xref: /illumos-gate/usr/src/cmd/mandoc/mdoc.c (revision 142d813a06c6f9a6142e2c276b62129a17a31a65)
1 /*	$Id: mdoc.c,v 1.274 2018/12/31 07:46:07 schwarze Exp $ */
2 /*
3  * Copyright (c) 2008, 2009, 2010, 2011 Kristaps Dzonsons <kristaps@bsd.lv>
4  * Copyright (c) 2010, 2012-2018 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 <stdarg.h>
25 #include <stdio.h>
26 #include <stdlib.h>
27 #include <string.h>
28 #include <time.h>
29 
30 #include "mandoc_aux.h"
31 #include "mandoc.h"
32 #include "roff.h"
33 #include "mdoc.h"
34 #include "libmandoc.h"
35 #include "roff_int.h"
36 #include "libmdoc.h"
37 
38 const	char *const __mdoc_argnames[MDOC_ARG_MAX] = {
39 	"split",		"nosplit",		"ragged",
40 	"unfilled",		"literal",		"file",
41 	"offset",		"bullet",		"dash",
42 	"hyphen",		"item",			"enum",
43 	"tag",			"diag",			"hang",
44 	"ohang",		"inset",		"column",
45 	"width",		"compact",		"std",
46 	"filled",		"words",		"emphasis",
47 	"symbolic",		"nested",		"centered"
48 };
49 const	char * const *mdoc_argnames = __mdoc_argnames;
50 
51 static	int		  mdoc_ptext(struct roff_man *, int, char *, int);
52 static	int		  mdoc_pmacro(struct roff_man *, int, char *, int);
53 
54 
55 /*
56  * Main parse routine.  Parses a single line -- really just hands off to
57  * the macro (mdoc_pmacro()) or text parser (mdoc_ptext()).
58  */
59 int
60 mdoc_parseln(struct roff_man *mdoc, int ln, char *buf, int offs)
61 {
62 
63 	if (mdoc->last->type != ROFFT_EQN || ln > mdoc->last->line)
64 		mdoc->flags |= MDOC_NEWLINE;
65 
66 	/*
67 	 * Let the roff nS register switch SYNOPSIS mode early,
68 	 * such that the parser knows at all times
69 	 * whether this mode is on or off.
70 	 * Note that this mode is also switched by the Sh macro.
71 	 */
72 	if (roff_getreg(mdoc->roff, "nS"))
73 		mdoc->flags |= MDOC_SYNOPSIS;
74 	else
75 		mdoc->flags &= ~MDOC_SYNOPSIS;
76 
77 	return roff_getcontrol(mdoc->roff, buf, &offs) ?
78 	    mdoc_pmacro(mdoc, ln, buf, offs) :
79 	    mdoc_ptext(mdoc, ln, buf, offs);
80 }
81 
82 void
83 mdoc_tail_alloc(struct roff_man *mdoc, int line, int pos, enum roff_tok tok)
84 {
85 	struct roff_node *p;
86 
87 	p = roff_node_alloc(mdoc, line, pos, ROFFT_TAIL, tok);
88 	roff_node_append(mdoc, p);
89 	mdoc->next = ROFF_NEXT_CHILD;
90 }
91 
92 struct roff_node *
93 mdoc_endbody_alloc(struct roff_man *mdoc, int line, int pos,
94     enum roff_tok tok, struct roff_node *body)
95 {
96 	struct roff_node *p;
97 
98 	body->flags |= NODE_ENDED;
99 	body->parent->flags |= NODE_ENDED;
100 	p = roff_node_alloc(mdoc, line, pos, ROFFT_BODY, tok);
101 	p->body = body;
102 	p->norm = body->norm;
103 	p->end = ENDBODY_SPACE;
104 	roff_node_append(mdoc, p);
105 	mdoc->next = ROFF_NEXT_SIBLING;
106 	return p;
107 }
108 
109 struct roff_node *
110 mdoc_block_alloc(struct roff_man *mdoc, int line, int pos,
111     enum roff_tok tok, struct mdoc_arg *args)
112 {
113 	struct roff_node *p;
114 
115 	p = roff_node_alloc(mdoc, line, pos, ROFFT_BLOCK, tok);
116 	p->args = args;
117 	if (p->args)
118 		(args->refcnt)++;
119 
120 	switch (tok) {
121 	case MDOC_Bd:
122 	case MDOC_Bf:
123 	case MDOC_Bl:
124 	case MDOC_En:
125 	case MDOC_Rs:
126 		p->norm = mandoc_calloc(1, sizeof(union mdoc_data));
127 		break;
128 	default:
129 		break;
130 	}
131 	roff_node_append(mdoc, p);
132 	mdoc->next = ROFF_NEXT_CHILD;
133 	return p;
134 }
135 
136 void
137 mdoc_elem_alloc(struct roff_man *mdoc, int line, int pos,
138      enum roff_tok tok, struct mdoc_arg *args)
139 {
140 	struct roff_node *p;
141 
142 	p = roff_node_alloc(mdoc, line, pos, ROFFT_ELEM, tok);
143 	p->args = args;
144 	if (p->args)
145 		(args->refcnt)++;
146 
147 	switch (tok) {
148 	case MDOC_An:
149 		p->norm = mandoc_calloc(1, sizeof(union mdoc_data));
150 		break;
151 	default:
152 		break;
153 	}
154 	roff_node_append(mdoc, p);
155 	mdoc->next = ROFF_NEXT_CHILD;
156 }
157 
158 /*
159  * Parse free-form text, that is, a line that does not begin with the
160  * control character.
