]> git.cameronkatri.com Git - mandoc.git/blob - roff.c
git.ckatri.com / mandoc.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
history | raw | HEAD
Initial tbl framework. Parse point is in libroff, which keeps a
[mandoc.git] / roff.c
1 /* $Id: roff.c,v 1.109 2010/12/28 10:51:03 kristaps Exp $ */
2 /*
3 * Copyright (c) 2010 Kristaps Dzonsons <kristaps@bsd.lv>
4 * Copyright (c) 2010 Ingo Schwarze <schwarze@openbsd.org>
5 *
6 * Permission to use, copy, modify, and distribute this software for any
7 * purpose with or without fee is hereby granted, provided that the above
8 * copyright notice and this permission notice appear in all copies.
9 *
10 * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHORS DISCLAIM ALL WARRANTIES
11 * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
12 * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHORS BE LIABLE FOR
13 * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
14 * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
15 * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
16 * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
17 */
18 #ifdef HAVE_CONFIG_H
19 #include "config.h"
20 #endif
21
22 #include <assert.h>
23 #include <errno.h>
24 #include <ctype.h>
25 #include <limits.h>
26 #include <stdlib.h>
27 #include <string.h>
28 #include <stdio.h>
29
30 #include "mandoc.h"
31 #include "roff.h"
32 #include "libroff.h"
33 #include "libmandoc.h"
34
35 #define RSTACK_MAX 128
36
37 #define ROFF_CTL(c) \
38 ('.' == (c) || '\'' == (c))
39
40 enum rofft {
41 ROFF_ad,
42 ROFF_am,
43 ROFF_ami,
44 ROFF_am1,
45 ROFF_de,
46 ROFF_dei,
47 ROFF_de1,
48 ROFF_ds,
49 ROFF_el,
50 ROFF_hy,
51 ROFF_ie,
52 ROFF_if,
53 ROFF_ig,
54 ROFF_ne,
55 ROFF_nh,
56 ROFF_nr,
57 ROFF_rm,
58 ROFF_so,
59 ROFF_tr,
60 ROFF_TS,
61 ROFF_TE,
62 ROFF_cblock,
63 ROFF_ccond, /* FIXME: remove this. */
64 ROFF_USERDEF,
65 ROFF_MAX
66 };
67
68 enum roffrule {
69 ROFFRULE_ALLOW,
70 ROFFRULE_DENY
71 };
72
73 struct roffstr {
74 char *name; /* key of symbol */
75 char *string; /* current value */
76 struct roffstr *next; /* next in list */
77 };
78
79 struct roff {
80 struct roffnode *last; /* leaf of stack */
81 mandocmsg msg; /* err/warn/fatal messages */
82 void *data; /* privdata for messages */
83 enum roffrule rstack[RSTACK_MAX]; /* stack of !`ie' rules */
84 int rstackpos; /* position in rstack */
85 struct regset *regs; /* read/writable registers */
86 struct roffstr *first_string; /* user-defined strings & macros */
87 const char *current_string; /* value of last called user macro */
88 struct tbl *tbl;
89 };
90
91 struct roffnode {
92 enum rofft tok; /* type of node */
93 struct roffnode *parent; /* up one in stack */
94 int line; /* parse line */
95 int col; /* parse col */
96 char *name; /* node name, e.g. macro name */
97 char *end; /* end-rules: custom token */
98 int endspan; /* end-rules: next-line or infty */
99 enum roffrule rule; /* current evaluation rule */
100 };
101
102 #define ROFF_ARGS struct roff *r, /* parse ctx */ \
103 enum rofft tok, /* tok of macro */ \
104 char **bufp, /* input buffer */ \
105 size_t *szp, /* size of input buffer */ \
106 int ln, /* parse line */ \
107 int ppos, /* original pos in buffer */ \
108 int pos, /* current pos in buffer */ \
109 int *offs /* reset offset of buffer data */
110
111 typedef enum rofferr (*roffproc)(ROFF_ARGS);
112
113 struct roffmac {
114 const char *name; /* macro name */
115 roffproc proc; /* process new macro */
116 roffproc text; /* process as child text of macro */
117 roffproc sub; /* process as child of macro */
118 int flags;
119 #define ROFFMAC_STRUCT (1 << 0) /* always interpret */
120 struct roffmac *next;
121 };
122
123 static enum rofferr roff_block(ROFF_ARGS);
124 static enum rofferr roff_block_text(ROFF_ARGS);
125 static enum rofferr roff_block_sub(ROFF_ARGS);
126 static enum rofferr roff_cblock(ROFF_ARGS);
127 static enum rofferr roff_ccond(ROFF_ARGS);
128 static enum rofferr roff_cond(ROFF_ARGS);
129 static enum rofferr roff_cond_text(ROFF_ARGS);
130 static enum rofferr roff_cond_sub(ROFF_ARGS);
131 static enum rofferr roff_ds(ROFF_ARGS);
132 static enum roffrule roff_evalcond(const char *, int *);
133 static void roff_freestr(struct roff *);
134 static const char *roff_getstrn(const struct roff *,
135 const char *, size_t);
136 static enum rofferr roff_line_ignore(ROFF_ARGS);
