Annotation of src/usr.bin/mandoc/read.c, Revision 1.66
1.66 ! schwarze 1: /* $OpenBSD: read.c,v 1.65 2014/10/11 21:14:11 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: */
1.58 schwarze 19: #include <sys/types.h>
20: #include <sys/mman.h>
1.1 schwarze 21: #include <sys/stat.h>
1.58 schwarze 22: #include <sys/wait.h>
1.1 schwarze 23:
24: #include <assert.h>
25: #include <ctype.h>
1.18 schwarze 26: #include <errno.h>
1.1 schwarze 27: #include <fcntl.h>
28: #include <stdarg.h>
29: #include <stdio.h>
30: #include <stdlib.h>
31: #include <string.h>
32: #include <unistd.h>
33:
34: #include "mandoc.h"
1.24 schwarze 35: #include "mandoc_aux.h"
1.1 schwarze 36: #include "libmandoc.h"
37: #include "mdoc.h"
38: #include "man.h"
39:
40: #define REPARSE_LIMIT 1000
41:
42: struct buf {
1.25 schwarze 43: char *buf; /* binary input buffer */
1.1 schwarze 44: size_t sz; /* size of binary buffer */
45: };
46:
47: struct mparse {
48: struct man *pman; /* persistent man parser */
49: struct mdoc *pmdoc; /* persistent mdoc parser */
50: struct man *man; /* man parser */
51: struct mdoc *mdoc; /* mdoc parser */
52: struct roff *roff; /* roff parser (!NULL) */
1.23 schwarze 53: char *sodest; /* filename pointed to by .so */
1.59 schwarze 54: const char *file; /* filename of current input file */
55: struct buf *primary; /* buffer currently being parsed */
56: struct buf *secondary; /* preprocessed copy of input */
57: const char *defos; /* default operating system */
58: mandocmsg mmsg; /* warning/error message handler */
59: enum mandoclevel file_status; /* status of current parse */
60: enum mandoclevel wlevel; /* ignore messages below this */
61: int options; /* parser options */
1.1 schwarze 62: int reparse_count; /* finite interp. stack */
1.59 schwarze 63: int line; /* line number in the file */
1.1 schwarze 64: };
65:
1.60 schwarze 66: static void choose_parser(struct mparse *);
1.1 schwarze 67: static void resize_buf(struct buf *, size_t);
68: static void mparse_buf_r(struct mparse *, struct buf, int);
1.18 schwarze 69: static int read_whole_file(struct mparse *, const char *, int,
70: struct buf *, int *);
1.1 schwarze 71: static void mparse_end(struct mparse *);
1.15 schwarze 72: static void mparse_parse_buffer(struct mparse *, struct buf,
73: const char *);
1.1 schwarze 74:
75: static const enum mandocerr mandoclimits[MANDOCLEVEL_MAX] = {
76: MANDOCERR_OK,
77: MANDOCERR_WARNING,
78: MANDOCERR_WARNING,
79: MANDOCERR_ERROR,
80: MANDOCERR_FATAL,
81: MANDOCERR_MAX,
82: MANDOCERR_MAX
83: };
84:
85: static const char * const mandocerrs[MANDOCERR_MAX] = {
86: "ok",
87:
88: "generic warning",
89:
90: /* related to the prologue */
1.57 schwarze 91: "missing manual title, using UNTITLED",
92: "missing manual title, using \"\"",
1.32 schwarze 93: "lower case character in document title",
1.57 schwarze 94: "missing manual section, using \"\"",
1.1 schwarze 95: "unknown manual section",
1.11 schwarze 96: "unknown manual volume or arch",
1.32 schwarze 97: "missing date, using today's date",
1.1 schwarze 98: "cannot parse date, using it verbatim",
1.57 schwarze 99: "missing Os macro, using \"\"",
100: "duplicate prologue macro",
101: "late prologue macro",
102: "skipping late title macro",
1.1 schwarze 103: "prologue macros out of order",
104:
105: /* related to document structure */
106: ".so is fragile, better use ln(1)",
1.28 schwarze 107: "no document body",
1.32 schwarze 108: "content before first section header",
109: "first section is not \"NAME\"",
1.1 schwarze 110: "bad NAME section contents",
111: "sections out of conventional order",
1.32 schwarze 112: "duplicate section title",
113: "unexpected section",
1.63 schwarze 114: "unusual Xr order",
115: "unusual Xr punctuation",
1.62 schwarze 116: "AUTHORS section without An macro",
1.1 schwarze 117:
118: /* related to macros and nesting */
1.33 schwarze 119: "obsolete macro",
1.1 schwarze 120: "skipping paragraph macro",
1.10 schwarze 121: "moving paragraph macro out of list",
1.1 schwarze 122: "skipping no-space macro",
123: "blocks badly nested",
124: "nested displays are not portable",
