xref: /openbsd-src/usr.bin/sed/compile.c (revision 58b8553b11de927ce5c25f31489cec554cac10de)
1 /*	$OpenBSD: compile.c,v 1.53 2024/07/17 20:57:15 millert Exp $	*/
2 
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/types.h>
37 #include <sys/stat.h>
38 
39 #include <ctype.h>
40 #include <err.h>
41 #include <fcntl.h>
42 #include <limits.h>
43 #include <regex.h>
44 #include <stdio.h>
45 #include <stdlib.h>
46 #include <string.h>
47 
48 #include "defs.h"
49 #include "extern.h"
50 
51 #define LHSZ	128
52 #define	LHMASK	(LHSZ - 1)
53 static struct labhash {
54 	struct	labhash *lh_next;
55 	u_int	lh_hash;
56 	struct	s_command *lh_cmd;
57 	int	lh_ref;
58 } *labels[LHSZ];
59 
60 static char	 *compile_addr(char *, struct s_addr *);
61 static char	 *compile_ccl(char **, char *);
62 static char	 *compile_delimited(char *, char *);
63 static char	 *compile_flags(char *, struct s_subst *);
64 static char	 *compile_re(char *, regex_t **);
65 static char	 *compile_subst(char *, struct s_subst *);
66 static char	 *compile_text(void);
67 static char	 *compile_tr(char *, char **);
68 static struct s_command
69 		**compile_stream(struct s_command **);
70 static char	 *duptoeol(char *, char *, char **);
71 static void	  enterlabel(struct s_command *);
72 static struct s_command
73 		 *findlabel(char *);
74 static void	  fixuplabel(struct s_command *, struct s_command *);
75 static void	  uselabel(void);
76 
77 /*
78  * Command specification.  This is used to drive the command parser.
79  */
80 struct s_format {
81 	char code;				/* Command code */
82 	int naddr;				/* Number of address args */
83 	enum e_args args;			/* Argument type */
84 };
85 
86 static struct s_format cmd_fmts[] = {
87 	{'{', 2, GROUP},
88 	{'}', 0, ENDGROUP},
89 	{'a', 1, TEXT},
90 	{'b', 2, BRANCH},
91 	{'c', 2, TEXT},
92 	{'d', 2, EMPTY},
93 	{'D', 2, EMPTY},
94 	{'g', 2, EMPTY},
95 	{'G', 2, EMPTY},
96 	{'h', 2, EMPTY},
97 	{'H', 2, EMPTY},
98 	{'i', 1, TEXT},
99 	{'l', 2, EMPTY},
100 	{'n', 2, EMPTY},
101 	{'N', 2, EMPTY},
102 	{'p', 2, EMPTY},
103 	{'P', 2, EMPTY},
104 	{'q', 1, EMPTY},
105 	{'r', 1, RFILE},
106 	{'s', 2, SUBST},
107 	{'t', 2, BRANCH},
108 	{'w', 2, WFILE},
109 	{'x', 2, EMPTY},
110 	{'y', 2, TR},
111 	{'!', 2, NONSEL},
112 	{':', 0, LABEL},
113 	{'#', 0, COMMENT},
114 	{'=', 1, EMPTY},
115 	{'\0', 0, COMMENT},
116 };
117 
118 /* The compiled program. */
119 struct s_command *prog;
120 
121 /*
122  * Compile the program into prog.
123  * Initialise appends.
