Annotation of mandoc/read.c, Revision 1.70
1.70 ! schwarze 1: /* $Id: read.c,v 1.69 2014/07/09 11:31:43 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"
1.46 schwarze 40: #include "mandoc_aux.h"
1.3 kristaps 41: #include "libmandoc.h"
1.1 kristaps 42: #include "mdoc.h"
43: #include "man.h"
1.28 joerg 44: #include "main.h"
1.1 kristaps 45:
46: #define REPARSE_LIMIT 1000
47:
48: struct buf {
1.47 schwarze 49: char *buf; /* binary input buffer */
1.1 kristaps 50: size_t sz; /* size of binary buffer */
51: };
52:
53: struct mparse {
54: enum mandoclevel file_status; /* status of current parse */
1.3 kristaps 55: enum mandoclevel wlevel; /* ignore messages below this */
1.1 kristaps 56: int line; /* line number in the file */
1.44 schwarze 57: int options; /* parser options */
1.1 kristaps 58: struct man *pman; /* persistent man parser */
59: struct mdoc *pmdoc; /* persistent mdoc parser */
60: struct man *man; /* man parser */
61: struct mdoc *mdoc; /* mdoc parser */
62: struct roff *roff; /* roff parser (!NULL) */
1.45 schwarze 63: char *sodest; /* filename pointed to by .so */
1.1 kristaps 64: int reparse_count; /* finite interp. stack */
65: mandocmsg mmsg; /* warning/error message handler */
1.47 schwarze 66: const char *file;
1.24 kristaps 67: struct buf *secondary;
1.69 schwarze 68: const char *defos; /* default operating system */
1.1 kristaps 69: };
70:
71: static void resize_buf(struct buf *, size_t);
72: static void mparse_buf_r(struct mparse *, struct buf, int);
73: static void pset(const char *, int, struct mparse *);
1.40 schwarze 74: static int read_whole_file(struct mparse *, const char *, int,
75: struct buf *, int *);
1.1 kristaps 76: static void mparse_end(struct mparse *);
1.37 schwarze 77: static void mparse_parse_buffer(struct mparse *, struct buf,
78: const char *);
1.1 kristaps 79:
1.3 kristaps 80: static const enum mandocerr mandoclimits[MANDOCLEVEL_MAX] = {
81: MANDOCERR_OK,
82: MANDOCERR_WARNING,
83: MANDOCERR_WARNING,
84: MANDOCERR_ERROR,
85: MANDOCERR_FATAL,
86: MANDOCERR_MAX,
87: MANDOCERR_MAX
88: };
89:
1.7 kristaps 90: static const char * const mandocerrs[MANDOCERR_MAX] = {
91: "ok",
92:
93: "generic warning",
94:
95: /* related to the prologue */
1.54 schwarze 96: "missing .TH macro, using \"unknown 1\"",
97: "lower case character in document title",
1.7 kristaps 98: "unknown manual section",
1.32 schwarze 99: "unknown manual volume or arch",
1.54 schwarze 100: "missing date, using today's date",
1.7 kristaps 101: "cannot parse date, using it verbatim",
102: "prologue macros out of order",
103: "duplicate prologue macro",
1.54 schwarze 104: "incomplete prologue, terminated by",
105: "skipping prologue macro in body",
1.7 kristaps 106:
107: /* related to document structure */
108: ".so is fragile, better use ln(1)",
1.50 schwarze 109: "no document body",
1.54 schwarze 110: "content before first section header",
111: "first section is not \"NAME\"",
1.7 kristaps 112: "bad NAME section contents",
113: "sections out of conventional order",
1.54 schwarze 114: "duplicate section title",
115: "unexpected section",
1.7 kristaps 116:
117: /* related to macros and nesting */
1.55 schwarze 118: "obsolete macro",
1.7 kristaps 119: "skipping paragraph macro",
1.31 schwarze 120: "moving paragraph macro out of list",
1.7 kristaps 121: "skipping no-space macro",
122: "blocks badly nested",
123: "nested displays are not portable",
1.57 schwarze 124: "moving content out of list",
125: ".Vt block has child macro",
1.56 schwarze 126: "fill mode already enabled, skipping .fi",
127: "fill mode already disabled, skipping .nf",
1.7 kristaps 128: "line scope broken",
129:
130: /* related to missing macro arguments */
1.58 schwarze 131: "skipping empty request",
132: "conditional request controls empty scope",
1.7 kristaps 133: "skipping empty macro",
1.62 schwarze 134: "empty argument, using 0n",
1.7 kristaps 135: "argument count wrong",
1.60 schwarze 136: "missing display type, using -ragged",
137: "list type is not the first argument",
