Annotation of src/usr.bin/mandoc/read.c, Revision 1.59
1.59 ! schwarze 1: /* $OpenBSD: read.c,v 1.58 2014/09/03 23:20:33 schwarze Exp $ */
1.1 schwarze 2: /*
3: * Copyright (c) 2008, 2009, 2010, 2011 Kristaps Dzonsons <kristaps@bsd.lv>
1.18 schwarze 4: * Copyright (c) 2010-2014 Ingo Schwarze <schwarze@openbsd.org>
1.20 schwarze 5: * Copyright (c) 2010, 2012 Joerg Sonnenberger <joerg@netbsd.org>
1.1 schwarze 6: *
7: * Permission to use, copy, modify, and distribute this software for any
8: * purpose with or without fee is hereby granted, provided that the above
9: * copyright notice and this permission notice appear in all copies.
10: *
11: * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
12: * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
13: * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
14: * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
15: * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
16: * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
17: * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
18: */
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:
66: static void resize_buf(struct buf *, size_t);
67: static void mparse_buf_r(struct mparse *, struct buf, int);
68: static void pset(const char *, int, struct mparse *);
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",
1.1 schwarze 114:
115: /* related to macros and nesting */
1.33 schwarze 116: "obsolete macro",
1.1 schwarze 117: "skipping paragraph macro",
1.10 schwarze 118: "moving paragraph macro out of list",
1.1 schwarze 119: "skipping no-space macro",
120: "blocks badly nested",
121: "nested displays are not portable",
1.35 schwarze 122: "moving content out of list",
123: ".Vt block has child macro",
1.56 schwarze 124: "fill mode already enabled, skipping",
125: "fill mode already disabled, skipping",
1.1 schwarze 126: "line scope broken",
127:
128: /* related to missing macro arguments */
1.36 schwarze 129: "skipping empty request",
130: "conditional request controls empty scope",
1.1 schwarze 131: "skipping empty macro",
1.40 schwarze 132: "empty argument, using 0n",
1.1 schwarze 133: "argument count wrong",
1.38 schwarze 134: "missing display type, using -ragged",
135: "list type is not the first argument",
136: "missing -width in -tag list, using 8n",
1.56 schwarze 137: "missing utility name, using \"\"",
1.38 schwarze 138: "empty head in list item",
139: "empty list item",
1.39 schwarze 140: "missing font type, using \\fR",
141: "unknown font type, using \\fR",
1.38 schwarze 142: "missing -std argument, adding it",
1.1 schwarze 143:
144: /* related to bad macro arguments */
1.42 schwarze 145: "unterminated quoted argument",
1.1 schwarze 146: "duplicate argument",
1.54 schwarze 147: "skipping duplicate argument",
1.41 schwarze 148: "skipping duplicate display type",
149: "skipping duplicate list type",
1.54 schwarze 150: "skipping -width argument",
1.1 schwarze 151: "unknown AT&T UNIX version",
1.45 schwarze 152: "invalid content in Rs block",
1.41 schwarze 153: "invalid Boolean argument",
154: "unknown font, skipping request",
1.1 schwarze 155:
156: /* related to plain text */
1.42 schwarze 157: "blank line in fill mode, using .sp",
158: "tab in filled text",
159: "whitespace at end of input line",
1.1 schwarze 160: "bad comment style",
1.42 schwarze 161: "invalid escape sequence",
162: "undefined string, using \"\"",
1.3 schwarze 163:
1.1 schwarze 164: "generic error",
165:
1.3 schwarze 166: /* related to equations */
167: "unexpected equation scope closure",
168: "equation scope open on exit",
169: "overlapping equation scopes",
170: "unexpected end of equation",
171: "equation syntax error",
172:
1.1 schwarze 173: /* related to tables */
174: "bad table syntax",
175: "bad table option",
176: "bad table layout",
177: "no table layout cells specified",
