]> git.cameronkatri.com Git - mandoc.git/blob - read.c
update after recent code changes
[mandoc.git] / read.c
1 /* $Id: read.c,v 1.69 2014/07/09 11:31:43 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 const 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 /* related to document structure and macros */
184 "input stack limit exceeded, infinite loop?",
185 "skipping bad character",
186 "skipping unknown macro",
187 "skipping column outside column list",
188 "skipping end of block that is not open",
189 "inserting missing end of block",
190 "appending missing end of block",
191
192 /* related to request and macro arguments */
193 "escaped character not allowed in a name",
194 "manual name not yet set",
195 "argument count wrong",
196 "unknown standard specifier",
197 "uname(3) system call failed",
198 "request requires a numeric argument",
199 "missing list type, using -item",
200 "skipping all arguments",
201 "skipping excess arguments",
202
203 "generic fatal error",
204
205 "input too large",
206 "not a manual",
207 "column syntax is inconsistent",
208 "NOT IMPLEMENTED: .Bd -file",
209 "child violates parent syntax",
210 "argument count wrong, violates syntax",
211 "NOT IMPLEMENTED: .so with absolute path or \"..\"",
212 ".so request failed",
213 "no document prologue",
214 "static buffer exhausted",
215
216 /* system errors */
217 NULL,
218 "cannot stat file",
219 "cannot read file",
220 };
221
222 static const char * const mandoclevels[MANDOCLEVEL_MAX] = {
223 "SUCCESS",
224 "RESERVED",
225 "WARNING",
226 "ERROR",
227 "FATAL",
228 "BADARG",
229 "SYSERR"
230 };
231
232
233 static void
234 resize_buf(struct buf *buf, size_t initial)
235 {
236
237 buf->sz = buf->sz > initial/2 ? 2 * buf->sz : initial;
238 buf->buf = mandoc_realloc(buf->buf, buf->sz);
239 }
240
241 static void
242 pset(const char *buf, int pos, struct mparse *curp)
243 {
244 int i;
245
246 /*
247 * Try to intuit which kind of manual parser should be used. If
248 * passed in by command-line (-man, -mdoc), then use that
249 * explicitly. If passed as -mandoc, then try to guess from the
250 * line: either skip dot-lines, use -mdoc when finding `.Dt', or
251 * default to -man, which is more lenient.
252 *
253 * Separate out pmdoc/pman from mdoc/man: the first persists
254 * through all parsers, while the latter is used per-parse.
255 */
256
257 if ('.' == buf[0] || '\'' == buf[0]) {
258 for (i = 1; buf[i]; i++)
259 if (' ' != buf[i] && '\t' != buf[i])
260 break;
261 if ('\0' == buf[i])
262 return;
263 }
264
265 if (MPARSE_MDOC & curp->options) {
266 if (NULL == curp->pmdoc)
267 curp->pmdoc = mdoc_alloc(
268 curp->roff, curp, curp->defos,
269 MPARSE_QUICK & curp->options ? 1 : 0);
270 assert(curp->pmdoc);
271 curp->mdoc = curp->pmdoc;
272 return;
273 } else if (MPARSE_MAN & curp->options) {
274 if (NULL == curp->pman)
275 curp->pman = man_alloc(curp->roff, curp,
276 MPARSE_QUICK & curp->options ? 1 : 0);
277 assert(curp->pman);
278 curp->man = curp->pman;
279 return;
280 }
281
282 if (pos >= 3 && 0 == memcmp(buf, ".Dd", 3)) {
283 if (NULL == curp->pmdoc)
284 curp->pmdoc = mdoc_alloc(
285 curp->roff, curp, curp->defos,
286 MPARSE_QUICK & curp->options ? 1 : 0);
287 assert(curp->pmdoc);
288 curp->mdoc = curp->pmdoc;
289 return;
290 }
291
292 if (NULL == curp->pman)
293 curp->pman = man_alloc(curp->roff, curp,
294 MPARSE_QUICK & curp->options ? 1 : 0);
295 assert(curp->pman);
296 curp->man = curp->pman;
297 }
298
299 /*
300 * Main parse routine for an opened file. This is called for each
301 * opened file and simply loops around the full input file, possibly
302 * nesting (i.e., with `so').
