Annotation of src/usr.bin/mandoc/read.c, Revision 1.57
1.57 ! schwarze 1: /* $Id: read.c,v 1.56 2014/08/08 15:54:10 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: */
19: #include <sys/stat.h>
20: #include <sys/mman.h>
21:
22: #include <assert.h>
23: #include <ctype.h>
1.18 schwarze 24: #include <errno.h>
1.1 schwarze 25: #include <fcntl.h>
26: #include <stdarg.h>
27: #include <stdio.h>
28: #include <stdlib.h>
29: #include <string.h>
30: #include <unistd.h>
31:
32: #include "mandoc.h"
1.24 schwarze 33: #include "mandoc_aux.h"
1.1 schwarze 34: #include "libmandoc.h"
35: #include "mdoc.h"
36: #include "man.h"
37:
38: #define REPARSE_LIMIT 1000
39:
40: struct buf {
1.25 schwarze 41: char *buf; /* binary input buffer */
1.1 schwarze 42: size_t sz; /* size of binary buffer */
43: };
44:
45: struct mparse {
46: enum mandoclevel file_status; /* status of current parse */
47: enum mandoclevel wlevel; /* ignore messages below this */
48: int line; /* line number in the file */
1.22 schwarze 49: int options; /* parser options */
1.1 schwarze 50: struct man *pman; /* persistent man parser */
51: struct mdoc *pmdoc; /* persistent mdoc parser */
52: struct man *man; /* man parser */
53: struct mdoc *mdoc; /* mdoc parser */
54: struct roff *roff; /* roff parser (!NULL) */
1.23 schwarze 55: char *sodest; /* filename pointed to by .so */
1.1 schwarze 56: int reparse_count; /* finite interp. stack */
57: mandocmsg mmsg; /* warning/error message handler */
1.25 schwarze 58: const char *file;
1.4 schwarze 59: struct buf *secondary;
1.47 schwarze 60: const char *defos; /* default operating system */
1.1 schwarze 61: };
62:
63: static void resize_buf(struct buf *, size_t);
64: static void mparse_buf_r(struct mparse *, struct buf, int);
65: static void pset(const char *, int, struct mparse *);
1.18 schwarze 66: static int read_whole_file(struct mparse *, const char *, int,
67: struct buf *, int *);
1.1 schwarze 68: static void mparse_end(struct mparse *);
1.15 schwarze 69: static void mparse_parse_buffer(struct mparse *, struct buf,
70: const char *);
1.1 schwarze 71:
72: static const enum mandocerr mandoclimits[MANDOCLEVEL_MAX] = {
73: MANDOCERR_OK,
74: MANDOCERR_WARNING,
75: MANDOCERR_WARNING,
76: MANDOCERR_ERROR,
77: MANDOCERR_FATAL,
78: MANDOCERR_MAX,
79: MANDOCERR_MAX
80: };
81:
82: static const char * const mandocerrs[MANDOCERR_MAX] = {
83: "ok",
84:
85: "generic warning",
86:
87: /* related to the prologue */
1.57 ! schwarze 88: "missing manual title, using UNTITLED",
! 89: "missing manual title, using \"\"",
1.32 schwarze 90: "lower case character in document title",
1.57 ! schwarze 91: "missing manual section, using \"\"",
1.1 schwarze 92: "unknown manual section",
1.11 schwarze 93: "unknown manual volume or arch",
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",
1.1 schwarze 111:
112: /* related to macros and nesting */
1.33 schwarze 113: "obsolete macro",
1.1 schwarze 114: "skipping paragraph macro",
1.10 schwarze 115: "moving paragraph macro out of list",
1.1 schwarze 116: "skipping no-space macro",
117: "blocks badly nested",
118: "nested displays are not portable",
1.35 schwarze 119: "moving content out of list",
120: ".Vt block has child macro",
1.56 schwarze 121: "fill mode already enabled, skipping",
122: "fill mode already disabled, skipping",
1.1 schwarze 123: "line scope broken",
124:
125: /* related to missing macro arguments */
1.36 schwarze 126: "skipping empty request",
127: "conditional request controls empty scope",
1.1 schwarze 128: "skipping empty macro",
1.40 schwarze 129: "empty argument, using 0n",
1.1 schwarze 130: "argument count wrong",
1.38 schwarze 131: "missing display type, using -ragged",
132: "list type is not the first argument",
133: "missing -width in -tag list, using 8n",
1.56 schwarze 134: "missing utility name, using \"\"",
