]> git.cameronkatri.com Git - mandoc.git/blob - read.c
Add support for `gsize' eqn token (introduced in second-edition troff).
[mandoc.git] / read.c
1 /* $Id: read.c,v 1.22 2011/07/22 14:55:07 kristaps Exp $ */
2 /*
3 * Copyright (c) 2008, 2009, 2010, 2011 Kristaps Dzonsons <kristaps@bsd.lv>
4 * Copyright (c) 2010, 2011 Ingo Schwarze <schwarze@openbsd.org>
5 *
6 * Permission to use, copy, modify, and distribute this software for any
7 * purpose with or without fee is hereby granted, provided that the above
8 * copyright notice and this permission notice appear in all copies.
9 *
10 * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
11 * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
12 * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
13 * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
14 * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
15 * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
16 * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
17 */
18 #ifdef HAVE_CONFIG_H
19 #include "config.h"
20 #endif
21
22 #ifdef HAVE_MMAP
23 # include <sys/stat.h>
24 # include <sys/mman.h>
25 #endif
26
27 #include <assert.h>
28 #include <ctype.h>
29 #include <fcntl.h>
30 #include <stdarg.h>
31 #include <stdio.h>
32 #include <stdlib.h>
33 #include <string.h>
34 #include <unistd.h>
35
36 #include "mandoc.h"
37 #include "libmandoc.h"
38 #include "mdoc.h"
39 #include "man.h"
40
41 #ifndef MAP_FILE
42 #define MAP_FILE 0
43 #endif
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 enum mparset inttype; /* which parser to use */
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 int reparse_count; /* finite interp. stack */
63 mandocmsg mmsg; /* warning/error message handler */
64 void *arg; /* argument to mmsg */
65 const char *file;
66 };
67
68 static void resize_buf(struct buf *, size_t);
69 static void mparse_buf_r(struct mparse *, struct buf, int);
70 static void mparse_readfd_r(struct mparse *, int, const char *, int);
71 static void pset(const char *, int, struct mparse *);
72 static void pdesc(struct mparse *, const char *, int);
73 static int read_whole_file(const char *, int, struct buf *, int *);
74 static void mparse_end(struct mparse *);
75
76 static const enum mandocerr mandoclimits[MANDOCLEVEL_MAX] = {
77 MANDOCERR_OK,
78 MANDOCERR_WARNING,
79 MANDOCERR_WARNING,
80 MANDOCERR_ERROR,
81 MANDOCERR_FATAL,
82 MANDOCERR_MAX,
83 MANDOCERR_MAX
84 };
85
86 static const char * const mandocerrs[MANDOCERR_MAX] = {
87 "ok",
88
89 "generic warning",
90
91 /* related to the prologue */
92 "no title in document",
93 "document title should be all caps",
94 "unknown manual section",
95 "date missing, using today's date",
96 "cannot parse date, using it verbatim",
97 "prologue macros out of order",
98 "duplicate prologue macro",
99 "macro not allowed in prologue",
100 "macro not allowed in body",
101
102 /* related to document structure */
103 ".so is fragile, better use ln(1)",
104 "NAME section must come first",
105 "bad NAME section contents",
106 "manual name not yet set",
107 "sections out of conventional order",
108 "duplicate section name",
109 "section not in conventional manual section",
110
111 /* related to macros and nesting */
112 "skipping obsolete macro",
113 "skipping paragraph macro",
114 "skipping no-space macro",
115 "blocks badly nested",
116 "child violates parent syntax",
117 "nested displays are not portable",
118 "already in literal mode",
