# define PERL_IN_XSUB_RE
# endif
/* need access to debugger hooks */
-# ifndef DEBUGGING
+# if defined(PERL_EXT_RE_DEBUG) && !defined(DEBUGGING)
# define DEBUGGING
# endif
#endif
# define Perl_regexec_flags my_regexec
# define Perl_regdump my_regdump
# define Perl_regprop my_regprop
+# define Perl_re_intuit_start my_re_intuit_start
/* *These* symbols are masked to allow static link. */
# define Perl_pregexec my_pregexec
# define Perl_reginitcolors my_reginitcolors
+
+# define PERL_NO_GET_CONTEXT
#endif
/*SUPPRESS 112*/
*
**** Alterations to Henry's code are...
****
- **** Copyright (c) 1991-1998, Larry Wall
+ **** Copyright (c) 1991-1999, Larry Wall
****
**** You may distribute under the terms of either the GNU General Public
**** License or the Artistic License, as specified in the README file.
* regular-expression syntax might require a total rethink.
*/
#include "EXTERN.h"
+#define PERL_IN_REGEXEC_C
#include "perl.h"
+#ifdef PERL_IN_XSUB_RE
+# if defined(PERL_CAPI) || defined(PERL_OBJECT)
+# include "XSUB.h"
+# endif
+#endif
+
#include "regcomp.h"
#define RF_tainted 1 /* tainted information used? */
#define STATIC static
#endif
-#ifndef PERL_OBJECT
-typedef I32 CHECKPOINT;
-
/*
* Forwards.
*/
-static I32 regmatch _((regnode *prog));
-static I32 regrepeat _((regnode *p, I32 max));
-static I32 regrepeat_hard _((regnode *p, I32 max, I32 *lp));
-static I32 regtry _((regexp *prog, char *startpos));
-
-static bool reginclass _((char *p, I32 c));
-static bool reginclassutf8 _((regnode *f, U8* p));
-static CHECKPOINT regcppush _((I32 parenfloor));
-static char * regcppop _((void));
-static char * regcp_set_to _((I32 ss));
-static void cache_re _((regexp *prog));
-static void restore_pos _((void *arg));
-#endif
-
-#define REGINCLASS(p,c) (*(p) ? reginclass(p,c) : ANYOF_TEST(p,c))
+#define REGINCLASS(p,c) (ANYOF_FLAGS(p) ? reginclass(p,c) : ANYOF_BITMAP_TEST(p,c))
#define REGINCLASSUTF8(f,p) (ARG1(f) ? reginclassutf8(f,p) : swash_fetch((SV*)PL_regdata->data[ARG2(f)],p))
#define CHR_SVLEN(sv) (UTF ? sv_len_utf8(sv) : SvCUR(sv))
#define CHR_DIST(a,b) (UTF ? utf8_distance(a,b) : a - b)
-#ifndef PERL_OBJECT
-static U8 * reghop _((U8 *pos, I32 off));
-static U8 * reghopmaybe _((U8 *pos, I32 off));
-#endif
#define reghop_c(pos,off) ((char*)reghop((U8*)pos, off))
#define reghopmaybe_c(pos,off) ((char*)reghopmaybe((U8*)pos, off))
#define HOP(pos,off) (UTF ? reghop((U8*)pos, off) : (U8*)(pos + off))
#define HOPc(pos,off) ((char*)HOP(pos,off))
#define HOPMAYBEc(pos,off) ((char*)HOPMAYBE(pos,off))
+static void restore_pos(pTHXo_ void *arg);
+
+
STATIC CHECKPOINT
-regcppush(I32 parenfloor)
+S_regcppush(pTHX_ I32 parenfloor)
{
dTHR;
int retval = PL_savestack_ix;
SSCHECK(i + 5);
for (p = PL_regsize; p > parenfloor; p--) {
- SSPUSHPTR(PL_regendp[p]);
- SSPUSHPTR(PL_regstartp[p]);
+ SSPUSHINT(PL_regendp[p]);
+ SSPUSHINT(PL_regstartp[p]);
SSPUSHPTR(PL_reg_start_tmp[p]);
SSPUSHINT(p);
}
/* These are needed since we do not localize EVAL nodes: */
# define REGCP_SET DEBUG_r(PerlIO_printf(Perl_debug_log, \
- " Setting an EVAL scope, savestack=%i\n", \
- PL_savestack_ix)); lastcp = PL_savestack_ix
+ " Setting an EVAL scope, savestack=%"IVdf"\n", \
+ (IV)PL_savestack_ix)); lastcp = PL_savestack_ix
# define REGCP_UNWIND DEBUG_r(lastcp != PL_savestack_ix ? \
PerlIO_printf(Perl_debug_log, \
- " Clearing an EVAL scope, savestack=%i..%i\n", \
- lastcp, PL_savestack_ix) : 0); regcpblow(lastcp)
+ " Clearing an EVAL scope, savestack=%"IVdf"..%"IVdf"\n", \
+ (IV)lastcp, (IV)PL_savestack_ix) : 0); regcpblow(lastcp)
STATIC char *
-regcppop(void)
+S_regcppop(pTHX)
{
dTHR;
I32 i = SSPOPINT;
U32 paren = 0;
char *input;
- char *tmps;
+ I32 tmps;
assert(i == SAVEt_REGCONTEXT);
i = SSPOPINT;
input = (char *) SSPOPPTR;
for (i -= 3; i > 0; i -= 4) {
paren = (U32)SSPOPINT;
PL_reg_start_tmp[paren] = (char *) SSPOPPTR;
- PL_regstartp[paren] = (char *) SSPOPPTR;
- tmps = (char*)SSPOPPTR;
+ PL_regstartp[paren] = SSPOPINT;
+ tmps = SSPOPINT;
if (paren <= *PL_reglastparen)
PL_regendp[paren] = tmps;
DEBUG_r(
PerlIO_printf(Perl_debug_log,
- " restoring \\%d to %d(%d)..%d%s\n",
- paren, PL_regstartp[paren] - PL_regbol,
- PL_reg_start_tmp[paren] - PL_regbol,
- PL_regendp[paren] - PL_regbol,
+ " restoring \\%"UVuf" to %"IVdf"(%"IVdf")..%"IVdf"%s\n",
+ (UV)paren, (IV)PL_regstartp[paren],
+ (IV)(PL_reg_start_tmp[paren] - PL_bostr),
+ (IV)PL_regendp[paren],
(paren > *PL_reglastparen ? "(no)" : ""));
);
}
DEBUG_r(
if (*PL_reglastparen + 1 <= PL_regnpar) {
PerlIO_printf(Perl_debug_log,
- " restoring \\%d..\\%d to undef\n",
- *PL_reglastparen + 1, PL_regnpar);
+ " restoring \\%"IVdf"..\\%"IVdf" to undef\n",
+ (IV)(*PL_reglastparen + 1), (IV)PL_regnpar);
}
);
for (paren = *PL_reglastparen + 1; paren <= PL_regnpar; paren++) {
if (paren > PL_regsize)
- PL_regstartp[paren] = Nullch;
- PL_regendp[paren] = Nullch;
+ PL_regstartp[paren] = -1;
+ PL_regendp[paren] = -1;
}
return input;
}
STATIC char *
-regcp_set_to(I32 ss)
+S_regcp_set_to(pTHX_ I32 ss)
{
dTHR;
I32 tmp = PL_savestack_ix;
- pregexec - match a regexp against a string
*/
I32
-pregexec(register regexp *prog, char *stringarg, register char *strend,
+Perl_pregexec(pTHX_ register regexp *prog, char *stringarg, register char *strend,
char *strbeg, I32 minend, SV *screamer, U32 nosave)
/* strend: pointer to null at end of string */
/* strbeg: real beginning of string */
}
STATIC void
-cache_re(regexp *prog)
+S_cache_re(pTHX_ regexp *prog)
{
dTHR;
PL_regprecomp = prog->precomp; /* Needed for FAIL. */
PL_reg_re = prog;
}
-STATIC void
-restore_pos(void *arg)
+/*
+ * Need to implement the following flags for reg_anch:
+ *
+ * USE_INTUIT_NOML - Useful to call re_intuit_start() first
+ * USE_INTUIT_ML
+ * INTUIT_AUTORITATIVE_NOML - Can trust a positive answer
+ * INTUIT_AUTORITATIVE_ML
+ * INTUIT_ONCE_NOML - Intuit can match in one location only.
+ * INTUIT_ONCE_ML
+ *
+ * Another flag for this function: SECOND_TIME (so that float substrs
+ * with giant delta may be not rechecked).
+ */
+
+/* Assumptions: if ANCH_GPOS, then strpos is anchored. XXXX Check GPOS logic */
+
+/* If SCREAM, then SvPVX(sv) should be compatible with strpos and strend.
+ Otherwise, only SvCUR(sv) is used to get strbeg. */
+
+/* XXXX We assume that strpos is strbeg unless sv. */
+
+/* A failure to find a constant substring means that there is no need to make
+ an expensive call to REx engine, thus we celebrate a failure. Similarly,
+ finding a substring too deep into the string means that less calls to
+ regtry() should be needed.
+
+ REx compiler's optimizer found 4 possible hints:
+ a) Anchored substring;
+ b) Fixed substring;
+ c) Whether we are anchored (beginning-of-line or \G);
+ d) First node (of those at offset 0) which may distingush positions;
+ We use 'a', 'b', multiline-part of 'c', and try to find a position in the
+ string which does not contradict any of them.
