lib/libc/regex/regcomp.c

   1 /*      $NetBSD: regcomp.c,v 1.29 2009/02/12 05:06:54 lukem Exp $       */
   2
   3 /*-
   4  * Copyright (c) 1992, 1993, 1994
   5  *      The Regents of the University of California.  All rights reserved.
   6  *
   7  * This code is derived from software contributed to Berkeley by
   8  * Henry Spencer.
   9  *
  10  * Redistribution and use in source and binary forms, with or without
  11  * modification, are permitted provided that the following conditions
  12  * are met:
  13  * 1. Redistributions of source code must retain the above copyright
  14  *    notice, this list of conditions and the following disclaimer.
  15  * 2. Redistributions in binary form must reproduce the above copyright
  16  *    notice, this list of conditions and the following disclaimer in the
  17  *    documentation and/or other materials provided with the distribution.
  18  * 3. Neither the name of the University nor the names of its contributors
  19  *    may be used to endorse or promote products derived from this software
  20  *    without specific prior written permission.
  21  *
  22  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  23  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  24  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  25  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  26  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  27  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  28  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  29  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  30  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  31  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  32  * SUCH DAMAGE.
  33  *
  34  *      @(#)regcomp.c   8.5 (Berkeley) 3/20/94
  35  */
  36
  37 /*-
  38  * Copyright (c) 1992, 1993, 1994 Henry Spencer.
  39  *
  40  * This code is derived from software contributed to Berkeley by
  41  * Henry Spencer.
  42  *
  43  * Redistribution and use in source and binary forms, with or without
  44  * modification, are permitted provided that the following conditions
  45  * are met:
  46  * 1. Redistributions of source code must retain the above copyright
  47  *    notice, this list of conditions and the following disclaimer.
  48  * 2. Redistributions in binary form must reproduce the above copyright
  49  *    notice, this list of conditions and the following disclaimer in the
  50  *    documentation and/or other materials provided with the distribution.
  51  * 3. All advertising materials mentioning features or use of this software
  52  *    must display the following acknowledgement:
  53  *      This product includes software developed by the University of
  54  *      California, Berkeley and its contributors.
  55  * 4. Neither the name of the University nor the names of its contributors
  56  *    may be used to endorse or promote products derived from this software
  57  *    without specific prior written permission.
  58  *
  59  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  60  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  61  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  62  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  63  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  64  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  65  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  66  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  67  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  68  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  69  * SUCH DAMAGE.
  70  *
  71  *      @(#)regcomp.c   8.5 (Berkeley) 3/20/94
  72  */
  73
  74 #include <sys/cdefs.h>
  75 #if defined(LIBC_SCCS) && !defined(lint)
  76 #if 0
  77 static char sccsid[] = "@(#)regcomp.c   8.5 (Berkeley) 3/20/94";
  78 #else
  79 __RCSID("$NetBSD: regcomp.c,v 1.29 2009/02/12 05:06:54 lukem Exp $");
  80 #endif
  81 #endif /* LIBC_SCCS and not lint */
  82
  83 #include "namespace.h"
  84 #include <sys/types.h>
  85
  86 #include <assert.h>
  87 #include <ctype.h>
  88 #include <limits.h>
  89 #include <stdio.h>
  90 #include <stdlib.h>
  91 #include <string.h>
  92 #include <regex.h>
  93
  94 #ifdef __weak_alias
  95 __weak_alias(regcomp,_regcomp)
  96 #endif
  97
  98 #include "utils.h"
  99 #include "regex2.h"
 100
 101 #include "cclass.h"
 102 #include "cname.h"
 103
 104 /*
 105  * parse structure, passed up and down to avoid global variables and
 106  * other clumsinesses
 107  */
 108 struct parse {
 109         const char *next;       /* next character in RE */
 110         const char *end;        /* end of string (-> NUL normally) */
 111         int error;              /* has an error been seen? */
 112         sop *strip;             /* malloced strip */
 113         sopno ssize;            /* malloced strip size (allocated) */
 114         sopno slen;             /* malloced strip length (used) */
 115         int ncsalloc;           /* number of csets allocated */
 116         struct re_guts *g;
 117 #       define  NPAREN  10      /* we need to remember () 1-9 for back refs */
 118         sopno pbegin[NPAREN];   /* -> ( ([0] unused) */
 119         sopno pend[NPAREN];     /* -> ) ([0] unused) */
 120 };
 121
 122 /* ========= begin header generated by ./mkh ========= */
 123 #ifdef __cplusplus
 124 extern "C" {
 125 #endif
 126
 127 /* === regcomp.c === */
 128 static void p_ere(struct parse *p, int stop);
 129 static void p_ere_exp(struct parse *p);
 130 static void p_str(struct parse *p);
 131 static void p_bre(struct parse *p, int end1, int end2);
 132 static int p_simp_re(struct parse *p, int starordinary);
 133 static int p_count(struct parse *p);
 134 static void p_bracket(struct parse *p);
 135 static void p_b_term(struct parse *p, cset *cs);
 136 static void p_b_cclass(struct parse *p, cset *cs);
 137 static void p_b_eclass(struct parse *p, cset *cs);
 138 static char p_b_symbol(struct parse *p);
 139 static char p_b_coll_elem(struct parse *p, int endc);
 140 static int othercase(int ch);
 141 static void bothcases(struct parse *p, int ch);
 142 static void ordinary(struct parse *p, int ch);
 143 static void nonnewline(struct parse *p);
 144 static void repeat(struct parse *p, sopno start, int from, int to);
 145 static int seterr(struct parse *p, int e);
 146 static cset *allocset(struct parse *p);
 147 static void freeset(struct parse *p, cset *cs);
 148 static int freezeset(struct parse *p, cset *cs);
 149 static int firstch(struct parse *p, cset *cs);
 150 static int nch(struct parse *p, cset *cs);
 151 static void mcadd(struct parse *p, cset *cs, const char *cp);
 152 #if 0
 153 static void mcsub(cset *cs, char *cp);
 154 static int mcin(cset *cs, char *cp);
 155 static char *mcfind(cset *cs, char *cp);
 156 #endif
 157 static void mcinvert(struct parse *p, cset *cs);
 158 static void mccase(struct parse *p, cset *cs);
 159 static int isinsets(struct re_guts *g, int c);
 160 static int samesets(struct re_guts *g, int c1, int c2);
 161 static void categorize(struct parse *p, struct re_guts *g);
 162 static sopno dupl(struct parse *p, sopno start, sopno finish);
 163 static void doemit(struct parse *p, sop op, sopno opnd);
 164 static void doinsert(struct parse *p, sop op, sopno opnd, sopno pos);
 165 static void dofwd(struct parse *p, sopno pos, sopno value);
 166 static void enlarge(struct parse *p, sopno size);
 167 static void stripsnug(struct parse *p, struct re_guts *g);
 168 static void findmust(struct parse *p, struct re_guts *g);
 169 static sopno pluscount(struct parse *p, struct re_guts *g);
 170
 171 #ifdef __cplusplus
 172 }
 173 #endif
 174 /* ========= end header generated by ./mkh ========= */
 175
 176 static char nuls[10];           /* place to point scanner in event of error */
 177
 178 /*
 179  * macros for use with parse structure
 180  * BEWARE:  these know that the parse structure is named `p' !!!
