Annotation of mandoc/read.c, Revision 1.83
1.83 ! schwarze 1: /* $Id: read.c,v 1.82 2014/09/03 23:21:47 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: struct man *pman; /* persistent man parser */
55: struct mdoc *pmdoc; /* persistent mdoc parser */
56: struct man *man; /* man parser */
57: struct mdoc *mdoc; /* mdoc parser */
58: struct roff *roff; /* roff parser (!NULL) */
1.45 schwarze 59: char *sodest; /* filename pointed to by .so */
1.83 ! schwarze 60: const char *file; /* filename of current input file */
! 61: struct buf *primary; /* buffer currently being parsed */
! 62: struct buf *secondary; /* preprocessed copy of input */
! 63: const char *defos; /* default operating system */
! 64: mandocmsg mmsg; /* warning/error message handler */
! 65: enum mandoclevel file_status; /* status of current parse */
! 66: enum mandoclevel wlevel; /* ignore messages below this */
! 67: int options; /* parser options */
1.1 kristaps 68: int reparse_count; /* finite interp. stack */
1.83 ! schwarze 69: int line; /* line number in the file */
1.1 kristaps 70: };
71:
72: static void resize_buf(struct buf *, size_t);
73: static void mparse_buf_r(struct mparse *, struct buf, int);
74: static void pset(const char *, int, struct mparse *);
1.40 schwarze 75: static int read_whole_file(struct mparse *, const char *, int,
76: struct buf *, int *);
1.1 kristaps 77: static void mparse_end(struct mparse *);
1.37 schwarze 78: static void mparse_parse_buffer(struct mparse *, struct buf,
79: const char *);
1.1 kristaps 80:
1.3 kristaps 81: static const enum mandocerr mandoclimits[MANDOCLEVEL_MAX] = {
82: MANDOCERR_OK,
83: MANDOCERR_WARNING,
84: MANDOCERR_WARNING,
85: MANDOCERR_ERROR,
86: MANDOCERR_FATAL,
87: MANDOCERR_MAX,
88: MANDOCERR_MAX
89: };
90:
1.7 kristaps 91: static const char * const mandocerrs[MANDOCERR_MAX] = {
92: "ok",
93:
94: "generic warning",
95:
96: /* related to the prologue */
1.79 schwarze 97: "missing manual title, using UNTITLED",
98: "missing manual title, using \"\"",
1.54 schwarze 99: "lower case character in document title",
1.79 schwarze 100: "missing manual section, using \"\"",
1.7 kristaps 101: "unknown manual section",
1.32 schwarze 102: "unknown manual volume or arch",
1.54 schwarze 103: "missing date, using today's date",
1.7 kristaps 104: "cannot parse date, using it verbatim",
1.79 schwarze 105: "missing Os macro, using \"\"",
106: "duplicate prologue macro",
107: "late prologue macro",
108: "skipping late title macro",
1.7 kristaps 109: "prologue macros out of order",
110:
111: /* related to document structure */
112: ".so is fragile, better use ln(1)",
1.50 schwarze 113: "no document body",
1.54 schwarze 114: "content before first section header",
115: "first section is not \"NAME\"",
1.7 kristaps 116: "bad NAME section contents",
117: "sections out of conventional order",
1.54 schwarze 118: "duplicate section title",
119: "unexpected section",
1.7 kristaps 120:
121: /* related to macros and nesting */
1.55 schwarze 122: "obsolete macro",
1.7 kristaps 123: "skipping paragraph macro",
1.31 schwarze 124: "moving paragraph macro out of list",
1.7 kristaps 125: "skipping no-space macro",
126: "blocks badly nested",
127: "nested displays are not portable",
1.57 schwarze 128: "moving content out of list",
129: ".Vt block has child macro",
1.78 schwarze 130: "fill mode already enabled, skipping",
131: "fill mode already disabled, skipping",
1.7 kristaps 132: "line scope broken",
133:
134: /* related to missing macro arguments */
1.58 schwarze 135: "skipping empty request",
136: "conditional request controls empty scope",
1.7 kristaps 137: "skipping empty macro",
1.62 schwarze 138: "empty argument, using 0n",
1.7 kristaps 139: "argument count wrong",
1.60 schwarze 140: "missing display type, using -ragged",
141: "list type is not the first argument",
142: "missing -width in -tag list, using 8n",
1.78 schwarze 143: "missing utility name, using \"\"",
1.60 schwarze 144: "empty head in list item",
145: "empty list item",
1.61 schwarze 146: "missing font type, using \\fR",
