1 #include <sys/types.h>
2 #include <stdio.h>
3 #include <string.h>
4 #include <ctype.h>
5 #include <limits.h>
6 #include <stdlib.h>
7
8 /* local headers */
9 #include "regex.h"
10 #include "utils.h"
11 #include "regex2.h"
12 #include "cclass.h"
13 #include "cname.h"
14
15 /*
16 * parse structure, passed up and down to avoid global variables and
17 * other clumsinesses
18 */
19 struct parse {
20 char *next; /* next character in RE */
21 char *end; /* end of string (-> NUL normally) */
22 int error; /* has an error been seen? */
23 sop *strip; /* malloced strip */
24 sopno ssize; /* malloced strip size (allocated) */
25 sopno slen; /* malloced strip length (used) */
26 int ncsalloc; /* number of csets allocated */
27 struct re_guts *g;
28 # define NPAREN 10 /* we need to remember () 1-9 for back refs */
29 sopno pbegin[NPAREN]; /* -> ( ([0] unused) */
30 sopno pend[NPAREN]; /* -> ) ([0] unused) */
31 };
32
33 /* #include "regcomp.ih" // why generate this */
34 /* forward references */
35 static void p_ere(register struct parse *p, int stop);
36 static void p_ere_exp(register struct parse *p);
37 static void p_str(register struct parse *p);
38 static void p_bre(register struct parse *p, register int end1, register int end2);
39 static int p_simp_re(register struct parse *p, int starordinary);
40 static int p_count(register struct parse *p);
41 static void p_bracket(register struct parse *p);
42 static void p_b_term(register struct parse *p, register cset *cs);
43 static void p_b_cclass(register struct parse *p, register cset *cs);
44 static void p_b_eclass(register struct parse *p, register cset *cs);
45 static char p_b_symbol(register struct parse *p);
46 static char p_b_coll_elem(register struct parse *p, int endc);
47 static char othercase(int ch);
48 static void bothcases(register struct parse *p, int ch);
49 static void ordinary(register struct parse *p, register int ch);
50 static void nonnewline(register struct parse *p);
51 static void repeat(register struct parse *p, sopno start, int from, int to);
52 static int seterr(register struct parse *p, int e);
53 static cset *allocset(register struct parse *p);
54 static void freeset(register struct parse *p, register cset *cs);
55 static int freezeset(register struct parse *p, register cset *cs);
56 static int firstch(register struct parse *p, register cset *cs);
57 static int nch(register struct parse *p, register cset *cs);
58 static void mcadd(register struct parse *p, register cset *cs, register char *cp);
59 static void mcsub(register cset *cs, register char *cp);
60 static int mcin(register cset *cs, register char *cp);
61 static char *mcfind(register cset *cs, register char *cp);
62 static void mcinvert(register struct parse *p, register cset *cs);
63 static void mccase(register struct parse *p, register cset *cs);
64 static int isinsets(register struct re_guts *g, int c);
65 static int samesets(register struct re_guts *g, int c1, int c2);
66 static void categorize(struct parse *p, register struct re_guts *g);
67 static sopno dupl(register struct parse *p, sopno start, sopno finish);
68 static void doemit(register struct parse *p, sop op, size_t opnd);
69 static void doinsert(register struct parse *p, sop op, size_t opnd, sopno pos);
70 static void dofwd(register struct parse *p, sopno pos, sop value);
71 static void enlarge(register struct parse *p, sopno size);
72 static void stripsnug(register struct parse *p, register struct re_guts *g);
73 static void findmust(register struct parse *p, register struct re_guts *g);
74 static sopno pluscount(register struct parse *p, register struct re_guts *g);
75
76 static char nuls[10]; /* place to point scanner in event of error */
77
78 /*
79 * macros for use with parse structure
80 * BEWARE: these know that the parse structure is named `p' !!!
81 */
82 #define PEEK() (*p->next)
83 #define PEEK2() (*(p->next+1))
84 #define MORE() (p->next < p->end)
85 #define MORE2() (p->next+1 < p->end)
86 #define SEE(c) (MORE() && PEEK() == (c))
87 #define SEETWO(a, b) (MORE() && MORE2() && PEEK() == (a) && PEEK2() == (b))
88 #define EAT(c) ((SEE(c)) ? (NEXT(), 1) : 0)
89 #define EATTWO(a, b) ((SEETWO(a, b)) ? (NEXT2(), 1) : 0)
90 #define NEXT() (p->next++)
91 #define NEXT2() (p->next += 2)
92 #define NEXTn(n) (p->next += (n))
93 #define GETNEXT() (*p->next++)
94 #define SETERROR(e) seterr(p, (e))
95 #define REQUIRE(co, e) ((co) || SETERROR(e))
96 #define MUSTSEE(c, e) (REQUIRE(MORE() && PEEK() == (c), e))
97 #define MUSTEAT(c, e) (REQUIRE(MORE() && GETNEXT() == (c), e))
98 #define MUSTNOTSEE(c, e) (REQUIRE(!MORE() || PEEK() != (c), e))
99 #define EMIT(op, sopnd) doemit(p, (sop)(op), (size_t)(sopnd))
100 #define INSERT(op, pos) doinsert(p, (sop)(op), HERE()-(pos)+1, pos)
101 #define AHEAD(pos) dofwd(p, pos, HERE()-(pos))
102 #define ASTERN(sop, pos) EMIT(sop, HERE()-pos)
103 #define HERE() (p->slen)
104 #define THERE() (p->slen - 1)
105 #define THERETHERE() (p->slen - 2)
106 #define DROP(n) (p->slen -= (n))
107
108 #ifndef NDEBUG
109 static int never = 0; /* for use in asserts; shuts lint up */
110 #else
111 #define never 0 /* some <assert.h>s have bugs too */
112 #endif
113
114 /*
115 - regcomp - interface for parser and compilation
116 = extern int regcomp(regex_t *, const char *, int);
117 = #define REG_BASIC 0000
118 = #define REG_EXTENDED 0001
