Home | History | Annotate | Download | only in include
      1 /* Definitions for data structures and routines for the regular
      2    expression library.
      3    Copyright (C) 1985,1989-93,1995-98,2000,2001,2002,2003,2005,2006
      4    Free Software Foundation, Inc.
      5    This file is part of the GNU C Library.
      6 
      7    The GNU C Library is free software; you can redistribute it and/or
      8    modify it under the terms of the GNU Lesser General Public
      9    License as published by the Free Software Foundation; either
     10    version 2.1 of the License, or (at your option) any later version.
     11 
     12    The GNU C Library is distributed in the hope that it will be useful,
     13    but WITHOUT ANY WARRANTY; without even the implied warranty of
     14    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
     15    Lesser General Public License for more details.
     16 
     17    You should have received a copy of the GNU Lesser General Public
     18    License along with the GNU C Library; if not, write to the Free
     19    Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
     20    02111-1307 USA.  */
     21 
     22 #ifndef _REGEX_H
     23 #define _REGEX_H 1
     24 
     25 #include <sys/types.h>
     26 
     27 /* Allow the use in C++ code.  */
     28 #ifdef __cplusplus
     29 extern "C" {
     30 #endif
     31 
     32 /* The following two types have to be signed and unsigned integer type
     33    wide enough to hold a value of a pointer.  For most ANSI compilers
     34    ptrdiff_t and size_t should be likely OK.  Still size of these two
     35    types is 2 for Microsoft C.  Ugh... */
     36 typedef long int s_reg_t;
     37 typedef unsigned long int active_reg_t;
     38 
     39 /* The following bits are used to determine the regexp syntax we
     40    recognize.  The set/not-set meanings are chosen so that Emacs syntax
     41    remains the value 0.  The bits are given in alphabetical order, and
     42    the definitions shifted by one from the previous bit; thus, when we
     43    add or remove a bit, only one other definition need change.  */
     44 typedef unsigned long int reg_syntax_t;
     45 
     46 /* If this bit is not set, then \ inside a bracket expression is literal.
     47    If set, then such a \ quotes the following character.  */
     48 #define RE_BACKSLASH_ESCAPE_IN_LISTS ((unsigned long int) 1)
     49 
     50 /* If this bit is not set, then + and ? are operators, and \+ and \? are
     51      literals.
     52    If set, then \+ and \? are operators and + and ? are literals.  */
     53 #define RE_BK_PLUS_QM (RE_BACKSLASH_ESCAPE_IN_LISTS << 1)
     54 
     55 /* If this bit is set, then character classes are supported.  They are:
     56      [:alpha:], [:upper:], [:lower:],  [:digit:], [:alnum:], [:xdigit:],
     57      [:space:], [:print:], [:punct:], [:graph:], and [:cntrl:].
     58    If not set, then character classes are not supported.  */
     59 #define RE_CHAR_CLASSES (RE_BK_PLUS_QM << 1)
     60 
     61 /* If this bit is set, then ^ and $ are always anchors (outside bracket
     62      expressions, of course).
     63    If this bit is not set, then it depends:
     64         ^  is an anchor if it is at the beginning of a regular
     65            expression or after an open-group or an alternation operator;
     66         $  is an anchor if it is at the end of a regular expression, or
     67            before a close-group or an alternation operator.
     68 
     69    This bit could be (re)combined with RE_CONTEXT_INDEP_OPS, because
     70    POSIX draft 11.2 says that * etc. in leading positions is undefined.
     71    We already implemented a previous draft which made those constructs
     72    invalid, though, so we haven't changed the code back.  */
     73 #define RE_CONTEXT_INDEP_ANCHORS (RE_CHAR_CLASSES << 1)
     74 
     75 /* If this bit is set, then special characters are always special
     76      regardless of where they are in the pattern.