138: "missing -width in -tag list, using 8n",
1.70 ! schwarze 139: "missing name for .Ex, using \"\"",
1.60 schwarze 140: "empty head in list item",
141: "empty list item",
1.61 schwarze 142: "missing font type, using \\fR",
143: "unknown font type, using \\fR",
1.60 schwarze 144: "missing -std argument, adding it",
1.7 kristaps 145:
146: /* related to bad macro arguments */
147: "skipping argument",
1.64 schwarze 148: "unterminated quoted argument",
1.7 kristaps 149: "duplicate argument",
1.63 schwarze 150: "skipping duplicate display type",
151: "skipping duplicate list type",
1.7 kristaps 152: "unknown AT&T UNIX version",
1.67 schwarze 153: "invalid content in Rs block",
1.63 schwarze 154: "invalid Boolean argument",
155: "unknown font, skipping request",
1.7 kristaps 156:
157: /* related to plain text */
1.64 schwarze 158: "blank line in fill mode, using .sp",
159: "tab in filled text",
160: "whitespace at end of input line",
1.7 kristaps 161: "bad comment style",
1.64 schwarze 162: "invalid escape sequence",
163: "undefined string, using \"\"",
1.16 kristaps 164:
1.7 kristaps 165: "generic error",
1.17 kristaps 166:
167: /* related to equations */
1.20 kristaps 168: "unexpected equation scope closure",
169: "equation scope open on exit",
1.21 kristaps 170: "overlapping equation scopes",
171: "unexpected end of equation",
172: "equation syntax error",
1.7 kristaps 173:
174: /* related to tables */
175: "bad table syntax",
176: "bad table option",
177: "bad table layout",
178: "no table layout cells specified",
179: "no table data cells specified",
180: "ignore data in cell",
181: "data block still open",
182: "ignoring extra data cells",
183:
1.68 schwarze 184: /* related to document structure and macros */
1.7 kristaps 185: "input stack limit exceeded, infinite loop?",
186: "skipping bad character",
1.68 schwarze 187: "skipping unknown macro",
188: "skipping column outside column list",
189: "skipping end of block that is not open",
190: "inserting missing end of block",
191: "appending missing end of block",
192:
193: /* related to request and macro arguments */
1.7 kristaps 194: "escaped character not allowed in a name",
195: "argument count wrong",
1.70 ! schwarze 196: "missing manual name, using \"\"",
1.63 schwarze 197: "unknown standard specifier",
1.7 kristaps 198: "uname(3) system call failed",
1.38 schwarze 199: "request requires a numeric argument",
1.60 schwarze 200: "missing list type, using -item",
1.61 schwarze 201: "skipping all arguments",
202: "skipping excess arguments",
1.7 kristaps 203:
204: "generic fatal error",
205:
1.40 schwarze 206: "input too large",
1.7 kristaps 207: "not a manual",
208: "column syntax is inconsistent",
209: "NOT IMPLEMENTED: .Bd -file",
210: "child violates parent syntax",
211: "argument count wrong, violates syntax",
212: "NOT IMPLEMENTED: .so with absolute path or \"..\"",
1.52 schwarze 213: ".so request failed",
1.7 kristaps 214: "no document prologue",
215: "static buffer exhausted",
1.40 schwarze 216:
217: /* system errors */
1.51 schwarze 218: NULL,
1.40 schwarze 219: "cannot stat file",
220: "cannot read file",
1.7 kristaps 221: };
222:
223: static const char * const mandoclevels[MANDOCLEVEL_MAX] = {
224: "SUCCESS",
225: "RESERVED",
226: "WARNING",
227: "ERROR",
228: "FATAL",
229: "BADARG",
230: "SYSERR"
231: };
232:
1.47 schwarze 233:
1.1 kristaps 234: static void
235: resize_buf(struct buf *buf, size_t initial)
236: {
237:
238: buf->sz = buf->sz > initial/2 ? 2 * buf->sz : initial;
239: buf->buf = mandoc_realloc(buf->buf, buf->sz);
240: }
241:
242: static void
243: pset(const char *buf, int pos, struct mparse *curp)
244: {
245: int i;
246:
247: /*
248: * Try to intuit which kind of manual parser should be used. If
249: * passed in by command-line (-man, -mdoc), then use that
250: * explicitly. If passed as -mandoc, then try to guess from the
251: * line: either skip dot-lines, use -mdoc when finding `.Dt', or
252: * default to -man, which is more lenient.
