xref: /illumos-gate/usr/src/cmd/mandoc/mdoc_macro.c (revision cec8643b)
1 /*	$Id: mdoc_macro.c,v 1.232 2019/01/07 07:26:29 schwarze Exp $ */
2 /*
3  * Copyright (c) 2008-2012 Kristaps Dzonsons <kristaps@bsd.lv>
4  * Copyright (c) 2010, 2012-2019 Ingo Schwarze <schwarze@openbsd.org>
5  *
6  * Permission to use, copy, modify, and distribute this software for any
7  * purpose with or without fee is hereby granted, provided that the above
8  * copyright notice and this permission notice appear in all copies.
9  *
10  * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHORS DISCLAIM ALL WARRANTIES
11  * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
12  * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHORS BE LIABLE FOR
13  * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
14  * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
15  * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
16  * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
17  */
18 #include "config.h"
19 
20 #include <sys/types.h>
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 "roff.h"
31 #include "mdoc.h"
32 #include "libmandoc.h"
33 #include "roff_int.h"
34 #include "libmdoc.h"
35 
36 static	void		blk_full(MACRO_PROT_ARGS);
37 static	void		blk_exp_close(MACRO_PROT_ARGS);
38 static	void		blk_part_exp(MACRO_PROT_ARGS);
39 static	void		blk_part_imp(MACRO_PROT_ARGS);
40 static	void		ctx_synopsis(MACRO_PROT_ARGS);
41 static	void		in_line_eoln(MACRO_PROT_ARGS);
42 static	void		in_line_argn(MACRO_PROT_ARGS);
43 static	void		in_line(MACRO_PROT_ARGS);
44 static	void		phrase_ta(MACRO_PROT_ARGS);
45 
46 static	void		append_delims(struct roff_man *, int, int *, char *);
47 static	void		dword(struct roff_man *, int, int, const char *,
48 				enum mdelim, int);
49 static	int		find_pending(struct roff_man *, enum roff_tok,
50 				int, int, struct roff_node *);
51 static	int		lookup(struct roff_man *, int, int, int, const char *);
52 static	int		macro_or_word(MACRO_PROT_ARGS, char *, int);
53 static	void		break_intermediate(struct roff_node *,
54 				struct roff_node *);
55 static	int		parse_rest(struct roff_man *, enum roff_tok,
56 				int, int *, char *);
57 static	enum roff_tok	rew_alt(enum roff_tok);
58 static	void		rew_elem(struct roff_man *, enum roff_tok);
59 static	void		rew_last(struct roff_man *, const struct roff_node *);
60 static	void		rew_pending(struct roff_man *,
61 				const struct roff_node *);
62 
63 static const struct mdoc_macro mdoc_macros[MDOC_MAX - MDOC_Dd] = {
64 	{ in_line_eoln, MDOC_PROLOGUE }, /* Dd */
65 	{ in_line_eoln, MDOC_PROLOGUE }, /* Dt */
66 	{ in_line_eoln, MDOC_PROLOGUE }, /* Os */
67 	{ blk_full, MDOC_PARSED | MDOC_JOIN }, /* Sh */
68 	{ blk_full, MDOC_PARSED | MDOC_JOIN }, /* Ss */
69 	{ in_line_eoln, 0 }, /* Pp */
70 	{ blk_part_imp, MDOC_PARSED | MDOC_JOIN }, /* D1 */
71 	{ blk_part_imp, MDOC_PARSED | MDOC_JOIN }, /* Dl */
72 	{ blk_full, MDOC_EXPLICIT }, /* Bd */
73 	{ blk_exp_close, MDOC_EXPLICIT | MDOC_JOIN }, /* Ed */
74 	{ blk_full, MDOC_EXPLICIT }, /* Bl */
75 	{ blk_exp_close, MDOC_EXPLICIT | MDOC_JOIN }, /* El */
76 	{ blk_full, MDOC_PARSED | MDOC_JOIN }, /* It */
77 	{ in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Ad */
78 	{ in_line, MDOC_CALLABLE | MDOC_PARSED | MDOC_JOIN }, /* An */
79 	{ in_line_argn, MDOC_CALLABLE | MDOC_PARSED |
80 			MDOC_IGNDELIM | MDOC_JOIN }, /* Ap */
81 	{ in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Ar */
82 	{ in_line, MDOC_CALLABLE | MDOC_PARSED | MDOC_JOIN }, /* Cd */
83 	{ in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Cm */
84 	{ in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Dv */
85 	{ in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Er */
86 	{ in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Ev */
87 	{ in_line_eoln, 0 }, /* Ex */
88 	{ in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Fa */
89 	{ in_line_eoln, 0 }, /* Fd */
90 	{ in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Fl */
91 	{ in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Fn */
92 	{ in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Ft */
93 	{ in_line, MDOC_CALLABLE | MDOC_PARSED | MDOC_JOIN }, /* Ic */
94 	{ in_line_argn, MDOC_CALLABLE | MDOC_PARSED }, /* In */
95 	{ in_line, MDOC_CALLABLE | MDOC_PARSED | MDOC_JOIN }, /* Li */
96 	{ blk_full, MDOC_JOIN }, /* Nd */
97 	{ ctx_synopsis, MDOC_CALLABLE | MDOC_PARSED }, /* Nm */
98 	{ blk_part_imp, MDOC_CALLABLE | MDOC_PARSED }, /* Op */
99 	{ in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Ot */
100 	{ in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Pa */
101 	{ in_line_eoln, 0 }, /* Rv */
102 	{ in_line_argn, MDOC_CALLABLE | MDOC_PARSED }, /* St */
103 	{ in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Va */
104 	{ ctx_synopsis, MDOC_CALLABLE | MDOC_PARSED }, /* Vt */
105 	{ in_line_argn, MDOC_CALLABLE | MDOC_PARSED }, /* Xr */
106 	{ in_line_eoln, MDOC_JOIN }, /* %A */
107 	{ in_line_eoln, MDOC_JOIN }, /* %B */
108 	{ in_line_eoln, MDOC_JOIN }, /* %D */
109 	{ in_line_eoln, MDOC_JOIN }, /* %I */
110 	{ in_line_eoln, MDOC_JOIN }, /* %J */
111 	{ in_line_eoln, 0 }, /* %N */
112 	{ in_line_eoln, MDOC_JOIN }, /* %O */
113 	{ in_line_eoln, 0 }, /* %P */
114 	{ in_line_eoln, MDOC_JOIN }, /* %R */
115 	{ in_line_eoln, MDOC_JOIN }, /* %T */
116 	{ in_line_eoln, 0 }, /* %V */
117 	{ blk_exp_close, MDOC_CALLABLE | MDOC_PARSED |
118 			 MDOC_EXPLICIT | MDOC_JOIN }, /* Ac */
119 	{ blk_part_exp, MDOC_CALLABLE | MDOC_PARSED |
120 			MDOC_EXPLICIT | MDOC_JOIN }, /* Ao */
121 	{ blk_part_imp, MDOC_CALLABLE | MDOC_PARSED | MDOC_JOIN }, /* Aq */
122 	{ in_line_argn, MDOC_CALLABLE | MDOC_PARSED }, /* At */
123 	{ blk_exp_close, MDOC_CALLABLE | MDOC_PARSED |
124 			 MDOC_EXPLICIT | MDOC_JOIN }, /* Bc */
125 	{ blk_full, MDOC_EXPLICIT }, /* Bf */
126 	{ blk_part_exp, MDOC_CALLABLE | MDOC_PARSED |
