Annotation of src/usr.bin/mandoc/read.c, Revision 1.64
1.64 ! schwarze 1: /* $OpenBSD: read.c,v 1.63 2014/09/11 23:52:47 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",
1.1 schwarze 146:
147: /* related to bad macro arguments */
1.42 schwarze 148: "unterminated quoted argument",
1.1 schwarze 149: "duplicate argument",
1.54 schwarze 150: "skipping duplicate argument",
1.41 schwarze 151: "skipping duplicate display type",
152: "skipping duplicate list type",
1.54 schwarze 153: "skipping -width argument",
1.1 schwarze 154: "unknown AT&T UNIX version",
1.64 ! schwarze 155: "comma in function argument",
1.45 schwarze 156: "invalid content in Rs block",
1.41 schwarze 157: "invalid Boolean argument",
158: "unknown font, skipping request",
1.1 schwarze 159:
160: /* related to plain text */
1.42 schwarze 161: "blank line in fill mode, using .sp",
162: "tab in filled text",
163: "whitespace at end of input line",
1.1 schwarze 164: "bad comment style",
1.42 schwarze 165: "invalid escape sequence",
166: "undefined string, using \"\"",
1.3 schwarze 167:
1.1 schwarze 168: "generic error",
169:
1.3 schwarze 170: /* related to equations */
171: "unexpected equation scope closure",
172: "equation scope open on exit",
173: "overlapping equation scopes",
174: "unexpected end of equation",
175: "equation syntax error",
176:
1.1 schwarze 177: /* related to tables */
178: "bad table syntax",
179: "bad table option",
180: "bad table layout",
181: "no table layout cells specified",
182: "no table data cells specified",
183: "ignore data in cell",
184: "data block still open",
185: "ignoring extra data cells",
186:
1.46 schwarze 187: /* related to document structure and macros */
1.1 schwarze 188: "input stack limit exceeded, infinite loop?",
189: "skipping bad character",
1.46 schwarze 190: "skipping unknown macro",
1.51 schwarze 191: "skipping item outside list",
1.46 schwarze 192: "skipping column outside column list",
193: "skipping end of block that is not open",
194: "inserting missing end of block",
195: "appending missing end of block",
196:
197: /* related to request and macro arguments */
1.1 schwarze 198: "escaped character not allowed in a name",
199: "argument count wrong",
1.49 schwarze 200: "missing list type, using -item",
1.48 schwarze 201: "missing manual name, using \"\"",
1.49 schwarze 202: "uname(3) system call failed, using UNKNOWN",
1.41 schwarze 203: "unknown standard specifier",
1.49 schwarze 204: "skipping request without numeric argument",
1.39 schwarze 205: "skipping all arguments",
206: "skipping excess arguments",
1.1 schwarze 207:
208: "generic fatal error",
209:
1.18 schwarze 210: "input too large",
1.56 schwarze 211: "NOT IMPLEMENTED: Bd -file",
1.1 schwarze 212: "NOT IMPLEMENTED: .so with absolute path or \"..\"",
1.30 schwarze 213: ".so request failed",
1.18 schwarze 214:
215: /* system errors */
1.58 schwarze 216: "cannot dup file descriptor",
217: "cannot exec",
218: "gunzip failed with code",
219: "cannot fork",
1.29 schwarze 220: NULL,
1.58 schwarze 221: "cannot open pipe",
222: "cannot read file",
223: "gunzip died from signal",
1.18 schwarze 224: "cannot stat file",
1.58 schwarze 225: "wait failed",
1.1 schwarze 226: };
227:
228: static const char * const mandoclevels[MANDOCLEVEL_MAX] = {
229: "SUCCESS",
230: "RESERVED",
231: "WARNING",
232: "ERROR",
233: "FATAL",
234: "BADARG",
235: "SYSERR"
236: };
237:
1.25 schwarze 238:
1.1 schwarze 239: static void
240: resize_buf(struct buf *buf, size_t initial)
241: {
242:
243: buf->sz = buf->sz > initial/2 ? 2 * buf->sz : initial;
244: buf->buf = mandoc_realloc(buf->buf, buf->sz);
245: }
246:
247: static void
1.60 schwarze 248: choose_parser(struct mparse *curp)
1.1 schwarze 249: {
1.59 schwarze 250: char *cp, *ep;
251: int format;
1.1 schwarze 252:
1.59 schwarze 253: /*
254: * If neither command line arguments -mdoc or -man select
255: * a parser nor the roff parser found a .Dd or .TH macro
256: * yet, look ahead in the main input buffer.
