Annotation of mandoc/read.c, Revision 1.79
1.79 ! schwarze 1: /* $Id: read.c,v 1.78 2014/08/01 17:27:44 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: #ifdef HAVE_CONFIG_H
20: #include "config.h"
21: #endif
22:
1.15 kristaps 23: #ifdef HAVE_MMAP
24: # include <sys/stat.h>
25: # include <sys/mman.h>
26: #endif
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: enum mandoclevel file_status; /* status of current parse */
1.3 kristaps 55: enum mandoclevel wlevel; /* ignore messages below this */
1.1 kristaps 56: int line; /* line number in the file */
1.44 schwarze 57: int options; /* parser options */
1.1 kristaps 58: struct man *pman; /* persistent man parser */
59: struct mdoc *pmdoc; /* persistent mdoc parser */
60: struct man *man; /* man parser */
61: struct mdoc *mdoc; /* mdoc parser */
62: struct roff *roff; /* roff parser (!NULL) */
1.45 schwarze 63: char *sodest; /* filename pointed to by .so */
1.1 kristaps 64: int reparse_count; /* finite interp. stack */
65: mandocmsg mmsg; /* warning/error message handler */
1.47 schwarze 66: const char *file;
1.24 kristaps 67: struct buf *secondary;
1.69 schwarze 68: const char *defos; /* default operating system */
1.1 kristaps 69: };
70:
71: static void resize_buf(struct buf *, size_t);
72: static void mparse_buf_r(struct mparse *, struct buf, int);
73: static void pset(const char *, int, struct mparse *);
1.40 schwarze 74: static int read_whole_file(struct mparse *, const char *, int,
75: struct buf *, int *);
1.1 kristaps 76: static void mparse_end(struct mparse *);
1.37 schwarze 77: static void mparse_parse_buffer(struct mparse *, struct buf,
78: const char *);
1.1 kristaps 79:
1.3 kristaps 80: static const enum mandocerr mandoclimits[MANDOCLEVEL_MAX] = {
81: MANDOCERR_OK,
82: MANDOCERR_WARNING,
83: MANDOCERR_WARNING,
84: MANDOCERR_ERROR,
85: MANDOCERR_FATAL,
86: MANDOCERR_MAX,
87: MANDOCERR_MAX
88: };
89:
1.7 kristaps 90: static const char * const mandocerrs[MANDOCERR_MAX] = {
91: "ok",
92:
93: "generic warning",
94:
95: /* related to the prologue */
1.79 ! schwarze 96: "missing manual title, using UNTITLED",
! 97: "missing manual title, using \"\"",
1.54 schwarze 98: "lower case character in document title",
1.79 ! schwarze 99: "missing manual section, using \"\"",
1.7 kristaps 100: "unknown manual section",
1.32 schwarze 101: "unknown manual volume or arch",
1.54 schwarze 102: "missing date, using today's date",
1.7 kristaps 103: "cannot parse date, using it verbatim",
1.79 ! schwarze 104: "missing Os macro, using \"\"",
! 105: "duplicate prologue macro",
! 106: "late prologue macro",
! 107: "skipping late title macro",
1.7 kristaps 108: "prologue macros out of order",
109:
110: /* related to document structure */
111: ".so is fragile, better use ln(1)",
1.50 schwarze 112: "no document body",
1.54 schwarze 113: "content before first section header",
114: "first section is not \"NAME\"",
1.7 kristaps 115: "bad NAME section contents",
116: "sections out of conventional order",
1.54 schwarze 117: "duplicate section title",
118: "unexpected section",
1.7 kristaps 119:
120: /* related to macros and nesting */
1.55 schwarze 121: "obsolete macro",
1.7 kristaps 122: "skipping paragraph macro",
1.31 schwarze 123: "moving paragraph macro out of list",
1.7 kristaps 124: "skipping no-space macro",
125: "blocks badly nested",
126: "nested displays are not portable",
1.57 schwarze 127: "moving content out of list",
128: ".Vt block has child macro",
1.78 schwarze 129: "fill mode already enabled, skipping",
130: "fill mode already disabled, skipping",
1.7 kristaps 131: "line scope broken",
132:
133: /* related to missing macro arguments */
1.58 schwarze 134: "skipping empty request",
135: "conditional request controls empty scope",
1.7 kristaps 136: "skipping empty macro",
1.62 schwarze 137: "empty argument, using 0n",
1.7 kristaps 138: "argument count wrong",