 181  */
 182 #define PEEK()  (*p->next)
 183 #define PEEK2() (*(p->next+1))
 184 #define MORE()  (p->next < p->end)
 185 #define MORE2() (p->next+1 < p->end)
 186 #define SEE(c)  (MORE() && PEEK() == (c))
 187 #define SEETWO(a, b)    (MORE() && MORE2() && PEEK() == (a) && PEEK2() == (b))
 188 #define EAT(c)  ((SEE(c)) ? (NEXT(), 1) : 0)
 189 #define EATTWO(a, b)    ((SEETWO(a, b)) ? (NEXT2(), 1) : 0)
 190 #define NEXT()  (p->next++)
 191 #define NEXT2() (p->next += 2)
 192 #define NEXTn(n)        (p->next += (n))
 193 #define GETNEXT()       (*p->next++)
 194 #define SETERROR(e)     seterr(p, (e))
 195 #define REQUIRE(co, e)  (void) ((co) || SETERROR(e))
 196 #define MUSTSEE(c, e)   (REQUIRE(MORE() && PEEK() == (c), e))
 197 #define MUSTEAT(c, e)   (void) (REQUIRE(MORE() && GETNEXT() == (c), e))
 198 #define MUSTNOTSEE(c, e)        (REQUIRE(!MORE() || PEEK() != (c), e))
 199 #define EMIT(op, sopnd) doemit(p, (sop)(op), sopnd)
 200 #define INSERT(op, pos) doinsert(p, (sop)(op), HERE()-(pos)+1, pos)
 201 #define AHEAD(pos)              dofwd(p, pos, HERE()-(pos))
 202 #define ASTERN(sop, pos)        EMIT(sop, HERE()-pos)
 203 #define HERE()          (p->slen)
 204 #define THERE()         (p->slen - 1)
 205 #define THERETHERE()    (p->slen - 2)
 206 #define DROP(n) (p->slen -= (n))
 207
 208 #ifndef NDEBUG
 209 static int never = 0;           /* for use in asserts; shuts lint up */
 210 #else
 211 #define never   0               /* some <assert.h>s have bugs too */
 212 #endif
 213
 214 /*
 215  - regcomp - interface for parser and compilation
 216  = extern int regcomp(regex_t *, const char *, int);
 217  = #define      REG_BASIC       0000
 218  = #define      REG_EXTENDED    0001
 219  = #define      REG_ICASE       0002
 220  = #define      REG_NOSUB       0004
 221  = #define      REG_NEWLINE     0010
 222  = #define      REG_NOSPEC      0020
 223  = #define      REG_PEND        0040
 224  = #define      REG_DUMP        0200
 225  */
 226 int                             /* 0 success, otherwise REG_something */
 227 regcomp(
 228     regex_t *preg,
 229     const char *pattern,
 230     int cflags)
 231 {
 232         struct parse pa;
 233         struct re_guts *g;
 234         struct parse *p = &pa;
 235         int i;
 236         size_t len;
 237 #ifdef REDEBUG
 238 #       define  GOODFLAGS(f)    (f)
 239 #else
 240 #       define  GOODFLAGS(f)    ((f)&~REG_DUMP)
 241 #endif
 242
 243         _DIAGASSERT(preg != NULL);
 244         _DIAGASSERT(pattern != NULL);
 245
 246         cflags = GOODFLAGS(cflags);
 247         if ((cflags&REG_EXTENDED) && (cflags&REG_NOSPEC))
 248                 return(REG_INVARG);
 249
 250         if (cflags&REG_PEND) {
 251                 if (preg->re_endp < pattern)
 252                         return(REG_INVARG);
 253                 len = preg->re_endp - pattern;
 254         } else
 255                 len = strlen(pattern);
 256
 257         /* do the mallocs early so failure handling is easy */
 258         g = (struct re_guts *)malloc(sizeof(struct re_guts) +
 259                                                         (NC-1)*sizeof(cat_t));
 260         if (g == NULL)
 261                 return(REG_ESPACE);
 262         p->ssize = len/(size_t)2*(size_t)3 + (size_t)1; /* ugh */
 263         p->strip = (sop *)malloc(p->ssize * sizeof(sop));
 264         p->slen = 0;
 265         if (p->strip == NULL) {
 266                 free(g);
 267                 return(REG_ESPACE);
 268         }
 269
 270         /* set things up */
 271         p->g = g;
 272         p->next = pattern;
 273         p->end = p->next + len;
 274         p->error = 0;
 275         p->ncsalloc = 0;
 276         for (i = 0; i < NPAREN; i++) {
 277                 p->pbegin[i] = 0;
 278                 p->pend[i] = 0;
 279         }
 280         g->csetsize = NC;
 281         g->sets = NULL;
 282         g->setbits = NULL;
 283         g->ncsets = 0;
 284         g->cflags = cflags;
 285         g->iflags = 0;
 286         g->nbol = 0;
 287         g->neol = 0;
 288         g->must = NULL;
 289         g->mlen = 0;
 290         g->nsub = 0;
 291         g->ncategories = 1;     /* category 0 is "everything else" */
 292         g->categories = &g->catspace[-(CHAR_MIN)];
 293         (void) memset((char *)g->catspace, 0, NC*sizeof(cat_t));
 294         g->backrefs = 0;
 295
 296         /* do it */
 297         EMIT(OEND, 0);
 298         g->firststate = THERE();
 299         if (cflags&REG_EXTENDED)
 300                 p_ere(p, OUT);
 301         else if (cflags&REG_NOSPEC)
 302                 p_str(p);
 303         else
 304                 p_bre(p, OUT, OUT);
 305         EMIT(OEND, 0);
 306         g->laststate = THERE();
 307
 308         /* tidy up loose ends and fill things in */
 309         categorize(p, g);
 310         stripsnug(p, g);
 311         findmust(p, g);
 312         g->nplus = pluscount(p, g);
 313         g->magic = MAGIC2;
 314         preg->re_nsub = g->nsub;
 315         preg->re_g = g;
 316         preg->re_magic = MAGIC1;
 317 #ifndef REDEBUG
 318         /* not debugging, so can't rely on the assert() in regexec() */
 319         if (g->iflags&BAD)
 320                 SETERROR(REG_ASSERT);
 321 #endif
 322
 323         /* win or lose, we're done */
 324         if (p->error != 0)      /* lose */
 325                 regfree(preg);
 326         return(p->error);
 327 }
 328
 329 /*
 330  - p_ere - ERE parser top level, concatenation and alternation
 331  == static void p_ere(struct parse *p, int stop);
 332  */
 333 static void
 334 p_ere(
 335     struct parse *p,
 336     int stop)                   /* character this ERE should end at */
 337 {
 338         char c;
 339         sopno prevback = 0;     /* pacify gcc */
 340         sopno prevfwd = 0;      /* pacify gcc */
 341         sopno conc;
 342         int first = 1;          /* is this the first alternative? */
 343
 344         _DIAGASSERT(p != NULL);
 345
 346         for (;;) {
 347                 /* do a bunch of concatenated expressions */
 348                 conc = HERE();
 349                 while (MORE() && (c = PEEK()) != '|' && c != stop)
 350                         p_ere_exp(p);
 351                 REQUIRE(HERE() != conc, REG_EMPTY);     /* require nonempty */
 352
 353                 if (!EAT('|'))
 354                         break;          /* NOTE BREAK OUT */
 355
 356                 if (first) {
 357                         INSERT(OCH_, conc);     /* offset is wrong */
 358                         prevfwd = conc;
 359                         prevback = conc;
 360                         first = 0;
 361                 }
 362                 ASTERN(OOR1, prevback);
 363                 prevback = THERE();
 364                 AHEAD(prevfwd);                 /* fix previous offset */
 365                 prevfwd = HERE();
 366                 EMIT(OOR2, 0);                  /* offset is very wrong */
 367         }
 368
 369         if (!first) {           /* tail-end fixups */
 370                 AHEAD(prevfwd);
 371                 ASTERN(O_CH, prevback);
 372         }
 373
 374         assert(!MORE() || SEE(stop));
 375 }
 376
 377 /*
 378  - p_ere_exp - parse one subERE, an atom possibly followed by a repetition op
 379  == static void p_ere_exp(struct parse *p);
 380  */
 381 static void
 382 p_ere_exp(
 383     struct parse *p)
 384 {
 385         char c;
 386         sopno pos;
 387         int count;
 388         int count2;
 389         sopno subno;
 390         int wascaret = 0;
 391
 392         _DIAGASSERT(p != NULL);
 393
 394         assert(MORE());         /* caller should have ensured this */
 395         c = GETNEXT();
 396
 397         pos = HERE();
 398         switch (c) {
 399         case '(':
 400                 REQUIRE(MORE(), REG_EPAREN);
 401                 p->g->nsub++;
 402                 subno = p->g->nsub;
 403                 if (subno < NPAREN)
 404                         p->pbegin[subno] = HERE();
 405                 EMIT(OLPAREN, subno);
 406                 if (!SEE(')'))
 407                         p_ere(p, ')');
 408                 if (subno < NPAREN) {
 409                         p->pend[subno] = HERE();
 410                         assert(p->pend[subno] != 0);
 411                 }
 412                 EMIT(ORPAREN, subno);
 413                 MUSTEAT(')', REG_EPAREN);
 414                 break;
 415 #ifndef POSIX_MISTAKE
 416         case ')':               /* happens only if no current unmatched ( */
 417                 /*
 418                  * You may ask, why the ifndef?  Because I didn't notice
 419                  * this until slightly too late for 1003.2, and none of the
 420                  * other 1003.2 regular-expression reviewers noticed it at
 421                  * all.  So an unmatched ) is legal POSIX, at least until
 422                  * we can get it fixed.
