]>
git.cameronkatri.com Git - mandoc.git/blob - read.c
1 /* $Id: read.c,v 1.29 2012/05/27 17:48:57 schwarze 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
;
69 char *defos
; /* default operating system */
72 static void resize_buf(struct buf
*, size_t);
73 static void mparse_buf_r(struct mparse
*, struct buf
, int);
74 static void mparse_readfd_r(struct mparse
*, int, const char *, int);
75 static void pset(const char *, int, struct mparse
*);
76 static int read_whole_file(const char *, int, struct buf
*, int *);
77 static void mparse_end(struct mparse
*);
79 static const enum mandocerr mandoclimits
[MANDOCLEVEL_MAX
] = {
89 static const char * const mandocerrs
[MANDOCERR_MAX
] = {
94 /* related to the prologue */
95 "no title in document",
96 "document title should be all caps",
97 "unknown manual section",
98 "date missing, using today's date",
99 "cannot parse date, using it verbatim",
100 "prologue macros out of order",
101 "duplicate prologue macro",
102 "macro not allowed in prologue",
103 "macro not allowed in body",
105 /* related to document structure */
106 ".so is fragile, better use ln(1)",
107 "NAME section must come first",
108 "bad NAME section contents",
109 "manual name not yet set",
110 "sections out of conventional order",
111 "duplicate section name",
112 "section not in conventional manual section",
114 /* related to macros and nesting */
115 "skipping obsolete macro",
116 "skipping paragraph macro",
117 "skipping no-space macro",
118 "blocks badly nested",
119 "child violates parent syntax",
120 "nested displays are not portable",
121 "already in literal mode",
124 /* related to missing macro arguments */
125 "skipping empty macro",
126 "argument count wrong",
127 "missing display type",
128 "list type must come first",
129 "tag lists require a width argument",
131 "skipping end of block that is not open",
133 /* related to bad macro arguments */
135 "duplicate argument",
136 "duplicate display type",
137 "duplicate list type",
138 "unknown AT&T UNIX version",
141 "unknown standard specifier",
142 "bad width argument",
144 /* related to plain text */
145 "blank line in non-literal context",
146 "tab in non-literal context",
147 "end of line whitespace",
149 "bad escape sequence",
150 "unterminated quoted string",
152 /* related to equations */
153 "unexpected literal in equation",
157 /* related to equations */
158 "unexpected equation scope closure",
159 "equation scope open on exit",
160 "overlapping equation scopes",
161 "unexpected end of equation",
162 "equation syntax error",
164 /* related to tables */
168 "no table layout cells specified",
169 "no table data cells specified",
170 "ignore data in cell",
171 "data block still open",
172 "ignoring extra data cells",
174 "input stack limit exceeded, infinite loop?",
175 "skipping bad character",
176 "escaped character not allowed in a name",
177 "skipping text before the first section header",
178 "skipping unknown macro",
179 "NOT IMPLEMENTED, please use groff: skipping request",
180 "argument count wrong",
181 "skipping end of block that is not open",
182 "missing end of block",
183 "scope open on exit",
184 "uname(3) system call failed",
185 "macro requires line argument(s)",
186 "macro requires body argument(s)",
187 "macro requires argument(s)",
189 "line argument(s) will be lost",
190 "body argument(s) will be lost",
192 "generic fatal error",
195 "column syntax is inconsistent",
196 "NOT IMPLEMENTED: .Bd -file",
197 "argument count wrong, violates syntax",
198 "child violates parent syntax",
199 "argument count wrong, violates syntax",
200 "NOT IMPLEMENTED: .so with absolute path or \"..\"",
202 "no document prologue",
203 "static buffer exhausted",
206 static const char * const mandoclevels
[MANDOCLEVEL_MAX
] = {
217 resize_buf(struct buf
*buf
, size_t initial
)
220 buf
->sz
= buf
->sz
> initial
/2 ? 2 * buf
->sz
: initial
;
221 buf
->buf
= mandoc_realloc(buf
->buf
, buf
->sz
);
225 pset(const char *buf
, int pos
, struct mparse
*curp
)
230 * Try to intuit which kind of manual parser should be used. If
231 * passed in by command-line (-man, -mdoc), then use that
232 * explicitly. If passed as -mandoc, then try to guess from the
233 * line: either skip dot-lines, use -mdoc when finding `.Dt', or
234 * default to -man, which is more lenient.
236 * Separate out pmdoc/pman from mdoc/man: the first persists
237 * through all parsers, while the latter is used per-parse.