137 static enum rofferr roff_line_error(ROFF_ARGS);
138 static enum rofferr roff_nr(ROFF_ARGS);
139 static int roff_res(struct roff *,
140 char **, size_t *, int);
141 static void roff_setstr(struct roff *,
142 const char *, const char *, int);
143 static enum rofferr roff_so(ROFF_ARGS);
144 static enum rofferr roff_TE(ROFF_ARGS);
145 static enum rofferr roff_TS(ROFF_ARGS);
146 static enum rofferr roff_userdef(ROFF_ARGS);
147
148 /* See roff_hash_find() */
149
150 #define ASCII_HI 126
151 #define ASCII_LO 33
152 #define HASHWIDTH (ASCII_HI - ASCII_LO + 1)
153
154 static struct roffmac *hash[HASHWIDTH];
155
156 static struct roffmac roffs[ROFF_MAX] = {
157 { "ad", roff_line_ignore, NULL, NULL, 0, NULL },
158 { "am", roff_block, roff_block_text, roff_block_sub, 0, NULL },
159 { "ami", roff_block, roff_block_text, roff_block_sub, 0, NULL },
160 { "am1", roff_block, roff_block_text, roff_block_sub, 0, NULL },
161 { "de", roff_block, roff_block_text, roff_block_sub, 0, NULL },
162 { "dei", roff_block, roff_block_text, roff_block_sub, 0, NULL },
163 { "de1", roff_block, roff_block_text, roff_block_sub, 0, NULL },
164 { "ds", roff_ds, NULL, NULL, 0, NULL },
165 { "el", roff_cond, roff_cond_text, roff_cond_sub, ROFFMAC_STRUCT, NULL },
166 { "hy", roff_line_ignore, NULL, NULL, 0, NULL },
167 { "ie", roff_cond, roff_cond_text, roff_cond_sub, ROFFMAC_STRUCT, NULL },
168 { "if", roff_cond, roff_cond_text, roff_cond_sub, ROFFMAC_STRUCT, NULL },
169 { "ig", roff_block, roff_block_text, roff_block_sub, 0, NULL },
170 { "ne", roff_line_ignore, NULL, NULL, 0, NULL },
171 { "nh", roff_line_ignore, NULL, NULL, 0, NULL },
172 { "nr", roff_nr, NULL, NULL, 0, NULL },
173 { "rm", roff_line_error, NULL, NULL, 0, NULL },
174 { "so", roff_so, NULL, NULL, 0, NULL },
175 { "tr", roff_line_ignore, NULL, NULL, 0, NULL },
176 { "TS", roff_TS, NULL, NULL, 0, NULL },
177 { "TE", roff_TE, NULL, NULL, 0, NULL },
178 { ".", roff_cblock, NULL, NULL, 0, NULL },
179 { "\\}", roff_ccond, NULL, NULL, 0, NULL },
180 { NULL, roff_userdef, NULL, NULL, 0, NULL },
181 };
182
183 static void roff_free1(struct roff *);
184 static enum rofft roff_hash_find(const char *, size_t);
185 static void roff_hash_init(void);
186 static void roffnode_cleanscope(struct roff *);
187 static void roffnode_push(struct roff *, enum rofft,
188 const char *, int, int);
189 static void roffnode_pop(struct roff *);
190 static enum rofft roff_parse(struct roff *, const char *, int *);
191 static int roff_parse_nat(const char *, unsigned int *);
192
193 /* See roff_hash_find() */
194 #define ROFF_HASH(p) (p[0] - ASCII_LO)
195
196 static void
197 roff_hash_init(void)
198 {
199 struct roffmac *n;
200 int buc, i;
201
202 for (i = 0; i < (int)ROFF_USERDEF; i++) {
203 assert(roffs[i].name[0] >= ASCII_LO);
204 assert(roffs[i].name[0] <= ASCII_HI);
205
206 buc = ROFF_HASH(roffs[i].name);
207
208 if (NULL != (n = hash[buc])) {
209 for ( ; n->next; n = n->next)
210 /* Do nothing. */ ;
211 n->next = &roffs[i];
212 } else
213 hash[buc] = &roffs[i];
214 }
215 }
216
217
218 /*
219 * Look up a roff token by its name. Returns ROFF_MAX if no macro by
220 * the nil-terminated string name could be found.
221 */
222 static enum rofft
223 roff_hash_find(const char *p, size_t s)
224 {
225 int buc;
226 struct roffmac *n;
227
228 /*
229 * libroff has an extremely simple hashtable, for the time
230 * being, which simply keys on the first character, which must
231 * be printable, then walks a chain. It works well enough until
232 * optimised.
233 */
234
235 if (p[0] < ASCII_LO || p[0] > ASCII_HI)
236 return(ROFF_MAX);
237
238 buc = ROFF_HASH(p);
239
240 if (NULL == (n = hash[buc]))
241 return(ROFF_MAX);
242 for ( ; n; n = n->next)
243 if (0 == strncmp(n->name, p, s) && '\0' == n->name[(int)s])
244 return((enum rofft)(n - roffs));
245
246 return(ROFF_MAX);
247 }
248
249
250 /*
251 * Pop the current node off of the stack of roff instructions currently
252 * pending.
253 */
254 static void
255 roffnode_pop(struct roff *r)
256 {
257 struct roffnode *p;
258
259 assert(r->last);
260 p = r->last;
261
262 if (ROFF_el == p->tok)
263 if (r->rstackpos > -1)
264 r->rstackpos--;
265
266 r->last = r->last->parent;
267 free(p->name);
268 free(p->end);
269 free(p);
270 }
271
272
273 /*
274 * Push a roff node onto the instruction stack. This must later be
275 * removed with roffnode_pop().
