Annotation of src/usr.bin/mandoc/read.c, Revision 1.40
1.40 ! schwarze 1: /* $Id: read.c,v 1.39 2014/07/04 16:11:41 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;
1.7 schwarze 60: 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 */
1.32 schwarze 88: "missing .TH macro, using \"unknown 1\"",
89: "lower case character in document title",
1.1 schwarze 90: "unknown manual section",
1.11 schwarze 91: "unknown manual volume or arch",
1.32 schwarze 92: "missing date, using today's date",
1.1 schwarze 93: "cannot parse date, using it verbatim",
94: "prologue macros out of order",
95: "duplicate prologue macro",
1.32 schwarze 96: "incomplete prologue, terminated by",
97: "skipping prologue macro in body",
1.1 schwarze 98:
99: /* related to document structure */
100: ".so is fragile, better use ln(1)",
1.28 schwarze 101: "no document body",
1.32 schwarze 102: "content before first section header",
103: "first section is not \"NAME\"",
1.1 schwarze 104: "bad NAME section contents",
105: "sections out of conventional order",
1.32 schwarze 106: "duplicate section title",
107: "unexpected section",
1.1 schwarze 108:
109: /* related to macros and nesting */
1.33 schwarze 110: "obsolete macro",
1.1 schwarze 111: "skipping paragraph macro",
1.10 schwarze 112: "moving paragraph macro out of list",
1.1 schwarze 113: "skipping no-space macro",
114: "blocks badly nested",
115: "nested displays are not portable",
1.35 schwarze 116: "moving content out of list",
117: ".Vt block has child macro",
1.34 schwarze 118: "fill mode already enabled, skipping .fi",
119: "fill mode already disabled, skipping .nf",
1.1 schwarze 120: "line scope broken",
121:
122: /* related to missing macro arguments */
1.36 schwarze 123: "skipping empty request",
124: "conditional request controls empty scope",
1.1 schwarze 125: "skipping empty macro",
1.40 ! schwarze 126: "empty argument, using 0n",
1.1 schwarze 127: "argument count wrong",
1.38 schwarze 128: "missing display type, using -ragged",
129: "list type is not the first argument",
130: "missing -width in -tag list, using 8n",
131: "empty head in list item",
132: "empty list item",
1.39 schwarze 133: "missing font type, using \\fR",
134: "unknown font type, using \\fR",
1.38 schwarze 135: "missing -std argument, adding it",
1.1 schwarze 136:
137: /* related to bad macro arguments */
138: "skipping argument",
139: "duplicate argument",
140: "duplicate display type",
141: "duplicate list type",
142: "unknown AT&T UNIX version",
143: "bad Boolean value",
144: "unknown font",
145: "unknown standard specifier",
146: "bad width argument",
147:
148: /* related to plain text */
149: "blank line in non-literal context",
150: "tab in non-literal context",
151: "end of line whitespace",
152: "bad comment style",
1.2 schwarze 153: "bad escape sequence",
1.1 schwarze 154: "unterminated quoted string",
1.3 schwarze 155:
1.1 schwarze 156: "generic error",
157:
1.3 schwarze 158: /* related to equations */
159: "unexpected equation scope closure",
160: "equation scope open on exit",
161: "overlapping equation scopes",
162: "unexpected end of equation",
163: "equation syntax error",
164:
1.1 schwarze 165: /* related to tables */
166: "bad table syntax",
167: "bad table option",
168: "bad table layout",
169: "no table layout cells specified",
170: "no table data cells specified",
171: "ignore data in cell",
172: "data block still open",
173: "ignoring extra data cells",
174:
175: "input stack limit exceeded, infinite loop?",
176: "skipping bad character",
177: "escaped character not allowed in a name",
1.9 schwarze 178: "manual name not yet set",
1.32 schwarze 179: "skipping text before first section header",
1.1 schwarze 180: "skipping unknown macro",
181: "NOT IMPLEMENTED, please use groff: skipping request",
182: "argument count wrong",
1.35 schwarze 183: "skipping invalid content in .Rs block",
1.12 schwarze 184: "skipping column outside column list",
1.1 schwarze 185: "skipping end of block that is not open",
186: "missing end of block",
187: "scope open on exit",
188: "uname(3) system call failed",
1.16 schwarze 189: "request requires a numeric argument",
