Annotation of src/usr.bin/mandoc/read.c, Revision 1.80
1.80 ! schwarze 1: /* $OpenBSD: read.c,v 1.79 2014/11/30 05:28:00 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 mparse {
43: struct man *pman; /* persistent man parser */
44: struct mdoc *pmdoc; /* persistent mdoc parser */
45: struct man *man; /* man parser */
46: struct mdoc *mdoc; /* mdoc parser */
47: struct roff *roff; /* roff parser (!NULL) */
1.69 schwarze 48: const struct mchars *mchars; /* character table */
1.23 schwarze 49: char *sodest; /* filename pointed to by .so */
1.59 schwarze 50: const char *file; /* filename of current input file */
51: struct buf *primary; /* buffer currently being parsed */
52: struct buf *secondary; /* preprocessed copy of input */
53: const char *defos; /* default operating system */
54: mandocmsg mmsg; /* warning/error message handler */
55: enum mandoclevel file_status; /* status of current parse */
56: enum mandoclevel wlevel; /* ignore messages below this */
57: int options; /* parser options */
1.70 schwarze 58: int filenc; /* encoding of the current file */
1.1 schwarze 59: int reparse_count; /* finite interp. stack */
1.59 schwarze 60: int line; /* line number in the file */
1.73 schwarze 61: pid_t child; /* the gunzip(1) process */
1.1 schwarze 62: };
63:
1.60 schwarze 64: static void choose_parser(struct mparse *);
1.1 schwarze 65: static void resize_buf(struct buf *, size_t);
1.71 schwarze 66: static void mparse_buf_r(struct mparse *, struct buf, size_t, int);
1.18 schwarze 67: static int read_whole_file(struct mparse *, const char *, int,
68: struct buf *, int *);
1.1 schwarze 69: static void mparse_end(struct mparse *);
1.15 schwarze 70: static void mparse_parse_buffer(struct mparse *, struct buf,
71: const char *);
1.1 schwarze 72:
73: static const enum mandocerr mandoclimits[MANDOCLEVEL_MAX] = {
74: MANDOCERR_OK,
75: MANDOCERR_WARNING,
76: MANDOCERR_WARNING,
77: MANDOCERR_ERROR,
78: MANDOCERR_FATAL,
79: MANDOCERR_MAX,
80: MANDOCERR_MAX
81: };
82:
83: static const char * const mandocerrs[MANDOCERR_MAX] = {
84: "ok",
85:
86: "generic warning",
87:
88: /* related to the prologue */
1.57 schwarze 89: "missing manual title, using UNTITLED",
90: "missing manual title, using \"\"",
1.32 schwarze 91: "lower case character in document title",
1.57 schwarze 92: "missing manual section, using \"\"",
1.1 schwarze 93: "unknown manual section",
1.32 schwarze 94: "missing date, using today's date",
1.1 schwarze 95: "cannot parse date, using it verbatim",
1.57 schwarze 96: "missing Os macro, using \"\"",
97: "duplicate prologue macro",
98: "late prologue macro",
99: "skipping late title macro",
1.1 schwarze 100: "prologue macros out of order",
101:
102: /* related to document structure */
103: ".so is fragile, better use ln(1)",
1.28 schwarze 104: "no document body",
1.32 schwarze 105: "content before first section header",
106: "first section is not \"NAME\"",
1.1 schwarze 107: "bad NAME section contents",
108: "sections out of conventional order",
1.32 schwarze 109: "duplicate section title",
110: "unexpected section",
1.63 schwarze 111: "unusual Xr order",
112: "unusual Xr punctuation",
1.62 schwarze 113: "AUTHORS section without An macro",
1.1 schwarze 114:
115: /* related to macros and nesting */
1.33 schwarze 116: "obsolete macro",
1.78 schwarze 117: "macro neither callable nor escaped",
1.1 schwarze 118: "skipping paragraph macro",
1.10 schwarze 119: "moving paragraph macro out of list",
1.1 schwarze 120: "skipping no-space macro",
121: "blocks badly nested",
122: "nested displays are not portable",
1.35 schwarze 123: "moving content out of list",
124: ".Vt block has child macro",
1.56 schwarze 125: "fill mode already enabled, skipping",
126: "fill mode already disabled, skipping",
1.1 schwarze 127: "line scope broken",
128:
129: /* related to missing macro arguments */
1.36 schwarze 130: "skipping empty request",
131: "conditional request controls empty scope",
1.1 schwarze 132: "skipping empty macro",
1.40 schwarze 133: "empty argument, using 0n",
1.1 schwarze 134: "argument count wrong",
1.38 schwarze 135: "missing display type, using -ragged",
