]> git.cameronkatri.com Git - mandoc.git/blob - read.c
a99ca0584ccfb1fffaf289b9b82d5d39c46da93c
[mandoc.git] / read.c
1 /* $Id: read.c,v 1.45 2014/03/19 22:20:43 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 "libmandoc.h"
41 #include "mdoc.h"
42 #include "man.h"
43 #include "main.h"
44
45 #define REPARSE_LIMIT 1000
46
47 struct buf {
48 char *buf; /* binary input buffer */
49 size_t sz; /* size of binary buffer */
50 };
51
52 struct mparse {
53 enum mandoclevel file_status; /* status of current parse */
54 enum mandoclevel wlevel; /* ignore messages below this */
55 int line; /* line number in the file */
56 int options; /* parser options */
57 struct man *pman; /* persistent man parser */
58 struct mdoc *pmdoc; /* persistent mdoc parser */
59 struct man *man; /* man parser */
60 struct mdoc *mdoc; /* mdoc parser */
61 struct roff *roff; /* roff parser (!NULL) */
62 char *sodest; /* filename pointed to by .so */
63 int reparse_count; /* finite interp. stack */
64 mandocmsg mmsg; /* warning/error message handler */
65 const char *file;
66 struct buf *secondary;
67 char *defos; /* default operating system */
68 };
69
70 static void resize_buf(struct buf *, size_t);
71 static void mparse_buf_r(struct mparse *, struct buf, int);
72 static void pset(const char *, int, struct mparse *);
73 static int read_whole_file(struct mparse *, const char *, int,
74 struct buf *, int *);
75 static void mparse_end(struct mparse *);
76 static void mparse_parse_buffer(struct mparse *, struct buf,
77 const char *);
78
79 static const enum mandocerr mandoclimits[MANDOCLEVEL_MAX] = {
80 MANDOCERR_OK,
81 MANDOCERR_WARNING,
82 MANDOCERR_WARNING,
83 MANDOCERR_ERROR,
84 MANDOCERR_FATAL,
85 MANDOCERR_MAX,
86 MANDOCERR_MAX
87 };
88
89 static const char * const mandocerrs[MANDOCERR_MAX] = {
90 "ok",
91
92 "generic warning",
93
94 /* related to the prologue */
95 "no title in document",
96 "document title should be all caps",
97 "unknown manual section",
98 "unknown manual volume or arch",
99 "date missing, using today's date",
100 "cannot parse date, using it verbatim",
101 "prologue macros out of order",
102 "duplicate prologue macro",
103 "macro not allowed in prologue",
104 "macro not allowed in body",
105
106 /* related to document structure */
107 ".so is fragile, better use ln(1)",
108 "NAME section must come first",
109 "bad NAME section contents",
110 "sections out of conventional order",
111 "duplicate section name",
112 "section header suited to sections 2, 3, and 9 only",
113
114 /* related to macros and nesting */
115 "skipping obsolete macro",
116 "skipping paragraph macro",
117 "moving paragraph macro out of list",
118 "skipping no-space macro",
119 "blocks badly nested",
120 "child violates parent syntax",
121 "nested displays are not portable",
122 "already in literal mode",
123 "line scope broken",
124
125 /* related to missing macro arguments */
126 "skipping empty macro",
127 "argument count wrong",
128 "missing display type",
129 "list type must come first",
130 "tag lists require a width argument",
131 "missing font type",
132 "skipping end of block that is not open",
133
134 /* related to bad macro arguments */
135 "skipping argument",
136 "duplicate argument",