257: */
258:
259: if ((format = roff_getformat(curp->roff)) == 0) {
260: cp = curp->primary->buf;
261: ep = cp + curp->primary->sz;
262: while (cp < ep) {
1.61 schwarze 263: if (*cp == '.' || *cp == '\'') {
1.59 schwarze 264: cp++;
265: if (cp[0] == 'D' && cp[1] == 'd') {
266: format = MPARSE_MDOC;
267: break;
268: }
269: if (cp[0] == 'T' && cp[1] == 'H') {
270: format = MPARSE_MAN;
271: break;
272: }
273: }
274: cp = memchr(cp, '\n', ep - cp);
275: if (cp == NULL)
276: break;
277: cp++;
278: }
1.1 schwarze 279: }
280:
1.59 schwarze 281: if (format == MPARSE_MDOC) {
1.25 schwarze 282: if (NULL == curp->pmdoc)
1.22 schwarze 283: curp->pmdoc = mdoc_alloc(
284: curp->roff, curp, curp->defos,
285: MPARSE_QUICK & curp->options ? 1 : 0);
1.1 schwarze 286: assert(curp->pmdoc);
287: curp->mdoc = curp->pmdoc;
288: return;
1.25 schwarze 289: }
1.1 schwarze 290:
1.59 schwarze 291: /* Fall back to man(7) as a last resort. */
292:
1.25 schwarze 293: if (NULL == curp->pman)
1.22 schwarze 294: curp->pman = man_alloc(curp->roff, curp,
295: MPARSE_QUICK & curp->options ? 1 : 0);
1.1 schwarze 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:
1.25 schwarze 318: lnn = curp->line;
319: pos = 0;
1.1 schwarze 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:
1.13 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:
1.25 schwarze 356: /*
1.1 schwarze 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,
1.6 schwarze 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 schwarze 363: */
364:
365: c = (unsigned char) blk.buf[i];
366:
1.25 schwarze 367: if ( ! (isascii(c) &&
368: (isgraph(c) || isblank(c)))) {
1.56 schwarze 369: mandoc_vmsg(MANDOCERR_BADCHAR, curp,
370: curp->line, pos, "0x%x", c);
1.1 schwarze 371: i++;
1.6 schwarze 372: ln.buf[pos++] = '?';
1.1 schwarze 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:
399: if ('"' == blk.buf[i + 1] || '#' == blk.buf[i + 1]) {
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:
1.13 schwarze 420: /* Catch escaped bogus characters. */
421:
422: c = (unsigned char) blk.buf[i+1];
423:
1.25 schwarze 424: if ( ! (isascii(c) &&
425: (isgraph(c) || isblank(c)))) {
1.56 schwarze 426: mandoc_vmsg(MANDOCERR_BADCHAR, curp,
427: curp->line, pos, "0x%x", c);
1.13 schwarze 428: i += 2;
429: ln.buf[pos++] = '?';
430: continue;
431: }
432:
1.1 schwarze 433: /* Some other escape sequence, copy & cont. */
434:
435: ln.buf[pos++] = blk.buf[i++];
436: ln.buf[pos++] = blk.buf[i++];
437: }
438:
1.25 schwarze 439: if (pos >= (int)ln.sz)
1.1 schwarze 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:
1.4 schwarze 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) {
1.25 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);
1.4 schwarze 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 schwarze 475: rerun:
1.25 schwarze 476: rr = roff_parseln(curp->roff, curp->line,
477: &ln.buf, &ln.sz, of, &of);
1.1 schwarze 478:
479: switch (rr) {
1.25 schwarze 480: case ROFF_REPARSE:
1.1 schwarze 481: if (REPARSE_LIMIT >= ++curp->reparse_count)
482: mparse_buf_r(curp, ln, 0);
483: else
484: mandoc_msg(MANDOCERR_ROFFLOOP, curp,
1.25 schwarze 485: curp->line, pos, NULL);
1.1 schwarze 486: pos = 0;
487: continue;
1.25 schwarze 488: case ROFF_APPEND:
1.1 schwarze 489: pos = (int)strlen(ln.buf);
490: continue;
1.25 schwarze 491: case ROFF_RERUN:
1.1 schwarze 492: goto rerun;
1.25 schwarze 493: case ROFF_IGN:
1.1 schwarze 494: pos = 0;
495: continue;
1.25 schwarze 496: case ROFF_ERR:
1.1 schwarze 497: assert(MANDOCLEVEL_FATAL <= curp->file_status);
498: break;
1.25 schwarze 499: case ROFF_SO:
1.23 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: }
1.4 schwarze 506: /*
507: * We remove `so' clauses from our lookaside
508: * buffer because we're going to descend into
509: * the file recursively.