178: "no table data cells specified",
179: "ignore data in cell",
180: "data block still open",
181: "ignoring extra data cells",
182:
1.46 schwarze 183: /* related to document structure and macros */
1.1 schwarze 184: "input stack limit exceeded, infinite loop?",
185: "skipping bad character",
1.46 schwarze 186: "skipping unknown macro",
1.51 schwarze 187: "skipping item outside list",
1.46 schwarze 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 */
1.1 schwarze 194: "escaped character not allowed in a name",
195: "argument count wrong",
1.49 schwarze 196: "missing list type, using -item",
1.48 schwarze 197: "missing manual name, using \"\"",
1.49 schwarze 198: "uname(3) system call failed, using UNKNOWN",
1.41 schwarze 199: "unknown standard specifier",
1.49 schwarze 200: "skipping request without numeric argument",
1.39 schwarze 201: "skipping all arguments",
202: "skipping excess arguments",
1.1 schwarze 203:
204: "generic fatal error",
205:
1.18 schwarze 206: "input too large",
1.56 schwarze 207: "NOT IMPLEMENTED: Bd -file",
1.1 schwarze 208: "NOT IMPLEMENTED: .so with absolute path or \"..\"",
1.30 schwarze 209: ".so request failed",
1.18 schwarze 210:
211: /* system errors */
1.58 schwarze 212: "cannot dup file descriptor",
213: "cannot exec",
214: "gunzip failed with code",
215: "cannot fork",
1.29 schwarze 216: NULL,
1.58 schwarze 217: "cannot open pipe",
218: "cannot read file",
219: "gunzip died from signal",
1.18 schwarze 220: "cannot stat file",
1.58 schwarze 221: "wait failed",
1.1 schwarze 222: };
223:
224: static const char * const mandoclevels[MANDOCLEVEL_MAX] = {
225: "SUCCESS",
226: "RESERVED",
227: "WARNING",
228: "ERROR",
229: "FATAL",
230: "BADARG",
231: "SYSERR"
232: };
233:
1.25 schwarze 234:
1.1 schwarze 235: static void
236: resize_buf(struct buf *buf, size_t initial)
237: {
238:
239: buf->sz = buf->sz > initial/2 ? 2 * buf->sz : initial;
240: buf->buf = mandoc_realloc(buf->buf, buf->sz);
241: }
242:
243: static void
244: pset(const char *buf, int pos, struct mparse *curp)
245: {
1.59 ! schwarze 246: char *cp, *ep;
! 247: int format;
1.1 schwarze 248: int i;
249:
250: if ('.' == buf[0] || '\'' == buf[0]) {
251: for (i = 1; buf[i]; i++)
252: if (' ' != buf[i] && '\t' != buf[i])
253: break;
254: if ('\0' == buf[i])
255: return;
256: }
257:
1.59 ! schwarze 258: /*
! 259: * If neither command line arguments -mdoc or -man select
! 260: * a parser nor the roff parser found a .Dd or .TH macro
! 261: * yet, look ahead in the main input buffer.
! 262: */
! 263:
! 264: if ((format = roff_getformat(curp->roff)) == 0) {
! 265: cp = curp->primary->buf;
! 266: ep = cp + curp->primary->sz;
! 267: while (cp < ep) {
! 268: if (*cp == '.' || *cp != '\'') {
! 269: cp++;
! 270: if (cp[0] == 'D' && cp[1] == 'd') {
! 271: format = MPARSE_MDOC;
! 272: break;
! 273: }
! 274: if (cp[0] == 'T' && cp[1] == 'H') {
! 275: format = MPARSE_MAN;
! 276: break;
! 277: }
! 278: }
! 279: cp = memchr(cp, '\n', ep - cp);
! 280: if (cp == NULL)
! 281: break;
! 282: cp++;
! 283: }
1.1 schwarze 284: }
285:
1.59 ! schwarze 286: if (format == MPARSE_MDOC) {
1.25 schwarze 287: if (NULL == curp->pmdoc)
1.22 schwarze 288: curp->pmdoc = mdoc_alloc(
289: curp->roff, curp, curp->defos,
290: MPARSE_QUICK & curp->options ? 1 : 0);
1.1 schwarze 291: assert(curp->pmdoc);
292: curp->mdoc = curp->pmdoc;
293: return;
1.25 schwarze 294: }
1.1 schwarze 295:
1.59 ! schwarze 296: /* Fall back to man(7) as a last resort. */
! 297:
1.25 schwarze 298: if (NULL == curp->pman)
1.22 schwarze 299: curp->pman = man_alloc(curp->roff, curp,
300: MPARSE_QUICK & curp->options ? 1 : 0);
1.1 schwarze 301: assert(curp->pman);
302: curp->man = curp->pman;
303: }
304:
305: /*
306: * Main parse routine for an opened file. This is called for each
307: * opened file and simply loops around the full input file, possibly
308: * nesting (i.e., with `so').