+ */
+
+char *
+Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos,
+ char *strend, U32 flags, re_scream_pos_data *data)
{
- dTHR;
- if (PL_reg_eval_set) {
- PL_reg_magic->mg_len = PL_reg_oldpos;
- PL_reg_eval_set = 0;
- PL_curpm = PL_reg_oldcurpm;
- }
-}
+ register I32 start_shift;
+ /* Should be nonnegative! */
+ register I32 end_shift;
+ register char *s;
+ register SV *check;
+ char *t;
+ I32 ml_anch;
+ char *tmp;
+ register char *other_last = Nullch;
+#ifdef DEBUGGING
+ char *i_strpos = strpos;
+#endif
+
+ DEBUG_r( if (!PL_colorset) reginitcolors() );
+ DEBUG_r(PerlIO_printf(Perl_debug_log,
+ "%sGuessing start of match, REx%s `%s%.60s%s%s' against `%s%.*s%s%s'...\n",
+ PL_colors[4],PL_colors[5],PL_colors[0],
+ prog->precomp,
+ PL_colors[1],
+ (strlen(prog->precomp) > 60 ? "..." : ""),
+ PL_colors[0],
+ (int)(strend - strpos > 60 ? 60 : strend - strpos),
+ strpos, PL_colors[1],
+ (strend - strpos > 60 ? "..." : ""))
+ );
+
+ if (prog->minlen > strend - strpos) {
+ DEBUG_r(PerlIO_printf(Perl_debug_log, "String too short...\n"));
+ goto fail;
+ }
+ if (prog->reganch & ROPT_ANCH) { /* Match at beg-of-str or after \n */
+ ml_anch = !( (prog->reganch & ROPT_ANCH_SINGLE)
+ || ( (prog->reganch & ROPT_ANCH_BOL)
+ && !PL_multiline ) ); /* Check after \n? */
+
+ if ((prog->check_offset_min == prog->check_offset_max) && !ml_anch) {
+ /* Substring at constant offset from beg-of-str... */
+ I32 slen;
+
+ if ( !(prog->reganch & ROPT_ANCH_GPOS) /* Checked by the caller */
+ && (sv && (strpos + SvCUR(sv) != strend)) ) {
+ DEBUG_r(PerlIO_printf(Perl_debug_log, "Not at start...\n"));
+ goto fail;
+ }
+ PL_regeol = strend; /* Used in HOP() */
+ s = HOPc(strpos, prog->check_offset_min);
+ if (SvTAIL(prog->check_substr)) {
+ slen = SvCUR(prog->check_substr); /* >= 1 */
+
+ if ( strend - s > slen || strend - s < slen - 1
+ || (strend - s == slen && strend[-1] != '\n')) {
+ DEBUG_r(PerlIO_printf(Perl_debug_log, "String too long...\n"));
+ goto fail_finish;
+ }
+ /* Now should match s[0..slen-2] */
+ slen--;
+ if (slen && (*SvPVX(prog->check_substr) != *s
+ || (slen > 1
+ && memNE(SvPVX(prog->check_substr), s, slen)))) {
+ report_neq:
+ DEBUG_r(PerlIO_printf(Perl_debug_log, "String not equal...\n"));
+ goto fail_finish;
+ }
+ }
+ else if (*SvPVX(prog->check_substr) != *s
+ || ((slen = SvCUR(prog->check_substr)) > 1
+ && memNE(SvPVX(prog->check_substr), s, slen)))
+ goto report_neq;
+ goto success_at_start;
+ }
+ /* Match is anchored, but substr is not anchored wrt beg-of-str. */
+ s = strpos;
+ start_shift = prog->check_offset_min; /* okay to underestimate on CC */
+ /* Should be nonnegative! */
+ end_shift = prog->minlen - start_shift -
+ CHR_SVLEN(prog->check_substr) + (SvTAIL(prog->check_substr) != 0);
+ if (!ml_anch) {
+ I32 end = prog->check_offset_max + CHR_SVLEN(prog->check_substr)
+ - (SvTAIL(prog->check_substr) != 0);
+ I32 eshift = strend - s - end;
+
+ if (end_shift < eshift)
+ end_shift = eshift;
+ }
+ }
+ else { /* Can match at random position */
+ ml_anch = 0;
+ s = strpos;
+ start_shift = prog->check_offset_min; /* okay to underestimate on CC */
+ /* Should be nonnegative! */
+ end_shift = prog->minlen - start_shift -
+ CHR_SVLEN(prog->check_substr) + (SvTAIL(prog->check_substr) != 0);
+ }
+
+#ifdef DEBUGGING /* 7/99: reports of failure (with the older version) */
+ if (end_shift < 0)
+ Perl_croak(aTHX_ "panic: end_shift");
+#endif
+ check = prog->check_substr;
+ restart:
+ /* Find a possible match in the region s..strend by looking for
+ the "check" substring in the region corrected by start/end_shift. */
+ if (flags & REXEC_SCREAM) {
+ char *strbeg = SvPVX(sv); /* XXXX Assume PV_force() on SCREAM! */
+ I32 p = -1; /* Internal iterator of scream. */
+ I32 *pp = data ? data->scream_pos : &p;
+
+ if (PL_screamfirst[BmRARE(check)] >= 0
+ || ( BmRARE(check) == '\n'
+ && (BmPREVIOUS(check) == SvCUR(check) - 1)
+ && SvTAIL(check) ))
+ s = screaminstr(sv, check,
+ start_shift + (s - strbeg), end_shift, pp, 0);
+ else
+ goto fail_finish;
+ if (data)
+ *data->scream_olds = s;
+ }
+ else
+ s = fbm_instr((unsigned char*)s + start_shift,
+ (unsigned char*)strend - end_shift,
+ check, PL_multiline ? FBMrf_MULTILINE : 0);
+
+ /* Update the count-of-usability, remove useless subpatterns,
+ unshift s. */
+
+ DEBUG_r(PerlIO_printf(Perl_debug_log, "%s %s substr `%s%.*s%s'%s%s",
+ (s ? "Found" : "Did not find"),
+ ((check == prog->anchored_substr) ? "anchored" : "floating"),
+ PL_colors[0],
+ (int)(SvCUR(check) - (SvTAIL(check)!=0)),
+ SvPVX(check),
+ PL_colors[1], (SvTAIL(check) ? "$" : ""),
+ (s ? " at offset " : "...\n") ) );
+
+ if (!s)
+ goto fail_finish;
+
+ /* Finish the diagnostic message */
+ DEBUG_r(PerlIO_printf(Perl_debug_log, "%ld...\n", (long)(s - i_strpos)) );
+
+ /* Got a candidate. Check MBOL anchoring, and the *other* substr.
+ Start with the other substr.
+ XXXX no SCREAM optimization yet - and a very coarse implementation
+ XXXX /ttx+/ results in anchored=`ttx', floating=`x'. floating will
+ *always* match. Probably should be marked during compile...
+ Probably it is right to do no SCREAM here...
+ */
+
+ if (prog->float_substr && prog->anchored_substr) {
+ /* Take into account the "other" substring. */
+ /* XXXX May be hopelessly wrong for UTF... */
+ if (!other_last)
+ other_last = strpos - 1;
+ if (check == prog->float_substr) {
+ do_other_anchored:
+ {
+ char *last = s - start_shift, *last1, *last2;
+ char *s1 = s;
+
+ tmp = PL_bostr;
+ t = s - prog->check_offset_max;
+ if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */
+ && (!(prog->reganch & ROPT_UTF8)
+ || (PL_bostr = strpos, /* Used in regcopmaybe() */
+ (t = reghopmaybe_c(s, -(prog->check_offset_max)))
+ && t > strpos)))
+ /* EMPTY */;
+ else
+ t = strpos;
+ t += prog->anchored_offset;
+ if (t <= other_last)
+ t = other_last + 1;
+ PL_bostr = tmp;
+ last2 = last1 = strend - prog->minlen;
+ if (last < last1)
+ last1 = last;
+ /* XXXX It is not documented what units *_offsets are in. Assume bytes. */
+ /* On end-of-str: see comment below. */
+ s = fbm_instr((unsigned char*)t,
+ (unsigned char*)last1 + prog->anchored_offset
+ + SvCUR(prog->anchored_substr)
+ - (SvTAIL(prog->anchored_substr)!=0),
+ prog->anchored_substr, PL_multiline ? FBMrf_MULTILINE : 0);
+ DEBUG_r(PerlIO_printf(Perl_debug_log, "%s anchored substr `%s%.*s%s'%s",
+ (s ? "Found" : "Contradicts"),
+ PL_colors[0],
+ (int)(SvCUR(prog->anchored_substr)
+ - (SvTAIL(prog->anchored_substr)!=0)),
+ SvPVX(prog->anchored_substr),
+ PL_colors[1], (SvTAIL(prog->anchored_substr) ? "$" : "")));
+ if (!s) {
+ if (last1 >= last2) {
+ DEBUG_r(PerlIO_printf(Perl_debug_log,
+ ", giving up...\n"));
+ goto fail_finish;
+ }
+ DEBUG_r(PerlIO_printf(Perl_debug_log,
+ ", trying floating at offset %ld...\n",
+ (long)(s1 + 1 - i_strpos)));
+ PL_regeol = strend; /* Used in HOP() */
+ other_last = last1 + prog->anchored_offset;
+ s = HOPc(last, 1);
+ goto restart;
+ }
+ else {
+ DEBUG_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n",
+ (long)(s - i_strpos)));
+ t = s - prog->anchored_offset;
+ other_last = s - 1;
+ s = s1;
+ if (t == strpos)
+ goto try_at_start;
+ goto try_at_offset;
+ }
+ }