119 "line scope broken",
120
121 /* related to missing macro arguments */
122 "skipping empty macro",
123 "argument count wrong",
124 "missing display type",
125 "list type must come first",
126 "tag lists require a width argument",
127 "missing font type",
128 "skipping end of block that is not open",
129
130 /* related to bad macro arguments */
131 "skipping argument",
132 "duplicate argument",
133 "duplicate display type",
134 "duplicate list type",
135 "unknown AT&T UNIX version",
136 "bad Boolean value",
137 "unknown font",
138 "unknown standard specifier",
139 "bad width argument",
140
141 /* related to plain text */
142 "blank line in non-literal context",
143 "tab in non-literal context",
144 "end of line whitespace",
145 "bad comment style",
146 "bad escape sequence",
147 "unterminated quoted string",
148
149 /* related to equations */
150 "unexpected literal in equation",
151
152 "generic error",
153
154 /* related to equations */
155 "too many nested equation defines",
156 "unexpected equation scope closure",
157 "equation scope open on exit",
158 "overlapping equation scopes",
159 "unexpected end of equation",
160 "equation syntax error",
161
162 /* related to tables */
163 "bad table syntax",
164 "bad table option",
165 "bad table layout",
166 "no table layout cells specified",
167 "no table data cells specified",
168 "ignore data in cell",
169 "data block still open",
170 "ignoring extra data cells",
171
172 "input stack limit exceeded, infinite loop?",
173 "skipping bad character",
174 "escaped character not allowed in a name",
175 "skipping text before the first section header",
176 "skipping unknown macro",
177 "NOT IMPLEMENTED, please use groff: skipping request",
178 "argument count wrong",
179 "skipping end of block that is not open",
180 "missing end of block",
181 "scope open on exit",
182 "uname(3) system call failed",
183 "macro requires line argument(s)",
184 "macro requires body argument(s)",
185 "macro requires argument(s)",
186 "missing list type",
187 "line argument(s) will be lost",
188 "body argument(s) will be lost",
189
190 "generic fatal error",
191
192 "not a manual",
193 "column syntax is inconsistent",
194 "NOT IMPLEMENTED: .Bd -file",
195 "line scope broken, syntax violated",
196 "argument count wrong, violates syntax",
197 "child violates parent syntax",
198 "argument count wrong, violates syntax",
199 "NOT IMPLEMENTED: .so with absolute path or \"..\"",
200 "no document body",
201 "no document prologue",
202 "static buffer exhausted",
203 };
204
205 static const char * const mandoclevels[MANDOCLEVEL_MAX] = {
206 "SUCCESS",
207 "RESERVED",
208 "WARNING",
209 "ERROR",
210 "FATAL",
211 "BADARG",
212 "SYSERR"
213 };
214
215 static void
216 resize_buf(struct buf *buf, size_t initial)
217 {
218
219 buf->sz = buf->sz > initial/2 ? 2 * buf->sz : initial;
220 buf->buf = mandoc_realloc(buf->buf, buf->sz);
221 }
222
223 static void
224 pset(const char *buf, int pos, struct mparse *curp)
225 {
226 int i;
227
228 /*
229 * Try to intuit which kind of manual parser should be used. If
230 * passed in by command-line (-man, -mdoc), then use that
231 * explicitly. If passed as -mandoc, then try to guess from the
232 * line: either skip dot-lines, use -mdoc when finding `.Dt', or
233 * default to -man, which is more lenient.
234 *
235 * Separate out pmdoc/pman from mdoc/man: the first persists
236 * through all parsers, while the latter is used per-parse.
