Import mdocml-1.10.9
[dragonfly.git] / contrib / mdocml / roff.c
CommitLineData
80387638
SW
1/* $Id: roff.c,v 1.120 2011/01/03 23:24:16 schwarze Exp $ */
2/*
3 * Copyright (c) 2010, 2011 Kristaps Dzonsons <kristaps@bsd.lv>
4 * Copyright (c) 2010, 2011 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
40enum 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_T_,
63 ROFF_cblock,
64 ROFF_ccond, /* FIXME: remove this. */
65 ROFF_USERDEF,
66 ROFF_MAX
67};
68
69enum roffrule {
70 ROFFRULE_ALLOW,
71 ROFFRULE_DENY
72};
73
74struct roffstr {
75 char *name; /* key of symbol */
76 char *string; /* current value */
77 struct roffstr *next; /* next in list */
78};
79
80struct roff {
81 struct roffnode *last; /* leaf of stack */
82 mandocmsg msg; /* err/warn/fatal messages */
83 void *data; /* privdata for messages */
84 enum roffrule rstack[RSTACK_MAX]; /* stack of !`ie' rules */
85 int rstackpos; /* position in rstack */
86 struct regset *regs; /* read/writable registers */
87 struct roffstr *first_string; /* user-defined strings & macros */
88 const char *current_string; /* value of last called user macro */
89 struct tbl_node *first_tbl; /* first table parsed */
90 struct tbl_node *last_tbl; /* last table parsed */
91 struct tbl_node *tbl; /* current table being parsed */
92};
93
94struct roffnode {
95 enum rofft tok; /* type of node */
96 struct roffnode *parent; /* up one in stack */
97 int line; /* parse line */
98 int col; /* parse col */
99 char *name; /* node name, e.g. macro name */
100 char *end; /* end-rules: custom token */
101 int endspan; /* end-rules: next-line or infty */
102 enum roffrule rule; /* current evaluation rule */
103};
104
105#define ROFF_ARGS struct roff *r, /* parse ctx */ \
106 enum rofft tok, /* tok of macro */ \
107 char **bufp, /* input buffer */ \
108 size_t *szp, /* size of input buffer */ \
109 int ln, /* parse line */ \
110 int ppos, /* original pos in buffer */ \
111 int pos, /* current pos in buffer */ \
112 int *offs /* reset offset of buffer data */
113
114typedef enum rofferr (*roffproc)(ROFF_ARGS);
115
116struct roffmac {
117 const char *name; /* macro name */
118 roffproc proc; /* process new macro */
119 roffproc text; /* process as child text of macro */
120 roffproc sub; /* process as child of macro */
121 int flags;
122#define ROFFMAC_STRUCT (1 << 0) /* always interpret */
123 struct roffmac *next;
124};
125
126static enum rofferr roff_block(ROFF_ARGS);
127static enum rofferr roff_block_text(ROFF_ARGS);
128static enum rofferr roff_block_sub(ROFF_ARGS);
129static enum rofferr roff_cblock(ROFF_ARGS);
130static enum rofferr roff_ccond(ROFF_ARGS);
131static enum rofferr roff_cond(ROFF_ARGS);
132static enum rofferr roff_cond_text(ROFF_ARGS);
133static enum rofferr roff_cond_sub(ROFF_ARGS);
134static enum rofferr roff_ds(ROFF_ARGS);
135static enum roffrule roff_evalcond(const char *, int *);
136static void roff_freestr(struct roff *);
137static const char *roff_getstrn(const struct roff *,
138 const char *, size_t);
139static enum rofferr roff_line_ignore(ROFF_ARGS);
140static enum rofferr roff_line_error(ROFF_ARGS);
141static enum rofferr roff_nr(ROFF_ARGS);
142static int roff_res(struct roff *,
143 char **, size_t *, int);
144static void roff_setstr(struct roff *,
145 const char *, const char *, int);
146static enum rofferr roff_so(ROFF_ARGS);
147static enum rofferr roff_TE(ROFF_ARGS);
148static enum rofferr roff_TS(ROFF_ARGS);
149static enum rofferr roff_T_(ROFF_ARGS);
150static enum rofferr roff_userdef(ROFF_ARGS);
151
152/* See roff_hash_find() */
153
154#define ASCII_HI 126
155#define ASCII_LO 33
156#define HASHWIDTH (ASCII_HI - ASCII_LO + 1)
157
158static struct roffmac *hash[HASHWIDTH];
159
160static struct roffmac roffs[ROFF_MAX] = {
161 { "ad", roff_line_ignore, NULL, NULL, 0, NULL },
162 { "am", roff_block, roff_block_text, roff_block_sub, 0, NULL },
163 { "ami", roff_block, roff_block_text, roff_block_sub, 0, NULL },
164 { "am1", roff_block, roff_block_text, roff_block_sub, 0, NULL },
165 { "de", roff_block, roff_block_text, roff_block_sub, 0, NULL },
166 { "dei", roff_block, roff_block_text, roff_block_sub, 0, NULL },
167 { "de1", roff_block, roff_block_text, roff_block_sub, 0, NULL },
168 { "ds", roff_ds, NULL, NULL, 0, NULL },
169 { "el", roff_cond, roff_cond_text, roff_cond_sub, ROFFMAC_STRUCT, NULL },
170 { "hy", roff_line_ignore, NULL, NULL, 0, NULL },
