xref: /freebsd/usr.bin/sed/compile.c (revision 2f513db7)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1992 Diomidis Spinellis.
5  * Copyright (c) 1992, 1993
6  *	The Regents of the University of California.  All rights reserved.
7  *
8  * This code is derived from software contributed to Berkeley by
9  * Diomidis Spinellis of Imperial College, University of London.
10  *
11  * Redistribution and use in source and binary forms, with or without
12  * modification, are permitted provided that the following conditions
13  * are met:
14  * 1. Redistributions of source code must retain the above copyright
15  *    notice, this list of conditions and the following disclaimer.
16  * 2. Redistributions in binary form must reproduce the above copyright
17  *    notice, this list of conditions and the following disclaimer in the
18  *    documentation and/or other materials provided with the distribution.
19  * 3. Neither the name of the University nor the names of its contributors
20  *    may be used to endorse or promote products derived from this software
21  *    without specific prior written permission.
22  *
23  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
24  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
25  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
26  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
27  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
28  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
29  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
30  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
31  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
32  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33  * SUCH DAMAGE.
34  */
35 
36 #include <sys/cdefs.h>
37 __FBSDID("$FreeBSD$");
38 
39 #ifndef lint
40 static const char sccsid[] = "@(#)compile.c	8.1 (Berkeley) 6/6/93";
41 #endif
42 
43 #include <sys/types.h>
44 #include <sys/stat.h>
45 
46 #include <ctype.h>
47 #include <err.h>
48 #include <errno.h>
49 #include <fcntl.h>
50 #include <limits.h>
51 #include <regex.h>
52 #include <stdio.h>
53 #include <stdlib.h>
54 #include <string.h>
55 #include <wchar.h>
56 
57 #include "defs.h"
58 #include "extern.h"
59 
60 #define LHSZ	128
61 #define	LHMASK	(LHSZ - 1)
62 static struct labhash {
63 	struct	labhash *lh_next;
64 	u_int	lh_hash;
65 	struct	s_command *lh_cmd;
66 	int	lh_ref;
67 } *labels[LHSZ];
68 
69 static char	 *compile_addr(char *, struct s_addr *);
70 static char	 *compile_ccl(char **, char *);
71 static char	 *compile_delimited(char *, char *, int);
72 static char	 *compile_flags(char *, struct s_subst *);
73 static regex_t	 *compile_re(char *, int);
74 static char	 *compile_subst(char *, struct s_subst *);
75 static char	 *compile_text(void);
76 static char	 *compile_tr(char *, struct s_tr **);
77 static struct s_command
78 		**compile_stream(struct s_command **);
79 static char	 *duptoeol(char *, const char *);
80 static void	  enterlabel(struct s_command *);
81 static struct s_command
82 		 *findlabel(char *);
83 static void	  fixuplabel(struct s_command *, struct s_command *);
84 static void	  uselabel(void);
85 
86 /*
87  * Command specification.  This is used to drive the command parser.
88  */
89 struct s_format {
90 	char code;				/* Command code */
91 	int naddr;				/* Number of address args */
92 	enum e_args args;			/* Argument type */
93 };
94 
95 static struct s_format cmd_fmts[] = {
96 	{'{', 2, GROUP},
97 	{'}', 0, ENDGROUP},
98 	{'a', 1, TEXT},
99 	{'b', 2, BRANCH},
100 	{'c', 2, TEXT},
101 	{'d', 2, EMPTY},
102 	{'D', 2, EMPTY},
103 	{'g', 2, EMPTY},
104 	{'G', 2, EMPTY},
105 	{'h', 2, EMPTY},
106 	{'H', 2, EMPTY},
107 	{'i', 1, TEXT},
108 	{'l', 2, EMPTY},
109 	{'n', 2, EMPTY},
110 	{'N', 2, EMPTY},
111 	{'p', 2, EMPTY},
112 	{'P', 2, EMPTY},
113 	{'q', 1, EMPTY},
114 	{'r', 1, RFILE},
115 	{'s', 2, SUBST},
116 	{'t', 2, BRANCH},
117 	{'w', 2, WFILE},
118 	{'x', 2, EMPTY},
119 	{'y', 2, TR},
120 	{'!', 2, NONSEL},
121 	{':', 0, LABEL},
122 	{'#', 0, COMMENT},
123 	{'=', 1, EMPTY},
124 	{'\0', 0, COMMENT},
125 };
126 
127 /* The compiled program. */
128 struct s_command *prog;
129 
130 /*
131  * Compile the program into prog.
