Annotation of mandoc/read.c, Revision 1.82
1.82 ! schwarze 1: /* $Id: read.c,v 1.81 2014/08/16 19:00:01 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: #include "config.h"
20:
1.80 schwarze 21: #include <sys/types.h>
1.81 schwarze 22: #if HAVE_MMAP
1.82 ! schwarze 23: #include <sys/mman.h>
1.80 schwarze 24: #include <sys/stat.h>
1.15 kristaps 25: #endif
1.82 ! schwarze 26: #include <sys/wait.h>
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 */
1.79 schwarze 96: "missing manual title, using UNTITLED",
97: "missing manual title, using \"\"",
1.54 schwarze 98: "lower case character in document title",
1.79 schwarze 99: "missing manual section, using \"\"",
1.7 kristaps 100: "unknown manual section",
1.32 schwarze 101: "unknown manual volume or arch",
1.54 schwarze 102: "missing date, using today's date",
1.7 kristaps 103: "cannot parse date, using it verbatim",
1.79 schwarze 104: "missing Os macro, using \"\"",
105: "duplicate prologue macro",
106: "late prologue macro",
107: "skipping late title macro",
1.7 kristaps 108: "prologue macros out of order",
109:
110: /* related to document structure */
111: ".so is fragile, better use ln(1)",
1.50 schwarze 112: "no document body",
1.54 schwarze 113: "content before first section header",
114: "first section is not \"NAME\"",
1.7 kristaps 115: "bad NAME section contents",
116: "sections out of conventional order",
1.54 schwarze 117: "duplicate section title",
118: "unexpected section",
1.7 kristaps 119:
120: /* related to macros and nesting */
1.55 schwarze 121: "obsolete macro",
1.7 kristaps 122: "skipping paragraph macro",
1.31 schwarze 123: "moving paragraph macro out of list",
1.7 kristaps 124: "skipping no-space macro",
125: "blocks badly nested",
126: "nested displays are not portable",
1.57 schwarze 127: "moving content out of list",
128: ".Vt block has child macro",
1.78 schwarze 129: "fill mode already enabled, skipping",
130: "fill mode already disabled, skipping",
1.7 kristaps 131: "line scope broken",
132:
133: /* related to missing macro arguments */
1.58 schwarze 134: "skipping empty request",
135: "conditional request controls empty scope",
1.7 kristaps 136: "skipping empty macro",
1.62 schwarze 137: "empty argument, using 0n",
1.7 kristaps 138: "argument count wrong",
1.60 schwarze 139: "missing display type, using -ragged",
140: "list type is not the first argument",
141: "missing -width in -tag list, using 8n",
1.78 schwarze 142: "missing utility name, using \"\"",
1.60 schwarze 143: "empty head in list item",
144: "empty list item",
1.61 schwarze 145: "missing font type, using \\fR",
146: "unknown font type, using \\fR",
1.60 schwarze 147: "missing -std argument, adding it",
1.7 kristaps 148:
149: /* related to bad macro arguments */
1.64 schwarze 150: "unterminated quoted argument",
1.7 kristaps 151: "duplicate argument",
1.76 schwarze 152: "skipping duplicate argument",
1.63 schwarze 153: "skipping duplicate display type",
154: "skipping duplicate list type",
1.76 schwarze 155: "skipping -width argument",
1.7 kristaps 156: "unknown AT&T UNIX version",
1.67 schwarze 157: "invalid content in Rs block",
1.63 schwarze 158: "invalid Boolean argument",
159: "unknown font, skipping request",
1.7 kristaps 160:
161: /* related to plain text */
1.64 schwarze 162: "blank line in fill mode, using .sp",
163: "tab in filled text",
164: "whitespace at end of input line",
1.7 kristaps 165: "bad comment style",
1.64 schwarze 166: "invalid escape sequence",
167: "undefined string, using \"\"",
1.16 kristaps 168:
1.7 kristaps 169: "generic error",
1.17 kristaps 170:
171: /* related to equations */
1.20 kristaps 172: "unexpected equation scope closure",
173: "equation scope open on exit",
1.21 kristaps 174: "overlapping equation scopes",
175: "unexpected end of equation",
176: "equation syntax error",
1.7 kristaps 177:
178: /* related to tables */
179: "bad table syntax",
180: "bad table option",
181: "bad table layout",
