Annotation of src/usr.bin/mandoc/read.c, Revision 1.51
1.51 ! schwarze 1: /* $Id: read.c,v 1.50 2014/08/08 15:21:17 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;
1.47 schwarze 60: const 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 */
1.32 schwarze 88: "missing .TH macro, using \"unknown 1\"",
89: "lower case character in document title",
1.1 schwarze 90: "unknown manual section",
1.11 schwarze 91: "unknown manual volume or arch",
1.32 schwarze 92: "missing date, using today's date",
1.1 schwarze 93: "cannot parse date, using it verbatim",
94: "prologue macros out of order",
95: "duplicate prologue macro",
1.32 schwarze 96: "incomplete prologue, terminated by",
97: "skipping prologue macro in body",
1.1 schwarze 98:
99: /* related to document structure */
100: ".so is fragile, better use ln(1)",
1.28 schwarze 101: "no document body",
1.32 schwarze 102: "content before first section header",
103: "first section is not \"NAME\"",
1.1 schwarze 104: "bad NAME section contents",
105: "sections out of conventional order",
1.32 schwarze 106: "duplicate section title",
107: "unexpected section",
1.1 schwarze 108:
109: /* related to macros and nesting */
1.33 schwarze 110: "obsolete macro",
1.1 schwarze 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: "nested displays are not portable",
1.35 schwarze 116: "moving content out of list",
117: ".Vt block has child macro",
1.34 schwarze 118: "fill mode already enabled, skipping .fi",
119: "fill mode already disabled, skipping .nf",
1.1 schwarze 120: "line scope broken",
121:
122: /* related to missing macro arguments */
1.36 schwarze 123: "skipping empty request",
124: "conditional request controls empty scope",
1.1 schwarze 125: "skipping empty macro",
1.40 schwarze 126: "empty argument, using 0n",
1.1 schwarze 127: "argument count wrong",
1.38 schwarze 128: "missing display type, using -ragged",
129: "list type is not the first argument",
130: "missing -width in -tag list, using 8n",
1.48 schwarze 131: "missing name for .Ex, using \"\"",
1.38 schwarze 132: "empty head in list item",
133: "empty list item",
1.39 schwarze 134: "missing font type, using \\fR",
135: "unknown font type, using \\fR",
1.38 schwarze 136: "missing -std argument, adding it",
1.1 schwarze 137:
138: /* related to bad macro arguments */
139: "skipping argument",
1.42 schwarze 140: "unterminated quoted argument",
1.1 schwarze 141: "duplicate argument",
1.41 schwarze 142: "skipping duplicate display type",
143: "skipping duplicate list type",
1.1 schwarze 144: "unknown AT&T UNIX version",
1.45 schwarze 145: "invalid content in Rs block",
1.41 schwarze 146: "invalid Boolean argument",
147: "unknown font, skipping request",
1.1 schwarze 148:
149: /* related to plain text */
1.42 schwarze 150: "blank line in fill mode, using .sp",
151: "tab in filled text",
152: "whitespace at end of input line",
1.1 schwarze 153: "bad comment style",
1.42 schwarze 154: "invalid escape sequence",
155: "undefined string, using \"\"",
1.3 schwarze 156:
1.1 schwarze 157: "generic error",
158:
1.3 schwarze 159: /* related to equations */
160: "unexpected equation scope closure",
161: "equation scope open on exit",
162: "overlapping equation scopes",
163: "unexpected end of equation",
164: "equation syntax error",
165:
1.1 schwarze 166: /* related to tables */
167: "bad table syntax",
168: "bad table option",
169: "bad table layout",
170: "no table layout cells specified",
171: "no table data cells specified",
172: "ignore data in cell",
173: "data block still open",
174: "ignoring extra data cells",
175:
1.46 schwarze 176: /* related to document structure and macros */
1.1 schwarze 177: "input stack limit exceeded, infinite loop?",
178: "skipping bad character",
1.46 schwarze 179: "skipping unknown macro",
1.51 ! schwarze 180: "skipping item outside list",
1.46 schwarze 181: "skipping column outside column list",
182: "skipping end of block that is not open",