124  */
125 void
126 compile(void)
127 {
128 	*compile_stream(&prog) = NULL;
129 	fixuplabel(prog, NULL);
130 	uselabel();
131 	appends = xreallocarray(NULL, appendnum, sizeof(struct s_appends));
132 	match = xreallocarray(NULL, maxnsub + 1, sizeof(regmatch_t));
133 }
134 
135 #define EATSPACE() do {							\
136 	if (p)								\
137 		while (isascii((unsigned char)*p) &&			\
138 		    isspace((unsigned char)*p))				\
139 			p++;						\
140 	} while (0)
141 
142 static struct s_command **
143 compile_stream(struct s_command **link)
144 {
145 	char *p;
146 	static char *lbuf;	/* To avoid excessive malloc calls */
147 	static size_t bufsize;
148 	struct s_command *cmd, *cmd2, *stack;
149 	struct s_format *fp;
150 	int naddr;				/* Number of addresses */
151 
152 	stack = 0;
153 	for (;;) {
154 		if ((p = cu_getline(&lbuf, &bufsize)) == NULL) {
155 			if (stack != 0)
156 				error("unexpected EOF (pending }'s)");
157 			return (link);
158 		}
159 
160 semicolon:	EATSPACE();
161 		if (*p == '#' || *p == '\0')
162 			continue;
163 		if (*p == ';') {
164 			p++;
165 			goto semicolon;
166 		}
167 		*link = cmd = xmalloc(sizeof(struct s_command));
168 		link = &cmd->next;
169 		cmd->nonsel = cmd->inrange = 0;
170 		/* First parse the addresses */
171 		naddr = 0;
172 
173 /* Valid characters to start an address */
174 #define	addrchar(c)	(strchr("0123456789/\\$", (c)))
175 		if (addrchar(*p)) {
176 			naddr++;
177 			cmd->a1 = xmalloc(sizeof(struct s_addr));
178 			p = compile_addr(p, cmd->a1);
179 			EATSPACE();				/* EXTENSION */
180 			if (*p == ',') {
181 				p++;
182 				EATSPACE();			/* EXTENSION */
183 				naddr++;
184 				cmd->a2 = xmalloc(sizeof(struct s_addr));
185 				p = compile_addr(p, cmd->a2);
186 				EATSPACE();
187 			} else {
188 				cmd->a2 = 0;
189 			}
190 		} else {
191 			cmd->a1 = cmd->a2 = 0;
192 		}
193 
194 nonsel:		/* Now parse the command */
195 		if (!*p)
196 			error("command expected");
197 		cmd->code = *p;
198 		for (fp = cmd_fmts; fp->code; fp++)
199 			if (fp->code == *p)
200 				break;
201 		if (!fp->code)
202 			error("invalid command code %c", *p);
203 		if (naddr > fp->naddr)
204 			error("command %c expects up to %d address(es),"
205 			    " found %d", *p, fp->naddr, naddr);
206 		switch (fp->args) {
207 		case NONSEL:			/* ! */
208 			p++;
209 			EATSPACE();
210 			cmd->nonsel = 1;
211 			goto nonsel;
212 		case GROUP:			/* { */
213 			p++;
214 			EATSPACE();
215 			cmd->next = stack;
216 			stack = cmd;
217 			link = &cmd->u.c;
218 			if (*p)
219 				goto semicolon;
220 			break;
221 		case ENDGROUP:
222 			/*
223 			 * Short-circuit command processing, since end of
224 			 * group is really just a noop.