182: "no table layout cells specified",
183: "no table data cells specified",
184: "ignore data in cell",
185: "data block still open",
186: "ignoring extra data cells",
187:
1.68 schwarze 188: /* related to document structure and macros */
1.7 kristaps 189: "input stack limit exceeded, infinite loop?",
190: "skipping bad character",
1.68 schwarze 191: "skipping unknown macro",
1.73 schwarze 192: "skipping item outside list",
1.68 schwarze 193: "skipping column outside column list",
194: "skipping end of block that is not open",
195: "inserting missing end of block",
196: "appending missing end of block",
197:
198: /* related to request and macro arguments */
1.7 kristaps 199: "escaped character not allowed in a name",
200: "argument count wrong",
1.71 schwarze 201: "missing list type, using -item",
1.70 schwarze 202: "missing manual name, using \"\"",
1.71 schwarze 203: "uname(3) system call failed, using UNKNOWN",
1.63 schwarze 204: "unknown standard specifier",
1.71 schwarze 205: "skipping request without numeric argument",
1.61 schwarze 206: "skipping all arguments",
207: "skipping excess arguments",
1.7 kristaps 208:
209: "generic fatal error",
210:
1.40 schwarze 211: "input too large",
1.78 schwarze 212: "NOT IMPLEMENTED: Bd -file",
1.7 kristaps 213: "NOT IMPLEMENTED: .so with absolute path or \"..\"",
1.52 schwarze 214: ".so request failed",
1.40 schwarze 215:
216: /* system errors */
1.82 ! schwarze 217: "cannot dup file descriptor",
! 218: "cannot exec",
! 219: "gunzip failed with code",
! 220: "cannot fork",
1.51 schwarze 221: NULL,
1.82 ! schwarze 222: "cannot open pipe",
! 223: "cannot read file",
! 224: "gunzip died from signal",
1.40 schwarze 225: "cannot stat file",
1.82 ! schwarze 226: "wait failed",
1.7 kristaps 227: };
228:
229: static const char * const mandoclevels[MANDOCLEVEL_MAX] = {
230: "SUCCESS",
231: "RESERVED",
232: "WARNING",
233: "ERROR",
234: "FATAL",
235: "BADARG",
236: "SYSERR"
237: };
238:
1.47 schwarze 239:
1.1 kristaps 240: static void
241: resize_buf(struct buf *buf, size_t initial)
242: {
243:
244: buf->sz = buf->sz > initial/2 ? 2 * buf->sz : initial;
245: buf->buf = mandoc_realloc(buf->buf, buf->sz);
246: }
247:
248: static void
249: pset(const char *buf, int pos, struct mparse *curp)
250: {
251: int i;
252:
253: /*
254: * Try to intuit which kind of manual parser should be used. If
255: * passed in by command-line (-man, -mdoc), then use that
256: * explicitly. If passed as -mandoc, then try to guess from the
257: * line: either skip dot-lines, use -mdoc when finding `.Dt', or
258: * default to -man, which is more lenient.
259: *
260: * Separate out pmdoc/pman from mdoc/man: the first persists
261: * through all parsers, while the latter is used per-parse.
262: */
263:
264: if ('.' == buf[0] || '\'' == buf[0]) {
265: for (i = 1; buf[i]; i++)
266: if (' ' != buf[i] && '\t' != buf[i])
267: break;
268: if ('\0' == buf[i])
269: return;
270: }
271:
1.44 schwarze 272: if (MPARSE_MDOC & curp->options) {
1.1 kristaps 273: curp->mdoc = curp->pmdoc;
274: return;
1.44 schwarze 275: } else if (MPARSE_MAN & curp->options) {
1.1 kristaps 276: curp->man = curp->pman;
277: return;
278: }
279:
280: if (pos >= 3 && 0 == memcmp(buf, ".Dd", 3)) {
1.47 schwarze 281: if (NULL == curp->pmdoc)
1.44 schwarze 282: curp->pmdoc = mdoc_alloc(
283: curp->roff, curp, curp->defos,
284: MPARSE_QUICK & curp->options ? 1 : 0);
1.1 kristaps 285: assert(curp->pmdoc);
286: curp->mdoc = curp->pmdoc;
287: return;
1.47 schwarze 288: }
1.1 kristaps 289:
1.47 schwarze 290: if (NULL == curp->pman)
1.44 schwarze 291: curp->pman = man_alloc(curp->roff, curp,
292: MPARSE_QUICK & curp->options ? 1 : 0);
1.1 kristaps 293: assert(curp->pman);
294: curp->man = curp->pman;
295: }
296:
297: /*
298: * Main parse routine for an opened file. This is called for each
299: * opened file and simply loops around the full input file, possibly
300: * nesting (i.e., with `so').