     77    If this bit is not set, then special characters are special only in
     78      some contexts; otherwise they are ordinary.  Specifically,
     79      * + ? and intervals are only special when not after the beginning,
     80      open-group, or alternation operator.  */
     81 #define RE_CONTEXT_INDEP_OPS (RE_CONTEXT_INDEP_ANCHORS << 1)
     82 
     83 /* If this bit is set, then *, +, ?, and { cannot be first in an re or
     84      immediately after an alternation or begin-group operator.  */
     85 #define RE_CONTEXT_INVALID_OPS (RE_CONTEXT_INDEP_OPS << 1)
     86 
     87 /* If this bit is set, then . matches newline.
     88    If not set, then it doesn't.  */
     89 #define RE_DOT_NEWLINE (RE_CONTEXT_INVALID_OPS << 1)
     90 
     91 /* If this bit is set, then . doesn't match NUL.
     92    If not set, then it does.  */
     93 #define RE_DOT_NOT_NULL (RE_DOT_NEWLINE << 1)
     94 
     95 /* If this bit is set, nonmatching lists [^...] do not match newline.
     96    If not set, they do.  */
     97 #define RE_HAT_LISTS_NOT_NEWLINE (RE_DOT_NOT_NULL << 1)
     98 
     99 /* If this bit is set, either \{...\} or {...} defines an
    100      interval, depending on RE_NO_BK_BRACES.
    101    If not set, \{, \}, {, and } are literals.  */
    102 #define RE_INTERVALS (RE_HAT_LISTS_NOT_NEWLINE << 1)
    103 
    104 /* If this bit is set, +, ? and | aren't recognized as operators.
    105    If not set, they are.  */
    106 #define RE_LIMITED_OPS (RE_INTERVALS << 1)
    107 
    108 /* If this bit is set, newline is an alternation operator.
    109    If not set, newline is literal.  */
    110 #define RE_NEWLINE_ALT (RE_LIMITED_OPS << 1)
    111 
    112 /* If this bit is set, then `{...}' defines an interval, and \{ and \}
    113      are literals.
    114   If not set, then `\{...\}' defines an interval.  */
    115 #define RE_NO_BK_BRACES (RE_NEWLINE_ALT << 1)
    116 
    117 /* If this bit is set, (...) defines a group, and \( and \) are literals.
    118    If not set, \(...\) defines a group, and ( and ) are literals.  */
    119 #define RE_NO_BK_PARENS (RE_NO_BK_BRACES << 1)
    120 
    121 /* If this bit is set, then \<digit> matches <digit>.
    122    If not set, then \<digit> is a back-reference.  */
    123 #define RE_NO_BK_REFS (RE_NO_BK_PARENS << 1)
    124 
    125 /* If this bit is set, then | is an alternation operator, and \| is literal.
    126    If not set, then \| is an alternation operator, and | is literal.  */
    127 #define RE_NO_BK_VBAR (RE_NO_BK_REFS << 1)
    128 
    129 /* If this bit is set, then an ending range point collating higher
    130      than the starting range point, as in [z-a], is invalid.
    131    If not set, then when ending range point collates higher than the
    132      starting range point, the range is ignored.  */
    133 #define RE_NO_EMPTY_RANGES (RE_NO_BK_VBAR << 1)
    134 
    135 /* If this bit is set, then an unmatched ) is ordinary.
    136    If not set, then an unmatched ) is invalid.  */
    137 #define RE_UNMATCHED_RIGHT_PAREN_ORD (RE_NO_EMPTY_RANGES << 1)
    138 
    139 /* If this bit is set, succeed as soon as we match the whole pattern,
    140    without further backtracking.  */
    141 #define RE_NO_POSIX_BACKTRACKING (RE_UNMATCHED_RIGHT_PAREN_ORD << 1)
    142 
    143 /* If this bit is set, do not process the GNU regex operators.
    144    If not set, then the GNU regex operators are recognized. */
    145 #define RE_NO_GNU_OPS (RE_NO_POSIX_BACKTRACKING << 1)
    146 
    147 /* If this bit is set, turn on internal regex debugging.
