Annotation of src/usr.bin/mandoc/read.c, Revision 1.28
1.28 ! schwarze 1: /* $Id: read.c,v 1.27 2014/06/20 17:23:09 schwarze Exp $ */
1.1 schwarze 2: /*
3: * Copyright (c) 2008, 2009, 2010, 2011 Kristaps Dzonsons <kristaps@bsd.lv>
1.18 schwarze 4: * Copyright (c) 2010-2014 Ingo Schwarze <schwarze@openbsd.org>
1.20 schwarze 5: * Copyright (c) 2010, 2012 Joerg Sonnenberger <joerg@netbsd.org>
1.1 schwarze 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: */
19: #include <sys/stat.h>
20: #include <sys/mman.h>
21:
22: #include <assert.h>
23: #include <ctype.h>
1.18 schwarze 24: #include <errno.h>
1.1 schwarze 25: #include <fcntl.h>
26: #include <stdarg.h>
27: #include <stdio.h>
28: #include <stdlib.h>
29: #include <string.h>
30: #include <unistd.h>
31:
32: #include "mandoc.h"
1.24 schwarze 33: #include "mandoc_aux.h"
1.1 schwarze 34: #include "libmandoc.h"
35: #include "mdoc.h"
36: #include "man.h"
37:
38: #define REPARSE_LIMIT 1000
39:
40: struct buf {
1.25 schwarze 41: char *buf; /* binary input buffer */
1.1 schwarze 42: size_t sz; /* size of binary buffer */
43: };
44:
45: struct mparse {
46: enum mandoclevel file_status; /* status of current parse */
47: enum mandoclevel wlevel; /* ignore messages below this */
48: int line; /* line number in the file */
1.22 schwarze 49: int options; /* parser options */
1.1 schwarze 50: struct man *pman; /* persistent man parser */
51: struct mdoc *pmdoc; /* persistent mdoc parser */
52: struct man *man; /* man parser */
53: struct mdoc *mdoc; /* mdoc parser */
54: struct roff *roff; /* roff parser (!NULL) */
1.23 schwarze 55: char *sodest; /* filename pointed to by .so */
1.1 schwarze 56: int reparse_count; /* finite interp. stack */
57: mandocmsg mmsg; /* warning/error message handler */
1.25 schwarze 58: const char *file;
1.4 schwarze 59: struct buf *secondary;
1.7 schwarze 60: char *defos; /* default operating system */
1.1 schwarze 61: };
62:
63: static void resize_buf(struct buf *, size_t);
64: static void mparse_buf_r(struct mparse *, struct buf, int);
65: static void pset(const char *, int, struct mparse *);
1.18 schwarze 66: static int read_whole_file(struct mparse *, const char *, int,
67: struct buf *, int *);
1.1 schwarze 68: static void mparse_end(struct mparse *);
1.15 schwarze 69: static void mparse_parse_buffer(struct mparse *, struct buf,
70: const char *);
1.1 schwarze 71:
72: static const enum mandocerr mandoclimits[MANDOCLEVEL_MAX] = {
73: MANDOCERR_OK,
74: MANDOCERR_WARNING,
75: MANDOCERR_WARNING,
76: MANDOCERR_ERROR,
77: MANDOCERR_FATAL,
78: MANDOCERR_MAX,
79: MANDOCERR_MAX
80: };
81:
82: static const char * const mandocerrs[MANDOCERR_MAX] = {
83: "ok",
84:
85: "generic warning",
86:
87: /* related to the prologue */
1.27 schwarze 88: "no TH macro in document",
1.1 schwarze 89: "document title should be all caps",
90: "unknown manual section",
1.11 schwarze 91: "unknown manual volume or arch",
1.1 schwarze 92: "date missing, using today's date",
93: "cannot parse date, using it verbatim",
94: "prologue macros out of order",
95: "duplicate prologue macro",
96: "macro not allowed in prologue",
97: "macro not allowed in body",
98:
99: /* related to document structure */
100: ".so is fragile, better use ln(1)",
1.28 ! schwarze 101: "no document body",
! 102: "content before the first section header",
1.1 schwarze 103: "NAME section must come first",
104: "bad NAME section contents",
105: "sections out of conventional order",
106: "duplicate section name",
1.17 schwarze 107: "section header suited to sections 2, 3, and 9 only",
1.1 schwarze 108:
109: /* related to macros and nesting */
110: "skipping obsolete macro",
111: "skipping paragraph macro",
1.10 schwarze 112: "moving paragraph macro out of list",
