xref: /freebsd/usr.bin/sed/compile.c (revision 1669d8afc64812c8d2d1d147ae1fd42ff441e1b1)
1 /*-
2  * Copyright (c) 1992 Diomidis Spinellis.
3  * Copyright (c) 1992, 1993
4  *	The Regents of the University of California.  All rights reserved.
5  *
6  * This code is derived from software contributed to Berkeley by
7  * Diomidis Spinellis of Imperial College, University of London.
8  *
9  * Redistribution and use in source and binary forms, with or without
10  * modification, are permitted provided that the following conditions
11  * are met:
12  * 1. Redistributions of source code must retain the above copyright
13  *    notice, this list of conditions and the following disclaimer.
14  * 2. Redistributions in binary form must reproduce the above copyright
15  *    notice, this list of conditions and the following disclaimer in the
16  *    documentation and/or other materials provided with the distribution.
17  * 4. Neither the name of the University nor the names of its contributors
18  *    may be used to endorse or promote products derived from this software
19  *    without specific prior written permission.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31  * SUCH DAMAGE.
32  */
33 
34 #include <sys/cdefs.h>
35 __FBSDID("$FreeBSD$");
36 
37 #ifndef lint
38 static const char sccsid[] = "@(#)compile.c	8.1 (Berkeley) 6/6/93";
39 #endif
40 
41 #include <sys/types.h>
42 #include <sys/stat.h>
43 
44 #include <ctype.h>
45 #include <err.h>
46 #include <errno.h>
47 #include <fcntl.h>
48 #include <limits.h>
49 #include <regex.h>
50 #include <stdio.h>
51 #include <stdlib.h>
52 #include <string.h>
53 #include <wchar.h>
54 
55 #include "defs.h"
56 #include "extern.h"
57 
58 #define LHSZ	128
59 #define	LHMASK	(LHSZ - 1)
60 static struct labhash {
61 	struct	labhash *lh_next;
62 	u_int	lh_hash;
63 	struct	s_command *lh_cmd;
64 	int	lh_ref;
65 } *labels[LHSZ];
66 
67 static char	 *compile_addr(char *, struct s_addr *);
68 static char	 *compile_ccl(char **, char *);
69 static char	 *compile_delimited(char *, char *);
70 static char	 *compile_flags(char *, struct s_subst *);
71 static regex_t	 *compile_re(char *, int);
72 static char	 *compile_subst(char *, struct s_subst *);
73 static char	 *compile_text(void);
74 static char	 *compile_tr(char *, struct s_tr **);
75 static struct s_command
76 		**compile_stream(struct s_command **);
77 static char	 *duptoeol(char *, const char *);
78 static void	  enterlabel(struct s_command *);
79 static struct s_command
80 		 *findlabel(char *);
81 static void	  fixuplabel(struct s_command *, struct s_command *);
82 static void	  uselabel(void);
83 
84 /*
85  * Command specification.  This is used to drive the command parser.
86  */
87 struct s_format {
88 	char code;				/* Command code */
89 	int naddr;				/* Number of address args */
90 	enum e_args args;			/* Argument type */
91 };
92 
93 static struct s_format cmd_fmts[] = {
94 	{'{', 2, GROUP},
95 	{'}', 0, ENDGROUP},
96 	{'a', 1, TEXT},
97 	{'b', 2, BRANCH},
98 	{'c', 2, TEXT},
99 	{'d', 2, EMPTY},
100 	{'D', 2, EMPTY},
101 	{'g', 2, EMPTY},
102 	{'G', 2, EMPTY},
103 	{'h', 2, EMPTY},
104 	{'H', 2, EMPTY},
105 	{'i', 1, TEXT},
106 	{'l', 2, EMPTY},
107 	{'n', 2, EMPTY},
108 	{'N', 2, EMPTY},
109 	{'p', 2, EMPTY},
110 	{'P', 2, EMPTY},
111 	{'q', 1, EMPTY},
112 	{'r', 1, RFILE},
113 	{'s', 2, SUBST},
114 	{'t', 2, BRANCH},
115 	{'w', 2, WFILE},
116 	{'x', 2, EMPTY},
117 	{'y', 2, TR},
118 	{'!', 2, NONSEL},
119 	{':', 0, LABEL},
120 	{'#', 0, COMMENT},
121 	{'=', 1, EMPTY},
122 	{'\0', 0, COMMENT},
123 };
124 
125 /* The compiled program. */
126 struct s_command *prog;
127 
128 /*
129  * Compile the program into prog.
