xref: /netbsd-src/external/bsd/mdocml/dist/man_macro.c (revision bbde328be4e75ea9ad02e9715ea13ca54b797ada)
1 /*	$Vendor-Id: man_macro.c,v 1.42 2010/03/29 10:10:35 kristaps Exp $ */
2 /*
3  * Copyright (c) 2008, 2009 Kristaps Dzonsons <kristaps@kth.se>
4  *
5  * Permission to use, copy, modify, and distribute this software for any
6  * purpose with or without fee is hereby granted, provided that the above
7  * copyright notice and this permission notice appear in all copies.
8  *
9  * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
10  * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
11  * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
12  * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
13  * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
14  * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
15  * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
16  */
17 #ifdef HAVE_CONFIG_H
18 #include "config.h"
19 #endif
20 
21 #include <assert.h>
22 #include <ctype.h>
23 #include <stdlib.h>
24 #include <string.h>
25 
26 #include "libman.h"
27 
28 enum	rew {
29 	REW_REWIND,
30 	REW_NOHALT,
31 	REW_HALT
32 };
33 
34 static	int		 blk_close(MACRO_PROT_ARGS);
35 static	int		 blk_dotted(MACRO_PROT_ARGS);
36 static	int		 blk_exp(MACRO_PROT_ARGS);
37 static	int		 blk_imp(MACRO_PROT_ARGS);
38 static	int		 in_line_eoln(MACRO_PROT_ARGS);
39 
40 static	int		 rew_scope(enum man_type,
41 				struct man *, enum mant);
42 static	enum rew	 rew_dohalt(enum mant, enum man_type,
43 				const struct man_node *);
44 static	enum rew	 rew_block(enum mant, enum man_type,
45 				const struct man_node *);
46 static	int		 rew_warn(struct man *,
47 				struct man_node *, enum merr);
48 
49 const	struct man_macro __man_macros[MAN_MAX] = {
50 	{ in_line_eoln, MAN_NSCOPED }, /* br */
51 	{ in_line_eoln, 0 }, /* TH */
52 	{ blk_imp, MAN_SCOPED }, /* SH */
53 	{ blk_imp, MAN_SCOPED }, /* SS */
54 	{ blk_imp, MAN_SCOPED | MAN_FSCOPED }, /* TP */
55 	{ blk_imp, 0 }, /* LP */
56 	{ blk_imp, 0 }, /* PP */
57 	{ blk_imp, 0 }, /* P */
58 	{ blk_imp, 0 }, /* IP */
59 	{ blk_imp, 0 }, /* HP */
60 	{ in_line_eoln, MAN_SCOPED }, /* SM */
61 	{ in_line_eoln, MAN_SCOPED }, /* SB */
62 	{ in_line_eoln, 0 }, /* BI */
63 	{ in_line_eoln, 0 }, /* IB */
64 	{ in_line_eoln, 0 }, /* BR */
65 	{ in_line_eoln, 0 }, /* RB */
66 	{ in_line_eoln, MAN_SCOPED }, /* R */
67 	{ in_line_eoln, MAN_SCOPED }, /* B */
68 	{ in_line_eoln, MAN_SCOPED }, /* I */
69 	{ in_line_eoln, 0 }, /* IR */
70 	{ in_line_eoln, 0 }, /* RI */
71 	{ in_line_eoln, MAN_NSCOPED }, /* na */
72 	{ in_line_eoln, 0 }, /* i */
73 	{ in_line_eoln, MAN_NSCOPED }, /* sp */
74 	{ in_line_eoln, 0 }, /* nf */
75 	{ in_line_eoln, 0 }, /* fi */
76 	{ in_line_eoln, 0 }, /* r */
77 	{ blk_close, 0 }, /* RE */
78 	{ blk_exp, MAN_EXPLICIT }, /* RS */
79 	{ in_line_eoln, 0 }, /* DT */
80 	{ in_line_eoln, 0 }, /* UC */
81 	{ in_line_eoln, 0 }, /* PD */
82 	{ in_line_eoln, MAN_NSCOPED }, /* Sp */
83 	{ in_line_eoln, 0 }, /* Vb */
84 	{ in_line_eoln, 0 }, /* Ve */
85 	{ blk_exp, MAN_EXPLICIT | MAN_NOCLOSE}, /* de */
86 	{ blk_exp, MAN_EXPLICIT | MAN_NOCLOSE}, /* dei */
87 	{ blk_exp, MAN_EXPLICIT | MAN_NOCLOSE}, /* am */
88 	{ blk_exp, MAN_EXPLICIT | MAN_NOCLOSE}, /* ami */
89 	{ blk_exp, MAN_EXPLICIT | MAN_NOCLOSE}, /* ig */
90 	{ blk_dotted, 0 }, /* . */
91 };
92 
93 const	struct man_macro * const man_macros = __man_macros;
94 
95 
96 /*
97  * Warn when "n" is an explicit non-roff macro.