1.1 schwarze 113: "skipping no-space macro",
114: "blocks badly nested",
115: "child violates parent syntax",
116: "nested displays are not portable",
117: "already in literal mode",
118: "line scope broken",
119:
120: /* related to missing macro arguments */
121: "skipping empty macro",
122: "argument count wrong",
123: "missing display type",
124: "list type must come first",
125: "tag lists require a width argument",
126: "missing font type",
127: "skipping end of block that is not open",
128:
129: /* related to bad macro arguments */
130: "skipping argument",
131: "duplicate argument",
132: "duplicate display type",
133: "duplicate list type",
134: "unknown AT&T UNIX version",
135: "bad Boolean value",
136: "unknown font",
137: "unknown standard specifier",
138: "bad width argument",
139:
140: /* related to plain text */
141: "blank line in non-literal context",
142: "tab in non-literal context",
143: "end of line whitespace",
144: "bad comment style",
1.2 schwarze 145: "bad escape sequence",
1.1 schwarze 146: "unterminated quoted string",
1.3 schwarze 147:
148: /* related to equations */
149: "unexpected literal in equation",
1.25 schwarze 150:
1.1 schwarze 151: "generic error",
152:
1.3 schwarze 153: /* related to equations */
154: "unexpected equation scope closure",
155: "equation scope open on exit",
156: "overlapping equation scopes",
157: "unexpected end of equation",
158: "equation syntax error",
159:
1.1 schwarze 160: /* related to tables */
161: "bad table syntax",
162: "bad table option",
163: "bad table layout",
164: "no table layout cells specified",
165: "no table data cells specified",
166: "ignore data in cell",
167: "data block still open",
168: "ignoring extra data cells",
169:
170: "input stack limit exceeded, infinite loop?",
171: "skipping bad character",
172: "escaped character not allowed in a name",
1.9 schwarze 173: "manual name not yet set",
1.1 schwarze 174: "skipping text before the first section header",
175: "skipping unknown macro",
176: "NOT IMPLEMENTED, please use groff: skipping request",
177: "argument count wrong",
1.12 schwarze 178: "skipping column outside column list",
1.1 schwarze 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)",
1.16 schwarze 186: "request requires a numeric argument",
1.1 schwarze 187: "missing list type",
188: "line argument(s) will be lost",
189: "body argument(s) will be lost",
190:
191: "generic fatal error",
192:
1.18 schwarze 193: "input too large",
1.1 schwarze 194: "not a manual",
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 \"..\"",
201: "no document prologue",
202: "static buffer exhausted",
1.18 schwarze 203:
204: /* system errors */
205: "cannot open file",
206: "cannot stat file",
207: "cannot read file",
1.1 schwarze 208: };
209:
210: static const char * const mandoclevels[MANDOCLEVEL_MAX] = {
211: "SUCCESS",
212: "RESERVED",
213: "WARNING",
214: "ERROR",
215: "FATAL",
216: "BADARG",
217: "SYSERR"
218: };
219:
1.25 schwarze 220:
1.1 schwarze 221: static void
222: resize_buf(struct buf *buf, size_t initial)
223: {
224:
225: buf->sz = buf->sz > initial/2 ? 2 * buf->sz : initial;
226: buf->buf = mandoc_realloc(buf->buf, buf->sz);
227: }
228:
229: static void
230: pset(const char *buf, int pos, struct mparse *curp)
231: {
232: int i;
233:
234: /*
235: * Try to intuit which kind of manual parser should be used. If
236: * passed in by command-line (-man, -mdoc), then use that
237: * explicitly. If passed as -mandoc, then try to guess from the
238: * line: either skip dot-lines, use -mdoc when finding `.Dt', or
239: * default to -man, which is more lenient.
240: *
241: * Separate out pmdoc/pman from mdoc/man: the first persists
242: * through all parsers, while the latter is used per-parse.