301: */
302: static void
303: mparse_buf_r(struct mparse *curp, struct buf blk, int start)
304: {
305: const struct tbl_span *span;
306: struct buf ln;
307: enum rofferr rr;
308: int i, of, rc;
309: int pos; /* byte number in the ln buffer */
310: int lnn; /* line number in the real file */
311: unsigned char c;
312:
313: memset(&ln, 0, sizeof(struct buf));
314:
1.47 schwarze 315: lnn = curp->line;
316: pos = 0;
1.1 kristaps 317:
318: for (i = 0; i < (int)blk.sz; ) {
319: if (0 == pos && '\0' == blk.buf[i])
320: break;
321:
322: if (start) {
323: curp->line = lnn;
324: curp->reparse_count = 0;
325: }
326:
327: while (i < (int)blk.sz && (start || '\0' != blk.buf[i])) {
328:
329: /*
330: * When finding an unescaped newline character,
331: * leave the character loop to process the line.
332: * Skip a preceding carriage return, if any.
333: */
334:
335: if ('\r' == blk.buf[i] && i + 1 < (int)blk.sz &&
336: '\n' == blk.buf[i + 1])
337: ++i;
338: if ('\n' == blk.buf[i]) {
339: ++i;
340: ++lnn;
341: break;
342: }
343:
1.35 schwarze 344: /*
345: * Make sure we have space for at least
346: * one backslash and one other character
347: * and the trailing NUL byte.
348: */
349:
350: if (pos + 2 >= (int)ln.sz)
351: resize_buf(&ln, 256);
352:
1.47 schwarze 353: /*
1.1 kristaps 354: * Warn about bogus characters. If you're using
355: * non-ASCII encoding, you're screwing your
356: * readers. Since I'd rather this not happen,
1.27 joerg 357: * I'll be helpful and replace these characters
358: * with "?", so we don't display gibberish.
359: * Note to manual writers: use special characters.
1.1 kristaps 360: */
361:
362: c = (unsigned char) blk.buf[i];
363:
1.47 schwarze 364: if ( ! (isascii(c) &&
365: (isgraph(c) || isblank(c)))) {
1.78 schwarze 366: mandoc_vmsg(MANDOCERR_BADCHAR, curp,
367: curp->line, pos, "0x%x", c);
1.1 kristaps 368: i++;
1.27 joerg 369: ln.buf[pos++] = '?';
1.1 kristaps 370: continue;
371: }
372:
373: /* Trailing backslash = a plain char. */
374:
375: if ('\\' != blk.buf[i] || i + 1 == (int)blk.sz) {
376: ln.buf[pos++] = blk.buf[i++];
377: continue;
378: }
379:
380: /*
381: * Found escape and at least one other character.
382: * When it's a newline character, skip it.
383: * When there is a carriage return in between,
384: * skip that one as well.
385: */
386:
387: if ('\r' == blk.buf[i + 1] && i + 2 < (int)blk.sz &&
388: '\n' == blk.buf[i + 2])
389: ++i;
390: if ('\n' == blk.buf[i + 1]) {
391: i += 2;
392: ++lnn;
393: continue;
394: }
395:
1.13 kristaps 396: if ('"' == blk.buf[i + 1] || '#' == blk.buf[i + 1]) {
1.1 kristaps 397: i += 2;
398: /* Comment, skip to end of line */
399: for (; i < (int)blk.sz; ++i) {
400: if ('\n' == blk.buf[i]) {
401: ++i;
402: ++lnn;
403: break;
404: }
405: }
406:
407: /* Backout trailing whitespaces */
408: for (; pos > 0; --pos) {
409: if (ln.buf[pos - 1] != ' ')
410: break;
411: if (pos > 2 && ln.buf[pos - 2] == '\\')
412: break;
413: }
414: break;
415: }
416:
1.35 schwarze 417: /* Catch escaped bogus characters. */
418:
419: c = (unsigned char) blk.buf[i+1];
420:
1.47 schwarze 421: if ( ! (isascii(c) &&
422: (isgraph(c) || isblank(c)))) {
1.78 schwarze 423: mandoc_vmsg(MANDOCERR_BADCHAR, curp,
424: curp->line, pos, "0x%x", c);
1.35 schwarze 425: i += 2;
426: ln.buf[pos++] = '?';
427: continue;
428: }
429:
1.1 kristaps 430: /* Some other escape sequence, copy & cont. */
431:
432: ln.buf[pos++] = blk.buf[i++];
433: ln.buf[pos++] = blk.buf[i++];
434: }
435:
1.47 schwarze 436: if (pos >= (int)ln.sz)
1.1 kristaps 437: resize_buf(&ln, 256);
438:
439: ln.buf[pos] = '\0';
440:
441: /*
442: * A significant amount of complexity is contained by
443: * the roff preprocessor. It's line-oriented but can be
444: * expressed on one line, so we need at times to
445: * readjust our starting point and re-run it. The roff
446: * preprocessor can also readjust the buffers with new
447: * data, so we pass them in wholesale.