510: */
1.25 schwarze 511: if (curp->secondary)
1.4 schwarze 512: curp->secondary->sz -= pos + 1;
1.14 schwarze 513: mparse_readfd(curp, -1, ln.buf + of);
1.30 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 schwarze 518: break;
1.30 schwarze 519: }
1.1 schwarze 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.
1.2 schwarze 536: * We keep these instanced between parsers, but set them
1.1 schwarze 537: * locally per parse routine since we can use different
538: * parsers with each one.
539: */
540:
541: if ( ! (curp->man || curp->mdoc))
1.60 schwarze 542: choose_parser(curp);
1.1 schwarze 543:
1.25 schwarze 544: /*
1.60 schwarze 545: * Lastly, push down into the parsers themselves.
1.1 schwarze 546: * If libroff returns ROFF_TBL, then add it to the
547: * currently open parse. Since we only get here if
548: * there does exist data (see tbl_data.c), we're
549: * guaranteed that something's been allocated.
550: * Do the same for ROFF_EQN.
551: */
552:
553: rc = -1;
554:
555: if (ROFF_TBL == rr)
556: while (NULL != (span = roff_span(curp->roff))) {
557: rc = curp->man ?
1.25 schwarze 558: man_addspan(curp->man, span) :
559: mdoc_addspan(curp->mdoc, span);
1.1 schwarze 560: if (0 == rc)
561: break;
562: }
563: else if (ROFF_EQN == rr)
1.25 schwarze 564: rc = curp->mdoc ?
565: mdoc_addeqn(curp->mdoc,
566: roff_eqn(curp->roff)) :
567: man_addeqn(curp->man,
568: roff_eqn(curp->roff));
1.1 schwarze 569: else if (curp->man || curp->mdoc)
570: rc = curp->man ?
1.25 schwarze 571: man_parseln(curp->man,
572: curp->line, ln.buf, of) :
573: mdoc_parseln(curp->mdoc,
574: curp->line, ln.buf, of);
1.1 schwarze 575:
576: if (0 == rc) {
577: assert(MANDOCLEVEL_FATAL <= curp->file_status);
578: break;
1.19 schwarze 579: } else if (2 == rc)
580: break;
1.1 schwarze 581:
582: /* Temporary buffers typically are not full. */
583:
584: if (0 == start && '\0' == blk.buf[i])
585: break;
586:
587: /* Start the next input line. */
588:
589: pos = 0;
590: }
591:
592: free(ln.buf);
593: }
594:
595: static int
1.18 schwarze 596: read_whole_file(struct mparse *curp, const char *file, int fd,
597: struct buf *fb, int *with_mmap)
1.1 schwarze 598: {
599: struct stat st;
600: size_t off;
601: ssize_t ssz;
602:
603: if (-1 == fstat(fd, &st)) {
1.18 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 schwarze 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)) {
1.18 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 schwarze 624: return(0);
625: }
626: *with_mmap = 1;
627: fb->sz = (size_t)st.st_size;
1.15 schwarze 628: fb->buf = mmap(NULL, fb->sz, PROT_READ, MAP_SHARED, fd, 0);
1.1 schwarze 629: if (fb->buf != MAP_FAILED)
630: return(1);
631: }
632:
633: /*
634: * If this isn't a regular file (like, say, stdin), then we must
635: * go the old way and just read things in bit by bit.