303 */
304 static void
305 mparse_buf_r(struct mparse *curp, struct buf blk, int start)
306 {
307 const struct tbl_span *span;
308 struct buf ln;
309 enum rofferr rr;
310 int i, of, rc;
311 int pos; /* byte number in the ln buffer */
312 int lnn; /* line number in the real file */
313 unsigned char c;
314
315 memset(&ln, 0, sizeof(struct buf));
316
317 lnn = curp->line;
318 pos = 0;
319
320 for (i = 0; i < (int)blk.sz; ) {
321 if (0 == pos && '\0' == blk.buf[i])
322 break;
323
324 if (start) {
325 curp->line = lnn;
326 curp->reparse_count = 0;
327 }
328
329 while (i < (int)blk.sz && (start || '\0' != blk.buf[i])) {
330
331 /*
332 * When finding an unescaped newline character,
333 * leave the character loop to process the line.
334 * Skip a preceding carriage return, if any.
335 */
336
337 if ('\r' == blk.buf[i] && i + 1 < (int)blk.sz &&
338 '\n' == blk.buf[i + 1])
339 ++i;
340 if ('\n' == blk.buf[i]) {
341 ++i;
342 ++lnn;
343 break;
344 }
345
346 /*
347 * Make sure we have space for at least
348 * one backslash and one other character
349 * and the trailing NUL byte.
350 */
351
352 if (pos + 2 >= (int)ln.sz)
353 resize_buf(&ln, 256);
354
355 /*
356 * Warn about bogus characters. If you're using
357 * non-ASCII encoding, you're screwing your
358 * readers. Since I'd rather this not happen,
359 * I'll be helpful and replace these characters
360 * with "?", so we don't display gibberish.
361 * Note to manual writers: use special characters.
362 */
363
364 c = (unsigned char) blk.buf[i];
365
366 if ( ! (isascii(c) &&
367 (isgraph(c) || isblank(c)))) {
368 mandoc_msg(MANDOCERR_BADCHAR, curp,
369 curp->line, pos, NULL);
370 i++;
371 ln.buf[pos++] = '?';
372 continue;
373 }
374
375 /* Trailing backslash = a plain char. */
376
377 if ('\\' != blk.buf[i] || i + 1 == (int)blk.sz) {
378 ln.buf[pos++] = blk.buf[i++];
379 continue;
380 }
381
382 /*
383 * Found escape and at least one other character.
384 * When it's a newline character, skip it.
385 * When there is a carriage return in between,
386 * skip that one as well.
387 */
388
389 if ('\r' == blk.buf[i + 1] && i + 2 < (int)blk.sz &&
390 '\n' == blk.buf[i + 2])
391 ++i;
392 if ('\n' == blk.buf[i + 1]) {
393 i += 2;
394 ++lnn;
395 continue;
396 }
397
398 if ('"' == blk.buf[i + 1] || '#' == blk.buf[i + 1]) {
399 i += 2;
400 /* Comment, skip to end of line */
401 for (; i < (int)blk.sz; ++i) {
402 if ('\n' == blk.buf[i]) {
403 ++i;
404 ++lnn;
405 break;
406 }
407 }
408
409 /* Backout trailing whitespaces */
410 for (; pos > 0; --pos) {
411 if (ln.buf[pos - 1] != ' ')
412 break;
413 if (pos > 2 && ln.buf[pos - 2] == '\\')
414 break;
415 }
416 break;
417 }
418
419 /* Catch escaped bogus characters. */
420
421 c = (unsigned char) blk.buf[i+1];
422
423 if ( ! (isascii(c) &&
424 (isgraph(c) || isblank(c)))) {
425 mandoc_msg(MANDOCERR_BADCHAR, curp,
426 curp->line, pos, NULL);
427 i += 2;
428 ln.buf[pos++] = '?';
429 continue;
430 }
431
432 /* Some other escape sequence, copy & cont. */
433
434 ln.buf[pos++] = blk.buf[i++];
435 ln.buf[pos++] = blk.buf[i++];
436 }
437
438 if (pos >= (int)ln.sz)
439 resize_buf(&ln, 256);
440
441 ln.buf[pos] = '\0';
442
443 /*
444 * A significant amount of complexity is contained by
445 * the roff preprocessor. It's line-oriented but can be
446 * expressed on one line, so we need at times to
447 * readjust our starting point and re-run it. The roff
448 * preprocessor can also readjust the buffers with new
449 * data, so we pass them in wholesale.
450 */
451
452 of = 0;
453
454 /*
455 * Maintain a lookaside buffer of all parsed lines. We
456 * only do this if mparse_keep() has been invoked (the
457 * buffer may be accessed with mparse_getkeep()).
