Import mdocml-1.11.3
[dragonfly.git] / contrib / mdocml / roff.c
CommitLineData
a4c7eb57 1/* $Id: roff.c,v 1.142 2011/05/26 11:58:25 kristaps Exp $ */
80387638
SW
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>
80387638 23#include <ctype.h>
80387638
SW
24#include <stdlib.h>
25#include <string.h>
80387638
SW
26
27#include "mandoc.h"
80387638
SW
28#include "libroff.h"
29#include "libmandoc.h"
30
a4c7eb57 31/* Maximum number of nested if-else conditionals. */
80387638
SW
32#define RSTACK_MAX 128
33
80387638
SW
34enum rofft {
35 ROFF_ad,
36 ROFF_am,
37 ROFF_ami,
38 ROFF_am1,
39 ROFF_de,
40 ROFF_dei,
41 ROFF_de1,
42 ROFF_ds,
43 ROFF_el,
44 ROFF_hy,
45 ROFF_ie,
46 ROFF_if,
47 ROFF_ig,
60e1e752 48 ROFF_it,
80387638
SW
49 ROFF_ne,
50 ROFF_nh,
51 ROFF_nr,
60e1e752
SW
52 ROFF_ns,
53 ROFF_ps,
80387638
SW
54 ROFF_rm,
55 ROFF_so,
60e1e752 56 ROFF_ta,
80387638
SW
57 ROFF_tr,
58 ROFF_TS,
59 ROFF_TE,
60 ROFF_T_,
60e1e752
SW
61 ROFF_EQ,
62 ROFF_EN,
80387638 63 ROFF_cblock,
a4c7eb57 64 ROFF_ccond,
80387638
SW
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 {
60e1e752 81 struct mparse *parse; /* parse point */
80387638 82 struct roffnode *last; /* leaf of stack */
80387638
SW
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_node *first_tbl; /* first table parsed */
89 struct tbl_node *last_tbl; /* last table parsed */
90 struct tbl_node *tbl; /* current table being parsed */
60e1e752
SW
91 struct eqn_node *last_eqn; /* last equation parsed */
92 struct eqn_node *first_eqn; /* first equation parsed */
93 struct eqn_node *eqn; /* current equation being parsed */
80387638
SW
94};
95
96struct roffnode {
97 enum rofft tok; /* type of node */
98 struct roffnode *parent; /* up one in stack */
99 int line; /* parse line */
100 int col; /* parse col */
101 char *name; /* node name, e.g. macro name */
102 char *end; /* end-rules: custom token */
103 int endspan; /* end-rules: next-line or infty */
104 enum roffrule rule; /* current evaluation rule */
105};
106
107#define ROFF_ARGS struct roff *r, /* parse ctx */ \
108 enum rofft tok, /* tok of macro */ \
109 char **bufp, /* input buffer */ \
110 size_t *szp, /* size of input buffer */ \
111 int ln, /* parse line */ \
112 int ppos, /* original pos in buffer */ \
113 int pos, /* current pos in buffer */ \
114 int *offs /* reset offset of buffer data */
115
116typedef enum rofferr (*roffproc)(ROFF_ARGS);
117
118struct roffmac {
119 const char *name; /* macro name */
120 roffproc proc; /* process new macro */
121 roffproc text; /* process as child text of macro */
122 roffproc sub; /* process as child of macro */
123 int flags;
124#define ROFFMAC_STRUCT (1 << 0) /* always interpret */
125 struct roffmac *next;
126};
127
a4c7eb57
SW
128struct predef {
129 const char *name; /* predefined input name */
130 const char *str; /* replacement symbol */
131};
132
133#define PREDEF(__name, __str) \
134 { (__name), (__str) },
135
80387638
SW
136static enum rofferr roff_block(ROFF_ARGS);
137static enum rofferr roff_block_text(ROFF_ARGS);
138static enum rofferr roff_block_sub(ROFF_ARGS);
139static enum rofferr roff_cblock(ROFF_ARGS);
140static enum rofferr roff_ccond(ROFF_ARGS);
141static enum rofferr roff_cond(ROFF_ARGS);
142static enum rofferr roff_cond_text(ROFF_ARGS);
143static enum rofferr roff_cond_sub(ROFF_ARGS);
144static enum rofferr roff_ds(ROFF_ARGS);
145static enum roffrule roff_evalcond(const char *, int *);
146static void roff_freestr(struct roff *);
60e1e752 147static char *roff_getname(struct roff *, char **, int, int);
80387638
SW
148static const char *roff_getstrn(const struct roff *,
149 const char *, size_t);
150static enum rofferr roff_line_ignore(ROFF_ARGS);
80387638
SW
151static enum rofferr roff_nr(ROFF_ARGS);
152static int roff_res(struct roff *,
a4c7eb57 153 char **, size_t *, int, int);
60e1e752 154static enum rofferr roff_rm(ROFF_ARGS);
80387638
SW
155static void roff_setstr(struct roff *,
156 const char *, const char *, int);
157static enum rofferr roff_so(ROFF_ARGS);
158static enum rofferr roff_TE(ROFF_ARGS);
159static enum rofferr roff_TS(ROFF_ARGS);
60e1e752
SW
160static enum rofferr roff_EQ(ROFF_ARGS);
161static enum rofferr roff_EN(ROFF_ARGS);
80387638
SW
162static enum rofferr roff_T_(ROFF_ARGS);
163static enum rofferr roff_userdef(ROFF_ARGS);
164
165/* See roff_hash_find() */
166
167#define ASCII_HI 126
