]>
git.cameronkatri.com Git - mandoc.git/blob - read.c
1 /* $Id: read.c,v 1.27 2012/02/05 16:46:15 joerg Exp $ */
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.
23 # include <sys/stat.h>
24 # include <sys/mman.h>
37 #include "libmandoc.h"
45 #define REPARSE_LIMIT 1000
48 char *buf
; /* binary input buffer */
49 size_t sz
; /* size of binary buffer */
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 */
66 struct buf
*secondary
;
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
*);
77 static const enum mandocerr mandoclimits
[MANDOCLEVEL_MAX
] = {
87 static const char * const mandocerrs
[MANDOCERR_MAX
] = {
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",
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",
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",
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",
129 "skipping end of block that is not open",
131 /* related to bad macro arguments */
133 "duplicate argument",
134 "duplicate display type",
135 "duplicate list type",
136 "unknown AT&T UNIX version",
139 "unknown standard specifier",
140 "bad width argument",
142 /* related to plain text */
143 "blank line in non-literal context",
144 "tab in non-literal context",
145 "end of line whitespace",
147 "bad escape sequence",
148 "unterminated quoted string",
150 /* related to equations */
151 "unexpected literal in equation",
155 /* related to equations */
156 "unexpected equation scope closure",
157 "equation scope open on exit",
158 "overlapping equation scopes",
159 "unexpected end of equation",
160 "equation syntax error",
162 /* related to tables */
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",
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)",
187 "line argument(s) will be lost",
188 "body argument(s) will be lost",
190 "generic fatal error",
193 "column syntax is inconsistent",
194 "NOT IMPLEMENTED: .Bd -file",
195 "argument count wrong, violates syntax",
196 "child violates parent syntax",
197 "argument count wrong, violates syntax",
198 "NOT IMPLEMENTED: .so with absolute path or \"..\"",
200 "no document prologue",
201 "static buffer exhausted",
204 static const char * const mandoclevels
[MANDOCLEVEL_MAX
] = {
215 resize_buf(struct buf
*buf
, size_t initial
)
218 buf
->sz
= buf
->sz
> initial
/2 ? 2 * buf
->sz
: initial
;
219 buf
->buf
= mandoc_realloc(buf
->buf
, buf
->sz
);
223 pset(const char *buf
, int pos
, struct mparse
*curp
)
228 * Try to intuit which kind of manual parser should be used. If
229 * passed in by command-line (-man, -mdoc), then use that
230 * explicitly. If passed as -mandoc, then try to guess from the
231 * line: either skip dot-lines, use -mdoc when finding `.Dt', or
232 * default to -man, which is more lenient.
234 * Separate out pmdoc/pman from mdoc/man: the first persists
235 * through all parsers, while the latter is used per-parse.
238 if ('.' == buf
[0] || '\'' == buf
[0]) {
239 for (i
= 1; buf
[i
]; i
++)
240 if (' ' != buf
[i
] && '\t' != buf
[i
])
246 switch (curp
->inttype
) {
248 if (NULL
== curp
->pmdoc
)
249 curp
->pmdoc
= mdoc_alloc(curp
->roff
, curp
);
251 curp
->mdoc
= curp
->pmdoc
;
254 if (NULL
== curp
->pman
)
255 curp
->pman
= man_alloc(curp
->roff
, curp
);
257 curp
->man
= curp
->pman
;
263 if (pos
>= 3 && 0 == memcmp(buf
, ".Dd", 3)) {
264 if (NULL
== curp
->pmdoc
)
265 curp
->pmdoc
= mdoc_alloc(curp
->roff
, curp
);
267 curp
->mdoc
= curp
->pmdoc
;
271 if (NULL
== curp
->pman
)
272 curp
->pman
= man_alloc(curp
->roff
, curp
);
274 curp
->man
= curp
->pman
;
278 * Main parse routine for an opened file. This is called for each
279 * opened file and simply loops around the full input file, possibly
280 * nesting (i.e., with `so').
