/* 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, 1994, 1995, 1996, 1997, 1998, 1999, 2000, 2001, 2002, 2003, 2004, 2005 Free Software Foundation, Inc. This program is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation; either version 2, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program; if not, write to the Free Software Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, 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 #ifdef HAVE_CONFIG_H # include #endif #ifndef INSIDE_RECURSION # include # 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 /* Tru64 with Desktop Toolkit C has a bug: must be included before . */ # include /* 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 # define iswctype __iswctype # define mbrtowc __mbrtowc # define wcslen __wcslen # define wcscoll __wcscoll # define wcrtomb __wcrtomb /* We are also using some library internals. */ # include # include # include # include # endif # ifdef _LIBC # include # undef gettext # define gettext(msgid) __dcgettext ("libc", msgid, LC_MESSAGES) /* This define is so xgettext can find the internationalizable strings. */ # define gettext_noop(msgid) msgid # else /* This is for other GNU distributions with internationalized messages. */ # include "gettext.h" # endif /* Support for bounded pointers. */ # if !defined _LIBC && !defined __BOUNDED_POINTERS__ # define __bounded /* nothing */ # define __unbounded /* nothing */ # define __ptrvalue /* nothing */ # 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 # include /* 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 # include # ifndef bzero # ifndef _LIBC # define bzero(s, n) (memset (s, '\0', n), (s)) # else # define bzero(s, n) __bzero (s, n) # 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 */ # include # ifndef MB_LEN_MAX # define MB_LEN_MAX 1 # endif /* 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. */ # if defined STDC_HEADERS || (!defined isascii && !defined HAVE_ISASCII) # define IN_CTYPE_DOMAIN(c) 1 # else # define IN_CTYPE_DOMAIN(c) isascii(c) # endif # ifdef isblank # define ISBLANK(c) (IN_CTYPE_DOMAIN (c) && isblank (c)) # else # define ISBLANK(c) ((c) == ' ' || (c) == '\t') # endif # ifdef isgraph # define ISGRAPH(c) (IN_CTYPE_DOMAIN (c) && isgraph (c)) # else # define ISGRAPH(c) (IN_CTYPE_DOMAIN (c) && isprint (c) && !isspace (c)) # endif # undef ISPRINT # define ISPRINT(c) (IN_CTYPE_DOMAIN (c) && isprint (c)) # define ISDIGIT(c) (IN_CTYPE_DOMAIN (c) && isdigit (c)) # define ISALNUM(c) (IN_CTYPE_DOMAIN (c) && isalnum (c)) # define ISALPHA(c) (IN_CTYPE_DOMAIN (c) && isalpha (c)) # define ISCNTRL(c) (IN_CTYPE_DOMAIN (c) && iscntrl (c)) # define ISLOWER(c) (IN_CTYPE_DOMAIN (c) && islower (c)) # define ISPUNCT(c) (IN_CTYPE_DOMAIN (c) && ispunct (c)) # define ISSPACE(c) (IN_CTYPE_DOMAIN (c) && isspace (c)) # define ISUPPER(c) (IN_CTYPE_DOMAIN (c) && isupper (c)) # define ISXDIGIT(c) (IN_CTYPE_DOMAIN (c) && isxdigit (c)) # ifdef _tolower # define TOLOWER(c) _tolower(c) # else # define TOLOWER(c) tolower(c) # 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 (void) { 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[(unsigned char) (c)] # 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. */ # include # 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 reg_errcode_t byte_regex_compile (const char *pattern, size_t size, reg_syntax_t syntax, struct re_pattern_buffer *bufp); static int byte_re_match_2_internal (struct re_pattern_buffer *bufp, const char *string1, int size1, const char *string2, int size2, int pos, struct re_registers *regs, int stop); static int byte_re_search_2 (struct re_pattern_buffer *bufp, const char *string1, int size1, const char *string2, int size2, int startpos, int range, struct re_registers *regs, int stop); static int byte_re_compile_fastmap (struct re_pattern_buffer *bufp); #ifdef MBS_SUPPORT static reg_errcode_t wcs_regex_compile (const char *pattern, size_t size, reg_syntax_t syntax, struct re_pattern_buffer *bufp); static int wcs_re_match_2_internal (struct re_pattern_buffer *bufp, const char *cstring1, int csize1, const char *cstring2, int csize2, int pos, struct re_registers *regs, int stop, wchar_t *string1, int size1, wchar_t *string2, int size2, int *mbs_offset1, int *mbs_offset2); static int wcs_re_search_2 (struct re_pattern_buffer *bufp, const char *string1, int size1, const char *string2, int size2, int startpos, int range, struct re_registers *regs, int stop); static int wcs_re_compile_fastmap (struct re_pattern_buffer *bufp); #endif /* 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, # ifdef MBS_SUPPORT /* Same as exactn, but contains binary data. */ exactn_bin, # endif /* 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. */ /* ifdef MBS_SUPPORT, following element is length of character classes, length of collating symbols, length of equivalence classes, length of character ranges, and length of characters. Next, character class element, collating symbols elements, equivalence class elements, range elements, and character elements follow. See regex_compile function. */ 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. */ /* ifdef MBS_SUPPORT, the size of address is 1. */ 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. */ /* ifdef MBS_SUPPORT, the size of address is 1. */ 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. */ /* ifdef MBS_SUPPORT, the size of address is 1. */ 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. */ /* ifdef MBS_SUPPORT, the size of address is 1. */ 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. */ /* ifdef MBS_SUPPORT, the size of address is 1. */ succeed_n, /* Followed by two-byte relative address, and two-byte number n. Jump to the address N times, then fail. */ /* ifdef MBS_SUPPORT, the size of address is 1. */ jump_n, /* Set the following two-byte relative address to the subsequent two-byte number. The address *includes* the two bytes of number. */ /* ifdef MBS_SUPPORT, the size of address is 1. */ 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; #endif /* not INSIDE_RECURSION */ #ifdef BYTE # define CHAR_T char # define UCHAR_T unsigned char # define COMPILED_BUFFER_VAR