Annotation of mandoc/read.c, Revision 1.84
1.84 ! schwarze 1: /* $Id: read.c,v 1.83 2014/09/06 22:39:36 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:
1.84 ! schwarze 72: static void choose_parser(struct mparse *);
1.1 kristaps 73: static void resize_buf(struct buf *, size_t);
74: static void mparse_buf_r(struct mparse *, struct buf, int);
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
1.84 ! schwarze 250: choose_parser(struct mparse *curp)
1.1 kristaps 251: {
1.83 schwarze 252: char *cp, *ep;
253: int format;
1.1 kristaps 254:
1.83 schwarze 255: /*
256: * If neither command line arguments -mdoc or -man select
257: * a parser nor the roff parser found a .Dd or .TH macro
258: * yet, look ahead in the main input buffer.
259: */
260:
261: if ((format = roff_getformat(curp->roff)) == 0) {
262: cp = curp->primary->buf;
263: ep = cp + curp->primary->sz;
264: while (cp < ep) {
265: if (*cp == '.' || *cp != '\'') {
266: cp++;
267: if (cp[0] == 'D' && cp[1] == 'd') {
268: format = MPARSE_MDOC;
269: break;
270: }
271: if (cp[0] == 'T' && cp[1] == 'H') {
272: format = MPARSE_MAN;
273: break;
274: }
275: }
276: cp = memchr(cp, '\n', ep - cp);
277: if (cp == NULL)
278: break;
279: cp++;
280: }
1.1 kristaps 281: }
282:
1.83 schwarze 283: if (format == MPARSE_MDOC) {
1.47 schwarze 284: if (NULL == curp->pmdoc)
1.44 schwarze 285: curp->pmdoc = mdoc_alloc(
286: curp->roff, curp, curp->defos,
287: MPARSE_QUICK & curp->options ? 1 : 0);
1.1 kristaps 288: assert(curp->pmdoc);
289: curp->mdoc = curp->pmdoc;
290: return;
1.47 schwarze 291: }
1.1 kristaps 292:
1.83 schwarze 293: /* Fall back to man(7) as a last resort. */
294:
1.47 schwarze 295: if (NULL == curp->pman)
1.44 schwarze 296: curp->pman = man_alloc(curp->roff, curp,
297: MPARSE_QUICK & curp->options ? 1 : 0);
1.1 kristaps 298: assert(curp->pman);
299: curp->man = curp->pman;
300: }
301:
302: /*
303: * Main parse routine for an opened file. This is called for each
304: * opened file and simply loops around the full input file, possibly
305: * nesting (i.e., with `so').
306: */
307: static void
308: mparse_buf_r(struct mparse *curp, struct buf blk, int start)
309: {
310: const struct tbl_span *span;
311: struct buf ln;
312: enum rofferr rr;
313: int i, of, rc;
314: int pos; /* byte number in the ln buffer */
315: int lnn; /* line number in the real file */
316: unsigned char c;
317:
318: memset(&ln, 0, sizeof(struct buf));
319:
1.47 schwarze 320: lnn = curp->line;
321: pos = 0;
1.1 kristaps 322:
323: for (i = 0; i < (int)blk.sz; ) {
324: if (0 == pos && '\0' == blk.buf[i])
325: break;
326:
327: if (start) {
328: curp->line = lnn;
329: curp->reparse_count = 0;
330: }
331:
332: while (i < (int)blk.sz && (start || '\0' != blk.buf[i])) {
333:
334: /*
335: * When finding an unescaped newline character,
336: * leave the character loop to process the line.
337: * Skip a preceding carriage return, if any.
338: */
339:
340: if ('\r' == blk.buf[i] && i + 1 < (int)blk.sz &&
341: '\n' == blk.buf[i + 1])
342: ++i;
343: if ('\n' == blk.buf[i]) {
344: ++i;
345: ++lnn;
346: break;
347: }
348:
1.35 schwarze 349: /*
350: * Make sure we have space for at least
351: * one backslash and one other character
352: * and the trailing NUL byte.
