Import mdocml-1.12.1
[dragonfly.git] / contrib / mdocml / roff.c
CommitLineData
36342e81 1/* $Id: roff.c,v 1.172 2011/10/24 21:41:45 schwarze 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
36342e81
SW
34/* Maximum number of string expansions per line, to break infinite loops. */
35#define EXPAND_LIMIT 1000
36
80387638
SW
37enum rofft {
38 ROFF_ad,
39 ROFF_am,
40 ROFF_ami,
41 ROFF_am1,
42 ROFF_de,
43 ROFF_dei,
44 ROFF_de1,
45 ROFF_ds,
46 ROFF_el,
47 ROFF_hy,
48 ROFF_ie,
49 ROFF_if,
50 ROFF_ig,
60e1e752 51 ROFF_it,
80387638
SW
52 ROFF_ne,
53 ROFF_nh,
54 ROFF_nr,
60e1e752
SW
55 ROFF_ns,
56 ROFF_ps,
80387638
SW
57 ROFF_rm,
58 ROFF_so,
60e1e752 59 ROFF_ta,
80387638
SW
60 ROFF_tr,
61 ROFF_TS,
62 ROFF_TE,
63 ROFF_T_,
60e1e752
SW
64 ROFF_EQ,
65 ROFF_EN,
80387638 66 ROFF_cblock,
a4c7eb57 67 ROFF_ccond,
80387638
SW
68 ROFF_USERDEF,
69 ROFF_MAX
70};
71
72enum roffrule {
73 ROFFRULE_ALLOW,
74 ROFFRULE_DENY
75};
76
36342e81
SW
77/*
78 * A single register entity. If "set" is zero, the value of the
79 * register should be the default one, which is per-register.
80 * Registers are assumed to be unsigned ints for now.
81 */
82struct reg {
83 int set; /* whether set or not */
84 unsigned int u; /* unsigned integer */
85};
86
87/*
88 * An incredibly-simple string buffer.
89 */
80387638 90struct roffstr {
36342e81
SW
91 char *p; /* nil-terminated buffer */
92 size_t sz; /* saved strlen(p) */
93};
94
95/*
96 * A key-value roffstr pair as part of a singly-linked list.
97 */
98struct roffkv {
99 struct roffstr key;
100 struct roffstr val;
101 struct roffkv *next; /* next in list */
80387638
SW
102};
103
104struct roff {
60e1e752 105 struct mparse *parse; /* parse point */
80387638 106 struct roffnode *last; /* leaf of stack */
80387638
SW
107 enum roffrule rstack[RSTACK_MAX]; /* stack of !`ie' rules */
108 int rstackpos; /* position in rstack */
36342e81
SW
109 struct reg regs[REG__MAX];
110 struct roffkv *strtab; /* user-defined strings & macros */
111 struct roffkv *xmbtab; /* multi-byte trans table (`tr') */
112 struct roffstr *xtab; /* single-byte trans table (`tr') */
80387638
SW
113 const char *current_string; /* value of last called user macro */
114 struct tbl_node *first_tbl; /* first table parsed */
115 struct tbl_node *last_tbl; /* last table parsed */
116 struct tbl_node *tbl; /* current table being parsed */
60e1e752
SW
117 struct eqn_node *last_eqn; /* last equation parsed */
118 struct eqn_node *first_eqn; /* first equation parsed */
119 struct eqn_node *eqn; /* current equation being parsed */
80387638
SW
120};
121
122struct roffnode {
123 enum rofft tok; /* type of node */
124 struct roffnode *parent; /* up one in stack */
125 int line; /* parse line */
126 int col; /* parse col */
127 char *name; /* node name, e.g. macro name */
128 char *end; /* end-rules: custom token */
129 int endspan; /* end-rules: next-line or infty */
130 enum roffrule rule; /* current evaluation rule */
131};
132
133#define ROFF_ARGS struct roff *r, /* parse ctx */ \
134 enum rofft tok, /* tok of macro */ \
135 char **bufp, /* input buffer */ \
136 size_t *szp, /* size of input buffer */ \
137 int ln, /* parse line */ \
138 int ppos, /* original pos in buffer */ \
139 int pos, /* current pos in buffer */ \
140 int *offs /* reset offset of buffer data */
141
142typedef enum rofferr (*roffproc)(ROFF_ARGS);
143
144struct roffmac {
145 const char *name; /* macro name */
146 roffproc proc; /* process new macro */
147 roffproc text; /* process as child text of macro */
148 roffproc sub; /* process as child of macro */
149 int flags;
150#define ROFFMAC_STRUCT (1 << 0) /* always interpret */
151 struct roffmac *next;
152};
153
a4c7eb57
SW
154struct predef {
155 const char *name; /* predefined input name */
156 const char *str; /* replacement symbol */
157};
158
159#define PREDEF(__name, __str) \
160 { (__name), (__str) },
161
36342e81
SW
162static enum rofft roffhash_find(const char *, size_t);
163static void roffhash_init(void);
164static void roffnode_cleanscope(struct roff *);
165static void roffnode_pop(struct roff *);
166static void roffnode_push(struct roff *, enum rofft,
167 const char *, int, int);
80387638
SW
168static enum rofferr roff_block(ROFF_ARGS);
169static enum rofferr roff_block_text(ROFF_ARGS);
170static enum rofferr roff_block_sub(ROFF_ARGS);
171static enum rofferr roff_cblock(ROFF_ARGS);
172static enum rofferr roff_ccond(ROFF_ARGS);
173static enum rofferr roff_cond(ROFF_ARGS);
174static enum rofferr roff_cond_text(ROFF_ARGS);
175static enum rofferr roff_cond_sub(ROFF_ARGS);
176static enum rofferr roff_ds(ROFF_ARGS);
177static enum roffrule roff_evalcond(const char *, int *);
36342e81
SW
178static void roff_free1(struct roff *);
179static void roff_freestr(struct roffkv *);
60e1e752 180static char *roff_getname(struct roff *, char **, int, int);
80387638
SW
181static const char *roff_getstrn(const struct roff *,
182 const char *, size_t);
183static enum rofferr roff_line_ignore(ROFF_ARGS);
80387638 184static enum rofferr roff_nr(ROFF_ARGS);
36342e81
SW
185static void roff_openeqn(struct roff *, const char *,
186 int, int, const char *);
187static enum rofft roff_parse(struct roff *, const char *, int *);
188static enum rofferr roff_parsetext(char *);
189static enum rofferr roff_res(struct roff *,
a4c7eb57 190 char **, size_t *, int, int);
60e1e752 191static enum rofferr roff_rm(ROFF_ARGS);
80387638
SW
192static void roff_setstr(struct roff *,
193 const char *, const char *, int);
36342e81
SW
194static void roff_setstrn(struct roffkv **, const char *,
195 size_t, const char *, size_t, int);
80387638 196static enum rofferr roff_so(ROFF_ARGS);
36342e81 197static enum rofferr roff_tr(ROFF_ARGS);
80387638
SW
198static enum rofferr roff_TE(ROFF_ARGS);
199static enum rofferr roff_TS(ROFF_ARGS);
60e1e752
SW
200static enum rofferr roff_EQ(ROFF_ARGS);
201static enum rofferr roff_EN(ROFF_ARGS);
80387638
SW
202static enum rofferr roff_T_(ROFF_ARGS);
203static enum rofferr roff_userdef(ROFF_ARGS);
204
36342e81 205/* See roffhash_find() */
80387638
SW
206
207#define ASCII_HI 126
208#define ASCII_LO 33
209#define HASHWIDTH (ASCII_HI - ASCII_LO + 1)