225 			 */
226 			cmd->nonsel = 1;
227 			if (stack == 0)
228 				error("unexpected }");
229 			cmd2 = stack;
230 			stack = cmd2->next;
231 			cmd2->next = cmd;
232 			/*FALLTHROUGH*/
233 		case EMPTY:		/* d D g G h H l n N p P q x = \0 */
234 			p++;
235 			EATSPACE();
236 			if (*p == ';') {
237 				p++;
238 				link = &cmd->next;
239 				goto semicolon;
240 			}
241 			if (*p)
242 				error("extra characters at the end of %c"
243 				    " command", cmd->code);
244 			break;
245 		case TEXT:			/* a c i */
246 			p++;
247 			EATSPACE();
248 			if (*p != '\\')
249 				error("command %c expects \\ followed by text",
250 				    cmd->code);
251 			p++;
252 			EATSPACE();
253 			if (*p)
254 				error("extra characters after \\ at the"
255 				    " end of %c command", cmd->code);
256 			cmd->t = compile_text();
257 			break;
258 		case COMMENT:			/* \0 # */
259 			break;
260 		case WFILE:			/* w */
261 			p++;
262 			EATSPACE();
263 			if (*p == '\0')
264 				error("filename expected");
265 			cmd->t = duptoeol(p, "w command", NULL);
266 			if (aflag) {
267 				cmd->u.fd = -1;
268 				pledge_wpath = 1;
269 			}
270 			else if ((cmd->u.fd = open(p,
271 			    O_WRONLY|O_APPEND|O_CREAT|O_TRUNC,
272 			    DEFFILEMODE)) == -1)
273 				err(1, "%s", p);
274 			break;
275 		case RFILE:			/* r */
276 			pledge_rpath = 1;
277 			p++;
278 			EATSPACE();
279 			if (*p == '\0')
280 				error("filename expected");
281 			cmd->t = duptoeol(p, "read command", NULL);
282 			break;
283 		case BRANCH:			/* b t */
284 			p++;
285 			EATSPACE();
286 			if (*p == '\0' || *p == ';')
287 				cmd->t = NULL;
288 			else
289 				cmd->t = duptoeol(p, "branch", &p);
290 			if (*p == ';') {
291 				p++;
292 				goto semicolon;
293 			}
294 			break;
295 		case LABEL:			/* : */
296 			p++;
297 			EATSPACE();
298 			cmd->t = duptoeol(p, "label", &p);
299 			if (strlen(cmd->t) == 0)
300 				error("empty label");
301 			enterlabel(cmd);
302 			if (*p == ';') {
303 				p++;
304 				goto semicolon;
305 			}
306 			break;
307 		case SUBST:			/* s */
308 			p++;
309 			if (*p == '\0' || *p == '\\')
310 				error("substitute pattern can not be"
311 				    " delimited by newline or backslash");
312 			cmd->u.s = xmalloc(sizeof(struct s_subst));
313 			p = compile_re(p, &cmd->u.s->re);
314 			if (p == NULL)
315 				error("unterminated substitute pattern");
316 			--p;
317 			p = compile_subst(p, cmd->u.s);
318 			p = compile_flags(p, cmd->u.s);
319 			EATSPACE();
320 			if (*p == ';') {
321 				p++;
322 				link = &cmd->next;
323 				goto semicolon;
324 			}
325 			break;
326 		case TR:			/* y */
327 			p++;
328 			p = compile_tr(p, (char **)&cmd->u.y);
329 			EATSPACE();
330 			if (*p == ';') {
331 				p++;
332 				link = &cmd->next;
333 				goto semicolon;
334 			}
335 			if (*p)
336 				error("extra text at the end of a"
337 				    " transform command");
338 			break;
339 		}
340 	}
341 }
342 
343 /*
344  * Get a delimited string.  P points to the delimiter of the string; d points
345  * to a buffer area.  Newline and delimiter escapes are processed; other
346  * escapes are ignored.
347  *
348  * Returns a pointer to the first character after the final delimiter or NULL
349  * in the case of a non-terminated string.  The character array d is filled
350  * with the processed string.
351  */
352 static char *
353 compile_delimited(char *p, char *d)
354 {
355 	char c;
356 
357 	c = *p++;
358 	if (c == '\0')
359 		return (NULL);
360 	else if (c == '\\')
361 		error("\\ can not be used as a string delimiter");
362 	else if (c == '\n')
363 		error("newline can not be used as a string delimiter");
364 
365 	while (p[0]) {
366 		/* Unescaped delimiter: We are done. */
367 		if (p[0] == c) {
368 			*d = '\0';
369 			return p + 1;
370 		}
371 		if (p[0] == '\\') {
372 			/* Escaped delimiter: Skip the backslash. */
373 			if (p[1] == c) {
374 				p++;
375 			} else {
376 				/* Backslash-n: Match linefeed. */
377 				if (p[1] == 'n') {
378 					*d++ = '\n';
379 					p += 2;
380 				/* Other escapes remain unchanged. */
381 				} else {
382 					*d++ = *p++;
383 					*d++ = *p++;
384 				}
385 				continue;
386 			}
387 		}
388 		if (p[0] != '[')
389 			*d++ = *p++;
390 		/*
391 		 * Bracket expression:
392 		 * It may contain the delimiter without escaping.
