From 82d766043c6a8dcf6283788419f110dd7ab52f80 Mon Sep 17 00:00:00 2001 From: Eric Andersen Date: Fri, 20 Oct 2000 03:48:11 +0000 Subject: A smaller, kinder, gentler regexp implementation. --- libc/misc/regex/Makefile | 2 +- libc/misc/regex/regex.c | 5725 ++++++++++++++++++++++++++++++++++++ libc/misc/regex/rx.c | 7273 ---------------------------------------------- 3 files changed, 5726 insertions(+), 7274 deletions(-) create mode 100644 libc/misc/regex/regex.c delete mode 100644 libc/misc/regex/rx.c (limited to 'libc') diff --git a/libc/misc/regex/Makefile b/libc/misc/regex/Makefile index c4c13f6cf..38b7e98bf 100644 --- a/libc/misc/regex/Makefile +++ b/libc/misc/regex/Makefile @@ -24,7 +24,7 @@ TOPDIR=../../ include $(TOPDIR)Rules.mak LIBC=$(TOPDIR)libc.a -CSRC=rx.c +CSRC=regex.c COBJS=$(patsubst %.c,%.o, $(CSRC)) OBJS=$(COBJS) diff --git a/libc/misc/regex/regex.c b/libc/misc/regex/regex.c new file mode 100644 index 000000000..64e754ee0 --- /dev/null +++ b/libc/misc/regex/regex.c @@ -0,0 +1,5725 @@ +/* Extended regular expression matching and search library, + version 0.12. + (Implements POSIX draft P1003.2/D11.2, except for some of the + internationalization features.) + Copyright (C) 1993-1999, 2000 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., 59 Temple Place - Suite 330, + Boston, MA 02111-1307, USA. */ + +/* AIX requires this to be the first thing in the file. */ +#if defined _AIX && !defined REGEX_MALLOC +#pragma alloca +#endif + +#undef _GNU_SOURCE +#define _GNU_SOURCE +#define STDC_HEADERS + +#ifdef HAVE_CONFIG_H +# include +#endif + +#ifndef PARAMS +# if defined __GNUC__ || (defined __STDC__ && __STDC__) +# define PARAMS(args) args +# else +# define PARAMS(args) () +# endif /* GCC. */ +#endif /* Not PARAMS. */ + +#if defined STDC_HEADERS && !defined emacs +# include +#else +/* We need this for `regex.h', and perhaps for the Emacs include files. */ +# include +#endif + +#define WIDE_CHAR_SUPPORT (HAVE_WCTYPE_H && HAVE_WCHAR_H && HAVE_BTOWC) + +/* For platform which support the ISO C amendement 1 functionality we + support user defined character classes. */ +#if defined _LIBC || WIDE_CHAR_SUPPORT +/* Solaris 2.5 has a bug: must be included before . */ +# include +# include +#endif + +#ifdef _LIBC +/* We have to keep the namespace clean. */ +# define regfree(preg) __regfree (preg) +# define regexec(pr, st, nm, pm, ef) __regexec (pr, st, nm, pm, ef) +# define regcomp(preg, pattern, cflags) __regcomp (preg, pattern, cflags) +# define regerror(errcode, preg, errbuf, errbuf_size) \ + __regerror(errcode, preg, errbuf, errbuf_size) +# define re_set_registers(bu, re, nu, st, en) \ + __re_set_registers (bu, re, nu, st, en) +# define re_match_2(bufp, string1, size1, string2, size2, pos, regs, stop) \ + __re_match_2 (bufp, string1, size1, string2, size2, pos, regs, stop) +# define re_match(bufp, string, size, pos, regs) \ + __re_match (bufp, string, size, pos, regs) +# define re_search(bufp, string, size, startpos, range, regs) \ + __re_search (bufp, string, size, startpos, range, regs) +# define re_compile_pattern(pattern, length, bufp) \ + __re_compile_pattern (pattern, length, bufp) +# define re_set_syntax(syntax) __re_set_syntax (syntax) +# define re_search_2(bufp, st1, s1, st2, s2, startpos, range, regs, stop) \ + __re_search_2 (bufp, st1, s1, st2, s2, startpos, range, regs, stop) +# define re_compile_fastmap(bufp) __re_compile_fastmap (bufp) + +#define btowc __btowc +#endif + +/* This is for other GNU distributions with internationalized messages. */ +#if HAVE_LIBINTL_H || defined _LIBC +# include +#else +# define gettext(msgid) (msgid) +#endif + +#ifndef gettext_noop +/* This define is so xgettext can find the internationalizable + strings. */ +# define gettext_noop(String) String +#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 +# else +char *malloc(); +char *realloc(); +# endif + +/* When used in Emacs's lib-src, we need to get bzero and bcopy somehow. + If nothing else has been done, use the method below. */ +# ifdef INHIBIT_STRING_HEADER +# if !(defined HAVE_BZERO && defined HAVE_BCOPY) +# if !defined bzero && !defined bcopy +# undef INHIBIT_STRING_HEADER +# endif +# endif +# endif + +/* This is the normal way of making sure we have a bcopy and a bzero. + This is used in most programs--a few other programs avoid this + by defining INHIBIT_STRING_HEADER. */ +# ifndef INHIBIT_STRING_HEADER +# if defined HAVE_STRING_H || defined STDC_HEADERS || defined _LIBC +# include +# ifndef bzero +# ifndef _LIBC +# define bzero(s, n) (memset (s, '\0', n), (s)) +# else +# define bzero(s, n) __bzero (s, n) +# endif +# endif +# else +# include +# ifndef memcmp +# define memcmp(s1, s2, n) bcmp (s1, s2, n) +# endif +# ifndef memcpy +# define memcpy(d, s, n) (bcopy (s, d, n), (d)) +# endif +# 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 + +#endif /* not emacs */ + +/* Get the interface, including the syntax bits. */ +#include + +/* isalpha etc. are used for the character classes. */ +#include + +/* 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." + Solaris defines some of these symbols so we must undefine them first. */ + +#undef ISASCII +#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 + +#undef ISPRINT +#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)) + +#ifdef _tolower +# define TOLOWER(c) _tolower(c) +#else +# define TOLOWER(c) tolower(c) +#endif + +#ifndef NULL +# define NULL (void *)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 + +#ifndef emacs +/* How many characters in the character set. */ +# define CHAR_SET_SIZE 256 + +# ifdef SYNTAX_TABLE + +extern char *re_syntax_table; + +# else /* not SYNTAX_TABLE */ + +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 = 0; c < CHAR_SET_SIZE; ++c) + if (ISALNUM(c)) + re_syntax_table[c] = Sword; + + re_syntax_table['_'] = Sword; + + done = 1; +} + +# endif /* not SYNTAX_TABLE */ + +# define SYNTAX(c) re_syntax_table[((c) & 0xFF)] + +#endif /* emacs */ + +/* 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 +# endif /* 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), \ + memcpy (destination, source, osize)) + +/* No need to do anything to free, after alloca. */ +# define REGEX_FREE(arg) ((void)0) /* Do nothing! But inhibit gcc warning. */ + +#endif /* not REGEX_MALLOC */ + +/* Define how to allocate the failure stack. */ + +#if defined REL_ALLOC && defined REGEX_MALLOC + +# 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 using relocating allocator */ + +# 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 using relocating allocator */ + + +/* 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 PARAMS((struct re_pattern_buffer * bufp, + const char *string1, int size1, + const char *string2, int size2, + int pos, + struct re_registers * regs, + + int stop)); + +/* 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 _RE_ARGS((int *dest, unsigned char *source)); +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 _RE_ARGS((int *destination, + unsigned char **source)); +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 + +/* It is useful to test things that ``must'' be true when debugging. */ +# include + +static int debug; + +# 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 *p1; + 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); + p1 = p + mcnt; + extract_number_and_incr(&mcnt2, &p); + printf("/succeed_n to %d, %d times", p1 - start, mcnt2); + break; + + case jump_n: + extract_number_and_incr(&mcnt, &p); + p1 = p + mcnt; + extract_number_and_incr(&mcnt2, &p); + printf("/jump_n to %d, %d times", p1 - start, mcnt2); + break; + + case set_number_at: + extract_number_and_incr(&mcnt, &p); + p1 = p + mcnt; + extract_number_and_incr(&mcnt2, &p); + printf("/set_number_at location %d to %d", p1 - 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("%ld bytes used/%ld 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: %lx\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; +{ + int 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]); + } +} + +void printchar(c) +int c; +{ + putc(c, stderr); +} + +#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; +#ifdef DEBUG + if (syntax & RE_DEBUG) + debug = 1; + else if (debug) /* was on but now is not */ + debug = 0; +#endif /* DEBUG */ + return ret; +} + +#ifdef _LIBC +weak_alias(__re_set_syntax, re_set_syntax) +#endif + /* This table gives an error message for each of the error codes listed + in regex.h. Obviously the order here has to be same as there. + POSIX doesn't require that we do anything for REG_NOERROR, + but why not be nice? */ +static const char re_error_msgid[] = { +#define REG_NOERROR_IDX 0 + gettext_noop("Success") /* REG_NOERROR */ + "\0" +#define REG_NOMATCH_IDX (REG_NOERROR_IDX + sizeof "Success") + gettext_noop("No match") /* REG_NOMATCH */ + "\0" +#define REG_BADPAT_IDX (REG_NOMATCH_IDX + sizeof "No match") + gettext_noop("Invalid regular expression") /* REG_BADPAT */ + "\0" +#define REG_ECOLLATE_IDX (REG_BADPAT_IDX + sizeof "Invalid regular expression") + gettext_noop("Invalid collation character") /* REG_ECOLLATE */ + "\0" +#define REG_ECTYPE_IDX (REG_ECOLLATE_IDX + sizeof "Invalid collation character") + gettext_noop("Invalid character class name") /* REG_ECTYPE */ + "\0" +#define REG_EESCAPE_IDX (REG_ECTYPE_IDX + sizeof "Invalid character class name") + gettext_noop("Trailing backslash") /* REG_EESCAPE */ + "\0" +#define REG_ESUBREG_IDX (REG_EESCAPE_IDX + sizeof "Trailing backslash") + gettext_noop("Invalid back reference") /* REG_ESUBREG */ + "\0" +#define REG_EBRACK_IDX (REG_ESUBREG_IDX + sizeof "Invalid back reference") + gettext_noop("Unmatched [ or [^") /* REG_EBRACK */ + "\0" +#define REG_EPAREN_IDX (REG_EBRACK_IDX + sizeof "Unmatched [ or [^") + gettext_noop("Unmatched ( or \\(") /* REG_EPAREN */ + "\0" +#define REG_EBRACE_IDX (REG_EPAREN_IDX + sizeof "Unmatched ( or \\(") + gettext_noop("Unmatched \\{") /* REG_EBRACE */ + "\0" +#define REG_BADBR_IDX (REG_EBRACE_IDX + sizeof "Unmatched \\{") + gettext_noop("Invalid content of \\{\\}") /* REG_BADBR */ + "\0" +#define REG_ERANGE_IDX (REG_BADBR_IDX + sizeof "Invalid content of \\{\\}") + gettext_noop("Invalid range end") /* REG_ERANGE */ + "\0" +#define REG_ESPACE_IDX (REG_ERANGE_IDX + sizeof "Invalid range end") + gettext_noop("Memory exhausted") /* REG_ESPACE */ + "\0" +#define REG_BADRPT_IDX (REG_ESPACE_IDX + sizeof "Memory exhausted") + gettext_noop("Invalid preceding regular expression") /* REG_BADRPT */ + "\0" +#define REG_EEND_IDX (REG_BADRPT_IDX + sizeof "Invalid preceding regular expression") + gettext_noop("Premature end of regular expression") /* REG_EEND */ + "\0" +#define REG_ESIZE_IDX (REG_EEND_IDX + sizeof "Premature end of regular expression") + gettext_noop("Regular expression too big") /* REG_ESIZE */ + "\0" +#define REG_ERPAREN_IDX (REG_ESIZE_IDX + sizeof "Regular expression too big") + gettext_noop("Unmatched ) or \\)") /* REG_ERPAREN */ +}; + +static const size_t re_error_msgid_idx[] = { + REG_NOERROR_IDX, + REG_NOMATCH_IDX, + REG_BADPAT_IDX, + REG_ECOLLATE_IDX, + REG_ECTYPE_IDX, + REG_EESCAPE_IDX, + REG_ESUBREG_IDX, + REG_EBRACK_IDX, + REG_EPAREN_IDX, + REG_EBRACE_IDX, + REG_BADBR_IDX, + REG_ERANGE_IDX, + REG_ESPACE_IDX, + REG_BADRPT_IDX, + REG_EEND_IDX, + REG_ESIZE_IDX, + REG_ERPAREN_IDX +}; + +/* 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_ITEMS items each time we failed. + This is a variable only so users of regex can assign to it; we never + change it ourselves. */ + +#ifdef INT_IS_16BIT + +# if defined MATCH_MAY_ALLOCATE +/* 4400 was enough to cause a crash on Alpha OSF/1, + whose default stack limit is 2mb. */ +long int re_max_failures = 4000; +# else +long int re_max_failures = 2000; +# endif + +union fail_stack_elt { + unsigned char *pointer; + long int integer; +}; + +typedef union fail_stack_elt fail_stack_elt_t; + +typedef struct { + fail_stack_elt_t *stack; + unsigned long int size; + unsigned long int avail; /* Offset of next open position. */ +} fail_stack_type; + +#else /* not INT_IS_16BIT */ + +# if defined MATCH_MAY_ALLOCATE +/* 4400 was enough to cause a crash on Alpha OSF/1, + whose default stack limit is 2mb. */ +int re_max_failures = 20000; +# 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; + +#endif /* INT_IS_16BIT */ + +#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 > (unsigned) (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_pushed 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. */ \ + /* Can't be int, since there is not a shred of a guarantee that int \ + is wide enough to hold a value of something to which pointer can \ + be assigned */ \ + active_reg_t 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: %ld\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"); \ + \ + if (1) \ + for (this_reg = lowest_active_reg; this_reg <= highest_active_reg; \ + this_reg++) \ + { \ + DEBUG_PRINT2 (" Pushing reg: %lu\n", this_reg); \ + DEBUG_STATEMENT (num_regs_pushed++); \ + \ + DEBUG_PRINT2 (" start: %p\n", regstart[this_reg]); \ + PUSH_FAILURE_POINTER (regstart[this_reg]); \ + \ + DEBUG_PRINT2 (" end: %p\n", regend[this_reg]); \ + PUSH_FAILURE_POINTER (regend[this_reg]); \ + \ + DEBUG_PRINT2 (" info: %p\n ", \ + reg_info[this_reg].word.pointer); \ + 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: %ld\n", lowest_active_reg);\ + PUSH_FAILURE_INT (lowest_active_reg); \ + \ + DEBUG_PRINT2 (" Pushing high active reg: %ld\n", highest_active_reg);\ + PUSH_FAILURE_INT (highest_active_reg); \ + \ + DEBUG_PRINT2 (" Pushing pattern %p:\n", pattern_place); \ + DEBUG_PRINT_COMPILED_PATTERN (bufp, pattern_place, pend); \ + PUSH_FAILURE_POINTER (pattern_place); \ + \ + DEBUG_PRINT2 (" Pushing string %p: `", 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. */ +/* We used to use (num_regs - 1), which is the number of registers + this regexp will save; but that was changed to 5 + to avoid stack overflow for a regexp with lots of parens. */ +#define MAX_FAILURE_ITEMS (5 * NUM_REG_ITEMS + NUM_NONREG_ITEMS) + +/* We actually push this many items. */ +#define NUM_FAILURE_ITEMS \ + (((0 \ + ? 0 : 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 (unsigned failure_id;) \ + active_reg_t 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 %p: `", str); \ + DEBUG_PRINT_DOUBLE_STRING (str, string1, size1, string2, size2); \ + DEBUG_PRINT1 ("'\n"); \ + \ + pat = (unsigned char *) POP_FAILURE_POINTER (); \ + DEBUG_PRINT2 (" Popping pattern %p:\n", pat); \ + DEBUG_PRINT_COMPILED_PATTERN (bufp, pat, pend); \ + \ + /* Restore register info. */ \ + high_reg = (active_reg_t) POP_FAILURE_INT (); \ + DEBUG_PRINT2 (" Popping high active reg: %ld\n", high_reg); \ + \ + low_reg = (active_reg_t) POP_FAILURE_INT (); \ + DEBUG_PRINT2 (" Popping low active reg: %ld\n", low_reg); \ + \ + if (1) \ + for (this_reg = high_reg; this_reg >= low_reg; this_reg--) \ + { \ + DEBUG_PRINT2 (" Popping reg: %ld\n", this_reg); \ + \ + reg_info[this_reg].word = POP_FAILURE_ELT (); \ + DEBUG_PRINT2 (" info: %p\n", \ + reg_info[this_reg].word.pointer); \ + \ + regend[this_reg] = (const char *) POP_FAILURE_POINTER (); \ + DEBUG_PRINT2 (" end: %p\n", regend[this_reg]); \ + \ + regstart[this_reg] = (const char *) POP_FAILURE_POINTER (); \ + DEBUG_PRINT2 (" start: %p\n", regstart[this_reg]); \ + } \ + else \ + { \ + for (this_reg = highest_active_reg; this_reg > high_reg; this_reg--) \ + { \ + reg_info[this_reg].word.integer = 0; \ + regend[this_reg] = 0; \ + regstart[this_reg] = 0; \ + } \ + highest_active_reg = high_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. */ + + +/* Declarations and macros for re_match_2. */ + +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) \ + { \ + active_reg_t 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 (®_unset_dummy) +#define REG_UNSET(e) ((e) == REG_UNSET_VALUE) + +/* Subroutine declarations and macros for regex_compile. */ + +static reg_errcode_t regex_compile +_RE_ARGS( + (const char *pattern, size_t size, reg_syntax_t syntax, + struct re_pattern_buffer * bufp)); +static void store_op1 + +_RE_ARGS((re_opcode_t op, unsigned char *loc, int arg)); +static void store_op2 +_RE_ARGS((re_opcode_t op, unsigned char *loc, int arg1, int arg2)); +static void insert_op1 +_RE_ARGS( + + (re_opcode_t op, unsigned char *loc, int arg, + unsigned char *end)); +static void insert_op2 +_RE_ARGS( + (re_opcode_t op, unsigned char *loc, int arg1, int arg2, + + unsigned char *end)); +static boolean at_begline_loc_p +_RE_ARGS((const char *pattern, const char *p, reg_syntax_t syntax)); +static boolean at_endline_loc_p +_RE_ARGS((const char *p, const char *pend, reg_syntax_t syntax)); +static reg_errcode_t compile_range +_RE_ARGS( + (const char **p_ptr, const char *pend, char *translate, + reg_syntax_t syntax, unsigned char *b)); + +/* 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'). */ +#ifndef PATFETCH +# define PATFETCH(c) \ + do {if (p == pend) return REG_EEND; \ + c = (unsigned char) *p++; \ + if (translate) c = (unsigned char) translate[c]; \ + } while (0) +#endif + +/* 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. */ +#ifndef TRANSLATE +# define TRANSLATE(d) \ + (translate ? (char) translate[(unsigned char) (d)] : (d)) +#endif + + +/* 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 ((unsigned long) (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, (int) ((to) - (loc) - 3)) + +/* Likewise, for a two-argument jump. */ +#define STORE_JUMP2(op, loc, to, arg) \ + store_op2 (op, loc, (int) ((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, (int) ((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, (int) ((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. */ +/* Any other compiler which, like MSC, has allocation limit below 2^16 + bytes will have to use approach similar to what was done below for + MSC and drop MAX_BUF_SIZE a bit. Otherwise you may end up + reallocating to 0 bytes. Such thing is not going to work too well. + You have been warned!! */ +#if defined _MSC_VER && !defined WIN32 +/* Microsoft C 16-bit versions limit malloc to approx 65512 bytes. + The REALLOC define eliminates a flurry of conversion warnings, + but is not required. */ +# define MAX_BUF_SIZE 65500L +# define REALLOC(p,s) realloc ((p), (size_t) (s)) +#else +# define MAX_BUF_SIZE (1L << 16) +# define REALLOC(p,s) realloc ((p), (s)) +#endif + +/* 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. */ +/* int may be not enough when sizeof(int) == 2. */ +typedef long 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 ('0' <= c && c <= '9') \ + { \ + if (num < 0) \ + num = 0; \ + num = num * 10 + c - '0'; \ + if (p == pend) \ + break; \ + PATFETCH (c); \ + } \ + } \ + } + +#if defined _LIBC || WIDE_CHAR_SUPPORT +/* The GNU C library provides support for user-defined character classes + and the functions from ISO C amendement 1. */ +# ifdef CHARCLASS_NAME_MAX +# define CHAR_CLASS_MAX_LENGTH CHARCLASS_NAME_MAX +# else +/* This shouldn't happen but some implementation might still have this + problem. Use a reasonable default value. */ +# define CHAR_CLASS_MAX_LENGTH 256 +# endif + +# ifdef _LIBC +# define IS_CHAR_CLASS(string) __wctype (string) +# else +# define IS_CHAR_CLASS(string) wctype (string) +# endif +#else +# 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")) +#endif + +#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 */ + +static boolean group_in_compile_stack _RE_ARGS((compile_stack_type + compile_stack, + + regnum_t regnum)); + +/* `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; +size_t 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. */ + RE_TRANSLATE_TYPE 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 == ':' && *p == ']') || p == pend) + break; + if (c1 < CHAR_CLASS_MAX_LENGTH) + str[c1++] = c; + else + /* This is in any case an invalid class name. */ + str[0] = '\0'; + } + 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 == ']') { +#if defined _LIBC || WIDE_CHAR_SUPPORT + boolean is_lower = STREQ(str, "lower"); + boolean is_upper = STREQ(str, "upper"); + wctype_t wt; + int ch; + + wt = IS_CHAR_CLASS(str); + if (wt == 0) + 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) { +# ifdef _LIBC + if (__iswctype(__btowc(ch), wt)) + SET_LIST_BIT(ch); +# else + if (iswctype(btowc(ch), wt)) + SET_LIST_BIT(ch); +# endif + + if (translate && (is_upper || is_lower) + && (ISUPPER(ch) || ISLOWER(ch))) + SET_LIST_BIT(ch); + } + + had_char_class = true; +#else + 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); + if (translate && (is_upper || is_lower) + && (ISUPPER(ch) || ISLOWER(ch))) + SET_LIST_BIT(ch); + } + had_char_class = true; +#endif /* libc || wctype.h */ + } 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_INTERVALS) + && (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 ((!(syntax & RE_NO_BK_BRACES) && c != '\\') + || ((syntax & RE_NO_BK_BRACES) && c != '}')) + FREE_STACK_RETURN(REG_BADBR); + + 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_INTERVALS) + && (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_INTERVALS) + && (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 + set_number_at + succeed_n + + jump_n + (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': + if (syntax & RE_NO_GNU_OPS) + goto normal_char; + laststart = b; + BUF_PUSH(wordchar); + break; + + + case 'W': + if (syntax & RE_NO_GNU_OPS) + goto normal_char; + laststart = b; + BUF_PUSH(notwordchar); + break; + + + case '<': + if (syntax & RE_NO_GNU_OPS) + goto normal_char; + BUF_PUSH(wordbeg); + break; + + case '>': + if (syntax & RE_NO_GNU_OPS) + goto normal_char; + BUF_PUSH(wordend); + break; + + case 'b': + if (syntax & RE_NO_GNU_OPS) + goto normal_char; + BUF_PUSH(wordbound); + break; + + case 'B': + if (syntax & RE_NO_GNU_OPS) + goto normal_char; + BUF_PUSH(notwordbound); + break; + + case '`': + if (syntax & RE_NO_GNU_OPS) + goto normal_char; + BUF_PUSH(begbuf); + break; + + case '\'': + if (syntax & RE_NO_GNU_OPS) + goto normal_char; + 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, (regnum_t) 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; +reg_syntax_t syntax; +{ + const char *next = p; + boolean next_backslash = *next == '\\'; + const char *next_next = p + 1 < pend ? p + 1 : 0; + + 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; +RE_TRANSLATE_TYPE translate; +reg_syntax_t syntax; +unsigned char *b; +{ + unsigned this_char; + + const char *p = *p_ptr; + reg_errcode_t ret; + char range_start[2]; + char range_end[2]; + char ch[2]; + + if (p == pend) + return REG_ERANGE; + + /* Fetch the endpoints without translating them; the + appropriate translation is done in the bit-setting loop below. */ + range_start[0] = p[-2]; + range_start[1] = '\0'; + range_end[0] = p[0]; + range_end[1] = '\0'; + + /* Have to increment the pointer into the pattern string, so the + caller isn't still at the ending character. */ + (*p_ptr)++; + + /* Report an error if the range is empty and the syntax prohibits this. */ + ret = syntax & RE_NO_EMPTY_RANGES ? REG_ERANGE : REG_NOERROR; + + /* Here we see why `this_char' has to be larger than an `unsigned + char' -- we would otherwise go into an infinite loop, since all + characters <= 0xff. */ + ch[1] = '\0'; + for (this_char = 0; this_char <= (unsigned char) -1; ++this_char) { + ch[0] = this_char; + if (strcoll(range_start, ch) <= 0 && strcoll(ch, range_end) <= 0) { + SET_LIST_BIT(TRANSLATE(this_char)); + ret = REG_NOERROR; + } + } + + return ret; +} + +/* 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 + + register char *fastmap = bufp->fastmap; + unsigned char *pattern = bufp->buffer; + unsigned char *p = pattern; + register unsigned char *pend = pattern + bufp->used; + +#ifdef REL_ALLOC + /* This holds the pointer to the failure stack, when + it is allocated relocatably. */ + fail_stack_elt_t *failure_stack_ptr; +#endif + + /* 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 /* 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 */ + +#ifdef _LIBC +weak_alias(__re_compile_fastmap, re_compile_fastmap) +#endif + /* 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; + } +} + +#ifdef _LIBC +weak_alias(__re_set_registers, re_set_registers) +#endif + /* 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); +} + +#ifdef _LIBC +weak_alias(__re_search, re_search) +#endif +/* 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 RE_TRANSLATE_TYPE 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. + Make sure we won't move STARTPOS below 0 or above TOTAL_SIZE. */ + if (endpos < 0) + range = 0 - 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 && range > 0 + && ((re_opcode_t) bufp->buffer[0] == begbuf + /* `begline' is like `begbuf' if it cannot match at newlines. */ + || ((re_opcode_t) bufp->buffer[0] == begline + && !bufp->newline_anchor))) { + if (startpos > 0) + return -1; + else + range = 1; + } +#ifdef emacs + /* In a forward search for something that starts with \=. + don't keep searching past point. */ + if (bufp->used > 0 && (re_opcode_t) bufp->buffer[0] == at_dot + && range > 0) { + range = PT - startpos; + if (range <= 0) + return -1; + } +#endif /* emacs */ + + /* 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 */ + +#ifdef _LIBC +weak_alias(__re_search_2, re_search_2) +#endif + /* 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) +/* Disabled due to a compiler bug -- see comment at case wordbound */ +#if 0 +/* 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)) +#endif +/* 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); + +# ifndef REGEX_MALLOC +# ifdef C_ALLOCA + alloca(0); +# endif +# endif + return result; +} + +# ifdef _LIBC +weak_alias(__re_match, re_match) +# endif +#endif /* not emacs */ +static boolean group_match_null_string_p _RE_ARGS((unsigned char **p, + unsigned char *end, + register_info_type * + + reg_info)); +static boolean alt_match_null_string_p +_RE_ARGS( + + (unsigned char *p, unsigned char *end, + register_info_type * reg_info)); +static boolean common_op_match_null_string_p +_RE_ARGS( + + (unsigned char **p, unsigned char *end, + register_info_type * reg_info)); +static int bcmp_translate +_RE_ARGS((const char *s1, const char *s2, int len, char *translate)); + +/* 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); + +#ifndef REGEX_MALLOC +# ifdef C_ALLOCA + alloca(0); +# endif +#endif + return result; +} + +#ifdef _LIBC +weak_alias(__re_match_2, re_match_2) +#endif +/* 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. */ + RE_TRANSLATE_TYPE 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; + unsigned nfailure_points_pushed = 0, nfailure_points_popped = 0; +#endif + +#ifdef REL_ALLOC + /* This holds the pointer to the failure stack, when + it is allocated relocatably. */ + fail_stack_elt_t *failure_stack_ptr; +#endif + + /* We fill all the registers internally, independent of what we + return, for use in backreferences. The number here includes + an element for register zero. */ + size_t num_regs = bufp->re_nsub + 1; + + /* The currently active registers. */ + active_reg_t lowest_active_reg = NO_LOWEST_ACTIVE_REG; + active_reg_t 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; (unsigned) 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:\n"); + 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 (;;) { +#ifdef _LIBC + DEBUG_PRINT2("\n%p: ", p); +#else + DEBUG_PRINT2("\n0x%x: ", p); +#endif + + 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; (unsigned) 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; (unsigned) 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; + (unsigned) 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; (unsigned) 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 ((unsigned char) translate[(unsigned char) *d++] + != (unsigned 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 < (unsigned) *p + (unsigned) *(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; + + + /* \ has been turned into a `duplicate' command which is + followed by the numeric value of 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) + : memcmp(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); +#ifdef _LIBC + DEBUG_PRINT3(" %d (to %p):\n", mcnt, p + mcnt); +#else + DEBUG_PRINT3(" %d (to 0x%x):\n", mcnt, p + mcnt); +#endif + + 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); +#ifdef _LIBC + DEBUG_PRINT3(" %d (to %p)", mcnt, p + mcnt); +#else + DEBUG_PRINT3(" %d (to 0x%x)", mcnt, p + mcnt); +#endif + + /* 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 \(zz\(a*\)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) { + /* We win if the first character of the loop is not part + of the charset. */ + if ((re_opcode_t) p1[3] == exactn + && !((int) p2[1] * BYTEWIDTH > (int) p1[5] + && (p2[2 + p1[5] / BYTEWIDTH] + & (1 << (p1[5] % BYTEWIDTH))))) { + p[-3] = (unsigned char) pop_failure_jump; + DEBUG_PRINT1(" No match => pop_failure_jump.\n"); + } + + 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'. */ + active_reg_t 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. */ + + unconditional_jump: +#ifdef _LIBC + DEBUG_PRINT2("\n%p: ", p); +#else + DEBUG_PRINT2("\n0x%x: ", p); +#endif + /* Note fall through. */ + + /* Unconditionally jump (without popping any failure points). */ + case jump: + EXTRACT_NUMBER_AND_INCR(mcnt, p); /* Get the amount to jump. */ + DEBUG_PRINT2("EXECUTING jump %d ", mcnt); + p += mcnt; /* Do the jump. */ +#ifdef _LIBC + DEBUG_PRINT2("(to %p).\n", p); +#else + DEBUG_PRINT2("(to 0x%x).\n", p); +#endif + 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(NULL, NULL, -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(NULL, NULL, -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); +#ifdef _LIBC + DEBUG_PRINT3(" Setting %p to %d.\n", p - 2, mcnt); +#else + DEBUG_PRINT3(" Setting 0x%x to %d.\n", p - 2, mcnt); +#endif + } else if (mcnt == 0) { +#ifdef _LIBC + DEBUG_PRINT2(" Setting two bytes from %p to no_op.\n", + p + 2); +#else + DEBUG_PRINT2(" Setting two bytes from 0x%x to no_op.\n", + p + 2); +#endif + 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); +#ifdef _LIBC + DEBUG_PRINT3(" Setting %p to %d.\n", p + 2, mcnt); +#else + DEBUG_PRINT3(" Setting 0x%x to %d.\n", p + 2, mcnt); +#endif + 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); +#ifdef _LIBC + DEBUG_PRINT3(" Setting %p to %d.\n", p1, mcnt); +#else + DEBUG_PRINT3(" Setting 0x%x to %d.\n", p1, mcnt); +#endif + STORE_NUMBER(p1, mcnt); + break; + } + +#if 0 + /* The DEC Alpha C compiler 3.x generates incorrect code for the + test WORDCHAR_P (d - 1) != WORDCHAR_P (d) in the expansion of + AT_WORD_BOUNDARY, so this code is disabled. Expanding the + macro and introducing temporary variables works around the bug. */ + + 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; +#else + case wordbound: + { + boolean prevchar, thischar; + + DEBUG_PRINT1("EXECUTING wordbound.\n"); + if (AT_STRINGS_BEG(d) || AT_STRINGS_END(d)) + break; + + prevchar = WORDCHAR_P(d - 1); + thischar = WORDCHAR_P(d); + if (prevchar != thischar) + break; + goto fail; + } + + case notwordbound: + { + boolean prevchar, thischar; + + DEBUG_PRINT1("EXECUTING notwordbound.\n"); + if (AT_STRINGS_BEG(d) || AT_STRINGS_END(d)) + goto fail; + + prevchar = WORDCHAR_P(d - 1); + thischar = WORDCHAR_P(d); + if (prevchar != thischar) + goto fail; + break; + } +#endif + + 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; + + 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) +const char *s1, *s2; +register int len; +RE_TRANSLATE_TYPE translate; +{ + register const unsigned char *p1 = (const unsigned char *) s1; + register const unsigned char *p2 = (const unsigned char *) 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; +size_t 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 + re_error_msgid_idx[(int) ret]); +} + +#ifdef _LIBC +weak_alias(__re_compile_pattern, re_compile_pattern) +#endif + /* Entry points compatible with 4.2 BSD regex library. We don't define + them unless specifically requested. */ +#if defined _REGEX_RE_COMP || defined _LIBC +/* BSD has one and only one pattern buffer. */ +static struct re_pattern_buffer re_comp_buf; + +char * +#ifdef _LIBC +/* Make these definitions weak in libc, so POSIX programs can redefine + these names if they don't use our functions, and still use + regcomp/regexec below without link errors. */ weak_function +#endif +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 (char *) gettext(re_error_msgid + + + re_error_msgid_idx[(int) REG_ESPACE]); + re_comp_buf.allocated = 200; + + re_comp_buf.fastmap = (char *) malloc(1 << BYTEWIDTH); + if (re_comp_buf.fastmap == NULL) + return (char *) gettext(re_error_msgid + + + re_error_msgid_idx[(int) REG_ESPACE]); + } + + /* Since `re_exec' always passes NULL for the `regs' argument, we + don't need to initialize the pattern buffer fields which affect it. */ + + /* Match anchors at newlines. */ + re_comp_buf.newline_anchor = 1; + + ret = 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 + + re_error_msgid_idx[(int) ret]); +} + + +int +#ifdef _LIBC + weak_function +#endif +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' to an allocated space for the fastmap; + `fastmap_accurate' to zero; + `re_nsub' to the number of subexpressions in PATTERN. + + PATTERN is the address of the pattern string. + + CFLAGS is a series of bits which affect compilation. + + If REG_EXTENDED is set, we use POSIX extended syntax; otherwise, we + use POSIX basic syntax. + + If REG_NEWLINE is set, then . and [^...] don't match newline. + Also, regexec will try a match beginning after every newline. + + If REG_ICASE is set, then we considers upper- and lowercase + versions of letters to be equivalent when matching. + + If REG_NOSUB is set, then when PREG is passed to regexec, that + routine will report only success or failure, and nothing about the + registers. + + It returns 0 if it succeeds, nonzero if it doesn't. (See regex.h for + the return codes and their meanings.) */ + +int regcomp(preg, pattern, cflags) +regex_t *preg; +const char *pattern; +int cflags; +{ + reg_errcode_t ret; + reg_syntax_t 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; + + /* Try to allocate space for the fastmap. */ + preg->fastmap = (char *) malloc(1 << BYTEWIDTH); + + if (cflags & REG_ICASE) { + unsigned i; + + preg->translate + = (RE_TRANSLATE_TYPE) malloc(CHAR_SET_SIZE + * sizeof(*(RE_TRANSLATE_TYPE) 0)); + 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; + + if (ret == REG_NOERROR && preg->fastmap) { + /* Compute the fastmap now, since regexec cannot modify the pattern + buffer. */ + if (re_compile_fastmap(preg) == -2) { + /* Some error occurred while computing the fastmap, just forget + about it. */ + free(preg->fastmap); + preg->fastmap = NULL; + } + } + + return (int) ret; +} + +#ifdef _LIBC +weak_alias(__regcomp, regcomp) +#endif +/* 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 * 2, regoff_t); + if (regs.start == NULL) + return (int) REG_NOMATCH; + regs.end = regs.start + nmatch; + } + + /* Perform the searching operation. */ + ret = re_search(&private_preg, string, len, + /* start: */ 0, /* range: */ len, + want_reg_info ? ®s : (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); + } + + /* We want zero return to mean success, unlike `re_search'. */ + return ret >= 0 ? (int) REG_NOERROR : (int) REG_NOMATCH; +} + +#ifdef _LIBC +weak_alias(__regexec, regexec) +#endif +/* Returns a message corresponding to an error code, ERRCODE, returned + from either regcomp or regexec. We don't use PREG here. */ + size_t regerror(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 >= (int) (sizeof(re_error_msgid_idx) + / sizeof(re_error_msgid_idx[0]))) + /* Only error codes returned by the rest of the code should be passed + to this routine. If we are given anything else, or if other regex + code generates an invalid error code, then the program has a bug. + Dump core so we can fix it. */ + abort(); + + msg = gettext(re_error_msgid + re_error_msgid_idx[errcode]); + + msg_size = strlen(msg) + 1; /* Includes the null. */ + + if (errbuf_size != 0) { + if (msg_size > errbuf_size) { +#if defined HAVE_MEMPCPY || defined _LIBC + *((char *) __mempcpy(errbuf, msg, errbuf_size - 1)) = '\0'; +#else + memcpy(errbuf, msg, errbuf_size - 1); + errbuf[errbuf_size - 1] = 0; +#endif + } else + memcpy(errbuf, msg, msg_size); + } + + return msg_size; +} + +#ifdef _LIBC +weak_alias(__regerror, regerror) +#endif +/* 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; +} + +#ifdef _LIBC +weak_alias(__regfree, regfree) +#endif +#endif /* not emacs */ diff --git a/libc/misc/regex/rx.c b/libc/misc/regex/rx.c deleted file mode 100644 index 39f77adb6..000000000 --- a/libc/misc/regex/rx.c +++ /dev/null @@ -1,7273 +0,0 @@ -/* Copyright (C) 1992, 1993, 1994, 1995 Free Software Foundation, Inc. - -This file is part of the librx library. - -Librx 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, or (at your option) -any later version. - -Librx is distributed in the hope that it will be useful, but WITHOUT -ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or -FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License -for more details. - -You should have received a copy of the GNU Library General Public -License along with this software; see the file COPYING.LIB. If not, -write to the Free Software Foundation, 675 Mass Ave, Cambridge, MA -02139, USA. */ - -/* NOTE!!! AIX is so losing it requires this to be the first thing in the - * file. - * Do not put ANYTHING before it! - */ -#if !defined (__GNUC__) && defined (_AIX) -#pragma alloca -#endif - -/* To make linux happy? */ -#ifndef _GNU_SOURCE -#define _GNU_SOURCE -#endif - - -#include -#include -#include -#include -#ifndef isgraph -#define isgraph(c) (isprint (c) && !isspace (c)) -#endif -#ifndef isblank -#define isblank(c) ((c) == ' ' || (c) == '\t') -#endif - -#include - -#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 - -#ifndef __GCC__ -#undef __inline__ -#define __inline__ -#endif - -/* 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 -#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 */ - -/* Memory management and stuff for emacs. */ - -#define CHARBITS 8 -#define remalloc(M, S) (M ? realloc (M, S) : malloc (S)) - - -/* Should we use malloc or alloca? If REGEX_MALLOC is not defined, we - * use `alloca' instead of `malloc' for the backtracking stack. - * - * Emacs will die miserably if we don't do this. - */ - -#ifdef REGEX_MALLOC -#define REGEX_ALLOCATE malloc -#else /* not REGEX_MALLOC */ -#define REGEX_ALLOCATE alloca -#endif /* not REGEX_MALLOC */ - - -#ifdef RX_WANT_RX_DEFS -#define RX_DECL extern -#define RX_DEF_QUAL -#else -#define RX_WANT_RX_DEFS -#define RX_DECL static -#define RX_DEF_QUAL static -#endif - -#include -#undef RX_DECL -#define RX_DECL RX_DEF_QUAL - - -/* - * Prototypes. - */ -#ifdef __STDC__ -RX_DECL struct rx_hash_item -*rx_hash_find(struct rx_hash *, unsigned long, - - void *, struct rx_hash_rules *); -RX_DECL struct rx_hash_item -*rx_hash_find(struct rx_hash *, unsigned long, - - void *, struct rx_hash_rules *); -RX_DECL struct rx_hash_item -*rx_hash_store(struct rx_hash *, unsigned long, - - void *, struct rx_hash_rules *); -RX_DECL void rx_hash_free(struct rx_hash_item *, struct rx_hash_rules *); -RX_DECL void rx_free_hash_table(struct rx_hash *, rx_hash_freefn, - - struct rx_hash_rules *); -RX_DECL rx_Bitset rx_cset(struct rx *); -RX_DECL rx_Bitset rx_copy_cset(struct rx *, rx_Bitset); -RX_DECL void rx_free_cset(struct rx *, rx_Bitset); -static struct rx_hash_item -*compiler_hash_item_alloc(struct rx_hash_rules *, void *); -static struct rx_hash -*compiler_hash_alloc(struct rx_hash_rules *); -static void compiler_free_hash(struct rx_hash *, struct rx_hash_rules *); -static void compiler_free_hash_item(struct rx_hash_item *, - - struct rx_hash_rules *); -RX_DECL struct rexp_node -*rexp_node(struct rx *, enum rexp_node_type); -RX_DECL struct rexp_node -*rx_mk_r_cset(struct rx *, rx_Bitset); -RX_DECL struct rexp_node -*rx_mk_r_concat(struct rx *, struct rexp_node *, struct rexp_node *); -RX_DECL struct rexp_node -*rx_mk_r_alternate(struct rx *, struct rexp_node *, struct rexp_node *); -RX_DECL struct rexp_node -*rx_mk_r_alternate(struct rx *, struct rexp_node *, struct rexp_node *); -RX_DECL struct rexp_node -*rx_mk_r_opt(struct rx *, struct rexp_node *); -RX_DECL struct rexp_node -*rx_mk_r_star(struct rx *, struct rexp_node *); -RX_DECL struct rexp_node -*rx_mk_r_2phase_star(struct rx *, struct rexp_node *, struct rexp_node *); -RX_DECL struct rexp_node -*rx_mk_r_side_effect(struct rx *, rx_side_effect); - -//RX_DECL struct rexp_node -// *rx_mk_r_data (struct rx *, void *); -RX_DECL void rx_free_rexp(struct rx *, struct rexp_node *); -RX_DECL struct rexp_node -*rx_copy_rexp(struct rx *, struct rexp_node *); -RX_DECL struct rx_nfa_state -*rx_nfa_state(struct rx *); -RX_DECL void rx_free_nfa_state(struct rx_nfa_state *); -RX_DECL struct rx_nfa_state -*rx_id_to_nfa_state(struct rx *, int); -RX_DECL struct rx_nfa_edge -*rx_nfa_edge(struct rx *, enum rx_nfa_etype, - - struct rx_nfa_state *, struct rx_nfa_state *); -RX_DECL void rx_free_nfa_edge(struct rx_nfa_edge *); -static struct rx_possible_future -*rx_possible_future(struct rx *, struct rx_se_list *); -static void rx_free_possible_future(struct rx_possible_future *); -RX_DECL void rx_free_nfa(struct rx *); -RX_DECL int rx_build_nfa(struct rx *, struct rexp_node *, - struct rx_nfa_state **, struct rx_nfa_state **); -RX_DECL void rx_name_nfa_states(struct rx *); -static int se_list_cmp(void *, void *); -static int se_list_equal(void *, void *); -static struct rx_se_list -*hash_cons_se_prog(struct rx *, struct rx_hash *, - - void *, struct rx_se_list *); -static struct rx_se_list -*hash_se_prog(struct rx *, struct rx_hash *, struct rx_se_list *); -static int nfa_set_cmp(void *, void *); -static int nfa_set_equal(void *, void *); -static struct rx_nfa_state_set -*nfa_set_cons(struct rx *, struct rx_hash *, - - struct rx_nfa_state *, struct rx_nfa_state_set *); -static struct rx_nfa_state_set -*nfa_set_enjoin(struct rx *, struct rx_hash *, - - struct rx_nfa_state *, struct rx_nfa_state_set *); -#endif - -#ifndef emacs - -#ifdef SYNTAX_TABLE -extern char *re_syntax_table; -#else /* not SYNTAX_TABLE */ - -#ifndef RX_WANT_RX_DEFS -RX_DECL char re_syntax_table[CHAR_SET_SIZE]; -#endif - -#ifdef __STDC__ -static void init_syntax_once(void) -#else -static void init_syntax_once() -#endif -{ - 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 */ -#endif /* not emacs */ - -/* Compile with `-DRX_DEBUG' and use the following flags. - * - * Debugging flags: - * rx_debug - print information as a regexp is compiled - * rx_debug_trace - print information as a regexp is executed - */ - -#ifdef RX_DEBUG - -int rx_debug_compile = 0; -int rx_debug_trace = 0; -static struct re_pattern_buffer *dbug_rxb = 0; - - -/* - * More Prototypes - */ -#ifdef __STDC__ -typedef void (*side_effect_printer) (struct rx *, void *, FILE *); -static void print_cset(struct rx *, rx_Bitset, FILE *); -static void print_rexp(struct rx *, struct rexp_node *, int, - side_effect_printer, FILE *); -static void print_nfa(struct rx *, struct rx_nfa_state *, - side_effect_printer, FILE *); -static void re_seprint(struct rx *, void *, FILE *); -void print_compiled_pattern(struct re_pattern_buffer *); -void print_fastmap(char *); -#else -typedef void (*side_effect_printer) (); -static void print_cset(); -#endif - -#ifdef __STDC__ -static void -print_rexp(struct rx *rx, - struct rexp_node *node, int depth, - side_effect_printer seprint, FILE * fp) -#else -static void print_rexp(rx, node, depth, seprint, fp) -struct rx *rx; -struct rexp_node *node; -int depth; -side_effect_printer seprint; -FILE *fp; -#endif -{ - if (!node) - return; - else { - switch (node->type) { - case r_cset: - { - fprintf(fp, "%*s", depth, ""); - print_cset(rx, node->params.cset, fp); - fputc('\n', fp); - break; - } - - case r_opt: - case r_star: - fprintf(fp, "%*s%s\n", depth, "", - node->type == r_opt ? "opt" : "star"); - print_rexp(rx, node->params.pair.left, depth + 3, seprint, fp); - break; - - case r_2phase_star: - fprintf(fp, "%*s2phase star\n", depth, ""); - print_rexp(rx, node->params.pair.right, depth + 3, seprint, - fp); - print_rexp(rx, node->params.pair.left, depth + 3, seprint, fp); - break; - - - case r_alternate: - case r_concat: - fprintf(fp, "%*s%s\n", depth, "", - node->type == r_alternate ? "alt" : "concat"); - print_rexp(rx, node->params.pair.left, depth + 3, seprint, fp); - print_rexp(rx, node->params.pair.right, depth + 3, seprint, - fp); - break; - case r_side_effect: - fprintf(fp, "%*sSide effect: ", depth, ""); - seprint(rx, node->params.side_effect, fp); - fputc('\n', fp); - } - } -} - -#ifdef __STDC__ -static void -print_nfa(struct rx *rx, - struct rx_nfa_state *n, side_effect_printer seprint, FILE * fp) -#else -static void print_nfa(rx, n, seprint, fp) -struct rx *rx; -struct rx_nfa_state *n; -side_effect_printer seprint; -FILE *fp; -#endif -{ - while (n) { - struct rx_nfa_edge *e = n->edges; - struct rx_possible_future *ec = n->futures; - - fprintf(fp, "node %d %s\n", n->id, - n->is_final ? "final" : (n->is_start ? "start" : "")); - while (e) { - fprintf(fp, " edge to %d, ", e->dest->id); - switch (e->type) { - case ne_epsilon: - fprintf(fp, "epsilon\n"); - break; - case ne_side_effect: - fprintf(fp, "side effect "); - seprint(rx, e->params.side_effect, fp); - fputc('\n', fp); - break; - case ne_cset: - fprintf(fp, "cset "); - print_cset(rx, e->params.cset, fp); - fputc('\n', fp); - break; - } - e = e->next; - } - - while (ec) { - int x; - struct rx_nfa_state_set *s; - struct rx_se_list *l; - - fprintf(fp, " eclosure to {"); - for (s = ec->destset; s; s = s->cdr) - fprintf(fp, "%d ", s->car->id); - fprintf(fp, "} ("); - for (l = ec->effects; l; l = l->cdr) { - seprint(rx, l->car, fp); - fputc(' ', fp); - } - fprintf(fp, ")\n"); - ec = ec->next; - } - n = n->next; - } -} - -static char *efnames[] = { - "bogon", - "re_se_try", - "re_se_pushback", - "re_se_push0", - "re_se_pushpos", - "re_se_chkpos", - "re_se_poppos", - "re_se_at_dot", - "re_se_syntax", - "re_se_not_syntax", - "re_se_begbuf", - "re_se_hat", - "re_se_wordbeg", - "re_se_wordbound", - "re_se_notwordbound", - "re_se_wordend", - "re_se_endbuf", - "re_se_dollar", - "re_se_fail", -}; - -static char *efnames2[] = { - "re_se_win", - "re_se_lparen", - "re_se_rparen", - "re_se_backref", - "re_se_iter", - "re_se_end_iter", - "re_se_tv" -}; - -static char *inx_names[] = { - "rx_backtrack_point", - "rx_do_side_effects", - "rx_cache_miss", - "rx_next_char", - "rx_backtrack", - "rx_error_inx", - "rx_num_instructions" -}; - - -#ifdef __STDC__ -static void re_seprint(struct rx *rx, void *effect, FILE * fp) -#else -static void re_seprint(rx, effect, fp) -struct rx *rx; -void *effect; -FILE *fp; -#endif -{ - if ((int) effect < 0) - fputs(efnames[-(int) effect], fp); - else if (dbug_rxb) { - struct re_se_params *p = &dbug_rxb->se_params[(int) effect]; - - fprintf(fp, "%s(%d,%d)", efnames2[p->se], p->op1, p->op2); - } else - fprintf(fp, "[complex op # %d]", (int) effect); -} - -/* These are so the regex.c regression tests will compile. */ -void print_compiled_pattern(rxb) -struct re_pattern_buffer *rxb; -{ -} - -void print_fastmap(fm) -char *fm; -{ -} - -#endif /* RX_DEBUG */ - - - -/* This page: Bitsets. Completely unintersting. */ - -//RX_DECL int rx_bitset_is_equal (int, rx_Bitset, rx_Bitset); -RX_DECL int rx_bitset_is_subset(int, rx_Bitset, rx_Bitset); - -//RX_DECL int rx_bitset_empty (int, rx_Bitset); -RX_DECL void rx_bitset_null(int, rx_Bitset); -RX_DECL void rx_bitset_complement(int, rx_Bitset); -RX_DECL void rx_bitset_complement(int, rx_Bitset); -RX_DECL void rx_bitset_assign(int, rx_Bitset, rx_Bitset); -RX_DECL void rx_bitset_union(int, rx_Bitset, rx_Bitset); -RX_DECL void rx_bitset_intersection(int, rx_Bitset, rx_Bitset); -RX_DECL void rx_bitset_difference(int, rx_Bitset, rx_Bitset); - -//RX_DECL void rx_bitset_revdifference (int, rx_Bitset, rx_Bitset); -#ifdef emacs -RX_DECL void rx_bitset_xor(int, rx_Bitset, rx_Bitset); -#endif -RX_DECL unsigned long rx_bitset_hash(int, rx_Bitset); - -#if 0 -#ifdef __STDC__ -RX_DECL int rx_bitset_is_equal(int size, rx_Bitset a, rx_Bitset b) -#else -RX_DECL int rx_bitset_is_equal(size, a, b) -int size; -rx_Bitset a; -rx_Bitset b; -#endif -{ - int x; - RX_subset s = b[0]; - - b[0] = ~a[0]; - - for (x = rx_bitset_numb_subsets(size) - 1; a[x] == b[x]; --x); - - b[0] = s; - return !x && s == a[0]; -} -#endif - -#ifdef __STDC__ -RX_DECL int rx_bitset_is_subset(int size, rx_Bitset a, rx_Bitset b) -#else -RX_DECL int rx_bitset_is_subset(size, a, b) -int size; -rx_Bitset a; -rx_Bitset b; -#endif -{ - int x = rx_bitset_numb_subsets(size) - 1; - - while (x-- && (a[x] & b[x]) == a[x]); - return x == -1; -} - -#if 0 -#ifdef __STDC__ -RX_DECL int rx_bitset_empty(int size, rx_Bitset set) -#else -RX_DECL int rx_bitset_empty(size, set) -int size; -rx_Bitset set; -#endif -{ - int x; - RX_subset s = set[0]; - - set[0] = 1; - for (x = rx_bitset_numb_subsets(size) - 1; !set[x]; --x); - set[0] = s; - return !s; -} -#endif - -#ifdef __STDC__ -RX_DECL void rx_bitset_null(int