171 { "ie", roff_cond, roff_cond_text, roff_cond_sub, ROFFMAC_STRUCT, NULL },
172 { "if", roff_cond, roff_cond_text, roff_cond_sub, ROFFMAC_STRUCT, NULL },
173 { "ig", roff_block, roff_block_text, roff_block_sub, 0, NULL },
174 { "ne", roff_line_ignore, NULL, NULL, 0, NULL },
175 { "nh", roff_line_ignore, NULL, NULL, 0, NULL },
176 { "nr", roff_nr, NULL, NULL, 0, NULL },
177 { "rm", roff_line_error, NULL, NULL, 0, NULL },
178 { "so", roff_so, NULL, NULL, 0, NULL },
179 { "tr", roff_line_ignore, NULL, NULL, 0, NULL },
180 { "TS", roff_TS, NULL, NULL, 0, NULL },
181 { "TE", roff_TE, NULL, NULL, 0, NULL },
182 { "T&", roff_T_, NULL, NULL, 0, NULL },
183 { ".", roff_cblock, NULL, NULL, 0, NULL },
184 { "\\}", roff_ccond, NULL, NULL, 0, NULL },
185 { NULL, roff_userdef, NULL, NULL, 0, NULL },
186};
187
188static void roff_free1(struct roff *);
189static enum rofft roff_hash_find(const char *, size_t);
190static void roff_hash_init(void);
191static void roffnode_cleanscope(struct roff *);
192static void roffnode_push(struct roff *, enum rofft,
193 const char *, int, int);
194static void roffnode_pop(struct roff *);
195static enum rofft roff_parse(struct roff *, const char *, int *);
196static int roff_parse_nat(const char *, unsigned int *);
197
198/* See roff_hash_find() */
199#define ROFF_HASH(p) (p[0] - ASCII_LO)
200
201static void
202roff_hash_init(void)
203{
204 struct roffmac *n;
205 int buc, i;
206
207 for (i = 0; i < (int)ROFF_USERDEF; i++) {
208 assert(roffs[i].name[0] >= ASCII_LO);
209 assert(roffs[i].name[0] <= ASCII_HI);
210
211 buc = ROFF_HASH(roffs[i].name);
212
213 if (NULL != (n = hash[buc])) {
214 for ( ; n->next; n = n->next)
215 /* Do nothing. */ ;
216 n->next = &roffs[i];
217 } else
218 hash[buc] = &roffs[i];
219 }
220}
221
222
223/*
224 * Look up a roff token by its name. Returns ROFF_MAX if no macro by
225 * the nil-terminated string name could be found.
226 */
227static enum rofft
228roff_hash_find(const char *p, size_t s)
229{
230 int buc;
231 struct roffmac *n;
232
233 /*
234 * libroff has an extremely simple hashtable, for the time
235 * being, which simply keys on the first character, which must
236 * be printable, then walks a chain. It works well enough until
237 * optimised.
238 */
239
240 if (p[0] < ASCII_LO || p[0] > ASCII_HI)
241 return(ROFF_MAX);
242
243 buc = ROFF_HASH(p);
244
245 if (NULL == (n = hash[buc]))
246 return(ROFF_MAX);
247 for ( ; n; n = n->next)
248 if (0 == strncmp(n->name, p, s) && '\0' == n->name[(int)s])
249 return((enum rofft)(n - roffs));
250
251 return(ROFF_MAX);
252}
253
254
255/*
256 * Pop the current node off of the stack of roff instructions currently
257 * pending.
258 */
259static void
260roffnode_pop(struct roff *r)
261{
262 struct roffnode *p;
263
264 assert(r->last);
265 p = r->last;
266
267 if (ROFF_el == p->tok)
268 if (r->rstackpos > -1)
269 r->rstackpos--;
270
271 r->last = r->last->parent;
272 free(p->name);
273 free(p->end);
274 free(p);
275}
276
277
278/*
279 * Push a roff node onto the instruction stack. This must later be
280 * removed with roffnode_pop().
281 */
282static void
283roffnode_push(struct roff *r, enum rofft tok, const char *name,
284 int line, int col)
285{
286 struct roffnode *p;
287
288 p = mandoc_calloc(1, sizeof(struct roffnode));
289 p->tok = tok;
290 if (name)
291 p->name = mandoc_strdup(name);
292 p->parent = r->last;
293 p->line = line;
294 p->col = col;
295 p->rule = p->parent ? p->parent->rule : ROFFRULE_DENY;
296
297 r->last = p;
298}
299
300
301static void
302roff_free1(struct roff *r)
303{
304 struct tbl_node *t;
305
306 while (r->first_tbl) {
307 t = r->first_tbl;
308 r->first_tbl = t->next;
309 tbl_free(t);
310 }
311
312 r->first_tbl = r->last_tbl = r->tbl = NULL;
313
314 while (r->last)
315 roffnode_pop(r);
316
317 roff_freestr(r);
318}
319
320
321void
322roff_reset(struct roff *r)
323{
324
325 roff_free1(r);
326}
327
328
329void
330roff_free(struct roff *r)
331{
332
333 roff_free1(r);
334 free(r);
335}
336
337
338struct roff *
339roff_alloc(struct regset *regs, void *data, const mandocmsg msg)
340{
341 struct roff *r;
342
343 r = mandoc_calloc(1, sizeof(struct roff));
344 r->regs = regs;
345 r->msg = msg;
346 r->data = data;
347 r->rstackpos = -1;
348
349 roff_hash_init();
350 return(r);
351}
352
353
354/*
355 * Pre-filter each and every line for reserved words (one beginning with
356 * `\*', e.g., `\*(ab'). These must be handled before the actual line
357 * is processed.