210
211static struct roffmac *hash[HASHWIDTH];
212
213static struct roffmac roffs[ROFF_MAX] = {
214 { "ad", roff_line_ignore, NULL, NULL, 0, NULL },
215 { "am", roff_block, roff_block_text, roff_block_sub, 0, NULL },
216 { "ami", roff_block, roff_block_text, roff_block_sub, 0, NULL },
217 { "am1", roff_block, roff_block_text, roff_block_sub, 0, NULL },
218 { "de", roff_block, roff_block_text, roff_block_sub, 0, NULL },
219 { "dei", roff_block, roff_block_text, roff_block_sub, 0, NULL },
220 { "de1", roff_block, roff_block_text, roff_block_sub, 0, NULL },
221 { "ds", roff_ds, NULL, NULL, 0, NULL },
222 { "el", roff_cond, roff_cond_text, roff_cond_sub, ROFFMAC_STRUCT, NULL },
223 { "hy", roff_line_ignore, NULL, NULL, 0, NULL },
224 { "ie", roff_cond, roff_cond_text, roff_cond_sub, ROFFMAC_STRUCT, NULL },
225 { "if", roff_cond, roff_cond_text, roff_cond_sub, ROFFMAC_STRUCT, NULL },
226 { "ig", roff_block, roff_block_text, roff_block_sub, 0, NULL },
60e1e752 227 { "it", roff_line_ignore, NULL, NULL, 0, NULL },
80387638
SW
228 { "ne", roff_line_ignore, NULL, NULL, 0, NULL },
229 { "nh", roff_line_ignore, NULL, NULL, 0, NULL },
230 { "nr", roff_nr, NULL, NULL, 0, NULL },
60e1e752
SW
231 { "ns", roff_line_ignore, NULL, NULL, 0, NULL },
232 { "ps", roff_line_ignore, NULL, NULL, 0, NULL },
233 { "rm", roff_rm, NULL, NULL, 0, NULL },
80387638 234 { "so", roff_so, NULL, NULL, 0, NULL },
60e1e752 235 { "ta", roff_line_ignore, NULL, NULL, 0, NULL },
36342e81 236 { "tr", roff_tr, NULL, NULL, 0, NULL },
80387638
SW
237 { "TS", roff_TS, NULL, NULL, 0, NULL },
238 { "TE", roff_TE, NULL, NULL, 0, NULL },
239 { "T&", roff_T_, NULL, NULL, 0, NULL },
60e1e752
SW
240 { "EQ", roff_EQ, NULL, NULL, 0, NULL },
241 { "EN", roff_EN, NULL, NULL, 0, NULL },
80387638
SW
242 { ".", roff_cblock, NULL, NULL, 0, NULL },
243 { "\\}", roff_ccond, NULL, NULL, 0, NULL },
244 { NULL, roff_userdef, NULL, NULL, 0, NULL },
245};
246
a4c7eb57
SW
247/* Array of injected predefined strings. */
248#define PREDEFS_MAX 38
249static const struct predef predefs[PREDEFS_MAX] = {
250#include "predefs.in"
251};
252
36342e81 253/* See roffhash_find() */
80387638
SW
254#define ROFF_HASH(p) (p[0] - ASCII_LO)
255
256static void
36342e81 257roffhash_init(void)
80387638
SW
258{
259 struct roffmac *n;
260 int buc, i;
261
262 for (i = 0; i < (int)ROFF_USERDEF; i++) {
263 assert(roffs[i].name[0] >= ASCII_LO);
264 assert(roffs[i].name[0] <= ASCII_HI);
265
266 buc = ROFF_HASH(roffs[i].name);
267
268 if (NULL != (n = hash[buc])) {
269 for ( ; n->next; n = n->next)
270 /* Do nothing. */ ;
271 n->next = &roffs[i];
272 } else
273 hash[buc] = &roffs[i];
274 }
275}
276
80387638
SW
277/*
278 * Look up a roff token by its name. Returns ROFF_MAX if no macro by
279 * the nil-terminated string name could be found.
280 */
281static enum rofft
36342e81 282roffhash_find(const char *p, size_t s)
80387638
SW
283{
284 int buc;
285 struct roffmac *n;
286
287 /*
288 * libroff has an extremely simple hashtable, for the time
289 * being, which simply keys on the first character, which must
290 * be printable, then walks a chain. It works well enough until
291 * optimised.
292 */
293
294 if (p[0] < ASCII_LO || p[0] > ASCII_HI)
295 return(ROFF_MAX);
296
297 buc = ROFF_HASH(p);
298
299 if (NULL == (n = hash[buc]))
300 return(ROFF_MAX);
301 for ( ; n; n = n->next)
302 if (0 == strncmp(n->name, p, s) && '\0' == n->name[(int)s])
303 return((enum rofft)(n - roffs));
304
305 return(ROFF_MAX);
306}
307
308
309/*
310 * Pop the current node off of the stack of roff instructions currently
311 * pending.
312 */
313static void
314roffnode_pop(struct roff *r)
315{
316 struct roffnode *p;
317
318 assert(r->last);
319 p = r->last;
320
80387638
SW
321 r->last = r->last->parent;
322 free(p->name);
323 free(p->end);
324 free(p);
325}
326
327
328/*
329 * Push a roff node onto the instruction stack. This must later be
330 * removed with roffnode_pop().
331 */
332static void
333roffnode_push(struct roff *r, enum rofft tok, const char *name,
334 int line, int col)
335{
336 struct roffnode *p;
337
338 p = mandoc_calloc(1, sizeof(struct roffnode));
339 p->tok = tok;
340 if (name)
341 p->name = mandoc_strdup(name);
342 p->parent = r->last;
343 p->line = line;
344 p->col = col;
345 p->rule = p->parent ? p->parent->rule : ROFFRULE_DENY;
346
347 r->last = p;
348}
349
350
351static void
352roff_free1(struct roff *r)
353{
354 struct tbl_node *t;
60e1e752 355 struct eqn_node *e;
36342e81 356 int i;
80387638 357
60e1e752 358 while (NULL != (t = r->first_tbl)) {
80387638
SW
359 r->first_tbl = t->next;
360 tbl_free(t);
361 }
362
363 r->first_tbl = r->last_tbl = r->tbl = NULL;
364
60e1e752
SW
365 while (NULL != (e = r->first_eqn)) {
366 r->first_eqn = e->next;
367 eqn_free(e);
368 }
369
370 r->first_eqn = r->last_eqn = r->eqn = NULL;
371
80387638
SW
372 while (r->last)
373 roffnode_pop(r);
374
36342e81
SW
375 roff_freestr(r->strtab);
376 roff_freestr(r->xmbtab);
377
378 r->strtab = r->xmbtab = NULL;
80387638 379
36342e81
SW
380 if (r->xtab)
381 for (i = 0; i < 128; i++)
382 free(r->xtab[i].p);
383
384 free(r->xtab);
385 r->xtab = NULL;
386}
80387638
SW
387
388void
389roff_reset(struct roff *r)
390{
36342e81 391 int i;
80387638
SW
392
393 roff_free1(r);
36342e81
SW
394
395 memset(&r->regs, 0, sizeof(struct reg) * REG__MAX);
396
397 for (i = 0; i < PREDEFS_MAX; i++)
398 roff_setstr(r, predefs[i].name, predefs[i].str, 0);
80387638
SW
399}
400
401
402void
403roff_free(struct roff *r)
404{
405
406 roff_free1(r);
407 free(r);
408}
409
410
411struct roff *
36342e81 412roff_alloc(struct mparse *parse)
80387638
SW
413{
414 struct roff *r;
a4c7eb57 415 int i;
80387638
SW
416
417 r = mandoc_calloc(1, sizeof(struct roff));
60e1e752 418 r->parse = parse;
80387638
SW
419 r->rstackpos = -1;
420
36342e81 421 roffhash_init();
a4c7eb57
SW
422
423 for (i = 0; i < PREDEFS_MAX; i++)
424 roff_setstr(r, predefs[i].name, predefs[i].str, 0);
425
80387638
SW
426 return(r);
427}
428
80387638
SW
429/*
430 * Pre-filter each and every line for reserved words (one beginning with
431 * `\*', e.g., `\*(ab'). These must be handled before the actual line
432 * is processed.