283 mparse_buf_r(struct mparse
*curp
, struct buf blk
, int start
)
285 const struct tbl_span
*span
;
289 int pos
; /* byte number in the ln buffer */
290 int lnn
; /* line number in the real file */
293 memset(&ln
, 0, sizeof(struct buf
));
298 for (i
= 0; i
< (int)blk
.sz
; ) {
299 if (0 == pos
&& '\0' == blk
.buf
[i
])
304 curp
->reparse_count
= 0;
307 while (i
< (int)blk
.sz
&& (start
|| '\0' != blk
.buf
[i
])) {
310 * When finding an unescaped newline character,
311 * leave the character loop to process the line.
312 * Skip a preceding carriage return, if any.
315 if ('\r' == blk
.buf
[i
] && i
+ 1 < (int)blk
.sz
&&
316 '\n' == blk
.buf
[i
+ 1])
318 if ('\n' == blk
.buf
[i
]) {
325 * Warn about bogus characters. If you're using
326 * non-ASCII encoding, you're screwing your
327 * readers. Since I'd rather this not happen,
328 * I'll be helpful and replace these characters
329 * with "?", so we don't display gibberish.
330 * Note to manual writers: use special characters.
333 c
= (unsigned char) blk
.buf
[i
];
335 if ( ! (isascii(c
) &&
336 (isgraph(c
) || isblank(c
)))) {
337 mandoc_msg(MANDOCERR_BADCHAR
, curp
,
338 curp
->line
, pos
, NULL
);
340 if (pos
>= (int)ln
.sz
)
341 resize_buf(&ln
, 256);
346 /* Trailing backslash = a plain char. */
348 if ('\\' != blk
.buf
[i
] || i
+ 1 == (int)blk
.sz
) {
349 if (pos
>= (int)ln
.sz
)
350 resize_buf(&ln
, 256);
351 ln
.buf
[pos
++] = blk
.buf
[i
++];
356 * Found escape and at least one other character.
357 * When it's a newline character, skip it.
358 * When there is a carriage return in between,
359 * skip that one as well.
362 if ('\r' == blk
.buf
[i
+ 1] && i
+ 2 < (int)blk
.sz
&&
363 '\n' == blk
.buf
[i
+ 2])
365 if ('\n' == blk
.buf
[i
+ 1]) {
371 if ('"' == blk
.buf
[i
+ 1] || '#' == blk
.buf
[i
+ 1]) {
373 /* Comment, skip to end of line */
374 for (; i
< (int)blk
.sz
; ++i
) {
375 if ('\n' == blk
.buf
[i
]) {
382 /* Backout trailing whitespaces */
383 for (; pos
> 0; --pos
) {
384 if (ln
.buf
[pos
- 1] != ' ')
386 if (pos
> 2 && ln
.buf
[pos
- 2] == '\\')
392 /* Some other escape sequence, copy & cont. */
394 if (pos
+ 1 >= (int)ln
.sz
)
395 resize_buf(&ln
, 256);
397 ln
.buf
[pos
++] = blk
.buf
[i
++];
398 ln
.buf
[pos
++] = blk
.buf
[i
++];
401 if (pos
>= (int)ln
.sz
)
402 resize_buf(&ln
, 256);
407 * A significant amount of complexity is contained by
408 * the roff preprocessor. It's line-oriented but can be
409 * expressed on one line, so we need at times to
410 * readjust our starting point and re-run it. The roff
411 * preprocessor can also readjust the buffers with new
412 * data, so we pass them in wholesale.
418 * Maintain a lookaside buffer of all parsed lines. We
419 * only do this if mparse_keep() has been invoked (the
420 * buffer may be accessed with mparse_getkeep()).
423 if (curp
->secondary
) {
424 curp
->secondary
->buf
=
426 (curp
->secondary
->buf
,
427 curp
->secondary
->sz
+ pos
+ 2);
428 memcpy(curp
->secondary
->buf
+
431 curp
->secondary
->sz
+= pos
;
433 [curp
->secondary
->sz
] = '\n';
434 curp
->secondary
->sz
++;
436 [curp
->secondary
->sz
] = '\0';
440 (curp
->roff
, curp
->line
,
441 &ln
.buf
, &ln
.sz
, of
, &of
);
445 if (REPARSE_LIMIT
>= ++curp
->reparse_count
)
446 mparse_buf_r(curp
, ln
, 0);
448 mandoc_msg(MANDOCERR_ROFFLOOP
, curp
,
449 curp
->line
, pos
, NULL
);
453 pos
= (int)strlen(ln
.buf
);
461 assert(MANDOCLEVEL_FATAL
<= curp
->file_status
);
465 * We remove `so' clauses from our lookaside
466 * buffer because we're going to descend into
467 * the file recursively.
