1/*	$Id: man_macro.c,v 1.71 2012/01/03 15:16:24 kristaps Exp $ */
2/*
3 * Copyright (c) 2008, 2009, 2010, 2011 Kristaps Dzonsons <kristaps@bsd.lv>
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 "man.h"
27#include "mandoc.h"
28#include "libmandoc.h"
29#include "libman.h"
30
31enum	rew {
32	REW_REWIND,
33	REW_NOHALT,
34	REW_HALT
35};
36
37static	int		 blk_close(MACRO_PROT_ARGS);
38static	int		 blk_exp(MACRO_PROT_ARGS);
39static	int		 blk_imp(MACRO_PROT_ARGS);
40static	int		 in_line_eoln(MACRO_PROT_ARGS);
41static	int		 man_args(struct man *, int,
42				int *, char *, char **);
43
44static	int		 rew_scope(enum man_type,
45				struct man *, enum mant);
46static	enum rew	 rew_dohalt(enum mant, enum man_type,
47				const struct man_node *);
48static	enum rew	 rew_block(enum mant, enum man_type,
49				const struct man_node *);
50static	void		 rew_warn(struct man *,
51				struct man_node *, enum mandocerr);
52
53const	struct man_macro __man_macros[MAN_MAX] = {
54	{ in_line_eoln, MAN_NSCOPED }, /* br */
55	{ in_line_eoln, MAN_BSCOPE }, /* TH */
56	{ blk_imp, MAN_BSCOPE | MAN_SCOPED }, /* SH */
57	{ blk_imp, MAN_BSCOPE | MAN_SCOPED }, /* SS */
58	{ blk_imp, MAN_BSCOPE | MAN_SCOPED | MAN_FSCOPED }, /* TP */
59	{ blk_imp, MAN_BSCOPE }, /* LP */
60	{ blk_imp, MAN_BSCOPE }, /* PP */
61	{ blk_imp, MAN_BSCOPE }, /* P */
62	{ blk_imp, MAN_BSCOPE }, /* IP */
63	{ blk_imp, MAN_BSCOPE }, /* HP */
64	{ in_line_eoln, MAN_SCOPED }, /* SM */
65	{ in_line_eoln, MAN_SCOPED }, /* SB */
66	{ in_line_eoln, 0 }, /* BI */
67	{ in_line_eoln, 0 }, /* IB */
68	{ in_line_eoln, 0 }, /* BR */
69	{ in_line_eoln, 0 }, /* RB */
70	{ in_line_eoln, MAN_SCOPED }, /* R */
71	{ in_line_eoln, MAN_SCOPED }, /* B */
72	{ in_line_eoln, MAN_SCOPED }, /* I */
73	{ in_line_eoln, 0 }, /* IR */
74	{ in_line_eoln, 0 }, /* RI */
75	{ in_line_eoln, MAN_NSCOPED }, /* na */
76	{ in_line_eoln, MAN_NSCOPED }, /* sp */
77	{ in_line_eoln, MAN_BSCOPE }, /* nf */
78	{ in_line_eoln, MAN_BSCOPE }, /* fi */
79	{ blk_close, 0 }, /* RE */
80	{ blk_exp, MAN_EXPLICIT }, /* RS */
81	{ in_line_eoln, 0 }, /* DT */
82	{ in_line_eoln, 0 }, /* UC */
83	{ in_line_eoln, 0 }, /* PD */
84	{ in_line_eoln, 0 }, /* AT */
85	{ in_line_eoln, 0 }, /* in */
86	{ in_line_eoln, 0 }, /* ft */
87	{ in_line_eoln, 0 }, /* OP */
88};
89
90const	struct man_macro * const man_macros = __man_macros;
91
92
93/*
94 * Warn when "n" is an explicit non-roff macro.
95 */
96static void
97rew_warn(struct man *m, struct man_node *n, enum mandocerr er)
98{
99
100	if (er == MANDOCERR_MAX || MAN_BLOCK != n->type)
101		return;
102	if (MAN_VALID & n->flags)
103		return;
104	if ( ! (MAN_EXPLICIT & man_macros[n->tok].flags))
105		return;
106
107	assert(er < MANDOCERR_FATAL);
108	man_nmsg(m, n, er);
109}
110
111
112/*
113 * Rewind scope.  If a code "er" != MANDOCERR_MAX has been provided, it
114 * will be used if an explicit block scope is being closed out.
115 */
116int
117man_unscope(struct man *m, const struct man_node *to,
118		enum mandocerr er)
119{
120	struct man_node	*n;
121
122	assert(to);
123
124	m->next = MAN_NEXT_SIBLING;
125
126	/* LINTED */
127	while (m->last != to) {
128		/*
129		 * Save the parent here, because we may delete the
130		 * m->last node in the post-validation phase and reset
131		 * it to m->last->parent, causing a step in the closing
132		 * out to be lost.
