Annotation of src/usr.bin/mandoc/read.c, Revision 1.41
1.41 ! schwarze 1: /* $Id: read.c,v 1.40 2014/07/05 01:11:33 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 */
1.32 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",
131: "empty head in list item",
132: "empty list item",
1.39 schwarze 133: "missing font type, using \\fR",
134: "unknown font type, using \\fR",
1.38 schwarze 135: "missing -std argument, adding it",
1.1 schwarze 136:
137: /* related to bad macro arguments */
138: "skipping argument",
139: "duplicate argument",
1.41 ! schwarze 140: "skipping duplicate display type",
! 141: "skipping duplicate list type",
1.1 schwarze 142: "unknown AT&T UNIX version",
1.41 ! schwarze 143: "invalid Boolean argument",
! 144: "unknown font, skipping request",
1.1 schwarze 145:
146: /* related to plain text */
147: "blank line in non-literal context",
148: "tab in non-literal context",
149: "end of line whitespace",
150: "bad comment style",
1.2 schwarze 151: "bad escape sequence",
1.1 schwarze 152: "unterminated quoted string",
1.3 schwarze 153:
1.1 schwarze 154: "generic error",
155:
1.3 schwarze 156: /* related to equations */
157: "unexpected equation scope closure",
158: "equation scope open on exit",
159: "overlapping equation scopes",
160: "unexpected end of equation",
161: "equation syntax error",
162:
1.1 schwarze 163: /* related to tables */
164: "bad table syntax",
165: "bad table option",
166: "bad table layout",
167: "no table layout cells specified",
168: "no table data cells specified",
169: "ignore data in cell",
170: "data block still open",
171: "ignoring extra data cells",
172:
173: "input stack limit exceeded, infinite loop?",
174: "skipping bad character",
175: "escaped character not allowed in a name",
1.9 schwarze 176: "manual name not yet set",
1.32 schwarze 177: "skipping text before first section header",
1.1 schwarze 178: "skipping unknown macro",
179: "NOT IMPLEMENTED, please use groff: skipping request",
180: "argument count wrong",
1.35 schwarze 181: "skipping invalid content in .Rs block",
1.41 ! schwarze 182: "unknown standard specifier",
1.12 schwarze 183: "skipping column outside column list",
1.1 schwarze 184: "skipping end of block that is not open",
185: "missing end of block",
186: "scope open on exit",
187: "uname(3) system call failed",
1.16 schwarze 188: "request requires a numeric argument",
1.38 schwarze 189: "missing list type, using -item",
1.39 schwarze 190: "skipping all arguments",
191: "skipping excess arguments",
1.1 schwarze 192:
193: "generic fatal error",
194:
1.18 schwarze 195: "input too large",
1.1 schwarze 196: "not a manual",
197: "column syntax is inconsistent",
198: "NOT IMPLEMENTED: .Bd -file",
199: "child violates parent syntax",
200: "argument count wrong, violates syntax",
201: "NOT IMPLEMENTED: .so with absolute path or \"..\"",
1.30 schwarze 202: ".so request failed",
1.1 schwarze 203: "no document prologue",
204: "static buffer exhausted",
1.18 schwarze 205:
206: /* system errors */
1.29 schwarze 207: NULL,
1.18 schwarze 208: "cannot stat file",
209: "cannot read file",
1.1 schwarze 210: };
211:
212: static const char * const mandoclevels[MANDOCLEVEL_MAX] = {
213: "SUCCESS",
214: "RESERVED",
215: "WARNING",
216: "ERROR",
217: "FATAL",
218: "BADARG",
219: "SYSERR"
220: };
221:
1.25 schwarze 222:
1.1 schwarze 223: static void
224: resize_buf(struct buf *buf, size_t initial)
225: {
226:
227: buf->sz = buf->sz > initial/2 ? 2 * buf->sz : initial;
228: buf->buf = mandoc_realloc(buf->buf, buf->sz);
229: }
230:
231: static void
232: pset(const char *buf, int pos, struct mparse *curp)
233: {
234: int i;
235:
236: /*
237: * Try to intuit which kind of manual parser should be used. If
238: * passed in by command-line (-man, -mdoc), then use that
239: * explicitly. If passed as -mandoc, then try to guess from the
240: * line: either skip dot-lines, use -mdoc when finding `.Dt', or
241: * default to -man, which is more lenient.