358 */
359static int
360roff_res(struct roff *r, char **bufp, size_t *szp, int pos)
361{
362 const char *stesc; /* start of an escape sequence ('\\') */
363 const char *stnam; /* start of the name, after "[(*" */
364 const char *cp; /* end of the name, e.g. before ']' */
365 const char *res; /* the string to be substituted */
366 int i, maxl;
367 size_t nsz;
368 char *n;
369
370 /* Search for a leading backslash and save a pointer to it. */
371
372 cp = *bufp + pos;
373 while (NULL != (cp = strchr(cp, '\\'))) {
374 stesc = cp++;
375
376 /*
377 * The second character must be an asterisk.
378 * If it isn't, skip it anyway: It is escaped,
379 * so it can't start another escape sequence.
380 */
381
382 if ('\0' == *cp)
383 return(1);
384 if ('*' != *cp++)
385 continue;
386
387 /*
388 * The third character decides the length
389 * of the name of the string.
390 * Save a pointer to the name.
391 */
392
393 switch (*cp) {
394 case ('\0'):
395 return(1);
396 case ('('):
397 cp++;
398 maxl = 2;
399 break;
400 case ('['):
401 cp++;
402 maxl = 0;
403 break;
404 default:
405 maxl = 1;
406 break;
407 }
408 stnam = cp;
409
410 /* Advance to the end of the name. */
411
412 for (i = 0; 0 == maxl || i < maxl; i++, cp++) {
413 if ('\0' == *cp)
414 return(1); /* Error. */
415 if (0 == maxl && ']' == *cp)
416 break;
417 }
418
419 /*
420 * Retrieve the replacement string; if it is
421 * undefined, resume searching for escapes.
422 */
423
424 res = roff_getstrn(r, stnam, (size_t)i);
425
426 if (NULL == res) {
427 cp -= maxl ? 1 : 0;
428 continue;
429 }
430
431 /* Replace the escape sequence by the string. */
432
433 nsz = *szp + strlen(res) + 1;
434 n = mandoc_malloc(nsz);
435
436 strlcpy(n, *bufp, (size_t)(stesc - *bufp + 1));
437 strlcat(n, res, nsz);
438 strlcat(n, cp + (maxl ? 0 : 1), nsz);
439
440 free(*bufp);
441
442 *bufp = n;
443 *szp = nsz;
444 return(0);
445 }
446
447 return(1);
448}
449
450
451enum rofferr
452roff_parseln(struct roff *r, int ln, char **bufp,
453 size_t *szp, int pos, int *offs)
454{
455 enum rofft t;
456 enum rofferr e;
457 int ppos;
458
459 /*
460 * Run the reserved-word filter only if we have some reserved
461 * words to fill in.
462 */
463
464 if (r->first_string && ! roff_res(r, bufp, szp, pos))
465 return(ROFF_REPARSE);
466
467 /*
468 * First, if a scope is open and we're not a macro, pass the
469 * text through the macro's filter. If a scope isn't open and
470 * we're not a macro, just let it through.
471 */
472
473 if (r->last && ! ROFF_CTL((*bufp)[pos])) {
474 t = r->last->tok;
475 assert(roffs[t].text);
476 e = (*roffs[t].text)
477 (r, t, bufp, szp, ln, pos, pos, offs);
478 assert(ROFF_IGN == e || ROFF_CONT == e);
479 if (ROFF_CONT == e && r->tbl)
480 return(tbl_read(r->tbl, ln, *bufp, *offs));
481 return(e);
482 } else if ( ! ROFF_CTL((*bufp)[pos])) {
483 if (r->tbl)
484 return(tbl_read(r->tbl, ln, *bufp, *offs));
485 return(ROFF_CONT);
486 }
487
488 /*
489 * If a scope is open, go to the child handler for that macro,
490 * as it may want to preprocess before doing anything with it.
491 */
492
493 if (r->last) {
494 t = r->last->tok;
495 assert(roffs[t].sub);
496 return((*roffs[t].sub)
497 (r, t, bufp, szp,
498 ln, pos, pos, offs));
499 }
500
501 /*
502 * Lastly, as we've no scope open, try to look up and execute
503 * the new macro. If no macro is found, simply return and let
504 * the compilers handle it.
