Annotation of mandoc/read.c, Revision 1.44
1.44 ! schwarze 1: /* $Id: read.c,v 1.43 2014/01/06 21:34:31 schwarze Exp $ */
1.1 kristaps 2: /*
3: * Copyright (c) 2008, 2009, 2010, 2011 Kristaps Dzonsons <kristaps@bsd.lv>
1.40 schwarze 4: * Copyright (c) 2010-2014 Ingo Schwarze <schwarze@openbsd.org>
1.42 schwarze 5: * Copyright (c) 2010, 2012 Joerg Sonnenberger <joerg@netbsd.org>
1.1 kristaps 6: *
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.
10: *
11: * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
12: * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
13: * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR 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.
18: */
1.11 kristaps 19: #ifdef HAVE_CONFIG_H
20: #include "config.h"
21: #endif
22:
1.15 kristaps 23: #ifdef HAVE_MMAP
24: # include <sys/stat.h>
25: # include <sys/mman.h>
26: #endif
1.1 kristaps 27:
28: #include <assert.h>
29: #include <ctype.h>
1.40 schwarze 30: #include <errno.h>
1.1 kristaps 31: #include <fcntl.h>
1.3 kristaps 32: #include <stdarg.h>
1.28 joerg 33: #include <stdint.h>
1.1 kristaps 34: #include <stdio.h>
35: #include <stdlib.h>
36: #include <string.h>
37: #include <unistd.h>
38:
39: #include "mandoc.h"
1.3 kristaps 40: #include "libmandoc.h"
1.1 kristaps 41: #include "mdoc.h"
42: #include "man.h"
1.28 joerg 43: #include "main.h"
1.1 kristaps 44:
45: #define REPARSE_LIMIT 1000
46:
47: struct buf {
48: char *buf; /* binary input buffer */
49: size_t sz; /* size of binary buffer */
50: };
51:
52: struct mparse {
53: enum mandoclevel file_status; /* status of current parse */
1.3 kristaps 54: enum mandoclevel wlevel; /* ignore messages below this */
1.1 kristaps 55: int line; /* line number in the file */
1.44 ! schwarze 56: int options; /* parser options */
1.1 kristaps 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 */
1.3 kristaps 64: const char *file;
1.24 kristaps 65: struct buf *secondary;
1.29 schwarze 66: char *defos; /* default operating system */
1.1 kristaps 67: };
68:
69: static void resize_buf(struct buf *, size_t);
70: static void mparse_buf_r(struct mparse *, struct buf, int);
71: static void pset(const char *, int, struct mparse *);
1.40 schwarze 72: static int read_whole_file(struct mparse *, const char *, int,
73: struct buf *, int *);
1.1 kristaps 74: static void mparse_end(struct mparse *);
1.37 schwarze 75: static void mparse_parse_buffer(struct mparse *, struct buf,
76: const char *);
1.1 kristaps 77:
1.3 kristaps 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
86: };
87:
1.7 kristaps 88: static const char * const mandocerrs[MANDOCERR_MAX] = {
89: "ok",
90:
91: "generic warning",
92:
93: /* related to the prologue */
94: "no title in document",
95: "document title should be all caps",
96: "unknown manual section",
1.32 schwarze 97: "unknown manual volume or arch",
1.7 kristaps 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",
104:
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: "sections out of conventional order",
110: "duplicate section name",
1.39 schwarze 111: "section header suited to sections 2, 3, and 9 only",
1.7 kristaps 112:
113: /* related to macros and nesting */
114: "skipping obsolete macro",
115: "skipping paragraph macro",
1.31 schwarze 116: "moving paragraph macro out of list",
1.7 kristaps 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",
122: "line scope broken",
123:
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",
130: "missing font type",
131: "skipping end of block that is not open",
132:
133: /* related to bad macro arguments */
134: "skipping argument",
135: "duplicate argument",
136: "duplicate display type",
137: "duplicate list type",
138: "unknown AT&T UNIX version",
139: "bad Boolean value",
140: "unknown font",
141: "unknown standard specifier",
142: "bad width argument",
143:
144: /* related to plain text */
145: "blank line in non-literal context",
146: "tab in non-literal context",
147: "end of line whitespace",
148: "bad comment style",
1.12 kristaps 149: "bad escape sequence",
1.7 kristaps 150: "unterminated quoted string",
1.16 kristaps 151:
152: /* related to equations */
153: "unexpected literal in equation",
1.7 kristaps 154:
155: "generic error",
1.17 kristaps 156:
157: /* related to equations */
1.20 kristaps 158: "unexpected equation scope closure",
159: "equation scope open on exit",
1.21 kristaps 160: "overlapping equation scopes",
161: "unexpected end of equation",
162: "equation syntax error",
1.7 kristaps 163:
164: /* related to tables */
165: "bad table syntax",
166: "bad table option",
167: "bad table layout",
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",
173:
174: "input stack limit exceeded, infinite loop?",
175: "skipping bad character",
176: "escaped character not allowed in a name",
1.30 schwarze 177: "manual name not yet set",
1.7 kristaps 178: "skipping text before the first section header",
179: "skipping unknown macro",
180: "NOT IMPLEMENTED, please use groff: skipping request",
181: "argument count wrong",
1.34 schwarze 182: "skipping column outside column list",
1.7 kristaps 183: "skipping end of block that is not open",
184: "missing end of block",
185: "scope open on exit",
186: "uname(3) system call failed",
187: "macro requires line argument(s)",
188: "macro requires body argument(s)",
189: "macro requires argument(s)",
1.38 schwarze 190: "request requires a numeric argument",
1.7 kristaps 191: "missing list type",
192: "line argument(s) will be lost",
193: "body argument(s) will be lost",
194:
195: "generic fatal error",
196:
1.40 schwarze 197: "input too large",
1.7 kristaps 198: "not a manual",
199: "column syntax is inconsistent",
200: "NOT IMPLEMENTED: .Bd -file",
201: "argument count wrong, violates syntax",
202: "child violates parent syntax",
203: "argument count wrong, violates syntax",
204: "NOT IMPLEMENTED: .so with absolute path or \"..\"",
205: "no document body",
206: "no document prologue",
207: "static buffer exhausted",
1.40 schwarze 208:
209: /* system errors */
210: "cannot open file",
211: "cannot stat file",
212: "cannot read file",
1.7 kristaps 213: };
214:
215: static const char * const mandoclevels[MANDOCLEVEL_MAX] = {
216: "SUCCESS",
217: "RESERVED",
218: "WARNING",
219: "ERROR",
220: "FATAL",
221: "BADARG",
222: "SYSERR"
223: };
224:
1.1 kristaps 225: static void
226: resize_buf(struct buf *buf, size_t initial)
227: {
228:
229: buf->sz = buf->sz > initial/2 ? 2 * buf->sz : initial;
230: buf->buf = mandoc_realloc(buf->buf, buf->sz);
231: }
232:
233: static void
234: pset(const char *buf, int pos, struct mparse *curp)
235: {
236: int i;
237:
238: /*
239: * Try to intuit which kind of manual parser should be used. If
240: * passed in by command-line (-man, -mdoc), then use that
241: * explicitly. If passed as -mandoc, then try to guess from the
242: * line: either skip dot-lines, use -mdoc when finding `.Dt', or
243: * default to -man, which is more lenient.
244: *
245: * Separate out pmdoc/pman from mdoc/man: the first persists
246: * through all parsers, while the latter is used per-parse.