98  */
99 static int
100 rew_warn(struct man *m, struct man_node *n, enum merr er)
101 {
102 
103 	if (er == WERRMAX || MAN_BLOCK != n->type)
104 		return(1);
105 	if (MAN_VALID & n->flags)
106 		return(1);
107 	if ( ! (MAN_EXPLICIT & man_macros[n->tok].flags))
108 		return(1);
109 	if (MAN_NOCLOSE & man_macros[n->tok].flags)
110 		return(1);
111 	return(man_nwarn(m, n, er));
112 }
113 
114 
115 /*
116  * Rewind scope.  If a code "er" != WERRMAX has been provided, it will
117  * be used if an explicit block scope is being closed out.
118  */
119 int
120 man_unscope(struct man *m, const struct man_node *n, enum merr er)
121 {
122 
123 	assert(n);
124 
125 	/* LINTED */
126 	while (m->last != n) {
127 		if ( ! rew_warn(m, m->last, er))
128 			return(0);
129 		if ( ! man_valid_post(m))
130 			return(0);
131 		if ( ! man_action_post(m))
132 			return(0);
133 		m->last = m->last->parent;
134 		assert(m->last);
135 	}
136 
137 	if ( ! rew_warn(m, m->last, er))
138 		return(0);
139 	if ( ! man_valid_post(m))
140 		return(0);
141 	if ( ! man_action_post(m))
142 		return(0);
143 
144 	m->next = MAN_ROOT == m->last->type ?
145 		MAN_NEXT_CHILD : MAN_NEXT_SIBLING;
146 
147 	return(1);
148 }
149 
150 
151 static enum rew
152 rew_block(enum mant ntok, enum man_type type, const struct man_node *n)
153 {
154 
155 	if (MAN_BLOCK == type && ntok == n->parent->tok &&
156 			MAN_BODY == n->parent->type)
157 		return(REW_REWIND);
158 	return(ntok == n->tok ? REW_HALT : REW_NOHALT);
159 }
160 
161 
162 /*
163  * There are three scope levels: scoped to the root (all), scoped to the
164  * section (all less sections), and scoped to subsections (all less
165  * sections and subsections).
166  */
167 static enum rew
168 rew_dohalt(enum mant tok, enum man_type type, const struct man_node *n)
169 {
170 	enum rew	 c;
171 
172 	/* We cannot progress beyond the root ever. */
173 	if (MAN_ROOT == n->type)
174 		return(REW_HALT);
175 
176 	assert(n->parent);
177 
178 	/* Normal nodes shouldn't go to the level of the root. */
179 	if (MAN_ROOT == n->parent->type)
180 		return(REW_REWIND);
181 
182 	/* Already-validated nodes should be closed out. */
183 	if (MAN_VALID & n->flags)
184 		return(REW_NOHALT);
185 
186 	/* First: rewind to ourselves. */
187 	if (type == n->type && tok == n->tok)
188 		return(REW_REWIND);
189 
190 	/*
191 	 * If we're a roff macro, then we can close out anything that
192 	 * stands between us and our parent context.