237 */
238
239 if ('.' == buf[0] || '\'' == buf[0]) {
240 for (i = 1; buf[i]; i++)
241 if (' ' != buf[i] && '\t' != buf[i])
242 break;
243 if ('\0' == buf[i])
244 return;
245 }
246
247 switch (curp->inttype) {
248 case (MPARSE_MDOC):
249 if (NULL == curp->pmdoc)
250 curp->pmdoc = mdoc_alloc(curp->roff, curp);
251 assert(curp->pmdoc);
252 curp->mdoc = curp->pmdoc;
253 return;
254 case (MPARSE_MAN):
255 if (NULL == curp->pman)
256 curp->pman = man_alloc(curp->roff, curp);
257 assert(curp->pman);
258 curp->man = curp->pman;
259 return;
260 default:
261 break;
262 }
263
264 if (pos >= 3 && 0 == memcmp(buf, ".Dd", 3)) {
265 if (NULL == curp->pmdoc)
266 curp->pmdoc = mdoc_alloc(curp->roff, curp);
267 assert(curp->pmdoc);
268 curp->mdoc = curp->pmdoc;
269 return;
270 }
271
272 if (NULL == curp->pman)
273 curp->pman = man_alloc(curp->roff, curp);
274 assert(curp->pman);
275 curp->man = curp->pman;
276 }
277
278 /*
279 * Main parse routine for an opened file. This is called for each
280 * opened file and simply loops around the full input file, possibly
281 * nesting (i.e., with `so').
282 */
283 static void
284 mparse_buf_r(struct mparse *curp, struct buf blk, int start)
285 {
286 const struct tbl_span *span;
287 struct buf ln;
288 enum rofferr rr;
289 int i, of, rc;
290 int pos; /* byte number in the ln buffer */
291 int lnn; /* line number in the real file */
292 unsigned char c;
293
294 memset(&ln, 0, sizeof(struct buf));
295
296 lnn = curp->line;
297 pos = 0;
298
299 for (i = 0; i < (int)blk.sz; ) {
300 if (0 == pos && '\0' == blk.buf[i])
301 break;
302
303 if (start) {
304 curp->line = lnn;
305 curp->reparse_count = 0;
306 }
307
308 while (i < (int)blk.sz && (start || '\0' != blk.buf[i])) {
309
310 /*
311 * When finding an unescaped newline character,
312 * leave the character loop to process the line.
313 * Skip a preceding carriage return, if any.
314 */
315
316 if ('\r' == blk.buf[i] && i + 1 < (int)blk.sz &&
317 '\n' == blk.buf[i + 1])
318 ++i;
319 if ('\n' == blk.buf[i]) {
320 ++i;
321 ++lnn;
322 break;
323 }
324
325 /*
326 * Warn about bogus characters. If you're using
327 * non-ASCII encoding, you're screwing your
328 * readers. Since I'd rather this not happen,
329 * I'll be helpful and drop these characters so
330 * we don't display gibberish. Note to manual
331 * writers: use special characters.
332 */
333
334 c = (unsigned char) blk.buf[i];
335
336 if ( ! (isascii(c) &&
337 (isgraph(c) || isblank(c)))) {
338 mandoc_msg(MANDOCERR_BADCHAR, curp,
339 curp->line, pos, "ignoring byte");
340 i++;
341 continue;
342 }
343
344 /* Trailing backslash = a plain char. */
345
346 if ('\\' != blk.buf[i] || i + 1 == (int)blk.sz) {
347 if (pos >= (int)ln.sz)
348 resize_buf(&ln, 256);
349 ln.buf[pos++] = blk.buf[i++];
350 continue;
351 }
352
353 /*
354 * Found escape and at least one other character.
355 * When it's a newline character, skip it.
356 * When there is a carriage return in between,
357 * skip that one as well.