247: */
248:
249: if ('.' == buf[0] || '\'' == buf[0]) {
250: for (i = 1; buf[i]; i++)
251: if (' ' != buf[i] && '\t' != buf[i])
252: break;
253: if ('\0' == buf[i])
254: return;
255: }
256:
1.44 ! schwarze 257: if (MPARSE_MDOC & curp->options) {
1.1 kristaps 258: if (NULL == curp->pmdoc)
1.44 ! schwarze 259: curp->pmdoc = mdoc_alloc(
! 260: curp->roff, curp, curp->defos,
! 261: MPARSE_QUICK & curp->options ? 1 : 0);
1.1 kristaps 262: assert(curp->pmdoc);
263: curp->mdoc = curp->pmdoc;
264: return;
1.44 ! schwarze 265: } else if (MPARSE_MAN & curp->options) {
1.1 kristaps 266: if (NULL == curp->pman)
1.41 schwarze 267: curp->pman = man_alloc(curp->roff, curp,
1.44 ! schwarze 268: MPARSE_QUICK & curp->options ? 1 : 0);
1.1 kristaps 269: assert(curp->pman);
270: curp->man = curp->pman;
271: return;
272: }
273:
274: if (pos >= 3 && 0 == memcmp(buf, ".Dd", 3)) {
275: if (NULL == curp->pmdoc)
1.44 ! schwarze 276: curp->pmdoc = mdoc_alloc(
! 277: curp->roff, curp, curp->defos,
! 278: MPARSE_QUICK & curp->options ? 1 : 0);
1.1 kristaps 279: assert(curp->pmdoc);
280: curp->mdoc = curp->pmdoc;
281: return;
282: }
283:
284: if (NULL == curp->pman)
1.44 ! schwarze 285: curp->pman = man_alloc(curp->roff, curp,
! 286: MPARSE_QUICK & curp->options ? 1 : 0);
1.1 kristaps 287: assert(curp->pman);
288: curp->man = curp->pman;
289: }
290:
291: /*
292: * Main parse routine for an opened file. This is called for each
293: * opened file and simply loops around the full input file, possibly
294: * nesting (i.e., with `so').
295: */
296: static void
297: mparse_buf_r(struct mparse *curp, struct buf blk, int start)
298: {
299: const struct tbl_span *span;
300: struct buf ln;
301: enum rofferr rr;
302: int i, of, rc;
303: int pos; /* byte number in the ln buffer */
304: int lnn; /* line number in the real file */
305: unsigned char c;
306:
307: memset(&ln, 0, sizeof(struct buf));
308:
309: lnn = curp->line;
310: pos = 0;
311:
312: for (i = 0; i < (int)blk.sz; ) {
313: if (0 == pos && '\0' == blk.buf[i])
314: break;
315:
316: if (start) {
317: curp->line = lnn;
318: curp->reparse_count = 0;
319: }
320:
321: while (i < (int)blk.sz && (start || '\0' != blk.buf[i])) {
322:
323: /*
324: * When finding an unescaped newline character,
325: * leave the character loop to process the line.
326: * Skip a preceding carriage return, if any.
327: */
328:
329: if ('\r' == blk.buf[i] && i + 1 < (int)blk.sz &&
330: '\n' == blk.buf[i + 1])
331: ++i;
332: if ('\n' == blk.buf[i]) {
333: ++i;
334: ++lnn;
335: break;
336: }
337:
1.35 schwarze 338: /*
339: * Make sure we have space for at least
340: * one backslash and one other character
341: * and the trailing NUL byte.
342: */
343:
344: if (pos + 2 >= (int)ln.sz)
345: resize_buf(&ln, 256);
346:
1.1 kristaps 347: /*
348: * Warn about bogus characters. If you're using
349: * non-ASCII encoding, you're screwing your
350: * readers. Since I'd rather this not happen,
1.27 joerg 351: * I'll be helpful and replace these characters
352: * with "?", so we don't display gibberish.
353: * Note to manual writers: use special characters.
1.1 kristaps 354: */
355:
356: c = (unsigned char) blk.buf[i];
357:
358: if ( ! (isascii(c) &&
359: (isgraph(c) || isblank(c)))) {
1.3 kristaps 360: mandoc_msg(MANDOCERR_BADCHAR, curp,
1.27 joerg 361: curp->line, pos, NULL);
1.1 kristaps 362: i++;
1.27 joerg 363: ln.buf[pos++] = '?';
1.1 kristaps 364: continue;
365: }
366:
367: /* Trailing backslash = a plain char. */
368:
369: if ('\\' != blk.buf[i] || i + 1 == (int)blk.sz) {
370: ln.buf[pos++] = blk.buf[i++];
371: continue;
372: }
373:
374: /*
375: * Found escape and at least one other character.
376: * When it's a newline character, skip it.
377: * When there is a carriage return in between,
378: * skip that one as well.
