case exactn:
+ mcnt = *p++;
+ DEBUG_PRINT2 ("EXECUTING exactn %d.\n", mcnt);
+
+ /* This is written out as an if-else so we don't waste time
+ testing `translate' inside the loop. */
+ if (translate)
+ {
+ do
+ {
+ PREFETCH ();
+ if (translate[(unsigned char) *d++] != (char) *p++)
+ goto fail;
+ }
+ while (--mcnt);
+ }
+ else
+ {
+ do
+ {
+ PREFETCH ();
+ if (*d++ != (char) *p++) goto fail;
+ }
+ while (--mcnt);
+ }
+ SET_REGS_MATCHED ();
+ break;
+
+
+ /* Match any character except possibly a newline or a null. */
+ case anychar:
+ DEBUG_PRINT1 ("EXECUTING anychar.\n");
+
+ PREFETCH ();
+
+ if ((!(bufp->syntax & RE_DOT_NEWLINE) && TRANSLATE (*d) == '\n')
+ || (bufp->syntax & RE_DOT_NOT_NULL && TRANSLATE (*d) == '\000'))
+ goto fail;
+
+ SET_REGS_MATCHED ();
+ DEBUG_PRINT2 (" Matched `%d'.\n", *d);
+ d++;
+ break;
+
+
+ case charset:
+ case charset_not:
+ {
+ register unsigned char c;
+ boolean not = (re_opcode_t) *(p - 1) == charset_not;
+
+ DEBUG_PRINT2 ("EXECUTING charset%s.\n", not ? "_not" : "");
+
+ PREFETCH ();
+ c = TRANSLATE (*d); /* The character to match. */
+
+ /* Cast to `unsigned' instead of `unsigned char' in case the
+ bit list is a full 32 bytes long. */
+ if (c < (unsigned) (*p * BYTEWIDTH)
+ && p[1 + c / BYTEWIDTH] & (1 << (c % BYTEWIDTH)))
+ not = !not;
+
+ p += 1 + *p;
+
+ if (!not) goto fail;
+
+ SET_REGS_MATCHED ();
+ d++;
+ break;
+ }
+
+
+ /* The beginning of a group is represented by start_memory.
+ The arguments are the register number in the next byte, and the
+ number of groups inner to this one in the next. The text
+ matched within the group is recorded (in the internal
+ registers data structure) under the register number. */
+ case start_memory:
+ DEBUG_PRINT3 ("EXECUTING start_memory %d (%d):\n", *p, p[1]);
+
+ /* Find out if this group can match the empty string. */
+ p1 = p; /* To send to group_match_null_string_p. */
+
+ if (REG_MATCH_NULL_STRING_P (reg_info[*p]) == MATCH_NULL_UNSET_VALUE)
+ REG_MATCH_NULL_STRING_P (reg_info[*p])
+ = group_match_null_string_p (&p1, pend, reg_info);
+
+ /* Save the position in the string where we were the last time
+ we were at this open-group operator in case the group is
+ operated upon by a repetition operator, e.g., with `(a*)*b'
+ against `ab'; then we want to ignore where we are now in
+ the string in case this attempt to match fails. */
+ old_regstart[*p] = REG_MATCH_NULL_STRING_P (reg_info[*p])
+ ? REG_UNSET (regstart[*p]) ? d : regstart[*p]
+ : regstart[*p];
+ DEBUG_PRINT2 (" old_regstart: %d\n",
+ POINTER_TO_OFFSET (old_regstart[*p]));
+
+ regstart[*p] = d;
+ DEBUG_PRINT2 (" regstart: %d\n", POINTER_TO_OFFSET (regstart[*p]));
+
+ IS_ACTIVE (reg_info[*p]) = 1;
+ MATCHED_SOMETHING (reg_info[*p]) = 0;
+
+ /* Clear this whenever we change the register activity status. */
+ set_regs_matched_done = 0;
+
+ /* This is the new highest active register. */
+ highest_active_reg = *p;
+
+ /* If nothing was active before, this is the new lowest active
+ register. */
+ if (lowest_active_reg == NO_LOWEST_ACTIVE_REG)
+ lowest_active_reg = *p;
+
+ /* Move past the register number and inner group count. */
+ p += 2;
+ just_past_start_mem = p;
+
+ break;
+
+
+ /* The stop_memory opcode represents the end of a group. Its
+ arguments are the same as start_memory's: the register
+ number, and the number of inner groups. */
+ case stop_memory:
+ DEBUG_PRINT3 ("EXECUTING stop_memory %d (%d):\n", *p, p[1]);
+
+ /* We need to save the string position the last time we were at
+ this close-group operator in case the group is operated
+ upon by a repetition operator, e.g., with `((a*)*(b*)*)*'
+ against `aba'; then we want to ignore where we are now in
+ the string in case this attempt to match fails. */
+ old_regend[*p] = REG_MATCH_NULL_STRING_P (reg_info[*p])
+ ? REG_UNSET (regend[*p]) ? d : regend[*p]
+ : regend[*p];
+ DEBUG_PRINT2 (" old_regend: %d\n",
+ POINTER_TO_OFFSET (old_regend[*p]));
+
+ regend[*p] = d;
+ DEBUG_PRINT2 (" regend: %d\n", POINTER_TO_OFFSET (regend[*p]));
+
+ /* This register isn't active anymore. */
+ IS_ACTIVE (reg_info[*p]) = 0;
+
+ /* Clear this whenever we change the register activity status. */
+ set_regs_matched_done = 0;
+
+ /* If this was the only register active, nothing is active
+ anymore. */
+ if (lowest_active_reg == highest_active_reg)
+ {
+ lowest_active_reg = NO_LOWEST_ACTIVE_REG;
+ highest_active_reg = NO_HIGHEST_ACTIVE_REG;
+ }
+ else
+ { /* We must scan for the new highest active register, since
+ it isn't necessarily one less than now: consider
+ (a(b)c(d(e)f)g). When group 3 ends, after the f), the
+ new highest active register is 1. */
+ unsigned char r = *p - 1;
+ while (r > 0 && !IS_ACTIVE (reg_info[r]))
+ r--;
+
+ /* If we end up at register zero, that means that we saved
+ the registers as the result of an `on_failure_jump', not
+ a `start_memory', and we jumped to past the innermost
+ `stop_memory'. For example, in ((.)*) we save
+ registers 1 and 2 as a result of the *, but when we pop
+ back to the second ), we are at the stop_memory 1.