505 */
506
507 ppos = pos;
508 if (ROFF_MAX == (t = roff_parse(r, *bufp, &pos)))
509 return(ROFF_CONT);
510
511 assert(roffs[t].proc);
512 return((*roffs[t].proc)
513 (r, t, bufp, szp,
514 ln, ppos, pos, offs));
515}
516
517
518void
519roff_endparse(struct roff *r)
520{
521
522 if (r->last)
523 (*r->msg)(MANDOCERR_SCOPEEXIT, r->data,
524 r->last->line, r->last->col, NULL);
525
526 if (r->tbl) {
527 (*r->msg)(MANDOCERR_SCOPEEXIT, r->data,
528 r->tbl->line, r->tbl->pos, NULL);
529 tbl_end(r->tbl);
530 r->tbl = NULL;
531 }
532}
533
534
535/*
536 * Parse a roff node's type from the input buffer. This must be in the
537 * form of ".foo xxx" in the usual way.
538 */
539static enum rofft
540roff_parse(struct roff *r, const char *buf, int *pos)
541{
542 const char *mac;
543 size_t maclen;
544 enum rofft t;
545
546 assert(ROFF_CTL(buf[*pos]));
547 (*pos)++;
548
549 while (' ' == buf[*pos] || '\t' == buf[*pos])
550 (*pos)++;
551
552 if ('\0' == buf[*pos])
553 return(ROFF_MAX);
554
555 mac = buf + *pos;
556 maclen = strcspn(mac, " \\\t\0");
557
558 t = (r->current_string = roff_getstrn(r, mac, maclen))
559 ? ROFF_USERDEF : roff_hash_find(mac, maclen);
560
561 *pos += maclen;
562 while (buf[*pos] && ' ' == buf[*pos])
563 (*pos)++;
564
565 return(t);
566}
567
568
569static int
570roff_parse_nat(const char *buf, unsigned int *res)
571{
572 char *ep;
573 long lval;
574
575 errno = 0;
576 lval = strtol(buf, &ep, 10);
577 if (buf[0] == '\0' || *ep != '\0')
578 return(0);
579 if ((errno == ERANGE &&
580 (lval == LONG_MAX || lval == LONG_MIN)) ||
581 (lval > INT_MAX || lval < 0))
582 return(0);
583
584 *res = (unsigned int)lval;
585 return(1);
586}
587
588
589/* ARGSUSED */
590static enum rofferr
591roff_cblock(ROFF_ARGS)
592{
593
594 /*
595 * A block-close `..' should only be invoked as a child of an
596 * ignore macro, otherwise raise a warning and just ignore it.
597 */
598
599 if (NULL == r->last) {
600 (*r->msg)(MANDOCERR_NOSCOPE, r->data, ln, ppos, NULL);
601 return(ROFF_IGN);
602 }
603
604 switch (r->last->tok) {
605 case (ROFF_am):
606 /* FALLTHROUGH */
607 case (ROFF_ami):
608 /* FALLTHROUGH */
609 case (ROFF_am1):
610 /* FALLTHROUGH */
611 case (ROFF_de):
612 /* ROFF_de1 is remapped to ROFF_de in roff_block(). */
613 /* FALLTHROUGH */
614 case (ROFF_dei):
615 /* FALLTHROUGH */
616 case (ROFF_ig):
617 break;
618 default:
619 (*r->msg)(MANDOCERR_NOSCOPE, r->data, ln, ppos, NULL);
620 return(ROFF_IGN);
621 }
622
623 if ((*bufp)[pos])
624 (*r->msg)(MANDOCERR_ARGSLOST, r->data, ln, pos, NULL);
625
626 roffnode_pop(r);
627 roffnode_cleanscope(r);
628 return(ROFF_IGN);
629
630}
631
632
633static void
634roffnode_cleanscope(struct roff *r)
635{
636
637 while (r->last) {
638 if (--r->last->endspan < 0)
639 break;
640 roffnode_pop(r);
641 }
642}
643
644
645/* ARGSUSED */
646static enum rofferr
647roff_ccond(ROFF_ARGS)
648{
649
650 if (NULL == r->last) {
651 (*r->msg)(MANDOCERR_NOSCOPE, r->data, ln, ppos, NULL);
652 return(ROFF_IGN);
653 }
654
655 switch (r->last->tok) {
656 case (ROFF_el):
657 /* FALLTHROUGH */
658 case (ROFF_ie):
659 /* FALLTHROUGH */
660 case (ROFF_if):
661 break;
662 default:
663 (*r->msg)(MANDOCERR_NOSCOPE, r->data, ln, ppos, NULL);
664 return(ROFF_IGN);
665 }
666
667 if (r->last->endspan > -1) {
668 (*r->msg)(MANDOCERR_NOSCOPE, r->data, ln, ppos, NULL);
669 return(ROFF_IGN);
670 }
671
672 if ((*bufp)[pos])
673 (*r->msg)(MANDOCERR_ARGSLOST, r->data, ln, pos, NULL);
674
675 roffnode_pop(r);
676 roffnode_cleanscope(r);
677 return(ROFF_IGN);
678}
679
680
681/* ARGSUSED */
682static enum rofferr
683roff_block(ROFF_ARGS)
684{
685 int sv;
686 size_t sz;
687 char *name;
688
689 name = NULL;
690
691 if (ROFF_ig != tok) {
692 if ('\0' == (*bufp)[pos]) {
693 (*r->msg)(MANDOCERR_NOARGS, r->data, ln, ppos, NULL);
694 return(ROFF_IGN);
695 }
696
697 /*
698 * Re-write `de1', since we don't really care about
699 * groff's strange compatibility mode, into `de'.
