]> git.cameronkatri.com Git - mandoc.git/blob - read.c
Warn if equation `define' key is quoted (groff-ism).
[mandoc.git] / read.c
1 /* $Id: read.c,v 1.16 2011/07/17 12:52:54 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 struct regset regs; /* roff registers */
63 int reparse_count; /* finite interp. stack */
64 mandocmsg mmsg; /* warning/error message handler */
65 void *arg; /* argument to mmsg */
66 const char *file;
67 };
68
69 static void resize_buf(struct buf *, size_t);
70 static void mparse_buf_r(struct mparse *, struct buf, int);
71 static void mparse_readfd_r(struct mparse *, int, const char *, int);
72 static void pset(const char *, int, struct mparse *);
73 static void pdesc(struct mparse *, const char *, int);
74 static int read_whole_file(const char *, int, struct buf *, int *);
75 static void mparse_end(struct mparse *);
76
77 static const enum mandocerr mandoclimits[MANDOCLEVEL_MAX] = {
78 MANDOCERR_OK,
79 MANDOCERR_WARNING,
80 MANDOCERR_WARNING,
81 MANDOCERR_ERROR,
82 MANDOCERR_FATAL,
83 MANDOCERR_MAX,
84 MANDOCERR_MAX
85 };
86
87 static const char * const mandocerrs[MANDOCERR_MAX] = {
88 "ok",
89
90 "generic warning",
91
92 /* related to the prologue */
93 "no title in document",
94 "document title should be all caps",
95 "unknown manual section",
96 "date missing, using today's date",
97 "cannot parse date, using it verbatim",
98 "prologue macros out of order",
99 "duplicate prologue macro",
100 "macro not allowed in prologue",
101 "macro not allowed in body",
102
103 /* related to document structure */
104 ".so is fragile, better use ln(1)",
105 "NAME section must come first",
106 "bad NAME section contents",
107 "manual name not yet set",
108 "sections out of conventional order",
109 "duplicate section name",
110 "section not in conventional manual section",
111
112 /* related to macros and nesting */
113 "skipping obsolete macro",
114 "skipping paragraph macro",
115 "skipping no-space macro",
116 "blocks badly nested",
117 "child violates parent syntax",
118 "nested displays are not portable",
119 "already in literal mode",
120 "line scope broken",
121
122 /* related to missing macro arguments */
123 "skipping empty macro",
124 "argument count wrong",
125 "missing display type",
126 "list type must come first",
127 "tag lists require a width argument",
128 "missing font type",
129 "skipping end of block that is not open",
130
131 /* related to bad macro arguments */
132 "skipping argument",
133 "duplicate argument",
134 "duplicate display type",
135 "duplicate list type",
136 "unknown AT&T UNIX version",
137 "bad Boolean value",
138 "unknown font",
139 "unknown standard specifier",
140 "bad width argument",
141
142 /* related to plain text */
143 "blank line in non-literal context",
144 "tab in non-literal context",
145 "end of line whitespace",
146 "bad comment style",
147 "bad escape sequence",
148 "unterminated quoted string",
149
150 /* related to equations */
151 "unexpected literal in equation",
152
153 "generic error",
154
155 /* related to tables */
156 "bad table syntax",
157 "bad table option",
158 "bad table layout",
159 "no table layout cells specified",
160 "no table data cells specified",
161 "ignore data in cell",
162 "data block still open",
163 "ignoring extra data cells",
164
165 "input stack limit exceeded, infinite loop?",
166 "skipping bad character",
167 "escaped character not allowed in a name",
168 "skipping text before the first section header",
169 "skipping unknown macro",
170 "NOT IMPLEMENTED, please use groff: skipping request",
171 "argument count wrong",
172 "skipping end of block that is not open",
173 "missing end of block",
174 "scope open on exit",
175 "uname(3) system call failed",
176 "macro requires line argument(s)",
177 "macro requires body argument(s)",
178 "macro requires argument(s)",
179 "missing list type",
180 "line argument(s) will be lost",
181 "body argument(s) will be lost",
182
183 "generic fatal error",
184
185 "not a manual",
186 "column syntax is inconsistent",
187 "NOT IMPLEMENTED: .Bd -file",
188 "line scope broken, syntax violated",
189 "argument count wrong, violates syntax",
190 "child violates parent syntax",
191 "argument count wrong, violates syntax",
192 "NOT IMPLEMENTED: .so with absolute path or \"..\"",
193 "no document body",
194 "no document prologue",
195 "static buffer exhausted",
196 };
197
198 static const char * const mandoclevels[MANDOCLEVEL_MAX] = {
199 "SUCCESS",
200 "RESERVED",
201 "WARNING",
202 "ERROR",
203 "FATAL",
204 "BADARG",
205 "SYSERR"
206 };
207
208 static void
209 resize_buf(struct buf *buf, size_t initial)
210 {
211
212 buf->sz = buf->sz > initial/2 ? 2 * buf->sz : initial;
213 buf->buf = mandoc_realloc(buf->buf, buf->sz);
214 }
215
216 static void
217 pset(const char *buf, int pos, struct mparse *curp)
218 {
219 int i;
220
221 /*
222 * Try to intuit which kind of manual parser should be used. If
223 * passed in by command-line (-man, -mdoc), then use that
224 * explicitly. If passed as -mandoc, then try to guess from the
225 * line: either skip dot-lines, use -mdoc when finding `.Dt', or
226 * default to -man, which is more lenient.