+ Thus, nothing is active. */
+ if (r == 0)
+ {
+ lowest_active_reg = NO_LOWEST_ACTIVE_REG;
+ highest_active_reg = NO_HIGHEST_ACTIVE_REG;
+ }
+ else
+ highest_active_reg = r;
+ }
+
+ /* If just failed to match something this time around with a
+ group that's operated on by a repetition operator, try to
+ force exit from the ``loop'', and restore the register
+ information for this group that we had before trying this
+ last match. */
+ if ((!MATCHED_SOMETHING (reg_info[*p])
+ || just_past_start_mem == p - 1)
+ && (p + 2) < pend)
+ {
+ boolean is_a_jump_n = false;
+
+ p1 = p + 2;
+ mcnt = 0;
+ switch ((re_opcode_t) *p1++)
+ {
+ case jump_n:
+ is_a_jump_n = true;
+ case pop_failure_jump:
+ case maybe_pop_jump:
+ case jump:
+ case dummy_failure_jump:
+ EXTRACT_NUMBER_AND_INCR (mcnt, p1);
+ if (is_a_jump_n)
+ p1 += 2;
+ break;
+
+ default:
+ /* do nothing */ ;
+ }
+ p1 += mcnt;
+
+ /* If the next operation is a jump backwards in the pattern
+ to an on_failure_jump right before the start_memory
+ corresponding to this stop_memory, exit from the loop
+ by forcing a failure after pushing on the stack the
+ on_failure_jump's jump in the pattern, and d. */
+ if (mcnt < 0 && (re_opcode_t) *p1 == on_failure_jump
+ && (re_opcode_t) p1[3] == start_memory && p1[4] == *p)
+ {
+ /* If this group ever matched anything, then restore
+ what its registers were before trying this last
+ failed match, e.g., with `(a*)*b' against `ab' for
+ regstart[1], and, e.g., with `((a*)*(b*)*)*'
+ against `aba' for regend[3].
+
+ Also restore the registers for inner groups for,
+ e.g., `((a*)(b*))*' against `aba' (register 3 would
+ otherwise get trashed). */
+
+ if (EVER_MATCHED_SOMETHING (reg_info[*p]))
+ {
+ unsigned r;
+
+ EVER_MATCHED_SOMETHING (reg_info[*p]) = 0;
+
+ /* Restore this and inner groups' (if any) registers. */
+ for (r = *p; r < *p + *(p + 1); r++)
+ {
+ regstart[r] = old_regstart[r];
+
+ /* xx why this test? */
+ if (old_regend[r] >= regstart[r])
+ regend[r] = old_regend[r];
+ }
+ }
+ p1++;
+ EXTRACT_NUMBER_AND_INCR (mcnt, p1);
+ PUSH_FAILURE_POINT (p1 + mcnt, d, -2);
+
+ goto fail;
+ }
+ }
+
+ /* Move past the register number and the inner group count. */
+ p += 2;
+ break;
+
+
+ /* \<digit> has been turned into a `duplicate' command which is
+ followed by the numeric value of <digit> as the register number. */
+ case duplicate:
+ {
+ register const char *d2, *dend2;
+ int regno = *p++; /* Get which register to match against. */
+ DEBUG_PRINT2 ("EXECUTING duplicate %d.\n", regno);
+
+ /* Can't back reference a group which we've never matched. */
+ if (REG_UNSET (regstart[regno]) || REG_UNSET (regend[regno]))
+ goto fail;
+
+ /* Where in input to try to start matching. */
+ d2 = regstart[regno];
+
+ /* Where to stop matching; if both the place to start and
+ the place to stop matching are in the same string, then
+ set to the place to stop, otherwise, for now have to use
+ the end of the first string. */
+
+ dend2 = ((FIRST_STRING_P (regstart[regno])
+ == FIRST_STRING_P (regend[regno]))
+ ? regend[regno] : end_match_1);
+ for (;;)
+ {
+ /* If necessary, advance to next segment in register
+ contents. */
+ while (d2 == dend2)
+ {
+ if (dend2 == end_match_2) break;
+ if (dend2 == regend[regno]) break;
+
+ /* End of string1 => advance to string2. */
+ d2 = string2;
+ dend2 = regend[regno];
+ }
+ /* At end of register contents => success */
+ if (d2 == dend2) break;
+
+ /* If necessary, advance to next segment in data. */
+ PREFETCH ();
+
+ /* How many characters left in this segment to match. */
+ mcnt = dend - d;
+
+ /* Want how many consecutive characters we can match in
+ one shot, so, if necessary, adjust the count. */
+ if (mcnt > dend2 - d2)
+ mcnt = dend2 - d2;
+
+ /* Compare that many; failure if mismatch, else move
+ past them. */
+ if (translate
+ ? bcmp_translate (d, d2, mcnt, translate)
+ : bcmp (d, d2, mcnt))
+ goto fail;
+ d += mcnt, d2 += mcnt;
+
+ /* Do this because we've match some characters. */
+ SET_REGS_MATCHED ();
+ }
+ }
+ break;
+
+
+ /* begline matches the empty string at the beginning of the string
+ (unless `not_bol' is set in `bufp'), and, if
+ `newline_anchor' is set, after newlines. */
+ case begline:
+ DEBUG_PRINT1 ("EXECUTING begline.\n");
+
+ if (AT_STRINGS_BEG (d))
+ {
+ if (!bufp->not_bol) break;
+ }
+ else if (d[-1] == '\n' && bufp->newline_anchor)
+ {
+ break;
+ }
+ /* In all other cases, we fail. */
+ goto fail;
+
+
+ /* endline is the dual of begline. */
+ case endline:
+ DEBUG_PRINT1 ("EXECUTING endline.\n");
+
+ if (AT_STRINGS_END (d))
+ {
+ if (!bufp->not_eol) break;
+ }
+
+ /* We have to ``prefetch'' the next character. */
+ else if ((d == end1 ? *string2 : *d) == '\n'
+ && bufp->newline_anchor)
+ {
+ break;
+ }
+ goto fail;
+
+
+ /* Match at the very beginning of the data. */
+ case begbuf:
+ DEBUG_PRINT1 ("EXECUTING begbuf.\n");
+ if (AT_STRINGS_BEG (d))
+ break;
+ goto fail;
+
+
+ /* Match at the very end of the data. */
+ case endbuf:
+ DEBUG_PRINT1 ("EXECUTING endbuf.\n");
+ if (AT_STRINGS_END (d))
+ break;
+ goto fail;
+
+
+ /* on_failure_keep_string_jump is used to optimize `.*\n'. It
+ pushes NULL as the value for the string on the stack. Then
+ `pop_failure_point' will keep the current value for the
+ string, instead of restoring it. To see why, consider
+ matching `foo\nbar' against `.*\n'. The .* matches the foo;
+ then the . fails against the \n. But the next thing we want
+ to do is match the \n against the \n; if we restored the
+ string value, we would be back at the foo.
+
+ Because this is used only in specific cases, we don't need to
+ check all the things that `on_failure_jump' does, to make
+ sure the right things get saved on the stack. Hence we don't
+ share its code. The only reason to push anything on the
+ stack at all is that otherwise we would have to change
+ `anychar's code to do something besides goto fail in this
+ case; that seems worse than this. */
+ case on_failure_keep_string_jump:
+ DEBUG_PRINT1 ("EXECUTING on_failure_keep_string_jump");
+
+ EXTRACT_NUMBER_AND_INCR (mcnt, p);
+ DEBUG_PRINT3 (" %d (to 0x%x):\n", mcnt, p + mcnt);
+
+ PUSH_FAILURE_POINT (p + mcnt, NULL, -2);
+ break;
+
+
+ /* Uses of on_failure_jump:
+
+ Each alternative starts with an on_failure_jump that points
+ to the beginning of the next alternative. Each alternative
+ except the last ends with a jump that in effect jumps past
+ the rest of the alternatives. (They really jump to the
+ ending jump of the following alternative, because tensioning
+ these jumps is a hassle.)