448: */
449:
450: of = 0;
451:
1.24 kristaps 452: /*
453: * Maintain a lookaside buffer of all parsed lines. We
454: * only do this if mparse_keep() has been invoked (the
455: * buffer may be accessed with mparse_getkeep()).
456: */
457:
458: if (curp->secondary) {
1.47 schwarze 459: curp->secondary->buf = mandoc_realloc(
460: curp->secondary->buf,
461: curp->secondary->sz + pos + 2);
462: memcpy(curp->secondary->buf +
463: curp->secondary->sz,
464: ln.buf, pos);
1.24 kristaps 465: curp->secondary->sz += pos;
466: curp->secondary->buf
467: [curp->secondary->sz] = '\n';
468: curp->secondary->sz++;
469: curp->secondary->buf
470: [curp->secondary->sz] = '\0';
471: }
1.1 kristaps 472: rerun:
1.47 schwarze 473: rr = roff_parseln(curp->roff, curp->line,
474: &ln.buf, &ln.sz, of, &of);
1.1 kristaps 475:
476: switch (rr) {
1.47 schwarze 477: case ROFF_REPARSE:
1.1 kristaps 478: if (REPARSE_LIMIT >= ++curp->reparse_count)
479: mparse_buf_r(curp, ln, 0);
480: else
1.3 kristaps 481: mandoc_msg(MANDOCERR_ROFFLOOP, curp,
1.47 schwarze 482: curp->line, pos, NULL);
1.1 kristaps 483: pos = 0;
484: continue;
1.47 schwarze 485: case ROFF_APPEND:
1.1 kristaps 486: pos = (int)strlen(ln.buf);
487: continue;
1.47 schwarze 488: case ROFF_RERUN:
1.1 kristaps 489: goto rerun;
1.47 schwarze 490: case ROFF_IGN:
1.1 kristaps 491: pos = 0;
492: continue;
1.47 schwarze 493: case ROFF_ERR:
1.1 kristaps 494: assert(MANDOCLEVEL_FATAL <= curp->file_status);
495: break;
1.47 schwarze 496: case ROFF_SO:
1.45 schwarze 497: if (0 == (MPARSE_SO & curp->options) &&
498: (i >= (int)blk.sz || '\0' == blk.buf[i])) {
499: curp->sodest = mandoc_strdup(ln.buf + of);
500: free(ln.buf);
501: return;
502: }
1.24 kristaps 503: /*
504: * We remove `so' clauses from our lookaside
505: * buffer because we're going to descend into
506: * the file recursively.
507: */
1.47 schwarze 508: if (curp->secondary)
1.25 kristaps 509: curp->secondary->sz -= pos + 1;
1.36 schwarze 510: mparse_readfd(curp, -1, ln.buf + of);
1.52 schwarze 511: if (MANDOCLEVEL_FATAL <= curp->file_status) {
512: mandoc_vmsg(MANDOCERR_SO_FAIL,
513: curp, curp->line, pos,
514: ".so %s", ln.buf + of);
1.1 kristaps 515: break;
1.52 schwarze 516: }
1.1 kristaps 517: pos = 0;
518: continue;
519: default:
520: break;
521: }
522:
523: /*
524: * If we encounter errors in the recursive parse, make
525: * sure we don't continue parsing.
526: */
527:
528: if (MANDOCLEVEL_FATAL <= curp->file_status)
529: break;
530:
531: /*
532: * If input parsers have not been allocated, do so now.
1.14 kristaps 533: * We keep these instanced between parsers, but set them
1.1 kristaps 534: * locally per parse routine since we can use different
535: * parsers with each one.