240 if ('.' == buf
[0] || '\'' == buf
[0]) {
241 for (i
= 1; buf
[i
]; i
++)
242 if (' ' != buf
[i
] && '\t' != buf
[i
])
248 switch (curp
->inttype
) {
250 if (NULL
== curp
->pmdoc
)
251 curp
->pmdoc
= mdoc_alloc(curp
->roff
, curp
,
254 curp
->mdoc
= curp
->pmdoc
;
257 if (NULL
== curp
->pman
)
258 curp
->pman
= man_alloc(curp
->roff
, curp
);
260 curp
->man
= curp
->pman
;
266 if (pos
>= 3 && 0 == memcmp(buf
, ".Dd", 3)) {
267 if (NULL
== curp
->pmdoc
)
268 curp
->pmdoc
= mdoc_alloc(curp
->roff
, curp
,
271 curp
->mdoc
= curp
->pmdoc
;
275 if (NULL
== curp
->pman
)
276 curp
->pman
= man_alloc(curp
->roff
, curp
);
278 curp
->man
= curp
->pman
;
282 * Main parse routine for an opened file. This is called for each
283 * opened file and simply loops around the full input file, possibly
284 * nesting (i.e., with `so').
287 mparse_buf_r(struct mparse
*curp
, struct buf blk
, int start
)
289 const struct tbl_span
*span
;
293 int pos
; /* byte number in the ln buffer */
294 int lnn
; /* line number in the real file */
297 memset(&ln
, 0, sizeof(struct buf
));
302 for (i
= 0; i
< (int)blk
.sz
; ) {
303 if (0 == pos
&& '\0' == blk
.buf
[i
])
308 curp
->reparse_count
= 0;
311 while (i
< (int)blk
.sz
&& (start
|| '\0' != blk
.buf
[i
])) {
314 * When finding an unescaped newline character,
315 * leave the character loop to process the line.
316 * Skip a preceding carriage return, if any.
319 if ('\r' == blk
.buf
[i
] && i
+ 1 < (int)blk
.sz
&&
320 '\n' == blk
.buf
[i
+ 1])
322 if ('\n' == blk
.buf
[i
]) {
329 * Warn about bogus characters. If you're using
330 * non-ASCII encoding, you're screwing your
331 * readers. Since I'd rather this not happen,
332 * I'll be helpful and replace these characters
333 * with "?", so we don't display gibberish.
334 * Note to manual writers: use special characters.
337 c
= (unsigned char) blk
.buf
[i
];
339 if ( ! (isascii(c
) &&
340 (isgraph(c
) || isblank(c
)))) {
341 mandoc_msg(MANDOCERR_BADCHAR
, curp
,
342 curp
->line
, pos
, NULL
);
344 if (pos
>= (int)ln
.sz
)
345 resize_buf(&ln
, 256);
350 /* Trailing backslash = a plain char. */
352 if ('\\' != blk
.buf
[i
] || i
+ 1 == (int)blk
.sz
) {
353 if (pos
>= (int)ln
.sz
)
354 resize_buf(&ln
, 256);
355 ln
.buf
[pos
++] = blk
.buf
[i
++];
360 * Found escape and at least one other character.
361 * When it's a newline character, skip it.
362 * When there is a carriage return in between,
363 * skip that one as well.
366 if ('\r' == blk
.buf
[i
+ 1] && i
+ 2 < (int)blk
.sz
&&
367 '\n' == blk
.buf
[i
+ 2])
369 if ('\n' == blk
.buf
[i
+ 1]) {
375 if ('"' == blk
.buf
[i
+ 1] || '#' == blk
.buf
[i
+ 1]) {
377 /* Comment, skip to end of line */
378 for (; i
< (int)blk
.sz
; ++i
) {
379 if ('\n' == blk
.buf
[i
]) {
386 /* Backout trailing whitespaces */
387 for (; pos
> 0; --pos
) {
388 if (ln
.buf
[pos
- 1] != ' ')
390 if (pos
> 2 && ln
.buf
[pos
- 2] == '\\')
396 /* Some other escape sequence, copy & cont. */
398 if (pos
+ 1 >= (int)ln
.sz
)
399 resize_buf(&ln
, 256);
401 ln
.buf
[pos
++] = blk
.buf
[i
++];
402 ln
.buf
[pos
++] = blk
.buf
[i
++];
405 if (pos
>= (int)ln
.sz
)
406 resize_buf(&ln
, 256);
411 * A significant amount of complexity is contained by
412 * the roff preprocessor. It's line-oriented but can be
413 * expressed on one line, so we need at times to
414 * readjust our starting point and re-run it. The roff
415 * preprocessor can also readjust the buffers with new
416 * data, so we pass them in wholesale.
