Annotation of mandoc/read.c, Revision 1.92
1.92 ! kristaps 1: /* $Id: read.c,v 1.91 2014/10/18 15:57:34 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",
1.87 schwarze 120: "unusual Xr order",
121: "unusual Xr punctuation",
1.86 schwarze 122: "AUTHORS section without An macro",
1.7 kristaps 123:
124: /* related to macros and nesting */
1.55 schwarze 125: "obsolete macro",
1.7 kristaps 126: "skipping paragraph macro",
1.31 schwarze 127: "moving paragraph macro out of list",
1.7 kristaps 128: "skipping no-space macro",
129: "blocks badly nested",
130: "nested displays are not portable",
1.57 schwarze 131: "moving content out of list",
132: ".Vt block has child macro",
1.78 schwarze 133: "fill mode already enabled, skipping",
134: "fill mode already disabled, skipping",
1.7 kristaps 135: "line scope broken",
136:
137: /* related to missing macro arguments */
1.58 schwarze 138: "skipping empty request",
139: "conditional request controls empty scope",
1.7 kristaps 140: "skipping empty macro",
1.62 schwarze 141: "empty argument, using 0n",
1.7 kristaps 142: "argument count wrong",
1.60 schwarze 143: "missing display type, using -ragged",
144: "list type is not the first argument",
145: "missing -width in -tag list, using 8n",
1.78 schwarze 146: "missing utility name, using \"\"",
1.60 schwarze 147: "empty head in list item",
148: "empty list item",
1.61 schwarze 149: "missing font type, using \\fR",
150: "unknown font type, using \\fR",
1.60 schwarze 151: "missing -std argument, adding it",
1.90 schwarze 152: "missing eqn box, using \"\"",
1.7 kristaps 153:
154: /* related to bad macro arguments */
1.64 schwarze 155: "unterminated quoted argument",
1.7 kristaps 156: "duplicate argument",
1.76 schwarze 157: "skipping duplicate argument",
1.63 schwarze 158: "skipping duplicate display type",
159: "skipping duplicate list type",
1.76 schwarze 160: "skipping -width argument",
1.7 kristaps 161: "unknown AT&T UNIX version",
1.88 schwarze 162: "comma in function argument",
1.89 schwarze 163: "parenthesis in function name",
1.67 schwarze 164: "invalid content in Rs block",
1.63 schwarze 165: "invalid Boolean argument",
166: "unknown font, skipping request",
1.7 kristaps 167:
168: /* related to plain text */
1.64 schwarze 169: "blank line in fill mode, using .sp",
170: "tab in filled text",
171: "whitespace at end of input line",
1.7 kristaps 172: "bad comment style",
1.64 schwarze 173: "invalid escape sequence",
174: "undefined string, using \"\"",
1.16 kristaps 175:
1.7 kristaps 176: "generic error",
1.17 kristaps 177:
178: /* related to equations */
1.20 kristaps 179: "unexpected equation scope closure",
180: "equation scope open on exit",
1.21 kristaps 181: "overlapping equation scopes",
182: "unexpected end of equation",
1.7 kristaps 183:
184: /* related to tables */
185: "bad table syntax",
186: "bad table option",
187: "bad table layout",
188: "no table layout cells specified",
189: "no table data cells specified",
190: "ignore data in cell",
191: "data block still open",
192: "ignoring extra data cells",
193:
1.68 schwarze 194: /* related to document structure and macros */
1.7 kristaps 195: "input stack limit exceeded, infinite loop?",
196: "skipping bad character",
1.68 schwarze 197: "skipping unknown macro",
1.73 schwarze 198: "skipping item outside list",
1.68 schwarze 199: "skipping column outside column list",
200: "skipping end of block that is not open",
201: "inserting missing end of block",
202: "appending missing end of block",
203:
204: /* related to request and macro arguments */
1.7 kristaps 205: "escaped character not allowed in a name",
206: "argument count wrong",
1.71 schwarze 207: "missing list type, using -item",
1.70 schwarze 208: "missing manual name, using \"\"",
1.71 schwarze 209: "uname(3) system call failed, using UNKNOWN",