243: */
244:
245: if ('.' == buf[0] || '\'' == buf[0]) {
246: for (i = 1; buf[i]; i++)
247: if (' ' != buf[i] && '\t' != buf[i])
248: break;
249: if ('\0' == buf[i])
250: return;
251: }
252:
1.22 schwarze 253: if (MPARSE_MDOC & curp->options) {
1.25 schwarze 254: if (NULL == curp->pmdoc)
1.22 schwarze 255: curp->pmdoc = mdoc_alloc(
256: curp->roff, curp, curp->defos,
257: MPARSE_QUICK & curp->options ? 1 : 0);
1.1 schwarze 258: assert(curp->pmdoc);
259: curp->mdoc = curp->pmdoc;
260: return;
1.22 schwarze 261: } else if (MPARSE_MAN & curp->options) {
1.25 schwarze 262: if (NULL == curp->pman)
1.19 schwarze 263: curp->pman = man_alloc(curp->roff, curp,
1.22 schwarze 264: MPARSE_QUICK & curp->options ? 1 : 0);
1.1 schwarze 265: assert(curp->pman);
266: curp->man = curp->pman;
267: return;
268: }
269:
270: if (pos >= 3 && 0 == memcmp(buf, ".Dd", 3)) {
1.25 schwarze 271: if (NULL == curp->pmdoc)
1.22 schwarze 272: curp->pmdoc = mdoc_alloc(
273: curp->roff, curp, curp->defos,
274: MPARSE_QUICK & curp->options ? 1 : 0);
1.1 schwarze 275: assert(curp->pmdoc);
276: curp->mdoc = curp->pmdoc;
277: return;
1.25 schwarze 278: }
1.1 schwarze 279:
1.25 schwarze 280: if (NULL == curp->pman)
1.22 schwarze 281: curp->pman = man_alloc(curp->roff, curp,
282: MPARSE_QUICK & curp->options ? 1 : 0);
1.1 schwarze 283: assert(curp->pman);
284: curp->man = curp->pman;
285: }
286:
287: /*
288: * Main parse routine for an opened file. This is called for each
289: * opened file and simply loops around the full input file, possibly
290: * nesting (i.e., with `so').
291: */
292: static void
293: mparse_buf_r(struct mparse *curp, struct buf blk, int start)
294: {
295: const struct tbl_span *span;
296: struct buf ln;
297: enum rofferr rr;
298: int i, of, rc;
299: int pos; /* byte number in the ln buffer */
300: int lnn; /* line number in the real file */
301: unsigned char c;
302:
303: memset(&ln, 0, sizeof(struct buf));
304:
1.25 schwarze 305: lnn = curp->line;
306: pos = 0;
1.1 schwarze 307:
308: for (i = 0; i < (int)blk.sz; ) {
309: if (0 == pos && '\0' == blk.buf[i])
310: break;
311:
312: if (start) {
313: curp->line = lnn;
314: curp->reparse_count = 0;
315: }
316:
317: while (i < (int)blk.sz && (start || '\0' != blk.buf[i])) {
318:
319: /*
320: * When finding an unescaped newline character,
321: * leave the character loop to process the line.
322: * Skip a preceding carriage return, if any.
323: */
324:
325: if ('\r' == blk.buf[i] && i + 1 < (int)blk.sz &&
326: '\n' == blk.buf[i + 1])
327: ++i;
328: if ('\n' == blk.buf[i]) {
329: ++i;
330: ++lnn;
331: break;
332: }
333:
1.13 schwarze 334: /*
335: * Make sure we have space for at least
336: * one backslash and one other character
337: * and the trailing NUL byte.
338: */
339:
340: if (pos + 2 >= (int)ln.sz)
341: resize_buf(&ln, 256);
342:
1.25 schwarze 343: /*
1.1 schwarze 344: * Warn about bogus characters. If you're using
345: * non-ASCII encoding, you're screwing your
346: * readers. Since I'd rather this not happen,
1.6 schwarze 347: * I'll be helpful and replace these characters
348: * with "?", so we don't display gibberish.
349: * Note to manual writers: use special characters.
