]> git.cameronkatri.com Git - mandoc.git/blob - read.c
no need to delete any content from .Rs blocks,
[mandoc.git] / read.c
1 /* $Id: read.c,v 1.67 2014/07/07 16:12:20 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 "unterminated quoted argument",
148 "duplicate argument",
149 "skipping duplicate display type",
150 "skipping duplicate list type",
151 "unknown AT&T UNIX version",
152 "invalid content in Rs block",
153 "invalid Boolean argument",
154 "unknown font, skipping request",
155
156 /* related to plain text */
157 "blank line in fill mode, using .sp",
158 "tab in filled text",
159 "whitespace at end of input line",
160 "bad comment style",
161 "invalid escape sequence",
162 "undefined string, using \"\"",
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 unknown macro",
188 "argument count wrong",
189 "unknown standard specifier",
190 "skipping column outside column list",
191 "skipping end of block that is not open",
192 "missing end of block",
193 "scope open on exit",
194 "uname(3) system call failed",
195 "request requires a numeric argument",
196 "missing list type, using -item",
197 "skipping all arguments",
198 "skipping excess arguments",
199
200 "generic fatal error",
201
202 "input too large",
203 "not a manual",
204 "column syntax is inconsistent",
205 "NOT IMPLEMENTED: .Bd -file",
206 "child violates parent syntax",
207 "argument count wrong, violates syntax",
208 "NOT IMPLEMENTED: .so with absolute path or \"..\"",
209 ".so request failed",
210 "no document prologue",
211 "static buffer exhausted",
212
213 /* system errors */
214 NULL,
215 "cannot stat file",
216 "cannot read file",
217 };
218
219 static const char * const mandoclevels[MANDOCLEVEL_MAX] = {
220 "SUCCESS",
221 "RESERVED",
222 "WARNING",
223 "ERROR",
224 "FATAL",
225 "BADARG",
226 "SYSERR"
227 };
228
229
230 static void
231 resize_buf(struct buf *buf, size_t initial)
232 {
233
234 buf->sz = buf->sz > initial/2 ? 2 * buf->sz : initial;
235 buf->buf = mandoc_realloc(buf->buf, buf->sz);
236 }
237
238 static void
239 pset(const char *buf, int pos, struct mparse *curp)
240 {
241 int i;
242
243 /*
244 * Try to intuit which kind of manual parser should be used. If
245 * passed in by command-line (-man, -mdoc), then use that
246 * explicitly. If passed as -mandoc, then try to guess from the
247 * line: either skip dot-lines, use -mdoc when finding `.Dt', or
248 * default to -man, which is more lenient.
249 *
250 * Separate out pmdoc/pman from mdoc/man: the first persists
251 * through all parsers, while the latter is used per-parse.
252 */
253
254 if ('.' == buf[0] || '\'' == buf[0]) {
255 for (i = 1; buf[i]; i++)
256 if (' ' != buf[i] && '\t' != buf[i])
257 break;
258 if ('\0' == buf[i])
259 return;
260 }
261
262 if (MPARSE_MDOC & curp->options) {
263 if (NULL == curp->pmdoc)
264 curp->pmdoc = mdoc_alloc(
265 curp->roff, curp, curp->defos,
266 MPARSE_QUICK & curp->options ? 1 : 0);
267 assert(curp->pmdoc);
268 curp->mdoc = curp->pmdoc;
269 return;
270 } else if (MPARSE_MAN & curp->options) {
271 if (NULL == curp->pman)
272 curp->pman = man_alloc(curp->roff, curp,
273 MPARSE_QUICK & curp->options ? 1 : 0);
274 assert(curp->pman);
275 curp->man = curp->pman;
276 return;
277 }
278
279 if (pos >= 3 && 0 == memcmp(buf, ".Dd", 3)) {
280 if (NULL == curp->pmdoc)
281 curp->pmdoc = mdoc_alloc(
282 curp->roff, curp, curp->defos,
283 MPARSE_QUICK & curp->options ? 1 : 0);
284 assert(curp->pmdoc);
285 curp->mdoc = curp->pmdoc;
286 return;
287 }
288
289 if (NULL == curp->pman)
290 curp->pman = man_alloc(curp->roff, curp,
291 MPARSE_QUICK & curp->options ? 1 : 0);
292 assert(curp->pman);
293 curp->man = curp->pman;
294 }
295
296 /*
297 * Main parse routine for an opened file. This is called for each
298 * opened file and simply loops around the full input file, possibly
299 * nesting (i.e., with `so').
