]> git.cameronkatri.com Git - mandoc.git/blob - read.c
Cleanup regarding -offset and -width:
[mandoc.git] / read.c
1 /* $Id: read.c,v 1.62 2014/07/05 01:12:20 schwarze Exp $ */
2 /*
3 * Copyright (c) 2008, 2009, 2010, 2011 Kristaps Dzonsons <kristaps@bsd.lv>
4 * Copyright (c) 2010-2014 Ingo Schwarze <schwarze@openbsd.org>
5 * Copyright (c) 2010, 2012 Joerg Sonnenberger <joerg@netbsd.org>
6 *
7 * Permission to use, copy, modify, and distribute this software for any
8 * purpose with or without fee is hereby granted, provided that the above
9 * copyright notice and this permission notice appear in all copies.
10 *
11 * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
12 * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
13 * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
14 * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
15 * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
16 * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
17 * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
18 */
19 #ifdef HAVE_CONFIG_H
20 #include "config.h"
21 #endif
22
23 #ifdef HAVE_MMAP
24 # include <sys/stat.h>
25 # include <sys/mman.h>
26 #endif
27
28 #include <assert.h>
29 #include <ctype.h>
30 #include <errno.h>
31 #include <fcntl.h>
32 #include <stdarg.h>
33 #include <stdint.h>
34 #include <stdio.h>
35 #include <stdlib.h>
36 #include <string.h>
37 #include <unistd.h>
38
39 #include "mandoc.h"
40 #include "mandoc_aux.h"
41 #include "libmandoc.h"
42 #include "mdoc.h"
43 #include "man.h"
44 #include "main.h"
45
46 #define REPARSE_LIMIT 1000
47
48 struct buf {
49 char *buf; /* binary input buffer */
50 size_t sz; /* size of binary buffer */
51 };
52
53 struct mparse {
54 enum mandoclevel file_status; /* status of current parse */
55 enum mandoclevel wlevel; /* ignore messages below this */
56 int line; /* line number in the file */
57 int options; /* parser options */
58 struct man *pman; /* persistent man parser */
59 struct mdoc *pmdoc; /* persistent mdoc parser */
60 struct man *man; /* man parser */
61 struct mdoc *mdoc; /* mdoc parser */
62 struct roff *roff; /* roff parser (!NULL) */
63 char *sodest; /* filename pointed to by .so */
64 int reparse_count; /* finite interp. stack */
65 mandocmsg mmsg; /* warning/error message handler */
66 const char *file;
67 struct buf *secondary;
68 char *defos; /* default operating system */
69 };
70
71 static void resize_buf(struct buf *, size_t);
72 static void mparse_buf_r(struct mparse *, struct buf, int);
73 static void pset(const char *, int, struct mparse *);
74 static int read_whole_file(struct mparse *, const char *, int,
75 struct buf *, int *);
76 static void mparse_end(struct mparse *);
77 static void mparse_parse_buffer(struct mparse *, struct buf,
78 const char *);
79
80 static const enum mandocerr mandoclimits[MANDOCLEVEL_MAX] = {
81 MANDOCERR_OK,
82 MANDOCERR_WARNING,
83 MANDOCERR_WARNING,
84 MANDOCERR_ERROR,
85 MANDOCERR_FATAL,
86 MANDOCERR_MAX,
87 MANDOCERR_MAX
88 };
89
90 static const char * const mandocerrs[MANDOCERR_MAX] = {
91 "ok",
92
93 "generic warning",
94
95 /* related to the prologue */
96 "missing .TH macro, using \"unknown 1\"",
97 "lower case character in document title",
98 "unknown manual section",
99 "unknown manual volume or arch",
100 "missing date, using today's date",
101 "cannot parse date, using it verbatim",
102 "prologue macros out of order",
103 "duplicate prologue macro",
104 "incomplete prologue, terminated by",
105 "skipping prologue macro in body",
106
107 /* related to document structure */
108 ".so is fragile, better use ln(1)",
109 "no document body",
110 "content before first section header",
111 "first section is not \"NAME\"",
112 "bad NAME section contents",
113 "sections out of conventional order",
114 "duplicate section title",
115 "unexpected section",
116
117 /* related to macros and nesting */
