xref: /netbsd/usr.bin/sed/compile.c (revision c4a72b64)
1 /*	$NetBSD: compile.c,v 1.24 2002/06/14 00:41:42 wiz 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. All advertising materials mentioning features or use of this software
20  *    must display the following acknowledgement:
21  *	This product includes software developed by the University of
22  *	California, Berkeley and its contributors.
23  * 4. Neither the name of the University nor the names of its contributors
24  *    may be used to endorse or promote products derived from this software
25  *    without specific prior written permission.
26  *
27  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
28  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
29  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
30  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
31  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
32  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
33  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
34  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
35  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
36  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
37  * SUCH DAMAGE.
38  */
39 
40 #include <sys/cdefs.h>
41 #ifndef lint
42 #if 0
43 static char sccsid[] = "@(#)compile.c	8.2 (Berkeley) 4/28/95";
44 #else
45 __RCSID("$NetBSD: compile.c,v 1.24 2002/06/14 00:41:42 wiz Exp $");
46 #endif
47 #endif /* not lint */
48 
49 #include <sys/types.h>
50 #include <sys/stat.h>
51 
52 #include <ctype.h>
53 #include <errno.h>
54 #include <fcntl.h>
55 #include <limits.h>
56 #include <regex.h>
57 #include <stdio.h>
58 #include <stdlib.h>
59 #include <string.h>
60 
61 #include "defs.h"
62 #include "extern.h"
63 
64 #define LHSZ	128
65 #define	LHMASK	(LHSZ - 1)
66 static struct labhash {
67 	struct	labhash *lh_next;
68 	u_int	lh_hash;
69 	struct	s_command *lh_cmd;
70 	int	lh_ref;
71 } *labels[LHSZ];
72 
73 static char	 *compile_addr(char *, struct s_addr *);
74 static char	 *compile_ccl(char **, char *);
75 static char	 *compile_delimited(char *, char *);
76 static char	 *compile_flags(char *, struct s_subst *);
77 static char	 *compile_re(char *, regex_t **);
78 static char	 *compile_subst(char *, struct s_subst *);
79 static char	 *compile_text(void);
80 static char	 *compile_tr(char *, char **);
81 static struct s_command
82 		**compile_stream(struct s_command **);
83 static char	 *duptoeol(char *, char *);
84 static void	  enterlabel(struct s_command *);
85 static struct s_command
86 		 *findlabel(char *);
87 static void	  fixuplabel(struct s_command *, struct s_command *);
88 static void	  uselabel(void);
89 
90 /*
91  * Command specification.  This is used to drive the command parser.
92  */
93 struct s_format {
94 	char code;				/* Command code */
95 	int naddr;				/* Number of address args */
96 	enum e_args args;			/* Argument type */
97 };
98 
99 static struct s_format cmd_fmts[] = {
100 	{'{', 2, GROUP},
101 	{'}', 0, ENDGROUP},
102 	{'a', 1, TEXT},
103 	{'b', 2, BRANCH},
104 	{'c', 2, TEXT},
105 	{'d', 2, EMPTY},
106 	{'D', 2, EMPTY},
107 	{'g', 2, EMPTY},
108 	{'G', 2, EMPTY},
109 	{'h', 2, EMPTY},
110 	{'H', 2, EMPTY},
111 	{'i', 1, TEXT},
112 	{'l', 2, EMPTY},
113 	{'n', 2, EMPTY},
114 	{'N', 2, EMPTY},
115 	{'p', 2, EMPTY},
116 	{'P', 2, EMPTY},
117 	{'q', 1, EMPTY},
118 	{'r', 1, RFILE},
119 	{'s', 2, SUBST},
120 	{'t', 2, BRANCH},
121 	{'w', 2, WFILE},
122 	{'x', 2, EMPTY},
123 	{'y', 2, TR},
124 	{'!', 2, NONSEL},
125 	{':', 0, LABEL},
126 	{'#', 0, COMMENT},
127 	{'=', 1, EMPTY},
128 	{'\0', 0, COMMENT},
129 };
130 
131 /* The compiled program. */
132 struct s_command *prog;
133 
134 /*
135  * Compile the program into prog.
136  * Initialise appends.