+ }
+ else { /* Take into account the floating substring. */
+ char *last, *last1;
+ char *s1 = s;
+
+ t = s - start_shift;
+ last1 = last = strend - prog->minlen + prog->float_min_offset;
+ if (last - t > prog->float_max_offset)
+ last = t + prog->float_max_offset;
+ s = t + prog->float_min_offset;
+ if (s <= other_last)
+ s = other_last + 1;
+ /* XXXX It is not documented what units *_offsets are in. Assume bytes. */
+ /* fbm_instr() takes into account exact value of end-of-str
+ if the check is SvTAIL(ed). Since false positives are OK,
+ and end-of-str is not later than strend we are OK. */
+ s = fbm_instr((unsigned char*)s,
+ (unsigned char*)last + SvCUR(prog->float_substr)
+ - (SvTAIL(prog->float_substr)!=0),
+ prog->float_substr, PL_multiline ? FBMrf_MULTILINE : 0);
+ DEBUG_r(PerlIO_printf(Perl_debug_log, "%s floating substr `%s%.*s%s'%s",
+ (s ? "Found" : "Contradicts"),
+ PL_colors[0],
+ (int)(SvCUR(prog->float_substr)
+ - (SvTAIL(prog->float_substr)!=0)),
+ SvPVX(prog->float_substr),
+ PL_colors[1], (SvTAIL(prog->float_substr) ? "$" : "")));
+ if (!s) {
+ if (last1 == last) {
+ DEBUG_r(PerlIO_printf(Perl_debug_log,
+ ", giving up...\n"));
+ goto fail_finish;
+ }
+ DEBUG_r(PerlIO_printf(Perl_debug_log,
+ ", trying anchored starting at offset %ld...\n",
+ (long)(s1 + 1 - i_strpos)));
+ other_last = last;
+ PL_regeol = strend; /* Used in HOP() */
+ s = HOPc(t, 1);
+ goto restart;
+ }
+ else {
+ DEBUG_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n",
+ (long)(s - i_strpos)));
+ other_last = s - 1;
+ s = s1;
+ if (t == strpos)
+ goto try_at_start;
+ goto try_at_offset;
+ }
+ }
+ }
+
+ t = s - prog->check_offset_max;
+ tmp = PL_bostr;
+ if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */
+ && (!(prog->reganch & ROPT_UTF8)
+ || (PL_bostr = strpos, /* Used in regcopmaybe() */
+ ((t = reghopmaybe_c(s, -(prog->check_offset_max)))
+ && t > strpos)))) {
+ PL_bostr = tmp;
+ /* Fixed substring is found far enough so that the match
+ cannot start at strpos. */
+ try_at_offset:
+ if (ml_anch && t[-1] != '\n') {
+ /* Eventually fbm_*() should handle this, but often
+ anchored_offset is not 0, so this check will not be wasted. */
+ /* XXXX In the code below we prefer to look for "^" even in
+ presence of anchored substrings. And we search even
+ beyond the found float position. These pessimizations
+ are historical artefacts only. */
+ find_anchor:
+ while (t < strend - prog->minlen) {
+ if (*t == '\n') {
+ if (t < s - prog->check_offset_min) {
+ if (prog->anchored_substr) {
+ /* We definitely contradict the found anchored
+ substr. Due to the above check we do not
+ contradict "check" substr.
+ Thus we can arrive here only if check substr
+ is float. Redo checking for "other"=="fixed".
+ */
+ strpos = t + 1;
+ DEBUG_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld, rescanning for anchored from offset %ld...\n",
+ PL_colors[0],PL_colors[1], (long)(strpos - i_strpos), (long)(strpos - i_strpos + prog->anchored_offset)));
+ goto do_other_anchored;
+ }
+ s = t + 1;
+ DEBUG_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld...\n",
+ PL_colors[0],PL_colors[1], (long)(s - i_strpos)));
+ goto set_useful;
+ }
+ DEBUG_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m, restarting at offset %ld...\n",
+ PL_colors[0],PL_colors[1], (long)(t + 1 - i_strpos)));
+ strpos = s = t + 1;
+ goto restart;
+ }
+ t++;
+ }
+ DEBUG_r(PerlIO_printf(Perl_debug_log, "Did not find /%s^%s/m...\n",
+ PL_colors[0],PL_colors[1]));
+ goto fail_finish;
+ }
+ s = t;
+ set_useful:
+ ++BmUSEFUL(prog->check_substr); /* hooray/5 */
+ }
+ else {
+ PL_bostr = tmp;
+ /* The found string does not prohibit matching at beg-of-str
+ - no optimization of calling REx engine can be performed,
+ unless it was an MBOL and we are not after MBOL. */
+ try_at_start:
+ /* Even in this situation we may use MBOL flag if strpos is offset
+ wrt the start of the string. */
+ if (ml_anch && sv
+ && (strpos + SvCUR(sv) != strend) && strpos[-1] != '\n') {
+ t = strpos;
+ goto find_anchor;
+ }
+ DEBUG_r( if (ml_anch)
+ PerlIO_printf(Perl_debug_log, "Does not contradict /%s^%s/m...\n",
+ PL_colors[0],PL_colors[1]);
+ );
+ success_at_start:
+ if (!(prog->reganch & ROPT_NAUGHTY) /* XXXX If strpos moved? */
+ && --BmUSEFUL(prog->check_substr) < 0
+ && prog->check_substr == prog->float_substr) { /* boo */
+ /* If flags & SOMETHING - do not do it many times on the same match */
+ SvREFCNT_dec(prog->check_substr);
+ prog->check_substr = Nullsv; /* disable */
+ prog->float_substr = Nullsv; /* clear */
+ s = strpos;
+ /* XXXX This is a remnant of the old implementation. It
+ looks wasteful, since now INTUIT can use many
+ other heuristics too. */
+ prog->reganch &= ~RE_USE_INTUIT;
+ }
+ else
+ s = strpos;
+ }
+
+ DEBUG_r(PerlIO_printf(Perl_debug_log, "%sGuessed:%s match at offset %ld\n",
+ PL_colors[4], PL_colors[5], (long)(s - i_strpos)) );
+ return s;
+
+ fail_finish: /* Substring not found */
+ BmUSEFUL(prog->check_substr) += 5; /* hooray */
+ fail:
+ DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch rejected by optimizer%s\n",
+ PL_colors[4],PL_colors[5]));
+ return Nullch;
+}
/*
- regexec_flags - match a regexp against a string
*/
I32
-regexec_flags(register regexp *prog, char *stringarg, register char *strend,
+Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char *strend,
char *strbeg, I32 minend, SV *sv, void *data, U32 flags)
/* strend: pointer to null at end of string */
/* strbeg: real beginning of string */
register I32 tmp;
I32 minlen; /* must match at least this many chars */
I32 dontbother = 0; /* how many characters not to try at end */
- CURCUR cc;
I32 start_shift = 0; /* Offset of the start to find
constant substr. */ /* CC */
I32 end_shift = 0; /* Same for the end. */ /* CC */
char *scream_olds;
SV* oreplsv = GvSV(PL_replgv);
- cc.cur = 0;
- cc.oldcc = 0;
- PL_regcc = &cc;
+ PL_regcc = 0;
cache_re(prog);
#ifdef DEBUGGING
/* Be paranoid... */
if (prog == NULL || startpos == NULL) {
- croak("NULL regexp parameter");
+ Perl_croak(aTHX_ "NULL regexp parameter");
return 0;
}
/* Check validity of program. */
if (UCHARAT(prog->program) != REG_MAGIC) {
- croak("corrupted regexp program");
+ Perl_croak(aTHX_ "corrupted regexp program");
}
PL_reg_flags = 0;
PL_reg_eval_set = 0;
+ PL_reg_maxiter = 0;
if (prog->reganch & ROPT_UTF8)
PL_reg_flags |= RF_utf8;
/* If there is a "must appear" string, look for it. */
s = startpos;
- if (!(flags & REXEC_CHECKED)
- && prog->check_substr != Nullsv &&
- !(prog->reganch & ROPT_ANCH_GPOS) &&
- (!(prog->reganch & (ROPT_ANCH_BOL | ROPT_ANCH_MBOL))
- || (PL_multiline && prog->check_substr == prog->anchored_substr)) )
- {
- char *t;
- start_shift = prog->check_offset_min; /* okay to underestimate on CC */
- /* Should be nonnegative! */
- end_shift = minlen - start_shift - CHR_SVLEN(prog->check_substr);
- if (flags & REXEC_SCREAM) {
- if (PL_screamfirst[BmRARE(prog->check_substr)] >= 0)
- s = screaminstr(sv, prog->check_substr,
- start_shift + (stringarg - strbeg),
- end_shift, &scream_pos, 0);
- else
- s = Nullch;
- scream_olds = s;
+
+ if (prog->reganch & ROPT_GPOS_SEEN) { /* Need to have PL_reg_ganch */