+
+ Repeats start with an on_failure_jump that points past both
+ the repetition text and either the following jump or
+ pop_failure_jump back to this on_failure_jump. */
+ case on_failure_jump:
+ on_failure:
+ DEBUG_PRINT1 ("EXECUTING on_failure_jump");
+
+ EXTRACT_NUMBER_AND_INCR (mcnt, p);
+ DEBUG_PRINT3 (" %d (to 0x%x)", mcnt, p + mcnt);
+
+ /* If this on_failure_jump comes right before a group (i.e.,
+ the original * applied to a group), save the information
+ for that group and all inner ones, so that if we fail back
+ to this point, the group's information will be correct.
+ For example, in \(a*\)*\1, we need the preceding group,
+ and in \(\(a*\)b*\)\2, we need the inner group. */
+
+ /* We can't use `p' to check ahead because we push
+ a failure point to `p + mcnt' after we do this. */
+ p1 = p;
+
+ /* We need to skip no_op's before we look for the
+ start_memory in case this on_failure_jump is happening as
+ the result of a completed succeed_n, as in \(a\)\{1,3\}b\1
+ against aba. */
+ while (p1 < pend && (re_opcode_t) *p1 == no_op)
+ p1++;
+
+ if (p1 < pend && (re_opcode_t) *p1 == start_memory)
+ {
+ /* We have a new highest active register now. This will
+ get reset at the start_memory we are about to get to,
+ but we will have saved all the registers relevant to
+ this repetition op, as described above. */
+ highest_active_reg = *(p1 + 1) + *(p1 + 2);
+ if (lowest_active_reg == NO_LOWEST_ACTIVE_REG)
+ lowest_active_reg = *(p1 + 1);
+ }
+
+ DEBUG_PRINT1 (":\n");
+ PUSH_FAILURE_POINT (p + mcnt, d, -2);
+ break;
+
+
+ /* A smart repeat ends with `maybe_pop_jump'.
+ We change it to either `pop_failure_jump' or `jump'. */
+ case maybe_pop_jump:
+ EXTRACT_NUMBER_AND_INCR (mcnt, p);
+ DEBUG_PRINT2 ("EXECUTING maybe_pop_jump %d.\n", mcnt);
+ {
+ register unsigned char *p2 = p;
+
+ /* Compare the beginning of the repeat with what in the
+ pattern follows its end. If we can establish that there
+ is nothing that they would both match, i.e., that we
+ would have to backtrack because of (as in, e.g., `a*a')
+ then we can change to pop_failure_jump, because we'll
+ never have to backtrack.
+
+ This is not true in the case of alternatives: in
+ `(a|ab)*' we do need to backtrack to the `ab' alternative
+ (e.g., if the string was `ab'). But instead of trying to
+ detect that here, the alternative has put on a dummy
+ failure point which is what we will end up popping. */
+
+ /* Skip over open/close-group commands.
+ If what follows this loop is a ...+ construct,
+ look at what begins its body, since we will have to
+ match at least one of that. */
+ while (1)
+ {
+ if (p2 + 2 < pend
+ && ((re_opcode_t) *p2 == stop_memory
+ || (re_opcode_t) *p2 == start_memory))
+ p2 += 3;
+ else if (p2 + 6 < pend
+ && (re_opcode_t) *p2 == dummy_failure_jump)
+ p2 += 6;
+ else
+ break;
+ }
+
+ p1 = p + mcnt;
+ /* p1[0] ... p1[2] are the `on_failure_jump' corresponding
+ to the `maybe_finalize_jump' of this case. Examine what
+ follows. */
+
+ /* If we're at the end of the pattern, we can change. */
+ if (p2 == pend)
+ {
+ /* Consider what happens when matching ":\(.*\)"
+ against ":/". I don't really understand this code
+ yet. */
+ p[-3] = (unsigned char) pop_failure_jump;
+ DEBUG_PRINT1
+ (" End of pattern: change to `pop_failure_jump'.\n");
+ }
+
+ else if ((re_opcode_t) *p2 == exactn
+ || (bufp->newline_anchor && (re_opcode_t) *p2 == endline))
+ {
+ register unsigned char c
+ = *p2 == (unsigned char) endline ? '\n' : p2[2];
+
+ if ((re_opcode_t) p1[3] == exactn && p1[5] != c)
+ {
+ p[-3] = (unsigned char) pop_failure_jump;
+ DEBUG_PRINT3 (" %c != %c => pop_failure_jump.\n",
+ c, p1[5]);
+ }
+
+ else if ((re_opcode_t) p1[3] == charset
+ || (re_opcode_t) p1[3] == charset_not)
+ {
+ int not = (re_opcode_t) p1[3] == charset_not;
+
+ if (c < (unsigned char) (p1[4] * BYTEWIDTH)
+ && p1[5 + c / BYTEWIDTH] & (1 << (c % BYTEWIDTH)))
+ not = !not;
+
+ /* `not' is equal to 1 if c would match, which means
+ that we can't change to pop_failure_jump. */
+ if (!not)
+ {
+ p[-3] = (unsigned char) pop_failure_jump;
+ DEBUG_PRINT1 (" No match => pop_failure_jump.\n");
+ }
+ }
+ }
+ else if ((re_opcode_t) *p2 == charset)
+ {
+#ifdef DEBUG
+ register unsigned char c
+ = *p2 == (unsigned char) endline ? '\n' : p2[2];
+#endif
+
+ if ((re_opcode_t) p1[3] == exactn
+ && ! ((int) p2[1] * BYTEWIDTH > (int) p1[4]
+ && (p2[1 + p1[4] / BYTEWIDTH]
+ & (1 << (p1[4] % BYTEWIDTH)))))
+ {
+ p[-3] = (unsigned char) pop_failure_jump;
+ DEBUG_PRINT3 (" %c != %c => pop_failure_jump.\n",
+ c, p1[5]);
+ }
+
+ else if ((re_opcode_t) p1[3] == charset_not)
+ {
+ int idx;
+ /* We win if the charset_not inside the loop
+ lists every character listed in the charset after. */
+ for (idx = 0; idx < (int) p2[1]; idx++)
+ if (! (p2[2 + idx] == 0