 423                  */
 424                 SETERROR(REG_EPAREN);
 425                 break;
 426 #endif
 427         case '^':
 428                 EMIT(OBOL, 0);
 429                 p->g->iflags |= USEBOL;
 430                 p->g->nbol++;
 431                 wascaret = 1;
 432                 break;
 433         case '$':
 434                 EMIT(OEOL, 0);
 435                 p->g->iflags |= USEEOL;
 436                 p->g->neol++;
 437                 break;
 438         case '|':
 439                 SETERROR(REG_EMPTY);
 440                 break;
 441         case '*':
 442         case '+':
 443         case '?':
 444                 SETERROR(REG_BADRPT);
 445                 break;
 446         case '.':
 447                 if (p->g->cflags&REG_NEWLINE)
 448                         nonnewline(p);
 449                 else
 450                         EMIT(OANY, 0);
 451                 break;
 452         case '[':
 453                 p_bracket(p);
 454                 break;
 455         case '\\':
 456                 REQUIRE(MORE(), REG_EESCAPE);
 457                 c = GETNEXT();
 458                 ordinary(p, c);
 459                 break;
 460         case '{':               /* okay as ordinary except if digit follows */
 461                 REQUIRE(!MORE() || !isdigit((unsigned char)PEEK()), REG_BADRPT);
 462                 /* FALLTHROUGH */
 463         default:
 464                 ordinary(p, c);
 465                 break;
 466         }
 467
 468         if (!MORE())
 469                 return;
 470         c = PEEK();
 471         /* we call { a repetition if followed by a digit */
 472         if (!( c == '*' || c == '+' || c == '?' ||
 473             (c == '{' && MORE2() && isdigit((unsigned char)PEEK2())) ))
 474                 return;         /* no repetition, we're done */
 475         NEXT();
 476
 477         REQUIRE(!wascaret, REG_BADRPT);
 478         switch (c) {
 479         case '*':       /* implemented as +? */
 480                 /* this case does not require the (y|) trick, noKLUDGE */
 481                 INSERT(OPLUS_, pos);
 482                 ASTERN(O_PLUS, pos);
 483                 INSERT(OQUEST_, pos);
 484                 ASTERN(O_QUEST, pos);
 485                 break;
 486         case '+':
 487                 INSERT(OPLUS_, pos);
 488                 ASTERN(O_PLUS, pos);
 489                 break;
 490         case '?':
 491                 /* KLUDGE: emit y? as (y|) until subtle bug gets fixed */
 492                 INSERT(OCH_, pos);              /* offset slightly wrong */
 493                 ASTERN(OOR1, pos);              /* this one's right */
 494                 AHEAD(pos);                     /* fix the OCH_ */
 495                 EMIT(OOR2, 0);                  /* offset very wrong... */
 496                 AHEAD(THERE());                 /* ...so fix it */
 497                 ASTERN(O_CH, THERETHERE());
 498                 break;
 499         case '{':
 500                 count = p_count(p);
 501                 if (EAT(',')) {
 502                         if (isdigit((unsigned char)PEEK())) {
 503                                 count2 = p_count(p);
 504                                 REQUIRE(count <= count2, REG_BADBR);
 505                         } else          /* single number with comma */
 506                                 count2 = INFINITY;
 507                 } else          /* just a single number */
 508                         count2 = count;
 509                 repeat(p, pos, count, count2);
 510                 if (!EAT('}')) {        /* error heuristics */
 511                         while (MORE() && PEEK() != '}')
 512                                 NEXT();
 513                         REQUIRE(MORE(), REG_EBRACE);
 514                         SETERROR(REG_BADBR);
 515                 }
 516                 break;
 517         }
 518
 519         if (!MORE())
 520                 return;
 521         c = PEEK();
 522         if (!( c == '*' || c == '+' || c == '?' ||
 523             (c == '{' && MORE2() && isdigit((unsigned char)PEEK2())) ) )
 524                 return;
 525         SETERROR(REG_BADRPT);
 526 }
 527
 528 /*
 529  - p_str - string (no metacharacters) "parser"
 530  == static void p_str(struct parse *p);
 531  */
 532 static void
 533 p_str(
 534     struct parse *p)
 535 {
 536
 537         _DIAGASSERT(p != NULL);
 538
 539         REQUIRE(MORE(), REG_EMPTY);
 540         while (MORE())
 541                 ordinary(p, GETNEXT());
 542 }
 543
 544 /*
 545  - p_bre - BRE parser top level, anchoring and concatenation
 546  == static void p_bre(struct parse *p, int end1, \
 547  ==     int end2);
 548  * Giving end1 as OUT essentially eliminates the end1/end2 check.
 549  *
 550  * This implementation is a bit of a kludge, in that a trailing $ is first
 551  * taken as an ordinary character and then revised to be an anchor.  The
 552  * only undesirable side effect is that '$' gets included as a character
 553  * category in such cases.  This is fairly harmless; not worth fixing.
 554  * The amount of lookahead needed to avoid this kludge is excessive.
 555  */
 556 static void
 557 p_bre(
 558     struct parse *p,
 559     int end1,           /* first terminating character */
 560     int end2)           /* second terminating character */
 561 {
 562         sopno start;
 563         int first = 1;                  /* first subexpression? */
 564         int wasdollar = 0;
 565
 566         _DIAGASSERT(p != NULL);
 567
 568         start = HERE();
 569
 570         if (EAT('^')) {
 571                 EMIT(OBOL, 0);
 572                 p->g->iflags |= USEBOL;
 573                 p->g->nbol++;
 574         }
 575         while (MORE() && !SEETWO(end1, end2)) {
 576                 wasdollar = p_simp_re(p, first);
 577                 first = 0;
 578         }
 579         if (wasdollar) {        /* oops, that was a trailing anchor */
 580                 DROP(1);
 581                 EMIT(OEOL, 0);
 582                 p->g->iflags |= USEEOL;
 583                 p->g->neol++;
 584         }
 585
 586         REQUIRE(HERE() != start, REG_EMPTY);    /* require nonempty */
 587 }
 588
 589 /*
 590  - p_simp_re - parse a simple RE, an atom possibly followed by a repetition
 591  == static int p_simp_re(struct parse *p, int starordinary);
 592  */
 593 static int                      /* was the simple RE an unbackslashed $? */
 594 p_simp_re(
 595     struct parse *p,
 596     int starordinary)           /* is a leading * an ordinary character? */
 597 {
 598         int c;
 599         int count;
 600         int count2;
 601         sopno pos;
 602         int i;
 603         sopno subno;
 604 #       define  BACKSL  (1<<CHAR_BIT)
 605
 606         _DIAGASSERT(p != NULL);
 607
 608         pos = HERE();           /* repetion op, if any, covers from here */
 609
 610         assert(MORE());         /* caller should have ensured this */
 611         c = GETNEXT();
 612         if (c == '\\') {
 613                 REQUIRE(MORE(), REG_EESCAPE);
 614                 c = BACKSL | (unsigned char)GETNEXT();
 615         }
 616         switch (c) {
 617         case '.':
 618                 if (p->g->cflags&REG_NEWLINE)
 619                         nonnewline(p);
 620                 else
 621                         EMIT(OANY, 0);
 622                 break;
 623         case '[':
 624                 p_bracket(p);
 625                 break;
 626         case BACKSL|'{':
 627                 SETERROR(REG_BADRPT);
 628                 break;
 629         case BACKSL|'(':