300 */
301 static void
302 mparse_buf_r(struct mparse *curp, struct buf blk, int start)
303 {
304 const struct tbl_span *span;
305 struct buf ln;
306 enum rofferr rr;
307 int i, of, rc;
308 int pos; /* byte number in the ln buffer */
309 int lnn; /* line number in the real file */
310 unsigned char c;
311
312 memset(&ln, 0, sizeof(struct buf));
313
314 lnn = curp->line;
315 pos = 0;
316
317 for (i = 0; i < (int)blk.sz; ) {
318 if (0 == pos && '\0' == blk.buf[i])
319 break;
320
321 if (start) {
322 curp->line = lnn;
323 curp->reparse_count = 0;
324 }
325
326 while (i < (int)blk.sz && (start || '\0' != blk.buf[i])) {
327
328 /*
329 * When finding an unescaped newline character,
330 * leave the character loop to process the line.
331 * Skip a preceding carriage return, if any.
332 */
333
334 if ('\r' == blk.buf[i] && i + 1 < (int)blk.sz &&
335 '\n' == blk.buf[i + 1])
336 ++i;
337 if ('\n' == blk.buf[i]) {
338 ++i;
339 ++lnn;
340 break;
341 }
342
343 /*
344 * Make sure we have space for at least
345 * one backslash and one other character
346 * and the trailing NUL byte.
347 */
348
349 if (pos + 2 >= (int)ln.sz)
350 resize_buf(&ln, 256);
351
352 /*
353 * Warn about bogus characters. If you're using
354 * non-ASCII encoding, you're screwing your
355 * readers. Since I'd rather this not happen,
356 * I'll be helpful and replace these characters
357 * with "?", so we don't display gibberish.
358 * Note to manual writers: use special characters.
359 */
360
361 c = (unsigned char) blk.buf[i];
362
363 if ( ! (isascii(c) &&
364 (isgraph(c) || isblank(c)))) {
365 mandoc_msg(MANDOCERR_BADCHAR, curp,
366 curp->line, pos, NULL);
367 i++;
368 ln.buf[pos++] = '?';
369 continue;
370 }
371
372 /* Trailing backslash = a plain char. */
373
374 if ('\\' != blk.buf[i] || i + 1 == (int)blk.sz) {
375 ln.buf[pos++] = blk.buf[i++];
376 continue;
377 }
378
379 /*
380 * Found escape and at least one other character.
381 * When it's a newline character, skip it.
382 * When there is a carriage return in between,
383 * skip that one as well.
384 */
385
386 if ('\r' == blk.buf[i + 1] && i + 2 < (int)blk.sz &&
387 '\n' == blk.buf[i + 2])
388 ++i;
389 if ('\n' == blk.buf[i + 1]) {
390 i += 2;
391 ++lnn;
392 continue;
393 }
394
395 if ('"' == blk.buf[i + 1] || '#' == blk.buf[i + 1]) {
396 i += 2;
397 /* Comment, skip to end of line */
398 for (; i < (int)blk.sz; ++i) {
399 if ('\n' == blk.buf[i]) {
400 ++i;
401 ++lnn;
402 break;
403 }
404 }
405
406 /* Backout trailing whitespaces */
407 for (; pos > 0; --pos) {
408 if (ln.buf[pos - 1] != ' ')
409 break;
410 if (pos > 2 && ln.buf[pos - 2] == '\\')
411 break;
412 }
413 break;
414 }
415
416 /* Catch escaped bogus characters. */
417
418 c = (unsigned char) blk.buf[i+1];
419
420 if ( ! (isascii(c) &&
421 (isgraph(c) || isblank(c)))) {
422 mandoc_msg(MANDOCERR_BADCHAR, curp,
423 curp->line, pos, NULL);
424 i += 2;
425 ln.buf[pos++] = '?';
426 continue;
427 }
428
429 /* Some other escape sequence, copy & cont. */
430
431 ln.buf[pos++] = blk.buf[i++];
432 ln.buf[pos++] = blk.buf[i++];
433 }
434
435 if (pos >= (int)ln.sz)
436 resize_buf(&ln, 256);
437
438 ln.buf[pos] = '\0';
439
440 /*
441 * A significant amount of complexity is contained by
442 * the roff preprocessor. It's line-oriented but can be
443 * expressed on one line, so we need at times to
444 * readjust our starting point and re-run it. The roff
445 * preprocessor can also readjust the buffers with new
446 * data, so we pass them in wholesale.
447 */
448
449 of = 0;
450
451 /*
452 * Maintain a lookaside buffer of all parsed lines. We
453 * only do this if mparse_keep() has been invoked (the
454 * buffer may be accessed with mparse_getkeep()).