130  * Initialise appends.
131  */
132 void
133 compile(void)
134 {
135 	*compile_stream(&prog) = NULL;
136 	fixuplabel(prog, NULL);
137 	uselabel();
138 	if (appendnum == 0)
139 		appends = NULL;
140 	else if ((appends = malloc(sizeof(struct s_appends) * appendnum)) ==
141 	    NULL)
142 		err(1, "malloc");
143 	if ((match = malloc((maxnsub + 1) * sizeof(regmatch_t))) == NULL)
144 		err(1, "malloc");
145 }
146 
147 #define EATSPACE() do {							\
148 	if (p)								\
149 		while (*p && isspace((unsigned char)*p))                \
150 			p++;						\
151 	} while (0)
152 
153 static struct s_command **
154 compile_stream(struct s_command **link)
155 {
156 	char *p;
157 	static char lbuf[_POSIX2_LINE_MAX + 1];	/* To save stack */
158 	struct s_command *cmd, *cmd2, *stack;
159 	struct s_format *fp;
160 	char re[_POSIX2_LINE_MAX + 1];
161 	int naddr;				/* Number of addresses */
162 
163 	stack = 0;
164 	for (;;) {
165 		if ((p = cu_fgets(lbuf, sizeof(lbuf), NULL)) == NULL) {
166 			if (stack != 0)
167 				errx(1, "%lu: %s: unexpected EOF (pending }'s)",
168 							linenum, fname);
169 			return (link);
170 		}
171 
172 semicolon:	EATSPACE();
173 		if (p) {
174 			if (*p == '#' || *p == '\0')
175 				continue;
176 			else if (*p == ';') {
177 				p++;
178 				goto semicolon;
179 			}
180 		}
181 		if ((*link = cmd = malloc(sizeof(struct s_command))) == NULL)
182 			err(1, "malloc");
183 		link = &cmd->next;
184 		cmd->nonsel = cmd->inrange = 0;
185 		/* First parse the addresses */
186 		naddr = 0;
187 
188 /* Valid characters to start an address */
189 #define	addrchar(c)	(strchr("0123456789/\\$", (c)))
190 		if (addrchar(*p)) {
191 			naddr++;
192 			if ((cmd->a1 = malloc(sizeof(struct s_addr))) == NULL)
193 				err(1, "malloc");
194 			p = compile_addr(p, cmd->a1);
195 			EATSPACE();				/* EXTENSION */
196 			if (*p == ',') {
197 				p++;
198 				EATSPACE();			/* EXTENSION */
199 				naddr++;
200 				if ((cmd->a2 = malloc(sizeof(struct s_addr)))
201 				    == NULL)
202 					err(1, "malloc");
203 				p = compile_addr(p, cmd->a2);
204 				EATSPACE();
205 			} else
206 				cmd->a2 = 0;
207 		} else
208 			cmd->a1 = cmd->a2 = 0;
209 
210 nonsel:		/* Now parse the command */
211 		if (!*p)
212 			errx(1, "%lu: %s: command expected", linenum, fname);
213 		cmd->code = *p;
214 		for (fp = cmd_fmts; fp->code; fp++)
215 			if (fp->code == *p)
216 				break;
217 		if (!fp->code)
218 			errx(1, "%lu: %s: invalid command code %c", linenum, fname, *p);
219 		if (naddr > fp->naddr)
220 			errx(1,
221 				"%lu: %s: command %c expects up to %d address(es), found %d",
222 				linenum, fname, *p, fp->naddr, naddr);
223 		switch (fp->args) {
224 		case NONSEL:			/* ! */
225 			p++;
226 			EATSPACE();
227 			cmd->nonsel = ! cmd->nonsel;
228 			goto nonsel;
229 		case GROUP:			/* { */
230 			p++;
231 			EATSPACE();
232 			cmd->next = stack;
233 			stack = cmd;
234 			link = &cmd->u.c;
235 			if (*p)
236 				goto semicolon;
237 			break;
238 		case ENDGROUP:
239 			/*
240 			 * Short-circuit command processing, since end of
241 			 * group is really just a noop.
