1 /* $Id: mdoc_macro.c,v 1.125 2013/12/24 20:45:27 schwarze Exp $ */
2 /*
3 * Copyright (c) 2008-2012 Kristaps Dzonsons <kristaps@bsd.lv>
4 * Copyright (c) 2010, 2012, 2013 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 "mdoc.h"
30 #include "mandoc.h"
31 #include "libmdoc.h"
32 #include "libmandoc.h"
33
34 enum rew { /* see rew_dohalt() */
35 REWIND_NONE,
36 REWIND_THIS,
37 REWIND_MORE,
38 REWIND_FORCE,
39 REWIND_LATER,
40 REWIND_ERROR
41 };
42
43 static int blk_full(MACRO_PROT_ARGS);
44 static int blk_exp_close(MACRO_PROT_ARGS);
45 static int blk_part_exp(MACRO_PROT_ARGS);
46 static int blk_part_imp(MACRO_PROT_ARGS);
47 static int ctx_synopsis(MACRO_PROT_ARGS);
48 static int in_line_eoln(MACRO_PROT_ARGS);
49 static int in_line_argn(MACRO_PROT_ARGS);
50 static int in_line(MACRO_PROT_ARGS);
51 static int obsolete(MACRO_PROT_ARGS);
52 static int phrase_ta(MACRO_PROT_ARGS);
53
54 static int dword(struct mdoc *, int, int, const char *,
55 enum mdelim, int);
56 static int append_delims(struct mdoc *,
57 int, int *, char *);
58 static enum mdoct lookup(enum mdoct, const char *);
59 static enum mdoct lookup_raw(const char *);
60 static int make_pending(struct mdoc_node *, enum mdoct,
61 struct mdoc *, int, int);
62 static int phrase(struct mdoc *, int, int, char *);
63 static enum mdoct rew_alt(enum mdoct);
64 static enum rew rew_dohalt(enum mdoct, enum mdoc_type,
65 const struct mdoc_node *);
66 static int rew_elem(struct mdoc *, enum mdoct);
67 static int rew_last(struct mdoc *,
68 const struct mdoc_node *);
69 static int rew_sub(enum mdoc_type, struct mdoc *,
70 enum mdoct, int, int);
71
72 const struct mdoc_macro __mdoc_macros[MDOC_MAX] = {
73 { in_line_argn, MDOC_CALLABLE | MDOC_PARSED | MDOC_JOIN }, /* Ap */
74 { in_line_eoln, MDOC_PROLOGUE }, /* Dd */
75 { in_line_eoln, MDOC_PROLOGUE }, /* Dt */
76 { in_line_eoln, MDOC_PROLOGUE }, /* Os */
77 { blk_full, MDOC_PARSED | MDOC_JOIN }, /* Sh */
78 { blk_full, MDOC_PARSED | MDOC_JOIN }, /* Ss */
79 { in_line_eoln, 0 }, /* Pp */
80 { blk_part_imp, MDOC_PARSED | MDOC_JOIN }, /* D1 */
81 { blk_part_imp, MDOC_PARSED | MDOC_JOIN }, /* Dl */
82 { blk_full, MDOC_EXPLICIT }, /* Bd */
83 { blk_exp_close, MDOC_EXPLICIT | MDOC_JOIN }, /* Ed */
84 { blk_full, MDOC_EXPLICIT }, /* Bl */
85 { blk_exp_close, MDOC_EXPLICIT | MDOC_JOIN }, /* El */
86 { blk_full, MDOC_PARSED | MDOC_JOIN }, /* It */
87 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Ad */
88 { in_line, MDOC_CALLABLE | MDOC_PARSED | MDOC_JOIN }, /* An */
89 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Ar */
90 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Cd */
91 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Cm */
92 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Dv */
93 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Er */
94 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Ev */
95 { in_line_eoln, 0 }, /* Ex */
96 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Fa */
97 { in_line_eoln, 0 }, /* Fd */
98 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Fl */
99 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Fn */
100 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Ft */
101 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Ic */
102 { in_line_argn, MDOC_CALLABLE | MDOC_PARSED }, /* In */
103 { in_line, MDOC_CALLABLE | MDOC_PARSED | MDOC_JOIN }, /* Li */
104 { blk_full, MDOC_JOIN }, /* Nd */
105 { ctx_synopsis, MDOC_CALLABLE | MDOC_PARSED }, /* Nm */
106 { blk_part_imp, MDOC_CALLABLE | MDOC_PARSED }, /* Op */
107 { obsolete, 0 }, /* Ot */
108 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Pa */
109 { in_line_eoln, 0 }, /* Rv */
110 { in_line_argn, MDOC_CALLABLE | MDOC_PARSED }, /* St */
111 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Va */
112 { ctx_synopsis, MDOC_CALLABLE | MDOC_PARSED }, /* Vt */
113 { in_line_argn, MDOC_CALLABLE | MDOC_PARSED }, /* Xr */
114 { in_line_eoln, MDOC_JOIN }, /* %A */
115 { in_line_eoln, MDOC_JOIN }, /* %B */
116 { in_line_eoln, MDOC_JOIN }, /* %D */
117 { in_line_eoln, MDOC_JOIN }, /* %I */
118 { in_line_eoln, MDOC_JOIN }, /* %J */
119 { in_line_eoln, 0 }, /* %N */
120 { in_line_eoln, MDOC_JOIN }, /* %O */
121 { in_line_eoln, 0 }, /* %P */
122 { in_line_eoln, MDOC_JOIN }, /* %R */
123 { in_line_eoln, MDOC_JOIN }, /* %T */
124 { in_line_eoln, 0 }, /* %V */
125 { blk_exp_close, MDOC_CALLABLE | MDOC_PARSED |
126 MDOC_EXPLICIT | MDOC_JOIN }, /* Ac */
127 { blk_part_exp, MDOC_CALLABLE | MDOC_PARSED |
128 MDOC_EXPLICIT | MDOC_JOIN }, /* Ao */
129 { blk_part_imp, MDOC_CALLABLE | MDOC_PARSED | MDOC_JOIN }, /* Aq */
130 { in_line_argn, MDOC_CALLABLE | MDOC_PARSED }, /* At */
131 { blk_exp_close, MDOC_CALLABLE | MDOC_PARSED |
132 MDOC_EXPLICIT | MDOC_JOIN }, /* Bc */
133 { blk_full, MDOC_EXPLICIT }, /* Bf */
134 { blk_part_exp, MDOC_CALLABLE | MDOC_PARSED |
135 MDOC_EXPLICIT | MDOC_JOIN }, /* Bo */
136 { blk_part_imp, MDOC_CALLABLE | MDOC_PARSED | MDOC_JOIN }, /* Bq */
137 { in_line_argn, MDOC_CALLABLE | MDOC_PARSED }, /* Bsx */
138 { in_line_argn, MDOC_CALLABLE | MDOC_PARSED }, /* Bx */
139 { in_line_eoln, 0 }, /* Db */
140 { blk_exp_close, MDOC_CALLABLE | MDOC_PARSED |
141 MDOC_EXPLICIT | MDOC_JOIN }, /* Dc */
142 { blk_part_exp, MDOC_CALLABLE | MDOC_PARSED |
143 MDOC_EXPLICIT | MDOC_JOIN }, /* Do */
144 { blk_part_imp, MDOC_CALLABLE | MDOC_PARSED | MDOC_JOIN }, /* Dq */
145 { blk_exp_close, MDOC_CALLABLE | MDOC_PARSED | MDOC_EXPLICIT }, /* Ec */
146 { blk_exp_close, MDOC_EXPLICIT | MDOC_JOIN }, /* Ef */
147 { in_line, MDOC_CALLABLE | MDOC_PARSED | MDOC_JOIN }, /* Em */
148 { blk_part_exp, MDOC_CALLABLE | MDOC_PARSED | MDOC_EXPLICIT }, /* Eo */
149 { in_line_argn, MDOC_CALLABLE | MDOC_PARSED }, /* Fx */
150 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Ms */
151 { in_line_argn, MDOC_CALLABLE | MDOC_PARSED |
152 MDOC_IGNDELIM | MDOC_JOIN }, /* No */
153 { in_line_argn, MDOC_CALLABLE | MDOC_PARSED |
154 MDOC_IGNDELIM | MDOC_JOIN }, /* Ns */
155 { in_line_argn, MDOC_CALLABLE | MDOC_PARSED }, /* Nx */