455 */
456
457 if (curp->secondary) {
458 curp->secondary->buf = mandoc_realloc(
459 curp->secondary->buf,
460 curp->secondary->sz + pos + 2);
461 memcpy(curp->secondary->buf +
462 curp->secondary->sz,
463 ln.buf, pos);
464 curp->secondary->sz += pos;
465 curp->secondary->buf
466 [curp->secondary->sz] = '\n';
467 curp->secondary->sz++;
468 curp->secondary->buf
469 [curp->secondary->sz] = '\0';
470 }
471 rerun:
472 rr = roff_parseln(curp->roff, curp->line,
473 &ln.buf, &ln.sz, of, &of);
474
475 switch (rr) {
476 case ROFF_REPARSE:
477 if (REPARSE_LIMIT >= ++curp->reparse_count)
478 mparse_buf_r(curp, ln, 0);
479 else
480 mandoc_msg(MANDOCERR_ROFFLOOP, curp,
481 curp->line, pos, NULL);
482 pos = 0;
483 continue;
484 case ROFF_APPEND:
485 pos = (int)strlen(ln.buf);
486 continue;
487 case ROFF_RERUN:
488 goto rerun;
489 case ROFF_IGN:
490 pos = 0;
491 continue;
492 case ROFF_ERR:
493 assert(MANDOCLEVEL_FATAL <= curp->file_status);
494 break;
495 case ROFF_SO:
496 if (0 == (MPARSE_SO & curp->options) &&
497 (i >= (int)blk.sz || '\0' == blk.buf[i])) {
498 curp->sodest = mandoc_strdup(ln.buf + of);
499 free(ln.buf);
500 return;
501 }
502 /*
503 * We remove `so' clauses from our lookaside
504 * buffer because we're going to descend into
505 * the file recursively.
506 */
507 if (curp->secondary)
508 curp->secondary->sz -= pos + 1;
509 mparse_readfd(curp, -1, ln.buf + of);
510 if (MANDOCLEVEL_FATAL <= curp->file_status) {
511 mandoc_vmsg(MANDOCERR_SO_FAIL,
512 curp, curp->line, pos,
513 ".so %s", ln.buf + of);
514 break;
515 }
516 pos = 0;
517 continue;
518 default:
519 break;
520 }
521
522 /*
523 * If we encounter errors in the recursive parse, make
524 * sure we don't continue parsing.
525 */
526
527 if (MANDOCLEVEL_FATAL <= curp->file_status)
528 break;
529
530 /*
531 * If input parsers have not been allocated, do so now.
532 * We keep these instanced between parsers, but set them
533 * locally per parse routine since we can use different
534 * parsers with each one.
535 */
536
537 if ( ! (curp->man || curp->mdoc))
538 pset(ln.buf + of, pos - of, curp);
539
540 /*
541 * Lastly, push down into the parsers themselves. One
542 * of these will have already been set in the pset()
543 * routine.
544 * If libroff returns ROFF_TBL, then add it to the
545 * currently open parse. Since we only get here if
546 * there does exist data (see tbl_data.c), we're
547 * guaranteed that something's been allocated.
548 * Do the same for ROFF_EQN.
549 */
550
551 rc = -1;
552
553 if (ROFF_TBL == rr)
554 while (NULL != (span = roff_span(curp->roff))) {
555 rc = curp->man ?
556 man_addspan(curp->man, span) :
557 mdoc_addspan(curp->mdoc, span);
558 if (0 == rc)
559 break;
560 }
561 else if (ROFF_EQN == rr)
562 rc = curp->mdoc ?
563 mdoc_addeqn(curp->mdoc,
564 roff_eqn(curp->roff)) :
565 man_addeqn(curp->man,
566 roff_eqn(curp->roff));
567 else if (curp->man || curp->mdoc)
568 rc = curp->man ?
569 man_parseln(curp->man,
570 curp->line, ln.buf, of) :
571 mdoc_parseln(curp->mdoc,
572 curp->line, ln.buf, of);
573
574 if (0 == rc) {
575 assert(MANDOCLEVEL_FATAL <= curp->file_status);
576 break;
577 } else if (2 == rc)
578 break;
579
580 /* Temporary buffers typically are not full. */
581
582 if (0 == start && '\0' == blk.buf[i])
583 break;
584
585 /* Start the next input line. */
586
587 pos = 0;
588 }
589
590 free(ln.buf);
591 }
592
593 static int
594 read_whole_file(struct mparse *curp, const char *file, int fd,
595 struct buf *fb, int *with_mmap)
596 {
597 size_t off;
598 ssize_t ssz;
599
600 #ifdef HAVE_MMAP
601 struct stat st;
602 if (-1 == fstat(fd, &st)) {
603 curp->file_status = MANDOCLEVEL_SYSERR;
604 if (curp->mmsg)
605 (*curp->mmsg)(MANDOCERR_SYSSTAT, curp->file_status,
606 file, 0, 0, strerror(errno));
607 return(0);
608 }
609
610 /*
611 * If we're a regular file, try just reading in the whole entry
612 * via mmap(). This is faster than reading it into blocks, and
613 * since each file is only a few bytes to begin with, I'm not
614 * concerned that this is going to tank any machines.
