]> git.cameronkatri.com Git - mandoc.git/blob - read.c
implement .dei and .ami
[mandoc.git] / read.c
1 /* $Id: read.c,v 1.65 2014/07/07 11:35:06 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 Boolean argument",
153 "unknown font, skipping request",
154
155 /* related to plain text */
156 "blank line in fill mode, using .sp",
157 "tab in filled text",
158 "whitespace at end of input line",
159 "bad comment style",
160 "invalid escape sequence",
161 "undefined string, using \"\"",
162
163 "generic error",
164
165 /* related to equations */
166 "unexpected equation scope closure",
167 "equation scope open on exit",
168 "overlapping equation scopes",
169 "unexpected end of equation",
170 "equation syntax error",
171
172 /* related to tables */
173 "bad table syntax",
174 "bad table option",
175 "bad table layout",
176 "no table layout cells specified",
177 "no table data cells specified",
178 "ignore data in cell",
179 "data block still open",
180 "ignoring extra data cells",
181
182 "input stack limit exceeded, infinite loop?",
183 "skipping bad character",
184 "escaped character not allowed in a name",
185 "manual name not yet set",
186 "skipping text before first section header",
187 "skipping unknown macro",
188 "argument count wrong",
189 "skipping invalid content in .Rs block",
190 "unknown standard specifier",
191 "skipping column outside column list",
192 "skipping end of block that is not open",
193 "missing end of block",
194 "scope open on exit",
195 "uname(3) system call failed",
196 "request requires a numeric argument",
197 "missing list type, using -item",
198 "skipping all arguments",
199 "skipping excess arguments",
200
201 "generic fatal error",
202
203 "input too large",
204 "not a manual",
205 "column syntax is inconsistent",
206 "NOT IMPLEMENTED: .Bd -file",
207 "child violates parent syntax",
208 "argument count wrong, violates syntax",
209 "NOT IMPLEMENTED: .so with absolute path or \"..\"",
210 ".so request failed",
211 "no document prologue",
212 "static buffer exhausted",
213
214 /* system errors */
215 NULL,
216 "cannot stat file",
217 "cannot read file",
218 };
219
220 static const char * const mandoclevels[MANDOCLEVEL_MAX] = {
221 "SUCCESS",
222 "RESERVED",
223 "WARNING",
224 "ERROR",
225 "FATAL",
226 "BADARG",
227 "SYSERR"
228 };
229
230
231 static void
232 resize_buf(struct buf *buf, size_t initial)
233 {
234
235 buf->sz = buf->sz > initial/2 ? 2 * buf->sz : initial;
236 buf->buf = mandoc_realloc(buf->buf, buf->sz);
237 }
238
239 static void
240 pset(const char *buf, int pos, struct mparse *curp)
241 {
242 int i;
243
244 /*
245 * Try to intuit which kind of manual parser should be used. If
246 * passed in by command-line (-man, -mdoc), then use that
247 * explicitly. If passed as -mandoc, then try to guess from the
248 * line: either skip dot-lines, use -mdoc when finding `.Dt', or
249 * default to -man, which is more lenient.
250 *
251 * Separate out pmdoc/pman from mdoc/man: the first persists
252 * through all parsers, while the latter is used per-parse.