1.63 schwarze 210: "unknown standard specifier",
1.71 schwarze 211: "skipping request without numeric argument",
1.61 schwarze 212: "skipping all arguments",
213: "skipping excess arguments",
1.92 ! kristaps 214: "divide by zero",
1.7 kristaps 215:
216: "generic fatal error",
217:
1.40 schwarze 218: "input too large",
1.78 schwarze 219: "NOT IMPLEMENTED: Bd -file",
1.7 kristaps 220: "NOT IMPLEMENTED: .so with absolute path or \"..\"",
1.52 schwarze 221: ".so request failed",
1.40 schwarze 222:
223: /* system errors */
1.82 schwarze 224: "cannot dup file descriptor",
225: "cannot exec",
226: "gunzip failed with code",
227: "cannot fork",
1.51 schwarze 228: NULL,
1.82 schwarze 229: "cannot open pipe",
230: "cannot read file",
231: "gunzip died from signal",
1.40 schwarze 232: "cannot stat file",
1.82 schwarze 233: "wait failed",
1.7 kristaps 234: };
235:
236: static const char * const mandoclevels[MANDOCLEVEL_MAX] = {
237: "SUCCESS",
238: "RESERVED",
239: "WARNING",
240: "ERROR",
241: "FATAL",
242: "BADARG",
243: "SYSERR"
244: };
245:
1.47 schwarze 246:
1.1 kristaps 247: static void
248: resize_buf(struct buf *buf, size_t initial)
249: {
250:
251: buf->sz = buf->sz > initial/2 ? 2 * buf->sz : initial;
252: buf->buf = mandoc_realloc(buf->buf, buf->sz);
253: }
254:
255: static void
1.84 schwarze 256: choose_parser(struct mparse *curp)
1.1 kristaps 257: {
1.83 schwarze 258: char *cp, *ep;
259: int format;
1.1 kristaps 260:
1.83 schwarze 261: /*
262: * If neither command line arguments -mdoc or -man select
263: * a parser nor the roff parser found a .Dd or .TH macro
264: * yet, look ahead in the main input buffer.
265: */
266:
267: if ((format = roff_getformat(curp->roff)) == 0) {
268: cp = curp->primary->buf;
269: ep = cp + curp->primary->sz;
270: while (cp < ep) {
1.85 schwarze 271: if (*cp == '.' || *cp == '\'') {
1.83 schwarze 272: cp++;
273: if (cp[0] == 'D' && cp[1] == 'd') {
274: format = MPARSE_MDOC;
275: break;
276: }
277: if (cp[0] == 'T' && cp[1] == 'H') {
278: format = MPARSE_MAN;
279: break;
280: }
281: }
282: cp = memchr(cp, '\n', ep - cp);
283: if (cp == NULL)
284: break;
285: cp++;
286: }
1.1 kristaps 287: }
288:
1.83 schwarze 289: if (format == MPARSE_MDOC) {
1.47 schwarze 290: if (NULL == curp->pmdoc)
1.44 schwarze 291: curp->pmdoc = mdoc_alloc(
292: curp->roff, curp, curp->defos,
293: MPARSE_QUICK & curp->options ? 1 : 0);
1.1 kristaps 294: assert(curp->pmdoc);
295: curp->mdoc = curp->pmdoc;
296: return;
1.47 schwarze 297: }
1.1 kristaps 298:
1.83 schwarze 299: /* Fall back to man(7) as a last resort. */
300:
1.47 schwarze 301: if (NULL == curp->pman)
1.44 schwarze 302: curp->pman = man_alloc(curp->roff, curp,
303: MPARSE_QUICK & curp->options ? 1 : 0);
1.1 kristaps 304: assert(curp->pman);
305: curp->man = curp->pman;
306: }
307:
308: /*
309: * Main parse routine for an opened file. This is called for each
310: * opened file and simply loops around the full input file, possibly
311: * nesting (i.e., with `so').
312: */
313: static void
314: mparse_buf_r(struct mparse *curp, struct buf blk, int start)
315: {
316: const struct tbl_span *span;
317: struct buf ln;
318: enum rofferr rr;
319: int i, of, rc;
320: int pos; /* byte number in the ln buffer */
321: int lnn; /* line number in the real file */
322: unsigned char c;
323:
324: memset(&ln, 0, sizeof(struct buf));
325:
1.47 schwarze 326: lnn = curp->line;
327: pos = 0;
1.1 kristaps 328:
329: for (i = 0; i < (int)blk.sz; ) {
330: if (0 == pos && '\0' == blk.buf[i])
331: break;
332:
333: if (start) {
334: curp->line = lnn;
335: curp->reparse_count = 0;
336: }
337:
338: while (i < (int)blk.sz && (start || '\0' != blk.buf[i])) {
339:
340: /*
341: * When finding an unescaped newline character,
342: * leave the character loop to process the line.
