release.sh changes & fixes
[minix3.git] / external / bsd / mdocml / dist / read.c
blobd68dc02cd0d45a0c5a3362a1fe20a86c61999c67
1 /* $Vendor-Id: read.c,v 1.28 2012/02/16 20:51:31 joerg Exp $ */
2 /*
3 * Copyright (c) 2008, 2009, 2010, 2011 Kristaps Dzonsons <kristaps@bsd.lv>
4 * Copyright (c) 2010, 2011 Ingo Schwarze <schwarze@openbsd.org>
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.
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.
18 #ifdef HAVE_CONFIG_H
19 #include "config.h"
20 #endif
22 #ifdef HAVE_MMAP
23 # include <sys/stat.h>
24 # include <sys/mman.h>
25 #endif
27 #include <assert.h>
28 #include <ctype.h>
29 #include <fcntl.h>
30 #include <stdarg.h>
31 #include <stdint.h>
32 #include <stdio.h>
33 #include <stdlib.h>
34 #include <string.h>
35 #include <unistd.h>
37 #include "mandoc.h"
38 #include "libmandoc.h"
39 #include "mdoc.h"
40 #include "man.h"
41 #include "main.h"
43 #ifndef MAP_FILE
44 #define MAP_FILE 0
45 #endif
47 #define REPARSE_LIMIT 1000
49 struct buf {
50 char *buf; /* binary input buffer */
51 size_t sz; /* size of binary buffer */
54 struct mparse {
55 enum mandoclevel file_status; /* status of current parse */
56 enum mandoclevel wlevel; /* ignore messages below this */
57 int line; /* line number in the file */
58 enum mparset inttype; /* which parser to use */
59 struct man *pman; /* persistent man parser */
60 struct mdoc *pmdoc; /* persistent mdoc parser */
61 struct man *man; /* man parser */
62 struct mdoc *mdoc; /* mdoc parser */
63 struct roff *roff; /* roff parser (!NULL) */
64 int reparse_count; /* finite interp. stack */
65 mandocmsg mmsg; /* warning/error message handler */
66 void *arg; /* argument to mmsg */
67 const char *file;
68 struct buf *secondary;
71 static void resize_buf(struct buf *, size_t);
72 static void mparse_buf_r(struct mparse *, struct buf, int);
73 static void mparse_readfd_r(struct mparse *, int, const char *, int);
74 static void pset(const char *, int, struct mparse *);
75 static int read_whole_file(const char *, int, struct buf *, int *);
76 static void mparse_end(struct mparse *);
78 static const enum mandocerr mandoclimits[MANDOCLEVEL_MAX] = {
79 MANDOCERR_OK,
80 MANDOCERR_WARNING,
81 MANDOCERR_WARNING,
82 MANDOCERR_ERROR,
83 MANDOCERR_FATAL,
84 MANDOCERR_MAX,
85 MANDOCERR_MAX
88 static const char * const mandocerrs[MANDOCERR_MAX] = {
89 "ok",
91 "generic warning",
93 /* related to the prologue */
94 "no title in document",
95 "document title should be all caps",
96 "unknown manual section",
97 "date missing, using today's date",
98 "cannot parse date, using it verbatim",
99 "prologue macros out of order",
100 "duplicate prologue macro",
101 "macro not allowed in prologue",
102 "macro not allowed in body",
104 /* related to document structure */
105 ".so is fragile, better use ln(1)",
106 "NAME section must come first",
107 "bad NAME section contents",
108 "manual name not yet set",
109 "sections out of conventional order",
110 "duplicate section name",
111 "section not in conventional manual section",
113 /* related to macros and nesting */
114 "skipping obsolete macro",
115 "skipping paragraph macro",
116 "skipping no-space macro",
117 "blocks badly nested",
118 "child violates parent syntax",
119 "nested displays are not portable",
120 "already in literal mode",
121 "line scope broken",
123 /* related to missing macro arguments */
124 "skipping empty macro",
125 "argument count wrong",
126 "missing display type",
127 "list type must come first",
128 "tag lists require a width argument",
129 "missing font type",
130 "skipping end of block that is not open",
132 /* related to bad macro arguments */
133 "skipping argument",
134 "duplicate argument",
135 "duplicate display type",
136 "duplicate list type",
137 "unknown AT&T UNIX version",
138 "bad Boolean value",
139 "unknown font",
140 "unknown standard specifier",
141 "bad width argument",
143 /* related to plain text */
144 "blank line in non-literal context",
145 "tab in non-literal context",
146 "end of line whitespace",
147 "bad comment style",
