]>
git.cameronkatri.com Git - mandoc.git/blob - read.c
1 /* $Id: read.c,v 1.152 2016/10/09 18:16:56 schwarze Exp $ */
3 * Copyright (c) 2008, 2009, 2010, 2011 Kristaps Dzonsons <kristaps@bsd.lv>
4 * Copyright (c) 2010-2016 Ingo Schwarze <schwarze@openbsd.org>
5 * Copyright (c) 2010, 2012 Joerg Sonnenberger <joerg@netbsd.org>
7 * Permission to use, copy, modify, and distribute this software for any
8 * purpose with or without fee is hereby granted, provided that the above
9 * copyright notice and this permission notice appear in all copies.
11 * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHORS DISCLAIM ALL WARRANTIES
12 * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
13 * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHORS BE LIABLE FOR
14 * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
15 * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
16 * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
17 * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
21 #include <sys/types.h>
40 #include "mandoc_aux.h"
45 #include "libmandoc.h"
48 #define REPARSE_LIMIT 1000
51 struct roff_man
*man
; /* man parser */
52 struct roff
*roff
; /* roff parser (!NULL) */
53 char *sodest
; /* filename pointed to by .so */
54 const char *file
; /* filename of current input file */
55 struct buf
*primary
; /* buffer currently being parsed */
56 struct buf
*secondary
; /* preprocessed copy of input */
57 const char *defos
; /* default operating system */
58 mandocmsg mmsg
; /* warning/error message handler */
59 enum mandoclevel file_status
; /* status of current parse */
60 enum mandoclevel wlevel
; /* ignore messages below this */
61 int options
; /* parser options */
62 int gzip
; /* current input file is gzipped */
63 int filenc
; /* encoding of the current file */
64 int reparse_count
; /* finite interp. stack */
65 int line
; /* line number in the file */
68 static void choose_parser(struct mparse
*);
69 static void resize_buf(struct buf
*, size_t);
70 static void mparse_buf_r(struct mparse
*, struct buf
, size_t, int);
71 static int read_whole_file(struct mparse
*, const char *, int,
73 static void mparse_end(struct mparse
*);
74 static void mparse_parse_buffer(struct mparse
*, struct buf
,
77 static const enum mandocerr mandoclimits
[MANDOCLEVEL_MAX
] = {
87 static const char * const mandocerrs
[MANDOCERR_MAX
] = {
92 /* related to the prologue */
93 "missing manual title, using UNTITLED",
94 "missing manual title, using \"\"",
95 "lower case character in document title",
96 "missing manual section, using \"\"",
97 "unknown manual section",
98 "missing date, using today's date",
99 "cannot parse date, using it verbatim",
100 "missing Os macro, using \"\"",
101 "duplicate prologue macro",
102 "late prologue macro",
103 "skipping late title macro",
104 "prologue macros out of order",
106 /* related to document structure */
107 ".so is fragile, better use ln(1)",
109 "content before first section header",
110 "first section is not \"NAME\"",
111 "NAME section without name",
112 "NAME section without description",
113 "description not at the end of NAME",
114 "bad NAME section content",
115 "missing description line, using \"\"",
116 "sections out of conventional order",
117 "duplicate section title",
118 "unexpected section",
120 "unusual Xr punctuation",
121 "AUTHORS section without An macro",
123 /* related to macros and nesting */
125 "macro neither callable nor escaped",
126 "skipping paragraph macro",
127 "moving paragraph macro out of list",
128 "skipping no-space macro",
129 "blocks badly nested",
130 "nested displays are not portable",
131 "moving content out of list",
132 "fill mode already enabled, skipping",
133 "fill mode already disabled, skipping",
136 /* related to missing macro arguments */
137 "skipping empty request",
138 "conditional request controls empty scope",
139 "skipping empty macro",
141 "empty argument, using 0n",
142 "missing display type, using -ragged",
143 "list type is not the first argument",
144 "missing -width in -tag list, using 6n",
145 "missing utility name, using \"\"",
146 "missing function name, using \"\"",
147 "empty head in list item",
149 "missing font type, using \\fR",
150 "unknown font type, using \\fR",
151 "nothing follows prefix",
152 "empty reference block",
153 "missing -std argument, adding it",
154 "missing option string, using \"\"",
155 "missing resource identifier, using \"\"",