700 */
701
702 if (ROFF_de1 == tok)
703 tok = ROFF_de;
704 if (ROFF_de == tok)
705 name = *bufp + pos;
706 else
707 (*r->msg)(MANDOCERR_REQUEST, r->data, ln, ppos,
708 roffs[tok].name);
709
710 while ((*bufp)[pos] && ' ' != (*bufp)[pos])
711 pos++;
712
713 while (' ' == (*bufp)[pos])
714 (*bufp)[pos++] = '\0';
715 }
716
717 roffnode_push(r, tok, name, ln, ppos);
718
719 /*
720 * At the beginning of a `de' macro, clear the existing string
721 * with the same name, if there is one. New content will be
722 * added from roff_block_text() in multiline mode.
723 */
724
725 if (ROFF_de == tok)
726 roff_setstr(r, name, "", 0);
727
728 if ('\0' == (*bufp)[pos])
729 return(ROFF_IGN);
730
731 /* If present, process the custom end-of-line marker. */
732
733 sv = pos;
734 while ((*bufp)[pos] &&
735 ' ' != (*bufp)[pos] &&
736 '\t' != (*bufp)[pos])
737 pos++;
738
739 /*
740 * Note: groff does NOT like escape characters in the input.
741 * Instead of detecting this, we're just going to let it fly and
742 * to hell with it.
743 */
744
745 assert(pos > sv);
746 sz = (size_t)(pos - sv);
747
748 if (1 == sz && '.' == (*bufp)[sv])
749 return(ROFF_IGN);
750
751 r->last->end = mandoc_malloc(sz + 1);
752
753 memcpy(r->last->end, *bufp + sv, sz);
754 r->last->end[(int)sz] = '\0';
755
756 if ((*bufp)[pos])
757 (*r->msg)(MANDOCERR_ARGSLOST, r->data, ln, pos, NULL);
758
759 return(ROFF_IGN);
760}
761
762
763/* ARGSUSED */
764static enum rofferr
765roff_block_sub(ROFF_ARGS)
766{
767 enum rofft t;
768 int i, j;
769
770 /*
771 * First check whether a custom macro exists at this level. If
772 * it does, then check against it. This is some of groff's
773 * stranger behaviours. If we encountered a custom end-scope
774 * tag and that tag also happens to be a "real" macro, then we
775 * need to try interpreting it again as a real macro. If it's
776 * not, then return ignore. Else continue.
777 */
778
779 if (r->last->end) {
780 i = pos + 1;
781 while (' ' == (*bufp)[i] || '\t' == (*bufp)[i])
782 i++;
783
784 for (j = 0; r->last->end[j]; j++, i++)
785 if ((*bufp)[i] != r->last->end[j])
786 break;
787
788 if ('\0' == r->last->end[j] &&
789 ('\0' == (*bufp)[i] ||
790 ' ' == (*bufp)[i] ||
791 '\t' == (*bufp)[i])) {
792 roffnode_pop(r);
793 roffnode_cleanscope(r);
794
795 if (ROFF_MAX != roff_parse(r, *bufp, &pos))
796 return(ROFF_RERUN);
797 return(ROFF_IGN);
798 }
799 }
800
801 /*
802 * If we have no custom end-query or lookup failed, then try
803 * pulling it out of the hashtable.
804 */
805
806 ppos = pos;
807 t = roff_parse(r, *bufp, &pos);
808
809 /*
810 * Macros other than block-end are only significant
811 * in `de' blocks; elsewhere, simply throw them away.
812 */
813 if (ROFF_cblock != t) {
814 if (ROFF_de == tok)
815 roff_setstr(r, r->last->name, *bufp + ppos, 1);
816 return(ROFF_IGN);
817 }
818
819 assert(roffs[t].proc);
820 return((*roffs[t].proc)(r, t, bufp, szp,
821 ln, ppos, pos, offs));
822}
823
824
825/* ARGSUSED */
826static enum rofferr
827roff_block_text(ROFF_ARGS)
828{
829
830 if (ROFF_de == tok)
831 roff_setstr(r, r->last->name, *bufp + pos, 1);
832
833 return(ROFF_IGN);
834}
835
836
837/* ARGSUSED */
838static enum rofferr
839roff_cond_sub(ROFF_ARGS)
840{
841 enum rofft t;
842 enum roffrule rr;
843
844 ppos = pos;
845 rr = r->last->rule;
846
847 /*
848 * Clean out scope. If we've closed ourselves, then don't
849 * continue.