156 { in_line_argn, MDOC_CALLABLE | MDOC_PARSED }, /* Ox */
157 { blk_exp_close, MDOC_CALLABLE | MDOC_PARSED |
158 MDOC_EXPLICIT | MDOC_JOIN }, /* Pc */
159 { in_line_argn, MDOC_CALLABLE | MDOC_PARSED | MDOC_IGNDELIM }, /* Pf */
160 { blk_part_exp, MDOC_CALLABLE | MDOC_PARSED |
161 MDOC_EXPLICIT | MDOC_JOIN }, /* Po */
162 { blk_part_imp, MDOC_CALLABLE | MDOC_PARSED | MDOC_JOIN }, /* Pq */
163 { blk_exp_close, MDOC_CALLABLE | MDOC_PARSED |
164 MDOC_EXPLICIT | MDOC_JOIN }, /* Qc */
165 { blk_part_imp, MDOC_CALLABLE | MDOC_PARSED | MDOC_JOIN }, /* Ql */
166 { blk_part_exp, MDOC_CALLABLE | MDOC_PARSED |
167 MDOC_EXPLICIT | MDOC_JOIN }, /* Qo */
168 { blk_part_imp, MDOC_CALLABLE | MDOC_PARSED | MDOC_JOIN }, /* Qq */
169 { blk_exp_close, MDOC_EXPLICIT | MDOC_JOIN }, /* Re */
170 { blk_full, MDOC_EXPLICIT }, /* Rs */
171 { blk_exp_close, MDOC_CALLABLE | MDOC_PARSED |
172 MDOC_EXPLICIT | MDOC_JOIN }, /* Sc */
173 { blk_part_exp, MDOC_CALLABLE | MDOC_PARSED |
174 MDOC_EXPLICIT | MDOC_JOIN }, /* So */
175 { blk_part_imp, MDOC_CALLABLE | MDOC_PARSED | MDOC_JOIN }, /* Sq */
176 { in_line_eoln, 0 }, /* Sm */
177 { in_line, MDOC_CALLABLE | MDOC_PARSED | MDOC_JOIN }, /* Sx */
178 { in_line, MDOC_CALLABLE | MDOC_PARSED | MDOC_JOIN }, /* Sy */
179 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Tn */
180 { in_line_argn, MDOC_CALLABLE | MDOC_PARSED | MDOC_JOIN }, /* Ux */
181 { blk_exp_close, MDOC_EXPLICIT | MDOC_CALLABLE | MDOC_PARSED }, /* Xc */
182 { blk_part_exp, MDOC_CALLABLE | MDOC_PARSED | MDOC_EXPLICIT }, /* Xo */
183 { blk_full, MDOC_EXPLICIT | MDOC_CALLABLE }, /* Fo */
184 { blk_exp_close, MDOC_CALLABLE | MDOC_PARSED |
185 MDOC_EXPLICIT | MDOC_JOIN }, /* Fc */
186 { blk_part_exp, MDOC_CALLABLE | MDOC_PARSED |
187 MDOC_EXPLICIT | MDOC_JOIN }, /* Oo */
188 { blk_exp_close, MDOC_CALLABLE | MDOC_PARSED |
189 MDOC_EXPLICIT | MDOC_JOIN }, /* Oc */
190 { blk_full, MDOC_EXPLICIT }, /* Bk */
191 { blk_exp_close, MDOC_EXPLICIT | MDOC_JOIN }, /* Ek */
192 { in_line_eoln, 0 }, /* Bt */
193 { in_line_eoln, 0 }, /* Hf */
194 { obsolete, 0 }, /* Fr */
195 { in_line_eoln, 0 }, /* Ud */
196 { in_line, 0 }, /* Lb */
197 { in_line_eoln, 0 }, /* Lp */
198 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Lk */
199 { in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Mt */
200 { blk_part_imp, MDOC_CALLABLE | MDOC_PARSED | MDOC_JOIN }, /* Brq */
201 { blk_part_exp, MDOC_CALLABLE | MDOC_PARSED |
202 MDOC_EXPLICIT | MDOC_JOIN }, /* Bro */
203 { blk_exp_close, MDOC_CALLABLE | MDOC_PARSED |
204 MDOC_EXPLICIT | MDOC_JOIN }, /* Brc */
205 { in_line_eoln, MDOC_JOIN }, /* %C */
206 { obsolete, 0 }, /* Es */
207 { obsolete, 0 }, /* En */
208 { in_line_argn, MDOC_CALLABLE | MDOC_PARSED }, /* Dx */
209 { in_line_eoln, MDOC_JOIN }, /* %Q */
210 { in_line_eoln, 0 }, /* br */
211 { in_line_eoln, 0 }, /* sp */
212 { in_line_eoln, 0 }, /* %U */
213 { phrase_ta, MDOC_CALLABLE | MDOC_PARSED | MDOC_JOIN }, /* Ta */
214 };
215
216 const struct mdoc_macro * const mdoc_macros = __mdoc_macros;
217
218
219 /*
220 * This is called at the end of parsing. It must traverse up the tree,
221 * closing out open [implicit] scopes. Obviously, open explicit scopes
222 * are errors.
223 */
224 int
mdoc_macroend(struct mdoc * mdoc)225 mdoc_macroend(struct mdoc *mdoc)
226 {
227 struct mdoc_node *n;
228
229 /* Scan for open explicit scopes. */
230
231 n = MDOC_VALID & mdoc->last->flags ?
232 mdoc->last->parent : mdoc->last;
233
234 for ( ; n; n = n->parent)
235 if (MDOC_BLOCK == n->type &&
236 MDOC_EXPLICIT & mdoc_macros[n->tok].flags)
237 mdoc_nmsg(mdoc, n, MANDOCERR_SCOPEEXIT);
238
239 /* Rewind to the first. */
240
241 return(rew_last(mdoc, mdoc->first));
242 }
243
244
245 /*
246 * Look up a macro from within a subsequent context.
247 */
248 static enum mdoct
lookup(enum mdoct from,const char * p)249 lookup(enum mdoct from, const char *p)
250 {
251
252 if ( ! (MDOC_PARSED & mdoc_macros[from].flags))
253 return(MDOC_MAX);
254 return(lookup_raw(p));
255 }
256
257
258 /*
259 * Lookup a macro following the initial line macro.
260 */
261 static enum mdoct
lookup_raw(const char * p)262 lookup_raw(const char *p)
263 {
264 enum mdoct res;
265
266 if (MDOC_MAX == (res = mdoc_hash_find(p)))
267 return(MDOC_MAX);
268 if (MDOC_CALLABLE & mdoc_macros[res].flags)
269 return(res);
270 return(MDOC_MAX);
271 }
272
273
274 static int
rew_last(struct mdoc * mdoc,const struct mdoc_node * to)275 rew_last(struct mdoc *mdoc, const struct mdoc_node *to)
276 {
277 struct mdoc_node *n, *np;
278
279 assert(to);
280 mdoc->next = MDOC_NEXT_SIBLING;
281
282 /* LINTED */
283 while (mdoc->last != to) {
284 /*
285 * Save the parent here, because we may delete the
286 * mdoc->last node in the post-validation phase and reset
287 * it to mdoc->last->parent, causing a step in the closing
288 * out to be lost.
289 */
290 np = mdoc->last->parent;
291 if ( ! mdoc_valid_post(mdoc))
292 return(0);
293 n = mdoc->last;
294 mdoc->last = np;
295 assert(mdoc->last);
296 mdoc->last->last = n;
297 }
298
299 return(mdoc_valid_post(mdoc));
300 }
301
302
303 /*
304 * For a block closing macro, return the corresponding opening one.
305 * Otherwise, return the macro itself.
306 */
307 static enum mdoct
rew_alt(enum mdoct tok)308 rew_alt(enum mdoct tok)
309 {
310 switch (tok) {
311 case (MDOC_Ac):
312 return(MDOC_Ao);
313 case (MDOC_Bc):
314 return(MDOC_Bo);
315 case (MDOC_Brc):
316 return(MDOC_Bro);
317 case (MDOC_Dc):
318 return(MDOC_Do);
319 case (MDOC_Ec):
320 return(MDOC_Eo);
321 case (MDOC_Ed):
322 return(MDOC_Bd);
323 case (MDOC_Ef):
324 return(MDOC_Bf);
325 case (MDOC_Ek):
326 return(MDOC_Bk);
327 case (MDOC_El):
328 return(MDOC_Bl);
329 case (MDOC_Fc):
330 return(MDOC_Fo);
331 case (MDOC_Oc):
332 return(MDOC_Oo);
333 case (MDOC_Pc):
334 return(MDOC_Po);
335 case (MDOC_Qc):
336 return(MDOC_Qo);
337 case (MDOC_Re):
338 return(MDOC_Rs);
339 case (MDOC_Sc):
340 return(MDOC_So);
341 case (MDOC_Xc):
342 return(MDOC_Xo);
343 default:
344 return(tok);
345 }
346 /* NOTREACHED */
347 }
348
349
350 /*
351 * Rewinding to tok, how do we have to handle *p?