343: * Skip a preceding carriage return, if any.
344: */
345:
346: if ('\r' == blk.buf[i] && i + 1 < (int)blk.sz &&
347: '\n' == blk.buf[i + 1])
348: ++i;
349: if ('\n' == blk.buf[i]) {
350: ++i;
351: ++lnn;
352: break;
353: }
354:
1.35 schwarze 355: /*
356: * Make sure we have space for at least
357: * one backslash and one other character
358: * and the trailing NUL byte.
359: */
360:
361: if (pos + 2 >= (int)ln.sz)
362: resize_buf(&ln, 256);
363:
1.47 schwarze 364: /*
1.1 kristaps 365: * Warn about bogus characters. If you're using
366: * non-ASCII encoding, you're screwing your
367: * readers. Since I'd rather this not happen,
1.27 joerg 368: * I'll be helpful and replace these characters
369: * with "?", so we don't display gibberish.
370: * Note to manual writers: use special characters.
1.1 kristaps 371: */
372:
373: c = (unsigned char) blk.buf[i];
374:
1.47 schwarze 375: if ( ! (isascii(c) &&
376: (isgraph(c) || isblank(c)))) {
1.78 schwarze 377: mandoc_vmsg(MANDOCERR_BADCHAR, curp,
378: curp->line, pos, "0x%x", c);
1.1 kristaps 379: i++;
1.27 joerg 380: ln.buf[pos++] = '?';
1.1 kristaps 381: continue;
382: }
383:
384: /* Trailing backslash = a plain char. */
385:
386: if ('\\' != blk.buf[i] || i + 1 == (int)blk.sz) {
387: ln.buf[pos++] = blk.buf[i++];
388: continue;
389: }
390:
391: /*
392: * Found escape and at least one other character.
393: * When it's a newline character, skip it.
394: * When there is a carriage return in between,
395: * skip that one as well.
396: */
397:
398: if ('\r' == blk.buf[i + 1] && i + 2 < (int)blk.sz &&
399: '\n' == blk.buf[i + 2])
400: ++i;
401: if ('\n' == blk.buf[i + 1]) {
402: i += 2;
403: ++lnn;
404: continue;
405: }
406:
1.13 kristaps 407: if ('"' == blk.buf[i + 1] || '#' == blk.buf[i + 1]) {
1.1 kristaps 408: i += 2;
409: /* Comment, skip to end of line */
410: for (; i < (int)blk.sz; ++i) {
411: if ('\n' == blk.buf[i]) {
412: ++i;
413: ++lnn;
414: break;
415: }
416: }
417:
418: /* Backout trailing whitespaces */
419: for (; pos > 0; --pos) {
420: if (ln.buf[pos - 1] != ' ')
421: break;
422: if (pos > 2 && ln.buf[pos - 2] == '\\')
423: break;
424: }
425: break;
426: }
427:
1.35 schwarze 428: /* Catch escaped bogus characters. */
429:
430: c = (unsigned char) blk.buf[i+1];
431:
1.47 schwarze 432: if ( ! (isascii(c) &&
433: (isgraph(c) || isblank(c)))) {
1.78 schwarze 434: mandoc_vmsg(MANDOCERR_BADCHAR, curp,
435: curp->line, pos, "0x%x", c);
1.35 schwarze 436: i += 2;
437: ln.buf[pos++] = '?';
438: continue;
439: }
440:
1.1 kristaps 441: /* Some other escape sequence, copy & cont. */
442:
443: ln.buf[pos++] = blk.buf[i++];
444: ln.buf[pos++] = blk.buf[i++];
445: }
446:
1.47 schwarze 447: if (pos >= (int)ln.sz)
1.1 kristaps 448: resize_buf(&ln, 256);
449:
450: ln.buf[pos] = '\0';
451:
452: /*
453: * A significant amount of complexity is contained by
454: * the roff preprocessor. It's line-oriented but can be
455: * expressed on one line, so we need at times to
456: * readjust our starting point and re-run it. The roff
457: * preprocessor can also readjust the buffers with new
458: * data, so we pass them in wholesale.
