regcomp.c

   1 /*    regcomp.c
   2  */
   3
   4 /*
   5  * "A fair jaw-cracker dwarf-language must be."  --Samwise Gamgee
   6  */
   7
   8 /* NOTE: this is derived from Henry Spencer's regexp code, and should not
   9  * confused with the original package (see point 3 below).  Thanks, Henry!
  10  */
  11
  12 /* Additional note: this code is very heavily munged from Henry's version
  13  * in places.  In some spots I've traded clarity for efficiency, so don't
  14  * blame Henry for some of the lack of readability.
  15  */
  16
  17 /* The names of the functions have been changed from regcomp and
  18  * regexec to  pregcomp and pregexec in order to avoid conflicts
  19  * with the POSIX routines of the same names.
  20 */
  21
  22 #ifdef PERL_EXT_RE_BUILD
  23 /* need to replace pregcomp et al, so enable that */
  24 #  ifndef PERL_IN_XSUB_RE
  25 #    define PERL_IN_XSUB_RE
  26 #  endif
  27 /* need access to debugger hooks */
  28 #  if defined(PERL_EXT_RE_DEBUG) && !defined(DEBUGGING)
  29 #    define DEBUGGING
  30 #  endif
  31 #endif
  32
  33 #ifdef PERL_IN_XSUB_RE
  34 /* We *really* need to overwrite these symbols: */
  35 #  define Perl_pregcomp my_regcomp
  36 #  define Perl_regdump my_regdump
  37 #  define Perl_regprop my_regprop
  38 #  define Perl_pregfree my_regfree
  39 #  define Perl_re_intuit_string my_re_intuit_string
  40 /* *These* symbols are masked to allow static link. */
  41 #  define Perl_regnext my_regnext
  42 #  define Perl_save_re_context my_save_re_context
  43 #  define Perl_reginitcolors my_reginitcolors
  44
  45 #  define PERL_NO_GET_CONTEXT
  46 #endif
  47
  48 /*SUPPRESS 112*/
  49 /*
  50  * pregcomp and pregexec -- regsub and regerror are not used in perl
  51  *
  52  *      Copyright (c) 1986 by University of Toronto.
  53  *      Written by Henry Spencer.  Not derived from licensed software.
  54  *
  55  *      Permission is granted to anyone to use this software for any
  56  *      purpose on any computer system, and to redistribute it freely,
  57  *      subject to the following restrictions:
  58  *
  59  *      1. The author is not responsible for the consequences of use of
  60  *              this software, no matter how awful, even if they arise
  61  *              from defects in it.
  62  *
  63  *      2. The origin of this software must not be misrepresented, either
  64  *              by explicit claim or by omission.
  65  *
  66  *      3. Altered versions must be plainly marked as such, and must not
  67  *              be misrepresented as being the original software.
  68  *
  69  *
  70  ****    Alterations to Henry's code are...
  71  ****
  72  ****    Copyright (c) 1991-2000, Larry Wall
  73  ****
  74  ****    You may distribute under the terms of either the GNU General Public
  75  ****    License or the Artistic License, as specified in the README file.
  76
  77  *
  78  * Beware that some of this code is subtly aware of the way operator
  79  * precedence is structured in regular expressions.  Serious changes in
  80  * regular-expression syntax might require a total rethink.
  81  */
  82 #include "EXTERN.h"
  83 #define PERL_IN_REGCOMP_C
  84 #include "perl.h"
  85
  86 #ifdef PERL_IN_XSUB_RE
  87 #  if defined(PERL_CAPI) || defined(PERL_OBJECT)
  88 #    include "XSUB.h"
  89 #  endif
  90 #else
  91 #  include "INTERN.h"
  92 #endif
  93
  94 #define REG_COMP_C
  95 #include "regcomp.h"
  96
  97 #ifdef op
  98 #undef op
  99 #endif /* op */
 100
 101 #ifdef MSDOS
 102 # if defined(BUGGY_MSC6)
 103  /* MSC 6.00A breaks on op/regexp.t test 85 unless we turn this off */
 104  # pragma optimize("a",off)
 105  /* But MSC 6.00A is happy with 'w', for aliases only across function calls*/
 106  # pragma optimize("w",on )
 107 # endif /* BUGGY_MSC6 */
 108 #endif /* MSDOS */
 109
 110 #ifndef STATIC
 111 #define STATIC  static
 112 #endif
 113
 114 typedef struct RExC_state_t {
 115     U16         flags16;                /* are we folding, multilining? */
 116     char        *precomp;               /* uncompiled string. */
 117     regexp      *rx;
 118     char        *end;                   /* End of input for compile */
 119     char        *parse;                 /* Input-scan pointer. */
 120     I32         whilem_seen;            /* number of WHILEM in this expr */
 121     regnode     *emit;                  /* Code-emit pointer; &regdummy = don't */
 122     I32         naughty;                /* How bad is this pattern? */
 123     I32         sawback;                /* Did we see \1, ...? */
 124     U32         seen;
 125     I32         size;                   /* Code size. */
 126     I32         npar;                   /* () count. */
 127     I32         extralen;
 128     I32         seen_zerolen;
 129     I32         seen_evals;
 130 #if ADD_TO_REGEXEC
 131     char        *starttry;              /* -Dr: where regtry was called. */
 132 #define RExC_starttry   (pRExC_state->starttry)
 133 #endif
 134 } RExC_state_t;
 135
 136 #define RExC_flags16    (pRExC_state->flags16)
 137 #define RExC_precomp    (pRExC_state->precomp)
 138 #define RExC_rx         (pRExC_state->rx)
 139 #define RExC_end        (pRExC_state->end)
 140 #define RExC_parse      (pRExC_state->parse)
 141 #define RExC_whilem_seen        (pRExC_state->whilem_seen)
 142 #define RExC_emit       (pRExC_state->emit)
 143 #define RExC_naughty    (pRExC_state->naughty)
 144 #define RExC_sawback    (pRExC_state->sawback)
 145 #define RExC_seen       (pRExC_state->seen)
 146 #define RExC_size       (pRExC_state->size)
 147 #define RExC_npar       (pRExC_state->npar)
 148 #define RExC_extralen   (pRExC_state->extralen)
 149 #define RExC_seen_zerolen       (pRExC_state->seen_zerolen)
 150 #define RExC_seen_evals (pRExC_state->seen_evals)
 151
 152 #define ISMULT1(c)      ((c) == '*' || (c) == '+' || (c) == '?')
 153 #define ISMULT2(s)      ((*s) == '*' || (*s) == '+' || (*s) == '?' || \
 154         ((*s) == '{' && regcurly(s)))
 155 #ifdef atarist
 156 #define PERL_META       "^$.[()|?+*\\"
 157 #else
 158 #define META    "^$.[()|?+*\\"
 159 #endif
 160
 161 #ifdef SPSTART
 162 #undef SPSTART          /* dratted cpp namespace... */
 163 #endif
 164 /*
 165  * Flags to be passed up and down.
 166  */
 167 #define WORST           0       /* Worst case. */
 168 #define HASWIDTH        0x1     /* Known to match non-null strings. */
 169 #define SIMPLE          0x2     /* Simple enough to be STAR/PLUS operand. */
 170 #define SPSTART         0x4     /* Starts with * or +. */
 171 #define TRYAGAIN        0x8     /* Weeded out a declaration. */
 172
 173 /* Length of a variant. */
 174
 175 typedef struct scan_data_t {
 176     I32 len_min;
 177     I32 len_delta;
 178     I32 pos_min;
 179     I32 pos_delta;
 180     SV *last_found;
 181     I32 last_end;                       /* min value, <0 unless valid. */
 182     I32 last_start_min;
 183     I32 last_start_max;
 184     SV **longest;                       /* Either &l_fixed, or &l_float. */
 185     SV *longest_fixed;
 186     I32 offset_fixed;
 187     SV *longest_float;
 188     I32 offset_float_min;
 189     I32 offset_float_max;
 190     I32 flags;
 191     I32 whilem_c;
 192     I32 *last_closep;
 193     struct regnode_charclass_class *start_class;
 194 } scan_data_t;
 195
 196 /*
 197  * Forward declarations for pregcomp()'s friends.
 198  */
 199
 200 static scan_data_t zero_scan_data = { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
 201                                       0, 0, 0, 0, 0, 0};
 202
 203 #define SF_BEFORE_EOL           (SF_BEFORE_SEOL|SF_BEFORE_MEOL)
 204 #define SF_BEFORE_SEOL          0x1
 205 #define SF_BEFORE_MEOL          0x2
 206 #define SF_FIX_BEFORE_EOL       (SF_FIX_BEFORE_SEOL|SF_FIX_BEFORE_MEOL)
 207 #define SF_FL_BEFORE_EOL        (SF_FL_BEFORE_SEOL|SF_FL_BEFORE_MEOL)
 208
 209 #ifdef NO_UNARY_PLUS
 210 #  define SF_FIX_SHIFT_EOL      (0+2)
 211 #  define SF_FL_SHIFT_EOL               (0+4)
 212 #else
 213 #  define SF_FIX_SHIFT_EOL      (+2)
 214 #  define SF_FL_SHIFT_EOL               (+4)
 215 #endif
 216
 217 #define SF_FIX_BEFORE_SEOL      (SF_BEFORE_SEOL << SF_FIX_SHIFT_EOL)
 218 #define SF_FIX_BEFORE_MEOL      (SF_BEFORE_MEOL << SF_FIX_SHIFT_EOL)
 219
 220 #define SF_FL_BEFORE_SEOL       (SF_BEFORE_SEOL << SF_FL_SHIFT_EOL)
 221 #define SF_FL_BEFORE_MEOL       (SF_BEFORE_MEOL << SF_FL_SHIFT_EOL) /* 0x20 */
 222 #define SF_IS_INF               0x40
 223 #define SF_HAS_PAR              0x80
 224 #define SF_IN_PAR               0x100
 225 #define SF_HAS_EVAL             0x200
 226 #define SCF_DO_SUBSTR           0x400
 227 #define SCF_DO_STCLASS_AND      0x0800
 228 #define SCF_DO_STCLASS_OR       0x1000
 229 #define SCF_DO_STCLASS          (SCF_DO_STCLASS_AND|SCF_DO_STCLASS_OR)
 230 #define SCF_WHILEM_VISITED_POS  0x2000
 231
 232 #define RF_utf8         8
 233 #define UTF (PL_reg_flags & RF_utf8)
 234 #define LOC (RExC_flags16 & PMf_LOCALE)
 235 #define FOLD (RExC_flags16 & PMf_FOLD)
 236
 237 #define OOB_CHAR8               1234
 238 #define OOB_UTF8                123456
 239 #define OOB_NAMEDCLASS          -1
 240
 241 #define CHR_SVLEN(sv) (UTF ? sv_len_utf8(sv) : SvCUR(sv))
 242 #define CHR_DIST(a,b) (UTF ? utf8_distance(a,b) : a - b)
 243
 244
 245 /* length of regex to show in messages that don't mark a position within */
 246 #define RegexLengthToShowInErrorMessages 127
 247
 248 /*
 249  * If MARKER[12] are adjusted, be sure to adjust the constants at the top
 250  * of t/op/regmesg.t, the tests in t/op/re_tests, and those in
 251  * op/pragma/warn/regcomp.
 252  */
 253 #define MARKER1 "HERE"      /* marker as it appears in the description */
 254 #define MARKER2 " << HERE "  /* marker as it appears within the regex */
 255
 256 #define REPORT_LOCATION " before " MARKER1 " mark in regex m/%.*s" MARKER2 "%s/"
 257
 258 /*
 259  * Calls SAVEDESTRUCTOR_X if needed, then calls Perl_croak with the given
 260  * arg. Show regex, up to a maximum length. If it's too long, chop and add
 261  * "...".
 262  */
 263 #define FAIL(msg)                                                             \
 264     STMT_START {                                                             \
 265         char *ellipses = "";                                                 \
 266         unsigned len = strlen(RExC_precomp);                                \
 267                                                                              \
 268         if (!SIZE_ONLY)                                                      \
 269             SAVEDESTRUCTOR_X(clear_re,(void*)RExC_rx);                 \
 270                                                                              \
 271         if (len > RegexLengthToShowInErrorMessages) {                        \
 272             /* chop 10 shorter than the max, to ensure meaning of "..." */   \
 273             len = RegexLengthToShowInErrorMessages - 10;                     \
 274             ellipses = "...";                                                \
 275         }                                                                    \
 276         Perl_croak(aTHX_ "%s in regex m/%.*s%s/",                            \
 277                    msg, (int)len, RExC_precomp, ellipses);                  \
 278     } STMT_END
 279
 280 /*
 281  * Calls SAVEDESTRUCTOR_X if needed, then calls Perl_croak with the given
 282  * args. Show regex, up to a maximum length. If it's too long, chop and add
 283  * "...".
 284  */
 285 #define FAIL2(pat,msg)                                                        \
 286     STMT_START {                                                             \
 287         char *ellipses = "";                                                 \
 288         unsigned len = strlen(RExC_precomp);                                \
 289                                                                              \
 290         if (!SIZE_ONLY)                                                      \
 291             SAVEDESTRUCTOR_X(clear_re,(void*)RExC_rx);                 \
 292                                                                              \
 293         if (len > RegexLengthToShowInErrorMessages) {                        \
 294             /* chop 10 shorter than the max, to ensure meaning of "..." */   \
 295             len = RegexLengthToShowInErrorMessages - 10;                     \
 296             ellipses = "...";                                                \
 297         }                                                                    \
 298         S_re_croak2(aTHX_ pat, " in regex m/%.*s%s/",                        \
 299                     msg, (int)len, RExC_precomp, ellipses);                \
 300     } STMT_END
 301
 302
 303 /*
 304  * Simple_vFAIL -- like FAIL, but marks the current location in the scan
 305  */
 306 #define Simple_vFAIL(m)                                                      \
 307     STMT_START {                                                             \
 308       unsigned offset = strlen(RExC_precomp)-(RExC_end-RExC_parse); \
 309                                                                              \
 310       Perl_croak(aTHX_ "%s" REPORT_LOCATION,               \
 311                  m, (int)offset, RExC_precomp, RExC_precomp + offset);     \
 312     } STMT_END
 313
 314 /*
 315  * Calls SAVEDESTRUCTOR_X if needed, then Simple_vFAIL()
 316  */
 317 #define vFAIL(m)                                                             \
 318     STMT_START {                                                             \
 319       if (!SIZE_ONLY)                                                        \
 320             SAVEDESTRUCTOR_X(clear_re,(void*)RExC_rx);                 \
 321       Simple_vFAIL(m);                                                       \
 322     } STMT_END
 323
 324 /*
 325  * Like Simple_vFAIL(), but accepts two arguments.
 326  */
 327 #define Simple_vFAIL2(m,a1)                                                  \
 328     STMT_START {                                                             \
 329       unsigned offset = strlen(RExC_precomp)-(RExC_end-RExC_parse); \
 330                                                                              \
 331       S_re_croak2(aTHX_ m, REPORT_LOCATION, a1,       \
 332                   (int)offset, RExC_precomp, RExC_precomp + offset);       \
 333     } STMT_END
 334
 335 /*
 336  * Calls SAVEDESTRUCTOR_X if needed, then Simple_vFAIL2().
 337  */
 338 #define vFAIL2(m,a1)                                                         \
 339     STMT_START {                                                             \
 340       if (!SIZE_ONLY)                                                        \
 341             SAVEDESTRUCTOR_X(clear_re,(void*)RExC_rx);                 \
 342       Simple_vFAIL2(m, a1);                                                  \
 343     } STMT_END
 344
 345
 346 /*
 347  * Like Simple_vFAIL(), but accepts three arguments.
 348  */
 349 #define Simple_vFAIL3(m, a1, a2)                                             \
 350     STMT_START {                                                             \
 351       unsigned offset = strlen(RExC_precomp)-(RExC_end-RExC_parse); \
 352                                                                              \
 353       S_re_croak2(aTHX_ m, REPORT_LOCATION, a1, a2,   \
 354                   (int)offset, RExC_precomp, RExC_precomp + offset);       \
 355     } STMT_END
 356
 357 /*
 358  * Calls SAVEDESTRUCTOR_X if needed, then Simple_vFAIL3().
 359  */
 360 #define vFAIL3(m,a1,a2)                                                      \
 361     STMT_START {                                                             \
 362       if (!SIZE_ONLY)                                                        \
 363             SAVEDESTRUCTOR_X(clear_re,(void*)RExC_rx);                 \
 364       Simple_vFAIL3(m, a1, a2);                                              \
 365     } STMT_END
 366
 367 /*
 368  * Like Simple_vFAIL(), but accepts four arguments.
 369  */
 370 #define Simple_vFAIL4(m, a1, a2, a3)                                         \
 371     STMT_START {                                                             \
 372       unsigned offset = strlen(RExC_precomp)-(RExC_end-RExC_parse); \
 373                                                                              \
 374       S_re_croak2(aTHX_ m, REPORT_LOCATION, a1, a2, a3,\
 375                   (int)offset, RExC_precomp, RExC_precomp + offset);       \
 376     } STMT_END
 377
 378 /*
 379  * Like Simple_vFAIL(), but accepts five arguments.
 380  */
 381 #define Simple_vFAIL5(m, a1, a2, a3, a4)                                     \
 382     STMT_START {                                                             \
 383       unsigned offset = strlen(RExC_precomp)-(RExC_end-RExC_parse); \
 384       S_re_croak2(aTHX_ m, REPORT_LOCATION, a1, a2, a3, a4,\
 385                   (int)offset, RExC_precomp, RExC_precomp + offset);       \
 386     } STMT_END
 387
 388
 389 #define vWARN(loc,m)                                                         \
 390     STMT_START {                                                             \
 391         unsigned offset = strlen(RExC_precomp)-(RExC_end-(loc));          \
 392         Perl_warner(aTHX_ WARN_REGEXP, "%s" REPORT_LOCATION,\
 393                  m, (int)offset, RExC_precomp, RExC_precomp + offset);          \
 394     } STMT_END                                                               \
 395
 396
 397 #define vWARN2(loc, m, a1)                                                   \
 398     STMT_START {                                                             \
 399         unsigned offset = strlen(RExC_precomp)-(RExC_end-(loc));          \
 400         Perl_warner(aTHX_ WARN_REGEXP, m REPORT_LOCATION,\
 401                  a1,                                                         \
 402                  (int)offset, RExC_precomp, RExC_precomp + offset);        \
 403     } STMT_END
 404
 405 #define vWARN3(loc, m, a1, a2)                                               \
 406     STMT_START {                                                             \
 407       unsigned offset = strlen(RExC_precomp) - (RExC_end - (loc));        \
 408         Perl_warner(aTHX_ WARN_REGEXP, m REPORT_LOCATION,                    \
 409                  a1, a2,                                                     \
 410                  (int)offset, RExC_precomp, RExC_precomp + offset);        \
 411     } STMT_END
 412
 413 #define vWARN4(loc, m, a1, a2, a3)                                           \
 414     STMT_START {                                                             \
 415       unsigned offset = strlen(RExC_precomp)-(RExC_end-(loc));            \
 416         Perl_warner(aTHX_ WARN_REGEXP, m REPORT_LOCATION,\
 417                  a1, a2, a3,                                                 \
 418                  (int)offset, RExC_precomp, RExC_precomp + offset);        \
 419     } STMT_END
 420
 421
 422 /* Allow for side effects in s */
 423 #define REGC(c,s) STMT_START { if (!SIZE_ONLY) *(s) = (c); else (s);} STMT_END
 424
 425 static void clear_re(pTHXo_ void *r);
 426
 427 /* Mark that we cannot extend a found fixed substring at this point.
 428    Updata the longest found anchored substring and the longest found
 429    floating substrings if needed. */
 430
 431 STATIC void
 432 S_scan_commit(pTHX_ RExC_state_t *pRExC_state, scan_data_t *data)
 433 {
 434     STRLEN l = CHR_SVLEN(data->last_found);
 435     STRLEN old_l = CHR_SVLEN(*data->longest);
 436
 437     if ((l >= old_l) && ((l > old_l) || (data->flags & SF_BEFORE_EOL))) {
 438         sv_setsv(*data->longest, data->last_found);
 439         if (*data->longest == data->longest_fixed) {
 440             data->offset_fixed = l ? data->last_start_min : data->pos_min;
 441             if (data->flags & SF_BEFORE_EOL)
 442                 data->flags
 443                     |= ((data->flags & SF_BEFORE_EOL) << SF_FIX_SHIFT_EOL);
 444             else
 445                 data->flags &= ~SF_FIX_BEFORE_EOL;
 446         }
 447         else {
 448             data->offset_float_min = l ? data->last_start_min : data->pos_min;
 449             data->offset_float_max = (l
 450                                       ? data->last_start_max
 451                                       : data->pos_min + data->pos_delta);
 452             if (data->flags & SF_BEFORE_EOL)
 453                 data->flags
 454                     |= ((data->flags & SF_BEFORE_EOL) << SF_FL_SHIFT_EOL);
 455             else
 456                 data->flags &= ~SF_FL_BEFORE_EOL;
 457         }
 458     }
 459     SvCUR_set(data->last_found, 0);
 460     data->last_end = -1;
 461     data->flags &= ~SF_BEFORE_EOL;
 462 }
 463
 464 /* Can match anything (initialization) */
 465 STATIC void
 466 S_cl_anything(pTHX_ RExC_state_t *pRExC_state, struct regnode_charclass_class *cl)
 467 {
 468     int value;
 469
 470     ANYOF_CLASS_ZERO(cl);
 471     for (value = 0; value < 256; ++value)
 472         ANYOF_BITMAP_SET(cl, value);
 473     cl->flags = ANYOF_EOS;
 474     if (LOC)
 475         cl->flags |= ANYOF_LOCALE;
 476 }
 477
 478 /* Can match anything (initialization) */
 479 STATIC int
 480 S_cl_is_anything(pTHX_ struct regnode_charclass_class *cl)
 481 {
 482     int value;
 483
 484     for (value = 0; value <= ANYOF_MAX; value += 2)
 485         if (ANYOF_CLASS_TEST(cl, value) && ANYOF_CLASS_TEST(cl, value + 1))
 486             return 1;
 487     for (value = 0; value < 256; ++value)
 488         if (!ANYOF_BITMAP_TEST(cl, value))
 489             return 0;
 490     return 1;
 491 }
 492
 493 /* Can match anything (initialization) */
 494 STATIC void
 495 S_cl_init(pTHX_ RExC_state_t *pRExC_state, struct regnode_charclass_class *cl)
 496 {
 497     Zero(cl, 1, struct regnode_charclass_class);
 498     cl->type = ANYOF;
 499     cl_anything(pRExC_state, cl);
 500 }
 501
 502 STATIC void
 503 S_cl_init_zero(pTHX_ RExC_state_t *pRExC_state, struct regnode_charclass_class *cl)
 504 {
 505     Zero(cl, 1, struct regnode_charclass_class);
 506     cl->type = ANYOF;
 507     cl_anything(pRExC_state, cl);
 508     if (LOC)
 509         cl->flags |= ANYOF_LOCALE;
 510 }
 511
 512 /* 'And' a given class with another one.  Can create false positives */
 513 /* We assume that cl is not inverted */
 514 STATIC void
 515 S_cl_and(pTHX_ struct regnode_charclass_class *cl,
 516          struct regnode_charclass_class *and_with)
 517 {
 518     if (!(and_with->flags & ANYOF_CLASS)
 519         && !(cl->flags & ANYOF_CLASS)
 520         && (and_with->flags & ANYOF_LOCALE) == (cl->flags & ANYOF_LOCALE)
 521         && !(and_with->flags & ANYOF_FOLD)
 522         && !(cl->flags & ANYOF_FOLD)) {
 523         int i;
 524
 525         if (and_with->flags & ANYOF_INVERT)
 526             for (i = 0; i < ANYOF_BITMAP_SIZE; i++)
 527                 cl->bitmap[i] &= ~and_with->bitmap[i];
 528         else
 529             for (i = 0; i < ANYOF_BITMAP_SIZE; i++)
 530                 cl->bitmap[i] &= and_with->bitmap[i];
 531     } /* XXXX: logic is complicated otherwise, leave it along for a moment. */
 532     if (!(and_with->flags & ANYOF_EOS))
 533         cl->flags &= ~ANYOF_EOS;
 534 }
 535
 536 /* 'OR' a given class with another one.  Can create false positives */
 537 /* We assume that cl is not inverted */
 538 STATIC void
 539 S_cl_or(pTHX_ RExC_state_t *pRExC_state, struct regnode_charclass_class *cl, struct regnode_charclass_class *or_with)
 540 {
 541     if (or_with->flags & ANYOF_INVERT) {
 542         /* We do not use
 543          * (B1 | CL1) | (!B2 & !CL2) = (B1 | !B2 & !CL2) | (CL1 | (!B2 & !CL2))
 544          *   <= (B1 | !B2) | (CL1 | !CL2)
 545          * which is wasteful if CL2 is small, but we ignore CL2:
 546          *   (B1 | CL1) | (!B2 & !CL2) <= (B1 | CL1) | !B2 = (B1 | !B2) | CL1
 547          * XXXX Can we handle case-fold?  Unclear:
 548          *   (OK1(i) | OK1(i')) | !(OK1(i) | OK1(i')) =
 549          *   (OK1(i) | OK1(i')) | (!OK1(i) & !OK1(i'))
 550          */
 551         if ( (or_with->flags & ANYOF_LOCALE) == (cl->flags & ANYOF_LOCALE)
 552              && !(or_with->flags & ANYOF_FOLD)
 553              && !(cl->flags & ANYOF_FOLD) ) {
 554             int i;
 555
 556             for (i = 0; i < ANYOF_BITMAP_SIZE; i++)
 557                 cl->bitmap[i] |= ~or_with->bitmap[i];
 558         } /* XXXX: logic is complicated otherwise */
 559         else {
 560             cl_anything(pRExC_state, cl);
 561         }
 562     } else {
 563         /* (B1 | CL1) | (B2 | CL2) = (B1 | B2) | (CL1 | CL2)) */
 564         if ( (or_with->flags & ANYOF_LOCALE) == (cl->flags & ANYOF_LOCALE)
 565              && (!(or_with->flags & ANYOF_FOLD)
 566                  || (cl->flags & ANYOF_FOLD)) ) {
 567             int i;
 568
 569             /* OR char bitmap and class bitmap separately */
 570             for (i = 0; i < ANYOF_BITMAP_SIZE; i++)
 571                 cl->bitmap[i] |= or_with->bitmap[i];
 572             if (or_with->flags & ANYOF_CLASS) {
 573                 for (i = 0; i < ANYOF_CLASSBITMAP_SIZE; i++)
 574                     cl->classflags[i] |= or_with->classflags[i];
 575                 cl->flags |= ANYOF_CLASS;
 576             }
 577         }
 578         else { /* XXXX: logic is complicated, leave it along for a moment. */
 579             cl_anything(pRExC_state, cl);
 580         }
 581     }
 582     if (or_with->flags & ANYOF_EOS)
 583         cl->flags |= ANYOF_EOS;
 584 }
 585
 586 /* REx optimizer.  Converts nodes into quickier variants "in place".