119 = #define REG_ICASE 0002
120 = #define REG_NOSUB 0004
121 = #define REG_NEWLINE 0010
122 = #define REG_NOSPEC 0020
123 = #define REG_PEND 0040
124 = #define REG_DUMP 0200
125 */
126 int /* 0 success, otherwise REG_something */
regcomp(preg,pattern,cflags)127 regcomp(preg, pattern, cflags)
128 regex_t *preg;
129 const char *pattern;
130 int cflags;
131 {
132 struct parse pa;
133 register struct re_guts *g;
134 register struct parse *p = &pa;
135 register int i;
136 register size_t len;
137 #ifdef REDEBUG
138 # define GOODFLAGS(f) (f)
139 #else
140 # define GOODFLAGS(f) ((f)&~REG_DUMP)
141 #endif
142
143 cflags = GOODFLAGS(cflags);
144 if ((cflags®_EXTENDED) && (cflags®_NOSPEC))
145 return(REG_INVARG);
146
147 if (cflags®_PEND) {
148 if (preg->re_endp < pattern)
149 return(REG_INVARG);
150 len = preg->re_endp - pattern;
151 } else
152 len = strlen((char *)pattern);
153
154 /* do the mallocs early so failure handling is easy */
155 g = (struct re_guts *)malloc(sizeof(struct re_guts) +
156 (NC-1)*sizeof(cat_t));
157 if (g == NULL)
158 return(REG_ESPACE);
159 {
160 /* Patched for CERT Vulnerability Note VU#695940, Feb 2015. */
161 size_t new_ssize = len/(size_t)2*(size_t)3 + (size_t)1; /* ugh */
162 if (new_ssize < len || new_ssize > LONG_MAX / sizeof(sop)) {
163 free((char *) g);
164 return REG_INVARG;
165 }
166 p->ssize = new_ssize;
167 }
168 p->strip = (sop *)malloc(p->ssize * sizeof(sop));
169 p->slen = 0;
170 if (p->strip == NULL) {
171 free((char *)g);
172 return(REG_ESPACE);
173 }
174
175 /* set things up */
176 p->g = g;
177 p->next = (char *)pattern; /* convenience; we do not modify it */
178 p->end = p->next + len;
179 p->error = 0;
180 p->ncsalloc = 0;
181 for (i = 0; i < NPAREN; i++) {
182 p->pbegin[i] = 0;
183 p->pend[i] = 0;
184 }
185 g->csetsize = NC;
186 g->sets = NULL;
187 g->setbits = NULL;
188 g->ncsets = 0;
189 g->cflags = cflags;
190 g->iflags = 0;
191 g->nbol = 0;
192 g->neol = 0;
193 g->must = NULL;
194 g->mlen = 0;
195 g->nsub = 0;
196 g->ncategories = 1; /* category 0 is "everything else" */
197 g->categories = &g->catspace[-(CHAR_MIN)];
198 (void) memset((char *)g->catspace, 0, NC*sizeof(cat_t));
199 g->backrefs = 0;
200
201 /* do it */
202 EMIT(OEND, 0);
203 g->firststate = THERE();
204 if (cflags®_EXTENDED)
205 p_ere(p, OUT);
206 else if (cflags®_NOSPEC)
207 p_str(p);
208 else
209 p_bre(p, OUT, OUT);
210 EMIT(OEND, 0);
211 g->laststate = THERE();
212
213 /* tidy up loose ends and fill things in */
214 categorize(p, g);
215 stripsnug(p, g);
216 findmust(p, g);
217 g->nplus = pluscount(p, g);
218 g->magic = MAGIC2;
219 preg->re_nsub = g->nsub;
220 preg->re_g = g;
221 preg->re_magic = MAGIC1;
222 #ifndef REDEBUG
223 /* not debugging, so can't rely on the assert() in regexec() */
224 if (g->iflags&BAD)
225 SETERROR(REG_ASSERT);
226 #endif
227
228 /* win or lose, we're done */
229 if (p->error != 0) /* lose */
230 regfree(preg);
231 return(p->error);
232 }
233
234 /*
235 - p_ere - ERE parser top level, concatenation and alternation
236 == static void p_ere(register struct parse *p, int stop);
237 */
238 static void
p_ere(p,stop)239 p_ere(p, stop)
240 register struct parse *p;
241 int stop; /* character this ERE should end at */
242 {
243 register char c;
244 register sopno prevback;
245 register sopno prevfwd;
246 register sopno conc;
247 register int first = 1; /* is this the first alternative? */
248
249 for (;;) {
250 /* do a bunch of concatenated expressions */
251 conc = HERE();
252 while (MORE() && (c = PEEK()) != '|' && c != stop)
253 p_ere_exp(p);
254 REQUIRE(HERE() != conc, REG_EMPTY); /* require nonempty */
255
256 if (!EAT('|'))
257 break; /* NOTE BREAK OUT */
258
259 if (first) {
260 INSERT(OCH_, conc); /* offset is wrong */
261 prevfwd = conc;
262 prevback = conc;
263 first = 0;
264 }
265 ASTERN(OOR1, prevback);
266 prevback = THERE();
267 AHEAD(prevfwd); /* fix previous offset */
268 prevfwd = HERE();
269 EMIT(OOR2, 0); /* offset is very wrong */
270 }
271
272 if (!first) { /* tail-end fixups */
273 AHEAD(prevfwd);
274 ASTERN(O_CH, prevback);
275 }
276
277 assert(!MORE() || SEE(stop));
278 }
279
280 /*
281 - p_ere_exp - parse one subERE, an atom possibly followed by a repetition op
282 == static void p_ere_exp(register struct parse *p);
283 */
284 static void
p_ere_exp(p)285 p_ere_exp(p)
286 register struct parse *p;
287 {
288 register char c;
289 register sopno pos;
290 register int count;
291 register int count2;
292 register sopno subno;
293 int wascaret = 0;
294
295 assert(MORE()); /* caller should have ensured this */
296 c = GETNEXT();
297
298 pos = HERE();
299 switch (c) {
300 case '(':
301 REQUIRE(MORE(), REG_EPAREN);
302 p->g->nsub++;
303 subno = p->g->nsub;
304 if (subno < NPAREN)
305 p->pbegin[subno] = HERE();
306 EMIT(OLPAREN, subno);
307 if (!SEE(')'))
308 p_ere(p, ')');
309 if (subno < NPAREN) {
310 p->pend[subno] = HERE();
311 assert(p->pend[subno] != 0);
312 }
313 EMIT(ORPAREN, subno);
314 MUSTEAT(')', REG_EPAREN);
315 break;
316 #ifndef POSIX_MISTAKE
317 case ')': /* happens only if no current unmatched ( */
318 /*
319 * You may ask, why the ifndef? Because I didn't notice
320 * this until slightly too late for 1003.2, and none of the
321 * other 1003.2 regular-expression reviewers noticed it at
322 * all. So an unmatched ) is legal POSIX, at least until
323 * we can get it fixed.