+ || (idx < (int) p1[4]
+ && ((p2[2 + idx] & ~ p1[5 + idx]) == 0))))
+ break;
+
+ if (idx == p2[1])
+ {
+ p[-3] = (unsigned char) pop_failure_jump;
+ DEBUG_PRINT1 (" No match => pop_failure_jump.\n");
+ }
+ }
+ else if ((re_opcode_t) p1[3] == charset)
+ {
+ int idx;
+ /* We win if the charset inside the loop
+ has no overlap with the one after the loop. */
+ for (idx = 0;
+ idx < (int) p2[1] && idx < (int) p1[4];
+ idx++)
+ if ((p2[2 + idx] & p1[5 + idx]) != 0)
+ break;
+
+ if (idx == p2[1] || idx == p1[4])
+ {
+ p[-3] = (unsigned char) pop_failure_jump;
+ DEBUG_PRINT1 (" No match => pop_failure_jump.\n");
+ }
+ }
+ }
+ }
+ p -= 2; /* Point at relative address again. */
+ if ((re_opcode_t) p[-1] != pop_failure_jump)
+ {
+ p[-1] = (unsigned char) jump;
+ DEBUG_PRINT1 (" Match => jump.\n");
+ goto unconditional_jump;
+ }
+ /* Note fall through. */
+
+
+ /* The end of a simple repeat has a pop_failure_jump back to
+ its matching on_failure_jump, where the latter will push a
+ failure point. The pop_failure_jump takes off failure
+ points put on by this pop_failure_jump's matching
+ on_failure_jump; we got through the pattern to here from the
+ matching on_failure_jump, so didn't fail. */
+ case pop_failure_jump:
+ {
+ /* We need to pass separate storage for the lowest and
+ highest registers, even though we don't care about the
+ actual values. Otherwise, we will restore only one
+ register from the stack, since lowest will == highest in
+ `pop_failure_point'. */
+ unsigned dummy_low_reg, dummy_high_reg;
+ unsigned char *pdummy;
+ const char *sdummy;
+
+ DEBUG_PRINT1 ("EXECUTING pop_failure_jump.\n");
+ POP_FAILURE_POINT (sdummy, pdummy,
+ dummy_low_reg, dummy_high_reg,
+ reg_dummy, reg_dummy, reg_info_dummy);
+ }
+ /* Note fall through. */
+
+
+ /* Unconditionally jump (without popping any failure points). */
+ case jump:
+ unconditional_jump:
+ EXTRACT_NUMBER_AND_INCR (mcnt, p); /* Get the amount to jump. */
+ DEBUG_PRINT2 ("EXECUTING jump %d ", mcnt);
+ p += mcnt; /* Do the jump. */
+ DEBUG_PRINT2 ("(to 0x%x).\n", p);
+ break;
+
+
+ /* We need this opcode so we can detect where alternatives end
+ in `group_match_null_string_p' et al. */
+ case jump_past_alt:
+ DEBUG_PRINT1 ("EXECUTING jump_past_alt.\n");
+ goto unconditional_jump;
+
+
+ /* Normally, the on_failure_jump pushes a failure point, which
+ then gets popped at pop_failure_jump. We will end up at
+ pop_failure_jump, also, and with a pattern of, say, `a+', we
+ are skipping over the on_failure_jump, so we have to push
+ something meaningless for pop_failure_jump to pop. */
+ case dummy_failure_jump:
+ DEBUG_PRINT1 ("EXECUTING dummy_failure_jump.\n");
+ /* It doesn't matter what we push for the string here. What
+ the code at `fail' tests is the value for the pattern. */
+ PUSH_FAILURE_POINT (0, 0, -2);
+ goto unconditional_jump;
+
+
+ /* At the end of an alternative, we need to push a dummy failure
+ point in case we are followed by a `pop_failure_jump', because
+ we don't want the failure point for the alternative to be
+ popped. For example, matching `(a|ab)*' against `aab'
+ requires that we match the `ab' alternative. */
+ case push_dummy_failure:
+ DEBUG_PRINT1 ("EXECUTING push_dummy_failure.\n");
+ /* See comments just above at `dummy_failure_jump' about the
+ two zeroes. */
+ PUSH_FAILURE_POINT (0, 0, -2);
+ break;
+
+ /* Have to succeed matching what follows at least n times.
+ After that, handle like `on_failure_jump'. */
+ case succeed_n:
+ EXTRACT_NUMBER (mcnt, p + 2);
+ DEBUG_PRINT2 ("EXECUTING succeed_n %d.\n", mcnt);
+
+ assert (mcnt >= 0);
+ /* Originally, this is how many times we HAVE to succeed. */
+ if (mcnt > 0)
+ {
+ mcnt--;
+ p += 2;
+ STORE_NUMBER_AND_INCR (p, mcnt);
+ DEBUG_PRINT3 (" Setting 0x%x to %d.\n", p, mcnt);
+ }
+ else if (mcnt == 0)
+ {
+ DEBUG_PRINT2 (" Setting two bytes from 0x%x to no_op.\n", p+2);
+ p[2] = (unsigned char) no_op;
+ p[3] = (unsigned char) no_op;
+ goto on_failure;
+ }
+ break;
+
+ case jump_n:
+ EXTRACT_NUMBER (mcnt, p + 2);
+ DEBUG_PRINT2 ("EXECUTING jump_n %d.\n", mcnt);
+
+ /* Originally, this is how many times we CAN jump. */
+ if (mcnt)
+ {
+ mcnt--;
+ STORE_NUMBER (p + 2, mcnt);
+ goto unconditional_jump;
+ }
+ /* If don't have to jump any more, skip over the rest of command. */
+ else
+ p += 4;
+ break;
+
+ case set_number_at:
+ {
+ DEBUG_PRINT1 ("EXECUTING set_number_at.\n");
+
+ EXTRACT_NUMBER_AND_INCR (mcnt, p);
+ p1 = p + mcnt;
+ EXTRACT_NUMBER_AND_INCR (mcnt, p);
+ DEBUG_PRINT3 (" Setting 0x%x to %d.\n", p1, mcnt);
+ STORE_NUMBER (p1, mcnt);
+ break;
+ }
+
+ case wordbound:
+ DEBUG_PRINT1 ("EXECUTING wordbound.\n");
+ if (AT_WORD_BOUNDARY (d))
+ break;
+ goto fail;
+
+ case notwordbound:
+ DEBUG_PRINT1 ("EXECUTING notwordbound.\n");