183: "inserting missing end of block",
184: "appending missing end of block",
185:
186: /* related to request and macro arguments */
1.1 schwarze 187: "escaped character not allowed in a name",
188: "argument count wrong",
1.49 schwarze 189: "missing list type, using -item",
1.48 schwarze 190: "missing manual name, using \"\"",
1.49 schwarze 191: "uname(3) system call failed, using UNKNOWN",
1.41 schwarze 192: "unknown standard specifier",
1.49 schwarze 193: "skipping request without numeric argument",
1.39 schwarze 194: "skipping all arguments",
195: "skipping excess arguments",
1.1 schwarze 196:
197: "generic fatal error",
198:
1.18 schwarze 199: "input too large",
1.1 schwarze 200: "column syntax is inconsistent",
201: "NOT IMPLEMENTED: .Bd -file",
202: "argument count wrong, violates syntax",
203: "NOT IMPLEMENTED: .so with absolute path or \"..\"",
1.30 schwarze 204: ".so request failed",
1.1 schwarze 205: "static buffer exhausted",
1.18 schwarze 206:
207: /* system errors */
1.29 schwarze 208: NULL,
1.18 schwarze 209: "cannot stat file",
210: "cannot read file",
1.1 schwarze 211: };
212:
213: static const char * const mandoclevels[MANDOCLEVEL_MAX] = {
214: "SUCCESS",
215: "RESERVED",
216: "WARNING",
217: "ERROR",
218: "FATAL",
219: "BADARG",
220: "SYSERR"
221: };
222:
1.25 schwarze 223:
1.1 schwarze 224: static void
225: resize_buf(struct buf *buf, size_t initial)
226: {
227:
228: buf->sz = buf->sz > initial/2 ? 2 * buf->sz : initial;
229: buf->buf = mandoc_realloc(buf->buf, buf->sz);
230: }
231:
232: static void
233: pset(const char *buf, int pos, struct mparse *curp)
234: {
235: int i;
236:
237: /*
238: * Try to intuit which kind of manual parser should be used. If
239: * passed in by command-line (-man, -mdoc), then use that
240: * explicitly. If passed as -mandoc, then try to guess from the
241: * line: either skip dot-lines, use -mdoc when finding `.Dt', or
242: * default to -man, which is more lenient.
243: *
244: * Separate out pmdoc/pman from mdoc/man: the first persists
245: * through all parsers, while the latter is used per-parse.
246: */
247:
248: if ('.' == buf[0] || '\'' == buf[0]) {
249: for (i = 1; buf[i]; i++)
250: if (' ' != buf[i] && '\t' != buf[i])
251: break;
252: if ('\0' == buf[i])
253: return;
254: }
255:
1.22 schwarze 256: if (MPARSE_MDOC & curp->options) {
1.1 schwarze 257: curp->mdoc = curp->pmdoc;
258: return;
1.22 schwarze 259: } else if (MPARSE_MAN & curp->options) {
1.1 schwarze 260: curp->man = curp->pman;
261: return;
262: }
263:
264: if (pos >= 3 && 0 == memcmp(buf, ".Dd", 3)) {
1.25 schwarze 265: if (NULL == curp->pmdoc)
1.22 schwarze 266: curp->pmdoc = mdoc_alloc(
267: curp->roff, curp, curp->defos,
268: MPARSE_QUICK & curp->options ? 1 : 0);
1.1 schwarze 269: assert(curp->pmdoc);
270: curp->mdoc = curp->pmdoc;
271: return;
1.25 schwarze 272: }
1.1 schwarze 273:
1.25 schwarze 274: if (NULL == curp->pman)
1.22 schwarze 275: curp->pman = man_alloc(curp->roff, curp,
276: MPARSE_QUICK & curp->options ? 1 : 0);
1.1 schwarze 277: assert(curp->pman);
278: curp->man = curp->pman;
279: }
280:
281: /*
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').
285: */
286: static void
287: mparse_buf_r(struct mparse *curp, struct buf blk, int start)
288: {
289: const struct tbl_span *span;
290: struct buf ln;
291: enum rofferr rr;
292: int i, of, rc;
293: int pos; /* byte number in the ln buffer */
294: int lnn; /* line number in the real file */
295: unsigned char c;
296:
297: memset(&ln, 0, sizeof(struct buf));
298:
1.25 schwarze 299: lnn = curp->line;
300: pos = 0;
1.1 schwarze 301:
302: for (i = 0; i < (int)blk.sz; ) {
303: if (0 == pos && '\0' == blk.buf[i])
304: break;
305:
306: if (start) {
307: curp->line = lnn;
308: curp->reparse_count = 0;
309: }
310:
311: while (i < (int)blk.sz && (start || '\0' != blk.buf[i])) {
312:
313: /*
314: * When finding an unescaped newline character,
315: * leave the character loop to process the line.