276 */
277 static void
278 roffnode_push(struct roff *r, enum rofft tok, const char *name,
279 int line, int col)
280 {
281 struct roffnode *p;
282
283 p = mandoc_calloc(1, sizeof(struct roffnode));
284 p->tok = tok;
285 if (name)
286 p->name = mandoc_strdup(name);
287 p->parent = r->last;
288 p->line = line;
289 p->col = col;
290 p->rule = p->parent ? p->parent->rule : ROFFRULE_DENY;
291
292 r->last = p;
293 }
294
295
296 static void
297 roff_free1(struct roff *r)
298 {
299
300 if (r->tbl) {
301 tbl_free(r->tbl);
302 r->tbl = NULL;
303 }
304
305 while (r->last)
306 roffnode_pop(r);
307
308 roff_freestr(r);
309 }
310
311
312 void
313 roff_reset(struct roff *r)
314 {
315
316 roff_free1(r);
317 }
318
319
320 void
321 roff_free(struct roff *r)
322 {
323
324 roff_free1(r);
325 free(r);
326 }
327
328
329 struct roff *
330 roff_alloc(struct regset *regs, void *data, const mandocmsg msg)
331 {
332 struct roff *r;
333
334 r = mandoc_calloc(1, sizeof(struct roff));
335 r->regs = regs;
336 r->msg = msg;
337 r->data = data;
338 r->rstackpos = -1;
339
340 roff_hash_init();
341 return(r);
342 }
343
344
345 /*
346 * Pre-filter each and every line for reserved words (one beginning with
347 * `\*', e.g., `\*(ab'). These must be handled before the actual line
348 * is processed.
349 */
350 static int
351 roff_res(struct roff *r, char **bufp, size_t *szp, int pos)
352 {
353 const char *stesc; /* start of an escape sequence ('\\') */
354 const char *stnam; /* start of the name, after "[(*" */
355 const char *cp; /* end of the name, e.g. before ']' */
356 const char *res; /* the string to be substituted */
357 int i, maxl;
358 size_t nsz;
359 char *n;
360
361 /* Search for a leading backslash and save a pointer to it. */
362
363 cp = *bufp + pos;
364 while (NULL != (cp = strchr(cp, '\\'))) {
365 stesc = cp++;
366
367 /*
368 * The second character must be an asterisk.
369 * If it isn't, skip it anyway: It is escaped,
370 * so it can't start another escape sequence.
371 */
372
373 if ('\0' == *cp)
374 return(1);
375 if ('*' != *cp++)
376 continue;
377
378 /*
379 * The third character decides the length
380 * of the name of the string.
381 * Save a pointer to the name.
382 */
383
384 switch (*cp) {
385 case ('\0'):
386 return(1);
387 case ('('):
388 cp++;
389 maxl = 2;
390 break;
391 case ('['):
392 cp++;
393 maxl = 0;
394 break;
395 default:
396 maxl = 1;
397 break;
398 }
399 stnam = cp;
400
401 /* Advance to the end of the name. */
402
403 for (i = 0; 0 == maxl || i < maxl; i++, cp++) {
404 if ('\0' == *cp)
405 return(1); /* Error. */
406 if (0 == maxl && ']' == *cp)
407 break;
408 }
409
410 /*
411 * Retrieve the replacement string; if it is
412 * undefined, resume searching for escapes.
413 */
414
415 res = roff_getstrn(r, stnam, (size_t)i);
416
417 if (NULL == res) {
418 cp -= maxl ? 1 : 0;
419 continue;
420 }
421
422 /* Replace the escape sequence by the string. */
423
424 nsz = *szp + strlen(res) + 1;
425 n = mandoc_malloc(nsz);
426
427 strlcpy(n, *bufp, (size_t)(stesc - *bufp + 1));
428 strlcat(n, res, nsz);
429 strlcat(n, cp + (maxl ? 0 : 1), nsz);
430
431 free(*bufp);
432
433 *bufp = n;
434 *szp = nsz;
435 return(0);
436 }
437
438 return(1);
439 }
440
441
442 enum rofferr
443 roff_parseln(struct roff *r, int ln, char **bufp,
444 size_t *szp, int pos, int *offs)
445 {
446 enum rofft t;
447 enum rofferr e;
448 int ppos;
449
450 /*
451 * Run the reserved-word filter only if we have some reserved
452 * words to fill in.
453 */
454
455 if (r->first_string && ! roff_res(r, bufp, szp, pos))
456 return(ROFF_REPARSE);
457
458 /*
459 * First, if a scope is open and we're not a macro, pass the
460 * text through the macro's filter. If a scope isn't open and
461 * we're not a macro, just let it through.
462 */
463
464 if (r->last && ! ROFF_CTL((*bufp)[pos])) {
465 t = r->last->tok;
466 assert(roffs[t].text);
467 e = (*roffs[t].text)
468 (r, t, bufp, szp, ln, pos, pos, offs);
469 assert(ROFF_IGN == e || ROFF_CONT == e);
470 if (ROFF_CONT == e && r->tbl)
471 return(tbl_read(r->tbl, ln, *bufp, *offs));
472 return(e);
473 } else if ( ! ROFF_CTL((*bufp)[pos])) {
474 if (r->tbl)
475 return(tbl_read(r->tbl, ln, *bufp, *offs));
476 return(ROFF_CONT);
477 }
478
479 /*
480 * If a scope is open, go to the child handler for that macro,
481 * as it may want to preprocess before doing anything with it.