1.1 schwarze 350: */
351:
352: c = (unsigned char) blk.buf[i];
353:
1.25 schwarze 354: if ( ! (isascii(c) &&
355: (isgraph(c) || isblank(c)))) {
1.1 schwarze 356: mandoc_msg(MANDOCERR_BADCHAR, curp,
1.25 schwarze 357: curp->line, pos, NULL);
1.1 schwarze 358: i++;
1.6 schwarze 359: ln.buf[pos++] = '?';
1.1 schwarze 360: continue;
361: }
362:
363: /* Trailing backslash = a plain char. */
364:
365: if ('\\' != blk.buf[i] || i + 1 == (int)blk.sz) {
366: ln.buf[pos++] = blk.buf[i++];
367: continue;
368: }
369:
370: /*
371: * Found escape and at least one other character.
372: * When it's a newline character, skip it.
373: * When there is a carriage return in between,
374: * skip that one as well.
375: */
376:
377: if ('\r' == blk.buf[i + 1] && i + 2 < (int)blk.sz &&
378: '\n' == blk.buf[i + 2])
379: ++i;
380: if ('\n' == blk.buf[i + 1]) {
381: i += 2;
382: ++lnn;
383: continue;
384: }
385:
386: if ('"' == blk.buf[i + 1] || '#' == blk.buf[i + 1]) {
387: i += 2;
388: /* Comment, skip to end of line */
389: for (; i < (int)blk.sz; ++i) {
390: if ('\n' == blk.buf[i]) {
391: ++i;
392: ++lnn;
393: break;
394: }
395: }
396:
397: /* Backout trailing whitespaces */
398: for (; pos > 0; --pos) {
399: if (ln.buf[pos - 1] != ' ')
400: break;
401: if (pos > 2 && ln.buf[pos - 2] == '\\')
402: break;
403: }
404: break;
405: }
406:
1.13 schwarze 407: /* Catch escaped bogus characters. */
408:
409: c = (unsigned char) blk.buf[i+1];
410:
1.25 schwarze 411: if ( ! (isascii(c) &&
412: (isgraph(c) || isblank(c)))) {
1.13 schwarze 413: mandoc_msg(MANDOCERR_BADCHAR, curp,
1.25 schwarze 414: curp->line, pos, NULL);
1.13 schwarze 415: i += 2;
416: ln.buf[pos++] = '?';
417: continue;
418: }
419:
1.1 schwarze 420: /* Some other escape sequence, copy & cont. */
421:
422: ln.buf[pos++] = blk.buf[i++];
423: ln.buf[pos++] = blk.buf[i++];
424: }
425:
1.25 schwarze 426: if (pos >= (int)ln.sz)
1.1 schwarze 427: resize_buf(&ln, 256);
428:
429: ln.buf[pos] = '\0';
430:
431: /*
432: * A significant amount of complexity is contained by
433: * the roff preprocessor. It's line-oriented but can be
434: * expressed on one line, so we need at times to
435: * readjust our starting point and re-run it. The roff
436: * preprocessor can also readjust the buffers with new
437: * data, so we pass them in wholesale.
438: */
439:
440: of = 0;
441:
1.4 schwarze 442: /*
443: * Maintain a lookaside buffer of all parsed lines. We
444: * only do this if mparse_keep() has been invoked (the
445: * buffer may be accessed with mparse_getkeep()).
446: */
447:
448: if (curp->secondary) {
1.25 schwarze 449: curp->secondary->buf = mandoc_realloc(
450: curp->secondary->buf,
451: curp->secondary->sz + pos + 2);
452: memcpy(curp->secondary->buf +
453: curp->secondary->sz,
454: ln.buf, pos);
1.4 schwarze 455: curp->secondary->sz += pos;
456: curp->secondary->buf
457: [curp->secondary->sz] = '\n';
458: curp->secondary->sz++;
459: curp->secondary->buf
460: [curp->secondary->sz] = '\0';
461: }
1.1 schwarze 462: rerun:
1.25 schwarze 463: rr = roff_parseln(curp->roff, curp->line,
464: &ln.buf, &ln.sz, of, &of);
1.1 schwarze 465:
466: switch (rr) {
1.25 schwarze 467: case ROFF_REPARSE:
1.1 schwarze 468: if (REPARSE_LIMIT >= ++curp->reparse_count)
469: mparse_buf_r(curp, ln, 0);
470: else
471: mandoc_msg(MANDOCERR_ROFFLOOP, curp,
1.25 schwarze 472: curp->line, pos, NULL);
1.1 schwarze 473: pos = 0;
474: continue;
1.25 schwarze 475: case ROFF_APPEND:
1.1 schwarze 476: pos = (int)strlen(ln.buf);
477: continue;
1.25 schwarze 478: case ROFF_RERUN:
1.1 schwarze 479: goto rerun;
1.25 schwarze 480: case ROFF_IGN:
1.1 schwarze 481: pos = 0;
482: continue;
1.25 schwarze 483: case ROFF_ERR:
1.1 schwarze 484: assert(MANDOCLEVEL_FATAL <= curp->file_status);
485: break;
1.25 schwarze 486: case ROFF_SO:
1.23 schwarze 487: if (0 == (MPARSE_SO & curp->options) &&
488: (i >= (int)blk.sz || '\0' == blk.buf[i])) {
489: curp->sodest = mandoc_strdup(ln.buf + of);
490: free(ln.buf);
491: return;
492: }
1.4 schwarze 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 schwarze 498: if (curp->secondary)
1.4 schwarze 499: curp->secondary->sz -= pos + 1;
1.14 schwarze 500: mparse_readfd(curp, -1, ln.buf + of);
1.1 schwarze 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.