1.35 schwarze 125: "moving content out of list",
126: ".Vt block has child macro",
1.56 schwarze 127: "fill mode already enabled, skipping",
128: "fill mode already disabled, skipping",
1.1 schwarze 129: "line scope broken",
130:
131: /* related to missing macro arguments */
1.36 schwarze 132: "skipping empty request",
133: "conditional request controls empty scope",
1.1 schwarze 134: "skipping empty macro",
1.40 schwarze 135: "empty argument, using 0n",
1.1 schwarze 136: "argument count wrong",
1.38 schwarze 137: "missing display type, using -ragged",
138: "list type is not the first argument",
139: "missing -width in -tag list, using 8n",
1.56 schwarze 140: "missing utility name, using \"\"",
1.38 schwarze 141: "empty head in list item",
142: "empty list item",
1.39 schwarze 143: "missing font type, using \\fR",
144: "unknown font type, using \\fR",
1.38 schwarze 145: "missing -std argument, adding it",
1.66 ! schwarze 146: "missing eqn box, using \"\"",
1.1 schwarze 147:
148: /* related to bad macro arguments */
1.42 schwarze 149: "unterminated quoted argument",
1.1 schwarze 150: "duplicate argument",
1.54 schwarze 151: "skipping duplicate argument",
1.41 schwarze 152: "skipping duplicate display type",
153: "skipping duplicate list type",
1.54 schwarze 154: "skipping -width argument",
1.1 schwarze 155: "unknown AT&T UNIX version",
1.64 schwarze 156: "comma in function argument",
1.65 schwarze 157: "parenthesis in function name",
1.45 schwarze 158: "invalid content in Rs block",
1.41 schwarze 159: "invalid Boolean argument",
160: "unknown font, skipping request",
1.1 schwarze 161:
162: /* related to plain text */
1.42 schwarze 163: "blank line in fill mode, using .sp",
164: "tab in filled text",
165: "whitespace at end of input line",
1.1 schwarze 166: "bad comment style",
1.42 schwarze 167: "invalid escape sequence",
168: "undefined string, using \"\"",
1.3 schwarze 169:
1.1 schwarze 170: "generic error",
171:
1.3 schwarze 172: /* related to equations */
173: "unexpected equation scope closure",
174: "equation scope open on exit",
175: "overlapping equation scopes",
176: "unexpected end of equation",
177:
1.1 schwarze 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:
1.46 schwarze 188: /* related to document structure and macros */
1.1 schwarze 189: "input stack limit exceeded, infinite loop?",
190: "skipping bad character",
1.46 schwarze 191: "skipping unknown macro",
1.51 schwarze 192: "skipping item outside list",
1.46 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 */
1.1 schwarze 199: "escaped character not allowed in a name",
200: "argument count wrong",
1.49 schwarze 201: "missing list type, using -item",
1.48 schwarze 202: "missing manual name, using \"\"",
1.49 schwarze 203: "uname(3) system call failed, using UNKNOWN",
1.41 schwarze 204: "unknown standard specifier",
1.49 schwarze 205: "skipping request without numeric argument",
1.39 schwarze 206: "skipping all arguments",
207: "skipping excess arguments",
1.1 schwarze 208:
209: "generic fatal error",
210:
1.18 schwarze 211: "input too large",
1.56 schwarze 212: "NOT IMPLEMENTED: Bd -file",
1.1 schwarze 213: "NOT IMPLEMENTED: .so with absolute path or \"..\"",
1.30 schwarze 214: ".so request failed",
1.18 schwarze 215:
216: /* system errors */
1.58 schwarze 217: "cannot dup file descriptor",
218: "cannot exec",
219: "gunzip failed with code",
220: "cannot fork",
1.29 schwarze 221: NULL,
1.58 schwarze 222: "cannot open pipe",
223: "cannot read file",
224: "gunzip died from signal",
1.18 schwarze 225: "cannot stat file",
1.58 schwarze 226: "wait failed",
1.1 schwarze 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:
1.25 schwarze 239:
1.1 schwarze 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
1.60 schwarze 249: choose_parser(struct mparse *curp)
1.1 schwarze 250: {
1.59 schwarze 251: char *cp, *ep;
252: int format;
1.1 schwarze 253:
1.59 schwarze 254: /*
255: * If neither command line arguments -mdoc or -man select
256: * a parser nor the roff parser found a .Dd or .TH macro
257: * yet, look ahead in the main input buffer.