bufp->buffer # define OFFSET_ADDRESS_SIZE 2 # define PREFIX(name) byte_##name # define ARG_PREFIX(name) name # define PUT_CHAR(c) putchar (c) #else # ifdef WCHAR # define CHAR_T wchar_t # define UCHAR_T wchar_t # define COMPILED_BUFFER_VAR wc_buffer # define OFFSET_ADDRESS_SIZE 1 /* the size which STORE_NUMBER macro use */ # define CHAR_CLASS_SIZE ((__alignof__(wctype_t)+sizeof(wctype_t))/sizeof(CHAR_T)+1) # define PREFIX(name) wcs_##name # define ARG_PREFIX(name) c##name /* Should we use wide stream?? */ # define PUT_CHAR(c) printf ("%C", c); # define TRUE 1 # define FALSE 0 # else # ifdef MBS_SUPPORT # define WCHAR # define INSIDE_RECURSION # include "regex.c" # undef INSIDE_RECURSION # endif # define BYTE # define INSIDE_RECURSION # include "regex.c" # undef INSIDE_RECURSION # endif #endif #if USE_UNLOCKED_IO # include "unlocked-io.h" #endif #ifdef INSIDE_RECURSION /* Common operations on the compiled pattern. */ /* Store NUMBER in two contiguous bytes starting at DESTINATION. */ /* ifdef MBS_SUPPORT, we store NUMBER in 1 element. */ # ifdef WCHAR # define STORE_NUMBER(destination, number) \ do { \ *(destination) = (UCHAR_T)(number); \ } while (0) # else /* BYTE */ # define STORE_NUMBER(destination, number) \ do { \ (destination)[0] = (number) & 0377; \ (destination)[1] = (number) >> 8; \ } while (0) # endif /* WCHAR */ /* Same as STORE_NUMBER, except increment DESTINATION to the byte after where the number is stored. Therefore, DESTINATION must be an lvalue. */ /* ifdef MBS_SUPPORT, we store NUMBER in 1 element. */ # define STORE_NUMBER_AND_INCR(destination, number) \ do { \ STORE_NUMBER (destination, number); \ (destination) += OFFSET_ADDRESS_SIZE; \ } while (0) /* Put into DESTINATION a number stored in two contiguous bytes starting at SOURCE. */ /* ifdef MBS_SUPPORT, we store NUMBER in 1 element. */ # ifdef WCHAR # define EXTRACT_NUMBER(destination, source) \ do { \ (destination) = *(source); \ } while (0) # else /* BYTE */ # define EXTRACT_NUMBER(destination, source) \ do { \ (destination) = *(source) & 0377; \ (destination) += (signed char) (*((source) + 1)) << 8; \ } while (0) # endif # ifdef DEBUG static void PREFIX(extract_number) (int *dest, UCHAR_T *source) { # ifdef WCHAR *dest = *source; # else /* BYTE */ signed char temp = source[1]; *dest = *source & 0377; *dest += temp << 8; # endif } # ifndef EXTRACT_MACROS /* To debug the macros. */ # undef EXTRACT_NUMBER # define EXTRACT_NUMBER(dest, src) PREFIX(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) += OFFSET_ADDRESS_SIZE; \ } while (0) # ifdef DEBUG static void PREFIX(extract_number_and_incr) (int *destination, UCHAR_T **source) { PREFIX(extract_number) (destination, *source); *source += OFFSET_ADDRESS_SIZE; } # ifndef EXTRACT_MACROS # undef EXTRACT_NUMBER_AND_INCR # define EXTRACT_NUMBER_AND_INCR(dest, src) \ PREFIX(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 # ifndef DEFINED_ONCE /* 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) # endif /* not DEFINED_ONCE */ # define DEBUG_PRINT_COMPILED_PATTERN(p, s, e) \ if (debug) PREFIX(print_partial_compiled_pattern) (s, e) # define DEBUG_PRINT_DOUBLE_STRING(w, s1, sz1, s2, sz2) \ if (debug) PREFIX(print_double_string) (w, s1, sz1, s2, sz2) /* Print the fastmap in human-readable form. */ # ifndef DEFINED_ONCE void print_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'); } # endif /* not DEFINED_ONCE */ /* Print a compiled pattern string in human-readable form, starting at the START pointer into it and ending just before the pointer END. */ void PREFIX(print_partial_compiled_pattern) (UCHAR_T *start, UCHAR_T *end) { int mcnt, mcnt2; UCHAR_T *p1; UCHAR_T *p = start; UCHAR_T *pend = end; if (start == NULL) { printf ("(null)\n"); return; } /* Loop over pattern commands. */ while (p < pend) { # ifdef _LIBC printf ("%td:\t", p - start); # else printf ("%ld:\t", (long int) (p - start)); # endif switch ((re_opcode_t) *p++) { case no_op: printf ("/no_op"); break; case exactn: mcnt = *p++; printf ("/exactn/%d", mcnt); do { putchar ('/'); PUT_CHAR (*p++); } while (--mcnt); break; # ifdef MBS_SUPPORT case exactn_bin: mcnt = *p++; printf ("/exactn_bin/%d", mcnt); do { printf("/%lx", (long int) *p++); } while (--mcnt); break; # endif /* MBS_SUPPORT */ case start_memory: mcnt = *p++; printf ("/start_memory/%d/%ld", mcnt, (long int) *p++); break; case stop_memory: mcnt = *p++; printf ("/stop_memory/%d/%ld", mcnt, (long int) *p++); break; case duplicate: printf ("/duplicate/%ld", (long int) *p++); break; case anychar: printf ("/anychar"); break; case charset: case charset_not: { # ifdef WCHAR int i, length; wchar_t *workp = p; printf ("/charset [%s", (re_opcode_t) *(workp - 1) == charset_not ? "^" : ""); p += 5; length = *workp++; /* the length of char_classes */ for (i=0 ; ibuffer; PREFIX(print_partial_compiled_pattern) (buffer, buffer + bufp->used / sizeof(UCHAR_T)); printf ("%ld bytes used/%ld bytes allocated.\n", bufp->used, bufp->allocated); if (bufp->fastmap_accurate && bufp->fastmap) { printf ("fastmap: "); print_fastmap (bufp->fastmap); } # ifdef _LIBC printf ("re_nsub: %Zd\t", bufp->re_nsub); # else printf ("re_nsub: %ld\t", (long int) bufp->re_nsub); # endif 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 PREFIX(print_double_string) (const CHAR_T *where, const CHAR_T *string1, const CHAR_T *string2, int size1, int size2) { int this_char; if (where == NULL) printf ("(null)"); else { int cnt; if (FIRST_STRING_P (where)) { for (this_char = where - string1; this_char < size1; this_char++) PUT_CHAR (string1[this_char]); where = string2; } cnt = 0; for (this_char = where - string2; this_char < size2; this_char++) { PUT_CHAR (string2[this_char]); if (++cnt > 100) { fputs ("...", stdout); break; } } } } # ifndef DEFINED_ONCE void printchar (c) int c; { putc (c, stderr); } # endif # else /* not DEBUG */ # ifndef DEFINED_ONCE # 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) # endif /* not DEFINED_ONCE */ # define DEBUG_PRINT_COMPILED_PATTERN(p, s, e) # define DEBUG_PRINT_DOUBLE_STRING(w, s1, sz1, s2, sz2) # endif /* not DEBUG */ # ifdef WCHAR /* This convert a multibyte string to a wide character string. And write their correspondances