393 		 */
394 		else if ((d = compile_ccl(&p, d)) == NULL)
395 			error("unbalanced brackets ([])");
396 	}
397 	return NULL;
398 }
399 
400 
401 /* compile_ccl: expand a POSIX character class */
402 static char *
403 compile_ccl(char **sp, char *t)
404 {
405 	int c, d;
406 	char *s = *sp;
407 
408 	*t++ = *s++;
409 	if (*s == '^')
410 		*t++ = *s++;
411 	if (*s == ']')
412 		*t++ = *s++;
413 	for (; *s && (*t = *s) != ']'; s++, t++)
414 		if (*s == '[' && ((d = *(s+1)) == '.' || d == ':' || d == '=')) {
415 			*++t = *++s, t++, s++;
416 			for (c = *s; (*t = *s) != ']' || c != d; s++, t++)
417 				if ((c = *s) == '\0')
418 					return NULL;
419 		} else if (*s == '\\' && s[1] == 'n') {
420 			*t = '\n';
421 			s++;
422 		}
423 	if (*s == ']') {
424 		*sp = ++s;
425 		return (++t);
426 	} else {
427 		return (NULL);
428 	}
429 }
430 
431 /*
432  * Get a regular expression.  P points to the delimiter of the regular
433  * expression; repp points to the address of a regexp pointer.  Newline
434  * and delimiter escapes are processed; other escapes are ignored.
435  * Returns a pointer to the first character after the final delimiter
436  * or NULL in the case of a non terminated regular expression.  The regexp
437  * pointer is set to the compiled regular expression.
438  * Cflags are passed to regcomp.
439  */
440 static char *
441 compile_re(char *p, regex_t **repp)
442 {
443 	int eval;
444 	char *re;
445 
446 	re = xmalloc(strlen(p) + 1); /* strlen(re) <= strlen(p) */
447 	p = compile_delimited(p, re);
448 	if (p && strlen(re) == 0) {
449 		*repp = NULL;
450 		free(re);
451 		return (p);
452 	}
453 	*repp = xmalloc(sizeof(regex_t));
454 	if (p && (eval = regcomp(*repp, re, Eflag ? REG_EXTENDED : 0)) != 0)
455 		error("RE error: %s", strregerror(eval, *repp));
456 	if (maxnsub < (*repp)->re_nsub)
457 		maxnsub = (*repp)->re_nsub;
458 	free(re);
459 	return (p);
460 }
461 
462 /*
463  * Compile the substitution string of a regular expression and set res to
464  * point to a saved copy of it.  Nsub is the number of parenthesized regular
465  * expressions.
466  */
467 static char *
468 compile_subst(char *p, struct s_subst *s)
469 {
470 	static char *lbuf;
471 	static size_t bufsize;
472 	size_t asize, ref, size;
473 	char c, *text, *op, *sp;
474 	int sawesc = 0;
475 
476 	c = *p++;			/* Terminator character */
477 	if (c == '\0')
478 		return (NULL);
479 
480 	s->maxbref = 0;
481 	s->linenum = linenum;
482 	text = NULL;
483 	asize = size = 0;
484 	do {
485 		size_t len = ROUNDLEN(strlen(p) + 1);
486 		if (asize - size < len) {
487 			do {
488 				asize += len;
489 			} while (asize - size < len);
490 			text = xrealloc(text, asize);
491 		}
492 		op = sp = text + size;
493 		for (; *p; p++) {
494 			if (*p == '\\' || sawesc) {
495 				/*
496 				 * If this is a continuation from the last
497 				 * buffer, we won't have a character to
498 				 * skip over.