309: */
310: static void
311: mparse_buf_r(struct mparse *curp, struct buf blk, int start)
312: {
313: const struct tbl_span *span;
314: struct buf ln;
315: enum rofferr rr;
316: int i, of, rc;
317: int pos; /* byte number in the ln buffer */
318: int lnn; /* line number in the real file */
319: unsigned char c;
320:
321: memset(&ln, 0, sizeof(struct buf));
322:
1.25 schwarze 323: lnn = curp->line;
324: pos = 0;
1.1 schwarze 325:
326: for (i = 0; i < (int)blk.sz; ) {
327: if (0 == pos && '\0' == blk.buf[i])
328: break;
329:
330: if (start) {
331: curp->line = lnn;
332: curp->reparse_count = 0;
333: }
334:
335: while (i < (int)blk.sz && (start || '\0' != blk.buf[i])) {
336:
337: /*
338: * When finding an unescaped newline character,
339: * leave the character loop to process the line.
340: * Skip a preceding carriage return, if any.
341: */
342:
343: if ('\r' == blk.buf[i] && i + 1 < (int)blk.sz &&
344: '\n' == blk.buf[i + 1])
345: ++i;
346: if ('\n' == blk.buf[i]) {
347: ++i;
348: ++lnn;
349: break;
350: }
351:
1.13 schwarze 352: /*
353: * Make sure we have space for at least
354: * one backslash and one other character
355: * and the trailing NUL byte.
356: */
357:
358: if (pos + 2 >= (int)ln.sz)
359: resize_buf(&ln, 256);
360:
1.25 schwarze 361: /*
1.1 schwarze 362: * Warn about bogus characters. If you're using
363: * non-ASCII encoding, you're screwing your
364: * readers. Since I'd rather this not happen,
1.6 schwarze 365: * I'll be helpful and replace these characters
366: * with "?", so we don't display gibberish.
367: * Note to manual writers: use special characters.
1.1 schwarze 368: */
369:
370: c = (unsigned char) blk.buf[i];
371:
1.25 schwarze 372: if ( ! (isascii(c) &&
373: (isgraph(c) || isblank(c)))) {
1.56 schwarze 374: mandoc_vmsg(MANDOCERR_BADCHAR, curp,
375: curp->line, pos, "0x%x", c);
1.1 schwarze 376: i++;
1.6 schwarze 377: ln.buf[pos++] = '?';
1.1 schwarze 378: continue;
379: }
380:
381: /* Trailing backslash = a plain char. */
382:
383: if ('\\' != blk.buf[i] || i + 1 == (int)blk.sz) {
384: ln.buf[pos++] = blk.buf[i++];
385: continue;
386: }
387:
388: /*
389: * Found escape and at least one other character.
390: * When it's a newline character, skip it.
391: * When there is a carriage return in between,
392: * skip that one as well.