137 "duplicate display type",
138 "duplicate list type",
139 "unknown AT&T UNIX version",
140 "bad Boolean value",
141 "unknown font",
142 "unknown standard specifier",
143 "bad width argument",
144
145 /* related to plain text */
146 "blank line in non-literal context",
147 "tab in non-literal context",
148 "end of line whitespace",
149 "bad comment style",
150 "bad escape sequence",
151 "unterminated quoted string",
152
153 /* related to equations */
154 "unexpected literal in equation",
155
156 "generic error",
157
158 /* related to equations */
159 "unexpected equation scope closure",
160 "equation scope open on exit",
161 "overlapping equation scopes",
162 "unexpected end of equation",
163 "equation syntax error",
164
165 /* related to tables */
166 "bad table syntax",
167 "bad table option",
168 "bad table layout",
169 "no table layout cells specified",
170 "no table data cells specified",
171 "ignore data in cell",
172 "data block still open",
173 "ignoring extra data cells",
174
175 "input stack limit exceeded, infinite loop?",
176 "skipping bad character",
177 "escaped character not allowed in a name",
178 "manual name not yet set",
179 "skipping text before the first section header",
180 "skipping unknown macro",
181 "NOT IMPLEMENTED, please use groff: skipping request",
182 "argument count wrong",
183 "skipping column outside column list",
184 "skipping end of block that is not open",
185 "missing end of block",
186 "scope open on exit",
187 "uname(3) system call failed",
188 "macro requires line argument(s)",
189 "macro requires body argument(s)",
190 "macro requires argument(s)",
191 "request requires a numeric argument",
192 "missing list type",
193 "line argument(s) will be lost",
194 "body argument(s) will be lost",
195
196 "generic fatal error",
197
198 "input too large",
199 "not a manual",
200 "column syntax is inconsistent",
201 "NOT IMPLEMENTED: .Bd -file",
202 "argument count wrong, violates syntax",
203 "child violates parent syntax",
204 "argument count wrong, violates syntax",
205 "NOT IMPLEMENTED: .so with absolute path or \"..\"",
206 "no document body",
207 "no document prologue",
208 "static buffer exhausted",
209
210 /* system errors */
211 "cannot open file",
212 "cannot stat file",
213 "cannot read file",
214 };
215
216 static const char * const mandoclevels[MANDOCLEVEL_MAX] = {
217 "SUCCESS",
218 "RESERVED",
219 "WARNING",
220 "ERROR",
221 "FATAL",
222 "BADARG",
223 "SYSERR"
224 };
225
226 static void
227 resize_buf(struct buf *buf, size_t initial)
228 {
229
230 buf->sz = buf->sz > initial/2 ? 2 * buf->sz : initial;
231 buf->buf = mandoc_realloc(buf->buf, buf->sz);
232 }
233
234 static void
235 pset(const char *buf, int pos, struct mparse *curp)
236 {
237 int i;
238
239 /*
240 * Try to intuit which kind of manual parser should be used. If
241 * passed in by command-line (-man, -mdoc), then use that
242 * explicitly. If passed as -mandoc, then try to guess from the
243 * line: either skip dot-lines, use -mdoc when finding `.Dt', or
244 * default to -man, which is more lenient.
245 *
246 * Separate out pmdoc/pman from mdoc/man: the first persists
247 * through all parsers, while the latter is used per-parse.