242 			 */
243 			cmd->nonsel = 1;
244 			if (stack == 0)
245 				errx(1, "%lu: %s: unexpected }", linenum, fname);
246 			cmd2 = stack;
247 			stack = cmd2->next;
248 			cmd2->next = cmd;
249 			/*FALLTHROUGH*/
250 		case EMPTY:		/* d D g G h H l n N p P q x = \0 */
251 			p++;
252 			EATSPACE();
253 			if (*p == ';') {
254 				p++;
255 				link = &cmd->next;
256 				goto semicolon;
257 			}
258 			if (*p)
259 				errx(1, "%lu: %s: extra characters at the end of %c command",
260 						linenum, fname, cmd->code);
261 			break;
262 		case TEXT:			/* a c i */
263 			p++;
264 			EATSPACE();
265 			if (*p != '\\')
266 				errx(1,
267 "%lu: %s: command %c expects \\ followed by text", linenum, fname, cmd->code);
268 			p++;
269 			EATSPACE();
270 			if (*p)
271 				errx(1,
272 				"%lu: %s: extra characters after \\ at the end of %c command",
273 				linenum, fname, cmd->code);
274 			cmd->t = compile_text();
275 			break;
276 		case COMMENT:			/* \0 # */
277 			break;
278 		case WFILE:			/* w */
279 			p++;
280 			EATSPACE();
281 			if (*p == '\0')
282 				errx(1, "%lu: %s: filename expected", linenum, fname);
283 			cmd->t = duptoeol(p, "w command");
284 			if (aflag)
285 				cmd->u.fd = -1;
286 			else if ((cmd->u.fd = open(p,
287 			    O_WRONLY|O_APPEND|O_CREAT|O_TRUNC,
288 			    DEFFILEMODE)) == -1)
289 				err(1, "%s", p);
290 			break;
291 		case RFILE:			/* r */
292 			p++;
293 			EATSPACE();
294 			if (*p == '\0')
295 				errx(1, "%lu: %s: filename expected", linenum, fname);
296 			else
297 				cmd->t = duptoeol(p, "read command");
298 			break;
299 		case BRANCH:			/* b t */
300 			p++;
301 			EATSPACE();
302 			if (*p == '\0')
303 				cmd->t = NULL;
304 			else
305 				cmd->t = duptoeol(p, "branch");
306 			break;
307 		case LABEL:			/* : */
308 			p++;
309 			EATSPACE();
310 			cmd->t = duptoeol(p, "label");
311 			if (strlen(p) == 0)
312 				errx(1, "%lu: %s: empty label", linenum, fname);
313 			enterlabel(cmd);
314 			break;
315 		case SUBST:			/* s */
316 			p++;
317 			if (*p == '\0' || *p == '\\')
318 				errx(1,
319 "%lu: %s: substitute pattern can not be delimited by newline or backslash",
320 					linenum, fname);
321 			if ((cmd->u.s = calloc(1, sizeof(struct s_subst))) == NULL)
322 				err(1, "malloc");
323 			p = compile_delimited(p, re);
324 			if (p == NULL)
325 				errx(1,
326 				"%lu: %s: unterminated substitute pattern", linenum, fname);
327 			--p;
328 			p = compile_subst(p, cmd->u.s);
329 			p = compile_flags(p, cmd->u.s);
330 			if (*re == '\0')
331 				cmd->u.s->re = NULL;
332 			else
333 				cmd->u.s->re = compile_re(re, cmd->u.s->icase);
334 			EATSPACE();
335 			if (*p == ';') {
336 				p++;
337 				link = &cmd->next;
338 				goto semicolon;
339 			}
340 			break;
341 		case TR:			/* y */
342 			p++;
343 			p = compile_tr(p, &cmd->u.y);
344 			EATSPACE();
345 			if (*p == ';') {
346 				p++;
347 				link = &cmd->next;
348 				goto semicolon;
349 			}
350 			if (*p)
351 				errx(1,
352 "%lu: %s: extra text at the end of a transform command", linenum, fname);
353 			break;
354 		}
355 	}
356 }
357 
358 /*
359  * Get a delimited string.  P points to the delimeter of the string; d points
360  * to a buffer area.  Newline and delimiter escapes are processed; other
361  * escapes are ignored.
