Annotation of src/usr.bin/mandoc/read.c, Revision 1.103
1.103 ! schwarze 1: /* $OpenBSD: read.c,v 1.102 2015/02/20 22:40:05 schwarze Exp $ */
1.1 schwarze 2: /*
3: * Copyright (c) 2008, 2009, 2010, 2011 Kristaps Dzonsons <kristaps@bsd.lv>
1.82 schwarze 4: * Copyright (c) 2010-2015 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,
1.87 schwarze 78: MANDOCERR_UNSUPP,
1.1 schwarze 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\"",
1.103 ! schwarze 107: "NAME section without name",
! 108: "NAME section without description",
! 109: "description not at the end of NAME",
! 110: "bad NAME section content",
1.96 schwarze 111: "missing description line, using \"\"",
1.1 schwarze 112: "sections out of conventional order",
1.32 schwarze 113: "duplicate section title",
114: "unexpected section",
1.63 schwarze 115: "unusual Xr order",
116: "unusual Xr punctuation",
1.62 schwarze 117: "AUTHORS section without An macro",
1.1 schwarze 118:
119: /* related to macros and nesting */
1.33 schwarze 120: "obsolete macro",
1.78 schwarze 121: "macro neither callable nor escaped",
1.1 schwarze 122: "skipping paragraph macro",
1.10 schwarze 123: "moving paragraph macro out of list",
1.1 schwarze 124: "skipping no-space macro",
125: "blocks badly nested",
126: "nested displays are not portable",
1.35 schwarze 127: "moving content out of list",
128: ".Vt block has child macro",
1.56 schwarze 129: "fill mode already enabled, skipping",
130: "fill mode already disabled, skipping",
1.1 schwarze 131: "line scope broken",
132:
133: /* related to missing macro arguments */
1.36 schwarze 134: "skipping empty request",
135: "conditional request controls empty scope",
1.1 schwarze 136: "skipping empty macro",
1.99 schwarze 137: "empty block",
1.40 schwarze 138: "empty argument, using 0n",
1.38 schwarze 139: "missing display type, using -ragged",
140: "list type is not the first argument",
141: "missing -width in -tag list, using 8n",
1.56 schwarze 142: "missing utility name, using \"\"",
1.98 schwarze 143: "missing function name, using \"\"",
1.38 schwarze 144: "empty head in list item",
145: "empty list item",
1.39 schwarze 146: "missing font type, using \\fR",
147: "unknown font type, using \\fR",
1.79 schwarze 148: "nothing follows prefix",
1.97 schwarze 149: "empty reference block",
1.38 schwarze 150: "missing -std argument, adding it",
1.100 schwarze 151: "missing option string, using \"\"",
152: "missing resource identifier, using \"\"",
1.66 schwarze 153: "missing eqn box, using \"\"",
1.1 schwarze 154:
155: /* related to bad macro arguments */
1.42 schwarze 156: "unterminated quoted argument",
1.1 schwarze 157: "duplicate argument",
1.54 schwarze 158: "skipping duplicate argument",
1.41 schwarze 159: "skipping duplicate display type",
160: "skipping duplicate list type",
1.54 schwarze 161: "skipping -width argument",
1.101 schwarze 162: "wrong number of cells",
1.1 schwarze 163: "unknown AT&T UNIX version",
1.64 schwarze 164: "comma in function argument",
1.65 schwarze 165: "parenthesis in function name",
1.45 schwarze 166: "invalid content in Rs block",
1.41 schwarze 167: "invalid Boolean argument",
168: "unknown font, skipping request",
1.101 schwarze 169: "odd number of characters in request",
1.1 schwarze 170:
171: /* related to plain text */
1.42 schwarze 172: "blank line in fill mode, using .sp",
173: "tab in filled text",
174: "whitespace at end of input line",
1.1 schwarze 175: "bad comment style",
1.42 schwarze 176: "invalid escape sequence",
177: "undefined string, using \"\"",
1.3 schwarze 178:
1.93 schwarze 179: /* related to tables */
180: "tbl line starts with span",
181: "tbl column starts with span",
182: "skipping vertical bar in tbl layout",
183:
1.1 schwarze 184: "generic error",
1.3 schwarze 185:
1.1 schwarze 186: /* related to tables */
1.91 schwarze 187: "non-alphabetic character in tbl options",
188: "skipping unknown tbl option",
189: "missing tbl option argument",
190: "wrong tbl option argument size",
