]>
git.cameronkatri.com Git - mandoc.git/blob - read.c
1 /* $Id: read.c,v 1.28 2012/02/16 20:51:31 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>
38 #include "libmandoc.h"
47 #define REPARSE_LIMIT 1000
50 char *buf
; /* binary input buffer */
51 size_t sz
; /* size of binary buffer */
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 */
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
] = {
88 static const char * const mandocerrs
[MANDOCERR_MAX
] = {
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",
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",
130 "skipping end of block that is not open",
132 /* related to bad macro arguments */
134 "duplicate argument",
135 "duplicate display type",
136 "duplicate list type",
137 "unknown AT&T UNIX version",
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",
148 "bad escape sequence",
149 "unterminated quoted string",
151 /* related to equations */
152 "unexpected literal in equation",
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 */
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)",
188 "line argument(s) will be lost",
189 "body argument(s) will be lost",
191 "generic fatal error",
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 \"..\"",
201 "no document prologue",
202 "static buffer exhausted",
205 static const char * const mandoclevels
[MANDOCLEVEL_MAX
] = {
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
);
224 pset(const char *buf
, int pos
, struct mparse
*curp
)
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
])
247 switch (curp
->inttype
) {
249 if (NULL
== curp
->pmdoc
)
250 curp
->pmdoc
= mdoc_alloc(curp
->roff
, curp
);
252 curp
->mdoc
= curp
->pmdoc
;
255 if (NULL
== curp
->pman
)
256 curp
->pman
= man_alloc(curp
->roff
, curp
);
258 curp
->man
= curp
->pman
;
264 if (pos
>= 3 && 0 == memcmp(buf
, ".Dd", 3)) {
265 if (NULL
== curp
->pmdoc
)
266 curp
->pmdoc
= mdoc_alloc(curp
->roff
, curp
);
268 curp
->mdoc
= curp
->pmdoc
;
272 if (NULL
== curp
->pman
)
273 curp
->pman
= man_alloc(curp
->roff
, curp
);
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').
284 mparse_buf_r(struct mparse
*curp
, struct buf blk
, int start
)
286 const struct tbl_span
*span
;
290 int pos
; /* byte number in the ln buffer */
291 int lnn
; /* line number in the real file */
294 memset(&ln
, 0, sizeof(struct buf
));
299 for (i
= 0; i
< (int)blk
.sz
; ) {
300 if (0 == pos
&& '\0' == blk
.buf
[i
])
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])
319 if ('\n' == blk
.buf
[i
]) {
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
);
341 if (pos
>= (int)ln
.sz
)
342 resize_buf(&ln
, 256);
347 /* Trailing backslash = a plain char. */
349 if ('\\' != blk
.buf
[i
] || i
+ 1 == (int)blk
.sz
) {
350 if (pos
>= (int)ln
.sz
)
351 resize_buf(&ln
, 256);
352 ln
.buf
[pos
++] = blk
.buf
[i
++];
357 * Found escape and at least one other character.
358 * When it's a newline character, skip it.
359 * When there is a carriage return in between,
360 * skip that one as well.
363 if ('\r' == blk
.buf
[i
+ 1] && i
+ 2 < (int)blk
.sz
&&
364 '\n' == blk
.buf
[i
+ 2])
366 if ('\n' == blk
.buf
[i
+ 1]) {
372 if ('"' == blk
.buf
[i
+ 1] || '#' == blk
.buf
[i
+ 1]) {
374 /* Comment, skip to end of line */
375 for (; i
< (int)blk
.sz
; ++i
) {
376 if ('\n' == blk
.buf
[i
]) {
383 /* Backout trailing whitespaces */
384 for (; pos
> 0; --pos
) {
385 if (ln
.buf
[pos
- 1] != ' ')
387 if (pos
> 2 && ln
.buf
[pos
- 2] == '\\')
393 /* Some other escape sequence, copy & cont. */
395 if (pos
+ 1 >= (int)ln
.sz
)
396 resize_buf(&ln
, 256);
398 ln
.buf
[pos
++] = blk
.buf
[i
++];
399 ln
.buf
[pos
++] = blk
.buf
[i
++];
402 if (pos
>= (int)ln
.sz
)
403 resize_buf(&ln
, 256);
408 * A significant amount of complexity is contained by
409 * the roff preprocessor. It's line-oriented but can be
410 * expressed on one line, so we need at times to
411 * readjust our starting point and re-run it. The roff
412 * preprocessor can also readjust the buffers with new
413 * data, so we pass them in wholesale.