379: */
380:
381: if ('\r' == blk.buf[i + 1] && i + 2 < (int)blk.sz &&
382: '\n' == blk.buf[i + 2])
383: ++i;
384: if ('\n' == blk.buf[i + 1]) {
385: i += 2;
386: ++lnn;
387: continue;
388: }
389:
1.13 kristaps 390: if ('"' == blk.buf[i + 1] || '#' == blk.buf[i + 1]) {
1.1 kristaps 391: i += 2;
392: /* Comment, skip to end of line */
393: for (; i < (int)blk.sz; ++i) {
394: if ('\n' == blk.buf[i]) {
395: ++i;
396: ++lnn;
397: break;
398: }
399: }
400:
401: /* Backout trailing whitespaces */
402: for (; pos > 0; --pos) {
403: if (ln.buf[pos - 1] != ' ')
404: break;
405: if (pos > 2 && ln.buf[pos - 2] == '\\')
406: break;
407: }
408: break;
409: }
410:
1.35 schwarze 411: /* Catch escaped bogus characters. */
412:
413: c = (unsigned char) blk.buf[i+1];
414:
415: if ( ! (isascii(c) &&
416: (isgraph(c) || isblank(c)))) {
417: mandoc_msg(MANDOCERR_BADCHAR, curp,
418: curp->line, pos, NULL);
419: i += 2;
420: ln.buf[pos++] = '?';
421: continue;
422: }
423:
1.1 kristaps 424: /* Some other escape sequence, copy & cont. */
425:
426: ln.buf[pos++] = blk.buf[i++];
427: ln.buf[pos++] = blk.buf[i++];
428: }
429:
430: if (pos >= (int)ln.sz)
431: resize_buf(&ln, 256);
432:
433: ln.buf[pos] = '\0';
434:
435: /*
436: * A significant amount of complexity is contained by
437: * the roff preprocessor. It's line-oriented but can be
438: * expressed on one line, so we need at times to
439: * readjust our starting point and re-run it. The roff
440: * preprocessor can also readjust the buffers with new
441: * data, so we pass them in wholesale.
442: */
443:
444: of = 0;
445:
1.24 kristaps 446: /*
447: * Maintain a lookaside buffer of all parsed lines. We
448: * only do this if mparse_keep() has been invoked (the
449: * buffer may be accessed with mparse_getkeep()).
450: */
451:
452: if (curp->secondary) {
453: curp->secondary->buf =
454: mandoc_realloc
455: (curp->secondary->buf,
456: curp->secondary->sz + pos + 2);
457: memcpy(curp->secondary->buf +
458: curp->secondary->sz,
459: ln.buf, pos);
460: curp->secondary->sz += pos;
461: curp->secondary->buf
462: [curp->secondary->sz] = '\n';
463: curp->secondary->sz++;
464: curp->secondary->buf
465: [curp->secondary->sz] = '\0';
466: }
1.1 kristaps 467: rerun:
468: rr = roff_parseln
469: (curp->roff, curp->line,
470: &ln.buf, &ln.sz, of, &of);
471:
472: switch (rr) {
473: case (ROFF_REPARSE):
474: if (REPARSE_LIMIT >= ++curp->reparse_count)
475: mparse_buf_r(curp, ln, 0);
476: else
1.3 kristaps 477: mandoc_msg(MANDOCERR_ROFFLOOP, curp,
1.1 kristaps 478: curp->line, pos, NULL);
479: pos = 0;
480: continue;
481: case (ROFF_APPEND):
482: pos = (int)strlen(ln.buf);
483: continue;
484: case (ROFF_RERUN):
485: goto rerun;
486: case (ROFF_IGN):
487: pos = 0;
488: continue;
489: case (ROFF_ERR):
490: assert(MANDOCLEVEL_FATAL <= curp->file_status);
491: break;
492: case (ROFF_SO):
1.24 kristaps 493: /*
494: * We remove `so' clauses from our lookaside
495: * buffer because we're going to descend into
496: * the file recursively.
497: */
1.25 kristaps 498: if (curp->secondary)
499: curp->secondary->sz -= pos + 1;
1.36 schwarze 500: mparse_readfd(curp, -1, ln.buf + of);
1.1 kristaps 501: if (MANDOCLEVEL_FATAL <= curp->file_status)
502: break;
503: pos = 0;
504: continue;
505: default:
506: break;
507: }
508:
509: /*
510: * If we encounter errors in the recursive parse, make
511: * sure we don't continue parsing.
512: */
513:
514: if (MANDOCLEVEL_FATAL <= curp->file_status)
515: break;
516:
517: /*
518: * If input parsers have not been allocated, do so now.
1.14 kristaps 519: * We keep these instanced between parsers, but set them
1.1 kristaps 520: * locally per parse routine since we can use different
521: * parsers with each one.