+ MAGIC *mg;
+
+ if (flags & REXEC_IGNOREPOS) /* Means: check only at start */
+ PL_reg_ganch = startpos;
+ else if (sv && SvTYPE(sv) >= SVt_PVMG
+ && SvMAGIC(sv)
+ && (mg = mg_find(sv, 'g')) && mg->mg_len >= 0) {
+ PL_reg_ganch = strbeg + mg->mg_len; /* Defined pos() */
+ if (prog->reganch & ROPT_ANCH_GPOS) {
+ if (s > PL_reg_ganch)
+ goto phooey;
+ s = PL_reg_ganch;
+ }
}
- else
- s = fbm_instr((unsigned char*)s + start_shift,
- (unsigned char*)strend - end_shift,
- prog->check_substr, 0);
- if (!s) {
- ++BmUSEFUL(prog->check_substr); /* hooray */
+ else /* pos() not defined */
+ PL_reg_ganch = strbeg;
+ }
+
+ if (!(flags & REXEC_CHECKED) && prog->check_substr != Nullsv) {
+ re_scream_pos_data d;
+
+ d.scream_olds = &scream_olds;
+ d.scream_pos = &scream_pos;
+ s = re_intuit_start(prog, sv, s, strend, flags, &d);
+ if (!s)
goto phooey; /* not present */
- }
- else if (s - stringarg > prog->check_offset_max &&
- (UTF
- ? ((t = reghopmaybe_c(s, -(prog->check_offset_max))) && t >= stringarg)
- : (t = s - prog->check_offset_max) != 0
- )
- )
- {
- ++BmUSEFUL(prog->check_substr); /* hooray/2 */
- s = t;
- }
- else if (!(prog->reganch & ROPT_NAUGHTY)
- && --BmUSEFUL(prog->check_substr) < 0
- && prog->check_substr == prog->float_substr) { /* boo */
- SvREFCNT_dec(prog->check_substr);
- prog->check_substr = Nullsv; /* disable */
- prog->float_substr = Nullsv; /* clear */
- s = startpos;
- }
- else
- s = startpos;
}
- DEBUG_r(if (!PL_colorset) reginitcolors());
- DEBUG_r(PerlIO_printf(Perl_debug_log,
- "%sMatching%s `%s%.60s%s%s' against `%s%.*s%s%s'\n",
+ DEBUG_r( if (!PL_colorset) reginitcolors() );
+ DEBUG_r(PerlIO_printf(Perl_debug_log,
+ "%sMatching REx%s `%s%.60s%s%s' against `%s%.*s%s%s'\n",
PL_colors[4],PL_colors[5],PL_colors[0],
prog->precomp,
PL_colors[1],
(strlen(prog->precomp) > 60 ? "..." : ""),
- PL_colors[0],
- (strend - startpos > 60 ? 60 : strend - startpos),
+ PL_colors[0],
+ (int)(strend - startpos > 60 ? 60 : strend - startpos),
startpos, PL_colors[1],
(strend - startpos > 60 ? "..." : ""))
);
- if (prog->reganch & ROPT_GPOS_SEEN) {
- MAGIC *mg;
-
- if (!(flags & REXEC_IGNOREPOS) && sv && SvTYPE(sv) >= SVt_PVMG
- && SvMAGIC(sv) && (mg = mg_find(sv, 'g')) && mg->mg_len >= 0)
- PL_reg_ganch = strbeg + mg->mg_len;
- else
- PL_reg_ganch = startpos;
- }
-
/* Simplest case: anchored match need be tried only once. */
/* [unless only anchor is BOL and multiline is set] */
if (prog->reganch & (ROPT_ANCH & ~ROPT_ANCH_GPOS)) {
- if (regtry(prog, startpos))
+ if (s == startpos && regtry(prog, startpos))
goto got_it;
else if (PL_multiline || (prog->reganch & ROPT_IMPLICIT)
|| (prog->reganch & ROPT_ANCH_MBOL)) /* XXXX SBOL? */
{
+ char *end;
+
if (minlen)
dontbother = minlen - 1;
- strend = HOPc(strend, -dontbother);
+ end = HOPc(strend, -dontbother) - 1;
/* for multiline we only have to try after newlines */
- if (s > startpos)
- s--;
- while (s < strend) {
- if (*s++ == '\n') { /* don't need PL_utf8skip here */
- if (s < strend && regtry(prog, s))
+ if (prog->check_substr) {
+ if (s == startpos)
+ goto after_try;
+ while (1) {
+ if (regtry(prog, s))
goto got_it;
- }
+ after_try:
+ if (s >= end)
+ goto phooey;
+ if (prog->reganch & RE_USE_INTUIT) {
+ s = re_intuit_start(prog, sv, s + 1, strend, flags, NULL);
+ if (!s)
+ goto phooey;
+ }
+ else
+ s++;
+ }
+ } else {
+ if (s > startpos)
+ s--;
+ while (s < end) {
+ if (*s++ == '\n') { /* don't need PL_utf8skip here */
+ if (regtry(prog, s))
+ goto got_it;
+ }
+ }
}
}
goto phooey;
/* Messy cases: unanchored match. */
if (prog->anchored_substr && prog->reganch & ROPT_SKIP) {
/* we have /x+whatever/ */
- /* it must be a one character string */
+ /* it must be a one character string (XXXX Except UTF?) */
char ch = SvPVX(prog->anchored_substr)[0];
if (UTF) {
while (s < strend) {
I32 back_min =
prog->anchored_substr ? prog->anchored_offset : prog->float_min_offset;
I32 delta = back_max - back_min;
- char *last = HOPc(strend, 0-(CHR_SVLEN(must) + back_min)); /* Cannot start after this */
+ char *last = HOPc(strend, /* Cannot start after this */
+ -(I32)(CHR_SVLEN(must)
+ - (SvTAIL(must) != 0) + back_min));
char *last1; /* Last position checked before */
if (s > PL_bostr)
? (s = screaminstr(sv, must, HOPc(s, back_min) - strbeg,
end_shift, &scream_pos, 0))
: (s = fbm_instr((unsigned char*)HOP(s, back_min),
- (unsigned char*)strend, must, 0))) ) {
+ (unsigned char*)strend, must,
+ PL_multiline ? FBMrf_MULTILINE : 0))) ) {
if (HOPc(s, -back_max) > last1) {
last1 = HOPc(s, -back_min);
s = HOPc(s, -back_max);
}
else if (c = prog->regstclass) {
I32 doevery = (prog->reganch & ROPT_SKIP) == 0;
- char *cc;
+ char *m;
+ int ln;
+ int c1;
+ int c2;
+ char *e;
if (minlen)
dontbother = minlen - 1;
/* We know what class it must start with. */
switch (OP(c)) {
case ANYOFUTF8:
- cc = (char *) OPERAND(c);
while (s < strend) {
if (REGINCLASSUTF8(c, (U8*)s)) {
if (tmp && regtry(prog, s))
}
break;
case ANYOF:
- cc = (char *) OPERAND(c);
while (s < strend) {
- if (REGINCLASS(cc, *s)) {
+ if (REGINCLASS(c, *s)) {
if (tmp && regtry(prog, s))
goto got_it;
else
s++;
}
break;
+ case EXACTF:
+ m = STRING(c);
+ ln = STR_LEN(c);
+ c1 = *m;
+ c2 = PL_fold[c1];
+ goto do_exactf;
+ case EXACTFL:
+ m = STRING(c);
+ ln = STR_LEN(c);
+ c1 = *m;
+ c2 = PL_fold_locale[c1];
+ do_exactf:
+ e = strend - ln;
+
+ /* Here it is NOT UTF! */
+ if (c1 == c2) {
+ while (s <= e) {
+ if ( *s == c1
+ && (ln == 1 || (OP(c) == EXACTF
+ ? ibcmp(s, m, ln)
+ : ibcmp_locale(s, m, ln)))
+ && regtry(prog, s) )
+ goto got_it;
+ s++;
+ }
+ } else {
+ while (s <= e) {
+ if ( (*s == c1 || *s == c2)
+ && (ln == 1 || (OP(c) == EXACTF
+ ? ibcmp(s, m, ln)
+ : ibcmp_locale(s, m, ln)))
+ && regtry(prog, s) )
+ goto got_it;
+ s++;
+ }
+ }
+ break;
case BOUNDL:
PL_reg_flags |= RF_tainted;
/* FALL THROUGH */
s += UTF8SKIP(s);
}
break;
+ case DIGITL:
+ PL_reg_flags |= RF_tainted;
+ while (s < strend) {
+ if (isDIGIT_LC(*s)) {
+ if (tmp && regtry(prog, s))
+ goto got_it;
+ else
+ tmp = doevery;
+ }
+ else
+ tmp = 1;
+ s++;
+ }
+ break;
+ case DIGITLUTF8:
+ PL_reg_flags |= RF_tainted;
+ while (s < strend) {
+ if (isDIGIT_LC_utf8((U8*)s)) {
+ if (tmp && regtry(prog, s))
+ goto got_it;
+ else
+ tmp = doevery;
+ }
+ else
+ tmp = 1;
+ s += UTF8SKIP(s);
+ }
+ break;
case NDIGIT:
while (s < strend) {
if (!isDIGIT(*s)) {
s += UTF8SKIP(s);
}
break;
+ case NDIGITL:
+ PL_reg_flags |= RF_tainted;
+ while (s < strend) {
+ if (!isDIGIT_LC(*s)) {
+ if (tmp && regtry(prog, s))
+ goto got_it;
+ else
+ tmp = doevery;
+ }
+ else
+ tmp = 1;
+ s++;
+ }
+ break;
+ case NDIGITLUTF8:
+ PL_reg_flags |= RF_tainted;
+ while (s < strend) {
+ if (!isDIGIT_LC_utf8((U8*)s)) {
+ if (tmp && regtry(prog, s))
+ goto got_it;
+ else
+ tmp = doevery;
+ }
+ else
+ tmp = 1;
+ s += UTF8SKIP(s);
+ }
+ break;
+ default:
+ Perl_croak(aTHX_ "panic: unknown regstclass %d", (int)OP(c));
+ break;
}
}
else {
if (flags & REXEC_SCREAM) {
last = screaminstr(sv, prog->float_substr, s - strbeg,
end_shift, &scream_pos, 1); /* last one */
- if (!last) {
+ if (!last)
last = scream_olds; /* Only one occurence. */
- }
}
else {
STRLEN len;
char *little = SvPV(prog->float_substr, len);
- if (len)
