]> git.cameronkatri.com Git - mandoc.git/blob - mdoc_macro.c
git.ckatri.com / mandoc.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
history | raw | HEAD
Fixed a goddamn subtle error causing MDOC_LITERAL to remain set after a
[mandoc.git] / mdoc_macro.c
1 /* $Id: mdoc_macro.c,v 1.97 2010/11/30 13:04:14 kristaps Exp $ */
2 /*
3 * Copyright (c) 2008, 2009, 2010 Kristaps Dzonsons <kristaps@bsd.lv>
4 * Copyright (c) 2010 Ingo Schwarze <schwarze@openbsd.org>
5 *
6 * Permission to use, copy, modify, and distribute this software for any
7 * purpose with or without fee is hereby granted, provided that the above
8 * copyright notice and this permission notice appear in all copies.
9 *
10 * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
11 * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
12 * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR 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 <ctype.h>
24 #include <stdlib.h>
25 #include <stdio.h>
26 #include <string.h>
27 #include <time.h>
28
29 #include "mandoc.h"
30 #include "libmdoc.h"
31 #include "libmandoc.h"
32
33 enum rew { /* see rew_dohalt() */
34 REWIND_NONE,
35 REWIND_THIS,
36 REWIND_MORE,
37 REWIND_LATER,
38 REWIND_ERROR
39 };
40
41 static int blk_full(MACRO_PROT_ARGS);
42 static int blk_exp_close(MACRO_PROT_ARGS);
43 static int blk_part_exp(MACRO_PROT_ARGS);
44 static int blk_part_imp(MACRO_PROT_ARGS);
45 static int ctx_synopsis(MACRO_PROT_ARGS);
46 static int in_line_eoln(MACRO_PROT_ARGS);
47 static int in_line_argn(MACRO_PROT_ARGS);
48 static int in_line(MACRO_PROT_ARGS);
49 static int obsolete(MACRO_PROT_ARGS);
50 static int phrase_ta(MACRO_PROT_ARGS);
51
52 static int append_delims(struct mdoc *,
53 int, int *, char *);
54 static enum mdoct lookup(enum mdoct, const char *);
55 static enum mdoct lookup_raw(const char *);
56 static int make_pending(struct mdoc_node *, enum mdoct,
57 struct mdoc *, int, int);
58 static int phrase(struct mdoc *, int, int, char *);
59 static enum mdoct rew_alt(enum mdoct);
60 static enum rew rew_dohalt(enum mdoct, enum mdoc_type,
61 const struct mdoc_node *);
62 static int rew_elem(struct mdoc *, enum mdoct);
63 static int rew_last(struct mdoc *,
64 const struct mdoc_node *);
65 static int rew_sub(enum mdoc_type, struct mdoc *,
66 enum mdoct, int, int);
67
68 const struct mdoc_macro __mdoc_macros[MDOC_MAX] = {
69 { in_line_argn, MDOC_CALLABLE | MDOC_PARSED }, /* Ap */
70 { in_line_eoln, MDOC_PROLOGUE }, /* Dd */
71 { in_line_eoln, MDOC_PROLOGUE }, /* Dt */
72 { in_line_eoln, MDOC_PROLOGUE }, /* Os */
73 { blk_full, 0 }, /* Sh */
74 { blk_full, 0 }, /* Ss */
75 { in_line_eoln, 0 }, /* Pp */
76 { blk_part_imp, MDOC_PARSED }, /* D1 */
77 { blk_part_imp, MDOC_PARSED }, /* Dl */
78 { blk_full, MDOC_EXPLICIT }, /* Bd */
79 { blk_exp_close, MDOC_EXPLICIT }, /* Ed */
80 { blk_full, MDOC_EXPLICIT }, /* Bl */
81 { blk_exp_close, MDOC_EXPLICIT }, /* El */
82 { blk_full, MDOC_PARSED }, /* It */
83 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Ad */
84 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* An */
85 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Ar */
86 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Cd */
87 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Cm */
88 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Dv */
89 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Er */
90 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Ev */
91 { in_line_eoln, 0 }, /* Ex */
92 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Fa */
93 { in_line_eoln, 0 }, /* Fd */
94 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Fl */
95 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Fn */
96 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Ft */
97 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Ic */
98 { in_line_argn, MDOC_CALLABLE | MDOC_PARSED }, /* In */
99 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Li */
100 { blk_full, 0 }, /* Nd */
101 { ctx_synopsis, MDOC_CALLABLE | MDOC_PARSED }, /* Nm */
102 { blk_part_imp, MDOC_CALLABLE | MDOC_PARSED }, /* Op */
103 { obsolete, 0 }, /* Ot */
104 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Pa */
105 { in_line_eoln, 0 }, /* Rv */
106 { in_line_argn, MDOC_CALLABLE | MDOC_PARSED }, /* St */
107 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Va */
108 { ctx_synopsis, MDOC_CALLABLE | MDOC_PARSED }, /* Vt */
109 { in_line_argn, MDOC_CALLABLE | MDOC_PARSED }, /* Xr */
110 { in_line_eoln, 0 }, /* %A */
111 { in_line_eoln, 0 }, /* %B */
112 { in_line_eoln, 0 }, /* %D */
113 { in_line_eoln, 0 }, /* %I */
114 { in_line_eoln, 0 }, /* %J */
115 { in_line_eoln, 0 }, /* %N */
116 { in_line_eoln, 0 }, /* %O */
117 { in_line_eoln, 0 }, /* %P */
118 { in_line_eoln, 0 }, /* %R */
119 { in_line_eoln, 0 }, /* %T */
120 { in_line_eoln, 0 }, /* %V */
121 { blk_exp_close, MDOC_EXPLICIT | MDOC_CALLABLE | MDOC_PARSED }, /* Ac */
122 { blk_part_exp, MDOC_CALLABLE | MDOC_PARSED | MDOC_EXPLICIT }, /* Ao */
123 { blk_part_imp, MDOC_CALLABLE | MDOC_PARSED }, /* Aq */
124 { in_line_argn, MDOC_CALLABLE | MDOC_PARSED }, /* At */
125 { blk_exp_close, MDOC_EXPLICIT | MDOC_CALLABLE | MDOC_PARSED }, /* Bc */
126 { blk_full, MDOC_EXPLICIT }, /* Bf */
127 { blk_part_exp, MDOC_CALLABLE | MDOC_PARSED | MDOC_EXPLICIT }, /* Bo */
128 { blk_part_imp, MDOC_CALLABLE | MDOC_PARSED }, /* Bq */
129 { in_line_argn, MDOC_CALLABLE | MDOC_PARSED }, /* Bsx */
130 { in_line_argn, MDOC_CALLABLE | MDOC_PARSED }, /* Bx */
131 { in_line_eoln, 0 }, /* Db */
132 { blk_exp_close, MDOC_EXPLICIT | MDOC_CALLABLE | MDOC_PARSED }, /* Dc */
133 { blk_part_exp, MDOC_CALLABLE | MDOC_PARSED | MDOC_EXPLICIT }, /* Do */
134 { blk_part_imp, MDOC_CALLABLE | MDOC_PARSED }, /* Dq */
135 { blk_exp_close, MDOC_EXPLICIT | MDOC_CALLABLE | MDOC_PARSED }, /* Ec */