127 			MDOC_EXPLICIT | MDOC_JOIN }, /* Bo */
128 	{ blk_part_imp, MDOC_CALLABLE | MDOC_PARSED | MDOC_JOIN }, /* 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_CALLABLE | MDOC_PARSED |
133 			 MDOC_EXPLICIT | MDOC_JOIN }, /* Dc */
134 	{ blk_part_exp, MDOC_CALLABLE | MDOC_PARSED |
135 			MDOC_EXPLICIT | MDOC_JOIN }, /* Do */
136 	{ blk_part_imp, MDOC_CALLABLE | MDOC_PARSED | MDOC_JOIN }, /* Dq */
137 	{ blk_exp_close, MDOC_CALLABLE | MDOC_PARSED | MDOC_EXPLICIT }, /* Ec */
138 	{ blk_exp_close, MDOC_EXPLICIT | MDOC_JOIN }, /* Ef */
139 	{ in_line, MDOC_CALLABLE | MDOC_PARSED | MDOC_JOIN }, /* Em */
140 	{ blk_part_exp, MDOC_CALLABLE | MDOC_PARSED | MDOC_EXPLICIT }, /* Eo */
141 	{ in_line_argn, MDOC_CALLABLE | MDOC_PARSED }, /* Fx */
142 	{ in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Ms */
143 	{ in_line, MDOC_CALLABLE | MDOC_PARSED | MDOC_JOIN }, /* No */
144 	{ in_line_argn, MDOC_CALLABLE | MDOC_PARSED |
145 			MDOC_IGNDELIM | MDOC_JOIN }, /* Ns */
146 	{ in_line_argn, MDOC_CALLABLE | MDOC_PARSED }, /* Nx */
147 	{ in_line_argn, MDOC_CALLABLE | MDOC_PARSED }, /* Ox */
148 	{ blk_exp_close, MDOC_CALLABLE | MDOC_PARSED |
149 			 MDOC_EXPLICIT | MDOC_JOIN }, /* Pc */
150 	{ in_line_argn, MDOC_CALLABLE | MDOC_PARSED | MDOC_IGNDELIM }, /* Pf */
151 	{ blk_part_exp, MDOC_CALLABLE | MDOC_PARSED |
152 			MDOC_EXPLICIT | MDOC_JOIN }, /* Po */
153 	{ blk_part_imp, MDOC_CALLABLE | MDOC_PARSED | MDOC_JOIN }, /* Pq */
154 	{ blk_exp_close, MDOC_CALLABLE | MDOC_PARSED |
155 			 MDOC_EXPLICIT | MDOC_JOIN }, /* Qc */
156 	{ blk_part_imp, MDOC_CALLABLE | MDOC_PARSED | MDOC_JOIN }, /* Ql */
157 	{ blk_part_exp, MDOC_CALLABLE | MDOC_PARSED |
158 			MDOC_EXPLICIT | MDOC_JOIN }, /* Qo */
159 	{ blk_part_imp, MDOC_CALLABLE | MDOC_PARSED | MDOC_JOIN }, /* Qq */
160 	{ blk_exp_close, MDOC_EXPLICIT | MDOC_JOIN }, /* Re */
161 	{ blk_full, MDOC_EXPLICIT }, /* Rs */
162 	{ blk_exp_close, MDOC_CALLABLE | MDOC_PARSED |
163 			 MDOC_EXPLICIT | MDOC_JOIN }, /* Sc */
164 	{ blk_part_exp, MDOC_CALLABLE | MDOC_PARSED |
165 			MDOC_EXPLICIT | MDOC_JOIN }, /* So */
166 	{ blk_part_imp, MDOC_CALLABLE | MDOC_PARSED | MDOC_JOIN }, /* Sq */
167 	{ in_line_argn, 0 }, /* Sm */
168 	{ in_line, MDOC_CALLABLE | MDOC_PARSED | MDOC_JOIN }, /* Sx */
169 	{ in_line, MDOC_CALLABLE | MDOC_PARSED | MDOC_JOIN }, /* Sy */
170 	{ in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Tn */
171 	{ in_line_argn, MDOC_CALLABLE | MDOC_PARSED | MDOC_JOIN }, /* Ux */
172 	{ blk_exp_close, MDOC_EXPLICIT | MDOC_CALLABLE | MDOC_PARSED }, /* Xc */
173 	{ blk_part_exp, MDOC_CALLABLE | MDOC_PARSED | MDOC_EXPLICIT }, /* Xo */
174 	{ blk_full, MDOC_EXPLICIT | MDOC_CALLABLE }, /* Fo */
175 	{ blk_exp_close, MDOC_CALLABLE | MDOC_PARSED |
176 			 MDOC_EXPLICIT | MDOC_JOIN }, /* Fc */
177 	{ blk_part_exp, MDOC_CALLABLE | MDOC_PARSED |
178 			MDOC_EXPLICIT | MDOC_JOIN }, /* Oo */
179 	{ blk_exp_close, MDOC_CALLABLE | MDOC_PARSED |
180 			 MDOC_EXPLICIT | MDOC_JOIN }, /* Oc */
181 	{ blk_full, MDOC_EXPLICIT }, /* Bk */
182 	{ blk_exp_close, MDOC_EXPLICIT | MDOC_JOIN }, /* Ek */
183 	{ in_line_eoln, 0 }, /* Bt */
184 	{ in_line_eoln, 0 }, /* Hf */
185 	{ in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Fr */
186 	{ in_line_eoln, 0 }, /* Ud */
187 	{ in_line, 0 }, /* Lb */
188 	{ in_line_eoln, 0 }, /* Lp */
189 	{ in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Lk */
190 	{ in_line, MDOC_CALLABLE | MDOC_PARSED }, /* Mt */
191 	{ blk_part_imp, MDOC_CALLABLE | MDOC_PARSED | MDOC_JOIN }, /* Brq */
192 	{ blk_part_exp, MDOC_CALLABLE | MDOC_PARSED |
193 			MDOC_EXPLICIT | MDOC_JOIN }, /* Bro */
194 	{ blk_exp_close, MDOC_CALLABLE | MDOC_PARSED |
195 			 MDOC_EXPLICIT | MDOC_JOIN }, /* Brc */
196 	{ in_line_eoln, MDOC_JOIN }, /* %C */
197 	{ in_line_argn, MDOC_CALLABLE | MDOC_PARSED }, /* Es */
198 	{ blk_part_imp, MDOC_CALLABLE | MDOC_PARSED | MDOC_JOIN }, /* En */
199 	{ in_line_argn, MDOC_CALLABLE | MDOC_PARSED }, /* Dx */
200 	{ in_line_eoln, MDOC_JOIN }, /* %Q */
201 	{ in_line_eoln, 0 }, /* %U */
202 	{ phrase_ta, MDOC_CALLABLE | MDOC_PARSED | MDOC_JOIN }, /* Ta */
203 };
204 
205 
206 const struct mdoc_macro *
mdoc_macro(enum roff_tok tok)207 mdoc_macro(enum roff_tok tok)
208 {
209 	assert(tok >= MDOC_Dd && tok < MDOC_MAX);
210 	return mdoc_macros + (tok - MDOC_Dd);
211 }
212 
213 /*
214  * This is called at the end of parsing.  It must traverse up the tree,
215  * closing out open [implicit] scopes.  Obviously, open explicit scopes
216  * are errors.
217  */
218 void
mdoc_endparse(struct roff_man * mdoc)219 mdoc_endparse(struct roff_man *mdoc)
220 {
221 	struct roff_node *n;
222 
223 	/* Scan for open explicit scopes. */
224 
225 	n = mdoc->last->flags & NODE_VALID ?
226 	    mdoc->last->parent : mdoc->last;
227 
228 	for ( ; n; n = n->parent)
229 		if (n->type == ROFFT_BLOCK &&
230 		    mdoc_macro(n->tok)->flags & MDOC_EXPLICIT)
231 			mandoc_msg(MANDOCERR_BLK_NOEND,
232 			    n->line, n->pos, "%s", roff_name[n->tok]);
233 
234 	/* Rewind to the first. */
235 
236 	rew_last(mdoc, mdoc->meta.first);
237 }
238 
239 /*
240  * Look up the macro at *p called by "from",
241  * or as a line macro if from == TOKEN_NONE.
242  */
243 static int
lookup(struct roff_man * mdoc,int from,int line,int ppos,const char * p)244 lookup(struct roff_man *mdoc, int from, int line, int ppos, const char *p)
245 {
246 	enum roff_tok	 res;
247 
248 	if (mdoc->flags & MDOC_PHRASEQF) {
249 		mdoc->flags &= ~MDOC_PHRASEQF;
250 		return TOKEN_NONE;
251 	}
252 	if (from == TOKEN_NONE || mdoc_macro(from)->flags & MDOC_PARSED) {
253 		res = roffhash_find(mdoc->mdocmac, p, 0);
254 		if (res != TOKEN_NONE) {
255 			if (mdoc_macro(res)->flags & MDOC_CALLABLE)
256 				return res;
257 			mandoc_msg(MANDOCERR_MACRO_CALL, line, ppos, "%s", p);
258 		}
259 	}
260 	return TOKEN_NONE;
261 }
262 
263 /*
264  * Rewind up to and including a specific node.