242: *
243: * Separate out pmdoc/pman from mdoc/man: the first persists
244: * through all parsers, while the latter is used per-parse.
245: */
246:
247: if ('.' == buf[0] || '\'' == buf[0]) {
248: for (i = 1; buf[i]; i++)
249: if (' ' != buf[i] && '\t' != buf[i])
250: break;
251: if ('\0' == buf[i])
252: return;
253: }
254:
1.22 schwarze 255: if (MPARSE_MDOC & curp->options) {
1.25 schwarze 256: if (NULL == curp->pmdoc)
1.22 schwarze 257: curp->pmdoc = mdoc_alloc(
258: curp->roff, curp, curp->defos,
259: MPARSE_QUICK & curp->options ? 1 : 0);
1.1 schwarze 260: assert(curp->pmdoc);
261: curp->mdoc = curp->pmdoc;
262: return;
1.22 schwarze 263: } else if (MPARSE_MAN & curp->options) {
1.25 schwarze 264: if (NULL == curp->pman)
1.19 schwarze 265: curp->pman = man_alloc(curp->roff, curp,
1.22 schwarze 266: MPARSE_QUICK & curp->options ? 1 : 0);
1.1 schwarze 267: assert(curp->pman);
268: curp->man = curp->pman;
269: return;
270: }
271:
272: if (pos >= 3 && 0 == memcmp(buf, ".Dd", 3)) {
1.25 schwarze 273: if (NULL == curp->pmdoc)
1.22 schwarze 274: curp->pmdoc = mdoc_alloc(
275: curp->roff, curp, curp->defos,
276: MPARSE_QUICK & curp->options ? 1 : 0);
1.1 schwarze 277: assert(curp->pmdoc);
278: curp->mdoc = curp->pmdoc;
279: return;
1.25 schwarze 280: }
1.1 schwarze 281:
1.25 schwarze 282: if (NULL == curp->pman)
1.22 schwarze 283: curp->pman = man_alloc(curp->roff, curp,
284: MPARSE_QUICK & curp->options ? 1 : 0);
1.1 schwarze 285: assert(curp->pman);
286: curp->man = curp->pman;
287: }
288:
289: /*
290: * Main parse routine for an opened file. This is called for each
291: * opened file and simply loops around the full input file, possibly
292: * nesting (i.e., with `so').
293: */
294: static void
295: mparse_buf_r(struct mparse *curp, struct buf blk, int start)
296: {
297: const struct tbl_span *span;
298: struct buf ln;
299: enum rofferr rr;
300: int i, of, rc;
301: int pos; /* byte number in the ln buffer */
302: int lnn; /* line number in the real file */
303: unsigned char c;
304:
305: memset(&ln, 0, sizeof(struct buf));
306:
1.25 schwarze 307: lnn = curp->line;
308: pos = 0;
1.1 schwarze 309:
310: for (i = 0; i < (int)blk.sz; ) {
311: if (0 == pos && '\0' == blk.buf[i])
312: break;
313:
314: if (start) {
315: curp->line = lnn;
316: curp->reparse_count = 0;
317: }
318:
319: while (i < (int)blk.sz && (start || '\0' != blk.buf[i])) {
320:
321: /*
322: * When finding an unescaped newline character,
323: * leave the character loop to process the line.
324: * Skip a preceding carriage return, if any.
325: */
326:
327: if ('\r' == blk.buf[i] && i + 1 < (int)blk.sz &&
328: '\n' == blk.buf[i + 1])
329: ++i;
330: if ('\n' == blk.buf[i]) {
331: ++i;
332: ++lnn;
333: break;
334: }
335:
1.13 schwarze 336: /*
337: * Make sure we have space for at least
338: * one backslash and one other character
339: * and the trailing NUL byte.