1.38 schwarze 135: "empty head in list item",
136: "empty list item",
1.39 schwarze 137: "missing font type, using \\fR",
138: "unknown font type, using \\fR",
1.38 schwarze 139: "missing -std argument, adding it",
1.1 schwarze 140:
141: /* related to bad macro arguments */
1.42 schwarze 142: "unterminated quoted argument",
1.1 schwarze 143: "duplicate argument",
1.54 schwarze 144: "skipping duplicate argument",
1.41 schwarze 145: "skipping duplicate display type",
146: "skipping duplicate list type",
1.54 schwarze 147: "skipping -width argument",
1.1 schwarze 148: "unknown AT&T UNIX version",
1.45 schwarze 149: "invalid content in Rs block",
1.41 schwarze 150: "invalid Boolean argument",
151: "unknown font, skipping request",
1.1 schwarze 152:
153: /* related to plain text */
1.42 schwarze 154: "blank line in fill mode, using .sp",
155: "tab in filled text",
156: "whitespace at end of input line",
1.1 schwarze 157: "bad comment style",
1.42 schwarze 158: "invalid escape sequence",
159: "undefined string, using \"\"",
1.3 schwarze 160:
1.1 schwarze 161: "generic error",
162:
1.3 schwarze 163: /* related to equations */
164: "unexpected equation scope closure",
165: "equation scope open on exit",
166: "overlapping equation scopes",
167: "unexpected end of equation",
168: "equation syntax error",
169:
1.1 schwarze 170: /* related to tables */
171: "bad table syntax",
172: "bad table option",
173: "bad table layout",
174: "no table layout cells specified",
175: "no table data cells specified",
176: "ignore data in cell",
177: "data block still open",
178: "ignoring extra data cells",
179:
1.46 schwarze 180: /* related to document structure and macros */
1.1 schwarze 181: "input stack limit exceeded, infinite loop?",
182: "skipping bad character",
1.46 schwarze 183: "skipping unknown macro",
1.51 schwarze 184: "skipping item outside list",
1.46 schwarze 185: "skipping column outside column list",
186: "skipping end of block that is not open",
187: "inserting missing end of block",
188: "appending missing end of block",
189:
190: /* related to request and macro arguments */
1.1 schwarze 191: "escaped character not allowed in a name",
192: "argument count wrong",
1.49 schwarze 193: "missing list type, using -item",
1.48 schwarze 194: "missing manual name, using \"\"",
1.49 schwarze 195: "uname(3) system call failed, using UNKNOWN",
1.41 schwarze 196: "unknown standard specifier",
1.49 schwarze 197: "skipping request without numeric argument",
1.39 schwarze 198: "skipping all arguments",
199: "skipping excess arguments",
1.1 schwarze 200:
201: "generic fatal error",
202:
1.18 schwarze 203: "input too large",
1.56 schwarze 204: "NOT IMPLEMENTED: Bd -file",
1.1 schwarze 205: "NOT IMPLEMENTED: .so with absolute path or \"..\"",
1.30 schwarze 206: ".so request failed",
1.18 schwarze 207:
208: /* system errors */
1.29 schwarze 209: NULL,
1.18 schwarze 210: "cannot stat file",
211: "cannot read file",
1.1 schwarze 212: };
213:
214: static const char * const mandoclevels[MANDOCLEVEL_MAX] = {
215: "SUCCESS",
216: "RESERVED",
217: "WARNING",
218: "ERROR",
219: "FATAL",
220: "BADARG",
221: "SYSERR"
222: };
223:
1.25 schwarze 224:
1.1 schwarze 225: static void
226: resize_buf(struct buf *buf, size_t initial)
227: {
228:
229: buf->sz = buf->sz > initial/2 ? 2 * buf->sz : initial;
230: buf->buf = mandoc_realloc(buf->buf, buf->sz);
231: }
232:
233: static void
234: pset(const char *buf, int pos, struct mparse *curp)
235: {
236: int i;
237:
238: /*
239: * Try to intuit which kind of manual parser should be used. If
240: * passed in by command-line (-man, -mdoc), then use that
241: * explicitly. If passed as -mandoc, then try to guess from the
242: * line: either skip dot-lines, use -mdoc when finding `.Dt', or
243: * default to -man, which is more lenient.