156 "missing eqn box, using \"\"",
158 /* related to bad macro arguments */
159 "unterminated quoted argument",
160 "duplicate argument",
161 "skipping duplicate argument",
162 "skipping duplicate display type",
163 "skipping duplicate list type",
164 "skipping -width argument",
165 "wrong number of cells",
166 "unknown AT&T UNIX version",
167 "comma in function argument",
168 "parenthesis in function name",
169 "invalid content in Rs block",
170 "invalid Boolean argument",
171 "unknown font, skipping request",
172 "odd number of characters in request",
174 /* related to plain text */
175 "blank line in fill mode, using .sp",
176 "tab in filled text",
177 "whitespace at end of input line",
179 "invalid escape sequence",
180 "undefined string, using \"\"",
182 /* related to tables */
183 "tbl line starts with span",
184 "tbl column starts with span",
185 "skipping vertical bar in tbl layout",
189 /* related to tables */
190 "non-alphabetic character in tbl options",
191 "skipping unknown tbl option",
192 "missing tbl option argument",
193 "wrong tbl option argument size",
195 "invalid character in tbl layout",
196 "unmatched parenthesis in tbl layout",
197 "tbl without any data cells",
198 "ignoring data in spanned tbl cell",
199 "ignoring extra tbl data cells",
200 "data block open at end of tbl",
202 /* related to document structure and macros */
204 "input stack limit exceeded, infinite loop?",
205 "skipping bad character",
206 "skipping unknown macro",
207 "skipping insecure request",
208 "skipping item outside list",
209 "skipping column outside column list",
210 "skipping end of block that is not open",
211 "fewer RS blocks open, skipping",
212 "inserting missing end of block",
213 "appending missing end of block",
215 /* related to request and macro arguments */
216 "escaped character not allowed in a name",
217 "NOT IMPLEMENTED: Bd -file",
218 "skipping display without arguments",
219 "missing list type, using -item",
220 "missing manual name, using \"\"",
221 "uname(3) system call failed, using UNKNOWN",
222 "unknown standard specifier",
223 "skipping request without numeric argument",
224 "NOT IMPLEMENTED: .so with absolute path or \"..\"",
225 ".so request failed",
226 "skipping all arguments",
227 "skipping excess arguments",
230 "unsupported feature",
232 "unsupported control character",
233 "unsupported roff request",
234 "eqn delim option in tbl",
235 "unsupported tbl layout modifier",
236 "ignoring macro in table",
239 static const char * const mandoclevels
[MANDOCLEVEL_MAX
] = {
251 resize_buf(struct buf
*buf
, size_t initial
)
254 buf
->sz
= buf
->sz
> initial
/2 ? 2 * buf
->sz
: initial
;
255 buf
->buf
= mandoc_realloc(buf
->buf
, buf
->sz
);
259 choose_parser(struct mparse
*curp
)
265 * If neither command line arguments -mdoc or -man select
266 * a parser nor the roff parser found a .Dd or .TH macro
267 * yet, look ahead in the main input buffer.
270 if ((format
= roff_getformat(curp
->roff
)) == 0) {
271 cp
= curp
->primary
->buf
;
272 ep
= cp
+ curp
->primary
->sz
;
274 if (*cp
== '.' || *cp
== '\'') {
276 if (cp
[0] == 'D' && cp
[1] == 'd') {
277 format
= MPARSE_MDOC
;
280 if (cp
[0] == 'T' && cp
[1] == 'H') {
285 cp
= memchr(cp
, '\n', ep
- cp
);
292 if (format
== MPARSE_MDOC
) {
294 curp
->man
->macroset
= MACROSET_MDOC
;
295 curp
->man
->first
->tok
= TOKEN_NONE
;
298 curp
->man
->macroset
= MACROSET_MAN
;
299 curp
->man
->first
->tok
= TOKEN_NONE
;
304 * Main parse routine for a buffer.
305 * It assumes encoding and line numbering are already set up.
306 * It can recurse directly (for invocations of user-defined
307 * macros, inline equations, and input line traps)
308 * and indirectly (for .so file inclusion).
311 mparse_buf_r(struct mparse
*curp
, struct buf blk
, size_t i
, int start
)
313 const struct tbl_span
*span
;
315 const char *save_file
;
317 size_t pos
; /* byte number in the ln buffer */
320 int lnn
; /* line number in the real file */
324 memset(&ln
, 0, sizeof(ln
));
330 if (0 == pos
&& '\0' == blk
.buf
[i
])
335 curp
->reparse_count
= 0;
338 curp
->filenc
& MPARSE_UTF8
&&
339 curp
->filenc
& MPARSE_LATIN1
)
340 curp
->filenc
= preconv_cue(&blk
, i
);
343 while (i
< blk
.sz
&& (start
|| blk
.buf
[i
] != '\0')) {
346 * When finding an unescaped newline character,
347 * leave the character loop to process the line.
348 * Skip a preceding carriage return, if any.