227 *
228 * Separate out pmdoc/pman from mdoc/man: the first persists
229 * through all parsers, while the latter is used per-parse.
230 */
231
232 if ('.' == buf[0] || '\'' == buf[0]) {
233 for (i = 1; buf[i]; i++)
234 if (' ' != buf[i] && '\t' != buf[i])
235 break;
236 if ('\0' == buf[i])
237 return;
238 }
239
240 switch (curp->inttype) {
241 case (MPARSE_MDOC):
242 if (NULL == curp->pmdoc)
243 curp->pmdoc = mdoc_alloc(&curp->regs, curp);
244 assert(curp->pmdoc);
245 curp->mdoc = curp->pmdoc;
246 return;
247 case (MPARSE_MAN):
248 if (NULL == curp->pman)
249 curp->pman = man_alloc(&curp->regs, curp);
250 assert(curp->pman);
251 curp->man = curp->pman;
252 return;
253 default:
254 break;
255 }
256
257 if (pos >= 3 && 0 == memcmp(buf, ".Dd", 3)) {
258 if (NULL == curp->pmdoc)
259 curp->pmdoc = mdoc_alloc(&curp->regs, curp);
260 assert(curp->pmdoc);
261 curp->mdoc = curp->pmdoc;
262 return;
263 }
264
265 if (NULL == curp->pman)
266 curp->pman = man_alloc(&curp->regs, curp);
267 assert(curp->pman);
268 curp->man = curp->pman;
269 }
270
271 /*
272 * Main parse routine for an opened file. This is called for each
273 * opened file and simply loops around the full input file, possibly
274 * nesting (i.e., with `so').
275 */
276 static void
277 mparse_buf_r(struct mparse *curp, struct buf blk, int start)
278 {
279 const struct tbl_span *span;
280 struct buf ln;
281 enum rofferr rr;
282 int i, of, rc;
283 int pos; /* byte number in the ln buffer */
284 int lnn; /* line number in the real file */
285 unsigned char c;
286
287 memset(&ln, 0, sizeof(struct buf));
288
289 lnn = curp->line;
290 pos = 0;
291
292 for (i = 0; i < (int)blk.sz; ) {
293 if (0 == pos && '\0' == blk.buf[i])
294 break;
295
296 if (start) {
297 curp->line = lnn;
298 curp->reparse_count = 0;
299 }
300
301 while (i < (int)blk.sz && (start || '\0' != blk.buf[i])) {
302
303 /*
304 * When finding an unescaped newline character,
305 * leave the character loop to process the line.
306 * Skip a preceding carriage return, if any.
307 */
308
309 if ('\r' == blk.buf[i] && i + 1 < (int)blk.sz &&
310 '\n' == blk.buf[i + 1])
311 ++i;
312 if ('\n' == blk.buf[i]) {
313 ++i;
314 ++lnn;
315 break;
316 }
317
318 /*
319 * Warn about bogus characters. If you're using
320 * non-ASCII encoding, you're screwing your
321 * readers. Since I'd rather this not happen,
322 * I'll be helpful and drop these characters so
323 * we don't display gibberish. Note to manual
324 * writers: use special characters.