168#define ASCII_LO 33
169#define HASHWIDTH (ASCII_HI - ASCII_LO + 1)
170
171static struct roffmac *hash[HASHWIDTH];
172
173static struct roffmac roffs[ROFF_MAX] = {
174 { "ad", roff_line_ignore, NULL, NULL, 0, NULL },
175 { "am", roff_block, roff_block_text, roff_block_sub, 0, NULL },
176 { "ami", roff_block, roff_block_text, roff_block_sub, 0, NULL },
177 { "am1", roff_block, roff_block_text, roff_block_sub, 0, NULL },
178 { "de", roff_block, roff_block_text, roff_block_sub, 0, NULL },
179 { "dei", roff_block, roff_block_text, roff_block_sub, 0, NULL },
180 { "de1", roff_block, roff_block_text, roff_block_sub, 0, NULL },
181 { "ds", roff_ds, NULL, NULL, 0, NULL },
182 { "el", roff_cond, roff_cond_text, roff_cond_sub, ROFFMAC_STRUCT, NULL },
183 { "hy", roff_line_ignore, NULL, NULL, 0, NULL },
184 { "ie", roff_cond, roff_cond_text, roff_cond_sub, ROFFMAC_STRUCT, NULL },
185 { "if", roff_cond, roff_cond_text, roff_cond_sub, ROFFMAC_STRUCT, NULL },
186 { "ig", roff_block, roff_block_text, roff_block_sub, 0, NULL },
60e1e752 187 { "it", roff_line_ignore, NULL, NULL, 0, NULL },
80387638
SW
188 { "ne", roff_line_ignore, NULL, NULL, 0, NULL },
189 { "nh", roff_line_ignore, NULL, NULL, 0, NULL },
190 { "nr", roff_nr, NULL, NULL, 0, NULL },
60e1e752
SW
191 { "ns", roff_line_ignore, NULL, NULL, 0, NULL },
192 { "ps", roff_line_ignore, NULL, NULL, 0, NULL },
193 { "rm", roff_rm, NULL, NULL, 0, NULL },
80387638 194 { "so", roff_so, NULL, NULL, 0, NULL },
60e1e752 195 { "ta", roff_line_ignore, NULL, NULL, 0, NULL },
80387638
SW
196 { "tr", roff_line_ignore, NULL, NULL, 0, NULL },
197 { "TS", roff_TS, NULL, NULL, 0, NULL },
198 { "TE", roff_TE, NULL, NULL, 0, NULL },
199 { "T&", roff_T_, NULL, NULL, 0, NULL },
60e1e752
SW
200 { "EQ", roff_EQ, NULL, NULL, 0, NULL },
201 { "EN", roff_EN, NULL, NULL, 0, NULL },
80387638
SW
202 { ".", roff_cblock, NULL, NULL, 0, NULL },
203 { "\\}", roff_ccond, NULL, NULL, 0, NULL },
204 { NULL, roff_userdef, NULL, NULL, 0, NULL },
205};
206
a4c7eb57
SW
207/* Array of injected predefined strings. */
208#define PREDEFS_MAX 38
209static const struct predef predefs[PREDEFS_MAX] = {
210#include "predefs.in"
211};
212
80387638
SW
213static void roff_free1(struct roff *);
214static enum rofft roff_hash_find(const char *, size_t);
215static void roff_hash_init(void);
216static void roffnode_cleanscope(struct roff *);
217static void roffnode_push(struct roff *, enum rofft,
218 const char *, int, int);
219static void roffnode_pop(struct roff *);
220static enum rofft roff_parse(struct roff *, const char *, int *);
80387638
SW
221
222/* See roff_hash_find() */
223#define ROFF_HASH(p) (p[0] - ASCII_LO)
224
225static void
226roff_hash_init(void)
227{
228 struct roffmac *n;
229 int buc, i;
230
231 for (i = 0; i < (int)ROFF_USERDEF; i++) {
232 assert(roffs[i].name[0] >= ASCII_LO);
233 assert(roffs[i].name[0] <= ASCII_HI);
234
235 buc = ROFF_HASH(roffs[i].name);
236
237 if (NULL != (n = hash[buc])) {
238 for ( ; n->next; n = n->next)
239 /* Do nothing. */ ;
240 n->next = &roffs[i];
241 } else
242 hash[buc] = &roffs[i];
243 }
244}
245
80387638
SW
246/*
247 * Look up a roff token by its name. Returns ROFF_MAX if no macro by
248 * the nil-terminated string name could be found.
249 */
250static enum rofft
251roff_hash_find(const char *p, size_t s)
252{
253 int buc;
254 struct roffmac *n;
255
256 /*
257 * libroff has an extremely simple hashtable, for the time
258 * being, which simply keys on the first character, which must
259 * be printable, then walks a chain. It works well enough until
260 * optimised.
261 */
262
263 if (p[0] < ASCII_LO || p[0] > ASCII_HI)
264 return(ROFF_MAX);
265
266 buc = ROFF_HASH(p);
267
268 if (NULL == (n = hash[buc]))
269 return(ROFF_MAX);
270 for ( ; n; n = n->next)
271 if (0 == strncmp(n->name, p, s) && '\0' == n->name[(int)s])
272 return((enum rofft)(n - roffs));
273
274 return(ROFF_MAX);
275}
276
277
278/*
279 * Pop the current node off of the stack of roff instructions currently
280 * pending.
281 */
282static void
283roffnode_pop(struct roff *r)
284{
285 struct roffnode *p;
286
287 assert(r->last);
288 p = r->last;
289
80387638
SW
290 r->last = r->last->parent;
291 free(p->name);
292 free(p->end);
293 free(p);
294}
295
296
297/*
298 * Push a roff node onto the instruction stack. This must later be
299 * removed with roffnode_pop().