458 */
459
460 if (curp->secondary) {
461 curp->secondary->buf = mandoc_realloc(
462 curp->secondary->buf,
463 curp->secondary->sz + pos + 2);
464 memcpy(curp->secondary->buf +
465 curp->secondary->sz,
466 ln.buf, pos);
467 curp->secondary->sz += pos;
468 curp->secondary->buf
469 [curp->secondary->sz] = '\n';
470 curp->secondary->sz++;
471 curp->secondary->buf
472 [curp->secondary->sz] = '\0';
473 }
474 rerun:
475 rr = roff_parseln(curp->roff, curp->line,
476 &ln.buf, &ln.sz, of, &of);
477
478 switch (rr) {
479 case ROFF_REPARSE:
480 if (REPARSE_LIMIT >= ++curp->reparse_count)
481 mparse_buf_r(curp, ln, 0);
482 else
483 mandoc_msg(MANDOCERR_ROFFLOOP, curp,
484 curp->line, pos, NULL);
485 pos = 0;
486 continue;
487 case ROFF_APPEND:
488 pos = (int)strlen(ln.buf);
489 continue;
490 case ROFF_RERUN:
491 goto rerun;
492 case ROFF_IGN:
493 pos = 0;
494 continue;
495 case ROFF_ERR:
496 assert(MANDOCLEVEL_FATAL <= curp->file_status);
497 break;
498 case ROFF_SO:
499 if (0 == (MPARSE_SO & curp->options) &&
500 (i >= (int)blk.sz || '\0' == blk.buf[i])) {
501 curp->sodest = mandoc_strdup(ln.buf + of);
502 free(ln.buf);
503 return;
504 }
505 /*
506 * We remove `so' clauses from our lookaside
507 * buffer because we're going to descend into
508 * the file recursively.
509 */
510 if (curp->secondary)
511 curp->secondary->sz -= pos + 1;
512 mparse_readfd(curp, -1, ln.buf + of);
513 if (MANDOCLEVEL_FATAL <= curp->file_status) {
514 mandoc_vmsg(MANDOCERR_SO_FAIL,
515 curp, curp->line, pos,
516 ".so %s", ln.buf + of);
517 break;
518 }
519 pos = 0;
520 continue;
521 default:
522 break;
523 }
524
525 /*
526 * If we encounter errors in the recursive parse, make
527 * sure we don't continue parsing.
528 */
529
530 if (MANDOCLEVEL_FATAL <= curp->file_status)
531 break;
532
533 /*
534 * If input parsers have not been allocated, do so now.
535 * We keep these instanced between parsers, but set them
536 * locally per parse routine since we can use different
537 * parsers with each one.
538 */
539
540 if ( ! (curp->man || curp->mdoc))
541 pset(ln.buf + of, pos - of, curp);
542
543 /*
544 * Lastly, push down into the parsers themselves. One
545 * of these will have already been set in the pset()
546 * routine.
547 * If libroff returns ROFF_TBL, then add it to the
548 * currently open parse. Since we only get here if
549 * there does exist data (see tbl_data.c), we're
550 * guaranteed that something's been allocated.
551 * Do the same for ROFF_EQN.
552 */
553
554 rc = -1;
555
556 if (ROFF_TBL == rr)
557 while (NULL != (span = roff_span(curp->roff))) {
558 rc = curp->man ?
559 man_addspan(curp->man, span) :
560 mdoc_addspan(curp->mdoc, span);
561 if (0 == rc)
562 break;
563 }
564 else if (ROFF_EQN == rr)
565 rc = curp->mdoc ?
566 mdoc_addeqn(curp->mdoc,
567 roff_eqn(curp->roff)) :
568 man_addeqn(curp->man,
569 roff_eqn(curp->roff));
570 else if (curp->man || curp->mdoc)
571 rc = curp->man ?
572 man_parseln(curp->man,
573 curp->line, ln.buf, of) :
574 mdoc_parseln(curp->mdoc,
575 curp->line, ln.buf, of);
576
577 if (0 == rc) {
578 assert(MANDOCLEVEL_FATAL <= curp->file_status);
579 break;
580 } else if (2 == rc)
581 break;
582
583 /* Temporary buffers typically are not full. */
584
585 if (0 == start && '\0' == blk.buf[i])
586 break;
587
588 /* Start the next input line. */
589
590 pos = 0;
591 }
592
593 free(ln.buf);
594 }
595
596 static int
597 read_whole_file(struct mparse *curp, const char *file, int fd,
598 struct buf *fb, int *with_mmap)
599 {
600 size_t off;
601 ssize_t ssz;
602
603 #ifdef HAVE_MMAP
604 struct stat st;
605 if (-1 == fstat(fd, &st)) {
606 curp->file_status = MANDOCLEVEL_SYSERR;
607 if (curp->mmsg)
608 (*curp->mmsg)(MANDOCERR_SYSSTAT, curp->file_status,
609 file, 0, 0, strerror(errno));
610 return(0);
611 }
612
613 /*
614 * If we're a regular file, try just reading in the whole entry
615 * via mmap(). This is faster than reading it into blocks, and
616 * since each file is only a few bytes to begin with, I'm not
617 * concerned that this is going to tank any machines.