253: *
254: * Separate out pmdoc/pman from mdoc/man: the first persists
255: * through all parsers, while the latter is used per-parse.
256: */
257:
258: if ('.' == buf[0] || '\'' == buf[0]) {
259: for (i = 1; buf[i]; i++)
260: if (' ' != buf[i] && '\t' != buf[i])
261: break;
262: if ('\0' == buf[i])
263: return;
264: }
265:
1.44 schwarze 266: if (MPARSE_MDOC & curp->options) {
1.47 schwarze 267: if (NULL == curp->pmdoc)
1.44 schwarze 268: curp->pmdoc = mdoc_alloc(
269: curp->roff, curp, curp->defos,
270: MPARSE_QUICK & curp->options ? 1 : 0);
1.1 kristaps 271: assert(curp->pmdoc);
272: curp->mdoc = curp->pmdoc;
273: return;
1.44 schwarze 274: } else if (MPARSE_MAN & curp->options) {
1.47 schwarze 275: if (NULL == curp->pman)
1.41 schwarze 276: curp->pman = man_alloc(curp->roff, curp,
1.44 schwarze 277: MPARSE_QUICK & curp->options ? 1 : 0);
1.1 kristaps 278: assert(curp->pman);
279: curp->man = curp->pman;
280: return;
281: }
282:
283: if (pos >= 3 && 0 == memcmp(buf, ".Dd", 3)) {
1.47 schwarze 284: if (NULL == curp->pmdoc)
1.44 schwarze 285: curp->pmdoc = mdoc_alloc(
286: curp->roff, curp, curp->defos,
287: MPARSE_QUICK & curp->options ? 1 : 0);
1.1 kristaps 288: assert(curp->pmdoc);
289: curp->mdoc = curp->pmdoc;
290: return;
1.47 schwarze 291: }
1.1 kristaps 292:
1.47 schwarze 293: if (NULL == curp->pman)
1.44 schwarze 294: curp->pman = man_alloc(curp->roff, curp,
295: MPARSE_QUICK & curp->options ? 1 : 0);
1.1 kristaps 296: assert(curp->pman);
297: curp->man = curp->pman;
298: }
299:
300: /*
301: * Main parse routine for an opened file. This is called for each
302: * opened file and simply loops around the full input file, possibly
303: * nesting (i.e., with `so').
304: */
305: static void
306: mparse_buf_r(struct mparse *curp, struct buf blk, int start)
307: {
308: const struct tbl_span *span;
309: struct buf ln;
310: enum rofferr rr;
311: int i, of, rc;
312: int pos; /* byte number in the ln buffer */
313: int lnn; /* line number in the real file */
314: unsigned char c;
315:
316: memset(&ln, 0, sizeof(struct buf));
317:
1.47 schwarze 318: lnn = curp->line;
319: pos = 0;
1.1 kristaps 320:
321: for (i = 0; i < (int)blk.sz; ) {
322: if (0 == pos && '\0' == blk.buf[i])
323: break;
324:
325: if (start) {
326: curp->line = lnn;
327: curp->reparse_count = 0;
328: }
329:
330: while (i < (int)blk.sz && (start || '\0' != blk.buf[i])) {
331:
332: /*
333: * When finding an unescaped newline character,
334: * leave the character loop to process the line.
335: * Skip a preceding carriage return, if any.