362  *
363  * Returns a pointer to the first character after the final delimiter or NULL
364  * in the case of a non-terminated string.  The character array d is filled
365  * with the processed string.
366  */
367 static char *
368 compile_delimited(char *p, char *d)
369 {
370 	char c;
371 
372 	c = *p++;
373 	if (c == '\0')
374 		return (NULL);
375 	else if (c == '\\')
376 		errx(1, "%lu: %s: \\ can not be used as a string delimiter",
377 				linenum, fname);
378 	else if (c == '\n')
379 		errx(1, "%lu: %s: newline can not be used as a string delimiter",
380 				linenum, fname);
381 	while (*p) {
382 		if (*p == '[') {
383 			if ((d = compile_ccl(&p, d)) == NULL)
384 				errx(1, "%lu: %s: unbalanced brackets ([])", linenum, fname);
385 			continue;
386 		} else if (*p == '\\' && p[1] == '[') {
387 			*d++ = *p++;
388 		} else if (*p == '\\' && p[1] == c)
389 			p++;
390 		else if (*p == '\\' && p[1] == 'n') {
391 			*d++ = '\n';
392 			p += 2;
393 			continue;
394 		} else if (*p == '\\' && p[1] == '\\')
395 			*d++ = *p++;
396 		else if (*p == c) {
397 			*d = '\0';
398 			return (p + 1);
399 		}
400 		*d++ = *p++;
401 	}
402 	return (NULL);
403 }
404 
405 
406 /* compile_ccl: expand a POSIX character class */
407 static char *
408 compile_ccl(char **sp, char *t)
409 {
410 	int c, d;
411 	char *s = *sp;
412 
413 	*t++ = *s++;
414 	if (*s == '^')
415 		*t++ = *s++;
416 	if (*s == ']')
417 		*t++ = *s++;
418 	for (; *s && (*t = *s) != ']'; s++, t++)
419 		if (*s == '[' && ((d = *(s+1)) == '.' || d == ':' || d == '=')) {
420 			*++t = *++s, t++, s++;
421 			for (c = *s; (*t = *s) != ']' || c != d; s++, t++)
422 				if ((c = *s) == '\0')
423 					return NULL;
424 		} else if (*s == '\\' && s[1] == 'n')
425 			    *t = '\n', s++;
426 	return (*s == ']') ? *sp = ++s, ++t : NULL;
427 }
428 
429 /*
430  * Compiles the regular expression in RE and returns a pointer to the compiled
431  * regular expression.
432  * Cflags are passed to regcomp.
433  */
434 static regex_t *
435 compile_re(char *re, int case_insensitive)
436 {
437 	regex_t *rep;
438 	int eval, flags;
439 
440 
441 	flags = rflags;
442 	if (case_insensitive)
443 		flags |= REG_ICASE;
444 	if ((rep = malloc(sizeof(regex_t))) == NULL)
445 		err(1, "malloc");
446 	if ((eval = regcomp(rep, re, flags)) != 0)
447 		errx(1, "%lu: %s: RE error: %s",
448 				linenum, fname, strregerror(eval, rep));
449 	if (maxnsub < rep->re_nsub)
450 		maxnsub = rep->re_nsub;
451 	return (rep);
452 }
453 
454 /*
455  * Compile the substitution string of a regular expression and set res to
456  * point to a saved copy of it.  Nsub is the number of parenthesized regular
457  * expressions.