1.38 schwarze 190: "missing list type, using -item",
1.39 schwarze 191: "skipping all arguments",
192: "skipping excess arguments",
1.1 schwarze 193:
194: "generic fatal error",
195:
1.18 schwarze 196: "input too large",
1.1 schwarze 197: "not a manual",
198: "column syntax is inconsistent",
199: "NOT IMPLEMENTED: .Bd -file",
200: "child violates parent syntax",
201: "argument count wrong, violates syntax",
202: "NOT IMPLEMENTED: .so with absolute path or \"..\"",
1.30 schwarze 203: ".so request failed",
1.1 schwarze 204: "no document prologue",
205: "static buffer exhausted",
1.18 schwarze 206:
207: /* system errors */
1.29 schwarze 208: NULL,
1.18 schwarze 209: "cannot stat file",
210: "cannot read file",
1.1 schwarze 211: };
212:
213: static const char * const mandoclevels[MANDOCLEVEL_MAX] = {
214: "SUCCESS",
215: "RESERVED",
216: "WARNING",
217: "ERROR",
218: "FATAL",
219: "BADARG",
220: "SYSERR"
221: };
222:
1.25 schwarze 223:
1.1 schwarze 224: static void
225: resize_buf(struct buf *buf, size_t initial)
226: {
227:
228: buf->sz = buf->sz > initial/2 ? 2 * buf->sz : initial;
229: buf->buf = mandoc_realloc(buf->buf, buf->sz);
230: }
231:
232: static void
233: pset(const char *buf, int pos, struct mparse *curp)
234: {
235: int i;
236:
237: /*
238: * Try to intuit which kind of manual parser should be used. If
239: * passed in by command-line (-man, -mdoc), then use that
240: * explicitly. If passed as -mandoc, then try to guess from the
241: * line: either skip dot-lines, use -mdoc when finding `.Dt', or
242: * default to -man, which is more lenient.
243: *
244: * Separate out pmdoc/pman from mdoc/man: the first persists
245: * through all parsers, while the latter is used per-parse.
246: */
247:
248: if ('.' == buf[0] || '\'' == buf[0]) {
249: for (i = 1; buf[i]; i++)
250: if (' ' != buf[i] && '\t' != buf[i])
251: break;
252: if ('\0' == buf[i])
253: return;
254: }
255:
1.22 schwarze 256: if (MPARSE_MDOC & curp->options) {
1.25 schwarze 257: if (NULL == curp->pmdoc)
1.22 schwarze 258: curp->pmdoc = mdoc_alloc(
259: curp->roff, curp, curp->defos,
260: MPARSE_QUICK & curp->options ? 1 : 0);
1.1 schwarze 261: assert(curp->pmdoc);
262: curp->mdoc = curp->pmdoc;
263: return;
1.22 schwarze 264: } else if (MPARSE_MAN & curp->options) {
1.25 schwarze 265: if (NULL == curp->pman)
1.19 schwarze 266: curp->pman = man_alloc(curp->roff, curp,
1.22 schwarze 267: MPARSE_QUICK & curp->options ? 1 : 0);
1.1 schwarze 268: assert(curp->pman);
269: curp->man = curp->pman;
270: return;
271: }
272:
273: if (pos >= 3 && 0 == memcmp(buf, ".Dd", 3)) {
1.25 schwarze 274: if (NULL == curp->pmdoc)
1.22 schwarze 275: curp->pmdoc = mdoc_alloc(
276: curp->roff, curp, curp->defos,
277: MPARSE_QUICK & curp->options ? 1 : 0);
1.1 schwarze 278: assert(curp->pmdoc);
279: curp->mdoc = curp->pmdoc;
280: return;
1.25 schwarze 281: }
1.1 schwarze 282:
1.25 schwarze 283: if (NULL == curp->pman)
1.22 schwarze 284: curp->pman = man_alloc(curp->roff, curp,
285: MPARSE_QUICK & curp->options ? 1 : 0);
1.1 schwarze 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:
1.25 schwarze 308: lnn = curp->line;
309: pos = 0;
1.1 schwarze 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:
1.13 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:
1.25 schwarze 346: /*
1.1 schwarze 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,
1.6 schwarze 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 schwarze 353: */
354:
355: c = (unsigned char) blk.buf[i];
356:
1.25 schwarze 357: if ( ! (isascii(c) &&
358: (isgraph(c) || isblank(c)))) {
1.1 schwarze 359: mandoc_msg(MANDOCERR_BADCHAR, curp,
1.25 schwarze 360: curp->line, pos, NULL);
1.1 schwarze 361: i++;
1.6 schwarze 362: ln.buf[pos++] = '?';
1.1 schwarze 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:
389: if ('"' == blk.buf[i + 1] || '#' == blk.buf[i + 1]) {
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:
1.13 schwarze 410: /* Catch escaped bogus characters. */
411:
412: c = (unsigned char) blk.buf[i+1];
413:
1.25 schwarze 414: if ( ! (isascii(c) &&
415: (isgraph(c) || isblank(c)))) {