358 */
359
360 if ('\r' == blk.buf[i + 1] && i + 2 < (int)blk.sz &&
361 '\n' == blk.buf[i + 2])
362 ++i;
363 if ('\n' == blk.buf[i + 1]) {
364 i += 2;
365 ++lnn;
366 continue;
367 }
368
369 if ('"' == blk.buf[i + 1] || '#' == blk.buf[i + 1]) {
370 i += 2;
371 /* Comment, skip to end of line */
372 for (; i < (int)blk.sz; ++i) {
373 if ('\n' == blk.buf[i]) {
374 ++i;
375 ++lnn;
376 break;
377 }
378 }
379
380 /* Backout trailing whitespaces */
381 for (; pos > 0; --pos) {
382 if (ln.buf[pos - 1] != ' ')
383 break;
384 if (pos > 2 && ln.buf[pos - 2] == '\\')
385 break;
386 }
387 break;
388 }
389
390 /* Some other escape sequence, copy & cont. */
391
392 if (pos + 1 >= (int)ln.sz)
393 resize_buf(&ln, 256);
394
395 ln.buf[pos++] = blk.buf[i++];
396 ln.buf[pos++] = blk.buf[i++];
397 }
398
399 if (pos >= (int)ln.sz)
400 resize_buf(&ln, 256);
401
402 ln.buf[pos] = '\0';
403
404 /*
405 * A significant amount of complexity is contained by
406 * the roff preprocessor. It's line-oriented but can be
407 * expressed on one line, so we need at times to
408 * readjust our starting point and re-run it. The roff
409 * preprocessor can also readjust the buffers with new
410 * data, so we pass them in wholesale.
411 */
412
413 of = 0;
414
415 rerun:
416 rr = roff_parseln
417 (curp->roff, curp->line,
418 &ln.buf, &ln.sz, of, &of);
419
420 switch (rr) {
421 case (ROFF_REPARSE):
422 if (REPARSE_LIMIT >= ++curp->reparse_count)
423 mparse_buf_r(curp, ln, 0);
424 else
425 mandoc_msg(MANDOCERR_ROFFLOOP, curp,
426 curp->line, pos, NULL);
427 pos = 0;
428 continue;
429 case (ROFF_APPEND):
430 pos = (int)strlen(ln.buf);
431 continue;
432 case (ROFF_RERUN):
433 goto rerun;
434 case (ROFF_IGN):
435 pos = 0;
436 continue;
437 case (ROFF_ERR):
438 assert(MANDOCLEVEL_FATAL <= curp->file_status);
439 break;
440 case (ROFF_SO):
441 mparse_readfd_r(curp, -1, ln.buf + of, 1);
442 if (MANDOCLEVEL_FATAL <= curp->file_status)
443 break;
444 pos = 0;
445 continue;
446 default:
447 break;
448 }
449
450 /*
451 * If we encounter errors in the recursive parse, make
452 * sure we don't continue parsing.
453 */
454
455 if (MANDOCLEVEL_FATAL <= curp->file_status)
456 break;
457
458 /*
459 * If input parsers have not been allocated, do so now.
460 * We keep these instanced between parsers, but set them
461 * locally per parse routine since we can use different
462 * parsers with each one.
463 */
464
465 if ( ! (curp->man || curp->mdoc))
466 pset(ln.buf + of, pos - of, curp);
467
468 /*
469 * Lastly, push down into the parsers themselves. One
470 * of these will have already been set in the pset()
471 * routine.
472 * If libroff returns ROFF_TBL, then add it to the
473 * currently open parse. Since we only get here if
474 * there does exist data (see tbl_data.c), we're
475 * guaranteed that something's been allocated.
476 * Do the same for ROFF_EQN.
477 */
478
479 rc = -1;
480
481 if (ROFF_TBL == rr)
482 while (NULL != (span = roff_span(curp->roff))) {
483 rc = curp->man ?
484 man_addspan(curp->man, span) :
485 mdoc_addspan(curp->mdoc, span);
486 if (0 == rc)
487 break;
488 }
489 else if (ROFF_EQN == rr)
490 rc = curp->mdoc ?
491 mdoc_addeqn(curp->mdoc,
492 roff_eqn(curp->roff)) :
493 man_addeqn(curp->man,
494 roff_eqn(curp->roff));
495 else if (curp->man || curp->mdoc)
496 rc = curp->man ?