36342e81 433 * This also checks the syntax of regular escapes.
80387638 434 */
36342e81 435static enum rofferr
a4c7eb57 436roff_res(struct roff *r, char **bufp, size_t *szp, int ln, int pos)
80387638 437{
36342e81 438 enum mandoc_esc esc;
80387638
SW
439 const char *stesc; /* start of an escape sequence ('\\') */
440 const char *stnam; /* start of the name, after "[(*" */
441 const char *cp; /* end of the name, e.g. before ']' */
442 const char *res; /* the string to be substituted */
36342e81 443 int i, maxl, expand_count;
80387638
SW
444 size_t nsz;
445 char *n;
446
36342e81 447 expand_count = 0;
80387638 448
36342e81 449again:
80387638
SW
450 cp = *bufp + pos;
451 while (NULL != (cp = strchr(cp, '\\'))) {
452 stesc = cp++;
453
454 /*
455 * The second character must be an asterisk.
456 * If it isn't, skip it anyway: It is escaped,
457 * so it can't start another escape sequence.
458 */
459
460 if ('\0' == *cp)
36342e81
SW
461 return(ROFF_CONT);
462
463 if ('*' != *cp) {
464 res = cp;
465 esc = mandoc_escape(&cp, NULL, NULL);
466 if (ESCAPE_ERROR != esc)
467 continue;
468 cp = res;
469 mandoc_msg
470 (MANDOCERR_BADESCAPE, r->parse,
471 ln, (int)(stesc - *bufp), NULL);
472 return(ROFF_CONT);
473 }
474
475 cp++;
80387638
SW
476
477 /*
478 * The third character decides the length
479 * of the name of the string.
480 * Save a pointer to the name.
481 */
482
483 switch (*cp) {
484 case ('\0'):
36342e81 485 return(ROFF_CONT);
80387638
SW
486 case ('('):
487 cp++;
488 maxl = 2;
489 break;
490 case ('['):
491 cp++;
492 maxl = 0;
493 break;
494 default:
495 maxl = 1;
496 break;
497 }
498 stnam = cp;
499
500 /* Advance to the end of the name. */
501
502 for (i = 0; 0 == maxl || i < maxl; i++, cp++) {
36342e81
SW
503 if ('\0' == *cp) {
504 mandoc_msg
505 (MANDOCERR_BADESCAPE,
506 r->parse, ln,
507 (int)(stesc - *bufp), NULL);
508 return(ROFF_CONT);
509 }
80387638
SW
510 if (0 == maxl && ']' == *cp)
511 break;
512 }
513
514 /*
515 * Retrieve the replacement string; if it is
516 * undefined, resume searching for escapes.
517 */
518
519 res = roff_getstrn(r, stnam, (size_t)i);
520
521 if (NULL == res) {
36342e81
SW
522 mandoc_msg
523 (MANDOCERR_BADESCAPE, r->parse,
524 ln, (int)(stesc - *bufp), NULL);
a4c7eb57 525 res = "";
80387638
SW
526 }
527
528 /* Replace the escape sequence by the string. */
529
36342e81
SW
530 pos = stesc - *bufp;
531
80387638
SW
532 nsz = *szp + strlen(res) + 1;
533 n = mandoc_malloc(nsz);
534
535 strlcpy(n, *bufp, (size_t)(stesc - *bufp + 1));
536 strlcat(n, res, nsz);
537 strlcat(n, cp + (maxl ? 0 : 1), nsz);
538
539 free(*bufp);
540
541 *bufp = n;
542 *szp = nsz;
80387638 543
36342e81
SW
544 if (EXPAND_LIMIT >= ++expand_count)
545 goto again;
546
547 /* Just leave the string unexpanded. */
548 mandoc_msg(MANDOCERR_ROFFLOOP, r->parse, ln, pos, NULL);
549 return(ROFF_IGN);
550 }
551 return(ROFF_CONT);
80387638
SW
552}
553
36342e81
SW
554/*
555 * Process text streams: convert all breakable hyphens into ASCII_HYPH.
556 */
557static enum rofferr
558roff_parsetext(char *p)
559{
560 size_t sz;
561 const char *start;
562 enum mandoc_esc esc;
563
564 start = p;
565
566 while ('\0' != *p) {
567 sz = strcspn(p, "-\\");
568 p += sz;
569
570 if ('\0' == *p)
571 break;
572
573 if ('\\' == *p) {
574 /* Skip over escapes. */
575 p++;
576 esc = mandoc_escape
577 ((const char **)&p, NULL, NULL);
578 if (ESCAPE_ERROR == esc)
579 break;
580 continue;
581 } else if (p == start) {
582 p++;
583 continue;
584 }
585
586 if (isalpha((unsigned char)p[-1]) &&
587 isalpha((unsigned char)p[1]))
588 *p = ASCII_HYPH;
589 p++;
590 }
591
592 return(ROFF_CONT);
593}
80387638
SW
594
595enum rofferr
596roff_parseln(struct roff *r, int ln, char **bufp,
597 size_t *szp, int pos, int *offs)
598{
599 enum rofft t;
600 enum rofferr e;
60e1e752 601 int ppos, ctl;
80387638
SW
602
603 /*
604 * Run the reserved-word filter only if we have some reserved
605 * words to fill in.
606 */
607
36342e81
SW
608 e = roff_res(r, bufp, szp, ln, pos);
609 if (ROFF_IGN == e)
610 return(e);
611 assert(ROFF_CONT == e);
80387638 612
60e1e752
SW
613 ppos = pos;
614 ctl = mandoc_getcontrol(*bufp, &pos);
615
80387638
SW
616 /*
617 * First, if a scope is open and we're not a macro, pass the
618 * text through the macro's filter. If a scope isn't open and
619 * we're not a macro, just let it through.
60e1e752
SW
620 * Finally, if there's an equation scope open, divert it into it
621 * no matter our state.
80387638
SW
622 */
623
60e1e752 624 if (r->last && ! ctl) {
80387638
SW
625 t = r->last->tok;
626 assert(roffs[t].text);
627 e = (*roffs[t].text)
628 (r, t, bufp, szp, ln, pos, pos, offs);
629 assert(ROFF_IGN == e || ROFF_CONT == e);
60e1e752
SW
630 if (ROFF_CONT != e)
631 return(e);
632 if (r->eqn)
36342e81 633 return(eqn_read(&r->eqn, ln, *bufp, pos, offs));
80387638 634 if (r->tbl)
60e1e752 635 return(tbl_read(r->tbl, ln, *bufp, pos));
36342e81 636 return(roff_parsetext(*bufp + pos));
60e1e752
SW
637 } else if ( ! ctl) {
638 if (r->eqn)
36342e81 639 return(eqn_read(&r->eqn, ln, *bufp, pos, offs));
60e1e752
SW
640 if (r->tbl)
641 return(tbl_read(r->tbl, ln, *bufp, pos));
36342e81 642 return(roff_parsetext(*bufp + pos));
60e1e752 643 } else if (r->eqn)
36342e81 644 return(eqn_read(&r->eqn, ln, *bufp, ppos, offs));
80387638
SW
645
646 /*
647 * If a scope is open, go to the child handler for that macro,
648 * as it may want to preprocess before doing anything with it.