 630                 p->g->nsub++;
 631                 subno = p->g->nsub;
 632                 if (subno < NPAREN)
 633                         p->pbegin[subno] = HERE();
 634                 EMIT(OLPAREN, subno);
 635                 /* the MORE here is an error heuristic */
 636                 if (MORE() && !SEETWO('\\', ')'))
 637                         p_bre(p, '\\', ')');
 638                 if (subno < NPAREN) {
 639                         p->pend[subno] = HERE();
 640                         assert(p->pend[subno] != 0);
 641                 }
 642                 EMIT(ORPAREN, subno);
 643                 REQUIRE(EATTWO('\\', ')'), REG_EPAREN);
 644                 break;
 645         case BACKSL|')':        /* should not get here -- must be user */
 646         case BACKSL|'}':
 647                 SETERROR(REG_EPAREN);
 648                 break;
 649         case BACKSL|'1':
 650         case BACKSL|'2':
 651         case BACKSL|'3':
 652         case BACKSL|'4':
 653         case BACKSL|'5':
 654         case BACKSL|'6':
 655         case BACKSL|'7':
 656         case BACKSL|'8':
 657         case BACKSL|'9':
 658                 i = (c&~BACKSL) - '0';
 659                 assert(i < NPAREN);
 660                 if (p->pend[i] != 0) {
 661                         assert(i <= p->g->nsub);
 662                         EMIT(OBACK_, i);
 663                         assert(p->pbegin[i] != 0);
 664                         assert(OP(p->strip[p->pbegin[i]]) == OLPAREN);
 665                         assert(OP(p->strip[p->pend[i]]) == ORPAREN);
 666                         (void) dupl(p, p->pbegin[i]+1, p->pend[i]);
 667                         EMIT(O_BACK, i);
 668                 } else
 669                         SETERROR(REG_ESUBREG);
 670                 p->g->backrefs = 1;
 671                 break;
 672         case '*':
 673                 REQUIRE(starordinary, REG_BADRPT);
 674                 /* FALLTHROUGH */
 675         default:
 676                 ordinary(p, c &~ BACKSL);
 677                 break;
 678         }
 679
 680         if (EAT('*')) {         /* implemented as +? */
 681                 /* this case does not require the (y|) trick, noKLUDGE */
 682                 INSERT(OPLUS_, pos);
 683                 ASTERN(O_PLUS, pos);
 684                 INSERT(OQUEST_, pos);
 685                 ASTERN(O_QUEST, pos);
 686         } else if (EATTWO('\\', '{')) {
 687                 count = p_count(p);
 688                 if (EAT(',')) {
 689                         if (MORE() && isdigit((unsigned char)PEEK())) {
 690                                 count2 = p_count(p);
 691                                 REQUIRE(count <= count2, REG_BADBR);
 692                         } else          /* single number with comma */
 693                                 count2 = INFINITY;
 694                 } else          /* just a single number */
 695                         count2 = count;
 696                 repeat(p, pos, count, count2);
 697                 if (!EATTWO('\\', '}')) {       /* error heuristics */
 698                         while (MORE() && !SEETWO('\\', '}'))
 699                                 NEXT();
 700                         REQUIRE(MORE(), REG_EBRACE);
 701                         SETERROR(REG_BADBR);
 702                 }
 703         } else if (c == (unsigned char)'$')     /* $ (but not \$) ends it */
 704                 return(1);
 705
 706         return(0);
 707 }
 708
 709 /*
 710  - p_count - parse a repetition count
 711  == static int p_count(struct parse *p);
 712  */
 713 static int                      /* the value */
 714 p_count(
 715     struct parse *p)
 716 {
 717         int count = 0;
 718         int ndigits = 0;
 719
 720         _DIAGASSERT(p != NULL);
 721
 722         while (MORE() && isdigit((unsigned char)PEEK()) && count <= DUPMAX) {
 723                 count = count*10 + (GETNEXT() - '0');
 724                 ndigits++;
 725         }
 726
 727         REQUIRE(ndigits > 0 && count <= DUPMAX, REG_BADBR);
 728         return(count);
 729 }
 730
 731 /*
 732  - p_bracket - parse a bracketed character list
 733  == static void p_bracket(struct parse *p);
 734  *
 735  * Note a significant property of this code:  if the allocset() did SETERROR,
 736  * no set operations are done.
 737  */
 738 static void
 739 p_bracket(
 740     struct parse *p)
 741 {
 742         cset *cs;
 743         int invert = 0;
 744
 745         _DIAGASSERT(p != NULL);
 746
 747         cs = allocset(p);
 748
 749         /* Dept of Truly Sickening Special-Case Kludges */
 750         if (p->next + 5 < p->end && strncmp(p->next, "[:<:]]",
 751                                             (size_t)6) == 0) {
 752                 EMIT(OBOW, 0);
 753                 NEXTn(6);
 754                 return;
 755         }
 756         if (p->next + 5 < p->end && strncmp(p->next, "[:>:]]",
 757                                             (size_t)6) == 0) {
 758                 EMIT(OEOW, 0);
 759                 NEXTn(6);
 760                 return;
 761         }
 762
 763         if (EAT('^'))
 764                 invert++;       /* make note to invert set at end */
 765         if (EAT(']'))
 766                 CHadd(cs, ']');
 767         else if (EAT('-'))
 768                 CHadd(cs, '-');
 769         while (MORE() && PEEK() != ']' && !SEETWO('-', ']'))
 770                 p_b_term(p, cs);
 771         if (EAT('-'))
 772                 CHadd(cs, '-');
 773         MUSTEAT(']', REG_EBRACK);
 774
 775         if (p->error != 0)      /* don't mess things up further */
 776                 return;
 777
 778         if (p->g->cflags&REG_ICASE) {
 779                 int i;
 780                 int ci;
 781
 782                 for (i = p->g->csetsize - 1; i >= 0; i--)
 783                         if (CHIN(cs, i) && isalpha(i)) {
 784                                 ci = othercase(i);
 785                                 if (ci != i)
 786                                         CHadd(cs, ci);
 787                         }
 788                 if (cs->multis != NULL)
 789                         mccase(p, cs);
 790         }
 791         if (invert) {
 792                 int i;
 793
 794                 for (i = p->g->csetsize - 1; i >= 0; i--)
 795                         if (CHIN(cs, i))
 796                                 CHsub(cs, i);
 797                         else
 798                                 CHadd(cs, i);
 799                 if (p->g->cflags&REG_NEWLINE)
 800                         CHsub(cs, '\n');
 801                 if (cs->multis != NULL)
 802                         mcinvert(p, cs);
 803         }
 804
 805         assert(cs->multis == NULL);             /* xxx */
 806
 807         if (nch(p, cs) == 1) {          /* optimize singleton sets */
 808                 ordinary(p, firstch(p, cs));
 809                 freeset(p, cs);
 810         } else
 811                 EMIT(OANYOF, freezeset(p, cs));
 812 }
 813
 814 /*
 815  - p_b_term - parse one term of a bracketed character list
 816  == static void p_b_term(struct parse *p, cset *cs);
 817  */
 818 static void
 819 p_b_term(
 820     struct parse *p,
 821     cset *cs)
 822 {
 823         char c;
 824         char start, finish;
 825         int i;
 826
 827         _DIAGASSERT(p != NULL);
 828         _DIAGASSERT(cs != NULL);
 829
 830         /* classify what we've got */
 831         switch ((MORE()) ? PEEK() : '\0') {
 832         case '[':
 833                 c = (MORE2()) ? PEEK2() : '\0';
 834                 break;
 835
 836         case '-':