482 */
483
484 if (r->last) {
485 t = r->last->tok;
486 assert(roffs[t].sub);
487 return((*roffs[t].sub)
488 (r, t, bufp, szp,
489 ln, pos, pos, offs));
490 }
491
492 /*
493 * Lastly, as we've no scope open, try to look up and execute
494 * the new macro. If no macro is found, simply return and let
495 * the compilers handle it.
496 */
497
498 ppos = pos;
499 if (ROFF_MAX == (t = roff_parse(r, *bufp, &pos)))
500 return(ROFF_CONT);
501
502 assert(roffs[t].proc);
503 return((*roffs[t].proc)
504 (r, t, bufp, szp,
505 ln, ppos, pos, offs));
506 }
507
508
509 int
510 roff_endparse(struct roff *r)
511 {
512
513 if (r->last || r->tbl)
514 (*r->msg)(MANDOCERR_SCOPEEXIT, r->data,
515 r->last->line, r->last->col, NULL);
516 return(1);
517 }
518
519
520 /*
521 * Parse a roff node's type from the input buffer. This must be in the
522 * form of ".foo xxx" in the usual way.
523 */
524 static enum rofft
525 roff_parse(struct roff *r, const char *buf, int *pos)
526 {
527 const char *mac;
528 size_t maclen;
529 enum rofft t;
530
531 assert(ROFF_CTL(buf[*pos]));
532 (*pos)++;
533
534 while (' ' == buf[*pos] || '\t' == buf[*pos])
535 (*pos)++;
536
537 if ('\0' == buf[*pos])
538 return(ROFF_MAX);
539
540 mac = buf + *pos;
541 maclen = strcspn(mac, " \\\t\0");
542
543 t = (r->current_string = roff_getstrn(r, mac, maclen))
544 ? ROFF_USERDEF : roff_hash_find(mac, maclen);
545
546 *pos += maclen;
547 while (buf[*pos] && ' ' == buf[*pos])
548 (*pos)++;
549
550 return(t);
551 }
552
553
554 static int
555 roff_parse_nat(const char *buf, unsigned int *res)
556 {
557 char *ep;
558 long lval;
559
560 errno = 0;
561 lval = strtol(buf, &ep, 10);
562 if (buf[0] == '\0' || *ep != '\0')
563 return(0);
564 if ((errno == ERANGE &&
565 (lval == LONG_MAX || lval == LONG_MIN)) ||
566 (lval > INT_MAX || lval < 0))
567 return(0);
568
569 *res = (unsigned int)lval;
570 return(1);
571 }
572
573
574 /* ARGSUSED */
575 static enum rofferr
576 roff_cblock(ROFF_ARGS)
577 {
578
579 /*
580 * A block-close `..' should only be invoked as a child of an
581 * ignore macro, otherwise raise a warning and just ignore it.
582 */
583
584 if (NULL == r->last) {
585 (*r->msg)(MANDOCERR_NOSCOPE, r->data, ln, ppos, NULL);
586 return(ROFF_IGN);
587 }
588
589 switch (r->last->tok) {
590 case (ROFF_am):
591 /* FALLTHROUGH */
592 case (ROFF_ami):
593 /* FALLTHROUGH */
594 case (ROFF_am1):
595 /* FALLTHROUGH */
596 case (ROFF_de):
597 /* ROFF_de1 is remapped to ROFF_de in roff_block(). */
598 /* FALLTHROUGH */
599 case (ROFF_dei):
600 /* FALLTHROUGH */
601 case (ROFF_ig):
602 break;
603 default:
604 (*r->msg)(MANDOCERR_NOSCOPE, r->data, ln, ppos, NULL);
605 return(ROFF_IGN);
606 }
607
608 if ((*bufp)[pos])
609 (*r->msg)(MANDOCERR_ARGSLOST, r->data, ln, pos, NULL);
610
611 roffnode_pop(r);
612 roffnode_cleanscope(r);
613 return(ROFF_IGN);
614
615 }
616
617
618 static void
619 roffnode_cleanscope(struct roff *r)
620 {
621
622 while (r->last) {
623 if (--r->last->endspan < 0)
624 break;
625 roffnode_pop(r);
626 }
627 }
628
629
630 /* ARGSUSED */
631 static enum rofferr
632 roff_ccond(ROFF_ARGS)
633 {
634
635 if (NULL == r->last) {
636 (*r->msg)(MANDOCERR_NOSCOPE, r->data, ln, ppos, NULL);
637 return(ROFF_IGN);
638 }
639
640 switch (r->last->tok) {
641 case (ROFF_el):
642 /* FALLTHROUGH */
643 case (ROFF_ie):
644 /* FALLTHROUGH */
645 case (ROFF_if):
646 break;
647 default:
648 (*r->msg)(MANDOCERR_NOSCOPE, r->data, ln, ppos, NULL);
649 return(ROFF_IGN);
650 }
651
652 if (r->last->endspan > -1) {
653 (*r->msg)(MANDOCERR_NOSCOPE, r->data, ln, ppos, NULL);
654 return(ROFF_IGN);
655 }
656
657 if ((*bufp)[pos])
658 (*r->msg)(MANDOCERR_ARGSLOST, r->data, ln, pos, NULL);
659
660 roffnode_pop(r);
661 roffnode_cleanscope(r);
662 return(ROFF_IGN);
663 }
664
665
666 /* ARGSUSED */
667 static enum rofferr
668 roff_block(ROFF_ARGS)
669 {
670 int sv;
671 size_t sz;
672 char *name;
673
674 name = NULL;
675
676 if (ROFF_ig != tok) {
677 if ('\0' == (*bufp)[pos]) {
678 (*r->msg)(MANDOCERR_NOARGS, r->data, ln, ppos, NULL);
679 return(ROFF_IGN);
680 }
681
682 /*
683 * Re-write `de1', since we don't really care about
684 * groff's strange compatibility mode, into `de'.