 587    Finds fixed substrings.  */
 588
 589 /* Stops at toplevel WHILEM as well as at `last'. At end *scanp is set
 590    to the position after last scanned or to NULL. */
 591
 592 STATIC I32
 593 S_study_chunk(pTHX_ RExC_state_t *pRExC_state, regnode **scanp, I32 *deltap, regnode *last, scan_data_t *data, U32 flags)
 594                         /* scanp: Start here (read-write). */
 595                         /* deltap: Write maxlen-minlen here. */
 596                         /* last: Stop before this one. */
 597 {
 598     I32 min = 0, pars = 0, code;
 599     regnode *scan = *scanp, *next;
 600     I32 delta = 0;
 601     int is_inf = (flags & SCF_DO_SUBSTR) && (data->flags & SF_IS_INF);
 602     int is_inf_internal = 0;            /* The studied chunk is infinite */
 603     I32 is_par = OP(scan) == OPEN ? ARG(scan) : 0;
 604     scan_data_t data_fake;
 605     struct regnode_charclass_class and_with; /* Valid if flags & SCF_DO_STCLASS_OR */
 606
 607     while (scan && OP(scan) != END && scan < last) {
 608         /* Peephole optimizer: */
 609
 610         if (PL_regkind[(U8)OP(scan)] == EXACT) {
 611             /* Merge several consecutive EXACTish nodes into one. */
 612             regnode *n = regnext(scan);
 613             U32 stringok = 1;
 614 #ifdef DEBUGGING
 615             regnode *stop = scan;
 616 #endif
 617
 618             next = scan + NODE_SZ_STR(scan);
 619             /* Skip NOTHING, merge EXACT*. */
 620             while (n &&
 621                    ( PL_regkind[(U8)OP(n)] == NOTHING ||
 622                      (stringok && (OP(n) == OP(scan))))
 623                    && NEXT_OFF(n)
 624                    && NEXT_OFF(scan) + NEXT_OFF(n) < I16_MAX) {
 625                 if (OP(n) == TAIL || n > next)
 626                     stringok = 0;
 627                 if (PL_regkind[(U8)OP(n)] == NOTHING) {
 628                     NEXT_OFF(scan) += NEXT_OFF(n);
 629                     next = n + NODE_STEP_REGNODE;
 630 #ifdef DEBUGGING
 631                     if (stringok)
 632                         stop = n;
 633 #endif
 634                     n = regnext(n);
 635                 }
 636                 else {
 637                     int oldl = STR_LEN(scan);
 638                     regnode *nnext = regnext(n);
 639
 640                     if (oldl + STR_LEN(n) > U8_MAX)
 641                         break;
 642                     NEXT_OFF(scan) += NEXT_OFF(n);
 643                     STR_LEN(scan) += STR_LEN(n);
 644                     next = n + NODE_SZ_STR(n);
 645                     /* Now we can overwrite *n : */
 646                     Move(STRING(n), STRING(scan) + oldl,
 647                          STR_LEN(n), char);
 648 #ifdef DEBUGGING
 649                     if (stringok)
 650                         stop = next - 1;
 651 #endif
 652                     n = nnext;
 653                 }
 654             }
 655 #ifdef DEBUGGING
 656             /* Allow dumping */
 657             n = scan + NODE_SZ_STR(scan);
 658             while (n <= stop) {
 659                 if (PL_regkind[(U8)OP(n)] != NOTHING || OP(n) == NOTHING) {
 660                     OP(n) = OPTIMIZED;
 661                     NEXT_OFF(n) = 0;
 662                 }
 663                 n++;
 664             }
 665 #endif
 666         }
 667         /* Follow the next-chain of the current node and optimize
 668            away all the NOTHINGs from it.  */
 669         if (OP(scan) != CURLYX) {
 670             int max = (reg_off_by_arg[OP(scan)]
 671                        ? I32_MAX
 672                        /* I32 may be smaller than U16 on CRAYs! */
 673                        : (I32_MAX < U16_MAX ? I32_MAX : U16_MAX));
 674             int off = (reg_off_by_arg[OP(scan)] ? ARG(scan) : NEXT_OFF(scan));
 675             int noff;
 676             regnode *n = scan;
 677
 678             /* Skip NOTHING and LONGJMP. */
 679             while ((n = regnext(n))
 680                    && ((PL_regkind[(U8)OP(n)] == NOTHING && (noff = NEXT_OFF(n)))
 681                        || ((OP(n) == LONGJMP) && (noff = ARG(n))))
 682                    && off + noff < max)
 683                 off += noff;
 684             if (reg_off_by_arg[OP(scan)])
 685                 ARG(scan) = off;
 686             else
 687                 NEXT_OFF(scan) = off;
 688         }
 689         /* The principal pseudo-switch.  Cannot be a switch, since we
 690            look into several different things.  */
 691         if (OP(scan) == BRANCH || OP(scan) == BRANCHJ
 692                    || OP(scan) == IFTHEN || OP(scan) == SUSPEND) {
 693             next = regnext(scan);
 694             code = OP(scan);
 695
 696             if (OP(next) == code || code == IFTHEN || code == SUSPEND) {
 697                 I32 max1 = 0, min1 = I32_MAX, num = 0;
 698                 struct regnode_charclass_class accum;
 699
 700                 if (flags & SCF_DO_SUBSTR) /* XXXX Add !SUSPEND? */
 701                     scan_commit(pRExC_state, data); /* Cannot merge strings after this. */
 702                 if (flags & SCF_DO_STCLASS)
 703                     cl_init_zero(pRExC_state, &accum);
 704                 while (OP(scan) == code) {
 705                     I32 deltanext, minnext, f = 0, fake;
 706                     struct regnode_charclass_class this_class;
 707
 708                     num++;
 709                     data_fake.flags = 0;
 710                     if (data) {
 711                         data_fake.whilem_c = data->whilem_c;
 712                         data_fake.last_closep = data->last_closep;
 713                     }
 714                     else
 715                         data_fake.last_closep = &fake;
 716                     next = regnext(scan);
 717                     scan = NEXTOPER(scan);
 718                     if (code != BRANCH)
 719                         scan = NEXTOPER(scan);
 720                     if (flags & SCF_DO_STCLASS) {
 721                         cl_init(pRExC_state, &this_class);
 722                         data_fake.start_class = &this_class;
 723                         f = SCF_DO_STCLASS_AND;
 724                     }
 725                     if (flags & SCF_WHILEM_VISITED_POS)
 726                         f |= SCF_WHILEM_VISITED_POS;
 727                     /* we suppose the run is continuous, last=next...*/
 728                     minnext = study_chunk(pRExC_state, &scan, &deltanext,
 729                                           next, &data_fake, f);
 730                     if (min1 > minnext)
 731                         min1 = minnext;
 732                     if (max1 < minnext + deltanext)
 733                         max1 = minnext + deltanext;
 734                     if (deltanext == I32_MAX)
 735                         is_inf = is_inf_internal = 1;
 736                     scan = next;
 737                     if (data_fake.flags & (SF_HAS_PAR|SF_IN_PAR))
 738                         pars++;
 739                     if (data && (data_fake.flags & SF_HAS_EVAL))
 740                         data->flags |= SF_HAS_EVAL;
 741                     if (data)
 742                         data->whilem_c = data_fake.whilem_c;
 743                     if (flags & SCF_DO_STCLASS)
 744                         cl_or(pRExC_state, &accum, &this_class);
 745                     if (code == SUSPEND)
 746                         break;
 747                 }
 748                 if (code == IFTHEN && num < 2) /* Empty ELSE branch */
 749                     min1 = 0;
 750                 if (flags & SCF_DO_SUBSTR) {
 751                     data->pos_min += min1;
 752                     data->pos_delta += max1 - min1;
 753                     if (max1 != min1 || is_inf)
 754                         data->longest = &(data->longest_float);
 755                 }
 756                 min += min1;
 757                 delta += max1 - min1;
 758                 if (flags & SCF_DO_STCLASS_OR) {
 759                     cl_or(pRExC_state, data->start_class, &accum);
 760                     if (min1) {
 761                         cl_and(data->start_class, &and_with);
 762                         flags &= ~SCF_DO_STCLASS;
 763                     }
 764                 }
 765                 else if (flags & SCF_DO_STCLASS_AND) {
 766                     if (min1) {
 767                         cl_and(data->start_class, &accum);
 768                         flags &= ~SCF_DO_STCLASS;
 769                     }
 770                     else {
 771                         /* Switch to OR mode: cache the old value of
 772                          * data->start_class */
 773                         StructCopy(data->start_class, &and_with,
 774                                    struct regnode_charclass_class);
 775                         flags &= ~SCF_DO_STCLASS_AND;
 776                         StructCopy(&accum, data->start_class,
 777                                    struct regnode_charclass_class);
 778                         flags |= SCF_DO_STCLASS_OR;
 779                         data->start_class->flags |= ANYOF_EOS;
 780                     }
 781                 }
 782             }
 783             else if (code == BRANCHJ)   /* single branch is optimized. */
 784                 scan = NEXTOPER(NEXTOPER(scan));
 785             else                        /* single branch is optimized. */
 786                 scan = NEXTOPER(scan);
 787             continue;
 788         }
 789         else if (OP(scan) == EXACT) {
 790             I32 l = STR_LEN(scan);
 791             if (UTF) {
 792                 unsigned char *s = (unsigned char *)STRING(scan);
 793                 unsigned char *e = s + l;
 794                 I32 newl = 0;
 795                 while (s < e) {
 796                     newl++;
 797                     s += UTF8SKIP(s);
 798                 }
 799                 l = newl;
 800             }
 801             min += l;
 802             if (flags & SCF_DO_SUBSTR) { /* Update longest substr. */
 803                 /* The code below prefers earlier match for fixed
 804                    offset, later match for variable offset.  */
 805                 if (data->last_end == -1) { /* Update the start info. */
 806                     data->last_start_min = data->pos_min;
 807                     data->last_start_max = is_inf
 808                         ? I32_MAX : data->pos_min + data->pos_delta;
 809                 }
 810                 sv_catpvn(data->last_found, STRING(scan), STR_LEN(scan));
 811                 data->last_end = data->pos_min + l;
 812                 data->pos_min += l; /* As in the first entry. */
 813                 data->flags &= ~SF_BEFORE_EOL;
 814             }
 815             if (flags & SCF_DO_STCLASS_AND) {
 816                 /* Check whether it is compatible with what we know already! */
 817                 int compat = 1;
 818
 819                 if (!(data->start_class->flags & (ANYOF_CLASS | ANYOF_LOCALE))
 820                     && !ANYOF_BITMAP_TEST(data->start_class, *STRING(scan))
 821                     && (!(data->start_class->flags & ANYOF_FOLD)
 822                         || !ANYOF_BITMAP_TEST(data->start_class,
 823                                               PL_fold[*(U8*)STRING(scan)])))
 824                     compat = 0;
 825                 ANYOF_CLASS_ZERO(data->start_class);
 826                 ANYOF_BITMAP_ZERO(data->start_class);
 827                 if (compat)
 828                     ANYOF_BITMAP_SET(data->start_class, *STRING(scan));
 829                 data->start_class->flags &= ~ANYOF_EOS;
 830             }
 831             else if (flags & SCF_DO_STCLASS_OR) {
 832                 /* false positive possible if the class is case-folded */
 833                 ANYOF_BITMAP_SET(data->start_class, *STRING(scan));
 834                 data->start_class->flags &= ~ANYOF_EOS;
 835                 cl_and(data->start_class, &and_with);
 836             }
 837             flags &= ~SCF_DO_STCLASS;
 838         }
 839         else if (PL_regkind[(U8)OP(scan)] == EXACT) { /* But OP != EXACT! */
 840             I32 l = STR_LEN(scan);
 841
 842             /* Search for fixed substrings supports EXACT only. */
 843             if (flags & SCF_DO_SUBSTR)
 844                 scan_commit(pRExC_state, data);
 845             if (UTF) {
 846                 unsigned char *s = (unsigned char *)STRING(scan);
 847                 unsigned char *e = s + l;
 848                 I32 newl = 0;
 849                 while (s < e) {
 850                     newl++;
 851                     s += UTF8SKIP(s);
 852                 }
 853                 l = newl;
 854             }
 855             min += l;
 856             if (data && (flags & SCF_DO_SUBSTR))
 857                 data->pos_min += l;
 858             if (flags & SCF_DO_STCLASS_AND) {
 859                 /* Check whether it is compatible with what we know already! */
 860                 int compat = 1;
 861
 862                 if (!(data->start_class->flags & (ANYOF_CLASS | ANYOF_LOCALE))
 863                     && !ANYOF_BITMAP_TEST(data->start_class, *STRING(scan))
 864                     && !ANYOF_BITMAP_TEST(data->start_class,
 865                                           PL_fold[*(U8*)STRING(scan)]))
 866                     compat = 0;
 867                 ANYOF_CLASS_ZERO(data->start_class);
 868                 ANYOF_BITMAP_ZERO(data->start_class);
 869                 if (compat) {
 870                     ANYOF_BITMAP_SET(data->start_class, *STRING(scan));
 871                     data->start_class->flags &= ~ANYOF_EOS;
 872                     data->start_class->flags |= ANYOF_FOLD;
 873                     if (OP(scan) == EXACTFL)
 874                         data->start_class->flags |= ANYOF_LOCALE;
 875                 }
 876             }
 877             else if (flags & SCF_DO_STCLASS_OR) {
 878                 if (data->start_class->flags & ANYOF_FOLD) {
 879                     /* false positive possible if the class is case-folded.
 880                        Assume that the locale settings are the same... */
 881                     ANYOF_BITMAP_SET(data->start_class, *STRING(scan));
 882                     data->start_class->flags &= ~ANYOF_EOS;
 883                 }
 884                 cl_and(data->start_class, &and_with);
 885             }
 886             flags &= ~SCF_DO_STCLASS;
 887         }
 888         else if (strchr((char*)PL_varies,OP(scan))) {
 889             I32 mincount, maxcount, minnext, deltanext, fl;
 890             I32 f = flags, pos_before = 0;
 891             regnode *oscan = scan;
 892             struct regnode_charclass_class this_class;
 893             struct regnode_charclass_class *oclass = NULL;
 894
 895             switch (PL_regkind[(U8)OP(scan)]) {
 896             case WHILEM:                /* End of (?:...)* . */
 897                 scan = NEXTOPER(scan);
 898                 goto finish;
 899             case PLUS:
 900                 if (flags & (SCF_DO_SUBSTR | SCF_DO_STCLASS)) {
 901                     next = NEXTOPER(scan);
 902                     if (OP(next) == EXACT || (flags & SCF_DO_STCLASS)) {
 903                         mincount = 1;
 904                         maxcount = REG_INFTY;
 905                         next = regnext(scan);
 906                         scan = NEXTOPER(scan);
 907                         goto do_curly;
 908                     }
 909                 }
 910                 if (flags & SCF_DO_SUBSTR)
 911                     data->pos_min++;
 912                 min++;
 913                 /* Fall through. */
 914             case STAR:
 915                 if (flags & SCF_DO_STCLASS) {
 916                     mincount = 0;
 917                     maxcount = REG_INFTY;
 918                     next = regnext(scan);
 919                     scan = NEXTOPER(scan);
 920                     goto do_curly;
 921                 }
 922                 is_inf = is_inf_internal = 1;
 923                 scan = regnext(scan);
 924                 if (flags & SCF_DO_SUBSTR) {
 925                     scan_commit(pRExC_state, data); /* Cannot extend fixed substrings */
 926                     data->longest = &(data->longest_float);
 927                 }
 928                 goto optimize_curly_tail;
 929             case CURLY:
 930                 mincount = ARG1(scan);
 931                 maxcount = ARG2(scan);
 932                 next = regnext(scan);
 933                 if (OP(scan) == CURLYX) {
 934                     I32 lp = (data ? *(data->last_closep) : 0);
 935
 936                     scan->flags = ((lp <= U8_MAX) ? lp : U8_MAX);
 937                 }
 938                 scan = NEXTOPER(scan) + EXTRA_STEP_2ARGS;
 939               do_curly:
 940                 if (flags & SCF_DO_SUBSTR) {
 941                     if (mincount == 0) scan_commit(pRExC_state,data); /* Cannot extend fixed substrings */
 942                     pos_before = data->pos_min;
 943                 }
 944                 if (data) {
 945                     fl = data->flags;
 946                     data->flags &= ~(SF_HAS_PAR|SF_IN_PAR|SF_HAS_EVAL);
 947                     if (is_inf)
 948                         data->flags |= SF_IS_INF;
 949                 }
 950                 if (flags & SCF_DO_STCLASS) {
 951                     cl_init(pRExC_state, &this_class);
 952                     oclass = data->start_class;
 953                     data->start_class = &this_class;
 954                     f |= SCF_DO_STCLASS_AND;
 955                     f &= ~SCF_DO_STCLASS_OR;
 956                 }
 957                 /* These are the cases when once a subexpression
 958                    fails at a particular position, it cannot succeed
 959                    even after backtracking at the enclosing scope.
 960
 961                    XXXX what if minimal match and we are at the
 962                         initial run of {n,m}? */
 963                 if ((mincount != maxcount - 1) && (maxcount != REG_INFTY))
 964                     f &= ~SCF_WHILEM_VISITED_POS;
 965
 966                 /* This will finish on WHILEM, setting scan, or on NULL: */
 967                 minnext = study_chunk(pRExC_state, &scan, &deltanext, last, data,
 968                                       mincount == 0
 969                                         ? (f & ~SCF_DO_SUBSTR) : f);
 970
 971                 if (flags & SCF_DO_STCLASS)
 972                     data->start_class = oclass;
 973                 if (mincount == 0 || minnext == 0) {
 974                     if (flags & SCF_DO_STCLASS_OR) {
 975                         cl_or(pRExC_state, data->start_class, &this_class);
 976                     }
 977                     else if (flags & SCF_DO_STCLASS_AND) {
 978                         /* Switch to OR mode: cache the old value of
 979                          * data->start_class */
 980                         StructCopy(data->start_class, &and_with,
 981                                    struct regnode_charclass_class);
 982                         flags &= ~SCF_DO_STCLASS_AND;
 983                         StructCopy(&this_class, data->start_class,
 984                                    struct regnode_charclass_class);
 985                         flags |= SCF_DO_STCLASS_OR;
 986                         data->start_class->flags |= ANYOF_EOS;
 987                     }
 988                 } else {                /* Non-zero len */
 989                     if (flags & SCF_DO_STCLASS_OR) {
 990                         cl_or(pRExC_state, data->start_class, &this_class);
 991                         cl_and(data->start_class, &and_with);
 992                     }
 993                     else if (flags & SCF_DO_STCLASS_AND)
 994                         cl_and(data->start_class, &this_class);
 995                     flags &= ~SCF_DO_STCLASS;
 996                 }
 997                 if (!scan)              /* It was not CURLYX, but CURLY. */
 998                     scan = next;
 999                 if (ckWARN(WARN_REGEXP) && (minnext + deltanext == 0)
1000                     && !(data->flags & (SF_HAS_PAR|SF_IN_PAR))
1001                     && maxcount <= REG_INFTY/3) /* Complement check for big count */
1002                 {
1003                     vWARN(RExC_parse,
1004                           "Quantifier unexpected on zero-length expression");
1005                 }
1006
1007                 min += minnext * mincount;
1008                 is_inf_internal |= ((maxcount == REG_INFTY
1009                                      && (minnext + deltanext) > 0)
1010                                     || deltanext == I32_MAX);
1011                 is_inf |= is_inf_internal;
1012                 delta += (minnext + deltanext) * maxcount - minnext * mincount;
1013
1014                 /* Try powerful optimization CURLYX => CURLYN. */
1015                 if (  OP(oscan) == CURLYX && data
1016                       && data->flags & SF_IN_PAR
1017                       && !(data->flags & SF_HAS_EVAL)
1018                       && !deltanext && minnext == 1 ) {
1019                     /* Try to optimize to CURLYN.  */
1020                     regnode *nxt = NEXTOPER(oscan) + EXTRA_STEP_2ARGS;
1021                     regnode *nxt1 = nxt, *nxt2;
1022
1023                     /* Skip open. */
1024                     nxt = regnext(nxt);
1025                     if (!strchr((char*)PL_simple,OP(nxt))
1026                         && !(PL_regkind[(U8)OP(nxt)] == EXACT
1027                              && STR_LEN(nxt) == 1))
1028                         goto nogo;
1029                     nxt2 = nxt;
1030                     nxt = regnext(nxt);
1031                     if (OP(nxt) != CLOSE)
1032                         goto nogo;
1033                     /* Now we know that nxt2 is the only contents: */
1034                     oscan->flags = ARG(nxt);
1035                     OP(oscan) = CURLYN;
1036                     OP(nxt1) = NOTHING; /* was OPEN. */
1037 #ifdef DEBUGGING
1038                     OP(nxt1 + 1) = OPTIMIZED; /* was count. */
1039                     NEXT_OFF(nxt1+ 1) = 0; /* just for consistancy. */
1040                     NEXT_OFF(nxt2) = 0; /* just for consistancy with CURLY. */
1041                     OP(nxt) = OPTIMIZED;        /* was CLOSE. */
1042                     OP(nxt + 1) = OPTIMIZED; /* was count. */
1043                     NEXT_OFF(nxt+ 1) = 0; /* just for consistancy. */
1044 #endif
1045                 }
1046               nogo:
1047
1048                 /* Try optimization CURLYX => CURLYM. */
1049                 if (  OP(oscan) == CURLYX && data
1050                       && !(data->flags & SF_HAS_PAR)
1051                       && !(data->flags & SF_HAS_EVAL)
1052                       && !deltanext  ) {
1053                     /* XXXX How to optimize if data == 0? */
1054                     /* Optimize to a simpler form.  */
1055                     regnode *nxt = NEXTOPER(oscan) + EXTRA_STEP_2ARGS; /* OPEN */
1056                     regnode *nxt2;
1057
1058                     OP(oscan) = CURLYM;
1059                     while ( (nxt2 = regnext(nxt)) /* skip over embedded stuff*/
1060                             && (OP(nxt2) != WHILEM))
1061                         nxt = nxt2;
1062                     OP(nxt2)  = SUCCEED; /* Whas WHILEM */
1063                     /* Need to optimize away parenths. */
1064                     if (data->flags & SF_IN_PAR) {
1065                         /* Set the parenth number.  */
1066                         regnode *nxt1 = NEXTOPER(oscan) + EXTRA_STEP_2ARGS; /* OPEN*/
1067
1068                         if (OP(nxt) != CLOSE)
1069                             FAIL("Panic opt close");
1070                         oscan->flags = ARG(nxt);
1071                         OP(nxt1) = OPTIMIZED;   /* was OPEN. */
1072                         OP(nxt) = OPTIMIZED;    /* was CLOSE. */
1073 #ifdef DEBUGGING
1074                         OP(nxt1 + 1) = OPTIMIZED; /* was count. */
1075                         OP(nxt + 1) = OPTIMIZED; /* was count. */
1076                         NEXT_OFF(nxt1 + 1) = 0; /* just for consistancy. */
1077                         NEXT_OFF(nxt + 1) = 0; /* just for consistancy. */
1078 #endif
1079 #if 0
1080                         while ( nxt1 && (OP(nxt1) != WHILEM)) {
1081                             regnode *nnxt = regnext(nxt1);
1082
1083                             if (nnxt == nxt) {
1084                                 if (reg_off_by_arg[OP(nxt1)])
1085                                     ARG_SET(nxt1, nxt2 - nxt1);
1086                                 else if (nxt2 - nxt1 < U16_MAX)
1087                                     NEXT_OFF(nxt1) = nxt2 - nxt1;
1088                                 else
1089                                     OP(nxt) = NOTHING;  /* Cannot beautify */
1090                             }
1091                             nxt1 = nnxt;
1092                         }
1093 #endif
1094                         /* Optimize again: */
1095                         study_chunk(pRExC_state, &nxt1, &deltanext, nxt,
1096                                     NULL, 0);
1097                     }
1098                     else
1099                         oscan->flags = 0;
1100                 }
1101                 else if ((OP(oscan) == CURLYX)
1102                          && (flags & SCF_WHILEM_VISITED_POS)
1103                          /* See the comment on a similar expression above.