265  */
266 static void
rew_last(struct roff_man * mdoc,const struct roff_node * to)267 rew_last(struct roff_man *mdoc, const struct roff_node *to)
268 {
269 
270 	if (to->flags & NODE_VALID)
271 		return;
272 
273 	while (mdoc->last != to) {
274 		mdoc_state(mdoc, mdoc->last);
275 		mdoc->last->flags |= NODE_VALID | NODE_ENDED;
276 		mdoc->last = mdoc->last->parent;
277 	}
278 	mdoc_state(mdoc, mdoc->last);
279 	mdoc->last->flags |= NODE_VALID | NODE_ENDED;
280 	mdoc->next = ROFF_NEXT_SIBLING;
281 }
282 
283 /*
284  * Rewind up to a specific block, including all blocks that broke it.
285  */
286 static void
rew_pending(struct roff_man * mdoc,const struct roff_node * n)287 rew_pending(struct roff_man *mdoc, const struct roff_node *n)
288 {
289 
290 	for (;;) {
291 		rew_last(mdoc, n);
292 
293 		if (mdoc->last == n) {
294 			switch (n->type) {
295 			case ROFFT_HEAD:
296 				roff_body_alloc(mdoc, n->line, n->pos,
297 				    n->tok);
298 				if (n->tok == MDOC_Ss)
299 					mdoc->flags &= ~ROFF_NONOFILL;
300 				break;
301 			case ROFFT_BLOCK:
302 				break;
303 			default:
304 				return;
305 			}
306 			if ( ! (n->flags & NODE_BROKEN))
307 				return;
308 		} else
309 			n = mdoc->last;
310 
311 		for (;;) {
312 			if ((n = n->parent) == NULL)
313 				return;
314 
315 			if (n->type == ROFFT_BLOCK ||
316 			    n->type == ROFFT_HEAD) {
317 				if (n->flags & NODE_ENDED)
318 					break;
319 				else
320 					return;
321 			}
322 		}
323 	}
324 }
325 
326 /*
327  * For a block closing macro, return the corresponding opening one.
328  * Otherwise, return the macro itself.
329  */
330 static enum roff_tok
rew_alt(enum roff_tok tok)331 rew_alt(enum roff_tok tok)
332 {
333 	switch (tok) {
334 	case MDOC_Ac:
335 		return MDOC_Ao;
336 	case MDOC_Bc:
337 		return MDOC_Bo;
338 	case MDOC_Brc:
339 		return MDOC_Bro;
340 	case MDOC_Dc:
341 		return MDOC_Do;
342 	case MDOC_Ec:
343 		return MDOC_Eo;
344 	case MDOC_Ed:
345 		return MDOC_Bd;
346 	case MDOC_Ef:
347 		return MDOC_Bf;
348 	case MDOC_Ek:
349 		return MDOC_Bk;
350 	case MDOC_El:
351 		return MDOC_Bl;
352 	case MDOC_Fc:
353 		return MDOC_Fo;
354 	case MDOC_Oc:
355 		return MDOC_Oo;
356 	case MDOC_Pc:
357 		return MDOC_Po;
358 	case MDOC_Qc:
359 		return MDOC_Qo;
360 	case MDOC_Re:
361 		return MDOC_Rs;
362 	case MDOC_Sc:
363 		return MDOC_So;
364 	case MDOC_Xc:
365 		return MDOC_Xo;
366 	default:
367 		return tok;
368 	}
369 }
370 
371 static void
rew_elem(struct roff_man * mdoc,enum roff_tok tok)372 rew_elem(struct roff_man *mdoc, enum roff_tok tok)
373 {
374 	struct roff_node *n;
375 
376 	n = mdoc->last;
377 	if (n->type != ROFFT_ELEM)
378 		n = n->parent;
379 	assert(n->type == ROFFT_ELEM);
380 	assert(tok == n->tok);
381 	rew_last(mdoc, n);
382 }
383 
384 static void
break_intermediate(struct roff_node * n,struct roff_node * breaker)385 break_intermediate(struct roff_node *n, struct roff_node *breaker)
386 {
387 	if (n != breaker &&
388 	    n->type != ROFFT_BLOCK && n->type != ROFFT_HEAD &&
389 	    (n->type != ROFFT_BODY || n->end != ENDBODY_NOT))
390 		n = n->parent;
391 	while (n != breaker) {
392 		if ( ! (n->flags & NODE_VALID))
393 			n->flags |= NODE_BROKEN;
394 		n = n->parent;
395 	}
396 }
397 
398 /*
399  * If there is an open sub-block of the target requiring
400  * explicit close-out, postpone closing out the target until
401  * the rew_pending() call closing out the sub-block.
402  */
403 static int
find_pending(struct roff_man * mdoc,enum roff_tok tok,int line,int ppos,struct roff_node * target)404 find_pending(struct roff_man *mdoc, enum roff_tok tok, int line, int ppos,
405 	struct roff_node *target)
406 {
407 	struct roff_node	*n;
408 	int			 irc;
409 
410 	if (target->flags & NODE_VALID)
411 		return 0;
412 
413 	irc = 0;
414 	for (n = mdoc->last; n != NULL && n != target; n = n->parent) {
415 		if (n->flags & NODE_ENDED)
416 			continue;
417 		if (n->type == ROFFT_BLOCK &&
418 		    mdoc_macro(n->tok)->flags & MDOC_EXPLICIT) {
419 			irc = 1;
420 			break_intermediate(mdoc->last, target);
421 			if (target->type == ROFFT_HEAD)
422 				target->flags |= NODE_ENDED;
423 			else if ( ! (target->flags & NODE_ENDED)) {
424 				mandoc_msg(MANDOCERR_BLK_NEST,
425 				    line, ppos, "%s breaks %s",
426 				    roff_name[tok], roff_name[n->tok]);
427 				mdoc_endbody_alloc(mdoc, line, ppos,
428 				    tok, target);
429 			}
430 		}
431 	}
432 	return irc;
433 }
434 
435 /*
436  * Allocate a word and check whether it's punctuation or not.
437  * Punctuation consists of those tokens found in mdoc_isdelim().
438  */
439 static void
dword(struct roff_man * mdoc,int line,int col,const char * p,enum mdelim d,int may_append)440 dword(struct roff_man *mdoc, int line, int col, const char *p,
441 		enum mdelim d, int may_append)
442 {
443 
444 	if (d == DELIM_MAX)
445 		d = mdoc_isdelim(p);
446 
447 	if (may_append &&
448 	    ! (mdoc->flags & (MDOC_SYNOPSIS | MDOC_KEEP | MDOC_SMOFF)) &&
449 	    d == DELIM_NONE && mdoc->last->type == ROFFT_TEXT &&
450 	    mdoc_isdelim(mdoc->last->string) == DELIM_NONE) {
451 		roff_word_append(mdoc, p);
452 		return;
453 	}
454 
455 	roff_word_alloc(mdoc, line, col, p);
456 
457 	/*
458 	 * If the word consists of a bare delimiter,
459 	 * flag the new node accordingly,
460 	 * unless doing so was vetoed by the invoking macro.
461 	 * Always clear the veto, it is only valid for one word.
462 	 */
463 
464 	if (d == DELIM_OPEN)
465 		mdoc->last->flags |= NODE_DELIMO;
466 	else if (d == DELIM_CLOSE &&
467 	    ! (mdoc->flags & MDOC_NODELIMC) &&
468 	    mdoc->last->parent->tok != MDOC_Fd)
469 		mdoc->last->flags |= NODE_DELIMC;
470 	mdoc->flags &= ~MDOC_NODELIMC;
471 }
472 
473 static void
append_delims(struct roff_man * mdoc,int line,int * pos,char * buf)474 append_delims(struct roff_man *mdoc, int line, int *pos, char *buf)
475 {
476 	char		*p;
477 	int		 la;
478 	enum margserr	 ac;
479 
480 	if (buf[*pos] == '\0')
481 		return;
482 
483 	for (;;) {
484 		la = *pos;
485 		ac = mdoc_args(mdoc, line, pos, buf, TOKEN_NONE, &p);
486 		if (ac == ARGS_EOLN)
487 			break;
488 		dword(mdoc, line, la, p, DELIM_MAX, 1);
489 
490 		/*
491 		 * If we encounter end-of-sentence symbols, then trigger
492 		 * the double-space.