340: */
341:
342: if (pos + 2 >= (int)ln.sz)
343: resize_buf(&ln, 256);
344:
1.25 schwarze 345: /*
1.1 schwarze 346: * Warn about bogus characters. If you're using
347: * non-ASCII encoding, you're screwing your
348: * readers. Since I'd rather this not happen,
1.6 schwarze 349: * I'll be helpful and replace these characters
350: * with "?", so we don't display gibberish.
351: * Note to manual writers: use special characters.
1.1 schwarze 352: */
353:
354: c = (unsigned char) blk.buf[i];
355:
1.25 schwarze 356: if ( ! (isascii(c) &&
357: (isgraph(c) || isblank(c)))) {
1.1 schwarze 358: mandoc_msg(MANDOCERR_BADCHAR, curp,
1.25 schwarze 359: curp->line, pos, NULL);
1.1 schwarze 360: i++;
1.6 schwarze 361: ln.buf[pos++] = '?';
1.1 schwarze 362: continue;
363: }
364:
365: /* Trailing backslash = a plain char. */
366:
367: if ('\\' != blk.buf[i] || i + 1 == (int)blk.sz) {
368: ln.buf[pos++] = blk.buf[i++];
369: continue;
370: }
371:
372: /*
373: * Found escape and at least one other character.
374: * When it's a newline character, skip it.
375: * When there is a carriage return in between,
376: * skip that one as well.
377: */
378:
379: if ('\r' == blk.buf[i + 1] && i + 2 < (int)blk.sz &&
380: '\n' == blk.buf[i + 2])
381: ++i;
382: if ('\n' == blk.buf[i + 1]) {
383: i += 2;
384: ++lnn;
385: continue;
386: }
387:
388: if ('"' == blk.buf[i + 1] || '#' == blk.buf[i + 1]) {
389: i += 2;
390: /* Comment, skip to end of line */
391: for (; i < (int)blk.sz; ++i) {
392: if ('\n' == blk.buf[i]) {
393: ++i;
394: ++lnn;
395: break;
396: }
397: }
398:
399: /* Backout trailing whitespaces */
400: for (; pos > 0; --pos) {
401: if (ln.buf[pos - 1] != ' ')
402: break;
403: if (pos > 2 && ln.buf[pos - 2] == '\\')
404: break;
405: }
406: break;
407: }
408:
1.13 schwarze 409: /* Catch escaped bogus characters. */
410:
411: c = (unsigned char) blk.buf[i+1];
412:
1.25 schwarze 413: if ( ! (isascii(c) &&
414: (isgraph(c) || isblank(c)))) {
1.13 schwarze 415: mandoc_msg(MANDOCERR_BADCHAR, curp,
1.25 schwarze 416: curp->line, pos, NULL);
1.13 schwarze 417: i += 2;
418: ln.buf[pos++] = '?';
419: continue;
420: }
421:
1.1 schwarze 422: /* Some other escape sequence, copy & cont. */
423:
424: ln.buf[pos++] = blk.buf[i++];
425: ln.buf[pos++] = blk.buf[i++];
426: }
427:
1.25 schwarze 428: if (pos >= (int)ln.sz)
1.1 schwarze 429: resize_buf(&ln, 256);
430:
431: ln.buf[pos] = '\0';
432:
433: /*
434: * A significant amount of complexity is contained by
435: * the roff preprocessor. It's line-oriented but can be
436: * expressed on one line, so we need at times to
437: * readjust our starting point and re-run it. The roff
438: * preprocessor can also readjust the buffers with new
439: * data, so we pass them in wholesale.
440: */
441:
442: of = 0;
443:
1.4 schwarze 444: /*
445: * Maintain a lookaside buffer of all parsed lines. We
446: * only do this if mparse_keep() has been invoked (the
447: * buffer may be accessed with mparse_getkeep()).