136 { blk_exp_close, MDOC_EXPLICIT }, /* Ef */
137 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Em */
138 { blk_part_exp, MDOC_CALLABLE | MDOC_PARSED | MDOC_EXPLICIT }, /* Eo */
139 { in_line_argn, MDOC_CALLABLE | MDOC_PARSED }, /* Fx */
140 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Ms */
141 { in_line_argn, MDOC_CALLABLE | MDOC_PARSED | MDOC_IGNDELIM }, /* No */
142 { in_line_argn, MDOC_CALLABLE | MDOC_PARSED | MDOC_IGNDELIM }, /* Ns */
143 { in_line_argn, MDOC_CALLABLE | MDOC_PARSED }, /* Nx */
144 { in_line_argn, MDOC_CALLABLE | MDOC_PARSED }, /* Ox */
145 { blk_exp_close, MDOC_EXPLICIT | MDOC_CALLABLE | MDOC_PARSED }, /* Pc */
146 { in_line_argn, MDOC_CALLABLE | MDOC_PARSED | MDOC_IGNDELIM }, /* Pf */
147 { blk_part_exp, MDOC_CALLABLE | MDOC_PARSED | MDOC_EXPLICIT }, /* Po */
148 { blk_part_imp, MDOC_CALLABLE | MDOC_PARSED }, /* Pq */
149 { blk_exp_close, MDOC_EXPLICIT | MDOC_CALLABLE | MDOC_PARSED }, /* Qc */
150 { blk_part_imp, MDOC_CALLABLE | MDOC_PARSED }, /* Ql */
151 { blk_part_exp, MDOC_CALLABLE | MDOC_PARSED | MDOC_EXPLICIT }, /* Qo */
152 { blk_part_imp, MDOC_CALLABLE | MDOC_PARSED }, /* Qq */
153 { blk_exp_close, MDOC_EXPLICIT }, /* Re */
154 { blk_full, MDOC_EXPLICIT }, /* Rs */
155 { blk_exp_close, MDOC_EXPLICIT | MDOC_CALLABLE | MDOC_PARSED }, /* Sc */
156 { blk_part_exp, MDOC_CALLABLE | MDOC_PARSED | MDOC_EXPLICIT }, /* So */
157 { blk_part_imp, MDOC_CALLABLE | MDOC_PARSED }, /* Sq */
158 { in_line_eoln, 0 }, /* Sm */
159 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Sx */
160 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Sy */
161 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Tn */
162 { in_line_argn, MDOC_CALLABLE | MDOC_PARSED }, /* Ux */
163 { blk_exp_close, MDOC_EXPLICIT | MDOC_CALLABLE | MDOC_PARSED }, /* Xc */
164 { blk_part_exp, MDOC_CALLABLE | MDOC_PARSED | MDOC_EXPLICIT }, /* Xo */
165 { blk_full, MDOC_EXPLICIT | MDOC_CALLABLE }, /* Fo */
166 { blk_exp_close, MDOC_EXPLICIT | MDOC_CALLABLE | MDOC_PARSED }, /* Fc */
167 { blk_part_exp, MDOC_CALLABLE | MDOC_PARSED | MDOC_EXPLICIT }, /* Oo */
168 { blk_exp_close, MDOC_EXPLICIT | MDOC_CALLABLE | MDOC_PARSED }, /* Oc */
169 { blk_full, MDOC_EXPLICIT }, /* Bk */
170 { blk_exp_close, MDOC_EXPLICIT }, /* Ek */
171 { in_line_eoln, 0 }, /* Bt */
172 { in_line_eoln, 0 }, /* Hf */
173 { obsolete, 0 }, /* Fr */
174 { in_line_eoln, 0 }, /* Ud */
175 { in_line, 0 }, /* Lb */
176 { in_line_eoln, 0 }, /* Lp */
177 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Lk */
178 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Mt */
179 { blk_part_imp, MDOC_CALLABLE | MDOC_PARSED }, /* Brq */
180 { blk_part_exp, MDOC_CALLABLE | MDOC_PARSED | MDOC_EXPLICIT }, /* Bro */
181 { blk_exp_close, MDOC_EXPLICIT | MDOC_CALLABLE | MDOC_PARSED }, /* Brc */
182 { in_line_eoln, 0 }, /* %C */
183 { obsolete, 0 }, /* Es */
184 { obsolete, 0 }, /* En */
185 { in_line_argn, MDOC_CALLABLE | MDOC_PARSED }, /* Dx */
186 { in_line_eoln, 0 }, /* %Q */
187 { in_line_eoln, 0 }, /* br */
188 { in_line_eoln, 0 }, /* sp */
189 { in_line_eoln, 0 }, /* %U */
190 { phrase_ta, MDOC_CALLABLE | MDOC_PARSED }, /* Ta */
191 };
192
193 const struct mdoc_macro * const mdoc_macros = __mdoc_macros;
194
195
196 /*
197 * This is called at the end of parsing. It must traverse up the tree,
198 * closing out open [implicit] scopes. Obviously, open explicit scopes
199 * are errors.
200 */
201 int
202 mdoc_macroend(struct mdoc *m)
203 {
204 struct mdoc_node *n;
205
206 /* Scan for open explicit scopes. */
207
208 n = MDOC_VALID & m->last->flags ? m->last->parent : m->last;
209
210 for ( ; n; n = n->parent)
211 if (MDOC_BLOCK == n->type &&
212 MDOC_EXPLICIT & mdoc_macros[n->tok].flags)
213 mdoc_nmsg(m, n, MANDOCERR_SCOPEEXIT);
214
215 /* Rewind to the first. */
216
217 return(rew_last(m, m->first));
218 }
219
220
221 /*
222 * Look up a macro from within a subsequent context.
223 */
224 static enum mdoct
225 lookup(enum mdoct from, const char *p)
226 {
227 /* FIXME: make -diag lists be un-PARSED. */
228
229 if ( ! (MDOC_PARSED & mdoc_macros[from].flags))
230 return(MDOC_MAX);
231 return(lookup_raw(p));
232 }
233
234
235 /*
236 * Lookup a macro following the initial line macro.
237 */
238 static enum mdoct
239 lookup_raw(const char *p)
240 {
241 enum mdoct res;
242
243 if (MDOC_MAX == (res = mdoc_hash_find(p)))
244 return(MDOC_MAX);
245 if (MDOC_CALLABLE & mdoc_macros[res].flags)
246 return(res);
247 return(MDOC_MAX);
248 }
249
250
251 static int
252 rew_last(struct mdoc *mdoc, const struct mdoc_node *to)
253 {
254
255 assert(to);
256 mdoc->next = MDOC_NEXT_SIBLING;
257
258 /* LINTED */
259 while (mdoc->last != to) {
260 if ( ! mdoc_valid_post(mdoc))
261 return(0);
262 mdoc->last = mdoc->last->parent;
263 assert(mdoc->last);
264 }
265
266 return(mdoc_valid_post(mdoc));
267 }
268
269
270 /*
271 * For a block closing macro, return the corresponding opening one.
272 * Otherwise, return the macro itself.
273 */
274 static enum mdoct
275 rew_alt(enum mdoct tok)
276 {
277 switch (tok) {
278 case (MDOC_Ac):
279 return(MDOC_Ao);
280 case (MDOC_Bc):
281 return(MDOC_Bo);
282 case (MDOC_Brc):
283 return(MDOC_Bro);
284 case (MDOC_Dc):
285 return(MDOC_Do);
286 case (MDOC_Ec):
287 return(MDOC_Eo);
288 case (MDOC_Ed):
289 return(MDOC_Bd);
290 case (MDOC_Ef):
291 return(MDOC_Bf);
292 case (MDOC_Ek):
293 return(MDOC_Bk);
294 case (MDOC_El):
295 return(MDOC_Bl);
296 case (MDOC_Fc):
297 return(MDOC_Fo);
298 case (MDOC_Oc):
299 return(MDOC_Oo);
300 case (MDOC_Pc):
301 return(MDOC_Po);
302 case (MDOC_Qc):
303 return(MDOC_Qo);
304 case (MDOC_Re):
305 return(MDOC_Rs);
306 case (MDOC_Sc):
307 return(MDOC_So);
308 case (MDOC_Xc):
309 return(MDOC_Xo);
310 default:
311 return(tok);
312 }
313 /* NOTREACHED */
314 }
315
316
317 /*
318 * Rewinding to tok, how do we have to handle *p?