253 */
254
255 if ('.' == buf[0] || '\'' == buf[0]) {
256 for (i = 1; buf[i]; i++)
257 if (' ' != buf[i] && '\t' != buf[i])
258 break;
259 if ('\0' == buf[i])
260 return;
261 }
262
263 if (MPARSE_MDOC & curp->options) {
264 if (NULL == curp->pmdoc)
265 curp->pmdoc = mdoc_alloc(
266 curp->roff, curp, curp->defos,
267 MPARSE_QUICK & curp->options ? 1 : 0);
268 assert(curp->pmdoc);
269 curp->mdoc = curp->pmdoc;
270 return;
271 } else if (MPARSE_MAN & curp->options) {
272 if (NULL == curp->pman)
273 curp->pman = man_alloc(curp->roff, curp,
274 MPARSE_QUICK & curp->options ? 1 : 0);
275 assert(curp->pman);
276 curp->man = curp->pman;
277 return;
278 }
279
280 if (pos >= 3 && 0 == memcmp(buf, ".Dd", 3)) {
281 if (NULL == curp->pmdoc)
282 curp->pmdoc = mdoc_alloc(
283 curp->roff, curp, curp->defos,
284 MPARSE_QUICK & curp->options ? 1 : 0);
285 assert(curp->pmdoc);
286 curp->mdoc = curp->pmdoc;
287 return;
288 }
289
290 if (NULL == curp->pman)
291 curp->pman = man_alloc(curp->roff, curp,
292 MPARSE_QUICK & curp->options ? 1 : 0);
293 assert(curp->pman);
294 curp->man = curp->pman;
295 }
296
297 /*
298 * Main parse routine for an opened file. This is called for each
299 * opened file and simply loops around the full input file, possibly
300 * nesting (i.e., with `so').
301 */
302 static void
303 mparse_buf_r(struct mparse *curp, struct buf blk, int start)
304 {
305 const struct tbl_span *span;
306 struct buf ln;
307 enum rofferr rr;
308 int i, of, rc;
309 int pos; /* byte number in the ln buffer */
310 int lnn; /* line number in the real file */
311 unsigned char c;
312
313 memset(&ln, 0, sizeof(struct buf));
314
315 lnn = curp->line;
316 pos = 0;
317
318 for (i = 0; i < (int)blk.sz; ) {
319 if (0 == pos && '\0' == blk.buf[i])
320 break;
321
322 if (start) {
323 curp->line = lnn;
324 curp->reparse_count = 0;
325 }
326
327 while (i < (int)blk.sz && (start || '\0' != blk.buf[i])) {
328
329 /*
330 * When finding an unescaped newline character,
331 * leave the character loop to process the line.
332 * Skip a preceding carriage return, if any.
333 */
334
335 if ('\r' == blk.buf[i] && i + 1 < (int)blk.sz &&
336 '\n' == blk.buf[i + 1])
337 ++i;
338 if ('\n' == blk.buf[i]) {
339 ++i;
340 ++lnn;
341 break;
342 }
343
344 /*
345 * Make sure we have space for at least
346 * one backslash and one other character
347 * and the trailing NUL byte.
348 */
349
350 if (pos + 2 >= (int)ln.sz)
351 resize_buf(&ln, 256);
352
353 /*
354 * Warn about bogus characters. If you're using
355 * non-ASCII encoding, you're screwing your
356 * readers. Since I'd rather this not happen,
357 * I'll be helpful and replace these characters
358 * with "?", so we don't display gibberish.
359 * Note to manual writers: use special characters.
360 */
361
362 c = (unsigned char) blk.buf[i];
363
364 if ( ! (isascii(c) &&
365 (isgraph(c) || isblank(c)))) {
366 mandoc_msg(MANDOCERR_BADCHAR, curp,
367 curp->line, pos, NULL);
368 i++;
369 ln.buf[pos++] = '?';
370 continue;
371 }
372
373 /* Trailing backslash = a plain char. */
374
375 if ('\\' != blk.buf[i] || i + 1 == (int)blk.sz) {
376 ln.buf[pos++] = blk.buf[i++];
377 continue;
378 }
379
380 /*
381 * Found escape and at least one other character.
382 * When it's a newline character, skip it.
383 * When there is a carriage return in between,
384 * skip that one as well.