636: */
637:
638: *with_mmap = 0;
639: off = 0;
640: fb->sz = 0;
641: fb->buf = NULL;
642: for (;;) {
643: if (off == fb->sz) {
644: if (fb->sz == (1U << 31)) {
1.18 schwarze 645: curp->file_status = MANDOCLEVEL_FATAL;
646: if (curp->mmsg)
647: (*curp->mmsg)(MANDOCERR_TOOLARGE,
648: curp->file_status,
649: file, 0, 0, NULL);
1.1 schwarze 650: break;
651: }
652: resize_buf(fb, 65536);
653: }
654: ssz = read(fd, fb->buf + (int)off, fb->sz - off);
655: if (ssz == 0) {
656: fb->sz = off;
657: return(1);
658: }
659: if (ssz == -1) {
1.18 schwarze 660: curp->file_status = MANDOCLEVEL_SYSERR;
661: if (curp->mmsg)
662: (*curp->mmsg)(MANDOCERR_SYSREAD,
663: curp->file_status, file, 0, 0,
664: strerror(errno));
1.1 schwarze 665: break;
666: }
667: off += (size_t)ssz;
668: }
669:
670: free(fb->buf);
671: fb->buf = NULL;
672: return(0);
673: }
674:
675: static void
676: mparse_end(struct mparse *curp)
677: {
678:
679: if (MANDOCLEVEL_FATAL <= curp->file_status)
680: return;
681:
1.50 schwarze 682: if (curp->mdoc == NULL &&
683: curp->man == NULL &&
684: curp->sodest == NULL) {
685: if (curp->options & MPARSE_MDOC)
686: curp->mdoc = curp->pmdoc;
687: else {
688: if (curp->pman == NULL)
689: curp->pman = man_alloc(curp->roff, curp,
690: curp->options & MPARSE_QUICK ? 1 : 0);
691: curp->man = curp->pman;
692: }
693: }
694:
1.1 schwarze 695: if (curp->mdoc && ! mdoc_endparse(curp->mdoc)) {
696: assert(MANDOCLEVEL_FATAL <= curp->file_status);
697: return;
698: }
699:
700: if (curp->man && ! man_endparse(curp->man)) {
701: assert(MANDOCLEVEL_FATAL <= curp->file_status);
702: return;
703: }
704:
705: roff_endparse(curp->roff);
706: }
707:
1.15 schwarze 708: static void
709: mparse_parse_buffer(struct mparse *curp, struct buf blk, const char *file)
1.1 schwarze 710: {
1.61 schwarze 711: struct buf *svprimary;
1.1 schwarze 712: const char *svfile;
1.14 schwarze 713: static int recursion_depth;
714:
715: if (64 < recursion_depth) {
716: mandoc_msg(MANDOCERR_ROFFLOOP, curp, curp->line, 0, NULL);
1.15 schwarze 717: return;
1.14 schwarze 718: }
1.1 schwarze 719:
1.15 schwarze 720: /* Line number is per-file. */
721: svfile = curp->file;
722: curp->file = file;
1.61 schwarze 723: svprimary = curp->primary;
1.59 schwarze 724: curp->primary = &blk;
1.15 schwarze 725: curp->line = 1;
726: recursion_depth++;
727:
728: mparse_buf_r(curp, blk, 1);
729:
730: if (0 == --recursion_depth && MANDOCLEVEL_FATAL > curp->file_status)
731: mparse_end(curp);
732:
1.61 schwarze 733: curp->primary = svprimary;
1.15 schwarze 734: curp->file = svfile;
735: }
736:
737: enum mandoclevel
738: mparse_readfd(struct mparse *curp, int fd, const char *file)
739: {
740: struct buf blk;
741: int with_mmap;
742:
1.18 schwarze 743: if (-1 == fd && -1 == (fd = open(file, O_RDONLY, 0))) {
744: curp->file_status = MANDOCLEVEL_SYSERR;
745: if (curp->mmsg)
746: (*curp->mmsg)(MANDOCERR_SYSOPEN,
747: curp->file_status,
748: file, 0, 0, strerror(errno));
749: goto out;
750: }
751:
1.15 schwarze 752: /*
753: * Run for each opened file; may be called more than once for
754: * each full parse sequence if the opened file is nested (i.e.,
755: * from `so'). Simply sucks in the whole file and moves into
756: * the parse phase for the file.