147: "unknown font type, using \\fR",
1.60 schwarze 148: "missing -std argument, adding it",
1.7 kristaps 149:
150: /* related to bad macro arguments */
1.64 schwarze 151: "unterminated quoted argument",
1.7 kristaps 152: "duplicate argument",
1.76 schwarze 153: "skipping duplicate argument",
1.63 schwarze 154: "skipping duplicate display type",
155: "skipping duplicate list type",
1.76 schwarze 156: "skipping -width argument",
1.7 kristaps 157: "unknown AT&T UNIX version",
1.67 schwarze 158: "invalid content in Rs block",
1.63 schwarze 159: "invalid Boolean argument",
160: "unknown font, skipping request",
1.7 kristaps 161:
162: /* related to plain text */
1.64 schwarze 163: "blank line in fill mode, using .sp",
164: "tab in filled text",
165: "whitespace at end of input line",
1.7 kristaps 166: "bad comment style",
1.64 schwarze 167: "invalid escape sequence",
168: "undefined string, using \"\"",
1.16 kristaps 169:
1.7 kristaps 170: "generic error",
1.17 kristaps 171:
172: /* related to equations */
1.20 kristaps 173: "unexpected equation scope closure",
174: "equation scope open on exit",
1.21 kristaps 175: "overlapping equation scopes",
176: "unexpected end of equation",
177: "equation syntax error",
1.7 kristaps 178:
179: /* related to tables */
180: "bad table syntax",
181: "bad table option",
182: "bad table layout",
183: "no table layout cells specified",
184: "no table data cells specified",
185: "ignore data in cell",
186: "data block still open",
187: "ignoring extra data cells",
188:
1.68 schwarze 189: /* related to document structure and macros */
1.7 kristaps 190: "input stack limit exceeded, infinite loop?",
191: "skipping bad character",
1.68 schwarze 192: "skipping unknown macro",
1.73 schwarze 193: "skipping item outside list",
1.68 schwarze 194: "skipping column outside column list",
195: "skipping end of block that is not open",
196: "inserting missing end of block",
197: "appending missing end of block",
198:
199: /* related to request and macro arguments */
1.7 kristaps 200: "escaped character not allowed in a name",
201: "argument count wrong",
1.71 schwarze 202: "missing list type, using -item",
1.70 schwarze 203: "missing manual name, using \"\"",
1.71 schwarze 204: "uname(3) system call failed, using UNKNOWN",
1.63 schwarze 205: "unknown standard specifier",
1.71 schwarze 206: "skipping request without numeric argument",
1.61 schwarze 207: "skipping all arguments",
208: "skipping excess arguments",
1.7 kristaps 209:
210: "generic fatal error",
211:
1.40 schwarze 212: "input too large",
1.78 schwarze 213: "NOT IMPLEMENTED: Bd -file",
1.7 kristaps 214: "NOT IMPLEMENTED: .so with absolute path or \"..\"",
1.52 schwarze 215: ".so request failed",
1.40 schwarze 216:
217: /* system errors */
1.82 schwarze 218: "cannot dup file descriptor",
219: "cannot exec",
220: "gunzip failed with code",
221: "cannot fork",
1.51 schwarze 222: NULL,
1.82 schwarze 223: "cannot open pipe",
224: "cannot read file",
225: "gunzip died from signal",
1.40 schwarze 226: "cannot stat file",
1.82 schwarze 227: "wait failed",
1.7 kristaps 228: };
229:
230: static const char * const mandoclevels[MANDOCLEVEL_MAX] = {
231: "SUCCESS",
232: "RESERVED",
233: "WARNING",
234: "ERROR",
235: "FATAL",
236: "BADARG",
237: "SYSERR"
238: };
239:
1.47 schwarze 240:
1.1 kristaps 241: static void
242: resize_buf(struct buf *buf, size_t initial)
243: {
244:
245: buf->sz = buf->sz > initial/2 ? 2 * buf->sz : initial;
246: buf->buf = mandoc_realloc(buf->buf, buf->sz);
247: }
248:
249: static void
250: pset(const char *buf, int pos, struct mparse *curp)
251: {
1.83 ! schwarze 252: char *cp, *ep;
! 253: int format;
1.1 kristaps 254: int i;
255:
256: if ('.' == buf[0] || '\'' == buf[0]) {
257: for (i = 1; buf[i]; i++)
258: if (' ' != buf[i] && '\t' != buf[i])
259: break;
260: if ('\0' == buf[i])
261: return;
262: }
263:
1.83 ! schwarze 264: /*
! 265: * If neither command line arguments -mdoc or -man select
! 266: * a parser nor the roff parser found a .Dd or .TH macro
! 267: * yet, look ahead in the main input buffer.