459: */
460:
461: of = 0;
462:
1.24 kristaps 463: /*
464: * Maintain a lookaside buffer of all parsed lines. We
465: * only do this if mparse_keep() has been invoked (the
466: * buffer may be accessed with mparse_getkeep()).
467: */
468:
469: if (curp->secondary) {
1.47 schwarze 470: curp->secondary->buf = mandoc_realloc(
471: curp->secondary->buf,
472: curp->secondary->sz + pos + 2);
473: memcpy(curp->secondary->buf +
474: curp->secondary->sz,
475: ln.buf, pos);
1.24 kristaps 476: curp->secondary->sz += pos;
477: curp->secondary->buf
478: [curp->secondary->sz] = '\n';
479: curp->secondary->sz++;
480: curp->secondary->buf
481: [curp->secondary->sz] = '\0';
482: }
1.1 kristaps 483: rerun:
1.47 schwarze 484: rr = roff_parseln(curp->roff, curp->line,
485: &ln.buf, &ln.sz, of, &of);
1.1 kristaps 486:
487: switch (rr) {
1.47 schwarze 488: case ROFF_REPARSE:
1.1 kristaps 489: if (REPARSE_LIMIT >= ++curp->reparse_count)
490: mparse_buf_r(curp, ln, 0);
491: else
1.3 kristaps 492: mandoc_msg(MANDOCERR_ROFFLOOP, curp,
1.47 schwarze 493: curp->line, pos, NULL);
1.1 kristaps 494: pos = 0;
495: continue;
1.47 schwarze 496: case ROFF_APPEND:
1.1 kristaps 497: pos = (int)strlen(ln.buf);
498: continue;
1.47 schwarze 499: case ROFF_RERUN:
1.1 kristaps 500: goto rerun;
1.47 schwarze 501: case ROFF_IGN:
1.1 kristaps 502: pos = 0;
503: continue;
1.47 schwarze 504: case ROFF_ERR:
1.1 kristaps 505: assert(MANDOCLEVEL_FATAL <= curp->file_status);
506: break;
1.47 schwarze 507: case ROFF_SO:
1.45 schwarze 508: if (0 == (MPARSE_SO & curp->options) &&
509: (i >= (int)blk.sz || '\0' == blk.buf[i])) {
510: curp->sodest = mandoc_strdup(ln.buf + of);
511: free(ln.buf);
512: return;
513: }
1.24 kristaps 514: /*
515: * We remove `so' clauses from our lookaside
516: * buffer because we're going to descend into
517: * the file recursively.
518: */
1.47 schwarze 519: if (curp->secondary)
1.25 kristaps 520: curp->secondary->sz -= pos + 1;
1.36 schwarze 521: mparse_readfd(curp, -1, ln.buf + of);
1.52 schwarze 522: if (MANDOCLEVEL_FATAL <= curp->file_status) {
523: mandoc_vmsg(MANDOCERR_SO_FAIL,
524: curp, curp->line, pos,
525: ".so %s", ln.buf + of);
1.1 kristaps 526: break;
1.52 schwarze 527: }
1.1 kristaps 528: pos = 0;
529: continue;
530: default:
531: break;
532: }
533:
534: /*
535: * If we encounter errors in the recursive parse, make
536: * sure we don't continue parsing.
537: */
538:
539: if (MANDOCLEVEL_FATAL <= curp->file_status)
540: break;
541:
542: /*
543: * If input parsers have not been allocated, do so now.
1.14 kristaps 544: * We keep these instanced between parsers, but set them
1.1 kristaps 545: * locally per parse routine since we can use different
546: * parsers with each one.
547: */
548:
549: if ( ! (curp->man || curp->mdoc))
1.84 schwarze 550: choose_parser(curp);
1.1 kristaps 551:
1.47 schwarze 552: /*
1.84 schwarze 553: * Lastly, push down into the parsers themselves.
1.1 kristaps 554: * If libroff returns ROFF_TBL, then add it to the
555: * currently open parse. Since we only get here if
556: * there does exist data (see tbl_data.c), we're
557: * guaranteed that something's been allocated.