324 */
325 SETERROR(REG_EPAREN);
326 break;
327 #endif
328 case '^':
329 EMIT(OBOL, 0);
330 p->g->iflags |= USEBOL;
331 p->g->nbol++;
332 wascaret = 1;
333 break;
334 case '$':
335 EMIT(OEOL, 0);
336 p->g->iflags |= USEEOL;
337 p->g->neol++;
338 break;
339 case '|':
340 SETERROR(REG_EMPTY);
341 break;
342 case '*':
343 case '+':
344 case '?':
345 SETERROR(REG_BADRPT);
346 break;
347 case '.':
348 if (p->g->cflags®_NEWLINE)
349 nonnewline(p);
350 else
351 EMIT(OANY, 0);
352 break;
353 case '[':
354 p_bracket(p);
355 break;
356 case '\\':
357 REQUIRE(MORE(), REG_EESCAPE);
358 c = GETNEXT();
359 ordinary(p, c);
360 break;
361 case '{': /* okay as ordinary except if digit follows */
362 REQUIRE(!MORE() || !isdigit(PEEK()), REG_BADRPT);
363 /* FALLTHROUGH */
364 default:
365 ordinary(p, c);
366 break;
367 }
368
369 if (!MORE())
370 return;
371 c = PEEK();
372 /* we call { a repetition if followed by a digit */
373 if (!( c == '*' || c == '+' || c == '?' ||
374 (c == '{' && MORE2() && isdigit(PEEK2())) ))
375 return; /* no repetition, we're done */
376 NEXT();
377
378 REQUIRE(!wascaret, REG_BADRPT);
379 switch (c) {
380 case '*': /* implemented as +? */
381 /* this case does not require the (y|) trick, noKLUDGE */
382 INSERT(OPLUS_, pos);
383 ASTERN(O_PLUS, pos);
384 INSERT(OQUEST_, pos);
385 ASTERN(O_QUEST, pos);
386 break;
387 case '+':
388 INSERT(OPLUS_, pos);
389 ASTERN(O_PLUS, pos);
390 break;
391 case '?':
392 /* KLUDGE: emit y? as (y|) until subtle bug gets fixed */
393 INSERT(OCH_, pos); /* offset slightly wrong */
394 ASTERN(OOR1, pos); /* this one's right */
395 AHEAD(pos); /* fix the OCH_ */
396 EMIT(OOR2, 0); /* offset very wrong... */
397 AHEAD(THERE()); /* ...so fix it */
398 ASTERN(O_CH, THERETHERE());
399 break;
400 case '{':
401 count = p_count(p);
402 if (EAT(',')) {
403 if (isdigit(PEEK())) {
404 count2 = p_count(p);
405 REQUIRE(count <= count2, REG_BADBR);
406 } else /* single number with comma */
407 count2 = INFINITY;
408 } else /* just a single number */
409 count2 = count;
410 repeat(p, pos, count, count2);
411 if (!EAT('}')) { /* error heuristics */
412 while (MORE() && PEEK() != '}')
413 NEXT();
414 REQUIRE(MORE(), REG_EBRACE);
415 SETERROR(REG_BADBR);
416 }
417 break;
418 }
419
420 if (!MORE())
421 return;
422 c = PEEK();
423 if (!( c == '*' || c == '+' || c == '?' ||
424 (c == '{' && MORE2() && isdigit(PEEK2())) ) )
425 return;
426 SETERROR(REG_BADRPT);
427 }
428
429 /*
430 - p_str - string (no metacharacters) "parser"
431 == static void p_str(register struct parse *p);
432 */
433 static void
p_str(p)434 p_str(p)
435 register struct parse *p;
436 {
437 REQUIRE(MORE(), REG_EMPTY);
438 while (MORE())
439 ordinary(p, GETNEXT());
440 }
441
442 /*
443 - p_bre - BRE parser top level, anchoring and concatenation
444 == static void p_bre(register struct parse *p, register int end1, \
445 == register int end2);
446 * Giving end1 as OUT essentially eliminates the end1/end2 check.
447 *
448 * This implementation is a bit of a kludge, in that a trailing $ is first
449 * taken as an ordinary character and then revised to be an anchor. The
450 * only undesirable side effect is that '$' gets included as a character
451 * category in such cases. This is fairly harmless; not worth fixing.
452 * The amount of lookahead needed to avoid this kludge is excessive.
453 */
454 static void
p_bre(p,end1,end2)455 p_bre(p, end1, end2)
456 register struct parse *p;
457 register int end1; /* first terminating character */
458 register int end2; /* second terminating character */
459 {
460 register sopno start = HERE();
461 register int first = 1; /* first subexpression? */
462 register int wasdollar = 0;
463
464 if (EAT('^')) {
465 EMIT(OBOL, 0);
466 p->g->iflags |= USEBOL;
467 p->g->nbol++;
468 }
469 while (MORE() && !SEETWO(end1, end2)) {
470 wasdollar = p_simp_re(p, first);
471 first = 0;
472 }
473 if (wasdollar) { /* oops, that was a trailing anchor */
474 DROP(1);
475 EMIT(OEOL, 0);
476 p->g->iflags |= USEEOL;
477 p->g->neol++;
478 }
479
480 REQUIRE(HERE() != start, REG_EMPTY); /* require nonempty */
481 }
482
483 /*
484 - p_simp_re - parse a simple RE, an atom possibly followed by a repetition
485 == static int p_simp_re(register struct parse *p, int starordinary);
486 */
487 static int /* was the simple RE an unbackslashed $? */
p_simp_re(p,starordinary)488 p_simp_re(p, starordinary)
489 register struct parse *p;
490 int starordinary; /* is a leading * an ordinary character? */
491 {
492 register int c;
493 register int count;
494 register int count2;
495 register sopno pos;
496 register int i;
497 register sopno subno;
498 # define BACKSL (1<<CHAR_BIT)
499
500 pos = HERE(); /* repetion op, if any, covers from here */
501
502 assert(MORE()); /* caller should have ensured this */
503 c = GETNEXT();
504 if (c == '\\') {
505 REQUIRE(MORE(), REG_EESCAPE);
506 c = BACKSL | (unsigned char)GETNEXT();
507 }
508 switch (c) {
509 case '.':
510 if (p->g->cflags®_NEWLINE)
511 nonnewline(p);
512 else
513 EMIT(OANY, 0);
514 break;
515 case '[':
516 p_bracket(p);
517 break;
518 case BACKSL|'{':
519 SETERROR(REG_BADRPT);
520 break;
521 case BACKSL|'(':
522 p->g->nsub++;
523 subno = p->g->nsub;
524 if (subno < NPAREN)
525 p->pbegin[subno] = HERE();
526 EMIT(OLPAREN, subno);
527 /* the MORE here is an error heuristic */
528 if (MORE() && !SEETWO('\\', ')'))
529 p_bre(p, '\\', ')');
530 if (subno < NPAREN) {
531 p->pend[subno] = HERE();
532 assert(p->pend[subno] != 0);
533 }
534 EMIT(ORPAREN, subno);
535 REQUIRE(EATTWO('\\', ')'), REG_EPAREN);
536 break;
537 case BACKSL|')': /* should not get here -- must be user */
538 case BACKSL|'}':
539 SETERROR(REG_EPAREN);
540 break;
541 case BACKSL|'1':
542 case BACKSL|'2':
543 case BACKSL|'3':
544 case BACKSL|'4':
545 case BACKSL|'5':
546 case BACKSL|'6':
547 case BACKSL|'7':
548 case BACKSL|'8':
549 case BACKSL|'9':
550 i = (c&~BACKSL) - '0';
551 assert(i < NPAREN);
552 if (p->pend[i] != 0) {
553 assert(i <= p->g->nsub);
554 EMIT(OBACK_, i);
555 assert(p->pbegin[i] != 0);
556 assert(OP(p->strip[p->pbegin[i]]) == OLPAREN);
557 assert(OP(p->strip[p->pend[i]]) == ORPAREN);
558 (void) dupl(p, p->pbegin[i]+1, p->pend[i]);
559 EMIT(O_BACK, i);
560 } else
561 SETERROR(REG_ESUBREG);
562 p->g->backrefs = 1;
563 break;
564 case '*':
565 REQUIRE(starordinary, REG_BADRPT);
566 /* FALLTHROUGH */
567 default:
568 ordinary(p, (char)c); /* takes off BACKSL, if any */
569 break;
570 }
571
572 if (EAT('*')) { /* implemented as +? */
573 /* this case does not require the (y|) trick, noKLUDGE */
574 INSERT(OPLUS_, pos);
575 ASTERN(O_PLUS, pos);
576 INSERT(OQUEST_, pos);
577 ASTERN(O_QUEST, pos);
578 } else if (EATTWO('\\', '{')) {
579 count = p_count(p);
580 if (EAT(',')) {
581 if (MORE() && isdigit(PEEK())) {
582 count2 = p_count(p);
583 REQUIRE(count <= count2, REG_BADBR);
584 } else /* single number with comma */
585 count2 = INFINITY;
586 } else /* just a single number */
587 count2 = count;
588 repeat(p, pos, count, count2);
589 if (!EATTWO('\\', '}')) { /* error heuristics */
590 while (MORE() && !SEETWO('\\', '}'))
591 NEXT();
592 REQUIRE(MORE(), REG_EBRACE);
593 SETERROR(REG_BADBR);
594 }
595 } else if (c == (unsigned char)'$') /* $ (but not \$) ends it */
596 return(1);
597
598 return(0);
599 }
600
601 /*
602 - p_count - parse a repetition count
603 == static int p_count(register struct parse *p);
604 */
605 static int /* the value */
p_count(p)606 p_count(p)
607 register struct parse *p;
608 {
609 register int count = 0;
610 register int ndigits = 0;
611
612 while (MORE() && isdigit(PEEK()) && count <= DUPMAX) {
613 count = count*10 + (GETNEXT() - '0');
614 ndigits++;
615 }
616
617 REQUIRE(ndigits > 0 && count <= DUPMAX, REG_BADBR);
618 return(count);
619 }
620
621 /*
622 - p_bracket - parse a bracketed character list
623 == static void p_bracket(register struct parse *p);
624 *
625 * Note a significant property of this code: if the allocset() did SETERROR,
626 * no set operations are done.