458  */
459 static char *
460 compile_subst(char *p, struct s_subst *s)
461 {
462 	static char lbuf[_POSIX2_LINE_MAX + 1];
463 	int asize, size;
464 	u_char ref;
465 	char c, *text, *op, *sp;
466 	int more = 1, sawesc = 0;
467 
468 	c = *p++;			/* Terminator character */
469 	if (c == '\0')
470 		return (NULL);
471 
472 	s->maxbref = 0;
473 	s->linenum = linenum;
474 	asize = 2 * _POSIX2_LINE_MAX + 1;
475 	if ((text = malloc(asize)) == NULL)
476 		err(1, "malloc");
477 	size = 0;
478 	do {
479 		op = sp = text + size;
480 		for (; *p; p++) {
481 			if (*p == '\\' || sawesc) {
482 				/*
483 				 * If this is a continuation from the last
484 				 * buffer, we won't have a character to
485 				 * skip over.
486 				 */
487 				if (sawesc)
488 					sawesc = 0;
489 				else
490 					p++;
491 
492 				if (*p == '\0') {
493 					/*
494 					 * This escaped character is continued
495 					 * in the next part of the line.  Note
496 					 * this fact, then cause the loop to
497 					 * exit w/ normal EOL case and reenter
498 					 * above with the new buffer.
499 					 */
500 					sawesc = 1;
501 					p--;
502 					continue;
503 				} else if (strchr("123456789", *p) != NULL) {
504 					*sp++ = '\\';
505 					ref = *p - '0';
506 					if (s->re != NULL &&
507 					    ref > s->re->re_nsub)
508 						errx(1, "%lu: %s: \\%c not defined in the RE",
509 								linenum, fname, *p);
510 					if (s->maxbref < ref)
511 						s->maxbref = ref;
512 				} else if (*p == '&' || *p == '\\')
513 					*sp++ = '\\';
514 			} else if (*p == c) {
515 				if (*++p == '\0' && more) {
516 					if (cu_fgets(lbuf, sizeof(lbuf), &more))
517 						p = lbuf;
518 				}
519 				*sp++ = '\0';
520 				size += sp - op;
521 				if ((s->new = realloc(text, size)) == NULL)
522 					err(1, "realloc");
523 				return (p);
524 			} else if (*p == '\n') {
525 				errx(1,
526 "%lu: %s: unescaped newline inside substitute pattern", linenum, fname);
527 				/* NOTREACHED */
528 			}
529 			*sp++ = *p;
530 		}
531 		size += sp - op;
532 		if (asize - size < _POSIX2_LINE_MAX + 1) {
533 			asize *= 2;
534 			if ((text = realloc(text, asize)) == NULL)
535 				err(1, "realloc");
536 		}
537 	} while (cu_fgets(p = lbuf, sizeof(lbuf), &more));
538 	errx(1, "%lu: %s: unterminated substitute in regular expression",
539 			linenum, fname);
540 	/* NOTREACHED */
541 }
542 
543 /*
544  * Compile the flags of the s command
545  */
546 static char *
547 compile_flags(char *p, struct s_subst *s)
548 {
549 	int gn;			/* True if we have seen g or n */
550 	unsigned long nval;
551 	char wfile[_POSIX2_LINE_MAX + 1], *q;
552 
553 	s->n = 1;				/* Default */
554 	s->p = 0;
555 	s->wfile = NULL;
556 	s->wfd = -1;
557 	s->icase = 0;
558 	for (gn = 0;;) {
559 		EATSPACE();			/* EXTENSION */
560 		switch (*p) {
561 		case 'g':
562 			if (gn)
563 				errx(1,
564 "%lu: %s: more than one number or 'g' in substitute flags", linenum, fname);
565 			gn = 1;
566 			s->n = 0;
567 			break;
568 		case '\0':
569 		case '\n':
570 		case ';':
571 			return (p);
572 		case 'p':
573 			s->p = 1;
574 			break;
575 		case 'I':
576 			s->icase = 1;
577 			break;