148 "bad escape sequence",
149 "unterminated quoted string",
151 /* related to equations */
152 "unexpected literal in equation",
154 "generic error",
156 /* related to equations */
157 "unexpected equation scope closure",
158 "equation scope open on exit",
159 "overlapping equation scopes",
160 "unexpected end of equation",
161 "equation syntax error",
163 /* related to tables */
164 "bad table syntax",
165 "bad table option",
166 "bad table layout",
167 "no table layout cells specified",
168 "no table data cells specified",
169 "ignore data in cell",
170 "data block still open",
171 "ignoring extra data cells",
173 "input stack limit exceeded, infinite loop?",
174 "skipping bad character",
175 "escaped character not allowed in a name",
176 "skipping text before the first section header",
177 "skipping unknown macro",
178 "NOT IMPLEMENTED, please use groff: skipping request",
179 "argument count wrong",
180 "skipping end of block that is not open",
181 "missing end of block",
182 "scope open on exit",
183 "uname(3) system call failed",
184 "macro requires line argument(s)",
185 "macro requires body argument(s)",
186 "macro requires argument(s)",
187 "missing list type",
188 "line argument(s) will be lost",
189 "body argument(s) will be lost",
191 "generic fatal error",
193 "not a manual",
194 "column syntax is inconsistent",
195 "NOT IMPLEMENTED: .Bd -file",
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",
205 static const char * const mandoclevels[MANDOCLEVEL_MAX] = {
206 "SUCCESS",
207 "RESERVED",
208 "WARNING",
209 "ERROR",
210 "FATAL",
211 "BADARG",
212 "SYSERR"
215 static void
216 resize_buf(struct buf *buf, size_t initial)
219 buf->sz = buf->sz > initial/2 ? 2 * buf->sz : initial;
220 buf->buf = mandoc_realloc(buf->buf, buf->sz);
223 static void
224 pset(const char *buf, int pos, struct mparse *curp)
226 int i;
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.
235 * Separate out pmdoc/pman from mdoc/man: the first persists
236 * through all parsers, while the latter is used per-parse.
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;
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;
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;
272 if (NULL == curp->pman)
273 curp->pman = man_alloc(curp->roff, curp);
274 assert(curp->pman);
275 curp->man = curp->pman;
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').
283 static void
284 mparse_buf_r(struct mparse *curp, struct buf blk, int start)
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;
294 memset(&ln, 0, sizeof(struct buf));
296 lnn = curp->line;
297 pos = 0;
299 for (i = 0; i < (int)blk.sz; ) {
300 if (0 == pos && '\0' == blk.buf[i])
301 break;
303 if (start) {
304 curp->line = lnn;
305 curp->reparse_count = 0;
308 while (i < (int)blk.sz && (start || '\0' != blk.buf[i])) {
311 * When finding an unescaped newline character,
312 * leave the character loop to process the line.
313 * Skip a preceding carriage return, if any.
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;
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 replace these characters
330 * with "?", so we don't display gibberish.
331 * Note to manual writers: use special characters.
334 c = (unsigned char) blk.buf[i];
336 if ( ! (isascii(c) &&
337 (isgraph(c) || isblank(c)))) {
338 mandoc_msg(MANDOCERR_BADCHAR, curp,
339 curp->line, pos, NULL);
340 i++;
341 if (pos >= (int)ln.sz)
342 resize_buf(&ln, 256);
343 ln.buf[pos++] = '?';
344 continue;
347 /* Expand registers inline */
348 if ('\\' == blk.buf[i] && 'n' == blk.buf[i + 1]) {
349 roff_expand_nr(curp->roff,
350 blk.buf, &i, blk.sz, &ln.buf, &pos, &ln.sz);
351 continue;
354 /* Trailing backslash = a plain char. */
355 if ('\\' != blk.buf[i] || i + 1 == (int)blk.sz) {
356 if (pos >= (int)ln.sz)
357 resize_buf(&ln, 256);
358 ln.buf[pos++] = blk.buf[i++];
359 continue;
362 * Found escape and at least one other character.
363 * When it's a newline character, skip it.
364 * When there is a carriage return in between,
365 * skip that one as well.