316: * Skip a preceding carriage return, if any.
317: */
318:
319: if ('\r' == blk.buf[i] && i + 1 < (int)blk.sz &&
320: '\n' == blk.buf[i + 1])
321: ++i;
322: if ('\n' == blk.buf[i]) {
323: ++i;
324: ++lnn;
325: break;
326: }
327:
1.13 schwarze 328: /*
329: * Make sure we have space for at least
330: * one backslash and one other character
331: * and the trailing NUL byte.
332: */
333:
334: if (pos + 2 >= (int)ln.sz)
335: resize_buf(&ln, 256);
336:
1.25 schwarze 337: /*
1.1 schwarze 338: * Warn about bogus characters. If you're using
339: * non-ASCII encoding, you're screwing your
340: * readers. Since I'd rather this not happen,
1.6 schwarze 341: * I'll be helpful and replace these characters
342: * with "?", so we don't display gibberish.
343: * Note to manual writers: use special characters.
1.1 schwarze 344: */
345:
346: c = (unsigned char) blk.buf[i];
347:
1.25 schwarze 348: if ( ! (isascii(c) &&
349: (isgraph(c) || isblank(c)))) {
1.1 schwarze 350: mandoc_msg(MANDOCERR_BADCHAR, curp,
1.25 schwarze 351: curp->line, pos, NULL);
1.1 schwarze 352: i++;
1.6 schwarze 353: ln.buf[pos++] = '?';
1.1 schwarze 354: continue;
355: }
356:
357: /* Trailing backslash = a plain char. */
358:
359: if ('\\' != blk.buf[i] || i + 1 == (int)blk.sz) {
360: ln.buf[pos++] = blk.buf[i++];
361: continue;
362: }
363:
364: /*
365: * Found escape and at least one other character.
366: * When it's a newline character, skip it.
367: * When there is a carriage return in between,
368: * skip that one as well.
369: */
370:
371: if ('\r' == blk.buf[i + 1] && i + 2 < (int)blk.sz &&
372: '\n' == blk.buf[i + 2])
373: ++i;
374: if ('\n' == blk.buf[i + 1]) {
375: i += 2;
376: ++lnn;
377: continue;
378: }
379:
380: if ('"' == blk.buf[i + 1] || '#' == blk.buf[i + 1]) {
381: i += 2;
382: /* Comment, skip to end of line */
383: for (; i < (int)blk.sz; ++i) {
384: if ('\n' == blk.buf[i]) {
385: ++i;
386: ++lnn;
387: break;
388: }
389: }
390:
391: /* Backout trailing whitespaces */
392: for (; pos > 0; --pos) {
393: if (ln.buf[pos - 1] != ' ')
394: break;
395: if (pos > 2 && ln.buf[pos - 2] == '\\')
396: break;
397: }
398: break;
399: }
400:
1.13 schwarze 401: /* Catch escaped bogus characters. */
402:
403: c = (unsigned char) blk.buf[i+1];
404:
1.25 schwarze 405: if ( ! (isascii(c) &&
406: (isgraph(c) || isblank(c)))) {
1.13 schwarze 407: mandoc_msg(MANDOCERR_BADCHAR, curp,
1.25 schwarze 408: curp->line, pos, NULL);
1.13 schwarze 409: i += 2;
410: ln.buf[pos++] = '?';
411: continue;
412: }
413:
1.1 schwarze 414: /* Some other escape sequence, copy & cont. */
415:
416: ln.buf[pos++] = blk.buf[i++];
417: ln.buf[pos++] = blk.buf[i++];
418: }
419:
1.25 schwarze 420: if (pos >= (int)ln.sz)
1.1 schwarze 421: resize_buf(&ln, 256);
422:
423: ln.buf[pos] = '\0';
424:
425: /*
426: * A significant amount of complexity is contained by
427: * the roff preprocessor. It's line-oriented but can be
428: * expressed on one line, so we need at times to
429: * readjust our starting point and re-run it. The roff
430: * preprocessor can also readjust the buffers with new
431: * data, so we pass them in wholesale.