353: */
354:
355: if (pos + 2 >= (int)ln.sz)
356: resize_buf(&ln, 256);
357:
1.47 schwarze 358: /*
1.1 kristaps 359: * Warn about bogus characters. If you're using
360: * non-ASCII encoding, you're screwing your
361: * readers. Since I'd rather this not happen,
1.27 joerg 362: * I'll be helpful and replace these characters
363: * with "?", so we don't display gibberish.
364: * Note to manual writers: use special characters.
1.1 kristaps 365: */
366:
367: c = (unsigned char) blk.buf[i];
368:
1.47 schwarze 369: if ( ! (isascii(c) &&
370: (isgraph(c) || isblank(c)))) {
1.78 schwarze 371: mandoc_vmsg(MANDOCERR_BADCHAR, curp,
372: curp->line, pos, "0x%x", c);
1.1 kristaps 373: i++;
1.27 joerg 374: ln.buf[pos++] = '?';
1.1 kristaps 375: continue;
376: }
377:
378: /* Trailing backslash = a plain char. */
379:
380: if ('\\' != blk.buf[i] || i + 1 == (int)blk.sz) {
381: ln.buf[pos++] = blk.buf[i++];
382: continue;
383: }
384:
385: /*
386: * Found escape and at least one other character.
387: * When it's a newline character, skip it.
388: * When there is a carriage return in between,
389: * skip that one as well.
390: */
391:
392: if ('\r' == blk.buf[i + 1] && i + 2 < (int)blk.sz &&
393: '\n' == blk.buf[i + 2])
394: ++i;
395: if ('\n' == blk.buf[i + 1]) {
396: i += 2;
397: ++lnn;
398: continue;
399: }
400:
1.13 kristaps 401: if ('"' == blk.buf[i + 1] || '#' == blk.buf[i + 1]) {
1.1 kristaps 402: i += 2;
403: /* Comment, skip to end of line */
404: for (; i < (int)blk.sz; ++i) {
405: if ('\n' == blk.buf[i]) {
406: ++i;
407: ++lnn;
408: break;
409: }
410: }
411:
412: /* Backout trailing whitespaces */
413: for (; pos > 0; --pos) {
414: if (ln.buf[pos - 1] != ' ')
415: break;
416: if (pos > 2 && ln.buf[pos - 2] == '\\')
417: break;
418: }
419: break;
420: }
421:
1.35 schwarze 422: /* Catch escaped bogus characters. */
423:
424: c = (unsigned char) blk.buf[i+1];
425:
1.47 schwarze 426: if ( ! (isascii(c) &&
427: (isgraph(c) || isblank(c)))) {
1.78 schwarze 428: mandoc_vmsg(MANDOCERR_BADCHAR, curp,
429: curp->line, pos, "0x%x", c);
1.35 schwarze 430: i += 2;
431: ln.buf[pos++] = '?';
432: continue;
433: }
434:
1.1 kristaps 435: /* Some other escape sequence, copy & cont. */
436:
437: ln.buf[pos++] = blk.buf[i++];
438: ln.buf[pos++] = blk.buf[i++];
439: }
440:
1.47 schwarze 441: if (pos >= (int)ln.sz)
1.1 kristaps 442: resize_buf(&ln, 256);
443:
444: ln.buf[pos] = '\0';
445:
446: /*
447: * A significant amount of complexity is contained by
448: * the roff preprocessor. It's line-oriented but can be
449: * expressed on one line, so we need at times to
450: * readjust our starting point and re-run it. The roff
451: * preprocessor can also readjust the buffers with new
452: * data, so we pass them in wholesale.
453: */
454:
455: of = 0;
456:
1.24 kristaps 457: /*
458: * Maintain a lookaside buffer of all parsed lines. We
459: * only do this if mparse_keep() has been invoked (the
460: * buffer may be accessed with mparse_getkeep()).