1.93 schwarze 191: "empty tbl layout",
192: "invalid character in tbl layout",
193: "unmatched parenthesis in tbl layout",
1.94 schwarze 194: "tbl without any data cells",
195: "ignoring data in spanned tbl cell",
196: "ignoring extra tbl data cells",
197: "data block open at end of tbl",
1.1 schwarze 198:
1.46 schwarze 199: /* related to document structure and macros */
1.83 schwarze 200: NULL,
1.1 schwarze 201: "input stack limit exceeded, infinite loop?",
202: "skipping bad character",
1.46 schwarze 203: "skipping unknown macro",
1.87 schwarze 204: "skipping insecure request",
1.51 schwarze 205: "skipping item outside list",
1.46 schwarze 206: "skipping column outside column list",
207: "skipping end of block that is not open",
1.90 schwarze 208: "fewer RS blocks open, skipping",
1.46 schwarze 209: "inserting missing end of block",
210: "appending missing end of block",
211:
212: /* related to request and macro arguments */
1.1 schwarze 213: "escaped character not allowed in a name",
1.75 schwarze 214: "NOT IMPLEMENTED: Bd -file",
1.49 schwarze 215: "missing list type, using -item",
1.48 schwarze 216: "missing manual name, using \"\"",
1.49 schwarze 217: "uname(3) system call failed, using UNKNOWN",
1.41 schwarze 218: "unknown standard specifier",
1.49 schwarze 219: "skipping request without numeric argument",
1.84 schwarze 220: "NOT IMPLEMENTED: .so with absolute path or \"..\"",
1.85 schwarze 221: ".so request failed",
1.39 schwarze 222: "skipping all arguments",
223: "skipping excess arguments",
1.68 schwarze 224: "divide by zero",
1.87 schwarze 225:
226: "unsupported feature",
227: "input too large",
1.89 schwarze 228: "unsupported control character",
1.87 schwarze 229: "unsupported roff request",
1.94 schwarze 230: "eqn delim option in tbl",
1.93 schwarze 231: "unsupported tbl layout modifier",
1.87 schwarze 232: "ignoring macro in table",
1.1 schwarze 233: };
234:
235: static const char * const mandoclevels[MANDOCLEVEL_MAX] = {
236: "SUCCESS",
237: "RESERVED",
238: "WARNING",
239: "ERROR",
1.87 schwarze 240: "UNSUPP",
1.1 schwarze 241: "BADARG",
242: "SYSERR"
243: };
244:
1.25 schwarze 245:
1.1 schwarze 246: static void
247: resize_buf(struct buf *buf, size_t initial)
248: {
249:
250: buf->sz = buf->sz > initial/2 ? 2 * buf->sz : initial;
251: buf->buf = mandoc_realloc(buf->buf, buf->sz);
252: }
253:
254: static void
1.60 schwarze 255: choose_parser(struct mparse *curp)
1.1 schwarze 256: {
1.59 schwarze 257: char *cp, *ep;
258: int format;
1.1 schwarze 259:
1.59 schwarze 260: /*
261: * If neither command line arguments -mdoc or -man select
262: * a parser nor the roff parser found a .Dd or .TH macro
263: * yet, look ahead in the main input buffer.
264: */
265:
266: if ((format = roff_getformat(curp->roff)) == 0) {
267: cp = curp->primary->buf;
268: ep = cp + curp->primary->sz;
269: while (cp < ep) {
1.61 schwarze 270: if (*cp == '.' || *cp == '\'') {
1.59 schwarze 271: cp++;
272: if (cp[0] == 'D' && cp[1] == 'd') {
273: format = MPARSE_MDOC;
274: break;
275: }
276: if (cp[0] == 'T' && cp[1] == 'H') {
277: format = MPARSE_MAN;
278: break;
279: }
280: }
281: cp = memchr(cp, '\n', ep - cp);
282: if (cp == NULL)
283: break;
284: cp++;
285: }
1.1 schwarze 286: }
287:
1.59 schwarze 288: if (format == MPARSE_MDOC) {
1.25 schwarze 289: if (NULL == curp->pmdoc)
1.22 schwarze 290: curp->pmdoc = mdoc_alloc(
291: curp->roff, curp, curp->defos,
292: MPARSE_QUICK & curp->options ? 1 : 0);
1.1 schwarze 293: assert(curp->pmdoc);
294: curp->mdoc = curp->pmdoc;
295: return;
1.25 schwarze 296: }
1.1 schwarze 297:
1.59 schwarze 298: /* Fall back to man(7) as a last resort. */
299:
1.25 schwarze 300: if (NULL == curp->pman)
1.81 schwarze 301: curp->pman = man_alloc(
302: curp->roff, curp, curp->defos,
1.22 schwarze 303: MPARSE_QUICK & curp->options ? 1 : 0);
1.1 schwarze 304: assert(curp->pman);
305: curp->man = curp->pman;
306: }
307:
308: /*
1.71 schwarze 309: * Main parse routine for a buffer.