558: * Do the same for ROFF_EQN.
559: */
560:
561: rc = -1;
562:
563: if (ROFF_TBL == rr)
564: while (NULL != (span = roff_span(curp->roff))) {
565: rc = curp->man ?
1.47 schwarze 566: man_addspan(curp->man, span) :
567: mdoc_addspan(curp->mdoc, span);
1.1 kristaps 568: if (0 == rc)
569: break;
570: }
571: else if (ROFF_EQN == rr)
1.47 schwarze 572: rc = curp->mdoc ?
573: mdoc_addeqn(curp->mdoc,
574: roff_eqn(curp->roff)) :
575: man_addeqn(curp->man,
576: roff_eqn(curp->roff));
1.1 kristaps 577: else if (curp->man || curp->mdoc)
578: rc = curp->man ?
1.47 schwarze 579: man_parseln(curp->man,
580: curp->line, ln.buf, of) :
581: mdoc_parseln(curp->mdoc,
582: curp->line, ln.buf, of);
1.1 kristaps 583:
584: if (0 == rc) {
585: assert(MANDOCLEVEL_FATAL <= curp->file_status);
586: break;
1.41 schwarze 587: } else if (2 == rc)
588: break;
1.1 kristaps 589:
590: /* Temporary buffers typically are not full. */
591:
592: if (0 == start && '\0' == blk.buf[i])
593: break;
594:
595: /* Start the next input line. */
596:
597: pos = 0;
598: }
599:
600: free(ln.buf);
601: }
602:
603: static int
1.40 schwarze 604: read_whole_file(struct mparse *curp, const char *file, int fd,
605: struct buf *fb, int *with_mmap)
1.1 kristaps 606: {
607: size_t off;
608: ssize_t ssz;
609:
1.81 schwarze 610: #if HAVE_MMAP
1.15 kristaps 611: struct stat st;
1.1 kristaps 612: if (-1 == fstat(fd, &st)) {
1.40 schwarze 613: curp->file_status = MANDOCLEVEL_SYSERR;
614: if (curp->mmsg)
615: (*curp->mmsg)(MANDOCERR_SYSSTAT, curp->file_status,
616: file, 0, 0, strerror(errno));
1.1 kristaps 617: return(0);
618: }
619:
620: /*
621: * If we're a regular file, try just reading in the whole entry
622: * via mmap(). This is faster than reading it into blocks, and
623: * since each file is only a few bytes to begin with, I'm not
624: * concerned that this is going to tank any machines.
625: */
626:
627: if (S_ISREG(st.st_mode)) {
628: if (st.st_size >= (1U << 31)) {
1.40 schwarze 629: curp->file_status = MANDOCLEVEL_FATAL;
630: if (curp->mmsg)
631: (*curp->mmsg)(MANDOCERR_TOOLARGE,
632: curp->file_status, file, 0, 0, NULL);
1.1 kristaps 633: return(0);
634: }
635: *with_mmap = 1;
636: fb->sz = (size_t)st.st_size;
1.37 schwarze 637: fb->buf = mmap(NULL, fb->sz, PROT_READ, MAP_SHARED, fd, 0);
1.1 kristaps 638: if (fb->buf != MAP_FAILED)
639: return(1);
640: }
1.15 kristaps 641: #endif
1.1 kristaps 642:
643: /*
644: * If this isn't a regular file (like, say, stdin), then we must
645: * go the old way and just read things in bit by bit.