352 * REWIND_NONE: *p would delimit tok, but no tok scope is open
353 * inside *p, so there is no need to rewind anything at all.
354 * REWIND_THIS: *p matches tok, so rewind *p and nothing else.
355 * REWIND_MORE: *p is implicit, rewind it and keep searching for tok.
356 * REWIND_FORCE: *p is explicit, but tok is full, force rewinding *p.
357 * REWIND_LATER: *p is explicit and still open, postpone rewinding.
358 * REWIND_ERROR: No tok block is open at all.
359 */
360 static enum rew
rew_dohalt(enum mdoct tok,enum mdoc_type type,const struct mdoc_node * p)361 rew_dohalt(enum mdoct tok, enum mdoc_type type,
362 const struct mdoc_node *p)
363 {
364
365 /*
366 * No matching token, no delimiting block, no broken block.
367 * This can happen when full implicit macros are called for
368 * the first time but try to rewind their previous
369 * instance anyway.
370 */
371 if (MDOC_ROOT == p->type)
372 return(MDOC_BLOCK == type &&
373 MDOC_EXPLICIT & mdoc_macros[tok].flags ?
374 REWIND_ERROR : REWIND_NONE);
375
376 /*
377 * When starting to rewind, skip plain text
378 * and nodes that have already been rewound.
379 */
380 if (MDOC_TEXT == p->type || MDOC_VALID & p->flags)
381 return(REWIND_MORE);
382
383 /*
384 * The easiest case: Found a matching token.
385 * This applies to both blocks and elements.
386 */
387 tok = rew_alt(tok);
388 if (tok == p->tok)
389 return(p->end ? REWIND_NONE :
390 type == p->type ? REWIND_THIS : REWIND_MORE);
391
392 /*
393 * While elements do require rewinding for themselves,
394 * they never affect rewinding of other nodes.
395 */
396 if (MDOC_ELEM == p->type)
397 return(REWIND_MORE);
398
399 /*
400 * Blocks delimited by our target token get REWIND_MORE.
401 * Blocks delimiting our target token get REWIND_NONE.
402 */
403 switch (tok) {
404 case (MDOC_Bl):
405 if (MDOC_It == p->tok)
406 return(REWIND_MORE);
407 break;
408 case (MDOC_It):
409 if (MDOC_BODY == p->type && MDOC_Bl == p->tok)
410 return(REWIND_NONE);
411 break;
412 /*
413 * XXX Badly nested block handling still fails badly
414 * when one block is breaking two blocks of the same type.
415 * This is an incomplete and extremely ugly workaround,
416 * required to let the OpenBSD tree build.
417 */
418 case (MDOC_Oo):
419 if (MDOC_Op == p->tok)
420 return(REWIND_MORE);
421 break;
422 case (MDOC_Nm):
423 return(REWIND_NONE);
424 case (MDOC_Nd):
425 /* FALLTHROUGH */
426 case (MDOC_Ss):
427 if (MDOC_BODY == p->type && MDOC_Sh == p->tok)
428 return(REWIND_NONE);
429 /* FALLTHROUGH */
430 case (MDOC_Sh):
431 if (MDOC_Nd == p->tok || MDOC_Ss == p->tok ||
432 MDOC_Sh == p->tok)
433 return(REWIND_MORE);
434 break;
435 default:
436 break;
437 }
438
439 /*
440 * Default block rewinding rules.
441 * In particular, always skip block end markers,
442 * and let all blocks rewind Nm children.
443 */
444 if (ENDBODY_NOT != p->end || MDOC_Nm == p->tok ||
445 (MDOC_BLOCK == p->type &&
446 ! (MDOC_EXPLICIT & mdoc_macros[tok].flags)))
447 return(REWIND_MORE);
448
449 /*
450 * By default, closing out full blocks
451 * forces closing of broken explicit blocks,
452 * while closing out partial blocks
453 * allows delayed rewinding by default.
454 */
455 return (&blk_full == mdoc_macros[tok].fp ?
456 REWIND_FORCE : REWIND_LATER);
457 }
458
459
460 static int
rew_elem(struct mdoc * mdoc,enum mdoct tok)461 rew_elem(struct mdoc *mdoc, enum mdoct tok)
462 {
463 struct mdoc_node *n;
464
465 n = mdoc->last;
466 if (MDOC_ELEM != n->type)
467 n = n->parent;
468 assert(MDOC_ELEM == n->type);
469 assert(tok == n->tok);
470
471 return(rew_last(mdoc, n));
472 }
473
474
475 /*
476 * We are trying to close a block identified by tok,
477 * but the child block *broken is still open.
478 * Thus, postpone closing the tok block
479 * until the rew_sub call closing *broken.
480 */
481 static int
make_pending(struct mdoc_node * broken,enum mdoct tok,struct mdoc * mdoc,int line,int ppos)482 make_pending(struct mdoc_node *broken, enum mdoct tok,
483 struct mdoc *mdoc, int line, int ppos)
484 {
485 struct mdoc_node *breaker;
486
487 /*
488 * Iterate backwards, searching for the block matching tok,
489 * that is, the block breaking the *broken block.
490 */
491 for (breaker = broken->parent; breaker; breaker = breaker->parent) {
492
493 /*
494 * If the *broken block had already been broken before
495 * and we encounter its breaker, make the tok block
496 * pending on the inner breaker.
497 * Graphically, "[A breaker=[B broken=[C->B B] tok=A] C]"
498 * becomes "[A broken=[B [C->B B] tok=A] C]"
499 * and finally "[A [B->A [C->B B] A] C]".
500 */
501 if (breaker == broken->pending) {
502 broken = breaker;
503 continue;
504 }
505
506 if (REWIND_THIS != rew_dohalt(tok, MDOC_BLOCK, breaker))
507 continue;
508 if (MDOC_BODY == broken->type)
509 broken = broken->parent;
510
511 /*
512 * Found the breaker.
513 * If another, outer breaker is already pending on
514 * the *broken block, we must not clobber the link
515 * to the outer breaker, but make it pending on the
516 * new, now inner breaker.
517 * Graphically, "[A breaker=[B broken=[C->A A] tok=B] C]"
518 * becomes "[A breaker=[B->A broken=[C A] tok=B] C]"
519 * and finally "[A [B->A [C->B A] B] C]".
520 */
521 if (broken->pending) {
522 struct mdoc_node *taker;
523
524 /*
525 * If the breaker had also been broken before,
526 * it cannot take on the outer breaker itself,
527 * but must hand it on to its own breakers.
528 * Graphically, this is the following situation:
529 * "[A [B breaker=[C->B B] broken=[D->A A] tok=C] D]"
530 * "[A taker=[B->A breaker=[C->B B] [D->C A] C] D]"
531 */
532 taker = breaker;
533 while (taker->pending)
534 taker = taker->pending;
535 taker->pending = broken->pending;
536 }
537 broken->pending = breaker;
538 mandoc_vmsg(MANDOCERR_SCOPENEST, mdoc->parse, line, ppos,
539 "%s breaks %s", mdoc_macronames[tok],
540 mdoc_macronames[broken->tok]);
541 return(1);
542 }
543
544 /*
545 * Found no matching block for tok.
546 * Are you trying to close a block that is not open?
547 */
548 return(0);
549 }
550
551
552 static int
rew_sub(enum mdoc_type t,struct mdoc * mdoc,enum mdoct tok,int line,int ppos)553 rew_sub(enum mdoc_type t, struct mdoc *mdoc,
554 enum mdoct tok, int line, int ppos)
555 {
556 struct mdoc_node *n;
557
558 n = mdoc->last;
559 while (n) {
560 switch (rew_dohalt(tok, t, n)) {
561 case (REWIND_NONE):
562 return(1);
563 case (REWIND_THIS):
564 n->lastline = line -
565 (MDOC_NEWLINE & mdoc->flags &&
566 ! (MDOC_EXPLICIT & mdoc_macros[tok].flags));
567 break;
568 case (REWIND_FORCE):
569 mandoc_vmsg(MANDOCERR_SCOPEBROKEN, mdoc->parse,
570 line, ppos, "%s breaks %s",
571 mdoc_macronames[tok],
572 mdoc_macronames[n->tok]);
573 /* FALLTHROUGH */
574 case (REWIND_MORE):
575 n->lastline = line -
576 (MDOC_NEWLINE & mdoc->flags ? 1 : 0);
577 n = n->parent;
578 continue;
579 case (REWIND_LATER):
580 if (make_pending(n, tok, mdoc, line, ppos) ||
581 MDOC_BLOCK != t)
582 return(1);
583 /* FALLTHROUGH */
584 case (REWIND_ERROR):
585 mdoc_pmsg(mdoc, line, ppos, MANDOCERR_NOSCOPE);
586 return(1);
587 }
588 break;
589 }
590
591 assert(n);
592 if ( ! rew_last(mdoc, n))
593 return(0);
594
595 /*
596 * The current block extends an enclosing block.