499 				 */
500 				if (sawesc)
501 					sawesc = 0;
502 				else
503 					p++;
504 
505 				if (*p == '\0') {
506 					/*
507 					 * This escaped character is continued
508 					 * in the next part of the line.  Note
509 					 * this fact, then cause the loop to
510 					 * exit w/ normal EOL case and reenter
511 					 * above with the new buffer.
512 					 */
513 					sawesc = 1;
514 					p--;
515 					continue;
516 				} else if (strchr("123456789", *p) != NULL) {
517 					*sp++ = '\\';
518 					ref = *p - '0';
519 					if (s->re != NULL &&
520 					    ref > s->re->re_nsub)
521 						error("\\%c not defined in the"
522 						    " RE", *p);
523 					if (s->maxbref < ref)
524 						s->maxbref = ref;
525 				} else if (*p == '&' || *p == '\\')
526 					*sp++ = '\\';
527 			} else if (*p == c) {
528 				p++;
529 				*sp++ = '\0';
530 				size += sp - op;
531 				s->new = xrealloc(text, size);
532 				return (p);
533 			} else if (*p == '\n') {
534 				error("unescaped newline inside substitute"
535 				    " pattern");
536 			}
537 			*sp++ = *p;
538 		}
539 		size += sp - op;
540 	} while ((p = cu_getline(&lbuf, &bufsize)));
541 	error("unterminated substitute in regular expression");
542 }
543 
544 /*
545  * Compile the flags of the s command
546  */
547 static char *
548 compile_flags(char *p, struct s_subst *s)
549 {
550 	int gn;			/* True if we have seen g or n */
551 	long l;
552 
553 	s->n = 1;				/* Default */
554 	s->p = 0;
555 	s->wfile = NULL;
556 	s->wfd = -1;
557 	for (gn = 0;;) {
558 		EATSPACE();			/* EXTENSION */
559 		switch (*p) {
560 		case 'g':
561 			if (gn)
562 				error("more than one number or 'g' in"
563 				    " substitute flags");
564 			gn = 1;
565 			s->n = 0;
566 			break;
567 		case '\0':
568 		case '\n':
569 		case ';':
570 			return (p);
571 		case 'p':
572 			s->p = 1;
573 			break;
574 		case '1': case '2': case '3':
575 		case '4': case '5': case '6':
576 		case '7': case '8': case '9':
577 			if (gn)
578 				error("more than one number or 'g' in"
579 				    " substitute flags");
580 			gn = 1;
581 			l = strtol(p, &p, 10);
582 			if (l <= 0 || l >= INT_MAX)
583 				error("number in substitute flags out of"
584 				    " range");
585 			s->n = (int)l;
586 			continue;
587 		case 'w':
588 			p++;
589 			EATSPACE();
590 			if (*p == '\0')
591 				error("filename expected");
592 			s->wfile = duptoeol(p, "s command w flag", NULL);
593 			*p = '\0';
594 			if (aflag)
595 				pledge_wpath = 1;
596 			else if ((s->wfd = open(s->wfile,
597 			    O_WRONLY|O_APPEND|O_CREAT|O_TRUNC,
598 			    DEFFILEMODE)) == -1)
599 				err(1, "%s", s->wfile);
600 			return (p);
601 		default:
602 			error("bad flag in substitute command: '%c'", *p);
603 			break;
604 		}
605 		p++;
606 	}
607 }
608 
609 /*
610  * Compile a translation set of strings into a lookup table.