351 if ('\r' == blk
.buf
[i
] && i
+ 1 < blk
.sz
&&
352 '\n' == blk
.buf
[i
+ 1])
354 if ('\n' == blk
.buf
[i
]) {
361 * Make sure we have space for the worst
362 * case of 11 bytes: "\\[u10ffff]\0"
365 if (pos
+ 11 > ln
.sz
)
366 resize_buf(&ln
, 256);
369 * Encode 8-bit input.
374 if ( ! (curp
->filenc
&& preconv_encode(
375 &blk
, &i
, &ln
, &pos
, &curp
->filenc
))) {
376 mandoc_vmsg(MANDOCERR_CHAR_BAD
, curp
,
377 curp
->line
, pos
, "0x%x", c
);
385 * Exclude control characters.
388 if (c
== 0x7f || (c
< 0x20 && c
!= 0x09)) {
389 mandoc_vmsg(c
== 0x00 || c
== 0x04 ||
390 c
> 0x0a ? MANDOCERR_CHAR_BAD
:
391 MANDOCERR_CHAR_UNSUPP
,
392 curp
, curp
->line
, pos
, "0x%x", c
);
399 /* Trailing backslash = a plain char. */
401 if (blk
.buf
[i
] != '\\' || i
+ 1 == blk
.sz
) {
402 ln
.buf
[pos
++] = blk
.buf
[i
++];
407 * Found escape and at least one other character.
408 * When it's a newline character, skip it.
409 * When there is a carriage return in between,
410 * skip that one as well.
413 if ('\r' == blk
.buf
[i
+ 1] && i
+ 2 < blk
.sz
&&
414 '\n' == blk
.buf
[i
+ 2])
416 if ('\n' == blk
.buf
[i
+ 1]) {
422 if ('"' == blk
.buf
[i
+ 1] || '#' == blk
.buf
[i
+ 1]) {
424 /* Comment, skip to end of line */
425 for (; i
< blk
.sz
; ++i
) {
426 if ('\n' == blk
.buf
[i
]) {
433 /* Backout trailing whitespaces */
434 for (; pos
> 0; --pos
) {
435 if (ln
.buf
[pos
- 1] != ' ')
437 if (pos
> 2 && ln
.buf
[pos
- 2] == '\\')
443 /* Catch escaped bogus characters. */
445 c
= (unsigned char) blk
.buf
[i
+1];
447 if ( ! (isascii(c
) &&
448 (isgraph(c
) || isblank(c
)))) {
449 mandoc_vmsg(MANDOCERR_CHAR_BAD
, curp
,
450 curp
->line
, pos
, "0x%x", c
);
456 /* Some other escape sequence, copy & cont. */
458 ln
.buf
[pos
++] = blk
.buf
[i
++];
459 ln
.buf
[pos
++] = blk
.buf
[i
++];
463 resize_buf(&ln
, 256);
468 * A significant amount of complexity is contained by
469 * the roff preprocessor. It's line-oriented but can be
470 * expressed on one line, so we need at times to
471 * readjust our starting point and re-run it. The roff
472 * preprocessor can also readjust the buffers with new
473 * data, so we pass them in wholesale.
479 * Maintain a lookaside buffer of all parsed lines. We
480 * only do this if mparse_keep() has been invoked (the
481 * buffer may be accessed with mparse_getkeep()).
484 if (curp
->secondary
) {
485 curp
->secondary
->buf
= mandoc_realloc(
486 curp
->secondary
->buf
,
487 curp
->secondary
->sz
+ pos
+ 2);
488 memcpy(curp
->secondary
->buf
+
491 curp
->secondary
->sz
+= pos
;
493 [curp
->secondary
->sz
] = '\n';
494 curp
->secondary
->sz
++;
496 [curp
->secondary
->sz
] = '\0';
499 rr
= roff_parseln(curp
->roff
, curp
->line
, &ln
, &of
);
503 if (REPARSE_LIMIT
>= ++curp
->reparse_count
)
504 mparse_buf_r(curp
, ln
, of
, 0);
506 mandoc_msg(MANDOCERR_ROFFLOOP
, curp
,
507 curp
->line
, pos
, NULL
);
511 pos
= strlen(ln
.buf
);
519 if ( ! (curp
->options
& MPARSE_SO
) &&
520 (i
>= blk
.sz
|| blk
.buf
[i
] == '\0')) {
521 curp
->sodest
= mandoc_strdup(ln
.buf
+ of
);
526 * We remove `so' clauses from our lookaside
527 * buffer because we're going to descend into
528 * the file recursively.