300 */
301static void
302roffnode_push(struct roff *r, enum rofft tok, const char *name,
303 int line, int col)
304{
305 struct roffnode *p;
306
307 p = mandoc_calloc(1, sizeof(struct roffnode));
308 p->tok = tok;
309 if (name)
310 p->name = mandoc_strdup(name);
311 p->parent = r->last;
312 p->line = line;
313 p->col = col;
314 p->rule = p->parent ? p->parent->rule : ROFFRULE_DENY;
315
316 r->last = p;
317}
318
319
320static void
321roff_free1(struct roff *r)
322{
323 struct tbl_node *t;
60e1e752 324 struct eqn_node *e;
80387638 325
60e1e752 326 while (NULL != (t = r->first_tbl)) {
80387638
SW
327 r->first_tbl = t->next;
328 tbl_free(t);
329 }
330
331 r->first_tbl = r->last_tbl = r->tbl = NULL;
332
60e1e752
SW
333 while (NULL != (e = r->first_eqn)) {
334 r->first_eqn = e->next;
335 eqn_free(e);
336 }
337
338 r->first_eqn = r->last_eqn = r->eqn = NULL;
339
80387638
SW
340 while (r->last)
341 roffnode_pop(r);
342
343 roff_freestr(r);
344}
345
346
347void
348roff_reset(struct roff *r)
349{
350
351 roff_free1(r);
352}
353
354
355void
356roff_free(struct roff *r)
357{
358
359 roff_free1(r);
360 free(r);
361}
362
363
364struct roff *
60e1e752 365roff_alloc(struct regset *regs, struct mparse *parse)
80387638
SW
366{
367 struct roff *r;
a4c7eb57 368 int i;
80387638
SW
369
370 r = mandoc_calloc(1, sizeof(struct roff));
371 r->regs = regs;
60e1e752 372 r->parse = parse;
80387638
SW
373 r->rstackpos = -1;
374
375 roff_hash_init();
a4c7eb57
SW
376
377 for (i = 0; i < PREDEFS_MAX; i++)
378 roff_setstr(r, predefs[i].name, predefs[i].str, 0);
379
80387638
SW
380 return(r);
381}
382
383
384/*
385 * Pre-filter each and every line for reserved words (one beginning with
386 * `\*', e.g., `\*(ab'). These must be handled before the actual line
387 * is processed.
388 */
389static int
a4c7eb57 390roff_res(struct roff *r, char **bufp, size_t *szp, int ln, int pos)
80387638
SW
391{
392 const char *stesc; /* start of an escape sequence ('\\') */
393 const char *stnam; /* start of the name, after "[(*" */
394 const char *cp; /* end of the name, e.g. before ']' */
395 const char *res; /* the string to be substituted */
396 int i, maxl;
397 size_t nsz;
398 char *n;
399
400 /* Search for a leading backslash and save a pointer to it. */
401
402 cp = *bufp + pos;
403 while (NULL != (cp = strchr(cp, '\\'))) {
404 stesc = cp++;
405
406 /*
407 * The second character must be an asterisk.
408 * If it isn't, skip it anyway: It is escaped,
409 * so it can't start another escape sequence.
410 */
411
412 if ('\0' == *cp)
413 return(1);
414 if ('*' != *cp++)
415 continue;
416
417 /*
418 * The third character decides the length
419 * of the name of the string.
420 * Save a pointer to the name.
421 */
422
423 switch (*cp) {
424 case ('\0'):
425 return(1);
426 case ('('):
427 cp++;
428 maxl = 2;
429 break;
430 case ('['):
431 cp++;
432 maxl = 0;
433 break;
434 default:
435 maxl = 1;
436 break;
437 }
438 stnam = cp;
439
440 /* Advance to the end of the name. */
441
442 for (i = 0; 0 == maxl || i < maxl; i++, cp++) {
443 if ('\0' == *cp)
444 return(1); /* Error. */
445 if (0 == maxl && ']' == *cp)
446 break;
447 }
448
449 /*
450 * Retrieve the replacement string; if it is
451 * undefined, resume searching for escapes.
452 */
453
454 res = roff_getstrn(r, stnam, (size_t)i);
455
456 if (NULL == res) {
a4c7eb57
SW
457 /* TODO: keep track of the correct position. */
458 mandoc_msg(MANDOCERR_BADESCAPE, r->parse, ln, pos, NULL);
459 res = "";
80387638
SW
460 }
461
462 /* Replace the escape sequence by the string. */
463
464 nsz = *szp + strlen(res) + 1;
465 n = mandoc_malloc(nsz);
466
467 strlcpy(n, *bufp, (size_t)(stesc - *bufp + 1));
468 strlcat(n, res, nsz);
469 strlcat(n, cp + (maxl ? 0 : 1), nsz);
470
471 free(*bufp);
472
473 *bufp = n;
474 *szp = nsz;
475 return(0);
476 }
477
478 return(1);
479}
480
481
482enum rofferr
483roff_parseln(struct roff *r, int ln, char **bufp,
484 size_t *szp, int pos, int *offs)
485{
486 enum rofft t;
487 enum rofferr e;
60e1e752 488 int ppos, ctl;
80387638
SW
489
490 /*
491 * Run the reserved-word filter only if we have some reserved
492 * words to fill in.
493 */
494
a4c7eb57 495 if (r->first_string && ! roff_res(r, bufp, szp, ln, pos))
80387638
SW
496 return(ROFF_REPARSE);
497
60e1e752
SW
498 ppos = pos;
499 ctl = mandoc_getcontrol(*bufp, &pos);
500
80387638
SW
501 /*
502 * First, if a scope is open and we're not a macro, pass the
503 * text through the macro's filter. If a scope isn't open and
504 * we're not a macro, just let it through.
60e1e752
SW
505 * Finally, if there's an equation scope open, divert it into it
506 * no matter our state.
80387638
SW
507 */
508
60e1e752 509 if (r->last && ! ctl) {
80387638
SW
510 t = r->last->tok;
511 assert(roffs[t].text);
512 e = (*roffs[t].text)
513 (r, t, bufp, szp, ln, pos, pos, offs);
514 assert(ROFF_IGN == e || ROFF_CONT == e);
60e1e752
SW
515 if (ROFF_CONT != e)
516 return(e);
517 if (r->eqn)
518 return(eqn_read(&r->eqn, ln, *bufp, pos));
80387638 519 if (r->tbl)
60e1e752 520 return(tbl_read(r->tbl, ln, *bufp, pos));
80387638 521 return(ROFF_CONT);
60e1e752
SW
522 } else if ( ! ctl) {
523 if (r->eqn)
524 return(eqn_read(&r->eqn, ln, *bufp, pos));
525 if (r->tbl)
526 return(tbl_read(r->tbl, ln, *bufp, pos));
527 return(ROFF_CONT);
528 } else if (r->eqn)
529 return(eqn_read(&r->eqn, ln, *bufp, ppos));
80387638
SW
530
531 /*
532 * If a scope is open, go to the child handler for that macro,
533 * as it may want to preprocess before doing anything with it.