325 */
326
327 c = (unsigned char) blk.buf[i];
328
329 if ( ! (isascii(c) &&
330 (isgraph(c) || isblank(c)))) {
331 mandoc_msg(MANDOCERR_BADCHAR, curp,
332 curp->line, pos, "ignoring byte");
333 i++;
334 continue;
335 }
336
337 /* Trailing backslash = a plain char. */
338
339 if ('\\' != blk.buf[i] || i + 1 == (int)blk.sz) {
340 if (pos >= (int)ln.sz)
341 resize_buf(&ln, 256);
342 ln.buf[pos++] = blk.buf[i++];
343 continue;
344 }
345
346 /*
347 * Found escape and at least one other character.
348 * When it's a newline character, skip it.
349 * When there is a carriage return in between,
350 * skip that one as well.
351 */
352
353 if ('\r' == blk.buf[i + 1] && i + 2 < (int)blk.sz &&
354 '\n' == blk.buf[i + 2])
355 ++i;
356 if ('\n' == blk.buf[i + 1]) {
357 i += 2;
358 ++lnn;
359 continue;
360 }
361
362 if ('"' == blk.buf[i + 1] || '#' == blk.buf[i + 1]) {
363 i += 2;
364 /* Comment, skip to end of line */
365 for (; i < (int)blk.sz; ++i) {
366 if ('\n' == blk.buf[i]) {
367 ++i;
368 ++lnn;
369 break;
370 }
371 }
372
373 /* Backout trailing whitespaces */
374 for (; pos > 0; --pos) {
375 if (ln.buf[pos - 1] != ' ')
376 break;
377 if (pos > 2 && ln.buf[pos - 2] == '\\')
378 break;
379 }
380 break;
381 }
382
383 /* Some other escape sequence, copy & cont. */
384
385 if (pos + 1 >= (int)ln.sz)
386 resize_buf(&ln, 256);
387
388 ln.buf[pos++] = blk.buf[i++];
389 ln.buf[pos++] = blk.buf[i++];
390 }
391
392 if (pos >= (int)ln.sz)
393 resize_buf(&ln, 256);
394
395 ln.buf[pos] = '\0';
396
397 /*
398 * A significant amount of complexity is contained by
399 * the roff preprocessor. It's line-oriented but can be
400 * expressed on one line, so we need at times to
401 * readjust our starting point and re-run it. The roff
402 * preprocessor can also readjust the buffers with new
403 * data, so we pass them in wholesale.
404 */
405
406 of = 0;
407
408 rerun:
409 rr = roff_parseln
410 (curp->roff, curp->line,
411 &ln.buf, &ln.sz, of, &of);
412
413 switch (rr) {
414 case (ROFF_REPARSE):
415 if (REPARSE_LIMIT >= ++curp->reparse_count)
416 mparse_buf_r(curp, ln, 0);
417 else
418 mandoc_msg(MANDOCERR_ROFFLOOP, curp,
419 curp->line, pos, NULL);
420 pos = 0;
421 continue;
422 case (ROFF_APPEND):
423 pos = (int)strlen(ln.buf);
424 continue;
425 case (ROFF_RERUN):
426 goto rerun;
427 case (ROFF_IGN):
428 pos = 0;
429 continue;
430 case (ROFF_ERR):
431 assert(MANDOCLEVEL_FATAL <= curp->file_status);
432 break;
433 case (ROFF_SO):
434 mparse_readfd_r(curp, -1, ln.buf + of, 1);
435 if (MANDOCLEVEL_FATAL <= curp->file_status)
436 break;
437 pos = 0;
438 continue;
439 default:
440 break;
441 }
442
443 /*
444 * If we encounter errors in the recursive parse, make
445 * sure we don't continue parsing.
446 */
447
448 if (MANDOCLEVEL_FATAL <= curp->file_status)
449 break;
450
451 /*
452 * If input parsers have not been allocated, do so now.
453 * We keep these instanced between parsers, but set them
454 * locally per parse routine since we can use different
455 * parsers with each one.
456 */
457
458 if ( ! (curp->man || curp->mdoc))
459 pset(ln.buf + of, pos - of, curp);
460
461 /*
462 * Lastly, push down into the parsers themselves. One
463 * of these will have already been set in the pset()
464 * routine.
465 * If libroff returns ROFF_TBL, then add it to the
466 * currently open parse. Since we only get here if
467 * there does exist data (see tbl_data.c), we're
468 * guaranteed that something's been allocated.