136: "list type is not the first argument",
137: "missing -width in -tag list, using 8n",
1.56 schwarze 138: "missing utility name, using \"\"",
1.38 schwarze 139: "empty head in list item",
140: "empty list item",
1.39 schwarze 141: "missing font type, using \\fR",
142: "unknown font type, using \\fR",
1.79 schwarze 143: "nothing follows prefix",
1.38 schwarze 144: "missing -std argument, adding it",
1.66 schwarze 145: "missing eqn box, using \"\"",
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",
1.65 schwarze 156: "parenthesis in function name",
1.45 schwarze 157: "invalid content in Rs block",
1.41 schwarze 158: "invalid Boolean argument",
159: "unknown font, skipping request",
1.1 schwarze 160:
161: /* related to plain text */
1.42 schwarze 162: "blank line in fill mode, using .sp",
163: "tab in filled text",
164: "whitespace at end of input line",
1.1 schwarze 165: "bad comment style",
1.42 schwarze 166: "invalid escape sequence",
167: "undefined string, using \"\"",
1.3 schwarze 168:
1.1 schwarze 169: "generic error",
170:
1.3 schwarze 171: /* related to equations */
172: "unexpected equation scope closure",
173: "equation scope open on exit",
174: "overlapping equation scopes",
175: "unexpected end of equation",
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",
1.80 ! schwarze 186: "ignoring macro in table",
1.1 schwarze 187:
1.46 schwarze 188: /* related to document structure and macros */
1.1 schwarze 189: "input stack limit exceeded, infinite loop?",
190: "skipping bad character",
1.46 schwarze 191: "skipping unknown macro",
1.51 schwarze 192: "skipping item outside list",
1.46 schwarze 193: "skipping column outside column list",
194: "skipping end of block that is not open",
195: "inserting missing end of block",
196: "appending missing end of block",
197:
198: /* related to request and macro arguments */
1.1 schwarze 199: "escaped character not allowed in a name",
200: "argument count wrong",
1.75 schwarze 201: "NOT IMPLEMENTED: Bd -file",
1.49 schwarze 202: "missing list type, using -item",
1.48 schwarze 203: "missing manual name, using \"\"",
1.49 schwarze 204: "uname(3) system call failed, using UNKNOWN",
1.41 schwarze 205: "unknown standard specifier",
1.49 schwarze 206: "skipping request without numeric argument",
1.39 schwarze 207: "skipping all arguments",
208: "skipping excess arguments",
1.68 schwarze 209: "divide by zero",
1.1 schwarze 210:
211: "generic fatal error",
212:
1.18 schwarze 213: "input too large",
1.1 schwarze 214: "NOT IMPLEMENTED: .so with absolute path or \"..\"",
1.30 schwarze 215: ".so request failed",
1.18 schwarze 216:
217: /* system errors */
1.58 schwarze 218: "cannot dup file descriptor",
219: "cannot exec",
220: "gunzip failed with code",
221: "cannot fork",
1.29 schwarze 222: NULL,
1.58 schwarze 223: "cannot open pipe",
224: "cannot read file",
225: "gunzip died from signal",
1.18 schwarze 226: "cannot stat file",
1.58 schwarze 227: "wait failed",
1.1 schwarze 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:
1.25 schwarze 240:
1.1 schwarze 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
1.60 schwarze 250: choose_parser(struct mparse *curp)
1.1 schwarze 251: {
1.59 schwarze 252: char *cp, *ep;
253: int format;
1.1 schwarze 254:
1.59 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) {
1.61 schwarze 265: if (*cp == '.' || *cp == '\'') {
1.59 schwarze 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 schwarze 281: }
282:
1.59 schwarze 283: if (format == MPARSE_MDOC) {
1.25 schwarze 284: if (NULL == curp->pmdoc)
1.22 schwarze 285: curp->pmdoc = mdoc_alloc(
286: curp->roff, curp, curp->defos,
287: MPARSE_QUICK & curp->options ? 1 : 0);
1.1 schwarze 288: assert(curp->pmdoc);
289: curp->mdoc = curp->pmdoc;
290: return;
1.25 schwarze 291: }
1.1 schwarze 292:
1.59 schwarze 293: /* Fall back to man(7) as a last resort. */
294:
1.25 schwarze 295: if (NULL == curp->pman)
1.22 schwarze 296: curp->pman = man_alloc(curp->roff, curp,
297: MPARSE_QUICK & curp->options ? 1 : 0);
1.1 schwarze 298: assert(curp->pman);
299: curp->man = curp->pman;
300: }
301:
302: /*
1.71 schwarze 303: * Main parse routine for a buffer.