258: */
259:
260: if ((format = roff_getformat(curp->roff)) == 0) {
261: cp = curp->primary->buf;
262: ep = cp + curp->primary->sz;
263: while (cp < ep) {
1.61 schwarze 264: if (*cp == '.' || *cp == '\'') {
1.59 schwarze 265: cp++;
266: if (cp[0] == 'D' && cp[1] == 'd') {
267: format = MPARSE_MDOC;
268: break;
269: }
270: if (cp[0] == 'T' && cp[1] == 'H') {
271: format = MPARSE_MAN;
272: break;
273: }
274: }
275: cp = memchr(cp, '\n', ep - cp);
276: if (cp == NULL)
277: break;
278: cp++;
279: }
1.1 schwarze 280: }
281:
1.59 schwarze 282: if (format == MPARSE_MDOC) {
1.25 schwarze 283: if (NULL == curp->pmdoc)
1.22 schwarze 284: curp->pmdoc = mdoc_alloc(
285: curp->roff, curp, curp->defos,
286: MPARSE_QUICK & curp->options ? 1 : 0);
1.1 schwarze 287: assert(curp->pmdoc);
288: curp->mdoc = curp->pmdoc;
289: return;
1.25 schwarze 290: }
1.1 schwarze 291:
1.59 schwarze 292: /* Fall back to man(7) as a last resort. */
293:
1.25 schwarze 294: if (NULL == curp->pman)
1.22 schwarze 295: curp->pman = man_alloc(curp->roff, curp,
296: MPARSE_QUICK & curp->options ? 1 : 0);
1.1 schwarze 297: assert(curp->pman);
298: curp->man = curp->pman;
299: }
300:
301: /*
302: * Main parse routine for an opened file. This is called for each
303: * opened file and simply loops around the full input file, possibly
304: * nesting (i.e., with `so').
305: */
306: static void
307: mparse_buf_r(struct mparse *curp, struct buf blk, int start)
308: {
309: const struct tbl_span *span;
310: struct buf ln;
311: enum rofferr rr;
312: int i, of, rc;
313: int pos; /* byte number in the ln buffer */
314: int lnn; /* line number in the real file */
315: unsigned char c;
316:
317: memset(&ln, 0, sizeof(struct buf));
318:
1.25 schwarze 319: lnn = curp->line;
320: pos = 0;
1.1 schwarze 321:
322: for (i = 0; i < (int)blk.sz; ) {
323: if (0 == pos && '\0' == blk.buf[i])
324: break;
325:
326: if (start) {
327: curp->line = lnn;
328: curp->reparse_count = 0;
329: }
330:
331: while (i < (int)blk.sz && (start || '\0' != blk.buf[i])) {
332:
333: /*
334: * When finding an unescaped newline character,
335: * leave the character loop to process the line.
336: * Skip a preceding carriage return, if any.
337: */
338:
339: if ('\r' == blk.buf[i] && i + 1 < (int)blk.sz &&
340: '\n' == blk.buf[i + 1])
341: ++i;
342: if ('\n' == blk.buf[i]) {
343: ++i;
344: ++lnn;
345: break;
346: }
347:
1.13 schwarze 348: /*
349: * Make sure we have space for at least
350: * one backslash and one other character
351: * and the trailing NUL byte.