497 man_parseln(curp->man,
498 curp->line, ln.buf, of) :
499 mdoc_parseln(curp->mdoc,
500 curp->line, ln.buf, of);
501
502 if (0 == rc) {
503 assert(MANDOCLEVEL_FATAL <= curp->file_status);
504 break;
505 }
506
507 /* Temporary buffers typically are not full. */
508
509 if (0 == start && '\0' == blk.buf[i])
510 break;
511
512 /* Start the next input line. */
513
514 pos = 0;
515 }
516
517 free(ln.buf);
518 }
519
520 static void
521 pdesc(struct mparse *curp, const char *file, int fd)
522 {
523 struct buf blk;
524 int with_mmap;
525
526 /*
527 * Run for each opened file; may be called more than once for
528 * each full parse sequence if the opened file is nested (i.e.,
529 * from `so'). Simply sucks in the whole file and moves into
530 * the parse phase for the file.
531 */
532
533 if ( ! read_whole_file(file, fd, &blk, &with_mmap)) {
534 curp->file_status = MANDOCLEVEL_SYSERR;
535 return;
536 }
537
538 /* Line number is per-file. */
539
540 curp->line = 1;
541
542 mparse_buf_r(curp, blk, 1);
543
544 #ifdef HAVE_MMAP
545 if (with_mmap)
546 munmap(blk.buf, blk.sz);
547 else
548 #endif
549 free(blk.buf);
550 }
551
552 static int
553 read_whole_file(const char *file, int fd, struct buf *fb, int *with_mmap)
554 {
555 size_t off;
556 ssize_t ssz;
557
558 #ifdef HAVE_MMAP
559 struct stat st;
560 if (-1 == fstat(fd, &st)) {
561 perror(file);
562 return(0);
563 }
564
565 /*
566 * If we're a regular file, try just reading in the whole entry
567 * via mmap(). This is faster than reading it into blocks, and
568 * since each file is only a few bytes to begin with, I'm not
569 * concerned that this is going to tank any machines.
570 */
571
572 if (S_ISREG(st.st_mode)) {
573 if (st.st_size >= (1U << 31)) {
574 fprintf(stderr, "%s: input too large\n", file);
575 return(0);
576 }
577 *with_mmap = 1;
578 fb->sz = (size_t)st.st_size;
579 fb->buf = mmap(NULL, fb->sz, PROT_READ,
580 MAP_FILE|MAP_SHARED, fd, 0);
581 if (fb->buf != MAP_FAILED)
582 return(1);
583 }
584 #endif
585
586 /*
587 * If this isn't a regular file (like, say, stdin), then we must
588 * go the old way and just read things in bit by bit.
589 */
590
591 *with_mmap = 0;
592 off = 0;
593 fb->sz = 0;
594 fb->buf = NULL;
595 for (;;) {
596 if (off == fb->sz) {
597 if (fb->sz == (1U << 31)) {
598 fprintf(stderr, "%s: input too large\n", file);
599 break;
600 }
601 resize_buf(fb, 65536);
602 }
603 ssz = read(fd, fb->buf + (int)off, fb->sz - off);
604 if (ssz == 0) {
605 fb->sz = off;
606 return(1);
607 }
608 if (ssz == -1) {
609 perror(file);
610 break;
611 }
612 off += (size_t)ssz;
613 }
614
615 free(fb->buf);
616 fb->buf = NULL;
617 return(0);
618 }
619
620 static void
621 mparse_end(struct mparse *curp)
622 {
623
624 if (MANDOCLEVEL_FATAL <= curp->file_status)
625 return;
626
627 if (curp->mdoc && ! mdoc_endparse(curp->mdoc)) {
628 assert(MANDOCLEVEL_FATAL <= curp->file_status);
629 return;
630 }
631
632 if (curp->man && ! man_endparse(curp->man)) {