161  */
162 static int
163 mdoc_ptext(struct roff_man *mdoc, int line, char *buf, int offs)
164 {
165 	struct roff_node *n;
166 	const char	 *cp, *sp;
167 	char		 *c, *ws, *end;
168 
169 	n = mdoc->last;
170 
171 	/*
172 	 * If a column list contains plain text, assume an implicit item
173 	 * macro.  This can happen one or more times at the beginning
174 	 * of such a list, intermixed with non-It mdoc macros and with
175 	 * nodes generated on the roff level, for example by tbl.
176 	 */
177 
178 	if ((n->tok == MDOC_Bl && n->type == ROFFT_BODY &&
179 	     n->end == ENDBODY_NOT && n->norm->Bl.type == LIST_column) ||
180 	    (n->parent != NULL && n->parent->tok == MDOC_Bl &&
181 	     n->parent->norm->Bl.type == LIST_column)) {
182 		mdoc->flags |= MDOC_FREECOL;
183 		(*mdoc_macro(MDOC_It)->fp)(mdoc, MDOC_It,
184 		    line, offs, &offs, buf);
185 		return 1;
186 	}
187 
188 	/*
189 	 * Search for the beginning of unescaped trailing whitespace (ws)
190 	 * and for the first character not to be output (end).
191 	 */
192 
193 	/* FIXME: replace with strcspn(). */
194 	ws = NULL;
195 	for (c = end = buf + offs; *c; c++) {
196 		switch (*c) {
197 		case ' ':
198 			if (NULL == ws)
199 				ws = c;
200 			continue;
201 		case '\t':
202 			/*
203 			 * Always warn about trailing tabs,
204 			 * even outside literal context,
205 			 * where they should be put on the next line.
206 			 */
207 			if (NULL == ws)
208 				ws = c;
209 			/*
210 			 * Strip trailing tabs in literal context only;
211 			 * outside, they affect the next line.
212 			 */
213 			if (mdoc->flags & ROFF_NOFILL)
214 				continue;
215 			break;
216 		case '\\':
217 			/* Skip the escaped character, too, if any. */
218 			if (c[1])
219 				c++;
220 			/* FALLTHROUGH */
221 		default:
222 			ws = NULL;
223 			break;
224 		}
225 		end = c + 1;
226 	}
227 	*end = '\0';
228 
229 	if (ws)
230 		mandoc_msg(MANDOCERR_SPACE_EOL, line, (int)(ws - buf), NULL);
231 
232 	/*
233 	 * Blank lines are allowed in no-fill mode
234 	 * and cancel preceding \c,
235 	 * but add a single vertical space elsewhere.
236 	 */
237 
238 	if (buf[offs] == '\0' && (mdoc->flags & ROFF_NOFILL) == 0) {
239 		switch (mdoc->last->type) {
240 		case ROFFT_TEXT:
241 			sp = mdoc->last->string;
242 			cp = end = strchr(sp, '\0') - 2;
243 			if (cp < sp || cp[0] != '\\' || cp[1] != 'c')
244 				break;
245 			while (cp > sp && cp[-1] == '\\')
246 				cp--;
247 			if ((end - cp) % 2)
248 				break;
249 			*end = '\0';
250 			return 1;
251 		default:
252 			break;
253 		}
254 		mandoc_msg(MANDOCERR_FI_BLANK, line, (int)(c - buf), NULL);
255 		roff_elem_alloc(mdoc, line, offs, ROFF_sp);
256 		mdoc->last->flags |= NODE_VALID | NODE_ENDED;
257 		mdoc->next = ROFF_NEXT_SIBLING;
258 		return 1;
259 	}
260 
261 	roff_word_alloc(mdoc, line, offs, buf+offs);
262 
263 	if (mdoc->flags & ROFF_NOFILL)
264 		return 1;
265 
266 	/*
267 	 * End-of-sentence check.  If the last character is an unescaped
268 	 * EOS character, then flag the node as being the end of a
269 	 * sentence.  The front-end will know how to interpret this.