310: * It assumes encoding and line numbering are already set up.
311: * It can recurse directly (for invocations of user-defined
312: * macros, inline equations, and input line traps)
313: * and indirectly (for .so file inclusion).
1.1 schwarze 314: */
315: static void
1.71 schwarze 316: mparse_buf_r(struct mparse *curp, struct buf blk, size_t i, int start)
1.1 schwarze 317: {
318: const struct tbl_span *span;
319: struct buf ln;
1.88 schwarze 320: const char *save_file;
1.85 schwarze 321: char *cp;
1.71 schwarze 322: size_t pos; /* byte number in the ln buffer */
1.1 schwarze 323: enum rofferr rr;
1.76 schwarze 324: int of;
1.1 schwarze 325: int lnn; /* line number in the real file */
1.85 schwarze 326: int fd;
327: pid_t save_child;
1.1 schwarze 328: unsigned char c;
329:
1.71 schwarze 330: memset(&ln, 0, sizeof(ln));
1.1 schwarze 331:
1.25 schwarze 332: lnn = curp->line;
333: pos = 0;
1.1 schwarze 334:
1.71 schwarze 335: while (i < blk.sz) {
1.1 schwarze 336: if (0 == pos && '\0' == blk.buf[i])
337: break;
338:
339: if (start) {
340: curp->line = lnn;
341: curp->reparse_count = 0;
1.70 schwarze 342:
343: if (lnn < 3 &&
344: curp->filenc & MPARSE_UTF8 &&
1.71 schwarze 345: curp->filenc & MPARSE_LATIN1)
346: curp->filenc = preconv_cue(&blk, i);
1.1 schwarze 347: }
348:
1.71 schwarze 349: while (i < blk.sz && (start || blk.buf[i] != '\0')) {
1.1 schwarze 350:
351: /*
352: * When finding an unescaped newline character,
353: * leave the character loop to process the line.
354: * Skip a preceding carriage return, if any.
355: */
356:
1.71 schwarze 357: if ('\r' == blk.buf[i] && i + 1 < blk.sz &&
1.1 schwarze 358: '\n' == blk.buf[i + 1])
359: ++i;
360: if ('\n' == blk.buf[i]) {
361: ++i;
362: ++lnn;
363: break;
364: }
365:
1.13 schwarze 366: /*
1.70 schwarze 367: * Make sure we have space for the worst
368: * case of 11 bytes: "\\[u10ffff]\0"
1.13 schwarze 369: */
370:
1.71 schwarze 371: if (pos + 11 > ln.sz)
1.13 schwarze 372: resize_buf(&ln, 256);
373:
1.25 schwarze 374: /*
1.70 schwarze 375: * Encode 8-bit input.
1.1 schwarze 376: */
377:
1.70 schwarze 378: c = blk.buf[i];
379: if (c & 0x80) {
1.71 schwarze 380: if ( ! (curp->filenc && preconv_encode(
381: &blk, &i, &ln, &pos, &curp->filenc))) {
1.89 schwarze 382: mandoc_vmsg(MANDOCERR_CHAR_BAD, curp,
383: curp->line, pos, "0x%x", c);
1.70 schwarze 384: ln.buf[pos++] = '?';
385: i++;
386: }
387: continue;
388: }
389:
390: /*
391: * Exclude control characters.
392: */
1.1 schwarze 393:
1.70 schwarze 394: if (c == 0x7f || (c < 0x20 && c != 0x09)) {
1.89 schwarze 395: mandoc_vmsg(c == 0x00 || c == 0x04 ||
396: c > 0x0a ? MANDOCERR_CHAR_BAD :
397: MANDOCERR_CHAR_UNSUPP,
398: curp, curp->line, pos, "0x%x", c);
1.1 schwarze 399: i++;
1.102 schwarze 400: if (c != '\r')
401: ln.buf[pos++] = '?';
1.1 schwarze 402: continue;
403: }
404:
405: /* Trailing backslash = a plain char. */
406:
1.71 schwarze 407: if (blk.buf[i] != '\\' || i + 1 == blk.sz) {
1.1 schwarze 408: ln.buf[pos++] = blk.buf[i++];
409: continue;
410: }
411:
412: /*
413: * Found escape and at least one other character.