244: *
245: * Separate out pmdoc/pman from mdoc/man: the first persists
246: * through all parsers, while the latter is used per-parse.
247: */
248:
249: if ('.' == buf[0] || '\'' == buf[0]) {
250: for (i = 1; buf[i]; i++)
251: if (' ' != buf[i] && '\t' != buf[i])
252: break;
253: if ('\0' == buf[i])
254: return;
255: }
256:
1.22 schwarze 257: if (MPARSE_MDOC & curp->options) {
1.1 schwarze 258: curp->mdoc = curp->pmdoc;
259: return;
1.22 schwarze 260: } else if (MPARSE_MAN & curp->options) {
1.1 schwarze 261: curp->man = curp->pman;
262: return;
263: }
264:
265: if (pos >= 3 && 0 == memcmp(buf, ".Dd", 3)) {
1.25 schwarze 266: if (NULL == curp->pmdoc)
1.22 schwarze 267: curp->pmdoc = mdoc_alloc(
268: curp->roff, curp, curp->defos,
269: MPARSE_QUICK & curp->options ? 1 : 0);
1.1 schwarze 270: assert(curp->pmdoc);
271: curp->mdoc = curp->pmdoc;
272: return;
1.25 schwarze 273: }
1.1 schwarze 274:
1.25 schwarze 275: if (NULL == curp->pman)
1.22 schwarze 276: curp->pman = man_alloc(curp->roff, curp,
277: MPARSE_QUICK & curp->options ? 1 : 0);
1.1 schwarze 278: assert(curp->pman);
279: curp->man = curp->pman;
280: }
281:
282: /*
283: * Main parse routine for an opened file. This is called for each
284: * opened file and simply loops around the full input file, possibly
285: * nesting (i.e., with `so').
286: */
287: static void
288: mparse_buf_r(struct mparse *curp, struct buf blk, int start)
289: {
290: const struct tbl_span *span;
291: struct buf ln;
292: enum rofferr rr;
293: int i, of, rc;
294: int pos; /* byte number in the ln buffer */
295: int lnn; /* line number in the real file */
296: unsigned char c;
297:
298: memset(&ln, 0, sizeof(struct buf));
299:
1.25 schwarze 300: lnn = curp->line;
301: pos = 0;
1.1 schwarze 302:
303: for (i = 0; i < (int)blk.sz; ) {
304: if (0 == pos && '\0' == blk.buf[i])
305: break;
306:
307: if (start) {
308: curp->line = lnn;
309: curp->reparse_count = 0;
310: }
311:
312: while (i < (int)blk.sz && (start || '\0' != blk.buf[i])) {
313:
314: /*
315: * When finding an unescaped newline character,
316: * leave the character loop to process the line.
317: * Skip a preceding carriage return, if any.
318: */
319:
320: if ('\r' == blk.buf[i] && i + 1 < (int)blk.sz &&
321: '\n' == blk.buf[i + 1])
322: ++i;
323: if ('\n' == blk.buf[i]) {
324: ++i;
325: ++lnn;
326: break;
327: }
328:
1.13 schwarze 329: /*
330: * Make sure we have space for at least
331: * one backslash and one other character
332: * and the trailing NUL byte.