1.2 schwarze 519: * We keep these instanced between parsers, but set them
1.1 schwarze 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:
1.25 schwarze 527: /*
1.1 schwarze 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 ?
1.25 schwarze 543: man_addspan(curp->man, span) :
544: mdoc_addspan(curp->mdoc, span);
1.1 schwarze 545: if (0 == rc)
546: break;
547: }
548: else if (ROFF_EQN == rr)
1.25 schwarze 549: rc = curp->mdoc ?
550: mdoc_addeqn(curp->mdoc,
551: roff_eqn(curp->roff)) :
552: man_addeqn(curp->man,
553: roff_eqn(curp->roff));
1.1 schwarze 554: else if (curp->man || curp->mdoc)
555: rc = curp->man ?
1.25 schwarze 556: man_parseln(curp->man,
557: curp->line, ln.buf, of) :
558: mdoc_parseln(curp->mdoc,
559: curp->line, ln.buf, of);
1.1 schwarze 560:
561: if (0 == rc) {
562: assert(MANDOCLEVEL_FATAL <= curp->file_status);
563: break;
1.19 schwarze 564: } else if (2 == rc)
565: break;
1.1 schwarze 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
1.18 schwarze 581: read_whole_file(struct mparse *curp, const char *file, int fd,
582: struct buf *fb, int *with_mmap)
1.1 schwarze 583: {
584: struct stat st;
585: size_t off;
586: ssize_t ssz;
587:
588: if (-1 == fstat(fd, &st)) {
1.18 schwarze 589: curp->file_status = MANDOCLEVEL_SYSERR;
590: if (curp->mmsg)
591: (*curp->mmsg)(MANDOCERR_SYSSTAT, curp->file_status,
592: file, 0, 0, strerror(errno));
1.1 schwarze 593: return(0);
594: }
595:
596: /*
597: * If we're a regular file, try just reading in the whole entry
598: * via mmap(). This is faster than reading it into blocks, and
599: * since each file is only a few bytes to begin with, I'm not
600: * concerned that this is going to tank any machines.
601: */
602:
603: if (S_ISREG(st.st_mode)) {
604: if (st.st_size >= (1U << 31)) {
1.18 schwarze 605: curp->file_status = MANDOCLEVEL_FATAL;
606: if (curp->mmsg)
607: (*curp->mmsg)(MANDOCERR_TOOLARGE,
608: curp->file_status, file, 0, 0, NULL);
1.1 schwarze 609: return(0);
610: }
611: *with_mmap = 1;
612: fb->sz = (size_t)st.st_size;
1.15 schwarze 613: fb->buf = mmap(NULL, fb->sz, PROT_READ, MAP_SHARED, fd, 0);
1.1 schwarze 614: if (fb->buf != MAP_FAILED)
615: return(1);
616: }
617:
618: /*
619: * If this isn't a regular file (like, say, stdin), then we must
620: * go the old way and just read things in bit by bit.