+ if (AT_WORD_BOUNDARY (d))
+ goto fail;
+ break;
+
+ case wordbeg:
+ DEBUG_PRINT1 ("EXECUTING wordbeg.\n");
+ if (WORDCHAR_P (d) && (AT_STRINGS_BEG (d) || !WORDCHAR_P (d - 1)))
+ break;
+ goto fail;
+
+ case wordend:
+ DEBUG_PRINT1 ("EXECUTING wordend.\n");
+ if (!AT_STRINGS_BEG (d) && WORDCHAR_P (d - 1)
+ && (!WORDCHAR_P (d) || AT_STRINGS_END (d)))
+ break;
+ goto fail;
+
+#ifdef emacs
+ case before_dot:
+ DEBUG_PRINT1 ("EXECUTING before_dot.\n");
+ if (PTR_CHAR_POS ((unsigned char *) d) >= point)
+ goto fail;
+ break;
+
+ case at_dot:
+ DEBUG_PRINT1 ("EXECUTING at_dot.\n");
+ if (PTR_CHAR_POS ((unsigned char *) d) != point)
+ goto fail;
+ break;
+
+ case after_dot:
+ DEBUG_PRINT1 ("EXECUTING after_dot.\n");
+ if (PTR_CHAR_POS ((unsigned char *) d) <= point)
+ goto fail;
+ break;
+#if 0 /* not emacs19 */
+ case at_dot:
+ DEBUG_PRINT1 ("EXECUTING at_dot.\n");
+ if (PTR_CHAR_POS ((unsigned char *) d) + 1 != point)
+ goto fail;
+ break;
+#endif /* not emacs19 */
+
+ case syntaxspec:
+ DEBUG_PRINT2 ("EXECUTING syntaxspec %d.\n", mcnt);
+ mcnt = *p++;
+ goto matchsyntax;
+
+ case wordchar:
+ DEBUG_PRINT1 ("EXECUTING Emacs wordchar.\n");
+ mcnt = (int) Sword;
+ matchsyntax:
+ PREFETCH ();
+ /* Can't use *d++ here; SYNTAX may be an unsafe macro. */
+ d++;
+ if (SYNTAX (d[-1]) != (enum syntaxcode) mcnt)
+ goto fail;
+ SET_REGS_MATCHED ();
+ break;
+
+ case notsyntaxspec:
+ DEBUG_PRINT2 ("EXECUTING notsyntaxspec %d.\n", mcnt);
+ mcnt = *p++;
+ goto matchnotsyntax;
+
+ case notwordchar:
+ DEBUG_PRINT1 ("EXECUTING Emacs notwordchar.\n");
+ mcnt = (int) Sword;
+ matchnotsyntax:
+ PREFETCH ();
+ /* Can't use *d++ here; SYNTAX may be an unsafe macro. */
+ d++;
+ if (SYNTAX (d[-1]) == (enum syntaxcode) mcnt)
+ goto fail;
+ SET_REGS_MATCHED ();
+ break;
+
+#else /* not emacs */
+ case wordchar:
+ DEBUG_PRINT1 ("EXECUTING non-Emacs wordchar.\n");
+ PREFETCH ();
+ if (!WORDCHAR_P (d))
+ goto fail;
+ SET_REGS_MATCHED ();
+ d++;
+ break;
+
+ case notwordchar:
+ DEBUG_PRINT1 ("EXECUTING non-Emacs notwordchar.\n");
+ PREFETCH ();
+ if (WORDCHAR_P (d))
+ goto fail;
+ SET_REGS_MATCHED ();
+ d++;
+ break;
+#endif /* not emacs */
+
+ default:
+ abort ();
+ }
+ continue; /* Successfully executed one pattern command; keep going. */
+
+
+ /* We goto here if a matching operation fails. */
+ fail:
+ if (!FAIL_STACK_EMPTY ())
+ { /* A restart point is known. Restore to that state. */
+ DEBUG_PRINT1 ("\nFAIL:\n");
+ POP_FAILURE_POINT (d, p,
+ lowest_active_reg, highest_active_reg,
+ regstart, regend, reg_info);
+
+ /* If this failure point is a dummy, try the next one. */
+ if (!p)
+ goto fail;
+
+ /* If we failed to the end of the pattern, don't examine *p. */
+ assert (p <= pend);
+ if (p < pend)
+ {
+ boolean is_a_jump_n = false;
+
+ /* If failed to a backwards jump that's part of a repetition
+ loop, need to pop this failure point and use the next one. */
+ switch ((re_opcode_t) *p)
+ {
+ case jump_n:
+ is_a_jump_n = true;
+ case maybe_pop_jump:
+ case pop_failure_jump:
+ case jump:
+ p1 = p + 1;
+ EXTRACT_NUMBER_AND_INCR (mcnt, p1);
+ p1 += mcnt;
+
+ if ((is_a_jump_n && (re_opcode_t) *p1 == succeed_n)
+ || (!is_a_jump_n
+ && (re_opcode_t) *p1 == on_failure_jump))
+ goto fail;
+ break;
+ default:
+ /* do nothing */ ;
+ }
+ }
+
+ if (d >= string1 && d <= end1)
+ dend = end_match_1;
+ }
+ else
+ break; /* Matching at this starting point really fails. */
+ } /* for (;;) */
+
+ if (best_regs_set)
+ goto restore_best_regs;
+
+ FREE_VARIABLES ();
+
+ return -1; /* Failure to match. */
+} /* re_match_2 */
+
+/* Subroutine definitions for re_match_2. */
+
+
+/* We are passed P pointing to a register number after a start_memory.
+
+ Return true if the pattern up to the corresponding stop_memory can
+ match the empty string, and false otherwise.
+
+ If we find the matching stop_memory, sets P to point to one past its number.
+ Otherwise, sets P to an undefined byte less than or equal to END.
+
+ We don't handle duplicates properly (yet). */
+
+static boolean
+group_match_null_string_p (p, end, reg_info)
+ unsigned char **p, *end;
+ register_info_type *reg_info;
+{
+ int mcnt;
+ /* Point to after the args to the start_memory. */
+ unsigned char *p1 = *p + 2;
+
+ while (p1 < end)
+ {
+ /* Skip over opcodes that can match nothing, and return true or
+ false, as appropriate, when we get to one that can't, or to the
+ matching stop_memory. */
+
+ switch ((re_opcode_t) *p1)
+ {
+ /* Could be either a loop or a series of alternatives. */
+ case on_failure_jump:
+ p1++;
+ EXTRACT_NUMBER_AND_INCR (mcnt, p1);
+
+ /* If the next operation is not a jump backwards in the
+ pattern. */
+
+ if (mcnt >= 0)
+ {
+ /* Go through the on_failure_jumps of the alternatives,
+ seeing if any of the alternatives cannot match nothing.