393: */
394:
395: if ('\r' == blk.buf[i + 1] && i + 2 < (int)blk.sz &&
396: '\n' == blk.buf[i + 2])
397: ++i;
398: if ('\n' == blk.buf[i + 1]) {
399: i += 2;
400: ++lnn;
401: continue;
402: }
403:
404: if ('"' == blk.buf[i + 1] || '#' == blk.buf[i + 1]) {
405: i += 2;
406: /* Comment, skip to end of line */
407: for (; i < (int)blk.sz; ++i) {
408: if ('\n' == blk.buf[i]) {
409: ++i;
410: ++lnn;
411: break;
412: }
413: }
414:
415: /* Backout trailing whitespaces */
416: for (; pos > 0; --pos) {
417: if (ln.buf[pos - 1] != ' ')
418: break;
419: if (pos > 2 && ln.buf[pos - 2] == '\\')
420: break;
421: }
422: break;
423: }
424:
1.13 schwarze 425: /* Catch escaped bogus characters. */
426:
427: c = (unsigned char) blk.buf[i+1];
428:
1.25 schwarze 429: if ( ! (isascii(c) &&
430: (isgraph(c) || isblank(c)))) {
1.56 schwarze 431: mandoc_vmsg(MANDOCERR_BADCHAR, curp,
432: curp->line, pos, "0x%x", c);
1.13 schwarze 433: i += 2;
434: ln.buf[pos++] = '?';
435: continue;
436: }
437:
1.1 schwarze 438: /* Some other escape sequence, copy & cont. */
439:
440: ln.buf[pos++] = blk.buf[i++];
441: ln.buf[pos++] = blk.buf[i++];
442: }
443:
1.25 schwarze 444: if (pos >= (int)ln.sz)
1.1 schwarze 445: resize_buf(&ln, 256);
446:
447: ln.buf[pos] = '\0';
448:
449: /*
450: * A significant amount of complexity is contained by
451: * the roff preprocessor. It's line-oriented but can be
452: * expressed on one line, so we need at times to
453: * readjust our starting point and re-run it. The roff
454: * preprocessor can also readjust the buffers with new
455: * data, so we pass them in wholesale.
456: */
457:
458: of = 0;
459:
1.4 schwarze 460: /*
461: * Maintain a lookaside buffer of all parsed lines. We
462: * only do this if mparse_keep() has been invoked (the
463: * buffer may be accessed with mparse_getkeep()).
464: */
465:
466: if (curp->secondary) {
1.25 schwarze 467: curp->secondary->buf = mandoc_realloc(
468: curp->secondary->buf,
469: curp->secondary->sz + pos + 2);
470: memcpy(curp->secondary->buf +
471: curp->secondary->sz,
472: ln.buf, pos);
1.4 schwarze 473: curp->secondary->sz += pos;
474: curp->secondary->buf
475: [curp->secondary->sz] = '\n';
476: curp->secondary->sz++;
477: curp->secondary->buf
478: [curp->secondary->sz] = '\0';
479: }
1.1 schwarze 480: rerun:
1.25 schwarze 481: rr = roff_parseln(curp->roff, curp->line,
482: &ln.buf, &ln.sz, of, &of);
1.1 schwarze 483:
484: switch (rr) {
1.25 schwarze 485: case ROFF_REPARSE:
1.1 schwarze 486: if (REPARSE_LIMIT >= ++curp->reparse_count)
487: mparse_buf_r(curp, ln, 0);
488: else
489: mandoc_msg(MANDOCERR_ROFFLOOP, curp,
1.25 schwarze 490: curp->line, pos, NULL);
1.1 schwarze 491: pos = 0;
492: continue;
1.25 schwarze 493: case ROFF_APPEND:
1.1 schwarze 494: pos = (int)strlen(ln.buf);
495: continue;
1.25 schwarze 496: case ROFF_RERUN:
1.1 schwarze 497: goto rerun;
1.25 schwarze 498: case ROFF_IGN:
1.1 schwarze 499: pos = 0;
500: continue;
1.25 schwarze 501: case ROFF_ERR:
1.1 schwarze 502: assert(MANDOCLEVEL_FATAL <= curp->file_status);
503: break;
1.25 schwarze 504: case ROFF_SO:
1.23 schwarze 505: if (0 == (MPARSE_SO & curp->options) &&
506: (i >= (int)blk.sz || '\0' == blk.buf[i])) {
507: curp->sodest = mandoc_strdup(ln.buf + of);
508: free(ln.buf);
509: return;
510: }
1.4 schwarze 511: /*
512: * We remove `so' clauses from our lookaside
513: * buffer because we're going to descend into
514: * the file recursively.
515: */
1.25 schwarze 516: if (curp->secondary)
1.4 schwarze 517: curp->secondary->sz -= pos + 1;
1.14 schwarze 518: mparse_readfd(curp, -1, ln.buf + of);
1.30 schwarze 519: if (MANDOCLEVEL_FATAL <= curp->file_status) {
520: mandoc_vmsg(MANDOCERR_SO_FAIL,
521: curp, curp->line, pos,
522: ".so %s", ln.buf + of);
1.1 schwarze 523: break;
1.30 schwarze 524: }
1.1 schwarze 525: pos = 0;
526: continue;
527: default:
528: break;
529: }
530:
531: /*
532: * If we encounter errors in the recursive parse, make
533: * sure we don't continue parsing.