333: */
334:
335: if (pos + 2 >= (int)ln.sz)
336: resize_buf(&ln, 256);
337:
1.25 schwarze 338: /*
1.1 schwarze 339: * Warn about bogus characters. If you're using
340: * non-ASCII encoding, you're screwing your
341: * readers. Since I'd rather this not happen,
1.6 schwarze 342: * I'll be helpful and replace these characters
343: * with "?", so we don't display gibberish.
344: * Note to manual writers: use special characters.
1.1 schwarze 345: */
346:
347: c = (unsigned char) blk.buf[i];
348:
1.25 schwarze 349: if ( ! (isascii(c) &&
350: (isgraph(c) || isblank(c)))) {
1.56 schwarze 351: mandoc_vmsg(MANDOCERR_BADCHAR, curp,
352: curp->line, pos, "0x%x", c);
1.1 schwarze 353: i++;
1.6 schwarze 354: ln.buf[pos++] = '?';
1.1 schwarze 355: continue;
356: }
357:
358: /* Trailing backslash = a plain char. */
359:
360: if ('\\' != blk.buf[i] || i + 1 == (int)blk.sz) {
361: ln.buf[pos++] = blk.buf[i++];
362: continue;
363: }
364:
365: /*
366: * Found escape and at least one other character.
367: * When it's a newline character, skip it.
368: * When there is a carriage return in between,
369: * skip that one as well.
370: */
371:
372: if ('\r' == blk.buf[i + 1] && i + 2 < (int)blk.sz &&
373: '\n' == blk.buf[i + 2])
374: ++i;
375: if ('\n' == blk.buf[i + 1]) {
376: i += 2;
377: ++lnn;
378: continue;
379: }
380:
381: if ('"' == blk.buf[i + 1] || '#' == blk.buf[i + 1]) {
382: i += 2;
383: /* Comment, skip to end of line */
384: for (; i < (int)blk.sz; ++i) {
385: if ('\n' == blk.buf[i]) {
386: ++i;
387: ++lnn;
388: break;
389: }
390: }
391:
392: /* Backout trailing whitespaces */
393: for (; pos > 0; --pos) {
394: if (ln.buf[pos - 1] != ' ')
395: break;
396: if (pos > 2 && ln.buf[pos - 2] == '\\')
397: break;
398: }
399: break;
400: }
401:
1.13 schwarze 402: /* Catch escaped bogus characters. */
403:
404: c = (unsigned char) blk.buf[i+1];
405:
1.25 schwarze 406: if ( ! (isascii(c) &&
407: (isgraph(c) || isblank(c)))) {
1.56 schwarze 408: mandoc_vmsg(MANDOCERR_BADCHAR, curp,
409: curp->line, pos, "0x%x", c);
1.13 schwarze 410: i += 2;
411: ln.buf[pos++] = '?';
412: continue;
413: }
414:
1.1 schwarze 415: /* Some other escape sequence, copy & cont. */
416:
417: ln.buf[pos++] = blk.buf[i++];
418: ln.buf[pos++] = blk.buf[i++];
419: }
420:
1.25 schwarze 421: if (pos >= (int)ln.sz)
1.1 schwarze 422: resize_buf(&ln, 256);
423:
424: ln.buf[pos] = '\0';
425:
426: /*
427: * A significant amount of complexity is contained by
428: * the roff preprocessor. It's line-oriented but can be
429: * expressed on one line, so we need at times to
430: * readjust our starting point and re-run it. The roff
431: * preprocessor can also readjust the buffers with new
432: * data, so we pass them in wholesale.
433: */
434:
435: of = 0;
436:
1.4 schwarze 437: /*
438: * Maintain a lookaside buffer of all parsed lines. We
439: * only do this if mparse_keep() has been invoked (the
440: * buffer may be accessed with mparse_getkeep()).