193 	 */
194 	if (MAN_NOCLOSE & man_macros[tok].flags)
195 		return(REW_NOHALT);
196 
197 	/*
198 	 * Don't clobber roff macros: this is a bit complicated.  If the
199 	 * current macro is a roff macro, halt immediately and don't
200 	 * rewind.  If it's not, and the parent is, then close out the
201 	 * current scope and halt at the parent.
202 	 */
203 	if (MAN_NOCLOSE & man_macros[n->tok].flags)
204 		return(REW_HALT);
205 	if (MAN_NOCLOSE & man_macros[n->parent->tok].flags)
206 		return(REW_REWIND);
207 
208 	/*
209 	 * Next follow the implicit scope-smashings as defined by man.7:
210 	 * section, sub-section, etc.
211 	 */
212 
213 	switch (tok) {
214 	case (MAN_SH):
215 		break;
216 	case (MAN_SS):
217 		/* Rewind to a section, if a block. */
218 		if (REW_NOHALT != (c = rew_block(MAN_SH, type, n)))
219 			return(c);
220 		break;
221 	case (MAN_RS):
222 		/* Rewind to a subsection, if a block. */
223 		if (REW_NOHALT != (c = rew_block(MAN_SS, type, n)))
224 			return(c);
225 		/* Rewind to a section, if a block. */
226 		if (REW_NOHALT != (c = rew_block(MAN_SH, type, n)))
227 			return(c);
228 		break;
229 	default:
230 		/* Rewind to an offsetter, if a block. */
231 		if (REW_NOHALT != (c = rew_block(MAN_RS, type, n)))
232 			return(c);
233 		/* Rewind to a subsection, if a block. */
234 		if (REW_NOHALT != (c = rew_block(MAN_SS, type, n)))
235 			return(c);
236 		/* Rewind to a section, if a block. */
237 		if (REW_NOHALT != (c = rew_block(MAN_SH, type, n)))
238 			return(c);
239 		break;
240 	}
241 
242 	return(REW_NOHALT);
243 }
244 
245 
246 /*
247  * Rewinding entails ascending the parse tree until a coherent point,
248  * for example, the `SH' macro will close out any intervening `SS'
249  * scopes.  When a scope is closed, it must be validated and actioned.
250  */
251 static int
252 rew_scope(enum man_type type, struct man *m, enum mant tok)
253 {
254 	struct man_node	*n;
255 	enum rew	 c;
256 
257 	/* LINTED */
258 	for (n = m->last; n; n = n->parent) {
259 		/*
260 		 * Whether we should stop immediately (REW_HALT), stop
261 		 * and rewind until this point (REW_REWIND), or keep
262 		 * rewinding (REW_NOHALT).
263 		 */
264 		c = rew_dohalt(tok, type, n);
265 		if (REW_HALT == c)
266 			return(1);
267 		if (REW_REWIND == c)
268 			break;
269 	}
270 
271 	/*
272 	 * Rewind until the current point.  Warn if we're a roff
273 	 * instruction that's mowing over explicit scopes.
274 	 */
275 	assert(n);
276 	if (MAN_NOCLOSE & man_macros[tok].flags)
277 		return(man_unscope(m, n, WROFFSCOPE));
278 
279 	return(man_unscope(m, n, WERRMAX));
280 }
281 
282 
283 /*
284  * Closure for dotted macros (de, dei, am, ami, ign).  This must handle
285  * any of these as the parent node, so it needs special handling.
286  * Beyond this, it's the same as blk_close().
287  */
288 /* ARGSUSED */
289 int
290 blk_dotted(MACRO_PROT_ARGS)
291 {
292 	enum mant	 ntok;
293 	struct man_node	*nn;
294 
295 	/* Check for any of the following parents... */
296 
297 	for (nn = m->last->parent; nn; nn = nn->parent)
298 		if (nn->tok == MAN_de || nn->tok == MAN_dei ||
299 				nn->tok == MAN_am ||
300 				nn->tok == MAN_ami ||
301 				nn->tok == MAN_ig) {
302 			ntok = nn->tok;
303 			break;
304 		}
305 
306 	if (NULL == nn) {
307 		if ( ! man_pwarn(m, line, ppos, WNOSCOPE))
308 			return(0);
309 		return(1);
310 	}
311 
312 	if ( ! rew_scope(MAN_BODY, m, ntok))
313 		return(0);
314 	if ( ! rew_scope(MAN_BLOCK, m, ntok))
315 		return(0);
316 
317 	/*
318 	 * Restore flags set when we got here and also stipulate that we
319 	 * don't post-process the line when exiting the macro op
320 	 * function in man_pmacro().  See blk_exp().