1.60 schwarze 139: "missing display type, using -ragged",
140: "list type is not the first argument",
141: "missing -width in -tag list, using 8n",
1.78 schwarze 142: "missing utility name, using \"\"",
1.60 schwarze 143: "empty head in list item",
144: "empty list item",
1.61 schwarze 145: "missing font type, using \\fR",
146: "unknown font type, using \\fR",
1.60 schwarze 147: "missing -std argument, adding it",
1.7 kristaps 148:
149: /* related to bad macro arguments */
1.64 schwarze 150: "unterminated quoted argument",
1.7 kristaps 151: "duplicate argument",
1.76 schwarze 152: "skipping duplicate argument",
1.63 schwarze 153: "skipping duplicate display type",
154: "skipping duplicate list type",
1.76 schwarze 155: "skipping -width argument",
1.7 kristaps 156: "unknown AT&T UNIX version",
1.67 schwarze 157: "invalid content in Rs block",
1.63 schwarze 158: "invalid Boolean argument",
159: "unknown font, skipping request",
1.7 kristaps 160:
161: /* related to plain text */
1.64 schwarze 162: "blank line in fill mode, using .sp",
163: "tab in filled text",
164: "whitespace at end of input line",
1.7 kristaps 165: "bad comment style",
1.64 schwarze 166: "invalid escape sequence",
167: "undefined string, using \"\"",
1.16 kristaps 168:
1.7 kristaps 169: "generic error",
1.17 kristaps 170:
171: /* related to equations */
1.20 kristaps 172: "unexpected equation scope closure",
173: "equation scope open on exit",
1.21 kristaps 174: "overlapping equation scopes",
175: "unexpected end of equation",
176: "equation syntax error",
1.7 kristaps 177:
178: /* related to tables */
179: "bad table syntax",
180: "bad table option",
181: "bad table layout",
182: "no table layout cells specified",
183: "no table data cells specified",
184: "ignore data in cell",
185: "data block still open",
186: "ignoring extra data cells",
187:
1.68 schwarze 188: /* related to document structure and macros */
1.7 kristaps 189: "input stack limit exceeded, infinite loop?",
190: "skipping bad character",
1.68 schwarze 191: "skipping unknown macro",
1.73 schwarze 192: "skipping item outside list",
1.68 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 */
1.7 kristaps 199: "escaped character not allowed in a name",
200: "argument count wrong",
1.71 schwarze 201: "missing list type, using -item",
1.70 schwarze 202: "missing manual name, using \"\"",
1.71 schwarze 203: "uname(3) system call failed, using UNKNOWN",
1.63 schwarze 204: "unknown standard specifier",
1.71 schwarze 205: "skipping request without numeric argument",
1.61 schwarze 206: "skipping all arguments",
207: "skipping excess arguments",
1.7 kristaps 208:
209: "generic fatal error",
210:
1.40 schwarze 211: "input too large",
1.78 schwarze 212: "NOT IMPLEMENTED: Bd -file",
1.7 kristaps 213: "NOT IMPLEMENTED: .so with absolute path or \"..\"",
1.52 schwarze 214: ".so request failed",
1.40 schwarze 215:
216: /* system errors */
1.51 schwarze 217: NULL,
1.40 schwarze 218: "cannot stat file",
219: "cannot read file",
1.7 kristaps 220: };
221:
222: static const char * const mandoclevels[MANDOCLEVEL_MAX] = {
223: "SUCCESS",
224: "RESERVED",
225: "WARNING",
226: "ERROR",
227: "FATAL",
228: "BADARG",
229: "SYSERR"
230: };
231:
1.47 schwarze 232:
1.1 kristaps 233: static void
234: resize_buf(struct buf *buf, size_t initial)
235: {
236:
237: buf->sz = buf->sz > initial/2 ? 2 * buf->sz : initial;
238: buf->buf = mandoc_realloc(buf->buf, buf->sz);
239: }
240:
241: static void
242: pset(const char *buf, int pos, struct mparse *curp)
243: {
244: int i;
245:
246: /*
247: * Try to intuit which kind of manual parser should be used. If
248: * passed in by command-line (-man, -mdoc), then use that
249: * explicitly. If passed as -mandoc, then try to guess from the
250: * line: either skip dot-lines, use -mdoc when finding `.Dt', or
251: * default to -man, which is more lenient.