385 */
386
387 if ('\r' == blk.buf[i + 1] && i + 2 < (int)blk.sz &&
388 '\n' == blk.buf[i + 2])
389 ++i;
390 if ('\n' == blk.buf[i + 1]) {
391 i += 2;
392 ++lnn;
393 continue;
394 }
395
396 if ('"' == blk.buf[i + 1] || '#' == blk.buf[i + 1]) {
397 i += 2;
398 /* Comment, skip to end of line */
399 for (; i < (int)blk.sz; ++i) {
400 if ('\n' == blk.buf[i]) {
401 ++i;
402 ++lnn;
403 break;
404 }
405 }
406
407 /* Backout trailing whitespaces */
408 for (; pos > 0; --pos) {
409 if (ln.buf[pos - 1] != ' ')
410 break;
411 if (pos > 2 && ln.buf[pos - 2] == '\\')
412 break;
413 }
414 break;
415 }
416
417 /* Catch escaped bogus characters. */
418
419 c = (unsigned char) blk.buf[i+1];
420
421 if ( ! (isascii(c) &&
422 (isgraph(c) || isblank(c)))) {
423 mandoc_msg(MANDOCERR_BADCHAR, curp,
424 curp->line, pos, NULL);
425 i += 2;
426 ln.buf[pos++] = '?';
427 continue;
428 }
429
430 /* Some other escape sequence, copy & cont. */
431
432 ln.buf[pos++] = blk.buf[i++];
433 ln.buf[pos++] = blk.buf[i++];
434 }
435
436 if (pos >= (int)ln.sz)
437 resize_buf(&ln, 256);
438
439 ln.buf[pos] = '\0';
440
441 /*
442 * A significant amount of complexity is contained by
443 * the roff preprocessor. It's line-oriented but can be
444 * expressed on one line, so we need at times to
445 * readjust our starting point and re-run it. The roff
446 * preprocessor can also readjust the buffers with new
447 * data, so we pass them in wholesale.
448 */
449
450 of = 0;
451
452 /*
453 * Maintain a lookaside buffer of all parsed lines. We
454 * only do this if mparse_keep() has been invoked (the
455 * buffer may be accessed with mparse_getkeep()).
456 */
457
458 if (curp->secondary) {
459 curp->secondary->buf = mandoc_realloc(
460 curp->secondary->buf,
461 curp->secondary->sz + pos + 2);
462 memcpy(curp->secondary->buf +
463 curp->secondary->sz,
464 ln.buf, pos);
465 curp->secondary->sz += pos;
466 curp->secondary->buf
467 [curp->secondary->sz] = '\n';
468 curp->secondary->sz++;
469 curp->secondary->buf
470 [curp->secondary->sz] = '\0';
471 }
472 rerun:
473 rr = roff_parseln(curp->roff, curp->line,
474 &ln.buf, &ln.sz, of, &of);
475
476 switch (rr) {
477 case ROFF_REPARSE:
478 if (REPARSE_LIMIT >= ++curp->reparse_count)
479 mparse_buf_r(curp, ln, 0);
480 else
481 mandoc_msg(MANDOCERR_ROFFLOOP, curp,
482 curp->line, pos, NULL);
483 pos = 0;
484 continue;
485 case ROFF_APPEND:
486 pos = (int)strlen(ln.buf);
487 continue;
488 case ROFF_RERUN:
489 goto rerun;
490 case ROFF_IGN:
491 pos = 0;
492 continue;
493 case ROFF_ERR:
494 assert(MANDOCLEVEL_FATAL <= curp->file_status);
495 break;
496 case ROFF_SO:
497 if (0 == (MPARSE_SO & curp->options) &&
498 (i >= (int)blk.sz || '\0' == blk.buf[i])) {
499 curp->sodest = mandoc_strdup(ln.buf + of);
500 free(ln.buf);
501 return;
502 }
503 /*
504 * We remove `so' clauses from our lookaside
505 * buffer because we're going to descend into
506 * the file recursively.
507 */
508 if (curp->secondary)
509 curp->secondary->sz -= pos + 1;
510 mparse_readfd(curp, -1, ln.buf + of);
511 if (MANDOCLEVEL_FATAL <= curp->file_status) {
512 mandoc_vmsg(MANDOCERR_SO_FAIL,
513 curp, curp->line, pos,
514 ".so %s", ln.buf + of);
515 break;
516 }
517 pos = 0;
518 continue;
519 default:
520 break;
521 }
522
523 /*
524 * If we encounter errors in the recursive parse, make
525 * sure we don't continue parsing.