470 curp
->secondary
->sz
-= pos
+ 1;
471 mparse_readfd_r(curp
, -1, ln
.buf
+ of
, 1);
472 if (MANDOCLEVEL_FATAL
<= curp
->file_status
)
481 * If we encounter errors in the recursive parse, make
482 * sure we don't continue parsing.
485 if (MANDOCLEVEL_FATAL
<= curp
->file_status
)
489 * If input parsers have not been allocated, do so now.
490 * We keep these instanced between parsers, but set them
491 * locally per parse routine since we can use different
492 * parsers with each one.
495 if ( ! (curp
->man
|| curp
->mdoc
))
496 pset(ln
.buf
+ of
, pos
- of
, curp
);
499 * Lastly, push down into the parsers themselves. One
500 * of these will have already been set in the pset()
502 * If libroff returns ROFF_TBL, then add it to the
503 * currently open parse. Since we only get here if
504 * there does exist data (see tbl_data.c), we're
505 * guaranteed that something's been allocated.
506 * Do the same for ROFF_EQN.
512 while (NULL
!= (span
= roff_span(curp
->roff
))) {
514 man_addspan(curp
->man
, span
) :
515 mdoc_addspan(curp
->mdoc
, span
);
519 else if (ROFF_EQN
== rr
)
521 mdoc_addeqn(curp
->mdoc
,
522 roff_eqn(curp
->roff
)) :
523 man_addeqn(curp
->man
,
524 roff_eqn(curp
->roff
));
525 else if (curp
->man
|| curp
->mdoc
)
527 man_parseln(curp
->man
,
528 curp
->line
, ln
.buf
, of
) :
529 mdoc_parseln(curp
->mdoc
,
530 curp
->line
, ln
.buf
, of
);
533 assert(MANDOCLEVEL_FATAL
<= curp
->file_status
);
537 /* Temporary buffers typically are not full. */
539 if (0 == start
&& '\0' == blk
.buf
[i
])
542 /* Start the next input line. */
551 pdesc(struct mparse
*curp
, const char *file
, int fd
)
557 * Run for each opened file; may be called more than once for
558 * each full parse sequence if the opened file is nested (i.e.,
559 * from `so'). Simply sucks in the whole file and moves into
560 * the parse phase for the file.
563 if ( ! read_whole_file(file
, fd
, &blk
, &with_mmap
)) {
564 curp
->file_status
= MANDOCLEVEL_SYSERR
;
568 /* Line number is per-file. */
572 mparse_buf_r(curp
, blk
, 1);
576 munmap(blk
.buf
, blk
.sz
);
583 read_whole_file(const char *file
, int fd
, struct buf
*fb
, int *with_mmap
)
590 if (-1 == fstat(fd
, &st
)) {
596 * If we're a regular file, try just reading in the whole entry
597 * via mmap(). This is faster than reading it into blocks, and
598 * since each file is only a few bytes to begin with, I'm not
599 * concerned that this is going to tank any machines.
602 if (S_ISREG(st
.st_mode
)) {
603 if (st
.st_size
>= (1U << 31)) {
604 fprintf(stderr
, "%s: input too large\n", file
);
608 fb
->sz
= (size_t)st
.st_size
;
609 fb
->buf
= mmap(NULL
, fb
->sz
, PROT_READ
,
610 MAP_FILE
|MAP_SHARED
, fd
, 0);
611 if (fb
->buf
!= MAP_FAILED
)
617 * If this isn't a regular file (like, say, stdin), then we must
618 * go the old way and just read things in bit by bit.