319 * REWIND_NONE: *p would delimit tok, but no tok scope is open
320 * inside *p, so there is no need to rewind anything at all.
321 * REWIND_THIS: *p matches tok, so rewind *p and nothing else.
322 * REWIND_MORE: *p is implicit, rewind it and keep searching for tok.
323 * REWIND_LATER: *p is explicit and still open, postpone rewinding.
324 * REWIND_ERROR: No tok block is open at all.
325 */
326 static enum rew
327 rew_dohalt(enum mdoct tok, enum mdoc_type type,
328 const struct mdoc_node *p)
329 {
330
331 /*
332 * No matching token, no delimiting block, no broken block.
333 * This can happen when full implicit macros are called for
334 * the first time but try to rewind their previous
335 * instance anyway.
336 */
337 if (MDOC_ROOT == p->type)
338 return(MDOC_BLOCK == type &&
339 MDOC_EXPLICIT & mdoc_macros[tok].flags ?
340 REWIND_ERROR : REWIND_NONE);
341
342 /*
343 * When starting to rewind, skip plain text
344 * and nodes that have already been rewound.
345 */
346 if (MDOC_TEXT == p->type || MDOC_VALID & p->flags)
347 return(REWIND_MORE);
348
349 /*
350 * The easiest case: Found a matching token.
351 * This applies to both blocks and elements.
352 */
353 tok = rew_alt(tok);
354 if (tok == p->tok)
355 return(p->end ? REWIND_NONE :
356 type == p->type ? REWIND_THIS : REWIND_MORE);
357
358 /*
359 * While elements do require rewinding for themselves,
360 * they never affect rewinding of other nodes.
361 */
362 if (MDOC_ELEM == p->type)
363 return(REWIND_MORE);
364
365 /*
366 * Blocks delimited by our target token get REWIND_MORE.
367 * Blocks delimiting our target token get REWIND_NONE.
368 */
369 switch (tok) {
370 case (MDOC_Bl):
371 if (MDOC_It == p->tok)
372 return(REWIND_MORE);
373 break;
374 case (MDOC_It):
375 if (MDOC_BODY == p->type && MDOC_Bl == p->tok)
376 return(REWIND_NONE);
377 break;
378 /*
379 * XXX Badly nested block handling still fails badly
380 * when one block is breaking two blocks of the same type.
381 * This is an incomplete and extremely ugly workaround,
382 * required to let the OpenBSD tree build.
383 */
384 case (MDOC_Oo):
385 if (MDOC_Op == p->tok)
386 return(REWIND_MORE);
387 break;
388 case (MDOC_Nm):
389 return(REWIND_NONE);
390 case (MDOC_Nd):
391 /* FALLTHROUGH */
392 case (MDOC_Ss):
393 if (MDOC_BODY == p->type && MDOC_Sh == p->tok)
394 return(REWIND_NONE);
395 /* FALLTHROUGH */
396 case (MDOC_Sh):
397 if (MDOC_Nd == p->tok || MDOC_Ss == p->tok ||
398 MDOC_Sh == p->tok)
399 return(REWIND_MORE);
400 break;
401 default:
402 break;
403 }
404
405 /*
406 * Default block rewinding rules.
407 * In particular, always skip block end markers,
408 * and let all blocks rewind Nm children.
409 */
410 if (ENDBODY_NOT != p->end || MDOC_Nm == p->tok ||
411 (MDOC_BLOCK == p->type &&
412 ! (MDOC_EXPLICIT & mdoc_macros[tok].flags)))
413 return(REWIND_MORE);
414
415 /*
416 * Partial blocks allow delayed rewinding by default.
417 */
418 if (&blk_full != mdoc_macros[tok].fp)
419 return (REWIND_LATER);
420
421 /*
422 * Full blocks can only be rewound when matching
423 * or when there is an explicit rule.
424 */
425 return(REWIND_ERROR);
426 }
427
428
429 static int
430 rew_elem(struct mdoc *mdoc, enum mdoct tok)
431 {
432 struct mdoc_node *n;
433
434 n = mdoc->last;
435 if (MDOC_ELEM != n->type)
436 n = n->parent;
437 assert(MDOC_ELEM == n->type);
438 assert(tok == n->tok);
439
440 return(rew_last(mdoc, n));
441 }
442
443
444 /*
445 * We are trying to close a block identified by tok,
446 * but the child block *broken is still open.
447 * Thus, postpone closing the tok block
448 * until the rew_sub call closing *broken.
449 */
450 static int
451 make_pending(struct mdoc_node *broken, enum mdoct tok,
452 struct mdoc *m, int line, int ppos)
453 {
454 struct mdoc_node *breaker;
455
456 /*
457 * Iterate backwards, searching for the block matching tok,
458 * that is, the block breaking the *broken block.
459 */
460 for (breaker = broken->parent; breaker; breaker = breaker->parent) {
461
462 /*
463 * If the *broken block had already been broken before
464 * and we encounter its breaker, make the tok block
465 * pending on the inner breaker.
466 * Graphically, "[A breaker=[B broken=[C->B B] tok=A] C]"
467 * becomes "[A broken=[B [C->B B] tok=A] C]"
468 * and finally "[A [B->A [C->B B] A] C]".
469 */
470 if (breaker == broken->pending) {
471 broken = breaker;
472 continue;
473 }
474
475 if (REWIND_THIS != rew_dohalt(tok, MDOC_BLOCK, breaker))
476 continue;
477 if (MDOC_BODY == broken->type)
478 broken = broken->parent;
479
480 /*
481 * Found the breaker.
482 * If another, outer breaker is already pending on
483 * the *broken block, we must not clobber the link
484 * to the outer breaker, but make it pending on the
485 * new, now inner breaker.
486 * Graphically, "[A breaker=[B broken=[C->A A] tok=B] C]"
487 * becomes "[A breaker=[B->A broken=[C A] tok=B] C]"
488 * and finally "[A [B->A [C->B A] B] C]".
489 */
490 if (broken->pending) {
491 struct mdoc_node *taker;
492
493 /*
494 * If the breaker had also been broken before,
495 * it cannot take on the outer breaker itself,
496 * but must hand it on to its own breakers.
497 * Graphically, this is the following situation:
498 * "[A [B breaker=[C->B B] broken=[D->A A] tok=C] D]"
499 * "[A taker=[B->A breaker=[C->B B] [D->C A] C] D]"
500 */
501 taker = breaker;
502 while (taker->pending)
503 taker = taker->pending;
504 taker->pending = broken->pending;
505 }
506 broken->pending = breaker;
507 mdoc_vmsg(m, MANDOCERR_SCOPENEST, line, ppos,
508 "%s breaks %s", mdoc_macronames[tok],
509 mdoc_macronames[broken->tok]);
510 return(1);
511 }
512
513 /*
514 * Found no matching block for tok.
515 * Are you trying to close a block that is not open?
516 * XXX Make this non-fatal.
517 */
518 mdoc_pmsg(m, line, ppos, MANDOCERR_SYNTNOSCOPE);
519 return(0);
520 }
521
522
523 static int
524 rew_sub(enum mdoc_type t, struct mdoc *m,
525 enum mdoct tok, int line, int ppos)
526 {
527 struct mdoc_node *n;
528
529 n = m->last;
530 while (n) {
531 switch (rew_dohalt(tok, t, n)) {
532 case (REWIND_NONE):
533 return(1);
534 case (REWIND_THIS):
535 break;
536 case (REWIND_MORE):
537 n = n->parent;
538 continue;
539 case (REWIND_LATER):
540 return(make_pending(n, tok, m, line, ppos));
541 case (REWIND_ERROR):
542 /* XXX Make this non-fatal. */
543 mdoc_vmsg(m, MANDOCERR_SCOPEFATAL, line, ppos,
544 "%s cannot break %s", mdoc_macronames[tok],
545 mdoc_macronames[n->tok]);
546 return 0;
547 }
548 break;
549 }
550
551 assert(n);
552 if ( ! rew_last(m, n))
553 return(0);
554
555 /*
556 * The current block extends an enclosing block.