1104                             However, this time it not a subexpression
1105                             we care about, but the expression itself. */
1106                          && (maxcount == REG_INFTY)
1107                          && data && ++data->whilem_c < 16) {
1108                     /* This stays as CURLYX, we can put the count/of pair. */
1109                     /* Find WHILEM (as in regexec.c) */
1110                     regnode *nxt = oscan + NEXT_OFF(oscan);
1111
1112                     if (OP(PREVOPER(nxt)) == NOTHING) /* LONGJMP */
1113                         nxt += ARG(nxt);
1114                     PREVOPER(nxt)->flags = data->whilem_c
1115                         | (RExC_whilem_seen << 4); /* On WHILEM */
1116                 }
1117                 if (data && fl & (SF_HAS_PAR|SF_IN_PAR))
1118                     pars++;
1119                 if (flags & SCF_DO_SUBSTR) {
1120                     SV *last_str = Nullsv;
1121                     int counted = mincount != 0;
1122
1123                     if (data->last_end > 0 && mincount != 0) { /* Ends with a string. */
1124                         I32 b = pos_before >= data->last_start_min
1125                             ? pos_before : data->last_start_min;
1126                         STRLEN l;
1127                         char *s = SvPV(data->last_found, l);
1128                         I32 old = b - data->last_start_min;
1129
1130                         if (UTF)
1131                             old = utf8_hop((U8*)s, old) - (U8*)s;
1132
1133                         l -= old;
1134                         /* Get the added string: */
1135                         last_str = newSVpvn(s  + old, l);
1136                         if (deltanext == 0 && pos_before == b) {
1137                             /* What was added is a constant string */
1138                             if (mincount > 1) {
1139                                 SvGROW(last_str, (mincount * l) + 1);
1140                                 repeatcpy(SvPVX(last_str) + l,
1141                                           SvPVX(last_str), l, mincount - 1);
1142                                 SvCUR(last_str) *= mincount;
1143                                 /* Add additional parts. */
1144                                 SvCUR_set(data->last_found,
1145                                           SvCUR(data->last_found) - l);
1146                                 sv_catsv(data->last_found, last_str);
1147                                 data->last_end += l * (mincount - 1);
1148                             }
1149                         } else {
1150                             /* start offset must point into the last copy */
1151                             data->last_start_min += minnext * (mincount - 1);
1152                             data->last_start_max += is_inf ? 0 : (maxcount - 1)
1153                                 * (minnext + data->pos_delta);
1154                         }
1155                     }
1156                     /* It is counted once already... */
1157                     data->pos_min += minnext * (mincount - counted);
1158                     data->pos_delta += - counted * deltanext +
1159                         (minnext + deltanext) * maxcount - minnext * mincount;
1160                     if (mincount != maxcount) {
1161                          /* Cannot extend fixed substrings found inside
1162                             the group.  */
1163                         scan_commit(pRExC_state,data);
1164                         if (mincount && last_str) {
1165                             sv_setsv(data->last_found, last_str);
1166                             data->last_end = data->pos_min;
1167                             data->last_start_min =
1168                                 data->pos_min - CHR_SVLEN(last_str);
1169                             data->last_start_max = is_inf
1170                                 ? I32_MAX
1171                                 : data->pos_min + data->pos_delta
1172                                 - CHR_SVLEN(last_str);
1173                         }
1174                         data->longest = &(data->longest_float);
1175                     }
1176                     SvREFCNT_dec(last_str);
1177                 }
1178                 if (data && (fl & SF_HAS_EVAL))
1179                     data->flags |= SF_HAS_EVAL;
1180               optimize_curly_tail:
1181                 if (OP(oscan) != CURLYX) {
1182                     while (PL_regkind[(U8)OP(next = regnext(oscan))] == NOTHING
1183                            && NEXT_OFF(next))
1184                         NEXT_OFF(oscan) += NEXT_OFF(next);
1185                 }
1186                 continue;
1187             default:                    /* REF and CLUMP only? */
1188                 if (flags & SCF_DO_SUBSTR) {
1189                     scan_commit(pRExC_state,data);      /* Cannot expect anything... */
1190                     data->longest = &(data->longest_float);
1191                 }
1192                 is_inf = is_inf_internal = 1;
1193                 if (flags & SCF_DO_STCLASS_OR)
1194                     cl_anything(pRExC_state, data->start_class);
1195                 flags &= ~SCF_DO_STCLASS;
1196                 break;
1197             }
1198         }
1199         else if (strchr((char*)PL_simple,OP(scan)) || PL_regkind[(U8)OP(scan)] == ANYUTF8) {
1200             int value;
1201
1202             if (flags & SCF_DO_SUBSTR) {
1203                 scan_commit(pRExC_state,data);
1204                 data->pos_min++;
1205             }
1206             min++;
1207             if (flags & SCF_DO_STCLASS) {
1208                 data->start_class->flags &= ~ANYOF_EOS; /* No match on empty */
1209
1210                 /* Some of the logic below assumes that switching
1211                    locale on will only add false positives. */
1212                 switch (PL_regkind[(U8)OP(scan)]) {
1213                 case ANYUTF8:
1214                 case SANY:
1215                 case SANYUTF8:
1216                 case ALNUMUTF8:
1217                 case ANYOFUTF8:
1218                 case ALNUMLUTF8:
1219                 case NALNUMUTF8:
1220                 case NALNUMLUTF8:
1221                 case SPACEUTF8:
1222                 case NSPACEUTF8:
1223                 case SPACELUTF8:
1224                 case NSPACELUTF8:
1225                 case DIGITUTF8:
1226                 case NDIGITUTF8:
1227                 default:
1228                   do_default:
1229                     /* Perl_croak(aTHX_ "panic: unexpected simple REx opcode %d", OP(scan)); */
1230                     if (flags & SCF_DO_STCLASS_OR) /* Allow everything */
1231                         cl_anything(pRExC_state, data->start_class);
1232                     break;
1233                 case REG_ANY:
1234                     if (OP(scan) == SANY)
1235                         goto do_default;
1236                     if (flags & SCF_DO_STCLASS_OR) { /* Everything but \n */
1237                         value = (ANYOF_BITMAP_TEST(data->start_class,'\n')
1238                                  || (data->start_class->flags & ANYOF_CLASS));
1239                         cl_anything(pRExC_state, data->start_class);
1240                     }
1241                     if (flags & SCF_DO_STCLASS_AND || !value)
1242                         ANYOF_BITMAP_CLEAR(data->start_class,'\n');
1243                     break;
1244                 case ANYOF:
1245                     if (flags & SCF_DO_STCLASS_AND)
1246                         cl_and(data->start_class,
1247                                (struct regnode_charclass_class*)scan);
1248                     else
1249                         cl_or(pRExC_state, data->start_class,
1250                               (struct regnode_charclass_class*)scan);
1251                     break;
1252                 case ALNUM:
1253                     if (flags & SCF_DO_STCLASS_AND) {
1254                         if (!(data->start_class->flags & ANYOF_LOCALE)) {
1255                             ANYOF_CLASS_CLEAR(data->start_class,ANYOF_NALNUM);
1256                             for (value = 0; value < 256; value++)
1257                                 if (!isALNUM(value))
1258                                     ANYOF_BITMAP_CLEAR(data->start_class, value);
1259                         }
1260                     }
1261                     else {
1262                         if (data->start_class->flags & ANYOF_LOCALE)
1263                             ANYOF_CLASS_SET(data->start_class,ANYOF_ALNUM);
1264                         else {
1265                             for (value = 0; value < 256; value++)
1266                                 if (isALNUM(value))
1267                                     ANYOF_BITMAP_SET(data->start_class, value);
1268                         }
1269                     }
1270                     break;
1271                 case ALNUML:
1272                     if (flags & SCF_DO_STCLASS_AND) {
1273                         if (data->start_class->flags & ANYOF_LOCALE)
1274                             ANYOF_CLASS_CLEAR(data->start_class,ANYOF_NALNUM);
1275                     }
1276                     else {
1277                         ANYOF_CLASS_SET(data->start_class,ANYOF_ALNUM);
1278                         data->start_class->flags |= ANYOF_LOCALE;
1279                     }
1280                     break;
1281                 case NALNUM:
1282                     if (flags & SCF_DO_STCLASS_AND) {
1283                         if (!(data->start_class->flags & ANYOF_LOCALE)) {
1284                             ANYOF_CLASS_CLEAR(data->start_class,ANYOF_ALNUM);
1285                             for (value = 0; value < 256; value++)
1286                                 if (isALNUM(value))
1287                                     ANYOF_BITMAP_CLEAR(data->start_class, value);
1288                         }
1289                     }
1290                     else {
1291                         if (data->start_class->flags & ANYOF_LOCALE)
1292                             ANYOF_CLASS_SET(data->start_class,ANYOF_NALNUM);
1293                         else {
1294                             for (value = 0; value < 256; value++)
1295                                 if (!isALNUM(value))
1296                                     ANYOF_BITMAP_SET(data->start_class, value);
1297                         }
1298                     }
1299                     break;
1300                 case NALNUML:
1301                     if (flags & SCF_DO_STCLASS_AND) {
1302                         if (data->start_class->flags & ANYOF_LOCALE)
1303                             ANYOF_CLASS_CLEAR(data->start_class,ANYOF_ALNUM);
1304                     }
1305                     else {
1306                         data->start_class->flags |= ANYOF_LOCALE;
1307                         ANYOF_CLASS_SET(data->start_class,ANYOF_NALNUM);
1308                     }
1309                     break;
1310                 case SPACE:
1311                     if (flags & SCF_DO_STCLASS_AND) {
1312                         if (!(data->start_class->flags & ANYOF_LOCALE)) {
1313                             ANYOF_CLASS_CLEAR(data->start_class,ANYOF_NSPACE);
1314                             for (value = 0; value < 256; value++)
1315                                 if (!isSPACE(value))
1316                                     ANYOF_BITMAP_CLEAR(data->start_class, value);
1317                         }
1318                     }
1319                     else {
1320                         if (data->start_class->flags & ANYOF_LOCALE)
1321                             ANYOF_CLASS_SET(data->start_class,ANYOF_SPACE);
1322                         else {
1323                             for (value = 0; value < 256; value++)
1324                                 if (isSPACE(value))
1325                                     ANYOF_BITMAP_SET(data->start_class, value);
1326                         }
1327                     }
1328                     break;
1329                 case SPACEL:
1330                     if (flags & SCF_DO_STCLASS_AND) {
1331                         if (data->start_class->flags & ANYOF_LOCALE)
1332                             ANYOF_CLASS_CLEAR(data->start_class,ANYOF_NSPACE);
1333                     }
1334                     else {
1335                         data->start_class->flags |= ANYOF_LOCALE;
1336                         ANYOF_CLASS_SET(data->start_class,ANYOF_SPACE);
1337                     }
1338                     break;
1339                 case NSPACE:
1340                     if (flags & SCF_DO_STCLASS_AND) {
1341                         if (!(data->start_class->flags & ANYOF_LOCALE)) {
1342                             ANYOF_CLASS_CLEAR(data->start_class,ANYOF_SPACE);
1343                             for (value = 0; value < 256; value++)
1344                                 if (isSPACE(value))
1345                                     ANYOF_BITMAP_CLEAR(data->start_class, value);
1346                         }
1347                     }
1348                     else {
1349                         if (data->start_class->flags & ANYOF_LOCALE)
1350                             ANYOF_CLASS_SET(data->start_class,ANYOF_NSPACE);
1351                         else {
1352                             for (value = 0; value < 256; value++)
1353                                 if (!isSPACE(value))
1354                                     ANYOF_BITMAP_SET(data->start_class, value);
1355                         }
1356                     }
1357                     break;
1358                 case NSPACEL:
1359                     if (flags & SCF_DO_STCLASS_AND) {
1360                         if (data->start_class->flags & ANYOF_LOCALE) {
1361                             ANYOF_CLASS_CLEAR(data->start_class,ANYOF_SPACE);
1362                             for (value = 0; value < 256; value++)
1363                                 if (!isSPACE(value))
1364                                     ANYOF_BITMAP_CLEAR(data->start_class, value);
1365                         }
1366                     }
1367                     else {
1368                         data->start_class->flags |= ANYOF_LOCALE;
1369                         ANYOF_CLASS_SET(data->start_class,ANYOF_NSPACE);
1370                     }
1371                     break;
1372                 case DIGIT:
1373                     if (flags & SCF_DO_STCLASS_AND) {
1374                         ANYOF_CLASS_CLEAR(data->start_class,ANYOF_NDIGIT);
1375                         for (value = 0; value < 256; value++)
1376                             if (!isDIGIT(value))
1377                                 ANYOF_BITMAP_CLEAR(data->start_class, value);
1378                     }
1379                     else {
1380                         if (data->start_class->flags & ANYOF_LOCALE)
1381                             ANYOF_CLASS_SET(data->start_class,ANYOF_DIGIT);
1382                         else {
1383                             for (value = 0; value < 256; value++)
1384                                 if (isDIGIT(value))
1385                                     ANYOF_BITMAP_SET(data->start_class, value);
1386                         }
1387                     }
1388                     break;
1389                 case NDIGIT:
1390                     if (flags & SCF_DO_STCLASS_AND) {
1391                         ANYOF_CLASS_CLEAR(data->start_class,ANYOF_DIGIT);
1392                         for (value = 0; value < 256; value++)
1393                             if (isDIGIT(value))
1394                                 ANYOF_BITMAP_CLEAR(data->start_class, value);
1395                     }
1396                     else {
1397                         if (data->start_class->flags & ANYOF_LOCALE)
1398                             ANYOF_CLASS_SET(data->start_class,ANYOF_NDIGIT);
1399                         else {
1400                             for (value = 0; value < 256; value++)
1401                                 if (!isDIGIT(value))
1402                                     ANYOF_BITMAP_SET(data->start_class, value);
1403                         }
1404                     }
1405                     break;
1406                 }
1407                 if (flags & SCF_DO_STCLASS_OR)
1408                     cl_and(data->start_class, &and_with);
1409                 flags &= ~SCF_DO_STCLASS;
1410             }
1411         }
1412         else if (PL_regkind[(U8)OP(scan)] == EOL && flags & SCF_DO_SUBSTR) {
1413             data->flags |= (OP(scan) == MEOL
1414                             ? SF_BEFORE_MEOL
1415                             : SF_BEFORE_SEOL);
1416         }
1417         else if (  PL_regkind[(U8)OP(scan)] == BRANCHJ
1418                  /* Lookbehind, or need to calculate parens/evals/stclass: */
1419                    && (scan->flags || data || (flags & SCF_DO_STCLASS))
1420                    && (OP(scan) == IFMATCH || OP(scan) == UNLESSM)) {
1421             /* Lookahead/lookbehind */
1422             I32 deltanext, minnext, fake = 0;
1423             regnode *nscan;
1424             struct regnode_charclass_class intrnl;
1425             int f = 0;
1426
1427             data_fake.flags = 0;
1428             if (data) {
1429                 data_fake.whilem_c = data->whilem_c;
1430                 data_fake.last_closep = data->last_closep;
1431             }
1432             else
1433                 data_fake.last_closep = &fake;
1434             if ( flags & SCF_DO_STCLASS && !scan->flags
1435                  && OP(scan) == IFMATCH ) { /* Lookahead */
1436                 cl_init(pRExC_state, &intrnl);
1437                 data_fake.start_class = &intrnl;
1438                 f |= SCF_DO_STCLASS_AND;
1439             }
1440             if (flags & SCF_WHILEM_VISITED_POS)
1441                 f |= SCF_WHILEM_VISITED_POS;
1442             next = regnext(scan);
1443             nscan = NEXTOPER(NEXTOPER(scan));
1444             minnext = study_chunk(pRExC_state, &nscan, &deltanext, last, &data_fake, f);
1445             if (scan->flags) {
1446                 if (deltanext) {
1447                     vFAIL("Variable length lookbehind not implemented");
1448                 }
1449                 else if (minnext > U8_MAX) {
1450                     vFAIL2("Lookbehind longer than %"UVuf" not implemented", (UV)U8_MAX);
1451                 }
1452                 scan->flags = minnext;
1453             }
1454             if (data && data_fake.flags & (SF_HAS_PAR|SF_IN_PAR))
1455                 pars++;
1456             if (data && (data_fake.flags & SF_HAS_EVAL))
1457                 data->flags |= SF_HAS_EVAL;
1458             if (data)
1459                 data->whilem_c = data_fake.whilem_c;
1460             if (f & SCF_DO_STCLASS_AND) {
1461                 int was = (data->start_class->flags & ANYOF_EOS);
1462
1463                 cl_and(data->start_class, &intrnl);
1464                 if (was)
1465                     data->start_class->flags |= ANYOF_EOS;
1466             }
1467         }
1468         else if (OP(scan) == OPEN) {
1469             pars++;
1470         }
1471         else if (OP(scan) == CLOSE) {
1472             if (ARG(scan) == is_par) {
1473                 next = regnext(scan);
1474
1475                 if ( next && (OP(next) != WHILEM) && next < last)
1476                     is_par = 0;         /* Disable optimization */
1477             }
1478             if (data)
1479                 *(data->last_closep) = ARG(scan);
1480         }
1481         else if (OP(scan) == EVAL) {
1482                 if (data)
1483                     data->flags |= SF_HAS_EVAL;
1484         }
1485         else if (OP(scan) == LOGICAL && scan->flags == 2) { /* Embedded follows */
1486                 if (flags & SCF_DO_SUBSTR) {
1487                     scan_commit(pRExC_state,data);
1488                     data->longest = &(data->longest_float);
1489                 }
1490                 is_inf = is_inf_internal = 1;
1491                 if (flags & SCF_DO_STCLASS_OR) /* Allow everything */
1492                     cl_anything(pRExC_state, data->start_class);
1493                 flags &= ~SCF_DO_STCLASS;
1494         }
1495         /* Else: zero-length, ignore. */
1496         scan = regnext(scan);
1497     }
1498
1499   finish:
1500     *scanp = scan;
1501     *deltap = is_inf_internal ? I32_MAX : delta;
1502     if (flags & SCF_DO_SUBSTR && is_inf)
1503         data->pos_delta = I32_MAX - data->pos_min;
1504     if (is_par > U8_MAX)
1505         is_par = 0;
1506     if (is_par && pars==1 && data) {
1507         data->flags |= SF_IN_PAR;
1508         data->flags &= ~SF_HAS_PAR;
1509     }
1510     else if (pars && data) {
1511         data->flags |= SF_HAS_PAR;
1512         data->flags &= ~SF_IN_PAR;
1513     }
1514     if (flags & SCF_DO_STCLASS_OR)
1515         cl_and(data->start_class, &and_with);
1516     return min;
1517 }
1518
1519 STATIC I32
1520 S_add_data(pTHX_ RExC_state_t *pRExC_state, I32 n, char *s)
1521 {
1522     if (RExC_rx->data) {
1523         Renewc(RExC_rx->data,
1524                sizeof(*RExC_rx->data) + sizeof(void*) * (RExC_rx->data->count + n - 1),
1525                char, struct reg_data);
1526         Renew(RExC_rx->data->what, RExC_rx->data->count + n, U8);
1527         RExC_rx->data->count += n;
1528     }
1529     else {
1530         Newc(1207, RExC_rx->data, sizeof(*RExC_rx->data) + sizeof(void*) * (n - 1),
1531              char, struct reg_data);
1532         New(1208, RExC_rx->data->what, n, U8);
1533         RExC_rx->data->count = n;
1534     }
1535     Copy(s, RExC_rx->data->what + RExC_rx->data->count - n, n, U8);
1536     return RExC_rx->data->count - n;
1537 }
1538
1539 void
1540 Perl_reginitcolors(pTHX)
1541 {
1542     int i = 0;
1543     char *s = PerlEnv_getenv("PERL_RE_COLORS");
1544
1545     if (s) {
1546         PL_colors[0] = s = savepv(s);
1547         while (++i < 6) {
1548             s = strchr(s, '\t');
1549             if (s) {
1550                 *s = '\0';
1551                 PL_colors[i] = ++s;
1552             }
1553             else
1554                 PL_colors[i] = s = "";
1555         }
1556     } else {
1557         while (i < 6)
1558             PL_colors[i++] = "";
1559     }
1560     PL_colorset = 1;
1561 }
1562
1563
1564 /*
1565  - pregcomp - compile a regular expression into internal code
1566  *
1567  * We can't allocate space until we know how big the compiled form will be,
1568  * but we can't compile it (and thus know how big it is) until we've got a
1569  * place to put the code.  So we cheat:  we compile it twice, once with code
1570  * generation turned off and size counting turned on, and once "for real".
1571  * This also means that we don't allocate space until we are sure that the
1572  * thing really will compile successfully, and we never have to move the
1573  * code and thus invalidate pointers into it.  (Note that it has to be in
1574  * one piece because free() must be able to free it all.) [NB: not true in perl]
1575  *
1576  * Beware that the optimization-preparation code in here knows about some
1577  * of the structure of the compiled regexp.  [I'll say.]
1578  */
1579 regexp *
1580 Perl_pregcomp(pTHX_ char *exp, char *xend, PMOP *pm)
1581 {
1582     register regexp *r;
1583     regnode *scan;
1584     regnode *first;
1585     I32 flags;
1586     I32 minlen = 0;
1587     I32 sawplus = 0;
1588     I32 sawopen = 0;
1589     scan_data_t data;
1590     RExC_state_t RExC_state;
1591     RExC_state_t *pRExC_state = &RExC_state;
1592
1593     if (exp == NULL)
1594         FAIL("NULL regexp argument");
1595
1596     /* XXXX This looks very suspicious... */
1597     if (pm->op_pmdynflags & PMdf_UTF8) {
1598         PL_reg_flags |= RF_utf8;
1599     }
1600     else
1601         PL_reg_flags = 0;
1602
1603     RExC_precomp = savepvn(exp, xend - exp);
1604     DEBUG_r(if (!PL_colorset) reginitcolors());
1605     DEBUG_r(PerlIO_printf(Perl_debug_log, "%sCompiling REx%s `%s%*s%s'\n",
1606                       PL_colors[4],PL_colors[5],PL_colors[0],
1607                       (int)(xend - exp), RExC_precomp, PL_colors[1]));
1608     RExC_flags16 = pm->op_pmflags;
1609     RExC_sawback = 0;
1610
1611     RExC_seen = 0;
1612     RExC_seen_zerolen = *exp == '^' ? -1 : 0;
1613     RExC_seen_evals = 0;
1614     RExC_extralen = 0;
1615
1616     /* First pass: determine size, legality. */
1617     RExC_parse = exp;
1618     RExC_end = xend;
1619     RExC_naughty = 0;
1620     RExC_npar = 1;
1621     RExC_size = 0L;
1622     RExC_emit = &PL_regdummy;
1623     RExC_whilem_seen = 0;
1624 #if 0 /* REGC() is (currently) a NOP at the first pass.