526 */
527
528 if (MANDOCLEVEL_FATAL <= curp->file_status)
529 break;
530
531 /*
532 * If input parsers have not been allocated, do so now.
533 * We keep these instanced between parsers, but set them
534 * locally per parse routine since we can use different
535 * parsers with each one.
536 */
537
538 if ( ! (curp->man || curp->mdoc))
539 pset(ln.buf + of, pos - of, curp);
540
541 /*
542 * Lastly, push down into the parsers themselves. One
543 * of these will have already been set in the pset()
544 * routine.
545 * If libroff returns ROFF_TBL, then add it to the
546 * currently open parse. Since we only get here if
547 * there does exist data (see tbl_data.c), we're
548 * guaranteed that something's been allocated.
549 * Do the same for ROFF_EQN.
550 */
551
552 rc = -1;
553
554 if (ROFF_TBL == rr)
555 while (NULL != (span = roff_span(curp->roff))) {
556 rc = curp->man ?
557 man_addspan(curp->man, span) :
558 mdoc_addspan(curp->mdoc, span);
559 if (0 == rc)
560 break;
561 }
562 else if (ROFF_EQN == rr)
563 rc = curp->mdoc ?
564 mdoc_addeqn(curp->mdoc,
565 roff_eqn(curp->roff)) :
566 man_addeqn(curp->man,
567 roff_eqn(curp->roff));
568 else if (curp->man || curp->mdoc)
569 rc = curp->man ?
570 man_parseln(curp->man,
571 curp->line, ln.buf, of) :
572 mdoc_parseln(curp->mdoc,
573 curp->line, ln.buf, of);
574
575 if (0 == rc) {
576 assert(MANDOCLEVEL_FATAL <= curp->file_status);
577 break;
578 } else if (2 == rc)
579 break;
580
581 /* Temporary buffers typically are not full. */
582
583 if (0 == start && '\0' == blk.buf[i])
584 break;
585
586 /* Start the next input line. */
587
588 pos = 0;
589 }
590
591 free(ln.buf);
592 }
593
594 static int
595 read_whole_file(struct mparse *curp, const char *file, int fd,
596 struct buf *fb, int *with_mmap)
597 {
598 size_t off;
599 ssize_t ssz;
600
601 #ifdef HAVE_MMAP
602 struct stat st;
603 if (-1 == fstat(fd, &st)) {
604 curp->file_status = MANDOCLEVEL_SYSERR;
605 if (curp->mmsg)
606 (*curp->mmsg)(MANDOCERR_SYSSTAT, curp->file_status,
607 file, 0, 0, strerror(errno));
608 return(0);
609 }
610
611 /*
612 * If we're a regular file, try just reading in the whole entry
613 * via mmap(). This is faster than reading it into blocks, and
614 * since each file is only a few bytes to begin with, I'm not
615 * concerned that this is going to tank any machines.
616 */
617
618 if (S_ISREG(st.st_mode)) {
619 if (st.st_size >= (1U << 31)) {
620 curp->file_status = MANDOCLEVEL_FATAL;
621 if (curp->mmsg)
622 (*curp->mmsg)(MANDOCERR_TOOLARGE,
623 curp->file_status, file, 0, 0, NULL);
624 return(0);
625 }
626 *with_mmap = 1;
627 fb->sz = (size_t)st.st_size;
628 fb->buf = mmap(NULL, fb->sz, PROT_READ, MAP_SHARED, fd, 0);
629 if (fb->buf != MAP_FAILED)
630 return(1);
631 }
632 #endif
633
634 /*
635 * If this isn't a regular file (like, say, stdin), then we must
636 * go the old way and just read things in bit by bit.