132  * Initialise appends.
133  */
134 void
135 compile(void)
136 {
137 	*compile_stream(&prog) = NULL;
138 	fixuplabel(prog, NULL);
139 	uselabel();
140 	if (appendnum == 0)
141 		appends = NULL;
142 	else if ((appends = malloc(sizeof(struct s_appends) * appendnum)) ==
143 	    NULL)
144 		err(1, "malloc");
145 	if ((match = malloc((maxnsub + 1) * sizeof(regmatch_t))) == NULL)
146 		err(1, "malloc");
147 }
148 
149 #define EATSPACE() do {							\
150 	if (p)								\
151 		while (*p && isspace((unsigned char)*p))                \
152 			p++;						\
153 	} while (0)
154 
155 static struct s_command **
156 compile_stream(struct s_command **link)
157 {
158 	char *p;
159 	static char lbuf[_POSIX2_LINE_MAX + 1];	/* To save stack */
160 	struct s_command *cmd, *cmd2, *stack;
161 	struct s_format *fp;
162 	char re[_POSIX2_LINE_MAX + 1];
163 	int naddr;				/* Number of addresses */
164 
165 	stack = NULL;
166 	for (;;) {
167 		if ((p = cu_fgets(lbuf, sizeof(lbuf), NULL)) == NULL) {
168 			if (stack != NULL)
169 				errx(1, "%lu: %s: unexpected EOF (pending }'s)",
170 							linenum, fname);
171 			return (link);
172 		}
173 
174 semicolon:	EATSPACE();
175 		if (p) {
176 			if (*p == '#' || *p == '\0')
177 				continue;
178 			else if (*p == ';') {
179 				p++;
180 				goto semicolon;
181 			}
182 		}
183 		if ((*link = cmd = malloc(sizeof(struct s_command))) == NULL)
184 			err(1, "malloc");
185 		link = &cmd->next;
186 		cmd->startline = cmd->nonsel = 0;
187 		/* First parse the addresses */
188 		naddr = 0;
189 
190 /* Valid characters to start an address */
191 #define	addrchar(c)	(strchr("0123456789/\\$", (c)))
192 		if (addrchar(*p)) {
193 			naddr++;
194 			if ((cmd->a1 = malloc(sizeof(struct s_addr))) == NULL)
195 				err(1, "malloc");
196 			p = compile_addr(p, cmd->a1);
197 			EATSPACE();				/* EXTENSION */
198 			if (*p == ',') {
199 				p++;
200 				EATSPACE();			/* EXTENSION */
201 				naddr++;
202 				if ((cmd->a2 = malloc(sizeof(struct s_addr)))
203 				    == NULL)
204 					err(1, "malloc");
205 				p = compile_addr(p, cmd->a2);
206 				EATSPACE();
207 			} else
208 				cmd->a2 = NULL;
209 		} else
210 			cmd->a1 = cmd->a2 = NULL;
211 
212 nonsel:		/* Now parse the command */
213 		if (!*p)
214 			errx(1, "%lu: %s: command expected", linenum, fname);
215 		cmd->code = *p;
216 		for (fp = cmd_fmts; fp->code; fp++)
217 			if (fp->code == *p)
218 				break;
219 		if (!fp->code)
220 			errx(1, "%lu: %s: invalid command code %c", linenum, fname, *p);
221 		if (naddr > fp->naddr)
222 			errx(1,
223 				"%lu: %s: command %c expects up to %d address(es), found %d",
224 				linenum, fname, *p, fp->naddr, naddr);
225 		switch (fp->args) {
226 		case NONSEL:			/* ! */
227 			p++;
228 			EATSPACE();
229 			cmd->nonsel = 1;
230 			goto nonsel;
231 		case GROUP:			/* { */
232 			p++;
233 			EATSPACE();
234 			cmd->next = stack;
235 			stack = cmd;
236 			link = &cmd->u.c;
237 			if (*p)
238 				goto semicolon;
239 			break;
240 		case ENDGROUP:
241 			/*
242 			 * Short-circuit command processing, since end of
243 			 * group is really just a noop.