557 * Now that the current block ends, close the enclosing block, too.
558 */
559 while (NULL != (n = n->pending)) {
560 if ( ! rew_last(m, n))
561 return(0);
562 if (MDOC_HEAD == n->type &&
563 ! mdoc_body_alloc(m, n->line, n->pos, n->tok))
564 return(0);
565 }
566
567 return(1);
568 }
569
570
571 static int
572 append_delims(struct mdoc *m, int line, int *pos, char *buf)
573 {
574 int la;
575 enum margserr ac;
576 char *p;
577
578 if ('\0' == buf[*pos])
579 return(1);
580
581 for (;;) {
582 la = *pos;
583 ac = mdoc_zargs(m, line, pos, buf, ARGS_NOWARN, &p);
584
585 if (ARGS_ERROR == ac)
586 return(0);
587 else if (ARGS_EOLN == ac)
588 break;
589
590 assert(DELIM_NONE != mdoc_isdelim(p));
591 if ( ! mdoc_word_alloc(m, line, la, p))
592 return(0);
593
594 /*
595 * If we encounter end-of-sentence symbols, then trigger
596 * the double-space.
597 *
598 * XXX: it's easy to allow this to propogate outward to
599 * the last symbol, such that `. )' will cause the
600 * correct double-spacing. However, (1) groff isn't
601 * smart enough to do this and (2) it would require
602 * knowing which symbols break this behaviour, for
603 * example, `. ;' shouldn't propogate the double-space.
604 */
605 if (mandoc_eos(p, strlen(p), 0))
606 m->last->flags |= MDOC_EOS;
607 }
608
609 return(1);
610 }
611
612
613 /*
614 * Close out block partial/full explicit.
615 */
616 static int
617 blk_exp_close(MACRO_PROT_ARGS)
618 {
619 struct mdoc_node *body; /* Our own body. */
620 struct mdoc_node *later; /* A sub-block starting later. */
621 struct mdoc_node *n; /* For searching backwards. */
622
623 int j, lastarg, maxargs, flushed, nl;
624 enum margserr ac;
625 enum mdoct atok, ntok;
626 char *p;
627
628 nl = MDOC_NEWLINE & m->flags;
629
630 switch (tok) {
631 case (MDOC_Ec):
632 maxargs = 1;
633 break;
634 default:
635 maxargs = 0;
636 break;
637 }
638
639 /*
640 * Search backwards for beginnings of blocks,
641 * both of our own and of pending sub-blocks.
642 */
643 atok = rew_alt(tok);
644 body = later = NULL;
645 for (n = m->last; n; n = n->parent) {
646 if (MDOC_VALID & n->flags)
647 continue;
648
649 /* Remember the start of our own body. */
650 if (MDOC_BODY == n->type && atok == n->tok) {
651 if (ENDBODY_NOT == n->end)
652 body = n;
653 continue;
654 }
655
656 if (MDOC_BLOCK != n->type || MDOC_Nm == n->tok)
657 continue;
658 if (atok == n->tok) {
659 assert(body);
660
661 /*
662 * Found the start of our own block.
663 * When there is no pending sub block,
664 * just proceed to closing out.
665 */
666 if (NULL == later)
667 break;
668
669 /*
670 * When there is a pending sub block,
671 * postpone closing out the current block
672 * until the rew_sub() closing out the sub-block.
673 */
674 if ( ! make_pending(later, tok, m, line, ppos))
675 return(0);
676
677 /*
678 * Mark the place where the formatting - but not
679 * the scope - of the current block ends.
680 */
681 if ( ! mdoc_endbody_alloc(m, line, ppos,
682 atok, body, ENDBODY_SPACE))
683 return(0);
684 break;
685 }
686
687 /*
688 * When finding an open sub block, remember the last
689 * open explicit block, or, in case there are only
690 * implicit ones, the first open implicit block.
691 */
692 if (later &&
693 MDOC_EXPLICIT & mdoc_macros[later->tok].flags)
694 continue;
695 if (MDOC_CALLABLE & mdoc_macros[n->tok].flags)
696 later = n;
697 }
698
699 if ( ! (MDOC_CALLABLE & mdoc_macros[tok].flags)) {
700 /* FIXME: do this in validate */
701 if (buf[*pos])
702 if ( ! mdoc_pmsg(m, line, ppos, MANDOCERR_ARGSLOST))
703 return(0);
704
705 if ( ! rew_sub(MDOC_BODY, m, tok, line, ppos))
706 return(0);
707 return(rew_sub(MDOC_BLOCK, m, tok, line, ppos));
708 }
709
710 if ( ! rew_sub(MDOC_BODY, m, tok, line, ppos))
711 return(0);
712
713 if (NULL == later && maxargs > 0)
714 if ( ! mdoc_tail_alloc(m, line, ppos, rew_alt(tok)))
715 return(0);
716
717 for (flushed = j = 0; ; j++) {
718 lastarg = *pos;
719
720 if (j == maxargs && ! flushed) {
721 if ( ! rew_sub(MDOC_BLOCK, m, tok, line, ppos))
722 return(0);
723 flushed = 1;
724 }
725
726 ac = mdoc_args(m, line, pos, buf, tok, &p);
727
728 if (ARGS_ERROR == ac)
729 return(0);
730 if (ARGS_PUNCT == ac)
731 break;
732 if (ARGS_EOLN == ac)
733 break;
734
735 ntok = ARGS_QWORD == ac ? MDOC_MAX : lookup(tok, p);
736
737 if (MDOC_MAX == ntok) {
738 if ( ! mdoc_word_alloc(m, line, lastarg, p))
739 return(0);
740 continue;
741 }
742
743 if ( ! flushed) {
744 if ( ! rew_sub(MDOC_BLOCK, m, tok, line, ppos))
745 return(0);
746 flushed = 1;
747 }
748 if ( ! mdoc_macro(m, ntok, line, lastarg, pos, buf))
749 return(0);
750 break;
751 }
752
753 if ( ! flushed && ! rew_sub(MDOC_BLOCK, m, tok, line, ppos))
754 return(0);
755
756 if ( ! nl)
757 return(1);
758 return(append_delims(m, line, pos, buf));
759 }
760
761
762 static int
763 in_line(MACRO_PROT_ARGS)
764 {
765 int la, scope, cnt, nc, nl;
766 enum margverr av;
767 enum mdoct ntok;
768 enum margserr ac;
769 enum mdelim d;
770 struct mdoc_arg *arg;
771 char *p;
772
773 nl = MDOC_NEWLINE & m->flags;
774
775 /*
776 * Whether we allow ignored elements (those without content,
777 * usually because of reserved words) to squeak by.