850 */
851
852 roffnode_cleanscope(r);
853
854 if (ROFF_MAX == (t = roff_parse(r, *bufp, &pos))) {
855 if ('\\' == (*bufp)[pos] && '}' == (*bufp)[pos + 1])
856 return(roff_ccond
857 (r, ROFF_ccond, bufp, szp,
858 ln, pos, pos + 2, offs));
859 return(ROFFRULE_DENY == rr ? ROFF_IGN : ROFF_CONT);
860 }
861
862 /*
863 * A denied conditional must evaluate its children if and only
864 * if they're either structurally required (such as loops and
865 * conditionals) or a closing macro.
866 */
867 if (ROFFRULE_DENY == rr)
868 if ( ! (ROFFMAC_STRUCT & roffs[t].flags))
869 if (ROFF_ccond != t)
870 return(ROFF_IGN);
871
872 assert(roffs[t].proc);
873 return((*roffs[t].proc)(r, t, bufp, szp,
874 ln, ppos, pos, offs));
875}
876
877
878/* ARGSUSED */
879static enum rofferr
880roff_cond_text(ROFF_ARGS)
881{
882 char *ep, *st;
883 enum roffrule rr;
884
885 rr = r->last->rule;
886
887 /*
888 * We display the value of the text if out current evaluation
889 * scope permits us to do so.
890 */
891
892 /* FIXME: use roff_ccond? */
893
894 st = &(*bufp)[pos];
895 if (NULL == (ep = strstr(st, "\\}"))) {
896 roffnode_cleanscope(r);
897 return(ROFFRULE_DENY == rr ? ROFF_IGN : ROFF_CONT);
898 }
899
900 if (ep == st || (ep > st && '\\' != *(ep - 1)))
901 roffnode_pop(r);
902
903 roffnode_cleanscope(r);
904 return(ROFFRULE_DENY == rr ? ROFF_IGN : ROFF_CONT);
905}
906
907
908static enum roffrule
909roff_evalcond(const char *v, int *pos)
910{
911
912 switch (v[*pos]) {
913 case ('n'):
914 (*pos)++;
915 return(ROFFRULE_ALLOW);
916 case ('e'):
917 /* FALLTHROUGH */
918 case ('o'):
919 /* FALLTHROUGH */
920 case ('t'):
921 (*pos)++;
922 return(ROFFRULE_DENY);
923 default:
924 break;
925 }
926
927 while (v[*pos] && ' ' != v[*pos])
928 (*pos)++;
929 return(ROFFRULE_DENY);
930}
931
932/* ARGSUSED */
933static enum rofferr
934roff_line_ignore(ROFF_ARGS)
935{
936
937 return(ROFF_IGN);
938}
939
940/* ARGSUSED */
941static enum rofferr
942roff_line_error(ROFF_ARGS)
943{
944
945 (*r->msg)(MANDOCERR_REQUEST, r->data, ln, ppos, roffs[tok].name);
946 return(ROFF_IGN);
947}
948
949/* ARGSUSED */
950static enum rofferr
951roff_cond(ROFF_ARGS)
952{
953 int sv;
954 enum roffrule rule;
955
956 /* Stack overflow! */
957
958 if (ROFF_ie == tok && r->rstackpos == RSTACK_MAX - 1) {
959 (*r->msg)(MANDOCERR_MEM, r->data, ln, ppos, NULL);
960 return(ROFF_ERR);
961 }
962
963 /* First, evaluate the conditional. */
964
965 if (ROFF_el == tok) {
966 /*
967 * An `.el' will get the value of the current rstack
968 * entry set in prior `ie' calls or defaults to DENY.
969 */
970 if (r->rstackpos < 0)
971 rule = ROFFRULE_DENY;
972 else
973 rule = r->rstack[r->rstackpos];
974 } else
975 rule = roff_evalcond(*bufp, &pos);
976
977 sv = pos;
978
979 while (' ' == (*bufp)[pos])
980 pos++;
981
982 /*
983 * Roff is weird. If we have just white-space after the
984 * conditional, it's considered the BODY and we exit without
985 * really doing anything. Warn about this. It's probably
986 * wrong.
987 */
988
989 if ('\0' == (*bufp)[pos] && sv != pos) {
990 (*r->msg)(MANDOCERR_NOARGS, r->data, ln, ppos, NULL);
991 return(ROFF_IGN);
992 }
993
994 roffnode_push(r, tok, NULL, ln, ppos);
995
996 r->last->rule = rule;
997
998 if (ROFF_ie == tok) {
999 /*
1000 * An if-else will put the NEGATION of the current
1001 * evaluated conditional into the stack.
1002 */
1003 r->rstackpos++;
1004 if (ROFFRULE_DENY == r->last->rule)
1005 r->rstack[r->rstackpos] = ROFFRULE_ALLOW;
1006 else
1007 r->rstack[r->rstackpos] = ROFFRULE_DENY;
1008 }
1009
1010 /* If the parent has false as its rule, then so do we. */
1011
1012 if (r->last->parent && ROFFRULE_DENY == r->last->parent->rule)
1013 r->last->rule = ROFFRULE_DENY;
1014
1015 /*
1016 * Determine scope. If we're invoked with "\{" trailing the
1017 * conditional, then we're in a multiline scope. Else our scope
1018 * expires on the next line.