60e1e752 534 * Don't do so if an equation is open.
80387638
SW
535 */
536
537 if (r->last) {
538 t = r->last->tok;
539 assert(roffs[t].sub);
540 return((*roffs[t].sub)
541 (r, t, bufp, szp,
60e1e752 542 ln, ppos, pos, offs));
80387638
SW
543 }
544
545 /*
546 * Lastly, as we've no scope open, try to look up and execute
547 * the new macro. If no macro is found, simply return and let
548 * the compilers handle it.
549 */
550
80387638
SW
551 if (ROFF_MAX == (t = roff_parse(r, *bufp, &pos)))
552 return(ROFF_CONT);
553
554 assert(roffs[t].proc);
555 return((*roffs[t].proc)
556 (r, t, bufp, szp,
557 ln, ppos, pos, offs));
558}
559
560
561void
562roff_endparse(struct roff *r)
563{
564
565 if (r->last)
60e1e752 566 mandoc_msg(MANDOCERR_SCOPEEXIT, r->parse,
80387638
SW
567 r->last->line, r->last->col, NULL);
568
60e1e752
SW
569 if (r->eqn) {
570 mandoc_msg(MANDOCERR_SCOPEEXIT, r->parse,
571 r->eqn->eqn.line, r->eqn->eqn.pos, NULL);
572 eqn_end(r->eqn);
573 r->eqn = NULL;
574 }
575
80387638 576 if (r->tbl) {
60e1e752 577 mandoc_msg(MANDOCERR_SCOPEEXIT, r->parse,
80387638
SW
578 r->tbl->line, r->tbl->pos, NULL);
579 tbl_end(r->tbl);
580 r->tbl = NULL;
581 }
582}
583
80387638
SW
584/*
585 * Parse a roff node's type from the input buffer. This must be in the
586 * form of ".foo xxx" in the usual way.
587 */
588static enum rofft
589roff_parse(struct roff *r, const char *buf, int *pos)
590{
591 const char *mac;
592 size_t maclen;
593 enum rofft t;
594
60e1e752 595 if ('\0' == buf[*pos] || '"' == buf[*pos])
80387638
SW
596 return(ROFF_MAX);
597
598 mac = buf + *pos;
599 maclen = strcspn(mac, " \\\t\0");
600
601 t = (r->current_string = roff_getstrn(r, mac, maclen))
602 ? ROFF_USERDEF : roff_hash_find(mac, maclen);
603
60e1e752
SW
604 *pos += (int)maclen;
605
80387638
SW
606 while (buf[*pos] && ' ' == buf[*pos])
607 (*pos)++;
608
609 return(t);
610}
611
80387638
SW
612/* ARGSUSED */
613static enum rofferr
614roff_cblock(ROFF_ARGS)
615{
616
617 /*
618 * A block-close `..' should only be invoked as a child of an
619 * ignore macro, otherwise raise a warning and just ignore it.
620 */
621
622 if (NULL == r->last) {
60e1e752 623 mandoc_msg(MANDOCERR_NOSCOPE, r->parse, ln, ppos, NULL);
80387638
SW
624 return(ROFF_IGN);
625 }
626
627 switch (r->last->tok) {
628 case (ROFF_am):
629 /* FALLTHROUGH */
630 case (ROFF_ami):
631 /* FALLTHROUGH */
632 case (ROFF_am1):
633 /* FALLTHROUGH */
634 case (ROFF_de):
635 /* ROFF_de1 is remapped to ROFF_de in roff_block(). */
636 /* FALLTHROUGH */
637 case (ROFF_dei):
638 /* FALLTHROUGH */
639 case (ROFF_ig):
640 break;
641 default:
60e1e752 642 mandoc_msg(MANDOCERR_NOSCOPE, r->parse, ln, ppos, NULL);
80387638
SW
643 return(ROFF_IGN);
644 }
645
646 if ((*bufp)[pos])
60e1e752 647 mandoc_msg(MANDOCERR_ARGSLOST, r->parse, ln, pos, NULL);
80387638
SW
648
649 roffnode_pop(r);
650 roffnode_cleanscope(r);
651 return(ROFF_IGN);
652
653}
654
655
656static void
657roffnode_cleanscope(struct roff *r)
658{
659
660 while (r->last) {
661 if (--r->last->endspan < 0)
662 break;
663 roffnode_pop(r);
664 }
665}
666
667
668/* ARGSUSED */
669static enum rofferr
670roff_ccond(ROFF_ARGS)
671{
672
673 if (NULL == r->last) {
60e1e752 674 mandoc_msg(MANDOCERR_NOSCOPE, r->parse, ln, ppos, NULL);
80387638
SW
675 return(ROFF_IGN);
676 }
677
678 switch (r->last->tok) {
679 case (ROFF_el):
680 /* FALLTHROUGH */
681 case (ROFF_ie):
682 /* FALLTHROUGH */
683 case (ROFF_if):
684 break;
685 default:
60e1e752 686 mandoc_msg(MANDOCERR_NOSCOPE, r->parse, ln, ppos, NULL);
80387638
SW
687 return(ROFF_IGN);
688 }
689
690 if (r->last->endspan > -1) {
60e1e752 691 mandoc_msg(MANDOCERR_NOSCOPE, r->parse, ln, ppos, NULL);
80387638
SW
692 return(ROFF_IGN);
693 }
694
695 if ((*bufp)[pos])
60e1e752 696 mandoc_msg(MANDOCERR_ARGSLOST, r->parse, ln, pos, NULL);
80387638
SW
697
698 roffnode_pop(r);
699 roffnode_cleanscope(r);
700 return(ROFF_IGN);
701}
702
703
704/* ARGSUSED */
705static enum rofferr
706roff_block(ROFF_ARGS)
707{
708 int sv;
709 size_t sz;
710 char *name;
711
712 name = NULL;
713
714 if (ROFF_ig != tok) {
715 if ('\0' == (*bufp)[pos]) {
60e1e752 716 mandoc_msg(MANDOCERR_NOARGS, r->parse, ln, ppos, NULL);
80387638
SW
717 return(ROFF_IGN);
718 }
719
720 /*
721 * Re-write `de1', since we don't really care about
722 * groff's strange compatibility mode, into `de'.