469 * Do the same for ROFF_EQN.
470 */
471
472 rc = -1;
473
474 if (ROFF_TBL == rr)
475 while (NULL != (span = roff_span(curp->roff))) {
476 rc = curp->man ?
477 man_addspan(curp->man, span) :
478 mdoc_addspan(curp->mdoc, span);
479 if (0 == rc)
480 break;
481 }
482 else if (ROFF_EQN == rr)
483 rc = curp->mdoc ?
484 mdoc_addeqn(curp->mdoc,
485 roff_eqn(curp->roff)) :
486 man_addeqn(curp->man,
487 roff_eqn(curp->roff));
488 else if (curp->man || curp->mdoc)
489 rc = curp->man ?
490 man_parseln(curp->man,
491 curp->line, ln.buf, of) :
492 mdoc_parseln(curp->mdoc,
493 curp->line, ln.buf, of);
494
495 if (0 == rc) {
496 assert(MANDOCLEVEL_FATAL <= curp->file_status);
497 break;
498 }
499
500 /* Temporary buffers typically are not full. */
501
502 if (0 == start && '\0' == blk.buf[i])
503 break;
504
505 /* Start the next input line. */
506
507 pos = 0;
508 }
509
510 free(ln.buf);
511 }
512
513 static void
514 pdesc(struct mparse *curp, const char *file, int fd)
515 {
516 struct buf blk;
517 int with_mmap;
518
519 /*
520 * Run for each opened file; may be called more than once for
521 * each full parse sequence if the opened file is nested (i.e.,
522 * from `so'). Simply sucks in the whole file and moves into
523 * the parse phase for the file.
524 */
525
526 if ( ! read_whole_file(file, fd, &blk, &with_mmap)) {
527 curp->file_status = MANDOCLEVEL_SYSERR;
528 return;
529 }
530
531 /* Line number is per-file. */
532
533 curp->line = 1;
534
535 mparse_buf_r(curp, blk, 1);
536
537 #ifdef HAVE_MMAP
538 if (with_mmap)
539 munmap(blk.buf, blk.sz);
540 else
541 #endif
542 free(blk.buf);
543 }
544
545 static int
546 read_whole_file(const char *file, int fd, struct buf *fb, int *with_mmap)
547 {
548 size_t off;
549 ssize_t ssz;
550
551 #ifdef HAVE_MMAP
552 struct stat st;
553 if (-1 == fstat(fd, &st)) {
554 perror(file);
555 return(0);
556 }
557
558 /*
559 * If we're a regular file, try just reading in the whole entry
560 * via mmap(). This is faster than reading it into blocks, and
561 * since each file is only a few bytes to begin with, I'm not
562 * concerned that this is going to tank any machines.
563 */
564
565 if (S_ISREG(st.st_mode)) {
566 if (st.st_size >= (1U << 31)) {
567 fprintf(stderr, "%s: input too large\n", file);
568 return(0);
569 }
570 *with_mmap = 1;
571 fb->sz = (size_t)st.st_size;
572 fb->buf = mmap(NULL, fb->sz, PROT_READ,
573 MAP_FILE|MAP_SHARED, fd, 0);
574 if (fb->buf != MAP_FAILED)
575 return(1);
576 }
577 #endif
578
579 /*
580 * If this isn't a regular file (like, say, stdin), then we must
581 * go the old way and just read things in bit by bit.