414: * When it's a newline character, skip it.
415: * When there is a carriage return in between,
416: * skip that one as well.
417: */
418:
1.71 schwarze 419: if ('\r' == blk.buf[i + 1] && i + 2 < blk.sz &&
1.1 schwarze 420: '\n' == blk.buf[i + 2])
421: ++i;
422: if ('\n' == blk.buf[i + 1]) {
423: i += 2;
424: ++lnn;
425: continue;
426: }
427:
428: if ('"' == blk.buf[i + 1] || '#' == blk.buf[i + 1]) {
429: i += 2;
430: /* Comment, skip to end of line */
1.71 schwarze 431: for (; i < blk.sz; ++i) {
1.1 schwarze 432: if ('\n' == blk.buf[i]) {
433: ++i;
434: ++lnn;
435: break;
436: }
437: }
438:
439: /* Backout trailing whitespaces */
440: for (; pos > 0; --pos) {
441: if (ln.buf[pos - 1] != ' ')
442: break;
443: if (pos > 2 && ln.buf[pos - 2] == '\\')
444: break;
445: }
446: break;
447: }
448:
1.13 schwarze 449: /* Catch escaped bogus characters. */
450:
451: c = (unsigned char) blk.buf[i+1];
452:
1.25 schwarze 453: if ( ! (isascii(c) &&
454: (isgraph(c) || isblank(c)))) {
1.89 schwarze 455: mandoc_vmsg(MANDOCERR_CHAR_BAD, curp,
1.56 schwarze 456: curp->line, pos, "0x%x", c);
1.13 schwarze 457: i += 2;
458: ln.buf[pos++] = '?';
459: continue;
460: }
461:
1.1 schwarze 462: /* Some other escape sequence, copy & cont. */
463:
464: ln.buf[pos++] = blk.buf[i++];
465: ln.buf[pos++] = blk.buf[i++];
466: }
467:
1.71 schwarze 468: if (pos >= ln.sz)
1.1 schwarze 469: resize_buf(&ln, 256);
470:
471: ln.buf[pos] = '\0';
472:
473: /*
474: * A significant amount of complexity is contained by
475: * the roff preprocessor. It's line-oriented but can be
476: * expressed on one line, so we need at times to
477: * readjust our starting point and re-run it. The roff
478: * preprocessor can also readjust the buffers with new
479: * data, so we pass them in wholesale.
480: */
481:
482: of = 0;
483:
1.4 schwarze 484: /*
485: * Maintain a lookaside buffer of all parsed lines. We
486: * only do this if mparse_keep() has been invoked (the
487: * buffer may be accessed with mparse_getkeep()).
488: */
489:
490: if (curp->secondary) {
1.25 schwarze 491: curp->secondary->buf = mandoc_realloc(
492: curp->secondary->buf,
493: curp->secondary->sz + pos + 2);
494: memcpy(curp->secondary->buf +
495: curp->secondary->sz,
496: ln.buf, pos);
1.4 schwarze 497: curp->secondary->sz += pos;
498: curp->secondary->buf
499: [curp->secondary->sz] = '\n';
500: curp->secondary->sz++;
501: curp->secondary->buf
502: [curp->secondary->sz] = '\0';
503: }
1.1 schwarze 504: rerun:
1.72 schwarze 505: rr = roff_parseln(curp->roff, curp->line, &ln, &of);
1.1 schwarze 506:
507: switch (rr) {
1.25 schwarze 508: case ROFF_REPARSE:
1.1 schwarze 509: if (REPARSE_LIMIT >= ++curp->reparse_count)
1.71 schwarze 510: mparse_buf_r(curp, ln, of, 0);
1.1 schwarze 511: else
512: mandoc_msg(MANDOCERR_ROFFLOOP, curp,
1.25 schwarze 513: curp->line, pos, NULL);
1.1 schwarze 514: pos = 0;
515: continue;
1.25 schwarze 516: case ROFF_APPEND:
1.71 schwarze 517: pos = strlen(ln.buf);
1.1 schwarze 518: continue;
1.25 schwarze 519: case ROFF_RERUN:
1.1 schwarze 520: goto rerun;
1.25 schwarze 521: case ROFF_IGN:
1.1 schwarze 522: pos = 0;
523: continue;
1.25 schwarze 524: case ROFF_SO:
1.71 schwarze 525: if ( ! (curp->options & MPARSE_SO) &&
526: (i >= blk.sz || blk.buf[i] == '\0')) {
1.23 schwarze 527: curp->sodest = mandoc_strdup(ln.buf + of);
528: free(ln.buf);
529: return;
530: }
1.4 schwarze 531: /*
532: * We remove `so' clauses from our lookaside
533: * buffer because we're going to descend into
534: * the file recursively.