to offset_buffer(see below) and write whether each wchar_t is binary data to is_binary. This assume invalid multibyte sequences as binary data. We assume offset_buffer and is_binary is already allocated enough space. */ static size_t convert_mbs_to_wcs (CHAR_T *dest, const unsigned char* src, /* The length of multibyte string. */ size_t len, /* Correspondences between src(char string) and dest(wchar_t string) for optimization. E.g.: src = "xxxyzz" dest = {'X', 'Y', 'Z'} (each "xxx", "y" and "zz" represent one multibyte character corresponding to 'X', 'Y' and 'Z'.) offset_buffer = {0, 0+3("xxx"), 0+3+1("y"), 0+3+1+2("zz")} = {0, 3, 4, 6} */ int *offset_buffer, char *is_binary) { wchar_t *pdest = dest; const unsigned char *psrc = src; size_t wc_count = 0; mbstate_t mbs; int i, consumed; size_t mb_remain = len; size_t mb_count = 0; /* Initialize the conversion state. */ memset (&mbs, 0, sizeof (mbstate_t)); offset_buffer[0] = 0; for( ; mb_remain > 0 ; ++wc_count, ++pdest, mb_remain -= consumed, psrc += consumed) { consumed = mbrtowc (pdest, psrc, mb_remain, &mbs); if (consumed <= 0) /* failed to convert. maybe src contains binary data. So we consume 1 byte manualy. */ { *pdest = *psrc; consumed = 1; is_binary[wc_count] = TRUE; } else is_binary[wc_count] = FALSE; /* In sjis encoding, we use yen sign as escape character in place of reverse solidus. So we convert 0x5c(yen sign in sjis) to not 0xa5(yen sign in UCS2) but 0x5c(reverse solidus in UCS2). */ if (consumed == 1 && (int) *psrc == 0x5c && (int) *pdest == 0xa5) *pdest = (wchar_t) *psrc; offset_buffer[wc_count + 1] = mb_count += consumed; } /* Fill remain of the buffer with sentinel. */ for (i = wc_count + 1 ; i <= len ; i++) offset_buffer[i] = mb_count + 1; return wc_count; } # endif /* WCHAR */ #else /* not INSIDE_RECURSION */ /* Set by `re_set_syntax' to the current regexp syntax to recognize. Can also be assigned to arbitrarily: each pattern buffer stores its own syntax, so it can be changed between regex compilations. */ /* This has no initializer because initialized variables in Emacs become read-only after dumping. */ reg_syntax_t re_syntax_options; /* Specify the precise syntax of regexps for compilation. This provides for compatibility for various utilities which historically have different, incompatible syntaxes. The argument SYNTAX is a bit mask comprised of the various bits defined in regex.h. We return the old syntax. */ reg_syntax_t re_set_syntax (reg_syntax_t syntax) { reg_syntax_t ret = re_syntax_options; re_syntax_options = syntax; # 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 }; #endif /* INSIDE_RECURSION */ #ifndef DEFINED_ONCE /* 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 #endif /* not DEFINED_ONCE */ #ifdef INSIDE_RECURSION /* 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 # ifndef DEFINED_ONCE # 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 # endif union PREFIX(fail_stack_elt) { UCHAR_T *pointer; long int integer; }; typedef union PREFIX(fail_stack_elt) PREFIX(fail_stack_elt_t); typedef struct { PREFIX(fail_stack_elt_t) *stack; unsigned long int size; unsigned long int avail; /* Offset of next open position. */ } PREFIX(fail_stack_type); # else /* not INT_IS_16BIT */ # ifndef DEFINED_ONCE # 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 = 4000; # else int re_max_failures = 2000; # endif # endif union PREFIX(fail_stack_elt) { UCHAR_T *pointer; int integer; }; typedef union PREFIX(fail_stack_elt) PREFIX(fail_stack_elt_t); typedef struct { PREFIX(fail_stack_elt_t) *stack; unsigned size; unsigned avail; /* Offset of next open position. */ } PREFIX(fail_stack_type); # endif /* INT_IS_16BIT */ # ifndef DEFINED_ONCE # 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) # endif /* 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 = (PREFIX(fail_stack_elt_t) *) \ REGEX_ALLOCATE_STACK (INIT_FAILURE_ALLOC * sizeof (PREFIX(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 = (PREFIX(fail_stack_elt_t) *) \ REGEX_REALLOCATE_STACK ((fail_stack).stack, \ (fail_stack).size * sizeof (PREFIX(fail_stack_elt_t)), \ ((fail_stack).size << 1) * sizeof (PREFIX(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 = (UCHAR_T *) (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) # ifndef DEFINED_ONCE /* 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) # endif /* not DEFINED_ONCE */ /* 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 UCHAR_T *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_T *) string_temp; \ \ DEBUG_PRINT2 (" Popping string %p: `", str); \ DEBUG_PRINT_DOUBLE_STRING (str, string1, size1, string2, size2); \ DEBUG_PRINT1 ("'\n"); \ \ pat = (UCHAR_T *) 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_T *) POP_FAILURE_POINTER (); \ DEBUG_PRINT2 (" end: %p\n", regend[this_reg]); \ \ regstart[this_reg] = (const CHAR_T *) 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 { PREFIX(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; } PREFIX(register_info_type); # ifndef DEFINED_ONCE # 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) # endif /* not DEFINED_ONCE */ /* Registers are set to a sentinel when they haven't yet matched. */ static CHAR_T PREFIX(reg_unset_dummy); # define REG_UNSET_VALUE (&PREFIX(reg_unset_dummy)) # define REG_UNSET(e) ((e) == REG_UNSET_VALUE) /* Subroutine declarations and macros for regex_compile. */ static void PREFIX(store_op1) (re_opcode_t op, UCHAR_T *loc, int arg); static void PREFIX(store_op2) (re_opcode_t op, UCHAR_T *loc, int arg1, int arg2); static void PREFIX(insert_op1) (re_opcode_t op, UCHAR_T *loc, int arg, UCHAR_T *end); static void PREFIX(insert_op2) (re_opcode_t op, UCHAR_T *loc, int arg1, int arg2, UCHAR_T *end); static boolean PREFIX(at_begline_loc_p) (const CHAR_T *pattern, const CHAR_T *p, reg_syntax_t syntax); static boolean PREFIX(at_endline_loc_p) (const CHAR_T *p, const CHAR_T *pend, reg_syntax_t syntax); # ifdef WCHAR static reg_errcode_t wcs_compile_range (CHAR_T range_start, const CHAR_T **p_ptr, const CHAR_T *pend, char *translate, reg_syntax_t syntax, UCHAR_T *b, CHAR_T *char_set); static void insert_space (int num, CHAR_T *loc, CHAR_T *end); # else /* BYTE */ static reg_errcode_t