618 */
619
620 if (S_ISREG(st.st_mode)) {
621 if (st.st_size >= (1U << 31)) {
622 curp->file_status = MANDOCLEVEL_FATAL;
623 if (curp->mmsg)
624 (*curp->mmsg)(MANDOCERR_TOOLARGE,
625 curp->file_status, file, 0, 0, NULL);
626 return(0);
627 }
628 *with_mmap = 1;
629 fb->sz = (size_t)st.st_size;
630 fb->buf = mmap(NULL, fb->sz, PROT_READ, MAP_SHARED, fd, 0);
631 if (fb->buf != MAP_FAILED)
632 return(1);
633 }
634 #endif
635
636 /*
637 * If this isn't a regular file (like, say, stdin), then we must
638 * go the old way and just read things in bit by bit.
639 */
640
641 *with_mmap = 0;
642 off = 0;
643 fb->sz = 0;
644 fb->buf = NULL;
645 for (;;) {
646 if (off == fb->sz) {
647 if (fb->sz == (1U << 31)) {
648 curp->file_status = MANDOCLEVEL_FATAL;
649 if (curp->mmsg)
650 (*curp->mmsg)(MANDOCERR_TOOLARGE,
651 curp->file_status,
652 file, 0, 0, NULL);
653 break;
654 }
655 resize_buf(fb, 65536);
656 }
657 ssz = read(fd, fb->buf + (int)off, fb->sz - off);
658 if (ssz == 0) {
659 fb->sz = off;
660 return(1);
661 }
662 if (ssz == -1) {
663 curp->file_status = MANDOCLEVEL_SYSERR;
664 if (curp->mmsg)
665 (*curp->mmsg)(MANDOCERR_SYSREAD,
666 curp->file_status, file, 0, 0,
667 strerror(errno));
668 break;
669 }
670 off += (size_t)ssz;
671 }
672
673 free(fb->buf);
674 fb->buf = NULL;
675 return(0);
676 }
677
678 static void
679 mparse_end(struct mparse *curp)
680 {
681
682 if (MANDOCLEVEL_FATAL <= curp->file_status)
683 return;
684
685 if (curp->mdoc && ! mdoc_endparse(curp->mdoc)) {
686 assert(MANDOCLEVEL_FATAL <= curp->file_status);
687 return;
688 }
689
690 if (curp->man && ! man_endparse(curp->man)) {
691 assert(MANDOCLEVEL_FATAL <= curp->file_status);
692 return;
693 }
694
695 if ( ! (curp->mdoc || curp->man || curp->sodest)) {
696 mandoc_msg(MANDOCERR_NOTMANUAL, curp, 0, 0, NULL);
697 curp->file_status = MANDOCLEVEL_FATAL;
698 return;
699 }
700
701 roff_endparse(curp->roff);
702 }
703
704 static void
705 mparse_parse_buffer(struct mparse *curp, struct buf blk, const char *file)
706 {
707 const char *svfile;
708 static int recursion_depth;
709
710 if (64 < recursion_depth) {
711 mandoc_msg(MANDOCERR_ROFFLOOP, curp, curp->line, 0, NULL);
712 return;
713 }
714
715 /* Line number is per-file. */
716 svfile = curp->file;
717 curp->file = file;
718 curp->line = 1;
719 recursion_depth++;
720
721 mparse_buf_r(curp, blk, 1);
722
723 if (0 == --recursion_depth && MANDOCLEVEL_FATAL > curp->file_status)
724 mparse_end(curp);
725
726 curp->file = svfile;
727 }
728
729 enum mandoclevel
730 mparse_readmem(struct mparse *curp, const void *buf, size_t len,
731 const char *file)
732 {
733 struct buf blk;
734
735 blk.buf = UNCONST(buf);
736 blk.sz = len;
737
738 mparse_parse_buffer(curp, blk, file);
739 return(curp->file_status);
740 }
741
742 enum mandoclevel
743 mparse_readfd(struct mparse *curp, int fd, const char *file)
744 {
745 struct buf blk;
746 int with_mmap;
747
748 if (-1 == fd && -1 == (fd = open(file, O_RDONLY, 0))) {
749 curp->file_status = MANDOCLEVEL_SYSERR;
750 if (curp->mmsg)
751 (*curp->mmsg)(MANDOCERR_SYSOPEN,
752 curp->file_status,
753 file, 0, 0, strerror(errno));
754 goto out;
755 }
756
757 /*
758 * Run for each opened file; may be called more than once for
759 * each full parse sequence if the opened file is nested (i.e.,
760 * from `so'). Simply sucks in the whole file and moves into
761 * the parse phase for the file.