252: *
253: * Separate out pmdoc/pman from mdoc/man: the first persists
254: * through all parsers, while the latter is used per-parse.
255: */
256:
257: if ('.' == buf[0] || '\'' == buf[0]) {
258: for (i = 1; buf[i]; i++)
259: if (' ' != buf[i] && '\t' != buf[i])
260: break;
261: if ('\0' == buf[i])
262: return;
263: }
264:
1.44 schwarze 265: if (MPARSE_MDOC & curp->options) {
1.1 kristaps 266: curp->mdoc = curp->pmdoc;
267: return;
1.44 schwarze 268: } else if (MPARSE_MAN & curp->options) {
1.1 kristaps 269: curp->man = curp->pman;
270: return;
271: }
272:
273: if (pos >= 3 && 0 == memcmp(buf, ".Dd", 3)) {
1.47 schwarze 274: if (NULL == curp->pmdoc)
1.44 schwarze 275: curp->pmdoc = mdoc_alloc(
276: curp->roff, curp, curp->defos,
277: MPARSE_QUICK & curp->options ? 1 : 0);
1.1 kristaps 278: assert(curp->pmdoc);
279: curp->mdoc = curp->pmdoc;
280: return;
1.47 schwarze 281: }
1.1 kristaps 282:
1.47 schwarze 283: if (NULL == curp->pman)
1.44 schwarze 284: curp->pman = man_alloc(curp->roff, curp,
285: MPARSE_QUICK & curp->options ? 1 : 0);
1.1 kristaps 286: assert(curp->pman);
287: curp->man = curp->pman;
288: }
289:
290: /*
291: * Main parse routine for an opened file. This is called for each
292: * opened file and simply loops around the full input file, possibly
293: * nesting (i.e., with `so').
294: */
295: static void
296: mparse_buf_r(struct mparse *curp, struct buf blk, int start)
297: {
298: const struct tbl_span *span;
299: struct buf ln;
300: enum rofferr rr;
301: int i, of, rc;
302: int pos; /* byte number in the ln buffer */
303: int lnn; /* line number in the real file */
304: unsigned char c;
305:
306: memset(&ln, 0, sizeof(struct buf));
307:
1.47 schwarze 308: lnn = curp->line;
309: pos = 0;
1.1 kristaps 310:
311: for (i = 0; i < (int)blk.sz; ) {
312: if (0 == pos && '\0' == blk.buf[i])
313: break;
314:
315: if (start) {
316: curp->line = lnn;
317: curp->reparse_count = 0;
318: }
319:
320: while (i < (int)blk.sz && (start || '\0' != blk.buf[i])) {
321:
322: /*
323: * When finding an unescaped newline character,
324: * leave the character loop to process the line.
325: * Skip a preceding carriage return, if any.
326: */
327:
328: if ('\r' == blk.buf[i] && i + 1 < (int)blk.sz &&
329: '\n' == blk.buf[i + 1])
330: ++i;
331: if ('\n' == blk.buf[i]) {
332: ++i;
333: ++lnn;
334: break;
335: }
336:
1.35 schwarze 337: /*
338: * Make sure we have space for at least
339: * one backslash and one other character
340: * and the trailing NUL byte.
341: */
342:
343: if (pos + 2 >= (int)ln.sz)
344: resize_buf(&ln, 256);
345:
1.47 schwarze 346: /*
1.1 kristaps 347: * Warn about bogus characters. If you're using
348: * non-ASCII encoding, you're screwing your
349: * readers. Since I'd rather this not happen,
1.27 joerg 350: * I'll be helpful and replace these characters
351: * with "?", so we don't display gibberish.
352: * Note to manual writers: use special characters.