531 curp
->secondary
->sz
-= pos
+ 1;
532 save_file
= curp
->file
;
533 if ((fd
= mparse_open(curp
, ln
.buf
+ of
)) != -1) {
534 mparse_readfd(curp
, fd
, ln
.buf
+ of
);
536 curp
->file
= save_file
;
538 curp
->file
= save_file
;
539 mandoc_vmsg(MANDOCERR_SO_FAIL
,
540 curp
, curp
->line
, pos
,
541 ".so %s", ln
.buf
+ of
);
542 ln
.sz
= mandoc_asprintf(&cp
,
543 ".sp\nSee the file %s.\n.sp",
548 mparse_buf_r(curp
, ln
, of
, 0);
556 if (curp
->man
->macroset
== MACROSET_NONE
)
560 * Lastly, push down into the parsers themselves.
561 * If libroff returns ROFF_TBL, then add it to the
562 * currently open parse. Since we only get here if
563 * there does exist data (see tbl_data.c), we're
564 * guaranteed that something's been allocated.
565 * Do the same for ROFF_EQN.
569 while ((span
= roff_span(curp
->roff
)) != NULL
)
570 roff_addtbl(curp
->man
, span
);
571 else if (rr
== ROFF_EQN
)
572 roff_addeqn(curp
->man
, roff_eqn(curp
->roff
));
573 else if ((curp
->man
->macroset
== MACROSET_MDOC
?
574 mdoc_parseln(curp
->man
, curp
->line
, ln
.buf
, of
) :
575 man_parseln(curp
->man
, curp
->line
, ln
.buf
, of
)) == 2)
578 /* Temporary buffers typically are not full. */
580 if (0 == start
&& '\0' == blk
.buf
[i
])
583 /* Start the next input line. */
592 read_whole_file(struct mparse
*curp
, const char *file
, int fd
,
593 struct buf
*fb
, int *with_mmap
)
601 if (fstat(fd
, &st
) == -1)
602 err((int)MANDOCLEVEL_SYSERR
, "%s", file
);
605 * If we're a regular file, try just reading in the whole entry
606 * via mmap(). This is faster than reading it into blocks, and
607 * since each file is only a few bytes to begin with, I'm not
608 * concerned that this is going to tank any machines.
611 if (curp
->gzip
== 0 && S_ISREG(st
.st_mode
)) {
612 if (st
.st_size
> 0x7fffffff) {
613 mandoc_msg(MANDOCERR_TOOLARGE
, curp
, 0, 0, NULL
);
617 fb
->sz
= (size_t)st
.st_size
;
618 fb
->buf
= mmap(NULL
, fb
->sz
, PROT_READ
, MAP_SHARED
, fd
, 0);
619 if (fb
->buf
!= MAP_FAILED
)
624 if ((gz
= gzdopen(fd
, "rb")) == NULL
)
625 err((int)MANDOCLEVEL_SYSERR
, "%s", file
);
630 * If this isn't a regular file (like, say, stdin), then we must
631 * go the old way and just read things in bit by bit.
640 if (fb
->sz
== (1U << 31)) {
641 mandoc_msg(MANDOCERR_TOOLARGE
, curp
,
645 resize_buf(fb
, 65536);
648 gzread(gz
, fb
->buf
+ (int)off
, fb
->sz
- off
) :
649 read(fd
, fb
->buf
+ (int)off
, fb
->sz
- off
);
655 err((int)MANDOCLEVEL_SYSERR
, "%s", file
);
665 mparse_end(struct mparse
*curp
)
667 if (curp
->man
->macroset
== MACROSET_NONE
)
668 curp
->man
->macroset
= MACROSET_MAN
;
669 if (curp
->man
->macroset
== MACROSET_MDOC
)
670 mdoc_endparse(curp
->man
);
672 man_endparse(curp
->man
);
673 roff_endparse(curp
->roff
);
677 mparse_parse_buffer(struct mparse
*curp
, struct buf blk
, const char *file
)
679 struct buf
*svprimary
;
682 static int recursion_depth
;
684 if (64 < recursion_depth
) {
685 mandoc_msg(MANDOCERR_ROFFLOOP
, curp
, curp
->line
, 0, NULL
);
689 /* Line number is per-file. */
692 svprimary
= curp
->primary
;
693 curp
->primary
= &blk
;
697 /* Skip an UTF-8 byte order mark. */
698 if (curp
->filenc
& MPARSE_UTF8
&& blk
.sz
> 2 &&
699 (unsigned char)blk
.buf
[0] == 0xef &&
700 (unsigned char)blk
.buf
[1] == 0xbb &&
701 (unsigned char)blk
.buf
[2] == 0xbf) {
703 curp
->filenc
&= ~MPARSE_LATIN1
;
707 mparse_buf_r(curp
, blk
, offset
, 1);
709 if (--recursion_depth
== 0)
712 curp
->primary
= svprimary
;
717 mparse_readmem(struct mparse
*curp
, void *buf
, size_t len
,
725 mparse_parse_buffer(curp
, blk
, file
);
726 return curp
->file_status
;
730 * Read the whole file into memory and call the parsers.