352: */
353:
354: if (pos + 2 >= (int)ln.sz)
355: resize_buf(&ln, 256);
356:
1.25 schwarze 357: /*
1.1 schwarze 358: * Warn about bogus characters. If you're using
359: * non-ASCII encoding, you're screwing your
360: * readers. Since I'd rather this not happen,
1.6 schwarze 361: * I'll be helpful and replace these characters
362: * with "?", so we don't display gibberish.
363: * Note to manual writers: use special characters.
1.1 schwarze 364: */
365:
366: c = (unsigned char) blk.buf[i];
367:
1.25 schwarze 368: if ( ! (isascii(c) &&
369: (isgraph(c) || isblank(c)))) {
1.56 schwarze 370: mandoc_vmsg(MANDOCERR_BADCHAR, curp,
371: curp->line, pos, "0x%x", c);
1.1 schwarze 372: i++;
1.6 schwarze 373: ln.buf[pos++] = '?';
1.1 schwarze 374: continue;
375: }
376:
377: /* Trailing backslash = a plain char. */
378:
379: if ('\\' != blk.buf[i] || i + 1 == (int)blk.sz) {
380: ln.buf[pos++] = blk.buf[i++];
381: continue;
382: }
383:
384: /*
385: * Found escape and at least one other character.
386: * When it's a newline character, skip it.
387: * When there is a carriage return in between,
388: * skip that one as well.
389: */
390:
391: if ('\r' == blk.buf[i + 1] && i + 2 < (int)blk.sz &&
392: '\n' == blk.buf[i + 2])
393: ++i;
394: if ('\n' == blk.buf[i + 1]) {
395: i += 2;
396: ++lnn;
397: continue;
398: }
399:
400: if ('"' == blk.buf[i + 1] || '#' == blk.buf[i + 1]) {
401: i += 2;
402: /* Comment, skip to end of line */
403: for (; i < (int)blk.sz; ++i) {
404: if ('\n' == blk.buf[i]) {
405: ++i;
406: ++lnn;
407: break;
408: }
409: }
410:
411: /* Backout trailing whitespaces */
412: for (; pos > 0; --pos) {
413: if (ln.buf[pos - 1] != ' ')
414: break;
415: if (pos > 2 && ln.buf[pos - 2] == '\\')
416: break;
417: }
418: break;
419: }
420:
1.13 schwarze 421: /* Catch escaped bogus characters. */
422:
423: c = (unsigned char) blk.buf[i+1];
424:
1.25 schwarze 425: if ( ! (isascii(c) &&
426: (isgraph(c) || isblank(c)))) {
1.56 schwarze 427: mandoc_vmsg(MANDOCERR_BADCHAR, curp,
428: curp->line, pos, "0x%x", c);
1.13 schwarze 429: i += 2;
430: ln.buf[pos++] = '?';
431: continue;
432: }
433:
1.1 schwarze 434: /* Some other escape sequence, copy & cont. */
435:
436: ln.buf[pos++] = blk.buf[i++];
437: ln.buf[pos++] = blk.buf[i++];
438: }
439:
1.25 schwarze 440: if (pos >= (int)ln.sz)
1.1 schwarze 441: resize_buf(&ln, 256);
442:
443: ln.buf[pos] = '\0';
444:
445: /*
446: * A significant amount of complexity is contained by
447: * the roff preprocessor. It's line-oriented but can be
448: * expressed on one line, so we need at times to
449: * readjust our starting point and re-run it. The roff
450: * preprocessor can also readjust the buffers with new
451: * data, so we pass them in wholesale.
452: */
453:
454: of = 0;
455:
1.4 schwarze 456: /*
457: * Maintain a lookaside buffer of all parsed lines. We
458: * only do this if mparse_keep() has been invoked (the
459: * buffer may be accessed with mparse_getkeep()).