size, rx_Bitset b) -#else -RX_DECL void rx_bitset_null(size, b) -int size; -rx_Bitset b; -#endif -{ - bzero(b, rx_sizeof_bitset(size)); -} - - -#ifdef __STDC__ -RX_DECL void rx_bitset_universe(int size, rx_Bitset b) -#else -RX_DECL void rx_bitset_universe(size, b) -int size; -rx_Bitset b; -#endif -{ - int x = rx_bitset_numb_subsets(size); - - while (x--) - *b++ = ~(RX_subset) 0; -} - - -#ifdef __STDC__ -RX_DECL void rx_bitset_complement(int size, rx_Bitset b) -#else -RX_DECL void rx_bitset_complement(size, b) -int size; -rx_Bitset b; -#endif -{ - int x = rx_bitset_numb_subsets(size); - - while (x--) { - *b = ~*b; - ++b; - } -} - - -#ifdef __STDC__ -RX_DECL void rx_bitset_assign(int size, rx_Bitset a, rx_Bitset b) -#else -RX_DECL void rx_bitset_assign(size, a, b) -int size; -rx_Bitset a; -rx_Bitset b; -#endif -{ - int x; - - for (x = rx_bitset_numb_subsets(size) - 1; x >= 0; --x) - a[x] = b[x]; -} - -#ifdef __STDC__ -RX_DECL void rx_bitset_union(int size, rx_Bitset a, rx_Bitset b) -#else -RX_DECL void rx_bitset_union(size, a, b) -int size; -rx_Bitset a; -rx_Bitset b; -#endif -{ - int x; - - for (x = rx_bitset_numb_subsets(size) - 1; x >= 0; --x) - a[x] |= b[x]; -} - - -#ifdef __STDC__ -RX_DECL void rx_bitset_intersection(int size, rx_Bitset a, rx_Bitset b) -#else -RX_DECL void rx_bitset_intersection(size, a, b) -int size; -rx_Bitset a; -rx_Bitset b; -#endif -{ - int x; - - for (x = rx_bitset_numb_subsets(size) - 1; x >= 0; --x) - a[x] &= b[x]; -} - - -#ifdef __STDC__ -RX_DECL void rx_bitset_difference(int size, rx_Bitset a, rx_Bitset b) -#else -RX_DECL void rx_bitset_difference(size, a, b) -int size; -rx_Bitset a; -rx_Bitset b; -#endif -{ - int x; - - for (x = rx_bitset_numb_subsets(size) - 1; x >= 0; --x) - a[x] &= ~b[x]; -} - - -#if 0 -#ifdef __STDC__ -RX_DECL void rx_bitset_revdifference(int size, rx_Bitset a, rx_Bitset b) -#else -RX_DECL void rx_bitset_revdifference(size, a, b) -int size; -rx_Bitset a; -rx_Bitset b; -#endif -{ - int x; - - for (x = rx_bitset_numb_subsets(size) - 1; x >= 0; --x) - a[x] = ~a[x] & b[x]; -} -#endif - - -#ifdef emacs -#ifdef __STDC__ -RX_DECL void rx_bitset_xor(int size, rx_Bitset a, rx_Bitset b) -#else -RX_DECL void rx_bitset_xor(size, a, b) -int size; -rx_Bitset a; -rx_Bitset b; -#endif -{ - int x; - - for (x = rx_bitset_numb_subsets(size) - 1; x >= 0; --x) - a[x] ^= b[x]; -} -#endif - - -#ifdef __STDC__ -RX_DECL unsigned long rx_bitset_hash(int size, rx_Bitset b) -#else -RX_DECL unsigned long rx_bitset_hash(size, b) -int size; -rx_Bitset b; -#endif -{ - int x; - unsigned long hash = (unsigned long) rx_bitset_hash; - - for (x = rx_bitset_numb_subsets(size) - 1; x >= 0; --x) - hash ^= rx_bitset_subset_val(b, x); - - return hash; -} - -RX_DECL RX_subset rx_subset_singletons[RX_subset_bits] = { - 0x1, - 0x2, - 0x4, - 0x8, - 0x10, - 0x20, - 0x40, - 0x80, - 0x100, - 0x200, - 0x400, - 0x800, - 0x1000, - 0x2000, - 0x4000, - 0x8000, - 0x10000, - 0x20000, - 0x40000, - 0x80000, - 0x100000, - 0x200000, - 0x400000, - 0x800000, - 0x1000000, - 0x2000000, - 0x4000000, - 0x8000000, - 0x10000000, - 0x20000000, - 0x40000000, - 0x80000000 -}; - -#ifdef RX_DEBUG - -#ifdef __STDC__ -static void print_cset(struct rx *rx, rx_Bitset cset, FILE * fp) -#else -static void print_cset(rx, cset, fp) -struct rx *rx; -rx_Bitset cset; -FILE *fp; -#endif -{ - int x; - - fputc('[', fp); - for (x = 0; x < rx->local_cset_size; ++x) - if (RX_bitset_member(cset, x)) { - if (isprint(x)) - fputc(x, fp); - else - fprintf(fp, "\\0%o ", x); - } - fputc(']', fp); -} - -#endif /* RX_DEBUG */ - - - -static unsigned long rx_hash_masks[4] = { - 0x12488421, - 0x96699669, - 0xbe7dd7eb, - 0xffffffff -}; - - -/* Hash tables */ -#ifdef __STDC__ -RX_DECL struct rx_hash_item *rx_hash_find(struct rx_hash *table, - unsigned long hash, - void *value, - struct rx_hash_rules *rules) -#else -RX_DECL struct rx_hash_item *rx_hash_find(table, hash, value, rules) -struct rx_hash *table; -unsigned long hash; -void *value; -struct rx_hash_rules *rules; -#endif -{ - rx_hash_eq eq = rules->eq; - int maskc = 0; - long mask = rx_hash_masks[0]; - int bucket = (hash & mask) % 13; - - while (table->children[bucket]) { - table = table->children[bucket]; - ++maskc; - mask = rx_hash_masks[maskc]; - bucket = (hash & mask) % 13; - } - - { - struct rx_hash_item *it = table->buckets[bucket]; - - while (it) - if (eq(it->data, value)) - return it; - else - it = it->next_same_hash; - } - - return 0; -} - -#ifdef __STDC__ -RX_DECL struct rx_hash_item *rx_hash_store(struct rx_hash *table, - unsigned long hash, - void *value, - struct rx_hash_rules *rules) -#else -RX_DECL struct rx_hash_item *rx_hash_store(table, hash, value, rules) -struct rx_hash *table; -unsigned long hash; -void *value; -struct rx_hash_rules *rules; -#endif -{ - rx_hash_eq eq = rules->eq; - int maskc = 0; - long mask = rx_hash_masks[0]; - int bucket = (hash & mask) % 13; - int depth = 0; - - while (table->children[bucket]) { - table = table->children[bucket]; - ++maskc; - mask = rx_hash_masks[maskc]; - bucket = (hash & mask) % 13; - ++depth; - } - - { - struct rx_hash_item *it = table->buckets[bucket]; - - while (it) - if (eq(it->data, value)) - return it; - else - it = it->next_same_hash; - } - - { - if ((depth < 3) - && (table->bucket_size[bucket] >= 4)) { - struct rx_hash *newtab = ((struct rx_hash *) - rules->hash_alloc(rules)); - - if (!newtab) - goto add_to_bucket; - bzero(newtab, sizeof(*newtab)); - newtab->parent = table; - { - struct rx_hash_item *them = table->buckets[bucket]; - unsigned long newmask = rx_hash_masks[maskc + 1]; - - while (them) { - struct rx_hash_item *save = them->next_same_hash; - int new_buck = (them->hash & newmask) % 13; - - them->next_same_hash = newtab->buckets[new_buck]; - newtab->buckets[new_buck] = them; - them->table = newtab; - them = save; - ++newtab->bucket_size[new_buck]; - ++newtab->refs; - } - table->refs = - (table->refs - table->bucket_size[bucket] + 1); - table->bucket_size[bucket] = 0; - table->buckets[bucket] = 0; - table->children[bucket] = newtab; - table = newtab; - bucket = (hash & newmask) % 13; - } - } - } - add_to_bucket: - { - struct rx_hash_item *it = ((struct rx_hash_item *) - rules->hash_item_alloc(rules, value)); - - if (!it) - return 0; - it->hash = hash; - it->table = table; - /* DATA and BINDING are to be set in hash_item_alloc */ - it->next_same_hash = table->buckets[bucket]; - table->buckets[bucket] = it; - ++table->bucket_size[bucket]; - ++table->refs; - return it; - } -} - - -#ifdef __STDC__ -RX_DECL void -rx_hash_free(struct rx_hash_item *it, struct rx_hash_rules *rules) -#else -RX_DECL void rx_hash_free(it, rules) -struct rx_hash_item *it; -struct rx_hash_rules *rules; -#endif -{ - if (it) { - struct rx_hash *table = it->table; - unsigned long hash = it->hash; - int depth = (table->parent - ? (table->parent->parent - ? (table->parent->parent->parent ? 3 : 2) - : 1) - : 0); - int bucket = (hash & rx_hash_masks[depth]) % 13; - struct rx_hash_item **pos = &table->buckets[bucket]; - - while (*pos != it) - pos = &(*pos)->next_same_hash; - *pos = it->next_same_hash; - rules->free_hash_item(it, rules); - --table->bucket_size[bucket]; - --table->refs; - while (!table->refs && depth) { - struct rx_hash *save = table; - - table = table->parent; - --depth; - bucket = (hash & rx_hash_masks[depth]) % 13; - --table->refs; - table->children[bucket] = 0; - rules->free_hash(save, rules); - } - } -} - -#ifdef __STDC__ -RX_DECL void -rx_free_hash_table(struct rx_hash *tab, rx_hash_freefn freefn, - struct rx_hash_rules *rules) -#else -RX_DECL void rx_free_hash_table(tab, freefn, rules) -struct rx_hash *tab; -rx_hash_freefn freefn; -struct rx_hash_rules *rules; -#endif -{ - int x; - - for (x = 0; x < 13; ++x) - if (tab->children[x]) { - rx_free_hash_table(tab->children[x], freefn, rules); - rules->free_hash(tab->children[x], rules); - } else { - struct rx_hash_item *them = tab->buckets[x]; - - while (them) { - struct rx_hash_item *that = them; - - them = that->next_same_hash; - freefn(that); - rules->free_hash_item(that, rules); - } - } -} - - - -/* Utilities for manipulating bitset represntations of characters sets. */ - -#ifdef __STDC__ -RX_DECL rx_Bitset rx_cset(struct rx *rx) -#else -RX_DECL rx_Bitset rx_cset(rx) -struct rx *rx; -#endif -{ - rx_Bitset b = - - (rx_Bitset) malloc(rx_sizeof_bitset(rx->local_cset_size)); - if (b) - rx_bitset_null(rx->local_cset_size, b); - return b; -} - - -#ifdef __STDC__ -RX_DECL rx_Bitset rx_copy_cset(struct rx * rx, rx_Bitset a) -#else -RX_DECL rx_Bitset rx_copy_cset(rx, a) -struct rx *rx; -rx_Bitset a; -#endif -{ - rx_Bitset cs = rx_cset(rx); - - if (cs) - rx_bitset_union(rx->local_cset_size, cs, a); - - return cs; -} - - -#ifdef __STDC__ -RX_DECL void rx_free_cset(struct rx *rx, rx_Bitset c) -#else -RX_DECL void rx_free_cset(rx, c) -struct rx *rx; -rx_Bitset c; -#endif -{ - if (c) - free((char *) c); -} - - -/* Hash table memory allocation policy for the regexp compiler */ - -#ifdef __STDC__ -static struct rx_hash *compiler_hash_alloc(struct rx_hash_rules *rules) -#else -static struct rx_hash *compiler_hash_alloc(rules) -struct rx_hash_rules *rules; -#endif -{ - return (struct rx_hash *) malloc(sizeof(struct rx_hash)); -} - - -#ifdef __STDC__ -static struct rx_hash_item *compiler_hash_item_alloc(struct rx_hash_rules - *rules, void *value) -#else -static struct rx_hash_item *compiler_hash_item_alloc(rules, value) -struct rx_hash_rules *rules; -void *value; -#endif -{ - struct rx_hash_item *it; - - it = (struct rx_hash_item *) malloc(sizeof(*it)); - if (it) { - it->data = value; - it->binding = 0; - } - return it; -} - -#ifdef __STDC__ -static void -compiler_free_hash(struct rx_hash *tab, struct rx_hash_rules *rules) -#else -static void compiler_free_hash(tab, rules) -struct rx_hash *tab; -struct rx_hash_rules *rules; -#endif -{ - free((char *) tab); -} - - -#ifdef __STDC__ -static void -compiler_free_hash_item(struct rx_hash_item *item, - struct rx_hash_rules *rules) -#else -static void compiler_free_hash_item(item, rules) -struct rx_hash_item *item; -struct rx_hash_rules *rules; -#endif -{ - free((char *) item); -} - - -/* This page: REXP_NODE (expression tree) structures. */ - -#ifdef __STDC__ -RX_DECL struct rexp_node *rexp_node(struct rx *rx, - enum rexp_node_type type) -#else -RX_DECL struct rexp_node *rexp_node(rx, type) -struct rx *rx; -enum rexp_node_type type; -#endif -{ - struct rexp_node *n; - - n = (struct rexp_node *) malloc(sizeof(*n)); - if (n) { - bzero(n, sizeof(*n)); - n->type = type; - } - return n; -} - - -/* free_rexp_node assumes that the bitset passed to rx_mk_r_cset - * can be freed using rx_free_cset. - */ -#ifdef __STDC__ -RX_DECL struct rexp_node *rx_mk_r_cset(struct rx *rx, rx_Bitset b) -#else -RX_DECL struct rexp_node *rx_mk_r_cset(rx, b) -struct rx *rx; -rx_Bitset b; -#endif -{ - struct rexp_node *n = rexp_node(rx, r_cset); - - if (n) - n->params.cset = b; - return n; -} - -#ifdef __STDC__ -RX_DECL struct rexp_node *rx_mk_r_concat(struct rx *rx, - struct rexp_node *a, - struct rexp_node *b) -#else -RX_DECL struct rexp_node *rx_mk_r_concat(rx, a, b) -struct rx *rx; -struct rexp_node *a; -struct rexp_node *b; -#endif -{ - struct rexp_node *n = rexp_node(rx, r_concat); - - if (n) { - n->params.pair.left = a; - n->params.pair.right = b; - } - return n; -} - - -#ifdef __STDC__ -RX_DECL struct rexp_node *rx_mk_r_alternate(struct rx *rx, - struct rexp_node *a, - struct rexp_node *b) -#else -RX_DECL struct rexp_node *rx_mk_r_alternate(rx, a, b) -struct rx *rx; -struct rexp_node *a; -struct rexp_node *b; -#endif -{ - struct rexp_node *n = rexp_node(rx, r_alternate); - - if (n) { - n->params.pair.left = a; - n->params.pair.right = b; - } - return n; -} - - -#ifdef __STDC__ -RX_DECL struct rexp_node *rx_mk_r_opt(struct rx *rx, struct rexp_node *a) -#else -RX_DECL struct rexp_node *rx_mk_r_opt(rx, a) -struct rx *rx; -struct rexp_node *a; -#endif -{ - struct rexp_node *n = rexp_node(rx, r_opt); - - if (n) { - n->params.pair.left = a; - n->params.pair.right = 0; - } - return n; -} - -#ifdef __STDC__ -RX_DECL struct rexp_node *rx_mk_r_star(struct rx *rx, struct rexp_node *a) -#else -RX_DECL struct rexp_node *rx_mk_r_star(rx, a) -struct rx *rx; -struct rexp_node *a; -#endif -{ - struct rexp_node *n = rexp_node(rx, r_star); - - if (n) { - n->params.pair.left = a; - n->params.pair.right = 0; - } - return n; -} - - -#ifdef __STDC__ -RX_DECL struct rexp_node *rx_mk_r_2phase_star(struct rx *rx, - struct rexp_node *a, - struct rexp_node *b) -#else -RX_DECL struct rexp_node *rx_mk_r_2phase_star(rx, a, b) -struct rx *rx; -struct rexp_node *a; -struct rexp_node *b; -#endif -{ - struct rexp_node *n = rexp_node(rx, r_2phase_star); - - if (n) { - n->params.pair.left = a; - n->params.pair.right = b; - } - return n; -} - -#ifdef __STDC__ -RX_DECL struct rexp_node *rx_mk_r_side_effect(struct rx *rx, - rx_side_effect a) -#else -RX_DECL struct rexp_node *rx_mk_r_side_effect(rx, a) -struct rx *rx; -rx_side_effect a; -#endif -{ - struct rexp_node *n = rexp_node(rx, r_side_effect); - - if (n) { - n->params.side_effect = a; - n->params.pair.right = 0; - } - return n; -} - - -#if 0 -#ifdef __STDC__ -RX_DECL struct rexp_node *rx_mk_r_data(struct rx *rx, void *a) -#else -RX_DECL struct rexp_node *rx_mk_r_data(rx, a) -struct rx *rx; -void *a; -#endif -{ - struct rexp_node *n = rexp_node(rx, r_data); - - if (n) { - n->params.pair.left = a; - n->params.pair.right = 0; - } - return n; -} -#endif - -#ifdef __STDC__ -RX_DECL void rx_free_rexp(struct rx *rx, struct rexp_node *node) -#else -RX_DECL void rx_free_rexp(rx, node) -struct rx *rx; -struct rexp_node *node; -#endif -{ - if (node) { - switch (node->type) { - case r_cset: - if (node->params.cset) - rx_free_cset(rx, node->params.cset); - - case r_side_effect: - break; - - case r_concat: - case r_alternate: - case r_2phase_star: - case r_opt: - case r_star: - rx_free_rexp(rx, node->params.pair.left); - rx_free_rexp(rx, node->params.pair.right); - break; - - case r_data: - /* This shouldn't occur. */ - break; - } - free((char *) node); - } -} - -#ifdef __STDC__ -RX_DECL struct rexp_node *rx_copy_rexp(struct rx *rx, - struct rexp_node *node) -#else -RX_DECL struct rexp_node *rx_copy_rexp(rx, node) -struct rx *rx; -struct rexp_node *node; -#endif -{ - if (!node) - return 0; - else { - struct rexp_node *n = rexp_node(rx, node->type); - - if (!n) - return 0; - switch (node->type) { - case r_cset: - n->params.cset = rx_copy_cset(rx, node->params.cset); - if (!n->params.cset) { - rx_free_rexp(rx, n); - return 0; - } - break; - - case r_side_effect: - n->params.side_effect = node->params.side_effect; - break; - - case r_concat: - case r_alternate: - case r_opt: - case r_2phase_star: - case r_star: - n->params.pair.left = rx_copy_rexp(rx, node->params.pair.left); - n->params.pair.right = - rx_copy_rexp(rx, node->params.pair.right); - if ((node->params.pair.left && !n->params.pair.left) - || (node->params.pair.right && !n->params.pair.right)) { - rx_free_rexp(rx, n); - return 0; - } - break; - case r_data: - /* shouldn't happen */ - break; - } - return n; - } -} - - - -/* This page: functions to build and destroy graphs that describe nfa's */ - -/* Constructs a new nfa node. */ -#ifdef __STDC__ -RX_DECL struct rx_nfa_state *rx_nfa_state(struct rx *rx) -#else -RX_DECL struct rx_nfa_state *rx_nfa_state(rx) -struct rx *rx; -#endif -{ - struct rx_nfa_state *n = (struct rx_nfa_state *) malloc(sizeof(*n)); - - if (!n) - return 0; - bzero(n, sizeof(*n)); - n->next = rx->nfa_states; - rx->nfa_states = n; - return n; -} - - -#ifdef __STDC__ -RX_DECL void rx_free_nfa_state(struct rx_nfa_state *n) -#else -RX_DECL void rx_free_nfa_state(n) -struct rx_nfa_state *n; -#endif -{ - free((char *) n); -} - - -/* This looks up an nfa node, given a numeric id. Numeric id's are - * assigned after the nfa has been built. - */ -#ifdef __STDC__ -RX_DECL struct rx_nfa_state *rx_id_to_nfa_state(struct rx *rx, int id) -#else -RX_DECL struct rx_nfa_state *rx_id_to_nfa_state(rx, id) -struct rx *rx; -int id; -#endif -{ - struct rx_nfa_state *n; - - for (n = rx->nfa_states; n; n = n->next) - if (n->id == id) - return n; - return 0; -} - - -/* This adds an edge between two nodes, but doesn't initialize the - * edge label. - */ - -#ifdef __STDC__ -RX_DECL struct rx_nfa_edge *rx_nfa_edge(struct rx *rx, - enum rx_nfa_etype type, - struct rx_nfa_state *start, - struct rx_nfa_state *dest) -#else -RX_DECL struct rx_nfa_edge *rx_nfa_edge(rx, type, start, dest) -struct rx *rx; -enum rx_nfa_etype type; -struct rx_nfa_state *start; -struct rx_nfa_state *dest; -#endif -{ - struct rx_nfa_edge *e; - - e = (struct rx_nfa_edge *) malloc(sizeof(*e)); - if (!e) - return 0; - e->next = start->edges; - start->edges = e; - e->type = type; - e->dest = dest; - return e; -} - - -#ifdef __STDC__ -RX_DECL void rx_free_nfa_edge(struct rx_nfa_edge *e) -#else -RX_DECL void rx_free_nfa_edge(e) -struct rx_nfa_edge *e; -#endif -{ - free((char *) e); -} - - -/* This constructs a POSSIBLE_FUTURE, which is a kind epsilon-closure - * of an NFA. These are added to an nfa automaticly by eclose_nfa. - */ - -#ifdef __STDC__ -static struct rx_possible_future *rx_possible_future(struct rx *rx, struct rx_se_list - *effects) -#else -static struct rx_possible_future *rx_possible_future(rx, effects) -struct rx *rx; -struct rx_se_list *effects; -#endif -{ - struct rx_possible_future *ec; - - ec = (struct rx_possible_future *) malloc(sizeof(*ec)); - if (!ec) - return 0; - ec->destset = 0; - ec->next = 0; - ec->effects = effects; - return ec; -} - - -#ifdef __STDC__ -static void rx_free_possible_future(struct rx_possible_future *pf) -#else -static void rx_free_possible_future(pf) -struct rx_possible_future *pf; -#endif -{ - free((char *) pf); -} - - -#ifdef __STDC__ -RX_DECL void rx_free_nfa(struct rx *rx) -#else -RX_DECL void rx_free_nfa(rx) -struct rx *rx; -#endif -{ - while (rx->nfa_states) { - while (rx->nfa_states->edges) { - switch (rx->nfa_states->edges->type) { - case ne_cset: - rx_free_cset(rx, rx->nfa_states->edges->params.cset); - break; - default: - break; - } - { - struct rx_nfa_edge *e; - - e = rx->nfa_states->edges; - rx->nfa_states->edges = rx->nfa_states->edges->next; - rx_free_nfa_edge(e); - } - } /* while (rx->nfa_states->edges) */ - { - /* Iterate over the partial epsilon closures of rx->nfa_states */ - struct rx_possible_future *pf = rx->nfa_states->futures; - - while (pf) { - struct rx_possible_future *pft = pf; - - pf = pf->next; - rx_free_possible_future(pft); - } - } - { - struct rx_nfa_state *n; - - n = rx->nfa_states; - rx->nfa_states = rx->nfa_states->next; - rx_free_nfa_state(n); - } - } -} - - - -/* This page: translating a pattern expression into an nfa and doing the - * static part of the nfa->super-nfa translation. - */ - -/* This is the thompson regexp->nfa algorithm. - * It is modified to allow for `side-effect epsilons.' Those are - * edges that are taken whenever a similar epsilon edge would be, - * but which imply that some side effect occurs when the edge - * is taken. - * - * Side effects are used to model parts of the pattern langauge - * that are not regular (in the formal sense). - */ - -#ifdef __STDC__ -RX_DECL int -rx_build_nfa(struct rx *rx, - struct rexp_node *rexp, - struct rx_nfa_state **start, struct rx_nfa_state **end) -#else -RX_DECL int rx_build_nfa(rx, rexp, start, end) -struct rx *rx; -struct rexp_node *rexp; -struct rx_nfa_state **start; -struct rx_nfa_state **end; -#endif -{ - struct rx_nfa_edge *edge; - - /* Start & end nodes may have been allocated by the caller. */ - *start = *start ? *start : rx_nfa_state(rx); - - if (!*start) - return 0; - - if (!rexp) { - *end = *start; - return 1; - } - - *end = *end ? *end : rx_nfa_state(rx); - - if (!*end) { - rx_free_nfa_state(*start); - return 0; - } - - switch (rexp->type) { - case r_data: - return 0; - - case r_cset: - edge = rx_nfa_edge(rx, ne_cset, *start, *end); - if (!edge) - return 0; - edge->params.cset = rx_copy_cset(rx, rexp->params.cset); - if (!edge->params.cset) { - rx_free_nfa_edge(edge); - return 0; - } - return 1; - - case r_opt: - return (rx_build_nfa(rx, rexp->params.pair.left, start, end) - && rx_nfa_edge(rx, ne_epsilon, *start, *end)); - - case r_star: - { - struct rx_nfa_state *star_start = 0; - struct rx_nfa_state *star_end = 0; - - return (rx_build_nfa(rx, rexp->params.pair.left, - &star_start, &star_end) - && star_start - && star_end - && rx_nfa_edge(rx, ne_epsilon, star_start, star_end) - && rx_nfa_edge(rx, ne_epsilon, *start, star_start) - && rx_nfa_edge(rx, ne_epsilon, star_end, *end) - - && rx_nfa_edge(rx, ne_epsilon, star_end, star_start)); - } - - case r_2phase_star: - { - struct rx_nfa_state *star_start = 0; - struct rx_nfa_state *star_end = 0; - struct rx_nfa_state *loop_exp_start = 0; - struct rx_nfa_state *loop_exp_end = 0; - - return (rx_build_nfa(rx, rexp->params.pair.left, - &star_start, &star_end) - && rx_build_nfa(rx, rexp->params.pair.right, - &loop_exp_start, &loop_exp_end) - && star_start - && star_end - && loop_exp_end - && loop_exp_start - && rx_nfa_edge(rx, ne_epsilon, star_start, *end) - && rx_nfa_edge(rx, ne_epsilon, *start, star_start) - && rx_nfa_edge(rx, ne_epsilon, star_end, *end) - - && rx_nfa_edge(rx, ne_epsilon, star_end, loop_exp_start) - && rx_nfa_edge(rx, ne_epsilon, loop_exp_end, star_start)); - } - - - case r_concat: - { - struct rx_nfa_state *shared = 0; - - return (rx_build_nfa(rx, rexp->params.pair.left, start, &shared) - && rx_build_nfa(rx, rexp->params.pair.right, &shared, - end)); - } - - case r_alternate: - { - struct rx_nfa_state *ls = 0; - struct rx_nfa_state *le = 0; - struct rx_nfa_state *rs = 0; - struct rx_nfa_state *re = 0; - - return (rx_build_nfa(rx, rexp->params.pair.left, &ls, &le) - && rx_build_nfa(rx, rexp->params.pair.right, &rs, &re) - && rx_nfa_edge(rx, ne_epsilon, *start, ls) - && rx_nfa_edge(rx, ne_epsilon, *start, rs) - && rx_nfa_edge(rx, ne_epsilon, le, *end) - && rx_nfa_edge(rx, ne_epsilon, re, *end)); - } - - case r_side_effect: - edge = rx_nfa_edge(rx, ne_side_effect, *start, *end); - if (!edge) - return 0; - edge->params.side_effect = rexp->params.side_effect; - return 1; - } - - /* this should never happen */ - return 0; -} - - -/* RX_NAME_NFA_STATES identifies all nodes with outgoing non-epsilon - * transitions. Only these nodes can occur in super-states. - * All nodes are given an integer id. - * The id is non-negative if the node has non-epsilon out-transitions, negative - * otherwise (this is because we want the non-negative ids to be used as - * array indexes in a few places). - */ - -#ifdef __STDC__ -RX_DECL void rx_name_nfa_states(struct rx *rx) -#else -RX_DECL void rx_name_nfa_states(rx) -struct rx *rx; -#endif -{ - struct rx_nfa_state *n = rx->nfa_states; - - rx->nodec = 0; - rx->epsnodec = -1; - - while (n) { - struct rx_nfa_edge *e = n->edges; - - if (n->is_start) - n->eclosure_needed = 1; - - while (e) { - switch (e->type) { - case ne_epsilon: - case ne_side_effect: - break; - - case ne_cset: - n->id = rx->nodec++; - { - struct rx_nfa_edge *from_n = n->edges; - - while (from_n) { - from_n->dest->eclosure_needed = 1; - from_n = from_n->next; - } - } - goto cont; - } - e = e->next; - } - n->id = rx->epsnodec--; - cont: - n = n->next; - } - rx->epsnodec = -rx->epsnodec; -} - - -/* This page: data structures for the static part of the nfa->supernfa - * translation. - * - * There are side effect lists -- lists of side effects occuring - * along an uninterrupted, acyclic path of side-effect epsilon edges. - * Such paths are collapsed to single edges in the course of computing - * epsilon closures. Such single edges are labled with a list of all - * the side effects entailed in crossing them. Like lists of side - * effects are made == by the constructors below. - * - * There are also nfa state sets. These are used to hold a list of all - * states reachable from a starting state for a given type of transition - * and side effect list. These are also hash-consed. - */ - -/* The next several functions compare, construct, etc. lists of side - * effects. See ECLOSE_NFA (below) for details. - */ - -/* Ordering of rx_se_list - * (-1, 0, 1 return value convention). - */ - -#ifdef __STDC__ -static int se_list_cmp(void *va, void *vb) -#else -static int se_list_cmp(va, vb) -void *va; -void *vb; -#endif -{ - struct rx_se_list *a = (struct rx_se_list *) va; - struct rx_se_list *b = (struct rx_se_list *) vb; - - return ((va == vb) - ? 0 - : (!va - ? -1 - : (!vb - ? 1 - : ((long) a->car < (long) b->car - ? 