646: */
647:
648: *with_mmap = 0;
649: off = 0;
650: fb->sz = 0;
651: fb->buf = NULL;
652: for (;;) {
653: if (off == fb->sz) {
654: if (fb->sz == (1U << 31)) {
1.40 schwarze 655: curp->file_status = MANDOCLEVEL_FATAL;
656: if (curp->mmsg)
657: (*curp->mmsg)(MANDOCERR_TOOLARGE,
658: curp->file_status,
659: file, 0, 0, NULL);
1.1 kristaps 660: break;
661: }
662: resize_buf(fb, 65536);
663: }
664: ssz = read(fd, fb->buf + (int)off, fb->sz - off);
665: if (ssz == 0) {
666: fb->sz = off;
667: return(1);
668: }
669: if (ssz == -1) {
1.40 schwarze 670: curp->file_status = MANDOCLEVEL_SYSERR;
671: if (curp->mmsg)
672: (*curp->mmsg)(MANDOCERR_SYSREAD,
673: curp->file_status, file, 0, 0,
674: strerror(errno));
1.1 kristaps 675: break;
676: }
677: off += (size_t)ssz;
678: }
679:
680: free(fb->buf);
681: fb->buf = NULL;
682: return(0);
683: }
684:
685: static void
686: mparse_end(struct mparse *curp)
687: {
688:
689: if (MANDOCLEVEL_FATAL <= curp->file_status)
690: return;
691:
1.72 schwarze 692: if (curp->mdoc == NULL &&
693: curp->man == NULL &&
694: curp->sodest == NULL) {
695: if (curp->options & MPARSE_MDOC)
696: curp->mdoc = curp->pmdoc;
697: else {
698: if (curp->pman == NULL)
699: curp->pman = man_alloc(curp->roff, curp,
700: curp->options & MPARSE_QUICK ? 1 : 0);
701: curp->man = curp->pman;
702: }
703: }
704:
1.1 kristaps 705: if (curp->mdoc && ! mdoc_endparse(curp->mdoc)) {
706: assert(MANDOCLEVEL_FATAL <= curp->file_status);
707: return;
708: }
709:
710: if (curp->man && ! man_endparse(curp->man)) {
711: assert(MANDOCLEVEL_FATAL <= curp->file_status);
712: return;
713: }
714:
715: roff_endparse(curp->roff);
716: }
717:
718: static void
1.36 schwarze 719: mparse_parse_buffer(struct mparse *curp, struct buf blk, const char *file)
1.28 joerg 720: {
1.85 schwarze 721: struct buf *svprimary;
1.28 joerg 722: const char *svfile;
1.36 schwarze 723: static int recursion_depth;
724:
725: if (64 < recursion_depth) {
726: mandoc_msg(MANDOCERR_ROFFLOOP, curp, curp->line, 0, NULL);
727: return;
728: }
1.28 joerg 729:
730: /* Line number is per-file. */
731: svfile = curp->file;
732: curp->file = file;
1.85 schwarze 733: svprimary = curp->primary;
1.83 schwarze 734: curp->primary = &blk;
1.28 joerg 735: curp->line = 1;
1.36 schwarze 736: recursion_depth++;
1.28 joerg 737:
738: mparse_buf_r(curp, blk, 1);
739:
1.36 schwarze 740: if (0 == --recursion_depth && MANDOCLEVEL_FATAL > curp->file_status)
1.28 joerg 741: mparse_end(curp);
742:
1.85 schwarze 743: curp->primary = svprimary;
1.28 joerg 744: curp->file = svfile;
745: }
746:
747: enum mandoclevel
748: mparse_readmem(struct mparse *curp, const void *buf, size_t len,
749: const char *file)
750: {
751: struct buf blk;
752:
753: blk.buf = UNCONST(buf);
754: blk.sz = len;
755:
1.36 schwarze 756: mparse_parse_buffer(curp, blk, file);
1.28 joerg 757: return(curp->file_status);
758: }
759:
1.36 schwarze 760: enum mandoclevel
761: mparse_readfd(struct mparse *curp, int fd, const char *file)
1.1 kristaps 762: {
1.28 joerg 763: struct buf blk;
764: int with_mmap;
1.1 kristaps 765:
1.40 schwarze 766: if (-1 == fd && -1 == (fd = open(file, O_RDONLY, 0))) {
767: curp->file_status = MANDOCLEVEL_SYSERR;
768: if (curp->mmsg)
769: (*curp->mmsg)(MANDOCERR_SYSOPEN,
770: curp->file_status,
771: file, 0, 0, strerror(errno));
1.91 schwarze 772: return(curp->file_status);
1.40 schwarze 773: }
774:
1.28 joerg 775: /*
776: * Run for each opened file; may be called more than once for
777: * each full parse sequence if the opened file is nested (i.e.,
778: * from `so'). Simply sucks in the whole file and moves into
779: * the parse phase for the file.