778 */
779
780 switch (tok) {
781 case (MDOC_An):
782 /* FALLTHROUGH */
783 case (MDOC_Ar):
784 /* FALLTHROUGH */
785 case (MDOC_Fl):
786 /* FALLTHROUGH */
787 case (MDOC_Mt):
788 /* FALLTHROUGH */
789 case (MDOC_Nm):
790 /* FALLTHROUGH */
791 case (MDOC_Pa):
792 nc = 1;
793 break;
794 default:
795 nc = 0;
796 break;
797 }
798
799 for (arg = NULL;; ) {
800 la = *pos;
801 av = mdoc_argv(m, line, tok, &arg, pos, buf);
802
803 if (ARGV_WORD == av) {
804 *pos = la;
805 break;
806 }
807 if (ARGV_EOLN == av)
808 break;
809 if (ARGV_ARG == av)
810 continue;
811
812 mdoc_argv_free(arg);
813 return(0);
814 }
815
816 for (cnt = scope = 0;; ) {
817 la = *pos;
818 ac = mdoc_args(m, line, pos, buf, tok, &p);
819
820 if (ARGS_ERROR == ac)
821 return(0);
822 if (ARGS_EOLN == ac)
823 break;
824 if (ARGS_PUNCT == ac)
825 break;
826
827 ntok = ARGS_QWORD == ac ? MDOC_MAX : lookup(tok, p);
828
829 /*
830 * In this case, we've located a submacro and must
831 * execute it. Close out scope, if open. If no
832 * elements have been generated, either create one (nc)
833 * or raise a warning.
834 */
835
836 if (MDOC_MAX != ntok) {
837 if (scope && ! rew_elem(m, tok))
838 return(0);
839 if (nc && 0 == cnt) {
840 if ( ! mdoc_elem_alloc(m, line, ppos, tok, arg))
841 return(0);
842 if ( ! rew_last(m, m->last))
843 return(0);
844 } else if ( ! nc && 0 == cnt) {
845 mdoc_argv_free(arg);
846 if ( ! mdoc_pmsg(m, line, ppos, MANDOCERR_MACROEMPTY))
847 return(0);
848 }
849 if ( ! mdoc_macro(m, ntok, line, la, pos, buf))
850 return(0);
851 if ( ! nl)
852 return(1);
853 return(append_delims(m, line, pos, buf));
854 }
855
856 /*
857 * Non-quote-enclosed punctuation. Set up our scope, if
858 * a word; rewind the scope, if a delimiter; then append
859 * the word.
860 */
861
862 d = ARGS_QWORD == ac ? DELIM_NONE : mdoc_isdelim(p);
863
864 if (DELIM_NONE != d) {
865 /*
866 * If we encounter closing punctuation, no word
867 * has been omitted, no scope is open, and we're
868 * allowed to have an empty element, then start
869 * a new scope. `Ar', `Fl', and `Li', only do
870 * this once per invocation. There may be more
871 * of these (all of them?).
872 */
873 if (0 == cnt && (nc || MDOC_Li == tok) &&
874 DELIM_CLOSE == d && ! scope) {
875 if ( ! mdoc_elem_alloc(m, line, ppos, tok, arg))
876 return(0);
877 if (MDOC_Ar == tok || MDOC_Li == tok ||
878 MDOC_Fl == tok)
879 cnt++;
880 scope = 1;
881 }
882 /*
883 * Close out our scope, if one is open, before
884 * any punctuation.
885 */
886 if (scope && ! rew_elem(m, tok))
887 return(0);
888 scope = 0;
889 } else if ( ! scope) {
890 if ( ! mdoc_elem_alloc(m, line, ppos, tok, arg))
891 return(0);
892 scope = 1;
893 }
894
895 if (DELIM_NONE == d)
896 cnt++;
897 if ( ! mdoc_word_alloc(m, line, la, p))
898 return(0);
899
900 /*
901 * `Fl' macros have their scope re-opened with each new
902 * word so that the `-' can be added to each one without
903 * having to parse out spaces.
904 */
905 if (scope && MDOC_Fl == tok) {
906 if ( ! rew_elem(m, tok))
907 return(0);
908 scope = 0;
909 }
910 }
911
912 if (scope && ! rew_elem(m, tok))
913 return(0);
914
915 /*
916 * If no elements have been collected and we're allowed to have
917 * empties (nc), open a scope and close it out. Otherwise,
918 * raise a warning.
919 */
920
921 if (nc && 0 == cnt) {
922 if ( ! mdoc_elem_alloc(m, line, ppos, tok, arg))
923 return(0);
924 if ( ! rew_last(m, m->last))
925 return(0);
926 } else if ( ! nc && 0 == cnt) {
927 mdoc_argv_free(arg);
928 if ( ! mdoc_pmsg(m, line, ppos, MANDOCERR_MACROEMPTY))
929 return(0);
930 }
931
932 if ( ! nl)
933 return(1);
934 return(append_delims(m, line, pos, buf));
935 }
936
937
938 static int
939 blk_full(MACRO_PROT_ARGS)
940 {
941 int la, nl;
942 struct mdoc_arg *arg;
943 struct mdoc_node *head; /* save of head macro */
944 struct mdoc_node *body; /* save of body macro */
945 struct mdoc_node *n;
946 enum mdoc_type mtt;
947 enum mdoct ntok;
948 enum margserr ac, lac;
949 enum margverr av;
950 char *p;
951
952 nl = MDOC_NEWLINE & m->flags;
953
954 /* Close out prior implicit scope. */
955
956 if ( ! (MDOC_EXPLICIT & mdoc_macros[tok].flags)) {
957 if ( ! rew_sub(MDOC_BODY, m, tok, line, ppos))
958 return(0);
959 if ( ! rew_sub(MDOC_BLOCK, m, tok, line, ppos))
960 return(0);
961 }
962
963 /*
964 * This routine accomodates implicitly- and explicitly-scoped
965 * macro openings. Implicit ones first close out prior scope
966 * (seen above). Delay opening the head until necessary to
967 * allow leading punctuation to print. Special consideration
968 * for `It -column', which has phrase-part syntax instead of
969 * regular child nodes.
970 */
971
972 for (arg = NULL;; ) {
973 la = *pos;
974 av = mdoc_argv(m, line, tok, &arg, pos, buf);
975
976 if (ARGV_WORD == av) {
977 *pos = la;
978 break;
979 }
980
981 if (ARGV_EOLN == av)
982 break;
983 if (ARGV_ARG == av)
984 continue;
985
986 mdoc_argv_free(arg);
987 return(0);
988 }
989
990 if ( ! mdoc_block_alloc(m, line, ppos, tok, arg))
991 return(0);
992
993 head = body = NULL;
994
995 /*
996 * The `Nd' macro has all arguments in its body: it's a hybrid
997 * of block partial-explicit and full-implicit. Stupid.
998 */
999
1000 if (MDOC_Nd == tok) {
1001 if ( ! mdoc_head_alloc(m, line, ppos, tok))
1002 return(0);
1003 head = m->last;
1004 if ( ! rew_sub(MDOC_HEAD, m, tok, line, ppos))
1005 return(0);
1006 if ( ! mdoc_body_alloc(m, line, ppos, tok))
1007 return(0);
1008 body = m->last;
1009 }
1010
1011 ac = ARGS_ERROR;
1012
1013 for ( ; ; ) {
1014 la = *pos;
1015 /* Initialise last-phrase-type with ARGS_PEND. */
1016 lac = ARGS_ERROR == ac ? ARGS_PEND : ac;
1017 ac = mdoc_args(m, line, pos, buf, tok, &p);
1018
1019 if (ARGS_PUNCT == ac)
1020 break;
1021
1022 if (ARGS_ERROR == ac)
1023 return(0);
1024
1025 if (ARGS_EOLN == ac) {
1026 if (ARGS_PPHRASE != lac && ARGS_PHRASE != lac)
1027 break;
1028 /*
1029 * This is necessary: if the last token on a
1030 * line is a `Ta' or tab, then we'll get
1031 * ARGS_EOLN, so we must be smart enough to
1032 * reopen our scope if the last parse was a
1033 * phrase or partial phrase.
1034 */
1035 if ( ! rew_sub(MDOC_BODY, m, tok, line, ppos))
1036 return(0);
1037 if ( ! mdoc_body_alloc(m, line, ppos, tok))
1038 return(0);
1039 body = m->last;
1040 break;
1041 }
1042
1043 /*
1044 * Emit leading punctuation (i.e., punctuation before
1045 * the MDOC_HEAD) for non-phrase types.