248 */
249
250 if ('.' == buf[0] || '\'' == buf[0]) {
251 for (i = 1; buf[i]; i++)
252 if (' ' != buf[i] && '\t' != buf[i])
253 break;
254 if ('\0' == buf[i])
255 return;
256 }
257
258 if (MPARSE_MDOC & curp->options) {
259 if (NULL == curp->pmdoc)
260 curp->pmdoc = mdoc_alloc(
261 curp->roff, curp, curp->defos,
262 MPARSE_QUICK & curp->options ? 1 : 0);
263 assert(curp->pmdoc);
264 curp->mdoc = curp->pmdoc;
265 return;
266 } else if (MPARSE_MAN & curp->options) {
267 if (NULL == curp->pman)
268 curp->pman = man_alloc(curp->roff, curp,
269 MPARSE_QUICK & curp->options ? 1 : 0);
270 assert(curp->pman);
271 curp->man = curp->pman;
272 return;
273 }
274
275 if (pos >= 3 && 0 == memcmp(buf, ".Dd", 3)) {
276 if (NULL == curp->pmdoc)
277 curp->pmdoc = mdoc_alloc(
278 curp->roff, curp, curp->defos,
279 MPARSE_QUICK & curp->options ? 1 : 0);
280 assert(curp->pmdoc);
281 curp->mdoc = curp->pmdoc;
282 return;
283 }
284
285 if (NULL == curp->pman)
286 curp->pman = man_alloc(curp->roff, curp,
287 MPARSE_QUICK & curp->options ? 1 : 0);
288 assert(curp->pman);
289 curp->man = curp->pman;
290 }
291
292 /*
293 * Main parse routine for an opened file. This is called for each
294 * opened file and simply loops around the full input file, possibly
295 * nesting (i.e., with `so').
296 */
297 static void
298 mparse_buf_r(struct mparse *curp, struct buf blk, int start)
299 {
300 const struct tbl_span *span;
301 struct buf ln;
302 enum rofferr rr;
303 int i, of, rc;
304 int pos; /* byte number in the ln buffer */
305 int lnn; /* line number in the real file */
306 unsigned char c;
307
308 memset(&ln, 0, sizeof(struct buf));
309
310 lnn = curp->line;
311 pos = 0;
312
313 for (i = 0; i < (int)blk.sz; ) {
314 if (0 == pos && '\0' == blk.buf[i])
315 break;
316
317 if (start) {
318 curp->line = lnn;
319 curp->reparse_count = 0;
320 }
321
322 while (i < (int)blk.sz && (start || '\0' != blk.buf[i])) {
323
324 /*
325 * When finding an unescaped newline character,
326 * leave the character loop to process the line.
327 * Skip a preceding carriage return, if any.
328 */
329
330 if ('\r' == blk.buf[i] && i + 1 < (int)blk.sz &&
331 '\n' == blk.buf[i + 1])
332 ++i;
333 if ('\n' == blk.buf[i]) {
334 ++i;
335 ++lnn;
336 break;
337 }
338
339 /*
340 * Make sure we have space for at least
341 * one backslash and one other character
342 * and the trailing NUL byte.
343 */
344
345 if (pos + 2 >= (int)ln.sz)
346 resize_buf(&ln, 256);
347
348 /*
349 * Warn about bogus characters. If you're using
350 * non-ASCII encoding, you're screwing your
351 * readers. Since I'd rather this not happen,
352 * I'll be helpful and replace these characters
353 * with "?", so we don't display gibberish.
354 * Note to manual writers: use special characters.
355 */
356
357 c = (unsigned char) blk.buf[i];
358
359 if ( ! (isascii(c) &&
360 (isgraph(c) || isblank(c)))) {
361 mandoc_msg(MANDOCERR_BADCHAR, curp,
362 curp->line, pos, NULL);
363 i++;
364 ln.buf[pos++] = '?';
365 continue;
366 }
367
368 /* Trailing backslash = a plain char. */
369
370 if ('\\' != blk.buf[i] || i + 1 == (int)blk.sz) {
371 ln.buf[pos++] = blk.buf[i++];
372 continue;
373 }
374
375 /*
376 * Found escape and at least one other character.
377 * When it's a newline character, skip it.
378 * When there is a carriage return in between,
379 * skip that one as well.