118 "obsolete macro",
119 "skipping paragraph macro",
120 "moving paragraph macro out of list",
121 "skipping no-space macro",
122 "blocks badly nested",
123 "nested displays are not portable",
124 "moving content out of list",
125 ".Vt block has child macro",
126 "fill mode already enabled, skipping .fi",
127 "fill mode already disabled, skipping .nf",
128 "line scope broken",
129
130 /* related to missing macro arguments */
131 "skipping empty request",
132 "conditional request controls empty scope",
133 "skipping empty macro",
134 "empty argument, using 0n",
135 "argument count wrong",
136 "missing display type, using -ragged",
137 "list type is not the first argument",
138 "missing -width in -tag list, using 8n",
139 "empty head in list item",
140 "empty list item",
141 "missing font type, using \\fR",
142 "unknown font type, using \\fR",
143 "missing -std argument, adding it",
144
145 /* related to bad macro arguments */
146 "skipping argument",
147 "duplicate argument",
148 "duplicate display type",
149 "duplicate list type",
150 "unknown AT&T UNIX version",
151 "bad Boolean value",
152 "unknown font",
153 "unknown standard specifier",
154 "bad width argument",
155
156 /* related to plain text */
157 "blank line in non-literal context",
158 "tab in non-literal context",
159 "end of line whitespace",
160 "bad comment style",
161 "bad escape sequence",
162 "unterminated quoted string",
163
164 "generic error",
165
166 /* related to equations */
167 "unexpected equation scope closure",
168 "equation scope open on exit",
169 "overlapping equation scopes",
170 "unexpected end of equation",
171 "equation syntax error",
172
173 /* related to tables */
174 "bad table syntax",
175 "bad table option",
176 "bad table layout",
177 "no table layout cells specified",
178 "no table data cells specified",
179 "ignore data in cell",
180 "data block still open",
181 "ignoring extra data cells",
182
183 "input stack limit exceeded, infinite loop?",
184 "skipping bad character",
185 "escaped character not allowed in a name",
186 "manual name not yet set",
187 "skipping text before first section header",
188 "skipping unknown macro",
189 "NOT IMPLEMENTED, please use groff: skipping request",
190 "argument count wrong",
191 "skipping invalid content in .Rs block",
192 "skipping column outside column list",
193 "skipping end of block that is not open",
194 "missing end of block",
195 "scope open on exit",
196 "uname(3) system call failed",
197 "request requires a numeric argument",
198 "missing list type, using -item",
199 "skipping all arguments",
200 "skipping excess arguments",
201
202 "generic fatal error",
203
204 "input too large",
205 "not a manual",
206 "column syntax is inconsistent",
207 "NOT IMPLEMENTED: .Bd -file",
208 "child violates parent syntax",
209 "argument count wrong, violates syntax",
210 "NOT IMPLEMENTED: .so with absolute path or \"..\"",
211 ".so request failed",
212 "no document prologue",
213 "static buffer exhausted",
214
215 /* system errors */
216 NULL,
217 "cannot stat file",
218 "cannot read file",
219 };
220
221 static const char * const mandoclevels[MANDOCLEVEL_MAX] = {
222 "SUCCESS",
223 "RESERVED",
224 "WARNING",
225 "ERROR",
226 "FATAL",
227 "BADARG",
228 "SYSERR"
229 };
230
231
232 static void
233 resize_buf(struct buf *buf, size_t initial)
234 {
235
236 buf->sz = buf->sz > initial/2 ? 2 * buf->sz : initial;
237 buf->buf = mandoc_realloc(buf->buf, buf->sz);
238 }
239
240 static void
241 pset(const char *buf, int pos, struct mparse *curp)
242 {
243 int i;
244
245 /*
246 * Try to intuit which kind of manual parser should be used. If
247 * passed in by command-line (-man, -mdoc), then use that
248 * explicitly. If passed as -mandoc, then try to guess from the
249 * line: either skip dot-lines, use -mdoc when finding `.Dt', or
250 * default to -man, which is more lenient.