1 - : ((long) a->car > (long) b->car - ? -1 - : se_list_cmp((void *) a->cdr, - (void *) b->cdr)))))); -} - - -#ifdef __STDC__ -static int se_list_equal(void *va, void *vb) -#else -static int se_list_equal(va, vb) -void *va; -void *vb; -#endif -{ - return !(se_list_cmp(va, vb)); -} - -static struct rx_hash_rules se_list_hash_rules = { - se_list_equal, - compiler_hash_alloc, - compiler_free_hash, - compiler_hash_item_alloc, - compiler_free_hash_item -}; - - -#ifdef __STDC__ -static struct rx_se_list *side_effect_cons(struct rx *rx, - void *se, - struct rx_se_list *list) -#else -static struct rx_se_list *side_effect_cons(rx, se, list) -struct rx *rx; -void *se; -struct rx_se_list *list; -#endif -{ - struct rx_se_list *l; - - l = ((struct rx_se_list *) malloc(sizeof(*l))); - if (!l) - return 0; - l->car = se; - l->cdr = list; - return l; -} - - -#ifdef __STDC__ -static struct rx_se_list *hash_cons_se_prog(struct rx *rx, - struct rx_hash *memo, - void *car, - struct rx_se_list *cdr) -#else -static struct rx_se_list *hash_cons_se_prog(rx, memo, car, cdr) -struct rx *rx; -struct rx_hash *memo; -void *car; -struct rx_se_list *cdr; -#endif -{ - long hash = (long) car ^ (long) cdr; - struct rx_se_list template; - - template.car = car; - template.cdr = cdr; - { - struct rx_hash_item *it = rx_hash_store(memo, hash, - (void *) &template, - &se_list_hash_rules); - - if (!it) - return 0; - if (it->data == (void *) &template) { - struct rx_se_list *consed; - - consed = (struct rx_se_list *) malloc(sizeof(*consed)); - if (!consed) { - free((char *) it); - return 0; - } - *consed = template; - it->data = (void *) consed; - } - return (struct rx_se_list *) it->data; - } -} - - -#ifdef __STDC__ -static struct rx_se_list *hash_se_prog(struct rx *rx, struct rx_hash *memo, - struct rx_se_list *prog) -#else -static struct rx_se_list *hash_se_prog(rx, memo, prog) -struct rx *rx; -struct rx_hash *memo; -struct rx_se_list *prog; -#endif -{ - struct rx_se_list *answer = 0; - - while (prog) { - answer = hash_cons_se_prog(rx, memo, prog->car, answer); - if (!answer) - return 0; - prog = prog->cdr; - } - return answer; -} - -#ifdef __STDC__ -static int nfa_set_cmp(void *va, void *vb) -#else -static int nfa_set_cmp(va, vb) -void *va; -void *vb; -#endif -{ - struct rx_nfa_state_set *a = (struct rx_nfa_state_set *) va; - struct rx_nfa_state_set *b = (struct rx_nfa_state_set *) vb; - - return ((va == vb) - ? 0 - : (!va - ? -1 - : (!vb - ? 1 - : (a->car->id < b->car->id - ? 1 - : (a->car->id > b->car->id - ? -1 - : nfa_set_cmp((void *) a->cdr, - (void *) b->cdr)))))); -} - -#ifdef __STDC__ -static int nfa_set_equal(void *va, void *vb) -#else -static int nfa_set_equal(va, vb) -void *va; -void *vb; -#endif -{ - return !nfa_set_cmp(va, vb); -} - -static struct rx_hash_rules nfa_set_hash_rules = { - nfa_set_equal, - compiler_hash_alloc, - compiler_free_hash, - compiler_hash_item_alloc, - compiler_free_hash_item -}; - - -#ifdef __STDC__ -static struct rx_nfa_state_set *nfa_set_cons(struct rx *rx, - struct rx_hash *memo, - struct rx_nfa_state *state, - struct rx_nfa_state_set *set) -#else -static struct rx_nfa_state_set *nfa_set_cons(rx, memo, state, set) -struct rx *rx; -struct rx_hash *memo; -struct rx_nfa_state *state; -struct rx_nfa_state_set *set; -#endif -{ - struct rx_nfa_state_set template; - struct rx_hash_item *node; - - template.car = state; - template.cdr = set; - node = rx_hash_store(memo, - (((long) state) >> 8) ^ (long) set, - &template, &nfa_set_hash_rules); - if (!node) - return 0; - if (node->data == &template) { - struct rx_nfa_state_set *l; - - l = (struct rx_nfa_state_set *) malloc(sizeof(*l)); - node->data = (void *) l; - if (!l) - return 0; - *l = template; - } - return (struct rx_nfa_state_set *) node->data; -} - -#ifdef __STDC__ -static struct rx_nfa_state_set *nfa_set_enjoin(struct rx *rx, - struct rx_hash *memo, - struct rx_nfa_state *state, - struct rx_nfa_state_set - *set) -#else -static struct rx_nfa_state_set *nfa_set_enjoin(rx, memo, state, set) -struct rx *rx; -struct rx_hash *memo; -struct rx_nfa_state *state; -struct rx_nfa_state_set *set; -#endif -{ - if (!set || state->id < set->car->id) - return nfa_set_cons(rx, memo, state, set); - if (state->id == set->car->id) - return set; - else { - struct rx_nfa_state_set *newcdr - - = nfa_set_enjoin(rx, memo, state, set->cdr); - if (newcdr != set->cdr) - set = nfa_set_cons(rx, memo, set->car, newcdr); - return set; - } -} - - - -/* This page: computing epsilon closures. The closures aren't total. - * Each node's closures are partitioned according to the side effects entailed - * along the epsilon edges. Return true on success. - */ - -struct eclose_frame { - struct rx_se_list *prog_backwards; -}; -static int eclose_node(struct rx *, struct rx_nfa_state *, - struct rx_nfa_state *, struct eclose_frame *); -RX_DECL int rx_eclose_nfa(struct rx *); -RX_DECL void rx_delete_epsilon_transitions(struct rx *); -static int nfacmp(void *, void *); -static int count_hash_nodes(struct rx_hash *); -static void nfa_set_freer(struct rx_hash_item *); -RX_DECL int rx_compactify_nfa(struct rx *, void **, unsigned long *); -static char *rx_cache_malloc(struct rx_cache *, int); -static void rx_cache_free(struct rx_cache *, - - struct rx_freelist **, char *); -static void install_transition(struct rx_superstate *, - - struct rx_inx *, rx_Bitset); -static int qlen(struct rx_superstate *); -static void check_cache(struct rx_cache *); -static void semifree_superstate(struct rx_cache *); -static void refresh_semifree_superstate - - (struct rx_cache *, struct rx_superstate *); -static void rx_refresh_this_superstate - - (struct rx_cache *, struct rx_superstate *); -static void release_superset_low(struct rx_cache *, struct rx_superset *); -RX_DECL void rx_release_superset(struct rx *, struct rx_superset *); -static int rx_really_free_superstate(struct rx_cache *); -static char *rx_cache_get(struct rx_cache *, struct rx_freelist **); -static char *rx_cache_malloc_or_get(struct rx_cache *, - struct rx_freelist **, int); -static char *rx_cache_get_superstate(struct rx_cache *); -static int supersetcmp(void *, void *); -static struct rx_hash_item -*superset_allocator(struct rx_hash_rules *, void *); -static struct rx_hash -*super_hash_allocator(struct rx_hash_rules *); -static void super_hash_liberator(struct rx_hash *, struct rx_hash_rules *); -static void superset_hash_item_liberator - - (struct rx_hash_item *, struct rx_hash_rules *); -static int bytes_for_cache_size(int, int); -static void rx_morecore(struct rx_cache *); -RX_DECL struct rx_superset -*rx_superset_cons(struct rx *, struct rx_nfa_state *, - - struct rx_superset *); -RX_DECL struct rx_superset -*rx_superstate_eclosure_union - - (struct rx *, struct rx_superset *, struct rx_nfa_state_set *); -static struct rx_distinct_future -*include_futures(struct rx *, - struct rx_distinct_future *, - - struct rx_nfa_state *, struct rx_superstate *); -RX_DECL struct rx_superstate -*rx_superstate(struct rx *, struct rx_superset *); -static int solve_destination(struct rx *, struct rx_distinct_future *); -static int compute_super_edge(struct rx *, - struct rx_distinct_future **, - - rx_Bitset, struct rx_superstate *, - unsigned char); -static struct rx_super_edge -*rx_super_edge(struct rx *, struct rx_superstate *, - - rx_Bitset, struct rx_distinct_future *); -static void install_partial_transition - (struct rx_superstate *, struct rx_inx *, RX_subset, int); -RX_DECL struct rx_inx -*rx_handle_cache_miss(struct rx *, struct rx_superstate *, - - unsigned char, void *); -static boolean - -at_begline_loc_p(__const__ char *, __const__ char *, reg_syntax_t); -static boolean at_endline_loc_p(__const__ char *, __const__ char *, int); -static rx_Bitset -inverse_translation(struct re_pattern_buffer *, char *, - rx_Bitset, unsigned char *, int); - - -#ifdef __STDC__ -static int -eclose_node(struct rx *rx, struct rx_nfa_state *outnode, - struct rx_nfa_state *node, struct eclose_frame *frame) -#else -static int eclose_node(rx, outnode, node, frame) -struct rx *rx; -struct rx_nfa_state *outnode; -struct rx_nfa_state *node; -struct eclose_frame *frame; -#endif -{ - struct rx_nfa_edge *e = node->edges; - - /* For each node, we follow all epsilon paths to build the closure. - * The closure omits nodes that have only epsilon edges. - * The closure is split into partial closures -- all the states in - * a partial closure are reached by crossing the same list of - * of side effects (though not necessarily the same path). - */ - if (node->mark) - return 1; - node->mark = 1; - - if (node->id >= 0 || node->is_final) { - struct rx_possible_future **ec; - struct rx_se_list *prog_in_order - = ((struct rx_se_list *) hash_se_prog(rx, - &rx->se_list_memo, - frame->prog_backwards)); - int cmp; - - ec = &outnode->futures; - - while (*ec) { - cmp = - se_list_cmp((void *) (*ec)->effects, - (void *) prog_in_order); - if (cmp <= 0) - break; - ec = &(*ec)->next; - } - if (!*ec || (cmp < 0)) { - struct rx_possible_future *saved = *ec; - - *ec = rx_possible_future(rx, prog_in_order); - (*ec)->next = saved; - if (!*ec) - return 0; - } - if (node->id >= 0) { - (*ec)->destset = nfa_set_enjoin(rx, &rx->set_list_memo, - node, (*ec)->destset); - if (!(*ec)->destset) - return 0; - } - } - - while (e) { - switch (e->type) { - case ne_epsilon: - if (!eclose_node(rx, outnode, e->dest, frame)) - return 0; - break; - case ne_side_effect: - { - frame->prog_backwards = side_effect_cons(rx, - e->params.side_effect, - frame->prog_backwards); - if (!frame->prog_backwards) - return 0; - if (!eclose_node(rx, outnode, e->dest, frame)) - return 0; - { - struct rx_se_list *dying = frame->prog_backwards; - - frame->prog_backwards = frame->prog_backwards->cdr; - free((char *) dying); - } - break; - } - default: - break; - } - e = e->next; - } - node->mark = 0; - return 1; -} - -#ifdef __STDC__ -RX_DECL int rx_eclose_nfa(struct rx *rx) -#else -RX_DECL int rx_eclose_nfa(rx) -struct rx *rx; -#endif -{ - struct rx_nfa_state *n = rx->nfa_states; - struct eclose_frame frame; - static int rx_id = 0; - - frame.prog_backwards = 0; - rx->rx_id = rx_id++; - bzero(&rx->se_list_memo, sizeof(rx->se_list_memo)); - bzero(&rx->set_list_memo, sizeof(rx->set_list_memo)); - while (n) { - n->futures = 0; - if (n->eclosure_needed && !eclose_node(rx, n, n, &frame)) - return 0; - /* clear_marks (rx); */ - n = n->next; - } - return 1; -} - - -/* This deletes epsilon edges from an NFA. After running eclose_node, - * we have no more need for these edges. They are removed to simplify - * further operations on the NFA. - */ - -#ifdef __STDC__ -RX_DECL void rx_delete_epsilon_transitions(struct rx *rx) -#else -RX_DECL void rx_delete_epsilon_transitions(rx) -struct rx *rx; -#endif -{ - struct rx_nfa_state *n = rx->nfa_states; - struct rx_nfa_edge **e; - - while (n) { - e = &n->edges; - while (*e) { - struct rx_nfa_edge *t; - - switch ((*e)->type) { - case ne_epsilon: - case ne_side_effect: - t = *e; - *e = t->next; - rx_free_nfa_edge(t); - break; - - default: - e = &(*e)->next; - break; - } - } - n = n->next; - } -} - - -/* This page: storing the nfa in a contiguous region of memory for - * subsequent conversion to a super-nfa. - */ - -/* This is for qsort on an array of nfa_states. The order - * is based on state ids and goes - * [0...MAX][MIN..-1] where (MAX>=0) and (MIN<0) - * This way, positive ids double as array indices. - */ - -#ifdef __STDC__ -static int nfacmp(void *va, void *vb) -#else -static int nfacmp(va, vb) -void *va; -void *vb; -#endif -{ - struct rx_nfa_state **a = (struct rx_nfa_state **) va; - struct rx_nfa_state **b = (struct rx_nfa_state **) vb; - - return (*a == *b /* &&&& 3.18 */ - ? 0 : (((*a)->id < 0) == ((*b)->id < 0) - ? (((*a)->id < (*b)->id) ? -1 : 1) - : (((*a)->id < 0) - ? 1 : -1))); -} - -#ifdef __STDC__ -static int count_hash_nodes(struct rx_hash *st) -#else -static int count_hash_nodes(st) -struct rx_hash *st; -#endif -{ - int x; - int count = 0; - - for (x = 0; x < 13; ++x) - count += ((st->children[x]) - ? count_hash_nodes(st->children[x]) - : st->bucket_size[x]); - - return count; -} - - -#ifdef __STDC__ -static void se_memo_freer(struct rx_hash_item *node) -#else -static void se_memo_freer(node) -struct rx_hash_item *node; -#endif -{ - free((char *) node->data); -} - - -#ifdef __STDC__ -static void nfa_set_freer(struct rx_hash_item *node) -#else -static void nfa_set_freer(node) -struct rx_hash_item *node; -#endif -{ - free((char *) node->data); -} - - -/* This copies an entire NFA into a single malloced block of memory. - * Mostly this is for compatability with regex.c, though it is convenient - * to have the nfa nodes in an array. - */ - -#ifdef __STDC__ -RX_DECL int -rx_compactify_nfa(struct rx *rx, void **mem, unsigned long *size) -#else -RX_DECL int rx_compactify_nfa(rx, mem, size) -struct rx *rx; -void **mem; -unsigned long *size; -#endif -{ - int total_nodec; - struct rx_nfa_state *n; - int edgec = 0; - int eclosec = 0; - int se_list_consc = count_hash_nodes(&rx->se_list_memo); - int nfa_setc = count_hash_nodes(&rx->set_list_memo); - unsigned long total_size; - - /* This takes place in two stages. First, the total size of the - * nfa is computed, then structures are copied. - */ - n = rx->nfa_states; - total_nodec = 0; - while (n) { - struct rx_nfa_edge *e = n->edges; - struct rx_possible_future *ec = n->futures; - - ++total_nodec; - while (e) { - ++edgec; - e = e->next; - } - while (ec) { - ++eclosec; - ec = ec->next; - } - n = n->next; - } - - total_size = (total_nodec * sizeof(struct rx_nfa_state) - + edgec * rx_sizeof_bitset(rx->local_cset_size) - + edgec * sizeof(struct rx_nfa_edge) - + nfa_setc * sizeof(struct rx_nfa_state_set) - + eclosec * sizeof(struct rx_possible_future) - + se_list_consc * sizeof(struct rx_se_list) - + rx->reserved); - - if (total_size > *size) { - *mem = remalloc(*mem, total_size); - if (*mem) - *size = total_size; - else - return 0; - } - /* Now we've allocated the memory; this copies the NFA. */ - { - static struct rx_nfa_state **scratch = 0; - static int scratch_alloc = 0; - struct rx_nfa_state *state_base = (struct rx_nfa_state *) *mem; - struct rx_nfa_state *new_state = state_base; - struct rx_nfa_edge *new_edge = (struct rx_nfa_edge *) - ((char *) state_base + total_nodec * sizeof(struct rx_nfa_state)); - struct rx_se_list *new_se_list = (struct rx_se_list *) - ((char *) new_edge + edgec * sizeof(struct rx_nfa_edge)); - struct rx_possible_future *new_close = - ((struct rx_possible_future *) - ((char *) new_se_list - - + se_list_consc * sizeof(struct rx_se_list))); - struct rx_nfa_state_set *new_nfa_set = ((struct rx_nfa_state_set *) - - ((char *) new_close + - eclosec * - - sizeof(struct - rx_possible_future))); - char *new_bitset = - - ((char *) new_nfa_set + - nfa_setc * sizeof(struct rx_nfa_state_set)); - int x; - struct rx_nfa_state *n; - - if (scratch_alloc < total_nodec) { - scratch = ((struct rx_nfa_state **) - remalloc(scratch, total_nodec * sizeof(*scratch))); - if (scratch) - scratch_alloc = total_nodec; - else { - scratch_alloc = 0; - return 0; - } - } - - for (x = 0, n = rx->nfa_states; n; n = n->next) - scratch[x++] = n; - - qsort(scratch, total_nodec, sizeof(struct rx_nfa_state *), - (__compar_fn_t) nfacmp); - - for (x = 0; x < total_nodec; ++x) { - struct rx_possible_future *eclose = scratch[x]->futures; - struct rx_nfa_edge *edge = scratch[x]->edges; - struct rx_nfa_state *cn = new_state++; - - cn->futures = 0; - cn->edges = 0; - cn->next = (x == total_nodec - 1) ? 0 : (cn + 1); - cn->id = scratch[x]->id; - cn->is_final = scratch[x]->is_final; - cn->is_start = scratch[x]->is_start; - cn->mark = 0; - while (edge) { - int indx = (edge->dest->id < 0 - ? (total_nodec + edge->dest->id) - - : edge->dest->id); - struct rx_nfa_edge *e = new_edge++; - rx_Bitset cset = (rx_Bitset) new_bitset; - - new_bitset += rx_sizeof_bitset(rx->local_cset_size); - rx_bitset_null(rx->local_cset_size, cset); - rx_bitset_union(rx->local_cset_size, cset, - edge->params.cset); - e->next = cn->edges; - cn->edges = e; - e->type = edge->type; - e->dest = state_base + indx; - e->params.cset = cset; - edge = edge->next; - } - while (eclose) { - struct rx_possible_future *ec = new_close++; - struct rx_hash_item *sp; - struct rx_se_list **sepos; - struct rx_se_list *sesrc; - struct rx_nfa_state_set *destlst; - struct rx_nfa_state_set **destpos; - - ec->next = cn->futures; - cn->futures = ec; - for (sepos = &ec->effects, sesrc = eclose->effects; - sesrc; sesrc = sesrc->cdr, sepos = &(*sepos)->cdr) { - sp = rx_hash_find(&rx->se_list_memo, - (long) sesrc-> - car ^ (long) sesrc->cdr, sesrc, - &se_list_hash_rules); - if (sp->binding) { - sesrc = (struct rx_se_list *) sp->binding; - break; - } - *new_se_list = *sesrc; - sp->binding = (void *) new_se_list; - *sepos = new_se_list; - ++new_se_list; - } - *sepos = sesrc; - for (destpos = &ec->destset, destlst = eclose->destset; - destlst; - destpos = &(*destpos)->cdr, destlst = destlst->cdr) { - sp = rx_hash_find(&rx->set_list_memo, - ((((long) destlst->car) >> 8) - ^ (long) destlst->cdr), - destlst, &nfa_set_hash_rules); - if (sp->binding) { - destlst = (struct rx_nfa_state_set *) sp->binding; - break; - } - *new_nfa_set = *destlst; - new_nfa_set->car = state_base + destlst->car->id; - sp->binding = (void *) new_nfa_set; - *destpos = new_nfa_set; - ++new_nfa_set; - } - *destpos = destlst; - eclose = eclose->next; - } - } - } - rx_free_hash_table(&rx->se_list_memo, se_memo_freer, - &se_list_hash_rules); - bzero(&rx->se_list_memo, sizeof(rx->se_list_memo)); - rx_free_hash_table(&rx->set_list_memo, nfa_set_freer, - &nfa_set_hash_rules); - bzero(&rx->set_list_memo, sizeof(rx->set_list_memo)); - - rx_free_nfa(rx); - rx->nfa_states = (struct rx_nfa_state *) *mem; - return 1; -} - - -/* The functions in the next several pages define the lazy-NFA-conversion used - * by matchers. The input to this construction is an NFA such as - * is built by compactify_nfa (rx.c). The output is the superNFA. - */ - -/* Match engines can use arbitrary values for opcodes. So, the parse tree - * is built using instructions names (enum rx_opcode), but the superstate - * nfa is populated with mystery opcodes (void *). - * - * For convenience, here is an id table. The opcodes are == to their inxs - * - * The lables in re_search_2 would make good values for instructions. - */ - -void *rx_id_instruction_table[rx_num_instructions] = { - (void *) rx_backtrack_point, - (void *) rx_do_side_effects, - (void *) rx_cache_miss, - (void *) rx_next_char, - (void *) rx_backtrack, - (void *) rx_error_inx -}; - - - -/* Memory mgt. for superstate graphs. */ - -#ifdef __STDC__ -static char *rx_cache_malloc(struct rx_cache *cache, int bytes) -#else -static char *rx_cache_malloc(cache, bytes) -struct rx_cache *cache; -int bytes; -#endif -{ - while (cache->bytes_left < bytes) { - if (cache->memory_pos) - cache->memory_pos = cache->memory_pos->next; - if (!cache->memory_pos) { - cache->morecore(cache); - if (!cache->memory_pos) - return 0; - } - cache->bytes_left = cache->memory_pos->bytes; - cache->memory_addr = ((char *) cache->memory_pos - - + sizeof(struct rx_blocklist)); - } - cache->bytes_left -= bytes; - { - char *addr = cache->memory_addr; - - cache->memory_addr += bytes; - return addr; - } -} - -#ifdef __STDC__ -static void -rx_cache_free(struct rx_cache *cache, - struct rx_freelist **freelist, char *mem) -#else -static void rx_cache_free(cache, freelist, mem) -struct rx_cache *cache; -struct rx_freelist **freelist; -char *mem; -#endif -{ - struct rx_freelist *it = (struct rx_freelist *) mem; - - it->next = *freelist; - *freelist = it; -} - -/* The partially instantiated superstate graph has a transition - * table at every node. There is one entry for every character. - * This fills in the transition for a set. - */ -#ifdef __STDC__ -static void -install_transition(struct rx_superstate *super, - struct rx_inx *answer, rx_Bitset trcset) -#else -static void install_transition(super, answer, trcset) -struct rx_superstate *super; -struct rx_inx *answer; -rx_Bitset trcset; -#endif -{ - struct rx_inx *transitions = super->transitions; - int chr; - - for (chr = 0; chr < 256;) - if (!*trcset) { - ++trcset; - chr += 32; - } else { - RX_subset sub = *trcset; - RX_subset mask = 1; - int bound = chr + 32; - - while (chr < bound) { - if (sub & mask) - transitions[chr] = *answer; - ++chr; - mask <<= 1; - } - ++trcset; - } -} - -#ifdef __STDC__ -static int qlen(struct rx_superstate *q) -#else -static int qlen(q) -struct rx_superstate *q; -#endif -{ - int count = 1; - struct rx_superstate *it; - - if (!q) - return 0; - for (it = q->next_recyclable; it != q; it = it->next_recyclable) - ++count; - return count; -} - -#ifdef __STDC__ -static void check_cache(struct rx_cache *cache) -#else -static void check_cache(cache) -struct rx_cache *cache; -#endif -{ - struct rx_cache *you_fucked_up = 0; - int total = cache->superstates; - int semi = cache->semifree_superstates; - - if (semi != qlen(cache->semifree_superstate)) - check_cache(you_fucked_up); - if ((total - semi) != qlen(cache->lru_superstate)) - check_cache(you_fucked_up); -} - -/* When a superstate is old and neglected, it can enter a - * semi-free state. A semi-free state is slated to die. - * Incoming transitions to a semi-free state are re-written - * to cause an (interpreted) fault when they are taken. - * The fault handler revives the semi-free state, patches - * incoming transitions back to normal, and continues. - * - * The idea is basicly to free in two stages, aborting - * between the two if the state turns out to be useful again. - * When a free is aborted, the rescued superstate is placed - * in the most-favored slot to maximize the time until it - * is next semi-freed. - */ - -#ifdef __STDC__ -static void semifree_superstate(struct rx_cache *cache) -#else -static void semifree_superstate(cache) -struct rx_cache *cache; -#endif -{ - int disqualified = cache->semifree_superstates; - - if (disqualified == cache->superstates) - return; - while (cache->lru_superstate->locks) { - cache->lru_superstate = cache->lru_superstate->next_recyclable; - ++disqualified; - if (disqualified == cache->superstates) - return; - } - { - struct rx_superstate *it = cache->lru_superstate; - - it->next_recyclable->prev_recyclable = it->prev_recyclable; - it->prev_recyclable->next_recyclable = it->next_recyclable; - cache->lru_superstate = (it == it->next_recyclable - ? 0 : it->next_recyclable); - if (!cache->semifree_superstate) { - cache->semifree_superstate = it; - it->next_recyclable = it; - it->prev_recyclable = it; - } else { - it->prev_recyclable = - cache->semifree_superstate->prev_recyclable; - it->next_recyclable = cache->semifree_superstate; - it->prev_recyclable->next_recyclable = it; - it->next_recyclable->prev_recyclable = it; - } - { - struct rx_distinct_future *df; - - it->is_semifree = 1; - ++cache->semifree_superstates; - df = it->transition_refs; - if (df) { - df->prev_same_dest->next_same_dest = 0; - for (df = it->transition_refs; df; df = df->next_same_dest) { - df->future_frame.inx = - cache->instruction_table[rx_cache_miss]; - df->future_frame.data = 0; - df->future_frame.data_2 = (void *) df; - /* If there are any NEXT-CHAR instruction frames that - * refer to this state, we convert them to CACHE-MISS frames. - */ - if (!df->effects - && (df->edge->options->next_same_super_edge[0] - == df->edge->options)) - install_transition(df->present, &df->future_frame, - df->edge->cset); - } - df = it->transition_refs; - df->prev_same_dest->next_same_dest = df; - } - } - } -} - -#ifdef __STDC__ -static void -refresh_semifree_superstate(struct rx_cache *cache, - struct rx_superstate *super) -#else -static void refresh_semifree_superstate(cache, super) -struct rx_cache *cache; -struct rx_superstate *super; -#endif -{ - struct rx_distinct_future *df; - - if (super->transition_refs) { - super->transition_refs->prev_same_dest->next_same_dest = 0; - for (df = super->transition_refs; df; df = df->next_same_dest) { - df->future_frame.inx = cache->instruction_table[rx_next_char]; - df->future_frame.data = (void *) super->transitions; - /* CACHE-MISS instruction frames that refer to this state, - * must be converted to NEXT-CHAR frames. - */ - if (!df->effects && (df->edge->options->next_same_super_edge[0] - == df->edge->options)) - install_transition(df->present, &df->future_frame, - df->edge->cset); - } - super->transition_refs->prev_same_dest->next_same_dest - = super->transition_refs; - } - if (cache->semifree_superstate == super) - cache->semifree_superstate = (super->prev_recyclable == super - ? 