461: */
462:
463: if (curp->secondary) {
1.47 schwarze 464: curp->secondary->buf = mandoc_realloc(
465: curp->secondary->buf,
466: curp->secondary->sz + pos + 2);
467: memcpy(curp->secondary->buf +
468: curp->secondary->sz,
469: ln.buf, pos);
1.24 kristaps 470: curp->secondary->sz += pos;
471: curp->secondary->buf
472: [curp->secondary->sz] = '\n';
473: curp->secondary->sz++;
474: curp->secondary->buf
475: [curp->secondary->sz] = '\0';
476: }
1.1 kristaps 477: rerun:
1.47 schwarze 478: rr = roff_parseln(curp->roff, curp->line,
479: &ln.buf, &ln.sz, of, &of);
1.1 kristaps 480:
481: switch (rr) {
1.47 schwarze 482: case ROFF_REPARSE:
1.1 kristaps 483: if (REPARSE_LIMIT >= ++curp->reparse_count)
484: mparse_buf_r(curp, ln, 0);
485: else
1.3 kristaps 486: mandoc_msg(MANDOCERR_ROFFLOOP, curp,
1.47 schwarze 487: curp->line, pos, NULL);
1.1 kristaps 488: pos = 0;
489: continue;
1.47 schwarze 490: case ROFF_APPEND:
1.1 kristaps 491: pos = (int)strlen(ln.buf);
492: continue;
1.47 schwarze 493: case ROFF_RERUN:
1.1 kristaps 494: goto rerun;
1.47 schwarze 495: case ROFF_IGN:
1.1 kristaps 496: pos = 0;
497: continue;
1.47 schwarze 498: case ROFF_ERR:
1.1 kristaps 499: assert(MANDOCLEVEL_FATAL <= curp->file_status);
500: break;
1.47 schwarze 501: case ROFF_SO:
1.45 schwarze 502: if (0 == (MPARSE_SO & curp->options) &&
503: (i >= (int)blk.sz || '\0' == blk.buf[i])) {
504: curp->sodest = mandoc_strdup(ln.buf + of);
505: free(ln.buf);
506: return;
507: }
1.24 kristaps 508: /*
509: * We remove `so' clauses from our lookaside
510: * buffer because we're going to descend into
511: * the file recursively.
512: */
1.47 schwarze 513: if (curp->secondary)
1.25 kristaps 514: curp->secondary->sz -= pos + 1;
1.36 schwarze 515: mparse_readfd(curp, -1, ln.buf + of);
1.52 schwarze 516: if (MANDOCLEVEL_FATAL <= curp->file_status) {
517: mandoc_vmsg(MANDOCERR_SO_FAIL,
518: curp, curp->line, pos,
519: ".so %s", ln.buf + of);
1.1 kristaps 520: break;
1.52 schwarze 521: }
1.1 kristaps 522: pos = 0;
523: continue;
524: default:
525: break;
526: }
527:
528: /*
529: * If we encounter errors in the recursive parse, make
530: * sure we don't continue parsing.
531: */
532:
533: if (MANDOCLEVEL_FATAL <= curp->file_status)
534: break;
535:
536: /*
537: * If input parsers have not been allocated, do so now.
1.14 kristaps 538: * We keep these instanced between parsers, but set them
1.1 kristaps 539: * locally per parse routine since we can use different
540: * parsers with each one.
541: */
542:
543: if ( ! (curp->man || curp->mdoc))
1.84 ! schwarze 544: choose_parser(curp);
1.1 kristaps 545:
1.47 schwarze 546: /*
1.84 ! schwarze 547: * Lastly, push down into the parsers themselves.
1.1 kristaps 548: * If libroff returns ROFF_TBL, then add it to the
549: * currently open parse. Since we only get here if
550: * there does exist data (see tbl_data.c), we're
551: * guaranteed that something's been allocated.
552: * Do the same for ROFF_EQN.
553: */
554:
555: rc = -1;
556:
557: if (ROFF_TBL == rr)
558: while (NULL != (span = roff_span(curp->roff))) {
559: rc = curp->man ?
1.47 schwarze 560: man_addspan(curp->man, span) :
561: mdoc_addspan(curp->mdoc, span);
1.1 kristaps 562: if (0 == rc)
563: break;
564: }
565: else if (ROFF_EQN == rr)
1.47 schwarze 566: rc = curp->mdoc ?