432: */
433:
434: of = 0;
435:
1.4 schwarze 436: /*
437: * Maintain a lookaside buffer of all parsed lines. We
438: * only do this if mparse_keep() has been invoked (the
439: * buffer may be accessed with mparse_getkeep()).
440: */
441:
442: if (curp->secondary) {
1.25 schwarze 443: curp->secondary->buf = mandoc_realloc(
444: curp->secondary->buf,
445: curp->secondary->sz + pos + 2);
446: memcpy(curp->secondary->buf +
447: curp->secondary->sz,
448: ln.buf, pos);
1.4 schwarze 449: curp->secondary->sz += pos;
450: curp->secondary->buf
451: [curp->secondary->sz] = '\n';
452: curp->secondary->sz++;
453: curp->secondary->buf
454: [curp->secondary->sz] = '\0';
455: }
1.1 schwarze 456: rerun:
1.25 schwarze 457: rr = roff_parseln(curp->roff, curp->line,
458: &ln.buf, &ln.sz, of, &of);
1.1 schwarze 459:
460: switch (rr) {
1.25 schwarze 461: case ROFF_REPARSE:
1.1 schwarze 462: if (REPARSE_LIMIT >= ++curp->reparse_count)
463: mparse_buf_r(curp, ln, 0);
464: else
465: mandoc_msg(MANDOCERR_ROFFLOOP, curp,
1.25 schwarze 466: curp->line, pos, NULL);
1.1 schwarze 467: pos = 0;
468: continue;
1.25 schwarze 469: case ROFF_APPEND:
1.1 schwarze 470: pos = (int)strlen(ln.buf);
471: continue;
1.25 schwarze 472: case ROFF_RERUN:
1.1 schwarze 473: goto rerun;
1.25 schwarze 474: case ROFF_IGN:
1.1 schwarze 475: pos = 0;
476: continue;
1.25 schwarze 477: case ROFF_ERR:
1.1 schwarze 478: assert(MANDOCLEVEL_FATAL <= curp->file_status);
479: break;
1.25 schwarze 480: case ROFF_SO:
1.23 schwarze 481: if (0 == (MPARSE_SO & curp->options) &&
482: (i >= (int)blk.sz || '\0' == blk.buf[i])) {
483: curp->sodest = mandoc_strdup(ln.buf + of);
484: free(ln.buf);
485: return;
486: }
1.4 schwarze 487: /*
488: * We remove `so' clauses from our lookaside
489: * buffer because we're going to descend into
490: * the file recursively.
491: */
1.25 schwarze 492: if (curp->secondary)
1.4 schwarze 493: curp->secondary->sz -= pos + 1;
1.14 schwarze 494: mparse_readfd(curp, -1, ln.buf + of);
1.30 schwarze 495: if (MANDOCLEVEL_FATAL <= curp->file_status) {
496: mandoc_vmsg(MANDOCERR_SO_FAIL,
497: curp, curp->line, pos,
498: ".so %s", ln.buf + of);
1.1 schwarze 499: break;
1.30 schwarze 500: }
1.1 schwarze 501: pos = 0;
502: continue;
503: default:
504: break;
505: }
506:
507: /*
508: * If we encounter errors in the recursive parse, make
509: * sure we don't continue parsing.
510: */
511:
512: if (MANDOCLEVEL_FATAL <= curp->file_status)
513: break;
514:
515: /*
516: * If input parsers have not been allocated, do so now.
1.2 schwarze 517: * We keep these instanced between parsers, but set them
1.1 schwarze 518: * locally per parse routine since we can use different
519: * parsers with each one.
520: */
521:
522: if ( ! (curp->man || curp->mdoc))
523: pset(ln.buf + of, pos - of, curp);
524:
1.25 schwarze 525: /*
1.1 schwarze 526: * Lastly, push down into the parsers themselves. One
527: * of these will have already been set in the pset()
528: * routine.
529: * If libroff returns ROFF_TBL, then add it to the
530: * currently open parse. Since we only get here if
531: * there does exist data (see tbl_data.c), we're
532: * guaranteed that something's been allocated.