byte_compile_range (unsigned int range_start, const char **p_ptr, const char *pend, char *translate, reg_syntax_t syntax, unsigned char *b); # endif /* WCHAR */ /* 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'). */ /* ifdef MBS_SUPPORT, we translate only if character <= 0xff, because it is impossible to allocate 4GB array for some encodings which have 4 byte character_set like UCS4. */ # ifndef PATFETCH # ifdef WCHAR # define PATFETCH(c) \ do {if (p == pend) return REG_EEND; \ c = (UCHAR_T) *p++; \ if (translate && (c <= 0xff)) c = (UCHAR_T) translate[c]; \ } while (0) # else /* BYTE */ # define PATFETCH(c) \ do {if (p == pend) return REG_EEND; \ c = (unsigned char) *p++; \ if (translate) c = (unsigned char) translate[c]; \ } while (0) # endif /* WCHAR */ # endif /* Fetch the next character in the uncompiled pattern, with no translation. */ # define PATFETCH_RAW(c) \ do {if (p == pend) return REG_EEND; \ c = (UCHAR_T) *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. */ /* ifdef MBS_SUPPORT, we translate only if character <= 0xff, because it is impossible to allocate 4GB array for some encodings which have 4 byte character_set like UCS4. */ # ifndef TRANSLATE # ifdef WCHAR # define TRANSLATE(d) \ ((translate && ((UCHAR_T) (d)) <= 0xff) \ ? (char) translate[(unsigned char) (d)] : (d)) # else /* BYTE */ # define TRANSLATE(d) \ (translate ? (char) translate[(unsigned char) (d)] : (d)) # endif /* WCHAR */ # 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 * sizeof(UCHAR_T)) /* Make sure we have at least N more bytes of space in buffer. */ # ifdef WCHAR # define GET_BUFFER_SPACE(n) \ while (((unsigned long)b - (unsigned long)COMPILED_BUFFER_VAR \ + (n)*sizeof(CHAR_T)) > bufp->allocated) \ EXTEND_BUFFER () # else /* BYTE */ # define GET_BUFFER_SPACE(n) \ while ((unsigned long) (b - bufp->buffer + (n)) > bufp->allocated) \ EXTEND_BUFFER () # endif /* WCHAR */ /* 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++ = (UCHAR_T) (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++ = (UCHAR_T) (c1); \ *b++ = (UCHAR_T) (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++ = (UCHAR_T) (c1); \ *b++ = (UCHAR_T) (c2); \ *b++ = (UCHAR_T) (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) \ PREFIX(store_op1) (op, loc, (int) ((to) - (loc) - (1 + OFFSET_ADDRESS_SIZE))) /* Likewise, for a two-argument jump. */ # define STORE_JUMP2(op, loc, to, arg) \ PREFIX(store_op2) (op, loc, (int) ((to) - (loc) - (1 + OFFSET_ADDRESS_SIZE)), arg) /* Like `STORE_JUMP', but for inserting. Assume `b' is the buffer end. */ # define INSERT_JUMP(op, loc, to) \ PREFIX(insert_op1) (op, loc, (int) ((to) - (loc) - (1 + OFFSET_ADDRESS_SIZE)), b) /* Like `STORE_JUMP2', but for inserting. Assume `b' is the buffer end. */ # define INSERT_JUMP2(op, loc, to, arg) \ PREFIX(insert_op2) (op, loc, (int) ((to) - (loc) - (1 + OFFSET_ADDRESS_SIZE)),\ 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!! */ # ifndef DEFINED_ONCE # 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. */ # if __BOUNDED_POINTERS__ # define SET_HIGH_BOUND(P) (__ptrhigh (P) = __ptrlow (P) + bufp->allocated) # define MOVE_BUFFER_POINTER(P) \ (__ptrlow (P) += incr, SET_HIGH_BOUND (P), __ptrvalue (P) += incr) # define ELSE_EXTEND_BUFFER_HIGH_BOUND \ else \ { \ SET_HIGH_BOUND (b); \ SET_HIGH_BOUND (begalt); \ if (fixup_alt_jump) \ SET_HIGH_BOUND (fixup_alt_jump); \ if (laststart) \ SET_HIGH_BOUND (laststart); \ if (pending_exact) \ SET_HIGH_BOUND (pending_exact); \ } # else # define MOVE_BUFFER_POINTER(P) (P) += incr # define ELSE_EXTEND_BUFFER_HIGH_BOUND # endif # endif /* not DEFINED_ONCE */ # ifdef WCHAR # define EXTEND_BUFFER() \ do { \ UCHAR_T *old_buffer = COMPILED_BUFFER_VAR; \ int wchar_count; \ if (bufp->allocated + sizeof(UCHAR_T) > MAX_BUF_SIZE) \ return REG_ESIZE; \ bufp->allocated <<= 1; \ if (bufp->allocated > MAX_BUF_SIZE) \ bufp->allocated = MAX_BUF_SIZE; \ /* How many characters the new buffer can have? */ \ wchar_count = bufp->allocated / sizeof(UCHAR_T); \ if (wchar_count == 0) wchar_count = 1; \ /* Truncate the buffer to CHAR_T align. */ \ bufp->allocated = wchar_count * sizeof(UCHAR_T); \ RETALLOC (COMPILED_BUFFER_VAR, wchar_count, UCHAR_T); \ bufp->buffer = (char*)COMPILED_BUFFER_VAR; \ if (COMPILED_BUFFER_VAR == NULL) \ return REG_ESPACE; \ /* If the buffer moved, move all the pointers into it. */ \ if (old_buffer != COMPILED_BUFFER_VAR) \ { \ int incr = COMPILED_BUFFER_VAR - old_buffer; \ MOVE_BUFFER_POINTER (b); \ MOVE_BUFFER_POINTER (begalt); \ if (fixup_alt_jump) \ MOVE_BUFFER_POINTER (fixup_alt_jump); \ if (laststart) \ MOVE_BUFFER_POINTER (laststart); \ if (pending_exact) \ MOVE_BUFFER_POINTER (pending_exact); \ } \ ELSE_EXTEND_BUFFER_HIGH_BOUND \ } while (0) # else /* BYTE */ # define EXTEND_BUFFER() \ do { \ UCHAR_T *old_buffer = COMPILED_BUFFER_VAR; \ 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 \ = (UCHAR_T *) REALLOC (COMPILED_BUFFER_VAR, bufp->allocated); \ if (COMPILED_BUFFER_VAR == NULL) \ return REG_ESPACE; \ /* If the buffer moved, move all the pointers into it. */ \ if (old_buffer != COMPILED_BUFFER_VAR) \ { \ int incr = COMPILED_BUFFER_VAR - old_buffer; \ MOVE_BUFFER_POINTER (b); \ MOVE_BUFFER_POINTER (begalt); \ if (fixup_alt_jump) \ MOVE_BUFFER_POINTER (fixup_alt_jump); \ if (laststart) \ MOVE_BUFFER_POINTER (laststart); \ if (pending_exact) \ MOVE_BUFFER_POINTER (pending_exact); \ } \ ELSE_EXTEND_BUFFER_HIGH_BOUND \ } while (0) # endif /* WCHAR */ # ifndef DEFINED_ONCE /* 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]) # endif /* not DEFINED_ONCE */ /* Set the bit for character C in a list. */ # ifndef DEFINED_ONCE # define SET_LIST_BIT(c) \ (b[((unsigned char) (c)) / BYTEWIDTH] \ |= 1 << (((unsigned char) c) % BYTEWIDTH)) # endif /* DEFINED_ONCE */ /* Get the next unsigned number in the uncompiled pattern. */ # define GET_UNSIGNED_NUMBER(num) \ { \ while (p != pend) \ { \ PATFETCH (c); \ if (c < '0' || c > '9') \ break; \ if (num <= RE_DUP_MAX) \ { \ if (num < 0) \ num = 0; \ num = num * 10 + c - '0'; \ } \ } \ } # ifndef DEFINED_ONCE # 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 # endif /* DEFINED_ONCE */ # 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 