244 			 */
245 			cmd->nonsel = 1;
246 			if (stack == NULL)
247 				errx(1, "%lu: %s: unexpected }", linenum, fname);
248 			cmd2 = stack;
249 			stack = cmd2->next;
250 			cmd2->next = cmd;
251 			/*FALLTHROUGH*/
252 		case EMPTY:		/* d D g G h H l n N p P q x = \0 */
253 			p++;
254 			EATSPACE();
255 			if (*p == ';') {
256 				p++;
257 				link = &cmd->next;
258 				goto semicolon;
259 			}
260 			if (*p)
261 				errx(1, "%lu: %s: extra characters at the end of %c command",
262 						linenum, fname, cmd->code);
263 			break;
264 		case TEXT:			/* a c i */
265 			p++;
266 			EATSPACE();
267 			if (*p != '\\')
268 				errx(1,
269 "%lu: %s: command %c expects \\ followed by text", linenum, fname, cmd->code);
270 			p++;
271 			EATSPACE();
272 			if (*p)
273 				errx(1,
274 				"%lu: %s: extra characters after \\ at the end of %c command",
275 				linenum, fname, cmd->code);
276 			cmd->t = compile_text();
277 			break;
278 		case COMMENT:			/* \0 # */
279 			break;
280 		case WFILE:			/* w */
281 			p++;
282 			EATSPACE();
283 			if (*p == '\0')
284 				errx(1, "%lu: %s: filename expected", linenum, fname);
285 			cmd->t = duptoeol(p, "w command");
286 			if (aflag)
287 				cmd->u.fd = -1;
288 			else if ((cmd->u.fd = open(p,
289 			    O_WRONLY|O_APPEND|O_CREAT|O_TRUNC,
290 			    DEFFILEMODE)) == -1)
291 				err(1, "%s", p);
292 			break;
293 		case RFILE:			/* r */
294 			p++;
295 			EATSPACE();
296 			if (*p == '\0')
297 				errx(1, "%lu: %s: filename expected", linenum, fname);
298 			else
299 				cmd->t = duptoeol(p, "read command");
300 			break;
301 		case BRANCH:			/* b t */
302 			p++;
303 			EATSPACE();
304 			if (*p == '\0')
305 				cmd->t = NULL;
306 			else
307 				cmd->t = duptoeol(p, "branch");
308 			break;
309 		case LABEL:			/* : */
310 			p++;
311 			EATSPACE();
312 			cmd->t = duptoeol(p, "label");
313 			if (strlen(p) == 0)
314 				errx(1, "%lu: %s: empty label", linenum, fname);
315 			enterlabel(cmd);
316 			break;
317 		case SUBST:			/* s */
318 			p++;
319 			if (*p == '\0' || *p == '\\')
320 				errx(1,
321 "%lu: %s: substitute pattern can not be delimited by newline or backslash",
322 					linenum, fname);
323 			if ((cmd->u.s = calloc(1, sizeof(struct s_subst))) == NULL)
324 				err(1, "malloc");
325 			p = compile_delimited(p, re, 0);
326 			if (p == NULL)
327 				errx(1,
328 				"%lu: %s: unterminated substitute pattern", linenum, fname);
329 
330 			/* Compile RE with no case sensitivity temporarily */
331 			if (*re == '\0')
332 				cmd->u.s->re = NULL;
333 			else
334 				cmd->u.s->re = compile_re(re, 0);
335 			--p;
336 			p = compile_subst(p, cmd->u.s);
337 			p = compile_flags(p, cmd->u.s);
338 
339 			/* Recompile RE with case sensitivity from "I" flag if any */
340 			if (*re == '\0')
341 				cmd->u.s->re = NULL;
342 			else
343 				cmd->u.s->re = compile_re(re, cmd->u.s->icase);
344 			EATSPACE();
345 			if (*p == ';') {
346 				p++;
347 				link = &cmd->next;
348 				goto semicolon;
349 			}
350 			break;
351 		case TR:			/* y */
352 			p++;
353 			p = compile_tr(p, &cmd->u.y);
354 			EATSPACE();
355 			if (*p == ';') {
356 				p++;
357 				link = &cmd->next;
358 				goto semicolon;
359 			}
360 			if (*p)
361 				errx(1,
362 "%lu: %s: extra text at the end of a transform command", linenum, fname);
363 			break;
364 		}
365 	}
366 }
367 
368 /*
369  * Get a delimited string.  P points to the delimiter of the string; d points
370  * to a buffer area.  Newline and delimiter escapes are processed; other
371  * escapes are ignored.