460: */
461:
462: if (curp->secondary) {
1.25 schwarze 463: curp->secondary->buf = mandoc_realloc(
464: curp->secondary->buf,
465: curp->secondary->sz + pos + 2);
466: memcpy(curp->secondary->buf +
467: curp->secondary->sz,
468: ln.buf, pos);
1.4 schwarze 469: curp->secondary->sz += pos;
470: curp->secondary->buf
471: [curp->secondary->sz] = '\n';
472: curp->secondary->sz++;
473: curp->secondary->buf
474: [curp->secondary->sz] = '\0';
475: }
1.1 schwarze 476: rerun:
1.25 schwarze 477: rr = roff_parseln(curp->roff, curp->line,
478: &ln.buf, &ln.sz, of, &of);
1.1 schwarze 479:
480: switch (rr) {
1.25 schwarze 481: case ROFF_REPARSE:
1.1 schwarze 482: if (REPARSE_LIMIT >= ++curp->reparse_count)
483: mparse_buf_r(curp, ln, 0);
484: else
485: mandoc_msg(MANDOCERR_ROFFLOOP, curp,
1.25 schwarze 486: curp->line, pos, NULL);
1.1 schwarze 487: pos = 0;
488: continue;
1.25 schwarze 489: case ROFF_APPEND:
1.1 schwarze 490: pos = (int)strlen(ln.buf);
491: continue;
1.25 schwarze 492: case ROFF_RERUN:
1.1 schwarze 493: goto rerun;
1.25 schwarze 494: case ROFF_IGN:
1.1 schwarze 495: pos = 0;
496: continue;
1.25 schwarze 497: case ROFF_ERR:
1.1 schwarze 498: assert(MANDOCLEVEL_FATAL <= curp->file_status);
499: break;
1.25 schwarze 500: case ROFF_SO:
1.23 schwarze 501: if (0 == (MPARSE_SO & curp->options) &&
502: (i >= (int)blk.sz || '\0' == blk.buf[i])) {
503: curp->sodest = mandoc_strdup(ln.buf + of);
504: free(ln.buf);
505: return;
506: }
1.4 schwarze 507: /*
508: * We remove `so' clauses from our lookaside
509: * buffer because we're going to descend into
510: * the file recursively.
511: */
1.25 schwarze 512: if (curp->secondary)
1.4 schwarze 513: curp->secondary->sz -= pos + 1;
1.14 schwarze 514: mparse_readfd(curp, -1, ln.buf + of);
1.30 schwarze 515: if (MANDOCLEVEL_FATAL <= curp->file_status) {
516: mandoc_vmsg(MANDOCERR_SO_FAIL,
517: curp, curp->line, pos,
518: ".so %s", ln.buf + of);
1.1 schwarze 519: break;
1.30 schwarze 520: }
1.1 schwarze 521: pos = 0;
522: continue;
523: default:
524: break;
525: }
526:
527: /*
528: * If we encounter errors in the recursive parse, make
529: * sure we don't continue parsing.
530: */
531:
532: if (MANDOCLEVEL_FATAL <= curp->file_status)
533: break;
534:
535: /*
536: * If input parsers have not been allocated, do so now.
1.2 schwarze 537: * We keep these instanced between parsers, but set them
1.1 schwarze 538: * locally per parse routine since we can use different
539: * parsers with each one.
540: */
541:
542: if ( ! (curp->man || curp->mdoc))
1.60 schwarze 543: choose_parser(curp);
1.1 schwarze 544:
1.25 schwarze 545: /*
1.60 schwarze 546: * Lastly, push down into the parsers themselves.
1.1 schwarze 547: * If libroff returns ROFF_TBL, then add it to the
548: * currently open parse. Since we only get here if
549: * there does exist data (see tbl_data.c), we're
550: * guaranteed that something's been allocated.
551: * Do the same for ROFF_EQN.
552: */
553:
554: rc = -1;
555:
556: if (ROFF_TBL == rr)
557: while (NULL != (span = roff_span(curp->roff))) {
558: rc = curp->man ?
1.25 schwarze 559: man_addspan(curp->man, span) :
560: mdoc_addspan(curp->mdoc, span);
1.1 schwarze 561: if (0 == rc)
562: break;
563: }
564: else if (ROFF_EQN == rr)
1.25 schwarze 565: rc = curp->mdoc ?
566: mdoc_addeqn(curp->mdoc,
567: roff_eqn(curp->roff)) :
568: man_addeqn(curp->man,
569: roff_eqn(curp->roff));
1.1 schwarze 570: else if (curp->man || curp->mdoc)
571: rc = curp->man ?