+ The last alternative starts with only a jump,
+ whereas the rest start with on_failure_jump and end
+ with a jump, e.g., here is the pattern for `a|b|c':
+
+ /on_failure_jump/0/6/exactn/1/a/jump_past_alt/0/6
+ /on_failure_jump/0/6/exactn/1/b/jump_past_alt/0/3
+ /exactn/1/c
+
+ So, we have to first go through the first (n-1)
+ alternatives and then deal with the last one separately. */
+
+
+ /* Deal with the first (n-1) alternatives, which start
+ with an on_failure_jump (see above) that jumps to right
+ past a jump_past_alt. */
+
+ while ((re_opcode_t) p1[mcnt-3] == jump_past_alt)
+ {
+ /* `mcnt' holds how many bytes long the alternative
+ is, including the ending `jump_past_alt' and
+ its number. */
+
+ if (!alt_match_null_string_p (p1, p1 + mcnt - 3,
+ reg_info))
+ return false;
+
+ /* Move to right after this alternative, including the
+ jump_past_alt. */
+ p1 += mcnt;
+
+ /* Break if it's the beginning of an n-th alternative
+ that doesn't begin with an on_failure_jump. */
+ if ((re_opcode_t) *p1 != on_failure_jump)
+ break;
+
+ /* Still have to check that it's not an n-th
+ alternative that starts with an on_failure_jump. */
+ p1++;
+ EXTRACT_NUMBER_AND_INCR (mcnt, p1);
+ if ((re_opcode_t) p1[mcnt-3] != jump_past_alt)
+ {
+ /* Get to the beginning of the n-th alternative. */
+ p1 -= 3;
+ break;
+ }
+ }
+
+ /* Deal with the last alternative: go back and get number
+ of the `jump_past_alt' just before it. `mcnt' contains
+ the length of the alternative. */
+ EXTRACT_NUMBER (mcnt, p1 - 2);
+
+ if (!alt_match_null_string_p (p1, p1 + mcnt, reg_info))
+ return false;
+
+ p1 += mcnt; /* Get past the n-th alternative. */
+ } /* if mcnt > 0 */
+ break;
+
+
+ case stop_memory:
+ assert (p1[1] == **p);
+ *p = p1 + 2;
+ return true;
+
+
+ default:
+ if (!common_op_match_null_string_p (&p1, end, reg_info))
+ return false;
+ }
+ } /* while p1 < end */
+
+ return false;
+} /* group_match_null_string_p */
+
+
+/* Similar to group_match_null_string_p, but doesn't deal with alternatives:
+ It expects P to be the first byte of a single alternative and END one
+ byte past the last. The alternative can contain groups. */
+
+static boolean
+alt_match_null_string_p (p, end, reg_info)
+ unsigned char *p, *end;
+ register_info_type *reg_info;
+{
+ int mcnt;
+ unsigned char *p1 = p;
+
+ while (p1 < end)
+ {
+ /* Skip over opcodes that can match nothing, and break when we get
+ to one that can't. */
+
+ switch ((re_opcode_t) *p1)
+ {
+ /* It's a loop. */
+ case on_failure_jump:
+ p1++;
+ EXTRACT_NUMBER_AND_INCR (mcnt, p1);
+ p1 += mcnt;
+ break;
+
+ default:
+ if (!common_op_match_null_string_p (&p1, end, reg_info))
+ return false;
+ }
+ } /* while p1 < end */
+
+ return true;
+} /* alt_match_null_string_p */
+
+
+/* Deals with the ops common to group_match_null_string_p and
+ alt_match_null_string_p.
+
+ Sets P to one after the op and its arguments, if any. */
+
+static boolean
+common_op_match_null_string_p (p, end, reg_info)
+ unsigned char **p, *end;
+ register_info_type *reg_info;
+{
+ int mcnt;
+ boolean ret;
+ int reg_no;
+ unsigned char *p1 = *p;
+
+ switch ((re_opcode_t) *p1++)
+ {
+ case no_op:
+ case begline:
+ case endline:
+ case begbuf:
+ case endbuf:
+ case wordbeg:
+ case wordend:
+ case wordbound:
+ case notwordbound:
+#ifdef emacs
+ case before_dot:
+ case at_dot:
+ case after_dot:
+#endif
+ break;
+
+ case start_memory:
+ reg_no = *p1;
+ assert (reg_no > 0 && reg_no <= MAX_REGNUM);
+ ret = group_match_null_string_p (&p1, end, reg_info);
+
+ /* Have to set this here in case we're checking a group which
+ contains a group and a back reference to it. */
+
+ if (REG_MATCH_NULL_STRING_P (reg_info[reg_no]) == MATCH_NULL_UNSET_VALUE)
+ REG_MATCH_NULL_STRING_P (reg_info[reg_no]) = ret;
+
+ if (!ret)
+ return false;
+ break;
+
+ /* If this is an optimized succeed_n for zero times, make the jump. */
+ case jump:
+ EXTRACT_NUMBER_AND_INCR (mcnt, p1);
+ if (mcnt >= 0)
+ p1 += mcnt;
+ else
+ return false;
+ break;
+
+ case succeed_n:
+ /* Get to the number of times to succeed. */
+ p1 += 2;
+ EXTRACT_NUMBER_AND_INCR (mcnt, p1);
+
+ if (mcnt == 0)
+ {
+ p1 -= 4;
+ EXTRACT_NUMBER_AND_INCR (mcnt, p1);
+ p1 += mcnt;
+ }
+ else
+ return false;
+ break;
+
+ case duplicate:
+ if (!REG_MATCH_NULL_STRING_P (reg_info[*p1]))
+ return false;
+ break;
+
+ case set_number_at:
+ p1 += 4;
+
+ default:
+ /* All other opcodes mean we cannot match the empty string. */
+ return false;
+ }
+
+ *p = p1;
+ return true;
+} /* common_op_match_null_string_p */
+
+
+/* Return zero if TRANSLATE[S1] and TRANSLATE[S2] are identical for LEN
+ bytes; nonzero otherwise. */
+
+static int
+bcmp_translate (s1, s2, len, translate)
+ unsigned char *s1, *s2;
+ register int len;
+ char *translate;
+{
+ register unsigned char *p1 = s1, *p2 = s2;
+ while (len)
+ {
+ if (translate[*p1++] != translate[*p2++]) return 1;
+ len--;
+ }
+ return 0;
+}
+
+/* Entry points for GNU code. */
+
+/* re_compile_pattern is the GNU regular expression compiler: it
+ compiles PATTERN (of length SIZE) 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.
+
+ We call regex_compile to do the actual compilation. */
+
+const char *
+re_compile_pattern (pattern, length, bufp)
+ const char *pattern;
+ int length;
+ struct re_pattern_buffer *bufp;
+{
+ reg_errcode_t ret;
+
+ /* GNU code is written to assume at least RE_NREGS registers will be set
+ (and at least one extra will be -1). */
+ bufp->regs_allocated = REGS_UNALLOCATED;
+
+ /* 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. */
+ bufp->no_sub = 0;
+
+ /* Match anchors at newline. */
+ bufp->newline_anchor = 1;
+
+ ret = regex_compile (pattern, length, re_syntax_options, bufp);
+
+ if (!ret)
+ return NULL;
+ return gettext (re_error_msgid[(int) ret]);
+}
+
+/* Entry points compatible with 4.2 BSD regex library. We don't define
+ them unless specifically requested. */
+
+#ifdef _REGEX_RE_COMP
+
+/* BSD has one and only one pattern buffer. */
+static struct re_pattern_buffer re_comp_buf;
+
+char *
+re_comp (s)
+ const char *s;
+{
+ reg_errcode_t ret;
+
+ if (!s)
+ {
+ if (!re_comp_buf.buffer)
+ return gettext ("No previous regular expression");
+ return 0;
+ }
+
+ if (!re_comp_buf.buffer)
+ {
+ re_comp_buf.buffer = (unsigned char *) malloc (200);
+ if (re_comp_buf.buffer == NULL)
+ return gettext (re_error_msgid[(int) REG_ESPACE]);
+ re_comp_buf.allocated = 200;
+
+ re_comp_buf.fastmap = (char *) malloc (1 << BYTEWIDTH);
+ if (re_comp_buf.fastmap == NULL)
+ return gettext (re_error_msgid[(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 = regex_compile (s, strlen (s), re_syntax_options, &re_comp_buf);
+
+ if (!ret)
+ return NULL;
+
+ /* Yes, we're discarding `const' here if !HAVE_LIBINTL. */
+ return (char *) gettext (re_error_msgid[(int) ret]);
+}
+
+
+int
+re_exec (s)
+ const char *s;
+{
+ const int len = strlen (s);
+ return
+ 0 <= re_search (&re_comp_buf, s, len, 0, len, (struct re_registers *) 0);
+}
+#endif /* _REGEX_RE_COMP */
+
+/* POSIX.2 functions. Don't define these for Emacs. */
+
+#ifndef emacs
+
+/* 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' and `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 (preg, pattern, cflags)
+ regex_t *preg;
+ const char *pattern;
+ int cflags;
+{
+ reg_errcode_t ret;
+ unsigned syntax
+ = (cflags & REG_EXTENDED) ?