535: */
1.25 schwarze 536: if (curp->secondary)
1.4 schwarze 537: curp->secondary->sz -= pos + 1;
1.88 schwarze 538: save_file = curp->file;
1.85 schwarze 539: save_child = curp->child;
540: if (mparse_open(curp, &fd, ln.buf + of) ==
1.88 schwarze 541: MANDOCLEVEL_OK) {
1.85 schwarze 542: mparse_readfd(curp, fd, ln.buf + of);
1.88 schwarze 543: curp->file = save_file;
544: } else {
545: curp->file = save_file;
1.30 schwarze 546: mandoc_vmsg(MANDOCERR_SO_FAIL,
547: curp, curp->line, pos,
548: ".so %s", ln.buf + of);
1.85 schwarze 549: ln.sz = mandoc_asprintf(&cp,
550: ".sp\nSee the file %s.\n.sp",
551: ln.buf + of);
552: free(ln.buf);
553: ln.buf = cp;
554: of = 0;
555: mparse_buf_r(curp, ln, of, 0);
1.30 schwarze 556: }
1.85 schwarze 557: curp->child = save_child;
1.1 schwarze 558: pos = 0;
559: continue;
560: default:
561: break;
562: }
563:
564: /*
565: * If input parsers have not been allocated, do so now.
1.2 schwarze 566: * We keep these instanced between parsers, but set them
1.1 schwarze 567: * locally per parse routine since we can use different
568: * parsers with each one.
569: */
570:
571: if ( ! (curp->man || curp->mdoc))
1.60 schwarze 572: choose_parser(curp);
1.1 schwarze 573:
1.25 schwarze 574: /*
1.60 schwarze 575: * Lastly, push down into the parsers themselves.
1.1 schwarze 576: * If libroff returns ROFF_TBL, then add it to the
577: * currently open parse. Since we only get here if
578: * there does exist data (see tbl_data.c), we're
579: * guaranteed that something's been allocated.
580: * Do the same for ROFF_EQN.
581: */
582:
1.76 schwarze 583: if (rr == ROFF_TBL) {
584: while ((span = roff_span(curp->roff)) != NULL)
585: if (curp->man == NULL)
586: mdoc_addspan(curp->mdoc, span);
587: else
588: man_addspan(curp->man, span);
589: } else if (rr == ROFF_EQN) {
590: if (curp->man == NULL)
591: mdoc_addeqn(curp->mdoc, roff_eqn(curp->roff));
592: else
593: man_addeqn(curp->man, roff_eqn(curp->roff));
594: } else if ((curp->man == NULL ?
595: mdoc_parseln(curp->mdoc, curp->line, ln.buf, of) :
596: man_parseln(curp->man, curp->line, ln.buf, of)) == 2)
597: break;
1.1 schwarze 598:
599: /* Temporary buffers typically are not full. */
600:
601: if (0 == start && '\0' == blk.buf[i])
602: break;
603:
604: /* Start the next input line. */
605:
606: pos = 0;
607: }
608:
609: free(ln.buf);
610: }
611:
612: static int
1.18 schwarze 613: read_whole_file(struct mparse *curp, const char *file, int fd,
614: struct buf *fb, int *with_mmap)
1.1 schwarze 615: {
616: struct stat st;
617: size_t off;
618: ssize_t ssz;
619:
620: if (-1 == fstat(fd, &st)) {
1.82 schwarze 621: perror(file);
622: exit((int)MANDOCLEVEL_SYSERR);
1.1 schwarze 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)) {
1.86 schwarze 634: mandoc_msg(MANDOCERR_TOOLARGE, curp, 0, 0, NULL);
1.1 schwarze 635: return(0);
636: }
637: *with_mmap = 1;
638: fb->sz = (size_t)st.st_size;
1.15 schwarze 639: fb->buf = mmap(NULL, fb->sz, PROT_READ, MAP_SHARED, fd, 0);
1.1 schwarze 640: if (fb->buf != MAP_FAILED)
641: return(1);
642: }
643:
644: /*
645: * If this isn't a regular file (like, say, stdin), then we must
646: * go the old way and just read things in bit by bit.