304: * It assumes encoding and line numbering are already set up.
305: * It can recurse directly (for invocations of user-defined
306: * macros, inline equations, and input line traps)
307: * and indirectly (for .so file inclusion).
1.1 schwarze 308: */
309: static void
1.71 schwarze 310: mparse_buf_r(struct mparse *curp, struct buf blk, size_t i, int start)
1.1 schwarze 311: {
312: const struct tbl_span *span;
313: struct buf ln;
1.71 schwarze 314: size_t pos; /* byte number in the ln buffer */
1.1 schwarze 315: enum rofferr rr;
1.76 schwarze 316: int of;
1.1 schwarze 317: int lnn; /* line number in the real file */
318: unsigned char c;
319:
1.71 schwarze 320: memset(&ln, 0, sizeof(ln));
1.1 schwarze 321:
1.25 schwarze 322: lnn = curp->line;
323: pos = 0;
1.1 schwarze 324:
1.71 schwarze 325: while (i < blk.sz) {
1.1 schwarze 326: if (0 == pos && '\0' == blk.buf[i])
327: break;
328:
329: if (start) {
330: curp->line = lnn;
331: curp->reparse_count = 0;
1.70 schwarze 332:
333: if (lnn < 3 &&
334: curp->filenc & MPARSE_UTF8 &&
1.71 schwarze 335: curp->filenc & MPARSE_LATIN1)
336: curp->filenc = preconv_cue(&blk, i);
1.1 schwarze 337: }
338:
1.71 schwarze 339: while (i < blk.sz && (start || blk.buf[i] != '\0')) {
1.1 schwarze 340:
341: /*
342: * When finding an unescaped newline character,
343: * leave the character loop to process the line.
344: * Skip a preceding carriage return, if any.
345: */
346:
1.71 schwarze 347: if ('\r' == blk.buf[i] && i + 1 < blk.sz &&
1.1 schwarze 348: '\n' == blk.buf[i + 1])
349: ++i;
350: if ('\n' == blk.buf[i]) {
351: ++i;
352: ++lnn;
353: break;
354: }
355:
1.13 schwarze 356: /*
1.70 schwarze 357: * Make sure we have space for the worst
358: * case of 11 bytes: "\\[u10ffff]\0"
1.13 schwarze 359: */
360:
1.71 schwarze 361: if (pos + 11 > ln.sz)
1.13 schwarze 362: resize_buf(&ln, 256);
363:
1.25 schwarze 364: /*
1.70 schwarze 365: * Encode 8-bit input.
1.1 schwarze 366: */
367:
1.70 schwarze 368: c = blk.buf[i];
369: if (c & 0x80) {
1.71 schwarze 370: if ( ! (curp->filenc && preconv_encode(
371: &blk, &i, &ln, &pos, &curp->filenc))) {
1.70 schwarze 372: mandoc_vmsg(MANDOCERR_BADCHAR,
373: curp, curp->line, pos,
374: "0x%x", c);
375: ln.buf[pos++] = '?';
376: i++;
377: }
378: continue;
379: }
380:
381: /*
382: * Exclude control characters.
383: */
1.1 schwarze 384:
1.70 schwarze 385: if (c == 0x7f || (c < 0x20 && c != 0x09)) {
1.56 schwarze 386: mandoc_vmsg(MANDOCERR_BADCHAR, curp,
387: curp->line, pos, "0x%x", c);
1.1 schwarze 388: i++;
1.6 schwarze 389: ln.buf[pos++] = '?';
1.1 schwarze 390: continue;
391: }
392:
393: /* Trailing backslash = a plain char. */
394:
1.71 schwarze 395: if (blk.buf[i] != '\\' || i + 1 == blk.sz) {
1.1 schwarze 396: ln.buf[pos++] = blk.buf[i++];
397: continue;
398: }
399:
400: /*
401: * Found escape and at least one other character.
402: * When it's a newline character, skip it.
403: * When there is a carriage return in between,
404: * skip that one as well.