1046 */
1047
1048 if (NULL == head &&
1049 ARGS_PEND != ac &&
1050 ARGS_PHRASE != ac &&
1051 ARGS_PPHRASE != ac &&
1052 ARGS_QWORD != ac &&
1053 DELIM_OPEN == mdoc_isdelim(p)) {
1054 if ( ! mdoc_word_alloc(m, line, la, p))
1055 return(0);
1056 continue;
1057 }
1058
1059 /* Open a head if one hasn't been opened. */
1060
1061 if (NULL == head) {
1062 if ( ! mdoc_head_alloc(m, line, ppos, tok))
1063 return(0);
1064 head = m->last;
1065 }
1066
1067 if (ARGS_PHRASE == ac ||
1068 ARGS_PEND == ac ||
1069 ARGS_PPHRASE == ac) {
1070 /*
1071 * If we haven't opened a body yet, rewind the
1072 * head; if we have, rewind that instead.
1073 */
1074
1075 mtt = body ? MDOC_BODY : MDOC_HEAD;
1076 if ( ! rew_sub(mtt, m, tok, line, ppos))
1077 return(0);
1078
1079 /* Then allocate our body context. */
1080
1081 if ( ! mdoc_body_alloc(m, line, ppos, tok))
1082 return(0);
1083 body = m->last;
1084
1085 /*
1086 * Process phrases: set whether we're in a
1087 * partial-phrase (this effects line handling)
1088 * then call down into the phrase parser.
1089 */
1090
1091 if (ARGS_PPHRASE == ac)
1092 m->flags |= MDOC_PPHRASE;
1093 if (ARGS_PEND == ac && ARGS_PPHRASE == lac)
1094 m->flags |= MDOC_PPHRASE;
1095
1096 if ( ! phrase(m, line, la, buf))
1097 return(0);
1098
1099 m->flags &= ~MDOC_PPHRASE;
1100 continue;
1101 }
1102
1103 ntok = ARGS_QWORD == ac ? MDOC_MAX : lookup(tok, p);
1104
1105 if (MDOC_MAX == ntok) {
1106 if ( ! mdoc_word_alloc(m, line, la, p))
1107 return(0);
1108 continue;
1109 }
1110
1111 if ( ! mdoc_macro(m, ntok, line, la, pos, buf))
1112 return(0);
1113 break;
1114 }
1115
1116 if (NULL == head) {
1117 if ( ! mdoc_head_alloc(m, line, ppos, tok))
1118 return(0);
1119 head = m->last;
1120 }
1121
1122 if (nl && ! append_delims(m, line, pos, buf))
1123 return(0);
1124
1125 /* If we've already opened our body, exit now. */
1126
1127 if (NULL != body)
1128 goto out;
1129
1130 /*
1131 * If there is an open (i.e., unvalidated) sub-block requiring
1132 * explicit close-out, postpone switching the current block from
1133 * head to body until the rew_sub() call closing out that
1134 * sub-block.
1135 */
1136 for (n = m->last; n && n != head; n = n->parent) {
1137 if (MDOC_BLOCK == n->type &&
1138 MDOC_EXPLICIT & mdoc_macros[n->tok].flags &&
1139 ! (MDOC_VALID & n->flags)) {
1140 n->pending = head;
1141 return(1);
1142 }
1143 }
1144
1145 /* Close out scopes to remain in a consistent state. */
1146
1147 if ( ! rew_sub(MDOC_HEAD, m, tok, line, ppos))
1148 return(0);
1149 if ( ! mdoc_body_alloc(m, line, ppos, tok))
1150 return(0);
1151
1152 out:
1153 if ( ! (MDOC_FREECOL & m->flags))
1154 return(1);
1155
1156 if ( ! rew_sub(MDOC_BODY, m, tok, line, ppos))
1157 return(0);
1158 if ( ! rew_sub(MDOC_BLOCK, m, tok, line, ppos))
1159 return(0);
1160
1161 m->flags &= ~MDOC_FREECOL;
1162 return(1);
1163 }
1164
1165
1166 static int
1167 blk_part_imp(MACRO_PROT_ARGS)
1168 {
1169 int la, nl;
1170 enum mdoct ntok;
1171 enum margserr ac;
1172 char *p;
1173 struct mdoc_node *blk; /* saved block context */
1174 struct mdoc_node *body; /* saved body context */
1175 struct mdoc_node *n;
1176
1177 nl = MDOC_NEWLINE & m->flags;
1178
1179 /*
1180 * A macro that spans to the end of the line. This is generally
1181 * (but not necessarily) called as the first macro. The block
1182 * has a head as the immediate child, which is always empty,
1183 * followed by zero or more opening punctuation nodes, then the
1184 * body (which may be empty, depending on the macro), then zero
1185 * or more closing punctuation nodes.
1186 */
1187
1188 if ( ! mdoc_block_alloc(m, line, ppos, tok, NULL))
1189 return(0);
1190
1191 blk = m->last;
1192
1193 if ( ! mdoc_head_alloc(m, line, ppos, tok))
1194 return(0);
1195 if ( ! rew_sub(MDOC_HEAD, m, tok, line, ppos))
1196 return(0);
1197
1198 /*
1199 * Open the body scope "on-demand", that is, after we've
1200 * processed all our the leading delimiters (open parenthesis,
1201 * etc.).
1202 */
1203
1204 for (body = NULL; ; ) {
1205 la = *pos;
1206 ac = mdoc_args(m, line, pos, buf, tok, &p);
1207
1208 if (ARGS_ERROR == ac)
1209 return(0);
1210 if (ARGS_EOLN == ac)
1211 break;
1212 if (ARGS_PUNCT == ac)
1213 break;
1214
1215 if (NULL == body && ARGS_QWORD != ac &&
1216 DELIM_OPEN == mdoc_isdelim(p)) {
1217 if ( ! mdoc_word_alloc(m, line, la, p))
1218 return(0);
1219 continue;
1220 }
1221
1222 if (NULL == body) {
1223 if ( ! mdoc_body_alloc(m, line, ppos, tok))
1224 return(0);
1225 body = m->last;
1226 }
1227
1228 ntok = ARGS_QWORD == ac ? MDOC_MAX : lookup(tok, p);
1229
1230 if (MDOC_MAX == ntok) {
1231 if ( ! mdoc_word_alloc(m, line, la, p))
1232 return(0);
1233 continue;
1234 }
1235
1236 if ( ! mdoc_macro(m, ntok, line, la, pos, buf))
1237 return(0);
1238 break;
1239 }
1240
1241 /* Clean-ups to leave in a consistent state. */
1242
1243 if (NULL == body) {
1244 if ( ! mdoc_body_alloc(m, line, ppos, tok))
1245 return(0);
1246 body = m->last;
1247 }
1248
1249 for (n = body->child; n && n->next; n = n->next)
1250 /* Do nothing. */ ;
1251
1252 /*
1253 * End of sentence spacing: if the last node is a text node and
1254 * has a trailing period, then mark it as being end-of-sentence.
1255 */
1256
1257 if (n && MDOC_TEXT == n->type && n->string)
1258 if (mandoc_eos(n->string, strlen(n->string), 1))
1259 n->flags |= MDOC_EOS;
1260
1261 /* Up-propogate the end-of-space flag. */
1262
1263 if (n && (MDOC_EOS & n->flags)) {
1264 body->flags |= MDOC_EOS;
1265 body->parent->flags |= MDOC_EOS;
1266 }
1267
1268 /*
1269 * If there is an open sub-block requiring explicit close-out,
1270 * postpone closing out the current block
1271 * until the rew_sub() call closing out the sub-block.