621: */
622:
623: *with_mmap = 0;
624: off = 0;
625: fb->sz = 0;
626: fb->buf = NULL;
627: for (;;) {
628: if (off == fb->sz) {
629: if (fb->sz == (1U << 31)) {
1.18 schwarze 630: curp->file_status = MANDOCLEVEL_FATAL;
631: if (curp->mmsg)
632: (*curp->mmsg)(MANDOCERR_TOOLARGE,
633: curp->file_status,
634: file, 0, 0, NULL);
1.1 schwarze 635: break;
636: }
637: resize_buf(fb, 65536);
638: }
639: ssz = read(fd, fb->buf + (int)off, fb->sz - off);
640: if (ssz == 0) {
641: fb->sz = off;
642: return(1);
643: }
644: if (ssz == -1) {
1.18 schwarze 645: curp->file_status = MANDOCLEVEL_SYSERR;
646: if (curp->mmsg)
647: (*curp->mmsg)(MANDOCERR_SYSREAD,
648: curp->file_status, file, 0, 0,
649: strerror(errno));
1.1 schwarze 650: break;
651: }
652: off += (size_t)ssz;
653: }
654:
655: free(fb->buf);
656: fb->buf = NULL;
657: return(0);
658: }
659:
660: static void
661: mparse_end(struct mparse *curp)
662: {
663:
664: if (MANDOCLEVEL_FATAL <= curp->file_status)
665: return;
666:
667: if (curp->mdoc && ! mdoc_endparse(curp->mdoc)) {
668: assert(MANDOCLEVEL_FATAL <= curp->file_status);
669: return;
670: }
671:
672: if (curp->man && ! man_endparse(curp->man)) {
673: assert(MANDOCLEVEL_FATAL <= curp->file_status);
674: return;
675: }
676:
1.23 schwarze 677: if ( ! (curp->mdoc || curp->man || curp->sodest)) {
1.28 ! schwarze 678: mandoc_msg(MANDOCERR_NOTMANUAL, curp, 0, 0, NULL);
1.1 schwarze 679: curp->file_status = MANDOCLEVEL_FATAL;
680: return;
681: }
682:
683: roff_endparse(curp->roff);
684: }
685:
1.15 schwarze 686: static void
687: mparse_parse_buffer(struct mparse *curp, struct buf blk, const char *file)
1.1 schwarze 688: {
689: const char *svfile;
1.14 schwarze 690: static int recursion_depth;
691:
692: if (64 < recursion_depth) {
693: mandoc_msg(MANDOCERR_ROFFLOOP, curp, curp->line, 0, NULL);
1.15 schwarze 694: return;
1.14 schwarze 695: }
1.1 schwarze 696:
1.15 schwarze 697: /* Line number is per-file. */
698: svfile = curp->file;
699: curp->file = file;
700: curp->line = 1;
701: recursion_depth++;
702:
703: mparse_buf_r(curp, blk, 1);
704:
705: if (0 == --recursion_depth && MANDOCLEVEL_FATAL > curp->file_status)
706: mparse_end(curp);
707:
708: curp->file = svfile;
709: }
710:
711: enum mandoclevel
712: mparse_readfd(struct mparse *curp, int fd, const char *file)
713: {
714: struct buf blk;
715: int with_mmap;
716:
1.18 schwarze 717: if (-1 == fd && -1 == (fd = open(file, O_RDONLY, 0))) {
718: curp->file_status = MANDOCLEVEL_SYSERR;
719: if (curp->mmsg)
720: (*curp->mmsg)(MANDOCERR_SYSOPEN,
721: curp->file_status,
722: file, 0, 0, strerror(errno));
723: goto out;
724: }
725:
1.15 schwarze 726: /*
727: * Run for each opened file; may be called more than once for
728: * each full parse sequence if the opened file is nested (i.e.,
729: * from `so'). Simply sucks in the whole file and moves into
730: * the parse phase for the file.