1.1 kristaps 353: */
354:
355: c = (unsigned char) blk.buf[i];
356:
1.47 schwarze 357: if ( ! (isascii(c) &&
358: (isgraph(c) || isblank(c)))) {
1.78 schwarze 359: mandoc_vmsg(MANDOCERR_BADCHAR, curp,
360: curp->line, pos, "0x%x", c);
1.1 kristaps 361: i++;
1.27 joerg 362: ln.buf[pos++] = '?';
1.1 kristaps 363: continue;
364: }
365:
366: /* Trailing backslash = a plain char. */
367:
368: if ('\\' != blk.buf[i] || i + 1 == (int)blk.sz) {
369: ln.buf[pos++] = blk.buf[i++];
370: continue;
371: }
372:
373: /*
374: * Found escape and at least one other character.
375: * When it's a newline character, skip it.
376: * When there is a carriage return in between,
377: * skip that one as well.
378: */
379:
380: if ('\r' == blk.buf[i + 1] && i + 2 < (int)blk.sz &&
381: '\n' == blk.buf[i + 2])
382: ++i;
383: if ('\n' == blk.buf[i + 1]) {
384: i += 2;
385: ++lnn;
386: continue;
387: }
388:
1.13 kristaps 389: if ('"' == blk.buf[i + 1] || '#' == blk.buf[i + 1]) {
1.1 kristaps 390: i += 2;
391: /* Comment, skip to end of line */
392: for (; i < (int)blk.sz; ++i) {
393: if ('\n' == blk.buf[i]) {
394: ++i;
395: ++lnn;
396: break;
397: }
398: }
399:
400: /* Backout trailing whitespaces */
401: for (; pos > 0; --pos) {
402: if (ln.buf[pos - 1] != ' ')
403: break;
404: if (pos > 2 && ln.buf[pos - 2] == '\\')
405: break;
406: }
407: break;
408: }
409:
1.35 schwarze 410: /* Catch escaped bogus characters. */
411:
412: c = (unsigned char) blk.buf[i+1];
413:
1.47 schwarze 414: if ( ! (isascii(c) &&
415: (isgraph(c) || isblank(c)))) {
1.78 schwarze 416: mandoc_vmsg(MANDOCERR_BADCHAR, curp,
417: curp->line, pos, "0x%x", c);
1.35 schwarze 418: i += 2;
419: ln.buf[pos++] = '?';
420: continue;
421: }
422:
1.1 kristaps 423: /* Some other escape sequence, copy & cont. */
424:
425: ln.buf[pos++] = blk.buf[i++];
426: ln.buf[pos++] = blk.buf[i++];
427: }
428:
1.47 schwarze 429: if (pos >= (int)ln.sz)
1.1 kristaps 430: resize_buf(&ln, 256);
431:
432: ln.buf[pos] = '\0';
433:
434: /*
435: * A significant amount of complexity is contained by
436: * the roff preprocessor. It's line-oriented but can be
437: * expressed on one line, so we need at times to
438: * readjust our starting point and re-run it. The roff
439: * preprocessor can also readjust the buffers with new
440: * data, so we pass them in wholesale.
441: */
442:
443: of = 0;
444:
1.24 kristaps 445: /*
446: * Maintain a lookaside buffer of all parsed lines. We
447: * only do this if mparse_keep() has been invoked (the
448: * buffer may be accessed with mparse_getkeep()).