368 if ('\r' == blk.buf[i + 1] && i + 2 < (int)blk.sz &&
369 '\n' == blk.buf[i + 2])
370 ++i;
371 if ('\n' == blk.buf[i + 1]) {
372 i += 2;
373 ++lnn;
374 continue;
377 if ('"' == blk.buf[i + 1] || '#' == blk.buf[i + 1]) {
378 i += 2;
379 /* Comment, skip to end of line */
380 for (; i < (int)blk.sz; ++i) {
381 if ('\n' == blk.buf[i]) {
382 ++i;
383 ++lnn;
384 break;
388 /* Backout trailing whitespaces */
389 for (; pos > 0; --pos) {
390 if (ln.buf[pos - 1] != ' ')
391 break;
392 if (pos > 2 && ln.buf[pos - 2] == '\\')
393 break;
395 break;
398 /* Some other escape sequence, copy & cont. */
400 if (pos + 1 >= (int)ln.sz)
401 resize_buf(&ln, 256);
403 ln.buf[pos++] = blk.buf[i++];
404 ln.buf[pos++] = blk.buf[i++];
407 if (pos >= (int)ln.sz)
408 resize_buf(&ln, 256);
410 ln.buf[pos] = '\0';
413 * A significant amount of complexity is contained by
414 * the roff preprocessor. It's line-oriented but can be
415 * expressed on one line, so we need at times to
416 * readjust our starting point and re-run it. The roff
417 * preprocessor can also readjust the buffers with new
418 * data, so we pass them in wholesale.
421 of = 0;
424 * Maintain a lookaside buffer of all parsed lines. We
425 * only do this if mparse_keep() has been invoked (the
426 * buffer may be accessed with mparse_getkeep()).
429 if (curp->secondary) {
430 curp->secondary->buf =
431 mandoc_realloc
432 (curp->secondary->buf,
433 curp->secondary->sz + pos + 2);
434 memcpy(curp->secondary->buf +
435 curp->secondary->sz,
436 ln.buf, pos);
437 curp->secondary->sz += pos;
438 curp->secondary->buf
439 [curp->secondary->sz] = '\n';
440 curp->secondary->sz++;
441 curp->secondary->buf
442 [curp->secondary->sz] = '\0';
444 rerun:
445 rr = roff_parseln
446 (curp->roff, curp->line,
447 &ln.buf, &ln.sz, of, &of);
449 switch (rr) {
450 case (ROFF_REPARSE):
451 if (REPARSE_LIMIT >= ++curp->reparse_count)
452 mparse_buf_r(curp, ln, 0);
453 else
454 mandoc_msg(MANDOCERR_ROFFLOOP, curp,
455 curp->line, pos, NULL);
456 pos = 0;
457 continue;
458 case (ROFF_APPEND):
459 pos = (int)strlen(ln.buf);
460 continue;
461 case (ROFF_RERUN):
462 goto rerun;
463 case (ROFF_IGN):
464 pos = 0;
465 continue;
466 case (ROFF_ERR):
467 assert(MANDOCLEVEL_FATAL <= curp->file_status);
468 break;
469 case (ROFF_SO):
471 * We remove `so' clauses from our lookaside
472 * buffer because we're going to descend into
473 * the file recursively.
475 if (curp->secondary)
476 curp->secondary->sz -= pos + 1;
477 mparse_readfd_r(curp, -1, ln.buf + of, 1);
478 if (MANDOCLEVEL_FATAL <= curp->file_status)
479 break;
480 pos = 0;
481 continue;
482 default:
483 break;
487 * If we encounter errors in the recursive parse, make
488 * sure we don't continue parsing.
491 if (MANDOCLEVEL_FATAL <= curp->file_status)
492 break;
495 * If input parsers have not been allocated, do so now.
496 * We keep these instanced between parsers, but set them
497 * locally per parse routine since we can use different
498 * parsers with each one.
501 if ( ! (curp->man || curp->mdoc))
502 pset(ln.buf + of, pos - of, curp);
505 * Lastly, push down into the parsers themselves. One
506 * of these will have already been set in the pset()
507 * routine.
508 * If libroff returns ROFF_TBL, then add it to the
509 * currently open parse. Since we only get here if
510 * there does exist data (see tbl_data.c), we're
511 * guaranteed that something's been allocated.