1019 */
1020
1021 r->last->endspan = 1;
1022
1023 if ('\\' == (*bufp)[pos] && '{' == (*bufp)[pos + 1]) {
1024 r->last->endspan = -1;
1025 pos += 2;
1026 }
1027
1028 /*
1029 * If there are no arguments on the line, the next-line scope is
1030 * assumed.
1031 */
1032
1033 if ('\0' == (*bufp)[pos])
1034 return(ROFF_IGN);
1035
1036 /* Otherwise re-run the roff parser after recalculating. */
1037
1038 *offs = pos;
1039 return(ROFF_RERUN);
1040}
1041
1042
1043/* ARGSUSED */
1044static enum rofferr
1045roff_ds(ROFF_ARGS)
1046{
1047 char *name, *string;
1048
1049 /*
1050 * A symbol is named by the first word following the macro
1051 * invocation up to a space. Its value is anything after the
1052 * name's trailing whitespace and optional double-quote. Thus,
1053 *
1054 * [.ds foo "bar " ]
1055 *
1056 * will have `bar " ' as its value.
1057 */
1058
1059 name = *bufp + pos;
1060 if ('\0' == *name)
1061 return(ROFF_IGN);
1062
1063 string = name;
1064 /* Read until end of name. */
1065 while (*string && ' ' != *string)
1066 string++;
1067
1068 /* Nil-terminate name. */
1069 if (*string)
1070 *(string++) = '\0';
1071
1072 /* Read past spaces. */
1073 while (*string && ' ' == *string)
1074 string++;
1075
1076 /* Read passed initial double-quote. */
1077 if (*string && '"' == *string)
1078 string++;
1079
1080 /* The rest is the value. */
1081 roff_setstr(r, name, string, 0);
1082 return(ROFF_IGN);
1083}
1084
1085
1086/* ARGSUSED */
1087static enum rofferr
1088roff_nr(ROFF_ARGS)
1089{
1090 const char *key, *val;
1091 struct reg *rg;
1092
1093 key = &(*bufp)[pos];
1094 rg = r->regs->regs;
1095
1096 /* Parse register request. */
1097 while ((*bufp)[pos] && ' ' != (*bufp)[pos])
1098 pos++;
1099
1100 /*
1101 * Set our nil terminator. Because this line is going to be
1102 * ignored anyway, we can munge it as we please.
1103 */
1104 if ((*bufp)[pos])
1105 (*bufp)[pos++] = '\0';
1106
1107 /* Skip whitespace to register token. */
1108 while ((*bufp)[pos] && ' ' == (*bufp)[pos])
1109 pos++;
1110
1111 val = &(*bufp)[pos];
1112
1113 /* Process register token. */
1114
1115 if (0 == strcmp(key, "nS")) {
1116 rg[(int)REG_nS].set = 1;
1117 if ( ! roff_parse_nat(val, &rg[(int)REG_nS].v.u))
1118 rg[(int)REG_nS].v.u = 0;
1119 }
1120
1121 return(ROFF_IGN);
1122}
1123
1124/* ARGSUSED */
1125static enum rofferr
1126roff_TE(ROFF_ARGS)
1127{
1128
1129 if (NULL == r->tbl)
1130 (*r->msg)(MANDOCERR_NOSCOPE, r->data, ln, ppos, NULL);
1131 else
1132 tbl_end(r->tbl);
1133
1134 r->tbl = NULL;
1135 return(ROFF_IGN);
1136}
1137
1138/* ARGSUSED */
1139static enum rofferr
1140roff_T_(ROFF_ARGS)
1141{
1142
1143 if (NULL == r->tbl)
1144 (*r->msg)(MANDOCERR_NOSCOPE, r->data, ln, ppos, NULL);
1145 else
1146 tbl_restart(ppos, ln, r->tbl);
1147
1148 return(ROFF_IGN);
1149}
1150
1151/* ARGSUSED */
1152static enum rofferr
1153roff_TS(ROFF_ARGS)
1154{
1155 struct tbl_node *t;
1156
1157 if (r->tbl) {
1158 (*r->msg)(MANDOCERR_SCOPEBROKEN, r->data, ln, ppos, NULL);
1159 tbl_end(r->tbl);
1160 }
1161
1162 t = tbl_alloc(ppos, ln, r->data, r->msg);
1163
1164 if (r->last_tbl)
1165 r->last_tbl->next = t;
1166 else
1167 r->first_tbl = r->last_tbl = t;
1168
1169 r->tbl = r->last_tbl = t;
1170 return(ROFF_IGN);
1171}
1172
1173/* ARGSUSED */
1174static enum rofferr
1175roff_so(ROFF_ARGS)
1176{
1177 char *name;
1178
1179 (*r->msg)(MANDOCERR_SO, r->data, ln, ppos, NULL);
1180
1181 /*
1182 * Handle `so'. Be EXTREMELY careful, as we shouldn't be
1183 * opening anything that's not in our cwd or anything beneath
1184 * it. Thus, explicitly disallow traversing up the file-system
1185 * or using absolute paths.