60e1e752 649 * Don't do so if an equation is open.
80387638
SW
650 */
651
652 if (r->last) {
653 t = r->last->tok;
654 assert(roffs[t].sub);
655 return((*roffs[t].sub)
656 (r, t, bufp, szp,
60e1e752 657 ln, ppos, pos, offs));
80387638
SW
658 }
659
660 /*
661 * Lastly, as we've no scope open, try to look up and execute
662 * the new macro. If no macro is found, simply return and let
663 * the compilers handle it.
664 */
665
80387638
SW
666 if (ROFF_MAX == (t = roff_parse(r, *bufp, &pos)))
667 return(ROFF_CONT);
668
669 assert(roffs[t].proc);
670 return((*roffs[t].proc)
671 (r, t, bufp, szp,
672 ln, ppos, pos, offs));
673}
674
675
676void
677roff_endparse(struct roff *r)
678{
679
680 if (r->last)
60e1e752 681 mandoc_msg(MANDOCERR_SCOPEEXIT, r->parse,
80387638
SW
682 r->last->line, r->last->col, NULL);
683
60e1e752
SW
684 if (r->eqn) {
685 mandoc_msg(MANDOCERR_SCOPEEXIT, r->parse,
36342e81
SW
686 r->eqn->eqn.ln, r->eqn->eqn.pos, NULL);
687 eqn_end(&r->eqn);
60e1e752
SW
688 }
689
80387638 690 if (r->tbl) {
60e1e752 691 mandoc_msg(MANDOCERR_SCOPEEXIT, r->parse,
80387638 692 r->tbl->line, r->tbl->pos, NULL);
36342e81 693 tbl_end(&r->tbl);
80387638
SW
694 }
695}
696
80387638
SW
697/*
698 * Parse a roff node's type from the input buffer. This must be in the
699 * form of ".foo xxx" in the usual way.
700 */
701static enum rofft
702roff_parse(struct roff *r, const char *buf, int *pos)
703{
704 const char *mac;
705 size_t maclen;
706 enum rofft t;
707
36342e81
SW
708 if ('\0' == buf[*pos] || '"' == buf[*pos] ||
709 '\t' == buf[*pos] || ' ' == buf[*pos])
80387638
SW
710 return(ROFF_MAX);
711
36342e81
SW
712 /*
713 * We stop the macro parse at an escape, tab, space, or nil.
714 * However, `\}' is also a valid macro, so make sure we don't
715 * clobber it by seeing the `\' as the end of token.
716 */
717
80387638 718 mac = buf + *pos;
36342e81 719 maclen = strcspn(mac + 1, " \\\t\0") + 1;
80387638
SW
720
721 t = (r->current_string = roff_getstrn(r, mac, maclen))
36342e81 722 ? ROFF_USERDEF : roffhash_find(mac, maclen);
80387638 723
60e1e752
SW
724 *pos += (int)maclen;
725
80387638
SW
726 while (buf[*pos] && ' ' == buf[*pos])
727 (*pos)++;
728
729 return(t);
730}
731
80387638
SW
732/* ARGSUSED */
733static enum rofferr
734roff_cblock(ROFF_ARGS)
735{
736
737 /*
738 * A block-close `..' should only be invoked as a child of an
739 * ignore macro, otherwise raise a warning and just ignore it.
740 */
741
742 if (NULL == r->last) {
60e1e752 743 mandoc_msg(MANDOCERR_NOSCOPE, r->parse, ln, ppos, NULL);
80387638
SW
744 return(ROFF_IGN);
745 }
746
747 switch (r->last->tok) {
748 case (ROFF_am):
749 /* FALLTHROUGH */
750 case (ROFF_ami):
751 /* FALLTHROUGH */
752 case (ROFF_am1):
753 /* FALLTHROUGH */
754 case (ROFF_de):
755 /* ROFF_de1 is remapped to ROFF_de in roff_block(). */
756 /* FALLTHROUGH */
757 case (ROFF_dei):
758 /* FALLTHROUGH */
759 case (ROFF_ig):
760 break;
761 default:
60e1e752 762 mandoc_msg(MANDOCERR_NOSCOPE, r->parse, ln, ppos, NULL);
80387638
SW
763 return(ROFF_IGN);
764 }
765
766 if ((*bufp)[pos])
60e1e752 767 mandoc_msg(MANDOCERR_ARGSLOST, r->parse, ln, pos, NULL);
80387638
SW
768
769 roffnode_pop(r);
770 roffnode_cleanscope(r);
771 return(ROFF_IGN);
772
773}
774
775
776static void
777roffnode_cleanscope(struct roff *r)
778{
779
780 while (r->last) {
781 if (--r->last->endspan < 0)
782 break;
783 roffnode_pop(r);
784 }
785}
786
787
788/* ARGSUSED */
789static enum rofferr
790roff_ccond(ROFF_ARGS)
791{
792
793 if (NULL == r->last) {
60e1e752 794 mandoc_msg(MANDOCERR_NOSCOPE, r->parse, ln, ppos, NULL);
80387638
SW
795 return(ROFF_IGN);
796 }
797
798 switch (r->last->tok) {
799 case (ROFF_el):
800 /* FALLTHROUGH */
801 case (ROFF_ie):
802 /* FALLTHROUGH */
803 case (ROFF_if):
804 break;
805 default:
60e1e752 806 mandoc_msg(MANDOCERR_NOSCOPE, r->parse, ln, ppos, NULL);
80387638
SW
807 return(ROFF_IGN);
808 }
809
810 if (r->last->endspan > -1) {
60e1e752 811 mandoc_msg(MANDOCERR_NOSCOPE, r->parse, ln, ppos, NULL);
80387638
SW
812 return(ROFF_IGN);
813 }
814
815 if ((*bufp)[pos])
60e1e752 816 mandoc_msg(MANDOCERR_ARGSLOST, r->parse, ln, pos, NULL);
80387638
SW
817
818 roffnode_pop(r);
819 roffnode_cleanscope(r);
820 return(ROFF_IGN);
821}
822
823
824/* ARGSUSED */
825static enum rofferr
826roff_block(ROFF_ARGS)
827{
828 int sv;
829 size_t sz;
830 char *name;
831
832 name = NULL;
833
834 if (ROFF_ig != tok) {
835 if ('\0' == (*bufp)[pos]) {
60e1e752 836 mandoc_msg(MANDOCERR_NOARGS, r->parse, ln, ppos, NULL);
80387638
SW
837 return(ROFF_IGN);
838 }
839
840 /*
841 * Re-write `de1', since we don't really care about
842 * groff's strange compatibility mode, into `de'.
843 */
844
845 if (ROFF_de1 == tok)
846 tok = ROFF_de;
847 if (ROFF_de == tok)
848 name = *bufp + pos;
849 else
60e1e752 850 mandoc_msg(MANDOCERR_REQUEST, r->parse, ln, ppos,
80387638
SW
851 roffs[tok].name);
852
a4c7eb57 853 while ((*bufp)[pos] && ! isspace((unsigned char)(*bufp)[pos]))
80387638
SW
854 pos++;
855
a4c7eb57 856 while (isspace((unsigned char)(*bufp)[pos]))
80387638
SW
857 (*bufp)[pos++] = '\0';
858 }
859
860 roffnode_push(r, tok, name, ln, ppos);
861
862 /*
863 * At the beginning of a `de' macro, clear the existing string
864 * with the same name, if there is one. New content will be
865 * added from roff_block_text() in multiline mode.