448: */
449:
450: if (curp->secondary) {
1.25 schwarze 451: curp->secondary->buf = mandoc_realloc(
452: curp->secondary->buf,
453: curp->secondary->sz + pos + 2);
454: memcpy(curp->secondary->buf +
455: curp->secondary->sz,
456: ln.buf, pos);
1.4 schwarze 457: curp->secondary->sz += pos;
458: curp->secondary->buf
459: [curp->secondary->sz] = '\n';
460: curp->secondary->sz++;
461: curp->secondary->buf
462: [curp->secondary->sz] = '\0';
463: }
1.1 schwarze 464: rerun:
1.25 schwarze 465: rr = roff_parseln(curp->roff, curp->line,
466: &ln.buf, &ln.sz, of, &of);
1.1 schwarze 467:
468: switch (rr) {
1.25 schwarze 469: case ROFF_REPARSE:
1.1 schwarze 470: if (REPARSE_LIMIT >= ++curp->reparse_count)
471: mparse_buf_r(curp, ln, 0);
472: else
473: mandoc_msg(MANDOCERR_ROFFLOOP, curp,
1.25 schwarze 474: curp->line, pos, NULL);
1.1 schwarze 475: pos = 0;
476: continue;
1.25 schwarze 477: case ROFF_APPEND:
1.1 schwarze 478: pos = (int)strlen(ln.buf);
479: continue;
1.25 schwarze 480: case ROFF_RERUN:
1.1 schwarze 481: goto rerun;
1.25 schwarze 482: case ROFF_IGN:
1.1 schwarze 483: pos = 0;
484: continue;
1.25 schwarze 485: case ROFF_ERR:
1.1 schwarze 486: assert(MANDOCLEVEL_FATAL <= curp->file_status);
487: break;
1.25 schwarze 488: case ROFF_SO:
1.23 schwarze 489: if (0 == (MPARSE_SO & curp->options) &&
490: (i >= (int)blk.sz || '\0' == blk.buf[i])) {
491: curp->sodest = mandoc_strdup(ln.buf + of);
492: free(ln.buf);
493: return;
494: }
1.4 schwarze 495: /*
496: * We remove `so' clauses from our lookaside
497: * buffer because we're going to descend into
498: * the file recursively.
499: */
1.25 schwarze 500: if (curp->secondary)
1.4 schwarze 501: curp->secondary->sz -= pos + 1;
1.14 schwarze 502: mparse_readfd(curp, -1, ln.buf + of);
1.30 schwarze 503: if (MANDOCLEVEL_FATAL <= curp->file_status) {
504: mandoc_vmsg(MANDOCERR_SO_FAIL,
505: curp, curp->line, pos,
506: ".so %s", ln.buf + of);
1.1 schwarze 507: break;
1.30 schwarze 508: }
1.1 schwarze 509: pos = 0;
510: continue;
511: default:
512: break;
513: }
514:
515: /*
516: * If we encounter errors in the recursive parse, make
517: * sure we don't continue parsing.
518: */
519:
520: if (MANDOCLEVEL_FATAL <= curp->file_status)
521: break;
522:
523: /*
524: * If input parsers have not been allocated, do so now.
1.2 schwarze 525: * We keep these instanced between parsers, but set them
1.1 schwarze 526: * locally per parse routine since we can use different
527: * parsers with each one.
528: */
529:
530: if ( ! (curp->man || curp->mdoc))
531: pset(ln.buf + of, pos - of, curp);
532:
1.25 schwarze 533: /*
1.1 schwarze 534: * Lastly, push down into the parsers themselves. One
535: * of these will have already been set in the pset()
536: * routine.
537: * If libroff returns ROFF_TBL, then add it to the
538: * currently open parse. Since we only get here if
539: * there does exist data (see tbl_data.c), we're
540: * guaranteed that something's been allocated.
541: * Do the same for ROFF_EQN.
542: */
543:
544: rc = -1;
545:
546: if (ROFF_TBL == rr)
547: while (NULL != (span = roff_span(curp->roff))) {
548: rc = curp->man ?