PREFIX(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. */ # ifdef DEFINED_ONCE 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 const char **reg_dummy; # endif /* DEFINED_ONCE */ static PREFIX(register_info_type) *PREFIX(reg_info); static PREFIX(register_info_type) *PREFIX(reg_info_dummy); /* Make the register vectors big enough for NUM_REGS registers, but don't make them smaller. */ static void PREFIX(regex_grow_registers) (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 (PREFIX(reg_info), num_regs, PREFIX(register_info_type)); RETALLOC_IF (reg_dummy, num_regs, const char *); RETALLOC_IF (PREFIX(reg_info_dummy), num_regs, PREFIX(register_info_type)); regs_allocated_size = num_regs; } } # endif /* not MATCH_MAY_ALLOCATE */ # ifndef DEFINED_ONCE static boolean group_in_compile_stack (compile_stack_type compile_stack, regnum_t regnum); # endif /* not DEFINED_ONCE */ /* `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. */ # ifdef WCHAR # define FREE_STACK_RETURN(value) \ return (free(pattern), free(mbs_offset), free(is_binary), free (compile_stack.stack), value) # else # define FREE_STACK_RETURN(value) \ return (free (compile_stack.stack), value) # endif /* WCHAR */ static reg_errcode_t PREFIX(regex_compile) (const char *ARG_PREFIX(pattern), size_t ARG_PREFIX(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 UCHAR_T c, c1; #ifdef WCHAR /* A temporary space to keep wchar_t pattern and compiled pattern. */ CHAR_T *pattern, *COMPILED_BUFFER_VAR; size_t size; /* offset buffer for optimization. See convert_mbs_to_wc. */ int *mbs_offset = NULL; /* It hold whether each wchar_t is binary data or not. */ char *is_binary = NULL; /* A flag whether exactn is handling binary data or not. */ char is_exactn_bin = FALSE; #endif /* WCHAR */ /* A random temporary spot in PATTERN. */ const CHAR_T *p1; /* Points to the end of the buffer, where we should append. */ register UCHAR_T *b; /* Keeps track of unclosed groups. */ compile_stack_type compile_stack; /* Points to the current (ending) position in the pattern. */ #ifdef WCHAR const CHAR_T *p; const CHAR_T *pend; #else /* BYTE */ const CHAR_T *p = pattern; const CHAR_T *pend = pattern + size; #endif /* WCHAR */ /* 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. */ UCHAR_T *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. */ UCHAR_T *laststart = 0; /* Address of beginning of regexp, or inside of last group. */ UCHAR_T *begalt; /* 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. */ UCHAR_T *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 WCHAR /* Initialize the wchar_t PATTERN and offset_buffer. */ p = pend = pattern = TALLOC(csize + 1, CHAR_T); mbs_offset = TALLOC(csize + 1, int); is_binary = TALLOC(csize + 1, char); if (pattern == NULL || mbs_offset == NULL || is_binary == NULL) { free(pattern); free(mbs_offset); free(is_binary); return REG_ESPACE; } pattern[csize] = L'\0'; /* sentinel */ size = convert_mbs_to_wcs(pattern, cpattern, csize, mbs_offset, is_binary); pend = p + size; if (size < 0) { free(pattern); free(mbs_offset); free(is_binary); return REG_BADPAT; } #endif #ifdef DEBUG DEBUG_PRINT1 ("\nCompiling pattern: "); if (debug) { unsigned debug_count; for (debug_count = 0; debug_count < size; debug_count++) PUT_CHAR (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) { #ifdef WCHAR free(pattern); free(mbs_offset); free(is_binary); #endif 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. */ #ifdef WCHAR /* Free bufp->buffer and allocate an array for wchar_t pattern buffer. */ free(bufp->buffer); COMPILED_BUFFER_VAR = TALLOC (INIT_BUF_SIZE/sizeof(UCHAR_T), UCHAR_T); #else RETALLOC (COMPILED_BUFFER_VAR, INIT_BUF_SIZE, UCHAR_T); #endif /* WCHAR */ } else { /* Caller did not allocate a buffer. Do it for them. */ COMPILED_BUFFER_VAR = TALLOC (INIT_BUF_SIZE / sizeof(UCHAR_T), UCHAR_T); } if (!COMPILED_BUFFER_VAR) FREE_STACK_RETURN (REG_ESPACE); #ifdef WCHAR bufp->buffer = (char*)COMPILED_BUFFER_VAR; #endif /* WCHAR */ bufp->allocated = INIT_BUF_SIZE; } #ifdef WCHAR else COMPILED_BUFFER_VAR = (UCHAR_T*) bufp->buffer; #endif begalt = b = COMPILED_BUFFER_VAR; /* 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. */ || PREFIX(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. */ || PREFIX(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 (1 + OFFSET_ADDRESS_SIZE); /* 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 - (1 + OFFSET_ADDRESS_SIZE)); /* We've added more stuff to the buffer. */ b += 1 + OFFSET_ADDRESS_SIZE; } /* On failure, jump from laststart to b + 3, which will be the end of the buffer after this jump is inserted. */ /* ifdef WCHAR, 'b + 1 + OFFSET_ADDRESS_SIZE' instead of 'b + 3'. */ GET_BUFFER_SPACE (1 + OFFSET_ADDRESS_SIZE); INSERT_JUMP (keep_string_p ? on_failure_keep_string_jump : on_failure_jump, laststart, b + 1 + OFFSET_ADDRESS_SIZE); pending_exact = 0; b += 1 + OFFSET_ADDRESS_SIZE; 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 (1 + OFFSET_ADDRESS_SIZE); INSERT_JUMP (dummy_failure_jump, laststart, laststart + 2 + 2 * OFFSET_ADDRESS_SIZE); b += 1 + OFFSET_ADDRESS_SIZE; } } break; case '.': laststart = b; BUF_PUSH (anychar); break; case '[': { boolean had_char_class = false; #ifdef WCHAR CHAR_T range_start = 0xffffffff; #else unsigned int range_start = 0xffffffff; #endif if (p == pend) FREE_STACK_RETURN (REG_EBRACK); #ifdef WCHAR /* We assume a charset(_not) structure as a wchar_t array. charset[0] = (re_opcode_t) charset(_not) charset[1] = l (= length of char_classes) charset[2] = m (= length of collating_symbols) charset[3] = n (= length of equivalence_classes) charset[4] = o (= length of char_ranges) charset[5] = p (= length of chars) charset[6] = char_class (wctype_t) charset[6+CHAR_CLASS_SIZE] = char_class (wctype_t) ... charset[l+5] = char_class (wctype_t) charset[l+6] = collating_symbol (wchar_t) ... charset[l+m+5] = collating_symbol (wchar_t) ifdef _LIBC we use the index if _NL_COLLATE_SYMB_EXTRAMB instead of wchar_t string. charset[l+m+6] = equivalence_classes (wchar_t) ... charset[l+m+n+5] = equivalence_classes (wchar_t) ifdef _LIBC we use the index in _NL_COLLATE_WEIGHT instead of wchar_t string. charset[l+m+n+6] = range_start