597 * Now that the current block ends, close the enclosing block, too.
598 */
599 while (NULL != (n = n->pending)) {
600 if ( ! rew_last(mdoc, n))
601 return(0);
602 if (MDOC_HEAD == n->type &&
603 ! mdoc_body_alloc(mdoc, n->line, n->pos, n->tok))
604 return(0);
605 }
606
607 return(1);
608 }
609
610 /*
611 * Allocate a word and check whether it's punctuation or not.
612 * Punctuation consists of those tokens found in mdoc_isdelim().
613 */
614 static int
dword(struct mdoc * mdoc,int line,int col,const char * p,enum mdelim d,int may_append)615 dword(struct mdoc *mdoc, int line, int col, const char *p,
616 enum mdelim d, int may_append)
617 {
618
619 if (DELIM_MAX == d)
620 d = mdoc_isdelim(p);
621
622 if (may_append &&
623 ! ((MDOC_SYNOPSIS | MDOC_KEEP | MDOC_SMOFF) & mdoc->flags) &&
624 DELIM_NONE == d && MDOC_TEXT == mdoc->last->type &&
625 DELIM_NONE == mdoc_isdelim(mdoc->last->string)) {
626 mdoc_word_append(mdoc, p);
627 return(1);
628 }
629
630 if ( ! mdoc_word_alloc(mdoc, line, col, p))
631 return(0);
632
633 if (DELIM_OPEN == d)
634 mdoc->last->flags |= MDOC_DELIMO;
635
636 /*
637 * Closing delimiters only suppress the preceding space
638 * when they follow something, not when they start a new
639 * block or element, and not when they follow `No'.
640 *
641 * XXX Explicitly special-casing MDOC_No here feels
642 * like a layering violation. Find a better way
643 * and solve this in the code related to `No'!
644 */
645
646 else if (DELIM_CLOSE == d && mdoc->last->prev &&
647 mdoc->last->prev->tok != MDOC_No &&
648 mdoc->last->parent->tok != MDOC_Fd)
649 mdoc->last->flags |= MDOC_DELIMC;
650
651 return(1);
652 }
653
654 static int
append_delims(struct mdoc * mdoc,int line,int * pos,char * buf)655 append_delims(struct mdoc *mdoc, int line, int *pos, char *buf)
656 {
657 int la;
658 enum margserr ac;
659 char *p;
660
661 if ('\0' == buf[*pos])
662 return(1);
663
664 for (;;) {
665 la = *pos;
666 ac = mdoc_zargs(mdoc, line, pos, buf, &p);
667
668 if (ARGS_ERROR == ac)
669 return(0);
670 else if (ARGS_EOLN == ac)
671 break;
672
673 dword(mdoc, line, la, p, DELIM_MAX, 1);
674
675 /*
676 * If we encounter end-of-sentence symbols, then trigger
677 * the double-space.
678 *
679 * XXX: it's easy to allow this to propagate outward to
680 * the last symbol, such that `. )' will cause the
681 * correct double-spacing. However, (1) groff isn't
682 * smart enough to do this and (2) it would require
683 * knowing which symbols break this behaviour, for
684 * example, `. ;' shouldn't propagate the double-space.
685 */
686 if (mandoc_eos(p, strlen(p), 0))
687 mdoc->last->flags |= MDOC_EOS;
688 }
689
690 return(1);
691 }
692
693
694 /*
695 * Close out block partial/full explicit.
696 */
697 static int
blk_exp_close(MACRO_PROT_ARGS)698 blk_exp_close(MACRO_PROT_ARGS)
699 {
700 struct mdoc_node *body; /* Our own body. */
701 struct mdoc_node *later; /* A sub-block starting later. */
702 struct mdoc_node *n; /* For searching backwards. */
703
704 int j, lastarg, maxargs, flushed, nl;
705 enum margserr ac;
706 enum mdoct atok, ntok;
707 char *p;
708
709 nl = MDOC_NEWLINE & mdoc->flags;
710
711 switch (tok) {
712 case (MDOC_Ec):
713 maxargs = 1;
714 break;
715 case (MDOC_Ek):
716 mdoc->flags &= ~MDOC_KEEP;
717 default:
718 maxargs = 0;
719 break;
720 }
721
722 /*
723 * Search backwards for beginnings of blocks,
724 * both of our own and of pending sub-blocks.
725 */
726 atok = rew_alt(tok);
727 body = later = NULL;
728 for (n = mdoc->last; n; n = n->parent) {
729 if (MDOC_VALID & n->flags)
730 continue;
731
732 /* Remember the start of our own body. */
733 if (MDOC_BODY == n->type && atok == n->tok) {
734 if (ENDBODY_NOT == n->end)
735 body = n;
736 continue;
737 }
738
739 if (MDOC_BLOCK != n->type || MDOC_Nm == n->tok)
740 continue;
741 if (atok == n->tok) {
742 assert(body);
743
744 /*
745 * Found the start of our own block.
746 * When there is no pending sub block,
747 * just proceed to closing out.
748 */
749 if (NULL == later)
750 break;
751
752 /*
753 * When there is a pending sub block,
754 * postpone closing out the current block
755 * until the rew_sub() closing out the sub-block.
756 */
757 make_pending(later, tok, mdoc, line, ppos);
758
759 /*
760 * Mark the place where the formatting - but not
761 * the scope - of the current block ends.
762 */
763 if ( ! mdoc_endbody_alloc(mdoc, line, ppos,
764 atok, body, ENDBODY_SPACE))
765 return(0);
766 break;
767 }
768
769 /*
770 * When finding an open sub block, remember the last
771 * open explicit block, or, in case there are only
772 * implicit ones, the first open implicit block.
773 */
774 if (later &&
775 MDOC_EXPLICIT & mdoc_macros[later->tok].flags)
776 continue;
777 if (MDOC_It != n->tok)
778 later = n;
779 }
780
781 if ( ! (MDOC_CALLABLE & mdoc_macros[tok].flags)) {
782 /* FIXME: do this in validate */
783 if (buf[*pos])
784 mdoc_pmsg(mdoc, line, ppos, MANDOCERR_ARGSLOST);
785
786 if ( ! rew_sub(MDOC_BODY, mdoc, tok, line, ppos))
787 return(0);
788 return(rew_sub(MDOC_BLOCK, mdoc, tok, line, ppos));
789 }
790
791 if ( ! rew_sub(MDOC_BODY, mdoc, tok, line, ppos))
792 return(0);
793
794 if (NULL == later && maxargs > 0)
795 if ( ! mdoc_tail_alloc(mdoc, line, ppos, rew_alt(tok)))
796 return(0);
797
798 for (flushed = j = 0; ; j++) {
799 lastarg = *pos;
800
801 if (j == maxargs && ! flushed) {
802 if ( ! rew_sub(MDOC_BLOCK, mdoc, tok, line, ppos))
803 return(0);
804 flushed = 1;
805 }
806
807 ac = mdoc_args(mdoc, line, pos, buf, tok, &p);
808
809 if (ARGS_ERROR == ac)
810 return(0);
811 if (ARGS_PUNCT == ac)
812 break;
813 if (ARGS_EOLN == ac)
814 break;
815
816 ntok = ARGS_QWORD == ac ? MDOC_MAX : lookup(tok, p);
817
818 if (MDOC_MAX == ntok) {
819 if ( ! dword(mdoc, line, lastarg, p, DELIM_MAX,
820 MDOC_JOIN & mdoc_macros[tok].flags))
821 return(0);
822 continue;
823 }
824
825 if ( ! flushed) {
826 if ( ! rew_sub(MDOC_BLOCK, mdoc, tok, line, ppos))
827 return(0);
828 flushed = 1;
829 }
830
831 mdoc->flags &= ~MDOC_NEWLINE;
832
833 if ( ! mdoc_macro(mdoc, ntok, line, lastarg, pos, buf))
834 return(0);
835 break;
836 }
837
838 if ( ! flushed && ! rew_sub(MDOC_BLOCK, mdoc, tok, line, ppos))
839 return(0);
840
841 if ( ! nl)
842 return(1);
843 return(append_delims(mdoc, line, pos, buf));
844 }
845
846
847 static int
in_line(MACRO_PROT_ARGS)848 in_line(MACRO_PROT_ARGS)
849 {
850 int la, scope, cnt, nc, nl;
851 enum margverr av;
852 enum mdoct ntok;
853 enum margserr ac;
854 enum mdelim d;
855 struct mdoc_arg *arg;
856 char *p;
857
858 nl = MDOC_NEWLINE & mdoc->flags;
859
860 /*
861 * Whether we allow ignored elements (those without content,
862 * usually because of reserved words) to squeak by.