449: */
450:
451: if (curp->secondary) {
1.47 schwarze 452: curp->secondary->buf = mandoc_realloc(
453: curp->secondary->buf,
454: curp->secondary->sz + pos + 2);
455: memcpy(curp->secondary->buf +
456: curp->secondary->sz,
457: ln.buf, pos);
1.24 kristaps 458: curp->secondary->sz += pos;
459: curp->secondary->buf
460: [curp->secondary->sz] = '\n';
461: curp->secondary->sz++;
462: curp->secondary->buf
463: [curp->secondary->sz] = '\0';
464: }
1.1 kristaps 465: rerun:
1.47 schwarze 466: rr = roff_parseln(curp->roff, curp->line,
467: &ln.buf, &ln.sz, of, &of);
1.1 kristaps 468:
469: switch (rr) {
1.47 schwarze 470: case ROFF_REPARSE:
1.1 kristaps 471: if (REPARSE_LIMIT >= ++curp->reparse_count)
472: mparse_buf_r(curp, ln, 0);
473: else
1.3 kristaps 474: mandoc_msg(MANDOCERR_ROFFLOOP, curp,
1.47 schwarze 475: curp->line, pos, NULL);
1.1 kristaps 476: pos = 0;
477: continue;
1.47 schwarze 478: case ROFF_APPEND:
1.1 kristaps 479: pos = (int)strlen(ln.buf);
480: continue;
1.47 schwarze 481: case ROFF_RERUN:
1.1 kristaps 482: goto rerun;
1.47 schwarze 483: case ROFF_IGN:
1.1 kristaps 484: pos = 0;
485: continue;
1.47 schwarze 486: case ROFF_ERR:
1.1 kristaps 487: assert(MANDOCLEVEL_FATAL <= curp->file_status);
488: break;
1.47 schwarze 489: case ROFF_SO:
1.45 schwarze 490: if (0 == (MPARSE_SO & curp->options) &&
491: (i >= (int)blk.sz || '\0' == blk.buf[i])) {
492: curp->sodest = mandoc_strdup(ln.buf + of);
493: free(ln.buf);
494: return;
495: }
1.24 kristaps 496: /*
497: * We remove `so' clauses from our lookaside
498: * buffer because we're going to descend into
499: * the file recursively.
500: */
1.47 schwarze 501: if (curp->secondary)
1.25 kristaps 502: curp->secondary->sz -= pos + 1;
1.36 schwarze 503: mparse_readfd(curp, -1, ln.buf + of);
1.52 schwarze 504: if (MANDOCLEVEL_FATAL <= curp->file_status) {
505: mandoc_vmsg(MANDOCERR_SO_FAIL,
506: curp, curp->line, pos,
507: ".so %s", ln.buf + of);
1.1 kristaps 508: break;
1.52 schwarze 509: }
1.1 kristaps 510: pos = 0;
511: continue;
512: default:
513: break;
514: }
515:
516: /*
517: * If we encounter errors in the recursive parse, make
518: * sure we don't continue parsing.
519: */
520:
521: if (MANDOCLEVEL_FATAL <= curp->file_status)
522: break;
523:
524: /*
525: * If input parsers have not been allocated, do so now.
1.14 kristaps 526: * We keep these instanced between parsers, but set them
1.1 kristaps 527: * locally per parse routine since we can use different
528: * parsers with each one.
529: */
530:
531: if ( ! (curp->man || curp->mdoc))
532: pset(ln.buf + of, pos - of, curp);
533:
1.47 schwarze 534: /*
1.1 kristaps 535: * Lastly, push down into the parsers themselves. One
536: * of these will have already been set in the pset()
537: * routine.
538: * If libroff returns ROFF_TBL, then add it to the
539: * currently open parse. Since we only get here if
540: * there does exist data (see tbl_data.c), we're
541: * guaranteed that something's been allocated.
542: * Do the same for ROFF_EQN.
543: */
544:
545: rc = -1;
546:
547: if (ROFF_TBL == rr)
548: while (NULL != (span = roff_span(curp->roff))) {
549: rc = curp->man ?
1.47 schwarze 550: man_addspan(curp->man, span) :
551: mdoc_addspan(curp->mdoc, span);
1.1 kristaps 552: if (0 == rc)
553: break;
554: }
555: else if (ROFF_EQN == rr)
1.47 schwarze 556: rc = curp->mdoc ?
557: mdoc_addeqn(curp->mdoc,
558: roff_eqn(curp->roff)) :
559: man_addeqn(curp->man,
560: roff_eqn(curp->roff));
1.1 kristaps 561: else if (curp->man || curp->mdoc)
562: rc = curp->man ?