    148    If not set, and debugging was on, turn it off.
    149    This only works if regex.c is compiled -DDEBUG.
    150    We define this bit always, so that all that's needed to turn on
    151    debugging is to recompile regex.c; the calling code can always have
    152    this bit set, and it won't affect anything in the normal case. */
    153 #define RE_DEBUG (RE_NO_GNU_OPS << 1)
    154 
    155 /* If this bit is set, a syntactically invalid interval is treated as
    156    a string of ordinary characters.  For example, the ERE 'a{1' is
    157    treated as 'a\{1'.  */
    158 #define RE_INVALID_INTERVAL_ORD (RE_DEBUG << 1)
    159 
    160 /* If this bit is set, then ignore case when matching.
    161    If not set, then case is significant.  */
    162 #define RE_ICASE (RE_INVALID_INTERVAL_ORD << 1)
    163 
    164 /* This bit is used internally like RE_CONTEXT_INDEP_ANCHORS but only
    165    for ^, because it is difficult to scan the regex backwards to find
    166    whether ^ should be special.  */
    167 #define RE_CARET_ANCHORS_HERE (RE_ICASE << 1)
    168 
    169 /* If this bit is set, then \{ cannot be first in an bre or
    170    immediately after an alternation or begin-group operator.  */
    171 #define RE_CONTEXT_INVALID_DUP (RE_CARET_ANCHORS_HERE << 1)
    172 
    173 /* If this bit is set, then no_sub will be set to 1 during
    174    re_compile_pattern.  */
    175 #define RE_NO_SUB (RE_CONTEXT_INVALID_DUP << 1)
    176 
    177 /* This global variable defines the particular regexp syntax to use (for
    178    some interfaces).  When a regexp is compiled, the syntax used is
    179    stored in the pattern buffer, so changing this does not affect
    180    already-compiled regexps.  */
    181 extern reg_syntax_t re_syntax_options;
    182 
    183 /* Define combinations of the above bits for the standard possibilities.
    185    (The [[[ comments delimit what gets put into the Texinfo file, so
    186    don't delete them!)  */
    187 /* [[[begin syntaxes]]] */
    188 #define RE_SYNTAX_EMACS 0
    189 
    190 #define RE_SYNTAX_AWK							\
    191   (RE_BACKSLASH_ESCAPE_IN_LISTS   | RE_DOT_NOT_NULL			\
    192    | RE_NO_BK_PARENS              | RE_NO_BK_REFS			\
    193    | RE_NO_BK_VBAR                | RE_NO_EMPTY_RANGES			\
    194    | RE_DOT_NEWLINE		  | RE_CONTEXT_INDEP_ANCHORS		\
    195    | RE_UNMATCHED_RIGHT_PAREN_ORD | RE_NO_GNU_OPS)
    196 
    197 #define RE_SYNTAX_GNU_AWK						\
    198   ((RE_SYNTAX_POSIX_EXTENDED | RE_BACKSLASH_ESCAPE_IN_LISTS | RE_DEBUG)	\
    199    & ~(RE_DOT_NOT_NULL | RE_INTERVALS | RE_CONTEXT_INDEP_OPS		\
    200        | RE_CONTEXT_INVALID_OPS ))
    201 
    202 #define RE_SYNTAX_POSIX_AWK						\
    203   (RE_SYNTAX_POSIX_EXTENDED | RE_BACKSLASH_ESCAPE_IN_LISTS		\
    204    | RE_INTERVALS	    | RE_NO_GNU_OPS)
    205 
    206 #define RE_SYNTAX_GREP							\
    207   (RE_BK_PLUS_QM              | RE_CHAR_CLASSES				\
    208    | RE_HAT_LISTS_NOT_NEWLINE | RE_INTERVALS				\
    209    | RE_NEWLINE_ALT)
    210 
    211 #define RE_SYNTAX_EGREP							\
    212   (RE_CHAR_CLASSES        | RE_CONTEXT_INDEP_ANCHORS			\