1625        * Clever compilers notice this and complain. --jhi */
1626     REGC((U8)REG_MAGIC, (char*)RExC_emit);
1627 #endif
1628     if (reg(pRExC_state, 0, &flags) == NULL) {
1629         Safefree(RExC_precomp);
1630         RExC_precomp = Nullch;
1631         return(NULL);
1632     }
1633     DEBUG_r(PerlIO_printf(Perl_debug_log, "size %"IVdf" ", (IV)RExC_size));
1634
1635     /* Small enough for pointer-storage convention?
1636        If extralen==0, this means that we will not need long jumps. */
1637     if (RExC_size >= 0x10000L && RExC_extralen)
1638         RExC_size += RExC_extralen;
1639     else
1640         RExC_extralen = 0;
1641     if (RExC_whilem_seen > 15)
1642         RExC_whilem_seen = 15;
1643
1644     /* Allocate space and initialize. */
1645     Newc(1001, r, sizeof(regexp) + (unsigned)RExC_size * sizeof(regnode),
1646          char, regexp);
1647     if (r == NULL)
1648         FAIL("Regexp out of space");
1649
1650 #ifdef DEBUGGING
1651     /* avoid reading uninitialized memory in DEBUGGING code in study_chunk() */
1652     Zero(r, sizeof(regexp) + (unsigned)RExC_size * sizeof(regnode), char);
1653 #endif
1654     r->refcnt = 1;
1655     r->prelen = xend - exp;
1656     r->precomp = RExC_precomp;
1657     r->subbeg = NULL;
1658     r->reganch = pm->op_pmflags & PMf_COMPILETIME;
1659     r->nparens = RExC_npar - 1; /* set early to validate backrefs */
1660
1661     r->substrs = 0;                     /* Useful during FAIL. */
1662     r->startp = 0;                      /* Useful during FAIL. */
1663     r->endp = 0;                        /* Useful during FAIL. */
1664
1665     RExC_rx = r;
1666
1667     /* Second pass: emit code. */
1668     RExC_parse = exp;
1669     RExC_end = xend;
1670     RExC_naughty = 0;
1671     RExC_npar = 1;
1672     RExC_emit = r->program;
1673     /* Store the count of eval-groups for security checks: */
1674     RExC_emit->next_off = ((RExC_seen_evals > U16_MAX) ? U16_MAX : RExC_seen_evals);
1675     REGC((U8)REG_MAGIC, (char*) RExC_emit++);
1676     r->data = 0;
1677     if (reg(pRExC_state, 0, &flags) == NULL)
1678         return(NULL);
1679
1680     /* Dig out information for optimizations. */
1681     r->reganch = pm->op_pmflags & PMf_COMPILETIME; /* Again? */
1682     pm->op_pmflags = RExC_flags16;
1683     if (UTF)
1684         r->reganch |= ROPT_UTF8;
1685     r->regstclass = NULL;
1686     if (RExC_naughty >= 10)     /* Probably an expensive pattern. */
1687         r->reganch |= ROPT_NAUGHTY;
1688     scan = r->program + 1;              /* First BRANCH. */
1689
1690     /* XXXX To minimize changes to RE engine we always allocate
1691        3-units-long substrs field. */
1692     Newz(1004, r->substrs, 1, struct reg_substr_data);
1693
1694     StructCopy(&zero_scan_data, &data, scan_data_t);
1695     /* XXXX Should not we check for something else?  Usually it is OPEN1... */
1696     if (OP(scan) != BRANCH) {   /* Only one top-level choice. */
1697         I32 fake;
1698         STRLEN longest_float_length, longest_fixed_length;
1699         struct regnode_charclass_class ch_class;
1700         int stclass_flag;
1701         I32 last_close = 0;
1702
1703         first = scan;
1704         /* Skip introductions and multiplicators >= 1. */
1705         while ((OP(first) == OPEN && (sawopen = 1)) ||
1706                /* An OR of *one* alternative - should not happen now. */
1707             (OP(first) == BRANCH && OP(regnext(first)) != BRANCH) ||
1708             (OP(first) == PLUS) ||
1709             (OP(first) == MINMOD) ||
1710                /* An {n,m} with n>0 */
1711             (PL_regkind[(U8)OP(first)] == CURLY && ARG1(first) > 0) ) {
1712                 if (OP(first) == PLUS)
1713                     sawplus = 1;
1714                 else
1715                     first += regarglen[(U8)OP(first)];
1716                 first = NEXTOPER(first);
1717         }
1718
1719         /* Starting-point info. */
1720       again:
1721         if (PL_regkind[(U8)OP(first)] == EXACT) {
1722             if (OP(first) == EXACT);    /* Empty, get anchored substr later. */
1723             else if ((OP(first) == EXACTF || OP(first) == EXACTFL)
1724                      && !UTF)
1725                 r->regstclass = first;
1726         }
1727         else if (strchr((char*)PL_simple,OP(first)))
1728             r->regstclass = first;
1729         else if (PL_regkind[(U8)OP(first)] == BOUND ||
1730                  PL_regkind[(U8)OP(first)] == NBOUND)
1731             r->regstclass = first;
1732         else if (PL_regkind[(U8)OP(first)] == BOL) {
1733             r->reganch |= (OP(first) == MBOL
1734                            ? ROPT_ANCH_MBOL
1735                            : (OP(first) == SBOL
1736                               ? ROPT_ANCH_SBOL
1737                               : ROPT_ANCH_BOL));
1738             first = NEXTOPER(first);
1739             goto again;
1740         }
1741         else if (OP(first) == GPOS) {
1742             r->reganch |= ROPT_ANCH_GPOS;
1743             first = NEXTOPER(first);
1744             goto again;
1745         }
1746         else if ((OP(first) == STAR &&
1747             PL_regkind[(U8)OP(NEXTOPER(first))] == REG_ANY) &&
1748             !(r->reganch & ROPT_ANCH) )
1749         {
1750             /* turn .* into ^.* with an implied $*=1 */
1751             int type = OP(NEXTOPER(first));
1752
1753             if (type == REG_ANY || type == ANYUTF8)
1754                 type = ROPT_ANCH_MBOL;
1755             else
1756                 type = ROPT_ANCH_SBOL;
1757
1758             r->reganch |= type | ROPT_IMPLICIT;
1759             first = NEXTOPER(first);
1760             goto again;
1761         }
1762         if (sawplus && (!sawopen || !RExC_sawback)
1763             && !(RExC_seen & REG_SEEN_EVAL)) /* May examine pos and $& */
1764             /* x+ must match at the 1st pos of run of x's */
1765             r->reganch |= ROPT_SKIP;
1766
1767         /* Scan is after the zeroth branch, first is atomic matcher. */
1768         DEBUG_r(PerlIO_printf(Perl_debug_log, "first at %"IVdf"\n",
1769                               (IV)(first - scan + 1)));
1770         /*
1771         * If there's something expensive in the r.e., find the
1772         * longest literal string that must appear and make it the
1773         * regmust.  Resolve ties in favor of later strings, since
1774         * the regstart check works with the beginning of the r.e.
1775         * and avoiding duplication strengthens checking.  Not a
1776         * strong reason, but sufficient in the absence of others.
1777         * [Now we resolve ties in favor of the earlier string if
1778         * it happens that c_offset_min has been invalidated, since the
1779         * earlier string may buy us something the later one won't.]
1780         */
1781         minlen = 0;
1782
1783         data.longest_fixed = newSVpvn("",0);
1784         data.longest_float = newSVpvn("",0);
1785         data.last_found = newSVpvn("",0);
1786         data.longest = &(data.longest_fixed);
1787         first = scan;
1788         if (!r->regstclass) {
1789             cl_init(pRExC_state, &ch_class);
1790             data.start_class = &ch_class;
1791             stclass_flag = SCF_DO_STCLASS_AND;
1792         } else                          /* XXXX Check for BOUND? */
1793             stclass_flag = 0;
1794         data.last_closep = &last_close;
1795
1796         minlen = study_chunk(pRExC_state, &first, &fake, scan + RExC_size, /* Up to end */
1797                              &data, SCF_DO_SUBSTR | SCF_WHILEM_VISITED_POS | stclass_flag);
1798         if ( RExC_npar == 1 && data.longest == &(data.longest_fixed)
1799              && data.last_start_min == 0 && data.last_end > 0
1800              && !RExC_seen_zerolen
1801              && (!(RExC_seen & REG_SEEN_GPOS) || (r->reganch & ROPT_ANCH_GPOS)))
1802             r->reganch |= ROPT_CHECK_ALL;
1803         scan_commit(pRExC_state, &data);
1804         SvREFCNT_dec(data.last_found);
1805
1806         longest_float_length = CHR_SVLEN(data.longest_float);
1807         if (longest_float_length
1808             || (data.flags & SF_FL_BEFORE_EOL
1809                 && (!(data.flags & SF_FL_BEFORE_MEOL)
1810                     || (RExC_flags16 & PMf_MULTILINE)))) {
1811             int t;
1812
1813             if (SvCUR(data.longest_fixed)                       /* ok to leave SvCUR */
1814                 && data.offset_fixed == data.offset_float_min
1815                 && SvCUR(data.longest_fixed) == SvCUR(data.longest_float))
1816                     goto remove_float;          /* As in (a)+. */
1817
1818             r->float_substr = data.longest_float;
1819             r->float_min_offset = data.offset_float_min;
1820             r->float_max_offset = data.offset_float_max;
1821             t = (data.flags & SF_FL_BEFORE_EOL /* Can't have SEOL and MULTI */
1822                        && (!(data.flags & SF_FL_BEFORE_MEOL)
1823                            || (RExC_flags16 & PMf_MULTILINE)));
1824             fbm_compile(r->float_substr, t ? FBMcf_TAIL : 0);
1825         }
1826         else {
1827           remove_float:
1828             r->float_substr = Nullsv;
1829             SvREFCNT_dec(data.longest_float);
1830             longest_float_length = 0;
1831         }
1832
1833         longest_fixed_length = CHR_SVLEN(data.longest_fixed);
1834         if (longest_fixed_length
1835             || (data.flags & SF_FIX_BEFORE_EOL /* Cannot have SEOL and MULTI */
1836                 && (!(data.flags & SF_FIX_BEFORE_MEOL)
1837                     || (RExC_flags16 & PMf_MULTILINE)))) {
1838             int t;
1839
1840             r->anchored_substr = data.longest_fixed;
1841             r->anchored_offset = data.offset_fixed;
1842             t = (data.flags & SF_FIX_BEFORE_EOL /* Can't have SEOL and MULTI */
1843                  && (!(data.flags & SF_FIX_BEFORE_MEOL)
1844                      || (RExC_flags16 & PMf_MULTILINE)));
1845             fbm_compile(r->anchored_substr, t ? FBMcf_TAIL : 0);
1846         }
1847         else {
1848             r->anchored_substr = Nullsv;
1849             SvREFCNT_dec(data.longest_fixed);
1850             longest_fixed_length = 0;
1851         }
1852         if (r->regstclass
1853             && (OP(r->regstclass) == REG_ANY || OP(r->regstclass) == ANYUTF8
1854                 || OP(r->regstclass) == SANYUTF8 || OP(r->regstclass) == SANY))
1855             r->regstclass = NULL;
1856         if ((!r->anchored_substr || r->anchored_offset) && stclass_flag
1857             && !(data.start_class->flags & ANYOF_EOS)
1858             && !cl_is_anything(data.start_class)) {
1859             SV *sv;
1860             I32 n = add_data(pRExC_state, 1, "f");
1861
1862             New(1006, RExC_rx->data->data[n], 1,
1863                 struct regnode_charclass_class);
1864             StructCopy(data.start_class,
1865                        (struct regnode_charclass_class*)RExC_rx->data->data[n],
1866                        struct regnode_charclass_class);
1867             r->regstclass = (regnode*)RExC_rx->data->data[n];
1868             r->reganch &= ~ROPT_SKIP;   /* Used in find_byclass(). */
1869             DEBUG_r((sv = sv_newmortal(),
1870                      regprop(sv, (regnode*)data.start_class),
1871                      PerlIO_printf(Perl_debug_log, "synthetic stclass `%s'.\n",
1872                                    SvPVX(sv))));
1873         }
1874
1875         /* A temporary algorithm prefers floated substr to fixed one to dig more info. */
1876         if (longest_fixed_length > longest_float_length) {
1877             r->check_substr = r->anchored_substr;
1878             r->check_offset_min = r->check_offset_max = r->anchored_offset;
1879             if (r->reganch & ROPT_ANCH_SINGLE)
1880                 r->reganch |= ROPT_NOSCAN;
1881         }
1882         else {
1883             r->check_substr = r->float_substr;
1884             r->check_offset_min = data.offset_float_min;
1885             r->check_offset_max = data.offset_float_max;
1886         }
1887         /* XXXX Currently intuiting is not compatible with ANCH_GPOS.
1888            This should be changed ASAP!  */
1889         if (r->check_substr && !(r->reganch & ROPT_ANCH_GPOS)) {
1890             r->reganch |= RE_USE_INTUIT;
1891             if (SvTAIL(r->check_substr))
1892                 r->reganch |= RE_INTUIT_TAIL;
1893         }
1894     }
1895     else {
1896         /* Several toplevels. Best we can is to set minlen. */
1897         I32 fake;
1898         struct regnode_charclass_class ch_class;
1899         I32 last_close = 0;
1900
1901         DEBUG_r(PerlIO_printf(Perl_debug_log, "\n"));
1902         scan = r->program + 1;
1903         cl_init(pRExC_state, &ch_class);
1904         data.start_class = &ch_class;
1905         data.last_closep = &last_close;
1906         minlen = study_chunk(pRExC_state, &scan, &fake, scan + RExC_size, &data, SCF_DO_STCLASS_AND|SCF_WHILEM_VISITED_POS);
1907         r->check_substr = r->anchored_substr = r->float_substr = Nullsv;
1908         if (!(data.start_class->flags & ANYOF_EOS)
1909             && !cl_is_anything(data.start_class)) {
1910             SV *sv;
1911             I32 n = add_data(pRExC_state, 1, "f");
1912
1913             New(1006, RExC_rx->data->data[n], 1,
1914                 struct regnode_charclass_class);
1915             StructCopy(data.start_class,
1916                        (struct regnode_charclass_class*)RExC_rx->data->data[n],
1917                        struct regnode_charclass_class);
1918             r->regstclass = (regnode*)RExC_rx->data->data[n];
1919             r->reganch &= ~ROPT_SKIP;   /* Used in find_byclass(). */
1920             DEBUG_r((sv = sv_newmortal(),
1921                      regprop(sv, (regnode*)data.start_class),
1922                      PerlIO_printf(Perl_debug_log, "synthetic stclass `%s'.\n",
1923                                    SvPVX(sv))));
1924         }
1925     }
1926
1927     r->minlen = minlen;
1928     if (RExC_seen & REG_SEEN_GPOS)
1929         r->reganch |= ROPT_GPOS_SEEN;
1930     if (RExC_seen & REG_SEEN_LOOKBEHIND)
1931         r->reganch |= ROPT_LOOKBEHIND_SEEN;
1932     if (RExC_seen & REG_SEEN_EVAL)
1933         r->reganch |= ROPT_EVAL_SEEN;
1934     Newz(1002, r->startp, RExC_npar, I32);
1935     Newz(1002, r->endp, RExC_npar, I32);
1936     PL_regdata = r->data; /* for regprop() ANYOFUTF8 */
1937     DEBUG_r(regdump(r));
1938     return(r);
1939 }
1940
1941 /*
1942  - reg - regular expression, i.e. main body or parenthesized thing
1943  *
1944  * Caller must absorb opening parenthesis.
1945  *
1946  * Combining parenthesis handling with the base level of regular expression
1947  * is a trifle forced, but the need to tie the tails of the branches to what
1948  * follows makes it hard to avoid.
1949  */
1950 STATIC regnode *
1951 S_reg(pTHX_ RExC_state_t *pRExC_state, I32 paren, I32 *flagp)
1952     /* paren: Parenthesized? 0=top, 1=(, inside: changed to letter. */
1953 {
1954     register regnode *ret;              /* Will be the head of the group. */
1955     register regnode *br;
1956     register regnode *lastbr;
1957     register regnode *ender = 0;
1958     register I32 parno = 0;
1959     I32 flags, oregflags = RExC_flags16, have_branch = 0, open = 0;
1960     char *oregcomp_parse = RExC_parse;
1961     char c;
1962
1963     *flagp = 0;                         /* Tentatively. */
1964
1965     /* Make an OPEN node, if parenthesized. */
1966     if (paren) {
1967         if (*RExC_parse == '?') {
1968             U16 posflags = 0, negflags = 0;
1969             U16 *flagsp = &posflags;
1970             int logical = 0;
1971             char *seqstart = RExC_parse;
1972
1973             RExC_parse++;
1974             paren = *RExC_parse++;
1975             ret = NULL;                 /* For look-ahead/behind. */
1976             switch (paren) {
1977             case '<':
1978                 RExC_seen |= REG_SEEN_LOOKBEHIND;
1979                 if (*RExC_parse == '!')
1980                     paren = ',';
1981                 if (*RExC_parse != '=' && *RExC_parse != '!')
1982                     goto unknown;
1983                 RExC_parse++;
1984             case '=':
1985             case '!':
1986                 RExC_seen_zerolen++;
1987             case ':':
1988             case '>':
1989                 break;
1990             case '$':
1991             case '@':
1992                 vFAIL2("Sequence (?%c...) not implemented", (int)paren);
1993                 break;
1994             case '#':
1995                 while (*RExC_parse && *RExC_parse != ')')
1996                     RExC_parse++;
1997                 if (*RExC_parse != ')')
1998                     FAIL("Sequence (?#... not terminated");
1999                 nextchar(pRExC_state);
2000                 *flagp = TRYAGAIN;
2001                 return NULL;
2002             case 'p':
2003                 if (SIZE_ONLY)
2004                     vWARN(RExC_parse, "(?p{}) is deprecated - use (??{})");
2005                 /* FALL THROUGH*/
2006             case '?':
2007                 logical = 1;
2008                 paren = *RExC_parse++;
2009                 /* FALL THROUGH */
2010             case '{':
2011             {
2012                 I32 count = 1, n = 0;
2013                 char c;
2014                 char *s = RExC_parse;
2015                 SV *sv;
2016                 OP_4tree *sop, *rop;
2017
2018                 RExC_seen_zerolen++;
2019                 RExC_seen |= REG_SEEN_EVAL;
2020                 while (count && (c = *RExC_parse)) {
2021                     if (c == '\\' && RExC_parse[1])
2022                         RExC_parse++;
2023                     else if (c == '{')
2024                         count++;
2025                     else if (c == '}')
2026                         count--;
2027                     RExC_parse++;
2028                 }
2029                 if (*RExC_parse != ')')
2030                 {
2031                     RExC_parse = s;
2032                     vFAIL("Sequence (?{...}) not terminated or not {}-balanced");
2033                 }
2034                 if (!SIZE_ONLY) {
2035                     AV *av;
2036
2037                     if (RExC_parse - 1 - s)
2038                         sv = newSVpvn(s, RExC_parse - 1 - s);
2039                     else
2040                         sv = newSVpvn("", 0);
2041
2042                     ENTER;
2043                     Perl_save_re_context(aTHX);
2044                     rop = sv_compile_2op(sv, &sop, "re", &av);
2045                     LEAVE;
2046
2047                     n = add_data(pRExC_state, 3, "nop");
2048                     RExC_rx->data->data[n] = (void*)rop;
2049                     RExC_rx->data->data[n+1] = (void*)sop;
2050                     RExC_rx->data->data[n+2] = (void*)av;
2051                     SvREFCNT_dec(sv);
2052                 }
2053                 else {                                          /* First pass */
2054                     if (PL_reginterp_cnt < ++RExC_seen_evals
2055                         && PL_curcop != &PL_compiling)
2056                         /* No compiled RE interpolated, has runtime
2057                            components ===> unsafe.  */
2058                         FAIL("Eval-group not allowed at runtime, use re 'eval'");
2059                     if (PL_tainted)
2060                         FAIL("Eval-group in insecure regular expression");
2061                 }
2062
2063                 nextchar(pRExC_state);
2064                 if (logical) {
2065                     ret = reg_node(pRExC_state, LOGICAL);
2066                     if (!SIZE_ONLY)
2067                         ret->flags = 2;
2068                     regtail(pRExC_state, ret, reganode(pRExC_state, EVAL, n));
2069                     return ret;
2070                 }
2071                 return reganode(pRExC_state, EVAL, n);
2072             }
2073             case '(':
2074             {
2075                 if (RExC_parse[0] == '?') {
2076                     if (RExC_parse[1] == '=' || RExC_parse[1] == '!'