 837                 SETERROR(REG_ERANGE);
 838                 return;                 /* NOTE RETURN */
 839
 840         default:
 841                 c = '\0';
 842                 break;
 843         }
 844
 845         switch (c) {
 846         case ':':               /* character class */
 847                 NEXT2();
 848                 REQUIRE(MORE(), REG_EBRACK);
 849                 c = PEEK();
 850                 REQUIRE(c != '-' && c != ']', REG_ECTYPE);
 851                 p_b_cclass(p, cs);
 852                 REQUIRE(MORE(), REG_EBRACK);
 853                 REQUIRE(EATTWO(':', ']'), REG_ECTYPE);
 854                 break;
 855         case '=':               /* equivalence class */
 856                 NEXT2();
 857                 REQUIRE(MORE(), REG_EBRACK);
 858                 c = PEEK();
 859                 REQUIRE(c != '-' && c != ']', REG_ECOLLATE);
 860                 p_b_eclass(p, cs);
 861                 REQUIRE(MORE(), REG_EBRACK);
 862                 REQUIRE(EATTWO('=', ']'), REG_ECOLLATE);
 863                 break;
 864         default:                /* symbol, ordinary character, or range */
 865 /* xxx revision needed for multichar stuff */
 866                 start = p_b_symbol(p);
 867                 if (SEE('-') && MORE2() && PEEK2() != ']') {
 868                         /* range */
 869                         NEXT();
 870                         if (EAT('-'))
 871                                 finish = '-';
 872                         else
 873                                 finish = p_b_symbol(p);
 874                 } else
 875                         finish = start;
 876 /* xxx what about signed chars here... */
 877                 REQUIRE(start <= finish, REG_ERANGE);
 878                 for (i = start; i <= finish; i++)
 879                         CHadd(cs, i);
 880                 break;
 881         }
 882 }
 883
 884 /*
 885  - p_b_cclass - parse a character-class name and deal with it
 886  == static void p_b_cclass(struct parse *p, cset *cs);
 887  */
 888 static void
 889 p_b_cclass(
 890     struct parse *p,
 891     cset *cs)
 892 {
 893         const char *sp;
 894         const struct cclass *cp;
 895         size_t len;
 896         const char *u;
 897         char c;
 898
 899         _DIAGASSERT(p != NULL);
 900         _DIAGASSERT(cs != NULL);
 901
 902         sp = p->next;
 903
 904         while (MORE() && isalpha((unsigned char)PEEK()))
 905                 NEXT();
 906         len = p->next - sp;
 907         for (cp = cclasses; cp->name != NULL; cp++)
 908                 if (strncmp(cp->name, sp, len) == 0 && cp->name[len] == '\0')
 909                         break;
 910         if (cp->name == NULL) {
 911                 /* oops, didn't find it */
 912                 SETERROR(REG_ECTYPE);
 913                 return;
 914         }
 915
 916         u = cp->chars;
 917         while ((c = *u++) != '\0')
 918                 CHadd(cs, c);
 919         for (u = cp->multis; *u != '\0'; u += strlen(u) + 1)
 920                 MCadd(p, cs, u);
 921 }
 922
 923 /*
 924  - p_b_eclass - parse an equivalence-class name and deal with it
 925  == static void p_b_eclass(struct parse *p, cset *cs);
 926  *
 927  * This implementation is incomplete. xxx
 928  */
 929 static void
 930 p_b_eclass(
 931     struct parse *p,
 932     cset *cs)
 933 {
 934         char c;
 935
 936         _DIAGASSERT(p != NULL);
 937         _DIAGASSERT(cs != NULL);
 938
 939         c = p_b_coll_elem(p, '=');
 940         CHadd(cs, c);
 941 }
 942
 943 /*
 944  - p_b_symbol - parse a character or [..]ed multicharacter collating symbol
 945  == static char p_b_symbol(struct parse *p);
 946  */
 947 static char                     /* value of symbol */
 948 p_b_symbol(
 949     struct parse *p)
 950 {
 951         char value;
 952
 953         _DIAGASSERT(p != NULL);
 954
 955         REQUIRE(MORE(), REG_EBRACK);
 956         if (!EATTWO('[', '.'))
 957                 return(GETNEXT());
 958
 959         /* collating symbol */
 960         value = p_b_coll_elem(p, '.');
 961         REQUIRE(EATTWO('.', ']'), REG_ECOLLATE);
 962         return(value);
 963 }
 964
 965 /*
 966  - p_b_coll_elem - parse a collating-element name and look it up
 967  == static char p_b_coll_elem(struct parse *p, int endc);
 968  */
 969 static char                     /* value of collating element */
 970 p_b_coll_elem(
 971     struct parse *p,
 972     int endc)                   /* name ended by endc,']' */
 973 {
 974         const char *sp;
 975         const struct cname *cp;
 976         size_t len;
 977
 978         _DIAGASSERT(p != NULL);
 979
 980         sp = p->next;
 981
 982         while (MORE() && !SEETWO(endc, ']'))
 983                 NEXT();
 984         if (!MORE()) {
 985                 SETERROR(REG_EBRACK);
 986                 return(0);
 987         }
 988         len = p->next - sp;
 989         for (cp = cnames; cp->name != NULL; cp++)
 990                 if (strncmp(cp->name, sp, len) == 0 && cp->name[len] == '\0')
 991                         return(cp->code);       /* known name */
 992         if (len == 1)
 993                 return(*sp);    /* single character */
 994         SETERROR(REG_ECOLLATE);                 /* neither */
 995         return(0);
 996 }
 997
 998 /*
 999  - othercase - return the case counterpart of an alphabetic
1000  == static int othercase(int ch);
1001  */
1002 static int                      /* if no counterpart, return ch */
1003 othercase(
1004     int ch)
1005 {
1006         assert(isalpha(ch));
1007         if (isupper(ch))
1008                 return(tolower(ch));
1009         else if (islower(ch))
1010                 return(toupper(ch));
1011         else                    /* peculiar, but could happen */
1012                 return(ch);
1013 }
1014
1015 /*
1016  - bothcases - emit a dualcase version of a two-case character
1017  == static void bothcases(struct parse *p, int ch);
1018  *
1019  * Boy, is this implementation ever a kludge...
1020  */
1021 static void
1022 bothcases(
1023     struct parse *p,
1024     int ch)
1025 {
1026         const char *oldnext;
1027         const char *oldend;
1028         char bracket[3];
1029
1030         _DIAGASSERT(p != NULL);
1031
1032         oldnext = p->next;
1033         oldend = p->end;
1034
1035         assert(othercase(ch) != ch);    /* p_bracket() would recurse */
1036         p->next = bracket;
1037         p->end = bracket+2;
1038         bracket[0] = ch;
1039         bracket[1] = ']';
1040         bracket[2] = '\0';
1041         p_bracket(p);
1042         assert(p->next == bracket+2);
1043         p->next = oldnext;
1044         p->end = oldend;
1045 }
1046
1047 /*
1048  - ordinary - emit an ordinary character
1049  == static void ordinary(struct parse *p, int ch);
1050  */
1051 static void
1052 ordinary(
1053     struct parse *p,
1054     int ch)
1055 {
1056         cat_t *cap;
1057
1058         _DIAGASSERT(p != NULL);
1059
1060         cap = p->g->categories;
1061         if ((p->g->cflags&REG_ICASE) && isalpha((unsigned char) ch)
1062             && othercase((unsigned char) ch) != (unsigned char) ch)
1063                 bothcases(p, (unsigned char) ch);
1064         else {
1065                 EMIT(OCHAR, (unsigned char)ch);
1066                 if (cap[ch] == 0)
1067                         cap[ch] = p->g->ncategories++;
1068         }
1069 }
1070
1071 /*
1072  - nonnewline - emit REG_NEWLINE version of OANY
1073  == static void nonnewline(struct parse *p);
1074  *
1075  * Boy, is this implementation ever a kludge...