582 */
583
584 *with_mmap = 0;
585 off = 0;
586 fb->sz = 0;
587 fb->buf = NULL;
588 for (;;) {
589 if (off == fb->sz) {
590 if (fb->sz == (1U << 31)) {
591 fprintf(stderr, "%s: input too large\n", file);
592 break;
593 }
594 resize_buf(fb, 65536);
595 }
596 ssz = read(fd, fb->buf + (int)off, fb->sz - off);
597 if (ssz == 0) {
598 fb->sz = off;
599 return(1);
600 }
601 if (ssz == -1) {
602 perror(file);
603 break;
604 }
605 off += (size_t)ssz;
606 }
607
608 free(fb->buf);
609 fb->buf = NULL;
610 return(0);
611 }
612
613 static void
614 mparse_end(struct mparse *curp)
615 {
616
617 if (MANDOCLEVEL_FATAL <= curp->file_status)
618 return;
619
620 if (curp->mdoc && ! mdoc_endparse(curp->mdoc)) {
621 assert(MANDOCLEVEL_FATAL <= curp->file_status);
622 return;
623 }
624
625 if (curp->man && ! man_endparse(curp->man)) {
626 assert(MANDOCLEVEL_FATAL <= curp->file_status);
627 return;
628 }
629
630 if ( ! (curp->man || curp->mdoc)) {
631 mandoc_msg(MANDOCERR_NOTMANUAL, curp, 1, 0, NULL);
632 curp->file_status = MANDOCLEVEL_FATAL;
633 return;
634 }
635
636 roff_endparse(curp->roff);
637 }
638
639 static void
640 mparse_readfd_r(struct mparse *curp, int fd, const char *file, int re)
641 {
642 const char *svfile;
643
644 if (-1 == fd)
645 if (-1 == (fd = open(file, O_RDONLY, 0))) {
646 perror(file);
647 curp->file_status = MANDOCLEVEL_SYSERR;
648 return;
649 }
650
651 svfile = curp->file;
652 curp->file = file;
653
654 pdesc(curp, file, fd);
655
656 if (0 == re && MANDOCLEVEL_FATAL > curp->file_status)
657 mparse_end(curp);
658
659 if (STDIN_FILENO != fd && -1 == close(fd))
660 perror(file);
661
662 curp->file = svfile;
663 }
664
665 enum mandoclevel
666 mparse_readfd(struct mparse *curp, int fd, const char *file)
667 {
668
669 mparse_readfd_r(curp, fd, file, 0);
670 return(curp->file_status);
671 }
672
673 struct mparse *
674 mparse_alloc(enum mparset inttype, enum mandoclevel wlevel, mandocmsg mmsg, void *arg)
675 {
676 struct mparse *curp;
677
678 assert(wlevel <= MANDOCLEVEL_FATAL);
679
680 curp = mandoc_calloc(1, sizeof(struct mparse));
681
682 curp->wlevel = wlevel;
683 curp->mmsg = mmsg;
684 curp->arg = arg;
685 curp->inttype = inttype;
686
687 curp->roff = roff_alloc(&curp->regs, curp);
688 return(curp);
689 }
690
691 void
692 mparse_reset(struct mparse *curp)
693 {
694
695 memset(&curp->regs, 0, sizeof(struct regset));
696
697 roff_reset(curp->roff);
698
699 if (curp->mdoc)
700 mdoc_reset(curp->mdoc);
701 if (curp->man)
702 man_reset(curp->man);
703
704 curp->file_status = MANDOCLEVEL_OK;
705 curp->mdoc = NULL;
706 curp->man = NULL;
707 }
708
709 void
710 mparse_free(struct mparse *curp)
711 {
712
713 if (curp->pmdoc)
714 mdoc_free(curp->pmdoc);
715 if (curp->pman)
716 man_free(curp->pman);
717 if (curp->roff)
718 roff_free(curp->roff);
719
720 free(curp);
721 }
722
723 void
724 mparse_result(struct mparse *curp, struct mdoc **mdoc, struct man **man)
725 {
726
727 if (mdoc)
728 *mdoc = curp->mdoc;
729 if (man)
730 *man = curp->man;
731 }
732
733 void
734 mandoc_vmsg(enum mandocerr t, struct mparse *m,
735 int ln, int pos, const char *fmt, ...)
736 {
737 char buf[256];
738 va_list ap;
739
740 va_start(ap, fmt);
741 vsnprintf(buf, sizeof(buf) - 1, fmt, ap);
742 va_end(ap);
743
744 mandoc_msg(t, m, ln, pos, buf);
745 }
746
747 void
748 mandoc_msg(enum mandocerr er, struct mparse *m,
749 int ln, int col, const char *msg)
750 {
751 enum mandoclevel level;
752
753 level = MANDOCLEVEL_FATAL;
754 while (er < mandoclimits[level])
755 level--;
756
757 if (level < m->wlevel)
758 return;
759
760 if (m->mmsg)
761 (*m->mmsg)(er, level, m->file, ln, col, msg);
762
763 if (m->file_status < level)
764 m->file_status = level;
765 }
766
767 const char *
768 mparse_strerror(enum mandocerr er)
769 {
770
771 return(mandocerrs[er]);
772 }
773
774 const char *
775 mparse_strlevel(enum mandoclevel lvl)
776 {
777 return(mandoclevels[lvl]);
778 }