512 * Do the same for ROFF_EQN.
515 rc = -1;
517 if (ROFF_TBL == rr)
518 while (NULL != (span = roff_span(curp->roff))) {
519 rc = curp->man ?
520 man_addspan(curp->man, span) :
521 mdoc_addspan(curp->mdoc, span);
522 if (0 == rc)
523 break;
525 else if (ROFF_EQN == rr)
526 rc = curp->mdoc ?
527 mdoc_addeqn(curp->mdoc,
528 roff_eqn(curp->roff)) :
529 man_addeqn(curp->man,
530 roff_eqn(curp->roff));
531 else if (curp->man || curp->mdoc)
532 rc = curp->man ?
533 man_parseln(curp->man,
534 curp->line, ln.buf, of) :
535 mdoc_parseln(curp->mdoc,
536 curp->line, ln.buf, of);
538 if (0 == rc) {
539 assert(MANDOCLEVEL_FATAL <= curp->file_status);
540 break;
543 /* Temporary buffers typically are not full. */
545 if (0 == start && '\0' == blk.buf[i])
546 break;
548 /* Start the next input line. */
550 pos = 0;
553 free(ln.buf);
556 static int
557 read_whole_file(const char *file, int fd, struct buf *fb, int *with_mmap)
559 size_t off;
560 ssize_t ssz;
562 #ifdef HAVE_MMAP
563 struct stat st;
564 if (-1 == fstat(fd, &st)) {
565 perror(file);
566 return(0);
570 * If we're a regular file, try just reading in the whole entry
571 * via mmap(). This is faster than reading it into blocks, and
572 * since each file is only a few bytes to begin with, I'm not
573 * concerned that this is going to tank any machines.
576 if (S_ISREG(st.st_mode)) {
577 if (st.st_size >= (1U << 31)) {
578 fprintf(stderr, "%s: input too large\n", file);
579 return(0);
581 *with_mmap = 1;
582 fb->sz = (size_t)st.st_size;
583 fb->buf = mmap(NULL, fb->sz, PROT_READ,
584 MAP_FILE|MAP_SHARED, fd, 0);
585 if (fb->buf != MAP_FAILED)
586 return(1);
588 #endif
591 * If this isn't a regular file (like, say, stdin), then we must
592 * go the old way and just read things in bit by bit.
595 *with_mmap = 0;
596 off = 0;
597 fb->sz = 0;
598 fb->buf = NULL;
599 for (;;) {
600 if (off == fb->sz) {
601 if (fb->sz == (1U << 31)) {
602 fprintf(stderr, "%s: input too large\n", file);
603 break;
605 resize_buf(fb, 65536);
607 ssz = read(fd, fb->buf + (int)off, fb->sz - off);
608 if (ssz == 0) {
609 fb->sz = off;
610 return(1);
612 if (ssz == -1) {
613 perror(file);
614 break;
616 off += (size_t)ssz;
619 free(fb->buf);
620 fb->buf = NULL;
621 return(0);
624 static void
625 mparse_end(struct mparse *curp)
628 if (MANDOCLEVEL_FATAL <= curp->file_status)
629 return;
631 if (curp->mdoc && ! mdoc_endparse(curp->mdoc)) {
632 assert(MANDOCLEVEL_FATAL <= curp->file_status);
633 return;
636 if (curp->man && ! man_endparse(curp->man)) {
637 assert(MANDOCLEVEL_FATAL <= curp->file_status);
638 return;
641 if ( ! (curp->man || curp->mdoc)) {
642 mandoc_msg(MANDOCERR_NOTMANUAL, curp, 1, 0, NULL);
643 curp->file_status = MANDOCLEVEL_FATAL;
644 return;
647 roff_endparse(curp->roff);
650 static void
651 mparse_parse_buffer(struct mparse *curp, struct buf blk, const char *file,
652 int re)
654 const char *svfile;
656 /* Line number is per-file. */
657 svfile = curp->file;
658 curp->file = file;
659 curp->line = 1;
661 mparse_buf_r(curp, blk, 1);
663 if (0 == re && MANDOCLEVEL_FATAL > curp->file_status)
664 mparse_end(curp);
666 curp->file = svfile;
669 enum mandoclevel
670 mparse_readmem(struct mparse *curp, const void *buf, size_t len,
671 const char *file)
673 struct buf blk;
675 blk.buf = UNCONST(buf);
676 blk.sz = len;
678 mparse_parse_buffer(curp, blk, file, 0);
679 return(curp->file_status);
682 static void
683 mparse_readfd_r(struct mparse *curp, int fd, const char *file, int re)
685 struct buf blk;
686 int with_mmap;
688 if (-1 == fd)
689 if (-1 == (fd = open(file, O_RDONLY, 0))) {
690 perror(file);
691 curp->file_status = MANDOCLEVEL_SYSERR;
692 return;
695 * Run for each opened file; may be called more than once for
696 * each full parse sequence if the opened file is nested (i.e.,
697 * from `so'). Simply sucks in the whole file and moves into
698 * the parse phase for the file.