336: */
337:
338: if ('\r' == blk.buf[i] && i + 1 < (int)blk.sz &&
339: '\n' == blk.buf[i + 1])
340: ++i;
341: if ('\n' == blk.buf[i]) {
342: ++i;
343: ++lnn;
344: break;
345: }
346:
1.35 schwarze 347: /*
348: * Make sure we have space for at least
349: * one backslash and one other character
350: * and the trailing NUL byte.
351: */
352:
353: if (pos + 2 >= (int)ln.sz)
354: resize_buf(&ln, 256);
355:
1.47 schwarze 356: /*
1.1 kristaps 357: * Warn about bogus characters. If you're using
358: * non-ASCII encoding, you're screwing your
359: * readers. Since I'd rather this not happen,
1.27 joerg 360: * I'll be helpful and replace these characters
361: * with "?", so we don't display gibberish.
362: * Note to manual writers: use special characters.
1.1 kristaps 363: */
364:
365: c = (unsigned char) blk.buf[i];
366:
1.47 schwarze 367: if ( ! (isascii(c) &&
368: (isgraph(c) || isblank(c)))) {
1.3 kristaps 369: mandoc_msg(MANDOCERR_BADCHAR, curp,
1.47 schwarze 370: curp->line, pos, NULL);
1.1 kristaps 371: i++;
1.27 joerg 372: ln.buf[pos++] = '?';
1.1 kristaps 373: continue;
374: }
375:
376: /* Trailing backslash = a plain char. */
377:
378: if ('\\' != blk.buf[i] || i + 1 == (int)blk.sz) {
379: ln.buf[pos++] = blk.buf[i++];
380: continue;
381: }
382:
383: /*
384: * Found escape and at least one other character.
385: * When it's a newline character, skip it.
386: * When there is a carriage return in between,
387: * skip that one as well.
388: */
389:
390: if ('\r' == blk.buf[i + 1] && i + 2 < (int)blk.sz &&
391: '\n' == blk.buf[i + 2])
392: ++i;
393: if ('\n' == blk.buf[i + 1]) {
394: i += 2;
395: ++lnn;
396: continue;
397: }
398:
1.13 kristaps 399: if ('"' == blk.buf[i + 1] || '#' == blk.buf[i + 1]) {
1.1 kristaps 400: i += 2;
401: /* Comment, skip to end of line */
402: for (; i < (int)blk.sz; ++i) {
403: if ('\n' == blk.buf[i]) {
404: ++i;
405: ++lnn;
406: break;
407: }
408: }
409:
410: /* Backout trailing whitespaces */
411: for (; pos > 0; --pos) {
412: if (ln.buf[pos - 1] != ' ')
413: break;
414: if (pos > 2 && ln.buf[pos - 2] == '\\')
415: break;
416: }
417: break;
418: }
419:
1.35 schwarze 420: /* Catch escaped bogus characters. */
421:
422: c = (unsigned char) blk.buf[i+1];
423:
1.47 schwarze 424: if ( ! (isascii(c) &&
425: (isgraph(c) || isblank(c)))) {
1.35 schwarze 426: mandoc_msg(MANDOCERR_BADCHAR, curp,
1.47 schwarze 427: curp->line, pos, NULL);
1.35 schwarze 428: i += 2;
429: ln.buf[pos++] = '?';
430: continue;
431: }
432:
1.1 kristaps 433: /* Some other escape sequence, copy & cont. */
434:
435: ln.buf[pos++] = blk.buf[i++];
436: ln.buf[pos++] = blk.buf[i++];
437: }
438:
1.47 schwarze 439: if (pos >= (int)ln.sz)
1.1 kristaps 440: resize_buf(&ln, 256);
441:
442: ln.buf[pos] = '\0';
443:
444: /*
445: * A significant amount of complexity is contained by
446: * the roff preprocessor. It's line-oriented but can be
447: * expressed on one line, so we need at times to
448: * readjust our starting point and re-run it. The roff
449: * preprocessor can also readjust the buffers with new
450: * data, so we pass them in wholesale.
451: */
452:
453: of = 0;
454:
1.24 kristaps 455: /*
456: * Maintain a lookaside buffer of all parsed lines. We
457: * only do this if mparse_keep() has been invoked (the
458: * buffer may be accessed with mparse_getkeep()).