372  *
373  * Returns a pointer to the first character after the final delimiter or NULL
374  * in the case of a non-terminated string.  The character array d is filled
375  * with the processed string.
376  */
377 static char *
378 compile_delimited(char *p, char *d, int is_tr)
379 {
380 	char c;
381 
382 	c = *p++;
383 	if (c == '\0')
384 		return (NULL);
385 	else if (c == '\\')
386 		errx(1, "%lu: %s: \\ can not be used as a string delimiter",
387 				linenum, fname);
388 	else if (c == '\n')
389 		errx(1, "%lu: %s: newline can not be used as a string delimiter",
390 				linenum, fname);
391 	while (*p) {
392 		if (*p == '[' && *p != c) {
393 			if ((d = compile_ccl(&p, d)) == NULL)
394 				errx(1, "%lu: %s: unbalanced brackets ([])", linenum, fname);
395 			continue;
396 		} else if (*p == '\\' && p[1] == '[') {
397 			*d++ = *p++;
398 		} else if (*p == '\\' && p[1] == c) {
399 			p++;
400 		} else if (*p == '\\' &&
401 		    (p[1] == 'n' || p[1] == 'r' || p[1] == 't')) {
402 			switch (p[1]) {
403 			case 'n':
404 				*d++ = '\n';
405 				break;
406 			case 'r':
407 				*d++ = '\r';
408 				break;
409 			case 't':
410 				*d++ = '\t';
411 				break;
412 			}
413 			p += 2;
414 			continue;
415 		} else if (*p == '\\' && p[1] == '\\') {
416 			if (is_tr)
417 				p++;
418 			else
419 				*d++ = *p++;
420 		} else if (*p == c) {
421 			*d = '\0';
422 			return (p + 1);
423 		}
424 		*d++ = *p++;
425 	}
426 	return (NULL);
427 }
428 
429 
430 /* compile_ccl: expand a POSIX character class */
431 static char *
432 compile_ccl(char **sp, char *t)
433 {
434 	int c, d;
435 	char *s = *sp;
436 
437 	*t++ = *s++;
438 	if (*s == '^')
439 		*t++ = *s++;
440 	if (*s == ']')
441 		*t++ = *s++;
442 	for (; *s && (*t = *s) != ']'; s++, t++) {
443 		if (*s == '[' && ((d = *(s+1)) == '.' || d == ':' || d == '=')) {
444 			*++t = *++s, t++, s++;
445 			for (c = *s; (*t = *s) != ']' || c != d; s++, t++)
446 				if ((c = *s) == '\0')
447 					return NULL;
448 		} else if (*s == '\\') {
449 			switch (s[1]) {
450 			case 'n':
451 				*t = '\n';
452 				s++;
453 				break;
454 			case 'r':
455 				*t = '\r';
456 				s++;
457 				break;
458 			case 't':
459 				*t = '\t';
460 				s++;
461 				break;
462 			}
463 		}
464 	}
465 	return (*s == ']') ? *sp = ++s, ++t : NULL;
466 }
467 
468 /*
469  * Compiles the regular expression in RE and returns a pointer to the compiled
470  * regular expression.
471  * Cflags are passed to regcomp.
472  */
473 static regex_t *
474 compile_re(char *re, int case_insensitive)
475 {
476 	regex_t *rep;
477 	int eval, flags;
478 
479 
480 	flags = rflags;
481 	if (case_insensitive)
482 		flags |= REG_ICASE;
483 	if ((rep = malloc(sizeof(regex_t))) == NULL)
484 		err(1, "malloc");
485 	if ((eval = regcomp(rep, re, flags)) != 0)
486 		errx(1, "%lu: %s: RE error: %s",
487 				linenum, fname, strregerror(eval, rep));
488 	if (maxnsub < rep->re_nsub)
489 		maxnsub = rep->re_nsub;
490 	return (rep);
491 }
492 
493 /*
494  * Compile the substitution string of a regular expression and set res to
495  * point to a saved copy of it.  Nsub is the number of parenthesized regular
496  * expressions.