493 		 *
494 		 * XXX: it's easy to allow this to propagate outward to
495 		 * the last symbol, such that `. )' will cause the
496 		 * correct double-spacing.  However, (1) groff isn't
497 		 * smart enough to do this and (2) it would require
498 		 * knowing which symbols break this behaviour, for
499 		 * example, `.  ;' shouldn't propagate the double-space.
500 		 */
501 
502 		if (mandoc_eos(p, strlen(p)))
503 			mdoc->last->flags |= NODE_EOS;
504 		if (ac == ARGS_ALLOC)
505 			free(p);
506 	}
507 }
508 
509 /*
510  * Parse one word.
511  * If it is a macro, call it and return 1.
512  * Otherwise, allocate it and return 0.
513  */
514 static int
macro_or_word(MACRO_PROT_ARGS,char * p,int parsed)515 macro_or_word(MACRO_PROT_ARGS, char *p, int parsed)
516 {
517 	int		 ntok;
518 
519 	ntok = buf[ppos] == '"' || parsed == 0 ||
520 	    mdoc->flags & MDOC_PHRASELIT ? TOKEN_NONE :
521 	    lookup(mdoc, tok, line, ppos, p);
522 
523 	if (ntok == TOKEN_NONE) {
524 		dword(mdoc, line, ppos, p, DELIM_MAX, tok == TOKEN_NONE ||
525 		    mdoc_macro(tok)->flags & MDOC_JOIN);
526 		return 0;
527 	} else {
528 		if (tok != TOKEN_NONE &&
529 		    mdoc_macro(tok)->fp == in_line_eoln)
530 			rew_elem(mdoc, tok);
531 		(*mdoc_macro(ntok)->fp)(mdoc, ntok, line, ppos, pos, buf);
532 		if (tok == TOKEN_NONE)
533 			append_delims(mdoc, line, pos, buf);
534 		return 1;
535 	}
536 }
537 
538 /*
539  * Close out block partial/full explicit.
540  */
541 static void
blk_exp_close(MACRO_PROT_ARGS)542 blk_exp_close(MACRO_PROT_ARGS)
543 {
544 	struct roff_node *body;		/* Our own body. */
545 	struct roff_node *endbody;	/* Our own end marker. */
546 	struct roff_node *itblk;	/* An It block starting later. */
547 	struct roff_node *later;	/* A sub-block starting later. */
548 	struct roff_node *n;		/* Search back to our block. */
549 	struct roff_node *target;	/* For find_pending(). */
550 
551 	int		 j, lastarg, maxargs, nl, pending;
552 	enum margserr	 ac;
553 	enum roff_tok	 atok, ntok;
554 	char		*p;
555 
556 	nl = MDOC_NEWLINE & mdoc->flags;
557 
558 	switch (tok) {
559 	case MDOC_Ec:
560 		maxargs = 1;
561 		break;
562 	case MDOC_Ek:
563 		mdoc->flags &= ~MDOC_KEEP;
564 		/* FALLTHROUGH */
565 	default:
566 		maxargs = 0;
567 		break;
568 	}
569 
570 	/* Search backwards for the beginning of our own body. */
571 
572 	atok = rew_alt(tok);
573 	body = NULL;
574 	for (n = mdoc->last; n; n = n->parent) {
575 		if (n->flags & NODE_ENDED || n->tok != atok ||
576 		    n->type != ROFFT_BODY || n->end != ENDBODY_NOT)
577 			continue;
578 		body = n;
579 		break;
580 	}
581 
582 	/*
583 	 * Search backwards for beginnings of blocks,
584 	 * both of our own and of pending sub-blocks.
585 	 */
586 
587 	endbody = itblk = later = NULL;
588 	for (n = mdoc->last; n; n = n->parent) {
589 		if (n->flags & NODE_ENDED)
590 			continue;
591 
592 		/*
593 		 * Mismatching end macros can never break anything
594 		 * and we only care about the breaking of BLOCKs.
595 		 */
596 
597 		if (body == NULL || n->type != ROFFT_BLOCK)
598 			continue;
599 
600 		/*
601 		 * SYNOPSIS name blocks can not be broken themselves,
602 		 * but they do get broken together with a broken child.
603 		 */
604 
605 		if (n->tok == MDOC_Nm) {
606 			if (later != NULL)
607 				n->flags |= NODE_BROKEN | NODE_ENDED;
608 			continue;
609 		}
610 
611 		if (n->tok == MDOC_It) {
612 			itblk = n;
613 			continue;
614 		}
615 
616 		if (atok == n->tok) {
617 
618 			/*
619 			 * Found the start of our own block.
620 			 * When there is no pending sub block,
621 			 * just proceed to closing out.
622 			 */
623 
624 			if (later == NULL ||
625 			    (tok == MDOC_El && itblk == NULL))
626 				break;
627 
628 			/*
629 			 * When there is a pending sub block, postpone
630 			 * closing out the current block until the
631 			 * rew_pending() closing out the sub-block.
632 			 * Mark the place where the formatting - but not
633 			 * the scope - of the current block ends.
634 			 */
635 
636 			mandoc_msg(MANDOCERR_BLK_NEST,
637 			    line, ppos, "%s breaks %s",
638 			    roff_name[atok], roff_name[later->tok]);
639 
640 			endbody = mdoc_endbody_alloc(mdoc, line, ppos,
641 			    atok, body);
642 
643 			if (tok == MDOC_El)
644 				itblk->flags |= NODE_ENDED | NODE_BROKEN;
645 
646 			/*
647 			 * If a block closing macro taking arguments
648 			 * breaks another block, put the arguments
649 			 * into the end marker.
650 			 */
651 
652 			if (maxargs)
653 				mdoc->next = ROFF_NEXT_CHILD;
654 			break;
655 		}
656 
657 		/*
658 		 * Explicit blocks close out description lines, but
659 		 * even those can get broken together with a child.
660 		 */
661 
662 		if (n->tok == MDOC_Nd) {
663 			if (later != NULL)
664 				n->flags |= NODE_BROKEN | NODE_ENDED;
665 			else
666 				rew_last(mdoc, n);
667 			continue;
668 		}
669 
670 		/* Breaking an open sub block. */
671 
672 		break_intermediate(mdoc->last, body);
673 		n->flags |= NODE_BROKEN;
674 		if (later == NULL)
675 			later = n;
676 	}
677 
678 	if (body == NULL) {
679 		mandoc_msg(MANDOCERR_BLK_NOTOPEN, line, ppos,
680 		    "%s", roff_name[tok]);
681 		if (maxargs && endbody == NULL) {
682 			/*
683 			 * Stray .Ec without previous .Eo:
684 			 * Break the output line, keep the arguments.
685 			 */
686 			roff_elem_alloc(mdoc, line, ppos, ROFF_br);
687 			rew_elem(mdoc, ROFF_br);
688 		}
689 	} else if (endbody == NULL) {
690 		rew_last(mdoc, body);
691 		if (maxargs)
692 			mdoc_tail_alloc(mdoc, line, ppos, atok);
693 	}
694 
695 	if ((mdoc_macro(tok)->flags & MDOC_PARSED) == 0) {
696 		if (buf[*pos] != '\0')
697 			mandoc_msg(MANDOCERR_ARG_SKIP, line, ppos,
698 			    "%s %s", roff_name[tok], buf + *pos);
699 		if (endbody == NULL && n != NULL)
700 			rew_pending(mdoc, n);
701 
702 		/*
703 		 * Restore the fill mode that was set before the display.
704 		 * This needs to be done here rather than during validation
705 		 * such that subsequent nodes get the right flags.