2077                         || RExC_parse[1] == '<'
2078                         || RExC_parse[1] == '{') { /* Lookahead or eval. */
2079                         I32 flag;
2080
2081                         ret = reg_node(pRExC_state, LOGICAL);
2082                         if (!SIZE_ONLY)
2083                             ret->flags = 1;
2084                         regtail(pRExC_state, ret, reg(pRExC_state, 1, &flag));
2085                         goto insert_if;
2086                     }
2087                 }
2088                 else if (RExC_parse[0] >= '1' && RExC_parse[0] <= '9' ) {
2089                     parno = atoi(RExC_parse++);
2090
2091                     while (isDIGIT(*RExC_parse))
2092                         RExC_parse++;
2093                     ret = reganode(pRExC_state, GROUPP, parno);
2094                     if ((c = *nextchar(pRExC_state)) != ')')
2095                         vFAIL("Switch condition not recognized");
2096                   insert_if:
2097                     regtail(pRExC_state, ret, reganode(pRExC_state, IFTHEN, 0));
2098                     br = regbranch(pRExC_state, &flags, 1);
2099                     if (br == NULL)
2100                         br = reganode(pRExC_state, LONGJMP, 0);
2101                     else
2102                         regtail(pRExC_state, br, reganode(pRExC_state, LONGJMP, 0));
2103                     c = *nextchar(pRExC_state);
2104                     if (flags&HASWIDTH)
2105                         *flagp |= HASWIDTH;
2106                     if (c == '|') {
2107                         lastbr = reganode(pRExC_state, IFTHEN, 0); /* Fake one for optimizer. */
2108                         regbranch(pRExC_state, &flags, 1);
2109                         regtail(pRExC_state, ret, lastbr);
2110                         if (flags&HASWIDTH)
2111                             *flagp |= HASWIDTH;
2112                         c = *nextchar(pRExC_state);
2113                     }
2114                     else
2115                         lastbr = NULL;
2116                     if (c != ')')
2117                         vFAIL("Switch (?(condition)... contains too many branches");
2118                     ender = reg_node(pRExC_state, TAIL);
2119                     regtail(pRExC_state, br, ender);
2120                     if (lastbr) {
2121                         regtail(pRExC_state, lastbr, ender);
2122                         regtail(pRExC_state, NEXTOPER(NEXTOPER(lastbr)), ender);
2123                     }
2124                     else
2125                         regtail(pRExC_state, ret, ender);
2126                     return ret;
2127                 }
2128                 else {
2129                     vFAIL2("Unknown switch condition (?(%.2s", RExC_parse);
2130                 }
2131             }
2132             case 0:
2133                 RExC_parse--; /* for vFAIL to print correctly */
2134                 vFAIL("Sequence (? incomplete");
2135                 break;
2136             default:
2137                 --RExC_parse;
2138               parse_flags:
2139                 while (*RExC_parse && strchr("iogcmsx", *RExC_parse)) {
2140                     if (*RExC_parse != 'o')
2141                         pmflag(flagsp, *RExC_parse);
2142                     ++RExC_parse;
2143                 }
2144                 if (*RExC_parse == '-') {
2145                     flagsp = &negflags;
2146                     ++RExC_parse;
2147                     goto parse_flags;
2148                 }
2149                 RExC_flags16 |= posflags;
2150                 RExC_flags16 &= ~negflags;
2151                 if (*RExC_parse == ':') {
2152                     RExC_parse++;
2153                     paren = ':';
2154                     break;
2155                 }
2156               unknown:
2157                 if (*RExC_parse != ')') {
2158                     RExC_parse++;
2159                     vFAIL3("Sequence (%.*s...) not recognized", RExC_parse-seqstart, seqstart);
2160                 }
2161                 nextchar(pRExC_state);
2162                 *flagp = TRYAGAIN;
2163                 return NULL;
2164             }
2165         }
2166         else {
2167             parno = RExC_npar;
2168             RExC_npar++;
2169             ret = reganode(pRExC_state, OPEN, parno);
2170             open = 1;
2171         }
2172     }
2173     else
2174         ret = NULL;
2175
2176     /* Pick up the branches, linking them together. */
2177     br = regbranch(pRExC_state, &flags, 1);
2178     if (br == NULL)
2179         return(NULL);
2180     if (*RExC_parse == '|') {
2181         if (!SIZE_ONLY && RExC_extralen) {
2182             reginsert(pRExC_state, BRANCHJ, br);
2183         }
2184         else
2185             reginsert(pRExC_state, BRANCH, br);
2186         have_branch = 1;
2187         if (SIZE_ONLY)
2188             RExC_extralen += 1;         /* For BRANCHJ-BRANCH. */
2189     }
2190     else if (paren == ':') {
2191         *flagp |= flags&SIMPLE;
2192     }
2193     if (open) {                         /* Starts with OPEN. */
2194         regtail(pRExC_state, ret, br);          /* OPEN -> first. */
2195     }
2196     else if (paren != '?')              /* Not Conditional */
2197         ret = br;
2198     if (flags&HASWIDTH)
2199         *flagp |= HASWIDTH;
2200     *flagp |= flags&SPSTART;
2201     lastbr = br;
2202     while (*RExC_parse == '|') {
2203         if (!SIZE_ONLY && RExC_extralen) {
2204             ender = reganode(pRExC_state, LONGJMP,0);
2205             regtail(pRExC_state, NEXTOPER(NEXTOPER(lastbr)), ender); /* Append to the previous. */
2206         }
2207         if (SIZE_ONLY)
2208             RExC_extralen += 2;         /* Account for LONGJMP. */
2209         nextchar(pRExC_state);
2210         br = regbranch(pRExC_state, &flags, 0);
2211         if (br == NULL)
2212             return(NULL);
2213         regtail(pRExC_state, lastbr, br);               /* BRANCH -> BRANCH. */
2214         lastbr = br;
2215         if (flags&HASWIDTH)
2216             *flagp |= HASWIDTH;
2217         *flagp |= flags&SPSTART;
2218     }
2219
2220     if (have_branch || paren != ':') {
2221         /* Make a closing node, and hook it on the end. */
2222         switch (paren) {
2223         case ':':
2224             ender = reg_node(pRExC_state, TAIL);
2225             break;
2226         case 1:
2227             ender = reganode(pRExC_state, CLOSE, parno);
2228             break;
2229         case '<':
2230         case ',':
2231         case '=':
2232         case '!':
2233             *flagp &= ~HASWIDTH;
2234             /* FALL THROUGH */
2235         case '>':
2236             ender = reg_node(pRExC_state, SUCCEED);
2237             break;
2238         case 0:
2239             ender = reg_node(pRExC_state, END);
2240             break;
2241         }
2242         regtail(pRExC_state, lastbr, ender);
2243
2244         if (have_branch) {
2245             /* Hook the tails of the branches to the closing node. */
2246             for (br = ret; br != NULL; br = regnext(br)) {
2247                 regoptail(pRExC_state, br, ender);
2248             }
2249         }
2250     }
2251
2252     {
2253         char *p;
2254         static char parens[] = "=!<,>";
2255
2256         if (paren && (p = strchr(parens, paren))) {
2257             int node = ((p - parens) % 2) ? UNLESSM : IFMATCH;
2258             int flag = (p - parens) > 1;
2259
2260             if (paren == '>')
2261                 node = SUSPEND, flag = 0;
2262             reginsert(pRExC_state, node,ret);
2263             ret->flags = flag;
2264             regtail(pRExC_state, ret, reg_node(pRExC_state, TAIL));
2265         }
2266     }
2267
2268     /* Check for proper termination. */
2269     if (paren) {
2270         RExC_flags16 = oregflags;
2271         if (RExC_parse >= RExC_end || *nextchar(pRExC_state) != ')') {
2272             RExC_parse = oregcomp_parse;
2273             vFAIL("Unmatched (");
2274         }
2275     }
2276     else if (!paren && RExC_parse < RExC_end) {
2277         if (*RExC_parse == ')') {
2278             RExC_parse++;
2279             vFAIL("Unmatched )");
2280         }
2281         else
2282             FAIL("Junk on end of regexp");      /* "Can't happen". */
2283         /* NOTREACHED */
2284     }
2285
2286     return(ret);
2287 }
2288
2289 /*
2290  - regbranch - one alternative of an | operator
2291  *
2292  * Implements the concatenation operator.
2293  */
2294 STATIC regnode *
2295 S_regbranch(pTHX_ RExC_state_t *pRExC_state, I32 *flagp, I32 first)
2296 {
2297     register regnode *ret;
2298     register regnode *chain = NULL;
2299     register regnode *latest;
2300     I32 flags = 0, c = 0;
2301
2302     if (first)
2303         ret = NULL;
2304     else {
2305         if (!SIZE_ONLY && RExC_extralen)
2306             ret = reganode(pRExC_state, BRANCHJ,0);
2307         else
2308             ret = reg_node(pRExC_state, BRANCH);
2309     }
2310
2311     if (!first && SIZE_ONLY)
2312         RExC_extralen += 1;                     /* BRANCHJ */
2313
2314     *flagp = WORST;                     /* Tentatively. */
2315
2316     RExC_parse--;
2317     nextchar(pRExC_state);
2318     while (RExC_parse < RExC_end && *RExC_parse != '|' && *RExC_parse != ')') {
2319         flags &= ~TRYAGAIN;
2320         latest = regpiece(pRExC_state, &flags);
2321         if (latest == NULL) {
2322             if (flags & TRYAGAIN)
2323                 continue;
2324             return(NULL);
2325         }
2326         else if (ret == NULL)
2327             ret = latest;
2328         *flagp |= flags&HASWIDTH;
2329         if (chain == NULL)      /* First piece. */
2330             *flagp |= flags&SPSTART;
2331         else {
2332             RExC_naughty++;
2333             regtail(pRExC_state, chain, latest);
2334         }
2335         chain = latest;
2336         c++;
2337     }
2338     if (chain == NULL) {        /* Loop ran zero times. */
2339         chain = reg_node(pRExC_state, NOTHING);
2340         if (ret == NULL)
2341             ret = chain;
2342     }
2343     if (c == 1) {
2344         *flagp |= flags&SIMPLE;
2345     }
2346
2347     return(ret);
2348 }
2349
2350 /*
2351  - regpiece - something followed by possible [*+?]
2352  *
2353  * Note that the branching code sequences used for ? and the general cases
2354  * of * and + are somewhat optimized:  they use the same NOTHING node as
2355  * both the endmarker for their branch list and the body of the last branch.
2356  * It might seem that this node could be dispensed with entirely, but the
2357  * endmarker role is not redundant.
2358  */
2359 STATIC regnode *
2360 S_regpiece(pTHX_ RExC_state_t *pRExC_state, I32 *flagp)
2361 {
2362     register regnode *ret;
2363     register char op;
2364     register char *next;
2365     I32 flags;
2366     char *origparse = RExC_parse;
2367     char *maxpos;
2368     I32 min;
2369     I32 max = REG_INFTY;
2370
2371     ret = regatom(pRExC_state, &flags);
2372     if (ret == NULL) {
2373         if (flags & TRYAGAIN)
2374             *flagp |= TRYAGAIN;
2375         return(NULL);
2376     }
2377
2378     op = *RExC_parse;
2379
2380     if (op == '{' && regcurly(RExC_parse)) {
2381         next = RExC_parse + 1;
2382         maxpos = Nullch;
2383         while (isDIGIT(*next) || *next == ',') {
2384             if (*next == ',') {
2385                 if (maxpos)
2386                     break;
2387                 else
2388                     maxpos = next;
2389             }
2390             next++;
2391         }
2392         if (*next == '}') {             /* got one */
2393             if (!maxpos)
2394                 maxpos = next;
2395             RExC_parse++;
2396             min = atoi(RExC_parse);
2397             if (*maxpos == ',')
2398                 maxpos++;
2399             else
2400                 maxpos = RExC_parse;
2401             max = atoi(maxpos);
2402             if (!max && *maxpos != '0')
2403                 max = REG_INFTY;                /* meaning "infinity" */
2404             else if (max >= REG_INFTY)
2405                 vFAIL2("Quantifier in {,} bigger than %d", REG_INFTY - 1);
2406             RExC_parse = next;
2407             nextchar(pRExC_state);
2408
2409         do_curly:
2410             if ((flags&SIMPLE)) {
2411                 RExC_naughty += 2 + RExC_naughty / 2;
2412                 reginsert(pRExC_state, CURLY, ret);
2413             }
2414             else {
2415                 regnode *w = reg_node(pRExC_state, WHILEM);
2416
2417                 w->flags = 0;
2418                 regtail(pRExC_state, ret, w);
2419                 if (!SIZE_ONLY && RExC_extralen) {
2420                     reginsert(pRExC_state, LONGJMP,ret);
2421                     reginsert(pRExC_state, NOTHING,ret);
2422                     NEXT_OFF(ret) = 3;  /* Go over LONGJMP. */
2423                 }
2424                 reginsert(pRExC_state, CURLYX,ret);
2425                 if (!SIZE_ONLY && RExC_extralen)
2426                     NEXT_OFF(ret) = 3;  /* Go over NOTHING to LONGJMP. */
2427                 regtail(pRExC_state, ret, reg_node(pRExC_state, NOTHING));
2428                 if (SIZE_ONLY)
2429                     RExC_whilem_seen++, RExC_extralen += 3;
2430                 RExC_naughty += 4 + RExC_naughty;       /* compound interest */
2431             }
2432             ret->flags = 0;
2433
2434             if (min > 0)
2435                 *flagp = WORST;
2436             if (max > 0)
2437                 *flagp |= HASWIDTH;
2438             if (max && max < min)
2439                 vFAIL("Can't do {n,m} with n > m");
2440             if (!SIZE_ONLY) {
2441                 ARG1_SET(ret, min);
2442                 ARG2_SET(ret, max);
2443             }
2444
2445             goto nest_check;
2446         }
2447     }
2448
2449     if (!ISMULT1(op)) {
2450         *flagp = flags;
2451         return(ret);
2452     }
2453
2454 #if 0                           /* Now runtime fix should be reliable. */
2455
2456     /* if this is reinstated, don't forget to put this back into perldiag:
2457
2458             =item Regexp *+ operand could be empty at {#} in regex m/%s/
2459
2460            (F) The part of the regexp subject to either the * or + quantifier
2461            could match an empty string. The {#} shows in the regular
2462            expression about where the problem was discovered.
2463
2464     */
2465
2466     if (!(flags&HASWIDTH) && op != '?')
2467       vFAIL("Regexp *+ operand could be empty");
2468 #endif
2469
2470     nextchar(pRExC_state);
2471
2472     *flagp = (op != '+') ? (WORST|SPSTART|HASWIDTH) : (WORST|HASWIDTH);
2473
2474     if (op == '*' && (flags&SIMPLE)) {
2475         reginsert(pRExC_state, STAR, ret);
2476         ret->flags = 0;
2477         RExC_naughty += 4;
2478     }
2479     else if (op == '*') {
2480         min = 0;
2481         goto do_curly;
2482     }
2483     else if (op == '+' && (flags&SIMPLE)) {
2484         reginsert(pRExC_state, PLUS, ret);
2485         ret->flags = 0;
2486         RExC_naughty += 3;
2487     }
2488     else if (op == '+') {
2489         min = 1;
2490         goto do_curly;
2491     }
2492     else if (op == '?') {
2493         min = 0; max = 1;
2494         goto do_curly;
2495     }
2496   nest_check:
2497     if (ckWARN(WARN_REGEXP) && !SIZE_ONLY && !(flags&HASWIDTH) && max > REG_INFTY/3) {
2498         vWARN3(RExC_parse,
2499                "%.*s matches null string many times",
2500                RExC_parse - origparse,
2501                origparse);
2502     }
2503
2504     if (*RExC_parse == '?') {
2505         nextchar(pRExC_state);
2506         reginsert(pRExC_state, MINMOD, ret);
2507         regtail(pRExC_state, ret, ret + NODE_STEP_REGNODE);
2508     }
2509     if (ISMULT2(RExC_parse)) {
2510         RExC_parse++;
2511         vFAIL("Nested quantifiers");
2512     }
2513
2514     return(ret);
2515 }
2516
2517 /*
2518  - regatom - the lowest level
2519  *
2520  * Optimization:  gobbles an entire sequence of ordinary characters so that
2521  * it can turn them into a single node, which is smaller to store and
2522  * faster to run.  Backslashed characters are exceptions, each becoming a
2523  * separate node; the code is simpler that way and it's not worth fixing.
2524  *
2525  * [Yes, it is worth fixing, some scripts can run twice the speed.] */
2526 STATIC regnode *
2527 S_regatom(pTHX_ RExC_state_t *pRExC_state, I32 *flagp)
2528 {
2529     register regnode *ret = 0;
2530     I32 flags;
2531
2532     *flagp = WORST;             /* Tentatively. */
2533
2534 tryagain:
2535     switch (*RExC_parse) {
2536     case '^':
2537         RExC_seen_zerolen++;
2538         nextchar(pRExC_state);
2539         if (RExC_flags16 & PMf_MULTILINE)
2540             ret = reg_node(pRExC_state, MBOL);
2541         else if (RExC_flags16 & PMf_SINGLELINE)
2542             ret = reg_node(pRExC_state, SBOL);
2543         else
2544             ret = reg_node(pRExC_state, BOL);
2545         break;
2546     case '$':
2547         nextchar(pRExC_state);
2548         if (*RExC_parse)
2549             RExC_seen_zerolen++;
2550         if (RExC_flags16 & PMf_MULTILINE)
2551             ret = reg_node(pRExC_state, MEOL);
2552         else if (RExC_flags16 & PMf_SINGLELINE)
2553             ret = reg_node(pRExC_state, SEOL);
2554         else
2555             ret = reg_node(pRExC_state, EOL);
2556         break;
2557     case '.':
2558         nextchar(pRExC_state);
2559         if (UTF) {
2560             if (RExC_flags16 & PMf_SINGLELINE)
2561                 ret = reg_node(pRExC_state, SANYUTF8);
2562             else
2563                 ret = reg_node(pRExC_state, ANYUTF8);
2564             *flagp |= HASWIDTH;
2565         }
2566         else {
2567             if (RExC_flags16 & PMf_SINGLELINE)
2568                 ret = reg_node(pRExC_state, SANY);
2569             else
2570                 ret = reg_node(pRExC_state, REG_ANY);
2571             *flagp |= HASWIDTH|SIMPLE;
2572         }
2573         RExC_naughty++;
2574         break;
2575     case '[':
2576     {
2577         char *oregcomp_parse = ++RExC_parse;
2578         ret = (UTF ? regclassutf8(pRExC_state) : regclass(pRExC_state));
2579         if (*RExC_parse != ']') {
2580             RExC_parse = oregcomp_parse;
2581             vFAIL("Unmatched [");
2582         }
2583         nextchar(pRExC_state);
2584         *flagp |= HASWIDTH|SIMPLE;
2585         break;
2586     }
2587     case '(':
2588         nextchar(pRExC_state);
2589         ret = reg(pRExC_state, 1, &flags);
2590         if (ret == NULL) {
2591                 if (flags & TRYAGAIN) {
2592                     if (RExC_parse == RExC_end) {
2593                          /* Make parent create an empty node if needed. */
2594                         *flagp |= TRYAGAIN;
2595                         return(NULL);
2596                     }
2597                     goto tryagain;
2598                 }
2599                 return(NULL);
2600         }
2601         *flagp |= flags&(HASWIDTH|SPSTART|SIMPLE);
2602         break;
2603     case '|':
2604     case ')':
2605         if (flags & TRYAGAIN) {
2606             *flagp |= TRYAGAIN;
2607             return NULL;
2608         }
2609         vFAIL("Internal urp");
2610                                 /* Supposed to be caught earlier. */
2611         break;
2612     case '{':
2613         if (!regcurly(RExC_parse)) {
2614             RExC_parse++;
2615             goto defchar;
2616         }
2617         /* FALL THROUGH */
2618     case '?':
2619     case '+':
2620     case '*':
2621         RExC_parse++;
2622         vFAIL("Quantifier follows nothing");
2623         break;
2624     case '\\':
2625         switch (*++RExC_parse) {
2626         case 'A':
2627             RExC_seen_zerolen++;
2628             ret = reg_node(pRExC_state, SBOL);
2629             *flagp |= SIMPLE;
2630             nextchar(pRExC_state);
2631             break;
2632         case 'G':
2633             ret = reg_node(pRExC_state, GPOS);
2634             RExC_seen |= REG_SEEN_GPOS;
2635             *flagp |= SIMPLE;
2636             nextchar(pRExC_state);
2637             break;
2638         case 'Z':
2639             ret = reg_node(pRExC_state, SEOL);
2640             *flagp |= SIMPLE;
2641             nextchar(pRExC_state);
2642             break;
2643         case 'z':
2644             ret = reg_node(pRExC_state, EOS);
2645             *flagp |= SIMPLE;
2646             RExC_seen_zerolen++;                /* Do not optimize RE away */
2647             nextchar(pRExC_state);
2648             break;
2649         case 'C':
2650             ret = reg_node(pRExC_state, SANY);
2651             *flagp |= HASWIDTH|SIMPLE;
2652             nextchar(pRExC_state);
2653             break;
2654         case 'X':
2655             ret = reg_node(pRExC_state, CLUMP);
2656             *flagp |= HASWIDTH;
2657             nextchar(pRExC_state);
2658             if (UTF && !PL_utf8_mark)
2659                 is_utf8_mark((U8*)"~");         /* preload table */
2660             break;
2661         case 'w':
2662             ret = reg_node(pRExC_state,
2663                 UTF
2664                     ? (LOC ? ALNUMLUTF8 : ALNUMUTF8)
2665                     : (LOC ? ALNUML     : ALNUM));
2666             *flagp |= HASWIDTH|SIMPLE;
2667             nextchar(pRExC_state);
2668             if (UTF && !PL_utf8_alnum)
2669                 is_utf8_alnum((U8*)"a");        /* preload table */
2670             break;
2671         case 'W':
2672             ret = reg_node(pRExC_state,
2673                 UTF
2674                     ? (LOC ? NALNUMLUTF8 : NALNUMUTF8)
2675                     : (LOC ? NALNUML     : NALNUM));
2676             *flagp |= HASWIDTH|SIMPLE;
2677             nextchar(pRExC_state);
2678             if (UTF && !PL_utf8_alnum)
2679                 is_utf8_alnum((U8*)"a");        /* preload table */
2680             break;
2681         case 'b':
2682             RExC_seen_zerolen++;
2683             RExC_seen |= REG_SEEN_LOOKBEHIND;
2684             ret = reg_node(pRExC_state,
2685                 UTF
2686                     ? (LOC ? BOUNDLUTF8 : BOUNDUTF8)
2687                     : (LOC ? BOUNDL     : BOUND));
2688             *flagp |= SIMPLE;
2689             nextchar(pRExC_state);
2690             if (UTF && !PL_utf8_alnum)
2691                 is_utf8_alnum((U8*)"a");        /* preload table */
2692             break;
2693         case 'B':
2694             RExC_seen_zerolen++;
2695             RExC_seen |= REG_SEEN_LOOKBEHIND;
2696             ret = reg_node(pRExC_state,
2697                 UTF
2698                     ? (LOC ? NBOUNDLUTF8 : NBOUNDUTF8)
2699                     : (LOC ? NBOUNDL     : NBOUND));
2700             *flagp |= SIMPLE;
2701             nextchar(pRExC_state);
2702             if (UTF && !PL_utf8_alnum)
2703                 is_utf8_alnum((U8*)"a");        /* preload table */
2704             break;
2705         case 's':
2706             ret = reg_node(pRExC_state,
2707                 UTF
2708                     ? (LOC ? SPACELUTF8 : SPACEUTF8)
2709                     : (LOC ? SPACEL     : SPACE));
2710             *flagp |= HASWIDTH|SIMPLE;
2711             nextchar(pRExC_state);
2712             if (UTF && !PL_utf8_space)
2713                 is_utf8_space((U8*)" ");        /* preload table */
2714             break;
2715         case 'S':
2716             ret = reg_node(pRExC_state,
2717                 UTF
2718                     ? (LOC ? NSPACELUTF8 : NSPACEUTF8)
2719                     : (LOC ? NSPACEL     : NSPACE));
2720             *flagp |= HASWIDTH|SIMPLE;
2721             nextchar(pRExC_state);
2722             if (UTF && !PL_utf8_space)
2723                 is_utf8_space((U8*)" ");        /* preload table */
2724             break;
2725         case 'd':
2726             ret = reg_node(pRExC_state, UTF ? DIGITUTF8 : DIGIT);
2727             *flagp |= HASWIDTH|SIMPLE;
2728             nextchar(pRExC_state);
2729             if (UTF && !PL_utf8_digit)
2730                 is_utf8_digit((U8*)"1");        /* preload table */
2731             break;
2732         case 'D':
2733             ret = reg_node(pRExC_state, UTF ? NDIGITUTF8 : NDIGIT);
2734             *flagp |= HASWIDTH|SIMPLE;
2735             nextchar(pRExC_state);
2736             if (UTF && !PL_utf8_digit)
2737                 is_utf8_digit((U8*)"1");        /* preload table */
2738             break;
2739         case 'p':
2740         case 'P':
2741             {   /* a lovely hack--pretend we saw [\pX] instead */
2742                 char* oldregxend = RExC_end;
2743
2744                 if (RExC_parse[1] == '{') {
2745                     RExC_end = strchr(RExC_parse, '}');
2746                     if (!RExC_end) {
2747                         RExC_parse += 2;
2748                         RExC_end = oldregxend;
2749                         vFAIL("Missing right brace on \\p{}");
2750                     }
2751                     RExC_end++;
2752                 }
2753                 else
2754                     RExC_end = RExC_parse + 2;
2755                 RExC_parse--;
2756
2757                 ret = regclassutf8(pRExC_state);
2758
2759                 RExC_end = oldregxend;
2760                 RExC_parse--;
2761                 nextchar(pRExC_state);
2762                 *flagp |= HASWIDTH|SIMPLE;
2763             }
2764             break;
2765         case 'n':
2766         case 'r':
2767         case 't':
2768         case 'f':
2769         case 'e':
2770         case 'a':
2771         case 'x':
2772         case 'c':
2773         case '0':
2774             goto defchar;
2775         case '1': case '2': case '3': case '4':
2776         case '5': case '6': case '7': case '8': case '9':
2777             {
2778                 I32 num = atoi(RExC_parse);
2779
2780                 if (num > 9 && num >= RExC_npar)
2781                     goto defchar;
2782                 else {
2783                     while (isDIGIT(*RExC_parse))
2784                         RExC_parse++;
2785
2786                     if (!SIZE_ONLY && num > RExC_rx->nparens)
2787                         vFAIL("Reference to nonexistent group");
2788                     RExC_sawback = 1;
2789                     ret = reganode(pRExC_state, FOLD
2790                                    ? (LOC ? REFFL : REFF)
2791                                    : REF, num);
2792                     *flagp |= HASWIDTH;
2793                     RExC_parse--;
2794                     nextchar(pRExC_state);
2795                 }
2796             }
2797             break;
2798         case '\0':
2799             if (RExC_parse >= RExC_end)
2800                 FAIL("Trailing \\");
2801             /* FALL THROUGH */
2802         default:
2803             /* Do not generate `unrecognized' warnings here, we fall
2804                back into the quick-grab loop below */
2805             goto defchar;
2806         }
2807         break;
2808
2809     case '#':
2810         if (RExC_flags16 & PMf_EXTENDED) {
2811             while (RExC_parse < RExC_end && *RExC_parse != '\n') RExC_parse++;
2812             if (RExC_parse < RExC_end)
2813                 goto tryagain;
2814         }
2815         /* FALL THROUGH */
2816
2817     default: {
2818             register STRLEN len;
2819             register UV ender;
2820             register char *p;
2821             char *oldp, *s;
2822             STRLEN numlen;
2823
2824             RExC_parse++;
2825
2826         defchar:
2827             ret = reg_node(pRExC_state, FOLD
2828                           ? (LOC ? EXACTFL : EXACTF)
2829                           : EXACT);
2830             s = STRING(ret);
2831             for (len = 0, p = RExC_parse - 1;
2832               len < 127 && p < RExC_end;
2833               len++)
2834             {
2835                 oldp = p;
2836
2837                 if (RExC_flags16 & PMf_EXTENDED)
2838                     p = regwhite(p, RExC_end);
2839                 switch (*p) {
2840                 case '^':
2841                 case '$':
2842                 case '.':
2843                 case '[':
2844                 case '(':
2845                 case ')':
2846                 case '|':
2847                     goto loopdone;
2848                 case '\\':
2849                     switch (*++p) {
2850                     case 'A':
2851                     case 'G':
2852                     case 'Z':
2853                     case 'z':
2854                     case 'w':
2855                     case 'W':
2856                     case 'b':
2857                     case 'B':
2858                     case 's':
2859                     case 'S':
2860                     case 'd':
2861                     case 'D':
2862                     case 'p':
2863                     case 'P':
2864                         --p;
2865                         goto loopdone;
2866                     case 'n':
2867                         ender = '\n';
2868                         p++;
2869                         break;
2870                     case 'r':
2871                         ender = '\r';
2872                         p++;
2873                         break;
2874                     case 't':
2875                         ender = '\t';
2876                         p++;
2877                         break;
2878                     case 'f':
2879                         ender = '\f';
2880                         p++;
2881                         break;
2882                     case 'e':
2883 #ifdef ASCIIish
2884                           ender = '\033';
2885 #else
2886                           ender = '\047';
2887 #endif
2888                         p++;
2889                         break;
2890                     case 'a':
2891 #ifdef ASCIIish
2892                           ender = '\007';
2893 #else
2894                           ender = '\057';
2895 #endif
2896                         p++;
2897                         break;
2898                     case 'x':
2899                         if (*++p == '{') {
2900                             char* e = strchr(p, '}');
2901
2902                             if (!e) {
2903                                 RExC_parse = p + 1;
2904                                 vFAIL("Missing right brace on \\x{}");
2905                             }
2906                             else {
2907                                 numlen = 1;     /* allow underscores */
2908                                 ender = (UV)scan_hex(p + 1, e - p - 1, &numlen);
2909                                 /* numlen is generous */
2910                                 if (numlen + len >= 127) {
2911                                     p--;
2912                                     goto loopdone;
2913                                 }
2914                                 p = e + 1;
2915                             }
2916                         }
2917                         else {
2918                             numlen = 0;         /* disallow underscores */
2919                             ender = (UV)scan_hex(p, 2, &numlen);
2920                             p += numlen;
2921                         }
2922                         break;
2923                     case 'c':
2924                         p++;
2925                         ender = UCHARAT(p++);
2926                         ender = toCTRL(ender);
2927                         break;
2928                     case '0': case '1': case '2': case '3':case '4':
2929                     case '5': case '6': case '7': case '8':case '9':
2930                         if (*p == '0' ||
2931                           (isDIGIT(p[1]) && atoi(p) >= RExC_npar) ) {
2932                             numlen = 0;         /* disallow underscores */
2933                             ender = (UV)scan_oct(p, 3, &numlen);
2934                             p += numlen;
2935                         }
2936                         else {
2937                             --p;
2938                             goto loopdone;
2939                         }
2940                         break;
2941                     case '\0':
2942                         if (p >= RExC_end)
2943                             FAIL("Trailing \\");
2944                         /* FALL THROUGH */
2945                     default:
2946                         if (!SIZE_ONLY && ckWARN(WARN_REGEXP) && isALPHA(*p))
2947                             vWARN2(p +1, "Unrecognized escape \\%c passed through", *p);
2948                         goto normal_default;
2949                     }
2950                     break;
2951                 default:
2952                   normal_default:
2953                     if ((*p & 0xc0) == 0xc0 && UTF) {
2954                         ender = utf8_to_uv((U8*)p, RExC_end - p,
2955                                                &numlen, 0);
2956                         p += numlen;
2957                     }
2958                     else
2959                         ender = *p++;
2960                     break;
2961                 }
2962                 if (RExC_flags16 & PMf_EXTENDED)
2963                     p = regwhite(p, RExC_end);
2964                 if (UTF && FOLD) {
2965                     if (LOC)
2966                         ender = toLOWER_LC_uni(ender);
2967                     else
2968                         ender = toLOWER_uni(ender);
2969                 }
2970                 if (ISMULT2(p)) { /* Back off on ?+*. */
2971                     if (len)
2972                         p = oldp;
2973                     else if (ender >= 0x80 && UTF) {
2974                         reguni(pRExC_state, ender, s, &numlen);
2975                         s += numlen;
2976                         len += numlen;
2977                     }
2978                     else {
2979                         len++;
2980                         REGC(ender, s++);
2981                     }
2982                     break;
2983                 }
2984                 if (ender >= 0x80 && UTF) {
2985                     reguni(pRExC_state, ender, s, &numlen);
2986                     s += numlen;
2987                     len += numlen - 1;
2988                 }
2989                 else
2990                     REGC(ender, s++);
2991             }
2992         loopdone:
2993             RExC_parse = p - 1;
2994             nextchar(pRExC_state);
2995             {
2996                 /* len is STRLEN which is unsigned, need to copy to signed */
2997                 IV iv = len;
2998                 if (iv < 0)
2999                     vFAIL("Internal disaster");
3000             }
3001             if (len > 0)
3002                 *flagp |= HASWIDTH;
3003             if (len == 1)
3004                 *flagp |= SIMPLE;
3005             if (!SIZE_ONLY)
3006                 STR_LEN(ret) = len;
3007             if (SIZE_ONLY)
3008                 RExC_size += STR_SZ(len);
3009             else
3010                 RExC_emit += STR_SZ(len);
3011         }
3012         break;
3013     }
3014
3015     return(ret);
3016 }
3017
3018 STATIC char *
3019 S_regwhite(pTHX_ char *p, char *e)
3020 {
3021     while (p < e) {
3022         if (isSPACE(*p))
3023             ++p;
3024         else if (*p == '#') {
3025             do {
3026                 p++;
3027             } while (p < e && *p != '\n');
3028         }
3029         else
3030             break;
3031     }
3032     return p;
3033 }
3034
3035 /* Parse POSIX character classes: [[:foo:]], [[=foo=]], [[.foo.]].