! 268: */
! 269:
! 270: if ((format = roff_getformat(curp->roff)) == 0) {
! 271: cp = curp->primary->buf;
! 272: ep = cp + curp->primary->sz;
! 273: while (cp < ep) {
! 274: if (*cp == '.' || *cp != '\'') {
! 275: cp++;
! 276: if (cp[0] == 'D' && cp[1] == 'd') {
! 277: format = MPARSE_MDOC;
! 278: break;
! 279: }
! 280: if (cp[0] == 'T' && cp[1] == 'H') {
! 281: format = MPARSE_MAN;
! 282: break;
! 283: }
! 284: }
! 285: cp = memchr(cp, '\n', ep - cp);
! 286: if (cp == NULL)
! 287: break;
! 288: cp++;
! 289: }
1.1 kristaps 290: }
291:
1.83 ! schwarze 292: if (format == MPARSE_MDOC) {
1.47 schwarze 293: if (NULL == curp->pmdoc)
1.44 schwarze 294: curp->pmdoc = mdoc_alloc(
295: curp->roff, curp, curp->defos,
296: MPARSE_QUICK & curp->options ? 1 : 0);
1.1 kristaps 297: assert(curp->pmdoc);
298: curp->mdoc = curp->pmdoc;
299: return;
1.47 schwarze 300: }
1.1 kristaps 301:
1.83 ! schwarze 302: /* Fall back to man(7) as a last resort. */
! 303:
1.47 schwarze 304: if (NULL == curp->pman)
1.44 schwarze 305: curp->pman = man_alloc(curp->roff, curp,
306: MPARSE_QUICK & curp->options ? 1 : 0);
1.1 kristaps 307: assert(curp->pman);
308: curp->man = curp->pman;
309: }
310:
311: /*
312: * Main parse routine for an opened file. This is called for each
313: * opened file and simply loops around the full input file, possibly
314: * nesting (i.e., with `so').
315: */
316: static void
317: mparse_buf_r(struct mparse *curp, struct buf blk, int start)
318: {
319: const struct tbl_span *span;
320: struct buf ln;
321: enum rofferr rr;
322: int i, of, rc;
323: int pos; /* byte number in the ln buffer */
324: int lnn; /* line number in the real file */
325: unsigned char c;
326:
327: memset(&ln, 0, sizeof(struct buf));
328:
1.47 schwarze 329: lnn = curp->line;
330: pos = 0;
1.1 kristaps 331:
332: for (i = 0; i < (int)blk.sz; ) {
333: if (0 == pos && '\0' == blk.buf[i])
334: break;
335:
336: if (start) {
337: curp->line = lnn;
338: curp->reparse_count = 0;
339: }
340:
341: while (i < (int)blk.sz && (start || '\0' != blk.buf[i])) {
342:
343: /*
344: * When finding an unescaped newline character,
345: * leave the character loop to process the line.
346: * Skip a preceding carriage return, if any.
347: */
348:
349: if ('\r' == blk.buf[i] && i + 1 < (int)blk.sz &&
350: '\n' == blk.buf[i + 1])
351: ++i;
352: if ('\n' == blk.buf[i]) {
353: ++i;
354: ++lnn;
355: break;
356: }
357:
1.35 schwarze 358: /*
359: * Make sure we have space for at least
360: * one backslash and one other character
361: * and the trailing NUL byte.
362: */
363:
364: if (pos + 2 >= (int)ln.sz)
365: resize_buf(&ln, 256);
366:
1.47 schwarze 367: /*
1.1 kristaps 368: * Warn about bogus characters. If you're using
369: * non-ASCII encoding, you're screwing your
370: * readers. Since I'd rather this not happen,
1.27 joerg 371: * I'll be helpful and replace these characters
372: * with "?", so we don't display gibberish.
373: * Note to manual writers: use special characters.
1.1 kristaps 374: */
375:
376: c = (unsigned char) blk.buf[i];
377:
1.47 schwarze 378: if ( ! (isascii(c) &&
379: (isgraph(c) || isblank(c)))) {
1.78 schwarze 380: mandoc_vmsg(MANDOCERR_BADCHAR, curp,
381: curp->line, pos, "0x%x", c);
1.1 kristaps 382: i++;
1.27 joerg 383: ln.buf[pos++] = '?';
1.1 kristaps 384: continue;
385: }
386:
387: /* Trailing backslash = a plain char. */
388:
389: if ('\\' != blk.buf[i] || i + 1 == (int)blk.sz) {
390: ln.buf[pos++] = blk.buf[i++];
391: continue;
392: }
393:
394: /*
395: * Found escape and at least one other character.