380 */
381
382 if ('\r' == blk.buf[i + 1] && i + 2 < (int)blk.sz &&
383 '\n' == blk.buf[i + 2])
384 ++i;
385 if ('\n' == blk.buf[i + 1]) {
386 i += 2;
387 ++lnn;
388 continue;
389 }
390
391 if ('"' == blk.buf[i + 1] || '#' == blk.buf[i + 1]) {
392 i += 2;
393 /* Comment, skip to end of line */
394 for (; i < (int)blk.sz; ++i) {
395 if ('\n' == blk.buf[i]) {
396 ++i;
397 ++lnn;
398 break;
399 }
400 }
401
402 /* Backout trailing whitespaces */
403 for (; pos > 0; --pos) {
404 if (ln.buf[pos - 1] != ' ')
405 break;
406 if (pos > 2 && ln.buf[pos - 2] == '\\')
407 break;
408 }
409 break;
410 }
411
412 /* Catch escaped bogus characters. */
413
414 c = (unsigned char) blk.buf[i+1];
415
416 if ( ! (isascii(c) &&
417 (isgraph(c) || isblank(c)))) {
418 mandoc_msg(MANDOCERR_BADCHAR, curp,
419 curp->line, pos, NULL);
420 i += 2;
421 ln.buf[pos++] = '?';
422 continue;
423 }
424
425 /* Some other escape sequence, copy & cont. */
426
427 ln.buf[pos++] = blk.buf[i++];
428 ln.buf[pos++] = blk.buf[i++];
429 }
430
431 if (pos >= (int)ln.sz)
432 resize_buf(&ln, 256);
433
434 ln.buf[pos] = '\0';
435
436 /*
437 * A significant amount of complexity is contained by
438 * the roff preprocessor. It's line-oriented but can be
439 * expressed on one line, so we need at times to
440 * readjust our starting point and re-run it. The roff
441 * preprocessor can also readjust the buffers with new
442 * data, so we pass them in wholesale.
443 */
444
445 of = 0;
446
447 /*
448 * Maintain a lookaside buffer of all parsed lines. We
449 * only do this if mparse_keep() has been invoked (the
450 * buffer may be accessed with mparse_getkeep()).
451 */
452
453 if (curp->secondary) {
454 curp->secondary->buf =
455 mandoc_realloc
456 (curp->secondary->buf,
457 curp->secondary->sz + pos + 2);
458 memcpy(curp->secondary->buf +
459 curp->secondary->sz,
460 ln.buf, pos);
461 curp->secondary->sz += pos;
462 curp->secondary->buf
463 [curp->secondary->sz] = '\n';
464 curp->secondary->sz++;
465 curp->secondary->buf
466 [curp->secondary->sz] = '\0';
467 }
468 rerun:
469 rr = roff_parseln
470 (curp->roff, curp->line,
471 &ln.buf, &ln.sz, of, &of);
472
473 switch (rr) {
474 case (ROFF_REPARSE):
475 if (REPARSE_LIMIT >= ++curp->reparse_count)
476 mparse_buf_r(curp, ln, 0);
477 else
478 mandoc_msg(MANDOCERR_ROFFLOOP, curp,
479 curp->line, pos, NULL);
480 pos = 0;
481 continue;
482 case (ROFF_APPEND):
483 pos = (int)strlen(ln.buf);
484 continue;
485 case (ROFF_RERUN):
486 goto rerun;
487 case (ROFF_IGN):
488 pos = 0;
489 continue;
490 case (ROFF_ERR):
491 assert(MANDOCLEVEL_FATAL <= curp->file_status);
492 break;
493 case (ROFF_SO):
494 if (0 == (MPARSE_SO & curp->options) &&
495 (i >= (int)blk.sz || '\0' == blk.buf[i])) {
496 curp->sodest = mandoc_strdup(ln.buf + of);
497 free(ln.buf);
498 return;
499 }
500 /*
501 * We remove `so' clauses from our lookaside
502 * buffer because we're going to descend into
503 * the file recursively.
504 */
505 if (curp->secondary)
506 curp->secondary->sz -= pos + 1;
507 mparse_readfd(curp, -1, ln.buf + of);
508 if (MANDOCLEVEL_FATAL <= curp->file_status)
509 break;
510 pos = 0;
511 continue;
512 default:
513 break;
514 }
515
516 /*
517 * If we encounter errors in the recursive parse, make
518 * sure we don't continue parsing.