1.25 schwarze 572: man_parseln(curp->man,
573: curp->line, ln.buf, of) :
574: mdoc_parseln(curp->mdoc,
575: curp->line, ln.buf, of);
1.1 schwarze 576:
577: if (0 == rc) {
578: assert(MANDOCLEVEL_FATAL <= curp->file_status);
579: break;
1.19 schwarze 580: } else if (2 == rc)
581: break;
1.1 schwarze 582:
583: /* Temporary buffers typically are not full. */
584:
585: if (0 == start && '\0' == blk.buf[i])
586: break;
587:
588: /* Start the next input line. */
589:
590: pos = 0;
591: }
592:
593: free(ln.buf);
594: }
595:
596: static int
1.18 schwarze 597: read_whole_file(struct mparse *curp, const char *file, int fd,
598: struct buf *fb, int *with_mmap)
1.1 schwarze 599: {
600: struct stat st;
601: size_t off;
602: ssize_t ssz;
603:
604: if (-1 == fstat(fd, &st)) {
1.18 schwarze 605: curp->file_status = MANDOCLEVEL_SYSERR;
606: if (curp->mmsg)
607: (*curp->mmsg)(MANDOCERR_SYSSTAT, curp->file_status,
608: file, 0, 0, strerror(errno));
1.1 schwarze 609: return(0);
610: }
611:
612: /*
613: * If we're a regular file, try just reading in the whole entry
614: * via mmap(). This is faster than reading it into blocks, and
615: * since each file is only a few bytes to begin with, I'm not
616: * concerned that this is going to tank any machines.
617: */
618:
619: if (S_ISREG(st.st_mode)) {
620: if (st.st_size >= (1U << 31)) {
1.18 schwarze 621: curp->file_status = MANDOCLEVEL_FATAL;
622: if (curp->mmsg)
623: (*curp->mmsg)(MANDOCERR_TOOLARGE,
624: curp->file_status, file, 0, 0, NULL);
1.1 schwarze 625: return(0);
626: }
627: *with_mmap = 1;
628: fb->sz = (size_t)st.st_size;
1.15 schwarze 629: fb->buf = mmap(NULL, fb->sz, PROT_READ, MAP_SHARED, fd, 0);
1.1 schwarze 630: if (fb->buf != MAP_FAILED)
631: return(1);
632: }
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)) {
1.18 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 schwarze 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) {
1.18 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 schwarze 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:
1.50 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 schwarze 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:
1.15 schwarze 709: static void
710: mparse_parse_buffer(struct mparse *curp, struct buf blk, const char *file)
1.1 schwarze 711: {
1.61 schwarze 712: struct buf *svprimary;
1.1 schwarze 713: const char *svfile;
1.14 schwarze 714: static int recursion_depth;
715:
716: if (64 < recursion_depth) {
717: mandoc_msg(MANDOCERR_ROFFLOOP, curp, curp->line, 0, NULL);
1.15 schwarze 718: return;
1.14 schwarze 719: }
1.1 schwarze 720:
1.15 schwarze 721: /* Line number is per-file. */
722: svfile = curp->file;
723: curp->file = file;
1.61 schwarze 724: svprimary = curp->primary;
1.59 schwarze 725: curp->primary = &blk;
1.15 schwarze 726: curp->line = 1;
727: recursion_depth++;
728:
729: mparse_buf_r(curp, blk, 1);
730:
731: if (0 == --recursion_depth && MANDOCLEVEL_FATAL > curp->file_status)
732: mparse_end(curp);
733:
1.61 schwarze 734: curp->primary = svprimary;
1.15 schwarze 735: curp->file = svfile;
736: }
737:
738: enum mandoclevel
739: mparse_readfd(struct mparse *curp, int fd, const char *file)
740: {
741: struct buf blk;
742: int with_mmap;
743:
1.18 schwarze 744: if (-1 == fd && -1 == (fd = open(file, O_RDONLY, 0))) {
745: curp->file_status = MANDOCLEVEL_SYSERR;
746: if (curp->mmsg)
747: (*curp->mmsg)(MANDOCERR_SYSOPEN,
748: curp->file_status,
749: file, 0, 0, strerror(errno));
750: goto out;
751: }
752:
1.15 schwarze 753: /*
754: * Run for each opened file; may be called more than once for
755: * each full parse sequence if the opened file is nested (i.e.,
756: * from `so'). Simply sucks in the whole file and moves into
757: * the parse phase for the file.