762 */
763
764 if ( ! read_whole_file(curp, file, fd, &blk, &with_mmap))
765 goto out;
766
767 mparse_parse_buffer(curp, blk, file);
768
769 #ifdef HAVE_MMAP
770 if (with_mmap)
771 munmap(blk.buf, blk.sz);
772 else
773 #endif
774 free(blk.buf);
775
776 if (STDIN_FILENO != fd && -1 == close(fd))
777 perror(file);
778 out:
779 return(curp->file_status);
780 }
781
782 struct mparse *
783 mparse_alloc(int options, enum mandoclevel wlevel,
784 mandocmsg mmsg, const char *defos)
785 {
786 struct mparse *curp;
787
788 assert(wlevel <= MANDOCLEVEL_FATAL);
789
790 curp = mandoc_calloc(1, sizeof(struct mparse));
791
792 curp->options = options;
793 curp->wlevel = wlevel;
794 curp->mmsg = mmsg;
795 curp->defos = defos;
796
797 curp->roff = roff_alloc(curp, options);
798 return(curp);
799 }
800
801 void
802 mparse_reset(struct mparse *curp)
803 {
804
805 roff_reset(curp->roff);
806
807 if (curp->mdoc)
808 mdoc_reset(curp->mdoc);
809 if (curp->man)
810 man_reset(curp->man);
811 if (curp->secondary)
812 curp->secondary->sz = 0;
813
814 curp->file_status = MANDOCLEVEL_OK;
815 curp->mdoc = NULL;
816 curp->man = NULL;
817
818 free(curp->sodest);
819 curp->sodest = NULL;
820 }
821
822 void
823 mparse_free(struct mparse *curp)
824 {
825
826 if (curp->pmdoc)
827 mdoc_free(curp->pmdoc);
828 if (curp->pman)
829 man_free(curp->pman);
830 if (curp->roff)
831 roff_free(curp->roff);
832 if (curp->secondary)
833 free(curp->secondary->buf);
834
835 free(curp->secondary);
836 free(curp->sodest);
837 free(curp);
838 }
839
840 void
841 mparse_result(struct mparse *curp,
842 struct mdoc **mdoc, struct man **man, char **sodest)
843 {
844
845 if (sodest && NULL != (*sodest = curp->sodest)) {
846 *mdoc = NULL;
847 *man = NULL;
848 return;
849 }
850 if (mdoc)
851 *mdoc = curp->mdoc;
852 if (man)
853 *man = curp->man;
854 }
855
856 void
857 mandoc_vmsg(enum mandocerr t, struct mparse *m,
858 int ln, int pos, const char *fmt, ...)
859 {
860 char buf[256];
861 va_list ap;
862
863 va_start(ap, fmt);
864 (void)vsnprintf(buf, sizeof(buf), fmt, ap);
865 va_end(ap);
866
867 mandoc_msg(t, m, ln, pos, buf);
868 }
869
870 void
871 mandoc_msg(enum mandocerr er, struct mparse *m,
872 int ln, int col, const char *msg)
873 {
874 enum mandoclevel level;
875
876 level = MANDOCLEVEL_FATAL;
877 while (er < mandoclimits[level])
878 level--;
879
880 if (level < m->wlevel)
881 return;
882
883 if (m->mmsg)
884 (*m->mmsg)(er, level, m->file, ln, col, msg);
885
886 if (m->file_status < level)
887 m->file_status = level;
888 }
889
890 const char *
891 mparse_strerror(enum mandocerr er)
892 {
893
894 return(mandocerrs[er]);
895 }
896
897 const char *
898 mparse_strlevel(enum mandoclevel lvl)
899 {
900 return(mandoclevels[lvl]);
901 }
902
903 void
904 mparse_keep(struct mparse *p)
905 {
906
907 assert(NULL == p->secondary);
908 p->secondary = mandoc_calloc(1, sizeof(struct buf));
909 }
910
911 const char *
912 mparse_getkeep(const struct mparse *p)
913 {
914
915 assert(p->secondary);
916 return(p->secondary->sz ? p->secondary->buf : NULL);
917 }