611  */
612 static char *
613 compile_tr(char *old, char **transtab)
614 {
615 	int i;
616 	char delimiter, check[UCHAR_MAX + 1];
617 	char *new, *end;
618 
619 	memset(check, 0, sizeof(check));
620 	delimiter = *old;
621 	if (delimiter == '\\')
622 		error("\\ can not be used as a string delimiter");
623 	else if (delimiter == '\n' || delimiter == '\0')
624 		error("newline can not be used as a string delimiter");
625 
626 	new = old++;
627 	do {
628 		if ((new = strchr(new + 1, delimiter)) == NULL)
629 			error("unterminated transform source string");
630 	} while (*(new - 1) == '\\' && *(new -2) != '\\');
631 	*new = '\0';
632 	end = new++;
633 	do {
634 		if ((end = strchr(end + 1, delimiter)) == NULL)
635 			error("unterminated transform target string");
636 	} while (*(end -1) == '\\' && *(end -2) != '\\');
637 	*end = '\0';
638 
639 	/* We assume characters are 8 bits */
640 	*transtab = xmalloc(UCHAR_MAX + 1);
641 	for (i = 0; i <= UCHAR_MAX; i++)
642 		(*transtab)[i] = (char)i;
643 
644 	while (*old != '\0' && *new != '\0') {
645 		if (*old == '\\') {
646 			old++;
647 			if (*old == 'n')
648 				*old = '\n';
649 			else if (*old != delimiter && *old != '\\')
650 				error("Unexpected character after backslash");
651 		}
652 		if (*new == '\\') {
653 			new++;
654 			if (*new == 'n')
655 				*new = '\n';
656 			else if (*new != delimiter && *new != '\\')
657 				error("Unexpected character after backslash");
658 		}
659 		if (check[(u_char) *old] == 1)
660 			error("Repeated character in source string");
661 		check[(u_char) *old] = 1;
662 		(*transtab)[(u_char) *old++] = *new++;
663 	}
664 	if (*old != '\0' || *new != '\0')
665 		error("transform strings are not the same length");
666 	return end + 1;
667 }
668 
669 /*
670  * Compile the text following an a, c, or i command.
671  */
672 static char *
673 compile_text(void)
674 {
675 	size_t asize, size, bufsize;
676 	char *lbuf, *text, *p, *op, *s;
677 	int esc_nl;
678 
679 	lbuf = text = NULL;
680 	asize = size = 0;
681 	while ((p = cu_getline(&lbuf, &bufsize))) {
682 		size_t len = ROUNDLEN(strlen(p) + 1);
683 		if (asize - size < len) {
684 			do {
685 				asize += len;
686 			} while (asize - size < len);
687 			text = xrealloc(text, asize);
688 		}
689 		op = s = text + size;
690 		for (esc_nl = 0; *p != '\0'; p++) {
691 			if (*p == '\\' && p[1] != '\0' && *++p == '\n')
692 				esc_nl = 1;
693 			*s++ = *p;
694 		}
695 		size += s - op;
696 		if (!esc_nl) {
697 			*s = '\0';
698 			break;
699 		}
700 	}
701 	free(lbuf);
702 	text = xrealloc(text, size + 1);
703 	text[size] = '\0';
704 	return (text);
705 }
706 
707 /*
708  * Get an address and return a pointer to the first character after
709  * it.  Fill the structure pointed to according to the address.
710  */
711 static char *
712 compile_addr(char *p, struct s_addr *a)
713 {
714 	char *end;
715 
716 	switch (*p) {
717 	case '\\':				/* Context address */
718 		++p;
719 		/* FALLTHROUGH */
720 	case '/':				/* Context address */
721 		p = compile_re(p, &a->u.r);
722 		if (p == NULL)
723 			error("unterminated regular expression");
724 		a->type = AT_RE;
725 		return (p);
726 
727 	case '$':				/* Last line */
728 		a->type = AT_LAST;
729 		return (p + 1);
730 						/* Line number */
731 	case '0': case '1': case '2': case '3': case '4':
732 	case '5': case '6': case '7': case '8': case '9':
733 		a->type = AT_LINE;
734 		a->u.l = strtoul(p, &end, 10);
735 		return (end);
736 	default:
737 		error("expected context address");
738 		return (NULL);
739 	}
740 }
741 
742 /*
743  * duptoeol --
744  *	Return a copy of all the characters up to \n or \0.