- last = rninstr(s, strend, little, little + len);
- else
- last = strend; /* matching `$' */
+
+ if (SvTAIL(prog->float_substr)) {
+ if (memEQ(strend - len + 1, little, len - 1))
+ last = strend - len + 1;
+ else if (!PL_multiline)
+ last = memEQ(strend - len, little, len)
+ ? strend - len : Nullch;
+ else
+ goto find_last;
+ } else {
+ find_last:
+ if (len)
+ last = rninstr(s, strend, little, little + len);
+ else
+ last = strend; /* matching `$' */
+ }
}
if (last == NULL) goto phooey; /* Should not happen! */
dontbother = strend - last + prog->float_min_offset;
goto phooey;
got_it:
- prog->subbeg = strbeg;
- prog->subend = PL_regeol; /* strend may have been modified */
RX_MATCH_TAINTED_set(prog, PL_reg_flags & RF_tainted);
- /* make sure $`, $&, $', and $digit will work later */
- if (strbeg != prog->subbase) { /* second+ //g match. */
- if (!(flags & REXEC_COPY_STR)) {
- if (prog->subbase) {
- Safefree(prog->subbase);
- prog->subbase = Nullch;
- }
- }
- else {
- I32 i = PL_regeol - startpos + (stringarg - strbeg);
- s = savepvn(strbeg, i);
- Safefree(prog->subbase);
- prog->subbase = s;
- prog->subbeg = prog->subbase;
- prog->subend = prog->subbase + i;
- s = prog->subbase + (stringarg - strbeg);
- for (i = 0; i <= prog->nparens; i++) {
- if (prog->endp[i]) {
- prog->startp[i] = s + (prog->startp[i] - startpos);
- prog->endp[i] = s + (prog->endp[i] - startpos);
- }
- }
- }
- }
if (PL_reg_eval_set) {
/* Preserve the current value of $^R */
if (oreplsv != GvSV(PL_replgv))
sv_setsv(oreplsv, GvSV(PL_replgv));/* So that when GvSV(replgv) is
restored, the value remains
the same. */
- restore_pos(0);
+ restore_pos(aTHXo_ 0);
+ }
+
+ /* make sure $`, $&, $', and $digit will work later */
+ if ( !(flags & REXEC_NOT_FIRST) ) {
+ if (RX_MATCH_COPIED(prog)) {
+ Safefree(prog->subbeg);
+ RX_MATCH_COPIED_off(prog);
+ }
+ if (flags & REXEC_COPY_STR) {
+ I32 i = PL_regeol - startpos + (stringarg - strbeg);
+
+ s = savepvn(strbeg, i);
+ prog->subbeg = s;
+ prog->sublen = i;
+ RX_MATCH_COPIED_on(prog);
+ }
+ else {
+ prog->subbeg = strbeg;
+ prog->sublen = PL_regeol - strbeg; /* strend may have been modified */
+ }
}
return 1;
phooey:
if (PL_reg_eval_set)
- restore_pos(0);
+ restore_pos(aTHXo_ 0);
return 0;
}
- regtry - try match at specific point
*/
STATIC I32 /* 0 failure, 1 success */
-regtry(regexp *prog, char *startpos)
+S_regtry(pTHX_ regexp *prog, char *startpos)
{
dTHR;
register I32 i;
- register char **sp;
- register char **ep;
+ register I32 *sp;
+ register I32 *ep;
CHECKPOINT lastcp;
if ((prog->reganch & ROPT_EVAL_SEEN) && !PL_reg_eval_set) {
PL_reg_eval_set = RS_init;
DEBUG_r(DEBUG_s(
- PerlIO_printf(Perl_debug_log, " setting stack tmpbase at %i\n",
- PL_stack_sp - PL_stack_base);
+ PerlIO_printf(Perl_debug_log, " setting stack tmpbase at %"IVdf"\n",
+ (IV)(PL_stack_sp - PL_stack_base));
));
SAVEINT(cxstack[cxstack_ix].blk_oldsp);
cxstack[cxstack_ix].blk_oldsp = PL_stack_sp - PL_stack_base;
}
PL_reg_magic = mg;
PL_reg_oldpos = mg->mg_len;
- SAVEDESTRUCTOR(restore_pos, 0);
+ SAVEDESTRUCTOR_X(restore_pos, 0);
}
if (!PL_reg_curpm)
New(22,PL_reg_curpm, 1, PMOP);
PL_reg_curpm->op_pmregexp = prog;
PL_reg_oldcurpm = PL_curpm;
PL_curpm = PL_reg_curpm;
+ if (RX_MATCH_COPIED(prog)) {
+ /* Here is a serious problem: we cannot rewrite subbeg,
+ since it may be needed if this match fails. Thus
+ $` inside (?{}) could fail... */
+ PL_reg_oldsaved = prog->subbeg;
+ PL_reg_oldsavedlen = prog->sublen;
+ RX_MATCH_COPIED_off(prog);
+ }
+ else
+ PL_reg_oldsaved = Nullch;
prog->subbeg = PL_bostr;
- prog->subend = PL_regeol; /* strend may have been modified */
+ prog->sublen = PL_regeol - PL_bostr; /* strend may have been modified */
}
- prog->startp[0] = startpos;
+ prog->startp[0] = startpos - PL_bostr;
PL_reginput = startpos;
PL_regstartp = prog->startp;
PL_regendp = prog->endp;
ep = prog->endp;
if (prog->nparens) {
for (i = prog->nparens; i >= 1; i--) {
- *++sp = NULL;
- *++ep = NULL;
+ *++sp = -1;
+ *++ep = -1;
}
}
REGCP_SET;
if (regmatch(prog->program + 1)) {
- prog->endp[0] = PL_reginput;
+ prog->endp[0] = PL_reginput - PL_bostr;
return 1;
}
REGCP_UNWIND;
* advantage of machines that use a register save mask on subroutine entry.
*/
STATIC I32 /* 0 failure, 1 success */
-regmatch(regnode *prog)
+S_regmatch(pTHX_ regnode *prog)
{
dTHR;
register regnode *scan; /* Current node. */
#ifdef DEBUGGING
# define sayYES goto yes
# define sayNO goto no
+# define sayYES_FINAL goto yes_final
+# define sayYES_LOUD goto yes_loud
+# define sayNO_FINAL goto no_final
+# define sayNO_SILENT goto do_no
# define saySAME(x) if (x) goto yes; else goto no
# define REPORT_CODE_OFF 24
#else
# define sayYES return 1
# define sayNO return 0
+# define sayYES_FINAL return 1
+# define sayYES_LOUD return 1
+# define sayNO_FINAL return 0
+# define sayNO_SILENT return 0
# define saySAME(x) return x
#endif
DEBUG_r( {
pref0_len = pref_len;
regprop(prop, scan);
PerlIO_printf(Perl_debug_log,
- "%4i <%s%.*s%s%s%.*s%s%s%s%.*s%s>%*s|%3d:%*s%s\n",
- locinput - PL_bostr,
+ "%4"IVdf" <%s%.*s%s%s%.*s%s%s%s%.*s%s>%*s|%3"IVdf":%*s%s\n",
+ (IV)(locinput - PL_bostr),
PL_colors[4], pref0_len,
locinput - pref_len, PL_colors[5],
PL_colors[2], pref_len - pref0_len,
PL_colors[0], l, locinput, PL_colors[1],
15 - l - pref_len + 1,
"",
- scan - PL_regprogram, PL_regindent*2, "",
+ (IV)(scan - PL_regprogram), PL_regindent*2, "",
SvPVX(prop));
} );
nextchr = UCHARAT(++locinput);
break;
case EXACT:
- s = (char *) OPERAND(scan);
- ln = UCHARAT(s++);
+ s = STRING(scan);
+ ln = STR_LEN(scan);
/* Inline the first character, for speed. */
if (UCHARAT(s) != nextchr)
sayNO;
PL_reg_flags |= RF_tainted;
/* FALL THROUGH */
case EXACTF:
- s = (char *) OPERAND(scan);
- ln = UCHARAT(s++);
+ s = STRING(scan);
+ ln = STR_LEN(scan);
if (UTF) {
char *l = locinput;
nextchr = UCHARAT(locinput);
break;
case ANYOFUTF8:
- s = (char *) OPERAND(scan);
if (!REGINCLASSUTF8(scan, (U8*)locinput))
sayNO;
if (locinput >= PL_regeol)
nextchr = UCHARAT(locinput);
break;
case ANYOF:
- s = (char *) OPERAND(scan);
if (nextchr < 0)
nextchr = UCHARAT(locinput);
- if (!REGINCLASS(s, nextchr))
+ if (!REGINCLASS(scan, nextchr))
sayNO;
if (!nextchr && locinput >= PL_regeol)
sayNO;
sayNO;
nextchr = UCHARAT(++locinput);
break;
+ case DIGITL:
+ PL_reg_flags |= RF_tainted;
+ /* FALL THROUGH */
case DIGIT:
- if (!isDIGIT(nextchr))
+ if (!nextchr && locinput >= PL_regeol)
+ sayNO;
+ if (!(OP(scan) == DIGIT
+ ? isDIGIT(nextchr) : isDIGIT_LC(nextchr)))
sayNO;
nextchr = UCHARAT(++locinput);
break;
+ case DIGITLUTF8:
+ PL_reg_flags |= RF_tainted;
+ /* FALL THROUGH */
case DIGITUTF8:
+ if (!nextchr)
+ sayNO;
if (nextchr & 0x80) {
- if (!(swash_fetch(PL_utf8_digit,(U8*)locinput)))
+ if (OP(scan) == NDIGITUTF8
+ ? swash_fetch(PL_utf8_digit,(U8*)locinput)
+ : isDIGIT_LC_utf8((U8*)locinput))
+ {
sayNO;
+ }
locinput += PL_utf8skip[nextchr];
nextchr = UCHARAT(locinput);
break;
sayNO;
nextchr = UCHARAT(++locinput);
break;
+ case NDIGITL:
+ PL_reg_flags |= RF_tainted;
+ /* FALL THROUGH */
case NDIGIT:
- if (!nextchr && locinput >= PL_regeol)
+ if (!nextchr)
sayNO;
- if (isDIGIT(nextchr))
+ if (OP(scan) == DIGIT
+ ? isDIGIT(nextchr) : isDIGIT_LC(nextchr))
