Annotation of src/usr.bin/mandoc/read.c, Revision 1.46
1.46 ! schwarze 1: /* $Id: read.c,v 1.45 2014/07/07 16:12:06 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",
1.42 schwarze 139: "unterminated quoted argument",
1.1 schwarze 140: "duplicate argument",
1.41 schwarze 141: "skipping duplicate display type",
142: "skipping duplicate list type",
1.1 schwarze 143: "unknown AT&T UNIX version",
1.45 schwarze 144: "invalid content in Rs block",
1.41 schwarze 145: "invalid Boolean argument",
146: "unknown font, skipping request",
1.1 schwarze 147:
148: /* related to plain text */
1.42 schwarze 149: "blank line in fill mode, using .sp",
150: "tab in filled text",
151: "whitespace at end of input line",
1.1 schwarze 152: "bad comment style",
1.42 schwarze 153: "invalid escape sequence",
154: "undefined string, using \"\"",
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:
1.46 ! schwarze 175: /* related to document structure and macros */
1.1 schwarze 176: "input stack limit exceeded, infinite loop?",
177: "skipping bad character",
1.46 ! schwarze 178: "skipping unknown macro",
! 179: "skipping column outside column list",
! 180: "skipping end of block that is not open",
! 181: "inserting missing end of block",
! 182: "appending missing end of block",
! 183:
! 184: /* related to request and macro arguments */
1.1 schwarze 185: "escaped character not allowed in a name",
1.9 schwarze 186: "manual name not yet set",
1.1 schwarze 187: "argument count wrong",
1.41 schwarze 188: "unknown standard specifier",
1.1 schwarze 189: "uname(3) system call failed",
1.16 schwarze 190: "request requires a numeric argument",
1.38 schwarze 191: "missing list type, using -item",
1.39 schwarze 192: "skipping all arguments",
193: "skipping excess arguments",
1.1 schwarze 194:
195: "generic fatal error",
196:
1.18 schwarze 197: "input too large",
1.1 schwarze 198: "not a manual",
199: "column syntax is inconsistent",
200: "NOT IMPLEMENTED: .Bd -file",
201: "child violates parent syntax",
202: "argument count wrong, violates syntax",
203: "NOT IMPLEMENTED: .so with absolute path or \"..\"",
1.30 schwarze 204: ".so request failed",
1.1 schwarze 205: "no document prologue",
206: "static buffer exhausted",
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) {
1.25 schwarze 258: if (NULL == curp->pmdoc)
1.22 schwarze 259: curp->pmdoc = mdoc_alloc(
260: curp->roff, curp, curp->defos,
261: MPARSE_QUICK & curp->options ? 1 : 0);
1.1 schwarze 262: assert(curp->pmdoc);
263: curp->mdoc = curp->pmdoc;
264: return;
1.22 schwarze 265: } else if (MPARSE_MAN & curp->options) {
1.25 schwarze 266: if (NULL == curp->pman)
1.19 schwarze 267: curp->pman = man_alloc(curp->roff, curp,
1.22 schwarze 268: MPARSE_QUICK & curp->options ? 1 : 0);
1.1 schwarze 269: assert(curp->pman);
270: curp->man = curp->pman;
271: return;
272: }
273:
274: if (pos >= 3 && 0 == memcmp(buf, ".Dd", 3)) {
1.25 schwarze 275: if (NULL == curp->pmdoc)
1.22 schwarze 276: curp->pmdoc = mdoc_alloc(
277: curp->roff, curp, curp->defos,
278: MPARSE_QUICK & curp->options ? 1 : 0);
1.1 schwarze 279: assert(curp->pmdoc);
280: curp->mdoc = curp->pmdoc;
281: return;
1.25 schwarze 282: }
1.1 schwarze 283:
1.25 schwarze 284: if (NULL == curp->pman)
1.22 schwarze 285: curp->pman = man_alloc(curp->roff, curp,
286: MPARSE_QUICK & curp->options ? 1 : 0);
1.1 schwarze 287: assert(curp->pman);
288: curp->man = curp->pman;
289: }
290:
291: /*
292: * Main parse routine for an opened file. This is called for each
293: * opened file and simply loops around the full input file, possibly
294: * nesting (i.e., with `so').
295: */
296: static void
297: mparse_buf_r(struct mparse *curp, struct buf blk, int start)
298: {
299: const struct tbl_span *span;
300: struct buf ln;
301: enum rofferr rr;
302: int i, of, rc;
303: int pos; /* byte number in the ln buffer */
304: int lnn; /* line number in the real file */
305: unsigned char c;
306:
307: memset(&ln, 0, sizeof(struct buf));
308:
1.25 schwarze 309: lnn = curp->line;
310: pos = 0;
1.1 schwarze 311:
312: for (i = 0; i < (int)blk.sz; ) {
313: if (0 == pos && '\0' == blk.buf[i])
314: break;
315:
316: if (start) {
317: curp->line = lnn;
318: curp->reparse_count = 0;
319: }
320:
321: while (i < (int)blk.sz && (start || '\0' != blk.buf[i])) {
322:
323: /*
324: * When finding an unescaped newline character,
325: * leave the character loop to process the line.