731 * Called recursively when an .so request is encountered.
734 mparse_readfd(struct mparse
*curp
, int fd
, const char *file
)
740 if (read_whole_file(curp
, file
, fd
, &blk
, &with_mmap
)) {
741 save_filenc
= curp
->filenc
;
742 curp
->filenc
= curp
->options
&
743 (MPARSE_UTF8
| MPARSE_LATIN1
);
744 mparse_parse_buffer(curp
, blk
, file
);
745 curp
->filenc
= save_filenc
;
747 munmap(blk
.buf
, blk
.sz
);
751 return curp
->file_status
;
755 mparse_open(struct mparse
*curp
, const char *file
)
761 cp
= strrchr(file
, '.');
762 curp
->gzip
= (cp
!= NULL
&& ! strcmp(cp
+ 1, "gz"));
764 /* First try to use the filename as it is. */
766 if ((fd
= open(file
, O_RDONLY
)) != -1)
770 * If that doesn't work and the filename doesn't
771 * already end in .gz, try appending .gz.
775 mandoc_asprintf(&cp
, "%s.gz", file
);
776 fd
= open(cp
, O_RDONLY
);
784 /* Neither worked, give up. */
786 mandoc_msg(MANDOCERR_FILE
, curp
, 0, 0, strerror(errno
));
791 mparse_alloc(int options
, enum mandoclevel wlevel
, mandocmsg mmsg
,
796 curp
= mandoc_calloc(1, sizeof(struct mparse
));
798 curp
->options
= options
;
799 curp
->wlevel
= wlevel
;
803 curp
->roff
= roff_alloc(curp
, options
);
804 curp
->man
= roff_man_alloc( curp
->roff
, curp
, curp
->defos
,
805 curp
->options
& MPARSE_QUICK
? 1 : 0);
806 if (curp
->options
& MPARSE_MDOC
) {
808 curp
->man
->macroset
= MACROSET_MDOC
;
809 } else if (curp
->options
& MPARSE_MAN
) {
811 curp
->man
->macroset
= MACROSET_MAN
;
813 curp
->man
->first
->tok
= TOKEN_NONE
;
818 mparse_reset(struct mparse
*curp
)
820 roff_reset(curp
->roff
);
821 roff_man_reset(curp
->man
);
823 curp
->secondary
->sz
= 0;
825 curp
->file_status
= MANDOCLEVEL_OK
;
832 mparse_free(struct mparse
*curp
)
835 roff_man_free(curp
->man
);
837 roff_free(curp
->roff
);
839 free(curp
->secondary
->buf
);
841 free(curp
->secondary
);
847 mparse_result(struct mparse
*curp
, struct roff_man
**man
,
851 if (sodest
&& NULL
!= (*sodest
= curp
->sodest
)) {
860 mandoc_vmsg(enum mandocerr t
, struct mparse
*m
,
861 int ln
, int pos
, const char *fmt
, ...)
867 (void)vsnprintf(buf
, sizeof(buf
), fmt
, ap
);
870 mandoc_msg(t
, m
, ln
, pos
, buf
);
874 mandoc_msg(enum mandocerr er
, struct mparse
*m
,
875 int ln
, int col
, const char *msg
)
877 enum mandoclevel level
;
879 level
= MANDOCLEVEL_UNSUPP
;
880 while (er
< mandoclimits
[level
])
883 if (level
< m
->wlevel
&& er
!= MANDOCERR_FILE
)
887 (*m
->mmsg
)(er
, level
, m
->file
, ln
, col
, msg
);
889 if (m
->file_status
< level
)
890 m
->file_status
= level
;
894 mparse_strerror(enum mandocerr er
)
897 return mandocerrs
[er
];
901 mparse_strlevel(enum mandoclevel lvl
)
903 return mandoclevels
[lvl
];
907 mparse_keep(struct mparse
*p
)
910 assert(NULL
== p
->secondary
);
911 p
->secondary
= mandoc_calloc(1, sizeof(struct buf
));
915 mparse_getkeep(const struct mparse
*p
)
918 assert(p
->secondary
);
919 return p
->secondary
->sz
? p
->secondary
->buf
: NULL
;