578 		case '1': case '2': case '3':
579 		case '4': case '5': case '6':
580 		case '7': case '8': case '9':
581 			if (gn)
582 				errx(1,
583 "%lu: %s: more than one number or 'g' in substitute flags", linenum, fname);
584 			gn = 1;
585 			errno = 0;
586 			nval = strtol(p, &p, 10);
587 			if (errno == ERANGE || nval > INT_MAX)
588 				errx(1,
589 "%lu: %s: overflow in the 'N' substitute flag", linenum, fname);
590 			s->n = nval;
591 			p--;
592 			break;
593 		case 'w':
594 			p++;
595 #ifdef HISTORIC_PRACTICE
596 			if (*p != ' ') {
597 				warnx("%lu: %s: space missing before w wfile", linenum, fname);
598 				return (p);
599 			}
600 #endif
601 			EATSPACE();
602 			q = wfile;
603 			while (*p) {
604 				if (*p == '\n')
605 					break;
606 				*q++ = *p++;
607 			}
608 			*q = '\0';
609 			if (q == wfile)
610 				errx(1, "%lu: %s: no wfile specified", linenum, fname);
611 			s->wfile = strdup(wfile);
612 			if (!aflag && (s->wfd = open(wfile,
613 			    O_WRONLY|O_APPEND|O_CREAT|O_TRUNC,
614 			    DEFFILEMODE)) == -1)
615 				err(1, "%s", wfile);
616 			return (p);
617 		default:
618 			errx(1, "%lu: %s: bad flag in substitute command: '%c'",
619 					linenum, fname, *p);
620 			break;
621 		}
622 		p++;
623 	}
624 }
625 
626 /*
627  * Compile a translation set of strings into a lookup table.
628  */
629 static char *
630 compile_tr(char *p, struct s_tr **py)
631 {
632 	struct s_tr *y;
633 	int i;
634 	const char *op, *np;
635 	char old[_POSIX2_LINE_MAX + 1];
636 	char new[_POSIX2_LINE_MAX + 1];
637 	size_t oclen, oldlen, nclen, newlen;
638 	mbstate_t mbs1, mbs2;
639 
640 	if ((*py = y = malloc(sizeof(*y))) == NULL)
641 		err(1, NULL);
642 	y->multis = NULL;
643 	y->nmultis = 0;
644 
645 	if (*p == '\0' || *p == '\\')
646 		errx(1,
647 	"%lu: %s: transform pattern can not be delimited by newline or backslash",
648 			linenum, fname);
649 	p = compile_delimited(p, old);
650 	if (p == NULL)
651 		errx(1, "%lu: %s: unterminated transform source string",
652 				linenum, fname);
653 	p = compile_delimited(p - 1, new);
654 	if (p == NULL)
655 		errx(1, "%lu: %s: unterminated transform target string",
656 				linenum, fname);
657 	EATSPACE();
658 	op = old;
659 	oldlen = mbsrtowcs(NULL, &op, 0, NULL);
660 	if (oldlen == (size_t)-1)
661 		err(1, NULL);
662 	np = new;
663 	newlen = mbsrtowcs(NULL, &np, 0, NULL);
664 	if (newlen == (size_t)-1)
665 		err(1, NULL);
666 	if (newlen != oldlen)
667 		errx(1, "%lu: %s: transform strings are not the same length",
668 				linenum, fname);
669 	if (MB_CUR_MAX == 1) {
670 		/*
671 		 * The single-byte encoding case is easy: generate a
672 		 * lookup table.
673 		 */
674 		for (i = 0; i <= UCHAR_MAX; i++)
675 			y->bytetab[i] = (char)i;
676 		for (; *op; op++, np++)
677 			y->bytetab[(u_char)*op] = *np;
678 	} else {
679 		/*
680 		 * Multi-byte encoding case: generate a lookup table as
681 		 * above, but only for single-byte characters. The first
682 		 * bytes of multi-byte characters have their lookup table
683 		 * entries set to 0, which causes do_tr() to search through
684 		 * an auxiliary vector of multi-byte mappings.