1.25 schwarze 549: man_addspan(curp->man, span) :
550: mdoc_addspan(curp->mdoc, span);
1.1 schwarze 551: if (0 == rc)
552: break;
553: }
554: else if (ROFF_EQN == rr)
1.25 schwarze 555: rc = curp->mdoc ?
556: mdoc_addeqn(curp->mdoc,
557: roff_eqn(curp->roff)) :
558: man_addeqn(curp->man,
559: roff_eqn(curp->roff));
1.1 schwarze 560: else if (curp->man || curp->mdoc)
561: rc = curp->man ?
1.25 schwarze 562: man_parseln(curp->man,
563: curp->line, ln.buf, of) :
564: mdoc_parseln(curp->mdoc,
565: curp->line, ln.buf, of);
1.1 schwarze 566:
567: if (0 == rc) {
568: assert(MANDOCLEVEL_FATAL <= curp->file_status);
569: break;
1.19 schwarze 570: } else if (2 == rc)
571: break;
1.1 schwarze 572:
573: /* Temporary buffers typically are not full. */
574:
575: if (0 == start && '\0' == blk.buf[i])
576: break;
577:
578: /* Start the next input line. */
579:
580: pos = 0;
581: }
582:
583: free(ln.buf);
584: }
585:
586: static int
1.18 schwarze 587: read_whole_file(struct mparse *curp, const char *file, int fd,
588: struct buf *fb, int *with_mmap)
1.1 schwarze 589: {
590: struct stat st;
591: size_t off;
592: ssize_t ssz;
593:
594: if (-1 == fstat(fd, &st)) {
1.18 schwarze 595: curp->file_status = MANDOCLEVEL_SYSERR;
596: if (curp->mmsg)
597: (*curp->mmsg)(MANDOCERR_SYSSTAT, curp->file_status,
598: file, 0, 0, strerror(errno));
1.1 schwarze 599: return(0);
600: }
601:
602: /*
603: * If we're a regular file, try just reading in the whole entry
604: * via mmap(). This is faster than reading it into blocks, and
605: * since each file is only a few bytes to begin with, I'm not
606: * concerned that this is going to tank any machines.
607: */
608:
609: if (S_ISREG(st.st_mode)) {
610: if (st.st_size >= (1U << 31)) {
1.18 schwarze 611: curp->file_status = MANDOCLEVEL_FATAL;
612: if (curp->mmsg)
613: (*curp->mmsg)(MANDOCERR_TOOLARGE,
614: curp->file_status, file, 0, 0, NULL);
1.1 schwarze 615: return(0);
616: }
617: *with_mmap = 1;
618: fb->sz = (size_t)st.st_size;
1.15 schwarze 619: fb->buf = mmap(NULL, fb->sz, PROT_READ, MAP_SHARED, fd, 0);
1.1 schwarze 620: if (fb->buf != MAP_FAILED)
621: return(1);
622: }
623:
624: /*
625: * If this isn't a regular file (like, say, stdin), then we must
626: * go the old way and just read things in bit by bit.