422 * Maintain a lookaside buffer of all parsed lines. We
423 * only do this if mparse_keep() has been invoked (the
424 * buffer may be accessed with mparse_getkeep()).
427 if (curp
->secondary
) {
428 curp
->secondary
->buf
=
430 (curp
->secondary
->buf
,
431 curp
->secondary
->sz
+ pos
+ 2);
432 memcpy(curp
->secondary
->buf
+
435 curp
->secondary
->sz
+= pos
;
437 [curp
->secondary
->sz
] = '\n';
438 curp
->secondary
->sz
++;
440 [curp
->secondary
->sz
] = '\0';
444 (curp
->roff
, curp
->line
,
445 &ln
.buf
, &ln
.sz
, of
, &of
);
449 if (REPARSE_LIMIT
>= ++curp
->reparse_count
)
450 mparse_buf_r(curp
, ln
, 0);
452 mandoc_msg(MANDOCERR_ROFFLOOP
, curp
,
453 curp
->line
, pos
, NULL
);
457 pos
= (int)strlen(ln
.buf
);
465 assert(MANDOCLEVEL_FATAL
<= curp
->file_status
);
469 * We remove `so' clauses from our lookaside
470 * buffer because we're going to descend into
471 * the file recursively.
474 curp
->secondary
->sz
-= pos
+ 1;
475 mparse_readfd_r(curp
, -1, ln
.buf
+ of
, 1);
476 if (MANDOCLEVEL_FATAL
<= curp
->file_status
)
485 * If we encounter errors in the recursive parse, make
486 * sure we don't continue parsing.
489 if (MANDOCLEVEL_FATAL
<= curp
->file_status
)
493 * If input parsers have not been allocated, do so now.
494 * We keep these instanced between parsers, but set them
495 * locally per parse routine since we can use different
496 * parsers with each one.
499 if ( ! (curp
->man
|| curp
->mdoc
))
500 pset(ln
.buf
+ of
, pos
- of
, curp
);
503 * Lastly, push down into the parsers themselves. One
504 * of these will have already been set in the pset()
506 * If libroff returns ROFF_TBL, then add it to the
507 * currently open parse. Since we only get here if
508 * there does exist data (see tbl_data.c), we're
509 * guaranteed that something's been allocated.
510 * Do the same for ROFF_EQN.
516 while (NULL
!= (span
= roff_span(curp
->roff
))) {
518 man_addspan(curp
->man
, span
) :
519 mdoc_addspan(curp
->mdoc
, span
);
523 else if (ROFF_EQN
== rr
)
525 mdoc_addeqn(curp
->mdoc
,
526 roff_eqn(curp
->roff
)) :
527 man_addeqn(curp
->man
,
528 roff_eqn(curp
->roff
));
529 else if (curp
->man
|| curp
->mdoc
)
531 man_parseln(curp
->man
,
532 curp
->line
, ln
.buf
, of
) :
533 mdoc_parseln(curp
->mdoc
,
534 curp
->line
, ln
.buf
, of
);
537 assert(MANDOCLEVEL_FATAL
<= curp
->file_status
);
541 /* Temporary buffers typically are not full. */
543 if (0 == start
&& '\0' == blk
.buf
[i
])
546 /* Start the next input line. */
555 read_whole_file(const char *file
, int fd
, struct buf
*fb
, int *with_mmap
)
562 if (-1 == fstat(fd
, &st
)) {
568 * If we're a regular file, try just reading in the whole entry
569 * via mmap(). This is faster than reading it into blocks, and
570 * since each file is only a few bytes to begin with, I'm not
571 * concerned that this is going to tank any machines.
574 if (S_ISREG(st
.st_mode
)) {
575 if (st
.st_size
>= (1U << 31)) {
576 fprintf(stderr
, "%s: input too large\n", file
);
580 fb
->sz
= (size_t)st
.st_size
;
581 fb
->buf
= mmap(NULL
, fb
->sz
, PROT_READ
,
582 MAP_FILE
|MAP_SHARED
, fd
, 0);
583 if (fb
->buf
!= MAP_FAILED
)
589 * If this isn't a regular file (like, say, stdin), then we must
590 * go the old way and just read things in bit by bit.