633 assert(MANDOCLEVEL_FATAL <= curp->file_status);
634 return;
635 }
636
637 if ( ! (curp->man || curp->mdoc)) {
638 mandoc_msg(MANDOCERR_NOTMANUAL, curp, 1, 0, NULL);
639 curp->file_status = MANDOCLEVEL_FATAL;
640 return;
641 }
642
643 roff_endparse(curp->roff);
644 }
645
646 static void
647 mparse_readfd_r(struct mparse *curp, int fd, const char *file, int re)
648 {
649 const char *svfile;
650
651 if (-1 == fd)
652 if (-1 == (fd = open(file, O_RDONLY, 0))) {
653 perror(file);
654 curp->file_status = MANDOCLEVEL_SYSERR;
655 return;
656 }
657
658 svfile = curp->file;
659 curp->file = file;
660
661 pdesc(curp, file, fd);
662
663 if (0 == re && MANDOCLEVEL_FATAL > curp->file_status)
664 mparse_end(curp);
665
666 if (STDIN_FILENO != fd && -1 == close(fd))
667 perror(file);
668
669 curp->file = svfile;
670 }
671
672 enum mandoclevel
673 mparse_readfd(struct mparse *curp, int fd, const char *file)
674 {
675
676 mparse_readfd_r(curp, fd, file, 0);
677 return(curp->file_status);
678 }
679
680 struct mparse *
681 mparse_alloc(enum mparset inttype, enum mandoclevel wlevel, mandocmsg mmsg, void *arg)
682 {
683 struct mparse *curp;
684
685 assert(wlevel <= MANDOCLEVEL_FATAL);
686
687 curp = mandoc_calloc(1, sizeof(struct mparse));
688
689 curp->wlevel = wlevel;
690 curp->mmsg = mmsg;
691 curp->arg = arg;
692 curp->inttype = inttype;
693
694 curp->roff = roff_alloc(curp);
695 return(curp);
696 }
697
698 void
699 mparse_reset(struct mparse *curp)
700 {
701
702 roff_reset(curp->roff);
703
704 if (curp->mdoc)
705 mdoc_reset(curp->mdoc);
706 if (curp->man)
707 man_reset(curp->man);
708
709 curp->file_status = MANDOCLEVEL_OK;
710 curp->mdoc = NULL;
711 curp->man = NULL;
712 }
713
714 void
715 mparse_free(struct mparse *curp)
716 {
717
718 if (curp->pmdoc)
719 mdoc_free(curp->pmdoc);
720 if (curp->pman)
721 man_free(curp->pman);
722 if (curp->roff)
723 roff_free(curp->roff);
724
725 free(curp);
726 }
727
728 void
729 mparse_result(struct mparse *curp, struct mdoc **mdoc, struct man **man)
730 {
731
732 if (mdoc)
733 *mdoc = curp->mdoc;
734 if (man)
735 *man = curp->man;
736 }
737
738 void
739 mandoc_vmsg(enum mandocerr t, struct mparse *m,
740 int ln, int pos, const char *fmt, ...)
741 {
742 char buf[256];
743 va_list ap;
744
745 va_start(ap, fmt);
746 vsnprintf(buf, sizeof(buf) - 1, fmt, ap);
747 va_end(ap);
748
749 mandoc_msg(t, m, ln, pos, buf);
750 }
751
752 void
753 mandoc_msg(enum mandocerr er, struct mparse *m,
754 int ln, int col, const char *msg)
755 {
756 enum mandoclevel level;
757
758 level = MANDOCLEVEL_FATAL;
759 while (er < mandoclimits[level])
760 level--;
761
762 if (level < m->wlevel)
763 return;
764
765 if (m->mmsg)
766 (*m->mmsg)(er, level, m->file, ln, col, msg);
767
768 if (m->file_status < level)
769 m->file_status = level;
770 }
771
772 const char *
773 mparse_strerror(enum mandocerr er)
774 {
775
776 return(mandocerrs[er]);
777 }
778
779 const char *
780 mparse_strlevel(enum mandoclevel lvl)
781 {
782 return(mandoclevels[lvl]);
783 }