627: */
628:
629: *with_mmap = 0;
630: off = 0;
631: fb->sz = 0;
632: fb->buf = NULL;
633: for (;;) {
634: if (off == fb->sz) {
635: if (fb->sz == (1U << 31)) {
1.18 schwarze 636: curp->file_status = MANDOCLEVEL_FATAL;
637: if (curp->mmsg)
638: (*curp->mmsg)(MANDOCERR_TOOLARGE,
639: curp->file_status,
640: file, 0, 0, NULL);
1.1 schwarze 641: break;
642: }
643: resize_buf(fb, 65536);
644: }
645: ssz = read(fd, fb->buf + (int)off, fb->sz - off);
646: if (ssz == 0) {
647: fb->sz = off;
648: return(1);
649: }
650: if (ssz == -1) {
1.18 schwarze 651: curp->file_status = MANDOCLEVEL_SYSERR;
652: if (curp->mmsg)
653: (*curp->mmsg)(MANDOCERR_SYSREAD,
654: curp->file_status, file, 0, 0,
655: strerror(errno));
1.1 schwarze 656: break;
657: }
658: off += (size_t)ssz;
659: }
660:
661: free(fb->buf);
662: fb->buf = NULL;
663: return(0);
664: }
665:
666: static void
667: mparse_end(struct mparse *curp)
668: {
669:
670: if (MANDOCLEVEL_FATAL <= curp->file_status)
671: return;
672:
673: if (curp->mdoc && ! mdoc_endparse(curp->mdoc)) {
674: assert(MANDOCLEVEL_FATAL <= curp->file_status);
675: return;
676: }
677:
678: if (curp->man && ! man_endparse(curp->man)) {
679: assert(MANDOCLEVEL_FATAL <= curp->file_status);
680: return;
681: }
682:
1.23 schwarze 683: if ( ! (curp->mdoc || curp->man || curp->sodest)) {
1.28 schwarze 684: mandoc_msg(MANDOCERR_NOTMANUAL, curp, 0, 0, NULL);
1.1 schwarze 685: curp->file_status = MANDOCLEVEL_FATAL;
686: return;
687: }
688:
689: roff_endparse(curp->roff);
690: }
691:
1.15 schwarze 692: static void
693: mparse_parse_buffer(struct mparse *curp, struct buf blk, const char *file)
1.1 schwarze 694: {
695: const char *svfile;
1.14 schwarze 696: static int recursion_depth;
697:
698: if (64 < recursion_depth) {
699: mandoc_msg(MANDOCERR_ROFFLOOP, curp, curp->line, 0, NULL);
1.15 schwarze 700: return;
1.14 schwarze 701: }
1.1 schwarze 702:
1.15 schwarze 703: /* Line number is per-file. */
704: svfile = curp->file;
705: curp->file = file;
706: curp->line = 1;
707: recursion_depth++;
708:
709: mparse_buf_r(curp, blk, 1);
710:
711: if (0 == --recursion_depth && MANDOCLEVEL_FATAL > curp->file_status)
712: mparse_end(curp);
713:
714: curp->file = svfile;
715: }
716:
717: enum mandoclevel
718: mparse_readfd(struct mparse *curp, int fd, const char *file)
719: {
720: struct buf blk;
721: int with_mmap;
722:
1.18 schwarze 723: if (-1 == fd && -1 == (fd = open(file, O_RDONLY, 0))) {
724: curp->file_status = MANDOCLEVEL_SYSERR;
725: if (curp->mmsg)
726: (*curp->mmsg)(MANDOCERR_SYSOPEN,
727: curp->file_status,
728: file, 0, 0, strerror(errno));
729: goto out;
730: }
731:
1.15 schwarze 732: /*
733: * Run for each opened file; may be called more than once for
734: * each full parse sequence if the opened file is nested (i.e.,
735: * from `so'). Simply sucks in the whole file and moves into
736: * the parse phase for the file.