637 */
638
639 *with_mmap = 0;
640 off = 0;
641 fb->sz = 0;
642 fb->buf = NULL;
643 for (;;) {
644 if (off == fb->sz) {
645 if (fb->sz == (1U << 31)) {
646 curp->file_status = MANDOCLEVEL_FATAL;
647 if (curp->mmsg)
648 (*curp->mmsg)(MANDOCERR_TOOLARGE,
649 curp->file_status,
650 file, 0, 0, NULL);
651 break;
652 }
653 resize_buf(fb, 65536);
654 }
655 ssz = read(fd, fb->buf + (int)off, fb->sz - off);
656 if (ssz == 0) {
657 fb->sz = off;
658 return(1);
659 }
660 if (ssz == -1) {
661 curp->file_status = MANDOCLEVEL_SYSERR;
662 if (curp->mmsg)
663 (*curp->mmsg)(MANDOCERR_SYSREAD,
664 curp->file_status, file, 0, 0,
665 strerror(errno));
666 break;
667 }
668 off += (size_t)ssz;
669 }
670
671 free(fb->buf);
672 fb->buf = NULL;
673 return(0);
674 }
675
676 static void
677 mparse_end(struct mparse *curp)
678 {
679
680 if (MANDOCLEVEL_FATAL <= curp->file_status)
681 return;
682
683 if (curp->mdoc && ! mdoc_endparse(curp->mdoc)) {
684 assert(MANDOCLEVEL_FATAL <= curp->file_status);
685 return;
686 }
687
688 if (curp->man && ! man_endparse(curp->man)) {
689 assert(MANDOCLEVEL_FATAL <= curp->file_status);
690 return;
691 }
692
693 if ( ! (curp->mdoc || curp->man || curp->sodest)) {
694 mandoc_msg(MANDOCERR_NOTMANUAL, curp, 0, 0, NULL);
695 curp->file_status = MANDOCLEVEL_FATAL;
696 return;
697 }
698
699 roff_endparse(curp->roff);
700 }
701
702 static void
703 mparse_parse_buffer(struct mparse *curp, struct buf blk, const char *file)
704 {
705 const char *svfile;
706 static int recursion_depth;
707
708 if (64 < recursion_depth) {
709 mandoc_msg(MANDOCERR_ROFFLOOP, curp, curp->line, 0, NULL);
710 return;
711 }
712
713 /* Line number is per-file. */
714 svfile = curp->file;
715 curp->file = file;
716 curp->line = 1;
717 recursion_depth++;
718
719 mparse_buf_r(curp, blk, 1);
720
721 if (0 == --recursion_depth && MANDOCLEVEL_FATAL > curp->file_status)
722 mparse_end(curp);
723
724 curp->file = svfile;
725 }
726
727 enum mandoclevel
728 mparse_readmem(struct mparse *curp, const void *buf, size_t len,
729 const char *file)
730 {
731 struct buf blk;
732
733 blk.buf = UNCONST(buf);
734 blk.sz = len;
735
736 mparse_parse_buffer(curp, blk, file);
737 return(curp->file_status);
738 }
739
740 enum mandoclevel
741 mparse_readfd(struct mparse *curp, int fd, const char *file)
742 {
743 struct buf blk;
744 int with_mmap;
745
746 if (-1 == fd && -1 == (fd = open(file, O_RDONLY, 0))) {
747 curp->file_status = MANDOCLEVEL_SYSERR;
748 if (curp->mmsg)
749 (*curp->mmsg)(MANDOCERR_SYSOPEN,
750 curp->file_status,
751 file, 0, 0, strerror(errno));
752 goto out;
753 }
754
755 /*
756 * Run for each opened file; may be called more than once for
757 * each full parse sequence if the opened file is nested (i.e.,
758 * from `so'). Simply sucks in the whole file and moves into
759 * the parse phase for the file.