647: */
648:
649: *with_mmap = 0;
650: off = 0;
651: fb->sz = 0;
652: fb->buf = NULL;
653: for (;;) {
654: if (off == fb->sz) {
655: if (fb->sz == (1U << 31)) {
1.86 schwarze 656: mandoc_msg(MANDOCERR_TOOLARGE, curp,
657: 0, 0, NULL);
1.1 schwarze 658: break;
659: }
660: resize_buf(fb, 65536);
661: }
662: ssz = read(fd, fb->buf + (int)off, fb->sz - off);
663: if (ssz == 0) {
664: fb->sz = off;
665: return(1);
666: }
667: if (ssz == -1) {
1.82 schwarze 668: perror(file);
669: exit((int)MANDOCLEVEL_SYSERR);
1.1 schwarze 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:
1.50 schwarze 683: if (curp->mdoc == NULL &&
684: curp->man == NULL &&
685: curp->sodest == NULL) {
686: if (curp->options & MPARSE_MDOC)
687: curp->mdoc = curp->pmdoc;
688: else {
689: if (curp->pman == NULL)
1.81 schwarze 690: curp->pman = man_alloc(
691: curp->roff, curp, curp->defos,
1.50 schwarze 692: curp->options & MPARSE_QUICK ? 1 : 0);
693: curp->man = curp->pman;
694: }
695: }
1.86 schwarze 696: if (curp->mdoc)
697: mdoc_endparse(curp->mdoc);
698: if (curp->man)
699: man_endparse(curp->man);
1.1 schwarze 700: roff_endparse(curp->roff);
701: }
702:
1.15 schwarze 703: static void
704: mparse_parse_buffer(struct mparse *curp, struct buf blk, const char *file)
1.1 schwarze 705: {
1.61 schwarze 706: struct buf *svprimary;
1.1 schwarze 707: const char *svfile;
1.71 schwarze 708: size_t offset;
1.14 schwarze 709: static int recursion_depth;
710:
711: if (64 < recursion_depth) {
712: mandoc_msg(MANDOCERR_ROFFLOOP, curp, curp->line, 0, NULL);
1.15 schwarze 713: return;
1.14 schwarze 714: }
1.1 schwarze 715:
1.15 schwarze 716: /* Line number is per-file. */
717: svfile = curp->file;
718: curp->file = file;
1.61 schwarze 719: svprimary = curp->primary;
1.59 schwarze 720: curp->primary = &blk;
1.15 schwarze 721: curp->line = 1;
722: recursion_depth++;
723:
1.70 schwarze 724: /* Skip an UTF-8 byte order mark. */
725: if (curp->filenc & MPARSE_UTF8 && blk.sz > 2 &&
726: (unsigned char)blk.buf[0] == 0xef &&
727: (unsigned char)blk.buf[1] == 0xbb &&
728: (unsigned char)blk.buf[2] == 0xbf) {
1.71 schwarze 729: offset = 3;
1.70 schwarze 730: curp->filenc &= ~MPARSE_LATIN1;
1.71 schwarze 731: } else
732: offset = 0;
1.70 schwarze 733:
1.71 schwarze 734: mparse_buf_r(curp, blk, offset, 1);
1.15 schwarze 735:
1.86 schwarze 736: if (--recursion_depth == 0)
1.15 schwarze 737: mparse_end(curp);
738:
1.61 schwarze 739: curp->primary = svprimary;
1.15 schwarze 740: curp->file = svfile;
741: }
742:
1.74 schwarze 743: /*
744: * Read the whole file into memory and call the parsers.
745: * Called recursively when an .so request is encountered.