534: */
535:
536: if (MANDOCLEVEL_FATAL <= curp->file_status)
537: break;
538:
539: /*
540: * If input parsers have not been allocated, do so now.
1.2 schwarze 541: * We keep these instanced between parsers, but set them
1.1 schwarze 542: * locally per parse routine since we can use different
543: * parsers with each one.
544: */
545:
546: if ( ! (curp->man || curp->mdoc))
547: pset(ln.buf + of, pos - of, curp);
548:
1.25 schwarze 549: /*
1.1 schwarze 550: * Lastly, push down into the parsers themselves. One
551: * of these will have already been set in the pset()
552: * routine.
553: * If libroff returns ROFF_TBL, then add it to the
554: * currently open parse. Since we only get here if
555: * there does exist data (see tbl_data.c), we're
556: * guaranteed that something's been allocated.
557: * Do the same for ROFF_EQN.
558: */
559:
560: rc = -1;
561:
562: if (ROFF_TBL == rr)
563: while (NULL != (span = roff_span(curp->roff))) {
564: rc = curp->man ?
1.25 schwarze 565: man_addspan(curp->man, span) :
566: mdoc_addspan(curp->mdoc, span);
1.1 schwarze 567: if (0 == rc)
568: break;
569: }
570: else if (ROFF_EQN == rr)
1.25 schwarze 571: rc = curp->mdoc ?
572: mdoc_addeqn(curp->mdoc,
573: roff_eqn(curp->roff)) :
574: man_addeqn(curp->man,
575: roff_eqn(curp->roff));
1.1 schwarze 576: else if (curp->man || curp->mdoc)
577: rc = curp->man ?
1.25 schwarze 578: man_parseln(curp->man,
579: curp->line, ln.buf, of) :
580: mdoc_parseln(curp->mdoc,
581: curp->line, ln.buf, of);
1.1 schwarze 582:
583: if (0 == rc) {
584: assert(MANDOCLEVEL_FATAL <= curp->file_status);
585: break;
1.19 schwarze 586: } else if (2 == rc)
587: break;
1.1 schwarze 588:
589: /* Temporary buffers typically are not full. */
590:
591: if (0 == start && '\0' == blk.buf[i])
592: break;
593:
594: /* Start the next input line. */
595:
596: pos = 0;
597: }
598:
599: free(ln.buf);
600: }
601:
602: static int
1.18 schwarze 603: read_whole_file(struct mparse *curp, const char *file, int fd,
604: struct buf *fb, int *with_mmap)
1.1 schwarze 605: {
606: struct stat st;
607: size_t off;
608: ssize_t ssz;
609:
610: if (-1 == fstat(fd, &st)) {
1.18 schwarze 611: curp->file_status = MANDOCLEVEL_SYSERR;
612: if (curp->mmsg)
613: (*curp->mmsg)(MANDOCERR_SYSSTAT, curp->file_status,
614: file, 0, 0, strerror(errno));
1.1 schwarze 615: return(0);
616: }
617:
618: /*
619: * If we're a regular file, try just reading in the whole entry
620: * via mmap(). This is faster than reading it into blocks, and
621: * since each file is only a few bytes to begin with, I'm not
622: * concerned that this is going to tank any machines.
623: */
624:
625: if (S_ISREG(st.st_mode)) {
626: if (st.st_size >= (1U << 31)) {
1.18 schwarze 627: curp->file_status = MANDOCLEVEL_FATAL;
628: if (curp->mmsg)
629: (*curp->mmsg)(MANDOCERR_TOOLARGE,
630: curp->file_status, file, 0, 0, NULL);
1.1 schwarze 631: return(0);
632: }
633: *with_mmap = 1;
634: fb->sz = (size_t)st.st_size;
1.15 schwarze 635: fb->buf = mmap(NULL, fb->sz, PROT_READ, MAP_SHARED, fd, 0);
1.1 schwarze 636: if (fb->buf != MAP_FAILED)
637: return(1);
638: }
639:
640: /*
641: * If this isn't a regular file (like, say, stdin), then we must
642: * go the old way and just read things in bit by bit.