459: */
460:
461: if (curp->secondary) {
1.47 schwarze 462: curp->secondary->buf = mandoc_realloc(
463: curp->secondary->buf,
464: curp->secondary->sz + pos + 2);
465: memcpy(curp->secondary->buf +
466: curp->secondary->sz,
467: ln.buf, pos);
1.24 kristaps 468: curp->secondary->sz += pos;
469: curp->secondary->buf
470: [curp->secondary->sz] = '\n';
471: curp->secondary->sz++;
472: curp->secondary->buf
473: [curp->secondary->sz] = '\0';
474: }
1.1 kristaps 475: rerun:
1.47 schwarze 476: rr = roff_parseln(curp->roff, curp->line,
477: &ln.buf, &ln.sz, of, &of);
1.1 kristaps 478:
479: switch (rr) {
1.47 schwarze 480: case ROFF_REPARSE:
1.1 kristaps 481: if (REPARSE_LIMIT >= ++curp->reparse_count)
482: mparse_buf_r(curp, ln, 0);
483: else
1.3 kristaps 484: mandoc_msg(MANDOCERR_ROFFLOOP, curp,
1.47 schwarze 485: curp->line, pos, NULL);
1.1 kristaps 486: pos = 0;
487: continue;
1.47 schwarze 488: case ROFF_APPEND:
1.1 kristaps 489: pos = (int)strlen(ln.buf);
490: continue;
1.47 schwarze 491: case ROFF_RERUN:
1.1 kristaps 492: goto rerun;
1.47 schwarze 493: case ROFF_IGN:
1.1 kristaps 494: pos = 0;
495: continue;
1.47 schwarze 496: case ROFF_ERR:
1.1 kristaps 497: assert(MANDOCLEVEL_FATAL <= curp->file_status);
498: break;
1.47 schwarze 499: case ROFF_SO:
1.45 schwarze 500: if (0 == (MPARSE_SO & curp->options) &&
501: (i >= (int)blk.sz || '\0' == blk.buf[i])) {
502: curp->sodest = mandoc_strdup(ln.buf + of);
503: free(ln.buf);
504: return;
505: }
1.24 kristaps 506: /*
507: * We remove `so' clauses from our lookaside
508: * buffer because we're going to descend into
509: * the file recursively.
510: */
1.47 schwarze 511: if (curp->secondary)
1.25 kristaps 512: curp->secondary->sz -= pos + 1;
1.36 schwarze 513: mparse_readfd(curp, -1, ln.buf + of);
1.52 schwarze 514: if (MANDOCLEVEL_FATAL <= curp->file_status) {
515: mandoc_vmsg(MANDOCERR_SO_FAIL,
516: curp, curp->line, pos,
517: ".so %s", ln.buf + of);
1.1 kristaps 518: break;
1.52 schwarze 519: }
1.1 kristaps 520: pos = 0;
521: continue;
522: default:
523: break;
524: }
525:
526: /*
527: * If we encounter errors in the recursive parse, make
528: * sure we don't continue parsing.
529: */
530:
531: if (MANDOCLEVEL_FATAL <= curp->file_status)
532: break;
533:
534: /*
535: * If input parsers have not been allocated, do so now.
1.14 kristaps 536: * We keep these instanced between parsers, but set them
1.1 kristaps 537: * locally per parse routine since we can use different
538: * parsers with each one.
539: */
540:
541: if ( ! (curp->man || curp->mdoc))
542: pset(ln.buf + of, pos - of, curp);
543:
1.47 schwarze 544: /*
1.1 kristaps 545: * Lastly, push down into the parsers themselves. One
546: * of these will have already been set in the pset()
547: * routine.
548: * If libroff returns ROFF_TBL, then add it to the
549: * currently open parse. Since we only get here if
550: * there does exist data (see tbl_data.c), we're
551: * guaranteed that something's been allocated.
552: * Do the same for ROFF_EQN.
553: */
554:
555: rc = -1;
556:
557: if (ROFF_TBL == rr)
558: while (NULL != (span = roff_span(curp->roff))) {
559: rc = curp->man ?
1.47 schwarze 560: man_addspan(curp->man, span) :
561: mdoc_addspan(curp->mdoc, span);
1.1 kristaps 562: if (0 == rc)
563: break;
564: }
565: else if (ROFF_EQN == rr)
1.47 schwarze 566: rc = curp->mdoc ?