863 */
864
865 switch (tok) {
866 case (MDOC_An):
867 /* FALLTHROUGH */
868 case (MDOC_Ar):
869 /* FALLTHROUGH */
870 case (MDOC_Fl):
871 /* FALLTHROUGH */
872 case (MDOC_Mt):
873 /* FALLTHROUGH */
874 case (MDOC_Nm):
875 /* FALLTHROUGH */
876 case (MDOC_Pa):
877 nc = 1;
878 break;
879 default:
880 nc = 0;
881 break;
882 }
883
884 for (arg = NULL;; ) {
885 la = *pos;
886 av = mdoc_argv(mdoc, line, tok, &arg, pos, buf);
887
888 if (ARGV_WORD == av) {
889 *pos = la;
890 break;
891 }
892 if (ARGV_EOLN == av)
893 break;
894 if (ARGV_ARG == av)
895 continue;
896
897 mdoc_argv_free(arg);
898 return(0);
899 }
900
901 for (cnt = scope = 0;; ) {
902 la = *pos;
903 ac = mdoc_args(mdoc, line, pos, buf, tok, &p);
904
905 if (ARGS_ERROR == ac)
906 return(0);
907 if (ARGS_EOLN == ac)
908 break;
909 if (ARGS_PUNCT == ac)
910 break;
911
912 ntok = ARGS_QWORD == ac ? MDOC_MAX : lookup(tok, p);
913
914 /*
915 * In this case, we've located a submacro and must
916 * execute it. Close out scope, if open. If no
917 * elements have been generated, either create one (nc)
918 * or raise a warning.
919 */
920
921 if (MDOC_MAX != ntok) {
922 if (scope && ! rew_elem(mdoc, tok))
923 return(0);
924 if (nc && 0 == cnt) {
925 if ( ! mdoc_elem_alloc(mdoc, line,
926 ppos, tok, arg))
927 return(0);
928 if ( ! rew_last(mdoc, mdoc->last))
929 return(0);
930 } else if ( ! nc && 0 == cnt) {
931 mdoc_argv_free(arg);
932 mdoc_pmsg(mdoc, line, ppos,
933 MANDOCERR_MACROEMPTY);
934 }
935
936 if ( ! mdoc_macro(mdoc, ntok, line, la, pos, buf))
937 return(0);
938 if ( ! nl)
939 return(1);
940 return(append_delims(mdoc, line, pos, buf));
941 }
942
943 /*
944 * Non-quote-enclosed punctuation. Set up our scope, if
945 * a word; rewind the scope, if a delimiter; then append
946 * the word.
947 */
948
949 d = ARGS_QWORD == ac ? DELIM_NONE : mdoc_isdelim(p);
950
951 if (DELIM_NONE != d) {
952 /*
953 * If we encounter closing punctuation, no word
954 * has been omitted, no scope is open, and we're
955 * allowed to have an empty element, then start
956 * a new scope. `Ar', `Fl', and `Li', only do
957 * this once per invocation. There may be more
958 * of these (all of them?).
959 */
960 if (0 == cnt && (nc || MDOC_Li == tok) &&
961 DELIM_CLOSE == d && ! scope) {
962 if ( ! mdoc_elem_alloc(mdoc, line,
963 ppos, tok, arg))
964 return(0);
965 if (MDOC_Ar == tok || MDOC_Li == tok ||
966 MDOC_Fl == tok)
967 cnt++;
968 scope = 1;
969 }
970 /*
971 * Close out our scope, if one is open, before
972 * any punctuation.
973 */
974 if (scope && ! rew_elem(mdoc, tok))
975 return(0);
976 scope = 0;
977 } else if ( ! scope) {
978 if ( ! mdoc_elem_alloc(mdoc, line, ppos, tok, arg))
979 return(0);
980 scope = 1;
981 }
982
983 if (DELIM_NONE == d)
984 cnt++;
985
986 if ( ! dword(mdoc, line, la, p, d,
987 MDOC_JOIN & mdoc_macros[tok].flags))
988 return(0);
989
990 /*
991 * `Fl' macros have their scope re-opened with each new
992 * word so that the `-' can be added to each one without
993 * having to parse out spaces.
994 */
995 if (scope && MDOC_Fl == tok) {
996 if ( ! rew_elem(mdoc, tok))
997 return(0);
998 scope = 0;
999 }
1000 }
1001
1002 if (scope && ! rew_elem(mdoc, tok))
1003 return(0);
1004
1005 /*
1006 * If no elements have been collected and we're allowed to have
1007 * empties (nc), open a scope and close it out. Otherwise,
1008 * raise a warning.
1009 */
1010
1011 if (nc && 0 == cnt) {
1012 if ( ! mdoc_elem_alloc(mdoc, line, ppos, tok, arg))
1013 return(0);
1014 if ( ! rew_last(mdoc, mdoc->last))
1015 return(0);
1016 } else if ( ! nc && 0 == cnt) {
1017 mdoc_argv_free(arg);
1018 mdoc_pmsg(mdoc, line, ppos, MANDOCERR_MACROEMPTY);
1019 }
1020
1021 if ( ! nl)
1022 return(1);
1023 return(append_delims(mdoc, line, pos, buf));
1024 }
1025
1026
1027 static int
blk_full(MACRO_PROT_ARGS)1028 blk_full(MACRO_PROT_ARGS)
1029 {
1030 int la, nl, nparsed;
1031 struct mdoc_arg *arg;
1032 struct mdoc_node *head; /* save of head macro */
1033 struct mdoc_node *body; /* save of body macro */
1034 struct mdoc_node *n;
1035 enum mdoc_type mtt;
1036 enum mdoct ntok;
1037 enum margserr ac, lac;
1038 enum margverr av;
1039 char *p;
1040
1041 nl = MDOC_NEWLINE & mdoc->flags;
1042
1043 /* Close out prior implicit scope. */
1044
1045 if ( ! (MDOC_EXPLICIT & mdoc_macros[tok].flags)) {
1046 if ( ! rew_sub(MDOC_BODY, mdoc, tok, line, ppos))
1047 return(0);
1048 if ( ! rew_sub(MDOC_BLOCK, mdoc, tok, line, ppos))
1049 return(0);
1050 }
1051
1052 /*
1053 * This routine accommodates implicitly- and explicitly-scoped
1054 * macro openings. Implicit ones first close out prior scope
1055 * (seen above). Delay opening the head until necessary to
1056 * allow leading punctuation to print. Special consideration
1057 * for `It -column', which has phrase-part syntax instead of
1058 * regular child nodes.
1059 */
1060
1061 for (arg = NULL;; ) {
1062 la = *pos;
1063 av = mdoc_argv(mdoc, line, tok, &arg, pos, buf);
1064
1065 if (ARGV_WORD == av) {
1066 *pos = la;
1067 break;
1068 }
1069
1070 if (ARGV_EOLN == av)
1071 break;
1072 if (ARGV_ARG == av)
1073 continue;
1074
1075 mdoc_argv_free(arg);
1076 return(0);
1077 }
1078
1079 if ( ! mdoc_block_alloc(mdoc, line, ppos, tok, arg))
1080 return(0);
1081
1082 head = body = NULL;
1083
1084 /*
1085 * Exception: Heads of `It' macros in `-diag' lists are not
1086 * parsed, even though `It' macros in general are parsed.
1087 */
1088 nparsed = MDOC_It == tok &&
1089 MDOC_Bl == mdoc->last->parent->tok &&
1090 LIST_diag == mdoc->last->parent->norm->Bl.type;
1091
1092 /*
1093 * The `Nd' macro has all arguments in its body: it's a hybrid
1094 * of block partial-explicit and full-implicit. Stupid.