sayNO;
nextchr = UCHARAT(++locinput);
break;
+ case NDIGITLUTF8:
+ PL_reg_flags |= RF_tainted;
+ /* FALL THROUGH */
case NDIGITUTF8:
if (!nextchr && locinput >= PL_regeol)
sayNO;
case REF:
case REFF:
n = ARG(scan); /* which paren pair */
- s = PL_regstartp[n];
- if (*PL_reglastparen < n || !s)
+ ln = PL_regstartp[n];
+ PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
+ if (*PL_reglastparen < n || ln == -1)
sayNO; /* Do not match unless seen CLOSEn. */
- if (s == PL_regendp[n])
+ if (ln == PL_regendp[n])
break;
+ s = PL_bostr + ln;
if (UTF && OP(scan) != REF) { /* REF can do byte comparison */
char *l = locinput;
- char *e = PL_regendp[n];
+ char *e = PL_bostr + PL_regendp[n];
/*
* Note that we can't do the "other character" lookup trick as
* in the 8-bit case (no pun intended) because in Unicode we
(UCHARAT(s) != ((OP(scan) == REFF
? PL_fold : PL_fold_locale)[nextchr]))))
sayNO;
- ln = PL_regendp[n] - s;
+ ln = PL_regendp[n] - ln;
if (locinput + ln > PL_regeol)
sayNO;
if (ln > 1 && (OP(scan) == REF
n = ARG(scan);
PL_op = (OP_4tree*)PL_regdata->data[n];
- DEBUG_r( PerlIO_printf(Perl_debug_log, " re_eval 0x%x\n", PL_op) );
- PL_curpad = AvARRAY((AV*)PL_regdata->data[n + 1]);
- PL_reg_magic->mg_len = locinput - PL_bostr;
- PL_regendp[0] = locinput;
+ DEBUG_r( PerlIO_printf(Perl_debug_log, " re_eval 0x%"UVxf"\n", PTR2UV(PL_op)) );
+ PL_curpad = AvARRAY((AV*)PL_regdata->data[n + 2]);
+ PL_regendp[0] = PL_reg_magic->mg_len = locinput - PL_bostr;
- CALLRUNOPS(); /* Scalar context. */
+ CALLRUNOPS(aTHX); /* Scalar context. */
SPAGAIN;
ret = POPs;
PUTBACK;
regexp *re;
MAGIC *mg = Null(MAGIC*);
re_cc_state state;
- CURCUR cctmp;
CHECKPOINT cp, lastcp;
if(SvROK(ret) || SvRMAGICAL(ret)) {
I32 onpar = PL_regnpar;
pm.op_pmflags = 0;
- re = CALLREGCOMP(t, t + len, &pm);
+ re = CALLREGCOMP(aTHX_ t, t + len, &pm);
if (!(SvFLAGS(ret)
& (SVs_TEMP | SVs_PADTMP | SVf_READONLY)))
sv_magic(ret,(SV*)ReREFCNT_inc(re),'r',0,0);
state.cc = PL_regcc;
state.re = PL_reg_re;
- cctmp.cur = 0;
- cctmp.oldcc = 0;
- PL_regcc = &cctmp;
+ PL_regcc = 0;
cp = regcppush(0); /* Save *all* the positions. */
REGCP_SET;
*PL_reglastparen = 0;
PL_reg_call_cc = &state;
PL_reginput = locinput;
+
+ /* XXXX This is too dramatic a measure... */
+ PL_reg_maxiter = 0;
+
if (regmatch(re->program + 1)) {
+ /* Even though we succeeded, we need to restore
+ global variables, since we may be wrapped inside
+ SUSPEND, thus the match may be not finished yet. */
+
+ /* XXXX Do this only if SUSPENDed? */
+ PL_reg_call_cc = state.prev;
+ PL_regcc = state.cc;
+ PL_reg_re = state.re;
+ cache_re(PL_reg_re);
+
+ /* XXXX This is too dramatic a measure... */
+ PL_reg_maxiter = 0;
+
+ /* These are needed even if not SUSPEND. */
ReREFCNT_dec(re);
regcpblow(cp);
sayYES;
}
- DEBUG_r(
- PerlIO_printf(Perl_debug_log,
- "%*s failed...\n",
- REPORT_CODE_OFF+PL_regindent*2, "")
- );
ReREFCNT_dec(re);
REGCP_UNWIND;
regcppop();
PL_regcc = state.cc;
PL_reg_re = state.re;
cache_re(PL_reg_re);
+
+ /* XXXX This is too dramatic a measure... */
+ PL_reg_maxiter = 0;
+
sayNO;
}
sw = SvTRUE(ret);
break;
case CLOSE:
n = ARG(scan); /* which paren pair */
- PL_regstartp[n] = PL_reg_start_tmp[n];
- PL_regendp[n] = locinput;
+ PL_regstartp[n] = PL_reg_start_tmp[n] - PL_bostr;
+ PL_regendp[n] = locinput - PL_bostr;
if (n > *PL_reglastparen)
*PL_reglastparen = n;
break;
case GROUPP:
n = ARG(scan); /* which paren pair */
- sw = (*PL_reglastparen >= n && PL_regendp[n] != NULL);
+ sw = (*PL_reglastparen >= n && PL_regendp[n] != -1);
break;
case IFTHEN:
+ PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
if (sw)
next = NEXTOPER(NEXTOPER(scan));
else {
case LOGICAL:
logical = scan->flags;
break;
+/*******************************************************************
+ PL_regcc contains infoblock about the innermost (...)* loop, and
+ a pointer to the next outer infoblock.
+
+ Here is how Y(A)*Z is processed (if it is compiled into CURLYX/WHILEM):
+
+ 1) After matching X, regnode for CURLYX is processed;
+
+ 2) This regnode creates infoblock on the stack, and calls
+ regmatch() recursively with the starting point at WHILEM node;
+
+ 3) Each hit of WHILEM node tries to match A and Z (in the order
+ depending on the current iteration, min/max of {min,max} and
+ greediness). The information about where are nodes for "A"
+ and "Z" is read from the infoblock, as is info on how many times "A"
+ was already matched, and greediness.
+
+ 4) After A matches, the same WHILEM node is hit again.
+
+ 5) Each time WHILEM is hit, PL_regcc is the infoblock created by CURLYX
+ of the same pair. Thus when WHILEM tries to match Z, it temporarily
+ resets PL_regcc, since this Y(A)*Z can be a part of some other loop:
+ as in (Y(A)*Z)*. If Z matches, the automaton will hit the WHILEM node
+ of the external loop.
+
+ Currently present infoblocks form a tree with a stem formed by PL_curcc
+ and whatever it mentions via ->next, and additional attached trees
+ corresponding to temporarily unset infoblocks as in "5" above.
+
+ In the following picture infoblocks for outer loop of
+ (Y(A)*?Z)*?T are denoted O, for inner I. NULL starting block
+ is denoted by x. The matched string is YAAZYAZT. Temporarily postponed
+ infoblocks are drawn below the "reset" infoblock.
+
+ In fact in the picture below we do not show failed matches for Z and T
+ by WHILEM blocks. [We illustrate minimal matches, since for them it is
+ more obvious *why* one needs to *temporary* unset infoblocks.]
+
+ Matched REx position InfoBlocks Comment
+ (Y(A)*?Z)*?T x
+ Y(A)*?Z)*?T x <- O
+ Y (A)*?Z)*?T x <- O
+ Y A)*?Z)*?T x <- O <- I
+ YA )*?Z)*?T x <- O <- I
+ YA A)*?Z)*?T x <- O <- I
+ YAA )*?Z)*?T x <- O <- I
+ YAA Z)*?T x <- O # Temporary unset I
+ I
+
+ YAAZ Y(A)*?Z)*?T x <- O
+ I
+
+ YAAZY (A)*?Z)*?T x <- O
+ I
+
+ YAAZY A)*?Z)*?T x <- O <- I
+ I
+
+ YAAZYA )*?Z)*?T x <- O <- I
+ I
+
+ YAAZYA Z)*?T x <- O # Temporary unset I
+ I,I
+
+ YAAZYAZ )*?T x <- O
+ I,I
+
+ YAAZYAZ T x # Temporary unset O
+ O
+ I,I
+
+ YAAZYAZT x
+ O
+ I,I
+ *******************************************************************/
case CURLYX: {
CURCUR cc;
CHECKPOINT cp = PL_savestack_ix;
/*
* This is really hard to understand, because after we match
* what we're trying to match, we must make sure the rest of
- * the RE is going to match for sure, and to do that we have
+ * the REx is going to match for sure, and to do that we have
* to go back UP the parse tree by recursing ever deeper. And
* if it fails, we have to reset our parent's current state
* that we can try again after backing off.
if (locinput == cc->lastloc && n >= cc->min) {
PL_regcc = cc->oldcc;
- ln = PL_regcc->cur;
+ if (PL_regcc)
+ ln = PL_regcc->cur;
DEBUG_r(
PerlIO_printf(Perl_debug_log,
"%*s empty match detected, try continuation...\n",
);
if (regmatch(cc->next))
sayYES;
- DEBUG_r(
- PerlIO_printf(Perl_debug_log,
- "%*s failed...\n",
- REPORT_CODE_OFF+PL_regindent*2, "")
- );
- PL_regcc->cur = ln;
+ if (PL_regcc)
+ PL_regcc->cur = ln;
PL_regcc = cc;
sayNO;
}
sayYES;
cc->cur = n - 1;
cc->lastloc = lastloc;
+ sayNO;
+ }
+
+ if (scan->flags) {
+ /* Check whether we already were at this position.