731: */
1.1 schwarze 732:
1.18 schwarze 733: if ( ! read_whole_file(curp, file, fd, &blk, &with_mmap))
1.15 schwarze 734: goto out;
1.1 schwarze 735:
1.15 schwarze 736: mparse_parse_buffer(curp, blk, file);
1.1 schwarze 737:
1.15 schwarze 738: if (with_mmap)
739: munmap(blk.buf, blk.sz);
740: else
741: free(blk.buf);
1.1 schwarze 742:
743: if (STDIN_FILENO != fd && -1 == close(fd))
744: perror(file);
1.14 schwarze 745: out:
1.1 schwarze 746: return(curp->file_status);
747: }
748:
749: struct mparse *
1.22 schwarze 750: mparse_alloc(int options, enum mandoclevel wlevel,
751: mandocmsg mmsg, char *defos)
1.1 schwarze 752: {
753: struct mparse *curp;
754:
755: assert(wlevel <= MANDOCLEVEL_FATAL);
756:
757: curp = mandoc_calloc(1, sizeof(struct mparse));
758:
1.22 schwarze 759: curp->options = options;
1.1 schwarze 760: curp->wlevel = wlevel;
761: curp->mmsg = mmsg;
1.7 schwarze 762: curp->defos = defos;
1.1 schwarze 763:
1.22 schwarze 764: curp->roff = roff_alloc(curp, options);
1.1 schwarze 765: return(curp);
766: }
767:
768: void
769: mparse_reset(struct mparse *curp)
770: {
771:
772: roff_reset(curp->roff);
773:
774: if (curp->mdoc)
775: mdoc_reset(curp->mdoc);
776: if (curp->man)
777: man_reset(curp->man);
1.4 schwarze 778: if (curp->secondary)
779: curp->secondary->sz = 0;
1.1 schwarze 780:
781: curp->file_status = MANDOCLEVEL_OK;
782: curp->mdoc = NULL;
783: curp->man = NULL;
1.23 schwarze 784:
785: free(curp->sodest);
786: curp->sodest = NULL;
1.1 schwarze 787: }
788:
789: void
790: mparse_free(struct mparse *curp)
791: {
792:
793: if (curp->pmdoc)
794: mdoc_free(curp->pmdoc);
795: if (curp->pman)
796: man_free(curp->pman);
797: if (curp->roff)
798: roff_free(curp->roff);
1.4 schwarze 799: if (curp->secondary)
800: free(curp->secondary->buf);
1.1 schwarze 801:
1.4 schwarze 802: free(curp->secondary);
1.23 schwarze 803: free(curp->sodest);
1.1 schwarze 804: free(curp);
805: }
806:
807: void
1.23 schwarze 808: mparse_result(struct mparse *curp,
809: struct mdoc **mdoc, struct man **man, char **sodest)
1.1 schwarze 810: {
811:
1.23 schwarze 812: if (sodest && NULL != (*sodest = curp->sodest)) {
813: *mdoc = NULL;
814: *man = NULL;
815: return;
816: }
1.1 schwarze 817: if (mdoc)
818: *mdoc = curp->mdoc;
819: if (man)
820: *man = curp->man;
821: }
822:
823: void
824: mandoc_vmsg(enum mandocerr t, struct mparse *m,
825: int ln, int pos, const char *fmt, ...)
826: {
827: char buf[256];
828: va_list ap;
829:
830: va_start(ap, fmt);
1.26 schwarze 831: (void)vsnprintf(buf, sizeof(buf), fmt, ap);
1.1 schwarze 832: va_end(ap);
833:
834: mandoc_msg(t, m, ln, pos, buf);
835: }
836:
837: void
1.25 schwarze 838: mandoc_msg(enum mandocerr er, struct mparse *m,
1.1 schwarze 839: int ln, int col, const char *msg)
840: {
841: enum mandoclevel level;
842:
843: level = MANDOCLEVEL_FATAL;
844: while (er < mandoclimits[level])
845: level--;
846:
847: if (level < m->wlevel)
848: return;
849:
850: if (m->mmsg)
851: (*m->mmsg)(er, level, m->file, ln, col, msg);
852:
853: if (m->file_status < level)
854: m->file_status = level;
855: }
856:
857: const char *
858: mparse_strerror(enum mandocerr er)
859: {
860:
861: return(mandocerrs[er]);
862: }
863:
864: const char *
865: mparse_strlevel(enum mandoclevel lvl)
866: {
867: return(mandoclevels[lvl]);
1.4 schwarze 868: }
869:
870: void
871: mparse_keep(struct mparse *p)
872: {
873:
874: assert(NULL == p->secondary);
875: p->secondary = mandoc_calloc(1, sizeof(struct buf));
876: }
877:
878: const char *
879: mparse_getkeep(const struct mparse *p)
880: {
881:
882: assert(p->secondary);
883: return(p->secondary->sz ? p->secondary->buf : NULL);
1.1 schwarze 884: }