charset[l+m+n+7] = range_end ... charset[l+m+n+2o+4] = range_start charset[l+m+n+2o+5] = range_end ifdef _LIBC we use the value looked up in _NL_COLLATE_COLLSEQ instead of wchar_t character. charset[l+m+n+2o+6] = char ... charset[l+m+n+2o+p+5] = char */ /* We need at least 6 spaces: the opcode, the length of char_classes, the length of collating_symbols, the length of equivalence_classes, the length of char_ranges, the length of chars. */ GET_BUFFER_SPACE (6); /* Save b as laststart. And We use laststart as the pointer to the first element of the charset here. In other words, laststart[i] indicates charset[i]. */ 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++; /* Push the length of char_classes, the length of collating_symbols, the length of equivalence_classes, the length of char_ranges and the length of chars. */ BUF_PUSH_3 (0, 0, 0); BUF_PUSH_2 (0, 0); /* Remember the first position in the bracket expression. */ p1 = p; /* charset_not matches newline according to a syntax bit. */ if ((re_opcode_t) b[-6] == charset_not && (syntax & RE_HAT_LISTS_NOT_NEWLINE)) { BUF_PUSH('\n'); laststart[5]++; /* Update the length of characters */ } /* 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); BUF_PUSH(c1); laststart[5]++; /* Update the length of chars */ range_start = 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; /* Allocate the space for range_start and range_end. */ GET_BUFFER_SPACE (2); /* Update the pointer to indicate end of buffer. */ b += 2; ret = wcs_compile_range (range_start, &p, pend, translate, syntax, b, laststart); if (ret != REG_NOERROR) FREE_STACK_RETURN (ret); range_start = 0xffffffff; } else if (p[0] == '-' && p[1] != ']') { /* This handles ranges made up of characters only. */ reg_errcode_t ret; /* Move past the `-'. */ PATFETCH (c1); /* Allocate the space for range_start and range_end. */ GET_BUFFER_SPACE (2); /* Update the pointer to indicate end of buffer. */ b += 2; ret = wcs_compile_range (c, &p, pend, translate, syntax, b, laststart); if (ret != REG_NOERROR) FREE_STACK_RETURN (ret); range_start = 0xffffffff; } /* 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 store them as character). */ if (c == ':' && *p == ']') { wctype_t wt; uintptr_t alignedp; /* Query the character class as wctype_t. */ 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); /* Allocate the space for character class. */ GET_BUFFER_SPACE(CHAR_CLASS_SIZE); /* Update the pointer to indicate end of buffer. */ b += CHAR_CLASS_SIZE; /* Move data which follow character classes not to violate the data. */ insert_space(CHAR_CLASS_SIZE, laststart + 6 + laststart[1], b - 1); alignedp = ((uintptr_t)(laststart + 6 + laststart[1]) + __alignof__(wctype_t) - 1) & ~(uintptr_t)(__alignof__(wctype_t) - 1); /* Store the character class. */ *((wctype_t*)alignedp) = wt; /* Update length of char_classes */ laststart[1] += CHAR_CLASS_SIZE; had_char_class = true; } else { c1++; while (c1--) PATUNFETCH; BUF_PUSH ('['); BUF_PUSH (':'); laststart[5] += 2; /* Update the length of characters */ range_start = ':'; had_char_class = false; } } else if (syntax & RE_CHAR_CLASSES && c == '[' && (*p == '=' || *p == '.')) { CHAR_T str[128]; /* Should be large enough. */ CHAR_T delim = *p; /* '=' or '.' */ # ifdef _LIBC uint32_t nrules = _NL_CURRENT_WORD (LC_COLLATE, _NL_COLLATE_NRULES); # endif PATFETCH (c); c1 = 0; /* If pattern is `[[=' or '[[.'. */ if (p == pend) FREE_STACK_RETURN (REG_EBRACK); for (;;) { PATFETCH (c); if ((c == delim && *p == ']') || p == pend) break; if (c1 < sizeof (str) - 1) str[c1++] = c; else /* This is in any case an invalid class name. */ str[0] = '\0'; } str[c1] = '\0'; if (c == delim && *p == ']' && str[0] != '\0') { unsigned int i, offset; /* If we have no collation data we use the default collation in which each character is in a class by itself. It also means that ASCII is the character set and therefore we cannot have character with more than one byte in the multibyte representation. */ /* If not defined _LIBC, we push the name and `\0' for the sake of matching performance. */ int datasize = c1 + 1; # ifdef _LIBC int32_t idx = 0; if (nrules == 0) # endif { if (c1 != 1) FREE_STACK_RETURN (REG_ECOLLATE); } # ifdef _LIBC else { const int32_t *table; const int32_t *weights; const int32_t *extra; const int32_t *indirect; wint_t *cp; /* This #include defines a local function! */ # include if(delim == '=') { /* We push the index for equivalence class. */ cp = (wint_t*)str; table = (const int32_t *) _NL_CURRENT (LC_COLLATE, _NL_COLLATE_TABLEWC); weights = (const int32_t *) _NL_CURRENT (LC_COLLATE, _NL_COLLATE_WEIGHTWC); extra = (const int32_t *) _NL_CURRENT (LC_COLLATE, _NL_COLLATE_EXTRAWC); indirect = (const int32_t *) _NL_CURRENT (LC_COLLATE, _NL_COLLATE_INDIRECTWC); idx = findidx ((const wint_t**)&cp); if (idx == 0 || cp < (wint_t*) str + c1) /* This is no valid character. */ FREE_STACK_RETURN (REG_ECOLLATE); str[0] = (wchar_t)idx; } else /* delim == '.' */ { /* We push collation sequence value for collating symbol. */ int32_t table_size; const int32_t *symb_table; const unsigned char *extra; int32_t idx; int32_t elem; int32_t second; int32_t hash; char char_str[c1]; /* We have to convert the name to a single-byte string. This is possible since the names consist of ASCII characters and the internal representation is UCS4. */ for (i = 0; i < c1; ++i) char_str[i] = str[i]; table_size = _NL_CURRENT_WORD (LC_COLLATE, _NL_COLLATE_SYMB_HASH_SIZEMB); symb_table = (const int32_t *) _NL_CURRENT (LC_COLLATE, _NL_COLLATE_SYMB_TABLEMB); extra = (const unsigned char *) _NL_CURRENT (LC_COLLATE, _NL_COLLATE_SYMB_EXTRAMB); /* Locate the character in the hashing table. */ hash = elem_hash (char_str, c1); idx = 0; elem = hash % table_size; second = hash % (table_size - 2); while (symb_table[2 * elem] != 0) { /* First compare the hashing value. */ if (symb_table[2 * elem] == hash && c1 == extra[symb_table[2 * elem + 1]] && memcmp (char_str, &extra[symb_table[2 * elem + 1] + 1], c1) == 0) { /* Yep, this is the entry. */ idx = symb_table[2 * elem + 1]; idx += 1 + extra[idx]; break; } /* Next entry. */ elem += second; } if (symb_table[2 * elem] != 0) { /* Compute the index of