251 *
252 * Separate out pmdoc/pman from mdoc/man: the first persists
253 * through all parsers, while the latter is used per-parse.
254 */
255
256 if ('.' == buf[0] || '\'' == buf[0]) {
257 for (i = 1; buf[i]; i++)
258 if (' ' != buf[i] && '\t' != buf[i])
259 break;
260 if ('\0' == buf[i])
261 return;
262 }
263
264 if (MPARSE_MDOC & curp->options) {
265 if (NULL == curp->pmdoc)
266 curp->pmdoc = mdoc_alloc(
267 curp->roff, curp, curp->defos,
268 MPARSE_QUICK & curp->options ? 1 : 0);
269 assert(curp->pmdoc);
270 curp->mdoc = curp->pmdoc;
271 return;
272 } else if (MPARSE_MAN & curp->options) {
273 if (NULL == curp->pman)
274 curp->pman = man_alloc(curp->roff, curp,
275 MPARSE_QUICK & curp->options ? 1 : 0);
276 assert(curp->pman);
277 curp->man = curp->pman;
278 return;
279 }
280
281 if (pos >= 3 && 0 == memcmp(buf, ".Dd", 3)) {
282 if (NULL == curp->pmdoc)
283 curp->pmdoc = mdoc_alloc(
284 curp->roff, curp, curp->defos,
285 MPARSE_QUICK & curp->options ? 1 : 0);
286 assert(curp->pmdoc);
287 curp->mdoc = curp->pmdoc;
288 return;
289 }
290
291 if (NULL == curp->pman)
292 curp->pman = man_alloc(curp->roff, curp,
293 MPARSE_QUICK & curp->options ? 1 : 0);
294 assert(curp->pman);
295 curp->man = curp->pman;
296 }
297
298 /*
299 * Main parse routine for an opened file. This is called for each
300 * opened file and simply loops around the full input file, possibly
301 * nesting (i.e., with `so').
302 */
303 static void
304 mparse_buf_r(struct mparse *curp, struct buf blk, int start)
305 {
306 const struct tbl_span *span;
307 struct buf ln;
308 enum rofferr rr;
309 int i, of, rc;
310 int pos; /* byte number in the ln buffer */
311 int lnn; /* line number in the real file */
312 unsigned char c;
313
314 memset(&ln, 0, sizeof(struct buf));
315
316 lnn = curp->line;
317 pos = 0;
318
319 for (i = 0; i < (int)blk.sz; ) {
320 if (0 == pos && '\0' == blk.buf[i])
321 break;
322
323 if (start) {
324 curp->line = lnn;
325 curp->reparse_count = 0;
326 }
327
328 while (i < (int)blk.sz && (start || '\0' != blk.buf[i])) {
329
330 /*
331 * When finding an unescaped newline character,
332 * leave the character loop to process the line.
333 * Skip a preceding carriage return, if any.
334 */
335
336 if ('\r' == blk.buf[i] && i + 1 < (int)blk.sz &&
337 '\n' == blk.buf[i + 1])
338 ++i;
339 if ('\n' == blk.buf[i]) {
340 ++i;
341 ++lnn;
342 break;
343 }
344
345 /*
346 * Make sure we have space for at least
347 * one backslash and one other character
348 * and the trailing NUL byte.
349 */
350
351 if (pos + 2 >= (int)ln.sz)
352 resize_buf(&ln, 256);
353
354 /*
355 * Warn about bogus characters. If you're using
356 * non-ASCII encoding, you're screwing your
357 * readers. Since I'd rather this not happen,
358 * I'll be helpful and replace these characters
359 * with "?", so we don't display gibberish.
360 * Note to manual writers: use special characters.
361 */
362
363 c = (unsigned char) blk.buf[i];
364
365 if ( ! (isascii(c) &&
366 (isgraph(c) || isblank(c)))) {
367 mandoc_msg(MANDOCERR_BADCHAR, curp,
368 curp->line, pos, NULL);
369 i++;
370 ln.buf[pos++] = '?';
371 continue;
372 }
373
374 /* Trailing backslash = a plain char. */
375
376 if ('\\' != blk.buf[i] || i + 1 == (int)blk.sz) {
377 ln.buf[pos++] = blk.buf[i++];
378 continue;
379 }
380
381 /*
382 * Found escape and at least one other character.