1076  */
1077 static void
1078 nonnewline(
1079     struct parse *p)
1080 {
1081         const char *oldnext;
1082         const char *oldend;
1083         char bracket[4];
1084
1085         _DIAGASSERT(p != NULL);
1086
1087         oldnext = p->next;
1088         oldend = p->end;
1089
1090         p->next = bracket;
1091         p->end = bracket+3;
1092         bracket[0] = '^';
1093         bracket[1] = '\n';
1094         bracket[2] = ']';
1095         bracket[3] = '\0';
1096         p_bracket(p);
1097         assert(p->next == bracket+3);
1098         p->next = oldnext;
1099         p->end = oldend;
1100 }
1101
1102 /*
1103  - repeat - generate code for a bounded repetition, recursively if needed
1104  == static void repeat(struct parse *p, sopno start, int from, int to);
1105  */
1106 static void
1107 repeat(
1108     struct parse *p,
1109     sopno start,                /* operand from here to end of strip */
1110     int from,                   /* repeated from this number */
1111     int to)                     /* to this number of times (maybe INFINITY) */
1112 {
1113         sopno finish;
1114 #       define  N       2
1115 #       define  INF     3
1116 #       define  REP(f, t)       ((f)*8 + (t))
1117 #       define  MAP(n)  (((n) <= 1) ? (n) : ((n) == INFINITY) ? INF : N)
1118         sopno copy;
1119
1120         _DIAGASSERT(p != NULL);
1121
1122         finish = HERE();
1123
1124         if (p->error != 0)      /* head off possible runaway recursion */
1125                 return;
1126
1127         assert(from <= to);
1128
1129         switch (REP(MAP(from), MAP(to))) {
1130         case REP(0, 0):                 /* must be user doing this */
1131                 DROP(finish-start);     /* drop the operand */
1132                 break;
1133         case REP(0, 1):                 /* as x{1,1}? */
1134         case REP(0, N):                 /* as x{1,n}? */
1135         case REP(0, INF):               /* as x{1,}? */
1136                 /* KLUDGE: emit y? as (y|) until subtle bug gets fixed */
1137                 INSERT(OCH_, start);            /* offset is wrong... */
1138                 repeat(p, start+1, 1, to);
1139                 ASTERN(OOR1, start);
1140                 AHEAD(start);                   /* ... fix it */
1141                 EMIT(OOR2, 0);
1142                 AHEAD(THERE());
1143                 ASTERN(O_CH, THERETHERE());
1144                 break;
1145         case REP(1, 1):                 /* trivial case */
1146                 /* done */
1147                 break;
1148         case REP(1, N):                 /* as x?x{1,n-1} */
1149                 /* KLUDGE: emit y? as (y|) until subtle bug gets fixed */
1150                 INSERT(OCH_, start);
1151                 ASTERN(OOR1, start);
1152                 AHEAD(start);
1153                 EMIT(OOR2, 0);                  /* offset very wrong... */
1154                 AHEAD(THERE());                 /* ...so fix it */
1155                 ASTERN(O_CH, THERETHERE());
1156                 copy = dupl(p, start+1, finish+1);
1157                 assert(copy == finish+4);
1158                 repeat(p, copy, 1, to-1);
1159                 break;
1160         case REP(1, INF):               /* as x+ */
1161                 INSERT(OPLUS_, start);
1162                 ASTERN(O_PLUS, start);
1163                 break;
1164         case REP(N, N):                 /* as xx{m-1,n-1} */
1165                 copy = dupl(p, start, finish);
1166                 repeat(p, copy, from-1, to-1);
1167                 break;
1168         case REP(N, INF):               /* as xx{n-1,INF} */
1169                 copy = dupl(p, start, finish);
1170                 repeat(p, copy, from-1, to);
1171                 break;
1172         default:                        /* "can't happen" */
1173                 SETERROR(REG_ASSERT);   /* just in case */
1174                 break;
1175         }
1176 }
1177
1178 /*
1179  - seterr - set an error condition
1180  == static int seterr(struct parse *p, int e);
1181  */
1182 static int                      /* useless but makes type checking happy */
1183 seterr(
1184     struct parse *p,
1185     int e)
1186 {
1187
1188         _DIAGASSERT(p != NULL);
1189
1190         if (p->error == 0)      /* keep earliest error condition */
1191                 p->error = e;
1192         p->next = nuls;         /* try to bring things to a halt */
1193         p->end = nuls;
1194         return(0);              /* make the return value well-defined */
1195 }
1196
1197 /*
1198  - allocset - allocate a set of characters for []
1199  == static cset *allocset(struct parse *p);
1200  */
1201 static cset *
1202 allocset(
1203     struct parse *p)
1204 {
1205         int no;
1206         size_t nc;
1207         size_t nbytes;
1208         cset *cs;
1209         size_t css;
1210         int i;
1211
1212         _DIAGASSERT(p != NULL);
1213
1214         no = p->g->ncsets++;
1215         css = (size_t)p->g->csetsize;
1216         if (no >= p->ncsalloc) {        /* need another column of space */
1217                 p->ncsalloc += CHAR_BIT;
1218                 nc = p->ncsalloc;
1219                 assert(nc % CHAR_BIT == 0);
1220                 nbytes = nc / CHAR_BIT * css;
1221                 if (p->g->sets == NULL)
1222                         p->g->sets = malloc(nc * sizeof(cset));
1223                 else
1224                         p->g->sets = realloc(p->g->sets, nc * sizeof(cset));
1225                 if (p->g->setbits == NULL)
1226                         p->g->setbits = malloc(nbytes);
1227                 else {
1228                         p->g->setbits = realloc(p->g->setbits, nbytes);
1229                         /* xxx this isn't right if setbits is now NULL */
1230                         for (i = 0; i < no; i++)
1231                                 p->g->sets[i].ptr = p->g->setbits + css*(i/CHAR_BIT);
1232                 }
1233                 if (p->g->sets != NULL && p->g->setbits != NULL)
1234                         (void) memset((char *)p->g->setbits + (nbytes - css),
1235                                                                 0, css);
1236                 else {
1237                         no = 0;
1238                         SETERROR(REG_ESPACE);
1239                         /* caller's responsibility not to do set ops */
1240                 }
1241         }
1242
1243         assert(p->g->sets != NULL);     /* xxx */
1244         cs = &p->g->sets[no];
1245         cs->ptr = p->g->setbits + css*((no)/CHAR_BIT);
1246         cs->mask = 1 << ((no) % CHAR_BIT);
1247         cs->hash = 0;
1248         cs->smultis = 0;
1249         cs->multis = NULL;
1250
1251         return(cs);
1252 }
1253
1254 /*
1255  - freeset - free a now-unused set
1256  == static void freeset(struct parse *p, cset *cs);
1257  */
1258 static void
1259 freeset(
1260     struct parse *p,
1261     cset *cs)
1262 {
1263         size_t i;
1264         cset *top;
1265         size_t css;
1266
1267         _DIAGASSERT(p != NULL);
1268         _DIAGASSERT(cs != NULL);
1269
1270         top = &p->g->sets[p->g->ncsets];
1271         css = (size_t)p->g->csetsize;
1272
1273         for (i = 0; i < css; i++)
1274                 CHsub(cs, i);
1275         if (cs == top-1)        /* recover only the easy case */
1276                 p->g->ncsets--;
1277 }
1278
1279 /*
1280  - freezeset - final processing on a set of characters
1281  == static int freezeset(struct parse *p, cset *cs);
1282  *
1283  * The main task here is merging identical sets.  This is usually a waste
1284  * of time (although the hash code minimizes the overhead), but can win
1285  * big if REG_ICASE is being used.  REG_ICASE, by the way, is why the hash
1286  * is done using addition rather than xor -- all ASCII [aA] sets xor to
1287  * the same value!