+ RE_SYNTAX_POSIX_EXTENDED : RE_SYNTAX_POSIX_BASIC;
+
+ /* regex_compile will allocate the space for the compiled pattern. */
+ preg->buffer = 0;
+ preg->allocated = 0;
+ preg->used = 0;
+
+ /* Don't bother to use a fastmap when searching. This simplifies the
+ REG_NEWLINE case: if we used a fastmap, we'd have to put all the
+ characters after newlines into the fastmap. This way, we just try
+ every character. */
+ preg->fastmap = 0;
+
+ if (cflags & REG_ICASE)
+ {
+ unsigned i;
+
+ preg->translate = (char *) malloc (CHAR_SET_SIZE);
+ if (preg->translate == NULL)
+ return (int) REG_ESPACE;
+
+ /* Map uppercase characters to corresponding lowercase ones. */
+ for (i = 0; i < CHAR_SET_SIZE; i++)
+ preg->translate[i] = ISUPPER (i) ? tolower (i) : i;
+ }
+ else
+ preg->translate = NULL;
+
+ /* 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);
+
+ /* POSIX says a null character in the pattern terminates it, so we
+ can use strlen here in compiling the pattern. */
+ ret = regex_compile (pattern, strlen (pattern), syntax, preg);
+
+ /* 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;
+
+ return (int) ret;
+}
+
+
+/* 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 (preg, string, nmatch, pmatch, eflags)
+ const regex_t *preg;
+ const char *string;
+ size_t nmatch;
+ regmatch_t pmatch[];
+ int eflags;
+{
+ int ret;
+ struct re_registers regs;
+ regex_t private_preg;
+ int len = strlen (string);
+ boolean want_reg_info = !preg->no_sub && nmatch > 0;
+
+ private_preg = *preg;
+
+ private_preg.not_bol = !!(eflags & REG_NOTBOL);
+ private_preg.not_eol = !!(eflags & REG_NOTEOL);
+
+ /* The user has told us exactly how many registers to return
+ information about, via `nmatch'. We have to pass that on to the
+ matching routines. */
+ private_preg.regs_allocated = REGS_FIXED;
+
+ if (want_reg_info)
+ {
+ regs.num_regs = nmatch;
+ regs.start = TALLOC (nmatch, regoff_t);
+ regs.end = TALLOC (nmatch, regoff_t);
+ if (regs.start == NULL || regs.end == NULL)
+ return (int) REG_NOMATCH;
+ }
+
+ /* Perform the searching operation. */
+ ret = re_search (&private_preg, string, len,
+ /* start: */ 0, /* range: */ len,
+ want_reg_info ? &regs : (struct re_registers *) 0);
+
+ /* Copy the register information to the POSIX structure. */
+ if (want_reg_info)
+ {
+ if (ret >= 0)
+ {
+ unsigned r;
+
+ for (r = 0; r < nmatch; r++)
+ {
+ pmatch[r].rm_so = regs.start[r];
+ pmatch[r].rm_eo = regs.end[r];
+ }
+ }
+
+ /* If we needed the temporary register info, free the space now. */
+ free (regs.start);
+ free (regs.end);
+ }
+
+ /* We want zero return to mean success, unlike `re_search'. */
+ return ret >= 0 ? (int) REG_NOERROR : (int) REG_NOMATCH;
+}
+
+
+/* Returns a message corresponding to an error code, ERRCODE, returned
+ from either regcomp or regexec. We don't use PREG here. */
+
+size_t
+regerror (errcode, preg, errbuf, errbuf_size)
+ int errcode;
+ const regex_t *preg;
+ char *errbuf;
+ size_t errbuf_size;
+{
+ const char *msg;
+ size_t msg_size;
+
+ if (errcode < 0
+ || errcode >= (sizeof (re_error_msgid) / sizeof (re_error_msgid[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[errcode]);
+
+ msg_size = strlen (msg) + 1; /* Includes the null. */
+
+ if (errbuf_size != 0)
+ {
+ if (msg_size > errbuf_size)
+ {
+ strncpy (errbuf, msg, errbuf_size - 1);
+ errbuf[errbuf_size - 1] = 0;
+ }
+ else
+ strcpy (errbuf, msg);
+ }
+
+ return msg_size;
+}
+
+
+/* Free dynamically allocated space used by PREG. */
+
+void
+regfree (preg)
+ regex_t *preg;
+{
+ if (preg->buffer != NULL)
+ free (preg->buffer);
+ preg->buffer = NULL;
+
+ preg->allocated = 0;
+ preg->used = 0;
+
+ if (preg->fastmap != NULL)
+ free (preg->fastmap);
+ preg->fastmap = NULL;
+ preg->fastmap_accurate = 0;
+
+ if (preg->translate != NULL)
+ free (preg->translate);
+ preg->translate = NULL;
+}
+
+#endif /* not emacs */
+
+/*
+Local variables:
+make-backup-files: t
+version-control: t
+trim-versions-without-asking: nil
+End:
+*/
diff --git a/posix/regex.h b/posix/regex.h
new file mode 100644
index 0000000000..42afd84881
--- /dev/null
+++ b/posix/regex.h
@@ -0,0 +1,495 @@
+/* Definitions for data structures and routines for the regular
+ expression library, version 0.12.
+
+ Copyright (C) 1985, 89, 90, 91, 92, 93, 95 Free Software Foundation, Inc.
+
+This file is part of the GNU C Library. Its master source is NOT part of
+the C library, however. The master source lives in /gd/gnu/lib.
+
+The GNU C Library is free software; you can redistribute it and/or
+modify it under the terms of the GNU Library General Public License as
+published by the Free Software Foundation; either version 2 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
+Library General Public License for more details.