396: * When it's a newline character, skip it.
397: * When there is a carriage return in between,
398: * skip that one as well.
399: */
400:
401: if ('\r' == blk.buf[i + 1] && i + 2 < (int)blk.sz &&
402: '\n' == blk.buf[i + 2])
403: ++i;
404: if ('\n' == blk.buf[i + 1]) {
405: i += 2;
406: ++lnn;
407: continue;
408: }
409:
1.13 kristaps 410: if ('"' == blk.buf[i + 1] || '#' == blk.buf[i + 1]) {
1.1 kristaps 411: i += 2;
412: /* Comment, skip to end of line */
413: for (; i < (int)blk.sz; ++i) {
414: if ('\n' == blk.buf[i]) {
415: ++i;
416: ++lnn;
417: break;
418: }
419: }
420:
421: /* Backout trailing whitespaces */
422: for (; pos > 0; --pos) {
423: if (ln.buf[pos - 1] != ' ')
424: break;
425: if (pos > 2 && ln.buf[pos - 2] == '\\')
426: break;
427: }
428: break;
429: }
430:
1.35 schwarze 431: /* Catch escaped bogus characters. */
432:
433: c = (unsigned char) blk.buf[i+1];
434:
1.47 schwarze 435: if ( ! (isascii(c) &&
436: (isgraph(c) || isblank(c)))) {
1.78 schwarze 437: mandoc_vmsg(MANDOCERR_BADCHAR, curp,
438: curp->line, pos, "0x%x", c);
1.35 schwarze 439: i += 2;
440: ln.buf[pos++] = '?';
441: continue;
442: }
443:
1.1 kristaps 444: /* Some other escape sequence, copy & cont. */
445:
446: ln.buf[pos++] = blk.buf[i++];
447: ln.buf[pos++] = blk.buf[i++];
448: }
449:
1.47 schwarze 450: if (pos >= (int)ln.sz)
1.1 kristaps 451: resize_buf(&ln, 256);
452:
453: ln.buf[pos] = '\0';
454:
455: /*
456: * A significant amount of complexity is contained by
457: * the roff preprocessor. It's line-oriented but can be
458: * expressed on one line, so we need at times to
459: * readjust our starting point and re-run it. The roff
460: * preprocessor can also readjust the buffers with new
461: * data, so we pass them in wholesale.
462: */
463:
464: of = 0;
465:
1.24 kristaps 466: /*
467: * Maintain a lookaside buffer of all parsed lines. We
468: * only do this if mparse_keep() has been invoked (the
469: * buffer may be accessed with mparse_getkeep()).
470: */
471:
472: if (curp->secondary) {
1.47 schwarze 473: curp->secondary->buf = mandoc_realloc(
474: curp->secondary->buf,
475: curp->secondary->sz + pos + 2);
476: memcpy(curp->secondary->buf +
477: curp->secondary->sz,
478: ln.buf, pos);
1.24 kristaps 479: curp->secondary->sz += pos;
480: curp->secondary->buf
481: [curp->secondary->sz] = '\n';
482: curp->secondary->sz++;
483: curp->secondary->buf
484: [curp->secondary->sz] = '\0';
485: }
1.1 kristaps 486: rerun:
1.47 schwarze 487: rr = roff_parseln(curp->roff, curp->line,
488: &ln.buf, &ln.sz, of, &of);
1.1 kristaps 489:
490: switch (rr) {
1.47 schwarze 491: case ROFF_REPARSE:
1.1 kristaps 492: if (REPARSE_LIMIT >= ++curp->reparse_count)
493: mparse_buf_r(curp, ln, 0);
494: else
1.3 kristaps 495: mandoc_msg(MANDOCERR_ROFFLOOP, curp,
1.47 schwarze 496: curp->line, pos, NULL);
1.1 kristaps 497: pos = 0;
498: continue;
1.47 schwarze 499: case ROFF_APPEND:
1.1 kristaps 500: pos = (int)strlen(ln.buf);
501: continue;
1.47 schwarze 502: case ROFF_RERUN:
1.1 kristaps 503: goto rerun;
1.47 schwarze 504: case ROFF_IGN:
1.1 kristaps 505: pos = 0;
506: continue;
1.47 schwarze 507: case ROFF_ERR:
1.1 kristaps 508: assert(MANDOCLEVEL_FATAL <= curp->file_status);
509: break;
1.47 schwarze 510: case ROFF_SO:
1.45 schwarze 511: if (0 == (MPARSE_SO & curp->options) &&
512: (i >= (int)blk.sz || '\0' == blk.buf[i])) {
513: curp->sodest = mandoc_strdup(ln.buf + of);
514: free(ln.buf);
515: return;
516: }
1.24 kristaps 517: /*
518: * We remove `so' clauses from our lookaside
519: * buffer because we're going to descend into
520: * the file recursively.