133		 */
134		n = m->last->parent;
135		rew_warn(m, m->last, er);
136		if ( ! man_valid_post(m))
137			return(0);
138		m->last = n;
139		assert(m->last);
140	}
141
142	rew_warn(m, m->last, er);
143	if ( ! man_valid_post(m))
144		return(0);
145
146	return(1);
147}
148
149
150static enum rew
151rew_block(enum mant ntok, enum man_type type, const struct man_node *n)
152{
153
154	if (MAN_BLOCK == type && ntok == n->parent->tok &&
155			MAN_BODY == n->parent->type)
156		return(REW_REWIND);
157	return(ntok == n->tok ? REW_HALT : REW_NOHALT);
158}
159
160
161/*
162 * There are three scope levels: scoped to the root (all), scoped to the
163 * section (all less sections), and scoped to subsections (all less
164 * sections and subsections).
165 */
166static enum rew
167rew_dohalt(enum mant tok, enum man_type type, const struct man_node *n)
168{
169	enum rew	 c;
170
171	/* We cannot progress beyond the root ever. */
172	if (MAN_ROOT == n->type)
173		return(REW_HALT);
174
175	assert(n->parent);
176
177	/* Normal nodes shouldn't go to the level of the root. */
178	if (MAN_ROOT == n->parent->type)
179		return(REW_REWIND);
180
181	/* Already-validated nodes should be closed out. */
182	if (MAN_VALID & n->flags)
183		return(REW_NOHALT);
184
185	/* First: rewind to ourselves. */
186	if (type == n->type && tok == n->tok)
187		return(REW_REWIND);
188
189	/*
190	 * Next follow the implicit scope-smashings as defined by man.7:
191	 * section, sub-section, etc.
192	 */
193
194	switch (tok) {
195	case (MAN_SH):
196		break;
197	case (MAN_SS):
198		/* Rewind to a section, if a block. */
199		if (REW_NOHALT != (c = rew_block(MAN_SH, type, n)))
200			return(c);
201		break;
202	case (MAN_RS):
203		/* Rewind to a subsection, if a block. */
204		if (REW_NOHALT != (c = rew_block(MAN_SS, type, n)))
205			return(c);
206		/* Rewind to a section, if a block. */
207		if (REW_NOHALT != (c = rew_block(MAN_SH, type, n)))
208			return(c);
209		break;
210	default:
211		/* Rewind to an offsetter, if a block. */
212		if (REW_NOHALT != (c = rew_block(MAN_RS, type, n)))
213			return(c);
214		/* Rewind to a subsection, if a block. */
215		if (REW_NOHALT != (c = rew_block(MAN_SS, type, n)))
216			return(c);
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	}
222
223	return(REW_NOHALT);
224}
225
226
227/*
228 * Rewinding entails ascending the parse tree until a coherent point,
229 * for example, the `SH' macro will close out any intervening `SS'
230 * scopes.  When a scope is closed, it must be validated and actioned.
231 */
232static int
233rew_scope(enum man_type type, struct man *m, enum mant tok)
234{
235	struct man_node	*n;
236	enum rew	 c;
237
238	/* LINTED */
239	for (n = m->last; n; n = n->parent) {
240		/*
241		 * Whether we should stop immediately (REW_HALT), stop
242		 * and rewind until this point (REW_REWIND), or keep
243		 * rewinding (REW_NOHALT).
244		 */
245		c = rew_dohalt(tok, type, n);
246		if (REW_HALT == c)
247			return(1);
248		if (REW_REWIND == c)
249			break;
250	}
251
252	/*
253	 * Rewind until the current point.  Warn if we're a roff
254	 * instruction that's mowing over explicit scopes.
255	 */
256	assert(n);
257
258	return(man_unscope(m, n, MANDOCERR_MAX));
259}
260
261
262/*
263 * Close out a generic explicit macro.
264 */
265/* ARGSUSED */
266int
267blk_close(MACRO_PROT_ARGS)
268{
269	enum mant	 	 ntok;
270	const struct man_node	*nn;
271
272	switch (tok) {
273	case (MAN_RE):
274		ntok = MAN_RS;
275		break;
276	default:
277		abort();
278		/* NOTREACHED */
279	}
280
281	for (nn = m->last->parent; nn; nn = nn->parent)
282		if (ntok == nn->tok)
283			break;
284
285	if (NULL == nn)
286		man_pmsg(m, line, ppos, MANDOCERR_NOSCOPE);
287
288	if ( ! rew_scope(MAN_BODY, m, ntok))
289		return(0);
290	if ( ! rew_scope(MAN_BLOCK, m, ntok))
291		return(0);
292
293	return(1);
294}
295
296
297/* ARGSUSED */
298int
299blk_exp(MACRO_PROT_ARGS)
300{
301	int		 la;
302	char		*p;
303
304	/*
305	 * Close out prior scopes.  "Regular" explicit macros cannot be
306	 * nested, but we allow roff macros to be placed just about
307	 * anywhere.