627 if (fb
->sz
== (1U << 31)) {
628 fprintf(stderr
, "%s: input too large\n", file
);
631 resize_buf(fb
, 65536);
633 ssz
= read(fd
, fb
->buf
+ (int)off
, fb
->sz
- off
);
651 mparse_end(struct mparse
*curp
)
654 if (MANDOCLEVEL_FATAL
<= curp
->file_status
)
657 if (curp
->mdoc
&& ! mdoc_endparse(curp
->mdoc
)) {
658 assert(MANDOCLEVEL_FATAL
<= curp
->file_status
);
662 if (curp
->man
&& ! man_endparse(curp
->man
)) {
663 assert(MANDOCLEVEL_FATAL
<= curp
->file_status
);
667 if ( ! (curp
->man
|| curp
->mdoc
)) {
668 mandoc_msg(MANDOCERR_NOTMANUAL
, curp
, 1, 0, NULL
);
669 curp
->file_status
= MANDOCLEVEL_FATAL
;
673 roff_endparse(curp
->roff
);
677 mparse_readfd_r(struct mparse
*curp
, int fd
, const char *file
, int re
)
682 if (-1 == (fd
= open(file
, O_RDONLY
, 0))) {
684 curp
->file_status
= MANDOCLEVEL_SYSERR
;
691 pdesc(curp
, file
, fd
);
693 if (0 == re
&& MANDOCLEVEL_FATAL
> curp
->file_status
)
696 if (STDIN_FILENO
!= fd
&& -1 == close(fd
))
703 mparse_readfd(struct mparse
*curp
, int fd
, const char *file
)
706 mparse_readfd_r(curp
, fd
, file
, 0);
707 return(curp
->file_status
);
711 mparse_alloc(enum mparset inttype
, enum mandoclevel wlevel
, mandocmsg mmsg
, void *arg
)
715 assert(wlevel
<= MANDOCLEVEL_FATAL
);
717 curp
= mandoc_calloc(1, sizeof(struct mparse
));
719 curp
->wlevel
= wlevel
;
722 curp
->inttype
= inttype
;
724 curp
->roff
= roff_alloc(curp
);
729 mparse_reset(struct mparse
*curp
)
732 roff_reset(curp
->roff
);
735 mdoc_reset(curp
->mdoc
);
737 man_reset(curp
->man
);
739 curp
->secondary
->sz
= 0;
741 curp
->file_status
= MANDOCLEVEL_OK
;
747 mparse_free(struct mparse
*curp
)
751 mdoc_free(curp
->pmdoc
);
753 man_free(curp
->pman
);
755 roff_free(curp
->roff
);
757 free(curp
->secondary
->buf
);
759 free(curp
->secondary
);
764 mparse_result(struct mparse
*curp
, struct mdoc
**mdoc
, struct man
**man
)
774 mandoc_vmsg(enum mandocerr t
, struct mparse
*m
,
775 int ln
, int pos
, const char *fmt
, ...)
781 vsnprintf(buf
, sizeof(buf
) - 1, fmt
, ap
);
784 mandoc_msg(t
, m
, ln
, pos
, buf
);
788 mandoc_msg(enum mandocerr er
, struct mparse
*m
,
789 int ln
, int col
, const char *msg
)
791 enum mandoclevel level
;
793 level
= MANDOCLEVEL_FATAL
;
794 while (er
< mandoclimits
[level
])
797 if (level
< m
->wlevel
)
801 (*m
->mmsg
)(er
, level
, m
->file
, ln
, col
, msg
);
803 if (m
->file_status
< level
)
804 m
->file_status
= level
;
808 mparse_strerror(enum mandocerr er
)
811 return(mandocerrs
[er
]);
815 mparse_strlevel(enum mandoclevel lvl
)
817 return(mandoclevels
[lvl
]);
821 mparse_keep(struct mparse
*p
)
824 assert(NULL
== p
->secondary
);
825 p
->secondary
= mandoc_calloc(1, sizeof(struct buf
));
829 mparse_getkeep(const struct mparse
*p
)
832 assert(p
->secondary
);
833 return(p
->secondary
->sz
? p
->secondary
->buf
: NULL
);