497  */
498 static char *
499 compile_subst(char *p, struct s_subst *s)
500 {
501 	static char lbuf[_POSIX2_LINE_MAX + 1];
502 	int asize, size;
503 	u_char ref;
504 	char c, *text, *op, *sp;
505 	int more = 1, sawesc = 0;
506 
507 	c = *p++;			/* Terminator character */
508 	if (c == '\0')
509 		return (NULL);
510 
511 	s->maxbref = 0;
512 	s->linenum = linenum;
513 	asize = 2 * _POSIX2_LINE_MAX + 1;
514 	if ((text = malloc(asize)) == NULL)
515 		err(1, "malloc");
516 	size = 0;
517 	do {
518 		op = sp = text + size;
519 		for (; *p; p++) {
520 			if (*p == '\\' || sawesc) {
521 				/*
522 				 * If this is a continuation from the last
523 				 * buffer, we won't have a character to
524 				 * skip over.
525 				 */
526 				if (sawesc)
527 					sawesc = 0;
528 				else
529 					p++;
530 
531 				if (*p == '\0') {
532 					/*
533 					 * This escaped character is continued
534 					 * in the next part of the line.  Note
535 					 * this fact, then cause the loop to
536 					 * exit w/ normal EOL case and reenter
537 					 * above with the new buffer.
538 					 */
539 					sawesc = 1;
540 					p--;
541 					continue;
542 				} else if (strchr("123456789", *p) != NULL) {
543 					*sp++ = '\\';
544 					ref = *p - '0';
545 					if (s->re != NULL &&
546 					    ref > s->re->re_nsub)
547 						errx(1, "%lu: %s: \\%c not defined in the RE",
548 								linenum, fname, *p);
549 					if (s->maxbref < ref)
550 						s->maxbref = ref;
551 				} else {
552 					switch (*p) {
553 					case '&':
554 					case '\\':
555 						*sp++ = '\\';
556 						break;
557 					case 'n':
558 						*p = '\n';
559 						break;
560 					case 'r':
561 						*p = '\r';
562 						break;
563 					case 't':
564 						*p = '\t';
565 						break;
566 					}
567 				}
568 			} else if (*p == c) {
569 				if (*++p == '\0' && more) {
570 					if (cu_fgets(lbuf, sizeof(lbuf), &more))
571 						p = lbuf;
572 				}
573 				*sp++ = '\0';
574 				size += sp - op;
575 				if ((s->new = realloc(text, size)) == NULL)
576 					err(1, "realloc");
577 				return (p);
578 			} else if (*p == '\n') {
579 				errx(1,
580 "%lu: %s: unescaped newline inside substitute pattern", linenum, fname);
581 				/* NOTREACHED */
582 			}
583 			*sp++ = *p;
584 		}
585 		size += sp - op;
586 		if (asize - size < _POSIX2_LINE_MAX + 1) {
587 			asize *= 2;
588 			if ((text = realloc(text, asize)) == NULL)
589 				err(1, "realloc");
590 		}
591 	} while (cu_fgets(p = lbuf, sizeof(lbuf), &more) != NULL);
592 	errx(1, "%lu: %s: unterminated substitute in regular expression",
593 			linenum, fname);
594 	/* NOTREACHED */
595 }
596 
597 /*
598  * Compile the flags of the s command
599  */
600 static char *
601 compile_flags(char *p, struct s_subst *s)
602 {
603 	int gn;			/* True if we have seen g or n */
604 	unsigned long nval;
605 	char wfile[_POSIX2_LINE_MAX + 1], *q, *eq;
606 
607 	s->n = 1;				/* Default */
608 	s->p = 0;
609 	s->wfile = NULL;
610 	s->wfd = -1;
611 	s->icase = 0;
612 	for (gn = 0;;) {
613 		EATSPACE();			/* EXTENSION */
614 		switch (*p) {
615 		case 'g':
616 			if (gn)
617 				errx(1,
618 "%lu: %s: more than one number or 'g' in substitute flags", linenum, fname);
619 			gn = 1;
620 			s->n = 0;
621 			break;
622 		case '\0':
623 		case '\n':
624 		case ';':
625 			return (p);
626 		case 'p':
627 			s->p = 1;