405: */
406:
1.71 schwarze 407: if ('\r' == blk.buf[i + 1] && i + 2 < blk.sz &&
1.1 schwarze 408: '\n' == blk.buf[i + 2])
409: ++i;
410: if ('\n' == blk.buf[i + 1]) {
411: i += 2;
412: ++lnn;
413: continue;
414: }
415:
416: if ('"' == blk.buf[i + 1] || '#' == blk.buf[i + 1]) {
417: i += 2;
418: /* Comment, skip to end of line */
1.71 schwarze 419: for (; i < blk.sz; ++i) {
1.1 schwarze 420: if ('\n' == blk.buf[i]) {
421: ++i;
422: ++lnn;
423: break;
424: }
425: }
426:
427: /* Backout trailing whitespaces */
428: for (; pos > 0; --pos) {
429: if (ln.buf[pos - 1] != ' ')
430: break;
431: if (pos > 2 && ln.buf[pos - 2] == '\\')
432: break;
433: }
434: break;
435: }
436:
1.13 schwarze 437: /* Catch escaped bogus characters. */
438:
439: c = (unsigned char) blk.buf[i+1];
440:
1.25 schwarze 441: if ( ! (isascii(c) &&
442: (isgraph(c) || isblank(c)))) {
1.56 schwarze 443: mandoc_vmsg(MANDOCERR_BADCHAR, curp,
444: curp->line, pos, "0x%x", c);
1.13 schwarze 445: i += 2;
446: ln.buf[pos++] = '?';
447: continue;
448: }
449:
1.1 schwarze 450: /* Some other escape sequence, copy & cont. */
451:
452: ln.buf[pos++] = blk.buf[i++];
453: ln.buf[pos++] = blk.buf[i++];
454: }
455:
1.71 schwarze 456: if (pos >= ln.sz)
1.1 schwarze 457: resize_buf(&ln, 256);
458:
459: ln.buf[pos] = '\0';
460:
461: /*
462: * A significant amount of complexity is contained by
463: * the roff preprocessor. It's line-oriented but can be
464: * expressed on one line, so we need at times to
465: * readjust our starting point and re-run it. The roff
466: * preprocessor can also readjust the buffers with new
467: * data, so we pass them in wholesale.
468: */
469:
470: of = 0;
471:
1.4 schwarze 472: /*
473: * Maintain a lookaside buffer of all parsed lines. We
474: * only do this if mparse_keep() has been invoked (the
475: * buffer may be accessed with mparse_getkeep()).
476: */
477:
478: if (curp->secondary) {
1.25 schwarze 479: curp->secondary->buf = mandoc_realloc(
480: curp->secondary->buf,
481: curp->secondary->sz + pos + 2);
482: memcpy(curp->secondary->buf +
483: curp->secondary->sz,
484: ln.buf, pos);
1.4 schwarze 485: curp->secondary->sz += pos;
486: curp->secondary->buf
487: [curp->secondary->sz] = '\n';
488: curp->secondary->sz++;
489: curp->secondary->buf
490: [curp->secondary->sz] = '\0';
491: }
1.1 schwarze 492: rerun:
1.72 schwarze 493: rr = roff_parseln(curp->roff, curp->line, &ln, &of);
1.1 schwarze 494:
495: switch (rr) {
1.25 schwarze 496: case ROFF_REPARSE:
1.1 schwarze 497: if (REPARSE_LIMIT >= ++curp->reparse_count)
1.71 schwarze 498: mparse_buf_r(curp, ln, of, 0);
1.1 schwarze 499: else
500: mandoc_msg(MANDOCERR_ROFFLOOP, curp,
1.25 schwarze 501: curp->line, pos, NULL);
1.1 schwarze 502: pos = 0;
503: continue;
1.25 schwarze 504: case ROFF_APPEND:
1.71 schwarze 505: pos = strlen(ln.buf);
1.1 schwarze 506: continue;
1.25 schwarze 507: case ROFF_RERUN:
1.1 schwarze 508: goto rerun;
1.25 schwarze 509: case ROFF_IGN:
1.1 schwarze 510: pos = 0;
511: continue;
1.25 schwarze 512: case ROFF_ERR:
1.1 schwarze 513: assert(MANDOCLEVEL_FATAL <= curp->file_status);
514: break;
1.25 schwarze 515: case ROFF_SO:
1.71 schwarze 516: if ( ! (curp->options & MPARSE_SO) &&
517: (i >= blk.sz || blk.buf[i] == '\0')) {
1.23 schwarze 518: curp->sodest = mandoc_strdup(ln.buf + of);
519: free(ln.buf);
520: return;
521: }
1.4 schwarze 522: /*
523: * We remove `so' clauses from our lookaside
524: * buffer because we're going to descend into
525: * the file recursively.