1.13 schwarze 416: mandoc_msg(MANDOCERR_BADCHAR, curp,
1.25 schwarze 417: curp->line, pos, NULL);
1.13 schwarze 418: i += 2;
419: ln.buf[pos++] = '?';
420: continue;
421: }
422:
1.1 schwarze 423: /* Some other escape sequence, copy & cont. */
424:
425: ln.buf[pos++] = blk.buf[i++];
426: ln.buf[pos++] = blk.buf[i++];
427: }
428:
1.25 schwarze 429: if (pos >= (int)ln.sz)
1.1 schwarze 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:
1.4 schwarze 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) {
1.25 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);
1.4 schwarze 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 schwarze 465: rerun:
1.25 schwarze 466: rr = roff_parseln(curp->roff, curp->line,
467: &ln.buf, &ln.sz, of, &of);
1.1 schwarze 468:
469: switch (rr) {
1.25 schwarze 470: case ROFF_REPARSE:
1.1 schwarze 471: if (REPARSE_LIMIT >= ++curp->reparse_count)
472: mparse_buf_r(curp, ln, 0);
473: else
474: mandoc_msg(MANDOCERR_ROFFLOOP, curp,
1.25 schwarze 475: curp->line, pos, NULL);
1.1 schwarze 476: pos = 0;
477: continue;
1.25 schwarze 478: case ROFF_APPEND:
1.1 schwarze 479: pos = (int)strlen(ln.buf);
480: continue;
1.25 schwarze 481: case ROFF_RERUN:
1.1 schwarze 482: goto rerun;
1.25 schwarze 483: case ROFF_IGN:
1.1 schwarze 484: pos = 0;
485: continue;
1.25 schwarze 486: case ROFF_ERR:
1.1 schwarze 487: assert(MANDOCLEVEL_FATAL <= curp->file_status);
488: break;
1.25 schwarze 489: case ROFF_SO:
1.23 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: }
1.4 schwarze 496: /*
497: * We remove `so' clauses from our lookaside
498: * buffer because we're going to descend into
499: * the file recursively.
500: */
1.25 schwarze 501: if (curp->secondary)
1.4 schwarze 502: curp->secondary->sz -= pos + 1;
1.14 schwarze 503: mparse_readfd(curp, -1, ln.buf + of);
1.30 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 schwarze 508: break;
1.30 schwarze 509: }
1.1 schwarze 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.
1.2 schwarze 526: * We keep these instanced between parsers, but set them
1.1 schwarze 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:
1.25 schwarze 534: /*
1.1 schwarze 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 ?
1.25 schwarze 550: man_addspan(curp->man, span) :
551: mdoc_addspan(curp->mdoc, span);
1.1 schwarze 552: if (0 == rc)
553: break;
554: }
555: else if (ROFF_EQN == rr)
1.25 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 schwarze 561: else if (curp->man || curp->mdoc)
562: rc = curp->man ?
1.25 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 schwarze 567:
568: if (0 == rc) {
569: assert(MANDOCLEVEL_FATAL <= curp->file_status);
570: break;
1.19 schwarze 571: } else if (2 == rc)
572: break;
1.1 schwarze 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
1.18 schwarze 588: read_whole_file(struct mparse *curp, const char *file, int fd,
589: struct buf *fb, int *with_mmap)
1.1 schwarze 590: {
591: struct stat st;
592: size_t off;
593: ssize_t ssz;
594:
595: if (-1 == fstat(fd, &st)) {
1.18 schwarze 596: curp->file_status = MANDOCLEVEL_SYSERR;
597: if (curp->mmsg)
598: (*curp->mmsg)(MANDOCERR_SYSSTAT, curp->file_status,
599: file, 0, 0, strerror(errno));
1.1 schwarze 600: return(0);
601: }
602:
603: /*
604: * If we're a regular file, try just reading in the whole entry
605: * via mmap(). This is faster than reading it into blocks, and
606: * since each file is only a few bytes to begin with, I'm not
607: * concerned that this is going to tank any machines.
608: */
609:
610: if (S_ISREG(st.st_mode)) {
611: if (st.st_size >= (1U << 31)) {
1.18 schwarze 612: curp->file_status = MANDOCLEVEL_FATAL;
613: if (curp->mmsg)
614: (*curp->mmsg)(MANDOCERR_TOOLARGE,
615: curp->file_status, file, 0, 0, NULL);
1.1 schwarze 616: return(0);
617: }
618: *with_mmap = 1;
619: fb->sz = (size_t)st.st_size;
1.15 schwarze 620: fb->buf = mmap(NULL, fb->sz, PROT_READ, MAP_SHARED, fd, 0);
1.1 schwarze 621: if (fb->buf != MAP_FAILED)
622: return(1);
623: }
624:
625: /*
626: * If this isn't a regular file (like, say, stdin), then we must
627: * go the old way and just read things in bit by bit.