1186 */
1187
1188 name = *bufp + pos;
1189 if ('/' == *name || strstr(name, "../") || strstr(name, "/..")) {
1190 (*r->msg)(MANDOCERR_SOPATH, r->data, ln, pos, NULL);
1191 return(ROFF_ERR);
1192 }
1193
1194 *offs = pos;
1195 return(ROFF_SO);
1196}
1197
1198/* ARGSUSED */
1199static enum rofferr
1200roff_userdef(ROFF_ARGS)
1201{
1202 const char *arg[9];
1203 char *cp, *n1, *n2;
1204 int i;
1205
1206 /*
1207 * Collect pointers to macro argument strings
1208 * and null-terminate them.
1209 */
1210 cp = *bufp + pos;
1211 for (i = 0; i < 9; i++)
1212 arg[i] = '\0' == *cp ? "" :
1213 mandoc_getarg(&cp, r->msg, r->data, ln, &pos);
1214
1215 /*
1216 * Expand macro arguments.
1217 */
1218 *szp = 0;
1219 n1 = cp = mandoc_strdup(r->current_string);
1220 while (NULL != (cp = strstr(cp, "\\$"))) {
1221 i = cp[2] - '1';
1222 if (0 > i || 8 < i) {
1223 /* Not an argument invocation. */
1224 cp += 2;
1225 continue;
1226 }
1227
1228 *szp = strlen(n1) - 3 + strlen(arg[i]) + 1;
1229 n2 = mandoc_malloc(*szp);
1230
1231 strlcpy(n2, n1, (size_t)(cp - n1 + 1));
1232 strlcat(n2, arg[i], *szp);
1233 strlcat(n2, cp + 3, *szp);
1234
1235 cp = n2 + (cp - n1);
1236 free(n1);
1237 n1 = n2;
1238 }
1239
1240 /*
1241 * Replace the macro invocation
1242 * by the expanded macro.
1243 */
1244 free(*bufp);
1245 *bufp = n1;
1246 if (0 == *szp)
1247 *szp = strlen(*bufp) + 1;
1248
1249 return(*szp > 1 && '\n' == (*bufp)[(int)*szp - 2] ?
1250 ROFF_REPARSE : ROFF_APPEND);
1251}
1252
1253/*
1254 * Store *string into the user-defined string called *name.
1255 * In multiline mode, append to an existing entry and append '\n';
1256 * else replace the existing entry, if there is one.
1257 * To clear an existing entry, call with (*r, *name, NULL, 0).
1258 */
1259static void
1260roff_setstr(struct roff *r, const char *name, const char *string,
1261 int multiline)
1262{
1263 struct roffstr *n;
1264 char *c;
1265 size_t oldch, newch;
1266
1267 /* Search for an existing string with the same name. */
1268 n = r->first_string;
1269 while (n && strcmp(name, n->name))
1270 n = n->next;
1271
1272 if (NULL == n) {
1273 /* Create a new string table entry. */
1274 n = mandoc_malloc(sizeof(struct roffstr));
1275 n->name = mandoc_strdup(name);
1276 n->string = NULL;
1277 n->next = r->first_string;
1278 r->first_string = n;
1279 } else if (0 == multiline) {
1280 /* In multiline mode, append; else replace. */
1281 free(n->string);
1282 n->string = NULL;
1283 }
1284
1285 if (NULL == string)
1286 return;
1287
1288 /*
1289 * One additional byte for the '\n' in multiline mode,
1290 * and one for the terminating '\0'.
1291 */
1292 newch = strlen(string) + (multiline ? 2 : 1);
1293 if (NULL == n->string) {
1294 n->string = mandoc_malloc(newch);
1295 *n->string = '\0';
1296 oldch = 0;
1297 } else {
1298 oldch = strlen(n->string);
1299 n->string = mandoc_realloc(n->string, oldch + newch);
1300 }
1301
1302 /* Skip existing content in the destination buffer. */
1303 c = n->string + oldch;
1304
1305 /* Append new content to the destination buffer. */
1306 while (*string) {
1307 /*
1308 * Rudimentary roff copy mode:
1309 * Handle escaped backslashes.
1310 */
1311 if ('\\' == *string && '\\' == *(string + 1))
1312 string++;
1313 *c++ = *string++;
1314 }
1315
1316 /* Append terminating bytes. */
1317 if (multiline)
1318 *c++ = '\n';
1319 *c = '\0';
1320}
1321
1322
1323static const char *
1324roff_getstrn(const struct roff *r, const char *name, size_t len)
1325{
1326 const struct roffstr *n;
1327
1328 n = r->first_string;
1329 while (n && (strncmp(name, n->name, len) || '\0' != n->name[(int)len]))
1330 n = n->next;
1331
1332 return(n ? n->string : NULL);
1333}
1334
1335
1336static void
1337roff_freestr(struct roff *r)
1338{
1339 struct roffstr *n, *nn;
1340
1341 for (n = r->first_string; n; n = nn) {
1342 free(n->name);
1343 free(n->string);
1344 nn = n->next;
1345 free(n);
1346 }
1347
1348 r->first_string = NULL;
1349}
1350
1351const struct tbl_span *
1352roff_span(const struct roff *r)
1353{
1354
1355 return(r->tbl ? tbl_span(r->tbl) : NULL);
1356}