441: */
442:
443: if (curp->secondary) {
1.25 schwarze 444: curp->secondary->buf = mandoc_realloc(
445: curp->secondary->buf,
446: curp->secondary->sz + pos + 2);
447: memcpy(curp->secondary->buf +
448: curp->secondary->sz,
449: ln.buf, pos);
1.4 schwarze 450: curp->secondary->sz += pos;
451: curp->secondary->buf
452: [curp->secondary->sz] = '\n';
453: curp->secondary->sz++;
454: curp->secondary->buf
455: [curp->secondary->sz] = '\0';
456: }
1.1 schwarze 457: rerun:
1.25 schwarze 458: rr = roff_parseln(curp->roff, curp->line,
459: &ln.buf, &ln.sz, of, &of);
1.1 schwarze 460:
461: switch (rr) {
1.25 schwarze 462: case ROFF_REPARSE:
1.1 schwarze 463: if (REPARSE_LIMIT >= ++curp->reparse_count)
464: mparse_buf_r(curp, ln, 0);
465: else
466: mandoc_msg(MANDOCERR_ROFFLOOP, curp,
1.25 schwarze 467: curp->line, pos, NULL);
1.1 schwarze 468: pos = 0;
469: continue;
1.25 schwarze 470: case ROFF_APPEND:
1.1 schwarze 471: pos = (int)strlen(ln.buf);
472: continue;
1.25 schwarze 473: case ROFF_RERUN:
1.1 schwarze 474: goto rerun;
1.25 schwarze 475: case ROFF_IGN:
1.1 schwarze 476: pos = 0;
477: continue;
1.25 schwarze 478: case ROFF_ERR:
1.1 schwarze 479: assert(MANDOCLEVEL_FATAL <= curp->file_status);
480: break;
1.25 schwarze 481: case ROFF_SO:
1.23 schwarze 482: if (0 == (MPARSE_SO & curp->options) &&
483: (i >= (int)blk.sz || '\0' == blk.buf[i])) {
484: curp->sodest = mandoc_strdup(ln.buf + of);
485: free(ln.buf);
486: return;
487: }
1.4 schwarze 488: /*
489: * We remove `so' clauses from our lookaside
490: * buffer because we're going to descend into
491: * the file recursively.
492: */
1.25 schwarze 493: if (curp->secondary)
1.4 schwarze 494: curp->secondary->sz -= pos + 1;
1.14 schwarze 495: mparse_readfd(curp, -1, ln.buf + of);
1.30 schwarze 496: if (MANDOCLEVEL_FATAL <= curp->file_status) {
497: mandoc_vmsg(MANDOCERR_SO_FAIL,
498: curp, curp->line, pos,
499: ".so %s", ln.buf + of);
1.1 schwarze 500: break;
1.30 schwarze 501: }
1.1 schwarze 502: pos = 0;
503: continue;
504: default:
505: break;
506: }
507:
508: /*
509: * If we encounter errors in the recursive parse, make
510: * sure we don't continue parsing.
511: */
512:
513: if (MANDOCLEVEL_FATAL <= curp->file_status)
514: break;
515:
516: /*
517: * If input parsers have not been allocated, do so now.
1.2 schwarze 518: * We keep these instanced between parsers, but set them
1.1 schwarze 519: * locally per parse routine since we can use different
520: * parsers with each one.
521: */
522:
523: if ( ! (curp->man || curp->mdoc))
524: pset(ln.buf + of, pos - of, curp);
525:
1.25 schwarze 526: /*
1.1 schwarze 527: * Lastly, push down into the parsers themselves. One
528: * of these will have already been set in the pset()
529: * routine.
530: * If libroff returns ROFF_TBL, then add it to the
531: * currently open parse. Since we only get here if
532: * there does exist data (see tbl_data.c), we're
533: * guaranteed that something's been allocated.
534: * Do the same for ROFF_EQN.
535: */
536:
537: rc = -1;
538:
539: if (ROFF_TBL == rr)
540: while (NULL != (span = roff_span(curp->roff))) {
541: rc = curp->man ?
1.25 schwarze 542: man_addspan(curp->man, span) :
543: mdoc_addspan(curp->mdoc, span);
1.1 schwarze 544: if (0 == rc)
545: break;
546: }
547: else if (ROFF_EQN == rr)
1.25 schwarze 548: rc = curp->mdoc ?