567: mdoc_addeqn(curp->mdoc,
568: roff_eqn(curp->roff)) :
569: man_addeqn(curp->man,
570: roff_eqn(curp->roff));
1.1 kristaps 571: else if (curp->man || curp->mdoc)
572: rc = curp->man ?
1.47 schwarze 573: man_parseln(curp->man,
574: curp->line, ln.buf, of) :
575: mdoc_parseln(curp->mdoc,
576: curp->line, ln.buf, of);
1.1 kristaps 577:
578: if (0 == rc) {
579: assert(MANDOCLEVEL_FATAL <= curp->file_status);
580: break;
1.41 schwarze 581: } else if (2 == rc)
582: break;
1.1 kristaps 583:
584: /* Temporary buffers typically are not full. */
585:
586: if (0 == start && '\0' == blk.buf[i])
587: break;
588:
589: /* Start the next input line. */
590:
591: pos = 0;
592: }
593:
594: free(ln.buf);
595: }
596:
597: static int
1.40 schwarze 598: read_whole_file(struct mparse *curp, const char *file, int fd,
599: struct buf *fb, int *with_mmap)
1.1 kristaps 600: {
601: size_t off;
602: ssize_t ssz;
603:
1.15 kristaps 604: #ifdef HAVE_MMAP
605: struct stat st;
1.1 kristaps 606: if (-1 == fstat(fd, &st)) {
1.40 schwarze 607: curp->file_status = MANDOCLEVEL_SYSERR;
608: if (curp->mmsg)
609: (*curp->mmsg)(MANDOCERR_SYSSTAT, curp->file_status,
610: file, 0, 0, strerror(errno));
1.1 kristaps 611: return(0);
612: }
613:
614: /*
615: * If we're a regular file, try just reading in the whole entry
616: * via mmap(). This is faster than reading it into blocks, and
617: * since each file is only a few bytes to begin with, I'm not
618: * concerned that this is going to tank any machines.
619: */
620:
621: if (S_ISREG(st.st_mode)) {
622: if (st.st_size >= (1U << 31)) {
1.40 schwarze 623: curp->file_status = MANDOCLEVEL_FATAL;
624: if (curp->mmsg)
625: (*curp->mmsg)(MANDOCERR_TOOLARGE,
626: curp->file_status, file, 0, 0, NULL);
1.1 kristaps 627: return(0);
628: }
629: *with_mmap = 1;
630: fb->sz = (size_t)st.st_size;
1.37 schwarze 631: fb->buf = mmap(NULL, fb->sz, PROT_READ, MAP_SHARED, fd, 0);
1.1 kristaps 632: if (fb->buf != MAP_FAILED)
633: return(1);
634: }
1.15 kristaps 635: #endif
1.1 kristaps 636:
637: /*
638: * If this isn't a regular file (like, say, stdin), then we must
639: * go the old way and just read things in bit by bit.