326: * Skip a preceding carriage return, if any.
327: */
328:
329: if ('\r' == blk.buf[i] && i + 1 < (int)blk.sz &&
330: '\n' == blk.buf[i + 1])
331: ++i;
332: if ('\n' == blk.buf[i]) {
333: ++i;
334: ++lnn;
335: break;
336: }
337:
1.13 schwarze 338: /*
339: * Make sure we have space for at least
340: * one backslash and one other character
341: * and the trailing NUL byte.
342: */
343:
344: if (pos + 2 >= (int)ln.sz)
345: resize_buf(&ln, 256);
346:
1.25 schwarze 347: /*
1.1 schwarze 348: * Warn about bogus characters. If you're using
349: * non-ASCII encoding, you're screwing your
350: * readers. Since I'd rather this not happen,
1.6 schwarze 351: * I'll be helpful and replace these characters
352: * with "?", so we don't display gibberish.
353: * Note to manual writers: use special characters.
1.1 schwarze 354: */
355:
356: c = (unsigned char) blk.buf[i];
357:
1.25 schwarze 358: if ( ! (isascii(c) &&
359: (isgraph(c) || isblank(c)))) {
1.1 schwarze 360: mandoc_msg(MANDOCERR_BADCHAR, curp,
1.25 schwarze 361: curp->line, pos, NULL);
1.1 schwarze 362: i++;
1.6 schwarze 363: ln.buf[pos++] = '?';
1.1 schwarze 364: continue;
365: }
366:
367: /* Trailing backslash = a plain char. */
368:
369: if ('\\' != blk.buf[i] || i + 1 == (int)blk.sz) {
370: ln.buf[pos++] = blk.buf[i++];
371: continue;
372: }
373:
374: /*
375: * Found escape and at least one other character.
376: * When it's a newline character, skip it.
377: * When there is a carriage return in between,
378: * skip that one as well.
379: */
380:
381: if ('\r' == blk.buf[i + 1] && i + 2 < (int)blk.sz &&
382: '\n' == blk.buf[i + 2])
383: ++i;
384: if ('\n' == blk.buf[i + 1]) {
385: i += 2;
386: ++lnn;
387: continue;
388: }
389:
390: if ('"' == blk.buf[i + 1] || '#' == blk.buf[i + 1]) {
391: i += 2;
392: /* Comment, skip to end of line */
393: for (; i < (int)blk.sz; ++i) {
394: if ('\n' == blk.buf[i]) {
395: ++i;
396: ++lnn;
397: break;
398: }
399: }
400:
401: /* Backout trailing whitespaces */
402: for (; pos > 0; --pos) {
403: if (ln.buf[pos - 1] != ' ')
404: break;
405: if (pos > 2 && ln.buf[pos - 2] == '\\')
406: break;
407: }
408: break;
409: }
410:
1.13 schwarze 411: /* Catch escaped bogus characters. */
412:
413: c = (unsigned char) blk.buf[i+1];
414:
1.25 schwarze 415: if ( ! (isascii(c) &&
416: (isgraph(c) || isblank(c)))) {
1.13 schwarze 417: mandoc_msg(MANDOCERR_BADCHAR, curp,
1.25 schwarze 418: curp->line, pos, NULL);
1.13 schwarze 419: i += 2;
420: ln.buf[pos++] = '?';
421: continue;
422: }
423:
1.1 schwarze 424: /* Some other escape sequence, copy & cont. */
425:
426: ln.buf[pos++] = blk.buf[i++];
427: ln.buf[pos++] = blk.buf[i++];
428: }
429:
1.25 schwarze 430: if (pos >= (int)ln.sz)
1.1 schwarze 431: resize_buf(&ln, 256);
432:
433: ln.buf[pos] = '\0';
434:
435: /*
436: * A significant amount of complexity is contained by
437: * the roff preprocessor. It's line-oriented but can be
438: * expressed on one line, so we need at times to
439: * readjust our starting point and re-run it. The roff
440: * preprocessor can also readjust the buffers with new
441: * data, so we pass them in wholesale.
442: */
443:
444: of = 0;
445:
1.4 schwarze 446: /*
447: * Maintain a lookaside buffer of all parsed lines. We
448: * only do this if mparse_keep() has been invoked (the
449: * buffer may be accessed with mparse_getkeep()).