1095 */
1096
1097 if (MDOC_Nd == tok) {
1098 if ( ! mdoc_head_alloc(mdoc, line, ppos, tok))
1099 return(0);
1100 head = mdoc->last;
1101 if ( ! rew_sub(MDOC_HEAD, mdoc, tok, line, ppos))
1102 return(0);
1103 if ( ! mdoc_body_alloc(mdoc, line, ppos, tok))
1104 return(0);
1105 body = mdoc->last;
1106 }
1107
1108 if (MDOC_Bk == tok)
1109 mdoc->flags |= MDOC_KEEP;
1110
1111 ac = ARGS_ERROR;
1112
1113 for ( ; ; ) {
1114 la = *pos;
1115 /* Initialise last-phrase-type with ARGS_PEND. */
1116 lac = ARGS_ERROR == ac ? ARGS_PEND : ac;
1117 ac = mdoc_args(mdoc, line, pos, buf, tok, &p);
1118
1119 if (ARGS_PUNCT == ac)
1120 break;
1121
1122 if (ARGS_ERROR == ac)
1123 return(0);
1124
1125 if (ARGS_EOLN == ac) {
1126 if (ARGS_PPHRASE != lac && ARGS_PHRASE != lac)
1127 break;
1128 /*
1129 * This is necessary: if the last token on a
1130 * line is a `Ta' or tab, then we'll get
1131 * ARGS_EOLN, so we must be smart enough to
1132 * reopen our scope if the last parse was a
1133 * phrase or partial phrase.
1134 */
1135 if ( ! rew_sub(MDOC_BODY, mdoc, tok, line, ppos))
1136 return(0);
1137 if ( ! mdoc_body_alloc(mdoc, line, ppos, tok))
1138 return(0);
1139 body = mdoc->last;
1140 break;
1141 }
1142
1143 /*
1144 * Emit leading punctuation (i.e., punctuation before
1145 * the MDOC_HEAD) for non-phrase types.
1146 */
1147
1148 if (NULL == head &&
1149 ARGS_PEND != ac &&
1150 ARGS_PHRASE != ac &&
1151 ARGS_PPHRASE != ac &&
1152 ARGS_QWORD != ac &&
1153 DELIM_OPEN == mdoc_isdelim(p)) {
1154 if ( ! dword(mdoc, line, la, p, DELIM_OPEN, 0))
1155 return(0);
1156 continue;
1157 }
1158
1159 /* Open a head if one hasn't been opened. */
1160
1161 if (NULL == head) {
1162 if ( ! mdoc_head_alloc(mdoc, line, ppos, tok))
1163 return(0);
1164 head = mdoc->last;
1165 }
1166
1167 if (ARGS_PHRASE == ac ||
1168 ARGS_PEND == ac ||
1169 ARGS_PPHRASE == ac) {
1170 /*
1171 * If we haven't opened a body yet, rewind the
1172 * head; if we have, rewind that instead.
1173 */
1174
1175 mtt = body ? MDOC_BODY : MDOC_HEAD;
1176 if ( ! rew_sub(mtt, mdoc, tok, line, ppos))
1177 return(0);
1178
1179 /* Then allocate our body context. */
1180
1181 if ( ! mdoc_body_alloc(mdoc, line, ppos, tok))
1182 return(0);
1183 body = mdoc->last;
1184
1185 /*
1186 * Process phrases: set whether we're in a
1187 * partial-phrase (this effects line handling)
1188 * then call down into the phrase parser.
1189 */
1190
1191 if (ARGS_PPHRASE == ac)
1192 mdoc->flags |= MDOC_PPHRASE;
1193 if (ARGS_PEND == ac && ARGS_PPHRASE == lac)
1194 mdoc->flags |= MDOC_PPHRASE;
1195
1196 if ( ! phrase(mdoc, line, la, buf))
1197 return(0);
1198
1199 mdoc->flags &= ~MDOC_PPHRASE;
1200 continue;
1201 }
1202
1203 ntok = nparsed || ARGS_QWORD == ac ?
1204 MDOC_MAX : lookup(tok, p);
1205
1206 if (MDOC_MAX == ntok) {
1207 if ( ! dword(mdoc, line, la, p, DELIM_MAX,
1208 MDOC_JOIN & mdoc_macros[tok].flags))
1209 return(0);
1210 continue;
1211 }
1212
1213 if ( ! mdoc_macro(mdoc, ntok, line, la, pos, buf))
1214 return(0);
1215 break;
1216 }
1217
1218 if (NULL == head) {
1219 if ( ! mdoc_head_alloc(mdoc, line, ppos, tok))
1220 return(0);
1221 head = mdoc->last;
1222 }
1223
1224 if (nl && ! append_delims(mdoc, line, pos, buf))
1225 return(0);
1226
1227 /* If we've already opened our body, exit now. */
1228
1229 if (NULL != body)
1230 goto out;
1231
1232 /*
1233 * If there is an open (i.e., unvalidated) sub-block requiring
1234 * explicit close-out, postpone switching the current block from
1235 * head to body until the rew_sub() call closing out that
1236 * sub-block.
1237 */
1238 for (n = mdoc->last; n && n != head; n = n->parent) {
1239 if (MDOC_BLOCK == n->type &&
1240 MDOC_EXPLICIT & mdoc_macros[n->tok].flags &&
1241 ! (MDOC_VALID & n->flags)) {
1242 n->pending = head;
1243 return(1);
1244 }
1245 }
1246
1247 /* Close out scopes to remain in a consistent state. */
1248
1249 if ( ! rew_sub(MDOC_HEAD, mdoc, tok, line, ppos))
1250 return(0);
1251 if ( ! mdoc_body_alloc(mdoc, line, ppos, tok))
1252 return(0);
1253
1254 out:
1255 if ( ! (MDOC_FREECOL & mdoc->flags))
1256 return(1);
1257
1258 if ( ! rew_sub(MDOC_BODY, mdoc, tok, line, ppos))
1259 return(0);
1260 if ( ! rew_sub(MDOC_BLOCK, mdoc, tok, line, ppos))
1261 return(0);
1262
1263 mdoc->flags &= ~MDOC_FREECOL;
1264 return(1);
1265 }
1266
1267
1268 static int
blk_part_imp(MACRO_PROT_ARGS)1269 blk_part_imp(MACRO_PROT_ARGS)
1270 {
1271 int la, nl;
1272 enum mdoct ntok;
1273 enum margserr ac;
1274 char *p;
1275 struct mdoc_node *blk; /* saved block context */
1276 struct mdoc_node *body; /* saved body context */
1277 struct mdoc_node *n;
1278
1279 nl = MDOC_NEWLINE & mdoc->flags;
1280
1281 /*
1282 * A macro that spans to the end of the line. This is generally
1283 * (but not necessarily) called as the first macro. The block
1284 * has a head as the immediate child, which is always empty,
1285 * followed by zero or more opening punctuation nodes, then the
1286 * body (which may be empty, depending on the macro), then zero
1287 * or more closing punctuation nodes.
1288 */
1289
1290 if ( ! mdoc_block_alloc(mdoc, line, ppos, tok, NULL))
1291 return(0);
1292
1293 blk = mdoc->last;
1294
1295 if ( ! mdoc_head_alloc(mdoc, line, ppos, tok))
1296 return(0);
1297 if ( ! rew_sub(MDOC_HEAD, mdoc, tok, line, ppos))
1298 return(0);
1299
1300 /*
1301 * Open the body scope "on-demand", that is, after we've
1302 * processed all our the leading delimiters (open parenthesis,
1303 * etc.).
1304 */
1305
1306 for (body = NULL; ; ) {
1307 la = *pos;
1308 ac = mdoc_args(mdoc, line, pos, buf, tok, &p);
1309
1310 if (ARGS_ERROR == ac)
1311 return(0);
1312 if (ARGS_EOLN == ac)
1313 break;
1314 if (ARGS_PUNCT == ac)
1315 break;
1316
1317 if (NULL == body && ARGS_QWORD != ac &&
1318 DELIM_OPEN == mdoc_isdelim(p)) {
1319 if ( ! dword(mdoc, line, la, p, DELIM_OPEN, 0))
1320 return(0);
1321 continue;
1322 }
1323
1324 if (NULL == body) {
1325 if ( ! mdoc_body_alloc(mdoc, line, ppos, tok))
1326 return(0);
1327 body = mdoc->last;
1328 }
1329
1330 ntok = ARGS_QWORD == ac ? MDOC_MAX : lookup(tok, p);
1331
1332 if (MDOC_MAX == ntok) {
1333 if ( ! dword(mdoc, line, la, p, DELIM_MAX,
1334 MDOC_JOIN & mdoc_macros[tok].flags))
1335 return(0);
1336 continue;
1337 }
1338
1339 if ( ! mdoc_macro(mdoc, ntok, line, la, pos, buf))
1340 return(0);
1341 break;
1342 }
1343
1344 /* Clean-ups to leave in a consistent state. */
1345
1346 if (NULL == body) {
1347 if ( ! mdoc_body_alloc(mdoc, line, ppos, tok))
1348 return(0);
1349 body = mdoc->last;
1350 }
1351
1352 for (n = body->child; n && n->next; n = n->next)
1353 /* Do nothing. */ ;
1354
1355 /*
1356 * End of sentence spacing: if the last node is a text node and
1357 * has a trailing period, then mark it as being end-of-sentence.