419 * Maintain a lookaside buffer of all parsed lines. We
420 * only do this if mparse_keep() has been invoked (the
421 * buffer may be accessed with mparse_getkeep()).
424 if (curp
->secondary
) {
425 curp
->secondary
->buf
=
427 (curp
->secondary
->buf
,
428 curp
->secondary
->sz
+ pos
+ 2);
429 memcpy(curp
->secondary
->buf
+
432 curp
->secondary
->sz
+= pos
;
434 [curp
->secondary
->sz
] = '\n';
435 curp
->secondary
->sz
++;
437 [curp
->secondary
->sz
] = '\0';
441 (curp
->roff
, curp
->line
,
442 &ln
.buf
, &ln
.sz
, of
, &of
);
446 if (REPARSE_LIMIT
>= ++curp
->reparse_count
)
447 mparse_buf_r(curp
, ln
, 0);
449 mandoc_msg(MANDOCERR_ROFFLOOP
, curp
,
450 curp
->line
, pos
, NULL
);
454 pos
= (int)strlen(ln
.buf
);
462 assert(MANDOCLEVEL_FATAL
<= curp
->file_status
);
466 * We remove `so' clauses from our lookaside
467 * buffer because we're going to descend into
468 * the file recursively.
471 curp
->secondary
->sz
-= pos
+ 1;
472 mparse_readfd_r(curp
, -1, ln
.buf
+ of
, 1);
473 if (MANDOCLEVEL_FATAL
<= curp
->file_status
)
482 * If we encounter errors in the recursive parse, make
483 * sure we don't continue parsing.
486 if (MANDOCLEVEL_FATAL
<= curp
->file_status
)
490 * If input parsers have not been allocated, do so now.
491 * We keep these instanced between parsers, but set them
492 * locally per parse routine since we can use different
493 * parsers with each one.
496 if ( ! (curp
->man
|| curp
->mdoc
))
497 pset(ln
.buf
+ of
, pos
- of
, curp
);
500 * Lastly, push down into the parsers themselves. One
501 * of these will have already been set in the pset()
503 * If libroff returns ROFF_TBL, then add it to the
504 * currently open parse. Since we only get here if
505 * there does exist data (see tbl_data.c), we're
506 * guaranteed that something's been allocated.
507 * Do the same for ROFF_EQN.
513 while (NULL
!= (span
= roff_span(curp
->roff
))) {
515 man_addspan(curp
->man
, span
) :
516 mdoc_addspan(curp
->mdoc
, span
);
520 else if (ROFF_EQN
== rr
)
522 mdoc_addeqn(curp
->mdoc
,
523 roff_eqn(curp
->roff
)) :
524 man_addeqn(curp
->man
,
525 roff_eqn(curp
->roff
));
526 else if (curp
->man
|| curp
->mdoc
)
528 man_parseln(curp
->man
,
529 curp
->line
, ln
.buf
, of
) :
530 mdoc_parseln(curp
->mdoc
,
531 curp
->line
, ln
.buf
, of
);
534 assert(MANDOCLEVEL_FATAL
<= curp
->file_status
);
538 /* Temporary buffers typically are not full. */
540 if (0 == start
&& '\0' == blk
.buf
[i
])
543 /* Start the next input line. */
552 read_whole_file(const char *file
, int fd
, struct buf
*fb
, int *with_mmap
)
559 if (-1 == fstat(fd
, &st
)) {
565 * If we're a regular file, try just reading in the whole entry
566 * via mmap(). This is faster than reading it into blocks, and
567 * since each file is only a few bytes to begin with, I'm not
568 * concerned that this is going to tank any machines.
571 if (S_ISREG(st
.st_mode
)) {
572 if (st
.st_size
>= (1U << 31)) {
573 fprintf(stderr
, "%s: input too large\n", file
);
577 fb
->sz
= (size_t)st
.st_size
;
578 fb
->buf
= mmap(NULL
, fb
->sz
, PROT_READ
,
579 MAP_FILE
|MAP_SHARED
, fd
, 0);
580 if (fb
->buf
!= MAP_FAILED
)
586 * If this isn't a regular file (like, say, stdin), then we must
587 * go the old way and just read things in bit by bit.