0 : super->prev_recyclable); - super->next_recyclable->prev_recyclable = super->prev_recyclable; - super->prev_recyclable->next_recyclable = super->next_recyclable; - - if (!cache->lru_superstate) - (cache->lru_superstate - = super->next_recyclable = super->prev_recyclable = super); - else { - super->next_recyclable = cache->lru_superstate; - super->prev_recyclable = cache->lru_superstate->prev_recyclable; - super->next_recyclable->prev_recyclable = super; - super->prev_recyclable->next_recyclable = super; - } - super->is_semifree = 0; - --cache->semifree_superstates; -} - -#ifdef __STDC__ -static void -rx_refresh_this_superstate(struct rx_cache *cache, - struct rx_superstate *superstate) -#else -static void rx_refresh_this_superstate(cache, superstate) -struct rx_cache *cache; -struct rx_superstate *superstate; -#endif -{ - if (superstate->is_semifree) - refresh_semifree_superstate(cache, superstate); - else if (cache->lru_superstate == superstate) - cache->lru_superstate = superstate->next_recyclable; - else if (superstate != cache->lru_superstate->prev_recyclable) { - superstate->next_recyclable->prev_recyclable - = superstate->prev_recyclable; - superstate->prev_recyclable->next_recyclable - = superstate->next_recyclable; - superstate->next_recyclable = cache->lru_superstate; - superstate->prev_recyclable = - cache->lru_superstate->prev_recyclable; - superstate->next_recyclable->prev_recyclable = superstate; - superstate->prev_recyclable->next_recyclable = superstate; - } -} - -#ifdef __STDC__ -static void -release_superset_low(struct rx_cache *cache, struct rx_superset *set) -#else -static void release_superset_low(cache, set) -struct rx_cache *cache; -struct rx_superset *set; -#endif -{ - if (!--set->refs) { - if (set->cdr) - release_superset_low(cache, set->cdr); - - set->starts_for = 0; - - rx_hash_free - (rx_hash_find - (&cache->superset_table, - (unsigned long) set->car ^ set-> - id ^ (unsigned long) set->cdr, (void *) set, - &cache->superset_hash_rules), &cache->superset_hash_rules); - rx_cache_free(cache, &cache->free_supersets, (char *) set); - } -} - -#ifdef __STDC__ -RX_DECL void rx_release_superset(struct rx *rx, struct rx_superset *set) -#else -RX_DECL void rx_release_superset(rx, set) -struct rx *rx; -struct rx_superset *set; -#endif -{ - release_superset_low(rx->cache, set); -} - -/* This tries to add a new superstate to the superstate freelist. - * It might, as a result, free some edge pieces or hash tables. - * If nothing can be freed because too many locks are being held, fail. - */ - -#ifdef __STDC__ -static int rx_really_free_superstate(struct rx_cache *cache) -#else -static int rx_really_free_superstate(cache) -struct rx_cache *cache; -#endif -{ - int locked_superstates = 0; - struct rx_superstate *it; - - if (!cache->superstates) - return 0; - - { - /* This is a total guess. The idea is that we should expect as - * many misses as we've recently experienced. I.e., cache->misses - * should be the same as cache->semifree_superstates. - */ - while ((cache->hits + cache->misses) > cache->superstates_allowed) { - cache->hits >>= 1; - cache->misses >>= 1; - } - if (((cache->hits + cache->misses) * cache->semifree_superstates) - < (cache->superstates * cache->misses)) { - semifree_superstate(cache); - semifree_superstate(cache); - } - } - - while (cache->semifree_superstate && cache->semifree_superstate->locks) { - refresh_semifree_superstate(cache, cache->semifree_superstate); - ++locked_superstates; - if (locked_superstates == cache->superstates) - return 0; - } - - if (cache->semifree_superstate) { - it = cache->semifree_superstate; - it->next_recyclable->prev_recyclable = it->prev_recyclable; - it->prev_recyclable->next_recyclable = it->next_recyclable; - cache->semifree_superstate = ((it == it->next_recyclable) - ? 0 : it->next_recyclable); - --cache->semifree_superstates; - } else { - while (cache->lru_superstate->locks) { - cache->lru_superstate = cache->lru_superstate->next_recyclable; - ++locked_superstates; - if (locked_superstates == cache->superstates) - return 0; - } - it = cache->lru_superstate; - it->next_recyclable->prev_recyclable = it->prev_recyclable; - it->prev_recyclable->next_recyclable = it->next_recyclable; - cache->lru_superstate = ((it == it->next_recyclable) - ? 0 : it->next_recyclable); - } - - if (it->transition_refs) { - struct rx_distinct_future *df; - - for (df = it->transition_refs, - df->prev_same_dest->next_same_dest = 0; - df; df = df->next_same_dest) { - df->future_frame.inx = cache->instruction_table[rx_cache_miss]; - df->future_frame.data = 0; - df->future_frame.data_2 = (void *) df; - df->future = 0; - } - it->transition_refs->prev_same_dest->next_same_dest = - it->transition_refs; - } - { - struct rx_super_edge *tc = it->edges; - - while (tc) { - struct rx_distinct_future *df; - struct rx_super_edge *tct = tc->next; - - df = tc->options; - df->next_same_super_edge[1]->next_same_super_edge[0] = 0; - while (df) { - struct rx_distinct_future *dft = df; - - df = df->next_same_super_edge[0]; - - - if (dft->future && dft->future->transition_refs == dft) { - dft->future->transition_refs = dft->next_same_dest; - if (dft->future->transition_refs == dft) - dft->future->transition_refs = 0; - } - dft->next_same_dest->prev_same_dest = dft->prev_same_dest; - dft->prev_same_dest->next_same_dest = dft->next_same_dest; - rx_cache_free(cache, &cache->free_discernable_futures, - (char *) dft); - } - rx_cache_free(cache, &cache->free_transition_classes, - (char *) tc); - tc = tct; - } - } - - if (it->contents->superstate == it) - it->contents->superstate = 0; - release_superset_low(cache, it->contents); - rx_cache_free(cache, &cache->free_superstates, (char *) it); - --cache->superstates; - return 1; -} - -#ifdef __STDC__ -static char *rx_cache_get(struct rx_cache *cache, - struct rx_freelist **freelist) -#else -static char *rx_cache_get(cache, freelist) -struct rx_cache *cache; -struct rx_freelist **freelist; -#endif -{ - while (!*freelist && rx_really_free_superstate(cache)); - if (!*freelist) - return 0; - { - struct rx_freelist *it = *freelist; - - *freelist = it->next; - return (char *) it; - } -} - -#ifdef __STDC__ -static char *rx_cache_malloc_or_get(struct rx_cache *cache, - struct rx_freelist **freelist, - int bytes) -#else -static char *rx_cache_malloc_or_get(cache, freelist, bytes) -struct rx_cache *cache; -struct rx_freelist **freelist; -int bytes; -#endif -{ - if (!*freelist) { - char *answer = rx_cache_malloc(cache, bytes); - - if (answer) - return answer; - } - - return rx_cache_get(cache, freelist); -} - -#ifdef __STDC__ -static char *rx_cache_get_superstate(struct rx_cache *cache) -#else -static char *rx_cache_get_superstate(cache) -struct rx_cache *cache; -#endif -{ - char *answer; - int bytes = (sizeof(struct rx_superstate) - + cache->local_cset_size * sizeof(struct rx_inx)); - - if (!cache->free_superstates - && (cache->superstates < cache->superstates_allowed)) { - answer = rx_cache_malloc(cache, bytes); - if (answer) { - ++cache->superstates; - return answer; - } - } - answer = rx_cache_get(cache, &cache->free_superstates); - if (!answer) { - answer = rx_cache_malloc(cache, bytes); - if (answer) - ++cache->superstates_allowed; - } - ++cache->superstates; - return answer; -} - - - -#ifdef __STDC__ -static int supersetcmp(void *va, void *vb) -#else -static int supersetcmp(va, vb) -void *va; -void *vb; -#endif -{ - struct rx_superset *a = (struct rx_superset *) va; - struct rx_superset *b = (struct rx_superset *) vb; - - return ((a == b) - || (a && b && (a->car == b->car) && (a->cdr == b->cdr))); -} - -#ifdef __STDC__ -static struct rx_hash_item *superset_allocator(struct rx_hash_rules *rules, - void *val) -#else -static struct rx_hash_item *superset_allocator(rules, val) -struct rx_hash_rules *rules; -void *val; -#endif -{ - struct rx_cache *cache = ((struct rx_cache *) - ((char *) rules - - - - (unsigned - long) (&((struct rx_cache *) - 0)->superset_hash_rules))); - struct rx_superset *template = (struct rx_superset *) val; - struct rx_superset *newset - = ((struct rx_superset *) rx_cache_malloc_or_get(cache, - &cache->free_supersets, - sizeof - - (*template))); - if (!newset) - return 0; - newset->refs = 0; - newset->car = template->car; - newset->id = template->car->id; - newset->cdr = template->cdr; - newset->superstate = 0; - rx_protect_superset(rx, template->cdr); - newset->hash_item.data = (void *) newset; - newset->hash_item.binding = 0; - return &newset->hash_item; -} - -#ifdef __STDC__ -static struct rx_hash *super_hash_allocator(struct rx_hash_rules *rules) -#else -static struct rx_hash *super_hash_allocator(rules) -struct rx_hash_rules *rules; -#endif -{ - struct rx_cache *cache = ((struct rx_cache *) - ((char *) rules - - - - (unsigned - long) (&((struct rx_cache *) - 0)->superset_hash_rules))); - return ((struct rx_hash *) - rx_cache_malloc_or_get(cache, &cache->free_hash, - - sizeof(struct rx_hash))); -} - - -#ifdef __STDC__ -static void -super_hash_liberator(struct rx_hash *hash, struct rx_hash_rules *rules) -#else -static void super_hash_liberator(hash, rules) -struct rx_hash *hash; -struct rx_hash_rules *rules; -#endif -{ - struct rx_cache *cache = ((struct rx_cache *) - - (char *) rules - - (long) (& - - ((struct rx_cache *) - 0)->superset_hash_rules)); - rx_cache_free(cache, &cache->free_hash, (char *) hash); -} - -#ifdef __STDC__ -static void -superset_hash_item_liberator(struct rx_hash_item *it, - struct rx_hash_rules *rules) -#else -static void superset_hash_item_liberator(it, rules) /* Well, it does ya know. */ -struct rx_hash_item *it; -struct rx_hash_rules *rules; -#endif -{ -} - -int rx_cache_bound = 128; -static int rx_default_cache_got = 0; - -#ifdef __STDC__ -static int bytes_for_cache_size(int supers, int cset_size) -#else -static int bytes_for_cache_size(supers, cset_size) -int supers; -int cset_size; -#endif -{ - /* What the hell is this? !!! */ - return (int) - ((float) supers * ((1.03 * (float) (rx_sizeof_bitset(cset_size) - + - sizeof(struct rx_super_edge))) - + - (1.80 * - (float) sizeof(struct rx_possible_future)) + - (float) (sizeof(struct rx_superstate) - + cset_size * sizeof(struct rx_inx)))); -} - -#ifdef __STDC__ -static void rx_morecore(struct rx_cache *cache) -#else -static void rx_morecore(cache) -struct rx_cache *cache; -#endif -{ - if (rx_default_cache_got >= rx_cache_bound) - return; - - rx_default_cache_got += 16; - cache->superstates_allowed = rx_cache_bound; - { - struct rx_blocklist **pos = &cache->memory; - int size = bytes_for_cache_size(16, cache->local_cset_size); - - while (*pos) - pos = &(*pos)->next; - *pos = ((struct rx_blocklist *) - malloc(size + sizeof(struct rx_blocklist))); - - if (!*pos) - return; - - (*pos)->next = 0; - (*pos)->bytes = size; - cache->memory_pos = *pos; - cache->memory_addr = (char *) *pos + sizeof(**pos); - cache->bytes_left = size; - } -} - -static struct rx_cache default_cache = { - { - supersetcmp, - super_hash_allocator, - super_hash_liberator, - superset_allocator, - superset_hash_item_liberator, - }, - 0, - 0, - 0, - 0, - rx_morecore, - - 0, - 0, - 0, - 0, - 0, - - 0, - 0, - - 0, - - 0, - 0, - 0, - 0, - 128, - - 256, - rx_id_instruction_table, - - { - 0, - 0, - {0}, - {0}, - {0} - } -}; - -/* This adds an element to a superstate set. These sets are lists, such - * that lists with == elements are ==. The empty set is returned by - * superset_cons (rx, 0, 0) and is NOT equivelent to - * (struct rx_superset)0. - */ - -#ifdef __STDC__ -RX_DECL struct rx_superset *rx_superset_cons(struct rx *rx, - struct rx_nfa_state *car, - struct rx_superset *cdr) -#else -RX_DECL struct rx_superset *rx_superset_cons(rx, car, cdr) -struct rx *rx; -struct rx_nfa_state *car; -struct rx_superset *cdr; -#endif -{ - struct rx_cache *cache = rx->cache; - - if (!car && !cdr) { - if (!cache->empty_superset) { - cache->empty_superset = ((struct rx_superset *) - rx_cache_malloc_or_get(cache, - &cache->free_supersets, - - sizeof(struct - rx_superset))); - if (!cache->empty_superset) - return 0; - bzero(cache->empty_superset, sizeof(struct rx_superset)); - - cache->empty_superset->refs = 1000; - } - return cache->empty_superset; - } - { - struct rx_superset template; - struct rx_hash_item *hit; - - template.car = car; - template.cdr = cdr; - template.id = car->id; - /* While hash_store will protect cdr itself it might first allocate hash - tables and stuff which might cause it to be garbage collected before - it's protected -- [gsstark:19961026.2155EST] */ - rx_protect_superset(rx, cdr); - hit = rx_hash_store(&cache->superset_table, - (unsigned long) car ^ car->id ^ (unsigned long) - cdr, (void *) &template, - &cache->superset_hash_rules); - rx_release_superset(rx, cdr); - return (hit ? (struct rx_superset *) hit->data : 0); - } -} - -/* This computes a union of two NFA state sets. The sets do not have the - * same representation though. One is a RX_SUPERSET structure (part - * of the superstate NFA) and the other is an NFA_STATE_SET (part of the NFA). - */ - -#ifdef __STDC__ -RX_DECL struct rx_superset *rx_superstate_eclosure_union - (struct rx *rx, struct rx_superset *set, struct rx_nfa_state_set *ecl) -#else -RX_DECL struct rx_superset *rx_superstate_eclosure_union(rx, set, ecl) -struct rx *rx; -struct rx_superset *set; -struct rx_nfa_state_set *ecl; -#endif -{ - if (!ecl) - return set; - - if (!set->car) - return rx_superset_cons(rx, ecl->car, - rx_superstate_eclosure_union(rx, set, - ecl->cdr)); - if (set->car == ecl->car) - return rx_superstate_eclosure_union(rx, set, ecl->cdr); - - { - struct rx_superset *tail; - struct rx_nfa_state *first; - - if (set->car > ecl->car) { - tail = rx_superstate_eclosure_union(rx, set->cdr, ecl); - first = set->car; - } else { - tail = rx_superstate_eclosure_union(rx, set, ecl->cdr); - first = ecl->car; - } - if (!tail) - return 0; - else { - struct rx_superset *answer; - - answer = rx_superset_cons(rx, first, tail); - if (!answer) { - rx_protect_superset(rx, tail); - rx_release_superset(rx, tail); - return 0; - } else - return answer; - } - } -} - - - - -/* - * This makes sure that a list of rx_distinct_futures contains - * a future for each possible set of side effects in the eclosure - * of a given state. This is some of the work of filling in a - * superstate transition. - */ - -#ifdef __STDC__ -static struct rx_distinct_future *include_futures(struct rx *rx, struct rx_distinct_future - *df, struct rx_nfa_state - *state, struct rx_superstate - *superstate) -#else -static struct rx_distinct_future *include_futures(rx, df, state, - superstate) -struct rx *rx; -struct rx_distinct_future *df; -struct rx_nfa_state *state; -struct rx_superstate *superstate; -#endif -{ - struct rx_possible_future *future; - struct rx_cache *cache = rx->cache; - - for (future = state->futures; future; future = future->next) { - struct rx_distinct_future *dfp; - struct rx_distinct_future *insert_before = 0; - - if (df) - df->next_same_super_edge[1]->next_same_super_edge[0] = 0; - for (dfp = df; dfp; dfp = dfp->next_same_super_edge[0]) - if (dfp->effects == future->effects) - break; - else { - int order = - - rx->se_list_cmp(rx, dfp->effects, future->effects); - if (order > 0) { - insert_before = dfp; - dfp = 0; - break; - } - } - if (df) - df->next_same_super_edge[1]->next_same_super_edge[0] = df; - if (!dfp) { - dfp = ((struct rx_distinct_future *) - rx_cache_malloc_or_get(cache, - &cache->free_discernable_futures, - - sizeof(struct - rx_distinct_future))); - if (!dfp) - return 0; - if (!df) { - df = insert_before = dfp; - df->next_same_super_edge[0] = df->next_same_super_edge[1] = - df; - } else if (!insert_before) - insert_before = df; - else if (insert_before == df) - df = dfp; - - dfp->next_same_super_edge[0] = insert_before; - dfp->next_same_super_edge[1] - = insert_before->next_same_super_edge[1]; - dfp->next_same_super_edge[1]->next_same_super_edge[0] = dfp; - dfp->next_same_super_edge[0]->next_same_super_edge[1] = dfp; - dfp->next_same_dest = dfp->prev_same_dest = dfp; - dfp->future = 0; - dfp->present = superstate; - dfp->future_frame.inx = rx->instruction_table[rx_cache_miss]; - dfp->future_frame.data = 0; - dfp->future_frame.data_2 = (void *) dfp; - dfp->side_effects_frame.inx - = rx->instruction_table[rx_do_side_effects]; - dfp->side_effects_frame.data = 0; - dfp->side_effects_frame.data_2 = (void *) dfp; - dfp->effects = future->effects; - } - } - return df; -} - - -/* This constructs a new superstate from its state set. The only - * complexity here is memory management. - */ -#ifdef __STDC__ -RX_DECL struct rx_superstate *rx_superstate(struct rx *rx, - struct rx_superset *set) -#else -RX_DECL struct rx_superstate *rx_superstate(rx, set) -struct rx *rx; -struct rx_superset *set; -#endif -{ - struct rx_cache *cache = rx->cache; - struct rx_superstate *superstate = 0; - - /* Does the superstate already exist in the cache? */ - if (set->superstate) { - if (set->superstate->rx_id != rx->rx_id) { - /* Aha. It is in the cache, but belongs to a superstate - * that refers to an NFA that no longer exists. - * (We know it no longer exists because it was evidently - * stored in the same region of memory as the current nfa - * yet it has a different id.) - */ - superstate = set->superstate; - if (!superstate->is_semifree) { - if (cache->lru_superstate == superstate) { - cache->lru_superstate = superstate->next_recyclable; - if (cache->lru_superstate == superstate) - cache->lru_superstate = 0; - } - { - superstate->next_recyclable->prev_recyclable - = superstate->prev_recyclable; - superstate->prev_recyclable->next_recyclable - = superstate->next_recyclable; - if (!cache->semifree_superstate) { - (cache->semifree_superstate - = superstate->next_recyclable - = superstate->prev_recyclable = superstate); - } else { - superstate->next_recyclable = - cache->semifree_superstate; - superstate->prev_recyclable = - cache->semifree_superstate->prev_recyclable; - superstate->next_recyclable->prev_recyclable = - superstate; - superstate->prev_recyclable->next_recyclable = - superstate; - cache->semifree_superstate = superstate; - } - ++cache->semifree_superstates; - } - } - set->superstate = 0; - goto handle_cache_miss; - } - ++cache->hits; - superstate = set->superstate; - - rx_refresh_this_superstate(cache, superstate); - return superstate; - } - - handle_cache_miss: - - /* This point reached only for cache misses. */ - ++cache->misses; -#if RX_DEBUG - if (rx_debug_trace > 1) { - struct rx_superset *setp = set; - - fprintf(stderr, "Building a superstet %d(%d): ", rx->rx_id, set); - while (setp) { - fprintf(stderr, "%d ", setp->id); - setp = setp->cdr; - } - fprintf(stderr, "(%d)\n", set); - } -#endif - superstate = (struct rx_superstate *) rx_cache_get_superstate(cache); - if (!superstate) - return 0; - - if (!cache->lru_superstate) - (cache->lru_superstate - = superstate->next_recyclable - = superstate->prev_recyclable = superstate); - else { - superstate->next_recyclable = cache->lru_superstate; - superstate->prev_recyclable = - cache->lru_superstate->prev_recyclable; - (superstate->prev_recyclable->next_recyclable = - superstate->next_recyclable->prev_recyclable = superstate); - } - superstate->rx_id = rx->rx_id; - superstate->transition_refs = 0; - superstate->locks = 0; - superstate->is_semifree = 0; - set->superstate = superstate; - superstate->contents = set; - rx_protect_superset(rx, set); - superstate->edges = 0; - { - int x; - - /* None of the transitions from this superstate are known yet. */ - for (x = 0; x < rx->local_cset_size; ++x) { /* &&&&& 3.8 % */ - struct rx_inx *ifr = &superstate->transitions[x]; - - ifr->inx = rx->instruction_table[rx_cache_miss]; - ifr->data = ifr->data_2 = 0; - } - } - return superstate; -} - - -/* This computes the destination set of one edge of the superstate NFA. - * Note that a RX_DISTINCT_FUTURE is a superstate edge. - * Returns 0 on an allocation failure. - */ - -#ifdef __STDC__ -static int solve_destination(struct rx *rx, struct rx_distinct_future *df) -#else -static int solve_destination(rx, df) -struct rx *rx; -struct rx_distinct_future *df; -#endif -{ - struct rx_super_edge *tc = df->edge; - struct rx_superset *nfa_state; - struct rx_superset *nil_set = rx_superset_cons(rx, 0, 0); - struct rx_superset *solution = nil_set; - struct rx_superstate *dest; - - rx_protect_superset(rx, solution); - /* Iterate over all NFA states in the state set of this superstate. */ - for (nfa_state = df->present->contents; - nfa_state->car; nfa_state = nfa_state->cdr) { - struct rx_nfa_edge *e; - - /* Iterate over all edges of each NFA state. */ - for (e = nfa_state->car->edges; e; e = e->next) - /* If we find an edge that is labeled with - * the characters we are solving for..... - */ - if (rx_bitset_is_subset(rx->local_cset_size, - tc->cset, e->params.cset)) { - struct rx_nfa_state *n = e->dest; - struct rx_possible_future *pf; - - /* ....search the partial epsilon closures of the destination - * of that edge for a path that involves the same set of - * side effects we are solving for. - * If we find such a RX_POSSIBLE_FUTURE, we add members to the - * stateset we are computing. - */ - for (pf = n->futures; pf; pf = pf->next) - if (pf->effects == df->effects) { - struct rx_superset *old_sol; - - old_sol = solution; - solution = - rx_superstate_eclosure_union(rx, solution, - pf->destset); - if (!solution) - return 0; - rx_protect_superset(rx, solution); - rx_release_superset(rx, old_sol); - } - } - } - /* It is possible that the RX_DISTINCT_FUTURE we are working on has - * the empty set of NFA states as its definition. In that case, this - * is a failure point. - */ - if (solution == nil_set) { - df->future_frame.inx = (void *) rx_backtrack; - df->future_frame.data = 0; - df->future_frame.data_2 = 0; - return 1; - } - dest = rx_superstate(rx, solution); - rx_release_superset(rx, solution); - if (!dest) - return 0; - - { - struct rx_distinct_future *dft; - - dft = df; - df->prev_same_dest->next_same_dest = 0; - while (dft) { - dft->future = dest; - dft->future_frame.inx = rx->instruction_table[rx_next_char]; - dft->future_frame.data = (void *) dest->transitions; - dft = dft->next_same_dest; - } - df->prev_same_dest->next_same_dest = df; - } - if (!dest->transition_refs) - dest->transition_refs = df; - else { - struct rx_distinct_future *dft = - - dest->transition_refs->next_same_dest; - dest->transition_refs->next_same_dest = df->next_same_dest; - df->next_same_dest->prev_same_dest = dest->transition_refs; - df->next_same_dest = dft; - dft->prev_same_dest = df; - } - return 1; -} - - -/* This takes a superstate and a character, and computes some edges - * from the superstate NFA. In particular, this computes all edges - * that lead from SUPERSTATE given CHR. This function also - * computes the set of characters that share this edge set. - * This returns 0 on allocation error. - * The character set and list of edges are returned through - * the paramters CSETOUT and DFOUT. -} */ - -#ifdef __STDC__ -static int -compute_super_edge(struct rx *rx, struct rx_distinct_future **dfout, - rx_Bitset csetout, struct rx_superstate *superstate, - unsigned char chr) -#else -static int compute_super_edge(rx, dfout, csetout, superstate, chr) -struct rx *rx; -struct rx_distinct_future **dfout; -rx_Bitset csetout; -struct rx_superstate *superstate; -unsigned char chr; -#endif -{ - struct rx_superset *stateset = superstate->contents; - - /* To compute the set of characters that share edges with CHR, - * we start with the full character set, and subtract. - */ - rx_bitset_universe(rx->local_cset_size, csetout); - *dfout = 0; - - /* Iterate over the NFA states in the superstate state-set. */ - while (stateset->car) { - struct rx_nfa_edge *e; - - for (e = stateset->car->edges; e; e = e->next) - if (RX_bitset_member(e->params.cset, chr)) { - /* If we find an NFA edge that applies, we make sure there - * are corresponding edges in the superstate NFA. - */ - { - struct rx_distinct_future *saved; - - saved = *dfout; - *dfout = - include_futures(rx, *dfout, e->dest, superstate); - if (!