+ Postpone detection until we know the match is not
+ *that* much linear. */
+ if (!PL_reg_maxiter) {
+ PL_reg_maxiter = (PL_regeol - PL_bostr + 1) * (scan->flags>>4);
+ PL_reg_leftiter = PL_reg_maxiter;
+ }
+ if (PL_reg_leftiter-- == 0) {
+ I32 size = (PL_reg_maxiter + 7)/8;
+ if (PL_reg_poscache) {
+ if (PL_reg_poscache_size < size) {
+ Renew(PL_reg_poscache, size, char);
+ PL_reg_poscache_size = size;
+ }
+ Zero(PL_reg_poscache, size, char);
+ }
+ else {
+ PL_reg_poscache_size = size;
+ Newz(29, PL_reg_poscache, size, char);
+ }
+ DEBUG_r(
+ PerlIO_printf(Perl_debug_log,
+ "%sDetected a super-linear match, switching on caching%s...\n",
+ PL_colors[4], PL_colors[5])
+ );
+ }
+ if (PL_reg_leftiter < 0) {
+ I32 o = locinput - PL_bostr, b;
+
+ o = (scan->flags & 0xf) - 1 + o * (scan->flags>>4);
+ b = o % 8;
+ o /= 8;
+ if (PL_reg_poscache[o] & (1<<b)) {
DEBUG_r(
PerlIO_printf(Perl_debug_log,
- "%*s failed...\n",
+ "%*s already tried at this position...\n",
REPORT_CODE_OFF+PL_regindent*2, "")
);
- sayNO;
+ sayNO_SILENT;
+ }
+ PL_reg_poscache[o] |= (1<<b);
+ }
}
/* Prefer next over scan for minimal matching. */
if (cc->minmod) {
PL_regcc = cc->oldcc;
- ln = PL_regcc->cur;
+ if (PL_regcc)
+ ln = PL_regcc->cur;
cp = regcppush(cc->parenfloor);
REGCP_SET;
if (regmatch(cc->next)) {
}
REGCP_UNWIND;
regcppop();
- PL_regcc->cur = ln;
+ if (PL_regcc)
+ PL_regcc->cur = ln;
PL_regcc = cc;
if (n >= cc->max) { /* Maximum greed exceeded? */
if (ckWARN(WARN_UNSAFE) && n >= REG_INFTY
&& !(PL_reg_flags & RF_warned)) {
PL_reg_flags |= RF_warned;
- warner(WARN_UNSAFE, "%s limit (%d) exceeded",
+ Perl_warner(aTHX_ WARN_UNSAFE, "%s limit (%d) exceeded",
"Complex regular subexpression recursion",
REG_INFTY - 1);
}
regcpblow(cp);
sayYES;
}
- DEBUG_r(
- PerlIO_printf(Perl_debug_log,
- "%*s failed...\n",
- REPORT_CODE_OFF+PL_regindent*2, "")
- );
REGCP_UNWIND;
regcppop();
cc->cur = n - 1;
if (ckWARN(WARN_UNSAFE) && n >= REG_INFTY
&& !(PL_reg_flags & RF_warned)) {
PL_reg_flags |= RF_warned;
- warner(WARN_UNSAFE, "%s limit (%d) exceeded",
+ Perl_warner(aTHX_ WARN_UNSAFE, "%s limit (%d) exceeded",
"Complex regular subexpression recursion",
REG_INFTY - 1);
}
/* Failed deeper matches of scan, so see if this one works. */
PL_regcc = cc->oldcc;
- ln = PL_regcc->cur;
+ if (PL_regcc)
+ ln = PL_regcc->cur;
if (regmatch(cc->next))
sayYES;
- DEBUG_r(
- PerlIO_printf(Perl_debug_log, "%*s failed...\n",
- REPORT_CODE_OFF+PL_regindent*2, "")
- );
- PL_regcc->cur = ln;
+ if (PL_regcc)
+ PL_regcc->cur = ln;
PL_regcc = cc;
cc->cur = n - 1;
cc->lastloc = lastloc;
sayYES;
REGCP_UNWIND;
for (n = *PL_reglastparen; n > lastparen; n--)
- PL_regendp[n] = 0;
+ PL_regendp[n] = -1;
*PL_reglastparen = n;
scan = next;
/*SUPPRESS 560*/
ln = n;
locinput = PL_reginput;
if (PL_regkind[(U8)OP(next)] == EXACT) {
- c1 = UCHARAT(OPERAND(next) + 1);
+ c1 = (U8)*STRING(next);
if (OP(next) == EXACTF)
c2 = PL_fold[c1];
else if (OP(next) == EXACTFL)
{
if (paren) {
if (n) {
- PL_regstartp[paren] = HOPc(PL_reginput, -l);
- PL_regendp[paren] = PL_reginput;
+ PL_regstartp[paren] =
+ HOPc(PL_reginput, -l) - PL_bostr;
+ PL_regendp[paren] = PL_reginput - PL_bostr;
}
else
- PL_regendp[paren] = NULL;
+ PL_regendp[paren] = -1;
}
if (regmatch(next))
sayYES;
locinput = PL_reginput;
DEBUG_r(
PerlIO_printf(Perl_debug_log,
- "%*s matched %ld times, len=%ld...\n",
- REPORT_CODE_OFF+PL_regindent*2, "", n, l)
+ "%*s matched %"IVdf" times, len=%"IVdf"...\n",
+ (int)(REPORT_CODE_OFF+PL_regindent*2), "",
+ (IV) n, (IV)l)
);
if (n >= ln) {
if (PL_regkind[(U8)OP(next)] == EXACT) {
- c1 = UCHARAT(OPERAND(next) + 1);
+ c1 = (U8)*STRING(next);
if (OP(next) == EXACTF)
c2 = PL_fold[c1];
else if (OP(next) == EXACTFL)
{
DEBUG_r(
PerlIO_printf(Perl_debug_log,
- "%*s trying tail with n=%ld...\n",
- REPORT_CODE_OFF+PL_regindent*2, "", n)
+ "%*s trying tail with n=%"IVdf"...\n",
+ (int)(REPORT_CODE_OFF+PL_regindent*2), "", (IV)n)
);
if (paren) {
if (n) {
- PL_regstartp[paren] = HOPc(PL_reginput, -l);
- PL_regendp[paren] = PL_reginput;
+ PL_regstartp[paren] = HOPc(PL_reginput, -l) - PL_bostr;
+ PL_regendp[paren] = PL_reginput - PL_bostr;
}
else
- PL_regendp[paren] = NULL;
+ PL_regendp[paren] = -1;
}
if (regmatch(next))
sayYES;
* when we know what character comes next.
*/
if (PL_regkind[(U8)OP(next)] == EXACT) {
- c1 = UCHARAT(OPERAND(next) + 1);
+ c1 = (U8)*STRING(next);
if (OP(next) == EXACTF)
c2 = PL_fold[c1];
else if (OP(next) == EXACTFL)
/* PL_reginput == locinput now */
if (paren) {
if (ln) {
- PL_regstartp[paren] = HOPc(locinput, -1);
- PL_regendp[paren] = locinput;
+ PL_regstartp[paren] = HOPc(locinput, -1) - PL_bostr;
+ PL_regendp[paren] = locinput - PL_bostr;
}
else
- PL_regendp[paren] = NULL;
+ PL_regendp[paren] = -1;
}
if (regmatch(next))
sayYES;
{
if (paren) {
if (n) {
- PL_regstartp[paren] = HOPc(PL_reginput, -1);
- PL_regendp[paren] = PL_reginput;
+ PL_regstartp[paren] = HOPc(PL_reginput, -1) - PL_bostr;
+ PL_regendp[paren] = PL_reginput - PL_bostr;
}
else
- PL_regendp[paren] = NULL;
+ PL_regendp[paren] = -1;
}
if (regmatch(next))
sayYES;
{
if (paren && n) {
if (n) {
- PL_regstartp[paren] = HOPc(PL_reginput, -1);
- PL_regendp[paren] = PL_reginput;
+ PL_regstartp[paren] = HOPc(PL_reginput, -1) - PL_bostr;
+ PL_regendp[paren] = PL_reginput - PL_bostr;
}
else
- PL_regendp[paren] = NULL;
+ PL_regendp[paren] = -1;
}
if (regmatch(next))
sayYES;
"%*s continuation failed...\n",
REPORT_CODE_OFF+PL_regindent*2, "")
);
- sayNO;
+ sayNO_SILENT;
}
- if (locinput < PL_regtill)
- sayNO; /* Cannot match: too short. */
- /* Fall through */
+ if (locinput < PL_regtill) {
+ DEBUG_r(PerlIO_printf(Perl_debug_log,
+ "%sMatch possible, but length=%ld is smaller than requested=%ld, failing!%s\n",
+ PL_colors[4],
+ (long)(locinput - PL_reg_starttry),
+ (long)(PL_regtill - PL_reg_starttry),
+ PL_colors[5]));
+ sayNO_FINAL; /* Cannot match: too short. */
+ }
+ PL_reginput = locinput; /* put where regtry can find it */
+ sayYES_FINAL; /* Success! */
case SUCCEED:
PL_reginput = locinput; /* put where regtry can find it */
- sayYES; /* Success! */
+ sayYES_LOUD; /* Success! */
case SUSPEND:
n = 1;
PL_reginput = locinput;
next = NULL;
break;
default:
- PerlIO_printf(PerlIO_stderr(), "%lx %d\n",
- (unsigned long)scan, OP(scan));
- croak("regexp memory corruption");
+ PerlIO_printf(Perl_error_log, "%"UVxf" %d\n",
+ PTR2UV(scan), OP(scan));
+ Perl_croak(aTHX_ "regexp memory corruption");
}
scan = next;
}
* We get here only if there's trouble -- normally "case END" is
* the terminating point.