270 	 */
271 
272 	assert(buf < end);
273 
274 	if (mandoc_eos(buf+offs, (size_t)(end-buf-offs)))
275 		mdoc->last->flags |= NODE_EOS;
276 
277 	for (c = buf + offs; c != NULL; c = strchr(c + 1, '.')) {
278 		if (c - buf < offs + 2)
279 			continue;
280 		if (end - c < 3)
281 			break;
282 		if (c[1] != ' ' ||
283 		    isalnum((unsigned char)c[-2]) == 0 ||
284 		    isalnum((unsigned char)c[-1]) == 0 ||
285 		    (c[-2] == 'n' && c[-1] == 'c') ||
286 		    (c[-2] == 'v' && c[-1] == 's'))
287 			continue;
288 		c += 2;
289 		if (*c == ' ')
290 			c++;
291 		if (*c == ' ')
292 			c++;
293 		if (isupper((unsigned char)(*c)))
294 			mandoc_msg(MANDOCERR_EOS, line, (int)(c - buf), NULL);
295 	}
296 
297 	return 1;
298 }
299 
300 /*
301  * Parse a macro line, that is, a line beginning with the control
302  * character.
303  */
304 static int
305 mdoc_pmacro(struct roff_man *mdoc, int ln, char *buf, int offs)
306 {
307 	struct roff_node *n;
308 	const char	 *cp;
309 	size_t		  sz;
310 	enum roff_tok	  tok;
311 	int		  sv;
312 
313 	/* Determine the line macro. */
314 
315 	sv = offs;
316 	tok = TOKEN_NONE;
317 	for (sz = 0; sz < 4 && strchr(" \t\\", buf[offs]) == NULL; sz++)
318 		offs++;
319 	if (sz == 2 || sz == 3)
320 		tok = roffhash_find(mdoc->mdocmac, buf + sv, sz);
321 	if (tok == TOKEN_NONE) {
322 		mandoc_msg(MANDOCERR_MACRO, ln, sv, "%s", buf + sv - 1);
323 		return 1;
324 	}
325 
326 	/* Skip a leading escape sequence or tab. */
327 
328 	switch (buf[offs]) {
329 	case '\\':
330 		cp = buf + offs + 1;
331 		mandoc_escape(&cp, NULL, NULL);
332 		offs = cp - buf;
333 		break;
334 	case '\t':
335 		offs++;
336 		break;
337 	default:
338 		break;
339 	}
340 
341 	/* Jump to the next non-whitespace word. */
342 
343 	while (buf[offs] == ' ')
344 		offs++;
345 
346 	/*
347 	 * Trailing whitespace.  Note that tabs are allowed to be passed
348 	 * into the parser as "text", so we only warn about spaces here.
349 	 */
350 
351 	if ('\0' == buf[offs] && ' ' == buf[offs - 1])
352 		mandoc_msg(MANDOCERR_SPACE_EOL, ln, offs - 1, NULL);
353 
354 	/*
355 	 * If an initial macro or a list invocation, divert directly
356 	 * into macro processing.
357 	 */
358 
359 	n = mdoc->last;
360 	if (n == NULL || tok == MDOC_It || tok == MDOC_El) {
361 		(*mdoc_macro(tok)->fp)(mdoc, tok, ln, sv, &offs, buf);
362 		return 1;
363 	}
364 
365 	/*
366 	 * If a column list contains a non-It macro, assume an implicit
367 	 * item macro.  This can happen one or more times at the
368 	 * beginning of such a list, intermixed with text lines and
369 	 * with nodes generated on the roff level, for example by tbl.
370 	 */
371 
372 	if ((n->tok == MDOC_Bl && n->type == ROFFT_BODY &&
373 	     n->end == ENDBODY_NOT && n->norm->Bl.type == LIST_column) ||
374 	    (n->parent != NULL && n->parent->tok == MDOC_Bl &&
375 	     n->parent->norm->Bl.type == LIST_column)) {
376 		mdoc->flags |= MDOC_FREECOL;
377 		(*mdoc_macro(MDOC_It)->fp)(mdoc, MDOC_It, ln, sv, &sv, buf);
378 		return 1;
379 	}
380 
381 	/* Normal processing of a macro. */
382 
383 	(*mdoc_macro(tok)->fp)(mdoc, tok, ln, sv, &offs, buf);
384 
385 	/* In quick mode (for mandocdb), abort after the NAME section. */
386 
387 	if (mdoc->quick && MDOC_Sh == tok &&
388 	    SEC_NAME != mdoc->last->sec)
389 		return 2;
390 
391 	return 1;
392 }
393 
394 enum mdelim
395 mdoc_isdelim(const char *p)
396 {
397 
398 	if ('\0' == p[0])
399 		return DELIM_NONE;
400 
401 	if ('\0' == p[1])
402 		switch (p[0]) {
403 		case '(':
404 		case '[':
405 			return DELIM_OPEN;
406 		case '|':
407 			return DELIM_MIDDLE;
408 		case '.':
409 		case ',':
410 		case ';':
411 		case ':':
412 		case '?':
413 		case '!':
414 		case ')':
415 		case ']':
416 			return DELIM_CLOSE;
417 		default:
418 			return DELIM_NONE;
419 		}
420 
421 	if ('\\' != p[0])
422 		return DELIM_NONE;
423 
424 	if (0 == strcmp(p + 1, "."))
425 		return DELIM_CLOSE;
426 	if (0 == strcmp(p + 1, "fR|\\fP"))
427 		return DELIM_MIDDLE;
428 
429 	return DELIM_NONE;
430 }
431