    213    | RE_CONTEXT_INDEP_OPS | RE_HAT_LISTS_NOT_NEWLINE			\
    214    | RE_NEWLINE_ALT       | RE_NO_BK_PARENS				\
    215    | RE_NO_BK_VBAR)
    216 
    217 #define RE_SYNTAX_POSIX_EGREP						\
    218   (RE_SYNTAX_EGREP | RE_INTERVALS | RE_NO_BK_BRACES			\
    219    | RE_INVALID_INTERVAL_ORD)
    220 
    221 /* P1003.2/D11.2, section 4.20.7.1, lines 5078ff.  */
    222 #define RE_SYNTAX_ED RE_SYNTAX_POSIX_BASIC
    223 
    224 #define RE_SYNTAX_SED RE_SYNTAX_POSIX_BASIC
    225 
    226 /* Syntax bits common to both basic and extended POSIX regex syntax.  */
    227 #define _RE_SYNTAX_POSIX_COMMON						\
    228   (RE_CHAR_CLASSES | RE_DOT_NEWLINE      | RE_DOT_NOT_NULL		\
    229    | RE_INTERVALS  | RE_NO_EMPTY_RANGES)
    230 
    231 #define RE_SYNTAX_POSIX_BASIC						\
    232   (_RE_SYNTAX_POSIX_COMMON | RE_BK_PLUS_QM | RE_CONTEXT_INVALID_DUP)
    233 
    234 /* Differs from ..._POSIX_BASIC only in that RE_BK_PLUS_QM becomes
    235    RE_LIMITED_OPS, i.e., \? \+ \| are not recognized.  Actually, this
    236    isn't minimal, since other operators, such as \`, aren't disabled.  */
    237 #define RE_SYNTAX_POSIX_MINIMAL_BASIC					\
    238   (_RE_SYNTAX_POSIX_COMMON | RE_LIMITED_OPS)
    239 
    240 #define RE_SYNTAX_POSIX_EXTENDED					\
    241   (_RE_SYNTAX_POSIX_COMMON  | RE_CONTEXT_INDEP_ANCHORS			\
    242    | RE_CONTEXT_INDEP_OPS   | RE_NO_BK_BRACES				\
    243    | RE_NO_BK_PARENS        | RE_NO_BK_VBAR				\
    244    | RE_CONTEXT_INVALID_OPS | RE_UNMATCHED_RIGHT_PAREN_ORD)
    245 
    246 /* Differs from ..._POSIX_EXTENDED in that RE_CONTEXT_INDEP_OPS is
    247    removed and RE_NO_BK_REFS is added.  */
    248 #define RE_SYNTAX_POSIX_MINIMAL_EXTENDED				\
    249   (_RE_SYNTAX_POSIX_COMMON  | RE_CONTEXT_INDEP_ANCHORS			\
    250    | RE_CONTEXT_INVALID_OPS | RE_NO_BK_BRACES				\
    251    | RE_NO_BK_PARENS        | RE_NO_BK_REFS				\
    252    | RE_NO_BK_VBAR	    | RE_UNMATCHED_RIGHT_PAREN_ORD)
    253 /* [[[end syntaxes]]] */
    254 
    255 /* Maximum number of duplicates an interval can allow.  Some systems
    257    (erroneously) define this in other header files, but we want our
    258    value, so remove any previous define.  */
    259 #ifdef RE_DUP_MAX
    260 # undef RE_DUP_MAX
    261 #endif
    262 /* If sizeof(int) == 2, then ((1 << 15) - 1) overflows.  */
    263 #define RE_DUP_MAX (0x7fff)
    264 
    265 
    266 /* POSIX `cflags' bits (i.e., information for `regcomp').  */
    267 
    268 /* If this bit is set, then use extended regular expression syntax.
    269    If not set, then use basic regular expression syntax.  */
    270 #define REG_EXTENDED 1
    271 
    272 /* If this bit is set, then ignore case when matching.
    273    If not set, then case is significant.  */
    274 #define REG_ICASE (REG_EXTENDED << 1)
    275 
    276 /* If this bit is set, then anchors do not match at newline
    277      characters in the string.
    278    If not set, then anchors do match at newlines.  */
    279 #define REG_NEWLINE (REG_ICASE << 1)
    280 
    281 /* If this bit is set, then report only success or fail in regexec.