521: */
1.47 schwarze 522: if (curp->secondary)
1.25 kristaps 523: curp->secondary->sz -= pos + 1;
1.36 schwarze 524: mparse_readfd(curp, -1, ln.buf + of);
1.52 schwarze 525: if (MANDOCLEVEL_FATAL <= curp->file_status) {
526: mandoc_vmsg(MANDOCERR_SO_FAIL,
527: curp, curp->line, pos,
528: ".so %s", ln.buf + of);
1.1 kristaps 529: break;
1.52 schwarze 530: }
1.1 kristaps 531: pos = 0;
532: continue;
533: default:
534: break;
535: }
536:
537: /*
538: * If we encounter errors in the recursive parse, make
539: * sure we don't continue parsing.
540: */
541:
542: if (MANDOCLEVEL_FATAL <= curp->file_status)
543: break;
544:
545: /*
546: * If input parsers have not been allocated, do so now.
1.14 kristaps 547: * We keep these instanced between parsers, but set them
1.1 kristaps 548: * locally per parse routine since we can use different
549: * parsers with each one.
550: */
551:
552: if ( ! (curp->man || curp->mdoc))
553: pset(ln.buf + of, pos - of, curp);
554:
1.47 schwarze 555: /*
1.1 kristaps 556: * Lastly, push down into the parsers themselves. One
557: * of these will have already been set in the pset()
558: * routine.
559: * If libroff returns ROFF_TBL, then add it to the
560: * currently open parse. Since we only get here if
561: * there does exist data (see tbl_data.c), we're
562: * guaranteed that something's been allocated.
563: * Do the same for ROFF_EQN.
564: */
565:
566: rc = -1;
567:
568: if (ROFF_TBL == rr)
569: while (NULL != (span = roff_span(curp->roff))) {
570: rc = curp->man ?
1.47 schwarze 571: man_addspan(curp->man, span) :
572: mdoc_addspan(curp->mdoc, span);
1.1 kristaps 573: if (0 == rc)
574: break;
575: }
576: else if (ROFF_EQN == rr)
1.47 schwarze 577: rc = curp->mdoc ?
578: mdoc_addeqn(curp->mdoc,
579: roff_eqn(curp->roff)) :
580: man_addeqn(curp->man,
581: roff_eqn(curp->roff));
1.1 kristaps 582: else if (curp->man || curp->mdoc)
583: rc = curp->man ?
1.47 schwarze 584: man_parseln(curp->man,
585: curp->line, ln.buf, of) :
586: mdoc_parseln(curp->mdoc,
587: curp->line, ln.buf, of);
1.1 kristaps 588:
589: if (0 == rc) {
590: assert(MANDOCLEVEL_FATAL <= curp->file_status);
591: break;
1.41 schwarze 592: } else if (2 == rc)
593: break;
1.1 kristaps 594:
595: /* Temporary buffers typically are not full. */
596:
597: if (0 == start && '\0' == blk.buf[i])
598: break;
599:
600: /* Start the next input line. */
601:
602: pos = 0;
603: }
604:
605: free(ln.buf);
606: }
607:
608: static int
1.40 schwarze 609: read_whole_file(struct mparse *curp, const char *file, int fd,
610: struct buf *fb, int *with_mmap)
1.1 kristaps 611: {
612: size_t off;
613: ssize_t ssz;
614:
1.81 schwarze 615: #if HAVE_MMAP
1.15 kristaps 616: struct stat st;
1.1 kristaps 617: if (-1 == fstat(fd, &st)) {
1.40 schwarze 618: curp->file_status = MANDOCLEVEL_SYSERR;
619: if (curp->mmsg)
620: (*curp->mmsg)(MANDOCERR_SYSSTAT, curp->file_status,
621: file, 0, 0, strerror(errno));
1.1 kristaps 622: return(0);
623: }
624:
625: /*
626: * If we're a regular file, try just reading in the whole entry
627: * via mmap(). This is faster than reading it into blocks, and
628: * since each file is only a few bytes to begin with, I'm not
629: * concerned that this is going to tank any machines.