640: */
641:
642: *with_mmap = 0;
643: off = 0;
644: fb->sz = 0;
645: fb->buf = NULL;
646: for (;;) {
647: if (off == fb->sz) {
648: if (fb->sz == (1U << 31)) {
1.40 schwarze 649: curp->file_status = MANDOCLEVEL_FATAL;
650: if (curp->mmsg)
651: (*curp->mmsg)(MANDOCERR_TOOLARGE,
652: curp->file_status,
653: file, 0, 0, NULL);
1.1 kristaps 654: break;
655: }
656: resize_buf(fb, 65536);
657: }
658: ssz = read(fd, fb->buf + (int)off, fb->sz - off);
659: if (ssz == 0) {
660: fb->sz = off;
661: return(1);
662: }
663: if (ssz == -1) {
1.40 schwarze 664: curp->file_status = MANDOCLEVEL_SYSERR;
665: if (curp->mmsg)
666: (*curp->mmsg)(MANDOCERR_SYSREAD,
667: curp->file_status, file, 0, 0,
668: strerror(errno));
1.1 kristaps 669: break;
670: }
671: off += (size_t)ssz;
672: }
673:
674: free(fb->buf);
675: fb->buf = NULL;
676: return(0);
677: }
678:
679: static void
680: mparse_end(struct mparse *curp)
681: {
682:
683: if (MANDOCLEVEL_FATAL <= curp->file_status)
684: return;
685:
686: if (curp->mdoc && ! mdoc_endparse(curp->mdoc)) {
687: assert(MANDOCLEVEL_FATAL <= curp->file_status);
688: return;
689: }
690:
691: if (curp->man && ! man_endparse(curp->man)) {
692: assert(MANDOCLEVEL_FATAL <= curp->file_status);
693: return;
694: }
695:
1.45 schwarze 696: if ( ! (curp->mdoc || curp->man || curp->sodest)) {
1.50 schwarze 697: mandoc_msg(MANDOCERR_NOTMANUAL, curp, 0, 0, NULL);
1.1 kristaps 698: curp->file_status = MANDOCLEVEL_FATAL;
1.6 kristaps 699: return;
1.1 kristaps 700: }
701:
702: roff_endparse(curp->roff);
703: }
704:
705: static void
1.36 schwarze 706: mparse_parse_buffer(struct mparse *curp, struct buf blk, const char *file)
1.28 joerg 707: {
708: const char *svfile;
1.36 schwarze 709: static int recursion_depth;
710:
711: if (64 < recursion_depth) {
712: mandoc_msg(MANDOCERR_ROFFLOOP, curp, curp->line, 0, NULL);
713: return;
714: }
1.28 joerg 715:
716: /* Line number is per-file. */
717: svfile = curp->file;
718: curp->file = file;
719: curp->line = 1;
1.36 schwarze 720: recursion_depth++;
1.28 joerg 721:
722: mparse_buf_r(curp, blk, 1);
723:
1.36 schwarze 724: if (0 == --recursion_depth && MANDOCLEVEL_FATAL > curp->file_status)
1.28 joerg 725: mparse_end(curp);
726:
727: curp->file = svfile;
728: }
729:
730: enum mandoclevel
731: mparse_readmem(struct mparse *curp, const void *buf, size_t len,
732: const char *file)
733: {
734: struct buf blk;
735:
736: blk.buf = UNCONST(buf);
737: blk.sz = len;
738:
1.36 schwarze 739: mparse_parse_buffer(curp, blk, file);
1.28 joerg 740: return(curp->file_status);
741: }
742:
1.36 schwarze 743: enum mandoclevel
744: mparse_readfd(struct mparse *curp, int fd, const char *file)
1.1 kristaps 745: {
1.28 joerg 746: struct buf blk;
747: int with_mmap;
1.1 kristaps 748:
1.40 schwarze 749: if (-1 == fd && -1 == (fd = open(file, O_RDONLY, 0))) {
750: curp->file_status = MANDOCLEVEL_SYSERR;
751: if (curp->mmsg)
752: (*curp->mmsg)(MANDOCERR_SYSOPEN,
753: curp->file_status,
754: file, 0, 0, strerror(errno));
755: goto out;
756: }
757:
1.28 joerg 758: /*
759: * Run for each opened file; may be called more than once for
760: * each full parse sequence if the opened file is nested (i.e.,
761: * from `so'). Simply sucks in the whole file and moves into
762: * the parse phase for the file.