685 */
686
687 if (ROFF_de1 == tok)
688 tok = ROFF_de;
689 if (ROFF_de == tok)
690 name = *bufp + pos;
691 else
692 (*r->msg)(MANDOCERR_REQUEST, r->data, ln, ppos,
693 roffs[tok].name);
694
695 while ((*bufp)[pos] && ' ' != (*bufp)[pos])
696 pos++;
697
698 while (' ' == (*bufp)[pos])
699 (*bufp)[pos++] = '\0';
700 }
701
702 roffnode_push(r, tok, name, ln, ppos);
703
704 /*
705 * At the beginning of a `de' macro, clear the existing string
706 * with the same name, if there is one. New content will be
707 * added from roff_block_text() in multiline mode.
708 */
709
710 if (ROFF_de == tok)
711 roff_setstr(r, name, "", 0);
712
713 if ('\0' == (*bufp)[pos])
714 return(ROFF_IGN);
715
716 /* If present, process the custom end-of-line marker. */
717
718 sv = pos;
719 while ((*bufp)[pos] &&
720 ' ' != (*bufp)[pos] &&
721 '\t' != (*bufp)[pos])
722 pos++;
723
724 /*
725 * Note: groff does NOT like escape characters in the input.
726 * Instead of detecting this, we're just going to let it fly and
727 * to hell with it.
728 */
729
730 assert(pos > sv);
731 sz = (size_t)(pos - sv);
732
733 if (1 == sz && '.' == (*bufp)[sv])
734 return(ROFF_IGN);
735
736 r->last->end = mandoc_malloc(sz + 1);
737
738 memcpy(r->last->end, *bufp + sv, sz);
739 r->last->end[(int)sz] = '\0';
740
741 if ((*bufp)[pos])
742 (*r->msg)(MANDOCERR_ARGSLOST, r->data, ln, pos, NULL);
743
744 return(ROFF_IGN);
745 }
746
747
748 /* ARGSUSED */
749 static enum rofferr
750 roff_block_sub(ROFF_ARGS)
751 {
752 enum rofft t;
753 int i, j;
754
755 /*
756 * First check whether a custom macro exists at this level. If
757 * it does, then check against it. This is some of groff's
758 * stranger behaviours. If we encountered a custom end-scope
759 * tag and that tag also happens to be a "real" macro, then we
760 * need to try interpreting it again as a real macro. If it's
761 * not, then return ignore. Else continue.
762 */
763
764 if (r->last->end) {
765 i = pos + 1;
766 while (' ' == (*bufp)[i] || '\t' == (*bufp)[i])
767 i++;
768
769 for (j = 0; r->last->end[j]; j++, i++)
770 if ((*bufp)[i] != r->last->end[j])
771 break;
772
773 if ('\0' == r->last->end[j] &&
774 ('\0' == (*bufp)[i] ||
775 ' ' == (*bufp)[i] ||
776 '\t' == (*bufp)[i])) {
777 roffnode_pop(r);
778 roffnode_cleanscope(r);
779
780 if (ROFF_MAX != roff_parse(r, *bufp, &pos))
781 return(ROFF_RERUN);
782 return(ROFF_IGN);
783 }
784 }
785
786 /*
787 * If we have no custom end-query or lookup failed, then try
788 * pulling it out of the hashtable.
789 */
790
791 ppos = pos;
792 t = roff_parse(r, *bufp, &pos);
793
794 /*
795 * Macros other than block-end are only significant
796 * in `de' blocks; elsewhere, simply throw them away.
797 */
798 if (ROFF_cblock != t) {
799 if (ROFF_de == tok)
800 roff_setstr(r, r->last->name, *bufp + ppos, 1);
801 return(ROFF_IGN);
802 }
803
804 assert(roffs[t].proc);
805 return((*roffs[t].proc)(r, t, bufp, szp,
806 ln, ppos, pos, offs));
807 }
808
809
810 /* ARGSUSED */
811 static enum rofferr
812 roff_block_text(ROFF_ARGS)
813 {
814
815 if (ROFF_de == tok)
816 roff_setstr(r, r->last->name, *bufp + pos, 1);
817
818 return(ROFF_IGN);
819 }
820
821
822 /* ARGSUSED */
823 static enum rofferr
824 roff_cond_sub(ROFF_ARGS)
825 {
826 enum rofft t;
827 enum roffrule rr;
828
829 ppos = pos;
830 rr = r->last->rule;
831
832 /*
833 * Clean out scope. If we've closed ourselves, then don't
834 * continue.