1.47 schwarze 563: man_parseln(curp->man,
564: curp->line, ln.buf, of) :
565: mdoc_parseln(curp->mdoc,
566: curp->line, ln.buf, of);
1.1 kristaps 567:
568: if (0 == rc) {
569: assert(MANDOCLEVEL_FATAL <= curp->file_status);
570: break;
1.41 schwarze 571: } else if (2 == rc)
572: break;
1.1 kristaps 573:
574: /* Temporary buffers typically are not full. */
575:
576: if (0 == start && '\0' == blk.buf[i])
577: break;
578:
579: /* Start the next input line. */
580:
581: pos = 0;
582: }
583:
584: free(ln.buf);
585: }
586:
587: static int
1.40 schwarze 588: read_whole_file(struct mparse *curp, const char *file, int fd,
589: struct buf *fb, int *with_mmap)
1.1 kristaps 590: {
591: size_t off;
592: ssize_t ssz;
593:
1.15 kristaps 594: #ifdef HAVE_MMAP
595: struct stat st;
1.1 kristaps 596: if (-1 == fstat(fd, &st)) {
1.40 schwarze 597: curp->file_status = MANDOCLEVEL_SYSERR;
598: if (curp->mmsg)
599: (*curp->mmsg)(MANDOCERR_SYSSTAT, curp->file_status,
600: file, 0, 0, strerror(errno));
1.1 kristaps 601: return(0);
602: }
603:
604: /*
605: * If we're a regular file, try just reading in the whole entry
606: * via mmap(). This is faster than reading it into blocks, and
607: * since each file is only a few bytes to begin with, I'm not
608: * concerned that this is going to tank any machines.
609: */
610:
611: if (S_ISREG(st.st_mode)) {
612: if (st.st_size >= (1U << 31)) {
1.40 schwarze 613: curp->file_status = MANDOCLEVEL_FATAL;
614: if (curp->mmsg)
615: (*curp->mmsg)(MANDOCERR_TOOLARGE,
616: curp->file_status, file, 0, 0, NULL);
1.1 kristaps 617: return(0);
618: }
619: *with_mmap = 1;
620: fb->sz = (size_t)st.st_size;
1.37 schwarze 621: fb->buf = mmap(NULL, fb->sz, PROT_READ, MAP_SHARED, fd, 0);
1.1 kristaps 622: if (fb->buf != MAP_FAILED)
623: return(1);
624: }
1.15 kristaps 625: #endif
1.1 kristaps 626:
627: /*
628: * If this isn't a regular file (like, say, stdin), then we must
629: * go the old way and just read things in bit by bit.
630: */
631:
632: *with_mmap = 0;
633: off = 0;
634: fb->sz = 0;
635: fb->buf = NULL;
636: for (;;) {
637: if (off == fb->sz) {
638: if (fb->sz == (1U << 31)) {
1.40 schwarze 639: curp->file_status = MANDOCLEVEL_FATAL;
640: if (curp->mmsg)
641: (*curp->mmsg)(MANDOCERR_TOOLARGE,
642: curp->file_status,
643: file, 0, 0, NULL);
1.1 kristaps 644: break;
645: }
646: resize_buf(fb, 65536);
647: }
648: ssz = read(fd, fb->buf + (int)off, fb->sz - off);
649: if (ssz == 0) {
650: fb->sz = off;
651: return(1);
652: }
653: if (ssz == -1) {
1.40 schwarze 654: curp->file_status = MANDOCLEVEL_SYSERR;
655: if (curp->mmsg)
656: (*curp->mmsg)(MANDOCERR_SYSREAD,
657: curp->file_status, file, 0, 0,
658: strerror(errno));
1.1 kristaps 659: break;
660: }
661: off += (size_t)ssz;
662: }
663:
664: free(fb->buf);
665: fb->buf = NULL;
666: return(0);
667: }
668:
669: static void
670: mparse_end(struct mparse *curp)
671: {
672:
673: if (MANDOCLEVEL_FATAL <= curp->file_status)
674: return;
675:
1.72 schwarze 676: if (curp->mdoc == NULL &&
677: curp->man == NULL &&
678: curp->sodest == NULL) {
679: if (curp->options & MPARSE_MDOC)
680: curp->mdoc = curp->pmdoc;
681: else {
682: if (curp->pman == NULL)