*dfout) { - struct rx_distinct_future *df; - - df = saved; - if (df) - df-> - next_same_super_edge - [1]->next_same_super_edge[0] = 0; - while (df) { - struct rx_distinct_future *dft; - - dft = df; - df = df->next_same_super_edge[0]; - - if (dft->future - && dft->future->transition_refs == dft) { - dft->future->transition_refs = - dft->next_same_dest; - if (dft->future->transition_refs == dft) - dft->future->transition_refs = 0; - } - dft->next_same_dest->prev_same_dest = - dft->prev_same_dest; - dft->prev_same_dest->next_same_dest = - dft->next_same_dest; - rx_cache_free(rx->cache, - &rx-> - cache->free_discernable_futures, - (char *) dft); - } - return 0; - } - } - /* We also trim the character set a bit. */ - rx_bitset_intersection(rx->local_cset_size, - csetout, e->params.cset); - } else - /* An edge that doesn't apply at least tells us some characters - * that don't share the same edge set as CHR. - */ - rx_bitset_difference(rx->local_cset_size, csetout, - e->params.cset); - stateset = stateset->cdr; - } - return 1; -} - - -/* This is a constructor for RX_SUPER_EDGE structures. These are - * wrappers for lists of superstate NFA edges that share character sets labels. - * If a transition class contains more than one rx_distinct_future (superstate - * edge), then it represents a non-determinism in the superstate NFA. - */ - - -#ifdef __STDC__ -static struct rx_super_edge *rx_super_edge(struct rx *rx, - struct rx_superstate *super, - rx_Bitset cset, - struct rx_distinct_future *df) -#else -static struct rx_super_edge *rx_super_edge(rx, super, cset, df) -struct rx *rx; -struct rx_superstate *super; -rx_Bitset cset; -struct rx_distinct_future *df; -#endif -{ - struct rx_super_edge *tc = - (struct rx_super_edge *) rx_cache_malloc_or_get - (rx->cache, &rx->cache->free_transition_classes, - sizeof(struct rx_super_edge) + - - rx_sizeof_bitset(rx->local_cset_size)); - - if (!tc) - return 0; - tc->next = super->edges; - super->edges = tc; - tc->rx_backtrack_frame.inx = rx->instruction_table[rx_backtrack_point]; - tc->rx_backtrack_frame.data = 0; - tc->rx_backtrack_frame.data_2 = (void *) tc; - tc->options = df; - tc->cset = (rx_Bitset) ((char *) tc + sizeof(*tc)); - rx_bitset_assign(rx->local_cset_size, tc->cset, cset); - if (df) { - struct rx_distinct_future *dfp = df; - - df->next_same_super_edge[1]->next_same_super_edge[0] = 0; - while (dfp) { - dfp->edge = tc; - dfp = dfp->next_same_super_edge[0]; - } - df->next_same_super_edge[1]->next_same_super_edge[0] = df; - } - return tc; -} - - -/* There are three kinds of cache miss. The first occurs when a - * transition is taken that has never been computed during the - * lifetime of the source superstate. That cache miss is handled by - * calling COMPUTE_SUPER_EDGE. The second kind of cache miss - * occurs when the destination superstate of a transition doesn't - * exist. SOLVE_DESTINATION is used to construct the destination superstate. - * Finally, the third kind of cache miss occurs when the destination - * superstate of a transition is in a `semi-free state'. That case is - * handled by UNFREE_SUPERSTATE. - * - * The function of HANDLE_CACHE_MISS is to figure out which of these - * cases applies. - */ - - -#ifdef __STDC__ -static void -install_partial_transition(struct rx_superstate *super, - struct rx_inx *answer, - RX_subset set, int offset) -#else -static void install_partial_transition(super, answer, set, offset) -struct rx_superstate *super; -struct rx_inx *answer; -RX_subset set; -int offset; -#endif -{ - int start = offset; - int end = start + 32; - RX_subset pos = 1; - struct rx_inx *transitions = super->transitions; - - while (start < end) { - if (set & pos) - transitions[start] = *answer; - pos <<= 1; - ++start; - } -} - -#ifdef __STDC__ -RX_DECL struct rx_inx *rx_handle_cache_miss - (struct rx *rx, struct rx_superstate *super, unsigned char chr, - void *data) -#else -RX_DECL struct rx_inx *rx_handle_cache_miss(rx, super, chr, data) -struct rx *rx; -struct rx_superstate *super; -unsigned char chr; -void *data; -#endif -{ - int offset = chr / RX_subset_bits; - struct rx_distinct_future *df = data; - - if (!df) { /* must be the shared_cache_miss_frame */ - /* Perhaps this is just a transition waiting to be filled. */ - struct rx_super_edge *tc; - RX_subset mask = rx_subset_singletons[chr % RX_subset_bits]; - - for (tc = super->edges; tc; tc = tc->next) - if (tc->cset[offset] & mask) { - struct rx_inx *answer; - - df = tc->options; - answer = - ((tc->options->next_same_super_edge[0] != - tc->options) ? &tc-> - rx_backtrack_frame : (df->effects ? - &df->side_effects_frame : - &df->future_frame)); - install_partial_transition(super, answer, tc->cset[offset], - offset * 32); - return answer; - } - /* Otherwise, it's a flushed or newly encountered edge. */ - { - char cset_space[1024]; /* this limit is far from unreasonable */ - rx_Bitset trcset; - struct rx_inx *answer; - - if (rx_sizeof_bitset(rx->local_cset_size) > sizeof(cset_space)) - return 0; /* If the arbitrary limit is hit, always fail */ - /* cleanly. */ - trcset = (rx_Bitset) cset_space; - rx_lock_superstate(rx, super); - if (!compute_super_edge(rx, &df, trcset, super, chr)) { - rx_unlock_superstate(rx, super); - return 0; - } - if (!df) { /* We just computed the fail transition. */ - static struct rx_inx - shared_fail_frame = { 0, 0, (void *) rx_backtrack, 0 }; - - answer = &shared_fail_frame; - } else { - tc = rx_super_edge(rx, super, trcset, df); - if (!tc) { - rx_unlock_superstate(rx, super); - return 0; - } - answer = - ((tc->options->next_same_super_edge[0] != - tc->options) ? &tc-> - rx_backtrack_frame : (df->effects ? - &df->side_effects_frame : - &df->future_frame)); - } - install_partial_transition(super, answer, - trcset[offset], offset * 32); - rx_unlock_superstate(rx, super); - return answer; - } - } else if (df->future) { /* A cache miss on an edge with a future? Must be - * a semi-free destination. */ - if (df->future->is_semifree) - refresh_semifree_superstate(rx->cache, df->future); - return &df->future_frame; - } else - /* no future superstate on an existing edge */ - { - rx_lock_superstate(rx, super); - if (!solve_destination(rx, df)) { - rx_unlock_superstate(rx, super); - return 0; - } - if (!df->effects - && (df->edge->options->next_same_super_edge[0] == - df->edge->options)) install_partial_transition(super, - &df->future_frame, - df-> - edge->cset - [offset], - offset * - 32); - rx_unlock_superstate(rx, super); - return &df->future_frame; - } -} - - - - -/* The rest of the code provides a regex.c compatable interface. */ - - -__const__ char *re_error_msg[] = { - 0, /* REG_NOUT */ - "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 */ -}; - - - -/* - * Macros used while compiling patterns. - * - * By convention, PEND points just past the end of the uncompiled pattern, - * P points to the read position in the pattern. `translate' is the name - * of the translation table (`TRANSLATE' is the name of a macro that looks - * things up in `translate'). - */ - - -/* - * 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++; \ - 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-- - - -#define TRANSLATE(d) translate[(unsigned char) (d)] - -typedef unsigned regnum_t; - -/* 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 { - struct rexp_node **top_expression; /* was begalt */ - struct rexp_node **last_expression; /* was laststart */ - pattern_offset_t inner_group_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; - -static boolean group_in_compile_stack(compile_stack_type, regnum_t); -static reg_errcode_t -compile_range(struct re_pattern_buffer *, rx_Bitset, - __const__ char **, __const__ char *, - unsigned char *, reg_syntax_t, rx_Bitset, char *); -static void find_backrefs(char *, struct rexp_node *, - - struct re_se_params *); -static int compute_fastset(struct re_pattern_buffer *, struct rexp_node *); -static int is_anchored(struct rexp_node *, rx_side_effect); -static struct rexp_node -*remove_unecessary_side_effects - - (struct rx *, char *, struct rexp_node *, struct re_se_params *); -static int pointless_if_repeated(struct rexp_node *, - - struct re_se_params *); -static int registers_on_stack(struct re_pattern_buffer *, - struct rexp_node *, - - int, struct re_se_params *); -static int has_any_se(struct rx *, struct rexp_node *); -static int has_non_idempotent_epsilon_path - - (struct rx *, struct rexp_node *, struct re_se_params *); -static int begins_with_complex_se(struct rx *, struct rexp_node *); -static void speed_up_alt(struct rx *, struct rexp_node *, int); -RX_DECL reg_errcode_t - -rx_compile(__const__ char *, int, reg_syntax_t, - struct re_pattern_buffer *); -RX_DECL void rx_blow_up_fastmap(struct re_pattern_buffer *); -static __inline__ enum rx_get_burst_return -re_search_2_get_burst(struct rx_string_position *, void *, int); -static __inline__ enum rx_back_check_return -re_search_2_back_check(struct rx_string_position *, int, - int, unsigned char *, void *, int); -static __inline__ int -re_search_2_fetch_char(struct rx_string_position *, int, void *, int); - - -#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)) / CHARBITS] \ - |= 1 << (((unsigned char) c) % CHARBITS)) - -/* 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) \ - (!strcmp (string, "alpha") || !strcmp (string, "upper") \ - || !strcmp (string, "lower") || !strcmp (string, "digit") \ - || !strcmp (string, "alnum") || !strcmp (string, "xdigit") \ - || !strcmp (string, "space") || !strcmp (string, "print") \ - || !strcmp (string, "punct") || !strcmp (string, "graph") \ - || !strcmp (string, "cntrl") || !strcmp (string, "blank")) - - -/* These predicates are used in regex_compile. */ - -/* 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 ^. - */ - -#ifdef __STDC__ -static boolean -at_begline_loc_p(__const__ char *pattern, __const__ char *p, - reg_syntax_t syntax) -#else -static boolean at_begline_loc_p(pattern, p, syntax) -__const__ char *pattern; -__const__ char *p; -reg_syntax_t syntax; -#endif -{ - __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'. - */ - -#ifdef __STDC__ -static boolean -at_endline_loc_p(__const__ char *p, __const__ char *pend, int syntax) -#else -static boolean at_endline_loc_p(p, pend, syntax) -__const__ char *p; -__const__ char *pend; -int syntax; -#endif -{ - __const__ char *next = p; - boolean next_backslash = (*next == '\\'); - __const__ char *next_next = (p + 1 < pend) ? (p + 1) : 0; - - 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 == '|'))) - ); -} - - -unsigned char rx_id_translation[256] = { - 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, - 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, - 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, - 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, - 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, - 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, - 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, - 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, - 80, 81, 82, 83, 84, 85, 86, 87, 88, 89, - 90, 91, 92, 93, 94, 95, 96, 97, 98, 99, - - 100, 101, 102, 103, 104, 105, 106, 107, 108, 109, - 110, 111, 112, 113, 114, 115, 116, 117, 118, 119, - 120, 121, 122, 123, 124, 125, 126, 127, 128, 129, - 130, 131, 132, 133, 134, 135, 136, 137, 138, 139, - 140, 141, 142, 143, 144, 145, 146, 147, 148, 149, - 150, 151, 152, 153, 154, 155, 156, 157, 158, 159, - 160, 161, 162, 163, 164, 165, 166, 167, 168, 169, - 170, 171, 172, 173, 174, 175, 176, 177, 178, 179, - 180, 181, 182, 183, 184, 185, 186, 187, 188, 189, - 190, 191, 192, 193, 194, 195, 196, 197, 198, 199, - - 200, 201, 202, 203, 204, 205, 206, 207, 208, 209, - 210, 211, 212, 213, 214, 215, 216, 217, 218, 219, - 220, 221, 222, 223, 224, 225, 226, 227, 228, 229, - 230, 231, 232, 233, 234, 235, 236, 237, 238, 239, - 240, 241, 242, 243, 244, 245, 246, 247, 248, 249, - 250, 251, 252, 253, 254, 255 -}; - -/* The compiler keeps an inverted translation table. - * This looks up/inititalize elements. - * VALID is an array of booleans that validate CACHE. - */ - -#ifdef __STDC__ -static rx_Bitset -inverse_translation(struct re_pattern_buffer *rxb, - char *valid, rx_Bitset cache, - unsigned char *translate, int c) -#else -static rx_Bitset inverse_translation(rxb, valid, cache, translate, c) -struct re_pattern_buffer *rxb; -char *valid; -rx_Bitset cache; -unsigned char *translate; -int c; -#endif -{ - rx_Bitset cs - - = cache + c * rx_bitset_numb_subsets(rxb->rx.local_cset_size); - - if (!valid[c]) { - int x; - int c_tr = TRANSLATE(c); - - rx_bitset_null(rxb->rx.local_cset_size, cs); - for (x = 0; x < 256; ++x) /* &&&& 13.37 */ - if (TRANSLATE(x) == c_tr) - RX_bitset_enjoin(cs, x); - valid[c] = 1; - } - return cs; -} - - - - -/* More subroutine declarations and macros for regex_compile. */ - -/* Returns true if REGNUM is in one of COMPILE_STACK's elements and - false if it's not. */ - -#ifdef __STDC__ -static boolean -group_in_compile_stack(compile_stack_type compile_stack, regnum_t regnum) -#else -static boolean group_in_compile_stack(compile_stack, regnum) -compile_stack_type compile_stack; -regnum_t regnum; -#endif -{ - 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. - */ - -#ifdef __STDC__ -static reg_errcode_t -compile_range(struct re_pattern_buffer *rxb, rx_Bitset cs, - __const__ char **p_ptr, __const__ char *pend, - unsigned char *translate, reg_syntax_t syntax, - rx_Bitset inv_tr, char *valid_inv_tr) -#else -static reg_errcode_t -compile_range(rxb, cs, p_ptr, pend, translate, syntax, inv_tr, - valid_inv_tr) -struct re_pattern_buffer *rxb; -rx_Bitset cs; -__const__ char **p_ptr; -__const__ char *pend; -unsigned char *translate; -reg_syntax_t syntax; -rx_Bitset inv_tr; -char *valid_inv_tr; -#endif -{ - unsigned this_char; - - __const__ char *p = *p_ptr; - - unsigned char range_end; - unsigned char range_start = TRANSLATE(p[-2]); - - if (p == pend) - return REG_ERANGE; - - PATFETCH(range_end); - - (*p_ptr)++; - - if (range_start > range_end) - return syntax & RE_NO_EMPTY_RANGES ? REG_ERANGE : REG_NOERROR; - - for (this_char = range_start; this_char <= range_end; this_char++) { - rx_Bitset it = - inverse_translation(rxb, valid_inv_tr, inv_tr, translate, - - this_char); - - rx_bitset_union(rxb->rx.local_cset_size, cs, it); - } - - return REG_NOERROR; -} - - -/* This searches a regexp for backreference side effects. - * It fills in the array OUT with 1 at the index of every register pair - * referenced by a backreference. - * - * This is used to help optimize patterns for searching. The information is - * useful because, if the caller doesn't want register values, backreferenced - * registers are the only registers for which we need rx_backtrack. - */ - -#ifdef __STDC__ -static void -find_backrefs(char *out, struct rexp_node *rexp, - struct re_se_params *params) -#else -static void find_backrefs(out, rexp, params) -char *out; -struct rexp_node *rexp; -struct re_se_params *params; -#endif -{ - if (rexp) - switch (rexp->type) { - case r_cset: - case r_data: - return; - case r_alternate: - case r_concat: - case r_opt: - case r_star: - case r_2phase_star: - find_backrefs(out, rexp->params.pair.left, params); - find_backrefs(out, rexp->params.pair.right, params); - return; - case r_side_effect: - if (((long) rexp->params.side_effect >= 0) - && (params[(long) rexp->params.side_effect].se == - re_se_backref)) - out[params[(long) rexp->params.side_effect].op1] = 1; - return; - } -} - - - -/* Returns 0 unless the pattern can match the empty string. */ - -#ifdef __STDC__ -static int -compute_fastset(struct re_pattern_buffer *rxb, struct rexp_node *rexp) -#else -static int compute_fastset(rxb, rexp) -struct re_pattern_buffer *rxb; -struct rexp_node *rexp; -#endif -{ - if (!rexp) - return 1; - switch (rexp->type) { - case r_data: - return 1; - case r_cset: - { - rx_bitset_union(rxb->rx.local_cset_size, - rxb->fastset, rexp->params.cset); - } - return 0; - case r_concat: - return (compute_fastset(rxb, rexp->params.pair.left) - && compute_fastset(rxb, rexp->params.pair.right)); - case r_2phase_star: - compute_fastset(rxb, rexp->params.pair.left); - /* compute_fastset (rxb, rexp->params.pair.right); nope... */ - return 1; - case r_alternate: - return !!(compute_fastset(rxb, rexp->params.pair.left) - + compute_fastset(rxb, rexp->params.pair.right)); - case r_opt: - case r_star: - compute_fastset(rxb, rexp->params.pair.left); - return 1; - case r_side_effect: - return 1; - } - - /* this should never happen */ - return 0; -} - - -/* returns - * 1 -- yes, definately anchored by the given side effect. - * 2 -- maybe anchored, maybe the empty string. - * 0 -- definately not anchored - * There is simply no other possibility. - */ - -#ifdef __STDC__ -static int is_anchored(struct rexp_node *rexp, rx_side_effect se) -#else -static int is_anchored(rexp, se) -struct rexp_node *rexp; -rx_side_effect se; -#endif -{ - if (!rexp) - return 2; - switch (rexp->type) { - case r_cset: - case r_data: - return 0; - case r_concat: - case r_2phase_star: - { - int l = is_anchored(rexp->params.pair.left, se); - - return (l == 2 ? is_anchored(rexp->params.pair.right, se) : l); - } - case r_alternate: - { - int l = is_anchored(rexp->params.pair.left, se); - int r = l ? is_anchored(rexp->params.pair.right, se) : 0; - - if (l == r) - return l; - else if ((l == 0) || (r == 0)) - return 0; - else - return 2; - } - case r_opt: - case r_star: - return is_anchored(rexp->params.pair.left, se) ? 2 : 0; - - case r_side_effect: - return ((rexp->params.side_effect == se) - ? 1 : 2); - } - - /* this should never happen */ - return 0; -} - - -/* This removes register assignments that aren't required by backreferencing. - * This can speed up explore_future, especially if it eliminates - * non-determinism in the superstate NFA. - * - * NEEDED is an array of characters, presumably filled in by FIND_BACKREFS. - * The non-zero elements of the array indicate which register assignments - * can NOT be removed from the expression. - */ - -#ifdef __STDC__ -static struct rexp_node *remove_unecessary_side_effects(struct rx *rx, - char *needed, - struct rexp_node - *rexp, - struct re_se_params - *params) -#else -static struct rexp_node *remove_unecessary_side_effects(rx, needed, rexp, - params) -struct rx *rx; -char *needed; -struct rexp_node *rexp; -struct re_se_params *params; -#endif -{ - struct rexp_node *l; - struct rexp_node *r; - - if (!rexp) - return 0; - else - switch (rexp->type) { - case r_cset: - case r_data: - return rexp; - case r_alternate: - case r_concat: - case r_2phase_star: - l = remove_unecessary_side_effects(rx, needed, - rexp->params.pair.left, - params); - r = - remove_unecessary_side_effects(rx, needed, - rexp->params.pair.right, - params); - if ((l && r) || (rexp->type != r_concat)) { - rexp->params.pair.left = l; - rexp->params.pair.right = r; - return rexp; - } else { - rexp->params.pair.left = rexp->params.pair.right = 0; - rx_free_rexp(rx, rexp); - return l ? l : r; - } - case r_opt: - case r_star: - l = remove_unecessary_side_effects(rx, needed, - rexp->params.pair.left, - params); - if (l) { - rexp->params.pair.left = l; - return rexp; - } else { - rexp->params.pair.left = 0; - rx_free_rexp(rx, rexp); - return 0; - } - case r_side_effect: - { - int se = (long) rexp->params.side_effect; - - if ((se >= 0) - && (((enum re_side_effects) params[se].se == re_se_lparen) - || ((enum re_side_effects) params[se].se == - re_se_rparen)) && (params[se].op1 > 0) - && (!needed[params[se].op1])) { - rx_free_rexp(rx, rexp); - return 0; - } else - return rexp; - } - } - - /* this should never happen */ - return 0; -} - - - -#ifdef __STDC__ -static int -pointless_if_repeated(struct rexp_node *node, struct re_se_params *params) -#else -static int pointless_if_repeated(node, params) -struct rexp_node *node; -struct re_se_params *params; -#endif -{ - if (!node) - return 1; - switch (node->type) { - case r_cset: - return 0; - case r_alternate: - case r_concat: - case r_2phase_star: - return (pointless_if_repeated(node->params.pair.left, params) - && pointless_if_repeated(node->params.pair.right, params)); - case r_opt: - case r_star: - return pointless_if_repeated(node->params.pair.left, params); - case r_side_effect: - switch (((long) node->params.side_effect < 0) - ? (enum re_side_effects) node->params.side_effect - : (enum re_side_effects) params[(long) node-> - params.side_effect].se) { - case re_se_try: - case re_se_at_dot: - case re_se_begbuf: - case re_se_hat: - case re_se_wordbeg: - case re_se_wordbound: - case re_se_notwordbound: - case re_se_wordend: - case re_se_endbuf: - case re_se_dollar: - case re_se_fail: - case re_se_win: - return 1; - case re_se_lparen: - case re_se_rparen: - case re_se_iter: - case re_se_end_iter: - case re_se_syntax: - case re_se_not_syntax: - case re_se_backref: - return 0; - } - case r_data: - default: - return 0; - } -} - - - -#ifdef __STDC__ -static int -registers_on_stack(struct re_pattern_buffer *rxb, - struct rexp_node *rexp, int in_danger, - struct re_se_params *params) -#else -static int registers_on_stack(rxb, rexp, in_danger, params) -struct re_pattern_buffer *rxb; -struct rexp_node *rexp; -int in_danger; -struct re_se_params *params; -#endif -{ - if (!rexp) - return 0; - else - switch (rexp->type) { - case r_cset: - case r_data: - return 0; - case r_alternate: - case r_concat: - return (registers_on_stack(rxb, rexp->params.pair.left, - in_danger, params) - || (registers_on_stack - (rxb, rexp->params.pair.right, - in_danger, params))); - case r_opt: - return registers_on_stack(rxb, rexp->params.pair.left, 0, - params); - case r_star: - return registers_on_stack(rxb, rexp->params.pair.left, 1, - params); - case r_2phase_star: - return - (registers_on_stack(rxb, rexp->params.pair.left, 1, params) - || registers_on_stack(rxb, rexp->params.pair.right, 1, - params)); - case r_side_effect: - { - int se = (long) rexp->params.side_effect; - - if (in_danger && (se >= 0) - && (params[se].op1 > 0) - && (((enum re_side_effects) params[se].se == re_se_lparen) - || ((enum re_side_effects) params[se].se == - re_se_rparen))) return 1; - else - return 0; - } - } - - /* this should never happen */ - return 0; -} - - - -static char idempotent_complex_se[] = { -#define RX_WANT_SE_DEFS 1 -#undef RX_DEF_SE -#undef RX_DEF_CPLX_SE -#define RX_DEF_SE(IDEM, NAME, VALUE) -#define RX_DEF_CPLX_SE(IDEM, NAME, VALUE) IDEM, -#include -#undef RX_DEF_SE -#undef RX_DEF_CPLX_SE -#undef RX_WANT_SE_DEFS - 23 -}; - -static char idempotent_se[] = { - 13, -#define RX_WANT_SE_DEFS 1 -#undef RX_DEF_SE -#undef RX_DEF_CPLX_SE -#define RX_DEF_SE(IDEM, NAME, VALUE) IDEM, -#define RX_DEF_CPLX_SE(IDEM, NAME, VALUE) -#include -#undef RX_DEF_SE -#undef RX_DEF_CPLX_SE -#undef RX_WANT_SE_DEFS - 42 -}; - - - -#ifdef __STDC__ -static int has_any_se(struct rx *rx, struct rexp_node *rexp) -#else -static int has_any_se(rx, rexp) -struct rx *rx; -struct rexp_node *rexp; -#endif -{ - if (!rexp) - return 0; - - switch (rexp->type) { - case r_cset: - case r_data: - return 0; - - case r_side_effect: - return 1; - - case r_2phase_star: - case r_concat: - case r_alternate: - return (has_any_se(rx, rexp->params.pair.left) - || has_any_se(rx, rexp->params.pair.right)); - - case r_opt: - case r_star: - return has_any_se(rx, rexp->params.pair.left); - } - - /* this should never happen */ - return 0; -} - - - -/* This must be called AFTER `convert_hard_loops' for a given REXP. */ -#ifdef __STDC__ -static int -has_non_idempotent_epsilon_path(struct rx *rx, - struct rexp_node *rexp, - struct re_se_params *params) -#else -static int has_non_idempotent_epsilon_path(rx, rexp, params) -struct rx *rx; -struct rexp_node *rexp; -struct re_se_params *params; -#endif -{ - if (!rexp) - return 0; - - switch (rexp->type) { - case r_cset: - case r_data: - case r_star: - return 0; - - case r_side_effect: - return - !