627 */
628 static void
p_bracket(p)629 p_bracket(p)
630 register struct parse *p;
631 {
632 register cset *cs = allocset(p);
633 register int invert = 0;
634
635 /* Dept of Truly Sickening Special-Case Kludges */
636 if (p->next + 5 < p->end && strncmp(p->next, "[:<:]]", 6) == 0) {
637 EMIT(OBOW, 0);
638 NEXTn(6);
639 return;
640 }
641 if (p->next + 5 < p->end && strncmp(p->next, "[:>:]]", 6) == 0) {
642 EMIT(OEOW, 0);
643 NEXTn(6);
644 return;
645 }
646
647 if (EAT('^'))
648 invert++; /* make note to invert set at end */
649 if (EAT(']'))
650 CHadd(cs, ']');
651 else if (EAT('-'))
652 CHadd(cs, '-');
653 while (MORE() && PEEK() != ']' && !SEETWO('-', ']'))
654 p_b_term(p, cs);
655 if (EAT('-'))
656 CHadd(cs, '-');
657 MUSTEAT(']', REG_EBRACK);
658
659 if (p->error != 0) /* don't mess things up further */
660 return;
661
662 if (p->g->cflags®_ICASE) {
663 register int i;
664 register int ci;
665
666 for (i = p->g->csetsize - 1; i >= 0; i--)
667 if (CHIN(cs, i) && isalpha(i)) {
668 ci = othercase(i);
669 if (ci != i)
670 CHadd(cs, ci);
671 }
672 if (cs->multis != NULL)
673 mccase(p, cs);
674 }
675 if (invert) {
676 register int i;
677
678 for (i = p->g->csetsize - 1; i >= 0; i--)
679 if (CHIN(cs, i))
680 CHsub(cs, i);
681 else
682 CHadd(cs, i);
683 if (p->g->cflags®_NEWLINE)
684 CHsub(cs, '\n');
685 if (cs->multis != NULL)
686 mcinvert(p, cs);
687 }
688
689 assert(cs->multis == NULL); /* xxx */
690
691 if (nch(p, cs) == 1) { /* optimize singleton sets */
692 ordinary(p, firstch(p, cs));
693 freeset(p, cs);
694 } else
695 EMIT(OANYOF, freezeset(p, cs));
696 }
697
698 /*
699 - p_b_term - parse one term of a bracketed character list
700 == static void p_b_term(register struct parse *p, register cset *cs);
701 */
702 static void
p_b_term(p,cs)703 p_b_term(p, cs)
704 register struct parse *p;
705 register cset *cs;
706 {
707 register char c;
708 register char start, finish;
709 register int i;
710
711 /* classify what we've got */
712 switch ((MORE()) ? PEEK() : '\0') {
713 case '[':
714 c = (MORE2()) ? PEEK2() : '\0';
715 break;
716 case '-':
717 SETERROR(REG_ERANGE);
718 return; /* NOTE RETURN */
719 break;
720 default:
721 c = '\0';
722 break;
723 }
724
725 switch (c) {
726 case ':': /* character class */
727 NEXT2();
728 REQUIRE(MORE(), REG_EBRACK);
729 c = PEEK();
730 REQUIRE(c != '-' && c != ']', REG_ECTYPE);
731 p_b_cclass(p, cs);
732 REQUIRE(MORE(), REG_EBRACK);
733 REQUIRE(EATTWO(':', ']'), REG_ECTYPE);
734 break;
735 case '=': /* equivalence class */
736 NEXT2();
737 REQUIRE(MORE(), REG_EBRACK);
738 c = PEEK();
739 REQUIRE(c != '-' && c != ']', REG_ECOLLATE);
740 p_b_eclass(p, cs);
741 REQUIRE(MORE(), REG_EBRACK);
742 REQUIRE(EATTWO('=', ']'), REG_ECOLLATE);
743 break;
744 default: /* symbol, ordinary character, or range */
745 /* xxx revision needed for multichar stuff */
746 start = p_b_symbol(p);
747 if (SEE('-') && MORE2() && PEEK2() != ']') {
748 /* range */
749 NEXT();
750 if (EAT('-'))
751 finish = '-';
752 else
753 finish = p_b_symbol(p);
754 } else
755 finish = start;
756 /* xxx what about signed chars here... */
757 REQUIRE(start <= finish, REG_ERANGE);
758 for (i = start; i <= finish; i++)
759 CHadd(cs, i);
760 break;
761 }
762 }
763
764 /*
765 - p_b_cclass - parse a character-class name and deal with it
766 == static void p_b_cclass(register struct parse *p, register cset *cs);
767 */
768 static void
p_b_cclass(p,cs)769 p_b_cclass(p, cs)
770 register struct parse *p;
771 register cset *cs;
772 {
773 register char *sp = p->next;
774 register struct cclass *cp;
775 register size_t len;
776 register char *u;
777 register char c;
778
779 while (MORE() && isalpha(PEEK()))
780 NEXT();
781 len = p->next - sp;
782 for (cp = cclasses; cp->name != NULL; cp++)
783 if (strncmp(cp->name, sp, len) == 0 && cp->name[len] == '\0')
784 break;
785 if (cp->name == NULL) {
786 /* oops, didn't find it */
787 SETERROR(REG_ECTYPE);
788 return;
789 }
790
791 u = cp->chars;
792 while ((c = *u++) != '\0')
793 CHadd(cs, c);
794 for (u = cp->multis; *u != '\0'; u += strlen(u) + 1)
795 MCadd(p, cs, u);
796 }
797
798 /*
799 - p_b_eclass - parse an equivalence-class name and deal with it
800 == static void p_b_eclass(register struct parse *p, register cset *cs);
801 *
802 * This implementation is incomplete. xxx
803 */
804 static void
p_b_eclass(p,cs)805 p_b_eclass(p, cs)
806 register struct parse *p;
807 register cset *cs;
808 {
809 register char c;
810
811 c = p_b_coll_elem(p, '=');
812 CHadd(cs, c);
813 }
814
815 /*
816 - p_b_symbol - parse a character or [..]ed multicharacter collating symbol
817 == static char p_b_symbol(register struct parse *p);
818 */
819 static char /* value of symbol */
p_b_symbol(p)820 p_b_symbol(p)
821 register struct parse *p;
822 {
823 register char value;
824
825 REQUIRE(MORE(), REG_EBRACK);
826 if (!EATTWO('[', '.'))