1272 */
1273 for (n = m->last; n && n != body && n != blk->parent; n = n->parent) {
1274 if (MDOC_BLOCK == n->type &&
1275 MDOC_EXPLICIT & mdoc_macros[n->tok].flags &&
1276 ! (MDOC_VALID & n->flags)) {
1277 if ( ! make_pending(n, tok, m, line, ppos))
1278 return(0);
1279 if ( ! mdoc_endbody_alloc(m, line, ppos,
1280 tok, body, ENDBODY_NOSPACE))
1281 return(0);
1282 return(1);
1283 }
1284 }
1285
1286 /*
1287 * If we can't rewind to our body, then our scope has already
1288 * been closed by another macro (like `Oc' closing `Op'). This
1289 * is ugly behaviour nodding its head to OpenBSD's overwhelming
1290 * crufty use of `Op' breakage.
1291 */
1292 if (n != body && ! mdoc_vmsg(m, MANDOCERR_SCOPENEST,
1293 line, ppos, "%s broken", mdoc_macronames[tok]))
1294 return(0);
1295
1296 if (n && ! rew_sub(MDOC_BODY, m, tok, line, ppos))
1297 return(0);
1298
1299 /* Standard appending of delimiters. */
1300
1301 if (nl && ! append_delims(m, line, pos, buf))
1302 return(0);
1303
1304 /* Rewind scope, if applicable. */
1305
1306 if (n && ! rew_sub(MDOC_BLOCK, m, tok, line, ppos))
1307 return(0);
1308
1309 return(1);
1310 }
1311
1312
1313 static int
1314 blk_part_exp(MACRO_PROT_ARGS)
1315 {
1316 int la, nl;
1317 enum margserr ac;
1318 struct mdoc_node *head; /* keep track of head */
1319 struct mdoc_node *body; /* keep track of body */
1320 char *p;
1321 enum mdoct ntok;
1322
1323 nl = MDOC_NEWLINE & m->flags;
1324
1325 /*
1326 * The opening of an explicit macro having zero or more leading
1327 * punctuation nodes; a head with optional single element (the
1328 * case of `Eo'); and a body that may be empty.
1329 */
1330
1331 if ( ! mdoc_block_alloc(m, line, ppos, tok, NULL))
1332 return(0);
1333
1334 for (head = body = NULL; ; ) {
1335 la = *pos;
1336 ac = mdoc_args(m, line, pos, buf, tok, &p);
1337
1338 if (ARGS_ERROR == ac)
1339 return(0);
1340 if (ARGS_PUNCT == ac)
1341 break;
1342 if (ARGS_EOLN == ac)
1343 break;
1344
1345 /* Flush out leading punctuation. */
1346
1347 if (NULL == head && ARGS_QWORD != ac &&
1348 DELIM_OPEN == mdoc_isdelim(p)) {
1349 assert(NULL == body);
1350 if ( ! mdoc_word_alloc(m, line, la, p))
1351 return(0);
1352 continue;
1353 }
1354
1355 if (NULL == head) {
1356 assert(NULL == body);
1357 if ( ! mdoc_head_alloc(m, line, ppos, tok))
1358 return(0);
1359 head = m->last;
1360 }
1361
1362 /*
1363 * `Eo' gobbles any data into the head, but most other
1364 * macros just immediately close out and begin the body.
1365 */
1366
1367 if (NULL == body) {
1368 assert(head);
1369 /* No check whether it's a macro! */
1370 if (MDOC_Eo == tok)
1371 if ( ! mdoc_word_alloc(m, line, la, p))
1372 return(0);
1373
1374 if ( ! rew_sub(MDOC_HEAD, m, tok, line, ppos))
1375 return(0);
1376 if ( ! mdoc_body_alloc(m, line, ppos, tok))
1377 return(0);
1378 body = m->last;
1379
1380 if (MDOC_Eo == tok)
1381 continue;
1382 }
1383
1384 assert(NULL != head && NULL != body);
1385
1386 ntok = ARGS_QWORD == ac ? MDOC_MAX : lookup(tok, p);
1387
1388 if (MDOC_MAX == ntok) {
1389 if ( ! mdoc_word_alloc(m, line, la, p))
1390 return(0);
1391 continue;
1392 }
1393
1394 if ( ! mdoc_macro(m, ntok, line, la, pos, buf))
1395 return(0);
1396 break;
1397 }
1398
1399 /* Clean-up to leave in a consistent state. */
1400
1401 if (NULL == head) {
1402 if ( ! mdoc_head_alloc(m, line, ppos, tok))
1403 return(0);
1404 head = m->last;
1405 }
1406
1407 if (NULL == body) {
1408 if ( ! rew_sub(MDOC_HEAD, m, tok, line, ppos))
1409 return(0);
1410 if ( ! mdoc_body_alloc(m, line, ppos, tok))
1411 return(0);
1412 body = m->last;
1413 }
1414
1415 /* Standard appending of delimiters. */
1416
1417 if ( ! nl)
1418 return(1);
1419 return(append_delims(m, line, pos, buf));
1420 }
1421
1422
1423 /* ARGSUSED */
1424 static int
1425 in_line_argn(MACRO_PROT_ARGS)
1426 {
1427 int la, flushed, j, maxargs, nl;
1428 enum margserr ac;
1429 enum margverr av;
1430 struct mdoc_arg *arg;
1431 char *p;
1432 enum mdoct ntok;
1433
1434 nl = MDOC_NEWLINE & m->flags;
1435
1436 /*
1437 * A line macro that has a fixed number of arguments (maxargs).
1438 * Only open the scope once the first non-leading-punctuation is
1439 * found (unless MDOC_IGNDELIM is noted, like in `Pf'), then
1440 * keep it open until the maximum number of arguments are
1441 * exhausted.
1442 */
1443
1444 switch (tok) {
1445 case (MDOC_Ap):
1446 /* FALLTHROUGH */
1447 case (MDOC_No):
1448 /* FALLTHROUGH */
1449 case (MDOC_Ns):
1450 /* FALLTHROUGH */
1451 case (MDOC_Ux):
1452 maxargs = 0;
1453 break;
1454 case (MDOC_Xr):
1455 maxargs = 2;
1456 break;
1457 default:
1458 maxargs = 1;
1459 break;
1460 }
1461
1462 for (arg = NULL; ; ) {
1463 la = *pos;
1464 av = mdoc_argv(m, line, tok, &arg, pos, buf);
1465
1466 if (ARGV_WORD == av) {
1467 *pos = la;
1468 break;
1469 }
1470
1471 if (ARGV_EOLN == av)
1472 break;
1473 if (ARGV_ARG == av)
1474 continue;
1475
1476 mdoc_argv_free(arg);
1477 return(0);
1478 }
1479
1480 for (flushed = j = 0; ; ) {
1481 la = *pos;
1482 ac = mdoc_args(m, line, pos, buf, tok, &p);
1483
1484 if (ARGS_ERROR == ac)
1485 return(0);
1486 if (ARGS_PUNCT == ac)
1487 break;
1488 if (ARGS_EOLN == ac)
1489 break;
1490
1491 if ( ! (MDOC_IGNDELIM & mdoc_macros[tok].flags) &&
1492 ARGS_QWORD != ac &&
1493 0 == j && DELIM_OPEN == mdoc_isdelim(p)) {
1494 if ( ! mdoc_word_alloc(m, line, la, p))
1495 return(0);
1496 continue;
1497 } else if (0 == j)
1498 if ( ! mdoc_elem_alloc(m, line, la, tok, arg))
1499 return(0);
1500
1501 if (j == maxargs && ! flushed) {
1502 if ( ! rew_elem(m, tok))
1503 return(0);
1504 flushed = 1;
1505 }
1506
1507 ntok = ARGS_QWORD == ac ? MDOC_MAX : lookup(tok, p);
1508
1509 if (MDOC_MAX != ntok) {
1510 if ( ! flushed && ! rew_elem(m, tok))
1511 return(0);
1512 flushed = 1;
1513 if ( ! mdoc_macro(m, ntok, line, la, pos, buf))
1514 return(0);
1515 j++;
1516 break;
1517 }
1518
1519 if ( ! (MDOC_IGNDELIM & mdoc_macros[tok].flags) &&
1520 ARGS_QWORD != ac &&
1521 ! flushed &&
1522 DELIM_NONE != mdoc_isdelim(p)) {
1523 if ( ! rew_elem(m, tok))
1524 return(0);
1525 flushed = 1;
1526 }
1527
1528 /*
1529 * XXX: this is a hack to work around groff's ugliness
1530 * as regards `Xr' and extraneous arguments. It should
1531 * ideally be deprecated behaviour, but because this is
1532 * code is no here, it's unlikely to be removed.