683: curp->pman = man_alloc(curp->roff, curp,
684: curp->options & MPARSE_QUICK ? 1 : 0);
685: curp->man = curp->pman;
686: }
687: }
688:
1.1 kristaps 689: if (curp->mdoc && ! mdoc_endparse(curp->mdoc)) {
690: assert(MANDOCLEVEL_FATAL <= curp->file_status);
691: return;
692: }
693:
694: if (curp->man && ! man_endparse(curp->man)) {
695: assert(MANDOCLEVEL_FATAL <= curp->file_status);
696: return;
697: }
698:
699: roff_endparse(curp->roff);
700: }
701:
702: static void
1.36 schwarze 703: mparse_parse_buffer(struct mparse *curp, struct buf blk, const char *file)
1.28 joerg 704: {
705: const char *svfile;
1.36 schwarze 706: static int recursion_depth;
707:
708: if (64 < recursion_depth) {
709: mandoc_msg(MANDOCERR_ROFFLOOP, curp, curp->line, 0, NULL);
710: return;
711: }
1.28 joerg 712:
713: /* Line number is per-file. */
714: svfile = curp->file;
715: curp->file = file;
716: curp->line = 1;
1.36 schwarze 717: recursion_depth++;
1.28 joerg 718:
719: mparse_buf_r(curp, blk, 1);
720:
1.36 schwarze 721: if (0 == --recursion_depth && MANDOCLEVEL_FATAL > curp->file_status)
1.28 joerg 722: mparse_end(curp);
723:
724: curp->file = svfile;
725: }
726:
727: enum mandoclevel
728: mparse_readmem(struct mparse *curp, const void *buf, size_t len,
729: const char *file)
730: {
731: struct buf blk;
732:
733: blk.buf = UNCONST(buf);
734: blk.sz = len;
735:
1.36 schwarze 736: mparse_parse_buffer(curp, blk, file);
1.28 joerg 737: return(curp->file_status);
738: }
739:
1.36 schwarze 740: enum mandoclevel
741: mparse_readfd(struct mparse *curp, int fd, const char *file)
1.1 kristaps 742: {
1.28 joerg 743: struct buf blk;
744: int with_mmap;
1.1 kristaps 745:
1.40 schwarze 746: if (-1 == fd && -1 == (fd = open(file, O_RDONLY, 0))) {
747: curp->file_status = MANDOCLEVEL_SYSERR;
748: if (curp->mmsg)
749: (*curp->mmsg)(MANDOCERR_SYSOPEN,
750: curp->file_status,
751: file, 0, 0, strerror(errno));
752: goto out;
753: }
754:
1.28 joerg 755: /*
756: * Run for each opened file; may be called more than once for
757: * each full parse sequence if the opened file is nested (i.e.,
758: * from `so'). Simply sucks in the whole file and moves into
759: * the parse phase for the file.
760: */
1.1 kristaps 761:
1.40 schwarze 762: if ( ! read_whole_file(curp, file, fd, &blk, &with_mmap))
1.36 schwarze 763: goto out;
1.1 kristaps 764:
1.36 schwarze 765: mparse_parse_buffer(curp, blk, file);
1.1 kristaps 766:
1.28 joerg 767: #ifdef HAVE_MMAP
768: if (with_mmap)
769: munmap(blk.buf, blk.sz);
770: else
771: #endif
772: free(blk.buf);
1.1 kristaps 773:
774: if (STDIN_FILENO != fd && -1 == close(fd))
775: perror(file);
1.36 schwarze 776: out:
1.1 kristaps 777: return(curp->file_status);
778: }
779:
780: struct mparse *
1.44 schwarze 781: mparse_alloc(int options, enum mandoclevel wlevel,
1.69 schwarze 782: mandocmsg mmsg, const char *defos)
1.1 kristaps 783: {
784: struct mparse *curp;
1.10 kristaps 785:
786: assert(wlevel <= MANDOCLEVEL_FATAL);
1.1 kristaps 787:
788: curp = mandoc_calloc(1, sizeof(struct mparse));
789:
1.44 schwarze 790: curp->options = options;
1.3 kristaps 791: curp->wlevel = wlevel;
1.1 kristaps 792: curp->mmsg = mmsg;