723 */
724
725 if (ROFF_de1 == tok)
726 tok = ROFF_de;
727 if (ROFF_de == tok)
728 name = *bufp + pos;
729 else
60e1e752 730 mandoc_msg(MANDOCERR_REQUEST, r->parse, ln, ppos,
80387638
SW
731 roffs[tok].name);
732
a4c7eb57 733 while ((*bufp)[pos] && ! isspace((unsigned char)(*bufp)[pos]))
80387638
SW
734 pos++;
735
a4c7eb57 736 while (isspace((unsigned char)(*bufp)[pos]))
80387638
SW
737 (*bufp)[pos++] = '\0';
738 }
739
740 roffnode_push(r, tok, name, ln, ppos);
741
742 /*
743 * At the beginning of a `de' macro, clear the existing string
744 * with the same name, if there is one. New content will be
745 * added from roff_block_text() in multiline mode.
746 */
747
748 if (ROFF_de == tok)
749 roff_setstr(r, name, "", 0);
750
751 if ('\0' == (*bufp)[pos])
752 return(ROFF_IGN);
753
754 /* If present, process the custom end-of-line marker. */
755
756 sv = pos;
a4c7eb57 757 while ((*bufp)[pos] && ! isspace((unsigned char)(*bufp)[pos]))
80387638
SW
758 pos++;
759
760 /*
761 * Note: groff does NOT like escape characters in the input.
762 * Instead of detecting this, we're just going to let it fly and
763 * to hell with it.
764 */
765
766 assert(pos > sv);
767 sz = (size_t)(pos - sv);
768
769 if (1 == sz && '.' == (*bufp)[sv])
770 return(ROFF_IGN);
771
772 r->last->end = mandoc_malloc(sz + 1);
773
774 memcpy(r->last->end, *bufp + sv, sz);
775 r->last->end[(int)sz] = '\0';
776
777 if ((*bufp)[pos])
60e1e752 778 mandoc_msg(MANDOCERR_ARGSLOST, r->parse, ln, pos, NULL);
80387638
SW
779
780 return(ROFF_IGN);
781}
782
783
784/* ARGSUSED */
785static enum rofferr
786roff_block_sub(ROFF_ARGS)
787{
788 enum rofft t;
789 int i, j;
790
791 /*
792 * First check whether a custom macro exists at this level. If
793 * it does, then check against it. This is some of groff's
794 * stranger behaviours. If we encountered a custom end-scope
795 * tag and that tag also happens to be a "real" macro, then we
796 * need to try interpreting it again as a real macro. If it's
797 * not, then return ignore. Else continue.
798 */
799
800 if (r->last->end) {
60e1e752 801 for (i = pos, j = 0; r->last->end[j]; j++, i++)
80387638
SW
802 if ((*bufp)[i] != r->last->end[j])
803 break;
804
805 if ('\0' == r->last->end[j] &&
806 ('\0' == (*bufp)[i] ||
807 ' ' == (*bufp)[i] ||
808 '\t' == (*bufp)[i])) {
809 roffnode_pop(r);
810 roffnode_cleanscope(r);
811
60e1e752
SW
812 while (' ' == (*bufp)[i] || '\t' == (*bufp)[i])
813 i++;
814
815 pos = i;
80387638
SW
816 if (ROFF_MAX != roff_parse(r, *bufp, &pos))
817 return(ROFF_RERUN);
818 return(ROFF_IGN);
819 }
820 }
821
822 /*
823 * If we have no custom end-query or lookup failed, then try
824 * pulling it out of the hashtable.
825 */
826
a4c7eb57 827 t = roff_parse(r, *bufp, &pos);
80387638
SW
828
829 /*
830 * Macros other than block-end are only significant
831 * in `de' blocks; elsewhere, simply throw them away.
832 */
833 if (ROFF_cblock != t) {
834 if (ROFF_de == tok)
835 roff_setstr(r, r->last->name, *bufp + ppos, 1);
836 return(ROFF_IGN);
837 }
838
839 assert(roffs[t].proc);
840 return((*roffs[t].proc)(r, t, bufp, szp,
841 ln, ppos, pos, offs));
842}
843
844
845/* ARGSUSED */
846static enum rofferr
847roff_block_text(ROFF_ARGS)
848{
849
850 if (ROFF_de == tok)
851 roff_setstr(r, r->last->name, *bufp + pos, 1);
852
853 return(ROFF_IGN);
854}
855
856
857/* ARGSUSED */
858static enum rofferr
859roff_cond_sub(ROFF_ARGS)
860{
861 enum rofft t;
862 enum roffrule rr;
a4c7eb57 863 char *ep;
80387638 864
80387638 865 rr = r->last->rule;
a4c7eb57 866 roffnode_cleanscope(r);
80387638 867
a4c7eb57
SW
868 /*
869 * If the macro is unknown, first check if it contains a closing
870 * delimiter `\}'. If it does, close out our scope and return
871 * the currently-scoped rule (ignore or continue). Else, drop
872 * into the currently-scoped rule.