526: */
1.25 schwarze 527: if (curp->secondary)
1.4 schwarze 528: curp->secondary->sz -= pos + 1;
1.14 schwarze 529: mparse_readfd(curp, -1, ln.buf + of);
1.30 schwarze 530: if (MANDOCLEVEL_FATAL <= curp->file_status) {
531: mandoc_vmsg(MANDOCERR_SO_FAIL,
532: curp, curp->line, pos,
533: ".so %s", ln.buf + of);
1.1 schwarze 534: break;
1.30 schwarze 535: }
1.1 schwarze 536: pos = 0;
537: continue;
538: default:
539: break;
540: }
541:
542: /*
543: * If we encounter errors in the recursive parse, make
544: * sure we don't continue parsing.
545: */
546:
547: if (MANDOCLEVEL_FATAL <= curp->file_status)
548: break;
549:
550: /*
551: * If input parsers have not been allocated, do so now.
1.2 schwarze 552: * We keep these instanced between parsers, but set them
1.1 schwarze 553: * locally per parse routine since we can use different
554: * parsers with each one.
555: */
556:
557: if ( ! (curp->man || curp->mdoc))
1.60 schwarze 558: choose_parser(curp);
1.1 schwarze 559:
1.25 schwarze 560: /*
1.60 schwarze 561: * Lastly, push down into the parsers themselves.
1.1 schwarze 562: * If libroff returns ROFF_TBL, then add it to the
563: * currently open parse. Since we only get here if
564: * there does exist data (see tbl_data.c), we're
565: * guaranteed that something's been allocated.
566: * Do the same for ROFF_EQN.
567: */
568:
1.76 schwarze 569: if (rr == ROFF_TBL) {
570: while ((span = roff_span(curp->roff)) != NULL)
571: if (curp->man == NULL)
572: mdoc_addspan(curp->mdoc, span);
573: else
574: man_addspan(curp->man, span);
575: } else if (rr == ROFF_EQN) {
576: if (curp->man == NULL)
577: mdoc_addeqn(curp->mdoc, roff_eqn(curp->roff));
578: else
579: man_addeqn(curp->man, roff_eqn(curp->roff));
580: } else if ((curp->man == NULL ?
581: mdoc_parseln(curp->mdoc, curp->line, ln.buf, of) :
582: man_parseln(curp->man, curp->line, ln.buf, of)) == 2)
583: break;
1.1 schwarze 584:
585: /* Temporary buffers typically are not full. */
586:
587: if (0 == start && '\0' == blk.buf[i])
588: break;
589:
590: /* Start the next input line. */
591:
592: pos = 0;
593: }
594:
595: free(ln.buf);
596: }
597:
598: static int
1.18 schwarze 599: read_whole_file(struct mparse *curp, const char *file, int fd,
600: struct buf *fb, int *with_mmap)
1.1 schwarze 601: {
602: struct stat st;
603: size_t off;
604: ssize_t ssz;
605:
606: if (-1 == fstat(fd, &st)) {
1.18 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 schwarze 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)) {
1.18 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 schwarze 627: return(0);
628: }
629: *with_mmap = 1;
630: fb->sz = (size_t)st.st_size;
1.15 schwarze 631: fb->buf = mmap(NULL, fb->sz, PROT_READ, MAP_SHARED, fd, 0);
1.1 schwarze 632: if (fb->buf != MAP_FAILED)
633: return(1);
634: }
635:
636: /*
637: * If this isn't a regular file (like, say, stdin), then we must
638: * go the old way and just read things in bit by bit.