533: * Do the same for ROFF_EQN.
534: */
535:
536: rc = -1;
537:
538: if (ROFF_TBL == rr)
539: while (NULL != (span = roff_span(curp->roff))) {
540: rc = curp->man ?
1.25 schwarze 541: man_addspan(curp->man, span) :
542: mdoc_addspan(curp->mdoc, span);
1.1 schwarze 543: if (0 == rc)
544: break;
545: }
546: else if (ROFF_EQN == rr)
1.25 schwarze 547: rc = curp->mdoc ?
548: mdoc_addeqn(curp->mdoc,
549: roff_eqn(curp->roff)) :
550: man_addeqn(curp->man,
551: roff_eqn(curp->roff));
1.1 schwarze 552: else if (curp->man || curp->mdoc)
553: rc = curp->man ?
1.25 schwarze 554: man_parseln(curp->man,
555: curp->line, ln.buf, of) :
556: mdoc_parseln(curp->mdoc,
557: curp->line, ln.buf, of);
1.1 schwarze 558:
559: if (0 == rc) {
560: assert(MANDOCLEVEL_FATAL <= curp->file_status);
561: break;
1.19 schwarze 562: } else if (2 == rc)
563: break;
1.1 schwarze 564:
565: /* Temporary buffers typically are not full. */
566:
567: if (0 == start && '\0' == blk.buf[i])
568: break;
569:
570: /* Start the next input line. */
571:
572: pos = 0;
573: }
574:
575: free(ln.buf);
576: }
577:
578: static int
1.18 schwarze 579: read_whole_file(struct mparse *curp, const char *file, int fd,
580: struct buf *fb, int *with_mmap)
1.1 schwarze 581: {
582: struct stat st;
583: size_t off;
584: ssize_t ssz;
585:
586: if (-1 == fstat(fd, &st)) {
1.18 schwarze 587: curp->file_status = MANDOCLEVEL_SYSERR;
588: if (curp->mmsg)
589: (*curp->mmsg)(MANDOCERR_SYSSTAT, curp->file_status,
590: file, 0, 0, strerror(errno));
1.1 schwarze 591: return(0);
592: }
593:
594: /*
595: * If we're a regular file, try just reading in the whole entry
596: * via mmap(). This is faster than reading it into blocks, and
597: * since each file is only a few bytes to begin with, I'm not
598: * concerned that this is going to tank any machines.
599: */
600:
601: if (S_ISREG(st.st_mode)) {
602: if (st.st_size >= (1U << 31)) {
1.18 schwarze 603: curp->file_status = MANDOCLEVEL_FATAL;
604: if (curp->mmsg)
605: (*curp->mmsg)(MANDOCERR_TOOLARGE,
606: curp->file_status, file, 0, 0, NULL);
1.1 schwarze 607: return(0);
608: }
609: *with_mmap = 1;
610: fb->sz = (size_t)st.st_size;
1.15 schwarze 611: fb->buf = mmap(NULL, fb->sz, PROT_READ, MAP_SHARED, fd, 0);
1.1 schwarze 612: if (fb->buf != MAP_FAILED)
613: return(1);
614: }
615:
616: /*
617: * If this isn't a regular file (like, say, stdin), then we must
618: * go the old way and just read things in bit by bit.