835 */
836
837 roffnode_cleanscope(r);
838
839 if (ROFF_MAX == (t = roff_parse(r, *bufp, &pos))) {
840 if ('\\' == (*bufp)[pos] && '}' == (*bufp)[pos + 1])
841 return(roff_ccond
842 (r, ROFF_ccond, bufp, szp,
843 ln, pos, pos + 2, offs));
844 return(ROFFRULE_DENY == rr ? ROFF_IGN : ROFF_CONT);
845 }
846
847 /*
848 * A denied conditional must evaluate its children if and only
849 * if they're either structurally required (such as loops and
850 * conditionals) or a closing macro.
851 */
852 if (ROFFRULE_DENY == rr)
853 if ( ! (ROFFMAC_STRUCT & roffs[t].flags))
854 if (ROFF_ccond != t)
855 return(ROFF_IGN);
856
857 assert(roffs[t].proc);
858 return((*roffs[t].proc)(r, t, bufp, szp,
859 ln, ppos, pos, offs));
860 }
861
862
863 /* ARGSUSED */
864 static enum rofferr
865 roff_cond_text(ROFF_ARGS)
866 {
867 char *ep, *st;
868 enum roffrule rr;
869
870 rr = r->last->rule;
871
872 /*
873 * We display the value of the text if out current evaluation
874 * scope permits us to do so.
875 */
876
877 /* FIXME: use roff_ccond? */
878
879 st = &(*bufp)[pos];
880 if (NULL == (ep = strstr(st, "\\}"))) {
881 roffnode_cleanscope(r);
882 return(ROFFRULE_DENY == rr ? ROFF_IGN : ROFF_CONT);
883 }
884
885 if (ep == st || (ep > st && '\\' != *(ep - 1)))
886 roffnode_pop(r);
887
888 roffnode_cleanscope(r);
889 return(ROFFRULE_DENY == rr ? ROFF_IGN : ROFF_CONT);
890 }
891
892
893 static enum roffrule
894 roff_evalcond(const char *v, int *pos)
895 {
896
897 switch (v[*pos]) {
898 case ('n'):
899 (*pos)++;
900 return(ROFFRULE_ALLOW);
901 case ('e'):
902 /* FALLTHROUGH */
903 case ('o'):
904 /* FALLTHROUGH */
905 case ('t'):
906 (*pos)++;
907 return(ROFFRULE_DENY);
908 default:
909 break;
910 }
911
912 while (v[*pos] && ' ' != v[*pos])
913 (*pos)++;
914 return(ROFFRULE_DENY);
915 }
916
917 /* ARGSUSED */
918 static enum rofferr
919 roff_line_ignore(ROFF_ARGS)
920 {
921
922 return(ROFF_IGN);
923 }
924
925 /* ARGSUSED */
926 static enum rofferr
927 roff_line_error(ROFF_ARGS)
928 {
929
930 (*r->msg)(MANDOCERR_REQUEST, r->data, ln, ppos, roffs[tok].name);
931 return(ROFF_IGN);
932 }
933
934 /* ARGSUSED */
935 static enum rofferr
936 roff_cond(ROFF_ARGS)
937 {
938 int sv;
939 enum roffrule rule;
940
941 /* Stack overflow! */
942
943 if (ROFF_ie == tok && r->rstackpos == RSTACK_MAX - 1) {
944 (*r->msg)(MANDOCERR_MEM, r->data, ln, ppos, NULL);
945 return(ROFF_ERR);
946 }
947
948 /* First, evaluate the conditional. */
949
950 if (ROFF_el == tok) {
951 /*
952 * An `.el' will get the value of the current rstack
953 * entry set in prior `ie' calls or defaults to DENY.
954 */
955 if (r->rstackpos < 0)
956 rule = ROFFRULE_DENY;
957 else
958 rule = r->rstack[r->rstackpos];
959 } else
960 rule = roff_evalcond(*bufp, &pos);
961
962 sv = pos;
963
964 while (' ' == (*bufp)[pos])
965 pos++;
966
967 /*
968 * Roff is weird. If we have just white-space after the
969 * conditional, it's considered the BODY and we exit without
970 * really doing anything. Warn about this. It's probably
971 * wrong.
972 */
973
974 if ('\0' == (*bufp)[pos] && sv != pos) {
975 (*r->msg)(MANDOCERR_NOARGS, r->data, ln, ppos, NULL);
976 return(ROFF_IGN);
977 }
978
979 roffnode_push(r, tok, NULL, ln, ppos);
980
981 r->last->rule = rule;
982
983 if (ROFF_ie == tok) {
984 /*
985 * An if-else will put the NEGATION of the current
986 * evaluated conditional into the stack.
987 */
988 r->rstackpos++;
989 if (ROFFRULE_DENY == r->last->rule)
990 r->rstack[r->rstackpos] = ROFFRULE_ALLOW;
991 else
992 r->rstack[r->rstackpos] = ROFFRULE_DENY;
993 }
994
995 /* If the parent has false as its rule, then so do we. */
996
997 if (r->last->parent && ROFFRULE_DENY == r->last->parent->rule)
998 r->last->rule = ROFFRULE_DENY;
999
1000 /*
1001 * Determine scope. If we're invoked with "\{" trailing the
1002 * conditional, then we're in a multiline scope. Else our scope
1003 * expires on the next line.