827 return(GETNEXT());
828
829 /* collating symbol */
830 value = p_b_coll_elem(p, '.');
831 REQUIRE(EATTWO('.', ']'), REG_ECOLLATE);
832 return(value);
833 }
834
835 /*
836 - p_b_coll_elem - parse a collating-element name and look it up
837 == static char p_b_coll_elem(register struct parse *p, int endc);
838 */
839 static char /* value of collating element */
p_b_coll_elem(p,endc)840 p_b_coll_elem(p, endc)
841 register struct parse *p;
842 int endc; /* name ended by endc,']' */
843 {
844 register char *sp = p->next;
845 register struct cname *cp;
846 register int len;
847
848 while (MORE() && !SEETWO(endc, ']'))
849 NEXT();
850 if (!MORE()) {
851 SETERROR(REG_EBRACK);
852 return(0);
853 }
854 len = p->next - sp;
855 for (cp = cnames; cp->name != NULL; cp++)
856 if (strncmp(cp->name, sp, len) == 0 && cp->name[len] == '\0')
857 return(cp->code); /* known name */
858 if (len == 1)
859 return(*sp); /* single character */
860 SETERROR(REG_ECOLLATE); /* neither */
861 return(0);
862 }
863
864 /*
865 - othercase - return the case counterpart of an alphabetic
866 == static char othercase(int ch);
867 */
868 static char /* if no counterpart, return ch */
othercase(ch)869 othercase(ch)
870 int ch;
871 {
872 assert(isalpha(ch));
873 if (isupper(ch))
874 return(tolower(ch));
875 else if (islower(ch))
876 return(toupper(ch));
877 else /* peculiar, but could happen */
878 return(ch);
879 }
880
881 /*
882 - bothcases - emit a dualcase version of a two-case character
883 == static void bothcases(register struct parse *p, int ch);
884 *
885 * Boy, is this implementation ever a kludge...
886 */
887 static void
bothcases(p,ch)888 bothcases(p, ch)
889 register struct parse *p;
890 int ch;
891 {
892 register char *oldnext = p->next;
893 register char *oldend = p->end;
894 char bracket[3];
895
896 assert(othercase(ch) != ch); /* p_bracket() would recurse */
897 p->next = bracket;
898 p->end = bracket+2;
899 bracket[0] = ch;
900 bracket[1] = ']';
901 bracket[2] = '\0';
902 p_bracket(p);
903 assert(p->next == bracket+2);
904 p->next = oldnext;
905 p->end = oldend;
906 }
907
908 /*
909 - ordinary - emit an ordinary character
910 == static void ordinary(register struct parse *p, register int ch);
911 */
912 static void
ordinary(p,ch)913 ordinary(p, ch)
914 register struct parse *p;
915 register int ch;
916 {
917 register cat_t *cap = p->g->categories;
918
919 if ((p->g->cflags®_ICASE) && isalpha(ch) && othercase(ch) != ch)
920 bothcases(p, ch);
921 else {
922 EMIT(OCHAR, (unsigned char)ch);
923 if (cap[ch] == 0)
924 cap[ch] = p->g->ncategories++;
925 }
926 }
927
928 /*
929 - nonnewline - emit REG_NEWLINE version of OANY
930 == static void nonnewline(register struct parse *p);
931 *
932 * Boy, is this implementation ever a kludge...
933 */
934 static void
nonnewline(p)935 nonnewline(p)
936 register struct parse *p;
937 {
938 register char *oldnext = p->next;
939 register char *oldend = p->end;
940 char bracket[4];
941
942 p->next = bracket;
943 p->end = bracket+3;
944 bracket[0] = '^';
945 bracket[1] = '\n';
946 bracket[2] = ']';
947 bracket[3] = '\0';
948 p_bracket(p);
949 assert(p->next == bracket+3);
950 p->next = oldnext;
951 p->end = oldend;
952 }
953
954 /*
955 - repeat - generate code for a bounded repetition, recursively if needed
956 == static void repeat(register struct parse *p, sopno start, int from, int to);
957 */
958 static void
repeat(p,start,from,to)959 repeat(p, start, from, to)
960 register struct parse *p;
961 sopno start; /* operand from here to end of strip */
962 int from; /* repeated from this number */
963 int to; /* to this number of times (maybe INFINITY) */
964 {
965 register sopno finish = HERE();
966 # define N 2
967 # define INF 3
968 # define REP(f, t) ((f)*8 + (t))
969 # define MAP(n) (((n) <= 1) ? (n) : ((n) == INFINITY) ? INF : N)
970 register sopno copy;
971
972 if (p->error != 0) /* head off possible runaway recursion */
973 return;
974
975 assert(from <= to);
976
977 switch (REP(MAP(from), MAP(to))) {
978 case REP(0, 0): /* must be user doing this */
979 DROP(finish-start); /* drop the operand */
980 break;
981 case REP(0, 1): /* as x{1,1}? */
982 case REP(0, N): /* as x{1,n}? */
983 case REP(0, INF): /* as x{1,}? */
984 /* KLUDGE: emit y? as (y|) until subtle bug gets fixed */
985 INSERT(OCH_, start); /* offset is wrong... */
986 repeat(p, start+1, 1, to);
987 ASTERN(OOR1, start);
988 AHEAD(start); /* ... fix it */
989 EMIT(OOR2, 0);
990 AHEAD(THERE());
991 ASTERN(O_CH, THERETHERE());
992 break;
993 case REP(1, 1): /* trivial case */
994 /* done */
995 break;
996 case REP(1, N): /* as x?x{1,n-1} */
997 /* KLUDGE: emit y? as (y|) until subtle bug gets fixed */
998 INSERT(OCH_, start);
999 ASTERN(OOR1, start);
1000 AHEAD(start);
1001 EMIT(OOR2, 0); /* offset very wrong... */
1002 AHEAD(THERE()); /* ...so fix it */
1003 ASTERN(O_CH, THERETHERE());
1004 copy = dupl(p, start+1, finish+1);
1005 assert(copy == finish+4);
1006 repeat(p, copy, 1, to-1);
1007 break;
1008 case REP(1, INF): /* as x+ */
1009 INSERT(OPLUS_, start);
1010 ASTERN(O_PLUS, start);
1011 break;
1012 case REP(N, N): /* as xx{m-1,n-1} */
1013 copy = dupl(p, start, finish);
1014 repeat(p, copy, from-1, to-1);
1015 break;
1016 case REP(N, INF): /* as xx{n-1,INF} */
1017 copy = dupl(p, start, finish);
1018 repeat(p, copy, from-1, to);
1019 break;
1020 default: /* "can't happen" */
1021 SETERROR(REG_ASSERT); /* just in case */
1022 break;
1023 }
1024 }
1025
1026 /*
1027 - seterr - set an error condition
1028 == static int seterr(register struct parse *p, int e);
1029 */
1030 static int /* useless but makes type checking happy */
seterr(p,e)1031 seterr(p, e)
1032 register struct parse *p;
1033 int e;
1034 {
1035 if (p->error == 0) /* keep earliest error condition */
1036 p->error = e;
1037 p->next = nuls; /* try to bring things to a halt */
1038 p->end = nuls;