536: */
537:
538: if ( ! (curp->man || curp->mdoc))
539: pset(ln.buf + of, pos - of, curp);
540:
1.47 schwarze 541: /*
1.1 kristaps 542: * Lastly, push down into the parsers themselves. One
543: * of these will have already been set in the pset()
544: * routine.
545: * If libroff returns ROFF_TBL, then add it to the
546: * currently open parse. Since we only get here if
547: * there does exist data (see tbl_data.c), we're
548: * guaranteed that something's been allocated.
549: * Do the same for ROFF_EQN.
550: */
551:
552: rc = -1;
553:
554: if (ROFF_TBL == rr)
555: while (NULL != (span = roff_span(curp->roff))) {
556: rc = curp->man ?
1.47 schwarze 557: man_addspan(curp->man, span) :
558: mdoc_addspan(curp->mdoc, span);
1.1 kristaps 559: if (0 == rc)
560: break;
561: }
562: else if (ROFF_EQN == rr)
1.47 schwarze 563: rc = curp->mdoc ?
564: mdoc_addeqn(curp->mdoc,
565: roff_eqn(curp->roff)) :
566: man_addeqn(curp->man,
567: roff_eqn(curp->roff));
1.1 kristaps 568: else if (curp->man || curp->mdoc)
569: rc = curp->man ?
1.47 schwarze 570: man_parseln(curp->man,
571: curp->line, ln.buf, of) :
572: mdoc_parseln(curp->mdoc,
573: curp->line, ln.buf, of);
1.1 kristaps 574:
575: if (0 == rc) {
576: assert(MANDOCLEVEL_FATAL <= curp->file_status);
577: break;
1.41 schwarze 578: } else if (2 == rc)
579: break;
1.1 kristaps 580:
581: /* Temporary buffers typically are not full. */
582:
583: if (0 == start && '\0' == blk.buf[i])
584: break;
585:
586: /* Start the next input line. */
587:
588: pos = 0;
589: }
590:
591: free(ln.buf);
592: }
593:
594: static int
1.40 schwarze 595: read_whole_file(struct mparse *curp, const char *file, int fd,
596: struct buf *fb, int *with_mmap)
1.1 kristaps 597: {
598: size_t off;
599: ssize_t ssz;
600:
1.81 schwarze 601: #if HAVE_MMAP
1.15 kristaps 602: struct stat st;
1.1 kristaps 603: if (-1 == fstat(fd, &st)) {
1.40 schwarze 604: curp->file_status = MANDOCLEVEL_SYSERR;
605: if (curp->mmsg)
606: (*curp->mmsg)(MANDOCERR_SYSSTAT, curp->file_status,
607: file, 0, 0, strerror(errno));
1.1 kristaps 608: return(0);
609: }
610:
611: /*
612: * If we're a regular file, try just reading in the whole entry
613: * via mmap(). This is faster than reading it into blocks, and
614: * since each file is only a few bytes to begin with, I'm not
615: * concerned that this is going to tank any machines.
616: */
617:
618: if (S_ISREG(st.st_mode)) {
619: if (st.st_size >= (1U << 31)) {
1.40 schwarze 620: curp->file_status = MANDOCLEVEL_FATAL;
621: if (curp->mmsg)
622: (*curp->mmsg)(MANDOCERR_TOOLARGE,
623: curp->file_status, file, 0, 0, NULL);
1.1 kristaps 624: return(0);
625: }
626: *with_mmap = 1;
627: fb->sz = (size_t)st.st_size;
1.37 schwarze 628: fb->buf = mmap(NULL, fb->sz, PROT_READ, MAP_SHARED, fd, 0);
1.1 kristaps 629: if (fb->buf != MAP_FAILED)
630: return(1);
631: }
1.15 kristaps 632: #endif
1.1 kristaps 633:
634: /*
635: * If this isn't a regular file (like, say, stdin), then we must
636: * go the old way and just read things in bit by bit.