763: */
1.1 kristaps 764:
1.40 schwarze 765: if ( ! read_whole_file(curp, file, fd, &blk, &with_mmap))
1.36 schwarze 766: goto out;
1.1 kristaps 767:
1.36 schwarze 768: mparse_parse_buffer(curp, blk, file);
1.1 kristaps 769:
1.28 joerg 770: #ifdef HAVE_MMAP
771: if (with_mmap)
772: munmap(blk.buf, blk.sz);
773: else
774: #endif
775: free(blk.buf);
1.1 kristaps 776:
777: if (STDIN_FILENO != fd && -1 == close(fd))
778: perror(file);
1.36 schwarze 779: out:
1.1 kristaps 780: return(curp->file_status);
781: }
782:
783: struct mparse *
1.44 schwarze 784: mparse_alloc(int options, enum mandoclevel wlevel,
1.69 schwarze 785: mandocmsg mmsg, const char *defos)
1.1 kristaps 786: {
787: struct mparse *curp;
1.10 kristaps 788:
789: assert(wlevel <= MANDOCLEVEL_FATAL);
1.1 kristaps 790:
791: curp = mandoc_calloc(1, sizeof(struct mparse));
792:
1.44 schwarze 793: curp->options = options;
1.3 kristaps 794: curp->wlevel = wlevel;
1.1 kristaps 795: curp->mmsg = mmsg;
1.29 schwarze 796: curp->defos = defos;
1.1 kristaps 797:
1.44 schwarze 798: curp->roff = roff_alloc(curp, options);
1.1 kristaps 799: return(curp);
800: }
801:
802: void
803: mparse_reset(struct mparse *curp)
804: {
805:
806: roff_reset(curp->roff);
807:
808: if (curp->mdoc)
809: mdoc_reset(curp->mdoc);
810: if (curp->man)
811: man_reset(curp->man);
1.24 kristaps 812: if (curp->secondary)
813: curp->secondary->sz = 0;
1.1 kristaps 814:
815: curp->file_status = MANDOCLEVEL_OK;
816: curp->mdoc = NULL;
817: curp->man = NULL;
1.45 schwarze 818:
819: free(curp->sodest);
820: curp->sodest = NULL;
1.1 kristaps 821: }
822:
823: void
824: mparse_free(struct mparse *curp)
825: {
826:
827: if (curp->pmdoc)
828: mdoc_free(curp->pmdoc);
829: if (curp->pman)
830: man_free(curp->pman);
831: if (curp->roff)
832: roff_free(curp->roff);
1.24 kristaps 833: if (curp->secondary)
834: free(curp->secondary->buf);
1.1 kristaps 835:
1.24 kristaps 836: free(curp->secondary);
1.45 schwarze 837: free(curp->sodest);
1.1 kristaps 838: free(curp);
839: }
840:
841: void
1.45 schwarze 842: mparse_result(struct mparse *curp,
843: struct mdoc **mdoc, struct man **man, char **sodest)
1.1 kristaps 844: {
845:
1.45 schwarze 846: if (sodest && NULL != (*sodest = curp->sodest)) {
847: *mdoc = NULL;
848: *man = NULL;
849: return;
850: }
1.9 kristaps 851: if (mdoc)
852: *mdoc = curp->mdoc;
853: if (man)
854: *man = curp->man;
1.3 kristaps 855: }
856:
857: void
858: mandoc_vmsg(enum mandocerr t, struct mparse *m,
859: int ln, int pos, const char *fmt, ...)
860: {
861: char buf[256];
862: va_list ap;
863:
864: va_start(ap, fmt);
1.48 schwarze 865: (void)vsnprintf(buf, sizeof(buf), fmt, ap);
1.3 kristaps 866: va_end(ap);
867:
868: mandoc_msg(t, m, ln, pos, buf);
869: }
870:
871: void
1.47 schwarze 872: mandoc_msg(enum mandocerr er, struct mparse *m,
1.3 kristaps 873: int ln, int col, const char *msg)
874: {
875: enum mandoclevel level;
876:
877: level = MANDOCLEVEL_FATAL;
878: while (er < mandoclimits[level])
879: level--;
880:
881: if (level < m->wlevel)
882: return;
883:
1.8 kristaps 884: if (m->mmsg)
885: (*m->mmsg)(er, level, m->file, ln, col, msg);
1.3 kristaps 886:
887: if (m->file_status < level)
888: m->file_status = level;
1.7 kristaps 889: }
890:
891: const char *
892: mparse_strerror(enum mandocerr er)
893: {
894:
895: return(mandocerrs[er]);
896: }
897:
898: const char *
899: mparse_strlevel(enum mandoclevel lvl)
900: {
901: return(mandoclevels[lvl]);
1.24 kristaps 902: }
903:
904: void
905: mparse_keep(struct mparse *p)
906: {
907:
908: assert(NULL == p->secondary);
909: p->secondary = mandoc_calloc(1, sizeof(struct buf));
910: }
911:
912: const char *
913: mparse_getkeep(const struct mparse *p)
914: {
915:
916: assert(p->secondary);
917: return(p->secondary->sz ? p->secondary->buf : NULL);
1.1 kristaps 918: }
CVSweb