639: */
640:
641: *with_mmap = 0;
642: off = 0;
643: fb->sz = 0;
644: fb->buf = NULL;
645: for (;;) {
646: if (off == fb->sz) {
647: if (fb->sz == (1U << 31)) {
1.18 schwarze 648: curp->file_status = MANDOCLEVEL_FATAL;
649: if (curp->mmsg)
650: (*curp->mmsg)(MANDOCERR_TOOLARGE,
651: curp->file_status,
652: file, 0, 0, NULL);
1.1 schwarze 653: break;
654: }
655: resize_buf(fb, 65536);
656: }
657: ssz = read(fd, fb->buf + (int)off, fb->sz - off);
658: if (ssz == 0) {
659: fb->sz = off;
660: return(1);
661: }
662: if (ssz == -1) {
1.18 schwarze 663: curp->file_status = MANDOCLEVEL_SYSERR;
664: if (curp->mmsg)
665: (*curp->mmsg)(MANDOCERR_SYSREAD,
666: curp->file_status, file, 0, 0,
667: strerror(errno));
1.1 schwarze 668: break;
669: }
670: off += (size_t)ssz;
671: }
672:
673: free(fb->buf);
674: fb->buf = NULL;
675: return(0);
676: }
677:
678: static void
679: mparse_end(struct mparse *curp)
680: {
681:
682: if (MANDOCLEVEL_FATAL <= curp->file_status)
683: return;
684:
1.50 schwarze 685: if (curp->mdoc == NULL &&
686: curp->man == NULL &&
687: curp->sodest == NULL) {
688: if (curp->options & MPARSE_MDOC)
689: curp->mdoc = curp->pmdoc;
690: else {
691: if (curp->pman == NULL)
692: curp->pman = man_alloc(curp->roff, curp,
693: curp->options & MPARSE_QUICK ? 1 : 0);
694: curp->man = curp->pman;
695: }
696: }
697:
1.1 schwarze 698: if (curp->mdoc && ! mdoc_endparse(curp->mdoc)) {
699: assert(MANDOCLEVEL_FATAL <= curp->file_status);
700: return;
701: }
702:
703: if (curp->man && ! man_endparse(curp->man)) {
704: assert(MANDOCLEVEL_FATAL <= curp->file_status);
705: return;
706: }
707:
708: roff_endparse(curp->roff);
709: }
710:
1.15 schwarze 711: static void
712: mparse_parse_buffer(struct mparse *curp, struct buf blk, const char *file)
1.1 schwarze 713: {
1.61 schwarze 714: struct buf *svprimary;
1.1 schwarze 715: const char *svfile;
1.71 schwarze 716: size_t offset;
1.14 schwarze 717: static int recursion_depth;
718:
719: if (64 < recursion_depth) {
720: mandoc_msg(MANDOCERR_ROFFLOOP, curp, curp->line, 0, NULL);
1.15 schwarze 721: return;
1.14 schwarze 722: }
1.1 schwarze 723:
1.15 schwarze 724: /* Line number is per-file. */
725: svfile = curp->file;
726: curp->file = file;
1.61 schwarze 727: svprimary = curp->primary;
1.59 schwarze 728: curp->primary = &blk;
1.15 schwarze 729: curp->line = 1;
730: recursion_depth++;
731:
1.70 schwarze 732: /* Skip an UTF-8 byte order mark. */
733: if (curp->filenc & MPARSE_UTF8 && blk.sz > 2 &&
734: (unsigned char)blk.buf[0] == 0xef &&
735: (unsigned char)blk.buf[1] == 0xbb &&
736: (unsigned char)blk.buf[2] == 0xbf) {
1.71 schwarze 737: offset = 3;
1.70 schwarze 738: curp->filenc &= ~MPARSE_LATIN1;
1.71 schwarze 739: } else
740: offset = 0;
1.70 schwarze 741:
1.71 schwarze 742: mparse_buf_r(curp, blk, offset, 1);
1.15 schwarze 743:
744: if (0 == --recursion_depth && MANDOCLEVEL_FATAL > curp->file_status)
745: mparse_end(curp);
746:
1.61 schwarze 747: curp->primary = svprimary;
1.15 schwarze 748: curp->file = svfile;
749: }
750:
1.74 schwarze 751: /*
752: * If a file descriptor is given, use it and assume it points
753: * to the named file. Otherwise, open the named file.
754: * Read the whole file into memory and call the parsers.
755: * Called recursively when an .so request is encountered.