    282    If not set, then returns differ between not matching and errors.  */
    283 #define REG_NOSUB (REG_NEWLINE << 1)
    284 
    285 
    286 /* POSIX `eflags' bits (i.e., information for regexec).  */
    287 
    288 /* If this bit is set, then the beginning-of-line operator doesn't match
    289      the beginning of the string (presumably because it's not the
    290      beginning of a line).
    291    If not set, then the beginning-of-line operator does match the
    292      beginning of the string.  */
    293 #define REG_NOTBOL 1
    294 
    295 /* Like REG_NOTBOL, except for the end-of-line.  */
    296 #define REG_NOTEOL (1 << 1)
    297 
    298 /* Use PMATCH[0] to delimit the start and end of the search in the
    299    buffer.  */
    300 #define REG_STARTEND (1 << 2)
    301 
    302 
    303 /* If any error codes are removed, changed, or added, update the
    304    `re_error_msg' table in regex.c.  */
    305 typedef enum
    306 {
    307 #ifdef _XOPEN_SOURCE
    308   REG_ENOSYS = -1,	/* This will never happen for this implementation.  */
    309 #endif
    310 
    311   REG_NOERROR = 0,	/* Success.  */
    312   REG_NOMATCH,		/* Didn't find a match (for regexec).  */
    313 
    314   /* POSIX regcomp return error codes.  (In the order listed in the
    315      standard.)  */
    316   REG_BADPAT,		/* Invalid pattern.  */
    317   REG_ECOLLATE,		/* Inalid collating element.  */
    318   REG_ECTYPE,		/* Invalid character class name.  */
    319   REG_EESCAPE,		/* Trailing backslash.  */
    320   REG_ESUBREG,		/* Invalid back reference.  */
    321   REG_EBRACK,		/* Unmatched left bracket.  */
    322   REG_EPAREN,		/* Parenthesis imbalance.  */
    323   REG_EBRACE,		/* Unmatched \{.  */
    324   REG_BADBR,		/* Invalid contents of \{\}.  */
    325   REG_ERANGE,		/* Invalid range end.  */
    326   REG_ESPACE,		/* Ran out of memory.  */
    327   REG_BADRPT,		/* No preceding re for repetition op.  */
    328 
    329   /* Error codes we've added.  */
    330   REG_EEND,		/* Premature end.  */
    331   REG_ESIZE,		/* Compiled pattern bigger than 2^16 bytes.  */
    332   REG_ERPAREN		/* Unmatched ) or \); not returned from regcomp.  */
    333 } reg_errcode_t;
    334 
    335 /* This data structure represents a compiled pattern.  Before calling
    337    the pattern compiler, the fields `buffer', `allocated', `fastmap',
    338    `translate', and `no_sub' can be set.  After the pattern has been
    339    compiled, the `re_nsub' field is available.  All other fields are
    340    private to the regex routines.  */
    341 
    342 #ifndef RE_TRANSLATE_TYPE
    343 # define RE_TRANSLATE_TYPE unsigned char *
    344 #endif
    345 
    346 struct re_pattern_buffer
    347 {
    348   /* Space that holds the compiled pattern.  It is declared as
    349      `unsigned char *' because its elements are sometimes used as
    350      array indexes.  */
    351   unsigned char *buffer;
    352 
    353   /* Number of bytes to which `buffer' points.  */
    354   unsigned long int allocated;
    355 
    356   /* Number of bytes actually used in `buffer'.  */
    357   unsigned long int used;
    358 
    359   /* Syntax setting with which the pattern was compiled.  */
    360   reg_syntax_t syntax;
    361 
    362   /* Pointer to a fastmap, if any, otherwise zero.  re_search uses the
    363      fastmap, if there is one, to skip over impossible starting points
    364      for matches.  */
    365   char *fastmap;
    366 
    367   /* Either a translate table to apply to all characters before
    368      comparing them, or zero for no translation.  The translation is
    369      applied to a pattern when it is compiled and to a string when it
    370      is matched.  */
    371   RE_TRANSLATE_TYPE translate;
    372 
    373   /* Number of subexpressions found by the compiler.  */
    374   size_t re_nsub;
    375 
    376   /* Zero if this pattern cannot match the empty string, one else.