((long) rexp->params.side_effect > 0 - ? - idempotent_complex_se[params - [(long) rexp->params. - side_effect].se] : - idempotent_se[-(long) rexp->params.side_effect]); - - case r_alternate: - return - (has_non_idempotent_epsilon_path(rx, - rexp->params.pair.left, - params) - || has_non_idempotent_epsilon_path(rx, - rexp->params.pair.right, - params)); - - case r_2phase_star: - case r_concat: - return - (has_non_idempotent_epsilon_path(rx, - rexp->params.pair.left, - params) - && has_non_idempotent_epsilon_path(rx, - rexp->params.pair.right, - params)); - - case r_opt: - return has_non_idempotent_epsilon_path(rx, - rexp->params.pair.left, - params); - } - - /* this should never happen */ - return 0; -} - - - -/* This computes rougly what it's name suggests. It can (and does) go wrong - * in the direction of returning spurious 0 without causing disasters. - */ -#ifdef __STDC__ -static int begins_with_complex_se(struct rx *rx, struct rexp_node *rexp) -#else -static int begins_with_complex_se(rx, rexp) -struct rx *rx; -struct rexp_node *rexp; -#endif -{ - if (!rexp) - return 0; - - switch (rexp->type) { - case r_cset: - case r_data: - return 0; - - case r_side_effect: - return ((long) rexp->params.side_effect >= 0); - - case r_alternate: - return (begins_with_complex_se(rx, rexp->params.pair.left) - && begins_with_complex_se(rx, rexp->params.pair.right)); - - - case r_concat: - return has_any_se(rx, rexp->params.pair.left); - case r_opt: - case r_star: - case r_2phase_star: - return 0; - } - - /* this should never happen */ - return 0; -} - - -/* This destructively removes some of the re_se_tv side effects from - * a rexp tree. In particular, during parsing re_se_tv was inserted on the - * right half of every | to guarantee that posix path preference could be - * honored. This function removes some which it can be determined aren't - * needed. - */ - -#ifdef __STDC__ -static void -speed_up_alt(struct rx *rx, struct rexp_node *rexp, int unposix) -#else -static void speed_up_alt(rx, rexp, unposix) -struct rx *rx; -struct rexp_node *rexp; -int unposix; -#endif -{ - if (!rexp) - return; - - switch (rexp->type) { - case r_cset: - case r_data: - case r_side_effect: - return; - - case r_opt: - case r_star: - speed_up_alt(rx, rexp->params.pair.left, unposix); - return; - - case r_2phase_star: - case r_concat: - speed_up_alt(rx, rexp->params.pair.left, unposix); - speed_up_alt(rx, rexp->params.pair.right, unposix); - return; - - case r_alternate: - /* the right child is guaranteed to be (concat re_se_tv ) */ - - speed_up_alt(rx, rexp->params.pair.left, unposix); - speed_up_alt(rx, rexp->params.pair.right->params.pair.right, - unposix); - - if (unposix - || (begins_with_complex_se - (rx, rexp->params.pair.right->params.pair.right)) - || !(has_any_se(rx, rexp->params.pair.right->params.pair.right) - || has_any_se(rx, rexp->params.pair.left))) { - struct rexp_node *conc = rexp->params.pair.right; - - rexp->params.pair.right = conc->params.pair.right; - conc->params.pair.right = 0; - rx_free_rexp(rx, conc); - } - } -} - - - - - -/* `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 set to zero; - `re_nsub' is set to the number of groups in PATTERN; - `not_bol' and `not_eol' are set to zero. - - The `fastmap' and `newline_anchor' fields are neither - examined nor set. */ - - -#ifdef __STDC__ -RX_DECL reg_errcode_t -rx_compile(__const__ char *pattern, int size, - reg_syntax_t syntax, struct re_pattern_buffer *rxb) -#else -RX_DECL reg_errcode_t rx_compile(pattern, size, syntax, rxb) -__const__ char *pattern; -int size; -reg_syntax_t syntax; -struct re_pattern_buffer *rxb; -#endif -{ - RX_subset - inverse_translate[CHAR_SET_SIZE * - rx_bitset_numb_subsets(CHAR_SET_SIZE)]; - char validate_inv_tr[CHAR_SET_SIZE * - - rx_bitset_numb_subsets(CHAR_SET_SIZE)]; - - /* 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 tempory spot in PATTERN. */ - __const__ char *p1; - - /* 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. */ - unsigned char *translate = (rxb->translate - - ? rxb->translate : rx_id_translation); - - /* When parsing is done, this will hold the expression tree. */ - struct rexp_node *rexp = 0; - - /* In the midst of compilation, this holds onto the regexp - * first parst while rexp goes on to aquire additional constructs. - */ - struct rexp_node *orig_rexp = 0; - struct rexp_node *fewer_side_effects = 0; - - /* This and top_expression are saved on the compile stack. */ - struct rexp_node **top_expression = &rexp; - struct rexp_node **last_expression = top_expression; - - /* Parameter to `goto append_node' */ - struct rexp_node *append; - - /* Counts open-groups as they are encountered. This is the index of the - * innermost group being compiled. - */ - regnum_t regnum = 0; - - /* Place in the uncompiled pattern (i.e., the {) to - * which to go back if the interval is invalid. - */ - __const__ char *beg_interval; - - struct re_se_params *params = 0; - int paramc = 0; /* How many complex side effects so far? */ - - rx_side_effect side; /* param to `goto add_side_effect' */ - - bzero(validate_inv_tr, sizeof(validate_inv_tr)); - - rxb->rx.instruction_table = rx_id_instruction_table; - - - /* Initialize the compile stack. */ - compile_stack.stack = ((compile_stack_elt_t *) - malloc((INIT_COMPILE_STACK_SIZE) * - sizeof(compile_stack_elt_t))); - if (compile_stack.stack == 0) - return REG_ESPACE; - - compile_stack.size = INIT_COMPILE_STACK_SIZE; - compile_stack.avail = 0; - - /* Initialize the pattern buffer. */ - rxb->rx.cache = &default_cache; - rxb->syntax = syntax; - rxb->fastmap_accurate = 0; - rxb->not_bol = rxb->not_eol = 0; - rxb->least_subs = 0; - - /* Always count groups, whether or not rxb->no_sub is set. - * The whole pattern is implicitly group 0, so counting begins - * with 1. - */ - rxb->re_nsub = 0; - -#if !defined (emacs) && !defined (SYNTAX_TABLE) - /* Initialize the syntax table. */ - init_syntax_once(); -#endif - - /* 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)) { - struct rexp_node *n = rx_mk_r_side_effect(&rxb->rx, - (rx_side_effect) re_se_hat); - - if (!n) - return REG_ESPACE; - append = n; - goto append_node; - } 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)) { - struct rexp_node *n = rx_mk_r_side_effect(&rxb->rx, - (rx_side_effect) re_se_dollar); - - if (!n) - return REG_ESPACE; - append = n; - goto append_node; - } 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 (pointless_if_repeated(*last_expression, params)) { - if (syntax & RE_CONTEXT_INVALID_OPS) - return REG_BADRPT; - else if (!(syntax & RE_CONTEXT_INDEP_OPS)) - goto normal_char; - } - - { - /* 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) - 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 (!last_expression) - break; - - /* Now we know whether or not zero matches is allowed - * and also whether or not two or more matches is allowed. - */ - - { - struct rexp_node *inner_exp = *last_expression; - int need_sync = 0; - - if (many_times_ok - && has_non_idempotent_epsilon_path(&rxb->rx, - inner_exp, - params)) { - struct rexp_node *pusher = - rx_mk_r_side_effect(&rxb->rx, - (rx_side_effect) re_se_pushpos); - struct rexp_node *checker - = rx_mk_r_side_effect(&rxb->rx, - (rx_side_effect) re_se_chkpos); - struct rexp_node *pushback - = rx_mk_r_side_effect(&rxb->rx, - (rx_side_effect) re_se_pushback); - rx_Bitset cs = rx_cset(&rxb->rx); - struct rexp_node *lit_t; - struct rexp_node *fake_state; - struct rexp_node *phase2; - struct rexp_node *popper; - struct rexp_node *star; - struct rexp_node *a; - struct rexp_node *whole_thing; - - if (!cs) - return REG_ESPACE; - lit_t = rx_mk_r_cset(&rxb->rx, cs); - fake_state = - rx_mk_r_concat(&rxb->rx, pushback, lit_t); - phase2 = - rx_mk_r_concat(&rxb->rx, checker, fake_state); - popper = - rx_mk_r_side_effect(&rxb->rx, - (rx_side_effect) - re_se_poppos); - star = - rx_mk_r_2phase_star(&rxb->rx, inner_exp, - phase2); - a = rx_mk_r_concat(&rxb->rx, pusher, star); - whole_thing = rx_mk_r_concat(&rxb->rx, a, popper); - - if (! - (pusher && star && pushback && lit_t - && fake_state && lit_t && phase2 && checker - && popper && a && whole_thing)) - return REG_ESPACE; - RX_bitset_enjoin(cs, 't'); - *last_expression = whole_thing; - } else { - struct rexp_node *star = - (many_times_ok ? rx_mk_r_star : rx_mk_r_opt) - (&rxb->rx, *last_expression); - - if (!star) - return REG_ESPACE; - *last_expression = star; - need_sync = has_any_se(&rxb->rx, *last_expression); - } - if (!zero_times_ok) { - struct rexp_node *concat - = rx_mk_r_concat(&rxb->rx, inner_exp, - rx_copy_rexp(&rxb->rx, - *last_expression)); - - if (!concat) - return REG_ESPACE; - *last_expression = concat; - } - if (need_sync) { - int sync_se = paramc; - - params = (params ? ((struct re_se_params *) - realloc(params, - sizeof(*params) * (1 + - paramc))) - : ((struct re_se_params *) - malloc(sizeof(*params)))); - if (!params) - return REG_ESPACE; - ++paramc; - params[sync_se].se = re_se_tv; - side = (rx_side_effect) sync_se; - goto add_side_effect; - } - } - /* The old regex.c used to optimize `.*\n'. - * Maybe rx should too? - */ - } - break; - - - case '.': - { - rx_Bitset cs = rx_cset(&rxb->rx); - struct rexp_node *n = rx_mk_r_cset(&rxb->rx, cs); - - if (!(cs && n)) - return REG_ESPACE; - - rx_bitset_universe(rxb->rx.local_cset_size, cs); - if (!(rxb->syntax & RE_DOT_NEWLINE)) - RX_bitset_remove(cs, '\n'); - if (!(rxb->syntax & RE_DOT_NOT_NULL)) - RX_bitset_remove(cs, 0); - - append = n; - goto append_node; - break; - } - - - case '[': - if (p == pend) - return REG_EBRACK; - { - boolean had_char_class = false; - rx_Bitset cs = rx_cset(&rxb->rx); - struct rexp_node *node = rx_mk_r_cset(&rxb->rx, cs); - int is_inverted = *p == '^'; - - if (!(node && cs)) - return REG_ESPACE; - - /* This branch of the switch is normally exited with - *`goto append_node' - */ - append = node; - - if (is_inverted) - p++; - - /* Remember the first position in the bracket expression. */ - p1 = p; - - /* Read in characters and ranges, setting map bits. */ - for (;;) { - if (p == pend) - return REG_EBRACK; - - PATFETCH(c); - - /* \ might escape characters inside [...] and [^...]. */ - if ((syntax & RE_BACKSLASH_ESCAPE_IN_LISTS) - && c == '\\') { - if (p == pend) - return REG_EESCAPE; - - PATFETCH(c1); - { - rx_Bitset it = inverse_translation(rxb, - validate_inv_tr, - inverse_translate, - translate, - c1); - - rx_bitset_union(rxb->rx.local_cset_size, cs, - it); - } - 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) - goto finalize_class_and_append; - - /* Look ahead to see if it's a range when the last thing - was a character class. */ - if (had_char_class && c == '-' && *p != ']') - 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(rxb, cs, &p, pend, translate, - syntax, - inverse_translate, - - validate_inv_tr); - - if (ret != REG_NOERROR) - 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(rxb, cs, &p, pend, translate, - syntax, inverse_translate, - validate_inv_tr); - if (ret != REG_NOERROR) - return ret; - } - - /* See if we're at the beginning of a possible character - class. */ - - else if ((syntax & RE_CHAR_CLASSES) - && (c == '[') && (*p == ':')) { - char str[CHAR_CLASS_MAX_LENGTH + 1]; - - PATFETCH(c); - c1 = 0; - - /* If pattern is `[[:'. */ - if (p == pend) - 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 = !strcmp(str, "alnum"); - boolean is_alpha = !strcmp(str, "alpha"); - boolean is_blank = !strcmp(str, "blank"); - boolean is_cntrl = !strcmp(str, "cntrl"); - boolean is_digit = !strcmp(str, "digit"); - boolean is_graph = !strcmp(str, "graph"); - boolean is_lower = !strcmp(str, "lower"); - boolean is_print = !strcmp(str, "print"); - boolean is_punct = !strcmp(str, "punct"); - boolean is_space = !strcmp(str, "space"); - boolean is_upper = !strcmp(str, "upper"); - boolean is_xdigit = !strcmp(str, "xdigit"); - - if (!IS_CHAR_CLASS(str)) - return REG_ECTYPE; - - /* Throw away the ] at the end of the character - class. */ - PATFETCH(c); - - if (p == pend) - return REG_EBRACK; - - for (ch = 0; ch < 1 << CHARBITS; ch++) { - if ((is_alnum && isalnum(ch)) - || (is_alpha && isalpha(ch)) - || (is_blank && isblank(ch)) - || (is_cntrl && iscntrl(ch)) - || (is_digit && isdigit(ch)) - || (is_graph && isgraph(ch)) - || (is_lower && islower(ch)) - || (is_print && isprint(ch)) - || (is_punct && ispunct(ch)) - || (is_space && isspace(ch)) - || (is_upper && isupper(ch)) - || (is_xdigit && isxdigit(ch))) { - rx_Bitset it = inverse_translation(rxb, - validate_inv_tr, - inverse_translate, - translate, - ch); - - rx_bitset_union(rxb-> - rx.local_cset_size, cs, - it); - } - } - had_char_class = true; - } else { - c1++; - while (c1--) - PATUNFETCH; - { - rx_Bitset it = inverse_translation(rxb, - validate_inv_tr, - inverse_translate, - translate, - '['); - - rx_bitset_union(rxb->rx.local_cset_size, - cs, it); - } - { - rx_Bitset it = inverse_translation(rxb, - validate_inv_tr, - inverse_translate, - translate, - ':'); - - rx_bitset_union(rxb->rx.local_cset_size, - cs, it); - } - had_char_class = false; - } - } else { - had_char_class = false; - { - rx_Bitset it = inverse_translation(rxb, - validate_inv_tr, - inverse_translate, - translate, - c); - - rx_bitset_union(rxb->rx.local_cset_size, cs, - it); - } - } - } - - finalize_class_and_append: - if (is_inverted) { - rx_bitset_complement(rxb->rx.local_cset_size, cs); - if (syntax & RE_HAT_LISTS_NOT_NEWLINE) - RX_bitset_remove(cs, '\n'); - } - goto append_node; - } - 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) - 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: - rxb->re_nsub++; - regnum++; - if (COMPILE_STACK_FULL) { - ((compile_stack.stack) = - (compile_stack_elt_t *) realloc(compile_stack.stack, - (compile_stack.size << - 1) * - sizeof - (compile_stack_elt_t))); - if (compile_stack.stack == 0) - return REG_ESPACE; - - compile_stack.size <<= 1; - } - - if (*last_expression) { - struct rexp_node *concat - = rx_mk_r_concat(&rxb->rx, *last_expression, 0); - - if (!concat) - return REG_ESPACE; - *last_expression = concat; - last_expression = &concat->params.pair.right; - } - - /* - * These are the values to restore when we hit end of this - * group. - */ - COMPILE_STACK_TOP.top_expression = top_expression; - COMPILE_STACK_TOP.last_expression = last_expression; - COMPILE_STACK_TOP.regnum = regnum; - - compile_stack.avail++; - - top_expression = last_expression; - break; - - - case ')': - if (syntax & RE_NO_BK_PARENS) - goto normal_backslash; - - handle_close: - /* See similar code for backslashed left paren above. */ - if (COMPILE_STACK_EMPTY) { - if (syntax & RE_UNMATCHED_RIGHT_PAREN_ORD) { - goto normal_char; - } else { - return REG_ERPAREN; - } - } - - /* Since we just checked for an empty stack above, this - ``can't happen''. */ - - { - /* 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; - struct rexp_node **inner = top_expression; - - compile_stack.avail--; - top_expression = COMPILE_STACK_TOP.top_expression; - last_expression = COMPILE_STACK_TOP.last_expression; - this_group_regnum = COMPILE_STACK_TOP.regnum; - { - int left_se = paramc; - int right_se = paramc + 1; - - params = (params ? ((struct re_se_params *) - realloc(params, - (paramc + - 2) * - sizeof(params[0]))) - : ((struct re_se_params *) - malloc(2 * sizeof(params[0])))); - if (!params) - return REG_ESPACE; - paramc += 2; - - params[left_se].se = re_se_lparen; - params[left_se].op1 = this_group_regnum; - params[right_se].se = re_se_rparen; - params[right_se].op1 = this_group_regnum; - { - struct rexp_node *left - = rx_mk_r_side_effect(&rxb->rx, - (rx_side_effect) left_se); - struct rexp_node *right - = rx_mk_r_side_effect(&rxb->rx, - (rx_side_effect) right_se); - struct rexp_node *c1 - = (*inner ? rx_mk_r_concat(&rxb->rx, left, - *inner) : left); - struct rexp_node *c2 = - rx_mk_r_concat(&rxb->rx, c1, right); - - if (!(left && right && c1 && c2)) - return REG_ESPACE; - *inner = c2; - } - } - 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; - - { - struct rexp_node *alt - = rx_mk_r_alternate(&rxb->rx, *top_expression, 0); - - if (!alt) - return REG_ESPACE; - *top_expression = alt; - last_expression = &alt->params.pair.right; - { - int sync_se = paramc; - - params = (params ? ((struct re_se_params *) - realloc(params, - (paramc + - 1) * - sizeof(params[0]))) - : ((struct re_se_params *) - malloc(sizeof(params[0])))); - if (!params) - return REG_ESPACE; - ++paramc; - - params[sync_se].se = re_se_tv; - { - struct rexp_node *sync - = rx_mk_r_side_effect(&rxb->rx, - (rx_side_effect) sync_se); - struct rexp_node *conc - = rx_mk_r_concat(&rxb->rx, sync, 0); - - if (!sync || !conc) - return REG_ESPACE; - - *last_expression = conc; - last_expression = &conc->params.pair.right; - } - } - } - 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 - 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 - return REG_BADBR; - } - - if (!(syntax & RE_NO_BK_BRACES)) { - if (c != '\\') - return REG_EBRACE; - PATFETCH(c); - } - - if (c != '}') { - if (syntax & RE_NO_BK_BRACES) - goto unfetch_interval; - else - return REG_BADBR; - } - - /* We just parsed a valid interval. */ - - /* If it's invalid to have no preceding re. */ - if (pointless_if_repeated(*last_expression, params)) { - if (syntax & RE_CONTEXT_INVALID_OPS) - return REG_BADRPT; - else if (!(syntax & RE_CONTEXT_INDEP_OPS)) - goto unfetch_interval; - /* was: else laststart = b; */ - } - - /* If the upper bound is zero, don't want to iterate - * at all. - */ - if (upper_bound == 0) { - if (*last_expression) { - rx_free_rexp(&rxb->rx, *last_expression); - *last_expression = 0; - } - } else - /* Otherwise, we have a nontrivial interval. */ - { - int iter_se = paramc; - int end_se = paramc + 1; - - params = (params ? ((struct re_se_params *) - realloc(params, - sizeof(*params) * (2 + - paramc))) - : ((struct re_se_params *) - malloc(2 * sizeof(*params)))); - if (!params) - return REG_ESPACE; - paramc += 2; - params[iter_se].se = re_se_iter; - params[iter_se].op1 = lower_bound; - params[iter_se].op2 = upper_bound; - - params[end_se].se = re_se_end_iter; - params[end_se].op1 = lower_bound; - params[end_se].op2 = upper_bound; - { - struct rexp_node *push0 - = rx_mk_r_side_effect(&rxb->rx, - (rx_side_effect) re_se_push0); - struct rexp_node *start_one_iter - = rx_mk_r_side_effect(&rxb->rx, - (rx_side_effect) iter_se); - struct rexp_node *phase1 - = rx_mk_r_concat(&rxb->rx, start_one_iter, - *last_expression); - struct rexp_node *pushback - = rx_mk_r_side_effect(&rxb->rx, - (rx_side_effect) re_se_pushback); - rx_Bitset cs = rx_cset(&rxb->rx); - struct rexp_node *lit_t; - struct rexp_node *phase2; - struct rexp_node *loop; - struct rexp_node *push_n_loop; - struct rexp_node *final_test; - struct rexp_node *full_exp; - - if (!cs) - return REG_ESPACE; - lit_t = rx_mk_r_cset(&rxb->rx, cs); - phase2 = - rx_mk_r_concat(&rxb->rx, pushback, lit_t); - loop = - rx_mk_r_2phase_star(&rxb->rx, phase1, - phase2); - push_n_loop = - rx_mk_r_concat(&rxb->rx, push0, loop); - final_test = - rx_mk_r_side_effect(&rxb->rx, - (rx_side_effect) - end_se); - full_exp = - rx_mk_r_concat(&rxb->rx, push_n_loop, - final_test); - - if (!(push0 && start_one_iter && phase1 - && pushback && lit_t && phase2 - && loop && push_n_loop && final_test - && full_exp)) return REG_ESPACE; - - RX_bitset_enjoin(cs, 't'); - - *last_expression = full_exp; - } - } - beg_interval = 0; - } - break; - - unfetch_interval: - /* If an invalid interval, match the characters as literals. */ - p = beg_interval; - beg_interval = 0; - - /* 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 '=': - side = (rx_side_effect) rx_se_at_dot; - goto add_side_effect; - break; - - case 's': - case 'S': - { - rx_Bitset cs = rx_cset(&rxb->rx); - struct rexp_node *set = rx_mk_r_cset(&rxb->rx, cs); - - if (!(cs && set)) - return REG_ESPACE; - if (c == 'S') - rx_bitset_universe(rxb->rx.local_cset_size, cs); - - PATFETCH(c); - { - int x; - enum syntaxcode code = syntax_spec_code[c]; - - for (x = 0; x < 256; ++x) { - - if (SYNTAX(x) == code) { - rx_Bitset it = - inverse_translation(rxb, validate_inv_tr, - inverse_translate, - translate, x); - - rx_bitset_xor(rxb->rx.local_cset_size, cs, it); - } - } - } - append = set; - goto append_node; - } - break; -#endif /* emacs */ - - - case 'w': - case 'W': - if (syntax & RE_NO_GNU_OPS) - goto normal_char; - { - rx_Bitset cs = rx_cset(&rxb->rx); - struct rexp_node *n = - (cs ? rx_mk_r_cset(&rxb->rx, cs) : 0); - - if (!(cs && n)) - return REG_ESPACE; - if (c == 'W') - rx_bitset_universe(rxb->rx.local_cset_size, cs); - { - int x; - - for (x = rxb->rx.local_cset_size - 1; x > 0; --x) - if (SYNTAX(x) & Sword) - RX_bitset_toggle(cs, x); - } - append = n; - goto append_node; - } - break; - -/* With a little extra work, some of these side effects could be optimized - * away (basicly by looking at what we already know about the surrounding - * chars). - */ - case '<': - if (syntax & RE_NO_GNU_OPS) - goto normal_char; - side = (rx_side_effect) re_se_wordbeg; - goto add_side_effect; - break; - - case '>': - if (syntax & RE_NO_GNU_OPS) - goto normal_char; - side = (rx_side_effect) re_se_wordend; - goto add_side_effect; - break; - - case 'b': - if (syntax & RE_NO_GNU_OPS) - goto normal_char; - side = (rx_side_effect) re_se_wordbound; - goto add_side_effect; - break; - - case 'B': - if (syntax & RE_NO_GNU_OPS) - goto normal_char; - side = (rx_side_effect) re_se_notwordbound; - goto add_side_effect; - break; - - case '`': - if (syntax & RE_NO_GNU_OPS) - goto normal_char; - side = (rx_side_effect) re_se_begbuf; - goto add_side_effect; - break; - - case '\'': - if (syntax & RE_NO_GNU_OPS) - goto normal_char; - side = (rx_side_effect) re_se_endbuf; - goto add_side_effect; - break; - - add_side_effect: - { - struct rexp_node *se - - = rx_mk_r_side_effect(&rxb->rx, side); - if (!se) - return REG_ESPACE; - append = se; - goto append_node; - } - 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) - return REG_ESUBREG; - - /* Can't back reference to a subexpression if inside of it. */ - if (group_in_compile_stack(compile_stack, c1)) - return REG_ESUBREG; - - { - int backref_se = paramc; - - params = (params ? ((struct re_se_params *) - realloc(params, - sizeof(*params) * (1 + - paramc))) - : ((struct re_se_params *) - malloc(sizeof(*params)))); - if (!params) - return REG_ESPACE; - ++paramc; - params[backref_se].se = re_se_backref; - params[backref_se].op1 = c1; - side = (rx_side_effect) backref_se; - goto add_side_effect; - } - 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: - { - rx_Bitset cs = rx_cset(&rxb->rx); - struct rexp_node *match = rx_mk_r_cset(&rxb->rx, cs); - rx_Bitset it; - - if (!(cs && match)) - return REG_ESPACE; - it = inverse_translation(rxb, validate_inv_tr, - inverse_translate, translate, c); - rx_bitset_union(CHAR_SET_SIZE, cs, it); - append = match; - - append_node: - /* This genericly appends the rexp APPEND to *LAST_EXPRESSION - * and then parses the next character normally. - */ - if (*last_expression) { - struct rexp_node *concat - = rx_mk_r_concat(&rxb->rx, *last_expression, append); - - if (!concat) - return REG_ESPACE; - *last_expression = concat; - last_expression = &concat->params.pair.right; - } else - *last_expression = append; - } - } /* switch (c) */ - } /* while p != pend */ - - - { - int win_se = paramc; - - params = (params ? ((struct re_se_params *) - realloc(params, - sizeof(*params) * (1 + paramc))) - : ((struct re_se_params *) - malloc(sizeof(*params)))); - if (!params) - return REG_ESPACE; - ++paramc; - params[win_se].se = re_se_win; - { - struct rexp_node *se - = rx_mk_r_side_effect(&rxb->rx, (rx_side_effect) win_se); - struct rexp_node *concat = rx_mk_r_concat(&rxb->rx, rexp, se); - - if (!(se && concat)) - return REG_ESPACE; - rexp = concat; - } - } - - - /* Through the pattern now. */ - - if (!COMPILE_STACK_EMPTY) - return REG_EPAREN; - - free(compile_stack.stack); - - orig_rexp = rexp; -#ifdef RX_DEBUG - if (rx_debug_compile) { - dbug_rxb = rxb; - fputs("\n\nCompiling ", stdout); - fwrite(pattern, 1, size, stdout); - fputs(":\n", stdout); - rxb->se_params = params; - print_rexp(&rxb->rx, orig_rexp, 2, re_seprint, stdout); - } -#endif - { - rx_Bitset cs = rx_cset(&rxb->rx); - rx_Bitset cs2 = rx_cset(&rxb->rx); - char *se_map = (char *) alloca(paramc); - struct rexp_node *new_rexp = 0; - - - bzero(se_map, paramc); - find_backrefs(se_map, rexp, params); - fewer_side_effects = - remove_unecessary_side_effects(&rxb->rx, se_map, - rx_copy_rexp(&rxb->rx, rexp), - params); - - speed_up_alt(&rxb->rx, rexp, 0); - speed_up_alt(&rxb->rx, fewer_side_effects, 1); - - { - char *syntax_parens = rxb->syntax_parens; - - if (syntax_parens == (char *) 0x1) - rexp = remove_unecessary_side_effects - (&rxb->rx, se_map, rexp, params); - else if (syntax_parens) { - int x; - - for (x = 0; x < paramc; ++x) - if (((params[x].se == re_se_lparen) - || (params[x].se == re_se_rparen)) - && (!syntax_parens[params[x].op1])) - se_map[x] = 1; - rexp = remove_unecessary_side_effects - (&rxb->rx, se_map, rexp, params); - } - } - - /* At least one more optimization would be nice to have here but i ran out - * of time. The idea would be to delay side effects. - * For examle, `(abc)' is the same thing as `abc()' except that the - * left paren is offset by 3 (which we know at compile time). - * (In this comment, write that second pattern `abc(:3:)' - * where `(:3:' is a syntactic unit.) - * - * Trickier: `(abc|defg)' is the same as `(abc(:3:|defg(:4:))' - * (The paren nesting may be hard to follow -- that's an alternation - * of `abc(:3:' and `defg(:4:' inside (purely syntactic) parens - * followed by the closing paren from the original expression.) - * - * Neither the expression tree representation nor the the nfa make - * this very easy to write. :( - */ - - /* What we compile is different than what the parser returns. - * Suppose the parser returns expression R. - * Let R' be R with unnecessary register assignments removed - * (see REMOVE_UNECESSARY_SIDE_EFFECTS, above). - * - * What we will compile is the expression: - * - * m{try}R{win}\|s{try}R'{win} - * - * {try} and {win} denote side effect epsilons (see EXPLORE_FUTURE). - * - * When trying a match, we insert an `m' at the beginning of the - * string if the user wants registers to be filled, `s' if not. - */ - new_rexp = - rx_mk_r_alternate - (&rxb->rx, - rx_mk_r_concat(&rxb->rx, rx_mk_r_cset(&rxb->rx, cs2), rexp), - rx_mk_r_concat(&rxb->rx, - rx_mk_r_cset(&rxb->rx, cs), - fewer_side_effects)); - - if (!