758: */
1.1 schwarze 759:
1.18 schwarze 760: if ( ! read_whole_file(curp, file, fd, &blk, &with_mmap))
1.15 schwarze 761: goto out;
1.1 schwarze 762:
1.15 schwarze 763: mparse_parse_buffer(curp, blk, file);
1.1 schwarze 764:
1.15 schwarze 765: if (with_mmap)
766: munmap(blk.buf, blk.sz);
767: else
768: free(blk.buf);
1.1 schwarze 769:
770: if (STDIN_FILENO != fd && -1 == close(fd))
771: perror(file);
1.14 schwarze 772: out:
1.1 schwarze 773: return(curp->file_status);
1.58 schwarze 774: }
775:
776: enum mandoclevel
777: mparse_open(struct mparse *curp, int *fd, const char *file,
778: pid_t *child_pid)
779: {
780: int pfd[2];
781: char *cp;
782: enum mandocerr err;
783:
784: pfd[1] = -1;
785: curp->file = file;
786: if ((cp = strrchr(file, '.')) == NULL ||
787: strcmp(cp + 1, "gz")) {
788: *child_pid = 0;
789: if ((*fd = open(file, O_RDONLY)) == -1) {
790: err = MANDOCERR_SYSOPEN;
791: goto out;
792: }
793: return(MANDOCLEVEL_OK);
794: }
795:
796: if (pipe(pfd) == -1) {
797: err = MANDOCERR_SYSPIPE;
798: goto out;
799: }
800:
801: switch (*child_pid = fork()) {
802: case -1:
803: err = MANDOCERR_SYSFORK;
804: close(pfd[0]);
805: close(pfd[1]);
806: pfd[1] = -1;
807: break;
808: case 0:
809: close(pfd[0]);
810: if (dup2(pfd[1], STDOUT_FILENO) == -1) {
811: err = MANDOCERR_SYSDUP;
812: break;
813: }
814: execlp("gunzip", "gunzip", "-c", file, NULL);
815: err = MANDOCERR_SYSEXEC;
816: break;
817: default:
818: close(pfd[1]);
819: *fd = pfd[0];
820: return(MANDOCLEVEL_OK);
821: }
822:
823: out:
824: *fd = -1;
825: *child_pid = 0;
826: curp->file_status = MANDOCLEVEL_SYSERR;
827: if (curp->mmsg)
828: (*curp->mmsg)(err, curp->file_status, file,
829: 0, 0, strerror(errno));
830: if (pfd[1] != -1)
831: exit(1);
832: return(curp->file_status);
833: }
834:
835: enum mandoclevel
836: mparse_wait(struct mparse *curp, pid_t child_pid)
837: {
838: int status;
839:
840: if (waitpid(child_pid, &status, 0) == -1) {
841: mandoc_msg(MANDOCERR_SYSWAIT, curp, 0, 0,
842: strerror(errno));
843: curp->file_status = MANDOCLEVEL_SYSERR;
844: return(curp->file_status);
845: }
846: if (WIFSIGNALED(status)) {
847: mandoc_vmsg(MANDOCERR_SYSSIG, curp, 0, 0,
848: "%d", WTERMSIG(status));
849: curp->file_status = MANDOCLEVEL_SYSERR;
850: return(curp->file_status);
851: }
852: if (WEXITSTATUS(status)) {
853: mandoc_vmsg(MANDOCERR_SYSEXIT, curp, 0, 0,
854: "%d", WEXITSTATUS(status));
855: curp->file_status = MANDOCLEVEL_SYSERR;
856: return(curp->file_status);
857: }
858: return(MANDOCLEVEL_OK);
1.1 schwarze 859: }
860:
861: struct mparse *
1.22 schwarze 862: mparse_alloc(int options, enum mandoclevel wlevel,
1.47 schwarze 863: mandocmsg mmsg, const char *defos)
1.1 schwarze 864: {
865: struct mparse *curp;
866:
867: assert(wlevel <= MANDOCLEVEL_FATAL);
868:
869: curp = mandoc_calloc(1, sizeof(struct mparse));
870:
1.22 schwarze 871: curp->options = options;
1.1 schwarze 872: curp->wlevel = wlevel;
873: curp->mmsg = mmsg;
1.7 schwarze 874: curp->defos = defos;
1.1 schwarze 875:
1.22 schwarze 876: curp->roff = roff_alloc(curp, options);
1.50 schwarze 877: if (curp->options & MPARSE_MDOC)
878: curp->pmdoc = mdoc_alloc(
879: curp->roff, curp, curp->defos,
880: curp->options & MPARSE_QUICK ? 1 : 0);
881: if (curp->options & MPARSE_MAN)
882: curp->pman = man_alloc(curp->roff, curp,
883: curp->options & MPARSE_QUICK ? 1 : 0);
884:
1.1 schwarze 885: return(curp);
886: }
887:
888: void
889: mparse_reset(struct mparse *curp)
890: {
891:
892: roff_reset(curp->roff);
893:
894: if (curp->mdoc)
895: mdoc_reset(curp->mdoc);
896: if (curp->man)
897: man_reset(curp->man);
1.4 schwarze 898: if (curp->secondary)
899: curp->secondary->sz = 0;
1.1 schwarze 900:
901: curp->file_status = MANDOCLEVEL_OK;
902: curp->mdoc = NULL;
903: curp->man = NULL;
1.23 schwarze 904:
905: free(curp->sodest);
906: curp->sodest = NULL;
1.1 schwarze 907: }
908:
909: void
910: mparse_free(struct mparse *curp)
911: {
912:
913: if (curp->pmdoc)
914: mdoc_free(curp->pmdoc);
915: if (curp->pman)
916: man_free(curp->pman);
917: if (curp->roff)
918: roff_free(curp->roff);
1.4 schwarze 919: if (curp->secondary)
920: free(curp->secondary->buf);
1.1 schwarze 921:
1.4 schwarze 922: free(curp->secondary);
1.23 schwarze 923: free(curp->sodest);
1.1 schwarze 924: free(curp);
925: }
926:
927: void
1.23 schwarze 928: mparse_result(struct mparse *curp,
929: struct mdoc **mdoc, struct man **man, char **sodest)
1.1 schwarze 930: {
931:
1.23 schwarze 932: if (sodest && NULL != (*sodest = curp->sodest)) {
933: *mdoc = NULL;
934: *man = NULL;
935: return;
936: }
1.1 schwarze 937: if (mdoc)
938: *mdoc = curp->mdoc;
939: if (man)
940: *man = curp->man;
941: }
942:
943: void
944: mandoc_vmsg(enum mandocerr t, struct mparse *m,
945: int ln, int pos, const char *fmt, ...)
946: {
947: char buf[256];
948: va_list ap;
949:
950: va_start(ap, fmt);
1.26 schwarze 951: (void)vsnprintf(buf, sizeof(buf), fmt, ap);
1.1 schwarze 952: va_end(ap);
953:
954: mandoc_msg(t, m, ln, pos, buf);
955: }
956:
957: void
1.25 schwarze 958: mandoc_msg(enum mandocerr er, struct mparse *m,
1.1 schwarze 959: int ln, int col, const char *msg)
960: {
961: enum mandoclevel level;
962:
963: level = MANDOCLEVEL_FATAL;
964: while (er < mandoclimits[level])
965: level--;
966:
967: if (level < m->wlevel)
968: return;
969:
970: if (m->mmsg)
971: (*m->mmsg)(er, level, m->file, ln, col, msg);
972:
973: if (m->file_status < level)
974: m->file_status = level;
975: }
976:
977: const char *
978: mparse_strerror(enum mandocerr er)
979: {
980:
981: return(mandocerrs[er]);
982: }
983:
984: const char *
985: mparse_strlevel(enum mandoclevel lvl)
986: {
987: return(mandoclevels[lvl]);
1.4 schwarze 988: }
989:
990: void
991: mparse_keep(struct mparse *p)
992: {
993:
994: assert(NULL == p->secondary);
995: p->secondary = mandoc_calloc(1, sizeof(struct buf));
996: }
997:
998: const char *
999: mparse_getkeep(const struct mparse *p)
1000: {
1001:
1002: assert(p->secondary);
1003: return(p->secondary->sz ? p->secondary->buf : NULL);
1.1 schwarze 1004: }