522: */
523:
524: if ( ! (curp->man || curp->mdoc))
525: pset(ln.buf + of, pos - of, curp);
526:
527: /*
528: * Lastly, push down into the parsers themselves. One
529: * of these will have already been set in the pset()
530: * routine.
531: * If libroff returns ROFF_TBL, then add it to the
532: * currently open parse. Since we only get here if
533: * there does exist data (see tbl_data.c), we're
534: * guaranteed that something's been allocated.
535: * Do the same for ROFF_EQN.
536: */
537:
538: rc = -1;
539:
540: if (ROFF_TBL == rr)
541: while (NULL != (span = roff_span(curp->roff))) {
542: rc = curp->man ?
543: man_addspan(curp->man, span) :
544: mdoc_addspan(curp->mdoc, span);
545: if (0 == rc)
546: break;
547: }
548: else if (ROFF_EQN == rr)
549: rc = curp->mdoc ?
550: mdoc_addeqn(curp->mdoc,
551: roff_eqn(curp->roff)) :
552: man_addeqn(curp->man,
553: roff_eqn(curp->roff));
554: else if (curp->man || curp->mdoc)
555: rc = curp->man ?
556: man_parseln(curp->man,
557: curp->line, ln.buf, of) :
558: mdoc_parseln(curp->mdoc,
559: curp->line, ln.buf, of);
560:
561: if (0 == rc) {
562: assert(MANDOCLEVEL_FATAL <= curp->file_status);
563: break;
1.41 schwarze 564: } else if (2 == rc)
565: break;
1.1 kristaps 566:
567: /* Temporary buffers typically are not full. */
568:
569: if (0 == start && '\0' == blk.buf[i])
570: break;
571:
572: /* Start the next input line. */
573:
574: pos = 0;
575: }
576:
577: free(ln.buf);
578: }
579:
580: static int
1.40 schwarze 581: read_whole_file(struct mparse *curp, const char *file, int fd,
582: struct buf *fb, int *with_mmap)
1.1 kristaps 583: {
584: size_t off;
585: ssize_t ssz;
586:
1.15 kristaps 587: #ifdef HAVE_MMAP
588: struct stat st;
1.1 kristaps 589: if (-1 == fstat(fd, &st)) {
1.40 schwarze 590: curp->file_status = MANDOCLEVEL_SYSERR;
591: if (curp->mmsg)
592: (*curp->mmsg)(MANDOCERR_SYSSTAT, curp->file_status,
593: file, 0, 0, strerror(errno));
1.1 kristaps 594: return(0);
595: }
596:
597: /*
598: * If we're a regular file, try just reading in the whole entry
599: * via mmap(). This is faster than reading it into blocks, and
600: * since each file is only a few bytes to begin with, I'm not
601: * concerned that this is going to tank any machines.
602: */
603:
604: if (S_ISREG(st.st_mode)) {
605: if (st.st_size >= (1U << 31)) {
1.40 schwarze 606: curp->file_status = MANDOCLEVEL_FATAL;
607: if (curp->mmsg)
608: (*curp->mmsg)(MANDOCERR_TOOLARGE,
609: curp->file_status, file, 0, 0, NULL);
1.1 kristaps 610: return(0);
611: }
612: *with_mmap = 1;
613: fb->sz = (size_t)st.st_size;
1.37 schwarze 614: fb->buf = mmap(NULL, fb->sz, PROT_READ, MAP_SHARED, fd, 0);
1.1 kristaps 615: if (fb->buf != MAP_FAILED)
616: return(1);
617: }
1.15 kristaps 618: #endif
1.1 kristaps 619:
620: /*
621: * If this isn't a regular file (like, say, stdin), then we must
622: * go the old way and just read things in bit by bit.