549: mdoc_addeqn(curp->mdoc,
550: roff_eqn(curp->roff)) :
551: man_addeqn(curp->man,
552: roff_eqn(curp->roff));
1.1 schwarze 553: else if (curp->man || curp->mdoc)
554: rc = curp->man ?
1.25 schwarze 555: man_parseln(curp->man,
556: curp->line, ln.buf, of) :
557: mdoc_parseln(curp->mdoc,
558: curp->line, ln.buf, of);
1.1 schwarze 559:
560: if (0 == rc) {
561: assert(MANDOCLEVEL_FATAL <= curp->file_status);
562: break;
1.19 schwarze 563: } else if (2 == rc)
564: break;
1.1 schwarze 565:
566: /* Temporary buffers typically are not full. */
567:
568: if (0 == start && '\0' == blk.buf[i])
569: break;
570:
571: /* Start the next input line. */
572:
573: pos = 0;
574: }
575:
576: free(ln.buf);
577: }
578:
579: static int
1.18 schwarze 580: read_whole_file(struct mparse *curp, const char *file, int fd,
581: struct buf *fb, int *with_mmap)
1.1 schwarze 582: {
583: struct stat st;
584: size_t off;
585: ssize_t ssz;
586:
587: if (-1 == fstat(fd, &st)) {
1.18 schwarze 588: curp->file_status = MANDOCLEVEL_SYSERR;
589: if (curp->mmsg)
590: (*curp->mmsg)(MANDOCERR_SYSSTAT, curp->file_status,
591: file, 0, 0, strerror(errno));
1.1 schwarze 592: return(0);
593: }
594:
595: /*
596: * If we're a regular file, try just reading in the whole entry
597: * via mmap(). This is faster than reading it into blocks, and
598: * since each file is only a few bytes to begin with, I'm not
599: * concerned that this is going to tank any machines.
600: */
601:
602: if (S_ISREG(st.st_mode)) {
603: if (st.st_size >= (1U << 31)) {
1.18 schwarze 604: curp->file_status = MANDOCLEVEL_FATAL;
605: if (curp->mmsg)
606: (*curp->mmsg)(MANDOCERR_TOOLARGE,
607: curp->file_status, file, 0, 0, NULL);
1.1 schwarze 608: return(0);
609: }
610: *with_mmap = 1;
611: fb->sz = (size_t)st.st_size;
1.15 schwarze 612: fb->buf = mmap(NULL, fb->sz, PROT_READ, MAP_SHARED, fd, 0);
1.1 schwarze 613: if (fb->buf != MAP_FAILED)
614: return(1);
615: }
616:
617: /*
618: * If this isn't a regular file (like, say, stdin), then we must
619: * go the old way and just read things in bit by bit.
620: */
621:
622: *with_mmap = 0;
623: off = 0;
624: fb->sz = 0;
625: fb->buf = NULL;
626: for (;;) {
627: if (off == fb->sz) {
628: if (fb->sz == (1U << 31)) {
1.18 schwarze 629: curp->file_status = MANDOCLEVEL_FATAL;
630: if (curp->mmsg)
631: (*curp->mmsg)(MANDOCERR_TOOLARGE,
632: curp->file_status,
633: file, 0, 0, NULL);
1.1 schwarze 634: break;
635: }
636: resize_buf(fb, 65536);
637: }
638: ssz = read(fd, fb->buf + (int)off, fb->sz - off);
639: if (ssz == 0) {
640: fb->sz = off;
641: return(1);
642: }
643: if (ssz == -1) {
1.18 schwarze 644: curp->file_status = MANDOCLEVEL_SYSERR;
645: if (curp->mmsg)
646: (*curp->mmsg)(MANDOCERR_SYSREAD,
647: curp->file_status, file, 0, 0,
648: strerror(errno));
1.1 schwarze 649: break;
650: }
651: off += (size_t)ssz;
652: }
653:
654: free(fb->buf);
655: fb->buf = NULL;
656: return(0);
657: }
658:
659: static void
660: mparse_end(struct mparse *curp)
661: {
662:
663: if (MANDOCLEVEL_FATAL <= curp->file_status)