567: mdoc_addeqn(curp->mdoc,
568: roff_eqn(curp->roff)) :
569: man_addeqn(curp->man,
570: roff_eqn(curp->roff));
1.1 kristaps 571: else if (curp->man || curp->mdoc)
572: rc = curp->man ?
1.47 schwarze 573: man_parseln(curp->man,
574: curp->line, ln.buf, of) :
575: mdoc_parseln(curp->mdoc,
576: curp->line, ln.buf, of);
1.1 kristaps 577:
578: if (0 == rc) {
579: assert(MANDOCLEVEL_FATAL <= curp->file_status);
580: break;
1.41 schwarze 581: } else if (2 == rc)
582: break;
1.1 kristaps 583:
584: /* Temporary buffers typically are not full. */
585:
586: if (0 == start && '\0' == blk.buf[i])
587: break;
588:
589: /* Start the next input line. */
590:
591: pos = 0;
592: }
593:
594: free(ln.buf);
595: }
596:
597: static int
1.40 schwarze 598: read_whole_file(struct mparse *curp, const char *file, int fd,
599: struct buf *fb, int *with_mmap)
1.1 kristaps 600: {
601: size_t off;
602: ssize_t ssz;
603:
1.81 schwarze 604: #if HAVE_MMAP
1.15 kristaps 605: struct stat st;
1.1 kristaps 606: if (-1 == fstat(fd, &st)) {
1.40 schwarze 607: curp->file_status = MANDOCLEVEL_SYSERR;
608: if (curp->mmsg)
609: (*curp->mmsg)(MANDOCERR_SYSSTAT, curp->file_status,
610: file, 0, 0, strerror(errno));
1.1 kristaps 611: return(0);
612: }
613:
614: /*
615: * If we're a regular file, try just reading in the whole entry
616: * via mmap(). This is faster than reading it into blocks, and
617: * since each file is only a few bytes to begin with, I'm not
618: * concerned that this is going to tank any machines.
619: */
620:
621: if (S_ISREG(st.st_mode)) {
622: if (st.st_size >= (1U << 31)) {
1.40 schwarze 623: curp->file_status = MANDOCLEVEL_FATAL;
624: if (curp->mmsg)
625: (*curp->mmsg)(MANDOCERR_TOOLARGE,
626: curp->file_status, file, 0, 0, NULL);
1.1 kristaps 627: return(0);
628: }
629: *with_mmap = 1;
630: fb->sz = (size_t)st.st_size;
1.37 schwarze 631: fb->buf = mmap(NULL, fb->sz, PROT_READ, MAP_SHARED, fd, 0);
1.1 kristaps 632: if (fb->buf != MAP_FAILED)
633: return(1);
634: }
1.15 kristaps 635: #endif
1.1 kristaps 636:
637: /*
638: * If this isn't a regular file (like, say, stdin), then we must
639: * go the old way and just read things in bit by bit.