643: */
644:
645: *with_mmap = 0;
646: off = 0;
647: fb->sz = 0;
648: fb->buf = NULL;
649: for (;;) {
650: if (off == fb->sz) {
651: if (fb->sz == (1U << 31)) {
1.18 schwarze 652: curp->file_status = MANDOCLEVEL_FATAL;
653: if (curp->mmsg)
654: (*curp->mmsg)(MANDOCERR_TOOLARGE,
655: curp->file_status,
656: file, 0, 0, NULL);
1.1 schwarze 657: break;
658: }
659: resize_buf(fb, 65536);
660: }
661: ssz = read(fd, fb->buf + (int)off, fb->sz - off);
662: if (ssz == 0) {
663: fb->sz = off;
664: return(1);
665: }
666: if (ssz == -1) {
1.18 schwarze 667: curp->file_status = MANDOCLEVEL_SYSERR;
668: if (curp->mmsg)
669: (*curp->mmsg)(MANDOCERR_SYSREAD,
670: curp->file_status, file, 0, 0,
671: strerror(errno));
1.1 schwarze 672: break;
673: }
674: off += (size_t)ssz;
675: }
676:
677: free(fb->buf);
678: fb->buf = NULL;
679: return(0);
680: }
681:
682: static void
683: mparse_end(struct mparse *curp)
684: {
685:
686: if (MANDOCLEVEL_FATAL <= curp->file_status)
687: return;
688:
1.50 schwarze 689: if (curp->mdoc == NULL &&
690: curp->man == NULL &&
691: curp->sodest == NULL) {
692: if (curp->options & MPARSE_MDOC)
693: curp->mdoc = curp->pmdoc;
694: else {
695: if (curp->pman == NULL)
696: curp->pman = man_alloc(curp->roff, curp,
697: curp->options & MPARSE_QUICK ? 1 : 0);
698: curp->man = curp->pman;
699: }
700: }
701:
1.1 schwarze 702: if (curp->mdoc && ! mdoc_endparse(curp->mdoc)) {
703: assert(MANDOCLEVEL_FATAL <= curp->file_status);
704: return;
705: }
706:
707: if (curp->man && ! man_endparse(curp->man)) {
708: assert(MANDOCLEVEL_FATAL <= curp->file_status);
709: return;
710: }
711:
712: roff_endparse(curp->roff);
713: }
714:
1.15 schwarze 715: static void
716: mparse_parse_buffer(struct mparse *curp, struct buf blk, const char *file)
1.1 schwarze 717: {
718: const char *svfile;
1.14 schwarze 719: static int recursion_depth;
720:
721: if (64 < recursion_depth) {
722: mandoc_msg(MANDOCERR_ROFFLOOP, curp, curp->line, 0, NULL);
1.15 schwarze 723: return;
1.14 schwarze 724: }
1.1 schwarze 725:
1.15 schwarze 726: /* Line number is per-file. */
727: svfile = curp->file;
728: curp->file = file;
1.59 ! schwarze 729: curp->primary = &blk;
1.15 schwarze 730: curp->line = 1;
731: recursion_depth++;
732:
733: mparse_buf_r(curp, blk, 1);
734:
735: if (0 == --recursion_depth && MANDOCLEVEL_FATAL > curp->file_status)
736: mparse_end(curp);
737:
738: curp->file = svfile;
739: }
740:
741: enum mandoclevel
742: mparse_readfd(struct mparse *curp, int fd, const char *file)
743: {
744: struct buf blk;
745: int with_mmap;
746:
1.18 schwarze 747: if (-1 == fd && -1 == (fd = open(file, O_RDONLY, 0))) {
748: curp->file_status = MANDOCLEVEL_SYSERR;
749: if (curp->mmsg)
750: (*curp->mmsg)(MANDOCERR_SYSOPEN,
751: curp->file_status,
752: file, 0, 0, strerror(errno));
753: goto out;
754: }
755:
1.15 schwarze 756: /*
757: * Run for each opened file; may be called more than once for
758: * each full parse sequence if the opened file is nested (i.e.,
759: * from `so'). Simply sucks in the whole file and moves into
760: * the parse phase for the file.