664: return;
665:
1.50 schwarze 666: if (curp->mdoc == NULL &&
667: curp->man == NULL &&
668: curp->sodest == NULL) {
669: if (curp->options & MPARSE_MDOC)
670: curp->mdoc = curp->pmdoc;
671: else {
672: if (curp->pman == NULL)
673: curp->pman = man_alloc(curp->roff, curp,
674: curp->options & MPARSE_QUICK ? 1 : 0);
675: curp->man = curp->pman;
676: }
677: }
678:
1.1 schwarze 679: if (curp->mdoc && ! mdoc_endparse(curp->mdoc)) {
680: assert(MANDOCLEVEL_FATAL <= curp->file_status);
681: return;
682: }
683:
684: if (curp->man && ! man_endparse(curp->man)) {
685: assert(MANDOCLEVEL_FATAL <= curp->file_status);
686: return;
687: }
688:
689: roff_endparse(curp->roff);
690: }
691:
1.15 schwarze 692: static void
693: mparse_parse_buffer(struct mparse *curp, struct buf blk, const char *file)
1.1 schwarze 694: {
695: const char *svfile;
1.14 schwarze 696: static int recursion_depth;
697:
698: if (64 < recursion_depth) {
699: mandoc_msg(MANDOCERR_ROFFLOOP, curp, curp->line, 0, NULL);
1.15 schwarze 700: return;
1.14 schwarze 701: }
1.1 schwarze 702:
1.15 schwarze 703: /* Line number is per-file. */
704: svfile = curp->file;
705: curp->file = file;
706: curp->line = 1;
707: recursion_depth++;
708:
709: mparse_buf_r(curp, blk, 1);
710:
711: if (0 == --recursion_depth && MANDOCLEVEL_FATAL > curp->file_status)
712: mparse_end(curp);
713:
714: curp->file = svfile;
715: }
716:
717: enum mandoclevel
718: mparse_readfd(struct mparse *curp, int fd, const char *file)
719: {
720: struct buf blk;
721: int with_mmap;
722:
1.18 schwarze 723: if (-1 == fd && -1 == (fd = open(file, O_RDONLY, 0))) {
724: curp->file_status = MANDOCLEVEL_SYSERR;
725: if (curp->mmsg)
726: (*curp->mmsg)(MANDOCERR_SYSOPEN,
727: curp->file_status,
728: file, 0, 0, strerror(errno));
729: goto out;
730: }
731:
1.15 schwarze 732: /*
733: * Run for each opened file; may be called more than once for
734: * each full parse sequence if the opened file is nested (i.e.,
735: * from `so'). Simply sucks in the whole file and moves into
736: * the parse phase for the file.
737: */
1.1 schwarze 738:
1.18 schwarze 739: if ( ! read_whole_file(curp, file, fd, &blk, &with_mmap))
1.15 schwarze 740: goto out;
1.1 schwarze 741:
1.15 schwarze 742: mparse_parse_buffer(curp, blk, file);
1.1 schwarze 743:
1.15 schwarze 744: if (with_mmap)
745: munmap(blk.buf, blk.sz);
746: else
747: free(blk.buf);
1.1 schwarze 748:
749: if (STDIN_FILENO != fd && -1 == close(fd))
750: perror(file);
1.14 schwarze 751: out:
1.1 schwarze 752: return(curp->file_status);
753: }
754:
755: struct mparse *
1.22 schwarze 756: mparse_alloc(int options, enum mandoclevel wlevel,
1.47 schwarze 757: mandocmsg mmsg, const char *defos)
1.1 schwarze 758: {
759: struct mparse *curp;
760:
761: assert(wlevel <= MANDOCLEVEL_FATAL);
762:
763: curp = mandoc_calloc(1, sizeof(struct mparse));
764:
1.22 schwarze 765: curp->options = options;
1.1 schwarze 766: curp->wlevel = wlevel;
767: curp->mmsg = mmsg;
1.7 schwarze 768: curp->defos = defos;
1.1 schwarze 769:
1.22 schwarze 770: curp->roff = roff_alloc(curp, options);
1.50 schwarze 771: if (curp->options & MPARSE_MDOC)
772: curp->pmdoc = mdoc_alloc(
773: curp->roff, curp, curp->defos,
774: curp->options & MPARSE_QUICK ? 1 : 0);
775: if (curp->options & MPARSE_MAN)
776: curp->pman = man_alloc(curp->roff, curp,
777: curp->options & MPARSE_QUICK ? 1 : 0);
778:
1.1 schwarze 779: return(curp);
780: }
781:
782: void
783: mparse_reset(struct mparse *curp)
784: {
785:
786: roff_reset(curp->roff);
787:
788: if (curp->mdoc)
789: mdoc_reset(curp->mdoc);
790: if (curp->man)
791: man_reset(curp->man);
1.4 schwarze 792: if (curp->secondary)
793: curp->secondary->sz = 0;
1.1 schwarze 794:
795: curp->file_status = MANDOCLEVEL_OK;
796: curp->mdoc = NULL;
797: curp->man = NULL;
1.23 schwarze 798:
799: free(curp->sodest);
800: curp->sodest = NULL;
1.1 schwarze 801: }
802:
803: void
804: mparse_free(struct mparse *curp)
805: {
806:
807: if (curp->pmdoc)
808: mdoc_free(curp->pmdoc);
809: if (curp->pman)
810: man_free(curp->pman);
811: if (curp->roff)
812: roff_free(curp->roff);
1.4 schwarze 813: if (curp->secondary)
814: free(curp->secondary->buf);
1.1 schwarze 815:
1.4 schwarze 816: free(curp->secondary);
1.23 schwarze 817: free(curp->sodest);
1.1 schwarze 818: free(curp);
819: }
820:
821: void
1.23 schwarze 822: mparse_result(struct mparse *curp,
823: struct mdoc **mdoc, struct man **man, char **sodest)
1.1 schwarze 824: {
825:
1.23 schwarze 826: if (sodest && NULL != (*sodest = curp->sodest)) {
827: *mdoc = NULL;
828: *man = NULL;
829: return;
830: }
1.1 schwarze 831: if (mdoc)
832: *mdoc = curp->mdoc;
833: if (man)
834: *man = curp->man;
835: }
836:
837: void
838: mandoc_vmsg(enum mandocerr t, struct mparse *m,
839: int ln, int pos, const char *fmt, ...)
840: {
841: char buf[256];
842: va_list ap;
843:
844: va_start(ap, fmt);
1.26 schwarze 845: (void)vsnprintf(buf, sizeof(buf), fmt, ap);
1.1 schwarze 846: va_end(ap);
847:
848: mandoc_msg(t, m, ln, pos, buf);
849: }
850:
851: void
1.25 schwarze 852: mandoc_msg(enum mandocerr er, struct mparse *m,
1.1 schwarze 853: int ln, int col, const char *msg)
854: {
855: enum mandoclevel level;
856:
857: level = MANDOCLEVEL_FATAL;
858: while (er < mandoclimits[level])
859: level--;
860:
861: if (level < m->wlevel)
862: return;
863:
864: if (m->mmsg)
865: (*m->mmsg)(er, level, m->file, ln, col, msg);
866:
867: if (m->file_status < level)
868: m->file_status = level;
869: }
870:
871: const char *
872: mparse_strerror(enum mandocerr er)
873: {
874:
875: return(mandocerrs[er]);
876: }
877:
878: const char *
879: mparse_strlevel(enum mandoclevel lvl)
880: {
881: return(mandoclevels[lvl]);
1.4 schwarze 882: }
883:
884: void
885: mparse_keep(struct mparse *p)
886: {
887:
888: assert(NULL == p->secondary);
889: p->secondary = mandoc_calloc(1, sizeof(struct buf));
890: }
891:
892: const char *
893: mparse_getkeep(const struct mparse *p)
894: {
895:
896: assert(p->secondary);
897: return(p->secondary->sz ? p->secondary->buf : NULL);
1.1 schwarze 898: }