760 */
761
762 if ( ! read_whole_file(curp, file, fd, &blk, &with_mmap))
763 goto out;
764
765 mparse_parse_buffer(curp, blk, file);
766
767 #ifdef HAVE_MMAP
768 if (with_mmap)
769 munmap(blk.buf, blk.sz);
770 else
771 #endif
772 free(blk.buf);
773
774 if (STDIN_FILENO != fd && -1 == close(fd))
775 perror(file);
776 out:
777 return(curp->file_status);
778 }
779
780 struct mparse *
781 mparse_alloc(int options, enum mandoclevel wlevel,
782 mandocmsg mmsg, char *defos)
783 {
784 struct mparse *curp;
785
786 assert(wlevel <= MANDOCLEVEL_FATAL);
787
788 curp = mandoc_calloc(1, sizeof(struct mparse));
789
790 curp->options = options;
791 curp->wlevel = wlevel;
792 curp->mmsg = mmsg;
793 curp->defos = defos;
794
795 curp->roff = roff_alloc(curp, options);
796 return(curp);
797 }
798
799 void
800 mparse_reset(struct mparse *curp)
801 {
802
803 roff_reset(curp->roff);
804
805 if (curp->mdoc)
806 mdoc_reset(curp->mdoc);
807 if (curp->man)
808 man_reset(curp->man);
809 if (curp->secondary)
810 curp->secondary->sz = 0;
811
812 curp->file_status = MANDOCLEVEL_OK;
813 curp->mdoc = NULL;
814 curp->man = NULL;
815
816 free(curp->sodest);
817 curp->sodest = NULL;
818 }
819
820 void
821 mparse_free(struct mparse *curp)
822 {
823
824 if (curp->pmdoc)
825 mdoc_free(curp->pmdoc);
826 if (curp->pman)
827 man_free(curp->pman);
828 if (curp->roff)
829 roff_free(curp->roff);
830 if (curp->secondary)
831 free(curp->secondary->buf);
832
833 free(curp->secondary);
834 free(curp->sodest);
835 free(curp);
836 }
837
838 void
839 mparse_result(struct mparse *curp,
840 struct mdoc **mdoc, struct man **man, char **sodest)
841 {
842
843 if (sodest && NULL != (*sodest = curp->sodest)) {
844 *mdoc = NULL;
845 *man = NULL;
846 return;
847 }
848 if (mdoc)
849 *mdoc = curp->mdoc;
850 if (man)
851 *man = curp->man;
852 }
853
854 void
855 mandoc_vmsg(enum mandocerr t, struct mparse *m,
856 int ln, int pos, const char *fmt, ...)
857 {
858 char buf[256];
859 va_list ap;
860
861 va_start(ap, fmt);
862 (void)vsnprintf(buf, sizeof(buf), fmt, ap);
863 va_end(ap);
864
865 mandoc_msg(t, m, ln, pos, buf);
866 }
867
868 void
869 mandoc_msg(enum mandocerr er, struct mparse *m,
870 int ln, int col, const char *msg)
871 {
872 enum mandoclevel level;
873
874 level = MANDOCLEVEL_FATAL;
875 while (er < mandoclimits[level])
876 level--;
877
878 if (level < m->wlevel)
879 return;
880
881 if (m->mmsg)
882 (*m->mmsg)(er, level, m->file, ln, col, msg);
883
884 if (m->file_status < level)
885 m->file_status = level;
886 }
887
888 const char *
889 mparse_strerror(enum mandocerr er)
890 {
891
892 return(mandocerrs[er]);
893 }
894
895 const char *
896 mparse_strlevel(enum mandoclevel lvl)
897 {
898 return(mandoclevels[lvl]);
899 }
900
901 void
902 mparse_keep(struct mparse *p)
903 {
904
905 assert(NULL == p->secondary);
906 p->secondary = mandoc_calloc(1, sizeof(struct buf));
907 }
908
909 const char *
910 mparse_getkeep(const struct mparse *p)
911 {
912
913 assert(p->secondary);
914 return(p->secondary->sz ? p->secondary->buf : NULL);
915 }