628 			break;
629 		case 'i':
630 		case 'I':
631 			s->icase = 1;
632 			break;
633 		case '1': case '2': case '3':
634 		case '4': case '5': case '6':
635 		case '7': case '8': case '9':
636 			if (gn)
637 				errx(1,
638 "%lu: %s: more than one number or 'g' in substitute flags", linenum, fname);
639 			gn = 1;
640 			errno = 0;
641 			nval = strtol(p, &p, 10);
642 			if (errno == ERANGE || nval > INT_MAX)
643 				errx(1,
644 "%lu: %s: overflow in the 'N' substitute flag", linenum, fname);
645 			s->n = nval;
646 			p--;
647 			break;
648 		case 'w':
649 			p++;
650 #ifdef HISTORIC_PRACTICE
651 			if (*p != ' ') {
652 				warnx("%lu: %s: space missing before w wfile", linenum, fname);
653 				return (p);
654 			}
655 #endif
656 			EATSPACE();
657 			q = wfile;
658 			eq = wfile + sizeof(wfile) - 1;
659 			while (*p) {
660 				if (*p == '\n')
661 					break;
662 				if (q >= eq)
663 					err(1, "wfile too long");
664 				*q++ = *p++;
665 			}
666 			*q = '\0';
667 			if (q == wfile)
668 				errx(1, "%lu: %s: no wfile specified", linenum, fname);
669 			s->wfile = strdup(wfile);
670 			if (!aflag && (s->wfd = open(wfile,
671 			    O_WRONLY|O_APPEND|O_CREAT|O_TRUNC,
672 			    DEFFILEMODE)) == -1)
673 				err(1, "%s", wfile);
674 			return (p);
675 		default:
676 			errx(1, "%lu: %s: bad flag in substitute command: '%c'",
677 					linenum, fname, *p);
678 			break;
679 		}
680 		p++;
681 	}
682 }
683 
684 /*
685  * Compile a translation set of strings into a lookup table.
686  */
687 static char *
688 compile_tr(char *p, struct s_tr **py)
689 {
690 	struct s_tr *y;
691 	int i;
692 	const char *op, *np;
693 	char old[_POSIX2_LINE_MAX + 1];
694 	char new[_POSIX2_LINE_MAX + 1];
695 	size_t oclen, oldlen, nclen, newlen;
696 	mbstate_t mbs1, mbs2;
697 
698 	if ((*py = y = malloc(sizeof(*y))) == NULL)
699 		err(1, NULL);
700 	y->multis = NULL;
701 	y->nmultis = 0;
702 
703 	if (*p == '\0' || *p == '\\')
704 		errx(1,
705 	"%lu: %s: transform pattern can not be delimited by newline or backslash",
706 			linenum, fname);
707 	p = compile_delimited(p, old, 1);
708 	if (p == NULL)
709 		errx(1, "%lu: %s: unterminated transform source string",
710 				linenum, fname);
711 	p = compile_delimited(p - 1, new, 1);
712 	if (p == NULL)
713 		errx(1, "%lu: %s: unterminated transform target string",
714 				linenum, fname);
715 	EATSPACE();
716 	op = old;
717 	oldlen = mbsrtowcs(NULL, &op, 0, NULL);
718 	if (oldlen == (size_t)-1)
719 		err(1, NULL);
720 	np = new;
721 	newlen = mbsrtowcs(NULL, &np, 0, NULL);
722 	if (newlen == (size_t)-1)
723 		err(1, NULL);
724 	if (newlen != oldlen)
725 		errx(1, "%lu: %s: transform strings are not the same length",
726 				linenum, fname);
727 	if (MB_CUR_MAX == 1) {
728 		/*
729 		 * The single-byte encoding case is easy: generate a
730 		 * lookup table.
731 		 */
732 		for (i = 0; i <= UCHAR_MAX; i++)
733 			y->bytetab[i] = (char)i;
734 		for (; *op; op++, np++)
735 			y->bytetab[(u_char)*op] = *np;
736 	} else {
737 		/*
738 		 * Multi-byte encoding case: generate a lookup table as
739 		 * above, but only for single-byte characters. The first
740 		 * bytes of multi-byte characters have their lookup table
741 		 * entries set to 0, which causes do_tr() to search through
742 		 * an auxiliary vector of multi-byte mappings.