640: */
641:
642: *with_mmap = 0;
643: off = 0;
644: fb->sz = 0;
645: fb->buf = NULL;
646: for (;;) {
647: if (off == fb->sz) {
648: if (fb->sz == (1U << 31)) {
1.40 schwarze 649: curp->file_status = MANDOCLEVEL_FATAL;
650: if (curp->mmsg)
651: (*curp->mmsg)(MANDOCERR_TOOLARGE,
652: curp->file_status,
653: file, 0, 0, NULL);
1.1 kristaps 654: break;
655: }
656: resize_buf(fb, 65536);
657: }
658: ssz = read(fd, fb->buf + (int)off, fb->sz - off);
659: if (ssz == 0) {
660: fb->sz = off;
661: return(1);
662: }
663: if (ssz == -1) {
1.40 schwarze 664: curp->file_status = MANDOCLEVEL_SYSERR;
665: if (curp->mmsg)
666: (*curp->mmsg)(MANDOCERR_SYSREAD,
667: curp->file_status, file, 0, 0,
668: strerror(errno));
1.1 kristaps 669: break;
670: }
671: off += (size_t)ssz;
672: }
673:
674: free(fb->buf);
675: fb->buf = NULL;
676: return(0);
677: }
678:
679: static void
680: mparse_end(struct mparse *curp)
681: {
682:
683: if (MANDOCLEVEL_FATAL <= curp->file_status)
684: return;
685:
1.72 schwarze 686: if (curp->mdoc == NULL &&
687: curp->man == NULL &&
688: curp->sodest == NULL) {
689: if (curp->options & MPARSE_MDOC)
690: curp->mdoc = curp->pmdoc;
691: else {
692: if (curp->pman == NULL)
693: curp->pman = man_alloc(curp->roff, curp,
694: curp->options & MPARSE_QUICK ? 1 : 0);
695: curp->man = curp->pman;
696: }
697: }
698:
1.1 kristaps 699: if (curp->mdoc && ! mdoc_endparse(curp->mdoc)) {
700: assert(MANDOCLEVEL_FATAL <= curp->file_status);
701: return;
702: }
703:
704: if (curp->man && ! man_endparse(curp->man)) {
705: assert(MANDOCLEVEL_FATAL <= curp->file_status);
706: return;
707: }
708:
709: roff_endparse(curp->roff);
710: }
711:
712: static void
1.36 schwarze 713: mparse_parse_buffer(struct mparse *curp, struct buf blk, const char *file)
1.28 joerg 714: {
715: const char *svfile;
1.36 schwarze 716: static int recursion_depth;
717:
718: if (64 < recursion_depth) {
719: mandoc_msg(MANDOCERR_ROFFLOOP, curp, curp->line, 0, NULL);
720: return;
721: }
1.28 joerg 722:
723: /* Line number is per-file. */
724: svfile = curp->file;
725: curp->file = file;
1.83 schwarze 726: curp->primary = &blk;
1.28 joerg 727: curp->line = 1;
1.36 schwarze 728: recursion_depth++;
1.28 joerg 729:
730: mparse_buf_r(curp, blk, 1);
731:
1.36 schwarze 732: if (0 == --recursion_depth && MANDOCLEVEL_FATAL > curp->file_status)
1.28 joerg 733: mparse_end(curp);
734:
735: curp->file = svfile;
736: }
737:
738: enum mandoclevel
739: mparse_readmem(struct mparse *curp, const void *buf, size_t len,
740: const char *file)
741: {
742: struct buf blk;
743:
744: blk.buf = UNCONST(buf);
745: blk.sz = len;
746:
1.36 schwarze 747: mparse_parse_buffer(curp, blk, file);
1.28 joerg 748: return(curp->file_status);
749: }
750:
1.36 schwarze 751: enum mandoclevel
752: mparse_readfd(struct mparse *curp, int fd, const char *file)
1.1 kristaps 753: {
1.28 joerg 754: struct buf blk;
755: int with_mmap;
1.1 kristaps 756:
1.40 schwarze 757: if (-1 == fd && -1 == (fd = open(file, O_RDONLY, 0))) {
758: curp->file_status = MANDOCLEVEL_SYSERR;
759: if (curp->mmsg)
760: (*curp->mmsg)(MANDOCERR_SYSOPEN,
761: curp->file_status,
762: file, 0, 0, strerror(errno));
763: goto out;
764: }
765:
1.28 joerg 766: /*
767: * Run for each opened file; may be called more than once for
768: * each full parse sequence if the opened file is nested (i.e.,
769: * from `so'). Simply sucks in the whole file and moves into
770: * the parse phase for the file.