780: */
1.1 kristaps 781:
1.91 schwarze 782: if (read_whole_file(curp, file, fd, &blk, &with_mmap)) {
783: mparse_parse_buffer(curp, blk, file);
1.81 schwarze 784: #if HAVE_MMAP
1.91 schwarze 785: if (with_mmap)
786: munmap(blk.buf, blk.sz);
787: else
1.28 joerg 788: #endif
1.91 schwarze 789: free(blk.buf);
790: }
1.1 kristaps 791:
792: if (STDIN_FILENO != fd && -1 == close(fd))
793: perror(file);
1.91 schwarze 794:
1.1 kristaps 795: return(curp->file_status);
1.82 schwarze 796: }
797:
798: enum mandoclevel
799: mparse_open(struct mparse *curp, int *fd, const char *file,
800: pid_t *child_pid)
801: {
802: int pfd[2];
803: char *cp;
804: enum mandocerr err;
805:
806: pfd[1] = -1;
807: curp->file = file;
808: if ((cp = strrchr(file, '.')) == NULL ||
809: strcmp(cp + 1, "gz")) {
810: *child_pid = 0;
811: if ((*fd = open(file, O_RDONLY)) == -1) {
812: err = MANDOCERR_SYSOPEN;
813: goto out;
814: }
815: return(MANDOCLEVEL_OK);
816: }
817:
818: if (pipe(pfd) == -1) {
819: err = MANDOCERR_SYSPIPE;
820: goto out;
821: }
822:
823: switch (*child_pid = fork()) {
824: case -1:
825: err = MANDOCERR_SYSFORK;
826: close(pfd[0]);
827: close(pfd[1]);
828: pfd[1] = -1;
829: break;
830: case 0:
831: close(pfd[0]);
832: if (dup2(pfd[1], STDOUT_FILENO) == -1) {
833: err = MANDOCERR_SYSDUP;
834: break;
835: }
836: execlp("gunzip", "gunzip", "-c", file, NULL);
837: err = MANDOCERR_SYSEXEC;
838: break;
839: default:
840: close(pfd[1]);
841: *fd = pfd[0];
842: return(MANDOCLEVEL_OK);
843: }
844:
845: out:
846: *fd = -1;
847: *child_pid = 0;
848: curp->file_status = MANDOCLEVEL_SYSERR;
849: if (curp->mmsg)
850: (*curp->mmsg)(err, curp->file_status, file,
851: 0, 0, strerror(errno));
852: if (pfd[1] != -1)
853: exit(1);
854: return(curp->file_status);
855: }
856:
857: enum mandoclevel
858: mparse_wait(struct mparse *curp, pid_t child_pid)
859: {
860: int status;
861:
862: if (waitpid(child_pid, &status, 0) == -1) {
863: mandoc_msg(MANDOCERR_SYSWAIT, curp, 0, 0,
864: strerror(errno));
865: curp->file_status = MANDOCLEVEL_SYSERR;
866: return(curp->file_status);
867: }
868: if (WIFSIGNALED(status)) {
869: mandoc_vmsg(MANDOCERR_SYSSIG, curp, 0, 0,
870: "%d", WTERMSIG(status));
871: curp->file_status = MANDOCLEVEL_SYSERR;
872: return(curp->file_status);
873: }
874: if (WEXITSTATUS(status)) {
875: mandoc_vmsg(MANDOCERR_SYSEXIT, curp, 0, 0,
876: "%d", WEXITSTATUS(status));
877: curp->file_status = MANDOCLEVEL_SYSERR;
878: return(curp->file_status);
879: }
880: return(MANDOCLEVEL_OK);
1.1 kristaps 881: }
882:
883: struct mparse *
1.44 schwarze 884: mparse_alloc(int options, enum mandoclevel wlevel,
1.69 schwarze 885: mandocmsg mmsg, const char *defos)
1.1 kristaps 886: {
887: struct mparse *curp;
1.10 kristaps 888:
889: assert(wlevel <= MANDOCLEVEL_FATAL);
1.1 kristaps 890:
891: curp = mandoc_calloc(1, sizeof(struct mparse));
892:
1.44 schwarze 893: curp->options = options;
1.3 kristaps 894: curp->wlevel = wlevel;
1.1 kristaps 895: curp->mmsg = mmsg;
1.29 schwarze 896: curp->defos = defos;
1.1 kristaps 897:
1.44 schwarze 898: curp->roff = roff_alloc(curp, options);