706 		 */
707 
708 		if (tok == MDOC_Ed && body != NULL) {
709 			if (body->flags & NODE_NOFILL)
710 				mdoc->flags |= ROFF_NOFILL;
711 			else
712 				mdoc->flags &= ~ROFF_NOFILL;
713 		}
714 		return;
715 	}
716 
717 	if (endbody != NULL)
718 		n = endbody;
719 
720 	ntok = TOKEN_NONE;
721 	for (j = 0; ; j++) {
722 		lastarg = *pos;
723 
724 		if (j == maxargs && n != NULL)
725 			rew_last(mdoc, n);
726 
727 		ac = mdoc_args(mdoc, line, pos, buf, tok, &p);
728 		if (ac == ARGS_PUNCT || ac == ARGS_EOLN)
729 			break;
730 
731 		ntok = lookup(mdoc, tok, line, lastarg, p);
732 
733 		if (ntok == TOKEN_NONE) {
734 			dword(mdoc, line, lastarg, p, DELIM_MAX,
735 			    mdoc_macro(tok)->flags & MDOC_JOIN);
736 			if (ac == ARGS_ALLOC)
737 				free(p);
738 			continue;
739 		}
740 		if (ac == ARGS_ALLOC)
741 			free(p);
742 
743 		if (n != NULL)
744 			rew_last(mdoc, n);
745 		mdoc->flags &= ~MDOC_NEWLINE;
746 		(*mdoc_macro(ntok)->fp)(mdoc, ntok, line, lastarg, pos, buf);
747 		break;
748 	}
749 
750 	if (n != NULL) {
751 		pending = 0;
752 		if (ntok != TOKEN_NONE && n->flags & NODE_BROKEN) {
753 			target = n;
754 			do
755 				target = target->parent;
756 			while ( ! (target->flags & NODE_ENDED));
757 			pending = find_pending(mdoc, ntok, line, ppos, target);
758 		}
759 		if ( ! pending)
760 			rew_pending(mdoc, n);
761 	}
762 	if (nl)
763 		append_delims(mdoc, line, pos, buf);
764 }
765 
766 static void
in_line(MACRO_PROT_ARGS)767 in_line(MACRO_PROT_ARGS)
768 {
769 	int		 la, scope, cnt, firstarg, mayopen, nc, nl;
770 	enum roff_tok	 ntok;
771 	enum margserr	 ac;
772 	enum mdelim	 d;
773 	struct mdoc_arg	*arg;
774 	char		*p;
775 
776 	nl = MDOC_NEWLINE & mdoc->flags;
777 
778 	/*
779 	 * Whether we allow ignored elements (those without content,
780 	 * usually because of reserved words) to squeak by.
781 	 */
782 
783 	switch (tok) {
784 	case MDOC_An:
785 	case MDOC_Ar:
786 	case MDOC_Fl:
787 	case MDOC_Mt:
788 	case MDOC_Nm:
789 	case MDOC_Pa:
790 		nc = 1;
791 		break;
792 	default:
793 		nc = 0;
794 		break;
795 	}
796 
797 	mdoc_argv(mdoc, line, tok, &arg, pos, buf);
798 
799 	d = DELIM_NONE;
800 	firstarg = 1;
801 	mayopen = 1;
802 	for (cnt = scope = 0;; ) {
803 		la = *pos;
804 		ac = mdoc_args(mdoc, line, pos, buf, tok, &p);
805 
806 		/*
807 		 * At the end of a macro line,
808 		 * opening delimiters do not suppress spacing.
809 		 */
810 
811 		if (ac == ARGS_EOLN) {
812 			if (d == DELIM_OPEN)
813 				mdoc->last->flags &= ~NODE_DELIMO;
814 			break;
815 		}
816 
817 		/*
818 		 * The rest of the macro line is only punctuation,
819 		 * to be handled by append_delims().
820 		 * If there were no other arguments,
821 		 * do not allow the first one to suppress spacing,
822 		 * even if it turns out to be a closing one.
823 		 */
824 
825 		if (ac == ARGS_PUNCT) {
826 			if (cnt == 0 && (nc == 0 || tok == MDOC_An))
827 				mdoc->flags |= MDOC_NODELIMC;
828 			break;
829 		}
830 
831 		ntok = (tok == MDOC_Fn && !cnt) ?
832 		    TOKEN_NONE : lookup(mdoc, tok, line, la, p);
833 
834 		/*
835 		 * In this case, we've located a submacro and must
836 		 * execute it.  Close out scope, if open.  If no
837 		 * elements have been generated, either create one (nc)
838 		 * or raise a warning.
839 		 */
840 
841 		if (ntok != TOKEN_NONE) {
842 			if (scope)
843 				rew_elem(mdoc, tok);
844 			if (nc && ! cnt) {
845 				mdoc_elem_alloc(mdoc, line, ppos, tok, arg);
846 				rew_last(mdoc, mdoc->last);
847 			} else if ( ! nc && ! cnt) {
848 				mdoc_argv_free(arg);
849 				mandoc_msg(MANDOCERR_MACRO_EMPTY,
850 				    line, ppos, "%s", roff_name[tok]);
851 			}
852 			(*mdoc_macro(ntok)->fp)(mdoc, ntok,
853 			    line, la, pos, buf);
854 			if (nl)
855 				append_delims(mdoc, line, pos, buf);
856 			if (ac == ARGS_ALLOC)
857 				free(p);
858 			return;
859 		}
860 
861 		/*
862 		 * Handle punctuation.  Set up our scope, if a word;
863 		 * rewind the scope, if a delimiter; then append the word.
864 		 */
865 
866 		if ((d = mdoc_isdelim(p)) != DELIM_NONE) {
867 			/*
868 			 * If we encounter closing punctuation, no word
869 			 * has been emitted, no scope is open, and we're
870 			 * allowed to have an empty element, then start
871 			 * a new scope.
872 			 */
873 			if ((d == DELIM_CLOSE ||
874 			     (d == DELIM_MIDDLE && tok == MDOC_Fl)) &&
875 			    !cnt && !scope && nc && mayopen) {
876 				mdoc_elem_alloc(mdoc, line, ppos, tok, arg);
877 				scope = 1;
878 				cnt++;
879 				if (tok == MDOC_Nm)
880 					mayopen = 0;
881 			}
882 			/*
883 			 * Close out our scope, if one is open, before
884 			 * any punctuation.
885 			 */
886 			if (scope && tok != MDOC_Lk) {
887 				rew_elem(mdoc, tok);
888 				scope = 0;
889 				if (tok == MDOC_Fn)
890 					mayopen = 0;
891 			}
892 		} else if (mayopen && !scope) {
893 			mdoc_elem_alloc(mdoc, line, ppos, tok, arg);
894 			scope = 1;
895 			cnt++;
896 		}
897 
898 		dword(mdoc, line, la, p, d,
899 		    mdoc_macro(tok)->flags & MDOC_JOIN);
900 
901 		if (ac == ARGS_ALLOC)
902 			free(p);
903 
904 		/*
905 		 * If the first argument is a closing delimiter,
906 		 * do not suppress spacing before it.
907 		 */
908 
909 		if (firstarg && d == DELIM_CLOSE && !nc)
910 			mdoc->last->flags &= ~NODE_DELIMC;
911 		firstarg = 0;
912 
913 		/*
914 		 * `Fl' macros have their scope re-opened with each new
915 		 * word so that the `-' can be added to each one without
916 		 * having to parse out spaces.
917 		 */
918 		if (scope && tok == MDOC_Fl) {
919 			rew_elem(mdoc, tok);
920 			scope = 0;
921 		}
922 	}
923 
924 	if (scope && tok != MDOC_Lk) {
925 		rew_elem(mdoc, tok);
926 		scope = 0;
927 	}
928 
929 	/*
930 	 * If no elements have been collected and we're allowed to have
931 	 * empties (nc), open a scope and close it out.  Otherwise,
932 	 * raise a warning.