137  */
138 void
139 compile(void)
140 {
141 	*compile_stream(&prog) = NULL;
142 	fixuplabel(prog, NULL);
143 	uselabel();
144 	if (appendnum > 0)
145 		appends = xmalloc(sizeof(struct s_appends) * appendnum);
146 	match = xmalloc((maxnsub + 1) * sizeof(regmatch_t));
147 }
148 
149 #define EATSPACE() do {							\
150 	if (p)								\
151 		while (*p && isascii((unsigned char)*p) &&		\
152 		    isspace((unsigned char)*p))				\
153 			p++;						\
154 	} while (0)
155 
156 static struct s_command **
157 compile_stream(struct s_command **link)
158 {
159 	char *p;
160 	static char lbuf[_POSIX2_LINE_MAX + 1];	/* To save stack */
161 	struct s_command *cmd, *cmd2, *stack;
162 	struct s_format *fp;
163 	int naddr;				/* Number of addresses */
164 
165 	stack = 0;
166 	for (;;) {
167 		if ((p = cu_fgets(lbuf, sizeof(lbuf))) == NULL) {
168 			if (stack != 0)
169 				err(COMPILE, "unexpected EOF (pending }'s)");
170 			return (link);
171 		}
172 
173 semicolon:	EATSPACE();
174 		if (p) {
175 			if (*p == '#' || *p == '\0')
176 				continue;
177 			else if (*p == ';') {
178 				p++;
179 				goto semicolon;
180 			}
181 		}
182 		*link = cmd = xmalloc(sizeof(struct s_command));
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 			cmd->a1 = xmalloc(sizeof(struct s_addr));
193 			p = compile_addr(p, cmd->a1);
194 			EATSPACE();				/* EXTENSION */
195 			if (*p == ',') {
196 				p++;
197 				EATSPACE();			/* EXTENSION */
198 				naddr++;
199 				cmd->a2 = xmalloc(sizeof(struct s_addr));
200 				p = compile_addr(p, cmd->a2);
201 				EATSPACE();
202 			} else
203 				cmd->a2 = 0;
204 		} else
205 			cmd->a1 = cmd->a2 = 0;
206 
207 nonsel:		/* Now parse the command */
208 		if (!*p)
209 			err(COMPILE, "command expected");
210 		cmd->code = *p;
211 		for (fp = cmd_fmts; fp->code; fp++)
212 			if (fp->code == *p)
213 				break;
214 		if (!fp->code)
215 			err(COMPILE, "invalid command code %c", *p);
216 		if (naddr > fp->naddr)
217 			err(COMPILE,
218 "command %c expects up to %d address(es), found %d", *p, fp->naddr, naddr);
219 		switch (fp->args) {
220 		case NONSEL:			/* ! */
221 			p++;
222 			EATSPACE();
223 			cmd->nonsel = ! cmd->nonsel;
224 			goto nonsel;
225 		case GROUP:			/* { */
226 			p++;
227 			EATSPACE();
228 			cmd->next = stack;
229 			stack = cmd;
230 			link = &cmd->u.c;
231 			if (*p)
232 				goto semicolon;
233 			break;
234 		case ENDGROUP:
235 			/*
236 			 * Short-circuit command processing, since end of
237 			 * group is really just a noop.
238 			 */
239 			cmd->nonsel = 1;
240 			if (stack == 0)
241 				err(COMPILE, "unexpected }");
242 			cmd2 = stack;
243 			stack = cmd2->next;
244 			cmd2->next = cmd;
245 			/*FALLTHROUGH*/
246 		case EMPTY:		/* d D g G h H l n N p P q x = \0 */
247 			p++;
248 			EATSPACE();
249 			if (*p == ';') {
250 				p++;
251 				link = &cmd->next;
252 				goto semicolon;
253 			}
254 			if (*p)
255 				err(COMPILE,
256 "extra characters at the end of %c command", cmd->code);
257 			break;
258 		case TEXT:			/* a c i */
259 			p++;
260 			EATSPACE();
261 			if (*p != '\\')
262 				err(COMPILE,
263 "command %c expects \\ followed by text", cmd->code);
264 			p++;
265 			EATSPACE();
266 			if (*p)
267 				err(COMPILE,
268 "extra characters after \\ at the end of %c command", cmd->code);
269 			cmd->t = compile_text();
270 			break;
271 		case COMMENT:			/* \0 # */
272 			break;
273 		case WFILE:			/* w */
274 			p++;
275 			EATSPACE();
276 			if (*p == '\0')
277 				err(COMPILE, "filename expected");
278 			cmd->t = duptoeol(p, "w command");
279 			if (aflag)
280 				cmd->u.fd = -1;
281 			else if ((cmd->u.fd = open(p,
282 			    O_WRONLY|O_APPEND|O_CREAT|O_TRUNC,
283 			    DEFFILEMODE)) == -1)