1.29 schwarze 793: curp->defos = defos;
1.1 kristaps 794:
1.44 schwarze 795: curp->roff = roff_alloc(curp, options);
1.72 schwarze 796: if (curp->options & MPARSE_MDOC)
797: curp->pmdoc = mdoc_alloc(
798: curp->roff, curp, curp->defos,
799: curp->options & MPARSE_QUICK ? 1 : 0);
800: if (curp->options & MPARSE_MAN)
801: curp->pman = man_alloc(curp->roff, curp,
802: curp->options & MPARSE_QUICK ? 1 : 0);
803:
1.1 kristaps 804: return(curp);
805: }
806:
807: void
808: mparse_reset(struct mparse *curp)
809: {
810:
811: roff_reset(curp->roff);
812:
813: if (curp->mdoc)
814: mdoc_reset(curp->mdoc);
815: if (curp->man)
816: man_reset(curp->man);
1.24 kristaps 817: if (curp->secondary)
818: curp->secondary->sz = 0;
1.1 kristaps 819:
820: curp->file_status = MANDOCLEVEL_OK;
821: curp->mdoc = NULL;
822: curp->man = NULL;
1.45 schwarze 823:
824: free(curp->sodest);
825: curp->sodest = NULL;
1.1 kristaps 826: }
827:
828: void
829: mparse_free(struct mparse *curp)
830: {
831:
832: if (curp->pmdoc)
833: mdoc_free(curp->pmdoc);
834: if (curp->pman)
835: man_free(curp->pman);
836: if (curp->roff)
837: roff_free(curp->roff);
1.24 kristaps 838: if (curp->secondary)
839: free(curp->secondary->buf);
1.1 kristaps 840:
1.24 kristaps 841: free(curp->secondary);
1.45 schwarze 842: free(curp->sodest);
1.1 kristaps 843: free(curp);
844: }
845:
846: void
1.45 schwarze 847: mparse_result(struct mparse *curp,
848: struct mdoc **mdoc, struct man **man, char **sodest)
1.1 kristaps 849: {
850:
1.45 schwarze 851: if (sodest && NULL != (*sodest = curp->sodest)) {
852: *mdoc = NULL;
853: *man = NULL;
854: return;
855: }
1.9 kristaps 856: if (mdoc)
857: *mdoc = curp->mdoc;
858: if (man)
859: *man = curp->man;
1.3 kristaps 860: }
861:
862: void
863: mandoc_vmsg(enum mandocerr t, struct mparse *m,
864: int ln, int pos, const char *fmt, ...)
865: {
866: char buf[256];
867: va_list ap;
868:
869: va_start(ap, fmt);
1.48 schwarze 870: (void)vsnprintf(buf, sizeof(buf), fmt, ap);
1.3 kristaps 871: va_end(ap);
872:
873: mandoc_msg(t, m, ln, pos, buf);
874: }
875:
876: void
1.47 schwarze 877: mandoc_msg(enum mandocerr er, struct mparse *m,
1.3 kristaps 878: int ln, int col, const char *msg)
879: {
880: enum mandoclevel level;
881:
882: level = MANDOCLEVEL_FATAL;
883: while (er < mandoclimits[level])
884: level--;
885:
886: if (level < m->wlevel)
887: return;
888:
1.8 kristaps 889: if (m->mmsg)
890: (*m->mmsg)(er, level, m->file, ln, col, msg);
1.3 kristaps 891:
892: if (m->file_status < level)
893: m->file_status = level;
1.7 kristaps 894: }
895:
896: const char *
897: mparse_strerror(enum mandocerr er)
898: {
899:
900: return(mandocerrs[er]);
901: }
902:
903: const char *
904: mparse_strlevel(enum mandoclevel lvl)
905: {
906: return(mandoclevels[lvl]);
1.24 kristaps 907: }
908:
909: void
910: mparse_keep(struct mparse *p)
911: {
912:
913: assert(NULL == p->secondary);
914: p->secondary = mandoc_calloc(1, sizeof(struct buf));
915: }
916:
917: const char *
918: mparse_getkeep(const struct mparse *p)
919: {
920:
921: assert(p->secondary);
922: return(p->secondary->sz ? p->secondary->buf : NULL);
1.1 kristaps 923: }
CVSweb