933 	 */
934 
935 	if ( ! cnt) {
936 		if (nc) {
937 			mdoc_elem_alloc(mdoc, line, ppos, tok, arg);
938 			rew_last(mdoc, mdoc->last);
939 		} else {
940 			mdoc_argv_free(arg);
941 			mandoc_msg(MANDOCERR_MACRO_EMPTY,
942 			    line, ppos, "%s", roff_name[tok]);
943 		}
944 	}
945 	if (nl)
946 		append_delims(mdoc, line, pos, buf);
947 	if (scope)
948 		rew_elem(mdoc, tok);
949 }
950 
951 static void
blk_full(MACRO_PROT_ARGS)952 blk_full(MACRO_PROT_ARGS)
953 {
954 	struct mdoc_arg	 *arg;
955 	struct roff_node *blk; /* Our own or a broken block. */
956 	struct roff_node *head; /* Our own head. */
957 	struct roff_node *body; /* Our own body. */
958 	struct roff_node *n;
959 	char		 *p;
960 	size_t		  iarg;
961 	int		  done, la, nl, parsed;
962 	enum margserr	  ac, lac;
963 
964 	nl = MDOC_NEWLINE & mdoc->flags;
965 
966 	if (buf[*pos] == '\0' && (tok == MDOC_Sh || tok == MDOC_Ss)) {
967 		mandoc_msg(MANDOCERR_MACRO_EMPTY,
968 		    line, ppos, "%s", roff_name[tok]);
969 		return;
970 	}
971 
972 	if ((mdoc_macro(tok)->flags & MDOC_EXPLICIT) == 0) {
973 
974 		/* Here, tok is one of Sh Ss Nm Nd It. */
975 
976 		blk = NULL;
977 		for (n = mdoc->last; n != NULL; n = n->parent) {
978 			if (n->flags & NODE_ENDED) {
979 				if ( ! (n->flags & NODE_VALID))
980 					n->flags |= NODE_BROKEN;
981 				continue;
982 			}
983 			if (n->type != ROFFT_BLOCK)
984 				continue;
985 
986 			if (tok == MDOC_It && n->tok == MDOC_Bl) {
987 				if (blk != NULL) {
988 					mandoc_msg(MANDOCERR_BLK_BROKEN,
989 					    line, ppos, "It breaks %s",
990 					    roff_name[blk->tok]);
991 					rew_pending(mdoc, blk);
992 				}
993 				break;
994 			}
995 
996 			if (mdoc_macro(n->tok)->flags & MDOC_EXPLICIT) {
997 				switch (tok) {
998 				case MDOC_Sh:
999 				case MDOC_Ss:
1000 					mandoc_msg(MANDOCERR_BLK_BROKEN,
1001 					    line, ppos,
1002 					    "%s breaks %s", roff_name[tok],
1003 					    roff_name[n->tok]);
1004 					rew_pending(mdoc, n);
1005 					n = mdoc->last;
1006 					continue;
1007 				case MDOC_It:
1008 					/* Delay in case it's astray. */
1009 					blk = n;
1010 					continue;
1011 				default:
1012 					break;
1013 				}
1014 				break;
1015 			}
1016 
1017 			/* Here, n is one of Sh Ss Nm Nd It. */
1018 
1019 			if (tok != MDOC_Sh && (n->tok == MDOC_Sh ||
1020 			    (tok != MDOC_Ss && (n->tok == MDOC_Ss ||
1021 			     (tok != MDOC_It && n->tok == MDOC_It)))))
1022 				break;
1023 
1024 			/* Item breaking an explicit block. */
1025 
1026 			if (blk != NULL) {
1027 				mandoc_msg(MANDOCERR_BLK_BROKEN, line, ppos,
1028 				    "It breaks %s", roff_name[blk->tok]);
1029 				rew_pending(mdoc, blk);
1030 				blk = NULL;
1031 			}
1032 
1033 			/* Close out prior implicit scopes. */
1034 
1035 			rew_pending(mdoc, n);
1036 		}
1037 
1038 		/* Skip items outside lists. */
1039 
1040 		if (tok == MDOC_It && (n == NULL || n->tok != MDOC_Bl)) {
1041 			mandoc_msg(MANDOCERR_IT_STRAY,
1042 			    line, ppos, "It %s", buf + *pos);
1043 			roff_elem_alloc(mdoc, line, ppos, ROFF_br);
1044 			rew_elem(mdoc, ROFF_br);
1045 			return;
1046 		}
1047 	}
1048 
1049 	/*
1050 	 * This routine accommodates implicitly- and explicitly-scoped
1051 	 * macro openings.  Implicit ones first close out prior scope
1052 	 * (seen above).  Delay opening the head until necessary to
1053 	 * allow leading punctuation to print.  Special consideration
1054 	 * for `It -column', which has phrase-part syntax instead of
1055 	 * regular child nodes.
1056 	 */
1057 
1058 	switch (tok) {
1059 	case MDOC_Sh:
1060 		mdoc->flags &= ~ROFF_NOFILL;
1061 		break;
1062 	case MDOC_Ss:
1063 		mdoc->flags |= ROFF_NONOFILL;
1064 		break;
1065 	default:
1066 		break;
1067 	}
1068 	mdoc_argv(mdoc, line, tok, &arg, pos, buf);
1069 	blk = mdoc_block_alloc(mdoc, line, ppos, tok, arg);
1070 	head = body = NULL;
1071 
1072 	/*
1073 	 * Exception: Heads of `It' macros in `-diag' lists are not
1074 	 * parsed, even though `It' macros in general are parsed.
1075 	 */
1076 
1077 	parsed = tok != MDOC_It ||
1078 	    mdoc->last->parent->tok != MDOC_Bl ||
1079 	    mdoc->last->parent->norm->Bl.type != LIST_diag;
1080 
1081 	/*
1082 	 * The `Nd' macro has all arguments in its body: it's a hybrid
1083 	 * of block partial-explicit and full-implicit.  Stupid.
1084 	 */
1085 
1086 	if (tok == MDOC_Nd) {
1087 		head = roff_head_alloc(mdoc, line, ppos, tok);
1088 		rew_last(mdoc, head);
1089 		body = roff_body_alloc(mdoc, line, ppos, tok);
1090 	}
1091 
1092 	if (tok == MDOC_Bk)
1093 		mdoc->flags |= MDOC_KEEP;
1094 
1095 	ac = ARGS_EOLN;
1096 	for (;;) {
1097 
1098 		/*
1099 		 * If we are right after a tab character,
1100 		 * do not parse the first word for macros.
1101 		 */
1102 
1103 		if (mdoc->flags & MDOC_PHRASEQN) {
1104 			mdoc->flags &= ~MDOC_PHRASEQN;
1105 			mdoc->flags |= MDOC_PHRASEQF;
1106 		}
1107 
1108 		la = *pos;
1109 		lac = ac;
1110 		ac = mdoc_args(mdoc, line, pos, buf, tok, &p);
1111 		if (ac == ARGS_EOLN) {
1112 			if (lac != ARGS_PHRASE ||
1113 			    ! (mdoc->flags & MDOC_PHRASEQF))
1114 				break;
1115 
1116 			/*
1117 			 * This line ends in a tab; start the next
1118 			 * column now, with a leading blank.
1119 			 */
1120 
1121 			if (body != NULL)
1122 				rew_last(mdoc, body);
1123 			body = roff_body_alloc(mdoc, line, ppos, tok);
1124 			roff_word_alloc(mdoc, line, ppos, "\\&");
1125 			break;
1126 		}
1127 
1128 		if (tok == MDOC_Bd || tok == MDOC_Bk) {
1129 			mandoc_msg(MANDOCERR_ARG_EXCESS, line, la,
1130 			    "%s ... %s", roff_name[tok], buf + la);
1131 			if (ac == ARGS_ALLOC)
1132 				free(p);
1133 			break;
1134 		}
1135 		if (tok == MDOC_Rs) {
1136 			mandoc_msg(MANDOCERR_ARG_SKIP,
1137 			    line, la, "Rs %s", buf + la);
1138 			if (ac == ARGS_ALLOC)
1139 				free(p);
1140 			break;
1141 		}
1142 		if (ac == ARGS_PUNCT)
1143 			break;
1144 
1145 		/*
1146 		 * Emit leading punctuation (i.e., punctuation before
1147 		 * the ROFFT_HEAD) for non-phrase types.
1148 		 */
1149 
1150 		if (head == NULL &&
1151 		    ac != ARGS_PHRASE &&
1152 		    mdoc_isdelim(p) == DELIM_OPEN) {
1153 			dword(mdoc, line, la, p, DELIM_OPEN, 0);
1154 			if (ac == ARGS_ALLOC)
1155 				free(p);
1156 			continue;
1157 		}
1158 
1159 		/* Open a head if one hasn't been opened. */
1160 
1161 		if (head == NULL)
1162 			head = roff_head_alloc(mdoc, line, ppos, tok);
1163 
1164 		if (ac == ARGS_PHRASE) {
1165 
1166 			/*
1167 			 * If we haven't opened a body yet, rewind the
1168 			 * head; if we have, rewind that instead.