757: */
1.1 schwarze 758:
1.18 schwarze 759: if ( ! read_whole_file(curp, file, fd, &blk, &with_mmap))
1.15 schwarze 760: goto out;
1.1 schwarze 761:
1.15 schwarze 762: mparse_parse_buffer(curp, blk, file);
1.1 schwarze 763:
1.15 schwarze 764: if (with_mmap)
765: munmap(blk.buf, blk.sz);
766: else
767: free(blk.buf);
1.1 schwarze 768:
769: if (STDIN_FILENO != fd && -1 == close(fd))
770: perror(file);
1.14 schwarze 771: out:
1.1 schwarze 772: return(curp->file_status);
1.58 schwarze 773: }
774:
775: enum mandoclevel
776: mparse_open(struct mparse *curp, int *fd, const char *file,
777: pid_t *child_pid)
778: {
779: int pfd[2];
780: char *cp;
781: enum mandocerr err;
782:
783: pfd[1] = -1;
784: curp->file = file;
785: if ((cp = strrchr(file, '.')) == NULL ||
786: strcmp(cp + 1, "gz")) {
787: *child_pid = 0;
788: if ((*fd = open(file, O_RDONLY)) == -1) {
789: err = MANDOCERR_SYSOPEN;
790: goto out;
791: }
792: return(MANDOCLEVEL_OK);
793: }
794:
795: if (pipe(pfd) == -1) {
796: err = MANDOCERR_SYSPIPE;
797: goto out;
798: }
799:
800: switch (*child_pid = fork()) {
801: case -1:
802: err = MANDOCERR_SYSFORK;
803: close(pfd[0]);
804: close(pfd[1]);
805: pfd[1] = -1;
806: break;
807: case 0:
808: close(pfd[0]);
809: if (dup2(pfd[1], STDOUT_FILENO) == -1) {
810: err = MANDOCERR_SYSDUP;
811: break;
812: }
813: execlp("gunzip", "gunzip", "-c", file, NULL);
814: err = MANDOCERR_SYSEXEC;
815: break;
816: default:
817: close(pfd[1]);
818: *fd = pfd[0];
819: return(MANDOCLEVEL_OK);
820: }
821:
822: out:
823: *fd = -1;
824: *child_pid = 0;
825: curp->file_status = MANDOCLEVEL_SYSERR;
826: if (curp->mmsg)
827: (*curp->mmsg)(err, curp->file_status, file,
828: 0, 0, strerror(errno));
829: if (pfd[1] != -1)
830: exit(1);
831: return(curp->file_status);
832: }
833:
834: enum mandoclevel
835: mparse_wait(struct mparse *curp, pid_t child_pid)
836: {
837: int status;
838:
839: if (waitpid(child_pid, &status, 0) == -1) {
840: mandoc_msg(MANDOCERR_SYSWAIT, curp, 0, 0,
841: strerror(errno));
842: curp->file_status = MANDOCLEVEL_SYSERR;
843: return(curp->file_status);
844: }
845: if (WIFSIGNALED(status)) {
846: mandoc_vmsg(MANDOCERR_SYSSIG, curp, 0, 0,
847: "%d", WTERMSIG(status));
848: curp->file_status = MANDOCLEVEL_SYSERR;
849: return(curp->file_status);
850: }
851: if (WEXITSTATUS(status)) {
852: mandoc_vmsg(MANDOCERR_SYSEXIT, curp, 0, 0,
853: "%d", WEXITSTATUS(status));
854: curp->file_status = MANDOCLEVEL_SYSERR;
855: return(curp->file_status);
856: }
857: return(MANDOCLEVEL_OK);
1.1 schwarze 858: }
859:
860: struct mparse *
1.22 schwarze 861: mparse_alloc(int options, enum mandoclevel wlevel,
1.47 schwarze 862: mandocmsg mmsg, const char *defos)
1.1 schwarze 863: {
864: struct mparse *curp;
865:
866: assert(wlevel <= MANDOCLEVEL_FATAL);
867:
868: curp = mandoc_calloc(1, sizeof(struct mparse));
869:
1.22 schwarze 870: curp->options = options;
1.1 schwarze 871: curp->wlevel = wlevel;
872: curp->mmsg = mmsg;
1.7 schwarze 873: curp->defos = defos;
1.1 schwarze 874:
1.22 schwarze 875: curp->roff = roff_alloc(curp, options);
1.50 schwarze 876: if (curp->options & MPARSE_MDOC)
877: curp->pmdoc = mdoc_alloc(
878: curp->roff, curp, curp->defos,
879: curp->options & MPARSE_QUICK ? 1 : 0);
880: if (curp->options & MPARSE_MAN)
881: curp->pman = man_alloc(curp->roff, curp,
882: curp->options & MPARSE_QUICK ? 1 : 0);
883:
1.1 schwarze 884: return(curp);
885: }
886:
887: void
888: mparse_reset(struct mparse *curp)
889: {
890:
891: roff_reset(curp->roff);
892:
893: if (curp->mdoc)
894: mdoc_reset(curp->mdoc);
895: if (curp->man)
896: man_reset(curp->man);
1.4 schwarze 897: if (curp->secondary)
898: curp->secondary->sz = 0;
1.1 schwarze 899:
900: curp->file_status = MANDOCLEVEL_OK;
901: curp->mdoc = NULL;
902: curp->man = NULL;
1.23 schwarze 903:
904: free(curp->sodest);
905: curp->sodest = NULL;
1.1 schwarze 906: }
907:
908: void
909: mparse_free(struct mparse *curp)
910: {
911:
912: if (curp->pmdoc)
913: mdoc_free(curp->pmdoc);
914: if (curp->pman)
915: man_free(curp->pman);
916: if (curp->roff)
917: roff_free(curp->roff);
1.4 schwarze 918: if (curp->secondary)
919: free(curp->secondary->buf);
1.1 schwarze 920:
1.4 schwarze 921: free(curp->secondary);
1.23 schwarze 922: free(curp->sodest);
1.1 schwarze 923: free(curp);
924: }
925:
926: void
1.23 schwarze 927: mparse_result(struct mparse *curp,
928: struct mdoc **mdoc, struct man **man, char **sodest)
1.1 schwarze 929: {
930:
1.23 schwarze 931: if (sodest && NULL != (*sodest = curp->sodest)) {
932: *mdoc = NULL;
933: *man = NULL;
934: return;
935: }
1.1 schwarze 936: if (mdoc)
937: *mdoc = curp->mdoc;
938: if (man)
939: *man = curp->man;
940: }
941:
942: void
943: mandoc_vmsg(enum mandocerr t, struct mparse *m,
944: int ln, int pos, const char *fmt, ...)
945: {
946: char buf[256];
947: va_list ap;
948:
949: va_start(ap, fmt);
1.26 schwarze 950: (void)vsnprintf(buf, sizeof(buf), fmt, ap);
1.1 schwarze 951: va_end(ap);
952:
953: mandoc_msg(t, m, ln, pos, buf);
954: }
955:
956: void
1.25 schwarze 957: mandoc_msg(enum mandocerr er, struct mparse *m,
1.1 schwarze 958: int ln, int col, const char *msg)
959: {
960: enum mandoclevel level;
961:
962: level = MANDOCLEVEL_FATAL;
963: while (er < mandoclimits[level])
964: level--;
965:
966: if (level < m->wlevel)
967: return;
968:
969: if (m->mmsg)
970: (*m->mmsg)(er, level, m->file, ln, col, msg);
971:
972: if (m->file_status < level)
973: m->file_status = level;
974: }
975:
976: const char *
977: mparse_strerror(enum mandocerr er)
978: {
979:
980: return(mandocerrs[er]);
981: }
982:
983: const char *
984: mparse_strlevel(enum mandoclevel lvl)
985: {
986: return(mandoclevels[lvl]);
1.4 schwarze 987: }
988:
989: void
990: mparse_keep(struct mparse *p)
991: {
992:
993: assert(NULL == p->secondary);
994: p->secondary = mandoc_calloc(1, sizeof(struct buf));
995: }
996:
997: const char *
998: mparse_getkeep(const struct mparse *p)
999: {
1000:
1001: assert(p->secondary);
1002: return(p->secondary->sz ? p->secondary->buf : NULL);
1.1 schwarze 1003: }