1533 */
1534
1535 #ifdef __OpenBSD__
1536 if (MDOC_Xr == tok && j == maxargs) {
1537 if ( ! mdoc_elem_alloc(m, line, la, MDOC_Ns, NULL))
1538 return(0);
1539 if ( ! rew_elem(m, MDOC_Ns))
1540 return(0);
1541 }
1542 #endif
1543
1544 if ( ! mdoc_word_alloc(m, line, la, p))
1545 return(0);
1546 j++;
1547 }
1548
1549 if (0 == j && ! mdoc_elem_alloc(m, line, la, tok, arg))
1550 return(0);
1551
1552 /* Close out in a consistent state. */
1553
1554 if ( ! flushed && ! rew_elem(m, tok))
1555 return(0);
1556 if ( ! nl)
1557 return(1);
1558 return(append_delims(m, line, pos, buf));
1559 }
1560
1561
1562 static int
1563 in_line_eoln(MACRO_PROT_ARGS)
1564 {
1565 int la;
1566 enum margserr ac;
1567 enum margverr av;
1568 struct mdoc_arg *arg;
1569 char *p;
1570 enum mdoct ntok;
1571
1572 assert( ! (MDOC_PARSED & mdoc_macros[tok].flags));
1573
1574 if (tok == MDOC_Pp)
1575 rew_sub(MDOC_BLOCK, m, MDOC_Nm, line, ppos);
1576
1577 /* Parse macro arguments. */
1578
1579 for (arg = NULL; ; ) {
1580 la = *pos;
1581 av = mdoc_argv(m, line, tok, &arg, pos, buf);
1582
1583 if (ARGV_WORD == av) {
1584 *pos = la;
1585 break;
1586 }
1587 if (ARGV_EOLN == av)
1588 break;
1589 if (ARGV_ARG == av)
1590 continue;
1591
1592 mdoc_argv_free(arg);
1593 return(0);
1594 }
1595
1596 /* Open element scope. */
1597
1598 if ( ! mdoc_elem_alloc(m, line, ppos, tok, arg))
1599 return(0);
1600
1601 /* Parse argument terms. */
1602
1603 for (;;) {
1604 la = *pos;
1605 ac = mdoc_args(m, line, pos, buf, tok, &p);
1606
1607 if (ARGS_ERROR == ac)
1608 return(0);
1609 if (ARGS_EOLN == ac)
1610 break;
1611
1612 ntok = ARGS_QWORD == ac ? MDOC_MAX : lookup(tok, p);
1613
1614 if (MDOC_MAX == ntok) {
1615 if ( ! mdoc_word_alloc(m, line, la, p))
1616 return(0);
1617 continue;
1618 }
1619
1620 if ( ! rew_elem(m, tok))
1621 return(0);
1622 return(mdoc_macro(m, ntok, line, la, pos, buf));
1623 }
1624
1625 /* Close out (no delimiters). */
1626
1627 return(rew_elem(m, tok));
1628 }
1629
1630
1631 /* ARGSUSED */
1632 static int
1633 ctx_synopsis(MACRO_PROT_ARGS)
1634 {
1635 int nl;
1636
1637 nl = MDOC_NEWLINE & m->flags;
1638
1639 /* If we're not in the SYNOPSIS, go straight to in-line. */
1640 if ( ! (MDOC_SYNOPSIS & m->flags))
1641 return(in_line(m, tok, line, ppos, pos, buf));
1642
1643 /* If we're a nested call, same place. */
1644 if ( ! nl)
1645 return(in_line(m, tok, line, ppos, pos, buf));
1646
1647 /*
1648 * XXX: this will open a block scope; however, if later we end
1649 * up formatting the block scope, then child nodes will inherit
1650 * the formatting. Be careful.
1651 */
1652 if (MDOC_Nm == tok)
1653 return(blk_full(m, tok, line, ppos, pos, buf));
1654 assert(MDOC_Vt == tok);
1655 return(blk_part_imp(m, tok, line, ppos, pos, buf));
1656 }
1657
1658
1659 /* ARGSUSED */
1660 static int
1661 obsolete(MACRO_PROT_ARGS)
1662 {
1663
1664 return(mdoc_pmsg(m, line, ppos, MANDOCERR_MACROOBS));
1665 }
1666
1667
1668 /*
1669 * Phrases occur within `Bl -column' entries, separated by `Ta' or tabs.
1670 * They're unusual because they're basically free-form text until a
1671 * macro is encountered.
1672 */
1673 static int
1674 phrase(struct mdoc *m, int line, int ppos, char *buf)
1675 {
1676 int la, pos;
1677 enum margserr ac;
1678 enum mdoct ntok;
1679 char *p;
1680
1681 for (pos = ppos; ; ) {
1682 la = pos;
1683
1684 ac = mdoc_zargs(m, line, &pos, buf, 0, &p);
1685
1686 if (ARGS_ERROR == ac)
1687 return(0);
1688 if (ARGS_EOLN == ac)
1689 break;
1690
1691 ntok = ARGS_QWORD == ac ? MDOC_MAX : lookup_raw(p);
1692
1693 if (MDOC_MAX == ntok) {
1694 if ( ! mdoc_word_alloc(m, line, la, p))
1695 return(0);
1696 continue;
1697 }
1698
1699 if ( ! mdoc_macro(m, ntok, line, la, &pos, buf))
1700 return(0);
1701 return(append_delims(m, line, &pos, buf));
1702 }
1703
1704 return(1);
1705 }
1706
1707
1708 /* ARGSUSED */
1709 static int
1710 phrase_ta(MACRO_PROT_ARGS)
1711 {
1712 int la;
1713 enum mdoct ntok;
1714 enum margserr ac;
1715 char *p;
1716
1717 /*
1718 * FIXME: this is overly restrictive: if the `Ta' is unexpected,
1719 * it should simply error out with ARGSLOST.
1720 */
1721
1722 if ( ! rew_sub(MDOC_BODY, m, MDOC_It, line, ppos))
1723 return(0);
1724 if ( ! mdoc_body_alloc(m, line, ppos, MDOC_It))
1725 return(0);
1726
1727 for (;;) {
1728 la = *pos;
1729 ac = mdoc_zargs(m, line, pos, buf, 0, &p);
1730
1731 if (ARGS_ERROR == ac)
1732 return(0);
1733 if (ARGS_EOLN == ac)
1734 break;
1735
1736 ntok = ARGS_QWORD == ac ? MDOC_MAX : lookup_raw(p);
1737
1738 if (MDOC_MAX == ntok) {
1739 if ( ! mdoc_word_alloc(m, line, la, p))
1740 return(0);
1741 continue;
1742 }
1743
1744 if ( ! mdoc_macro(m, ntok, line, la, pos, buf))
1745 return(0);
1746 return(append_delims(m, line, pos, buf));
1747 }
1748
1749 return(1);
1750 }
A git conversion mirror of mandoc.bsd.lv