637: */
638:
639: *with_mmap = 0;
640: off = 0;
641: fb->sz = 0;
642: fb->buf = NULL;
643: for (;;) {
644: if (off == fb->sz) {
645: if (fb->sz == (1U << 31)) {
1.40 schwarze 646: curp->file_status = MANDOCLEVEL_FATAL;
647: if (curp->mmsg)
648: (*curp->mmsg)(MANDOCERR_TOOLARGE,
649: curp->file_status,
650: file, 0, 0, NULL);
1.1 kristaps 651: break;
652: }
653: resize_buf(fb, 65536);
654: }
655: ssz = read(fd, fb->buf + (int)off, fb->sz - off);
656: if (ssz == 0) {
657: fb->sz = off;
658: return(1);
659: }
660: if (ssz == -1) {
1.40 schwarze 661: curp->file_status = MANDOCLEVEL_SYSERR;
662: if (curp->mmsg)
663: (*curp->mmsg)(MANDOCERR_SYSREAD,
664: curp->file_status, file, 0, 0,
665: strerror(errno));
1.1 kristaps 666: break;
667: }
668: off += (size_t)ssz;
669: }
670:
671: free(fb->buf);
672: fb->buf = NULL;
673: return(0);
674: }
675:
676: static void
677: mparse_end(struct mparse *curp)
678: {
679:
680: if (MANDOCLEVEL_FATAL <= curp->file_status)
681: return;
682:
1.72 schwarze 683: if (curp->mdoc == NULL &&
684: curp->man == NULL &&
685: curp->sodest == NULL) {
686: if (curp->options & MPARSE_MDOC)
687: curp->mdoc = curp->pmdoc;
688: else {
689: if (curp->pman == NULL)
690: curp->pman = man_alloc(curp->roff, curp,
691: curp->options & MPARSE_QUICK ? 1 : 0);
692: curp->man = curp->pman;
693: }
694: }
695:
1.1 kristaps 696: if (curp->mdoc && ! mdoc_endparse(curp->mdoc)) {
697: assert(MANDOCLEVEL_FATAL <= curp->file_status);
698: return;
699: }
700:
701: if (curp->man && ! man_endparse(curp->man)) {
702: assert(MANDOCLEVEL_FATAL <= curp->file_status);
703: return;
704: }
705:
706: roff_endparse(curp->roff);
707: }
708:
709: static void
1.36 schwarze 710: mparse_parse_buffer(struct mparse *curp, struct buf blk, const char *file)
1.28 joerg 711: {
712: const char *svfile;
1.36 schwarze 713: static int recursion_depth;
714:
715: if (64 < recursion_depth) {
716: mandoc_msg(MANDOCERR_ROFFLOOP, curp, curp->line, 0, NULL);
717: return;
718: }
1.28 joerg 719:
720: /* Line number is per-file. */
721: svfile = curp->file;
722: curp->file = file;
723: curp->line = 1;
1.36 schwarze 724: recursion_depth++;
1.28 joerg 725:
726: mparse_buf_r(curp, blk, 1);
727:
1.36 schwarze 728: if (0 == --recursion_depth && MANDOCLEVEL_FATAL > curp->file_status)
1.28 joerg 729: mparse_end(curp);
730:
731: curp->file = svfile;
732: }
733:
734: enum mandoclevel
735: mparse_readmem(struct mparse *curp, const void *buf, size_t len,
736: const char *file)
737: {
738: struct buf blk;
739:
740: blk.buf = UNCONST(buf);
741: blk.sz = len;
742:
1.36 schwarze 743: mparse_parse_buffer(curp, blk, file);
1.28 joerg 744: return(curp->file_status);
745: }
746:
1.36 schwarze 747: enum mandoclevel
748: mparse_readfd(struct mparse *curp, int fd, const char *file)
1.1 kristaps 749: {
1.28 joerg 750: struct buf blk;
751: int with_mmap;
1.1 kristaps 752:
1.40 schwarze 753: if (-1 == fd && -1 == (fd = open(file, O_RDONLY, 0))) {
754: curp->file_status = MANDOCLEVEL_SYSERR;
755: if (curp->mmsg)
756: (*curp->mmsg)(MANDOCERR_SYSOPEN,
757: curp->file_status,
758: file, 0, 0, strerror(errno));
759: goto out;
760: }
761:
1.28 joerg 762: /*
763: * Run for each opened file; may be called more than once for
764: * each full parse sequence if the opened file is nested (i.e.,
765: * from `so'). Simply sucks in the whole file and moves into
766: * the parse phase for the file.