756: */
1.15 schwarze 757: enum mandoclevel
758: mparse_readfd(struct mparse *curp, int fd, const char *file)
759: {
760: struct buf blk;
761: int with_mmap;
1.70 schwarze 762: int save_filenc;
1.74 schwarze 763: pid_t save_child;
1.15 schwarze 764:
1.74 schwarze 765: save_child = curp->child;
766: if (fd != -1)
767: curp->child = 0;
768: else if (mparse_open(curp, &fd, file) >= MANDOCLEVEL_SYSERR)
769: goto out;
1.1 schwarze 770:
1.67 schwarze 771: if (read_whole_file(curp, file, fd, &blk, &with_mmap)) {
1.70 schwarze 772: save_filenc = curp->filenc;
773: curp->filenc = curp->options &
774: (MPARSE_UTF8 | MPARSE_LATIN1);
1.67 schwarze 775: mparse_parse_buffer(curp, blk, file);
1.70 schwarze 776: curp->filenc = save_filenc;
1.67 schwarze 777: if (with_mmap)
778: munmap(blk.buf, blk.sz);
779: else
780: free(blk.buf);
781: }
1.1 schwarze 782:
1.74 schwarze 783: if (fd != STDIN_FILENO && close(fd) == -1)
1.1 schwarze 784: perror(file);
1.67 schwarze 785:
1.74 schwarze 786: mparse_wait(curp);
787: out:
788: curp->child = save_child;
1.1 schwarze 789: return(curp->file_status);
1.58 schwarze 790: }
791:
792: enum mandoclevel
1.73 schwarze 793: mparse_open(struct mparse *curp, int *fd, const char *file)
1.58 schwarze 794: {
795: int pfd[2];
1.74 schwarze 796: int save_errno;
1.58 schwarze 797: char *cp;
798: enum mandocerr err;
799:
800: pfd[1] = -1;
801: curp->file = file;
1.74 schwarze 802:
803: /* Unless zipped, try to just open the file. */
804:
1.58 schwarze 805: if ((cp = strrchr(file, '.')) == NULL ||
806: strcmp(cp + 1, "gz")) {
1.73 schwarze 807: curp->child = 0;
1.74 schwarze 808: if ((*fd = open(file, O_RDONLY)) != -1)
809: return(MANDOCLEVEL_OK);
810:
811: /* Open failed; try to append ".gz". */
812:
813: mandoc_asprintf(&cp, "%s.gz", file);
814: file = cp;
815: } else
816: cp = NULL;
817:
818: /* Before forking, make sure the file can be read. */
819:
820: save_errno = errno;
821: if (access(file, R_OK) == -1) {
822: if (cp != NULL)
823: errno = save_errno;
824: err = MANDOCERR_SYSOPEN;
825: goto out;
1.58 schwarze 826: }
827:
1.74 schwarze 828: /* Run gunzip(1). */
829:
1.58 schwarze 830: if (pipe(pfd) == -1) {
831: err = MANDOCERR_SYSPIPE;
832: goto out;
833: }
834:
1.73 schwarze 835: switch (curp->child = fork()) {
1.58 schwarze 836: case -1:
837: err = MANDOCERR_SYSFORK;
838: close(pfd[0]);
839: close(pfd[1]);
840: pfd[1] = -1;
841: break;
842: case 0:
843: close(pfd[0]);
844: if (dup2(pfd[1], STDOUT_FILENO) == -1) {
845: err = MANDOCERR_SYSDUP;
846: break;
847: }
848: execlp("gunzip", "gunzip", "-c", file, NULL);
849: err = MANDOCERR_SYSEXEC;
850: break;
851: default:
852: close(pfd[1]);
853: *fd = pfd[0];
854: return(MANDOCLEVEL_OK);
855: }
856:
857: out:
1.74 schwarze 858: free(cp);
1.58 schwarze 859: *fd = -1;
1.73 schwarze 860: curp->child = 0;
1.58 schwarze 861: curp->file_status = MANDOCLEVEL_SYSERR;
862: if (curp->mmsg)
1.74 schwarze 863: (*curp->mmsg)(err, curp->file_status, curp->file,
1.58 schwarze 864: 0, 0, strerror(errno));
865: if (pfd[1] != -1)
866: exit(1);
867: return(curp->file_status);
868: }
869:
870: enum mandoclevel
1.73 schwarze 871: mparse_wait(struct mparse *curp)
1.58 schwarze 872: {
873: int status;
874:
1.73 schwarze 875: if (curp->child == 0)
876: return(MANDOCLEVEL_OK);
877:
878: if (waitpid(curp->child, &status, 0) == -1) {
1.58 schwarze 879: mandoc_msg(MANDOCERR_SYSWAIT, curp, 0, 0,
880: strerror(errno));
881: curp->file_status = MANDOCLEVEL_SYSERR;
882: return(curp->file_status);
883: }
884: if (WIFSIGNALED(status)) {
885: mandoc_vmsg(MANDOCERR_SYSSIG, curp, 0, 0,
886: "%d", WTERMSIG(status));
887: curp->file_status = MANDOCLEVEL_SYSERR;
888: return(curp->file_status);
889: }
890: if (WEXITSTATUS(status)) {
891: mandoc_vmsg(MANDOCERR_SYSEXIT, curp, 0, 0,