623: */
624:
625: *with_mmap = 0;
626: off = 0;
627: fb->sz = 0;
628: fb->buf = NULL;
629: for (;;) {
630: if (off == fb->sz) {
631: if (fb->sz == (1U << 31)) {
1.40 schwarze 632: curp->file_status = MANDOCLEVEL_FATAL;
633: if (curp->mmsg)
634: (*curp->mmsg)(MANDOCERR_TOOLARGE,
635: curp->file_status,
636: file, 0, 0, NULL);
1.1 kristaps 637: break;
638: }
639: resize_buf(fb, 65536);
640: }
641: ssz = read(fd, fb->buf + (int)off, fb->sz - off);
642: if (ssz == 0) {
643: fb->sz = off;
644: return(1);
645: }
646: if (ssz == -1) {
1.40 schwarze 647: curp->file_status = MANDOCLEVEL_SYSERR;
648: if (curp->mmsg)
649: (*curp->mmsg)(MANDOCERR_SYSREAD,
650: curp->file_status, file, 0, 0,
651: strerror(errno));
1.1 kristaps 652: break;
653: }
654: off += (size_t)ssz;
655: }
656:
657: free(fb->buf);
658: fb->buf = NULL;
659: return(0);
660: }
661:
662: static void
663: mparse_end(struct mparse *curp)
664: {
665:
666: if (MANDOCLEVEL_FATAL <= curp->file_status)
667: return;
668:
669: if (curp->mdoc && ! mdoc_endparse(curp->mdoc)) {
670: assert(MANDOCLEVEL_FATAL <= curp->file_status);
671: return;
672: }
673:
674: if (curp->man && ! man_endparse(curp->man)) {
675: assert(MANDOCLEVEL_FATAL <= curp->file_status);
676: return;
677: }
678:
679: if ( ! (curp->man || curp->mdoc)) {
1.6 kristaps 680: mandoc_msg(MANDOCERR_NOTMANUAL, curp, 1, 0, NULL);
1.1 kristaps 681: curp->file_status = MANDOCLEVEL_FATAL;
1.6 kristaps 682: return;
1.1 kristaps 683: }
684:
685: roff_endparse(curp->roff);
686: }
687:
688: static void
1.36 schwarze 689: mparse_parse_buffer(struct mparse *curp, struct buf blk, const char *file)
1.28 joerg 690: {
691: const char *svfile;
1.36 schwarze 692: static int recursion_depth;
693:
694: if (64 < recursion_depth) {
695: mandoc_msg(MANDOCERR_ROFFLOOP, curp, curp->line, 0, NULL);
696: return;
697: }
1.28 joerg 698:
699: /* Line number is per-file. */
700: svfile = curp->file;
701: curp->file = file;
702: curp->line = 1;
1.36 schwarze 703: recursion_depth++;
1.28 joerg 704:
705: mparse_buf_r(curp, blk, 1);
706:
1.36 schwarze 707: if (0 == --recursion_depth && MANDOCLEVEL_FATAL > curp->file_status)
1.28 joerg 708: mparse_end(curp);
709:
710: curp->file = svfile;
711: }
712:
713: enum mandoclevel
714: mparse_readmem(struct mparse *curp, const void *buf, size_t len,
715: const char *file)
716: {
717: struct buf blk;
718:
719: blk.buf = UNCONST(buf);
720: blk.sz = len;
721:
1.36 schwarze 722: mparse_parse_buffer(curp, blk, file);
1.28 joerg 723: return(curp->file_status);
724: }
725:
1.36 schwarze 726: enum mandoclevel
727: mparse_readfd(struct mparse *curp, int fd, const char *file)
1.1 kristaps 728: {
1.28 joerg 729: struct buf blk;
730: int with_mmap;
1.1 kristaps 731:
1.40 schwarze 732: if (-1 == fd && -1 == (fd = open(file, O_RDONLY, 0))) {
733: curp->file_status = MANDOCLEVEL_SYSERR;
734: if (curp->mmsg)
735: (*curp->mmsg)(MANDOCERR_SYSOPEN,
736: curp->file_status,
737: file, 0, 0, strerror(errno));
738: goto out;
739: }
740:
1.28 joerg 741: /*
742: * Run for each opened file; may be called more than once for
743: * each full parse sequence if the opened file is nested (i.e.,
744: * from `so'). Simply sucks in the whole file and moves into
745: * the parse phase for the file.