1169 			 */
1170 
1171 			rew_last(mdoc, body == NULL ? head : body);
1172 			body = roff_body_alloc(mdoc, line, ppos, tok);
1173 
1174 			/* Process to the tab or to the end of the line. */
1175 
1176 			mdoc->flags |= MDOC_PHRASE;
1177 			parse_rest(mdoc, TOKEN_NONE, line, &la, buf);
1178 			mdoc->flags &= ~MDOC_PHRASE;
1179 
1180 			/* There may have been `Ta' macros. */
1181 
1182 			while (body->next != NULL)
1183 				body = body->next;
1184 			continue;
1185 		}
1186 
1187 		done = macro_or_word(mdoc, tok, line, la, pos, buf, p, parsed);
1188 		if (ac == ARGS_ALLOC)
1189 			free(p);
1190 		if (done)
1191 			break;
1192 	}
1193 
1194 	if (blk->flags & NODE_VALID)
1195 		return;
1196 	if (head == NULL)
1197 		head = roff_head_alloc(mdoc, line, ppos, tok);
1198 	if (nl && tok != MDOC_Bd && tok != MDOC_Bl && tok != MDOC_Rs)
1199 		append_delims(mdoc, line, pos, buf);
1200 	if (body != NULL)
1201 		goto out;
1202 	if (find_pending(mdoc, tok, line, ppos, head))
1203 		return;
1204 
1205 	/* Close out scopes to remain in a consistent state. */
1206 
1207 	rew_last(mdoc, head);
1208 	body = roff_body_alloc(mdoc, line, ppos, tok);
1209 	if (tok == MDOC_Ss)
1210 		mdoc->flags &= ~ROFF_NONOFILL;
1211 
1212 	/*
1213 	 * Set up fill mode for display blocks.
1214 	 * This needs to be done here up front rather than during
1215 	 * validation such that child nodes get the right flags.
1216 	 */
1217 
1218 	if (tok == MDOC_Bd && arg != NULL) {
1219 		for (iarg = 0; iarg < arg->argc; iarg++) {
1220 			switch (arg->argv[iarg].arg) {
1221 			case MDOC_Unfilled:
1222 			case MDOC_Literal:
1223 				mdoc->flags |= ROFF_NOFILL;
1224 				break;
1225 			case MDOC_Filled:
1226 			case MDOC_Ragged:
1227 			case MDOC_Centred:
1228 				mdoc->flags &= ~ROFF_NOFILL;
1229 				break;
1230 			default:
1231 				continue;
1232 			}
1233 			break;
1234 		}
1235 	}
1236 out:
1237 	if (mdoc->flags & MDOC_FREECOL) {
1238 		rew_last(mdoc, body);
1239 		rew_last(mdoc, blk);
1240 		mdoc->flags &= ~MDOC_FREECOL;
1241 	}
1242 }
1243 
1244 static void
blk_part_imp(MACRO_PROT_ARGS)1245 blk_part_imp(MACRO_PROT_ARGS)
1246 {
1247 	int		  done, la, nl;
1248 	enum margserr	  ac;
1249 	char		 *p;
1250 	struct roff_node *blk; /* saved block context */
1251 	struct roff_node *body; /* saved body context */
1252 	struct roff_node *n;
1253 
1254 	nl = MDOC_NEWLINE & mdoc->flags;
1255 
1256 	/*
1257 	 * A macro that spans to the end of the line.  This is generally
1258 	 * (but not necessarily) called as the first macro.  The block
1259 	 * has a head as the immediate child, which is always empty,
1260 	 * followed by zero or more opening punctuation nodes, then the
1261 	 * body (which may be empty, depending on the macro), then zero
1262 	 * or more closing punctuation nodes.
1263 	 */
1264 
1265 	blk = mdoc_block_alloc(mdoc, line, ppos, tok, NULL);
1266 	rew_last(mdoc, roff_head_alloc(mdoc, line, ppos, tok));
1267 
1268 	/*
1269 	 * Open the body scope "on-demand", that is, after we've
1270 	 * processed all our the leading delimiters (open parenthesis,
1271 	 * etc.).
1272 	 */
1273 
1274 	for (body = NULL; ; ) {
1275 		la = *pos;
1276 		ac = mdoc_args(mdoc, line, pos, buf, tok, &p);
1277 		if (ac == ARGS_EOLN || ac == ARGS_PUNCT)
1278 			break;
1279 
1280 		if (body == NULL && mdoc_isdelim(p) == DELIM_OPEN) {
1281 			dword(mdoc, line, la, p, DELIM_OPEN, 0);
1282 			if (ac == ARGS_ALLOC)
1283 				free(p);
1284 			continue;
1285 		}
1286 
1287 		if (body == NULL)
1288 			body = roff_body_alloc(mdoc, line, ppos, tok);
1289 
1290 		done = macro_or_word(mdoc, tok, line, la, pos, buf, p, 1);
1291 		if (ac == ARGS_ALLOC)
1292 			free(p);
1293 		if (done)
1294 			break;
1295 	}
1296 	if (body == NULL)
1297 		body = roff_body_alloc(mdoc, line, ppos, tok);
1298 
1299 	if (find_pending(mdoc, tok, line, ppos, body))
1300 		return;
1301 
1302 	rew_last(mdoc, body);
1303 	if (nl)
1304 		append_delims(mdoc, line, pos, buf);
1305 	rew_pending(mdoc, blk);
1306 
1307 	/* Move trailing .Ns out of scope. */
1308 
1309 	for (n = body->child; n && n->next; n = n->next)
1310 		/* Do nothing. */ ;
1311 	if (n && n->tok == MDOC_Ns)
1312 		roff_node_relink(mdoc, n);
1313 }
1314 
1315 static void
blk_part_exp(MACRO_PROT_ARGS)1316 blk_part_exp(MACRO_PROT_ARGS)
1317 {
1318 	int		  done, la, nl;
1319 	enum margserr	  ac;
1320 	struct roff_node *head; /* keep track of head */
1321 	char		 *p;
1322 
1323 	nl = MDOC_NEWLINE & mdoc->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 	roff_block_alloc(mdoc, line, ppos, tok);
1332 	head = NULL;
1333 	for (;;) {
1334 		la = *pos;
1335 		ac = mdoc_args(mdoc, line, pos, buf, tok, &p);
1336 		if (ac == ARGS_PUNCT || ac == ARGS_EOLN)
1337 			break;
1338 
1339 		/* Flush out leading punctuation. */
1340 
1341 		if (head == NULL && mdoc_isdelim(p) == DELIM_OPEN) {
1342 			dword(mdoc, line, la, p, DELIM_OPEN, 0);
1343 			if (ac == ARGS_ALLOC)
1344 				free(p);
1345 			continue;
1346 		}
1347 
1348 		if (head == NULL) {
1349 			head = roff_head_alloc(mdoc, line, ppos, tok);
1350 			if (tok == MDOC_Eo)  /* Not parsed. */
1351 				dword(mdoc, line, la, p, DELIM_MAX, 0);
1352 			rew_last(mdoc, head);
1353 			roff_body_alloc(mdoc, line, ppos, tok);
1354 			if (tok == MDOC_Eo) {
1355 				if (ac == ARGS_ALLOC)
1356 					free(p);
1357 				continue;
1358 			}
1359 		}
1360 
1361 		done = macro_or_word(mdoc, tok, line, la, pos, buf, p, 1);
1362 		if (ac == ARGS_ALLOC)
1363 			free(p);
1364 		if (done)
1365 			break;
1366 	}
1367 
1368 	/* Clean-up to leave in a consistent state. */
1369 
1370 	if (head == NULL) {
1371 		rew_last(mdoc, roff_head_alloc(mdoc, line, ppos, tok));
1372 		roff_body_alloc(mdoc, line, ppos, tok);
1373 	}
1374 	if (nl)
1375 		append_delims(mdoc, line, pos, buf);
1376 }
1377 
1378 static void
in_line_argn(MACRO_PROT_ARGS)1379 in_line_argn(MACRO_PROT_ARGS)
1380 {
1381 	struct mdoc_arg	*arg;
1382 	char		*p;
1383 	enum margserr	 ac;
1384 	enum roff_tok	 ntok;
1385 	int		 state; /* arg#; -1: not yet open; -2: closed */
1386 	int		 la, maxargs, nl;
1387 
1388 	nl = mdoc->flags & MDOC_NEWLINE;
1389 
1390 	/*
1391 	 * A line macro that has a fixed number of arguments (maxargs).
1392 	 * Only open the scope once the first non-leading-punctuation is
1393 	 * found (unless MDOC_IGNDELIM is noted, like in `Pf'), then
1394 	 * keep it open until the maximum number of arguments are
1395 	 * exhausted.