615 */
616
617 if (S_ISREG(st.st_mode)) {
618 if (st.st_size >= (1U << 31)) {
619 curp->file_status = MANDOCLEVEL_FATAL;
620 if (curp->mmsg)
621 (*curp->mmsg)(MANDOCERR_TOOLARGE,
622 curp->file_status, file, 0, 0, NULL);
623 return(0);
624 }
625 *with_mmap = 1;
626 fb->sz = (size_t)st.st_size;
627 fb->buf = mmap(NULL, fb->sz, PROT_READ, MAP_SHARED, fd, 0);
628 if (fb->buf != MAP_FAILED)
629 return(1);
630 }
631 #endif
632
633 /*
634 * If this isn't a regular file (like, say, stdin), then we must
635 * go the old way and just read things in bit by bit.
636 */
637
638 *with_mmap = 0;
639 off = 0;
640 fb->sz = 0;
641 fb->buf = NULL;
642 for (;;) {
643 if (off == fb->sz) {
644 if (fb->sz == (1U << 31)) {
645 curp->file_status = MANDOCLEVEL_FATAL;
646 if (curp->mmsg)
647 (*curp->mmsg)(MANDOCERR_TOOLARGE,
648 curp->file_status,
649 file, 0, 0, NULL);
650 break;
651 }
652 resize_buf(fb, 65536);
653 }
654 ssz = read(fd, fb->buf + (int)off, fb->sz - off);
655 if (ssz == 0) {
656 fb->sz = off;
657 return(1);
658 }
659 if (ssz == -1) {
660 curp->file_status = MANDOCLEVEL_SYSERR;
661 if (curp->mmsg)
662 (*curp->mmsg)(MANDOCERR_SYSREAD,
663 curp->file_status, file, 0, 0,
664 strerror(errno));
665 break;
666 }
667 off += (size_t)ssz;
668 }
669
670 free(fb->buf);
671 fb->buf = NULL;
672 return(0);
673 }
674
675 static void
676 mparse_end(struct mparse *curp)
677 {
678
679 if (MANDOCLEVEL_FATAL <= curp->file_status)
680 return;
681
682 if (curp->mdoc && ! mdoc_endparse(curp->mdoc)) {
683 assert(MANDOCLEVEL_FATAL <= curp->file_status);
684 return;
685 }
686
687 if (curp->man && ! man_endparse(curp->man)) {
688 assert(MANDOCLEVEL_FATAL <= curp->file_status);
689 return;
690 }
691
692 if ( ! (curp->mdoc || curp->man || curp->sodest)) {
693 mandoc_msg(MANDOCERR_NOTMANUAL, curp, 0, 0, NULL);
694 curp->file_status = MANDOCLEVEL_FATAL;
695 return;
696 }
697
698 roff_endparse(curp->roff);
699 }
700
701 static void
702 mparse_parse_buffer(struct mparse *curp, struct buf blk, const char *file)
703 {
704 const char *svfile;
705 static int recursion_depth;
706
707 if (64 < recursion_depth) {
708 mandoc_msg(MANDOCERR_ROFFLOOP, curp, curp->line, 0, NULL);
709 return;
710 }
711
712 /* Line number is per-file. */
713 svfile = curp->file;
714 curp->file = file;
715 curp->line = 1;
716 recursion_depth++;
717
718 mparse_buf_r(curp, blk, 1);
719
720 if (0 == --recursion_depth && MANDOCLEVEL_FATAL > curp->file_status)
721 mparse_end(curp);
722
723 curp->file = svfile;
724 }
725
726 enum mandoclevel
727 mparse_readmem(struct mparse *curp, const void *buf, size_t len,
728 const char *file)
729 {
730 struct buf blk;
731
732 blk.buf = UNCONST(buf);
733 blk.sz = len;
734
735 mparse_parse_buffer(curp, blk, file);
736 return(curp->file_status);
737 }
738
739 enum mandoclevel
740 mparse_readfd(struct mparse *curp, int fd, const char *file)
741 {
742 struct buf blk;
743 int with_mmap;
744
745 if (-1 == fd && -1 == (fd = open(file, O_RDONLY, 0))) {
746 curp->file_status = MANDOCLEVEL_SYSERR;
747 if (curp->mmsg)
748 (*curp->mmsg)(MANDOCERR_SYSOPEN,
749 curp->file_status,
750 file, 0, 0, strerror(errno));
751 goto out;
752 }
753
754 /*
755 * Run for each opened file; may be called more than once for
756 * each full parse sequence if the opened file is nested (i.e.,
757 * from `so'). Simply sucks in the whole file and moves into
758 * the parse phase for the file.