866 */
867
868 if (ROFF_de == tok)
869 roff_setstr(r, name, "", 0);
870
871 if ('\0' == (*bufp)[pos])
872 return(ROFF_IGN);
873
874 /* If present, process the custom end-of-line marker. */
875
876 sv = pos;
a4c7eb57 877 while ((*bufp)[pos] && ! isspace((unsigned char)(*bufp)[pos]))
80387638
SW
878 pos++;
879
880 /*
881 * Note: groff does NOT like escape characters in the input.
882 * Instead of detecting this, we're just going to let it fly and
883 * to hell with it.
884 */
885
886 assert(pos > sv);
887 sz = (size_t)(pos - sv);
888
889 if (1 == sz && '.' == (*bufp)[sv])
890 return(ROFF_IGN);
891
892 r->last->end = mandoc_malloc(sz + 1);
893
894 memcpy(r->last->end, *bufp + sv, sz);
895 r->last->end[(int)sz] = '\0';
896
897 if ((*bufp)[pos])
60e1e752 898 mandoc_msg(MANDOCERR_ARGSLOST, r->parse, ln, pos, NULL);
80387638
SW
899
900 return(ROFF_IGN);
901}
902
903
904/* ARGSUSED */
905static enum rofferr
906roff_block_sub(ROFF_ARGS)
907{
908 enum rofft t;
909 int i, j;
910
911 /*
912 * First check whether a custom macro exists at this level. If
913 * it does, then check against it. This is some of groff's
914 * stranger behaviours. If we encountered a custom end-scope
915 * tag and that tag also happens to be a "real" macro, then we
916 * need to try interpreting it again as a real macro. If it's
917 * not, then return ignore. Else continue.
918 */
919
920 if (r->last->end) {
60e1e752 921 for (i = pos, j = 0; r->last->end[j]; j++, i++)
80387638
SW
922 if ((*bufp)[i] != r->last->end[j])
923 break;
924
925 if ('\0' == r->last->end[j] &&
926 ('\0' == (*bufp)[i] ||
927 ' ' == (*bufp)[i] ||
928 '\t' == (*bufp)[i])) {
929 roffnode_pop(r);
930 roffnode_cleanscope(r);
931
60e1e752
SW
932 while (' ' == (*bufp)[i] || '\t' == (*bufp)[i])
933 i++;
934
935 pos = i;
80387638
SW
936 if (ROFF_MAX != roff_parse(r, *bufp, &pos))
937 return(ROFF_RERUN);
938 return(ROFF_IGN);
939 }
940 }
941
942 /*
943 * If we have no custom end-query or lookup failed, then try
944 * pulling it out of the hashtable.
945 */
946
a4c7eb57 947 t = roff_parse(r, *bufp, &pos);
80387638
SW
948
949 /*
950 * Macros other than block-end are only significant
951 * in `de' blocks; elsewhere, simply throw them away.
952 */
953 if (ROFF_cblock != t) {
954 if (ROFF_de == tok)
955 roff_setstr(r, r->last->name, *bufp + ppos, 1);
956 return(ROFF_IGN);
957 }
958
959 assert(roffs[t].proc);
960 return((*roffs[t].proc)(r, t, bufp, szp,
961 ln, ppos, pos, offs));
962}
963
964
965/* ARGSUSED */
966static enum rofferr
967roff_block_text(ROFF_ARGS)
968{
969
970 if (ROFF_de == tok)
971 roff_setstr(r, r->last->name, *bufp + pos, 1);
972
973 return(ROFF_IGN);
974}
975
976
977/* ARGSUSED */
978static enum rofferr
979roff_cond_sub(ROFF_ARGS)
980{
981 enum rofft t;
982 enum roffrule rr;
a4c7eb57 983 char *ep;
80387638 984
80387638 985 rr = r->last->rule;
a4c7eb57 986 roffnode_cleanscope(r);
80387638 987
a4c7eb57
SW
988 /*
989 * If the macro is unknown, first check if it contains a closing
990 * delimiter `\}'. If it does, close out our scope and return
991 * the currently-scoped rule (ignore or continue). Else, drop
992 * into the currently-scoped rule.
80387638
SW
993 */
994
80387638 995 if (ROFF_MAX == (t = roff_parse(r, *bufp, &pos))) {
a4c7eb57
SW
996 ep = &(*bufp)[pos];
997 for ( ; NULL != (ep = strchr(ep, '\\')); ep++) {
998 ep++;
999 if ('}' != *ep)
1000 continue;
36342e81
SW
1001
1002 /*
1003 * Make the \} go away.
1004 * This is a little haphazard, as it's not quite
1005 * clear how nroff does this.
1006 * If we're at the end of line, then just chop
1007 * off the \} and resize the buffer.
1008 * If we aren't, then conver it to spaces.
1009 */
1010
1011 if ('\0' == *(ep + 1)) {
1012 *--ep = '\0';
1013 *szp -= 2;
1014 } else
1015 *(ep - 1) = *ep = ' ';
1016
a4c7eb57
SW
1017 roff_ccond(r, ROFF_ccond, bufp, szp,
1018 ln, pos, pos + 2, offs);
1019 break;
1020 }
80387638
SW
1021 return(ROFFRULE_DENY == rr ? ROFF_IGN : ROFF_CONT);
1022 }
1023
1024 /*
1025 * A denied conditional must evaluate its children if and only
1026 * if they're either structurally required (such as loops and
1027 * conditionals) or a closing macro.
1028 */
a4c7eb57 1029
80387638
SW
1030 if (ROFFRULE_DENY == rr)
1031 if ( ! (ROFFMAC_STRUCT & roffs[t].flags))
1032 if (ROFF_ccond != t)
1033 return(ROFF_IGN);
1034
1035 assert(roffs[t].proc);
1036 return((*roffs[t].proc)(r, t, bufp, szp,
1037 ln, ppos, pos, offs));
1038}
1039
80387638
SW
1040/* ARGSUSED */
1041static enum rofferr
1042roff_cond_text(ROFF_ARGS)
1043{
a4c7eb57 1044 char *ep;
80387638
SW
1045 enum roffrule rr;
1046
1047 rr = r->last->rule;
a4c7eb57 1048 roffnode_cleanscope(r);
80387638 1049
a4c7eb57
SW
1050 ep = &(*bufp)[pos];
1051 for ( ; NULL != (ep = strchr(ep, '\\')); ep++) {
1052 ep++;
1053 if ('}' != *ep)
1054 continue;
1055 *ep = '&';
1056 roff_ccond(r, ROFF_ccond, bufp, szp,
1057 ln, pos, pos + 2, offs);
80387638 1058 }
80387638
SW
1059 return(ROFFRULE_DENY == rr ? ROFF_IGN : ROFF_CONT);
1060}
1061
80387638
SW
1062static enum roffrule
1063roff_evalcond(const char *v, int *pos)
1064{
1065
1066 switch (v[*pos]) {
1067 case ('n'):
1068 (*pos)++;
1069 return(ROFFRULE_ALLOW);
1070 case ('e'):
1071 /* FALLTHROUGH */
1072 case ('o'):
1073 /* FALLTHROUGH */
1074 case ('t'):
1075 (*pos)++;
1076 return(ROFFRULE_DENY);
1077 default:
1078 break;
1079 }
1080
1081 while (v[*pos] && ' ' != v[*pos])
1082 (*pos)++;
1083 return(ROFFRULE_DENY);
1084}
1085
1086/* ARGSUSED */
1087static enum rofferr
1088roff_line_ignore(ROFF_ARGS)
1089{
1090
60e1e752
SW
1091 if (ROFF_it == tok)
1092 mandoc_msg(MANDOCERR_REQUEST, r->parse, ln, ppos, "it");
80387638 1093
80387638
SW
1094 return(ROFF_IGN);
1095}
1096
1097/* ARGSUSED */
1098static enum rofferr
1099roff_cond(ROFF_ARGS)
1100{
1101 int sv;
1102 enum roffrule rule;
1103
a4c7eb57
SW
1104 /*
1105 * An `.el' has no conditional body: it will consume the value
1106 * of the current rstack entry set in prior `ie' calls or
1107 * defaults to DENY.