746: */
1.15 schwarze 747: enum mandoclevel
748: mparse_readfd(struct mparse *curp, int fd, const char *file)
749: {
750: struct buf blk;
751: int with_mmap;
1.70 schwarze 752: int save_filenc;
1.1 schwarze 753:
1.67 schwarze 754: if (read_whole_file(curp, file, fd, &blk, &with_mmap)) {
1.70 schwarze 755: save_filenc = curp->filenc;
756: curp->filenc = curp->options &
757: (MPARSE_UTF8 | MPARSE_LATIN1);
1.67 schwarze 758: mparse_parse_buffer(curp, blk, file);
1.70 schwarze 759: curp->filenc = save_filenc;
1.67 schwarze 760: if (with_mmap)
761: munmap(blk.buf, blk.sz);
762: else
763: free(blk.buf);
764: }
1.1 schwarze 765:
1.74 schwarze 766: if (fd != STDIN_FILENO && close(fd) == -1)
1.1 schwarze 767: perror(file);
1.67 schwarze 768:
1.74 schwarze 769: mparse_wait(curp);
1.1 schwarze 770: return(curp->file_status);
1.58 schwarze 771: }
772:
773: enum mandoclevel
1.73 schwarze 774: mparse_open(struct mparse *curp, int *fd, const char *file)
1.58 schwarze 775: {
776: int pfd[2];
1.74 schwarze 777: int save_errno;
1.58 schwarze 778: char *cp;
779:
780: curp->file = file;
1.74 schwarze 781:
782: /* Unless zipped, try to just open the file. */
783:
1.58 schwarze 784: if ((cp = strrchr(file, '.')) == NULL ||
785: strcmp(cp + 1, "gz")) {
1.73 schwarze 786: curp->child = 0;
1.74 schwarze 787: if ((*fd = open(file, O_RDONLY)) != -1)
788: return(MANDOCLEVEL_OK);
789:
790: /* Open failed; try to append ".gz". */
791:
792: mandoc_asprintf(&cp, "%s.gz", file);
793: file = cp;
794: } else
795: cp = NULL;
796:
797: /* Before forking, make sure the file can be read. */
798:
799: save_errno = errno;
800: if (access(file, R_OK) == -1) {
801: if (cp != NULL)
802: errno = save_errno;
1.83 schwarze 803: free(cp);
804: *fd = -1;
805: curp->child = 0;
806: mandoc_msg(MANDOCERR_FILE, curp, 0, 0, strerror(errno));
807: return(MANDOCLEVEL_ERROR);
1.58 schwarze 808: }
809:
1.74 schwarze 810: /* Run gunzip(1). */
811:
1.58 schwarze 812: if (pipe(pfd) == -1) {
1.82 schwarze 813: perror("pipe");
814: exit((int)MANDOCLEVEL_SYSERR);
1.58 schwarze 815: }
816:
1.73 schwarze 817: switch (curp->child = fork()) {
1.58 schwarze 818: case -1:
1.82 schwarze 819: perror("fork");
820: exit((int)MANDOCLEVEL_SYSERR);
1.58 schwarze 821: case 0:
822: close(pfd[0]);
823: if (dup2(pfd[1], STDOUT_FILENO) == -1) {
1.82 schwarze 824: perror("dup");
825: exit((int)MANDOCLEVEL_SYSERR);
1.58 schwarze 826: }
827: execlp("gunzip", "gunzip", "-c", file, NULL);
1.82 schwarze 828: perror("exec");
829: exit((int)MANDOCLEVEL_SYSERR);
1.58 schwarze 830: default:
831: close(pfd[1]);
832: *fd = pfd[0];
833: return(MANDOCLEVEL_OK);
834: }
835: }
836:
837: enum mandoclevel
1.73 schwarze 838: mparse_wait(struct mparse *curp)
1.58 schwarze 839: {
840: int status;
841:
1.73 schwarze 842: if (curp->child == 0)
843: return(MANDOCLEVEL_OK);
844:
845: if (waitpid(curp->child, &status, 0) == -1) {
1.82 schwarze 846: perror("wait");
847: exit((int)MANDOCLEVEL_SYSERR);
1.58 schwarze 848: }
849: if (WIFSIGNALED(status)) {
1.83 schwarze 850: mandoc_vmsg(MANDOCERR_FILE, curp, 0, 0,
851: "gunzip died from signal %d", WTERMSIG(status));
852: return(MANDOCLEVEL_ERROR);
1.58 schwarze 853: }
854: if (WEXITSTATUS(status)) {
1.83 schwarze 855: mandoc_vmsg(MANDOCERR_FILE, curp, 0, 0,
856: "gunzip failed with code %d", WEXITSTATUS(status));
857: return(MANDOCLEVEL_ERROR);
1.58 schwarze 858: }
859: return(MANDOCLEVEL_OK);
1.1 schwarze 860: }
861:
862: struct mparse *
1.69 schwarze 863: mparse_alloc(int options, enum mandoclevel wlevel, mandocmsg mmsg,