+
+You should have received a copy of the GNU Library General Public
+License along with the GNU C Library; see the file COPYING.LIB. If
+not, write to the Free Software Foundation, Inc., 675 Mass Ave,
+Cambridge, MA 02139, USA. */
+
+#ifndef __REGEXP_LIBRARY_H__
+#define __REGEXP_LIBRARY_H__
+
+/* POSIX says that <sys/types.h> must be included (by the caller) before
+ <regex.h>. */
+
+#if !defined (_POSIX_C_SOURCE) && !defined (_POSIX_SOURCE) && defined (VMS)
+/* VMS doesn't have `size_t' in <sys/types.h>, even though POSIX says it
+ should be there. */
+#include <stddef.h>
+#endif
+
+
+/* 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 reg_syntax_t;
+
+/* 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 (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)
+
+/* 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;
+
+/* 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_UNMATCHED_RIGHT_PAREN_ORD)
+
+#define RE_SYNTAX_POSIX_AWK \
+ (RE_SYNTAX_POSIX_EXTENDED | RE_BACKSLASH_ESCAPE_IN_LISTS)
+
+#define RE_SYNTAX_GREP \
+ (RE_BK_PLUS_QM | RE_CHAR_CLASSES \
+ | RE_HAT_LISTS_NOT_NEWLINE | RE_INTERVALS \
+ | RE_NEWLINE_ALT)
+
+#define RE_SYNTAX_EGREP \
+ (RE_CHAR_CLASSES | RE_CONTEXT_INDEP_ANCHORS \
+ | RE_CONTEXT_INDEP_OPS | RE_HAT_LISTS_NOT_NEWLINE \
+ | RE_NEWLINE_ALT | RE_NO_BK_PARENS \
+ | RE_NO_BK_VBAR)
+
+#define RE_SYNTAX_POSIX_EGREP \
+ (RE_SYNTAX_EGREP | RE_INTERVALS | RE_NO_BK_BRACES)
+
+/* 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)
+
+/* 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_UNMATCHED_RIGHT_PAREN_ORD)
+
+/* Differs from ..._POSIX_EXTENDED in that RE_CONTEXT_INVALID_OPS
+ replaces RE_CONTEXT_INDEP_OPS 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. Some systems
+ (erroneously) define this in other header files, but we want our
+ value, so remove any previous define. */
+#ifdef RE_DUP_MAX
+#undef RE_DUP_MAX
+#endif
+#define RE_DUP_MAX ((1 << 15) - 1)
+
+
+/* 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 (REG_EXTENDED << 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 (REG_ICASE << 1)
+
+/* 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 (REG_NEWLINE << 1)
+
+
+/* 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)
+
+
+/* If any error codes are removed, changed, or added, update the
+ `re_error_msg' table in regex.c. */
+typedef enum
+{
+ 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, /* Not implemented. */
+ 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, /* Compiled pattern bigger than 2^16 bytes. */
+ REG_ERPAREN /* Unmatched ) or \); not returned from regcomp. */
+} reg_errcode_t;
+
+/* 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
+ compiled, the `re_nsub' field is available. All other fields are
+ private to the regex routines. */
+
+struct re_pattern_buffer
+{
+/* [[[begin pattern_buffer]]] */
+ /* Space that holds the compiled pattern. It is declared as
+ `unsigned char *' because its elements are
+ sometimes used as array indexes. */
+ unsigned char *buffer;
+
+ /* Number of bytes to which `buffer' points. */
+ unsigned long allocated;
+
+ /* Number of bytes actually used in `buffer'. */
+ unsigned long used;
+
+ /* Syntax setting with which the pattern was compiled. */
+ reg_syntax_t 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 *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. */
+ char *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 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. */
+#define REGS_UNALLOCATED 0
+#define REGS_REALLOCATE 1
+#define REGS_FIXED 2
+ unsigned regs_allocated : 2;
+
+ /* Set to zero when `regex_compile' compiles a pattern; set to one
+ by `re_compile_fastmap' if it updates the fastmap. */
+ unsigned fastmap_accurate : 1;
+
+ /* If set, `re_match_2' does not return information about
+ subexpressions. */
+ unsigned no_sub : 1;
+
+ /* If set, a beginning-of-line anchor doesn't match at the
+ beginning of the string. */
+ unsigned not_bol : 1;
+
+ /* Similarly for an end-of-line anchor. */
+ unsigned not_eol : 1;
+
+ /* If true, an anchor at a newline matches. */
+ unsigned newline_anchor : 1;
+
+/* [[[end pattern_buffer]]] */
+};
+
+typedef struct re_pattern_buffer regex_t;
+
+/* Type for byte offsets within the string. POSIX mandates this. */
+typedef int regoff_t;
+
+
+/* This is the structure we store register match data in. See
+ regex.texinfo for a full description of what registers match. */
+struct re_registers
+{
+ unsigned 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
+
+
+/* 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. */
+
+/* To avoid duplicating every routine declaration -- once with a
+ prototype (if we are ANSI), and once without (if we aren't) -- we
+ use the following macro to declare argument types. This
+ unfortunately clutters up the declarations a bit, but I think it's
+ worth it. */
+
+#if __STDC__
+
+#define _RE_ARGS(args) args
+
+#else /* not __STDC__ */
+
+#define _RE_ARGS(args) ()
+
+#endif /* not __STDC__ */
+
+/* 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 _RE_ARGS ((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. */
+extern const char *re_compile_pattern
+ _RE_ARGS ((const char *pattern, int 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 _RE_ARGS ((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 int re_search
+ _RE_ARGS ((struct re_pattern_buffer *buffer, const char *string,
+ int length, int start, int 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 int re_search_2
+ _RE_ARGS ((struct re_pattern_buffer *buffer, const char *string1,
+ int length1, const char *string2, int length2,
+ int start, int range, struct re_registers *regs, int stop));
+
+
+/* Like `re_search', but return how many characters in STRING the regexp
+ in BUFFER matched, starting at position START. */
+extern int re_match
+ _RE_ARGS ((struct re_pattern_buffer *buffer, const char *string,
+ int length, int start, struct re_registers *regs));
+
+
+/* Relates to `re_match' as `re_search_2' relates to `re_search'. */
+extern int re_match_2
+ _RE_ARGS ((struct re_pattern_buffer *buffer, const char *string1,
+ int length1, const char *string2, int length2,
+ int start, struct re_registers *regs, int 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
+ PATTERN_BUFFER will allocate its own register data, without
+ freeing the old data. */
+extern void re_set_registers
+ _RE_ARGS ((struct re_pattern_buffer *buffer, struct re_registers *regs,
+ unsigned num_regs, regoff_t *starts, regoff_t *ends));
+
+#ifdef _REGEX_RE_COMP
+/* 4.2 bsd compatibility. */
+extern char *re_comp _RE_ARGS ((const char *));
+extern int re_exec _RE_ARGS ((const char *));
+#endif
+
+/* POSIX compatibility. */
+extern int regcomp _RE_ARGS ((regex_t *preg, const char *pattern, int cflags));
+extern int regexec
+ _RE_ARGS ((const regex_t *preg, const char *string, size_t nmatch,
+ regmatch_t pmatch[], int eflags));
+extern size_t regerror
+ _RE_ARGS ((int errcode, const regex_t *preg, char *errbuf,
+ size_t errbuf_size));
+extern void regfree _RE_ARGS ((regex_t *preg));
+
+#endif /* not __REGEXP_LIBRARY_H__ */
+
+/*
+Local variables:
+make-backup-files: t
+version-control: t
+trim-versions-without-asking: nil
+End:
+*/
diff --git a/regex.h b/regex.h
new file mode 100644
index 0000000000..22bfc89e77
--- /dev/null
+++ b/regex.h
@@ -0,0 +1 @@
+#include <posix/regex.h>
generated by cgit v1.2.3 (git 2.25.1) at 2024-11-08 20:02:57 +0000