1004 */
1005
1006 r->last->endspan = 1;
1007
1008 if ('\\' == (*bufp)[pos] && '{' == (*bufp)[pos + 1]) {
1009 r->last->endspan = -1;
1010 pos += 2;
1011 }
1012
1013 /*
1014 * If there are no arguments on the line, the next-line scope is
1015 * assumed.
1016 */
1017
1018 if ('\0' == (*bufp)[pos])
1019 return(ROFF_IGN);
1020
1021 /* Otherwise re-run the roff parser after recalculating. */
1022
1023 *offs = pos;
1024 return(ROFF_RERUN);
1025 }
1026
1027
1028 /* ARGSUSED */
1029 static enum rofferr
1030 roff_ds(ROFF_ARGS)
1031 {
1032 char *name, *string;
1033
1034 /*
1035 * A symbol is named by the first word following the macro
1036 * invocation up to a space. Its value is anything after the
1037 * name's trailing whitespace and optional double-quote. Thus,
1038 *
1039 * [.ds foo "bar " ]
1040 *
1041 * will have `bar " ' as its value.
1042 */
1043
1044 name = *bufp + pos;
1045 if ('\0' == *name)
1046 return(ROFF_IGN);
1047
1048 string = name;
1049 /* Read until end of name. */
1050 while (*string && ' ' != *string)
1051 string++;
1052
1053 /* Nil-terminate name. */
1054 if (*string)
1055 *(string++) = '\0';
1056
1057 /* Read past spaces. */
1058 while (*string && ' ' == *string)
1059 string++;
1060
1061 /* Read passed initial double-quote. */
1062 if (*string && '"' == *string)
1063 string++;
1064
1065 /* The rest is the value. */
1066 roff_setstr(r, name, string, 0);
1067 return(ROFF_IGN);
1068 }
1069
1070
1071 /* ARGSUSED */
1072 static enum rofferr
1073 roff_nr(ROFF_ARGS)
1074 {
1075 const char *key, *val;
1076 struct reg *rg;
1077
1078 key = &(*bufp)[pos];
1079 rg = r->regs->regs;
1080
1081 /* Parse register request. */
1082 while ((*bufp)[pos] && ' ' != (*bufp)[pos])
1083 pos++;
1084
1085 /*
1086 * Set our nil terminator. Because this line is going to be
1087 * ignored anyway, we can munge it as we please.
1088 */
1089 if ((*bufp)[pos])
1090 (*bufp)[pos++] = '\0';
1091
1092 /* Skip whitespace to register token. */
1093 while ((*bufp)[pos] && ' ' == (*bufp)[pos])
1094 pos++;
1095
1096 val = &(*bufp)[pos];
1097
1098 /* Process register token. */
1099
1100 if (0 == strcmp(key, "nS")) {
1101 rg[(int)REG_nS].set = 1;
1102 if ( ! roff_parse_nat(val, &rg[(int)REG_nS].v.u))
1103 rg[(int)REG_nS].v.u = 0;
1104 }
1105
1106 return(ROFF_IGN);
1107 }
1108
1109 /* ARGSUSED */
1110 static enum rofferr
1111 roff_TE(ROFF_ARGS)
1112 {
1113
1114 if (NULL == r->tbl)
1115 (*r->msg)(MANDOCERR_NOSCOPE, r->data, ln, ppos, NULL);
1116 else
1117 tbl_free(r->tbl);
1118
1119 r->tbl = NULL;
1120 return(ROFF_IGN);
1121 }
1122
1123 /* ARGSUSED */
1124 static enum rofferr
1125 roff_TS(ROFF_ARGS)
1126 {
1127
1128 if (r->tbl) {
1129 (*r->msg)(MANDOCERR_SCOPEBROKEN, r->data, ln, ppos, NULL);
1130 tbl_reset(r->tbl);
1131 } else
1132 r->tbl = tbl_alloc();
1133
1134 return(ROFF_IGN);
1135 }
1136
1137 /* ARGSUSED */
1138 static enum rofferr
1139 roff_so(ROFF_ARGS)
1140 {
1141 char *name;
1142
1143 (*r->msg)(MANDOCERR_SO, r->data, ln, ppos, NULL);
1144
1145 /*
1146 * Handle `so'. Be EXTREMELY careful, as we shouldn't be
1147 * opening anything that's not in our cwd or anything beneath
1148 * it. Thus, explicitly disallow traversing up the file-system
1149 * or using absolute paths.
1150 */
1151
1152 name = *bufp + pos;
1153 if ('/' == *name || strstr(name, "../") || strstr(name, "/..")) {
1154 (*r->msg)(MANDOCERR_SOPATH, r->data, ln, pos, NULL);
1155 return(ROFF_ERR);
1156 }
1157
1158 *offs = pos;
1159 return(ROFF_SO);
1160 }
1161
1162 /* ARGSUSED */
1163 static enum rofferr
1164 roff_userdef(ROFF_ARGS)
1165 {
1166 const char *arg[9];
1167 char *cp, *n1, *n2;
1168 int i, quoted, pairs;
1169
1170 /*
1171 * Collect pointers to macro argument strings
1172 * and null-terminate them.