the byte sequence in the table. */ idx += 1 + extra[idx]; /* Adjust for the alignment. */ idx = (idx + 3) & ~3; str[0] = (wchar_t) idx + 4; } else if (symb_table[2 * elem] == 0 && c1 == 1) { /* No valid character. Match it as a single byte character. */ had_char_class = false; BUF_PUSH(str[0]); /* Update the length of characters */ laststart[5]++; range_start = str[0]; /* Throw away the ] at the end of the collating symbol. */ PATFETCH (c); /* exit from the switch block. */ continue; } else FREE_STACK_RETURN (REG_ECOLLATE); } datasize = 1; } # endif /* Throw away the ] at the end of the equivalence class (or collating symbol). */ PATFETCH (c); /* Allocate the space for the equivalence class (or collating symbol) (and '\0' if needed). */ GET_BUFFER_SPACE(datasize); /* Update the pointer to indicate end of buffer. */ b += datasize; if (delim == '=') { /* equivalence class */ /* Calculate the offset of char_ranges, which is next to equivalence_classes. */ offset = laststart[1] + laststart[2] + laststart[3] +6; /* Insert space. */ insert_space(datasize, laststart + offset, b - 1); /* Write the equivalence_class and \0. */ for (i = 0 ; i < datasize ; i++) laststart[offset + i] = str[i]; /* Update the length of equivalence_classes. */ laststart[3] += datasize; had_char_class = true; } else /* delim == '.' */ { /* collating symbol */ /* Calculate the offset of the equivalence_classes, which is next to collating_symbols. */ offset = laststart[1] + laststart[2] + 6; /* Insert space and write the collationg_symbol and \0. */ insert_space(datasize, laststart + offset, b-1); for (i = 0 ; i < datasize ; i++) laststart[offset + i] = str[i]; /* In re_match_2_internal if range_start < -1, we assume -range_start is the offset of the collating symbol which is specified as the character of the range start. So we assign -(laststart[1] + laststart[2] + 6) to range_start. */ range_start = -(laststart[1] + laststart[2] + 6); /* Update the length of collating_symbol. */ laststart[2] += datasize; had_char_class = false; } } else { c1++; while (c1--) PATUNFETCH; BUF_PUSH ('['); BUF_PUSH (delim); laststart[5] += 2; /* Update the length of characters */ range_start = delim; had_char_class = false; } } else { had_char_class = false; BUF_PUSH(c); laststart[5]++; /* Update the length of characters */ range_start = c; } } #else /* BYTE */ /* 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); range_start = 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 = byte_compile_range (range_start, &p, pend, translate, syntax, b); if (ret != REG_NOERROR) FREE_STACK_RETURN (ret); range_start = 0xffffffff; } else if (p[0] == '-' && p[1] != ']') { /* This handles ranges made up of characters only. */ reg_errcode_t ret; /* Move past the `-'. */ PATFETCH (c1); ret = byte_compile_range (c, &p, pend, translate, syntax, b); if (ret != REG_NOERROR) FREE_STACK_RETURN (ret); range_start = 0xffffffff; } /* 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) { if (iswctype (btowc (ch), wt)) SET_LIST_BIT (ch); 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 (':'); range_start = ':'; had_char_class = false; } } else if (syntax & RE_CHAR_CLASSES && c == '[' && *p == '=') { unsigned char str[MB_LEN_MAX + 1]; # ifdef _LIBC uint32_t nrules = _NL_CURRENT_WORD (LC_COLLATE, _NL_COLLATE_NRULES); # endif 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 < MB_LEN_MAX) str[c1++] = c; else /* This is in any case an invalid class name. */ str[0] = '\0'; } str[c1] = '\0'; if (c == '=' && *p == ']' && str[0] != '\0') { /* If we have no collation data we use the default collation in which each character is in a class by itself. It also means that ASCII is the character set and therefore we cannot have character with more than one byte in the multibyte representation. */ # ifdef _LIBC if (nrules == 0) # endif { if (c1 != 1) FREE_STACK_RETURN (REG_ECOLLATE); /* Throw away the ] at the end of the equivalence class. */ PATFETCH (c); /* Set the bit for the character. */ SET_LIST_BIT (str[0]); } # ifdef _LIBC else { /* Try to match the byte sequence in `str' against those known to the collate implementation. First find out whether the bytes in `str' are actually from exactly one character. */ const int32_t *table; const unsigned char *weights; const unsigned char *extra; const int32_t *indirect; int32_t idx; const unsigned char *cp = str; int ch; /* This #include defines a local function! */ # include table = (const int32_t *) _NL_CURRENT (LC_COLLATE, _NL_COLLATE_TABLEMB); weights = (const unsigned char *) _NL_CURRENT (LC_COLLATE, _NL_COLLATE_WEIGHTMB); extra = (const unsigned char *) _NL_CURRENT (LC_COLLATE, _NL_COLLATE_EXTRAMB); indirect = (const int32_t *) _NL_CURRENT (LC_COLLATE, _NL_COLLATE_INDIRECTMB); idx = findidx (&cp); if (idx == 0 || cp < str + c1) /* This is no valid character. */ FREE_STACK_RETURN (REG_ECOLLATE); /* Throw away the ] at the end of the equivalence class. */ PATFETCH (c); /* Now we have to go throught the whole table and find all characters which have the same first level weight. XXX Note that this is not entirely correct. we would have to match multibyte sequences but this is not possible with the current implementation. */ for (ch = 1; ch < 256; ++ch) /* XXX This test would have to be changed if we would allow matching multibyte sequences. */ if (table[ch] > 0) { int32_t idx2 = table[ch]; size_t len = weights[idx2]; /* Test whether the lenghts match. */ if (weights[idx] == len) { /* They do. New compare the bytes of the weight. */ size_t cnt = 0; while (cnt < len && (weights[idx + 1 + cnt] == weights[idx2 + 1 + cnt])) ++cnt; if (cnt == len) /* They match. Mark the character as acceptable. */ SET_LIST_BIT (ch); } } } # endif had_char_class = true; } else { c1++; while (c1--) PATUNFETCH; SET_LIST_BIT ('['); SET_LIST_BIT ('='); range_start = '='; had_char_class = false; } } else if (syntax & RE_CHAR_CLASSES && c == '[' && *p == '.') { unsigned char str[128]; /* Should be large enough. */ # ifdef _LIBC uint32_t nrules = _NL_CURRENT_WORD (LC_COLLATE, _NL_COLLATE_NRULES); # endif 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 < sizeof (str)) str[c1++] = c; else /* This is in any case an invalid class