737: */
1.1 schwarze 738:
1.18 schwarze 739: if ( ! read_whole_file(curp, file, fd, &blk, &with_mmap))
1.15 schwarze 740: goto out;
1.1 schwarze 741:
1.15 schwarze 742: mparse_parse_buffer(curp, blk, file);
1.1 schwarze 743:
1.15 schwarze 744: if (with_mmap)
745: munmap(blk.buf, blk.sz);
746: else
747: free(blk.buf);
1.1 schwarze 748:
749: if (STDIN_FILENO != fd && -1 == close(fd))
750: perror(file);
1.14 schwarze 751: out:
1.1 schwarze 752: return(curp->file_status);
753: }
754:
755: struct mparse *
1.22 schwarze 756: mparse_alloc(int options, enum mandoclevel wlevel,
757: mandocmsg mmsg, char *defos)
1.1 schwarze 758: {
759: struct mparse *curp;
760:
761: assert(wlevel <= MANDOCLEVEL_FATAL);
762:
763: curp = mandoc_calloc(1, sizeof(struct mparse));
764:
1.22 schwarze 765: curp->options = options;
1.1 schwarze 766: curp->wlevel = wlevel;
767: curp->mmsg = mmsg;
1.7 schwarze 768: curp->defos = defos;
1.1 schwarze 769:
1.22 schwarze 770: curp->roff = roff_alloc(curp, options);
1.1 schwarze 771: return(curp);
772: }
773:
774: void
775: mparse_reset(struct mparse *curp)
776: {
777:
778: roff_reset(curp->roff);
779:
780: if (curp->mdoc)
781: mdoc_reset(curp->mdoc);
782: if (curp->man)
783: man_reset(curp->man);
1.4 schwarze 784: if (curp->secondary)
785: curp->secondary->sz = 0;
1.1 schwarze 786:
787: curp->file_status = MANDOCLEVEL_OK;
788: curp->mdoc = NULL;
789: curp->man = NULL;
1.23 schwarze 790:
791: free(curp->sodest);
792: curp->sodest = NULL;
1.1 schwarze 793: }
794:
795: void
796: mparse_free(struct mparse *curp)
797: {
798:
799: if (curp->pmdoc)
800: mdoc_free(curp->pmdoc);
801: if (curp->pman)
802: man_free(curp->pman);
803: if (curp->roff)
804: roff_free(curp->roff);
1.4 schwarze 805: if (curp->secondary)
806: free(curp->secondary->buf);
1.1 schwarze 807:
1.4 schwarze 808: free(curp->secondary);
1.23 schwarze 809: free(curp->sodest);
1.1 schwarze 810: free(curp);
811: }
812:
813: void
1.23 schwarze 814: mparse_result(struct mparse *curp,
815: struct mdoc **mdoc, struct man **man, char **sodest)
1.1 schwarze 816: {
817:
1.23 schwarze 818: if (sodest && NULL != (*sodest = curp->sodest)) {
819: *mdoc = NULL;
820: *man = NULL;
821: return;
822: }
1.1 schwarze 823: if (mdoc)
824: *mdoc = curp->mdoc;
825: if (man)
826: *man = curp->man;
827: }
828:
829: void
830: mandoc_vmsg(enum mandocerr t, struct mparse *m,
831: int ln, int pos, const char *fmt, ...)
832: {
833: char buf[256];
834: va_list ap;
835:
836: va_start(ap, fmt);
1.26 schwarze 837: (void)vsnprintf(buf, sizeof(buf), fmt, ap);
1.1 schwarze 838: va_end(ap);
839:
840: mandoc_msg(t, m, ln, pos, buf);
841: }
842:
843: void
1.25 schwarze 844: mandoc_msg(enum mandocerr er, struct mparse *m,
1.1 schwarze 845: int ln, int col, const char *msg)
846: {
847: enum mandoclevel level;
848:
849: level = MANDOCLEVEL_FATAL;
850: while (er < mandoclimits[level])
851: level--;
852:
853: if (level < m->wlevel)
854: return;
855:
856: if (m->mmsg)
857: (*m->mmsg)(er, level, m->file, ln, col, msg);
858:
859: if (m->file_status < level)
860: m->file_status = level;
861: }
862:
863: const char *
864: mparse_strerror(enum mandocerr er)
865: {
866:
867: return(mandocerrs[er]);
868: }
869:
870: const char *
871: mparse_strlevel(enum mandoclevel lvl)
872: {
873: return(mandoclevels[lvl]);
1.4 schwarze 874: }
875:
876: void
877: mparse_keep(struct mparse *p)
878: {
879:
880: assert(NULL == p->secondary);
881: p->secondary = mandoc_calloc(1, sizeof(struct buf));
882: }
883:
884: const char *
885: mparse_getkeep(const struct mparse *p)
886: {
887:
888: assert(p->secondary);
889: return(p->secondary->sz ? p->secondary->buf : NULL);
1.1 schwarze 890: }