746: */
1.1 kristaps 747:
1.40 schwarze 748: if ( ! read_whole_file(curp, file, fd, &blk, &with_mmap))
1.36 schwarze 749: goto out;
1.1 kristaps 750:
1.36 schwarze 751: mparse_parse_buffer(curp, blk, file);
1.1 kristaps 752:
1.28 joerg 753: #ifdef HAVE_MMAP
754: if (with_mmap)
755: munmap(blk.buf, blk.sz);
756: else
757: #endif
758: free(blk.buf);
1.1 kristaps 759:
760: if (STDIN_FILENO != fd && -1 == close(fd))
761: perror(file);
1.36 schwarze 762: out:
1.1 kristaps 763: return(curp->file_status);
764: }
765:
766: struct mparse *
1.44 ! schwarze 767: mparse_alloc(int options, enum mandoclevel wlevel,
! 768: mandocmsg mmsg, char *defos)
1.1 kristaps 769: {
770: struct mparse *curp;
1.10 kristaps 771:
772: assert(wlevel <= MANDOCLEVEL_FATAL);
1.1 kristaps 773:
774: curp = mandoc_calloc(1, sizeof(struct mparse));
775:
1.44 ! schwarze 776: curp->options = options;
1.3 kristaps 777: curp->wlevel = wlevel;
1.1 kristaps 778: curp->mmsg = mmsg;
1.29 schwarze 779: curp->defos = defos;
1.1 kristaps 780:
1.44 ! schwarze 781: curp->roff = roff_alloc(curp, options);
1.1 kristaps 782: return(curp);
783: }
784:
785: void
786: mparse_reset(struct mparse *curp)
787: {
788:
789: roff_reset(curp->roff);
790:
791: if (curp->mdoc)
792: mdoc_reset(curp->mdoc);
793: if (curp->man)
794: man_reset(curp->man);
1.24 kristaps 795: if (curp->secondary)
796: curp->secondary->sz = 0;
1.1 kristaps 797:
798: curp->file_status = MANDOCLEVEL_OK;
799: curp->mdoc = NULL;
800: curp->man = NULL;
801: }
802:
803: void
804: mparse_free(struct mparse *curp)
805: {
806:
807: if (curp->pmdoc)
808: mdoc_free(curp->pmdoc);
809: if (curp->pman)
810: man_free(curp->pman);
811: if (curp->roff)
812: roff_free(curp->roff);
1.24 kristaps 813: if (curp->secondary)
814: free(curp->secondary->buf);
1.1 kristaps 815:
1.24 kristaps 816: free(curp->secondary);
1.1 kristaps 817: free(curp);
818: }
819:
820: void
821: mparse_result(struct mparse *curp, struct mdoc **mdoc, struct man **man)
822: {
823:
1.9 kristaps 824: if (mdoc)
825: *mdoc = curp->mdoc;
826: if (man)
827: *man = curp->man;
1.3 kristaps 828: }
829:
830: void
831: mandoc_vmsg(enum mandocerr t, struct mparse *m,
832: int ln, int pos, const char *fmt, ...)
833: {
834: char buf[256];
835: va_list ap;
836:
837: va_start(ap, fmt);
838: vsnprintf(buf, sizeof(buf) - 1, fmt, ap);
839: va_end(ap);
840:
841: mandoc_msg(t, m, ln, pos, buf);
842: }
843:
844: void
845: mandoc_msg(enum mandocerr er, struct mparse *m,
846: int ln, int col, const char *msg)
847: {
848: enum mandoclevel level;
849:
850: level = MANDOCLEVEL_FATAL;
851: while (er < mandoclimits[level])
852: level--;
853:
854: if (level < m->wlevel)
855: return;
856:
1.8 kristaps 857: if (m->mmsg)
858: (*m->mmsg)(er, level, m->file, ln, col, msg);
1.3 kristaps 859:
860: if (m->file_status < level)
861: m->file_status = level;
1.7 kristaps 862: }
863:
864: const char *
865: mparse_strerror(enum mandocerr er)
866: {
867:
868: return(mandocerrs[er]);
869: }
870:
871: const char *
872: mparse_strlevel(enum mandoclevel lvl)
873: {
874: return(mandoclevels[lvl]);
1.24 kristaps 875: }
876:
877: void
878: mparse_keep(struct mparse *p)
879: {
880:
881: assert(NULL == p->secondary);
882: p->secondary = mandoc_calloc(1, sizeof(struct buf));
883: }
884:
885: const char *
886: mparse_getkeep(const struct mparse *p)
887: {
888:
889: assert(p->secondary);
890: return(p->secondary->sz ? p->secondary->buf : NULL);
1.1 kristaps 891: }
CVSweb