1173 */
1174 cp = *bufp + pos;
1175 for (i = 0; i < 9; i++) {
1176 /* Quoting can only start with a new word. */
1177 if ('"' == *cp) {
1178 quoted = 1;
1179 cp++;
1180 } else
1181 quoted = 0;
1182 arg[i] = cp;
1183 for (pairs = 0; '\0' != *cp; cp++) {
1184 /* Unquoted arguments end at blanks. */
1185 if (0 == quoted) {
1186 if (' ' == *cp)
1187 break;
1188 continue;
1189 }
1190 /* After pairs of quotes, move left. */
1191 if (pairs)
1192 cp[-pairs] = cp[0];
1193 /* Pairs of quotes do not end words, ... */
1194 if ('"' == cp[0] && '"' == cp[1]) {
1195 pairs++;
1196 cp++;
1197 continue;
1198 }
1199 /* ... but solitary quotes do. */
1200 if ('"' != *cp)
1201 continue;
1202 if (pairs)
1203 cp[-pairs] = '\0';
1204 *cp = ' ';
1205 break;
1206 }
1207 /* Last argument; the remaining ones are empty strings. */
1208 if ('\0' == *cp)
1209 continue;
1210 /* Null-terminate argument and move to the next one. */
1211 *cp++ = '\0';
1212 while (' ' == *cp)
1213 cp++;
1214 }
1215
1216 /*
1217 * Expand macro arguments.
1218 */
1219 *szp = 0;
1220 n1 = cp = mandoc_strdup(r->current_string);
1221 while (NULL != (cp = strstr(cp, "\\$"))) {
1222 i = cp[2] - '1';
1223 if (0 > i || 8 < i) {
1224 /* Not an argument invocation. */
1225 cp += 2;
1226 continue;
1227 }
1228
1229 *szp = strlen(n1) - 3 + strlen(arg[i]) + 1;
1230 n2 = mandoc_malloc(*szp);
1231
1232 strlcpy(n2, n1, (size_t)(cp - n1 + 1));
1233 strlcat(n2, arg[i], *szp);
1234 strlcat(n2, cp + 3, *szp);
1235
1236 cp = n2 + (cp - n1);
1237 free(n1);
1238 n1 = n2;
1239 }
1240
1241 /*
1242 * Replace the macro invocation
1243 * by the expanded macro.
1244 */
1245 free(*bufp);
1246 *bufp = n1;
1247 if (0 == *szp)
1248 *szp = strlen(*bufp) + 1;
1249
1250 return(*szp > 1 && '\n' == (*bufp)[(int)*szp - 2] ?
1251 ROFF_REPARSE : ROFF_APPEND);
1252 }
1253
1254 /*
1255 * Store *string into the user-defined string called *name.
1256 * In multiline mode, append to an existing entry and append '\n';
1257 * else replace the existing entry, if there is one.
1258 * To clear an existing entry, call with (*r, *name, NULL, 0).
1259 */
1260 static void
1261 roff_setstr(struct roff *r, const char *name, const char *string,
1262 int multiline)
1263 {
1264 struct roffstr *n;
1265 char *c;
1266 size_t oldch, newch;
1267
1268 /* Search for an existing string with the same name. */
1269 n = r->first_string;
1270 while (n && strcmp(name, n->name))
1271 n = n->next;
1272
1273 if (NULL == n) {
1274 /* Create a new string table entry. */
1275 n = mandoc_malloc(sizeof(struct roffstr));
1276 n->name = mandoc_strdup(name);
1277 n->string = NULL;
1278 n->next = r->first_string;
1279 r->first_string = n;
1280 } else if (0 == multiline) {
1281 /* In multiline mode, append; else replace. */
1282 free(n->string);
1283 n->string = NULL;
1284 }
1285
1286 if (NULL == string)
1287 return;
1288
1289 /*
1290 * One additional byte for the '\n' in multiline mode,
1291 * and one for the terminating '\0'.
1292 */
1293 newch = strlen(string) + (multiline ? 2 : 1);
1294 if (NULL == n->string) {
1295 n->string = mandoc_malloc(newch);
1296 *n->string = '\0';
1297 oldch = 0;
1298 } else {
1299 oldch = strlen(n->string);
1300 n->string = mandoc_realloc(n->string, oldch + newch);
1301 }
1302
1303 /* Skip existing content in the destination buffer. */
1304 c = n->string + oldch;
1305
1306 /* Append new content to the destination buffer. */
1307 while (*string) {
1308 /*
1309 * Rudimentary roff copy mode:
1310 * Handle escaped backslashes.
1311 */
1312 if ('\\' == *string && '\\' == *(string + 1))
1313 string++;
1314 *c++ = *string++;
1315 }
1316
1317 /* Append terminating bytes. */
1318 if (multiline)
1319 *c++ = '\n';
1320 *c = '\0';
1321 }
1322
1323
1324 static const char *
1325 roff_getstrn(const struct roff *r, const char *name, size_t len)
1326 {
1327 const struct roffstr *n;
1328
1329 n = r->first_string;
1330 while (n && (strncmp(name, n->name, len) || '\0' != n->name[(int)len]))
1331 n = n->next;
1332
1333 return(n ? n->string : NULL);
1334 }
1335
1336
1337 static void
1338 roff_freestr(struct roff *r)
1339 {
1340 struct roffstr *n, *nn;
1341
1342 for (n = r->first_string; n; n = nn) {
1343 free(n->name);
1344 free(n->string);
1345 nn = n->next;
1346 free(n);
1347 }
1348
1349 r->first_string = NULL;
1350 }
A git conversion mirror of mandoc.bsd.lv