519 */
520
521 if (MANDOCLEVEL_FATAL <= curp->file_status)
522 break;
523
524 /*
525 * If input parsers have not been allocated, do so now.
526 * We keep these instanced between parsers, but set them
527 * locally per parse routine since we can use different
528 * parsers with each one.
529 */
530
531 if ( ! (curp->man || curp->mdoc))
532 pset(ln.buf + of, pos - of, curp);
533
534 /*
535 * Lastly, push down into the parsers themselves. One
536 * of these will have already been set in the pset()
537 * routine.
538 * If libroff returns ROFF_TBL, then add it to the
539 * currently open parse. Since we only get here if
540 * there does exist data (see tbl_data.c), we're
541 * guaranteed that something's been allocated.
542 * Do the same for ROFF_EQN.
543 */
544
545 rc = -1;
546
547 if (ROFF_TBL == rr)
548 while (NULL != (span = roff_span(curp->roff))) {
549 rc = curp->man ?
550 man_addspan(curp->man, span) :
551 mdoc_addspan(curp->mdoc, span);
552 if (0 == rc)
553 break;
554 }
555 else if (ROFF_EQN == rr)
556 rc = curp->mdoc ?
557 mdoc_addeqn(curp->mdoc,
558 roff_eqn(curp->roff)) :
559 man_addeqn(curp->man,
560 roff_eqn(curp->roff));
561 else if (curp->man || curp->mdoc)
562 rc = curp->man ?
563 man_parseln(curp->man,
564 curp->line, ln.buf, of) :
565 mdoc_parseln(curp->mdoc,
566 curp->line, ln.buf, of);
567
568 if (0 == rc) {
569 assert(MANDOCLEVEL_FATAL <= curp->file_status);
570 break;
571 } else if (2 == rc)
572 break;
573
574 /* Temporary buffers typically are not full. */
575
576 if (0 == start && '\0' == blk.buf[i])
577 break;
578
579 /* Start the next input line. */
580
581 pos = 0;
582 }
583
584 free(ln.buf);
585 }
586
587 static int
588 read_whole_file(struct mparse *curp, const char *file, int fd,
589 struct buf *fb, int *with_mmap)
590 {
591 size_t off;
592 ssize_t ssz;
593
594 #ifdef HAVE_MMAP
595 struct stat st;
596 if (-1 == fstat(fd, &st)) {
597 curp->file_status = MANDOCLEVEL_SYSERR;
598 if (curp->mmsg)
599 (*curp->mmsg)(MANDOCERR_SYSSTAT, curp->file_status,
600 file, 0, 0, strerror(errno));
601 return(0);
602 }
603
604 /*
605 * If we're a regular file, try just reading in the whole entry
606 * via mmap(). This is faster than reading it into blocks, and
607 * since each file is only a few bytes to begin with, I'm not
608 * concerned that this is going to tank any machines.
609 */
610
611 if (S_ISREG(st.st_mode)) {
612 if (st.st_size >= (1U << 31)) {
613 curp->file_status = MANDOCLEVEL_FATAL;
614 if (curp->mmsg)
615 (*curp->mmsg)(MANDOCERR_TOOLARGE,
616 curp->file_status, file, 0, 0, NULL);
617 return(0);
618 }
619 *with_mmap = 1;
620 fb->sz = (size_t)st.st_size;
621 fb->buf = mmap(NULL, fb->sz, PROT_READ, MAP_SHARED, fd, 0);
622 if (fb->buf != MAP_FAILED)
623 return(1);
624 }
625 #endif
626
627 /*
628 * If this isn't a regular file (like, say, stdin), then we must
629 * go the old way and just read things in bit by bit.