1396 	 */
1397 
1398 	switch (tok) {
1399 	case MDOC_Ap:
1400 	case MDOC_Ns:
1401 	case MDOC_Ux:
1402 		maxargs = 0;
1403 		break;
1404 	case MDOC_Bx:
1405 	case MDOC_Es:
1406 	case MDOC_Xr:
1407 		maxargs = 2;
1408 		break;
1409 	default:
1410 		maxargs = 1;
1411 		break;
1412 	}
1413 
1414 	mdoc_argv(mdoc, line, tok, &arg, pos, buf);
1415 
1416 	state = -1;
1417 	p = NULL;
1418 	for (;;) {
1419 		la = *pos;
1420 		ac = mdoc_args(mdoc, line, pos, buf, tok, &p);
1421 
1422 		if ((ac == ARGS_WORD || ac == ARGS_ALLOC) && state == -1 &&
1423 		    (mdoc_macro(tok)->flags & MDOC_IGNDELIM) == 0 &&
1424 		    mdoc_isdelim(p) == DELIM_OPEN) {
1425 			dword(mdoc, line, la, p, DELIM_OPEN, 0);
1426 			if (ac == ARGS_ALLOC)
1427 				free(p);
1428 			continue;
1429 		}
1430 
1431 		if (state == -1 && tok != MDOC_In &&
1432 		    tok != MDOC_St && tok != MDOC_Xr) {
1433 			mdoc_elem_alloc(mdoc, line, ppos, tok, arg);
1434 			state = 0;
1435 		}
1436 
1437 		if (ac == ARGS_PUNCT || ac == ARGS_EOLN) {
1438 			if (abs(state) < 2 && tok == MDOC_Pf)
1439 				mandoc_msg(MANDOCERR_PF_SKIP,
1440 				    line, ppos, "Pf %s",
1441 				    p == NULL ? "at eol" : p);
1442 			break;
1443 		}
1444 
1445 		if (state == maxargs) {
1446 			rew_elem(mdoc, tok);
1447 			state = -2;
1448 		}
1449 
1450 		ntok = (tok == MDOC_Pf && state == 0) ?
1451 		    TOKEN_NONE : lookup(mdoc, tok, line, la, p);
1452 
1453 		if (ntok != TOKEN_NONE) {
1454 			if (state >= 0) {
1455 				rew_elem(mdoc, tok);
1456 				state = -2;
1457 			}
1458 			(*mdoc_macro(ntok)->fp)(mdoc, ntok,
1459 			    line, la, pos, buf);
1460 			if (ac == ARGS_ALLOC)
1461 				free(p);
1462 			break;
1463 		}
1464 
1465 		if (mdoc_macro(tok)->flags & MDOC_IGNDELIM ||
1466 		    mdoc_isdelim(p) == DELIM_NONE) {
1467 			if (state == -1) {
1468 				mdoc_elem_alloc(mdoc, line, ppos, tok, arg);
1469 				state = 1;
1470 			} else if (state >= 0)
1471 				state++;
1472 		} else if (state >= 0) {
1473 			rew_elem(mdoc, tok);
1474 			state = -2;
1475 		}
1476 
1477 		dword(mdoc, line, la, p, DELIM_MAX,
1478 		    mdoc_macro(tok)->flags & MDOC_JOIN);
1479 		if (ac == ARGS_ALLOC)
1480 			free(p);
1481 		p = mdoc->last->string;
1482 	}
1483 
1484 	if (state == -1) {
1485 		mandoc_msg(MANDOCERR_MACRO_EMPTY,
1486 		    line, ppos, "%s", roff_name[tok]);
1487 		return;
1488 	}
1489 
1490 	if (state == 0 && tok == MDOC_Pf)
1491 		append_delims(mdoc, line, pos, buf);
1492 	if (state >= 0)
1493 		rew_elem(mdoc, tok);
1494 	if (nl)
1495 		append_delims(mdoc, line, pos, buf);
1496 }
1497 
1498 static void
in_line_eoln(MACRO_PROT_ARGS)1499 in_line_eoln(MACRO_PROT_ARGS)
1500 {
1501 	struct roff_node	*n;
1502 	struct mdoc_arg		*arg;
1503 
1504 	if ((tok == MDOC_Pp || tok == MDOC_Lp) &&
1505 	    ! (mdoc->flags & MDOC_SYNOPSIS)) {
1506 		n = mdoc->last;
1507 		if (mdoc->next == ROFF_NEXT_SIBLING)
1508 			n = n->parent;
1509 		if (n->tok == MDOC_Nm)
1510 			rew_last(mdoc, n->parent);
1511 	}
1512 
1513 	if (buf[*pos] == '\0' &&
1514 	    (tok == MDOC_Fd || *roff_name[tok] == '%')) {
1515 		mandoc_msg(MANDOCERR_MACRO_EMPTY,
1516 		    line, ppos, "%s", roff_name[tok]);
1517 		return;
1518 	}
1519 
1520 	mdoc_argv(mdoc, line, tok, &arg, pos, buf);
1521 	mdoc_elem_alloc(mdoc, line, ppos, tok, arg);
1522 	if (parse_rest(mdoc, tok, line, pos, buf))
1523 		return;
1524 	rew_elem(mdoc, tok);
1525 }
1526 
1527 /*
1528  * The simplest argument parser available: Parse the remaining
1529  * words until the end of the phrase or line and return 0
1530  * or until the next macro, call that macro, and return 1.
1531  */
1532 static int
parse_rest(struct roff_man * mdoc,enum roff_tok tok,int line,int * pos,char * buf)1533 parse_rest(struct roff_man *mdoc, enum roff_tok tok,
1534     int line, int *pos, char *buf)
1535 {
1536 	char		*p;
1537 	int		 done, la;
1538 	enum margserr	 ac;
1539 
1540 	for (;;) {
1541 		la = *pos;
1542 		ac = mdoc_args(mdoc, line, pos, buf, tok, &p);
1543 		if (ac == ARGS_EOLN)
1544 			return 0;
1545 		done = macro_or_word(mdoc, tok, line, la, pos, buf, p, 1);
1546 		if (ac == ARGS_ALLOC)
1547 			free(p);
1548 		if (done)
1549 			return 1;
1550 	}
1551 }
1552 
1553 static void
ctx_synopsis(MACRO_PROT_ARGS)1554 ctx_synopsis(MACRO_PROT_ARGS)
1555 {
1556 
1557 	if (~mdoc->flags & (MDOC_SYNOPSIS | MDOC_NEWLINE))
1558 		in_line(mdoc, tok, line, ppos, pos, buf);
1559 	else if (tok == MDOC_Nm)
1560 		blk_full(mdoc, tok, line, ppos, pos, buf);
1561 	else {
1562 		assert(tok == MDOC_Vt);
1563 		blk_part_imp(mdoc, tok, line, ppos, pos, buf);
1564 	}
1565 }
1566 
1567 /*
1568  * Phrases occur within `Bl -column' entries, separated by `Ta' or tabs.
1569  * They're unusual because they're basically free-form text until a
1570  * macro is encountered.
1571  */
1572 static void
phrase_ta(MACRO_PROT_ARGS)1573 phrase_ta(MACRO_PROT_ARGS)
1574 {
1575 	struct roff_node *body, *n;
1576 
1577 	/* Make sure we are in a column list or ignore this macro. */
1578 
1579 	body = NULL;
1580 	for (n = mdoc->last; n != NULL; n = n->parent) {
1581 		if (n->flags & NODE_ENDED)
1582 			continue;
1583 		if (n->tok == MDOC_It && n->type == ROFFT_BODY)
1584 			body = n;
1585 		if (n->tok == MDOC_Bl && n->end == ENDBODY_NOT)
1586 			break;
1587 	}
1588 
1589 	if (n == NULL || n->norm->Bl.type != LIST_column) {
1590 		mandoc_msg(MANDOCERR_TA_STRAY, line, ppos, "Ta");
1591 		return;
1592 	}
1593 
1594 	/* Advance to the next column. */
1595 
1596 	rew_last(mdoc, body);
1597 	roff_body_alloc(mdoc, line, ppos, MDOC_It);
1598 	parse_rest(mdoc, TOKEN_NONE, line, pos, buf);
1599 }
1600