    377      Well, in truth it's used only in `re_search_2', to see whether or
    378      not we should use the fastmap, so we don't set this absolutely
    379      perfectly; see `re_compile_fastmap' (the `duplicate' case).  */
    380   unsigned can_be_null : 1;
    381 
    382   /* If REGS_UNALLOCATED, allocate space in the `regs' structure
    383      for `max (RE_NREGS, re_nsub + 1)' groups.
    384      If REGS_REALLOCATE, reallocate space if necessary.
    385      If REGS_FIXED, use what's there.  */
    386 #define REGS_UNALLOCATED 0
    387 #define REGS_REALLOCATE 1
    388 #define REGS_FIXED 2
    389   unsigned regs_allocated : 2;
    390 
    391   /* Set to zero when `regex_compile' compiles a pattern; set to one
    392      by `re_compile_fastmap' if it updates the fastmap.  */
    393   unsigned fastmap_accurate : 1;
    394 
    395   /* If set, `re_match_2' does not return information about
    396      subexpressions.  */
    397   unsigned no_sub : 1;
    398 
    399   /* If set, a beginning-of-line anchor doesn't match at the beginning
    400      of the string.  */
    401   unsigned not_bol : 1;
    402 
    403   /* Similarly for an end-of-line anchor.  */
    404   unsigned not_eol : 1;
    405 
    406   /* If true, an anchor at a newline matches.  */
    407   unsigned newline_anchor : 1;
    408 };
    409 
    410 typedef struct re_pattern_buffer regex_t;
    411 
    412 /* Type for byte offsets within the string.  POSIX mandates this.  */
    414 typedef int regoff_t;
    415 
    416 
    417 /* This is the structure we store register match data in.  See
    418    regex.texinfo for a full description of what registers match.  */
    419 struct re_registers
    420 {
    421   unsigned num_regs;
    422   regoff_t *start;
    423   regoff_t *end;
    424 };
    425 
    426 
    427 /* If `regs_allocated' is REGS_UNALLOCATED in the pattern buffer,
    428    `re_match_2' returns information about at least this many registers
    429    the first time a `regs' structure is passed.  */
    430 #ifndef RE_NREGS
    431 # define RE_NREGS 30
    432 #endif
    433 
    434 
    435 /* POSIX specification for registers.  Aside from the different names than
    436    `re_registers', POSIX uses an array of structures, instead of a
    437    structure of arrays.  */
    438 typedef struct
    439 {
    440   regoff_t rm_so;  /* Byte offset from string's start to substring's start.  */
    441   regoff_t rm_eo;  /* Byte offset from string's start to substring's end.  */
    442 } regmatch_t;
    443 
    444 /* Declarations for routines.  */
    446 
    447 /* Sets the current default syntax to SYNTAX, and return the old syntax.