3036    Character classes ([:foo:]) can also be negated ([:^foo:]).
3037    Returns a named class id (ANYOF_XXX) if successful, -1 otherwise.
3038    Equivalence classes ([=foo=]) and composites ([.foo.]) are parsed,
3039    but trigger warnings because they are currently unimplemented. */
3040 STATIC I32
3041 S_regpposixcc(pTHX_ RExC_state_t *pRExC_state, I32 value)
3042 {
3043     char *posixcc = 0;
3044     I32 namedclass = OOB_NAMEDCLASS;
3045
3046     if (value == '[' && RExC_parse + 1 < RExC_end &&
3047         /* I smell either [: or [= or [. -- POSIX has been here, right? */
3048         (*RExC_parse == ':' ||
3049          *RExC_parse == '=' ||
3050          *RExC_parse == '.')) {
3051         char  c = *RExC_parse;
3052         char* s = RExC_parse++;
3053
3054         while (RExC_parse < RExC_end && *RExC_parse != c)
3055             RExC_parse++;
3056         if (RExC_parse == RExC_end)
3057             /* Grandfather lone [:, [=, [. */
3058             RExC_parse = s;
3059         else {
3060             char* t = RExC_parse++; /* skip over the c */
3061
3062             if (*RExC_parse == ']') {
3063                 RExC_parse++; /* skip over the ending ] */
3064                 posixcc = s + 1;
3065                 if (*s == ':') {
3066                     I32 complement = *posixcc == '^' ? *posixcc++ : 0;
3067                     I32 skip = 5; /* the most common skip */
3068
3069                     switch (*posixcc) {
3070                     case 'a':
3071                         if (strnEQ(posixcc, "alnum", 5))
3072                             namedclass =
3073                                 complement ? ANYOF_NALNUMC : ANYOF_ALNUMC;
3074                         else if (strnEQ(posixcc, "alpha", 5))
3075                             namedclass =
3076                                 complement ? ANYOF_NALPHA : ANYOF_ALPHA;
3077                         else if (strnEQ(posixcc, "ascii", 5))
3078                             namedclass =
3079                                 complement ? ANYOF_NASCII : ANYOF_ASCII;
3080                         break;
3081                     case 'b':
3082                         if (strnEQ(posixcc, "blank", 5))
3083                             namedclass =
3084                                 complement ? ANYOF_NBLANK : ANYOF_BLANK;
3085                         break;
3086                     case 'c':
3087                         if (strnEQ(posixcc, "cntrl", 5))
3088                             namedclass =
3089                                 complement ? ANYOF_NCNTRL : ANYOF_CNTRL;
3090                         break;
3091                     case 'd':
3092                         if (strnEQ(posixcc, "digit", 5))
3093                             namedclass =
3094                                 complement ? ANYOF_NDIGIT : ANYOF_DIGIT;
3095                         break;
3096                     case 'g':
3097                         if (strnEQ(posixcc, "graph", 5))
3098                             namedclass =
3099                                 complement ? ANYOF_NGRAPH : ANYOF_GRAPH;
3100                         break;
3101                     case 'l':
3102                         if (strnEQ(posixcc, "lower", 5))
3103                             namedclass =
3104                                 complement ? ANYOF_NLOWER : ANYOF_LOWER;
3105                         break;
3106                     case 'p':
3107                         if (strnEQ(posixcc, "print", 5))
3108                             namedclass =
3109                                 complement ? ANYOF_NPRINT : ANYOF_PRINT;
3110                         else if (strnEQ(posixcc, "punct", 5))
3111                             namedclass =
3112                                 complement ? ANYOF_NPUNCT : ANYOF_PUNCT;
3113                         break;
3114                     case 's':
3115                         if (strnEQ(posixcc, "space", 5))
3116                             namedclass =
3117                                 complement ? ANYOF_NPSXSPC : ANYOF_PSXSPC;
3118                         break;
3119                     case 'u':
3120                         if (strnEQ(posixcc, "upper", 5))
3121                             namedclass =
3122                                 complement ? ANYOF_NUPPER : ANYOF_UPPER;
3123                         break;
3124                     case 'w': /* this is not POSIX, this is the Perl \w */
3125                         if (strnEQ(posixcc, "word", 4)) {
3126                             namedclass =
3127                                 complement ? ANYOF_NALNUM : ANYOF_ALNUM;
3128                             skip = 4;
3129                         }
3130                         break;
3131                     case 'x':
3132                         if (strnEQ(posixcc, "xdigit", 6)) {
3133                             namedclass =
3134                                 complement ? ANYOF_NXDIGIT : ANYOF_XDIGIT;
3135                             skip = 6;
3136                         }
3137                         break;
3138                     }
3139                     if (namedclass == OOB_NAMEDCLASS ||
3140                         posixcc[skip] != ':' ||
3141                         posixcc[skip+1] != ']')
3142                     {
3143                         Simple_vFAIL3("POSIX class [:%.*s:] unknown",
3144                                       t - s - 1, s + 1);
3145                     }
3146                 } else if (!SIZE_ONLY) {
3147                     /* [[=foo=]] and [[.foo.]] are still future. */
3148
3149                     /* adjust RExC_parse so the warning shows after
3150                        the class closes */
3151                     while (*RExC_parse && *RExC_parse != ']')
3152                         RExC_parse++;
3153                     Simple_vFAIL3("POSIX syntax [%c %c] is reserved for future extensions", c, c);
3154                 }
3155             } else {
3156                 /* Maternal grandfather:
3157                  * "[:" ending in ":" but not in ":]" */
3158                 RExC_parse = s;
3159             }
3160         }
3161     }
3162
3163     return namedclass;
3164 }
3165
3166 STATIC void
3167 S_checkposixcc(pTHX_ RExC_state_t *pRExC_state)
3168 {
3169     if (!SIZE_ONLY && ckWARN(WARN_REGEXP) &&
3170         (*RExC_parse == ':' ||
3171          *RExC_parse == '=' ||
3172          *RExC_parse == '.')) {
3173         char *s = RExC_parse;
3174         char  c = *s++;
3175
3176         while(*s && isALNUM(*s))
3177             s++;
3178         if (*s && c == *s && s[1] == ']') {
3179             vWARN3(s+2, "POSIX syntax [%c %c] belongs inside character classes", c, c);
3180
3181             /* [[=foo=]] and [[.foo.]] are still future. */
3182             if (c == '=' || c == '.')
3183             {
3184                 /* adjust RExC_parse so the error shows after
3185                    the class closes */
3186                 while (*RExC_parse && *RExC_parse++ != ']')
3187                     ;
3188                 Simple_vFAIL3("POSIX syntax [%c %c] is reserved for future extensions", c, c);
3189             }
3190         }
3191     }
3192 }
3193
3194 STATIC regnode *
3195 S_regclass(pTHX_ RExC_state_t *pRExC_state)
3196 {
3197     register U32 value;
3198     register I32 lastvalue = OOB_CHAR8;
3199     register I32 range = 0;
3200     register regnode *ret;
3201     STRLEN numlen;
3202     I32 namedclass;
3203     char *rangebegin;
3204     bool need_class = 0;
3205
3206     ret = reg_node(pRExC_state, ANYOF);
3207     if (SIZE_ONLY)
3208         RExC_size += ANYOF_SKIP;
3209     else {
3210         ret->flags = 0;
3211         ANYOF_BITMAP_ZERO(ret);
3212         RExC_emit += ANYOF_SKIP;
3213         if (FOLD)
3214             ANYOF_FLAGS(ret) |= ANYOF_FOLD;
3215         if (LOC)
3216             ANYOF_FLAGS(ret) |= ANYOF_LOCALE;
3217     }
3218     if (*RExC_parse == '^') {   /* Complement of range. */
3219         RExC_naughty++;
3220         RExC_parse++;
3221         if (!SIZE_ONLY)
3222             ANYOF_FLAGS(ret) |= ANYOF_INVERT;
3223     }
3224
3225     if (!SIZE_ONLY && ckWARN(WARN_REGEXP))
3226         checkposixcc(pRExC_state);
3227
3228     if (*RExC_parse == ']' || *RExC_parse == '-')
3229         goto skipcond;          /* allow 1st char to be ] or - */
3230     while (RExC_parse < RExC_end && *RExC_parse != ']') {
3231        skipcond:
3232         namedclass = OOB_NAMEDCLASS;
3233         if (!range)
3234             rangebegin = RExC_parse;
3235         value = UCHARAT(RExC_parse++);
3236         if (value == '[')
3237             namedclass = regpposixcc(pRExC_state, value);
3238         else if (value == '\\') {
3239             value = UCHARAT(RExC_parse++);
3240             /* Some compilers cannot handle switching on 64-bit integer
3241              * values, therefore the 'value' cannot be an UV. --jhi */
3242             switch (value) {
3243             case 'w':   namedclass = ANYOF_ALNUM;       break;
3244             case 'W':   namedclass = ANYOF_NALNUM;      break;
3245             case 's':   namedclass = ANYOF_SPACE;       break;
3246             case 'S':   namedclass = ANYOF_NSPACE;      break;
3247             case 'd':   namedclass = ANYOF_DIGIT;       break;
3248             case 'D':   namedclass = ANYOF_NDIGIT;      break;
3249             case 'n':   value = '\n';                   break;
3250             case 'r':   value = '\r';                   break;
3251             case 't':   value = '\t';                   break;
3252             case 'f':   value = '\f';                   break;
3253             case 'b':   value = '\b';                   break;
3254 #ifdef ASCIIish
3255             case 'e':   value = '\033';                 break;
3256             case 'a':   value = '\007';                 break;
3257 #else
3258             case 'e':   value = '\047';                 break;
3259             case 'a':   value = '\057';                 break;
3260 #endif
3261             case 'x':
3262                 numlen = 0;             /* disallow underscores */
3263                 value = (UV)scan_hex(RExC_parse, 2, &numlen);
3264                 RExC_parse += numlen;
3265                 break;
3266             case 'c':
3267                 value = UCHARAT(RExC_parse++);
3268                 value = toCTRL(value);
3269                 break;
3270             case '0': case '1': case '2': case '3': case '4':
3271             case '5': case '6': case '7': case '8': case '9':
3272                 numlen = 0;             /* disallow underscores */
3273                 value = (UV)scan_oct(--RExC_parse, 3, &numlen);
3274                 RExC_parse += numlen;
3275                 break;
3276             default:
3277                 if (!SIZE_ONLY && ckWARN(WARN_REGEXP) && isALPHA(value))
3278
3279                     vWARN2(RExC_parse, "Unrecognized escape \\%c in character class passed through", (int)value);
3280                 break;
3281             }
3282         }
3283         if (namedclass > OOB_NAMEDCLASS) {
3284             if (!need_class && !SIZE_ONLY)
3285                 ANYOF_CLASS_ZERO(ret);
3286             need_class = 1;
3287             if (range) { /* a-\d, a-[:digit:] */
3288                 if (!SIZE_ONLY) {
3289                     if (ckWARN(WARN_REGEXP))
3290                         vWARN4(RExC_parse,
3291                                "False [] range \"%*.*s\"",
3292                                RExC_parse - rangebegin,
3293                                RExC_parse - rangebegin,
3294                                rangebegin);
3295                     ANYOF_BITMAP_SET(ret, lastvalue);
3296                     ANYOF_BITMAP_SET(ret, '-');
3297                 }
3298                 range = 0; /* this is not a true range */
3299             }
3300             if (!SIZE_ONLY) {
3301                 switch (namedclass) {
3302                 case ANYOF_ALNUM:
3303                     if (LOC)
3304                         ANYOF_CLASS_SET(ret, ANYOF_ALNUM);
3305                     else {
3306                         for (value = 0; value < 256; value++)
3307                             if (isALNUM(value))
3308                                 ANYOF_BITMAP_SET(ret, value);
3309                     }
3310                     break;
3311                 case ANYOF_NALNUM:
3312                     if (LOC)
3313                         ANYOF_CLASS_SET(ret, ANYOF_NALNUM);
3314                     else {
3315                         for (value = 0; value < 256; value++)
3316                             if (!isALNUM(value))
3317                                 ANYOF_BITMAP_SET(ret, value);
3318                     }
3319                     break;
3320                 case ANYOF_SPACE:
3321                     if (LOC)
3322                         ANYOF_CLASS_SET(ret, ANYOF_SPACE);
3323                     else {
3324                         for (value = 0; value < 256; value++)
3325                             if (isSPACE(value))
3326                                 ANYOF_BITMAP_SET(ret, value);
3327                     }
3328                     break;
3329                 case ANYOF_NSPACE:
3330                     if (LOC)
3331                         ANYOF_CLASS_SET(ret, ANYOF_NSPACE);
3332                     else {
3333                         for (value = 0; value < 256; value++)
3334                             if (!isSPACE(value))
3335                                 ANYOF_BITMAP_SET(ret, value);
3336                     }
3337                     break;
3338                 case ANYOF_DIGIT:
3339                     if (LOC)
3340                         ANYOF_CLASS_SET(ret, ANYOF_DIGIT);
3341                     else {
3342                         for (value = '0'; value <= '9'; value++)
3343                             ANYOF_BITMAP_SET(ret, value);
3344                     }
3345                     break;
3346                 case ANYOF_NDIGIT:
3347                     if (LOC)
3348                         ANYOF_CLASS_SET(ret, ANYOF_NDIGIT);
3349                     else {
3350                         for (value = 0; value < '0'; value++)
3351                             ANYOF_BITMAP_SET(ret, value);
3352                         for (value = '9' + 1; value < 256; value++)
3353                             ANYOF_BITMAP_SET(ret, value);
3354                     }
3355                     break;
3356                 case ANYOF_NALNUMC:
3357                     if (LOC)
3358                         ANYOF_CLASS_SET(ret, ANYOF_NALNUMC);
3359                     else {
3360                         for (value = 0; value < 256; value++)
3361                             if (!isALNUMC(value))
3362                                 ANYOF_BITMAP_SET(ret, value);
3363                     }
3364                     break;
3365                 case ANYOF_ALNUMC:
3366                     if (LOC)
3367                         ANYOF_CLASS_SET(ret, ANYOF_ALNUMC);
3368                     else {
3369                         for (value = 0; value < 256; value++)
3370                             if (isALNUMC(value))
3371                                 ANYOF_BITMAP_SET(ret, value);
3372                     }
3373                     break;
3374                 case ANYOF_ALPHA:
3375                     if (LOC)
3376                         ANYOF_CLASS_SET(ret, ANYOF_ALPHA);
3377                     else {
3378                         for (value = 0; value < 256; value++)
3379                             if (isALPHA(value))
3380                                 ANYOF_BITMAP_SET(ret, value);
3381                     }
3382                     break;
3383                 case ANYOF_NALPHA:
3384                     if (LOC)
3385                         ANYOF_CLASS_SET(ret, ANYOF_NALPHA);
3386                     else {
3387                         for (value = 0; value < 256; value++)
3388                             if (!isALPHA(value))
3389                                 ANYOF_BITMAP_SET(ret, value);
3390                     }
3391                     break;
3392                 case ANYOF_ASCII:
3393                     if (LOC)
3394                         ANYOF_CLASS_SET(ret, ANYOF_ASCII);
3395                     else {
3396 #ifdef ASCIIish
3397                         for (value = 0; value < 128; value++)
3398                             ANYOF_BITMAP_SET(ret, value);
3399 #else  /* EBCDIC */
3400                         for (value = 0; value < 256; value++)
3401                             if (isASCII(value))
3402                                 ANYOF_BITMAP_SET(ret, value);
3403 #endif /* EBCDIC */
3404                     }
3405                     break;
3406                 case ANYOF_NASCII:
3407                     if (LOC)
3408                         ANYOF_CLASS_SET(ret, ANYOF_NASCII);
3409                     else {
3410 #ifdef ASCIIish
3411                         for (value = 128; value < 256; value++)
3412                             ANYOF_BITMAP_SET(ret, value);
3413 #else  /* EBCDIC */
3414                         for (value = 0; value < 256; value++)
3415                             if (!isASCII(value))
3416                                 ANYOF_BITMAP_SET(ret, value);
3417 #endif /* EBCDIC */
3418                     }
3419                     break;
3420                 case ANYOF_BLANK:
3421                     if (LOC)
3422                         ANYOF_CLASS_SET(ret, ANYOF_BLANK);
3423                     else {
3424                         for (value = 0; value < 256; value++)
3425                             if (isBLANK(value))
3426                                 ANYOF_BITMAP_SET(ret, value);
3427                     }
3428                     break;
3429                 case ANYOF_NBLANK:
3430                     if (LOC)
3431                         ANYOF_CLASS_SET(ret, ANYOF_NBLANK);
3432                     else {
3433                         for (value = 0; value < 256; value++)
3434                             if (!isBLANK(value))
3435                                 ANYOF_BITMAP_SET(ret, value);
3436                     }
3437                     break;
3438                 case ANYOF_CNTRL:
3439                     if (LOC)
3440                         ANYOF_CLASS_SET(ret, ANYOF_CNTRL);
3441                     else {
3442                         for (value = 0; value < 256; value++)
3443                             if (isCNTRL(value))
3444                                 ANYOF_BITMAP_SET(ret, value);
3445                     }
3446                     lastvalue = OOB_CHAR8;
3447                     break;
3448                 case ANYOF_NCNTRL:
3449                     if (LOC)
3450                         ANYOF_CLASS_SET(ret, ANYOF_NCNTRL);
3451                     else {
3452                         for (value = 0; value < 256; value++)
3453                             if (!isCNTRL(value))
3454                                 ANYOF_BITMAP_SET(ret, value);
3455                     }
3456                     break;
3457                 case ANYOF_GRAPH:
3458                     if (LOC)
3459                         ANYOF_CLASS_SET(ret, ANYOF_GRAPH);
3460                     else {
3461                         for (value = 0; value < 256; value++)
3462                             if (isGRAPH(value))
3463                                 ANYOF_BITMAP_SET(ret, value);
3464                     }
3465                     break;
3466                 case ANYOF_NGRAPH:
3467                     if (LOC)
3468                         ANYOF_CLASS_SET(ret, ANYOF_NGRAPH);
3469                     else {
3470                         for (value = 0; value < 256; value++)
3471                             if (!isGRAPH(value))
3472                                 ANYOF_BITMAP_SET(ret, value);
3473                     }
3474                     break;
3475                 case ANYOF_LOWER:
3476                     if (LOC)
3477                         ANYOF_CLASS_SET(ret, ANYOF_LOWER);
3478                     else {
3479                         for (value = 0; value < 256; value++)
3480                             if (isLOWER(value))
3481                                 ANYOF_BITMAP_SET(ret, value);
3482                     }
3483                     break;
3484                 case ANYOF_NLOWER:
3485                     if (LOC)
3486                         ANYOF_CLASS_SET(ret, ANYOF_NLOWER);
3487                     else {
3488                         for (value = 0; value < 256; value++)
3489                             if (!isLOWER(value))
3490                                 ANYOF_BITMAP_SET(ret, value);
3491                     }
3492                     break;
3493                 case ANYOF_PRINT:
3494                     if (LOC)
3495                         ANYOF_CLASS_SET(ret, ANYOF_PRINT);
3496                     else {
3497                         for (value = 0; value < 256; value++)
3498                             if (isPRINT(value))
3499                                 ANYOF_BITMAP_SET(ret, value);
3500                     }
3501                     break;
3502                 case ANYOF_NPRINT:
3503                     if (LOC)
3504                         ANYOF_CLASS_SET(ret, ANYOF_NPRINT);
3505                     else {
3506                         for (value = 0; value < 256; value++)
3507                             if (!isPRINT(value))
3508                                 ANYOF_BITMAP_SET(ret, value);
3509                     }
3510                     break;
3511                 case ANYOF_PSXSPC:
3512                     if (LOC)
3513                         ANYOF_CLASS_SET(ret, ANYOF_PSXSPC);
3514                     else {
3515                         for (value = 0; value < 256; value++)
3516                             if (isPSXSPC(value))
3517                                 ANYOF_BITMAP_SET(ret, value);
3518                     }
3519                     break;
3520                 case ANYOF_NPSXSPC:
3521                     if (LOC)
3522                         ANYOF_CLASS_SET(ret, ANYOF_NPSXSPC);
3523                     else {
3524                         for (value = 0; value < 256; value++)
3525                             if (!isPSXSPC(value))
3526                                 ANYOF_BITMAP_SET(ret, value);
3527                     }
3528                     break;