599 if (fb
->sz
== (1U << 31)) {
600 fprintf(stderr
, "%s: input too large\n", file
);
603 resize_buf(fb
, 65536);
605 ssz
= read(fd
, fb
->buf
+ (int)off
, fb
->sz
- off
);
623 mparse_end(struct mparse
*curp
)
626 if (MANDOCLEVEL_FATAL
<= curp
->file_status
)
629 if (curp
->mdoc
&& ! mdoc_endparse(curp
->mdoc
)) {
630 assert(MANDOCLEVEL_FATAL
<= curp
->file_status
);
634 if (curp
->man
&& ! man_endparse(curp
->man
)) {
635 assert(MANDOCLEVEL_FATAL
<= curp
->file_status
);
639 if ( ! (curp
->man
|| curp
->mdoc
)) {
640 mandoc_msg(MANDOCERR_NOTMANUAL
, curp
, 1, 0, NULL
);
641 curp
->file_status
= MANDOCLEVEL_FATAL
;
645 roff_endparse(curp
->roff
);
649 mparse_parse_buffer(struct mparse
*curp
, struct buf blk
, const char *file
,
654 /* Line number is per-file. */
659 mparse_buf_r(curp
, blk
, 1);
661 if (0 == re
&& MANDOCLEVEL_FATAL
> curp
->file_status
)
668 mparse_readmem(struct mparse
*curp
, const void *buf
, size_t len
,
673 blk
.buf
= UNCONST(buf
);
676 mparse_parse_buffer(curp
, blk
, file
, 0);
677 return(curp
->file_status
);
681 mparse_readfd_r(struct mparse
*curp
, int fd
, const char *file
, int re
)
687 if (-1 == (fd
= open(file
, O_RDONLY
, 0))) {
689 curp
->file_status
= MANDOCLEVEL_SYSERR
;
693 * Run for each opened file; may be called more than once for
694 * each full parse sequence if the opened file is nested (i.e.,
695 * from `so'). Simply sucks in the whole file and moves into
696 * the parse phase for the file.
699 if ( ! read_whole_file(file
, fd
, &blk
, &with_mmap
)) {
700 curp
->file_status
= MANDOCLEVEL_SYSERR
;
704 mparse_parse_buffer(curp
, blk
, file
, re
);
708 munmap(blk
.buf
, blk
.sz
);
713 if (STDIN_FILENO
!= fd
&& -1 == close(fd
))
718 mparse_readfd(struct mparse
*curp
, int fd
, const char *file
)
721 mparse_readfd_r(curp
, fd
, file
, 0);
722 return(curp
->file_status
);
726 mparse_alloc(enum mparset inttype
, enum mandoclevel wlevel
,
727 mandocmsg mmsg
, void *arg
, char *defos
)
731 assert(wlevel
<= MANDOCLEVEL_FATAL
);
733 curp
= mandoc_calloc(1, sizeof(struct mparse
));
735 curp
->wlevel
= wlevel
;
738 curp
->inttype
= inttype
;
741 curp
->roff
= roff_alloc(curp
);
746 mparse_reset(struct mparse
*curp
)
749 roff_reset(curp
->roff
);
752 mdoc_reset(curp
->mdoc
);
754 man_reset(curp
->man
);
756 curp
->secondary
->sz
= 0;
758 curp
->file_status
= MANDOCLEVEL_OK
;
764 mparse_free(struct mparse
*curp
)
768 mdoc_free(curp
->pmdoc
);
770 man_free(curp
->pman
);
772 roff_free(curp
->roff
);
774 free(curp
->secondary
->buf
);
776 free(curp
->secondary
);
781 mparse_result(struct mparse
*curp
, struct mdoc
**mdoc
, struct man
**man
)
791 mandoc_vmsg(enum mandocerr t
, struct mparse
*m
,
792 int ln
, int pos
, const char *fmt
, ...)
798 vsnprintf(buf
, sizeof(buf
) - 1, fmt
, ap
);
801 mandoc_msg(t
, m
, ln
, pos
, buf
);
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
])
814 if (level
< m
->wlevel
)
818 (*m
->mmsg
)(er
, level
, m
->file
, ln
, col
, msg
);
820 if (m
->file_status
< level
)
821 m
->file_status
= level
;
825 mparse_strerror(enum mandocerr er
)
828 return(mandocerrs
[er
]);
832 mparse_strlevel(enum mandoclevel lvl
)
834 return(mandoclevels
[lvl
]);
838 mparse_keep(struct mparse
*p
)
841 assert(NULL
== p
->secondary
);
842 p
->secondary
= mandoc_calloc(1, sizeof(struct buf
));
846 mparse_getkeep(const struct mparse
*p
)
849 assert(p
->secondary
);
850 return(p
->secondary
->sz
? p
->secondary
->buf
: NULL
);