    448    You can also simply assign to the `re_syntax_options' variable.  */
    449 extern reg_syntax_t re_set_syntax (reg_syntax_t __syntax);
    450 
    451 /* Compile the regular expression PATTERN, with length LENGTH
    452    and syntax given by the global `re_syntax_options', into the buffer
    453    BUFFER.  Return NULL if successful, and an error string if not.  */
    454 extern const char *re_compile_pattern (const char *__pattern, size_t __length,
    455 				       struct re_pattern_buffer *__buffer);
    456 
    457 
    458 /* Compile a fastmap for the compiled pattern in BUFFER; used to
    459    accelerate searches.  Return 0 if successful and -2 if was an
    460    internal error.  */
    461 extern int re_compile_fastmap (struct re_pattern_buffer *__buffer);
    462 
    463 
    464 /* Search in the string STRING (with length LENGTH) for the pattern
    465    compiled into BUFFER.  Start searching at position START, for RANGE
    466    characters.  Return the starting position of the match, -1 for no
    467    match, or -2 for an internal error.  Also return register
    468    information in REGS (if REGS and BUFFER->no_sub are nonzero).  */
    469 extern int re_search (struct re_pattern_buffer *__buffer, const char *__string,
    470 		      int __length, int __start, int __range,
    471 		      struct re_registers *__regs);
    472 
    473 
    474 /* Like `re_search', but search in the concatenation of STRING1 and
    475    STRING2.  Also, stop searching at index START + STOP.  */
    476 extern int re_search_2 (struct re_pattern_buffer *__buffer,
    477 			const char *__string1, int __length1,
    478 			const char *__string2, int __length2, int __start,
    479 			int __range, struct re_registers *__regs, int __stop);
    480 
    481 
    482 /* Like `re_search', but return how many characters in STRING the regexp
    483    in BUFFER matched, starting at position START.  */
    484 extern int re_match (struct re_pattern_buffer *__buffer, const char *__string,
    485 		     int __length, int __start, struct re_registers *__regs);
    486 
    487 
    488 /* Relates to `re_match' as `re_search_2' relates to `re_search'.  */
    489 extern int re_match_2 (struct re_pattern_buffer *__buffer,
    490 		       const char *__string1, int __length1,
    491 		       const char *__string2, int __length2, int __start,
    492 		       struct re_registers *__regs, int __stop);
    493 
    494 
    495 /* Set REGS to hold NUM_REGS registers, storing them in STARTS and
    496    ENDS.  Subsequent matches using BUFFER and REGS will use this memory
    497    for recording register information.  STARTS and ENDS must be
    498    allocated with malloc, and must each be at least `NUM_REGS * sizeof
    499    (regoff_t)' bytes long.
    500 
    501    If NUM_REGS == 0, then subsequent matches should allocate their own
    502    register data.
    503 
    504    Unless this function is called, the first search or match using
    505    PATTERN_BUFFER will allocate its own register data, without
    506    freeing the old data.  */
    507 extern void re_set_registers (struct re_pattern_buffer *__buffer,
    508 			      struct re_registers *__regs,
    509 			      unsigned int __num_regs,
    510 			      regoff_t *__starts, regoff_t *__ends);
    511 
    512 #if defined _REGEX_RE_COMP || defined _LIBC
    513 # ifndef _CRAY
    514 /* 4.2 bsd compatibility.  */
    515 extern char *re_comp (const char *);
    516 extern int re_exec (const char *);
    517 # endif
    518 #endif
    519 
    520 /* GCC 2.95 and later have "__restrict"; C99 compilers have
    521    "restrict", and "configure" may have defined "restrict".  */
    522 #ifndef __restrict
    523 # if ! (2 < __GNUC__ || (2 == __GNUC__ && 95 <= __GNUC_MINOR__))
    524 #  if defined restrict || 199901L <= __STDC_VERSION__
    525 #   define __restrict restrict
    526 #  else
    527 #   define __restrict
    528 #  endif
    529 # endif
    530 #endif
    531 /* gcc 3.1 and up support the [restrict] syntax.  */
    532 #ifndef __restrict_arr
    533 # if (__GNUC__ > 3 || (__GNUC__ == 3 && __GNUC_MINOR__ >= 1)) \
    534      && !defined __GNUG__
    535 #  define __restrict_arr __restrict
    536 # else
    537 #  define __restrict_arr
    538 # endif
    539 #endif
    540 
    541 /* POSIX compatibility.  */
    542 extern int regcomp (regex_t *__restrict __preg,
    543 		    const char *__restrict __pattern,
    544 		    int __cflags);
    545 
    546 extern int regexec (const regex_t *__restrict __preg,
    547 		    const char *__restrict __string, size_t __nmatch,
    548 		    regmatch_t __pmatch[__restrict_arr],
    549 		    int __eflags);
    550 
    551 extern size_t regerror (int __errcode, const regex_t *__restrict __preg,
    552 			char *__restrict __errbuf, size_t __errbuf_size);
    553 
    554 extern void regfree (regex_t *__preg);
    555 
    556 
    557 #ifdef __cplusplus
    558 }
    559 #endif	/* C++ */
    560 
    561 #endif /* regex.h */
    562