628: */
629:
630: *with_mmap = 0;
631: off = 0;
632: fb->sz = 0;
633: fb->buf = NULL;
634: for (;;) {
635: if (off == fb->sz) {
636: if (fb->sz == (1U << 31)) {
1.18 schwarze 637: curp->file_status = MANDOCLEVEL_FATAL;
638: if (curp->mmsg)
639: (*curp->mmsg)(MANDOCERR_TOOLARGE,
640: curp->file_status,
641: file, 0, 0, NULL);
1.1 schwarze 642: break;
643: }
644: resize_buf(fb, 65536);
645: }
646: ssz = read(fd, fb->buf + (int)off, fb->sz - off);
647: if (ssz == 0) {
648: fb->sz = off;
649: return(1);
650: }
651: if (ssz == -1) {
1.18 schwarze 652: curp->file_status = MANDOCLEVEL_SYSERR;
653: if (curp->mmsg)
654: (*curp->mmsg)(MANDOCERR_SYSREAD,
655: curp->file_status, file, 0, 0,
656: strerror(errno));
1.1 schwarze 657: break;
658: }
659: off += (size_t)ssz;
660: }
661:
662: free(fb->buf);
663: fb->buf = NULL;
664: return(0);
665: }
666:
667: static void
668: mparse_end(struct mparse *curp)
669: {
670:
671: if (MANDOCLEVEL_FATAL <= curp->file_status)
672: return;
673:
674: if (curp->mdoc && ! mdoc_endparse(curp->mdoc)) {
675: assert(MANDOCLEVEL_FATAL <= curp->file_status);
676: return;
677: }
678:
679: if (curp->man && ! man_endparse(curp->man)) {
680: assert(MANDOCLEVEL_FATAL <= curp->file_status);
681: return;
682: }
683:
1.23 schwarze 684: if ( ! (curp->mdoc || curp->man || curp->sodest)) {
1.28 schwarze 685: mandoc_msg(MANDOCERR_NOTMANUAL, curp, 0, 0, NULL);
1.1 schwarze 686: curp->file_status = MANDOCLEVEL_FATAL;
687: return;
688: }
689:
690: roff_endparse(curp->roff);
691: }
692:
1.15 schwarze 693: static void
694: mparse_parse_buffer(struct mparse *curp, struct buf blk, const char *file)
1.1 schwarze 695: {
696: const char *svfile;
1.14 schwarze 697: static int recursion_depth;
698:
699: if (64 < recursion_depth) {
700: mandoc_msg(MANDOCERR_ROFFLOOP, curp, curp->line, 0, NULL);
1.15 schwarze 701: return;
1.14 schwarze 702: }
1.1 schwarze 703:
1.15 schwarze 704: /* Line number is per-file. */
705: svfile = curp->file;
706: curp->file = file;
707: curp->line = 1;
708: recursion_depth++;
709:
710: mparse_buf_r(curp, blk, 1);
711:
712: if (0 == --recursion_depth && MANDOCLEVEL_FATAL > curp->file_status)
713: mparse_end(curp);
714:
715: curp->file = svfile;
716: }
717:
718: enum mandoclevel
719: mparse_readfd(struct mparse *curp, int fd, const char *file)
720: {
721: struct buf blk;
722: int with_mmap;
723:
1.18 schwarze 724: if (-1 == fd && -1 == (fd = open(file, O_RDONLY, 0))) {
725: curp->file_status = MANDOCLEVEL_SYSERR;
726: if (curp->mmsg)
727: (*curp->mmsg)(MANDOCERR_SYSOPEN,
728: curp->file_status,
729: file, 0, 0, strerror(errno));
730: goto out;
731: }
732:
1.15 schwarze 733: /*
734: * Run for each opened file; may be called more than once for
735: * each full parse sequence if the opened file is nested (i.e.,
736: * from `so'). Simply sucks in the whole file and moves into
737: * the parse phase for the file.