3529                 case ANYOF_PUNCT:
3530                     if (LOC)
3531                         ANYOF_CLASS_SET(ret, ANYOF_PUNCT);
3532                     else {
3533                         for (value = 0; value < 256; value++)
3534                             if (isPUNCT(value))
3535                                 ANYOF_BITMAP_SET(ret, value);
3536                     }
3537                     break;
3538                 case ANYOF_NPUNCT:
3539                     if (LOC)
3540                         ANYOF_CLASS_SET(ret, ANYOF_NPUNCT);
3541                     else {
3542                         for (value = 0; value < 256; value++)
3543                             if (!isPUNCT(value))
3544                                 ANYOF_BITMAP_SET(ret, value);
3545                     }
3546                     break;
3547                 case ANYOF_UPPER:
3548                     if (LOC)
3549                         ANYOF_CLASS_SET(ret, ANYOF_UPPER);
3550                     else {
3551                         for (value = 0; value < 256; value++)
3552                             if (isUPPER(value))
3553                                 ANYOF_BITMAP_SET(ret, value);
3554                     }
3555                     break;
3556                 case ANYOF_NUPPER:
3557                     if (LOC)
3558                         ANYOF_CLASS_SET(ret, ANYOF_NUPPER);
3559                     else {
3560                         for (value = 0; value < 256; value++)
3561                             if (!isUPPER(value))
3562                                 ANYOF_BITMAP_SET(ret, value);
3563                     }
3564                     break;
3565                 case ANYOF_XDIGIT:
3566                     if (LOC)
3567                         ANYOF_CLASS_SET(ret, ANYOF_XDIGIT);
3568                     else {
3569                         for (value = 0; value < 256; value++)
3570                             if (isXDIGIT(value))
3571                                 ANYOF_BITMAP_SET(ret, value);
3572                     }
3573                     break;
3574                 case ANYOF_NXDIGIT:
3575                     if (LOC)
3576                         ANYOF_CLASS_SET(ret, ANYOF_NXDIGIT);
3577                     else {
3578                         for (value = 0; value < 256; value++)
3579                             if (!isXDIGIT(value))
3580                                 ANYOF_BITMAP_SET(ret, value);
3581                     }
3582                     break;
3583                 default:
3584                     vFAIL("Invalid [::] class");
3585                     break;
3586                 }
3587                 if (LOC)
3588                     ANYOF_FLAGS(ret) |= ANYOF_CLASS;
3589                 continue;
3590             }
3591         }
3592         if (range) {
3593             if (lastvalue > value) /* b-a */ {
3594                 Simple_vFAIL4("Invalid [] range \"%*.*s\"",
3595                               RExC_parse - rangebegin,
3596                               RExC_parse - rangebegin,
3597                               rangebegin);
3598             }
3599             range = 0;
3600         }
3601         else {
3602             lastvalue = value;
3603             if (*RExC_parse == '-' && RExC_parse+1 < RExC_end &&
3604                 RExC_parse[1] != ']') {
3605                 RExC_parse++;
3606                 if (namedclass > OOB_NAMEDCLASS) { /* \w-, [:word:]- */
3607                     if (ckWARN(WARN_REGEXP))
3608                         vWARN4(RExC_parse,
3609                                "False [] range \"%*.*s\"",
3610                                RExC_parse - rangebegin,
3611                                RExC_parse - rangebegin,
3612                                rangebegin);
3613                     if (!SIZE_ONLY)
3614                         ANYOF_BITMAP_SET(ret, '-');
3615                 } else
3616                     range = 1;
3617                 continue;       /* do it next time */
3618             }
3619         }
3620         /* now is the next time */
3621         if (!SIZE_ONLY) {
3622 #ifndef ASCIIish /* EBCDIC, for example. */
3623             if ((isLOWER(lastvalue) && isLOWER(value)) ||
3624                 (isUPPER(lastvalue) && isUPPER(value)))
3625             {
3626                 I32 i;
3627                 if (isLOWER(lastvalue)) {
3628                     for (i = lastvalue; i <= value; i++)
3629                         if (isLOWER(i))
3630                             ANYOF_BITMAP_SET(ret, i);
3631                 } else {
3632                     for (i = lastvalue; i <= value; i++)
3633                         if (isUPPER(i))
3634                             ANYOF_BITMAP_SET(ret, i);
3635                 }
3636             }
3637             else
3638 #endif
3639                 for ( ; lastvalue <= value; lastvalue++)
3640                     ANYOF_BITMAP_SET(ret, lastvalue);
3641         }
3642         range = 0;
3643     }
3644     if (need_class) {
3645         if (SIZE_ONLY)
3646             RExC_size += ANYOF_CLASS_ADD_SKIP;
3647         else
3648             RExC_emit += ANYOF_CLASS_ADD_SKIP;
3649     }
3650     /* optimize case-insensitive simple patterns (e.g. /[a-z]/i) */
3651     if (!SIZE_ONLY &&
3652         (ANYOF_FLAGS(ret) & (ANYOF_FLAGS_ALL ^ ANYOF_INVERT)) == ANYOF_FOLD) {
3653         for (value = 0; value < 256; ++value) {
3654             if (ANYOF_BITMAP_TEST(ret, value)) {
3655                 I32 cf = PL_fold[value];
3656                 ANYOF_BITMAP_SET(ret, cf);
3657             }
3658         }
3659         ANYOF_FLAGS(ret) &= ~ANYOF_FOLD;
3660     }
3661     /* optimize inverted simple patterns (e.g. [^a-z]) */
3662     if (!SIZE_ONLY && (ANYOF_FLAGS(ret) & ANYOF_FLAGS_ALL) == ANYOF_INVERT) {
3663         for (value = 0; value < ANYOF_BITMAP_SIZE; ++value)
3664             ANYOF_BITMAP(ret)[value] ^= ANYOF_FLAGS_ALL;
3665         ANYOF_FLAGS(ret) = 0;
3666     }
3667     return ret;
3668 }
3669
3670 STATIC regnode *
3671 S_regclassutf8(pTHX_ RExC_state_t *pRExC_state)
3672 {
3673     register char *e;
3674     register U32 value;
3675     register U32 lastvalue = OOB_UTF8;
3676     register I32 range = 0;
3677     register regnode *ret;
3678     STRLEN numlen;
3679     I32 n;
3680     SV *listsv;
3681     U8 flags = 0;
3682     I32 namedclass;
3683     char *rangebegin;
3684
3685     if (*RExC_parse == '^') {   /* Complement of range. */
3686         RExC_naughty++;
3687         RExC_parse++;
3688         if (!SIZE_ONLY)
3689             flags |= ANYOF_INVERT;
3690     }
3691     if (!SIZE_ONLY) {
3692         if (FOLD)
3693             flags |= ANYOF_FOLD;
3694         if (LOC)
3695             flags |= ANYOF_LOCALE;
3696         listsv = newSVpvn("# comment\n", 10);
3697     }
3698
3699     if (!SIZE_ONLY && ckWARN(WARN_REGEXP))
3700         checkposixcc(pRExC_state);
3701
3702     if (*RExC_parse == ']' || *RExC_parse == '-')
3703         goto skipcond;          /* allow 1st char to be ] or - */
3704
3705     while (RExC_parse < RExC_end && *RExC_parse != ']') {
3706        skipcond:
3707         namedclass = OOB_NAMEDCLASS;
3708         if (!range)
3709             rangebegin = RExC_parse;
3710         value = utf8_to_uv((U8*)RExC_parse,
3711                                RExC_end - RExC_parse,
3712                                &numlen, 0);
3713         RExC_parse += numlen;
3714         if (value == '[')
3715             namedclass = regpposixcc(pRExC_state, value);
3716         else if (value == '\\') {
3717             value = (U32)utf8_to_uv((U8*)RExC_parse,
3718                                         RExC_end - RExC_parse,
3719                                         &numlen, 0);
3720             RExC_parse += numlen;
3721             /* Some compilers cannot handle switching on 64-bit integer
3722              * values, therefore value cannot be an UV.  Yes, this will
3723              * be a problem later if we want switch on Unicode.  --jhi */
3724             switch (value) {
3725             case 'w':           namedclass = ANYOF_ALNUM;               break;
3726             case 'W':           namedclass = ANYOF_NALNUM;              break;
3727             case 's':           namedclass = ANYOF_SPACE;               break;
3728             case 'S':           namedclass = ANYOF_NSPACE;              break;
3729             case 'd':           namedclass = ANYOF_DIGIT;               break;
3730             case 'D':           namedclass = ANYOF_NDIGIT;              break;
3731             case 'p':
3732             case 'P':
3733                 if (*RExC_parse == '{') {
3734                     e = strchr(RExC_parse++, '}');
3735                     if (!e)
3736                         vFAIL("Missing right brace on \\p{}");
3737                     n = e - RExC_parse;
3738                 }
3739                 else {
3740                     e = RExC_parse;
3741                     n = 1;
3742                 }
3743                 if (!SIZE_ONLY) {
3744                     if (value == 'p')
3745                         Perl_sv_catpvf(aTHX_ listsv,
3746                                        "+utf8::%.*s\n", (int)n, RExC_parse);
3747                     else
3748                         Perl_sv_catpvf(aTHX_ listsv,
3749                                        "!utf8::%.*s\n", (int)n, RExC_parse);
3750                 }
3751                 RExC_parse = e + 1;
3752                 lastvalue = OOB_UTF8;
3753                 continue;
3754             case 'n':           value = '\n';           break;
3755             case 'r':           value = '\r';           break;
3756             case 't':           value = '\t';           break;
3757             case 'f':           value = '\f';           break;
3758             case 'b':           value = '\b';           break;
3759 #ifdef ASCIIish
3760             case 'e':           value = '\033';         break;
3761             case 'a':           value = '\007';         break;
3762 #else
3763             case 'e':           value = '\047';         break;
3764             case 'a':           value = '\057';         break;
3765 #endif
3766             case 'x':
3767                 if (*RExC_parse == '{') {
3768                     e = strchr(RExC_parse++, '}');
3769                     if (!e)
3770                         vFAIL("Missing right brace on \\x{}");
3771                     numlen = 1;         /* allow underscores */
3772                     value = (UV)scan_hex(RExC_parse,
3773                                      e - RExC_parse,
3774                                      &numlen);
3775                     RExC_parse = e + 1;
3776                 }
3777                 else {
3778                     numlen = 0;         /* disallow underscores */
3779                     value = (UV)scan_hex(RExC_parse, 2, &numlen);
3780                     RExC_parse += numlen;
3781                 }
3782                 break;
3783             case 'c':
3784                 value = UCHARAT(RExC_parse++);
3785                 value = toCTRL(value);
3786                 break;
3787             case '0': case '1': case '2': case '3': case '4':
3788             case '5': case '6': case '7': case '8': case '9':
3789                 numlen = 0;             /* disallow underscores */
3790                 value = (UV)scan_oct(--RExC_parse, 3, &numlen);
3791                 RExC_parse += numlen;
3792                 break;
3793             default:
3794                 if (!SIZE_ONLY && ckWARN(WARN_REGEXP) && isALPHA(value))
3795                     vWARN2(RExC_parse,
3796                            "Unrecognized escape \\%c in character class passed through",
3797                            (int)value);
3798                 break;
3799             }
3800         }
3801         if (namedclass > OOB_NAMEDCLASS) {
3802             if (range) { /* a-\d, a-[:digit:] */
3803                 if (!SIZE_ONLY) {
3804                     if (ckWARN(WARN_REGEXP))
3805                         vWARN4(RExC_parse,
3806                                "False [] range \"%*.*s\"",
3807                                RExC_parse - rangebegin,
3808                                RExC_parse - rangebegin,
3809                                rangebegin);
3810                     Perl_sv_catpvf(aTHX_ listsv,
3811                                    /* 0x002D is Unicode for '-' */
3812                                    "%04"UVxf"\n002D\n", (UV)lastvalue);
3813                 }
3814                 range = 0;
3815             }
3816             if (!SIZE_ONLY) {
3817                 switch (namedclass) {
3818                 case ANYOF_ALNUM:
3819                     Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsWord\n");    break;
3820                 case ANYOF_NALNUM:
3821                     Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsWord\n");    break;
3822                 case ANYOF_ALNUMC:
3823                     Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsAlnum\n");   break;
3824                 case ANYOF_NALNUMC:
3825                     Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsAlnum\n");   break;
3826                 case ANYOF_ALPHA:
3827                     Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsAlpha\n");   break;
3828                 case ANYOF_NALPHA:
3829                     Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsAlpha\n");   break;
3830                 case ANYOF_ASCII:
3831                     Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsASCII\n");   break;
3832                 case ANYOF_NASCII:
3833                     Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsASCII\n");   break;
3834                 case ANYOF_CNTRL:
3835                     Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsCntrl\n");   break;
3836                 case ANYOF_NCNTRL:
3837                     Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsCntrl\n");   break;
3838                 case ANYOF_GRAPH:
3839                     Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsGraph\n");   break;
3840                 case ANYOF_NGRAPH:
3841                     Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsGraph\n");   break;
3842                 case ANYOF_DIGIT:
3843                     Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsDigit\n");   break;
3844                 case ANYOF_NDIGIT:
3845                     Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsDigit\n");   break;
3846                 case ANYOF_LOWER:
3847                     Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsLower\n");   break;
3848                 case ANYOF_NLOWER:
3849                     Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsLower\n");   break;
3850                 case ANYOF_PRINT:
3851                     Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsPrint\n");   break;
3852                 case ANYOF_NPRINT:
3853                     Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsPrint\n");   break;
3854                 case ANYOF_PUNCT:
3855                     Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsPunct\n");   break;
3856                 case ANYOF_NPUNCT:
3857                     Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsPunct\n");   break;
3858                 case ANYOF_SPACE:
3859                     Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsSpacePerl\n");break;
3860                 case ANYOF_NSPACE:
3861                     Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsSpacePerl\n");break;
3862                 case ANYOF_BLANK:
3863                     Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsBlank\n");   break;
3864                 case ANYOF_NBLANK:
3865                     Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsBlank\n");   break;
3866                 case ANYOF_PSXSPC:
3867                     Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsSpace\n");   break;
3868                 case ANYOF_NPSXSPC:
3869                     Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsSpace\n");   break;
3870                 case ANYOF_UPPER:
3871                     Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsUpper\n");   break;
3872                 case ANYOF_NUPPER:
3873                     Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsUpper\n");   break;
3874                 case ANYOF_XDIGIT:
3875                     Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsXDigit\n");  break;
3876                 case ANYOF_NXDIGIT:
3877                     Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsXDigit\n");  break;
3878                 }
3879                 continue;
3880             }
3881         }
3882         if (range) {
3883             if (lastvalue > value) { /* b-a */
3884                 Simple_vFAIL4("Invalid [] range \"%*.*s\"",
3885                               RExC_parse - rangebegin,
3886                               RExC_parse - rangebegin,
3887                               rangebegin);
3888             }
3889             range = 0;
3890         }
3891         else {
3892             lastvalue = value;
3893             if (*RExC_parse == '-' && RExC_parse+1 < RExC_end &&
3894                 RExC_parse[1] != ']') {
3895                 RExC_parse++;
3896                 if (namedclass > OOB_NAMEDCLASS) { /* \w-, [:word:]- */
3897                     if (ckWARN(WARN_REGEXP))
3898                         vWARN4(RExC_parse,
3899                                "False [] range \"%*.*s\"",
3900                                RExC_parse - rangebegin,
3901                                RExC_parse - rangebegin,
3902                                rangebegin);
3903                     if (!SIZE_ONLY)
3904                         Perl_sv_catpvf(aTHX_ listsv,
3905                                        /* 0x002D is Unicode for '-' */
3906                                        "002D\n");
3907                 } else
3908                     range = 1;
3909                 continue;       /* do it next time */
3910             }
3911         }
3912         /* now is the next time */
3913         if (!SIZE_ONLY)
3914             Perl_sv_catpvf(aTHX_ listsv, "%04"UVxf"\t%04"UVxf"\n",
3915                            (UV)lastvalue, (UV)value);
3916         range = 0;
3917     }
3918
3919     ret = reganode(pRExC_state, ANYOFUTF8, 0);
3920
3921     if (!SIZE_ONLY) {
3922         SV *rv = swash_init("utf8", "", listsv, 1, 0);
3923 #ifdef DEBUGGING
3924         AV *av = newAV();
3925         av_push(av, rv);
3926         av_push(av, listsv);
3927         rv = newRV_inc((SV*)av);
3928 #else
3929         SvREFCNT_dec(listsv);
3930 #endif
3931         n = add_data(pRExC_state, 1, "s");
3932         RExC_rx->data->data[n] = (void*)rv;
3933         ARG1_SET(ret, flags);
3934         ARG2_SET(ret, n);
3935     }
3936
3937     return ret;
3938 }
3939
3940 STATIC char*
3941 S_nextchar(pTHX_ RExC_state_t *pRExC_state)
3942 {
3943     char* retval = RExC_parse++;
3944
3945     for (;;) {
3946         if (*RExC_parse == '(' && RExC_parse[1] == '?' &&
3947                 RExC_parse[2] == '#') {
3948             while (*RExC_parse && *RExC_parse != ')')
3949                 RExC_parse++;
3950             RExC_parse++;
3951             continue;
3952         }
3953         if (RExC_flags16 & PMf_EXTENDED) {
3954             if (isSPACE(*RExC_parse)) {
3955                 RExC_parse++;
3956                 continue;
3957             }
3958             else if (*RExC_parse == '#') {
3959                 while (*RExC_parse && *RExC_parse != '\n')
3960                     RExC_parse++;
3961                 RExC_parse++;
3962                 continue;
3963             }
3964         }
3965         return retval;
3966     }
3967 }
3968
3969 /*
3970 - reg_node - emit a node
3971 */
3972 STATIC regnode *                        /* Location. */
3973 S_reg_node(pTHX_ RExC_state_t *pRExC_state, U8 op)
3974 {
3975     register regnode *ret;
3976     register regnode *ptr;
3977
3978     ret = RExC_emit;
3979     if (SIZE_ONLY) {
3980         SIZE_ALIGN(RExC_size);
3981         RExC_size += 1;
3982         return(ret);
3983     }
3984
3985     NODE_ALIGN_FILL(ret);
3986     ptr = ret;
3987     FILL_ADVANCE_NODE(ptr, op);
3988     RExC_emit = ptr;
3989
3990     return(ret);
3991 }
3992
3993 /*
3994 - reganode - emit a node with an argument
3995 */
3996 STATIC regnode *                        /* Location. */
3997 S_reganode(pTHX_ RExC_state_t *pRExC_state, U8 op, U32 arg)
3998 {
3999     register regnode *ret;
4000     register regnode *ptr;
4001
4002     ret = RExC_emit;
4003     if (SIZE_ONLY) {
4004         SIZE_ALIGN(RExC_size);
4005         RExC_size += 2;
4006         return(ret);
4007     }
4008
4009     NODE_ALIGN_FILL(ret);
4010     ptr = ret;
4011     FILL_ADVANCE_NODE_ARG(ptr, op, arg);
4012     RExC_emit = ptr;
4013
4014     return(ret);
4015 }
4016
4017 /*
4018 - reguni - emit (if appropriate) a Unicode character
4019 */
4020 STATIC void
4021 S_reguni(pTHX_ RExC_state_t *pRExC_state, UV uv, char* s, STRLEN* lenp)
4022 {
4023     *lenp = SIZE_ONLY ? UNISKIP(uv) : (uv_to_utf8((U8*)s, uv) - (U8*)s);
4024 }
4025
4026 /*
4027 - reginsert - insert an operator in front of already-emitted operand
4028 *
4029 * Means relocating the operand.
4030 */
4031 STATIC void
4032 S_reginsert(pTHX_ RExC_state_t *pRExC_state, U8 op, regnode *opnd)
4033 {
4034     register regnode *src;
4035     register regnode *dst;
4036     register regnode *place;
4037     register int offset = regarglen[(U8)op];
4038
4039 /* (PL_regkind[(U8)op] == CURLY ? EXTRA_STEP_2ARGS : 0); */
4040
4041     if (SIZE_ONLY) {
4042         RExC_size += NODE_STEP_REGNODE + offset;
4043         return;
4044     }
4045
4046     src = RExC_emit;
4047     RExC_emit += NODE_STEP_REGNODE + offset;
4048     dst = RExC_emit;
4049     while (src > opnd)
4050         StructCopy(--src, --dst, regnode);
4051
4052     place = opnd;               /* Op node, where operand used to be. */
4053     src = NEXTOPER(place);
4054     FILL_ADVANCE_NODE(place, op);
4055     Zero(src, offset, regnode);
4056 }
4057
4058 /*
4059 - regtail - set the next-pointer at the end of a node chain of p to val.