80387638
SW
873 */
874
80387638 875 if (ROFF_MAX == (t = roff_parse(r, *bufp, &pos))) {
a4c7eb57
SW
876 ep = &(*bufp)[pos];
877 for ( ; NULL != (ep = strchr(ep, '\\')); ep++) {
878 ep++;
879 if ('}' != *ep)
880 continue;
881 *ep = '&';
882 roff_ccond(r, ROFF_ccond, bufp, szp,
883 ln, pos, pos + 2, offs);
884 break;
885 }
80387638
SW
886 return(ROFFRULE_DENY == rr ? ROFF_IGN : ROFF_CONT);
887 }
888
889 /*
890 * A denied conditional must evaluate its children if and only
891 * if they're either structurally required (such as loops and
892 * conditionals) or a closing macro.
893 */
a4c7eb57 894
80387638
SW
895 if (ROFFRULE_DENY == rr)
896 if ( ! (ROFFMAC_STRUCT & roffs[t].flags))
897 if (ROFF_ccond != t)
898 return(ROFF_IGN);
899
900 assert(roffs[t].proc);
901 return((*roffs[t].proc)(r, t, bufp, szp,
902 ln, ppos, pos, offs));
903}
904
80387638
SW
905/* ARGSUSED */
906static enum rofferr
907roff_cond_text(ROFF_ARGS)
908{
a4c7eb57 909 char *ep;
80387638
SW
910 enum roffrule rr;
911
912 rr = r->last->rule;
a4c7eb57 913 roffnode_cleanscope(r);
80387638 914
a4c7eb57
SW
915 ep = &(*bufp)[pos];
916 for ( ; NULL != (ep = strchr(ep, '\\')); ep++) {
917 ep++;
918 if ('}' != *ep)
919 continue;
920 *ep = '&';
921 roff_ccond(r, ROFF_ccond, bufp, szp,
922 ln, pos, pos + 2, offs);
80387638 923 }
80387638
SW
924 return(ROFFRULE_DENY == rr ? ROFF_IGN : ROFF_CONT);
925}
926
80387638
SW
927static enum roffrule
928roff_evalcond(const char *v, int *pos)
929{
930
931 switch (v[*pos]) {
932 case ('n'):
933 (*pos)++;
934 return(ROFFRULE_ALLOW);
935 case ('e'):
936 /* FALLTHROUGH */
937 case ('o'):
938 /* FALLTHROUGH */
939 case ('t'):
940 (*pos)++;
941 return(ROFFRULE_DENY);
942 default:
943 break;
944 }
945
946 while (v[*pos] && ' ' != v[*pos])
947 (*pos)++;
948 return(ROFFRULE_DENY);
949}
950
951/* ARGSUSED */
952static enum rofferr
953roff_line_ignore(ROFF_ARGS)
954{
955
60e1e752
SW
956 if (ROFF_it == tok)
957 mandoc_msg(MANDOCERR_REQUEST, r->parse, ln, ppos, "it");
80387638 958
80387638
SW
959 return(ROFF_IGN);
960}
961
962/* ARGSUSED */
963static enum rofferr
964roff_cond(ROFF_ARGS)
965{
966 int sv;
967 enum roffrule rule;
968
a4c7eb57
SW
969 /*
970 * An `.el' has no conditional body: it will consume the value
971 * of the current rstack entry set in prior `ie' calls or
972 * defaults to DENY.
973 *
974 * If we're not an `el', however, then evaluate the conditional.
975 */
80387638 976
a4c7eb57
SW
977 rule = ROFF_el == tok ?
978 (r->rstackpos < 0 ?
979 ROFFRULE_DENY : r->rstack[r->rstackpos--]) :
980 roff_evalcond(*bufp, &pos);
80387638
SW
981
982 sv = pos;
80387638
SW
983 while (' ' == (*bufp)[pos])
984 pos++;
985
986 /*
987 * Roff is weird. If we have just white-space after the
988 * conditional, it's considered the BODY and we exit without
989 * really doing anything. Warn about this. It's probably
990 * wrong.
991 */
992
993 if ('\0' == (*bufp)[pos] && sv != pos) {
60e1e752 994 mandoc_msg(MANDOCERR_NOARGS, r->parse, ln, ppos, NULL);
80387638
SW
995 return(ROFF_IGN);
996 }
997
998 roffnode_push(r, tok, NULL, ln, ppos);
999
1000 r->last->rule = rule;
1001
a4c7eb57
SW
1002 /*
1003 * An if-else will put the NEGATION of the current evaluated
1004 * conditional into the stack of rules.
1005 */
1006
80387638 1007 if (ROFF_ie == tok) {
a4c7eb57
SW
1008 if (r->rstackpos == RSTACK_MAX - 1) {
1009 mandoc_msg(MANDOCERR_MEM,
1010 r->parse, ln, ppos, NULL);
1011 return(ROFF_ERR);
1012 }
1013 r->rstack[++r->rstackpos] =
1014 ROFFRULE_DENY == r->last->rule ?
1015 ROFFRULE_ALLOW : ROFFRULE_DENY;
80387638
SW
1016 }
1017
1018 /* If the parent has false as its rule, then so do we. */
1019
1020 if (r->last->parent && ROFFRULE_DENY == r->last->parent->rule)
1021 r->last->rule = ROFFRULE_DENY;
1022
1023 /*
1024 * Determine scope. If we're invoked with "\{" trailing the
1025 * conditional, then we're in a multiline scope. Else our scope
1026 * expires on the next line.
1027 */
1028
1029 r->last->endspan = 1;
1030
1031 if ('\\' == (*bufp)[pos] && '{' == (*bufp)[pos + 1]) {
1032 r->last->endspan = -1;
1033 pos += 2;
1034 }
1035
1036 /*
1037 * If there are no arguments on the line, the next-line scope is
1038 * assumed.
1039 */
1040
1041 if ('\0' == (*bufp)[pos])
1042 return(ROFF_IGN);
1043
1044 /* Otherwise re-run the roff parser after recalculating. */
1045
1046 *offs = pos;
1047 return(ROFF_RERUN);
1048}
1049
1050
1051/* ARGSUSED */
1052static enum rofferr
1053roff_ds(ROFF_ARGS)
1054{
1055 char *name, *string;
1056
1057 /*
1058 * A symbol is named by the first word following the macro
1059 * invocation up to a space. Its value is anything after the
1060 * name's trailing whitespace and optional double-quote. Thus,
1061 *
1062 * [.ds foo "bar " ]
1063 *
1064 * will have `bar " ' as its value.