284 				err(FATAL, "%s: %s", p, strerror(errno));
285 			break;
286 		case RFILE:			/* r */
287 			p++;
288 			EATSPACE();
289 			if (*p == '\0')
290 				err(COMPILE, "filename expected");
291 			else
292 				cmd->t = duptoeol(p, "read command");
293 			break;
294 		case BRANCH:			/* b t */
295 			p++;
296 			EATSPACE();
297 			if (*p == '\0')
298 				cmd->t = NULL;
299 			else
300 				cmd->t = duptoeol(p, "branch");
301 			break;
302 		case LABEL:			/* : */
303 			p++;
304 			EATSPACE();
305 			cmd->t = duptoeol(p, "label");
306 			if (strlen(p) == 0)
307 				err(COMPILE, "empty label");
308 			enterlabel(cmd);
309 			break;
310 		case SUBST:			/* s */
311 			p++;
312 			if (*p == '\0' || *p == '\\')
313 				err(COMPILE,
314 "substitute pattern can not be delimited by newline or backslash");
315 			cmd->u.s = xmalloc(sizeof(struct s_subst));
316 			p = compile_re(p, &cmd->u.s->re);
317 			if (p == NULL)
318 				err(COMPILE, "unterminated substitute pattern");
319 			--p;
320 			p = compile_subst(p, cmd->u.s);
321 			p = compile_flags(p, cmd->u.s);
322 			EATSPACE();
323 			if (*p == ';') {
324 				p++;
325 				link = &cmd->next;
326 				goto semicolon;
327 			}
328 			break;
329 		case TR:			/* y */
330 			p++;
331 			p = compile_tr(p, (char **)&cmd->u.y);
332 			EATSPACE();
333 			if (*p == ';') {
334 				p++;
335 				link = &cmd->next;
336 				goto semicolon;
337 			}
338 			if (*p)
339 				err(COMPILE,
340 "extra text at the end of a transform command");
341 			break;
342 		}
343 	}
344 }
345 
346 /*
347  * Get a delimited string.  P points to the delimiter of the string; d points
348  * to a buffer area.  Newline and delimiter escapes are processed; other
349  * escapes are ignored.
350  *
351  * Returns a pointer to the first character after the final delimiter or NULL
352  * in the case of a non-terminated string.  The character array d is filled
353  * with the processed string.
354  */
355 static char *
356 compile_delimited(char *p, char *d)
357 {
358 	char c;
359 
360 	c = *p++;
361 	if (c == '\0')
362 		return (NULL);
363 	else if (c == '\\')
364 		err(COMPILE, "\\ can not be used as a string delimiter");
365 	else if (c == '\n')
366 		err(COMPILE, "newline can not be used as a string delimiter");
367 	while (*p) {
368 		if (*p == '[') {
369 			if ((d = compile_ccl(&p, d)) == NULL)
370 				err(COMPILE, "unbalanced brackets ([])");
371 			continue;
372 		} else if (*p == '\\' && p[1] == '[') {
373 			*d++ = *p++;
374 		} else if (*p == '\\' && p[1] == c)
375 			p++;
376 		else if (*p == '\\' && p[1] == 'n') {
377 			*d++ = '\n';
378 			p += 2;
379 			continue;
380 		} else if (*p == '\\' && p[1] == '\\')
381 			*d++ = *p++;
382 		else if (*p == c) {
383 			*d = '\0';
384 			return (p + 1);
385 		}
386 		*d++ = *p++;
387 	}
388 	return (NULL);
389 }
390 
391 
392 /* compile_ccl: expand a POSIX character class */
393 static char *
394 compile_ccl(char **sp, char *t)
395 {
396 	int c, d;
397 	char *s = *sp;
398 
399 	*t++ = *s++;
400 	if (*s == '^')
401 		*t++ = *s++;
402 	if (*s == ']')
403 		*t++ = *s++;
404 	for (; *s && (*t = *s) != ']'; s++, t++)
405 		if (*s == '[' && ((d = *(s+1)) == '.' || d == ':' || d == '=')) {
406 			*++t = *++s, t++, s++;
407 			for (c = *s; (*t = *s) != ']' || c != d; s++, t++)
408 				if ((c = *s) == '\0')
409 					return NULL;
410 		} else if (*s == '\\' && s[1] == 'n')
411 			    *t = '\n', s++;
412 	return (*s == ']') ? *sp = ++s, ++t : NULL;
413 }
414 
415 /*
416  * Get a regular expression.  P points to the delimiter of the regular
417  * expression; repp points to the address of a regexp pointer.  Newline
418  * and delimiter escapes are processed; other escapes are ignored.