745  */
746 static char *
747 duptoeol(char *s, char *ctype, char **semi)
748 {
749 	size_t len;
750 	int ws;
751 	char *start;
752 
753 	ws = 0;
754 	if (semi) {
755 		for (start = s; *s != '\0' && *s != '\n' && *s != ';'; ++s)
756 			ws = isspace((unsigned char)*s);
757 	} else {
758 		for (start = s; *s != '\0' && *s != '\n'; ++s)
759 			ws = isspace((unsigned char)*s);
760 		*s = '\0';
761 	}
762 	if (ws)
763 		warning("whitespace after %s", ctype);
764 	len = s - start + 1;
765 	if (semi)
766 		*semi = s;
767 	s = xmalloc(len);
768 	strlcpy(s, start, len);
769 	return (s);
770 }
771 
772 /*
773  * Convert goto label names to addresses, and count a and r commands, in
774  * the given subset of the script.  Free the memory used by labels in b
775  * and t commands (but not by :).
776  *
777  * TODO: Remove } nodes
778  */
779 static void
780 fixuplabel(struct s_command *cp, struct s_command *end)
781 {
782 
783 	for (; cp != end; cp = cp->next)
784 		switch (cp->code) {
785 		case 'a':
786 		case 'r':
787 			appendnum++;
788 			break;
789 		case 'b':
790 		case 't':
791 			/* Resolve branch target. */
792 			if (cp->t == NULL) {
793 				cp->u.c = NULL;
794 				break;
795 			}
796 			if ((cp->u.c = findlabel(cp->t)) == NULL)
797 				error("undefined label '%s'", cp->t);
798 			free(cp->t);
799 			break;
800 		case '{':
801 			/* Do interior commands. */
802 			fixuplabel(cp->u.c, cp->next);
803 			break;
804 		}
805 }
806 
807 /*
808  * Associate the given command label for later lookup.
809  */
810 static void
811 enterlabel(struct s_command *cp)
812 {
813 	struct labhash **lhp, *lh;
814 	u_char *p;
815 	u_int h, c;
816 
817 	for (h = 0, p = (u_char *)cp->t; (c = *p) != 0; p++)
818 		h = (h << 5) + h + c;
819 	lhp = &labels[h & LHMASK];
820 	for (lh = *lhp; lh != NULL; lh = lh->lh_next)
821 		if (lh->lh_hash == h && strcmp(cp->t, lh->lh_cmd->t) == 0)
822 			error("duplicate label '%s'", cp->t);
823 	lh = xmalloc(sizeof *lh);
824 	lh->lh_next = *lhp;
825 	lh->lh_hash = h;
826 	lh->lh_cmd = cp;
827 	lh->lh_ref = 0;
828 	*lhp = lh;
829 }
830 
831 /*
832  * Find the label contained in the command l in the command linked
833  * list cp.  L is excluded from the search.  Return NULL if not found.
834  */
835 static struct s_command *
836 findlabel(char *name)
837 {
838 	struct labhash *lh;
839 	u_char *p;
840 	u_int h, c;
841 
842 	for (h = 0, p = (u_char *)name; (c = *p) != 0; p++)
843 		h = (h << 5) + h + c;
844 	for (lh = labels[h & LHMASK]; lh != NULL; lh = lh->lh_next) {
845 		if (lh->lh_hash == h && strcmp(name, lh->lh_cmd->t) == 0) {
846 			lh->lh_ref = 1;
847 			return (lh->lh_cmd);
848 		}
849 	}
850 	return (NULL);
851 }
852 
853 /*
854  * Warn about any unused labels.  As a side effect, release the label hash
855  * table space.
856  */
857 static void
858 uselabel(void)
859 {
860 	struct labhash *lh, *next;
861 	int i;
862 
863 	for (i = 0; i < LHSZ; i++) {
864 		for (lh = labels[i]; lh != NULL; lh = next) {
865 			next = lh->lh_next;
866 			if (!lh->lh_ref)
867 				warning("unused label '%s'",
868 				    lh->lh_cmd->t);
869 			free(lh);
870 		}
871 	}
872 }
873