596 if (fb
->sz
== (1U << 31)) {
597 fprintf(stderr
, "%s: input too large\n", file
);
600 resize_buf(fb
, 65536);
602 ssz
= read(fd
, fb
->buf
+ (int)off
, fb
->sz
- off
);
620 mparse_end(struct mparse
*curp
)
623 if (MANDOCLEVEL_FATAL
<= curp
->file_status
)
626 if (curp
->mdoc
&& ! mdoc_endparse(curp
->mdoc
)) {
627 assert(MANDOCLEVEL_FATAL
<= curp
->file_status
);
631 if (curp
->man
&& ! man_endparse(curp
->man
)) {
632 assert(MANDOCLEVEL_FATAL
<= curp
->file_status
);
636 if ( ! (curp
->man
|| curp
->mdoc
)) {
637 mandoc_msg(MANDOCERR_NOTMANUAL
, curp
, 1, 0, NULL
);
638 curp
->file_status
= MANDOCLEVEL_FATAL
;
642 roff_endparse(curp
->roff
);
646 mparse_parse_buffer(struct mparse
*curp
, struct buf blk
, const char *file
,
651 /* Line number is per-file. */
656 mparse_buf_r(curp
, blk
, 1);
658 if (0 == re
&& MANDOCLEVEL_FATAL
> curp
->file_status
)
665 mparse_readmem(struct mparse
*curp
, const void *buf
, size_t len
,
670 blk
.buf
= UNCONST(buf
);
673 mparse_parse_buffer(curp
, blk
, file
, 0);
674 return(curp
->file_status
);
678 mparse_readfd_r(struct mparse
*curp
, int fd
, const char *file
, int re
)
684 if (-1 == (fd
= open(file
, O_RDONLY
, 0))) {
686 curp
->file_status
= MANDOCLEVEL_SYSERR
;
690 * Run for each opened file; may be called more than once for
691 * each full parse sequence if the opened file is nested (i.e.,
692 * from `so'). Simply sucks in the whole file and moves into
693 * the parse phase for the file.
696 if ( ! read_whole_file(file
, fd
, &blk
, &with_mmap
)) {
697 curp
->file_status
= MANDOCLEVEL_SYSERR
;
701 mparse_parse_buffer(curp
, blk
, file
, re
);
705 munmap(blk
.buf
, blk
.sz
);
710 if (STDIN_FILENO
!= fd
&& -1 == close(fd
))
715 mparse_readfd(struct mparse
*curp
, int fd
, const char *file
)
718 mparse_readfd_r(curp
, fd
, file
, 0);
719 return(curp
->file_status
);
723 mparse_alloc(enum mparset inttype
, enum mandoclevel wlevel
, mandocmsg mmsg
, void *arg
)
727 assert(wlevel
<= MANDOCLEVEL_FATAL
);
729 curp
= mandoc_calloc(1, sizeof(struct mparse
));
731 curp
->wlevel
= wlevel
;
734 curp
->inttype
= inttype
;
736 curp
->roff
= roff_alloc(curp
);
741 mparse_reset(struct mparse
*curp
)
744 roff_reset(curp
->roff
);
747 mdoc_reset(curp
->mdoc
);
749 man_reset(curp
->man
);
751 curp
->secondary
->sz
= 0;
753 curp
->file_status
= MANDOCLEVEL_OK
;
759 mparse_free(struct mparse
*curp
)
763 mdoc_free(curp
->pmdoc
);
765 man_free(curp
->pman
);
767 roff_free(curp
->roff
);
769 free(curp
->secondary
->buf
);
771 free(curp
->secondary
);
776 mparse_result(struct mparse
*curp
, struct mdoc
**mdoc
, struct man
**man
)
786 mandoc_vmsg(enum mandocerr t
, struct mparse
*m
,
787 int ln
, int pos
, const char *fmt
, ...)
793 vsnprintf(buf
, sizeof(buf
) - 1, fmt
, ap
);
796 mandoc_msg(t
, m
, ln
, pos
, buf
);
800 mandoc_msg(enum mandocerr er
, struct mparse
*m
,
801 int ln
, int col
, const char *msg
)
803 enum mandoclevel level
;
805 level
= MANDOCLEVEL_FATAL
;
806 while (er
< mandoclimits
[level
])
809 if (level
< m
->wlevel
)
813 (*m
->mmsg
)(er
, level
, m
->file
, ln
, col
, msg
);
815 if (m
->file_status
< level
)
816 m
->file_status
= level
;
820 mparse_strerror(enum mandocerr er
)
823 return(mandocerrs
[er
]);
827 mparse_strlevel(enum mandoclevel lvl
)
829 return(mandoclevels
[lvl
]);
833 mparse_keep(struct mparse
*p
)
836 assert(NULL
== p
->secondary
);
837 p
->secondary
= mandoc_calloc(1, sizeof(struct buf
));
841 mparse_getkeep(const struct mparse
*p
)
844 assert(p
->secondary
);
845 return(p
->secondary
->sz
? p
->secondary
->buf
: NULL
);