1108 *
1109 * If we're not an `el', however, then evaluate the conditional.
1110 */
80387638 1111
a4c7eb57
SW
1112 rule = ROFF_el == tok ?
1113 (r->rstackpos < 0 ?
1114 ROFFRULE_DENY : r->rstack[r->rstackpos--]) :
1115 roff_evalcond(*bufp, &pos);
80387638
SW
1116
1117 sv = pos;
80387638
SW
1118 while (' ' == (*bufp)[pos])
1119 pos++;
1120
1121 /*
1122 * Roff is weird. If we have just white-space after the
1123 * conditional, it's considered the BODY and we exit without
1124 * really doing anything. Warn about this. It's probably
1125 * wrong.
1126 */
1127
1128 if ('\0' == (*bufp)[pos] && sv != pos) {
60e1e752 1129 mandoc_msg(MANDOCERR_NOARGS, r->parse, ln, ppos, NULL);
80387638
SW
1130 return(ROFF_IGN);
1131 }
1132
1133 roffnode_push(r, tok, NULL, ln, ppos);
1134
1135 r->last->rule = rule;
1136
a4c7eb57
SW
1137 /*
1138 * An if-else will put the NEGATION of the current evaluated
1139 * conditional into the stack of rules.
1140 */
1141
80387638 1142 if (ROFF_ie == tok) {
a4c7eb57
SW
1143 if (r->rstackpos == RSTACK_MAX - 1) {
1144 mandoc_msg(MANDOCERR_MEM,
1145 r->parse, ln, ppos, NULL);
1146 return(ROFF_ERR);
1147 }
1148 r->rstack[++r->rstackpos] =
1149 ROFFRULE_DENY == r->last->rule ?
1150 ROFFRULE_ALLOW : ROFFRULE_DENY;
80387638
SW
1151 }
1152
1153 /* If the parent has false as its rule, then so do we. */
1154
1155 if (r->last->parent && ROFFRULE_DENY == r->last->parent->rule)
1156 r->last->rule = ROFFRULE_DENY;
1157
1158 /*
1159 * Determine scope. If we're invoked with "\{" trailing the
1160 * conditional, then we're in a multiline scope. Else our scope
1161 * expires on the next line.
1162 */
1163
1164 r->last->endspan = 1;
1165
1166 if ('\\' == (*bufp)[pos] && '{' == (*bufp)[pos + 1]) {
1167 r->last->endspan = -1;
1168 pos += 2;
1169 }
1170
1171 /*
1172 * If there are no arguments on the line, the next-line scope is
1173 * assumed.
1174 */
1175
1176 if ('\0' == (*bufp)[pos])
1177 return(ROFF_IGN);
1178
1179 /* Otherwise re-run the roff parser after recalculating. */
1180
1181 *offs = pos;
1182 return(ROFF_RERUN);
1183}
1184
1185
1186/* ARGSUSED */
1187static enum rofferr
1188roff_ds(ROFF_ARGS)
1189{
1190 char *name, *string;
1191
1192 /*
1193 * A symbol is named by the first word following the macro
1194 * invocation up to a space. Its value is anything after the
1195 * name's trailing whitespace and optional double-quote. Thus,
1196 *
1197 * [.ds foo "bar " ]
1198 *
1199 * will have `bar " ' as its value.
1200 */
1201
60e1e752
SW
1202 string = *bufp + pos;
1203 name = roff_getname(r, &string, ln, pos);
80387638
SW
1204 if ('\0' == *name)
1205 return(ROFF_IGN);
1206
60e1e752
SW
1207 /* Read past initial double-quote. */
1208 if ('"' == *string)
80387638
SW
1209 string++;
1210
1211 /* The rest is the value. */
1212 roff_setstr(r, name, string, 0);
1213 return(ROFF_IGN);
1214}
1215
36342e81
SW
1216int
1217roff_regisset(const struct roff *r, enum regs reg)
1218{
1219
1220 return(r->regs[(int)reg].set);
1221}
1222
1223unsigned int
1224roff_regget(const struct roff *r, enum regs reg)
1225{
1226
1227 return(r->regs[(int)reg].u);
1228}
1229
1230void
1231roff_regunset(struct roff *r, enum regs reg)
1232{
1233
1234 r->regs[(int)reg].set = 0;
1235}
80387638
SW
1236
1237/* ARGSUSED */
1238static enum rofferr
1239roff_nr(ROFF_ARGS)
1240{
60e1e752
SW
1241 const char *key;
1242 char *val;
a4c7eb57 1243 int iv;
80387638 1244
60e1e752
SW
1245 val = *bufp + pos;
1246 key = roff_getname(r, &val, ln, pos);
80387638 1247
80387638 1248 if (0 == strcmp(key, "nS")) {
36342e81
SW
1249 r->regs[(int)REG_nS].set = 1;
1250 if ((iv = mandoc_strntoi(val, strlen(val), 10)) >= 0)
1251 r->regs[(int)REG_nS].u = (unsigned)iv;
a4c7eb57 1252 else
36342e81 1253 r->regs[(int)REG_nS].u = 0u;
80387638
SW
1254 }
1255
1256 return(ROFF_IGN);
1257}
1258
60e1e752
SW
1259/* ARGSUSED */
1260static enum rofferr
1261roff_rm(ROFF_ARGS)
1262{
1263 const char *name;
1264 char *cp;
1265
1266 cp = *bufp + pos;
1267 while ('\0' != *cp) {
1268 name = roff_getname(r, &cp, ln, (int)(cp - *bufp));
1269 if ('\0' != *name)
1270 roff_setstr(r, name, NULL, 0);
1271 }
1272 return(ROFF_IGN);
1273}
1274
80387638
SW
1275/* ARGSUSED */
1276static enum rofferr
1277roff_TE(ROFF_ARGS)
1278{
1279
1280 if (NULL == r->tbl)
60e1e752 1281 mandoc_msg(MANDOCERR_NOSCOPE, r->parse, ln, ppos, NULL);
80387638 1282 else
36342e81 1283 tbl_end(&r->tbl);
80387638 1284
80387638
SW
1285 return(ROFF_IGN);
1286}
1287
1288/* ARGSUSED */
1289static enum rofferr
1290roff_T_(ROFF_ARGS)
1291{
1292
1293 if (NULL == r->tbl)
60e1e752 1294 mandoc_msg(MANDOCERR_NOSCOPE, r->parse, ln, ppos, NULL);
80387638
SW
1295 else
1296 tbl_restart(ppos, ln, r->tbl);
1297
1298 return(ROFF_IGN);
1299}
1300
36342e81
SW
1301#if 0
1302static int
1303roff_closeeqn(struct roff *r)
60e1e752 1304{
36342e81
SW
1305
1306 return(r->eqn && ROFF_EQN == eqn_end(&r->eqn) ? 1 : 0);
1307}
1308#endif
1309
1310static void
1311roff_openeqn(struct roff *r, const char *name, int line,
1312 int offs, const char *buf)
1313{
1314 struct eqn_node *e;
1315 int poff;
60e1e752
SW
1316
1317 assert(NULL == r->eqn);
36342e81 1318 e = eqn_alloc(name, offs, line, r->parse);
60e1e752
SW
1319
1320 if (r->last_eqn)
1321 r->last_eqn->next = e;