1039 return(0); /* make the return value well-defined */
1040 }
1041
1042 /*
1043 - allocset - allocate a set of characters for []
1044 == static cset *allocset(register struct parse *p);
1045 */
1046 static cset *
allocset(p)1047 allocset(p)
1048 register struct parse *p;
1049 {
1050 register int no = p->g->ncsets++;
1051 register size_t nc;
1052 register size_t nbytes;
1053 register cset *cs;
1054 register size_t css = (size_t)p->g->csetsize;
1055 register int i;
1056
1057 if (no >= p->ncsalloc) { /* need another column of space */
1058 p->ncsalloc += CHAR_BIT;
1059 nc = p->ncsalloc;
1060 assert(nc % CHAR_BIT == 0);
1061 nbytes = nc / CHAR_BIT * css;
1062 if (p->g->sets == NULL)
1063 p->g->sets = (cset *)malloc(nc * sizeof(cset));
1064 else
1065 p->g->sets = (cset *)realloc((char *)p->g->sets,
1066 nc * sizeof(cset));
1067 if (p->g->setbits == NULL)
1068 p->g->setbits = (uch *)malloc(nbytes);
1069 else {
1070 p->g->setbits = (uch *)realloc((char *)p->g->setbits,
1071 nbytes);
1072 /* xxx this isn't right if setbits is now NULL */
1073 for (i = 0; i < no; i++)
1074 p->g->sets[i].ptr = p->g->setbits + css*(i/CHAR_BIT);
1075 }
1076 if (p->g->sets != NULL && p->g->setbits != NULL)
1077 (void) memset((char *)p->g->setbits + (nbytes - css),
1078 0, css);
1079 else {
1080 no = 0;
1081 SETERROR(REG_ESPACE);
1082 /* caller's responsibility not to do set ops */
1083 }
1084 }
1085
1086 assert(p->g->sets != NULL); /* xxx */
1087 cs = &p->g->sets[no];
1088 cs->ptr = p->g->setbits + css*((no)/CHAR_BIT);
1089 cs->mask = 1 << ((no) % CHAR_BIT);
1090 cs->hash = 0;
1091 cs->smultis = 0;
1092 cs->multis = NULL;
1093
1094 return(cs);
1095 }
1096
1097 /*
1098 - freeset - free a now-unused set
1099 == static void freeset(register struct parse *p, register cset *cs);
1100 */
1101 static void
freeset(p,cs)1102 freeset(p, cs)
1103 register struct parse *p;
1104 register cset *cs;
1105 {
1106 register int i;
1107 register cset *top = &p->g->sets[p->g->ncsets];
1108 register size_t css = (size_t)p->g->csetsize;
1109
1110 for (i = 0; (size_t)i < css; i++)
1111 CHsub(cs, i);
1112 if (cs == top-1) /* recover only the easy case */
1113 p->g->ncsets--;
1114 }
1115
1116 /*
1117 - freezeset - final processing on a set of characters
1118 == static int freezeset(register struct parse *p, register cset *cs);
1119 *
1120 * The main task here is merging identical sets. This is usually a waste
1121 * of time (although the hash code minimizes the overhead), but can win
1122 * big if REG_ICASE is being used. REG_ICASE, by the way, is why the hash
1123 * is done using addition rather than xor -- all ASCII [aA] sets xor to
1124 * the same value!
1125 */
1126 static int /* set number */
freezeset(p,cs)1127 freezeset(p, cs)
1128 register struct parse *p;
1129 register cset *cs;
1130 {
1131 register uch h = cs->hash;
1132 register int i;
1133 register cset *top = &p->g->sets[p->g->ncsets];
1134 register cset *cs2;
1135 register size_t css = (size_t)p->g->csetsize;
1136
1137 /* look for an earlier one which is the same */
1138 for (cs2 = &p->g->sets[0]; cs2 < top; cs2++)
1139 if (cs2->hash == h && cs2 != cs) {
1140 /* maybe */
1141 for (i = 0; (size_t)i < css; i++)
1142 if (!!CHIN(cs2, i) != !!CHIN(cs, i))
1143 break; /* no */
1144 if (i == css)
1145 break; /* yes */
1146 }
1147
1148 if (cs2 < top) { /* found one */
1149 freeset(p, cs);
1150 cs = cs2;
1151 }
1152
1153 return((int)(cs - p->g->sets));
1154 }
1155
1156 /*
1157 - firstch - return first character in a set (which must have at least one)
1158 == static int firstch(register struct parse *p, register cset *cs);
1159 */
1160 static int /* character; there is no "none" value */
firstch(p,cs)1161 firstch(p, cs)
1162 register struct parse *p;
1163 register cset *cs;
1164 {
1165 register int i;
1166 register size_t css = (size_t)p->g->csetsize;
1167
1168 for (i = 0; (size_t)i < css; i++)
1169 if (CHIN(cs, i))
1170 return((char)i);
1171 assert(never);
1172 return(0); /* arbitrary */
1173 }
1174
1175 /*
1176 - nch - number of characters in a set
1177 == static int nch(register struct parse *p, register cset *cs);
1178 */
1179 static int
nch(p,cs)1180 nch(p, cs)
1181 register struct parse *p;
1182 register cset *cs;
1183 {
1184 register int i;
1185 register size_t css = (size_t)p->g->csetsize;
1186 register int n = 0;
1187
1188 for (i = 0; (size_t)i < css; i++)
1189 if (CHIN(cs, i))
1190 n++;
1191 return(n);
1192 }
1193
1194 /*
1195 - mcadd - add a collating element to a cset
1196 == static void mcadd(register struct parse *p, register cset *cs, \
1197 == register char *cp);
1198 */
1199 static void
mcadd(p,cs,cp)1200 mcadd(p, cs, cp)
1201 register struct parse *p;
1202 register cset *cs;
1203 register char *cp;
1204 {
1205 register size_t oldend = cs->smultis;
1206
1207 cs->smultis += strlen(cp) + 1;
1208 if (cs->multis == NULL)
1209 cs->multis = malloc(cs->smultis);
1210 else
1211 cs->multis = realloc(cs->multis, cs->smultis);
1212 if (cs->multis == NULL) {
1213 SETERROR(REG_ESPACE);
1214 return;
1215 }
1216
1217 (void) strcpy(cs->multis + oldend - 1, cp);
1218 cs->multis[cs->smultis - 1] = '\0';
1219 }
1220
1221 /*
1222 - mcsub - subtract a collating element from a cset
1223 == static void mcsub(register cset *cs, register char *cp);
1224 */
1225 static void
mcsub(cs,cp)1226 mcsub(cs, cp)
1227 register cset *cs;
1228 register char *cp;
1229 {
1230 register char *fp = mcfind(cs, cp);
1231 register size_t len = strlen(fp);
1232
1233 assert(fp != NULL);
1234 (void) memmove(fp, fp + len + 1,
1235 cs->smultis - (fp + len + 1 - cs->multis));
1236 cs->smultis -= len;
1237
1238 if (cs->smultis == 0) {
1239 free(cs->multis);
1240 cs->multis = NULL;
1241 return;
1242 }
1243
1244 cs->multis = realloc(cs->multis, cs->smultis);
1245 assert(cs->multis != NULL);
1246 }
1247
1248 /*
1249 - mcin - is a collating element in a cset?