name. */ str[0] = '\0'; } str[c1] = '\0'; if (c == '.' && *p == ']' && str[0] != '\0') { /* If we have no collation data we use the default collation in which each character is the name for its own class which contains only the one character. It also means that ASCII is the character set and therefore we cannot have character with more than one byte in the multibyte representation. */ # ifdef _LIBC if (nrules == 0) # endif { if (c1 != 1) FREE_STACK_RETURN (REG_ECOLLATE); /* Throw away the ] at the end of the equivalence class. */ PATFETCH (c); /* Set the bit for the character. */ SET_LIST_BIT (str[0]); range_start = ((const unsigned char *) str)[0]; } # ifdef _LIBC else { /* Try to match the byte sequence in `str' against those known to the collate implementation. First find out whether the bytes in `str' are actually from exactly one character. */ int32_t table_size; const int32_t *symb_table; const unsigned char *extra; int32_t idx; int32_t elem; int32_t second; int32_t hash; table_size = _NL_CURRENT_WORD (LC_COLLATE, _NL_COLLATE_SYMB_HASH_SIZEMB); symb_table = (const int32_t *) _NL_CURRENT (LC_COLLATE, _NL_COLLATE_SYMB_TABLEMB); extra = (const unsigned char *) _NL_CURRENT (LC_COLLATE, _NL_COLLATE_SYMB_EXTRAMB); /* Locate the character in the hashing table. */ hash = elem_hash (str, c1); idx = 0; elem = hash % table_size; second = hash % (table_size - 2); while (symb_table[2 * elem] != 0) { /* First compare the hashing value. */ if (symb_table[2 * elem] == hash && c1 == extra[symb_table[2 * elem + 1]] && memcmp (str, &extra[symb_table[2 * elem + 1] + 1], c1) == 0) { /* Yep, this is the entry. */ idx = symb_table[2 * elem + 1]; idx += 1 + extra[idx]; break; } /* Next entry. */ elem += second; } if (symb_table[2 * elem] == 0) /* This is no valid character. */ FREE_STACK_RETURN (REG_ECOLLATE); /* Throw away the ] at the end of the equivalence class. */ PATFETCH (c); /* Now add the multibyte character(s) we found to the accept list. XXX Note that this is not entirely correct. we would have to match multibyte sequences but this is not possible with the current implementation. Also, we have to match collating symbols, which expand to more than one file, as a whole and not allow the individual bytes. */ c1 = extra[idx++]; if (c1 == 1) range_start = extra[idx]; while (c1-- > 0) { SET_LIST_BIT (extra[idx]); ++idx; } } # endif had_char_class = false; } else { c1++; while (c1--) PATUNFETCH; SET_LIST_BIT ('['); SET_LIST_BIT ('.'); range_start = '.'; had_char_class = false; } } else { had_char_class = false; SET_LIST_BIT (c); range_start = 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]; #endif /* WCHAR */ } 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 - COMPILED_BUFFER_VAR; COMPILE_STACK_TOP.fixup_alt_jump = fixup_alt_jump ? fixup_alt_jump - COMPILED_BUFFER_VAR + 1 : 0; COMPILE_STACK_TOP.laststart_offset = b - COMPILED_BUFFER_VAR; 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 - COMPILED_BUFFER_VAR + 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 = COMPILED_BUFFER_VAR + COMPILE_STACK_TOP.begalt_offset; fixup_alt_jump = COMPILE_STACK_TOP.fixup_alt_jump ? COMPILED_BUFFER_VAR + COMPILE_STACK_TOP.fixup_alt_jump - 1 : 0; laststart = COMPILED_BUFFER_VAR + 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) { UCHAR_T *inner_group_loc = COMPILED_BUFFER_VAR + 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 (1 + OFFSET_ADDRESS_SIZE); INSERT_JUMP (on_failure_jump, begalt, b + 2 + 2 * OFFSET_ADDRESS_SIZE); pending_exact = 0; b += 1 + OFFSET_ADDRESS_SIZE; /* 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 (1 + OFFSET_ADDRESS_SIZE); b += 1 + OFFSET_ADDRESS_SIZE; 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_NO_BK_BRACES)) 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; /* Place in the uncompiled pattern (i.e., just after the '{') to go back to if the interval is invalid. */ const CHAR_T *beg_interval = p; if (p == pend) goto invalid_interval; 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 (! (0 <= lower_bound && lower_bound <= upper_bound)) goto invalid_interval; if (!(syntax & RE_NO_BK_BRACES)) { if (c != '\\' || p == pend) goto invalid_interval; PATFETCH (c); } if (c != '}') goto invalid_interval; /* If it's invalid to have no preceding re. */ if (!laststart) { if (syntax & RE_CONTEXT_INVALID_OPS && !(syntax & RE_INVALID_INTERVAL_ORD)) FREE_STACK_RETURN (REG_BADRPT); else if (syntax & RE_CONTEXT_INDEP_OPS) laststart = b; else goto unfetch_interval; } /* We just parsed a valid interval. */ if (RE_DUP_MAX < upper_bound) FREE_STACK_RETURN (REG_BADBR); /* 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. */ /* ifdef WCHAR, 'b + 1 + OFFSET_ADDRESS_SIZE' instead of 'b + 3'. */ if (upper_bound == 0) { GET_BUFFER_SPACE (1 + OFFSET_ADDRESS_SIZE); INSERT_JUMP (jump, laststart, b + 1 + OFFSET_ADDRESS_SIZE); b += 1 + OFFSET_ADDRESS_SIZE; } /* 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 = 2 + 4 * OFFSET_ADDRESS_SIZE + (upper_bound > 1) * (2 + 4 * OFFSET_ADDRESS_SIZE); 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 + 1 + 2 * OFFSET_ADDRESS_SIZE + (upper_bound > 1) * (1 + 2 * OFFSET_ADDRESS_SIZE) , lower_bound); b += 1 + 2 * OFFSET_ADDRESS_SIZE; /* 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'. */ /* ifdef WCHAR, The '1+2*OFFSET_ADDRESS_SIZE' is the 'set_number_at', plus '1+OFFSET_ADDRESS_SIZE' of the following `succeed_n'. */ PREFIX(insert_op2) (set_number_at, laststart, 1 + 2 * OFFSET_ADDRESS_SIZE, lower_bound, b); b += 1 + 2 * OFFSET_ADDRESS_SIZE; 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 + 2 * OFFSET_ADDRESS_SIZE + 1, upper_bound - 1); b += 1 + 2 * OFFSET_ADDRESS_SIZE; /* 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. */ PREFIX(insert_op2) (set_number_at, laststart, b - laststart, upper_bound - 1, b); b += 1 + 2 * OFFSET_ADDRESS_SIZE; } } pending_exact = 0; break; invalid_interval: if (!(syntax & RE_INVALID_INTERVAL_ORD)) FREE_STACK_RETURN (p == pend ? REG_EBRACE : REG_BADBR); unfetch_interval: /* Match the characters as literals. */ p = beg_interval; c = '{'; if (syntax & RE_NO_BK_BRACES) goto normal_char; else goto normal_backslash; } #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);