(new_rexp && cs && cs2)) - return REG_ESPACE; - RX_bitset_enjoin(cs2, '\0'); /* prefixed to the rexp used for matching. */ - RX_bitset_enjoin(cs, '\1'); /* prefixed to the rexp used for searching. */ - rexp = new_rexp; - } - -#ifdef RX_DEBUG - if (rx_debug_compile) { - fputs("\n...which is compiled as:\n", stdout); - print_rexp(&rxb->rx, rexp, 2, re_seprint, stdout); - } -#endif - { - struct rx_nfa_state *start = 0; - struct rx_nfa_state *end = 0; - - if (!rx_build_nfa(&rxb->rx, rexp, &start, &end)) - return REG_ESPACE; /* */ - else { - void *mem = (void *) rxb->buffer; - unsigned long size = rxb->allocated; - int start_id; - char *perm_mem; - int iterator_size = paramc * sizeof(params[0]); - - end->is_final = 1; - start->is_start = 1; - rx_name_nfa_states(&rxb->rx); - start_id = start->id; -#ifdef RX_DEBUG - if (rx_debug_compile) { - fputs("...giving the NFA: \n", stdout); - dbug_rxb = rxb; - print_nfa(&rxb->rx, rxb->rx.nfa_states, re_seprint, - stdout); - } -#endif - if (!rx_eclose_nfa(&rxb->rx)) - return REG_ESPACE; - else { - rx_delete_epsilon_transitions(&rxb->rx); - - /* For compatability reasons, we need to shove the - * compiled nfa into one chunk of malloced memory. - */ - rxb->rx.reserved = (sizeof(params[0]) * paramc - + - rx_sizeof_bitset(rxb-> - rx.local_cset_size)); -#ifdef RX_DEBUG - if (rx_debug_compile) { - dbug_rxb = rxb; - fputs("...which cooks down (uncompactified) to: \n", - stdout); - print_nfa(&rxb->rx, rxb->rx.nfa_states, re_seprint, - stdout); - } -#endif - if (!rx_compactify_nfa(&rxb->rx, &mem, &size)) - return REG_ESPACE; - rxb->buffer = mem; - rxb->allocated = size; - rxb->rx.buffer = mem; - rxb->rx.allocated = size; - perm_mem = ((char *) rxb->rx.buffer - + rxb->rx.allocated - rxb->rx.reserved); - rxb->se_params = ((struct re_se_params *) perm_mem); - bcopy(params, rxb->se_params, iterator_size); - perm_mem += iterator_size; - rxb->fastset = (rx_Bitset) perm_mem; - rxb->start = rx_id_to_nfa_state(&rxb->rx, start_id); - } - rx_bitset_null(rxb->rx.local_cset_size, rxb->fastset); - rxb->can_match_empty = compute_fastset(rxb, orig_rexp); - rxb->match_regs_on_stack = - registers_on_stack(rxb, orig_rexp, 0, params); - rxb->search_regs_on_stack = - registers_on_stack(rxb, fewer_side_effects, 0, params); - if (rxb->can_match_empty) - rx_bitset_universe(rxb->rx.local_cset_size, rxb->fastset); - rxb->is_anchored = - is_anchored(orig_rexp, (rx_side_effect) re_se_hat); - rxb->begbuf_only = - is_anchored(orig_rexp, (rx_side_effect) re_se_begbuf); - } - rx_free_rexp(&rxb->rx, rexp); - if (params) - free(params); -#ifdef RX_DEBUG - if (rx_debug_compile) { - dbug_rxb = rxb; - fputs("...which cooks down to: \n", stdout); - print_nfa(&rxb->rx, rxb->rx.nfa_states, re_seprint, stdout); - } -#endif - } - return REG_NOERROR; -} - - - -/* 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. */ - -__const__ char *rx_error_msg[] = { 0, /* 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 */ -}; - - - - -char rx_slowmap[256] = { - 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, - 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, - 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, - 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, - 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, - 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, - 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, - 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, - 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, - 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, - 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, - 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, - 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, - 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, - 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, - 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, -}; - -#ifdef __STDC__ -RX_DECL void rx_blow_up_fastmap(struct re_pattern_buffer *rxb) -#else -RX_DECL void rx_blow_up_fastmap(rxb) -struct re_pattern_buffer *rxb; -#endif -{ - int x; - - for (x = 0; x < 256; ++x) /* &&&& 3.6 % */ - rxb->fastmap[x] = !!RX_bitset_member(rxb->fastset, x); - rxb->fastmap_accurate = 1; -} - - - - -#if !defined(REGEX_MALLOC) && !defined(__GNUC__) -#define RE_SEARCH_2_FN inner_re_search_2 -#define RE_S2_QUAL static -#else -#define RE_SEARCH_2_FN re_search_2 -#define RE_S2_QUAL -#endif - -struct re_search_2_closure { - __const__ char *string1; - int size1; - __const__ char *string2; - int size2; -}; - -RE_S2_QUAL int -RE_SEARCH_2_FN(struct re_pattern_buffer *, - __const__ char *, - int, __const__ char *, int, int, - - int, struct re_registers *, int); -int re_rx_search(struct re_pattern_buffer *, int, - int, int, int, rx_get_burst_fn, - rx_back_check_fn, rx_fetch_char_fn, - void *, struct re_registers *, - - struct rx_search_state *, struct rx_search_state *); -#if !defined(REGEX_MALLOC) && !defined(__GNUC__) -int re_search_2(struct re_pattern_buffer *, - __const__ char *, int, - __const__ char *, int, - - int, int, struct re_registers *, int); -#endif -int re_search(struct re_pattern_buffer *, - - __const__ char *, int, int, int, struct re_registers *); -int re_match_2(struct re_pattern_buffer *, - __const__ char *, int, - __const__ char *, int, int, struct re_registers *, int); -int re_match(struct re_pattern_buffer *, - - __const__ char *, int, int, struct re_registers *); -reg_syntax_t re_set_syntax(reg_syntax_t); -void re_set_registers(struct re_pattern_buffer *, - struct re_registers *, unsigned, - regoff_t *, regoff_t *); -static int cplx_se_sublist_len(struct rx_se_list *); -static int posix_se_list_order(struct rx *, struct rx_se_list *, - - struct rx_se_list *); -__const__ char -*re_compile_pattern(__const__ char *, int, struct re_pattern_buffer *); -int re_compile_fastmap(struct re_pattern_buffer *); -char *re_comp(__const__ char *); -int re_exec(__const__ char *); -int regcomp(regex_t *, __const__ char *, int); -int regexec(__const__ regex_t *, - __const__ char *, size_t, regmatch_t pmatch[], int); -size_t regerror(int, __const__ regex_t *, char *, size_t); - -#ifdef __STDC__ -static __inline__ enum rx_get_burst_return -re_search_2_get_burst(struct rx_string_position *pos, - void *vclosure, int stop) -#else -static __inline__ enum rx_get_burst_return -re_search_2_get_burst(pos, vclosure, stop) -struct rx_string_position *pos; -void *vclosure; -int stop; -#endif -{ - struct re_search_2_closure *closure; - - closure = (struct re_search_2_closure *) vclosure; - if (!closure->string2) { - int inset; - - inset = pos->pos - pos->string; - if ((inset < -1) || (inset > closure->size1)) - return rx_get_burst_no_more; - else { - pos->pos = - (__const__ unsigned char *) closure->string1 + inset; - pos->string = (__const__ unsigned char *) closure->string1; - pos->size = closure->size1; - pos->end = ((__const__ unsigned char *) - MIN(closure->string1 + closure->size1, - closure->string1 + stop)); - pos->offset = 0; - return ((pos->pos < pos->end) - ? rx_get_burst_ok : rx_get_burst_no_more); - } - } else if (!closure->string1) { - int inset; - - inset = pos->pos - pos->string; - pos->pos = (__const__ unsigned char *) closure->string2 + inset; - pos->string = (__const__ unsigned char *) closure->string2; - pos->size = closure->size2; - pos->end = ((__const__ unsigned char *) - MIN(closure->string2 + closure->size2, - closure->string2 + stop)); - pos->offset = 0; - return ((pos->pos < pos->end) - ? rx_get_burst_ok : rx_get_burst_no_more); - } else { - int inset; - - inset = pos->pos - pos->string + pos->offset; - if (inset < closure->size1) { - pos->pos = - (__const__ unsigned char *) closure->string1 + inset; - pos->string = (__const__ unsigned char *) closure->string1; - pos->size = closure->size1; - pos->end = ((__const__ unsigned char *) - MIN(closure->string1 + closure->size1, - closure->string1 + stop)); - pos->offset = 0; - return rx_get_burst_ok; - } else { - pos->pos = ((__const__ unsigned char *) - closure->string2 + inset - closure->size1); - pos->string = (__const__ unsigned char *) closure->string2; - pos->size = closure->size2; - pos->end = ((__const__ unsigned char *) - MIN(closure->string2 + closure->size2, - closure->string2 + stop - closure->size1)); - pos->offset = closure->size1; - return ((pos->pos < pos->end) - ? rx_get_burst_ok : rx_get_burst_no_more); - } - } -} - - -#ifdef __STDC__ -static __inline__ enum rx_back_check_return -re_search_2_back_check(struct rx_string_position *pos, - int lparen, int rparen, unsigned char *translate, - void *vclosure, int stop) -#else -static __inline__ enum rx_back_check_return -re_search_2_back_check(pos, lparen, rparen, translate, vclosure, stop) -struct rx_string_position *pos; -int lparen; -int rparen; -unsigned char *translate; -void *vclosure; -int stop; -#endif -{ - struct rx_string_position there; - struct rx_string_position past; - - there = *pos; - there.pos = there.string + lparen - there.offset; - re_search_2_get_burst(&there, vclosure, stop); - - past = *pos; - past.pos = past.string + rparen - there.offset; - re_search_2_get_burst(&past, vclosure, stop); - - ++pos->pos; - re_search_2_get_burst(pos, vclosure, stop); - - while ((there.pos != past.pos) - && (pos->pos != pos->end)) - if (TRANSLATE(*there.pos) != TRANSLATE(*pos->pos)) - return rx_back_check_fail; - else { - ++there.pos; - ++pos->pos; - if (there.pos == there.end) - re_search_2_get_burst(&there, vclosure, stop); - if (pos->pos == pos->end) - re_search_2_get_burst(pos, vclosure, stop); - } - - if (there.pos != past.pos) - return rx_back_check_fail; - --pos->pos; - re_search_2_get_burst(pos, vclosure, stop); - return rx_back_check_pass; -} - -#ifdef __STDC__ -static __inline__ int -re_search_2_fetch_char(struct rx_string_position *pos, int offset, - void *app_closure, int stop) -#else -static __inline__ int -re_search_2_fetch_char(pos, offset, app_closure, stop) -struct rx_string_position *pos; -int offset; -void *app_closure; -int stop; -#endif -{ - struct re_search_2_closure *closure; - - closure = (struct re_search_2_closure *) app_closure; - if (offset == 0) { - if (pos->pos >= pos->string) - return *pos->pos; - else { - if ( - (pos->string == - (__const__ unsigned char *) closure->string2) - && (closure->string1) && (closure->size1)) - return closure->string1[closure->size1 - 1]; - else - return 0; /* sure, why not. */ - } - } - if (pos->pos == pos->end) - return *closure->string2; - else -#if 0 - return pos->pos[1]; -#else - return pos->pos[offset]; /* FIXME */ -#endif -} - -#ifdef __STDC__ -RE_S2_QUAL int -RE_SEARCH_2_FN(struct re_pattern_buffer *rxb, - __const__ char *string1, int size1, - __const__ char *string2, int size2, - int startpos, int range, - struct re_registers *regs, int stop) -#else -RE_S2_QUAL int -RE_SEARCH_2_FN(rxb, - string1, size1, string2, size2, startpos, range, regs, stop) -struct re_pattern_buffer *rxb; -__const__ char *string1; -int size1; -__const__ char *string2; -int size2; -int startpos; -int range; -struct re_registers *regs; -int stop; -#endif -{ - int answer; - struct re_search_2_closure closure; - - closure.string1 = string1; - closure.size1 = size1; - closure.string2 = string2; - closure.size2 = size2; - answer = rx_search(rxb, startpos, range, stop, size1 + size2, - re_search_2_get_burst, - re_search_2_back_check, - re_search_2_fetch_char, - (void *) &closure, regs, 0, 0); - switch (answer) { - case rx_search_continuation: - abort(); - case rx_search_error: - return -2; - case rx_search_soft_fail: - case rx_search_fail: - return -1; - default: - return answer; - } -} - -/* Export rx_search to callers outside this file. */ - -#ifdef __STDC__ -int -re_rx_search(struct re_pattern_buffer *rxb, int startpos, int range, - int stop, int total_size, rx_get_burst_fn get_burst, - rx_back_check_fn back_check, rx_fetch_char_fn fetch_char, - void *app_closure, struct re_registers *regs, - struct rx_search_state *resume_state, - struct rx_search_state *save_state) -#else -int -re_rx_search(rxb, startpos, range, stop, total_size, - get_burst, back_check, fetch_char, - app_closure, regs, resume_state, save_state) -struct re_pattern_buffer *rxb; -int startpos; -int range; -int stop; -int total_size; -rx_get_burst_fn get_burst; -rx_back_check_fn back_check; -rx_fetch_char_fn fetch_char; -void *app_closure; -struct re_registers *regs; -struct rx_search_state *resume_state; -struct rx_search_state *save_state; -#endif -{ - return rx_search(rxb, startpos, range, stop, total_size, - get_burst, back_check, fetch_char, app_closure, - regs, resume_state, save_state); -} - -#if !defined(REGEX_MALLOC) && !defined(__GNUC__) -#ifdef __STDC__ -int -re_search_2(struct re_pattern_buffer *rxb, - __const__ char *string1, int size1, - __const__ char *string2, int size2, - int startpos, int range, struct re_registers *regs, int stop) -#else -int -re_search_2(rxb, string1, size1, string2, size2, startpos, range, regs, - stop) -struct re_pattern_buffer *rxb; -__const__ char *string1; -int size1; -__const__ char *string2; -int size2; -int startpos; -int range; -struct re_registers *regs; -int stop; -#endif -{ - int ret; - - ret = inner_re_search_2(rxb, string1, size1, string2, size2, startpos, - range, regs, stop); - alloca(0); - return ret; -} -#endif - - -/* Like re_search_2, above, but only one string is specified, and - * doesn't let you say where to stop matching. - */ - -#ifdef __STDC__ -int -re_search(struct re_pattern_buffer *rxb, __const__ char *string, - int size, int startpos, int range, struct re_registers *regs) -#else -int re_search(rxb, string, size, startpos, range, regs) -struct re_pattern_buffer *rxb; -__const__ char *string; -int size; -int startpos; -int range; -struct re_registers *regs; -#endif -{ - return re_search_2(rxb, 0, 0, string, size, startpos, range, regs, - size); -} - -#ifdef __STDC__ -int -re_match_2(struct re_pattern_buffer *rxb, - __const__ char *string1, int size1, - __const__ char *string2, int size2, - int pos, struct re_registers *regs, int stop) -#else -int re_match_2(rxb, string1, size1, string2, size2, pos, regs, stop) -struct re_pattern_buffer *rxb; -__const__ char *string1; -int size1; -__const__ char *string2; -int size2; -int pos; -struct re_registers *regs; -int stop; -#endif -{ - struct re_registers some_regs; - regoff_t start; - regoff_t end; - int srch; - int save = rxb->regs_allocated; - struct re_registers *regs_to_pass = regs; - char *old_fastmap = rxb->fastmap; - - if (!regs) { - some_regs.start = &start; - some_regs.end = &end; - some_regs.num_regs = 1; - regs_to_pass = &some_regs; - rxb->regs_allocated = REGS_FIXED; - } - - rxb->fastmap = NULL; - srch = re_search_2(rxb, string1, size1, string2, size2, - pos, 1, regs_to_pass, stop); - rxb->fastmap = old_fastmap; - if (regs_to_pass != regs) - rxb->regs_allocated = save; - if (srch < 0) - return srch; - return regs_to_pass->end[0] - regs_to_pass->start[0]; -} - -/* re_match is like re_match_2 except it takes only a single string. */ - -#ifdef __STDC__ -int -re_match(struct re_pattern_buffer *rxb, - __const__ char *string, - int size, int pos, struct re_registers *regs) -#else -int re_match(rxb, string, size, pos, regs) -struct re_pattern_buffer *rxb; -__const__ char *string; -int size; -int pos; -struct re_registers *regs; -#endif -{ - return re_match_2(rxb, string, size, 0, 0, pos, regs, size); -} - - - -/* 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. */ -reg_syntax_t re_syntax_options = RE_SYNTAX_EMACS; - - -/* 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. */ - -#ifdef __STDC__ -reg_syntax_t re_set_syntax(reg_syntax_t syntax) -#else -reg_syntax_t re_set_syntax(syntax) -reg_syntax_t syntax; -#endif -{ - reg_syntax_t ret = re_syntax_options; - - re_syntax_options = syntax; - return ret; -} - - -/* 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. */ - -#ifdef __STDC__ -void -re_set_registers(struct re_pattern_buffer *bufp, - struct re_registers *regs, - unsigned num_regs, regoff_t * starts, regoff_t * ends) -#else -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; -regoff_t *ends; -#endif -{ - 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; - } -} - - - - -#ifdef __STDC__ -static int cplx_se_sublist_len(struct rx_se_list *list) -#else -static int cplx_se_sublist_len(list) -struct rx_se_list *list; -#endif -{ - int x = 0; - - while (list) { - if ((long) list->car >= 0) - ++x; - list = list->cdr; - } - return x; -} - - -/* For rx->se_list_cmp */ - -#ifdef __STDC__ -static int -posix_se_list_order(struct rx *rx, - struct rx_se_list *a, struct rx_se_list *b) -#else -static int posix_se_list_order(rx, a, b) -struct rx *rx; -struct rx_se_list *a; -struct rx_se_list *b; -#endif -{ - int al = cplx_se_sublist_len(a); - int bl = cplx_se_sublist_len(b); - - if (!al && !bl) - return ((a == b) - ? 0 : ((a < b) ? -1 : 1)); - - else if (!al) - return -1; - - else if (!bl) - return 1; - - else { - rx_side_effect *av = ((rx_side_effect *) - alloca(sizeof(rx_side_effect) * (al + 1))); - rx_side_effect *bv = ((rx_side_effect *) - alloca(sizeof(rx_side_effect) * (bl + 1))); - struct rx_se_list *ap = a; - struct rx_se_list *bp = b; - int ai, bi; - - for (ai = al - 1; ai >= 0; --ai) { - while ((long) ap->car < 0) - ap = ap->cdr; - av[ai] = ap->car; - ap = ap->cdr; - } - av[al] = (rx_side_effect) - 2; - for (bi = bl - 1; bi >= 0; --bi) { - while ((long) bp->car < 0) - bp = bp->cdr; - bv[bi] = bp->car; - bp = bp->cdr; - } - bv[bl] = (rx_side_effect) - 1; - - { - int ret; - int x = 0; - - while (av[x] == bv[x]) - ++x; - ret = (((unsigned *) (av[x]) < (unsigned *) (bv[x])) ? -1 : 1); - return ret; - } - } -} - - - - -/* re_compile_pattern is the GNU regular expression compiler: it - compiles PATTERN (of length SIZE) and puts the result in RXB. - Returns 0 if the pattern was valid, otherwise an error string. - - Assumes the `allocated' (and perhaps `buffer') and `translate' fields - are set in RXB on entry. - - We call rx_compile to do the actual compilation. */ - -#ifdef __STDC__ -__const__ char *re_compile_pattern(__const__ char *pattern, - int length, - struct re_pattern_buffer *rxb) -#else -__const__ char *re_compile_pattern(pattern, length, rxb) -__const__ char *pattern; -int length; -struct re_pattern_buffer *rxb; -#endif -{ - 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). */ - rxb->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. */ - rxb->no_sub = 0; - - rxb->rx.local_cset_size = 256; - - /* Match anchors at newline. */ - rxb->newline_anchor = 1; - - rxb->re_nsub = 0; - rxb->start = 0; - rxb->se_params = 0; - rxb->rx.nodec = 0; - rxb->rx.epsnodec = 0; - rxb->rx.instruction_table = 0; - rxb->rx.nfa_states = 0; - rxb->rx.se_list_cmp = posix_se_list_order; - rxb->rx.start_set = 0; - - ret = rx_compile(pattern, length, re_syntax_options, rxb); - alloca(0); - return rx_error_msg[(int) ret]; -} - - -#ifdef __STDC__ -int re_compile_fastmap(struct re_pattern_buffer *rxb) -#else -int re_compile_fastmap(rxb) -struct re_pattern_buffer *rxb; -#endif -{ - rx_blow_up_fastmap(rxb); - return 0; -} - - - - -/* Entry points compatible with 4.2 BSD regex library. We don't define - them if this is an Emacs or POSIX compilation. */ - -#if (!defined (emacs) && !defined (_POSIX_SOURCE)) || defined(USE_BSD_REGEX) - -/* BSD has one and only one pattern buffer. */ -static struct re_pattern_buffer rx_comp_buf; - -#ifdef __STDC__ -char *re_comp(__const__ char *s) -#else -char *re_comp(s) -__const__ char *s; -#endif -{ - reg_errcode_t ret; - - if (!s || (*s == '\0')) { - if (!rx_comp_buf.buffer) - return "No previous regular expression"; - return 0; - } - - if (!rx_comp_buf.fastmap) { - rx_comp_buf.fastmap = (char *) malloc(1 << CHARBITS); - if (!rx_comp_buf.fastmap) - return "Memory exhausted"; - } - - /* Since `rx_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. */ - rx_comp_buf.newline_anchor = 1; - - rx_comp_buf.re_nsub = 0; - rx_comp_buf.start = 0; - rx_comp_buf.se_params = 0; - rx_comp_buf.rx.nodec = 0; - rx_comp_buf.rx.epsnodec = 0; - rx_comp_buf.rx.instruction_table = 0; - rx_comp_buf.rx.nfa_states = 0; - rx_comp_buf.rx.start = 0; - rx_comp_buf.rx.se_list_cmp = posix_se_list_order; - rx_comp_buf.rx.start_set = 0; - rx_comp_buf.rx.local_cset_size = 256; - - ret = rx_compile(s, strlen(s), re_syntax_options, &rx_comp_buf); - alloca(0); - - /* Yes, we're discarding `__const__' here. */ - return (char *) rx_error_msg[(int) ret]; -} - - -#ifdef __STDC__ -int re_exec(__const__ char *s) -#else -int re_exec(s) -__const__ char *s; -#endif -{ - __const__ int len = strlen(s); - - return - 0 <= re_search(&rx_comp_buf, s, len, 0, len, - (struct re_registers *) 0); -} -#endif /* not emacs and not _POSIX_SOURCE */ - - - -/* POSIX.2 functions. Don't define these for Emacs. */ - -#if !defined(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.) */ - - -#ifdef __STDC__ -int regcomp(regex_t * preg, __const__ char *pattern, int cflags) -#else -int regcomp(preg, pattern, cflags) -regex_t *preg; -__const__ char *pattern; -int cflags; -#endif -{ - 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->fastmap = malloc(256); - if (!preg->fastmap) - return REG_ESPACE; - preg->fastmap_accurate = 0; - - if (cflags & REG_ICASE) { - unsigned i; - - preg->translate = (unsigned char *) malloc(256); - if (!preg->translate) - 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 = 0; - - /* If REG_NEWLINE is set, newlines are treated differently. */ - if (cflags & REG_NEWLINE) { /* REG_NEWLINE implies neither . nor [^...] match newline. */ - syntax &= ~RE_DOT_NEWLINE; - syntax |= RE_HAT_LISTS_NOT_NEWLINE; - /* It also changes the matching behavior. */ - preg->newline_anchor = 1; - } else - preg->newline_anchor = 0; - - preg->no_sub = !!(cflags & REG_NOSUB); - - /* POSIX says a null character in the pattern terminates it, so we - can use strlen here in compiling the pattern. */ - preg->re_nsub = 0; - preg->start = 0; - preg->se_params = 0; - preg->syntax_parens = 0; - preg->rx.nodec = 0; - preg->rx.epsnodec = 0; - preg->rx.instruction_table = 0; - preg->rx.nfa_states = 0; - preg->rx.local_cset_size = 256; - preg->rx.start = 0; - preg->rx.se_list_cmp = posix_se_list_order; - preg->rx.start_set = 0; - ret = rx_compile(pattern, strlen(pattern), syntax, preg); - alloca(0); - - /* 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. */ - -#ifdef __STDC__ -int -regexec(__const__ regex_t * preg, __const__ char *string, - size_t nmatch, regmatch_t pmatch[], int eflags) -#else -int regexec(preg, string, nmatch, pmatch, eflags) -__const__ regex_t *preg; -__const__ char *string; -size_t nmatch; -regmatch_t pmatch[]; -int eflags; -#endif -{ - 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 = ((regoff_t *) malloc((nmatch) * sizeof(regoff_t))); - regs.end = ((regoff_t *) malloc((nmatch) * sizeof(regoff_t))); - if (regs.start == 0 || regs.end == 0) - return (int) REG_NOMATCH; - } - - /* Perform the searching operation. */ - ret = re_search(&private_preg, string, len, - /* start: */ 0, - /* range: */ len, - want_reg_info ? ®s : (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. */ - -#ifdef __STDC__ -size_t -regerror(int errcode, __const__ regex_t * preg, - char *errbuf, size_t errbuf_size) -#else -size_t regerror(errcode, preg, errbuf, errbuf_size) -int errcode; -__const__ regex_t *preg; -char *errbuf; -size_t errbuf_size; -#endif -{ - __const__ char *msg - = rx_error_msg[errcode] == 0 ? "Success" : rx_error_msg[errcode]; - size_t msg_size = strlen(msg) + 1; /* Includes the 0. */ - - 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. */ - -#ifdef __STDC__ -void regfree(regex_t * preg) -#else -void regfree(preg) -regex_t *preg; -#endif -{ - if (preg->buffer != 0) - free(preg->buffer); - preg->buffer = 0; - preg->allocated = 0; - - if (preg->fastmap != 0) - free(preg->fastmap); - preg->fastmap = 0; - preg->fastmap_accurate = 0; - - if (preg->translate != 0) - free(preg->translate); - preg->translate = 0; -} - -#endif /* not emacs */ -- cgit v1.2.3