630: */
631:
632: if (S_ISREG(st.st_mode)) {
633: if (st.st_size >= (1U << 31)) {
1.40 schwarze 634: curp->file_status = MANDOCLEVEL_FATAL;
635: if (curp->mmsg)
636: (*curp->mmsg)(MANDOCERR_TOOLARGE,
637: curp->file_status, file, 0, 0, NULL);
1.1 kristaps 638: return(0);
639: }
640: *with_mmap = 1;
641: fb->sz = (size_t)st.st_size;
1.37 schwarze 642: fb->buf = mmap(NULL, fb->sz, PROT_READ, MAP_SHARED, fd, 0);
1.1 kristaps 643: if (fb->buf != MAP_FAILED)
644: return(1);
645: }
1.15 kristaps 646: #endif
1.1 kristaps 647:
648: /*
649: * If this isn't a regular file (like, say, stdin), then we must
650: * go the old way and just read things in bit by bit.
651: */
652:
653: *with_mmap = 0;
654: off = 0;
655: fb->sz = 0;
656: fb->buf = NULL;
657: for (;;) {
658: if (off == fb->sz) {
659: if (fb->sz == (1U << 31)) {
1.40 schwarze 660: curp->file_status = MANDOCLEVEL_FATAL;
661: if (curp->mmsg)
662: (*curp->mmsg)(MANDOCERR_TOOLARGE,
663: curp->file_status,
664: file, 0, 0, NULL);
1.1 kristaps 665: break;
666: }
667: resize_buf(fb, 65536);
668: }
669: ssz = read(fd, fb->buf + (int)off, fb->sz - off);
670: if (ssz == 0) {
671: fb->sz = off;
672: return(1);
673: }
674: if (ssz == -1) {
1.40 schwarze 675: curp->file_status = MANDOCLEVEL_SYSERR;
676: if (curp->mmsg)
677: (*curp->mmsg)(MANDOCERR_SYSREAD,
678: curp->file_status, file, 0, 0,
679: strerror(errno));
1.1 kristaps 680: break;
681: }
682: off += (size_t)ssz;
683: }
684:
685: free(fb->buf);
686: fb->buf = NULL;
687: return(0);
688: }
689:
690: static void
691: mparse_end(struct mparse *curp)
692: {
693:
694: if (MANDOCLEVEL_FATAL <= curp->file_status)
695: return;
696:
1.72 schwarze 697: if (curp->mdoc == NULL &&
698: curp->man == NULL &&
699: curp->sodest == NULL) {
700: if (curp->options & MPARSE_MDOC)
701: curp->mdoc = curp->pmdoc;
702: else {
703: if (curp->pman == NULL)
704: curp->pman = man_alloc(curp->roff, curp,
705: curp->options & MPARSE_QUICK ? 1 : 0);
706: curp->man = curp->pman;
707: }
708: }
709:
1.1 kristaps 710: if (curp->mdoc && ! mdoc_endparse(curp->mdoc)) {
711: assert(MANDOCLEVEL_FATAL <= curp->file_status);
712: return;
713: }
714:
715: if (curp->man && ! man_endparse(curp->man)) {
716: assert(MANDOCLEVEL_FATAL <= curp->file_status);
717: return;
718: }
719:
720: roff_endparse(curp->roff);
721: }
722:
723: static void
1.36 schwarze 724: mparse_parse_buffer(struct mparse *curp, struct buf blk, const char *file)
1.28 joerg 725: {
726: const char *svfile;
1.36 schwarze 727: static int recursion_depth;
728:
729: if (64 < recursion_depth) {
730: mandoc_msg(MANDOCERR_ROFFLOOP, curp, curp->line, 0, NULL);
731: return;
732: }
1.28 joerg 733:
734: /* Line number is per-file. */
735: svfile = curp->file;
736: curp->file = file;
1.83 ! schwarze 737: curp->primary = &blk;
1.28 joerg 738: curp->line = 1;
1.36 schwarze 739: recursion_depth++;
1.28 joerg 740:
741: mparse_buf_r(curp, blk, 1);
742:
1.36 schwarze 743: if (0 == --recursion_depth && MANDOCLEVEL_FATAL > curp->file_status)
1.28 joerg 744: mparse_end(curp);
745:
746: curp->file = svfile;
747: }
748:
749: enum mandoclevel
750: mparse_readmem(struct mparse *curp, const void *buf, size_t len,
751: const char *file)
752: {
753: struct buf blk;
754:
755: blk.buf = UNCONST(buf);
756: blk.sz = len;
757:
1.36 schwarze 758: mparse_parse_buffer(curp, blk, file);
1.28 joerg 759: return(curp->file_status);
760: }
761:
1.36 schwarze 762: enum mandoclevel
763: mparse_readfd(struct mparse *curp, int fd, const char *file)
1.1 kristaps 764: {
1.28 joerg 765: struct buf blk;
766: int with_mmap;
1.1 kristaps 767:
1.40 schwarze 768: if (-1 == fd && -1 == (fd = open(file, O_RDONLY, 0))) {
769: curp->file_status = MANDOCLEVEL_SYSERR;
770: if (curp->mmsg)
771: (*curp->mmsg)(MANDOCERR_SYSOPEN,
772: curp->file_status,
773: file, 0, 0, strerror(errno));
774: goto out;
775: }
776:
1.28 joerg 777: /*
778: * Run for each opened file; may be called more than once for
779: * each full parse sequence if the opened file is nested (i.e.,
780: * from `so'). Simply sucks in the whole file and moves into
781: * the parse phase for the file.