1.72 schwarze 899: if (curp->options & MPARSE_MDOC)
900: curp->pmdoc = mdoc_alloc(
901: curp->roff, curp, curp->defos,
902: curp->options & MPARSE_QUICK ? 1 : 0);
903: if (curp->options & MPARSE_MAN)
904: curp->pman = man_alloc(curp->roff, curp,
905: curp->options & MPARSE_QUICK ? 1 : 0);
906:
1.1 kristaps 907: return(curp);
908: }
909:
910: void
911: mparse_reset(struct mparse *curp)
912: {
913:
914: roff_reset(curp->roff);
915:
916: if (curp->mdoc)
917: mdoc_reset(curp->mdoc);
918: if (curp->man)
919: man_reset(curp->man);
1.24 kristaps 920: if (curp->secondary)
921: curp->secondary->sz = 0;
1.1 kristaps 922:
923: curp->file_status = MANDOCLEVEL_OK;
924: curp->mdoc = NULL;
925: curp->man = NULL;
1.45 schwarze 926:
927: free(curp->sodest);
928: curp->sodest = NULL;
1.1 kristaps 929: }
930:
931: void
932: mparse_free(struct mparse *curp)
933: {
934:
935: if (curp->pmdoc)
936: mdoc_free(curp->pmdoc);
937: if (curp->pman)
938: man_free(curp->pman);
939: if (curp->roff)
940: roff_free(curp->roff);
1.24 kristaps 941: if (curp->secondary)
942: free(curp->secondary->buf);
1.1 kristaps 943:
1.24 kristaps 944: free(curp->secondary);
1.45 schwarze 945: free(curp->sodest);
1.1 kristaps 946: free(curp);
947: }
948:
949: void
1.45 schwarze 950: mparse_result(struct mparse *curp,
951: struct mdoc **mdoc, struct man **man, char **sodest)
1.1 kristaps 952: {
953:
1.45 schwarze 954: if (sodest && NULL != (*sodest = curp->sodest)) {
955: *mdoc = NULL;
956: *man = NULL;
957: return;
958: }
1.9 kristaps 959: if (mdoc)
960: *mdoc = curp->mdoc;
961: if (man)
962: *man = curp->man;
1.3 kristaps 963: }
964:
965: void
966: mandoc_vmsg(enum mandocerr t, struct mparse *m,
967: int ln, int pos, const char *fmt, ...)
968: {
969: char buf[256];
970: va_list ap;
971:
972: va_start(ap, fmt);
1.48 schwarze 973: (void)vsnprintf(buf, sizeof(buf), fmt, ap);
1.3 kristaps 974: va_end(ap);
975:
976: mandoc_msg(t, m, ln, pos, buf);
977: }
978:
979: void
1.47 schwarze 980: mandoc_msg(enum mandocerr er, struct mparse *m,
1.3 kristaps 981: int ln, int col, const char *msg)
982: {
983: enum mandoclevel level;
984:
985: level = MANDOCLEVEL_FATAL;
986: while (er < mandoclimits[level])
987: level--;
988:
989: if (level < m->wlevel)
990: return;
991:
1.8 kristaps 992: if (m->mmsg)
993: (*m->mmsg)(er, level, m->file, ln, col, msg);
1.3 kristaps 994:
995: if (m->file_status < level)
996: m->file_status = level;
1.7 kristaps 997: }
998:
999: const char *
1000: mparse_strerror(enum mandocerr er)
1001: {
1002:
1003: return(mandocerrs[er]);
1004: }
1005:
1006: const char *
1007: mparse_strlevel(enum mandoclevel lvl)
1008: {
1009: return(mandoclevels[lvl]);
1.24 kristaps 1010: }
1011:
1012: void
1013: mparse_keep(struct mparse *p)
1014: {
1015:
1016: assert(NULL == p->secondary);
1017: p->secondary = mandoc_calloc(1, sizeof(struct buf));
1018: }
1019:
1020: const char *
1021: mparse_getkeep(const struct mparse *p)
1022: {
1023:
1024: assert(p->secondary);
1025: return(p->secondary->sz ? p->secondary->buf : NULL);
1.1 kristaps 1026: }
CVSweb