419  * Returns a pointer to the first character after the final delimiter
420  * or NULL in the case of a non terminated regular expression.  The regexp
421  * pointer is set to the compiled regular expression.
422  * Cflags are passed to regcomp.
423  */
424 static char *
425 compile_re(char *p, regex_t **repp)
426 {
427 	int eval;
428 	char re[_POSIX2_LINE_MAX + 1];
429 
430 	p = compile_delimited(p, re);
431 	if (p && strlen(re) == 0) {
432 		*repp = NULL;
433 		return (p);
434 	}
435 	*repp = xmalloc(sizeof(regex_t));
436 	if (p && (eval = regcomp(*repp, re, ere)) != 0)
437 		err(COMPILE, "RE error: %s", strregerror(eval, *repp));
438 	if (maxnsub < (*repp)->re_nsub)
439 		maxnsub = (*repp)->re_nsub;
440 	return (p);
441 }
442 
443 /*
444  * Compile the substitution string of a regular expression and set res to
445  * point to a saved copy of it.  Nsub is the number of parenthesized regular
446  * expressions.
447  */
448 static char *
449 compile_subst(char *p, struct s_subst *s)
450 {
451 	static char lbuf[_POSIX2_LINE_MAX + 1];
452 	int asize, ref, size;
453 	char c, *text, *op, *sp;
454 
455 	c = *p++;			/* Terminator character */
456 	if (c == '\0')
457 		return (NULL);
458 
459 	s->maxbref = 0;
460 	s->linenum = linenum;
461 	asize = 2 * _POSIX2_LINE_MAX + 1;
462 	text = xmalloc(asize);
463 	size = 0;
464 	do {
465 		op = sp = text + size;
466 		for (; *p; p++) {
467 			if (*p == '\\') {
468 				p++;
469 				if (strchr("123456789", *p) != NULL) {
470 					*sp++ = '\\';
471 					ref = *p - '0';
472 					if (s->re != NULL &&
473 					    ref > s->re->re_nsub)
474 						err(COMPILE,
475 "\\%c not defined in the RE", *p);
476 					if (s->maxbref < ref)
477 						s->maxbref = ref;
478 				} else if (*p == '&' || *p == '\\')
479 					*sp++ = '\\';
480 			} else if (*p == c) {
481 				p++;
482 				*sp++ = '\0';
483 				size += sp - op;
484 				s->new = xrealloc(text, size);
485 				return (p);
486 			} else if (*p == '\n') {
487 				err(COMPILE,
488 "unescaped newline inside substitute pattern");
489 				/* NOTREACHED */
490 			}
491 			*sp++ = *p;
492 		}
493 		size += sp - op;
494 		if (asize - size < _POSIX2_LINE_MAX + 1) {
495 			asize *= 2;
496 			text = xmalloc(asize);
497 		}
498 	} while (cu_fgets(p = lbuf, sizeof(lbuf)));
499 	err(COMPILE, "unterminated substitute in regular expression");
500 	/* NOTREACHED */
501 	return (NULL);
502 }
503 
504 /*
505  * Compile the flags of the s command
506  */
507 static char *
508 compile_flags(char *p, struct s_subst *s)
509 {
510 	int gn;			/* True if we have seen g or n */
511 	char wfile[_POSIX2_LINE_MAX + 1], *q;
512 
513 	s->n = 1;				/* Default */
514 	s->p = 0;
515 	s->wfile = NULL;
516 	s->wfd = -1;
517 	for (gn = 0;;) {
518 		EATSPACE();			/* EXTENSION */
519 		switch (*p) {
520 		case 'g':
521 			if (gn)
522 				err(COMPILE,
523 "more than one number or 'g' in substitute flags");
524 			gn = 1;
525 			s->n = 0;
526 			break;
527 		case '\0':
528 		case '\n':
529 		case ';':
530 			return (p);
531 		case 'p':
532 			s->p = 1;