1250 == static int mcin(register cset *cs, register char *cp);
1251 */
1252 static int
mcin(cs,cp)1253 mcin(cs, cp)
1254 register cset *cs;
1255 register char *cp;
1256 {
1257 return(mcfind(cs, cp) != NULL);
1258 }
1259
1260 /*
1261 - mcfind - find a collating element in a cset
1262 == static char *mcfind(register cset *cs, register char *cp);
1263 */
1264 static char *
mcfind(cs,cp)1265 mcfind(cs, cp)
1266 register cset *cs;
1267 register char *cp;
1268 {
1269 register char *p;
1270
1271 if (cs->multis == NULL)
1272 return(NULL);
1273 for (p = cs->multis; *p != '\0'; p += strlen(p) + 1)
1274 if (strcmp(cp, p) == 0)
1275 return(p);
1276 return(NULL);
1277 }
1278
1279 /*
1280 - mcinvert - invert the list of collating elements in a cset
1281 == static void mcinvert(register struct parse *p, register cset *cs);
1282 *
1283 * This would have to know the set of possibilities. Implementation
1284 * is deferred.
1285 */
1286 static void
mcinvert(p,cs)1287 mcinvert(p, cs)
1288 register struct parse *p;
1289 register cset *cs;
1290 {
1291 assert(cs->multis == NULL); /* xxx */
1292 }
1293
1294 /*
1295 - mccase - add case counterparts of the list of collating elements in a cset
1296 == static void mccase(register struct parse *p, register cset *cs);
1297 *
1298 * This would have to know the set of possibilities. Implementation
1299 * is deferred.
1300 */
1301 static void
mccase(p,cs)1302 mccase(p, cs)
1303 register struct parse *p;
1304 register cset *cs;
1305 {
1306 assert(cs->multis == NULL); /* xxx */
1307 }
1308
1309 /*
1310 - isinsets - is this character in any sets?
1311 == static int isinsets(register struct re_guts *g, int c);
1312 */
1313 static int /* predicate */
isinsets(g,c)1314 isinsets(g, c)
1315 register struct re_guts *g;
1316 int c;
1317 {
1318 register uch *col;
1319 register int i;
1320 register int ncols = (g->ncsets+(CHAR_BIT-1)) / CHAR_BIT;
1321 register unsigned uc = (unsigned char)c;
1322
1323 for (i = 0, col = g->setbits; i < ncols; i++, col += g->csetsize)
1324 if (col[uc] != 0)
1325 return(1);
1326 return(0);
1327 }
1328
1329 /*
1330 - samesets - are these two characters in exactly the same sets?
1331 == static int samesets(register struct re_guts *g, int c1, int c2);
1332 */
1333 static int /* predicate */
samesets(g,c1,c2)1334 samesets(g, c1, c2)
1335 register struct re_guts *g;
1336 int c1;
1337 int c2;
1338 {
1339 register uch *col;
1340 register int i;
1341 register int ncols = (g->ncsets+(CHAR_BIT-1)) / CHAR_BIT;
1342 register unsigned uc1 = (unsigned char)c1;
1343 register unsigned uc2 = (unsigned char)c2;
1344
1345 for (i = 0, col = g->setbits; i < ncols; i++, col += g->csetsize)
1346 if (col[uc1] != col[uc2])
1347 return(0);
1348 return(1);
1349 }
1350
1351 /*
1352 - categorize - sort out character categories
1353 == static void categorize(struct parse *p, register struct re_guts *g);
1354 */
1355 static void
categorize(p,g)1356 categorize(p, g)
1357 struct parse *p;
1358 register struct re_guts *g;
1359 {
1360 register cat_t *cats = g->categories;
1361 register int c;
1362 register int c2;
1363 register cat_t cat;
1364
1365 /* avoid making error situations worse */
1366 if (p->error != 0)
1367 return;
1368
1369 for (c = CHAR_MIN; c <= CHAR_MAX; c++)
1370 if (cats[c] == 0 && isinsets(g, c)) {
1371 cat = g->ncategories++;
1372 cats[c] = cat;
1373 for (c2 = c+1; c2 <= CHAR_MAX; c2++)
1374 if (cats[c2] == 0 && samesets(g, c, c2))
1375 cats[c2] = cat;
1376 }
1377 }
1378
1379 /*
1380 - dupl - emit a duplicate of a bunch of sops
1381 == static sopno dupl(register struct parse *p, sopno start, sopno finish);
1382 */
1383 static sopno /* start of duplicate */
dupl(p,start,finish)1384 dupl(p, start, finish)
1385 register struct parse *p;
1386 sopno start; /* from here */
1387 sopno finish; /* to this less one */
1388 {
1389 register sopno ret = HERE();
1390 register sopno len = finish - start;
1391
1392 assert(finish >= start);
1393 if (len == 0)
1394 return(ret);
1395 enlarge(p, p->ssize + len); /* this many unexpected additions */
1396 assert(p->ssize >= p->slen + len);
1397 (void) memcpy((char *)(p->strip + p->slen),
1398 (char *)(p->strip + start), (size_t)len*sizeof(sop));
1399 p->slen += len;
1400 return(ret);
1401 }
1402
1403 /*
1404 - doemit - emit a strip operator
1405 == static void doemit(register struct parse *p, sop op, size_t opnd);
1406 *
1407 * It might seem better to implement this as a macro with a function as
1408 * hard-case backup, but it's just too big and messy unless there are
1409 * some changes to the data structures. Maybe later.