685 		 */
686 		memset(&mbs1, 0, sizeof(mbs1));
687 		memset(&mbs2, 0, sizeof(mbs2));
688 		for (i = 0; i <= UCHAR_MAX; i++)
689 			y->bytetab[i] = (btowc(i) != WEOF) ? i : 0;
690 		while (*op != '\0') {
691 			oclen = mbrlen(op, MB_LEN_MAX, &mbs1);
692 			if (oclen == (size_t)-1 || oclen == (size_t)-2)
693 				errc(1, EILSEQ, NULL);
694 			nclen = mbrlen(np, MB_LEN_MAX, &mbs2);
695 			if (nclen == (size_t)-1 || nclen == (size_t)-2)
696 				errc(1, EILSEQ, NULL);
697 			if (oclen == 1 && nclen == 1)
698 				y->bytetab[(u_char)*op] = *np;
699 			else {
700 				y->bytetab[(u_char)*op] = 0;
701 				y->multis = realloc(y->multis,
702 				    (y->nmultis + 1) * sizeof(*y->multis));
703 				if (y->multis == NULL)
704 					err(1, NULL);
705 				i = y->nmultis++;
706 				y->multis[i].fromlen = oclen;
707 				memcpy(y->multis[i].from, op, oclen);
708 				y->multis[i].tolen = nclen;
709 				memcpy(y->multis[i].to, np, nclen);
710 			}
711 			op += oclen;
712 			np += nclen;
713 		}
714 	}
715 	return (p);
716 }
717 
718 /*
719  * Compile the text following an a or i command.
720  */
721 static char *
722 compile_text(void)
723 {
724 	int asize, esc_nl, size;
725 	char *text, *p, *op, *s;
726 	char lbuf[_POSIX2_LINE_MAX + 1];
727 
728 	asize = 2 * _POSIX2_LINE_MAX + 1;
729 	if ((text = malloc(asize)) == NULL)
730 		err(1, "malloc");
731 	size = 0;
732 	while (cu_fgets(lbuf, sizeof(lbuf), NULL)) {
733 		op = s = text + size;
734 		p = lbuf;
735 		EATSPACE();
736 		for (esc_nl = 0; *p != '\0'; p++) {
737 			if (*p == '\\' && p[1] != '\0' && *++p == '\n')
738 				esc_nl = 1;
739 			*s++ = *p;
740 		}
741 		size += s - op;
742 		if (!esc_nl) {
743 			*s = '\0';
744 			break;
745 		}
746 		if (asize - size < _POSIX2_LINE_MAX + 1) {
747 			asize *= 2;
748 			if ((text = realloc(text, asize)) == NULL)
749 				err(1, "realloc");
750 		}
751 	}
752 	text[size] = '\0';
753 	if ((p = realloc(text, size + 1)) == NULL)
754 		err(1, "realloc");
755 	return (p);
756 }
757 
758 /*
759  * Get an address and return a pointer to the first character after
760  * it.  Fill the structure pointed to according to the address.
761  */
762 static char *
763 compile_addr(char *p, struct s_addr *a)
764 {
765 	char *end, re[_POSIX2_LINE_MAX + 1];
766 	int icase;
767 
768 	icase = 0;
769 
770 	switch (*p) {
771 	case '\\':				/* Context address */
772 		++p;
773 		/* FALLTHROUGH */
774 	case '/':				/* Context address */
775 		p = compile_delimited(p, re);
776 		if (p == NULL)
777 			errx(1, "%lu: %s: unterminated regular expression", linenum, fname);
778 		/* Check for case insensitive regexp flag */
779 		if (*p == 'I') {
780 			icase = 1;
781 			p++;
782 		}
783 		if (*re == '\0')
784 			a->u.r = NULL;
785 		else
786 			a->u.r = compile_re(re, icase);
787 		a->type = AT_RE;
788 		return (p);
789 
790 	case '$':				/* Last line */
791 		a->type = AT_LAST;
792 		return (p + 1);
793 						/* Line number */
794 	case '0': case '1': case '2': case '3': case '4':
795 	case '5': case '6': case '7': case '8': case '9':
796 		a->type = AT_LINE;
797 		a->u.l = strtol(p, &end, 10);
798 		return (end);
799 	default:
800 		errx(1, "%lu: %s: expected context address", linenum, fname);
801 		return (NULL);
802 	}
803 }
804 
805 /*
806  * duptoeol --
807  *	Return a copy of all the characters up to \n or \0.