1065 */
1066
60e1e752
SW
1067 string = *bufp + pos;
1068 name = roff_getname(r, &string, ln, pos);
80387638
SW
1069 if ('\0' == *name)
1070 return(ROFF_IGN);
1071
60e1e752
SW
1072 /* Read past initial double-quote. */
1073 if ('"' == *string)
80387638
SW
1074 string++;
1075
1076 /* The rest is the value. */
1077 roff_setstr(r, name, string, 0);
1078 return(ROFF_IGN);
1079}
1080
1081
1082/* ARGSUSED */
1083static enum rofferr
1084roff_nr(ROFF_ARGS)
1085{
60e1e752
SW
1086 const char *key;
1087 char *val;
a4c7eb57 1088 int iv;
80387638
SW
1089 struct reg *rg;
1090
60e1e752
SW
1091 val = *bufp + pos;
1092 key = roff_getname(r, &val, ln, pos);
80387638
SW
1093 rg = r->regs->regs;
1094
80387638
SW
1095 if (0 == strcmp(key, "nS")) {
1096 rg[(int)REG_nS].set = 1;
a4c7eb57
SW
1097 if ((iv = mandoc_strntou(val, strlen(val), 10)) >= 0)
1098 rg[REG_nS].v.u = (unsigned)iv;
1099 else
1100 rg[(int)REG_nS].v.u = 0u;
80387638
SW
1101 }
1102
1103 return(ROFF_IGN);
1104}
1105
60e1e752
SW
1106/* ARGSUSED */
1107static enum rofferr
1108roff_rm(ROFF_ARGS)
1109{
1110 const char *name;
1111 char *cp;
1112
1113 cp = *bufp + pos;
1114 while ('\0' != *cp) {
1115 name = roff_getname(r, &cp, ln, (int)(cp - *bufp));
1116 if ('\0' != *name)
1117 roff_setstr(r, name, NULL, 0);
1118 }
1119 return(ROFF_IGN);
1120}
1121
80387638
SW
1122/* ARGSUSED */
1123static enum rofferr
1124roff_TE(ROFF_ARGS)
1125{
1126
1127 if (NULL == r->tbl)
60e1e752 1128 mandoc_msg(MANDOCERR_NOSCOPE, r->parse, ln, ppos, NULL);
80387638
SW
1129 else
1130 tbl_end(r->tbl);
1131
1132 r->tbl = NULL;
1133 return(ROFF_IGN);
1134}
1135
1136/* ARGSUSED */
1137static enum rofferr
1138roff_T_(ROFF_ARGS)
1139{
1140
1141 if (NULL == r->tbl)
60e1e752 1142 mandoc_msg(MANDOCERR_NOSCOPE, r->parse, ln, ppos, NULL);
80387638
SW
1143 else
1144 tbl_restart(ppos, ln, r->tbl);
1145
1146 return(ROFF_IGN);
1147}
1148
60e1e752
SW
1149/* ARGSUSED */
1150static enum rofferr
1151roff_EQ(ROFF_ARGS)
1152{
1153 struct eqn_node *e;
1154
1155 assert(NULL == r->eqn);
1156 e = eqn_alloc(ppos, ln);
1157
1158 if (r->last_eqn)
1159 r->last_eqn->next = e;
1160 else
1161 r->first_eqn = r->last_eqn = e;
1162
1163 r->eqn = r->last_eqn = e;
1164 return(ROFF_IGN);
1165}
1166
1167/* ARGSUSED */
1168static enum rofferr
1169roff_EN(ROFF_ARGS)
1170{
1171
1172 mandoc_msg(MANDOCERR_NOSCOPE, r->parse, ln, ppos, NULL);
1173 return(ROFF_IGN);
1174}
1175
80387638
SW
1176/* ARGSUSED */
1177static enum rofferr
1178roff_TS(ROFF_ARGS)
1179{
1180 struct tbl_node *t;
1181
1182 if (r->tbl) {
60e1e752 1183 mandoc_msg(MANDOCERR_SCOPEBROKEN, r->parse, ln, ppos, NULL);
80387638
SW
1184 tbl_end(r->tbl);
1185 }
1186
60e1e752 1187 t = tbl_alloc(ppos, ln, r->parse);
80387638
SW
1188
1189 if (r->last_tbl)
1190 r->last_tbl->next = t;
1191 else
1192 r->first_tbl = r->last_tbl = t;
1193
1194 r->tbl = r->last_tbl = t;
1195 return(ROFF_IGN);
1196}
1197
1198/* ARGSUSED */
1199static enum rofferr
1200roff_so(ROFF_ARGS)
1201{
1202 char *name;
1203
60e1e752 1204 mandoc_msg(MANDOCERR_SO, r->parse, ln, ppos, NULL);
80387638
SW
1205
1206 /*
1207 * Handle `so'. Be EXTREMELY careful, as we shouldn't be
1208 * opening anything that's not in our cwd or anything beneath
1209 * it. Thus, explicitly disallow traversing up the file-system
1210 * or using absolute paths.
1211 */
1212
1213 name = *bufp + pos;
1214 if ('/' == *name || strstr(name, "../") || strstr(name, "/..")) {
60e1e752 1215 mandoc_msg(MANDOCERR_SOPATH, r->parse, ln, pos, NULL);
80387638
SW
1216 return(ROFF_ERR);
1217 }
1218
1219 *offs = pos;
1220 return(ROFF_SO);
1221}
1222
1223/* ARGSUSED */
1224static enum rofferr
1225roff_userdef(ROFF_ARGS)
1226{
1227 const char *arg[9];
1228 char *cp, *n1, *n2;
1229 int i;
1230
1231 /*
1232 * Collect pointers to macro argument strings
1233 * and null-terminate them.