1288  */
1289 static int                      /* set number */
1290 freezeset(
1291     struct parse *p,
1292     cset *cs)
1293 {
1294         uch h;
1295         size_t i;
1296         cset *top;
1297         cset *cs2;
1298         size_t css;
1299
1300         _DIAGASSERT(p != NULL);
1301         _DIAGASSERT(cs != NULL);
1302
1303         h = cs->hash;
1304         top = &p->g->sets[p->g->ncsets];
1305         css = (size_t)p->g->csetsize;
1306
1307         /* look for an earlier one which is the same */
1308         for (cs2 = &p->g->sets[0]; cs2 < top; cs2++)
1309                 if (cs2->hash == h && cs2 != cs) {
1310                         /* maybe */
1311                         for (i = 0; i < css; i++)
1312                                 if (!!CHIN(cs2, i) != !!CHIN(cs, i))
1313                                         break;          /* no */
1314                         if (i == css)
1315                                 break;                  /* yes */
1316                 }
1317
1318         if (cs2 < top) {        /* found one */
1319                 freeset(p, cs);
1320                 cs = cs2;
1321         }
1322
1323         return((int)(cs - p->g->sets));
1324 }
1325
1326 /*
1327  - firstch - return first character in a set (which must have at least one)
1328  == static int firstch(struct parse *p, cset *cs);
1329  */
1330 static int                      /* character; there is no "none" value */
1331 firstch(
1332     struct parse *p,
1333     cset *cs)
1334 {
1335         size_t i;
1336         size_t css;
1337
1338         _DIAGASSERT(p != NULL);
1339         _DIAGASSERT(cs != NULL);
1340
1341         css = (size_t)p->g->csetsize;
1342
1343         for (i = 0; i < css; i++)
1344                 if (CHIN(cs, i))
1345                         return((char)i);
1346         assert(never);
1347         return(0);              /* arbitrary */
1348 }
1349
1350 /*
1351  - nch - number of characters in a set
1352  == static int nch(struct parse *p, cset *cs);
1353  */
1354 static int
1355 nch(
1356     struct parse *p,
1357     cset *cs)
1358 {
1359         size_t i;
1360         size_t css;
1361         int n = 0;
1362
1363         _DIAGASSERT(p != NULL);
1364         _DIAGASSERT(cs != NULL);
1365
1366         css = (size_t)p->g->csetsize;
1367
1368         for (i = 0; i < css; i++)
1369                 if (CHIN(cs, i))
1370                         n++;
1371         return(n);
1372 }
1373
1374 /*
1375  - mcadd - add a collating element to a cset
1376  == static void mcadd(struct parse *p, cset *cs, \
1377  ==     char *cp);
1378  */
1379 static void
1380 mcadd(
1381     struct parse *p,
1382     cset *cs,
1383     const char *cp)
1384 {
1385         size_t oldend;
1386
1387         _DIAGASSERT(p != NULL);
1388         _DIAGASSERT(cs != NULL);
1389         _DIAGASSERT(cp != NULL);
1390
1391         oldend = cs->smultis;
1392
1393         cs->smultis += strlen(cp) + 1;
1394         if (cs->multis == NULL)
1395                 cs->multis = malloc(cs->smultis);
1396         else
1397                 cs->multis = realloc(cs->multis, cs->smultis);
1398         if (cs->multis == NULL) {
1399                 SETERROR(REG_ESPACE);
1400                 return;
1401         }
1402
1403         (void) strcpy(cs->multis + oldend - 1, cp);
1404         cs->multis[cs->smultis - 1] = '\0';
1405 }
1406
1407 #if 0
1408 /*
1409  - mcsub - subtract a collating element from a cset
1410  == static void mcsub(cset *cs, char *cp);
1411  */
1412 static void
1413 mcsub(
1414     cset *cs,
1415     char *cp)
1416 {
1417         char *fp;
1418         size_t len;
1419
1420         _DIAGASSERT(cs != NULL);
1421         _DIAGASSERT(cp != NULL);
1422
1423         fp = mcfind(cs, cp);
1424         len = strlen(fp);
1425
1426         assert(fp != NULL);
1427         (void) memmove(fp, fp + len + 1,
1428                                 cs->smultis - (fp + len + 1 - cs->multis));
1429         cs->smultis -= len;
1430
1431         if (cs->smultis == 0) {
1432                 free(cs->multis);
1433                 cs->multis = NULL;
1434                 return;
1435         }
1436
1437         cs->multis = realloc(cs->multis, cs->smultis);
1438         assert(cs->multis != NULL);
1439 }
1440
1441 /*
1442  - mcin - is a collating element in a cset?
1443  == static int mcin(cset *cs, char *cp);
1444  */
1445 static int
1446 mcin(
1447     cset *cs,
1448     char *cp)
1449 {
1450
1451         _DIAGASSERT(cs != NULL);
1452         _DIAGASSERT(cp != NULL);
1453
1454         return(mcfind(cs, cp) != NULL);
1455 }
1456
1457 /*
1458  - mcfind - find a collating element in a cset
1459  == static char *mcfind(cset *cs, char *cp);
1460  */
1461 static char *
1462 mcfind(
1463     cset *cs,
1464     char *cp)
1465 {
1466         char *p;
1467
1468         _DIAGASSERT(cs != NULL);
1469         _DIAGASSERT(cp != NULL);
1470
1471         if (cs->multis == NULL)
1472                 return(NULL);
1473         for (p = cs->multis; *p != '\0'; p += strlen(p) + 1)
1474                 if (strcmp(cp, p) == 0)
1475                         return(p);
1476         return(NULL);
1477 }
1478 #endif
1479
1480 /*
1481  - mcinvert - invert the list of collating elements in a cset
1482  == static void mcinvert(struct parse *p, cset *cs);
1483  *
1484  * This would have to know the set of possibilities.  Implementation
1485  * is deferred.
1486  */
1487 /* ARGSUSED */
1488 static void
1489 mcinvert(
1490     struct parse *p,
1491     cset *cs)
1492 {
1493
1494         _DIAGASSERT(p != NULL);
1495         _DIAGASSERT(cs != NULL);
1496
1497         assert(cs->multis == NULL);     /* xxx */
1498 }
1499
1500 /*
1501  - mccase - add case counterparts of the list of collating elements in a cset
1502  == static void mccase(struct parse *p, cset *cs);
1503  *
1504  * This would have to know the set of possibilities.  Implementation
1505  * is deferred.
1506  */
1507 /* ARGSUSED */
1508 static void
1509 mccase(
1510     struct parse *p,
1511     cset *cs)
1512 {
1513
1514         _DIAGASSERT(p != NULL);
1515         _DIAGASSERT(cs != NULL);
1516
1517         assert(cs->multis == NULL);     /* xxx */
1518 }
1519
1520 /*
1521  - isinsets - is this character in any sets?
1522  == static int isinsets(struct re_guts *g, int c);
1523  */
1524 static int                      /* predicate */
1525 isinsets(
1526     struct re_guts *g,
1527     int c)
1528 {
1529         uch *col;
1530         int i;
1531         int ncols;
1532         unsigned uc = (unsigned char)c;
1533
1534         _DIAGASSERT(g != NULL);
1535
1536         ncols = (g->ncsets+(CHAR_BIT-1)) / CHAR_BIT;
1537
1538         for (i = 0, col = g->setbits; i < ncols; i++, col += g->csetsize)
1539                 if (col[uc] != 0)
1540                         return(1);
1541         return(0);
1542 }
1543
1544 /*
1545  - samesets - are these two characters in exactly the same sets?
1546  == static int samesets(struct re_guts *g, int c1, int c2);
1547  */
1548 static int                      /* predicate */
1549 samesets(
1550     struct re_guts *g,
1551     int c1,
1552     int c2)
1553 {
1554         uch *col;
1555         int i;
1556         int ncols;
1557         unsigned uc1 = (unsigned char)c1;
1558         unsigned uc2 = (unsigned char)c2;
1559
1560         _DIAGASSERT(g != NULL);
1561
1562         ncols = (g->ncsets+(CHAR_BIT-1)) / CHAR_BIT;
1563
1564         for (i = 0, col = g->setbits; i < ncols; i++, col += g->csetsize)
1565                 if (col[uc1] != col[uc2])
1566                         return(0);
1567         return(1);
1568 }
1569
1570 /*
1571  - categorize - sort out character categories
1572  == static void categorize(struct parse *p, struct re_guts *g);
1573  */
1574 static void
1575 categorize(
1576     struct parse *p,
1577     struct re_guts *g)
1578 {
1579         cat_t *cats;
1580         int c;
1581         int c2;
1582         cat_t cat;
1583
1584         _DIAGASSERT(p != NULL);
1585         _DIAGASSERT(g != NULL);
1586
1587         cats = g->categories;
1588
1589         /* avoid making error situations worse */
1590         if (p->error != 0)
1591                 return;
1592
1593         for (c = CHAR_MIN; c <= CHAR_MAX; c++)
1594                 if (cats[c] == 0 && isinsets(g, c)) {
1595                         cat = g->ncategories++;
1596                         cats[c] = cat;
1597                         for (c2 = c+1; c2 <= CHAR_MAX; c2++)
1598                                 if (cats[c2] == 0 && samesets(g, c, c2))
1599                                         cats[c2] = cat;
1600                 }
1601 }
1602
1603 /*
1604  - dupl - emit a duplicate of a bunch of sops
1605  == static sopno dupl(struct parse *p, sopno start, sopno finish);
1606  */
1607 static sopno                    /* start of duplicate */
1608 dupl(
1609     struct parse *p,
1610     sopno start,                        /* from here */
1611     sopno finish)                       /* to this less one */
1612 {
1613         sopno ret;
1614         sopno len = finish - start;
1615
1616         _DIAGASSERT(p != NULL);
1617
1618         ret = HERE();
1619
1620         assert(finish >= start);
1621         if (len == 0)
1622                 return(ret);
1623         enlarge(p, p->ssize + len);     /* this many unexpected additions */
1624         assert(p->ssize >= p->slen + len);
1625         (void)memcpy(p->strip + p->slen, p->strip + start,
1626             (size_t)len * sizeof(sop));
1627         p->slen += len;
1628         return(ret);
1629 }
1630
1631 /*
1632  - doemit - emit a strip operator
1633  == static void doemit(struct parse *p, sop op, size_t opnd);
1634  *
1635  * It might seem better to implement this as a macro with a function as
1636  * hard-case backup, but it's just too big and messy unless there are
1637  * some changes to the data structures.  Maybe later.