1358 */
1359
1360 if (n && MDOC_TEXT == n->type && n->string)
1361 if (mandoc_eos(n->string, strlen(n->string), 1))
1362 n->flags |= MDOC_EOS;
1363
1364 /* Up-propagate the end-of-space flag. */
1365
1366 if (n && (MDOC_EOS & n->flags)) {
1367 body->flags |= MDOC_EOS;
1368 body->parent->flags |= MDOC_EOS;
1369 }
1370
1371 /*
1372 * If there is an open sub-block requiring explicit close-out,
1373 * postpone closing out the current block
1374 * until the rew_sub() call closing out the sub-block.
1375 */
1376 for (n = mdoc->last; n && n != body && n != blk->parent;
1377 n = n->parent) {
1378 if (MDOC_BLOCK == n->type &&
1379 MDOC_EXPLICIT & mdoc_macros[n->tok].flags &&
1380 ! (MDOC_VALID & n->flags)) {
1381 make_pending(n, tok, mdoc, line, ppos);
1382 if ( ! mdoc_endbody_alloc(mdoc, line, ppos,
1383 tok, body, ENDBODY_NOSPACE))
1384 return(0);
1385 return(1);
1386 }
1387 }
1388
1389 /*
1390 * If we can't rewind to our body, then our scope has already
1391 * been closed by another macro (like `Oc' closing `Op'). This
1392 * is ugly behaviour nodding its head to OpenBSD's overwhelming
1393 * crufty use of `Op' breakage.
1394 */
1395 if (n != body)
1396 mandoc_vmsg(MANDOCERR_SCOPENEST, mdoc->parse, line, ppos,
1397 "%s broken", mdoc_macronames[tok]);
1398
1399 if (n && ! rew_sub(MDOC_BODY, mdoc, tok, line, ppos))
1400 return(0);
1401
1402 /* Standard appending of delimiters. */
1403
1404 if (nl && ! append_delims(mdoc, line, pos, buf))
1405 return(0);
1406
1407 /* Rewind scope, if applicable. */
1408
1409 if (n && ! rew_sub(MDOC_BLOCK, mdoc, tok, line, ppos))
1410 return(0);
1411
1412 /* Move trailing .Ns out of scope. */
1413
1414 for (n = body->child; n && n->next; n = n->next)
1415 /* Do nothing. */ ;
1416 if (n && MDOC_Ns == n->tok)
1417 mdoc_node_relink(mdoc, n);
1418
1419 return(1);
1420 }
1421
1422
1423 static int
blk_part_exp(MACRO_PROT_ARGS)1424 blk_part_exp(MACRO_PROT_ARGS)
1425 {
1426 int la, nl;
1427 enum margserr ac;
1428 struct mdoc_node *head; /* keep track of head */
1429 struct mdoc_node *body; /* keep track of body */
1430 char *p;
1431 enum mdoct ntok;
1432
1433 nl = MDOC_NEWLINE & mdoc->flags;
1434
1435 /*
1436 * The opening of an explicit macro having zero or more leading
1437 * punctuation nodes; a head with optional single element (the
1438 * case of `Eo'); and a body that may be empty.
1439 */
1440
1441 if ( ! mdoc_block_alloc(mdoc, line, ppos, tok, NULL))
1442 return(0);
1443
1444 for (head = body = NULL; ; ) {
1445 la = *pos;
1446 ac = mdoc_args(mdoc, line, pos, buf, tok, &p);
1447
1448 if (ARGS_ERROR == ac)
1449 return(0);
1450 if (ARGS_PUNCT == ac)
1451 break;
1452 if (ARGS_EOLN == ac)
1453 break;
1454
1455 /* Flush out leading punctuation. */
1456
1457 if (NULL == head && ARGS_QWORD != ac &&
1458 DELIM_OPEN == mdoc_isdelim(p)) {
1459 assert(NULL == body);
1460 if ( ! dword(mdoc, line, la, p, DELIM_OPEN, 0))
1461 return(0);
1462 continue;
1463 }
1464
1465 if (NULL == head) {
1466 assert(NULL == body);
1467 if ( ! mdoc_head_alloc(mdoc, line, ppos, tok))
1468 return(0);
1469 head = mdoc->last;
1470 }
1471
1472 /*
1473 * `Eo' gobbles any data into the head, but most other
1474 * macros just immediately close out and begin the body.
1475 */
1476
1477 if (NULL == body) {
1478 assert(head);
1479 /* No check whether it's a macro! */
1480 if (MDOC_Eo == tok)
1481 if ( ! dword(mdoc, line, la, p, DELIM_MAX, 0))
1482 return(0);
1483
1484 if ( ! rew_sub(MDOC_HEAD, mdoc, tok, line, ppos))
1485 return(0);
1486 if ( ! mdoc_body_alloc(mdoc, line, ppos, tok))
1487 return(0);
1488 body = mdoc->last;
1489
1490 if (MDOC_Eo == tok)
1491 continue;
1492 }
1493
1494 assert(NULL != head && NULL != body);
1495
1496 ntok = ARGS_QWORD == ac ? MDOC_MAX : lookup(tok, p);
1497
1498 if (MDOC_MAX == ntok) {
1499 if ( ! dword(mdoc, line, la, p, DELIM_MAX,
1500 MDOC_JOIN & mdoc_macros[tok].flags))
1501 return(0);
1502 continue;
1503 }
1504
1505 if ( ! mdoc_macro(mdoc, ntok, line, la, pos, buf))
1506 return(0);
1507 break;
1508 }
1509
1510 /* Clean-up to leave in a consistent state. */
1511
1512 if (NULL == head)
1513 if ( ! mdoc_head_alloc(mdoc, line, ppos, tok))
1514 return(0);
1515
1516 if (NULL == body) {
1517 if ( ! rew_sub(MDOC_HEAD, mdoc, tok, line, ppos))
1518 return(0);
1519 if ( ! mdoc_body_alloc(mdoc, line, ppos, tok))
1520 return(0);
1521 }
1522
1523 /* Standard appending of delimiters. */
1524
1525 if ( ! nl)
1526 return(1);
1527 return(append_delims(mdoc, line, pos, buf));
1528 }
1529
1530
1531 /* ARGSUSED */
1532 static int
in_line_argn(MACRO_PROT_ARGS)1533 in_line_argn(MACRO_PROT_ARGS)
1534 {
1535 int la, flushed, j, maxargs, nl;
1536 enum margserr ac;
1537 enum margverr av;
1538 struct mdoc_arg *arg;
1539 char *p;
1540 enum mdoct ntok;
1541
1542 nl = MDOC_NEWLINE & mdoc->flags;
1543
1544 /*
1545 * A line macro that has a fixed number of arguments (maxargs).
1546 * Only open the scope once the first non-leading-punctuation is
1547 * found (unless MDOC_IGNDELIM is noted, like in `Pf'), then
1548 * keep it open until the maximum number of arguments are
1549 * exhausted.