738: */
1.1 schwarze 739:
1.18 schwarze 740: if ( ! read_whole_file(curp, file, fd, &blk, &with_mmap))
1.15 schwarze 741: goto out;
1.1 schwarze 742:
1.15 schwarze 743: mparse_parse_buffer(curp, blk, file);
1.1 schwarze 744:
1.15 schwarze 745: if (with_mmap)
746: munmap(blk.buf, blk.sz);
747: else
748: free(blk.buf);
1.1 schwarze 749:
750: if (STDIN_FILENO != fd && -1 == close(fd))
751: perror(file);
1.14 schwarze 752: out:
1.1 schwarze 753: return(curp->file_status);
754: }
755:
756: struct mparse *
1.22 schwarze 757: mparse_alloc(int options, enum mandoclevel wlevel,
758: mandocmsg mmsg, char *defos)
1.1 schwarze 759: {
760: struct mparse *curp;
761:
762: assert(wlevel <= MANDOCLEVEL_FATAL);
763:
764: curp = mandoc_calloc(1, sizeof(struct mparse));
765:
1.22 schwarze 766: curp->options = options;
1.1 schwarze 767: curp->wlevel = wlevel;
768: curp->mmsg = mmsg;
1.7 schwarze 769: curp->defos = defos;
1.1 schwarze 770:
1.22 schwarze 771: curp->roff = roff_alloc(curp, options);
1.1 schwarze 772: return(curp);
773: }
774:
775: void
776: mparse_reset(struct mparse *curp)
777: {
778:
779: roff_reset(curp->roff);
780:
781: if (curp->mdoc)
782: mdoc_reset(curp->mdoc);
783: if (curp->man)
784: man_reset(curp->man);
1.4 schwarze 785: if (curp->secondary)
786: curp->secondary->sz = 0;
1.1 schwarze 787:
788: curp->file_status = MANDOCLEVEL_OK;
789: curp->mdoc = NULL;
790: curp->man = NULL;
1.23 schwarze 791:
792: free(curp->sodest);
793: curp->sodest = NULL;
1.1 schwarze 794: }
795:
796: void
797: mparse_free(struct mparse *curp)
798: {
799:
800: if (curp->pmdoc)
801: mdoc_free(curp->pmdoc);
802: if (curp->pman)
803: man_free(curp->pman);
804: if (curp->roff)
805: roff_free(curp->roff);
1.4 schwarze 806: if (curp->secondary)
807: free(curp->secondary->buf);
1.1 schwarze 808:
1.4 schwarze 809: free(curp->secondary);
1.23 schwarze 810: free(curp->sodest);
1.1 schwarze 811: free(curp);
812: }
813:
814: void
1.23 schwarze 815: mparse_result(struct mparse *curp,
816: struct mdoc **mdoc, struct man **man, char **sodest)
1.1 schwarze 817: {
818:
1.23 schwarze 819: if (sodest && NULL != (*sodest = curp->sodest)) {
820: *mdoc = NULL;
821: *man = NULL;
822: return;
823: }
1.1 schwarze 824: if (mdoc)
825: *mdoc = curp->mdoc;
826: if (man)
827: *man = curp->man;
828: }
829:
830: void
831: mandoc_vmsg(enum mandocerr t, struct mparse *m,
832: int ln, int pos, const char *fmt, ...)
833: {
834: char buf[256];
835: va_list ap;
836:
837: va_start(ap, fmt);
1.26 schwarze 838: (void)vsnprintf(buf, sizeof(buf), fmt, ap);
1.1 schwarze 839: va_end(ap);
840:
841: mandoc_msg(t, m, ln, pos, buf);
842: }
843:
844: void
1.25 schwarze 845: mandoc_msg(enum mandocerr er, struct mparse *m,
1.1 schwarze 846: int ln, int col, const char *msg)
847: {
848: enum mandoclevel level;
849:
850: level = MANDOCLEVEL_FATAL;
851: while (er < mandoclimits[level])
852: level--;
853:
854: if (level < m->wlevel)
855: return;
856:
857: if (m->mmsg)
858: (*m->mmsg)(er, level, m->file, ln, col, msg);
859:
860: if (m->file_status < level)
861: m->file_status = level;
862: }
863:
864: const char *
865: mparse_strerror(enum mandocerr er)
866: {
867:
868: return(mandocerrs[er]);
869: }
870:
871: const char *
872: mparse_strlevel(enum mandoclevel lvl)
873: {
874: return(mandoclevels[lvl]);
1.4 schwarze 875: }
876:
877: void
878: mparse_keep(struct mparse *p)
879: {
880:
881: assert(NULL == p->secondary);
882: p->secondary = mandoc_calloc(1, sizeof(struct buf));
883: }
884:
885: const char *
886: mparse_getkeep(const struct mparse *p)
887: {
888:
889: assert(p->secondary);
890: return(p->secondary->sz ? p->secondary->buf : NULL);
1.1 schwarze 891: }