*/
- croak("corrupted regexp pointers");
+ Perl_croak(aTHX_ "corrupted regexp pointers");
/*NOTREACHED*/
sayNO;
+yes_loud:
+ DEBUG_r(
+ PerlIO_printf(Perl_debug_log,
+ "%*s %scould match...%s\n",
+ REPORT_CODE_OFF+PL_regindent*2, "", PL_colors[4],PL_colors[5])
+ );
+ goto yes;
+yes_final:
+ DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch successful!%s\n",
+ PL_colors[4],PL_colors[5]));
yes:
#ifdef DEBUGGING
PL_regindent--;
return 1;
no:
+ DEBUG_r(
+ PerlIO_printf(Perl_debug_log,
+ "%*s %sfailed...%s\n",
+ REPORT_CODE_OFF+PL_regindent*2, "",PL_colors[4],PL_colors[5])
+ );
+ goto do_no;
+no_final:
+do_no:
#ifdef DEBUGGING
PL_regindent--;
#endif
* rather than incrementing count on every character. [Er, except utf8.]]
*/
STATIC I32
-regrepeat(regnode *p, I32 max)
+S_regrepeat(pTHX_ regnode *p, I32 max)
{
dTHR;
register char *scan;
- register char *opnd;
register I32 c;
register char *loceol = PL_regeol;
register I32 hardcount = 0;
scan = PL_reginput;
if (max != REG_INFTY && max < loceol - scan)
loceol = scan + max;
- opnd = (char *) OPERAND(p);
switch (OP(p)) {
case REG_ANY:
while (scan < loceol && *scan != '\n')
}
break;
case EXACT: /* length of string is 1 */
- c = UCHARAT(++opnd);
+ c = (U8)*STRING(p);
while (scan < loceol && UCHARAT(scan) == c)
scan++;
break;
case EXACTF: /* length of string is 1 */
- c = UCHARAT(++opnd);
+ c = (U8)*STRING(p);
while (scan < loceol &&
(UCHARAT(scan) == c || UCHARAT(scan) == PL_fold[c]))
scan++;
break;
case EXACTFL: /* length of string is 1 */
PL_reg_flags |= RF_tainted;
- c = UCHARAT(++opnd);
+ c = (U8)*STRING(p);
while (scan < loceol &&
(UCHARAT(scan) == c || UCHARAT(scan) == PL_fold_locale[c]))
scan++;
}
break;
case ANYOF:
- while (scan < loceol && REGINCLASS(opnd, *scan))
+ while (scan < loceol && REGINCLASS(p, *scan))
scan++;
break;
case ALNUM:
regprop(prop, p);
PerlIO_printf(Perl_debug_log,
- "%*s %s can match %ld times out of %ld...\n",
- REPORT_CODE_OFF+1, "", SvPVX(prop),c,max);
+ "%*s %s can match %"IVdf" times out of %"IVdf"...\n",
+ REPORT_CODE_OFF+1, "", SvPVX(prop),(IV)c,(IV)max);
});
return(c);
*/
STATIC I32
-regrepeat_hard(regnode *p, I32 max, I32 *lp)
+S_regrepeat_hard(pTHX_ regnode *p, I32 max, I32 *lp)
{
dTHR;
register char *scan;
*/
STATIC bool
-reginclass(register char *p, register I32 c)
+S_reginclass(pTHX_ register regnode *p, register I32 c)
{
dTHR;
- char flags = *p;
+ char flags = ANYOF_FLAGS(p);
bool match = FALSE;
c &= 0xFF;
- if (ANYOF_TEST(p, c))
+ if (ANYOF_BITMAP_TEST(p, c))
match = TRUE;
else if (flags & ANYOF_FOLD) {
I32 cf;
}
else
cf = PL_fold[c];
- if (ANYOF_TEST(p, cf))
+ if (ANYOF_BITMAP_TEST(p, cf))
match = TRUE;
}
- if (!match && (flags & ANYOF_ISA)) {
+ if (!match && (flags & ANYOF_CLASS)) {
PL_reg_flags |= RF_tainted;
-
- if (((flags & ANYOF_ALNUML) && isALNUM_LC(c)) ||
- ((flags & ANYOF_NALNUML) && !isALNUM_LC(c)) ||
- ((flags & ANYOF_SPACEL) && isSPACE_LC(c)) ||
- ((flags & ANYOF_NSPACEL) && !isSPACE_LC(c)))
+ if (
+ (ANYOF_CLASS_TEST(p, ANYOF_ALNUM) && isALNUM_LC(c)) ||
+ (ANYOF_CLASS_TEST(p, ANYOF_NALNUM) && !isALNUM_LC(c)) ||
+ (ANYOF_CLASS_TEST(p, ANYOF_SPACE) && isSPACE_LC(c)) ||
+ (ANYOF_CLASS_TEST(p, ANYOF_NSPACE) && !isSPACE_LC(c)) ||
+ (ANYOF_CLASS_TEST(p, ANYOF_DIGIT) && isDIGIT_LC(c)) ||
+ (ANYOF_CLASS_TEST(p, ANYOF_NDIGIT) && !isDIGIT_LC(c)) ||
+ (ANYOF_CLASS_TEST(p, ANYOF_ALNUMC) && isALNUMC_LC(c)) ||
+ (ANYOF_CLASS_TEST(p, ANYOF_NALNUMC) && !isALNUMC_LC(c)) ||
+ (ANYOF_CLASS_TEST(p, ANYOF_ALPHA) && isALPHA_LC(c)) ||
+ (ANYOF_CLASS_TEST(p, ANYOF_NALPHA) && !isALPHA_LC(c)) ||
+ (ANYOF_CLASS_TEST(p, ANYOF_ASCII) && isASCII(c)) ||
+ (ANYOF_CLASS_TEST(p, ANYOF_NASCII) && !isASCII(c)) ||
+ (ANYOF_CLASS_TEST(p, ANYOF_CNTRL) && isCNTRL_LC(c)) ||
+ (ANYOF_CLASS_TEST(p, ANYOF_NCNTRL) && !isCNTRL_LC(c)) ||
+ (ANYOF_CLASS_TEST(p, ANYOF_GRAPH) && isGRAPH_LC(c)) ||
+ (ANYOF_CLASS_TEST(p, ANYOF_NGRAPH) && !isGRAPH_LC(c)) ||
+ (ANYOF_CLASS_TEST(p, ANYOF_LOWER) && isLOWER_LC(c)) ||
+ (ANYOF_CLASS_TEST(p, ANYOF_NLOWER) && !isLOWER_LC(c)) ||
+ (ANYOF_CLASS_TEST(p, ANYOF_PRINT) && isPRINT_LC(c)) ||
+ (ANYOF_CLASS_TEST(p, ANYOF_NPRINT) && !isPRINT_LC(c)) ||
+ (ANYOF_CLASS_TEST(p, ANYOF_PUNCT) && isPUNCT_LC(c)) ||
+ (ANYOF_CLASS_TEST(p, ANYOF_NPUNCT) && !isPUNCT_LC(c)) ||
+ (ANYOF_CLASS_TEST(p, ANYOF_UPPER) && isUPPER_LC(c)) ||
+ (ANYOF_CLASS_TEST(p, ANYOF_NUPPER) && !isUPPER_LC(c)) ||
+ (ANYOF_CLASS_TEST(p, ANYOF_XDIGIT) && isXDIGIT(c)) ||
+ (ANYOF_CLASS_TEST(p, ANYOF_NXDIGIT) && !isXDIGIT(c))
+ ) /* How's that for a conditional? */
{
match = TRUE;
}
}
STATIC bool
-reginclassutf8(regnode *f, U8 *p)
+S_reginclassutf8(pTHX_ regnode *f, U8 *p)
{
dTHR;
char flags = ARG1(f);
match = TRUE;
}
- if (!match && (flags & ANYOF_ISA)) {
- PL_reg_flags |= RF_tainted;
-
- if (((flags & ANYOF_ALNUML) && isALNUM_LC_utf8(p)) ||
- ((flags & ANYOF_NALNUML) && !isALNUM_LC_utf8(p)) ||
- ((flags & ANYOF_SPACEL) && isSPACE_LC_utf8(p)) ||
- ((flags & ANYOF_NSPACEL) && !isSPACE_LC_utf8(p)))
- {
- match = TRUE;
- }
- }
+ /* UTF8 combined with ANYOF_CLASS is ill-defined. */
return (flags & ANYOF_INVERT) ? !match : match;
}
STATIC U8 *
-reghop(U8 *s, I32 off)
+S_reghop(pTHX_ U8 *s, I32 off)
{
dTHR;
if (off >= 0) {
}
STATIC U8 *
-reghopmaybe(U8* s, I32 off)
+S_reghopmaybe(pTHX_ U8* s, I32 off)
{
dTHR;
if (off >= 0) {
}
return s;
}
+
+#ifdef PERL_OBJECT
+#include "XSUB.h"
+#endif
+
+static void
+restore_pos(pTHXo_ void *arg)
+{
+ dTHR;
+ if (PL_reg_eval_set) {
+ if (PL_reg_oldsaved) {
+ PL_reg_re->subbeg = PL_reg_oldsaved;
+ PL_reg_re->sublen = PL_reg_oldsavedlen;
+ RX_MATCH_COPIED_on(PL_reg_re);
+ }
+ PL_reg_magic->mg_len = PL_reg_oldpos;
+ PL_reg_eval_set = 0;
+ PL_curpm = PL_reg_oldcurpm;
+ }
+}
+