383 * When it's a newline character, skip it.
384 * When there is a carriage return in between,
385 * skip that one as well.
386 */
387
388 if ('\r' == blk.buf[i + 1] && i + 2 < (int)blk.sz &&
389 '\n' == blk.buf[i + 2])
390 ++i;
391 if ('\n' == blk.buf[i + 1]) {
392 i += 2;
393 ++lnn;
394 continue;
395 }
396
397 if ('"' == blk.buf[i + 1] || '#' == blk.buf[i + 1]) {
398 i += 2;
399 /* Comment, skip to end of line */
400 for (; i < (int)blk.sz; ++i) {
401 if ('\n' == blk.buf[i]) {
402 ++i;
403 ++lnn;
404 break;
405 }
406 }
407
408 /* Backout trailing whitespaces */
409 for (; pos > 0; --pos) {
410 if (ln.buf[pos - 1] != ' ')
411 break;
412 if (pos > 2 && ln.buf[pos - 2] == '\\')
413 break;
414 }
415 break;
416 }
417
418 /* Catch escaped bogus characters. */
419
420 c = (unsigned char) blk.buf[i+1];
421
422 if ( ! (isascii(c) &&
423 (isgraph(c) || isblank(c)))) {
424 mandoc_msg(MANDOCERR_BADCHAR, curp,
425 curp->line, pos, NULL);
426 i += 2;
427 ln.buf[pos++] = '?';
428 continue;
429 }
430
431 /* Some other escape sequence, copy & cont. */
432
433 ln.buf[pos++] = blk.buf[i++];
434 ln.buf[pos++] = blk.buf[i++];
435 }
436
437 if (pos >= (int)ln.sz)
438 resize_buf(&ln, 256);
439
440 ln.buf[pos] = '\0';
441
442 /*
443 * A significant amount of complexity is contained by
444 * the roff preprocessor. It's line-oriented but can be
445 * expressed on one line, so we need at times to
446 * readjust our starting point and re-run it. The roff
447 * preprocessor can also readjust the buffers with new
448 * data, so we pass them in wholesale.
449 */
450
451 of = 0;
452
453 /*
454 * Maintain a lookaside buffer of all parsed lines. We
455 * only do this if mparse_keep() has been invoked (the
456 * buffer may be accessed with mparse_getkeep()).
457 */
458
459 if (curp->secondary) {
460 curp->secondary->buf = mandoc_realloc(
461 curp->secondary->buf,
462 curp->secondary->sz + pos + 2);
463 memcpy(curp->secondary->buf +
464 curp->secondary->sz,
465 ln.buf, pos);
466 curp->secondary->sz += pos;
467 curp->secondary->buf
468 [curp->secondary->sz] = '\n';
469 curp->secondary->sz++;
470 curp->secondary->buf
471 [curp->secondary->sz] = '\0';
472 }
473 rerun:
474 rr = roff_parseln(curp->roff, curp->line,
475 &ln.buf, &ln.sz, of, &of);
476
477 switch (rr) {
478 case ROFF_REPARSE:
479 if (REPARSE_LIMIT >= ++curp->reparse_count)
480 mparse_buf_r(curp, ln, 0);
481 else
482 mandoc_msg(MANDOCERR_ROFFLOOP, curp,
483 curp->line, pos, NULL);
484 pos = 0;
485 continue;
486 case ROFF_APPEND:
487 pos = (int)strlen(ln.buf);
488 continue;
489 case ROFF_RERUN:
490 goto rerun;
491 case ROFF_IGN:
492 pos = 0;
493 continue;
494 case ROFF_ERR:
495 assert(MANDOCLEVEL_FATAL <= curp->file_status);
496 break;
497 case ROFF_SO:
498 if (0 == (MPARSE_SO & curp->options) &&
499 (i >= (int)blk.sz || '\0' == blk.buf[i])) {
500 curp->sodest = mandoc_strdup(ln.buf + of);
501 free(ln.buf);
502 return;
503 }
504 /*
505 * We remove `so' clauses from our lookaside
506 * buffer because we're going to descend into
507 * the file recursively.