782: */
1.1 kristaps 783:
1.40 schwarze 784: if ( ! read_whole_file(curp, file, fd, &blk, &with_mmap))
1.36 schwarze 785: goto out;
1.1 kristaps 786:
1.36 schwarze 787: mparse_parse_buffer(curp, blk, file);
1.1 kristaps 788:
1.81 schwarze 789: #if HAVE_MMAP
1.28 joerg 790: if (with_mmap)
791: munmap(blk.buf, blk.sz);
792: else
793: #endif
794: free(blk.buf);
1.1 kristaps 795:
796: if (STDIN_FILENO != fd && -1 == close(fd))
797: perror(file);
1.36 schwarze 798: out:
1.1 kristaps 799: return(curp->file_status);
1.82 schwarze 800: }
801:
802: enum mandoclevel
803: mparse_open(struct mparse *curp, int *fd, const char *file,
804: pid_t *child_pid)
805: {
806: int pfd[2];
807: char *cp;
808: enum mandocerr err;
809:
810: pfd[1] = -1;
811: curp->file = file;
812: if ((cp = strrchr(file, '.')) == NULL ||
813: strcmp(cp + 1, "gz")) {
814: *child_pid = 0;
815: if ((*fd = open(file, O_RDONLY)) == -1) {
816: err = MANDOCERR_SYSOPEN;
817: goto out;
818: }
819: return(MANDOCLEVEL_OK);
820: }
821:
822: if (pipe(pfd) == -1) {
823: err = MANDOCERR_SYSPIPE;
824: goto out;
825: }
826:
827: switch (*child_pid = fork()) {
828: case -1:
829: err = MANDOCERR_SYSFORK;
830: close(pfd[0]);
831: close(pfd[1]);
832: pfd[1] = -1;
833: break;
834: case 0:
835: close(pfd[0]);
836: if (dup2(pfd[1], STDOUT_FILENO) == -1) {
837: err = MANDOCERR_SYSDUP;
838: break;
839: }
840: execlp("gunzip", "gunzip", "-c", file, NULL);
841: err = MANDOCERR_SYSEXEC;
842: break;
843: default:
844: close(pfd[1]);
845: *fd = pfd[0];
846: return(MANDOCLEVEL_OK);
847: }
848:
849: out:
850: *fd = -1;
851: *child_pid = 0;
852: curp->file_status = MANDOCLEVEL_SYSERR;
853: if (curp->mmsg)
854: (*curp->mmsg)(err, curp->file_status, file,
855: 0, 0, strerror(errno));
856: if (pfd[1] != -1)
857: exit(1);
858: return(curp->file_status);
859: }
860:
861: enum mandoclevel
862: mparse_wait(struct mparse *curp, pid_t child_pid)
863: {
864: int status;
865:
866: if (waitpid(child_pid, &status, 0) == -1) {
867: mandoc_msg(MANDOCERR_SYSWAIT, curp, 0, 0,
868: strerror(errno));
869: curp->file_status = MANDOCLEVEL_SYSERR;
870: return(curp->file_status);
871: }
872: if (WIFSIGNALED(status)) {
873: mandoc_vmsg(MANDOCERR_SYSSIG, curp, 0, 0,
874: "%d", WTERMSIG(status));
875: curp->file_status = MANDOCLEVEL_SYSERR;
876: return(curp->file_status);
877: }
878: if (WEXITSTATUS(status)) {
879: mandoc_vmsg(MANDOCERR_SYSEXIT, curp, 0, 0,
880: "%d", WEXITSTATUS(status));
881: curp->file_status = MANDOCLEVEL_SYSERR;
882: return(curp->file_status);
883: }
884: return(MANDOCLEVEL_OK);
1.1 kristaps 885: }
886:
887: struct mparse *
1.44 schwarze 888: mparse_alloc(int options, enum mandoclevel wlevel,
1.69 schwarze 889: mandocmsg mmsg, const char *defos)
1.1 kristaps 890: {
891: struct mparse *curp;
1.10 kristaps 892:
893: assert(wlevel <= MANDOCLEVEL_FATAL);
1.1 kristaps 894:
895: curp = mandoc_calloc(1, sizeof(struct mparse));
896:
1.44 schwarze 897: curp->options = options;
1.3 kristaps 898: curp->wlevel = wlevel;
1.1 kristaps 899: curp->mmsg = mmsg;
1.29 schwarze 900: curp->defos = defos;
1.1 kristaps 901:
1.44 schwarze 902: curp->roff = roff_alloc(curp, options);
1.72 schwarze 903: if (curp->options & MPARSE_MDOC)
904: curp->pmdoc = mdoc_alloc(
905: curp->roff, curp, curp->defos,
906: curp->options & MPARSE_QUICK ? 1 : 0);
907: if (curp->options & MPARSE_MAN)
908: curp->pman = man_alloc(curp->roff, curp,
909: curp->options & MPARSE_QUICK ? 1 : 0);
910:
1.1 kristaps 911: return(curp);
912: }
913:
914: void
915: mparse_reset(struct mparse *curp)
916: {
917:
918: roff_reset(curp->roff);
919:
920: if (curp->mdoc)
921: mdoc_reset(curp->mdoc);
922: if (curp->man)
923: man_reset(curp->man);
1.24 kristaps 924: if (curp->secondary)
925: curp->secondary->sz = 0;
1.1 kristaps 926:
927: curp->file_status = MANDOCLEVEL_OK;
928: curp->mdoc = NULL;
929: curp->man = NULL;
1.45 schwarze 930:
931: free(curp->sodest);
932: curp->sodest = NULL;
1.1 kristaps 933: }
934:
935: void
936: mparse_free(struct mparse *curp)
937: {
938:
939: if (curp->pmdoc)
940: mdoc_free(curp->pmdoc);
941: if (curp->pman)
942: man_free(curp->pman);
943: if (curp->roff)
944: roff_free(curp->roff);
1.24 kristaps 945: if (curp->secondary)
946: free(curp->secondary->buf);
1.1 kristaps 947:
1.24 kristaps 948: free(curp->secondary);
1.45 schwarze 949: free(curp->sodest);
1.1 kristaps 950: free(curp);
951: }
952:
953: void
1.45 schwarze 954: mparse_result(struct mparse *curp,
955: struct mdoc **mdoc, struct man **man, char **sodest)
1.1 kristaps 956: {
957:
1.45 schwarze 958: if (sodest && NULL != (*sodest = curp->sodest)) {
959: *mdoc = NULL;
960: *man = NULL;
961: return;
962: }
1.9 kristaps 963: if (mdoc)
964: *mdoc = curp->mdoc;
965: if (man)
966: *man = curp->man;
1.3 kristaps 967: }
968:
969: void
970: mandoc_vmsg(enum mandocerr t, struct mparse *m,
971: int ln, int pos, const char *fmt, ...)
972: {
973: char buf[256];
974: va_list ap;
975:
976: va_start(ap, fmt);
1.48 schwarze 977: (void)vsnprintf(buf, sizeof(buf), fmt, ap);
1.3 kristaps 978: va_end(ap);
979:
980: mandoc_msg(t, m, ln, pos, buf);
981: }
982:
983: void
1.47 schwarze 984: mandoc_msg(enum mandocerr er, struct mparse *m,
1.3 kristaps 985: int ln, int col, const char *msg)
986: {
987: enum mandoclevel level;
988:
989: level = MANDOCLEVEL_FATAL;
990: while (er < mandoclimits[level])
991: level--;
992:
993: if (level < m->wlevel)
994: return;
995:
1.8 kristaps 996: if (m->mmsg)
997: (*m->mmsg)(er, level, m->file, ln, col, msg);
1.3 kristaps 998:
999: if (m->file_status < level)
1000: m->file_status = level;
1.7 kristaps 1001: }
1002:
1003: const char *
1004: mparse_strerror(enum mandocerr er)
1005: {
1006:
1007: return(mandocerrs[er]);
1008: }
1009:
1010: const char *
1011: mparse_strlevel(enum mandoclevel lvl)
1012: {
1013: return(mandoclevels[lvl]);
1.24 kristaps 1014: }
1015:
1016: void
1017: mparse_keep(struct mparse *p)
1018: {
1019:
1020: assert(NULL == p->secondary);
1021: p->secondary = mandoc_calloc(1, sizeof(struct buf));
1022: }
1023:
1024: const char *
1025: mparse_getkeep(const struct mparse *p)
1026: {
1027:
1028: assert(p->secondary);
1029: return(p->secondary->sz ? p->secondary->buf : NULL);
1.1 kristaps 1030: }
CVSweb