321 	 */
322 
323 	m->flags = m->svflags | MAN_ILINE;
324 	m->next = m->svnext;
325 	return(1);
326 }
327 
328 
329 /*
330  * Close out a generic explicit macro.
331  */
332 /* ARGSUSED */
333 int
334 blk_close(MACRO_PROT_ARGS)
335 {
336 	enum mant	 	 ntok;
337 	const struct man_node	*nn;
338 
339 	switch (tok) {
340 	case (MAN_RE):
341 		ntok = MAN_RS;
342 		break;
343 	default:
344 		abort();
345 		/* NOTREACHED */
346 	}
347 
348 	for (nn = m->last->parent; nn; nn = nn->parent)
349 		if (ntok == nn->tok)
350 			break;
351 
352 	if (NULL == nn)
353 		if ( ! man_pwarn(m, line, ppos, WNOSCOPE))
354 			return(0);
355 
356 	if ( ! rew_scope(MAN_BODY, m, ntok))
357 		return(0);
358 	if ( ! rew_scope(MAN_BLOCK, m, ntok))
359 		return(0);
360 
361 	return(1);
362 }
363 
364 
365 int
366 blk_exp(MACRO_PROT_ARGS)
367 {
368 	int		 w, la;
369 	char		*p;
370 
371 	/*
372 	 * Close out prior scopes.  "Regular" explicit macros cannot be
373 	 * nested, but we allow roff macros to be placed just about
374 	 * anywhere.
375 	 */
376 
377 	if ( ! (MAN_NOCLOSE & man_macros[tok].flags)) {
378 		if ( ! rew_scope(MAN_BODY, m, tok))
379 			return(0);
380 		if ( ! rew_scope(MAN_BLOCK, m, tok))
381 			return(0);
382 	} else {
383 		/*
384 		 * Save our state and next-scope indicator; we restore
385 		 * it when exiting from the roff instruction block.  See
386 		 * blk_dotted().
387 		 */
388 		m->svflags = m->flags;
389 		m->svnext = m->next;
390 
391 		/* Make sure we drop any line modes. */
392 		m->flags = 0;
393 	}
394 
395 	if ( ! man_block_alloc(m, line, ppos, tok))
396 		return(0);
397 	if ( ! man_head_alloc(m, line, ppos, tok))
398 		return(0);
399 
400 	for (;;) {
401 		la = *pos;
402 		w = man_args(m, line, pos, buf, &p);
403 
404 		if (-1 == w)
405 			return(0);
406 		if (0 == w)
407 			break;
408 
409 		if ( ! man_word_alloc(m, line, la, p))
410 			return(0);
411 	}
412 
413 	assert(m);
414 	assert(tok != MAN_MAX);
415 
416 	if ( ! rew_scope(MAN_HEAD, m, tok))
417 		return(0);
418 	return(man_body_alloc(m, line, ppos, tok));
419 }
420 
421 
422 
423 /*
424  * Parse an implicit-block macro.  These contain a MAN_HEAD and a
425  * MAN_BODY contained within a MAN_BLOCK.  Rules for closing out other
426  * scopes, such as `SH' closing out an `SS', are defined in the rew
427  * routines.