771: */
1.1 kristaps 772:
1.40 schwarze 773: if ( ! read_whole_file(curp, file, fd, &blk, &with_mmap))
1.36 schwarze 774: goto out;
1.1 kristaps 775:
1.36 schwarze 776: mparse_parse_buffer(curp, blk, file);
1.1 kristaps 777:
1.81 schwarze 778: #if HAVE_MMAP
1.28 joerg 779: if (with_mmap)
780: munmap(blk.buf, blk.sz);
781: else
782: #endif
783: free(blk.buf);
1.1 kristaps 784:
785: if (STDIN_FILENO != fd && -1 == close(fd))
786: perror(file);
1.36 schwarze 787: out:
1.1 kristaps 788: return(curp->file_status);
1.82 schwarze 789: }
790:
791: enum mandoclevel
792: mparse_open(struct mparse *curp, int *fd, const char *file,
793: pid_t *child_pid)
794: {
795: int pfd[2];
796: char *cp;
797: enum mandocerr err;
798:
799: pfd[1] = -1;
800: curp->file = file;
801: if ((cp = strrchr(file, '.')) == NULL ||
802: strcmp(cp + 1, "gz")) {
803: *child_pid = 0;
804: if ((*fd = open(file, O_RDONLY)) == -1) {
805: err = MANDOCERR_SYSOPEN;
806: goto out;
807: }
808: return(MANDOCLEVEL_OK);
809: }
810:
811: if (pipe(pfd) == -1) {
812: err = MANDOCERR_SYSPIPE;
813: goto out;
814: }
815:
816: switch (*child_pid = fork()) {
817: case -1:
818: err = MANDOCERR_SYSFORK;
819: close(pfd[0]);
820: close(pfd[1]);
821: pfd[1] = -1;
822: break;
823: case 0:
824: close(pfd[0]);
825: if (dup2(pfd[1], STDOUT_FILENO) == -1) {
826: err = MANDOCERR_SYSDUP;
827: break;
828: }
829: execlp("gunzip", "gunzip", "-c", file, NULL);
830: err = MANDOCERR_SYSEXEC;
831: break;
832: default:
833: close(pfd[1]);
834: *fd = pfd[0];
835: return(MANDOCLEVEL_OK);
836: }
837:
838: out:
839: *fd = -1;
840: *child_pid = 0;
841: curp->file_status = MANDOCLEVEL_SYSERR;
842: if (curp->mmsg)
843: (*curp->mmsg)(err, curp->file_status, file,
844: 0, 0, strerror(errno));
845: if (pfd[1] != -1)
846: exit(1);
847: return(curp->file_status);
848: }
849:
850: enum mandoclevel
851: mparse_wait(struct mparse *curp, pid_t child_pid)
852: {
853: int status;
854:
855: if (waitpid(child_pid, &status, 0) == -1) {
856: mandoc_msg(MANDOCERR_SYSWAIT, curp, 0, 0,
857: strerror(errno));
858: curp->file_status = MANDOCLEVEL_SYSERR;
859: return(curp->file_status);
860: }
861: if (WIFSIGNALED(status)) {
862: mandoc_vmsg(MANDOCERR_SYSSIG, curp, 0, 0,
863: "%d", WTERMSIG(status));
864: curp->file_status = MANDOCLEVEL_SYSERR;
865: return(curp->file_status);
866: }
867: if (WEXITSTATUS(status)) {
868: mandoc_vmsg(MANDOCERR_SYSEXIT, curp, 0, 0,
869: "%d", WEXITSTATUS(status));
870: curp->file_status = MANDOCLEVEL_SYSERR;
871: return(curp->file_status);
872: }
873: return(MANDOCLEVEL_OK);
1.1 kristaps 874: }
875:
876: struct mparse *
1.44 schwarze 877: mparse_alloc(int options, enum mandoclevel wlevel,
1.69 schwarze 878: mandocmsg mmsg, const char *defos)
1.1 kristaps 879: {
880: struct mparse *curp;
1.10 kristaps 881:
882: assert(wlevel <= MANDOCLEVEL_FATAL);
1.1 kristaps 883:
884: curp = mandoc_calloc(1, sizeof(struct mparse));
885:
1.44 schwarze 886: curp->options = options;
1.3 kristaps 887: curp->wlevel = wlevel;
1.1 kristaps 888: curp->mmsg = mmsg;
1.29 schwarze 889: curp->defos = defos;
1.1 kristaps 890:
1.44 schwarze 891: curp->roff = roff_alloc(curp, options);
1.72 schwarze 892: if (curp->options & MPARSE_MDOC)
893: curp->pmdoc = mdoc_alloc(
894: curp->roff, curp, curp->defos,
895: curp->options & MPARSE_QUICK ? 1 : 0);
896: if (curp->options & MPARSE_MAN)
897: curp->pman = man_alloc(curp->roff, curp,
898: curp->options & MPARSE_QUICK ? 1 : 0);
899:
1.1 kristaps 900: return(curp);
901: }
902:
903: void
904: mparse_reset(struct mparse *curp)
905: {
906:
907: roff_reset(curp->roff);
908:
909: if (curp->mdoc)
910: mdoc_reset(curp->mdoc);
911: if (curp->man)
912: man_reset(curp->man);
1.24 kristaps 913: if (curp->secondary)
914: curp->secondary->sz = 0;
1.1 kristaps 915:
916: curp->file_status = MANDOCLEVEL_OK;
917: curp->mdoc = NULL;
918: curp->man = NULL;
1.45 schwarze 919:
920: free(curp->sodest);
921: curp->sodest = NULL;
1.1 kristaps 922: }
923:
924: void
925: mparse_free(struct mparse *curp)
926: {
927:
928: if (curp->pmdoc)
929: mdoc_free(curp->pmdoc);
930: if (curp->pman)
931: man_free(curp->pman);
932: if (curp->roff)
933: roff_free(curp->roff);
1.24 kristaps 934: if (curp->secondary)
935: free(curp->secondary->buf);
1.1 kristaps 936:
1.24 kristaps 937: free(curp->secondary);
1.45 schwarze 938: free(curp->sodest);
1.1 kristaps 939: free(curp);
940: }
941:
942: void
1.45 schwarze 943: mparse_result(struct mparse *curp,
944: struct mdoc **mdoc, struct man **man, char **sodest)
1.1 kristaps 945: {
946:
1.45 schwarze 947: if (sodest && NULL != (*sodest = curp->sodest)) {
948: *mdoc = NULL;
949: *man = NULL;
950: return;
951: }
1.9 kristaps 952: if (mdoc)
953: *mdoc = curp->mdoc;
954: if (man)
955: *man = curp->man;
1.3 kristaps 956: }
957:
958: void
959: mandoc_vmsg(enum mandocerr t, struct mparse *m,
960: int ln, int pos, const char *fmt, ...)
961: {
962: char buf[256];
963: va_list ap;
964:
965: va_start(ap, fmt);
1.48 schwarze 966: (void)vsnprintf(buf, sizeof(buf), fmt, ap);
1.3 kristaps 967: va_end(ap);
968:
969: mandoc_msg(t, m, ln, pos, buf);
970: }
971:
972: void
1.47 schwarze 973: mandoc_msg(enum mandocerr er, struct mparse *m,
1.3 kristaps 974: int ln, int col, const char *msg)
975: {
976: enum mandoclevel level;
977:
978: level = MANDOCLEVEL_FATAL;
979: while (er < mandoclimits[level])
980: level--;
981:
982: if (level < m->wlevel)
983: return;
984:
1.8 kristaps 985: if (m->mmsg)
986: (*m->mmsg)(er, level, m->file, ln, col, msg);
1.3 kristaps 987:
988: if (m->file_status < level)
989: m->file_status = level;
1.7 kristaps 990: }
991:
992: const char *
993: mparse_strerror(enum mandocerr er)
994: {
995:
996: return(mandocerrs[er]);
997: }
998:
999: const char *
1000: mparse_strlevel(enum mandoclevel lvl)
1001: {
1002: return(mandoclevels[lvl]);
1.24 kristaps 1003: }
1004:
1005: void
1006: mparse_keep(struct mparse *p)
1007: {
1008:
1009: assert(NULL == p->secondary);
1010: p->secondary = mandoc_calloc(1, sizeof(struct buf));
1011: }
1012:
1013: const char *
1014: mparse_getkeep(const struct mparse *p)
1015: {
1016:
1017: assert(p->secondary);
1018: return(p->secondary->sz ? p->secondary->buf : NULL);
1.1 kristaps 1019: }
CVSweb