1638  */
1639 static void
1640 doemit(
1641     struct parse *p,
1642     sop op,
1643     sopno opnd)
1644 {
1645
1646         _DIAGASSERT(p != NULL);
1647
1648         /* avoid making error situations worse */
1649         if (p->error != 0)
1650                 return;
1651
1652         /* deal with oversize operands ("can't happen", more or less) */
1653         assert(opnd < 1<<OPSHIFT);
1654
1655         /* deal with undersized strip */
1656         if (p->slen >= p->ssize)
1657                 enlarge(p, (p->ssize+1) / 2 * 3);       /* +50% */
1658         assert(p->slen < p->ssize);
1659
1660         /* finally, it's all reduced to the easy case */
1661         p->strip[p->slen++] = SOP(op, opnd);
1662 }
1663
1664 /*
1665  - doinsert - insert a sop into the strip
1666  == static void doinsert(struct parse *p, sop op, size_t opnd, sopno pos);
1667  */
1668 static void
1669 doinsert(
1670     struct parse *p,
1671     sop op,
1672     sopno opnd,
1673     sopno pos)
1674 {
1675         sopno sn;
1676         sop s;
1677         int i;
1678
1679         _DIAGASSERT(p != NULL);
1680
1681         /* avoid making error situations worse */
1682         if (p->error != 0)
1683                 return;
1684
1685         sn = HERE();
1686         EMIT(op, opnd);         /* do checks, ensure space */
1687         assert(HERE() == sn+1);
1688         s = p->strip[sn];
1689
1690         /* adjust paren pointers */
1691         assert(pos > 0);
1692         for (i = 1; i < NPAREN; i++) {
1693                 if (p->pbegin[i] >= pos) {
1694                         p->pbegin[i]++;
1695                 }
1696                 if (p->pend[i] >= pos) {
1697                         p->pend[i]++;
1698                 }
1699         }
1700
1701         memmove(&p->strip[pos+1], &p->strip[pos], (HERE()-pos-1)*sizeof(sop));
1702         p->strip[pos] = s;
1703 }
1704
1705 /*
1706  - dofwd - complete a forward reference
1707  == static void dofwd(struct parse *p, sopno pos, sop value);
1708  */
1709 static void
1710 dofwd(
1711     struct parse *p,
1712     sopno pos,
1713     sopno value)
1714 {
1715
1716         _DIAGASSERT(p != NULL);
1717
1718         /* avoid making error situations worse */
1719         if (p->error != 0)
1720                 return;
1721
1722         assert(value < 1<<OPSHIFT);
1723         p->strip[pos] = OP(p->strip[pos]) | value;
1724 }
1725
1726 /*
1727  - enlarge - enlarge the strip
1728  == static void enlarge(struct parse *p, sopno size);
1729  */
1730 static void
1731 enlarge(
1732     struct parse *p,
1733     sopno size)
1734 {
1735         sop *sp;
1736
1737         _DIAGASSERT(p != NULL);
1738
1739         if (p->ssize >= size)
1740                 return;
1741
1742         sp = (sop *)realloc(p->strip, size*sizeof(sop));
1743         if (sp == NULL) {
1744                 SETERROR(REG_ESPACE);
1745                 return;
1746         }
1747         p->strip = sp;
1748         p->ssize = size;
1749 }
1750
1751 /*
1752  - stripsnug - compact the strip
1753  == static void stripsnug(struct parse *p, struct re_guts *g);
1754  */
1755 static void
1756 stripsnug(
1757     struct parse *p,
1758     struct re_guts *g)
1759 {
1760
1761         _DIAGASSERT(p != NULL);
1762         _DIAGASSERT(g != NULL);
1763
1764         g->nstates = p->slen;
1765         g->strip = realloc(p->strip, p->slen * sizeof(sop));
1766         if (g->strip == NULL) {
1767                 SETERROR(REG_ESPACE);
1768                 g->strip = p->strip;
1769         }
1770 }
1771
1772 /*
1773  - findmust - fill in must and mlen with longest mandatory literal string
1774  == static void findmust(struct parse *p, struct re_guts *g);
1775  *
1776  * This algorithm could do fancy things like analyzing the operands of |
1777  * for common subsequences.  Someday.  This code is simple and finds most
1778  * of the interesting cases.
1779  *
1780  * Note that must and mlen got initialized during setup.
1781  */
1782 static void
1783 findmust(
1784     struct parse *p,
1785     struct re_guts *g)
1786 {
1787         sop *scan;
1788         sop *start = NULL;
1789         sop *newstart = NULL;
1790         sopno newlen;
1791         sop s;
1792         char *cp;
1793         sopno i;
1794
1795         _DIAGASSERT(p != NULL);
1796         _DIAGASSERT(g != NULL);
1797
1798         /* avoid making error situations worse */
1799         if (p->error != 0)
1800                 return;
1801
1802         /* find the longest OCHAR sequence in strip */
1803         newlen = 0;
1804         scan = g->strip + 1;
1805         do {
1806                 s = *scan++;
1807                 switch (OP(s)) {
1808                 case OCHAR:             /* sequence member */
1809                         if (newlen == 0)                /* new sequence */
1810                                 newstart = scan - 1;
1811                         newlen++;
1812                         break;
1813                 case OPLUS_:            /* things that don't break one */
1814                 case OLPAREN:
1815                 case ORPAREN:
1816                         break;
1817                 case OQUEST_:           /* things that must be skipped */
1818                 case OCH_:
1819                         scan--;
1820                         do {
1821                                 scan += OPND(s);
1822                                 s = *scan;
1823                                 /* assert() interferes w debug printouts */
1824                                 if (OP(s) != O_QUEST && OP(s) != O_CH &&
1825                                                         OP(s) != OOR2) {
1826                                         g->iflags |= BAD;
1827                                         return;
1828                                 }
1829                         } while (OP(s) != O_QUEST && OP(s) != O_CH);
1830                         /* FALLTHROUGH */
1831                 default:                /* things that break a sequence */
1832                         if (newlen > g->mlen) {         /* ends one */
1833                                 start = newstart;
1834                                 g->mlen = newlen;
1835                         }
1836                         newlen = 0;
1837                         break;
1838                 }
1839         } while (OP(s) != OEND);
1840
1841         if (start == NULL)
1842                 g->mlen = 0;
1843
1844         if (g->mlen == 0)       /* there isn't one */
1845                 return;
1846
1847         /* turn it into a character string */
1848         g->must = malloc((size_t)g->mlen + 1);
1849         if (g->must == NULL) {          /* argh; just forget it */
1850                 g->mlen = 0;
1851                 return;
1852         }
1853         cp = g->must;
1854         scan = start;
1855         for (i = g->mlen; i > 0; i--) {
1856                 while (OP(s = *scan++) != OCHAR)
1857                         continue;
1858                 assert(cp < g->must + g->mlen);
1859                 *cp++ = (char)OPND(s);
1860         }
1861         assert(cp == g->must + g->mlen);
1862         *cp++ = '\0';           /* just on general principles */
1863 }
1864
1865 /*
1866  - pluscount - count + nesting
1867  == static sopno pluscount(struct parse *p, struct re_guts *g);
1868  */
1869 static sopno                    /* nesting depth */
1870 pluscount(
1871     struct parse *p,
1872     struct re_guts *g)
1873 {
1874         sop *scan;
1875         sop s;
1876         sopno plusnest = 0;
1877         sopno maxnest = 0;
1878
1879         _DIAGASSERT(p != NULL);
1880         _DIAGASSERT(g != NULL);
1881
1882         if (p->error != 0)
1883                 return(0);      /* there may not be an OEND */
1884
1885         scan = g->strip + 1;
1886         do {
1887                 s = *scan++;
1888                 switch (OP(s)) {
1889                 case OPLUS_:
1890                         plusnest++;
1891                         break;
1892                 case O_PLUS:
1893                         if (plusnest > maxnest)
1894                                 maxnest = plusnest;
1895                         plusnest--;
1896                         break;
1897                 }
1898         } while (OP(s) != OEND);
1899         if (plusnest != 0)
1900                 g->iflags |= BAD;
1901         return(maxnest);
1902 }