428  */
429 int
430 blk_imp(MACRO_PROT_ARGS)
431 {
432 	int		 w, la;
433 	char		*p;
434 	struct man_node	*n;
435 
436 	/* Close out prior scopes. */
437 
438 	if ( ! rew_scope(MAN_BODY, m, tok))
439 		return(0);
440 	if ( ! rew_scope(MAN_BLOCK, m, tok))
441 		return(0);
442 
443 	/* Allocate new block & head scope. */
444 
445 	if ( ! man_block_alloc(m, line, ppos, tok))
446 		return(0);
447 	if ( ! man_head_alloc(m, line, ppos, tok))
448 		return(0);
449 
450 	n = m->last;
451 
452 	/* Add line arguments. */
453 
454 	for (;;) {
455 		la = *pos;
456 		w = man_args(m, line, pos, buf, &p);
457 
458 		if (-1 == w)
459 			return(0);
460 		if (0 == w)
461 			break;
462 
463 		if ( ! man_word_alloc(m, line, la, p))
464 			return(0);
465 	}
466 
467 	/* Close out head and open body (unless MAN_SCOPE). */
468 
469 	if (MAN_SCOPED & man_macros[tok].flags) {
470 		/* If we're forcing scope (`TP'), keep it open. */
471 		if (MAN_FSCOPED & man_macros[tok].flags) {
472 			m->flags |= MAN_BLINE;
473 			return(1);
474 		} else if (n == m->last) {
475 			m->flags |= MAN_BLINE;
476 			return(1);
477 		}
478 	}
479 
480 	if ( ! rew_scope(MAN_HEAD, m, tok))
481 		return(0);
482 	return(man_body_alloc(m, line, ppos, tok));
483 }
484 
485 
486 int
487 in_line_eoln(MACRO_PROT_ARGS)
488 {
489 	int		 w, la;
490 	char		*p;
491 	struct man_node	*n;
492 
493 	if ( ! man_elem_alloc(m, line, ppos, tok))
494 		return(0);
495 
496 	n = m->last;
497 
498 	for (;;) {
499 		la = *pos;
500 		w = man_args(m, line, pos, buf, &p);
501 
502 		if (-1 == w)
503 			return(0);
504 		if (0 == w)
505 			break;
506 		if ( ! man_word_alloc(m, line, la, p))
507 			return(0);
508 	}
509 
510 	/*
511 	 * If no arguments are specified and this is MAN_SCOPED (i.e.,
512 	 * next-line scoped), then set our mode to indicate that we're
513 	 * waiting for terms to load into our context.
514 	 */
515 
516 	if (n == m->last && MAN_SCOPED & man_macros[tok].flags) {
517 		assert( ! (MAN_NSCOPED & man_macros[tok].flags));
518 		m->flags |= MAN_ELINE;
519 		return(1);
520 	}
521 
522 	/* Set ignorable context, if applicable. */
523 
524 	if (MAN_NSCOPED & man_macros[tok].flags) {
525 		assert( ! (MAN_SCOPED & man_macros[tok].flags));
526 		m->flags |= MAN_ILINE;
527 	}
528 
529 	/*
530 	 * Rewind our element scope.  Note that when TH is pruned, we'll
531 	 * be back at the root, so make sure that we don't clobber as
532 	 * its sibling.
533 	 */
534 
535 	for ( ; m->last; m->last = m->last->parent) {
536 		if (m->last == n)
537 			break;
538 		if (m->last->type == MAN_ROOT)
539 			break;
540 		if ( ! man_valid_post(m))
541 			return(0);
542 		if ( ! man_action_post(m))
543 			return(0);
544 	}
545 
546 	assert(m->last);
547 
548 	/*
549 	 * Same here regarding whether we're back at the root.
550 	 */
551 
552 	if (m->last->type != MAN_ROOT && ! man_valid_post(m))
553 		return(0);
554 	if (m->last->type != MAN_ROOT && ! man_action_post(m))
555 		return(0);
556 
557 	m->next = MAN_ROOT == m->last->type ?
558 		MAN_NEXT_CHILD : MAN_NEXT_SIBLING;
559 
560 	return(1);
561 }
562 
563 
564 int
565 man_macroend(struct man *m)
566 {
567 
568 	return(man_unscope(m, m->first, WEXITSCOPE));
569 }
570 
571