767: */
1.1 kristaps 768:
1.40 schwarze 769: if ( ! read_whole_file(curp, file, fd, &blk, &with_mmap))
1.36 schwarze 770: goto out;
1.1 kristaps 771:
1.36 schwarze 772: mparse_parse_buffer(curp, blk, file);
1.1 kristaps 773:
1.81 schwarze 774: #if HAVE_MMAP
1.28 joerg 775: if (with_mmap)
776: munmap(blk.buf, blk.sz);
777: else
778: #endif
779: free(blk.buf);
1.1 kristaps 780:
781: if (STDIN_FILENO != fd && -1 == close(fd))
782: perror(file);
1.36 schwarze 783: out:
1.1 kristaps 784: return(curp->file_status);
1.82 ! schwarze 785: }
! 786:
! 787: enum mandoclevel
! 788: mparse_open(struct mparse *curp, int *fd, const char *file,
! 789: pid_t *child_pid)
! 790: {
! 791: int pfd[2];
! 792: char *cp;
! 793: enum mandocerr err;
! 794:
! 795: pfd[1] = -1;
! 796: curp->file = file;
! 797: if ((cp = strrchr(file, '.')) == NULL ||
! 798: strcmp(cp + 1, "gz")) {
! 799: *child_pid = 0;
! 800: if ((*fd = open(file, O_RDONLY)) == -1) {
! 801: err = MANDOCERR_SYSOPEN;
! 802: goto out;
! 803: }
! 804: return(MANDOCLEVEL_OK);
! 805: }
! 806:
! 807: if (pipe(pfd) == -1) {
! 808: err = MANDOCERR_SYSPIPE;
! 809: goto out;
! 810: }
! 811:
! 812: switch (*child_pid = fork()) {
! 813: case -1:
! 814: err = MANDOCERR_SYSFORK;
! 815: close(pfd[0]);
! 816: close(pfd[1]);
! 817: pfd[1] = -1;
! 818: break;
! 819: case 0:
! 820: close(pfd[0]);
! 821: if (dup2(pfd[1], STDOUT_FILENO) == -1) {
! 822: err = MANDOCERR_SYSDUP;
! 823: break;
! 824: }
! 825: execlp("gunzip", "gunzip", "-c", file, NULL);
! 826: err = MANDOCERR_SYSEXEC;
! 827: break;
! 828: default:
! 829: close(pfd[1]);
! 830: *fd = pfd[0];
! 831: return(MANDOCLEVEL_OK);
! 832: }
! 833:
! 834: out:
! 835: *fd = -1;
! 836: *child_pid = 0;
! 837: curp->file_status = MANDOCLEVEL_SYSERR;
! 838: if (curp->mmsg)
! 839: (*curp->mmsg)(err, curp->file_status, file,
! 840: 0, 0, strerror(errno));
! 841: if (pfd[1] != -1)
! 842: exit(1);
! 843: return(curp->file_status);
! 844: }
! 845:
! 846: enum mandoclevel
! 847: mparse_wait(struct mparse *curp, pid_t child_pid)
! 848: {
! 849: int status;
! 850:
! 851: if (waitpid(child_pid, &status, 0) == -1) {
! 852: mandoc_msg(MANDOCERR_SYSWAIT, curp, 0, 0,
! 853: strerror(errno));
! 854: curp->file_status = MANDOCLEVEL_SYSERR;
! 855: return(curp->file_status);
! 856: }
! 857: if (WIFSIGNALED(status)) {
! 858: mandoc_vmsg(MANDOCERR_SYSSIG, curp, 0, 0,
! 859: "%d", WTERMSIG(status));
! 860: curp->file_status = MANDOCLEVEL_SYSERR;
! 861: return(curp->file_status);
! 862: }
! 863: if (WEXITSTATUS(status)) {
! 864: mandoc_vmsg(MANDOCERR_SYSEXIT, curp, 0, 0,
! 865: "%d", WEXITSTATUS(status));
! 866: curp->file_status = MANDOCLEVEL_SYSERR;
! 867: return(curp->file_status);
! 868: }
! 869: return(MANDOCLEVEL_OK);
1.1 kristaps 870: }
871:
872: struct mparse *
1.44 schwarze 873: mparse_alloc(int options, enum mandoclevel wlevel,
1.69 schwarze 874: mandocmsg mmsg, const char *defos)
1.1 kristaps 875: {
876: struct mparse *curp;
1.10 kristaps 877:
878: assert(wlevel <= MANDOCLEVEL_FATAL);
1.1 kristaps 879:
880: curp = mandoc_calloc(1, sizeof(struct mparse));
881:
1.44 schwarze 882: curp->options = options;
1.3 kristaps 883: curp->wlevel = wlevel;