533 			break;
534 		case '1': case '2': case '3':
535 		case '4': case '5': case '6':
536 		case '7': case '8': case '9':
537 			if (gn)
538 				err(COMPILE,
539 "more than one number or 'g' in substitute flags");
540 			gn = 1;
541 			/* XXX Check for overflow */
542 			s->n = (int)strtol(p, &p, 10);
543 			break;
544 		case 'w':
545 			p++;
546 #ifdef HISTORIC_PRACTICE
547 			if (*p != ' ') {
548 				err(WARNING, "space missing before w wfile");
549 				return (p);
550 			}
551 #endif
552 			EATSPACE();
553 			q = wfile;
554 			while (*p) {
555 				if (*p == '\n')
556 					break;
557 				*q++ = *p++;
558 			}
559 			*q = '\0';
560 			if (q == wfile)
561 				err(COMPILE, "no wfile specified");
562 			s->wfile = strdup(wfile);
563 			if (!aflag && (s->wfd = open(wfile,
564 			    O_WRONLY|O_APPEND|O_CREAT|O_TRUNC,
565 			    DEFFILEMODE)) == -1)
566 				err(FATAL, "%s: %s", wfile, strerror(errno));
567 			return (p);
568 		default:
569 			err(COMPILE,
570 			    "bad flag in substitute command: '%c'", *p);
571 			break;
572 		}
573 		p++;
574 	}
575 }
576 
577 /*
578  * Compile a translation set of strings into a lookup table.
579  */
580 static char *
581 compile_tr(char *p, char **transtab)
582 {
583 	int i;
584 	char *lt, *op, *np;
585 	char old[_POSIX2_LINE_MAX + 1];
586 	char new[_POSIX2_LINE_MAX + 1];
587 
588 	if (*p == '\0' || *p == '\\')
589 		err(COMPILE,
590 "transform pattern can not be delimited by newline or backslash");
591 	p = compile_delimited(p, old);
592 	if (p == NULL) {
593 		err(COMPILE, "unterminated transform source string");
594 		return (NULL);
595 	}
596 	p = compile_delimited(--p, new);
597 	if (p == NULL) {
598 		err(COMPILE, "unterminated transform target string");
599 		return (NULL);
600 	}
601 	EATSPACE();
602 	if (strlen(new) != strlen(old)) {
603 		err(COMPILE, "transform strings are not the same length");
604 		return (NULL);
605 	}
606 	/* We assume characters are 8 bits */
607 	lt = xmalloc(UCHAR_MAX);
608 	for (i = 0; i <= UCHAR_MAX; i++)
609 		lt[i] = (char)i;
610 	for (op = old, np = new; *op; op++, np++)
611 		lt[(u_char)*op] = *np;
612 	*transtab = lt;
613 	return (p);
614 }
615 
616 /*
617  * Compile the text following an a or i command.
618  */
619 static char *
620 compile_text(void)
621 {
622 	int asize, size;
623 	char *text, *p, *op, *s;
624 	char lbuf[_POSIX2_LINE_MAX + 1];
625 
626 	asize = 2 * _POSIX2_LINE_MAX + 1;
627 	text = xmalloc(asize);
628 	size = 0;
629 	while (cu_fgets(lbuf, sizeof(lbuf))) {
630 		op = s = text + size;
631 		p = lbuf;
632 		EATSPACE();
633 		for (; *p; p++) {
634 			if (*p == '\\')
635 				p++;
636 			*s++ = *p;
637 		}
638 		size += s - op;
639 		if (p[-2] != '\\') {
640 			*s = '\0';
641 			break;
642 		}
643 		if (asize - size < _POSIX2_LINE_MAX + 1) {
644 			asize *= 2;
645 			text = xmalloc(asize);
646 		}
647 	}
648 	return (xrealloc(text, size + 1));
649 }
650 
651 /*
652  * Get an address and return a pointer to the first character after
653  * it.  Fill the structure pointed to according to the address.