308	 */
309
310	if ( ! man_block_alloc(m, line, ppos, tok))
311		return(0);
312	if ( ! man_head_alloc(m, line, ppos, tok))
313		return(0);
314
315	for (;;) {
316		la = *pos;
317		if ( ! man_args(m, line, pos, buf, &p))
318			break;
319		if ( ! man_word_alloc(m, line, la, p))
320			return(0);
321	}
322
323	assert(m);
324	assert(tok != MAN_MAX);
325
326	if ( ! rew_scope(MAN_HEAD, m, tok))
327		return(0);
328	return(man_body_alloc(m, line, ppos, tok));
329}
330
331
332
333/*
334 * Parse an implicit-block macro.  These contain a MAN_HEAD and a
335 * MAN_BODY contained within a MAN_BLOCK.  Rules for closing out other
336 * scopes, such as `SH' closing out an `SS', are defined in the rew
337 * routines.
338 */
339/* ARGSUSED */
340int
341blk_imp(MACRO_PROT_ARGS)
342{
343	int		 la;
344	char		*p;
345	struct man_node	*n;
346
347	/* Close out prior scopes. */
348
349	if ( ! rew_scope(MAN_BODY, m, tok))
350		return(0);
351	if ( ! rew_scope(MAN_BLOCK, m, tok))
352		return(0);
353
354	/* Allocate new block & head scope. */
355
356	if ( ! man_block_alloc(m, line, ppos, tok))
357		return(0);
358	if ( ! man_head_alloc(m, line, ppos, tok))
359		return(0);
360
361	n = m->last;
362
363	/* Add line arguments. */
364
365	for (;;) {
366		la = *pos;
367		if ( ! man_args(m, line, pos, buf, &p))
368			break;
369		if ( ! man_word_alloc(m, line, la, p))
370			return(0);
371	}
372
373	/* Close out head and open body (unless MAN_SCOPE). */
374
375	if (MAN_SCOPED & man_macros[tok].flags) {
376		/* If we're forcing scope (`TP'), keep it open. */
377		if (MAN_FSCOPED & man_macros[tok].flags) {
378			m->flags |= MAN_BLINE;
379			return(1);
380		} else if (n == m->last) {
381			m->flags |= MAN_BLINE;
382			return(1);
383		}
384	}
385
386	if ( ! rew_scope(MAN_HEAD, m, tok))
387		return(0);
388	return(man_body_alloc(m, line, ppos, tok));
389}
390
391
392/* ARGSUSED */
393int
394in_line_eoln(MACRO_PROT_ARGS)
395{
396	int		 la;
397	char		*p;
398	struct man_node	*n;
399
400	if ( ! man_elem_alloc(m, line, ppos, tok))
401		return(0);
402
403	n = m->last;
404
405	for (;;) {
406		la = *pos;
407		if ( ! man_args(m, line, pos, buf, &p))
408			break;
409		if ( ! man_word_alloc(m, line, la, p))
410			return(0);
411	}
412
413	/*
414	 * If no arguments are specified and this is MAN_SCOPED (i.e.,
415	 * next-line scoped), then set our mode to indicate that we're
416	 * waiting for terms to load into our context.
417	 */
418
419	if (n == m->last && MAN_SCOPED & man_macros[tok].flags) {
420		assert( ! (MAN_NSCOPED & man_macros[tok].flags));
421		m->flags |= MAN_ELINE;
422		return(1);
423	}
424
425	/* Set ignorable context, if applicable. */
426
427	if (MAN_NSCOPED & man_macros[tok].flags) {
428		assert( ! (MAN_SCOPED & man_macros[tok].flags));
429		m->flags |= MAN_ILINE;
430	}
431
432	assert(MAN_ROOT != m->last->type);
433	m->next = MAN_NEXT_SIBLING;
434
435	/*
436	 * Rewind our element scope.  Note that when TH is pruned, we'll
437	 * be back at the root, so make sure that we don't clobber as
438	 * its sibling.
439	 */
440
441	for ( ; m->last; m->last = m->last->parent) {
442		if (m->last == n)
443			break;
444		if (m->last->type == MAN_ROOT)
445			break;
446		if ( ! man_valid_post(m))
447			return(0);
448	}
449
450	assert(m->last);
451
452	/*
453	 * Same here regarding whether we're back at the root.
454	 */
455
456	if (m->last->type != MAN_ROOT && ! man_valid_post(m))
457		return(0);
458
459	return(1);
460}
461
462
463int
464man_macroend(struct man *m)
465{
466
467	return(man_unscope(m, m->first, MANDOCERR_SCOPEEXIT));
468}
469
470static int
471man_args(struct man *m, int line, int *pos, char *buf, char **v)
472{
473	char	 *start;
474
475	assert(*pos);
476	*v = start = buf + *pos;
477	assert(' ' != *start);
478
479	if ('\0' == *start)
480		return(0);
481
482	*v = mandoc_getarg(m->parse, v, line, pos);
483	return(1);
484}
485