619: */
620:
621: *with_mmap = 0;
622: off = 0;
623: fb->sz = 0;
624: fb->buf = NULL;
625: for (;;) {
626: if (off == fb->sz) {
627: if (fb->sz == (1U << 31)) {
1.18 schwarze 628: curp->file_status = MANDOCLEVEL_FATAL;
629: if (curp->mmsg)
630: (*curp->mmsg)(MANDOCERR_TOOLARGE,
631: curp->file_status,
632: file, 0, 0, NULL);
1.1 schwarze 633: break;
634: }
635: resize_buf(fb, 65536);
636: }
637: ssz = read(fd, fb->buf + (int)off, fb->sz - off);
638: if (ssz == 0) {
639: fb->sz = off;
640: return(1);
641: }
642: if (ssz == -1) {
1.18 schwarze 643: curp->file_status = MANDOCLEVEL_SYSERR;
644: if (curp->mmsg)
645: (*curp->mmsg)(MANDOCERR_SYSREAD,
646: curp->file_status, file, 0, 0,
647: strerror(errno));
1.1 schwarze 648: break;
649: }
650: off += (size_t)ssz;
651: }
652:
653: free(fb->buf);
654: fb->buf = NULL;
655: return(0);
656: }
657:
658: static void
659: mparse_end(struct mparse *curp)
660: {
661:
662: if (MANDOCLEVEL_FATAL <= curp->file_status)
663: return;
664:
1.50 schwarze 665: if (curp->mdoc == NULL &&
666: curp->man == NULL &&
667: curp->sodest == NULL) {
668: if (curp->options & MPARSE_MDOC)
669: curp->mdoc = curp->pmdoc;
670: else {
671: if (curp->pman == NULL)
672: curp->pman = man_alloc(curp->roff, curp,
673: curp->options & MPARSE_QUICK ? 1 : 0);
674: curp->man = curp->pman;
675: }
676: }
677:
1.1 schwarze 678: if (curp->mdoc && ! mdoc_endparse(curp->mdoc)) {
679: assert(MANDOCLEVEL_FATAL <= curp->file_status);
680: return;
681: }
682:
683: if (curp->man && ! man_endparse(curp->man)) {
684: assert(MANDOCLEVEL_FATAL <= curp->file_status);
685: return;
686: }
687:
688: roff_endparse(curp->roff);
689: }
690:
1.15 schwarze 691: static void
692: mparse_parse_buffer(struct mparse *curp, struct buf blk, const char *file)
1.1 schwarze 693: {
694: const char *svfile;
1.14 schwarze 695: static int recursion_depth;
696:
697: if (64 < recursion_depth) {
698: mandoc_msg(MANDOCERR_ROFFLOOP, curp, curp->line, 0, NULL);
1.15 schwarze 699: return;
1.14 schwarze 700: }
1.1 schwarze 701:
1.15 schwarze 702: /* Line number is per-file. */
703: svfile = curp->file;
704: curp->file = file;
705: curp->line = 1;
706: recursion_depth++;
707:
708: mparse_buf_r(curp, blk, 1);
709:
710: if (0 == --recursion_depth && MANDOCLEVEL_FATAL > curp->file_status)
711: mparse_end(curp);
712:
713: curp->file = svfile;
714: }
715:
716: enum mandoclevel
717: mparse_readfd(struct mparse *curp, int fd, const char *file)
718: {
719: struct buf blk;
720: int with_mmap;
721:
1.18 schwarze 722: if (-1 == fd && -1 == (fd = open(file, O_RDONLY, 0))) {
723: curp->file_status = MANDOCLEVEL_SYSERR;
724: if (curp->mmsg)
725: (*curp->mmsg)(MANDOCERR_SYSOPEN,
726: curp->file_status,
727: file, 0, 0, strerror(errno));
728: goto out;
729: }
730:
1.15 schwarze 731: /*
732: * Run for each opened file; may be called more than once for
733: * each full parse sequence if the opened file is nested (i.e.,
734: * from `so'). Simply sucks in the whole file and moves into
735: * the parse phase for the file.