1322 else
1323 r->first_eqn = r->last_eqn = e;
1324
1325 r->eqn = r->last_eqn = e;
36342e81
SW
1326
1327 if (buf) {
1328 poff = 0;
1329 eqn_read(&r->eqn, line, buf, offs, &poff);
1330 }
1331}
1332
1333/* ARGSUSED */
1334static enum rofferr
1335roff_EQ(ROFF_ARGS)
1336{
1337
1338 roff_openeqn(r, *bufp + pos, ln, ppos, NULL);
60e1e752
SW
1339 return(ROFF_IGN);
1340}
1341
1342/* ARGSUSED */
1343static enum rofferr
1344roff_EN(ROFF_ARGS)
1345{
1346
1347 mandoc_msg(MANDOCERR_NOSCOPE, r->parse, ln, ppos, NULL);
1348 return(ROFF_IGN);
1349}
1350
80387638
SW
1351/* ARGSUSED */
1352static enum rofferr
1353roff_TS(ROFF_ARGS)
1354{
1355 struct tbl_node *t;
1356
1357 if (r->tbl) {
60e1e752 1358 mandoc_msg(MANDOCERR_SCOPEBROKEN, r->parse, ln, ppos, NULL);
36342e81 1359 tbl_end(&r->tbl);
80387638
SW
1360 }
1361
60e1e752 1362 t = tbl_alloc(ppos, ln, r->parse);
80387638
SW
1363
1364 if (r->last_tbl)
1365 r->last_tbl->next = t;
1366 else
1367 r->first_tbl = r->last_tbl = t;
1368
1369 r->tbl = r->last_tbl = t;
1370 return(ROFF_IGN);
1371}
1372
36342e81
SW
1373/* ARGSUSED */
1374static enum rofferr
1375roff_tr(ROFF_ARGS)
1376{
1377 const char *p, *first, *second;
1378 size_t fsz, ssz;
1379 enum mandoc_esc esc;
1380
1381 p = *bufp + pos;
1382
1383 if ('\0' == *p) {
1384 mandoc_msg(MANDOCERR_ARGCOUNT, r->parse, ln, ppos, NULL);
1385 return(ROFF_IGN);
1386 }
1387
1388 while ('\0' != *p) {
1389 fsz = ssz = 1;
1390
1391 first = p++;
1392 if ('\\' == *first) {
1393 esc = mandoc_escape(&p, NULL, NULL);
1394 if (ESCAPE_ERROR == esc) {
1395 mandoc_msg
1396 (MANDOCERR_BADESCAPE, r->parse,
1397 ln, (int)(p - *bufp), NULL);
1398 return(ROFF_IGN);
1399 }
1400 fsz = (size_t)(p - first);
1401 }
1402
1403 second = p++;
1404 if ('\\' == *second) {
1405 esc = mandoc_escape(&p, NULL, NULL);
1406 if (ESCAPE_ERROR == esc) {
1407 mandoc_msg
1408 (MANDOCERR_BADESCAPE, r->parse,
1409 ln, (int)(p - *bufp), NULL);
1410 return(ROFF_IGN);
1411 }
1412 ssz = (size_t)(p - second);
1413 } else if ('\0' == *second) {
1414 mandoc_msg(MANDOCERR_ARGCOUNT, r->parse,
1415 ln, (int)(p - *bufp), NULL);
1416 second = " ";
1417 p--;
1418 }
1419
1420 if (fsz > 1) {
1421 roff_setstrn(&r->xmbtab, first,
1422 fsz, second, ssz, 0);
1423 continue;
1424 }
1425
1426 if (NULL == r->xtab)
1427 r->xtab = mandoc_calloc
1428 (128, sizeof(struct roffstr));
1429
1430 free(r->xtab[(int)*first].p);
1431 r->xtab[(int)*first].p = mandoc_strndup(second, ssz);
1432 r->xtab[(int)*first].sz = ssz;
1433 }
1434
1435 return(ROFF_IGN);
1436}
1437
80387638
SW
1438/* ARGSUSED */
1439static enum rofferr
1440roff_so(ROFF_ARGS)
1441{
1442 char *name;
1443
60e1e752 1444 mandoc_msg(MANDOCERR_SO, r->parse, ln, ppos, NULL);
80387638
SW
1445
1446 /*
1447 * Handle `so'. Be EXTREMELY careful, as we shouldn't be
1448 * opening anything that's not in our cwd or anything beneath
1449 * it. Thus, explicitly disallow traversing up the file-system
1450 * or using absolute paths.
1451 */
1452
1453 name = *bufp + pos;
1454 if ('/' == *name || strstr(name, "../") || strstr(name, "/..")) {
60e1e752 1455 mandoc_msg(MANDOCERR_SOPATH, r->parse, ln, pos, NULL);
80387638
SW
1456 return(ROFF_ERR);
1457 }
1458
1459 *offs = pos;
1460 return(ROFF_SO);
1461}
1462
1463/* ARGSUSED */
1464static enum rofferr
1465roff_userdef(ROFF_ARGS)
1466{
1467 const char *arg[9];
1468 char *cp, *n1, *n2;
1469 int i;
1470
1471 /*
1472 * Collect pointers to macro argument strings
1473 * and null-terminate them.
1474 */
1475 cp = *bufp + pos;
1476 for (i = 0; i < 9; i++)
1477 arg[i] = '\0' == *cp ? "" :
60e1e752 1478 mandoc_getarg(r->parse, &cp, ln, &pos);
80387638
SW
1479
1480 /*
1481 * Expand macro arguments.
1482 */
1483 *szp = 0;
1484 n1 = cp = mandoc_strdup(r->current_string);
1485 while (NULL != (cp = strstr(cp, "\\$"))) {
1486 i = cp[2] - '1';
1487 if (0 > i || 8 < i) {
1488 /* Not an argument invocation. */
1489 cp += 2;
1490 continue;
1491 }
1492
1493 *szp = strlen(n1) - 3 + strlen(arg[i]) + 1;
1494 n2 = mandoc_malloc(*szp);
1495
1496 strlcpy(n2, n1, (size_t)(cp - n1 + 1));
1497 strlcat(n2, arg[i], *szp);
1498 strlcat(n2, cp + 3, *szp);
1499
1500 cp = n2 + (cp - n1);
1501 free(n1);
1502 n1 = n2;
1503 }
1504
1505 /*
1506 * Replace the macro invocation
1507 * by the expanded macro.
1508 */
1509 free(*bufp);
1510 *bufp = n1;
1511 if (0 == *szp)
1512 *szp = strlen(*bufp) + 1;
1513
1514 return(*szp > 1 && '\n' == (*bufp)[(int)*szp - 2] ?
1515 ROFF_REPARSE : ROFF_APPEND);
1516}
1517
60e1e752
SW
1518static char *
1519roff_getname(struct roff *r, char **cpp, int ln, int pos)
1520{
1521 char *name, *cp;
1522
1523 name = *cpp;
1524 if ('\0' == *name)
1525 return(name);
1526
1527 /* Read until end of name. */
1528 for (cp = name; '\0' != *cp && ' ' != *cp; cp++) {
1529 if ('\\' != *cp)
1530 continue;
1531 cp++;
1532 if ('\\' == *cp)
1533 continue;
1534 mandoc_msg(MANDOCERR_NAMESC, r->parse, ln, pos, NULL);
1535 *cp = '\0';
1536 name = cp;
1537 }
1538
1539 /* Nil-terminate name. */
1540 if ('\0' != *cp)
1541 *(cp++) = '\0';
1542
1543 /* Read past spaces. */
1544 while (' ' == *cp)
1545 cp++;
1546
1547 *cpp = cp;
1548 return(name);
1549}
1550
80387638
SW
1551/*
1552 * Store *string into the user-defined string called *name.