701 if ( ! read_whole_file(file, fd, &blk, &with_mmap)) {
702 curp->file_status = MANDOCLEVEL_SYSERR;
703 return;
706 mparse_parse_buffer(curp, blk, file, re);
708 #ifdef HAVE_MMAP
709 if (with_mmap)
710 munmap(blk.buf, blk.sz);
711 else
712 #endif
713 free(blk.buf);
715 if (STDIN_FILENO != fd && -1 == close(fd))
716 perror(file);
719 enum mandoclevel
720 mparse_readfd(struct mparse *curp, int fd, const char *file)
723 mparse_readfd_r(curp, fd, file, 0);
724 return(curp->file_status);
727 struct mparse *
728 mparse_alloc(enum mparset inttype, enum mandoclevel wlevel, mandocmsg mmsg, void *arg)
730 struct mparse *curp;
732 assert(wlevel <= MANDOCLEVEL_FATAL);
734 curp = mandoc_calloc(1, sizeof(struct mparse));
736 curp->wlevel = wlevel;
737 curp->mmsg = mmsg;
738 curp->arg = arg;
739 curp->inttype = inttype;
741 curp->roff = roff_alloc(curp);
742 return(curp);
745 void
746 mparse_reset(struct mparse *curp)
749 roff_reset(curp->roff);
751 if (curp->mdoc)
752 mdoc_reset(curp->mdoc);
753 if (curp->man)
754 man_reset(curp->man);
755 if (curp->secondary)
756 curp->secondary->sz = 0;
758 curp->file_status = MANDOCLEVEL_OK;
759 curp->mdoc = NULL;
760 curp->man = NULL;
763 void
764 mparse_free(struct mparse *curp)
767 if (curp->pmdoc)
768 mdoc_free(curp->pmdoc);
769 if (curp->pman)
770 man_free(curp->pman);
771 if (curp->roff)
772 roff_free(curp->roff);
773 if (curp->secondary)
774 free(curp->secondary->buf);
776 free(curp->secondary);
777 free(curp);
780 void
781 mparse_result(struct mparse *curp, struct mdoc **mdoc, struct man **man)
784 if (mdoc)
785 *mdoc = curp->mdoc;
786 if (man)
787 *man = curp->man;
790 void
791 mandoc_vmsg(enum mandocerr t, struct mparse *m,
792 int ln, int pos, const char *fmt, ...)
794 char buf[256];
795 va_list ap;
797 va_start(ap, fmt);
798 vsnprintf(buf, sizeof(buf) - 1, fmt, ap);
799 va_end(ap);
801 mandoc_msg(t, m, ln, pos, buf);
804 void
805 mandoc_msg(enum mandocerr er, struct mparse *m,
806 int ln, int col, const char *msg)
808 enum mandoclevel level;
810 level = MANDOCLEVEL_FATAL;
811 while (er < mandoclimits[level])
812 level--;
814 if (level < m->wlevel)
815 return;
817 if (m->mmsg)
818 (*m->mmsg)(er, level, m->file, ln, col, msg);
820 if (m->file_status < level)
821 m->file_status = level;
824 const char *
825 mparse_strerror(enum mandocerr er)
828 return(mandocerrs[er]);
831 const char *
832 mparse_strlevel(enum mandoclevel lvl)
834 return(mandoclevels[lvl]);
837 void
838 mparse_keep(struct mparse *p)
841 assert(NULL == p->secondary);
842 p->secondary = mandoc_calloc(1, sizeof(struct buf));
845 const char *
846 mparse_getkeep(const struct mparse *p)
849 assert(p->secondary);
850 return(p->secondary->sz ? p->secondary->buf : NULL);