508 */
509 if (curp->secondary)
510 curp->secondary->sz -= pos + 1;
511 mparse_readfd(curp, -1, ln.buf + of);
512 if (MANDOCLEVEL_FATAL <= curp->file_status) {
513 mandoc_vmsg(MANDOCERR_SO_FAIL,
514 curp, curp->line, pos,
515 ".so %s", ln.buf + of);
516 break;
517 }
518 pos = 0;
519 continue;
520 default:
521 break;
522 }
523
524 /*
525 * If we encounter errors in the recursive parse, make
526 * sure we don't continue parsing.
527 */
528
529 if (MANDOCLEVEL_FATAL <= curp->file_status)
530 break;
531
532 /*
533 * If input parsers have not been allocated, do so now.
534 * We keep these instanced between parsers, but set them
535 * locally per parse routine since we can use different
536 * parsers with each one.
537 */
538
539 if ( ! (curp->man || curp->mdoc))
540 pset(ln.buf + of, pos - of, curp);
541
542 /*
543 * Lastly, push down into the parsers themselves. One
544 * of these will have already been set in the pset()
545 * routine.
546 * If libroff returns ROFF_TBL, then add it to the
547 * currently open parse. Since we only get here if
548 * there does exist data (see tbl_data.c), we're
549 * guaranteed that something's been allocated.
550 * Do the same for ROFF_EQN.
551 */
552
553 rc = -1;
554
555 if (ROFF_TBL == rr)
556 while (NULL != (span = roff_span(curp->roff))) {
557 rc = curp->man ?
558 man_addspan(curp->man, span) :
559 mdoc_addspan(curp->mdoc, span);
560 if (0 == rc)
561 break;
562 }
563 else if (ROFF_EQN == rr)
564 rc = curp->mdoc ?
565 mdoc_addeqn(curp->mdoc,
566 roff_eqn(curp->roff)) :
567 man_addeqn(curp->man,
568 roff_eqn(curp->roff));
569 else if (curp->man || curp->mdoc)
570 rc = curp->man ?
571 man_parseln(curp->man,
572 curp->line, ln.buf, of) :
573 mdoc_parseln(curp->mdoc,
574 curp->line, ln.buf, of);
575
576 if (0 == rc) {
577 assert(MANDOCLEVEL_FATAL <= curp->file_status);
578 break;
579 } else if (2 == rc)
580 break;
581
582 /* Temporary buffers typically are not full. */
583
584 if (0 == start && '\0' == blk.buf[i])
585 break;
586
587 /* Start the next input line. */
588
589 pos = 0;
590 }
591
592 free(ln.buf);
593 }
594
595 static int
596 read_whole_file(struct mparse *curp, const char *file, int fd,
597 struct buf *fb, int *with_mmap)
598 {
599 size_t off;
600 ssize_t ssz;
601
602 #ifdef HAVE_MMAP
603 struct stat st;
604 if (-1 == fstat(fd, &st)) {
605 curp->file_status = MANDOCLEVEL_SYSERR;
606 if (curp->mmsg)
607 (*curp->mmsg)(MANDOCERR_SYSSTAT, curp->file_status,
608 file, 0, 0, strerror(errno));
609 return(0);
610 }
611
612 /*
613 * If we're a regular file, try just reading in the whole entry
614 * via mmap(). This is faster than reading it into blocks, and
615 * since each file is only a few bytes to begin with, I'm not
616 * concerned that this is going to tank any machines.
617 */
618
619 if (S_ISREG(st.st_mode)) {
620 if (st.st_size >= (1U << 31)) {
621 curp->file_status = MANDOCLEVEL_FATAL;
622 if (curp->mmsg)
623 (*curp->mmsg)(MANDOCERR_TOOLARGE,
624 curp->file_status, file, 0, 0, NULL);
625 return(0);
626 }
627 *with_mmap = 1;
628 fb->sz = (size_t)st.st_size;
629 fb->buf = mmap(NULL, fb->sz, PROT_READ, MAP_SHARED, fd, 0);
630 if (fb->buf != MAP_FAILED)
631 return(1);
632 }
633 #endif
634
635 /*
636 * If this isn't a regular file (like, say, stdin), then we must
637 * go the old way and just read things in bit by bit.