743 		 */
744 		memset(&mbs1, 0, sizeof(mbs1));
745 		memset(&mbs2, 0, sizeof(mbs2));
746 		for (i = 0; i <= UCHAR_MAX; i++)
747 			y->bytetab[i] = (btowc(i) != WEOF) ? i : 0;
748 		while (*op != '\0') {
749 			oclen = mbrlen(op, MB_LEN_MAX, &mbs1);
750 			if (oclen == (size_t)-1 || oclen == (size_t)-2)
751 				errc(1, EILSEQ, NULL);
752 			nclen = mbrlen(np, MB_LEN_MAX, &mbs2);
753 			if (nclen == (size_t)-1 || nclen == (size_t)-2)
754 				errc(1, EILSEQ, NULL);
755 			if (oclen == 1 && nclen == 1)
756 				y->bytetab[(u_char)*op] = *np;
757 			else {
758 				y->bytetab[(u_char)*op] = 0;
759 				y->multis = realloc(y->multis,
760 				    (y->nmultis + 1) * sizeof(*y->multis));
761 				if (y->multis == NULL)
762 					err(1, NULL);
763 				i = y->nmultis++;
764 				y->multis[i].fromlen = oclen;
765 				memcpy(y->multis[i].from, op, oclen);
766 				y->multis[i].tolen = nclen;
767 				memcpy(y->multis[i].to, np, nclen);
768 			}
769 			op += oclen;
770 			np += nclen;
771 		}
772 	}
773 	return (p);
774 }
775 
776 /*
777  * Compile the text following an a, c, or i command.
778  */
779 static char *
780 compile_text(void)
781 {
782 	int asize, esc_nl, size;
783 	char *text, *p, *op, *s;
784 	char lbuf[_POSIX2_LINE_MAX + 1];
785 
786 	asize = 2 * _POSIX2_LINE_MAX + 1;
787 	if ((text = malloc(asize)) == NULL)
788 		err(1, "malloc");
789 	size = 0;
790 	while (cu_fgets(lbuf, sizeof(lbuf), NULL) != NULL) {
791 		op = s = text + size;
792 		p = lbuf;
793 #ifdef LEGACY_BSDSED_COMPAT
794 		EATSPACE();
795 #endif
796 		for (esc_nl = 0; *p != '\0'; p++) {
797 			if (*p == '\\' && p[1] != '\0' && *++p == '\n')
798 				esc_nl = 1;
799 			*s++ = *p;
800 		}
801 		size += s - op;
802 		if (!esc_nl) {
803 			*s = '\0';
804 			break;
805 		}
806 		if (asize - size < _POSIX2_LINE_MAX + 1) {
807 			asize *= 2;
808 			if ((text = realloc(text, asize)) == NULL)
809 				err(1, "realloc");
810 		}
811 	}
812 	text[size] = '\0';
813 	if ((p = realloc(text, size + 1)) == NULL)
814 		err(1, "realloc");
815 	return (p);
816 }
817 
818 /*
819  * Get an address and return a pointer to the first character after
820  * it.  Fill the structure pointed to according to the address.
821  */
822 static char *
823 compile_addr(char *p, struct s_addr *a)
824 {
825 	char *end, re[_POSIX2_LINE_MAX + 1];
826 	int icase;
827 
828 	icase = 0;
829 
830 	a->type = 0;
831 	switch (*p) {
832 	case '\\':				/* Context address */
833 		++p;
834 		/* FALLTHROUGH */
835 	case '/':				/* Context address */
836 		p = compile_delimited(p, re, 0);
837 		if (p == NULL)
838 			errx(1, "%lu: %s: unterminated regular expression", linenum, fname);
839 		/* Check for case insensitive regexp flag */
840 		if (*p == 'I') {
841 			icase = 1;
842 			p++;
843 		}
844 		if (*re == '\0')
845 			a->u.r = NULL;
846 		else
847 			a->u.r = compile_re(re, icase);
848 		a->type = AT_RE;
849 		return (p);
850 
851 	case '$':				/* Last line */
852 		a->type = AT_LAST;
853 		return (p + 1);
854 
855 	case '+':				/* Relative line number */
856 		a->type = AT_RELLINE;
857 		p++;
858 		/* FALLTHROUGH */
859 						/* Line number */
860 	case '0': case '1': case '2': case '3': case '4':
861 	case '5': case '6': case '7': case '8': case '9':
862 		if (a->type == 0)
863 			a->type = AT_LINE;
864 		a->u.l = strtol(p, &end, 10);
865 		return (end);
866 	default:
867 		errx(1, "%lu: %s: expected context address", linenum, fname);
868 		return (NULL);
869 	}
870 }
871 
872 /*
873  * duptoeol --
874  *	Return a copy of all the characters up to \n or \0.