759 */
760
761 if ( ! read_whole_file(curp, file, fd, &blk, &with_mmap))
762 goto out;
763
764 mparse_parse_buffer(curp, blk, file);
765
766 #ifdef HAVE_MMAP
767 if (with_mmap)
768 munmap(blk.buf, blk.sz);
769 else
770 #endif
771 free(blk.buf);
772
773 if (STDIN_FILENO != fd && -1 == close(fd))
774 perror(file);
775 out:
776 return(curp->file_status);
777 }
778
779 struct mparse *
780 mparse_alloc(int options, enum mandoclevel wlevel,
781 mandocmsg mmsg, char *defos)
782 {
783 struct mparse *curp;
784
785 assert(wlevel <= MANDOCLEVEL_FATAL);
786
787 curp = mandoc_calloc(1, sizeof(struct mparse));
788
789 curp->options = options;
790 curp->wlevel = wlevel;
791 curp->mmsg = mmsg;
792 curp->defos = defos;
793
794 curp->roff = roff_alloc(curp, options);
795 return(curp);
796 }
797
798 void
799 mparse_reset(struct mparse *curp)
800 {
801
802 roff_reset(curp->roff);
803
804 if (curp->mdoc)
805 mdoc_reset(curp->mdoc);
806 if (curp->man)
807 man_reset(curp->man);
808 if (curp->secondary)
809 curp->secondary->sz = 0;
810
811 curp->file_status = MANDOCLEVEL_OK;
812 curp->mdoc = NULL;
813 curp->man = NULL;
814
815 free(curp->sodest);
816 curp->sodest = NULL;
817 }
818
819 void
820 mparse_free(struct mparse *curp)
821 {
822
823 if (curp->pmdoc)
824 mdoc_free(curp->pmdoc);
825 if (curp->pman)
826 man_free(curp->pman);
827 if (curp->roff)
828 roff_free(curp->roff);
829 if (curp->secondary)
830 free(curp->secondary->buf);
831
832 free(curp->secondary);
833 free(curp->sodest);
834 free(curp);
835 }
836
837 void
838 mparse_result(struct mparse *curp,
839 struct mdoc **mdoc, struct man **man, char **sodest)
840 {
841
842 if (sodest && NULL != (*sodest = curp->sodest)) {
843 *mdoc = NULL;
844 *man = NULL;
845 return;
846 }
847 if (mdoc)
848 *mdoc = curp->mdoc;
849 if (man)
850 *man = curp->man;
851 }
852
853 void
854 mandoc_vmsg(enum mandocerr t, struct mparse *m,
855 int ln, int pos, const char *fmt, ...)
856 {
857 char buf[256];
858 va_list ap;
859
860 va_start(ap, fmt);
861 (void)vsnprintf(buf, sizeof(buf), fmt, ap);
862 va_end(ap);
863
864 mandoc_msg(t, m, ln, pos, buf);
865 }
866
867 void
868 mandoc_msg(enum mandocerr er, struct mparse *m,
869 int ln, int col, const char *msg)
870 {
871 enum mandoclevel level;
872
873 level = MANDOCLEVEL_FATAL;
874 while (er < mandoclimits[level])
875 level--;
876
877 if (level < m->wlevel)
878 return;
879
880 if (m->mmsg)
881 (*m->mmsg)(er, level, m->file, ln, col, msg);
882
883 if (m->file_status < level)
884 m->file_status = level;
885 }
886
887 const char *
888 mparse_strerror(enum mandocerr er)
889 {
890
891 return(mandocerrs[er]);
892 }
893
894 const char *
895 mparse_strlevel(enum mandoclevel lvl)
896 {
897 return(mandoclevels[lvl]);
898 }
899
900 void
901 mparse_keep(struct mparse *p)
902 {
903
904 assert(NULL == p->secondary);
905 p->secondary = mandoc_calloc(1, sizeof(struct buf));
906 }
907
908 const char *
909 mparse_getkeep(const struct mparse *p)
910 {
911
912 assert(p->secondary);
913 return(p->secondary->sz ? p->secondary->buf : NULL);
914 }