1550 */
1551
1552 switch (tok) {
1553 case (MDOC_Ap):
1554 /* FALLTHROUGH */
1555 case (MDOC_No):
1556 /* FALLTHROUGH */
1557 case (MDOC_Ns):
1558 /* FALLTHROUGH */
1559 case (MDOC_Ux):
1560 maxargs = 0;
1561 break;
1562 case (MDOC_Bx):
1563 /* FALLTHROUGH */
1564 case (MDOC_Xr):
1565 maxargs = 2;
1566 break;
1567 default:
1568 maxargs = 1;
1569 break;
1570 }
1571
1572 for (arg = NULL; ; ) {
1573 la = *pos;
1574 av = mdoc_argv(mdoc, line, tok, &arg, pos, buf);
1575
1576 if (ARGV_WORD == av) {
1577 *pos = la;
1578 break;
1579 }
1580
1581 if (ARGV_EOLN == av)
1582 break;
1583 if (ARGV_ARG == av)
1584 continue;
1585
1586 mdoc_argv_free(arg);
1587 return(0);
1588 }
1589
1590 for (flushed = j = 0; ; ) {
1591 la = *pos;
1592 ac = mdoc_args(mdoc, line, pos, buf, tok, &p);
1593
1594 if (ARGS_ERROR == ac)
1595 return(0);
1596 if (ARGS_PUNCT == ac)
1597 break;
1598 if (ARGS_EOLN == ac)
1599 break;
1600
1601 if ( ! (MDOC_IGNDELIM & mdoc_macros[tok].flags) &&
1602 ARGS_QWORD != ac && 0 == j &&
1603 DELIM_OPEN == mdoc_isdelim(p)) {
1604 if ( ! dword(mdoc, line, la, p, DELIM_OPEN, 0))
1605 return(0);
1606 continue;
1607 } else if (0 == j)
1608 if ( ! mdoc_elem_alloc(mdoc, line, la, tok, arg))
1609 return(0);
1610
1611 if (j == maxargs && ! flushed) {
1612 if ( ! rew_elem(mdoc, tok))
1613 return(0);
1614 flushed = 1;
1615 }
1616
1617 ntok = ARGS_QWORD == ac ? MDOC_MAX : lookup(tok, p);
1618
1619 if (MDOC_MAX != ntok) {
1620 if ( ! flushed && ! rew_elem(mdoc, tok))
1621 return(0);
1622 flushed = 1;
1623 if ( ! mdoc_macro(mdoc, ntok, line, la, pos, buf))
1624 return(0);
1625 j++;
1626 break;
1627 }
1628
1629 if ( ! (MDOC_IGNDELIM & mdoc_macros[tok].flags) &&
1630 ARGS_QWORD != ac &&
1631 ! flushed &&
1632 DELIM_NONE != mdoc_isdelim(p)) {
1633 if ( ! rew_elem(mdoc, tok))
1634 return(0);
1635 flushed = 1;
1636 }
1637
1638 if ( ! dword(mdoc, line, la, p, DELIM_MAX,
1639 MDOC_JOIN & mdoc_macros[tok].flags))
1640 return(0);
1641 j++;
1642 }
1643
1644 if (0 == j && ! mdoc_elem_alloc(mdoc, line, la, tok, arg))
1645 return(0);
1646
1647 /* Close out in a consistent state. */
1648
1649 if ( ! flushed && ! rew_elem(mdoc, tok))
1650 return(0);
1651 if ( ! nl)
1652 return(1);
1653 return(append_delims(mdoc, line, pos, buf));
1654 }
1655
1656
1657 static int
in_line_eoln(MACRO_PROT_ARGS)1658 in_line_eoln(MACRO_PROT_ARGS)
1659 {
1660 int la;
1661 enum margserr ac;
1662 enum margverr av;
1663 struct mdoc_arg *arg;
1664 char *p;
1665 enum mdoct ntok;
1666
1667 assert( ! (MDOC_PARSED & mdoc_macros[tok].flags));
1668
1669 if (tok == MDOC_Pp)
1670 rew_sub(MDOC_BLOCK, mdoc, MDOC_Nm, line, ppos);
1671
1672 /* Parse macro arguments. */
1673
1674 for (arg = NULL; ; ) {
1675 la = *pos;
1676 av = mdoc_argv(mdoc, line, tok, &arg, pos, buf);
1677
1678 if (ARGV_WORD == av) {
1679 *pos = la;
1680 break;
1681 }
1682 if (ARGV_EOLN == av)
1683 break;
1684 if (ARGV_ARG == av)
1685 continue;
1686
1687 mdoc_argv_free(arg);
1688 return(0);
1689 }
1690
1691 /* Open element scope. */
1692
1693 if ( ! mdoc_elem_alloc(mdoc, line, ppos, tok, arg))
1694 return(0);
1695
1696 /* Parse argument terms. */
1697
1698 for (;;) {
1699 la = *pos;
1700 ac = mdoc_args(mdoc, line, pos, buf, tok, &p);
1701
1702 if (ARGS_ERROR == ac)
1703 return(0);
1704 if (ARGS_EOLN == ac)
1705 break;
1706
1707 ntok = ARGS_QWORD == ac ? MDOC_MAX : lookup(tok, p);
1708
1709 if (MDOC_MAX == ntok) {
1710 if ( ! dword(mdoc, line, la, p, DELIM_MAX,
1711 MDOC_JOIN & mdoc_macros[tok].flags))
1712 return(0);
1713 continue;
1714 }
1715
1716 if ( ! rew_elem(mdoc, tok))
1717 return(0);
1718 return(mdoc_macro(mdoc, ntok, line, la, pos, buf));
1719 }
1720
1721 /* Close out (no delimiters). */
1722
1723 return(rew_elem(mdoc, tok));
1724 }
1725
1726
1727 /* ARGSUSED */
1728 static int
ctx_synopsis(MACRO_PROT_ARGS)1729 ctx_synopsis(MACRO_PROT_ARGS)
1730 {
1731 int nl;
1732
1733 nl = MDOC_NEWLINE & mdoc->flags;
1734
1735 /* If we're not in the SYNOPSIS, go straight to in-line. */
1736 if ( ! (MDOC_SYNOPSIS & mdoc->flags))
1737 return(in_line(mdoc, tok, line, ppos, pos, buf));
1738
1739 /* If we're a nested call, same place. */
1740 if ( ! nl)
1741 return(in_line(mdoc, tok, line, ppos, pos, buf));
1742
1743 /*
1744 * XXX: this will open a block scope; however, if later we end
1745 * up formatting the block scope, then child nodes will inherit
1746 * the formatting. Be careful.
1747 */
1748 if (MDOC_Nm == tok)
1749 return(blk_full(mdoc, tok, line, ppos, pos, buf));
1750 assert(MDOC_Vt == tok);
1751 return(blk_part_imp(mdoc, tok, line, ppos, pos, buf));
1752 }
1753
1754
1755 /* ARGSUSED */
1756 static int
obsolete(MACRO_PROT_ARGS)1757 obsolete(MACRO_PROT_ARGS)
1758 {
1759
1760 mdoc_pmsg(mdoc, line, ppos, MANDOCERR_MACROOBS);
1761 return(1);
1762 }
1763
1764
1765 /*
1766 * Phrases occur within `Bl -column' entries, separated by `Ta' or tabs.
1767 * They're unusual because they're basically free-form text until a
1768 * macro is encountered.
1769 */
1770 static int
phrase(struct mdoc * mdoc,int line,int ppos,char * buf)1771 phrase(struct mdoc *mdoc, int line, int ppos, char *buf)
1772 {
1773 int la, pos;
1774 enum margserr ac;
1775 enum mdoct ntok;
1776 char *p;
1777
1778 for (pos = ppos; ; ) {
1779 la = pos;
1780
1781 ac = mdoc_zargs(mdoc, line, &pos, buf, &p);
1782
1783 if (ARGS_ERROR == ac)
1784 return(0);
1785 if (ARGS_EOLN == ac)
1786 break;
1787
1788 ntok = ARGS_QWORD == ac ? MDOC_MAX : lookup_raw(p);
1789
1790 if (MDOC_MAX == ntok) {
1791 if ( ! dword(mdoc, line, la, p, DELIM_MAX, 1))
1792 return(0);
1793 continue;
1794 }
1795
1796 if ( ! mdoc_macro(mdoc, ntok, line, la, &pos, buf))
1797 return(0);
1798 return(append_delims(mdoc, line, &pos, buf));
1799 }
1800
1801 return(1);
1802 }
1803
1804
1805 /* ARGSUSED */
1806 static int
phrase_ta(MACRO_PROT_ARGS)1807 phrase_ta(MACRO_PROT_ARGS)
1808 {
1809 struct mdoc_node *n;
1810 int la;
1811 enum mdoct ntok;
1812 enum margserr ac;
1813 char *p;
1814
1815 /* Make sure we are in a column list or ignore this macro. */
1816 n = mdoc->last;
1817 while (NULL != n && MDOC_Bl != n->tok)
1818 n = n->parent;
1819 if (NULL == n || LIST_column != n->norm->Bl.type) {
1820 mdoc_pmsg(mdoc, line, ppos, MANDOCERR_STRAYTA);
1821 return(1);
1822 }
1823
1824 /* Advance to the next column. */
1825 if ( ! rew_sub(MDOC_BODY, mdoc, MDOC_It, line, ppos))
1826 return(0);
1827 if ( ! mdoc_body_alloc(mdoc, line, ppos, MDOC_It))
1828 return(0);
1829
1830 for (;;) {
1831 la = *pos;
1832 ac = mdoc_zargs(mdoc, line, pos, buf, &p);
1833
1834 if (ARGS_ERROR == ac)
1835 return(0);
1836 if (ARGS_EOLN == ac)
1837 break;
1838
1839 ntok = ARGS_QWORD == ac ? MDOC_MAX : lookup_raw(p);
1840
1841 if (MDOC_MAX == ntok) {
1842 if ( ! dword(mdoc, line, la, p, DELIM_MAX,
1843 MDOC_JOIN & mdoc_macros[tok].flags))
1844 return(0);
1845 continue;
1846 }
1847
1848 if ( ! mdoc_macro(mdoc, ntok, line, la, pos, buf))
1849 return(0);
1850 return(append_delims(mdoc, line, pos, buf));
1851 }
1852
1853 return(1);
1854 }
1855