1553 * In multiline mode, append to an existing entry and append '\n';
1554 * else replace the existing entry, if there is one.
1555 * To clear an existing entry, call with (*r, *name, NULL, 0).
1556 */
1557static void
1558roff_setstr(struct roff *r, const char *name, const char *string,
1559 int multiline)
1560{
36342e81
SW
1561
1562 roff_setstrn(&r->strtab, name, strlen(name), string,
1563 string ? strlen(string) : 0, multiline);
1564}
1565
1566static void
1567roff_setstrn(struct roffkv **r, const char *name, size_t namesz,
1568 const char *string, size_t stringsz, int multiline)
1569{
1570 struct roffkv *n;
1571 char *c;
1572 int i;
1573 size_t oldch, newch;
80387638
SW
1574
1575 /* Search for an existing string with the same name. */
36342e81
SW
1576 n = *r;
1577
1578 while (n && strcmp(name, n->key.p))
80387638
SW
1579 n = n->next;
1580
1581 if (NULL == n) {
1582 /* Create a new string table entry. */
36342e81
SW
1583 n = mandoc_malloc(sizeof(struct roffkv));
1584 n->key.p = mandoc_strndup(name, namesz);
1585 n->key.sz = namesz;
1586 n->val.p = NULL;
1587 n->val.sz = 0;
1588 n->next = *r;
1589 *r = n;
80387638
SW
1590 } else if (0 == multiline) {
1591 /* In multiline mode, append; else replace. */
36342e81
SW
1592 free(n->val.p);
1593 n->val.p = NULL;
1594 n->val.sz = 0;
80387638
SW
1595 }
1596
1597 if (NULL == string)
1598 return;
1599
1600 /*
1601 * One additional byte for the '\n' in multiline mode,
1602 * and one for the terminating '\0'.
1603 */
36342e81
SW
1604 newch = stringsz + (multiline ? 2u : 1u);
1605
1606 if (NULL == n->val.p) {
1607 n->val.p = mandoc_malloc(newch);
1608 *n->val.p = '\0';
80387638
SW
1609 oldch = 0;
1610 } else {
36342e81
SW
1611 oldch = n->val.sz;
1612 n->val.p = mandoc_realloc(n->val.p, oldch + newch);
80387638
SW
1613 }
1614
1615 /* Skip existing content in the destination buffer. */
36342e81 1616 c = n->val.p + (int)oldch;
80387638
SW
1617
1618 /* Append new content to the destination buffer. */
36342e81
SW
1619 i = 0;
1620 while (i < (int)stringsz) {
80387638
SW
1621 /*
1622 * Rudimentary roff copy mode:
1623 * Handle escaped backslashes.
1624 */
36342e81
SW
1625 if ('\\' == string[i] && '\\' == string[i + 1])
1626 i++;
1627 *c++ = string[i++];
80387638
SW
1628 }
1629
1630 /* Append terminating bytes. */
1631 if (multiline)
1632 *c++ = '\n';
36342e81 1633
80387638 1634 *c = '\0';
36342e81 1635 n->val.sz = (int)(c - n->val.p);
80387638
SW
1636}
1637
80387638
SW
1638static const char *
1639roff_getstrn(const struct roff *r, const char *name, size_t len)
1640{
36342e81 1641 const struct roffkv *n;
80387638 1642
36342e81
SW
1643 for (n = r->strtab; n; n = n->next)
1644 if (0 == strncmp(name, n->key.p, len) &&
1645 '\0' == n->key.p[(int)len])
1646 return(n->val.p);
80387638 1647
36342e81 1648 return(NULL);
80387638
SW
1649}
1650
80387638 1651static void
36342e81 1652roff_freestr(struct roffkv *r)
80387638 1653{
36342e81 1654 struct roffkv *n, *nn;
80387638 1655
36342e81
SW
1656 for (n = r; n; n = nn) {
1657 free(n->key.p);
1658 free(n->val.p);
80387638
SW
1659 nn = n->next;
1660 free(n);
1661 }
80387638
SW
1662}
1663
1664const struct tbl_span *
1665roff_span(const struct roff *r)
1666{
1667
1668 return(r->tbl ? tbl_span(r->tbl) : NULL);
1669}
60e1e752
SW
1670
1671const struct eqn *
1672roff_eqn(const struct roff *r)
1673{
1674
1675 return(r->last_eqn ? &r->last_eqn->eqn : NULL);
1676}
36342e81
SW
1677
1678/*
1679 * Duplicate an input string, making the appropriate character
1680 * conversations (as stipulated by `tr') along the way.
1681 * Returns a heap-allocated string with all the replacements made.
1682 */
1683char *
1684roff_strdup(const struct roff *r, const char *p)
1685{
1686 const struct roffkv *cp;
1687 char *res;
1688 const char *pp;
1689 size_t ssz, sz;
1690 enum mandoc_esc esc;
1691
1692 if (NULL == r->xmbtab && NULL == r->xtab)
1693 return(mandoc_strdup(p));
1694 else if ('\0' == *p)
1695 return(mandoc_strdup(""));
1696
1697 /*
1698 * Step through each character looking for term matches
1699 * (remember that a `tr' can be invoked with an escape, which is
1700 * a glyph but the escape is multi-character).
1701 * We only do this if the character hash has been initialised
1702 * and the string is >0 length.
1703 */
1704
1705 res = NULL;
1706 ssz = 0;
1707
1708 while ('\0' != *p) {
1709 if ('\\' != *p && r->xtab && r->xtab[(int)*p].p) {
1710 sz = r->xtab[(int)*p].sz;
1711 res = mandoc_realloc(res, ssz + sz + 1);
1712 memcpy(res + ssz, r->xtab[(int)*p].p, sz);
1713 ssz += sz;
1714 p++;
1715 continue;
1716 } else if ('\\' != *p) {
1717 res = mandoc_realloc(res, ssz + 2);
1718 res[ssz++] = *p++;
1719 continue;
1720 }
1721
1722 /* Search for term matches. */
1723 for (cp = r->xmbtab; cp; cp = cp->next)
1724 if (0 == strncmp(p, cp->key.p, cp->key.sz))
1725 break;
1726
1727 if (NULL != cp) {
1728 /*
1729 * A match has been found.
1730 * Append the match to the array and move
1731 * forward by its keysize.
1732 */
1733 res = mandoc_realloc
1734 (res, ssz + cp->val.sz + 1);
1735 memcpy(res + ssz, cp->val.p, cp->val.sz);
1736 ssz += cp->val.sz;
1737 p += (int)cp->key.sz;
1738 continue;
1739 }
1740
1741 /*
1742 * Handle escapes carefully: we need to copy
1743 * over just the escape itself, or else we might
1744 * do replacements within the escape itself.
1745 * Make sure to pass along the bogus string.
1746 */
1747 pp = p++;
1748 esc = mandoc_escape(&p, NULL, NULL);
1749 if (ESCAPE_ERROR == esc) {
1750 sz = strlen(pp);
1751 res = mandoc_realloc(res, ssz + sz + 1);
1752 memcpy(res + ssz, pp, sz);
1753 break;
1754 }
1755 /*
1756 * We bail out on bad escapes.
1757 * No need to warn: we already did so when
1758 * roff_res() was called.
1759 */
1760 sz = (int)(p - pp);
1761 res = mandoc_realloc(res, ssz + sz + 1);
1762 memcpy(res + ssz, pp, sz);
1763 ssz += sz;
1764 }
1765
1766 res[(int)ssz] = '\0';
1767 return(res);
1768}