1234 */
1235 cp = *bufp + pos;
1236 for (i = 0; i < 9; i++)
1237 arg[i] = '\0' == *cp ? "" :
60e1e752 1238 mandoc_getarg(r->parse, &cp, ln, &pos);
80387638
SW
1239
1240 /*
1241 * Expand macro arguments.
1242 */
1243 *szp = 0;
1244 n1 = cp = mandoc_strdup(r->current_string);
1245 while (NULL != (cp = strstr(cp, "\\$"))) {
1246 i = cp[2] - '1';
1247 if (0 > i || 8 < i) {
1248 /* Not an argument invocation. */
1249 cp += 2;
1250 continue;
1251 }
1252
1253 *szp = strlen(n1) - 3 + strlen(arg[i]) + 1;
1254 n2 = mandoc_malloc(*szp);
1255
1256 strlcpy(n2, n1, (size_t)(cp - n1 + 1));
1257 strlcat(n2, arg[i], *szp);
1258 strlcat(n2, cp + 3, *szp);
1259
1260 cp = n2 + (cp - n1);
1261 free(n1);
1262 n1 = n2;
1263 }
1264
1265 /*
1266 * Replace the macro invocation
1267 * by the expanded macro.
1268 */
1269 free(*bufp);
1270 *bufp = n1;
1271 if (0 == *szp)
1272 *szp = strlen(*bufp) + 1;
1273
1274 return(*szp > 1 && '\n' == (*bufp)[(int)*szp - 2] ?
1275 ROFF_REPARSE : ROFF_APPEND);
1276}
1277
60e1e752
SW
1278static char *
1279roff_getname(struct roff *r, char **cpp, int ln, int pos)
1280{
1281 char *name, *cp;
1282
1283 name = *cpp;
1284 if ('\0' == *name)
1285 return(name);
1286
1287 /* Read until end of name. */
1288 for (cp = name; '\0' != *cp && ' ' != *cp; cp++) {
1289 if ('\\' != *cp)
1290 continue;
1291 cp++;
1292 if ('\\' == *cp)
1293 continue;
1294 mandoc_msg(MANDOCERR_NAMESC, r->parse, ln, pos, NULL);
1295 *cp = '\0';
1296 name = cp;
1297 }
1298
1299 /* Nil-terminate name. */
1300 if ('\0' != *cp)
1301 *(cp++) = '\0';
1302
1303 /* Read past spaces. */
1304 while (' ' == *cp)
1305 cp++;
1306
1307 *cpp = cp;
1308 return(name);
1309}
1310
80387638
SW
1311/*
1312 * Store *string into the user-defined string called *name.
1313 * In multiline mode, append to an existing entry and append '\n';
1314 * else replace the existing entry, if there is one.
1315 * To clear an existing entry, call with (*r, *name, NULL, 0).
1316 */
1317static void
1318roff_setstr(struct roff *r, const char *name, const char *string,
1319 int multiline)
1320{
1321 struct roffstr *n;
1322 char *c;
1323 size_t oldch, newch;
1324
1325 /* Search for an existing string with the same name. */
1326 n = r->first_string;
1327 while (n && strcmp(name, n->name))
1328 n = n->next;
1329
1330 if (NULL == n) {
1331 /* Create a new string table entry. */
1332 n = mandoc_malloc(sizeof(struct roffstr));
1333 n->name = mandoc_strdup(name);
1334 n->string = NULL;
1335 n->next = r->first_string;
1336 r->first_string = n;
1337 } else if (0 == multiline) {
1338 /* In multiline mode, append; else replace. */
1339 free(n->string);
1340 n->string = NULL;
1341 }
1342
1343 if (NULL == string)
1344 return;
1345
1346 /*
1347 * One additional byte for the '\n' in multiline mode,
1348 * and one for the terminating '\0'.
1349 */
60e1e752 1350 newch = strlen(string) + (multiline ? 2u : 1u);
80387638
SW
1351 if (NULL == n->string) {
1352 n->string = mandoc_malloc(newch);
1353 *n->string = '\0';
1354 oldch = 0;
1355 } else {
1356 oldch = strlen(n->string);
1357 n->string = mandoc_realloc(n->string, oldch + newch);
1358 }
1359
1360 /* Skip existing content in the destination buffer. */
60e1e752 1361 c = n->string + (int)oldch;
80387638
SW
1362
1363 /* Append new content to the destination buffer. */
1364 while (*string) {
1365 /*
1366 * Rudimentary roff copy mode:
1367 * Handle escaped backslashes.
1368 */
1369 if ('\\' == *string && '\\' == *(string + 1))
1370 string++;
1371 *c++ = *string++;
1372 }
1373
1374 /* Append terminating bytes. */
1375 if (multiline)
1376 *c++ = '\n';
1377 *c = '\0';
1378}
1379
80387638
SW
1380static const char *
1381roff_getstrn(const struct roff *r, const char *name, size_t len)
1382{
1383 const struct roffstr *n;
1384
1385 n = r->first_string;
1386 while (n && (strncmp(name, n->name, len) || '\0' != n->name[(int)len]))
1387 n = n->next;
1388
1389 return(n ? n->string : NULL);
1390}
1391
80387638
SW
1392static void
1393roff_freestr(struct roff *r)
1394{
1395 struct roffstr *n, *nn;
1396
1397 for (n = r->first_string; n; n = nn) {
1398 free(n->name);
1399 free(n->string);
1400 nn = n->next;
1401 free(n);
1402 }
1403
1404 r->first_string = NULL;
1405}
1406
1407const struct tbl_span *
1408roff_span(const struct roff *r)
1409{
1410
1411 return(r->tbl ? tbl_span(r->tbl) : NULL);
1412}
60e1e752
SW
1413
1414const struct eqn *
1415roff_eqn(const struct roff *r)
1416{
1417
1418 return(r->last_eqn ? &r->last_eqn->eqn : NULL);
1419}