1.1 kristaps 884: curp->mmsg = mmsg;
1.29 schwarze 885: curp->defos = defos;
1.1 kristaps 886:
1.44 schwarze 887: curp->roff = roff_alloc(curp, options);
1.72 schwarze 888: if (curp->options & MPARSE_MDOC)
889: curp->pmdoc = mdoc_alloc(
890: curp->roff, curp, curp->defos,
891: curp->options & MPARSE_QUICK ? 1 : 0);
892: if (curp->options & MPARSE_MAN)
893: curp->pman = man_alloc(curp->roff, curp,
894: curp->options & MPARSE_QUICK ? 1 : 0);
895:
1.1 kristaps 896: return(curp);
897: }
898:
899: void
900: mparse_reset(struct mparse *curp)
901: {
902:
903: roff_reset(curp->roff);
904:
905: if (curp->mdoc)
906: mdoc_reset(curp->mdoc);
907: if (curp->man)
908: man_reset(curp->man);
1.24 kristaps 909: if (curp->secondary)
910: curp->secondary->sz = 0;
1.1 kristaps 911:
912: curp->file_status = MANDOCLEVEL_OK;
913: curp->mdoc = NULL;
914: curp->man = NULL;
1.45 schwarze 915:
916: free(curp->sodest);
917: curp->sodest = NULL;
1.1 kristaps 918: }
919:
920: void
921: mparse_free(struct mparse *curp)
922: {
923:
924: if (curp->pmdoc)
925: mdoc_free(curp->pmdoc);
926: if (curp->pman)
927: man_free(curp->pman);
928: if (curp->roff)
929: roff_free(curp->roff);
1.24 kristaps 930: if (curp->secondary)
931: free(curp->secondary->buf);
1.1 kristaps 932:
1.24 kristaps 933: free(curp->secondary);
1.45 schwarze 934: free(curp->sodest);
1.1 kristaps 935: free(curp);
936: }
937:
938: void
1.45 schwarze 939: mparse_result(struct mparse *curp,
940: struct mdoc **mdoc, struct man **man, char **sodest)
1.1 kristaps 941: {
942:
1.45 schwarze 943: if (sodest && NULL != (*sodest = curp->sodest)) {
944: *mdoc = NULL;
945: *man = NULL;
946: return;
947: }
1.9 kristaps 948: if (mdoc)
949: *mdoc = curp->mdoc;
950: if (man)
951: *man = curp->man;
1.3 kristaps 952: }
953:
954: void
955: mandoc_vmsg(enum mandocerr t, struct mparse *m,
956: int ln, int pos, const char *fmt, ...)
957: {
958: char buf[256];
959: va_list ap;
960:
961: va_start(ap, fmt);
1.48 schwarze 962: (void)vsnprintf(buf, sizeof(buf), fmt, ap);
1.3 kristaps 963: va_end(ap);
964:
965: mandoc_msg(t, m, ln, pos, buf);
966: }
967:
968: void
1.47 schwarze 969: mandoc_msg(enum mandocerr er, struct mparse *m,
1.3 kristaps 970: int ln, int col, const char *msg)
971: {
972: enum mandoclevel level;
973:
974: level = MANDOCLEVEL_FATAL;
975: while (er < mandoclimits[level])
976: level--;
977:
978: if (level < m->wlevel)
979: return;
980:
1.8 kristaps 981: if (m->mmsg)
982: (*m->mmsg)(er, level, m->file, ln, col, msg);
1.3 kristaps 983:
984: if (m->file_status < level)
985: m->file_status = level;
1.7 kristaps 986: }
987:
988: const char *
989: mparse_strerror(enum mandocerr er)
990: {
991:
992: return(mandocerrs[er]);
993: }
994:
995: const char *
996: mparse_strlevel(enum mandoclevel lvl)
997: {
998: return(mandoclevels[lvl]);
1.24 kristaps 999: }
1000:
1001: void
1002: mparse_keep(struct mparse *p)
1003: {
1004:
1005: assert(NULL == p->secondary);
1006: p->secondary = mandoc_calloc(1, sizeof(struct buf));
1007: }
1008:
1009: const char *
1010: mparse_getkeep(const struct mparse *p)
1011: {
1012:
1013: assert(p->secondary);
1014: return(p->secondary->sz ? p->secondary->buf : NULL);
1.1 kristaps 1015: }
CVSweb