638 */
639
640 *with_mmap = 0;
641 off = 0;
642 fb->sz = 0;
643 fb->buf = NULL;
644 for (;;) {
645 if (off == fb->sz) {
646 if (fb->sz == (1U << 31)) {
647 curp->file_status = MANDOCLEVEL_FATAL;
648 if (curp->mmsg)
649 (*curp->mmsg)(MANDOCERR_TOOLARGE,
650 curp->file_status,
651 file, 0, 0, NULL);
652 break;
653 }
654 resize_buf(fb, 65536);
655 }
656 ssz = read(fd, fb->buf + (int)off, fb->sz - off);
657 if (ssz == 0) {
658 fb->sz = off;
659 return(1);
660 }
661 if (ssz == -1) {
662 curp->file_status = MANDOCLEVEL_SYSERR;
663 if (curp->mmsg)
664 (*curp->mmsg)(MANDOCERR_SYSREAD,
665 curp->file_status, file, 0, 0,
666 strerror(errno));
667 break;
668 }
669 off += (size_t)ssz;
670 }
671
672 free(fb->buf);
673 fb->buf = NULL;
674 return(0);
675 }
676
677 static void
678 mparse_end(struct mparse *curp)
679 {
680
681 if (MANDOCLEVEL_FATAL <= curp->file_status)
682 return;
683
684 if (curp->mdoc && ! mdoc_endparse(curp->mdoc)) {
685 assert(MANDOCLEVEL_FATAL <= curp->file_status);
686 return;
687 }
688
689 if (curp->man && ! man_endparse(curp->man)) {
690 assert(MANDOCLEVEL_FATAL <= curp->file_status);
691 return;
692 }
693
694 if ( ! (curp->mdoc || curp->man || curp->sodest)) {
695 mandoc_msg(MANDOCERR_NOTMANUAL, curp, 0, 0, NULL);
696 curp->file_status = MANDOCLEVEL_FATAL;
697 return;
698 }
699
700 roff_endparse(curp->roff);
701 }
702
703 static void
704 mparse_parse_buffer(struct mparse *curp, struct buf blk, const char *file)
705 {
706 const char *svfile;
707 static int recursion_depth;
708
709 if (64 < recursion_depth) {
710 mandoc_msg(MANDOCERR_ROFFLOOP, curp, curp->line, 0, NULL);
711 return;
712 }
713
714 /* Line number is per-file. */
715 svfile = curp->file;
716 curp->file = file;
717 curp->line = 1;
718 recursion_depth++;
719
720 mparse_buf_r(curp, blk, 1);
721
722 if (0 == --recursion_depth && MANDOCLEVEL_FATAL > curp->file_status)
723 mparse_end(curp);
724
725 curp->file = svfile;
726 }
727
728 enum mandoclevel
729 mparse_readmem(struct mparse *curp, const void *buf, size_t len,
730 const char *file)
731 {
732 struct buf blk;
733
734 blk.buf = UNCONST(buf);
735 blk.sz = len;
736
737 mparse_parse_buffer(curp, blk, file);
738 return(curp->file_status);
739 }
740
741 enum mandoclevel
742 mparse_readfd(struct mparse *curp, int fd, const char *file)
743 {
744 struct buf blk;
745 int with_mmap;
746
747 if (-1 == fd && -1 == (fd = open(file, O_RDONLY, 0))) {
748 curp->file_status = MANDOCLEVEL_SYSERR;
749 if (curp->mmsg)
750 (*curp->mmsg)(MANDOCERR_SYSOPEN,
751 curp->file_status,
752 file, 0, 0, strerror(errno));
753 goto out;
754 }
755
756 /*
757 * Run for each opened file; may be called more than once for
758 * each full parse sequence if the opened file is nested (i.e.,
759 * from `so'). Simply sucks in the whole file and moves into
760 * the parse phase for the file.