654  */
655 static char *
656 compile_addr(char *p, struct s_addr *a)
657 {
658 	char *end;
659 
660 	switch (*p) {
661 	case '\\':				/* Context address */
662 		++p;
663 		/* FALLTHROUGH */
664 	case '/':				/* Context address */
665 		p = compile_re(p, &a->u.r);
666 		if (p == NULL)
667 			err(COMPILE, "unterminated regular expression");
668 		a->type = AT_RE;
669 		return (p);
670 
671 	case '$':				/* Last line */
672 		a->type = AT_LAST;
673 		return (p + 1);
674 						/* Line number */
675 	case '0': case '1': case '2': case '3': case '4':
676 	case '5': case '6': case '7': case '8': case '9':
677 		a->type = AT_LINE;
678 		a->u.l = strtol(p, &end, 10);
679 		return (end);
680 	default:
681 		err(COMPILE, "expected context address");
682 		return (NULL);
683 	}
684 }
685 
686 /*
687  * duptoeol --
688  *	Return a copy of all the characters up to \n or \0.
689  */
690 static char *
691 duptoeol(char *s, char *ctype)
692 {
693 	size_t len;
694 	int ws;
695 	char *start;
696 
697 	ws = 0;
698 	for (start = s; *s != '\0' && *s != '\n'; ++s)
699 		ws = isspace((unsigned char)*s);
700 	*s = '\0';
701 	if (ws)
702 		err(WARNING, "whitespace after %s", ctype);
703 	len = s - start + 1;
704 	return (memmove(xmalloc(len), start, len));
705 }
706 
707 /*
708  * Convert goto label names to addresses, and count a and r commands, in
709  * the given subset of the script.  Free the memory used by labels in b
710  * and t commands (but not by :).
711  *
712  * TODO: Remove } nodes
713  */
714 static void
715 fixuplabel(struct s_command *cp, struct s_command *end)
716 {
717 
718 	for (; cp != end; cp = cp->next)
719 		switch (cp->code) {
720 		case 'a':
721 		case 'r':
722 			appendnum++;
723 			break;
724 		case 'b':
725 		case 't':
726 			/* Resolve branch target. */
727 			if (cp->t == NULL) {
728 				cp->u.c = NULL;
729 				break;
730 			}
731 			if ((cp->u.c = findlabel(cp->t)) == NULL)
732 				err(COMPILE2, "undefined label '%s'", cp->t);
733 			free(cp->t);
734 			break;
735 		case '{':
736 			/* Do interior commands. */
737 			fixuplabel(cp->u.c, cp->next);
738 			break;
739 		}
740 }
741 
742 /*
743  * Associate the given command label for later lookup.
744  */
745 static void
746 enterlabel(struct s_command *cp)
747 {
748 	struct labhash **lhp, *lh;
749 	u_char *p;
750 	u_int h, c;
751 
752 	for (h = 0, p = (u_char *)cp->t; (c = *p) != 0; p++)
753 		h = (h << 5) + h + c;
754 	lhp = &labels[h & LHMASK];
755 	for (lh = *lhp; lh != NULL; lh = lh->lh_next)
756 		if (lh->lh_hash == h && strcmp(cp->t, lh->lh_cmd->t) == 0)
757 			err(COMPILE2, "duplicate label '%s'", cp->t);
758 	lh = xmalloc(sizeof *lh);
759 	lh->lh_next = *lhp;
760 	lh->lh_hash = h;
761 	lh->lh_cmd = cp;
762 	lh->lh_ref = 0;
763 	*lhp = lh;
764 }
765 
766 /*
767  * Find the label contained in the command l in the command linked
768  * list cp.  L is excluded from the search.  Return NULL if not found.
769  */
770 static struct s_command *
771 findlabel(char *name)
772 {
773 	struct labhash *lh;
774 	u_char *p;
775 	u_int h, c;
776 
777 	for (h = 0, p = (u_char *)name; (c = *p) != 0; p++)
778 		h = (h << 5) + h + c;
779 	for (lh = labels[h & LHMASK]; lh != NULL; lh = lh->lh_next) {
780 		if (lh->lh_hash == h && strcmp(name, lh->lh_cmd->t) == 0) {
781 			lh->lh_ref = 1;
782 			return (lh->lh_cmd);
783 		}
784 	}
785 	return (NULL);
786 }
787 
788 /*
789  * Warn about any unused labels.  As a side effect, release the label hash
790  * table space.
791  */
792 static void
793 uselabel(void)
794 {
795 	struct labhash *lh, *next;
796 	int i;
797 
798 	for (i = 0; i < LHSZ; i++) {
799 		for (lh = labels[i]; lh != NULL; lh = next) {
800 			next = lh->lh_next;
801 			if (!lh->lh_ref)
802 				err(WARNING, "unused label '%s'",
803 				    lh->lh_cmd->t);
804 			free(lh);
805 		}
806 	}
807 }
808