4060 */
4061 STATIC void
4062 S_regtail(pTHX_ RExC_state_t *pRExC_state, regnode *p, regnode *val)
4063 {
4064     register regnode *scan;
4065     register regnode *temp;
4066
4067     if (SIZE_ONLY)
4068         return;
4069
4070     /* Find last node. */
4071     scan = p;
4072     for (;;) {
4073         temp = regnext(scan);
4074         if (temp == NULL)
4075             break;
4076         scan = temp;
4077     }
4078
4079     if (reg_off_by_arg[OP(scan)]) {
4080         ARG_SET(scan, val - scan);
4081     }
4082     else {
4083         NEXT_OFF(scan) = val - scan;
4084     }
4085 }
4086
4087 /*
4088 - regoptail - regtail on operand of first argument; nop if operandless
4089 */
4090 STATIC void
4091 S_regoptail(pTHX_ RExC_state_t *pRExC_state, regnode *p, regnode *val)
4092 {
4093     /* "Operandless" and "op != BRANCH" are synonymous in practice. */
4094     if (p == NULL || SIZE_ONLY)
4095         return;
4096     if (PL_regkind[(U8)OP(p)] == BRANCH) {
4097         regtail(pRExC_state, NEXTOPER(p), val);
4098     }
4099     else if ( PL_regkind[(U8)OP(p)] == BRANCHJ) {
4100         regtail(pRExC_state, NEXTOPER(NEXTOPER(p)), val);
4101     }
4102     else
4103         return;
4104 }
4105
4106 /*
4107  - regcurly - a little FSA that accepts {\d+,?\d*}
4108  */
4109 STATIC I32
4110 S_regcurly(pTHX_ register char *s)
4111 {
4112     if (*s++ != '{')
4113         return FALSE;
4114     if (!isDIGIT(*s))
4115         return FALSE;
4116     while (isDIGIT(*s))
4117         s++;
4118     if (*s == ',')
4119         s++;
4120     while (isDIGIT(*s))
4121         s++;
4122     if (*s != '}')
4123         return FALSE;
4124     return TRUE;
4125 }
4126
4127
4128 STATIC regnode *
4129 S_dumpuntil(pTHX_ regnode *start, regnode *node, regnode *last, SV* sv, I32 l)
4130 {
4131 #ifdef DEBUGGING
4132     register U8 op = EXACT;     /* Arbitrary non-END op. */
4133     register regnode *next;
4134
4135     while (op != END && (!last || node < last)) {
4136         /* While that wasn't END last time... */
4137
4138         NODE_ALIGN(node);
4139         op = OP(node);
4140         if (op == CLOSE)
4141             l--;
4142         next = regnext(node);
4143         /* Where, what. */
4144         if (OP(node) == OPTIMIZED)
4145             goto after_print;
4146         regprop(sv, node);
4147         PerlIO_printf(Perl_debug_log, "%4"IVdf":%*s%s", (IV)(node - start),
4148                       (int)(2*l + 1), "", SvPVX(sv));
4149         if (next == NULL)               /* Next ptr. */
4150             PerlIO_printf(Perl_debug_log, "(0)");
4151         else
4152             PerlIO_printf(Perl_debug_log, "(%"IVdf")", (IV)(next - start));
4153         (void)PerlIO_putc(Perl_debug_log, '\n');
4154       after_print:
4155         if (PL_regkind[(U8)op] == BRANCHJ) {
4156             register regnode *nnode = (OP(next) == LONGJMP
4157                                        ? regnext(next)
4158                                        : next);
4159             if (last && nnode > last)
4160                 nnode = last;
4161             node = dumpuntil(start, NEXTOPER(NEXTOPER(node)), nnode, sv, l + 1);
4162         }
4163         else if (PL_regkind[(U8)op] == BRANCH) {
4164             node = dumpuntil(start, NEXTOPER(node), next, sv, l + 1);
4165         }
4166         else if ( op == CURLY) {   /* `next' might be very big: optimizer */
4167             node = dumpuntil(start, NEXTOPER(node) + EXTRA_STEP_2ARGS,
4168                              NEXTOPER(node) + EXTRA_STEP_2ARGS + 1, sv, l + 1);
4169         }
4170         else if (PL_regkind[(U8)op] == CURLY && op != CURLYX) {
4171             node = dumpuntil(start, NEXTOPER(node) + EXTRA_STEP_2ARGS,
4172                              next, sv, l + 1);
4173         }
4174         else if ( op == PLUS || op == STAR) {
4175             node = dumpuntil(start, NEXTOPER(node), NEXTOPER(node) + 1, sv, l + 1);
4176         }
4177         else if (op == ANYOF) {
4178             node = NEXTOPER(node);
4179             node += ANYOF_SKIP;
4180         }
4181         else if (PL_regkind[(U8)op] == EXACT) {
4182             /* Literal string, where present. */
4183             node += NODE_SZ_STR(node) - 1;
4184             node = NEXTOPER(node);
4185         }
4186         else {
4187             node = NEXTOPER(node);
4188             node += regarglen[(U8)op];
4189         }
4190         if (op == CURLYX || op == OPEN)
4191             l++;
4192         else if (op == WHILEM)
4193             l--;
4194     }
4195 #endif  /* DEBUGGING */
4196     return node;
4197 }
4198
4199 /*
4200  - regdump - dump a regexp onto Perl_debug_log in vaguely comprehensible form
4201  */
4202 void
4203 Perl_regdump(pTHX_ regexp *r)
4204 {
4205 #ifdef DEBUGGING
4206     SV *sv = sv_newmortal();
4207
4208     (void)dumpuntil(r->program, r->program + 1, NULL, sv, 0);
4209
4210     /* Header fields of interest. */
4211     if (r->anchored_substr)
4212         PerlIO_printf(Perl_debug_log,
4213                       "anchored `%s%.*s%s'%s at %"IVdf" ",
4214                       PL_colors[0],
4215                       (int)(SvCUR(r->anchored_substr) - (SvTAIL(r->anchored_substr)!=0)),
4216                       SvPVX(r->anchored_substr),
4217                       PL_colors[1],
4218                       SvTAIL(r->anchored_substr) ? "$" : "",
4219                       (IV)r->anchored_offset);
4220     if (r->float_substr)
4221         PerlIO_printf(Perl_debug_log,
4222                       "floating `%s%.*s%s'%s at %"IVdf"..%"UVuf" ",
4223                       PL_colors[0],
4224                       (int)(SvCUR(r->float_substr) - (SvTAIL(r->float_substr)!=0)),
4225                       SvPVX(r->float_substr),
4226                       PL_colors[1],
4227                       SvTAIL(r->float_substr) ? "$" : "",
4228                       (IV)r->float_min_offset, (UV)r->float_max_offset);
4229     if (r->check_substr)
4230         PerlIO_printf(Perl_debug_log,
4231                       r->check_substr == r->float_substr
4232                       ? "(checking floating" : "(checking anchored");
4233     if (r->reganch & ROPT_NOSCAN)
4234         PerlIO_printf(Perl_debug_log, " noscan");
4235     if (r->reganch & ROPT_CHECK_ALL)
4236         PerlIO_printf(Perl_debug_log, " isall");
4237     if (r->check_substr)
4238         PerlIO_printf(Perl_debug_log, ") ");
4239
4240     if (r->regstclass) {
4241         regprop(sv, r->regstclass);
4242         PerlIO_printf(Perl_debug_log, "stclass `%s' ", SvPVX(sv));
4243     }
4244     if (r->reganch & ROPT_ANCH) {
4245         PerlIO_printf(Perl_debug_log, "anchored");
4246         if (r->reganch & ROPT_ANCH_BOL)
4247             PerlIO_printf(Perl_debug_log, "(BOL)");
4248         if (r->reganch & ROPT_ANCH_MBOL)
4249             PerlIO_printf(Perl_debug_log, "(MBOL)");
4250         if (r->reganch & ROPT_ANCH_SBOL)
4251             PerlIO_printf(Perl_debug_log, "(SBOL)");
4252         if (r->reganch & ROPT_ANCH_GPOS)
4253             PerlIO_printf(Perl_debug_log, "(GPOS)");
4254         PerlIO_putc(Perl_debug_log, ' ');
4255     }
4256     if (r->reganch & ROPT_GPOS_SEEN)
4257         PerlIO_printf(Perl_debug_log, "GPOS ");
4258     if (r->reganch & ROPT_SKIP)
4259         PerlIO_printf(Perl_debug_log, "plus ");
4260     if (r->reganch & ROPT_IMPLICIT)
4261         PerlIO_printf(Perl_debug_log, "implicit ");
4262     PerlIO_printf(Perl_debug_log, "minlen %ld ", (long) r->minlen);
4263     if (r->reganch & ROPT_EVAL_SEEN)
4264         PerlIO_printf(Perl_debug_log, "with eval ");
4265     PerlIO_printf(Perl_debug_log, "\n");
4266 #endif  /* DEBUGGING */
4267 }
4268
4269 STATIC void
4270 S_put_byte(pTHX_ SV *sv, int c)
4271 {
4272     if (isCNTRL(c) || c == 127 || c == 255)
4273         Perl_sv_catpvf(aTHX_ sv, "\\%o", c);
4274     else if (c == '-' || c == ']' || c == '\\' || c == '^')
4275         Perl_sv_catpvf(aTHX_ sv, "\\%c", c);
4276     else
4277         Perl_sv_catpvf(aTHX_ sv, "%c", c);
4278 }
4279
4280 /*
4281 - regprop - printable representation of opcode
4282 */
4283 void
4284 Perl_regprop(pTHX_ SV *sv, regnode *o)
4285 {
4286 #ifdef DEBUGGING
4287     register int k;
4288
4289     sv_setpvn(sv, "", 0);
4290     if (OP(o) >= reg_num)               /* regnode.type is unsigned */
4291         /* It would be nice to FAIL() here, but this may be called from
4292            regexec.c, and it would be hard to supply pRExC_state. */
4293         Perl_croak(aTHX_ "Corrupted regexp opcode");
4294     sv_catpv(sv, (char*)reg_name[OP(o)]); /* Take off const! */
4295
4296     k = PL_regkind[(U8)OP(o)];
4297
4298     if (k == EXACT)
4299         Perl_sv_catpvf(aTHX_ sv, " <%s%.*s%s>", PL_colors[0],
4300                        STR_LEN(o), STRING(o), PL_colors[1]);
4301     else if (k == CURLY) {
4302         if (OP(o) == CURLYM || OP(o) == CURLYN || OP(o) == CURLYX)
4303             Perl_sv_catpvf(aTHX_ sv, "[%d]", o->flags); /* Parenth number */
4304         Perl_sv_catpvf(aTHX_ sv, " {%d,%d}", ARG1(o), ARG2(o));
4305     }
4306     else if (k == WHILEM && o->flags)                   /* Ordinal/of */
4307         Perl_sv_catpvf(aTHX_ sv, "[%d/%d]", o->flags & 0xf, o->flags>>4);
4308     else if (k == REF || k == OPEN || k == CLOSE || k == GROUPP )
4309         Perl_sv_catpvf(aTHX_ sv, "%d", (int)ARG(o));    /* Parenth number */
4310     else if (k == LOGICAL)
4311         Perl_sv_catpvf(aTHX_ sv, "[%d]", o->flags);     /* 2: embedded, otherwise 1 */
4312     else if (k == ANYOF) {
4313         int i, rangestart = -1;
4314         bool anyofutf8 = OP(o) == ANYOFUTF8;
4315         U8 flags = anyofutf8 ? ARG1(o) : o->flags;
4316         const char * const anyofs[] = { /* Should be syncronized with
4317                                          * ANYOF_ #xdefines in regcomp.h */
4318             "\\w",
4319             "\\W",
4320             "\\s",
4321             "\\S",
4322             "\\d",
4323             "\\D",
4324             "[:alnum:]",
4325             "[:^alnum:]",
4326             "[:alpha:]",
4327             "[:^alpha:]",
4328             "[:ascii:]",
4329             "[:^ascii:]",
4330             "[:ctrl:]",
4331             "[:^ctrl:]",
4332             "[:graph:]",
4333             "[:^graph:]",
4334             "[:lower:]",
4335             "[:^lower:]",
4336             "[:print:]",
4337             "[:^print:]",
4338             "[:punct:]",
4339             "[:^punct:]",
4340             "[:upper:]",
4341             "[:^upper:]",
4342             "[:xdigit:]",
4343             "[:^xdigit:]",
4344             "[:space:]",
4345             "[:^space:]",
4346             "[:blank:]",
4347             "[:^blank:]"
4348         };
4349
4350         if (flags & ANYOF_LOCALE)
4351             sv_catpv(sv, "{loc}");
4352         if (flags & ANYOF_FOLD)
4353             sv_catpv(sv, "{i}");
4354         Perl_sv_catpvf(aTHX_ sv, "[%s", PL_colors[0]);
4355         if (flags & ANYOF_INVERT)
4356             sv_catpv(sv, "^");
4357         if (OP(o) == ANYOF) {
4358             for (i = 0; i <= 256; i++) {
4359                 if (i < 256 && ANYOF_BITMAP_TEST(o,i)) {
4360                     if (rangestart == -1)
4361                         rangestart = i;
4362                 } else if (rangestart != -1) {
4363                     if (i <= rangestart + 3)
4364                         for (; rangestart < i; rangestart++)
4365                             put_byte(sv, rangestart);
4366                     else {
4367                         put_byte(sv, rangestart);
4368                         sv_catpv(sv, "-");
4369                         put_byte(sv, i - 1);
4370                     }
4371                     rangestart = -1;
4372                 }
4373             }
4374             if (o->flags & ANYOF_CLASS)
4375                 for (i = 0; i < sizeof(anyofs)/sizeof(char*); i++)
4376                     if (ANYOF_CLASS_TEST(o,i))
4377                         sv_catpv(sv, anyofs[i]);
4378         }
4379         else {
4380             SV *rv = (SV*)PL_regdata->data[ARG2(o)];
4381             AV *av = (AV*)SvRV((SV*)rv);
4382             SV *sw = *av_fetch(av, 0, FALSE);
4383             SV *lv = *av_fetch(av, 1, FALSE);
4384             UV i;
4385             U8 s[UTF8_MAXLEN+1];
4386             for (i = 0; i <= 256; i++) { /* just the first 256 */
4387                 U8 *e = uv_to_utf8(s, i);
4388                 if (i < 256 && swash_fetch(sw, s)) {
4389                     if (rangestart == -1)
4390                         rangestart = i;
4391                 } else if (rangestart != -1) {
4392                     U8 *p;
4393
4394                     if (i <= rangestart + 3)
4395                         for (; rangestart < i; rangestart++) {
4396                             for(e = uv_to_utf8(s, rangestart), p = s; p < e; p++)
4397                                 put_byte(sv, *p);
4398                         }
4399                     else {
4400                         for (e = uv_to_utf8(s, rangestart), p = s; p < e; p++)
4401                             put_byte(sv, *p);
4402                         sv_catpv(sv, "-");
4403                         for (e = uv_to_utf8(s, i - 1), p = s; p < e; p++)
4404                             put_byte(sv, *p);
4405                     }
4406                     rangestart = -1;
4407                 }
4408             }
4409             sv_catpv(sv, "...");
4410             {
4411                 char *s = savepv(SvPVX(lv));
4412
4413                 while(*s && *s != '\n') s++;
4414                 if (*s == '\n') {
4415                     char *t = ++s;
4416
4417                     while (*s) {
4418                         if (*s == '\n')
4419                             *s = ' ';
4420                         s++;
4421                     }
4422                     if (s[-1] == ' ')
4423                         s[-1] = 0;
4424
4425                     sv_catpv(sv, t);
4426                 }
4427             }
4428         }
4429         Perl_sv_catpvf(aTHX_ sv, "%s]", PL_colors[1]);
4430     }
4431     else if (k == BRANCHJ && (OP(o) == UNLESSM || OP(o) == IFMATCH))
4432         Perl_sv_catpvf(aTHX_ sv, "[-%d]", o->flags);
4433 #endif  /* DEBUGGING */
4434 }
4435
4436 SV *
4437 Perl_re_intuit_string(pTHX_ regexp *prog)
4438 {                               /* Assume that RE_INTUIT is set */
4439     DEBUG_r(
4440         {   STRLEN n_a;
4441             char *s = SvPV(prog->check_substr,n_a);
4442
4443             if (!PL_colorset) reginitcolors();
4444             PerlIO_printf(Perl_debug_log,
4445                       "%sUsing REx substr:%s `%s%.60s%s%s'\n",
4446                       PL_colors[4],PL_colors[5],PL_colors[0],
4447                       s,
4448                       PL_colors[1],
4449                       (strlen(s) > 60 ? "..." : ""));
4450         } );
4451
4452     return prog->check_substr;
4453 }
4454
4455 void
4456 Perl_pregfree(pTHX_ struct regexp *r)
4457 {
4458     DEBUG_r(if (!PL_colorset) reginitcolors());
4459
4460     if (!r || (--r->refcnt > 0))
4461         return;
4462     DEBUG_r(PerlIO_printf(Perl_debug_log,
4463                       "%sFreeing REx:%s `%s%.60s%s%s'\n",
4464                       PL_colors[4],PL_colors[5],PL_colors[0],
4465                       r->precomp,
4466                       PL_colors[1],
4467                       (strlen(r->precomp) > 60 ? "..." : "")));
4468
4469     if (r->precomp)
4470         Safefree(r->precomp);
4471     if (RX_MATCH_COPIED(r))
4472         Safefree(r->subbeg);
4473     if (r->substrs) {
4474         if (r->anchored_substr)
4475             SvREFCNT_dec(r->anchored_substr);
4476         if (r->float_substr)
4477             SvREFCNT_dec(r->float_substr);
4478         Safefree(r->substrs);
4479     }
4480     if (r->data) {
4481         int n = r->data->count;
4482         AV* new_comppad = NULL;
4483         AV* old_comppad;
4484         SV** old_curpad;
4485
4486         while (--n >= 0) {
4487             switch (r->data->what[n]) {
4488             case 's':
4489 #ifdef DEBUGGING
4490                 {
4491                     SV *rv = (SV*)r->data->data[n];
4492                     AV *av = (AV*)SvRV((SV*)rv);
4493                     SV *sw = *av_fetch(av, 0, FALSE);
4494                     SV *lv = *av_fetch(av, 1, FALSE);
4495                     SvREFCNT_dec(sw);
4496                     SvREFCNT_dec(lv);
4497                 }
4498 #endif
4499                 SvREFCNT_dec((SV*)r->data->data[n]);
4500                 break;
4501             case 'f':
4502                 Safefree(r->data->data[n]);
4503                 break;
4504             case 'p':
4505                 new_comppad = (AV*)r->data->data[n];
4506                 break;
4507             case 'o':
4508                 if (new_comppad == NULL)
4509                     Perl_croak(aTHX_ "panic: pregfree comppad");
4510                 old_comppad = PL_comppad;
4511                 old_curpad = PL_curpad;
4512                 /* Watch out for global destruction's random ordering. */
4513                 if (SvTYPE(new_comppad) == SVt_PVAV) {
4514                     PL_comppad = new_comppad;
4515                     PL_curpad = AvARRAY(new_comppad);
4516                 }
4517                 else
4518                     PL_curpad = NULL;
4519                 op_free((OP_4tree*)r->data->data[n]);
4520                 PL_comppad = old_comppad;
4521                 PL_curpad = old_curpad;
4522                 SvREFCNT_dec((SV*)new_comppad);
4523                 new_comppad = NULL;
4524                 break;
4525             case 'n':
4526                 break;
4527             default:
4528                 Perl_croak(aTHX_ "panic: regfree data code '%c'", r->data->what[n]);
4529             }
4530         }
4531         Safefree(r->data->what);
4532         Safefree(r->data);
4533     }
4534     Safefree(r->startp);
4535     Safefree(r->endp);
4536     Safefree(r);
4537 }
4538
4539 /*
4540  - regnext - dig the "next" pointer out of a node
4541  *
4542  * [Note, when REGALIGN is defined there are two places in regmatch()
4543  * that bypass this code for speed.]
4544  */
4545 regnode *
4546 Perl_regnext(pTHX_ register regnode *p)
4547 {
4548     register I32 offset;
4549
4550     if (p == &PL_regdummy)
4551         return(NULL);
4552
4553     offset = (reg_off_by_arg[OP(p)] ? ARG(p) : NEXT_OFF(p));
4554     if (offset == 0)
4555         return(NULL);
4556
4557     return(p+offset);
4558 }
4559
4560 STATIC void
4561 S_re_croak2(pTHX_ const char* pat1,const char* pat2,...)
4562 {
4563     va_list args;
4564     STRLEN l1 = strlen(pat1);
4565     STRLEN l2 = strlen(pat2);
4566     char buf[512];
4567     SV *msv;
4568     char *message;
4569
4570     if (l1 > 510)
4571         l1 = 510;
4572     if (l1 + l2 > 510)
4573         l2 = 510 - l1;
4574     Copy(pat1, buf, l1 , char);
4575     Copy(pat2, buf + l1, l2 , char);
4576     buf[l1 + l2] = '\n';
4577     buf[l1 + l2 + 1] = '\0';
4578 #ifdef I_STDARG
4579     /* ANSI variant takes additional second argument */
4580     va_start(args, pat2);
4581 #else
4582     va_start(args);
4583 #endif
4584     msv = vmess(buf, &args);
4585     va_end(args);
4586     message = SvPV(msv,l1);
4587     if (l1 > 512)
4588         l1 = 512;
4589     Copy(message, buf, l1 , char);
4590     buf[l1] = '\0';                     /* Overwrite \n */
4591     Perl_croak(aTHX_ "%s", buf);
4592 }
4593
4594 /* XXX Here's a total kludge.  But we need to re-enter for swash routines. */
4595
4596 void
4597 Perl_save_re_context(pTHX)
4598 {
4599 #if 0
4600     SAVEPPTR(RExC_precomp);             /* uncompiled string. */
4601     SAVEI32(RExC_npar);         /* () count. */
4602     SAVEI32(RExC_size);         /* Code size. */
4603     SAVEI16(RExC_flags16);              /* are we folding, multilining? */
4604     SAVEVPTR(RExC_rx);          /* from regcomp.c */
4605     SAVEI32(RExC_seen);         /* from regcomp.c */
4606     SAVEI32(RExC_sawback);              /* Did we see \1, ...? */
4607     SAVEI32(RExC_naughty);              /* How bad is this pattern? */
4608     SAVEVPTR(RExC_emit);                /* Code-emit pointer; &regdummy = don't */
4609     SAVEPPTR(RExC_end);         /* End of input for compile */
4610     SAVEPPTR(RExC_parse);               /* Input-scan pointer. */
4611 #endif
4612
4613     SAVEI32(PL_reg_flags);              /* from regexec.c */
4614     SAVEPPTR(PL_bostr);
4615     SAVEPPTR(PL_reginput);              /* String-input pointer. */
4616     SAVEPPTR(PL_regbol);                /* Beginning of input, for ^ check. */
4617     SAVEPPTR(PL_regeol);                /* End of input, for $ check. */
4618     SAVEVPTR(PL_regstartp);             /* Pointer to startp array. */
4619     SAVEVPTR(PL_regendp);               /* Ditto for endp. */
4620     SAVEVPTR(PL_reglastparen);          /* Similarly for lastparen. */
4621     SAVEPPTR(PL_regtill);               /* How far we are required to go. */
4622     SAVEI8(PL_regprev);                 /* char before regbol, \n if none */
4623     SAVEVPTR(PL_reg_start_tmp);         /* from regexec.c */
4624     PL_reg_start_tmp = 0;
4625     SAVEFREEPV(PL_reg_start_tmp);
4626     SAVEI32(PL_reg_start_tmpl);         /* from regexec.c */
4627     PL_reg_start_tmpl = 0;
4628     SAVEVPTR(PL_regdata);
4629     SAVEI32(PL_reg_eval_set);           /* from regexec.c */
4630     SAVEI32(PL_regnarrate);             /* from regexec.c */
4631     SAVEVPTR(PL_regprogram);            /* from regexec.c */
4632     SAVEINT(PL_regindent);              /* from regexec.c */
4633     SAVEVPTR(PL_regcc);                 /* from regexec.c */
4634     SAVEVPTR(PL_curcop);
4635     SAVEVPTR(PL_reg_call_cc);           /* from regexec.c */
4636     SAVEVPTR(PL_reg_re);                /* from regexec.c */
4637     SAVEPPTR(PL_reg_ganch);             /* from regexec.c */
4638     SAVESPTR(PL_reg_sv);                /* from regexec.c */
4639     SAVEVPTR(PL_reg_magic);             /* from regexec.c */
4640     SAVEI32(PL_reg_oldpos);                     /* from regexec.c */
4641     SAVEVPTR(PL_reg_oldcurpm);          /* from regexec.c */
4642     SAVEVPTR(PL_reg_curpm);             /* from regexec.c */
4643     SAVEI32(PL_regnpar);                /* () count. */
4644 #ifdef DEBUGGING
4645     SAVEPPTR(PL_reg_starttry);          /* from regexec.c */
4646 #endif
4647 }
4648
4649 #ifdef PERL_OBJECT
4650 #include "XSUB.h"
4651 #undef this
4652 #define this pPerl
4653 #endif
4654
4655 static void
4656 clear_re(pTHXo_ void *r)
4657 {
4658     ReREFCNT_dec((regexp *)r);
4659 }
4660