1410 */
1411 static void
doemit(p,op,opnd)1412 doemit(p, op, opnd)
1413 register struct parse *p;
1414 sop op;
1415 size_t opnd;
1416 {
1417 /* avoid making error situations worse */
1418 if (p->error != 0)
1419 return;
1420
1421 /* deal with oversize operands ("can't happen", more or less) */
1422 assert(opnd < 1<<OPSHIFT);
1423
1424 /* deal with undersized strip */
1425 if (p->slen >= p->ssize)
1426 enlarge(p, (p->ssize+1) / 2 * 3); /* +50% */
1427 assert(p->slen < p->ssize);
1428
1429 /* finally, it's all reduced to the easy case */
1430 p->strip[p->slen++] = SOP(op, opnd);
1431 }
1432
1433 /*
1434 - doinsert - insert a sop into the strip
1435 == static void doinsert(register struct parse *p, sop op, size_t opnd, sopno pos);
1436 */
1437 static void
doinsert(p,op,opnd,pos)1438 doinsert(p, op, opnd, pos)
1439 register struct parse *p;
1440 sop op;
1441 size_t opnd;
1442 sopno pos;
1443 {
1444 register sopno sn;
1445 register sop s;
1446 register int i;
1447
1448 /* avoid making error situations worse */
1449 if (p->error != 0)
1450 return;
1451
1452 sn = HERE();
1453 EMIT(op, opnd); /* do checks, ensure space */
1454 assert(HERE() == sn+1);
1455 s = p->strip[sn];
1456
1457 /* adjust paren pointers */
1458 assert(pos > 0);
1459 for (i = 1; i < NPAREN; i++) {
1460 if (p->pbegin[i] >= pos) {
1461 p->pbegin[i]++;
1462 }
1463 if (p->pend[i] >= pos) {
1464 p->pend[i]++;
1465 }
1466 }
1467
1468 memmove((char *)&p->strip[pos+1], (char *)&p->strip[pos],
1469 (HERE()-pos-1)*sizeof(sop));
1470 p->strip[pos] = s;
1471 }
1472
1473 /*
1474 - dofwd - complete a forward reference
1475 == static void dofwd(register struct parse *p, sopno pos, sop value);
1476 */
1477 static void
dofwd(p,pos,value)1478 dofwd(p, pos, value)
1479 register struct parse *p;
1480 register sopno pos;
1481 sop value;
1482 {
1483 /* avoid making error situations worse */
1484 if (p->error != 0)
1485 return;
1486
1487 assert(value < 1<<OPSHIFT);
1488 p->strip[pos] = OP(p->strip[pos]) | value;
1489 }
1490
1491 /*
1492 - enlarge - enlarge the strip
1493 == static void enlarge(register struct parse *p, sopno size);
1494 */
1495 static void
enlarge(p,size)1496 enlarge(p, size)
1497 register struct parse *p;
1498 register sopno size;
1499 {
1500 register sop *sp;
1501
1502 if (p->ssize >= size)
1503 return;
1504
1505 sp = (sop *)realloc(p->strip, size*sizeof(sop));
1506 if (sp == NULL) {
1507 SETERROR(REG_ESPACE);
1508 return;
1509 }
1510 p->strip = sp;
1511 p->ssize = size;
1512 }
1513
1514 /*
1515 - stripsnug - compact the strip
1516 == static void stripsnug(register struct parse *p, register struct re_guts *g);
1517 */
1518 static void
stripsnug(p,g)1519 stripsnug(p, g)
1520 register struct parse *p;
1521 register struct re_guts *g;
1522 {
1523 g->nstates = p->slen;
1524 g->strip = (sop *)realloc((char *)p->strip, p->slen * sizeof(sop));
1525 if (g->strip == NULL) {
1526 SETERROR(REG_ESPACE);
1527 g->strip = p->strip;
1528 }
1529 }
1530
1531 /*
1532 - findmust - fill in must and mlen with longest mandatory literal string
1533 == static void findmust(register struct parse *p, register struct re_guts *g);
1534 *
1535 * This algorithm could do fancy things like analyzing the operands of |
1536 * for common subsequences. Someday. This code is simple and finds most
1537 * of the interesting cases.
1538 *
1539 * Note that must and mlen got initialized during setup.
1540 */
1541 static void
findmust(p,g)1542 findmust(p, g)
1543 struct parse *p;
1544 register struct re_guts *g;
1545 {
1546 register sop *scan;
1547 sop *start = NULL;
1548 register sop *newstart = NULL;
1549 register sopno newlen;
1550 register sop s;
1551 register char *cp;
1552 register sopno i;
1553
1554 /* avoid making error situations worse */
1555 if (p->error != 0)
1556 return;
1557
1558 /* find the longest OCHAR sequence in strip */
1559 newlen = 0;
1560 scan = g->strip + 1;
1561 do {
1562 s = *scan++;
1563 switch (OP(s)) {
1564 case OCHAR: /* sequence member */
1565 if (newlen == 0) /* new sequence */
1566 newstart = scan - 1;
1567 newlen++;
1568 break;
1569 case OPLUS_: /* things that don't break one */
1570 case OLPAREN:
1571 case ORPAREN:
1572 break;
1573 case OQUEST_: /* things that must be skipped */
1574 case OCH_:
1575 scan--;
1576 do {
1577 scan += OPND(s);
1578 s = *scan;
1579 /* assert() interferes w debug printouts */
1580 if (OP(s) != O_QUEST && OP(s) != O_CH &&
1581 OP(s) != OOR2) {
1582 g->iflags |= BAD;
1583 return;
1584 }
1585 } while (OP(s) != O_QUEST && OP(s) != O_CH);
1586 /* fallthrough */
1587 default: /* things that break a sequence */
1588 if (newlen > g->mlen) { /* ends one */
1589 start = newstart;
1590 g->mlen = newlen;
1591 }
1592 newlen = 0;
1593 break;
1594 }
1595 } while (OP(s) != OEND);
1596
1597 if (g->mlen == 0) /* there isn't one */
1598 return;
1599
1600 /* turn it into a character string */
1601 g->must = malloc((size_t)g->mlen + 1);
1602 if (g->must == NULL) { /* argh; just forget it */
1603 g->mlen = 0;
1604 return;
1605 }
1606 cp = g->must;
1607 scan = start;
1608 for (i = g->mlen; i > 0; i--) {
1609 while (OP(s = *scan++) != OCHAR)
1610 continue;
1611 assert(cp < g->must + g->mlen);
1612 *cp++ = (char)OPND(s);
1613 }
1614 assert(cp == g->must + g->mlen);
1615 *cp++ = '\0'; /* just on general principles */
1616 }
1617
1618 /*
1619 - pluscount - count + nesting
1620 == static sopno pluscount(register struct parse *p, register struct re_guts *g);
1621 */
1622 static sopno /* nesting depth */
pluscount(p,g)1623 pluscount(p, g)
1624 struct parse *p;
1625 register struct re_guts *g;
1626 {
1627 register sop *scan;
1628 register sop s;
1629 register sopno plusnest = 0;
1630 register sopno maxnest = 0;
1631
1632 if (p->error != 0)
1633 return(0); /* there may not be an OEND */
1634
1635 scan = g->strip + 1;
1636 do {
1637 s = *scan++;
1638 switch (OP(s)) {
1639 case OPLUS_:
1640 plusnest++;
1641 break;
1642 case O_PLUS:
1643 if (plusnest > maxnest)
1644 maxnest = plusnest;
1645 plusnest--;
1646 break;
1647 }
1648 } while (OP(s) != OEND);
1649 if (plusnest != 0)
1650 g->iflags |= BAD;
1651 return(maxnest);
1652 }
1653