630 */
631
632 *with_mmap = 0;
633 off = 0;
634 fb->sz = 0;
635 fb->buf = NULL;
636 for (;;) {
637 if (off == fb->sz) {
638 if (fb->sz == (1U << 31)) {
639 curp->file_status = MANDOCLEVEL_FATAL;
640 if (curp->mmsg)
641 (*curp->mmsg)(MANDOCERR_TOOLARGE,
642 curp->file_status,
643 file, 0, 0, NULL);
644 break;
645 }
646 resize_buf(fb, 65536);
647 }
648 ssz = read(fd, fb->buf + (int)off, fb->sz - off);
649 if (ssz == 0) {
650 fb->sz = off;
651 return(1);
652 }
653 if (ssz == -1) {
654 curp->file_status = MANDOCLEVEL_SYSERR;
655 if (curp->mmsg)
656 (*curp->mmsg)(MANDOCERR_SYSREAD,
657 curp->file_status, file, 0, 0,
658 strerror(errno));
659 break;
660 }
661 off += (size_t)ssz;
662 }
663
664 free(fb->buf);
665 fb->buf = NULL;
666 return(0);
667 }
668
669 static void
670 mparse_end(struct mparse *curp)
671 {
672
673 if (MANDOCLEVEL_FATAL <= curp->file_status)
674 return;
675
676 if (curp->mdoc && ! mdoc_endparse(curp->mdoc)) {
677 assert(MANDOCLEVEL_FATAL <= curp->file_status);
678 return;
679 }
680
681 if (curp->man && ! man_endparse(curp->man)) {
682 assert(MANDOCLEVEL_FATAL <= curp->file_status);
683 return;
684 }
685
686 if ( ! (curp->mdoc || curp->man || curp->sodest)) {
687 mandoc_msg(MANDOCERR_NOTMANUAL, curp, 1, 0, NULL);
688 curp->file_status = MANDOCLEVEL_FATAL;
689 return;
690 }
691
692 roff_endparse(curp->roff);
693 }
694
695 static void
696 mparse_parse_buffer(struct mparse *curp, struct buf blk, const char *file)
697 {
698 const char *svfile;
699 static int recursion_depth;
700
701 if (64 < recursion_depth) {
702 mandoc_msg(MANDOCERR_ROFFLOOP, curp, curp->line, 0, NULL);
703 return;
704 }
705
706 /* Line number is per-file. */
707 svfile = curp->file;
708 curp->file = file;
709 curp->line = 1;
710 recursion_depth++;
711
712 mparse_buf_r(curp, blk, 1);
713
714 if (0 == --recursion_depth && MANDOCLEVEL_FATAL > curp->file_status)
715 mparse_end(curp);
716
717 curp->file = svfile;
718 }
719
720 enum mandoclevel
721 mparse_readmem(struct mparse *curp, const void *buf, size_t len,
722 const char *file)
723 {
724 struct buf blk;
725
726 blk.buf = UNCONST(buf);
727 blk.sz = len;
728
729 mparse_parse_buffer(curp, blk, file);
730 return(curp->file_status);
731 }
732
733 enum mandoclevel
734 mparse_readfd(struct mparse *curp, int fd, const char *file)
735 {
736 struct buf blk;
737 int with_mmap;
738
739 if (-1 == fd && -1 == (fd = open(file, O_RDONLY, 0))) {
740 curp->file_status = MANDOCLEVEL_SYSERR;
741 if (curp->mmsg)
742 (*curp->mmsg)(MANDOCERR_SYSOPEN,
743 curp->file_status,
744 file, 0, 0, strerror(errno));
745 goto out;
746 }
747
748 /*
749 * Run for each opened file; may be called more than once for
750 * each full parse sequence if the opened file is nested (i.e.,
751 * from `so'). Simply sucks in the whole file and moves into
752 * the parse phase for the file.