892: "%d", WEXITSTATUS(status));
893: curp->file_status = MANDOCLEVEL_SYSERR;
894: return(curp->file_status);
895: }
896: return(MANDOCLEVEL_OK);
1.1 schwarze 897: }
898:
899: struct mparse *
1.69 schwarze 900: mparse_alloc(int options, enum mandoclevel wlevel, mandocmsg mmsg,
901: const struct mchars *mchars, const char *defos)
1.1 schwarze 902: {
903: struct mparse *curp;
904:
905: assert(wlevel <= MANDOCLEVEL_FATAL);
906:
907: curp = mandoc_calloc(1, sizeof(struct mparse));
908:
1.22 schwarze 909: curp->options = options;
1.1 schwarze 910: curp->wlevel = wlevel;
911: curp->mmsg = mmsg;
1.7 schwarze 912: curp->defos = defos;
1.1 schwarze 913:
1.69 schwarze 914: curp->mchars = mchars;
915: curp->roff = roff_alloc(curp, curp->mchars, options);
1.50 schwarze 916: if (curp->options & MPARSE_MDOC)
917: curp->pmdoc = mdoc_alloc(
918: curp->roff, curp, curp->defos,
919: curp->options & MPARSE_QUICK ? 1 : 0);
920: if (curp->options & MPARSE_MAN)
921: curp->pman = man_alloc(curp->roff, curp,
922: curp->options & MPARSE_QUICK ? 1 : 0);
923:
1.1 schwarze 924: return(curp);
925: }
926:
927: void
928: mparse_reset(struct mparse *curp)
929: {
930:
931: roff_reset(curp->roff);
932:
933: if (curp->mdoc)
934: mdoc_reset(curp->mdoc);
935: if (curp->man)
936: man_reset(curp->man);
1.4 schwarze 937: if (curp->secondary)
938: curp->secondary->sz = 0;
1.1 schwarze 939:
940: curp->file_status = MANDOCLEVEL_OK;
941: curp->mdoc = NULL;
942: curp->man = NULL;
1.23 schwarze 943:
944: free(curp->sodest);
945: curp->sodest = NULL;
1.1 schwarze 946: }
947:
948: void
949: mparse_free(struct mparse *curp)
950: {
951:
952: if (curp->pmdoc)
953: mdoc_free(curp->pmdoc);
954: if (curp->pman)
955: man_free(curp->pman);
956: if (curp->roff)
957: roff_free(curp->roff);
1.4 schwarze 958: if (curp->secondary)
959: free(curp->secondary->buf);
1.1 schwarze 960:
1.4 schwarze 961: free(curp->secondary);
1.23 schwarze 962: free(curp->sodest);
1.1 schwarze 963: free(curp);
964: }
965:
966: void
1.23 schwarze 967: mparse_result(struct mparse *curp,
968: struct mdoc **mdoc, struct man **man, char **sodest)
1.1 schwarze 969: {
970:
1.23 schwarze 971: if (sodest && NULL != (*sodest = curp->sodest)) {
972: *mdoc = NULL;
973: *man = NULL;
974: return;
975: }
1.1 schwarze 976: if (mdoc)
977: *mdoc = curp->mdoc;
978: if (man)
979: *man = curp->man;
980: }
981:
982: void
983: mandoc_vmsg(enum mandocerr t, struct mparse *m,
984: int ln, int pos, const char *fmt, ...)
985: {
986: char buf[256];
987: va_list ap;
988:
989: va_start(ap, fmt);
1.26 schwarze 990: (void)vsnprintf(buf, sizeof(buf), fmt, ap);
1.1 schwarze 991: va_end(ap);
992:
993: mandoc_msg(t, m, ln, pos, buf);
994: }
995:
996: void
1.25 schwarze 997: mandoc_msg(enum mandocerr er, struct mparse *m,
1.1 schwarze 998: int ln, int col, const char *msg)
999: {
1000: enum mandoclevel level;
1001:
1002: level = MANDOCLEVEL_FATAL;
1003: while (er < mandoclimits[level])
1004: level--;
1005:
1006: if (level < m->wlevel)
1007: return;
1008:
1009: if (m->mmsg)
1010: (*m->mmsg)(er, level, m->file, ln, col, msg);
1011:
1012: if (m->file_status < level)
1013: m->file_status = level;
1014: }
1015:
1016: const char *
1017: mparse_strerror(enum mandocerr er)
1018: {
1019:
1020: return(mandocerrs[er]);
1021: }
1022:
1023: const char *
1024: mparse_strlevel(enum mandoclevel lvl)
1025: {
1026: return(mandoclevels[lvl]);
1.4 schwarze 1027: }
1028:
1029: void
1030: mparse_keep(struct mparse *p)
1031: {
1032:
1033: assert(NULL == p->secondary);
1034: p->secondary = mandoc_calloc(1, sizeof(struct buf));
1035: }
1036:
1037: const char *
1038: mparse_getkeep(const struct mparse *p)
1039: {
1040:
1041: assert(p->secondary);
1042: return(p->secondary->sz ? p->secondary->buf : NULL);
1.1 schwarze 1043: }