808  */
809 static char *
810 duptoeol(char *s, const char *ctype)
811 {
812 	size_t len;
813 	int ws;
814 	char *p, *start;
815 
816 	ws = 0;
817 	for (start = s; *s != '\0' && *s != '\n'; ++s)
818 		ws = isspace((unsigned char)*s);
819 	*s = '\0';
820 	if (ws)
821 		warnx("%lu: %s: whitespace after %s", linenum, fname, ctype);
822 	len = s - start + 1;
823 	if ((p = malloc(len)) == NULL)
824 		err(1, "malloc");
825 	return (memmove(p, start, len));
826 }
827 
828 /*
829  * Convert goto label names to addresses, and count a and r commands, in
830  * the given subset of the script.  Free the memory used by labels in b
831  * and t commands (but not by :).
832  *
833  * TODO: Remove } nodes
834  */
835 static void
836 fixuplabel(struct s_command *cp, struct s_command *end)
837 {
838 
839 	for (; cp != end; cp = cp->next)
840 		switch (cp->code) {
841 		case 'a':
842 		case 'r':
843 			appendnum++;
844 			break;
845 		case 'b':
846 		case 't':
847 			/* Resolve branch target. */
848 			if (cp->t == NULL) {
849 				cp->u.c = NULL;
850 				break;
851 			}
852 			if ((cp->u.c = findlabel(cp->t)) == NULL)
853 				errx(1, "%lu: %s: undefined label '%s'", linenum, fname, cp->t);
854 			free(cp->t);
855 			break;
856 		case '{':
857 			/* Do interior commands. */
858 			fixuplabel(cp->u.c, cp->next);
859 			break;
860 		}
861 }
862 
863 /*
864  * Associate the given command label for later lookup.
865  */
866 static void
867 enterlabel(struct s_command *cp)
868 {
869 	struct labhash **lhp, *lh;
870 	u_char *p;
871 	u_int h, c;
872 
873 	for (h = 0, p = (u_char *)cp->t; (c = *p) != 0; p++)
874 		h = (h << 5) + h + c;
875 	lhp = &labels[h & LHMASK];
876 	for (lh = *lhp; lh != NULL; lh = lh->lh_next)
877 		if (lh->lh_hash == h && strcmp(cp->t, lh->lh_cmd->t) == 0)
878 			errx(1, "%lu: %s: duplicate label '%s'", linenum, fname, cp->t);
879 	if ((lh = malloc(sizeof *lh)) == NULL)
880 		err(1, "malloc");
881 	lh->lh_next = *lhp;
882 	lh->lh_hash = h;
883 	lh->lh_cmd = cp;
884 	lh->lh_ref = 0;
885 	*lhp = lh;
886 }
887 
888 /*
889  * Find the label contained in the command l in the command linked
890  * list cp.  L is excluded from the search.  Return NULL if not found.
891  */
892 static struct s_command *
893 findlabel(char *name)
894 {
895 	struct labhash *lh;
896 	u_char *p;
897 	u_int h, c;
898 
899 	for (h = 0, p = (u_char *)name; (c = *p) != 0; p++)
900 		h = (h << 5) + h + c;
901 	for (lh = labels[h & LHMASK]; lh != NULL; lh = lh->lh_next) {
902 		if (lh->lh_hash == h && strcmp(name, lh->lh_cmd->t) == 0) {
903 			lh->lh_ref = 1;
904 			return (lh->lh_cmd);
905 		}
906 	}
907 	return (NULL);
908 }
909 
910 /*
911  * Warn about any unused labels.  As a side effect, release the label hash
912  * table space.
913  */
914 static void
915 uselabel(void)
916 {
917 	struct labhash *lh, *next;
918 	int i;
919 
920 	for (i = 0; i < LHSZ; i++) {
921 		for (lh = labels[i]; lh != NULL; lh = next) {
922 			next = lh->lh_next;
923 			if (!lh->lh_ref)
924 				warnx("%lu: %s: unused label '%s'",
925 				    linenum, fname, lh->lh_cmd->t);
926 			free(lh);
927 		}
928 	}
929 }
930