875  */
876 static char *
877 duptoeol(char *s, const char *ctype)
878 {
879 	size_t len;
880 	int ws;
881 	char *p, *start;
882 
883 	ws = 0;
884 	for (start = s; *s != '\0' && *s != '\n'; ++s)
885 		ws = isspace((unsigned char)*s);
886 	*s = '\0';
887 	if (ws)
888 		warnx("%lu: %s: whitespace after %s", linenum, fname, ctype);
889 	len = s - start + 1;
890 	if ((p = malloc(len)) == NULL)
891 		err(1, "malloc");
892 	return (memmove(p, start, len));
893 }
894 
895 /*
896  * Convert goto label names to addresses, and count a and r commands, in
897  * the given subset of the script.  Free the memory used by labels in b
898  * and t commands (but not by :).
899  *
900  * TODO: Remove } nodes
901  */
902 static void
903 fixuplabel(struct s_command *cp, struct s_command *end)
904 {
905 
906 	for (; cp != end; cp = cp->next)
907 		switch (cp->code) {
908 		case 'a':
909 		case 'r':
910 			appendnum++;
911 			break;
912 		case 'b':
913 		case 't':
914 			/* Resolve branch target. */
915 			if (cp->t == NULL) {
916 				cp->u.c = NULL;
917 				break;
918 			}
919 			if ((cp->u.c = findlabel(cp->t)) == NULL)
920 				errx(1, "%lu: %s: undefined label '%s'", linenum, fname, cp->t);
921 			free(cp->t);
922 			break;
923 		case '{':
924 			/* Do interior commands. */
925 			fixuplabel(cp->u.c, cp->next);
926 			break;
927 		}
928 }
929 
930 /*
931  * Associate the given command label for later lookup.
932  */
933 static void
934 enterlabel(struct s_command *cp)
935 {
936 	struct labhash **lhp, *lh;
937 	u_char *p;
938 	u_int h, c;
939 
940 	for (h = 0, p = (u_char *)cp->t; (c = *p) != 0; p++)
941 		h = (h << 5) + h + c;
942 	lhp = &labels[h & LHMASK];
943 	for (lh = *lhp; lh != NULL; lh = lh->lh_next)
944 		if (lh->lh_hash == h && strcmp(cp->t, lh->lh_cmd->t) == 0)
945 			errx(1, "%lu: %s: duplicate label '%s'", linenum, fname, cp->t);
946 	if ((lh = malloc(sizeof *lh)) == NULL)
947 		err(1, "malloc");
948 	lh->lh_next = *lhp;
949 	lh->lh_hash = h;
950 	lh->lh_cmd = cp;
951 	lh->lh_ref = 0;
952 	*lhp = lh;
953 }
954 
955 /*
956  * Find the label contained in the command l in the command linked
957  * list cp.  L is excluded from the search.  Return NULL if not found.
958  */
959 static struct s_command *
960 findlabel(char *name)
961 {
962 	struct labhash *lh;
963 	u_char *p;
964 	u_int h, c;
965 
966 	for (h = 0, p = (u_char *)name; (c = *p) != 0; p++)
967 		h = (h << 5) + h + c;
968 	for (lh = labels[h & LHMASK]; lh != NULL; lh = lh->lh_next) {
969 		if (lh->lh_hash == h && strcmp(name, lh->lh_cmd->t) == 0) {
970 			lh->lh_ref = 1;
971 			return (lh->lh_cmd);
972 		}
973 	}
974 	return (NULL);
975 }
976 
977 /*
978  * Warn about any unused labels.  As a side effect, release the label hash
979  * table space.
980  */
981 static void
982 uselabel(void)
983 {
984 	struct labhash *lh, *next;
985 	int i;
986 
987 	for (i = 0; i < LHSZ; i++) {
988 		for (lh = labels[i]; lh != NULL; lh = next) {
989 			next = lh->lh_next;
990 			if (!lh->lh_ref)
991 				warnx("%lu: %s: unused label '%s'",
992 				    linenum, fname, lh->lh_cmd->t);
993 			free(lh);
994 		}
995 	}
996 }
997