736: */
1.1 schwarze 737:
1.18 schwarze 738: if ( ! read_whole_file(curp, file, fd, &blk, &with_mmap))
1.15 schwarze 739: goto out;
1.1 schwarze 740:
1.15 schwarze 741: mparse_parse_buffer(curp, blk, file);
1.1 schwarze 742:
1.15 schwarze 743: if (with_mmap)
744: munmap(blk.buf, blk.sz);
745: else
746: free(blk.buf);
1.1 schwarze 747:
748: if (STDIN_FILENO != fd && -1 == close(fd))
749: perror(file);
1.14 schwarze 750: out:
1.1 schwarze 751: return(curp->file_status);
752: }
753:
754: struct mparse *
1.22 schwarze 755: mparse_alloc(int options, enum mandoclevel wlevel,
1.47 schwarze 756: mandocmsg mmsg, const char *defos)
1.1 schwarze 757: {
758: struct mparse *curp;
759:
760: assert(wlevel <= MANDOCLEVEL_FATAL);
761:
762: curp = mandoc_calloc(1, sizeof(struct mparse));
763:
1.22 schwarze 764: curp->options = options;
1.1 schwarze 765: curp->wlevel = wlevel;
766: curp->mmsg = mmsg;
1.7 schwarze 767: curp->defos = defos;
1.1 schwarze 768:
1.22 schwarze 769: curp->roff = roff_alloc(curp, options);
1.50 schwarze 770: if (curp->options & MPARSE_MDOC)
771: curp->pmdoc = mdoc_alloc(
772: curp->roff, curp, curp->defos,
773: curp->options & MPARSE_QUICK ? 1 : 0);
774: if (curp->options & MPARSE_MAN)
775: curp->pman = man_alloc(curp->roff, curp,
776: curp->options & MPARSE_QUICK ? 1 : 0);
777:
1.1 schwarze 778: return(curp);
779: }
780:
781: void
782: mparse_reset(struct mparse *curp)
783: {
784:
785: roff_reset(curp->roff);
786:
787: if (curp->mdoc)
788: mdoc_reset(curp->mdoc);
789: if (curp->man)
790: man_reset(curp->man);
1.4 schwarze 791: if (curp->secondary)
792: curp->secondary->sz = 0;
1.1 schwarze 793:
794: curp->file_status = MANDOCLEVEL_OK;
795: curp->mdoc = NULL;
796: curp->man = NULL;
1.23 schwarze 797:
798: free(curp->sodest);
799: curp->sodest = NULL;
1.1 schwarze 800: }
801:
802: void
803: mparse_free(struct mparse *curp)
804: {
805:
806: if (curp->pmdoc)
807: mdoc_free(curp->pmdoc);
808: if (curp->pman)
809: man_free(curp->pman);
810: if (curp->roff)
811: roff_free(curp->roff);
1.4 schwarze 812: if (curp->secondary)
813: free(curp->secondary->buf);
1.1 schwarze 814:
1.4 schwarze 815: free(curp->secondary);
1.23 schwarze 816: free(curp->sodest);
1.1 schwarze 817: free(curp);
818: }
819:
820: void
1.23 schwarze 821: mparse_result(struct mparse *curp,
822: struct mdoc **mdoc, struct man **man, char **sodest)
1.1 schwarze 823: {
824:
1.23 schwarze 825: if (sodest && NULL != (*sodest = curp->sodest)) {
826: *mdoc = NULL;
827: *man = NULL;
828: return;
829: }
1.1 schwarze 830: if (mdoc)
831: *mdoc = curp->mdoc;
832: if (man)
833: *man = curp->man;
834: }
835:
836: void
837: mandoc_vmsg(enum mandocerr t, struct mparse *m,
838: int ln, int pos, const char *fmt, ...)
839: {
840: char buf[256];
841: va_list ap;
842:
843: va_start(ap, fmt);
1.26 schwarze 844: (void)vsnprintf(buf, sizeof(buf), fmt, ap);
1.1 schwarze 845: va_end(ap);
846:
847: mandoc_msg(t, m, ln, pos, buf);
848: }
849:
850: void
1.25 schwarze 851: mandoc_msg(enum mandocerr er, struct mparse *m,
1.1 schwarze 852: int ln, int col, const char *msg)
853: {
854: enum mandoclevel level;
855:
856: level = MANDOCLEVEL_FATAL;
857: while (er < mandoclimits[level])
858: level--;
859:
860: if (level < m->wlevel)
861: return;
862:
863: if (m->mmsg)
864: (*m->mmsg)(er, level, m->file, ln, col, msg);
865:
866: if (m->file_status < level)
867: m->file_status = level;
868: }
869:
870: const char *
871: mparse_strerror(enum mandocerr er)
872: {
873:
874: return(mandocerrs[er]);
875: }
876:
877: const char *
878: mparse_strlevel(enum mandoclevel lvl)
879: {
880: return(mandoclevels[lvl]);
1.4 schwarze 881: }
882:
883: void
884: mparse_keep(struct mparse *p)
885: {
886:
887: assert(NULL == p->secondary);
888: p->secondary = mandoc_calloc(1, sizeof(struct buf));
889: }
890:
891: const char *
892: mparse_getkeep(const struct mparse *p)
893: {
894:
895: assert(p->secondary);
896: return(p->secondary->sz ? p->secondary->buf : NULL);
1.1 schwarze 897: }