864: const struct mchars *mchars, const char *defos)
1.1 schwarze 865: {
866: struct mparse *curp;
867:
868: curp = mandoc_calloc(1, sizeof(struct mparse));
869:
1.22 schwarze 870: curp->options = options;
1.1 schwarze 871: curp->wlevel = wlevel;
872: curp->mmsg = mmsg;
1.7 schwarze 873: curp->defos = defos;
1.1 schwarze 874:
1.69 schwarze 875: curp->mchars = mchars;
876: curp->roff = roff_alloc(curp, curp->mchars, options);
1.50 schwarze 877: if (curp->options & MPARSE_MDOC)
878: curp->pmdoc = mdoc_alloc(
879: curp->roff, curp, curp->defos,
880: curp->options & MPARSE_QUICK ? 1 : 0);
881: if (curp->options & MPARSE_MAN)
1.81 schwarze 882: curp->pman = man_alloc(
883: curp->roff, curp, curp->defos,
1.50 schwarze 884: curp->options & MPARSE_QUICK ? 1 : 0);
885:
1.1 schwarze 886: return(curp);
887: }
888:
889: void
890: mparse_reset(struct mparse *curp)
891: {
892:
893: roff_reset(curp->roff);
894:
895: if (curp->mdoc)
896: mdoc_reset(curp->mdoc);
897: if (curp->man)
898: man_reset(curp->man);
1.4 schwarze 899: if (curp->secondary)
900: curp->secondary->sz = 0;
1.1 schwarze 901:
902: curp->file_status = MANDOCLEVEL_OK;
903: curp->mdoc = NULL;
904: curp->man = NULL;
1.23 schwarze 905:
906: free(curp->sodest);
907: curp->sodest = NULL;
1.1 schwarze 908: }
909:
910: void
911: mparse_free(struct mparse *curp)
912: {
913:
914: if (curp->pmdoc)
915: mdoc_free(curp->pmdoc);
916: if (curp->pman)
917: man_free(curp->pman);
918: if (curp->roff)
919: roff_free(curp->roff);
1.4 schwarze 920: if (curp->secondary)
921: free(curp->secondary->buf);
1.1 schwarze 922:
1.4 schwarze 923: free(curp->secondary);
1.23 schwarze 924: free(curp->sodest);
1.1 schwarze 925: free(curp);
926: }
927:
928: void
1.23 schwarze 929: mparse_result(struct mparse *curp,
930: struct mdoc **mdoc, struct man **man, char **sodest)
1.1 schwarze 931: {
932:
1.23 schwarze 933: if (sodest && NULL != (*sodest = curp->sodest)) {
934: *mdoc = NULL;
935: *man = NULL;
936: return;
937: }
1.1 schwarze 938: if (mdoc)
939: *mdoc = curp->mdoc;
940: if (man)
941: *man = curp->man;
942: }
943:
944: void
945: mandoc_vmsg(enum mandocerr t, struct mparse *m,
946: int ln, int pos, const char *fmt, ...)
947: {
948: char buf[256];
949: va_list ap;
950:
951: va_start(ap, fmt);
1.26 schwarze 952: (void)vsnprintf(buf, sizeof(buf), fmt, ap);
1.1 schwarze 953: va_end(ap);
954:
955: mandoc_msg(t, m, ln, pos, buf);
956: }
957:
958: void
1.25 schwarze 959: mandoc_msg(enum mandocerr er, struct mparse *m,
1.1 schwarze 960: int ln, int col, const char *msg)
961: {
962: enum mandoclevel level;
963:
1.87 schwarze 964: level = MANDOCLEVEL_UNSUPP;
1.1 schwarze 965: while (er < mandoclimits[level])
966: level--;
967:
1.83 schwarze 968: if (level < m->wlevel && er != MANDOCERR_FILE)
1.1 schwarze 969: return;
970:
971: if (m->mmsg)
972: (*m->mmsg)(er, level, m->file, ln, col, msg);
973:
974: if (m->file_status < level)
975: m->file_status = level;
976: }
977:
978: const char *
979: mparse_strerror(enum mandocerr er)
980: {
981:
982: return(mandocerrs[er]);
983: }
984:
985: const char *
986: mparse_strlevel(enum mandoclevel lvl)
987: {
988: return(mandoclevels[lvl]);
1.4 schwarze 989: }
990:
991: void
992: mparse_keep(struct mparse *p)
993: {
994:
995: assert(NULL == p->secondary);
996: p->secondary = mandoc_calloc(1, sizeof(struct buf));
997: }
998:
999: const char *
1000: mparse_getkeep(const struct mparse *p)
1001: {
1002:
1003: assert(p->secondary);
1004: return(p->secondary->sz ? p->secondary->buf : NULL);
1.1 schwarze 1005: }