761 */
762
763 if ( ! read_whole_file(curp, file, fd, &blk, &with_mmap))
764 goto out;
765
766 mparse_parse_buffer(curp, blk, file);
767
768 #ifdef HAVE_MMAP
769 if (with_mmap)
770 munmap(blk.buf, blk.sz);
771 else
772 #endif
773 free(blk.buf);
774
775 if (STDIN_FILENO != fd && -1 == close(fd))
776 perror(file);
777 out:
778 return(curp->file_status);
779 }
780
781 struct mparse *
782 mparse_alloc(int options, enum mandoclevel wlevel,
783 mandocmsg mmsg, char *defos)
784 {
785 struct mparse *curp;
786
787 assert(wlevel <= MANDOCLEVEL_FATAL);
788
789 curp = mandoc_calloc(1, sizeof(struct mparse));
790
791 curp->options = options;
792 curp->wlevel = wlevel;
793 curp->mmsg = mmsg;
794 curp->defos = defos;
795
796 curp->roff = roff_alloc(curp, options);
797 return(curp);
798 }
799
800 void
801 mparse_reset(struct mparse *curp)
802 {
803
804 roff_reset(curp->roff);
805
806 if (curp->mdoc)
807 mdoc_reset(curp->mdoc);
808 if (curp->man)
809 man_reset(curp->man);
810 if (curp->secondary)
811 curp->secondary->sz = 0;
812
813 curp->file_status = MANDOCLEVEL_OK;
814 curp->mdoc = NULL;
815 curp->man = NULL;
816
817 free(curp->sodest);
818 curp->sodest = NULL;
819 }
820
821 void
822 mparse_free(struct mparse *curp)
823 {
824
825 if (curp->pmdoc)
826 mdoc_free(curp->pmdoc);
827 if (curp->pman)
828 man_free(curp->pman);
829 if (curp->roff)
830 roff_free(curp->roff);
831 if (curp->secondary)
832 free(curp->secondary->buf);
833
834 free(curp->secondary);
835 free(curp->sodest);
836 free(curp);
837 }
838
839 void
840 mparse_result(struct mparse *curp,
841 struct mdoc **mdoc, struct man **man, char **sodest)
842 {
843
844 if (sodest && NULL != (*sodest = curp->sodest)) {
845 *mdoc = NULL;
846 *man = NULL;
847 return;
848 }
849 if (mdoc)
850 *mdoc = curp->mdoc;
851 if (man)
852 *man = curp->man;
853 }
854
855 void
856 mandoc_vmsg(enum mandocerr t, struct mparse *m,
857 int ln, int pos, const char *fmt, ...)
858 {
859 char buf[256];
860 va_list ap;
861
862 va_start(ap, fmt);
863 (void)vsnprintf(buf, sizeof(buf), fmt, ap);
864 va_end(ap);
865
866 mandoc_msg(t, m, ln, pos, buf);
867 }
868
869 void
870 mandoc_msg(enum mandocerr er, struct mparse *m,
871 int ln, int col, const char *msg)
872 {
873 enum mandoclevel level;
874
875 level = MANDOCLEVEL_FATAL;
876 while (er < mandoclimits[level])
877 level--;
878
879 if (level < m->wlevel)
880 return;
881
882 if (m->mmsg)
883 (*m->mmsg)(er, level, m->file, ln, col, msg);
884
885 if (m->file_status < level)
886 m->file_status = level;
887 }
888
889 const char *
890 mparse_strerror(enum mandocerr er)
891 {
892
893 return(mandocerrs[er]);
894 }
895
896 const char *
897 mparse_strlevel(enum mandoclevel lvl)
898 {
899 return(mandoclevels[lvl]);
900 }
901
902 void
903 mparse_keep(struct mparse *p)
904 {
905
906 assert(NULL == p->secondary);
907 p->secondary = mandoc_calloc(1, sizeof(struct buf));
908 }
909
910 const char *
911 mparse_getkeep(const struct mparse *p)
912 {
913
914 assert(p->secondary);
915 return(p->secondary->sz ? p->secondary->buf : NULL);
916 }