753 */
754
755 if ( ! read_whole_file(curp, file, fd, &blk, &with_mmap))
756 goto out;
757
758 mparse_parse_buffer(curp, blk, file);
759
760 #ifdef HAVE_MMAP
761 if (with_mmap)
762 munmap(blk.buf, blk.sz);
763 else
764 #endif
765 free(blk.buf);
766
767 if (STDIN_FILENO != fd && -1 == close(fd))
768 perror(file);
769 out:
770 return(curp->file_status);
771 }
772
773 struct mparse *
774 mparse_alloc(int options, enum mandoclevel wlevel,
775 mandocmsg mmsg, char *defos)
776 {
777 struct mparse *curp;
778
779 assert(wlevel <= MANDOCLEVEL_FATAL);
780
781 curp = mandoc_calloc(1, sizeof(struct mparse));
782
783 curp->options = options;
784 curp->wlevel = wlevel;
785 curp->mmsg = mmsg;
786 curp->defos = defos;
787
788 curp->roff = roff_alloc(curp, options);
789 return(curp);
790 }
791
792 void
793 mparse_reset(struct mparse *curp)
794 {
795
796 roff_reset(curp->roff);
797
798 if (curp->mdoc)
799 mdoc_reset(curp->mdoc);
800 if (curp->man)
801 man_reset(curp->man);
802 if (curp->secondary)
803 curp->secondary->sz = 0;
804
805 curp->file_status = MANDOCLEVEL_OK;
806 curp->mdoc = NULL;
807 curp->man = NULL;
808
809 free(curp->sodest);
810 curp->sodest = NULL;
811 }
812
813 void
814 mparse_free(struct mparse *curp)
815 {
816
817 if (curp->pmdoc)
818 mdoc_free(curp->pmdoc);
819 if (curp->pman)
820 man_free(curp->pman);
821 if (curp->roff)
822 roff_free(curp->roff);
823 if (curp->secondary)
824 free(curp->secondary->buf);
825
826 free(curp->secondary);
827 free(curp->sodest);
828 free(curp);
829 }
830
831 void
832 mparse_result(struct mparse *curp,
833 struct mdoc **mdoc, struct man **man, char **sodest)
834 {
835
836 if (sodest && NULL != (*sodest = curp->sodest)) {
837 *mdoc = NULL;
838 *man = NULL;
839 return;
840 }
841 if (mdoc)
842 *mdoc = curp->mdoc;
843 if (man)
844 *man = curp->man;
845 }
846
847 void
848 mandoc_vmsg(enum mandocerr t, struct mparse *m,
849 int ln, int pos, const char *fmt, ...)
850 {
851 char buf[256];
852 va_list ap;
853
854 va_start(ap, fmt);
855 vsnprintf(buf, sizeof(buf) - 1, fmt, ap);
856 va_end(ap);
857
858 mandoc_msg(t, m, ln, pos, buf);
859 }
860
861 void
862 mandoc_msg(enum mandocerr er, struct mparse *m,
863 int ln, int col, const char *msg)
864 {
865 enum mandoclevel level;
866
867 level = MANDOCLEVEL_FATAL;
868 while (er < mandoclimits[level])
869 level--;
870
871 if (level < m->wlevel)
872 return;
873
874 if (m->mmsg)
875 (*m->mmsg)(er, level, m->file, ln, col, msg);
876
877 if (m->file_status < level)
878 m->file_status = level;
879 }
880
881 const char *
882 mparse_strerror(enum mandocerr er)
883 {
884
885 return(mandocerrs[er]);
886 }
887
888 const char *
889 mparse_strlevel(enum mandoclevel lvl)
890 {
891 return(mandoclevels[lvl]);
892 }
893
894 void
895 mparse_keep(struct mparse *p)
896 {
897
898 assert(NULL == p->secondary);
899 p->secondary = mandoc_calloc(1, sizeof(struct buf));
900 }
901
902 const char *
903 mparse_getkeep(const struct mparse *p)
904 {
905
906 assert(p->secondary);
907 return(p->secondary->sz ? p->secondary->buf : NULL);
908 }