761: */
1.1 schwarze 762:
1.18 schwarze 763: if ( ! read_whole_file(curp, file, fd, &blk, &with_mmap))
1.15 schwarze 764: goto out;
1.1 schwarze 765:
1.15 schwarze 766: mparse_parse_buffer(curp, blk, file);
1.1 schwarze 767:
1.15 schwarze 768: if (with_mmap)
769: munmap(blk.buf, blk.sz);
770: else
771: free(blk.buf);
1.1 schwarze 772:
773: if (STDIN_FILENO != fd && -1 == close(fd))
774: perror(file);
1.14 schwarze 775: out:
1.1 schwarze 776: return(curp->file_status);
1.58 schwarze 777: }
778:
779: enum mandoclevel
780: mparse_open(struct mparse *curp, int *fd, const char *file,
781: pid_t *child_pid)
782: {
783: int pfd[2];
784: char *cp;
785: enum mandocerr err;
786:
787: pfd[1] = -1;
788: curp->file = file;
789: if ((cp = strrchr(file, '.')) == NULL ||
790: strcmp(cp + 1, "gz")) {
791: *child_pid = 0;
792: if ((*fd = open(file, O_RDONLY)) == -1) {
793: err = MANDOCERR_SYSOPEN;
794: goto out;
795: }
796: return(MANDOCLEVEL_OK);
797: }
798:
799: if (pipe(pfd) == -1) {
800: err = MANDOCERR_SYSPIPE;
801: goto out;
802: }
803:
804: switch (*child_pid = fork()) {
805: case -1:
806: err = MANDOCERR_SYSFORK;
807: close(pfd[0]);
808: close(pfd[1]);
809: pfd[1] = -1;
810: break;
811: case 0:
812: close(pfd[0]);
813: if (dup2(pfd[1], STDOUT_FILENO) == -1) {
814: err = MANDOCERR_SYSDUP;
815: break;
816: }
817: execlp("gunzip", "gunzip", "-c", file, NULL);
818: err = MANDOCERR_SYSEXEC;
819: break;
820: default:
821: close(pfd[1]);
822: *fd = pfd[0];
823: return(MANDOCLEVEL_OK);
824: }
825:
826: out:
827: *fd = -1;
828: *child_pid = 0;
829: curp->file_status = MANDOCLEVEL_SYSERR;
830: if (curp->mmsg)
831: (*curp->mmsg)(err, curp->file_status, file,
832: 0, 0, strerror(errno));
833: if (pfd[1] != -1)
834: exit(1);
835: return(curp->file_status);
836: }
837:
838: enum mandoclevel
839: mparse_wait(struct mparse *curp, pid_t child_pid)
840: {
841: int status;
842:
843: if (waitpid(child_pid, &status, 0) == -1) {
844: mandoc_msg(MANDOCERR_SYSWAIT, curp, 0, 0,
845: strerror(errno));
846: curp->file_status = MANDOCLEVEL_SYSERR;
847: return(curp->file_status);
848: }
849: if (WIFSIGNALED(status)) {
850: mandoc_vmsg(MANDOCERR_SYSSIG, curp, 0, 0,
851: "%d", WTERMSIG(status));
852: curp->file_status = MANDOCLEVEL_SYSERR;
853: return(curp->file_status);
854: }
855: if (WEXITSTATUS(status)) {
856: mandoc_vmsg(MANDOCERR_SYSEXIT, curp, 0, 0,
857: "%d", WEXITSTATUS(status));
858: curp->file_status = MANDOCLEVEL_SYSERR;
859: return(curp->file_status);
860: }
861: return(MANDOCLEVEL_OK);
1.1 schwarze 862: }
863:
864: struct mparse *
1.22 schwarze 865: mparse_alloc(int options, enum mandoclevel wlevel,
1.47 schwarze 866: mandocmsg mmsg, const char *defos)
1.1 schwarze 867: {
868: struct mparse *curp;
869:
870: assert(wlevel <= MANDOCLEVEL_FATAL);
871:
872: curp = mandoc_calloc(1, sizeof(struct mparse));
873:
1.22 schwarze 874: curp->options = options;
1.1 schwarze 875: curp->wlevel = wlevel;
876: curp->mmsg = mmsg;
1.7 schwarze 877: curp->defos = defos;
1.1 schwarze 878:
1.22 schwarze 879: curp->roff = roff_alloc(curp, options);
1.50 schwarze 880: if (curp->options & MPARSE_MDOC)
881: curp->pmdoc = mdoc_alloc(
882: curp->roff, curp, curp->defos,
883: curp->options & MPARSE_QUICK ? 1 : 0);
884: if (curp->options & MPARSE_MAN)
885: curp->pman = man_alloc(curp->roff, curp,
886: curp->options & MPARSE_QUICK ? 1 : 0);
887:
1.1 schwarze 888: return(curp);
889: }
890:
891: void
892: mparse_reset(struct mparse *curp)
893: {
894:
895: roff_reset(curp->roff);
896:
897: if (curp->mdoc)
898: mdoc_reset(curp->mdoc);
899: if (curp->man)
900: man_reset(curp->man);
1.4 schwarze 901: if (curp->secondary)
902: curp->secondary->sz = 0;
1.1 schwarze 903:
904: curp->file_status = MANDOCLEVEL_OK;
905: curp->mdoc = NULL;
906: curp->man = NULL;
1.23 schwarze 907:
908: free(curp->sodest);
909: curp->sodest = NULL;
1.1 schwarze 910: }
911:
912: void
913: mparse_free(struct mparse *curp)
914: {
915:
916: if (curp->pmdoc)
917: mdoc_free(curp->pmdoc);
918: if (curp->pman)
919: man_free(curp->pman);
920: if (curp->roff)
921: roff_free(curp->roff);
1.4 schwarze 922: if (curp->secondary)
923: free(curp->secondary->buf);
1.1 schwarze 924:
1.4 schwarze 925: free(curp->secondary);
1.23 schwarze 926: free(curp->sodest);
1.1 schwarze 927: free(curp);
928: }
929:
930: void
1.23 schwarze 931: mparse_result(struct mparse *curp,
932: struct mdoc **mdoc, struct man **man, char **sodest)
1.1 schwarze 933: {
934:
1.23 schwarze 935: if (sodest && NULL != (*sodest = curp->sodest)) {
936: *mdoc = NULL;
937: *man = NULL;
938: return;
939: }
1.1 schwarze 940: if (mdoc)
941: *mdoc = curp->mdoc;
942: if (man)
943: *man = curp->man;
944: }
945:
946: void
947: mandoc_vmsg(enum mandocerr t, struct mparse *m,
948: int ln, int pos, const char *fmt, ...)
949: {
950: char buf[256];
951: va_list ap;
952:
953: va_start(ap, fmt);
1.26 schwarze 954: (void)vsnprintf(buf, sizeof(buf), fmt, ap);
1.1 schwarze 955: va_end(ap);
956:
957: mandoc_msg(t, m, ln, pos, buf);
958: }
959:
960: void
1.25 schwarze 961: mandoc_msg(enum mandocerr er, struct mparse *m,
1.1 schwarze 962: int ln, int col, const char *msg)
963: {
964: enum mandoclevel level;
965:
966: level = MANDOCLEVEL_FATAL;
967: while (er < mandoclimits[level])
968: level--;
969:
970: if (level < m->wlevel)
971: return;
972:
973: if (m->mmsg)
974: (*m->mmsg)(er, level, m->file, ln, col, msg);
975:
976: if (m->file_status < level)
977: m->file_status = level;
978: }
979:
980: const char *
981: mparse_strerror(enum mandocerr er)
982: {
983:
984: return(mandocerrs[er]);
985: }
986:
987: const char *
988: mparse_strlevel(enum mandoclevel lvl)
989: {
990: return(mandoclevels[lvl]);
1.4 schwarze 991: }
992:
993: void
994: mparse_keep(struct mparse *p)
995: {
996:
997: assert(NULL == p->secondary);
998: p->secondary = mandoc_calloc(1, sizeof(struct buf));
999: }
1000:
1001: const char *
1002: mparse_getkeep(const struct mparse *p)
1003: {
1004:
1005: assert(p->secondary);
1006: return(p->secondary->sz ? p->secondary->buf : NULL);
1.1 schwarze 1007: }