450: */
451:
452: if (curp->secondary) {
1.25 schwarze 453: curp->secondary->buf = mandoc_realloc(
454: curp->secondary->buf,
455: curp->secondary->sz + pos + 2);
456: memcpy(curp->secondary->buf +
457: curp->secondary->sz,
458: ln.buf, pos);
1.4 schwarze 459: curp->secondary->sz += pos;
460: curp->secondary->buf
461: [curp->secondary->sz] = '\n';
462: curp->secondary->sz++;
463: curp->secondary->buf
464: [curp->secondary->sz] = '\0';
465: }
1.1 schwarze 466: rerun:
1.25 schwarze 467: rr = roff_parseln(curp->roff, curp->line,
468: &ln.buf, &ln.sz, of, &of);
1.1 schwarze 469:
470: switch (rr) {
1.25 schwarze 471: case ROFF_REPARSE:
1.1 schwarze 472: if (REPARSE_LIMIT >= ++curp->reparse_count)
473: mparse_buf_r(curp, ln, 0);
474: else
475: mandoc_msg(MANDOCERR_ROFFLOOP, curp,
1.25 schwarze 476: curp->line, pos, NULL);
1.1 schwarze 477: pos = 0;
478: continue;
1.25 schwarze 479: case ROFF_APPEND:
1.1 schwarze 480: pos = (int)strlen(ln.buf);
481: continue;
1.25 schwarze 482: case ROFF_RERUN:
1.1 schwarze 483: goto rerun;
1.25 schwarze 484: case ROFF_IGN:
1.1 schwarze 485: pos = 0;
486: continue;
1.25 schwarze 487: case ROFF_ERR:
1.1 schwarze 488: assert(MANDOCLEVEL_FATAL <= curp->file_status);
489: break;
1.25 schwarze 490: case ROFF_SO:
1.23 schwarze 491: if (0 == (MPARSE_SO & curp->options) &&
492: (i >= (int)blk.sz || '\0' == blk.buf[i])) {
493: curp->sodest = mandoc_strdup(ln.buf + of);
494: free(ln.buf);
495: return;
496: }
1.4 schwarze 497: /*
498: * We remove `so' clauses from our lookaside
499: * buffer because we're going to descend into
500: * the file recursively.
501: */
1.25 schwarze 502: if (curp->secondary)
1.4 schwarze 503: curp->secondary->sz -= pos + 1;
1.14 schwarze 504: mparse_readfd(curp, -1, ln.buf + of);
1.30 schwarze 505: if (MANDOCLEVEL_FATAL <= curp->file_status) {
506: mandoc_vmsg(MANDOCERR_SO_FAIL,
507: curp, curp->line, pos,
508: ".so %s", ln.buf + of);
1.1 schwarze 509: break;
1.30 schwarze 510: }
1.1 schwarze 511: pos = 0;
512: continue;
513: default:
514: break;
515: }
516:
517: /*
518: * If we encounter errors in the recursive parse, make
519: * sure we don't continue parsing.
520: */
521:
522: if (MANDOCLEVEL_FATAL <= curp->file_status)
523: break;
524:
525: /*
526: * If input parsers have not been allocated, do so now.
1.2 schwarze 527: * We keep these instanced between parsers, but set them
1.1 schwarze 528: * locally per parse routine since we can use different
529: * parsers with each one.
530: */
531:
532: if ( ! (curp->man || curp->mdoc))
533: pset(ln.buf + of, pos - of, curp);
534:
1.25 schwarze 535: /*
1.1 schwarze 536: * Lastly, push down into the parsers themselves. One
537: * of these will have already been set in the pset()
538: * routine.
539: * If libroff returns ROFF_TBL, then add it to the
540: * currently open parse. Since we only get here if
541: * there does exist data (see tbl_data.c), we're
542: * guaranteed that something's been allocated.
543: * Do the same for ROFF_EQN.
544: */
545:
546: rc = -1;
547:
548: if (ROFF_TBL == rr)
549: while (NULL != (span = roff_span(curp->roff))) {
550: rc = curp->man ?
1.25 schwarze 551: man_addspan(curp->man, span) :
552: mdoc_addspan(curp->mdoc, span);
1.1 schwarze 553: if (0 == rc)
554: break;
555: }
556: else if (ROFF_EQN == rr)
1.25 schwarze 557: rc = curp->mdoc ?
558: mdoc_addeqn(curp->mdoc,
559: roff_eqn(curp->roff)) :
560: man_addeqn(curp->man,
561: roff_eqn(curp->roff));
1.1 schwarze 562: else if (curp->man || curp->mdoc)
563: rc = curp->man ?
1.25 schwarze 564: man_parseln(curp->man,
565: curp->line, ln.buf, of) :
566: mdoc_parseln(curp->mdoc,
567: curp->line, ln.buf, of);
1.1 schwarze 568:
569: if (0 == rc) {
570: assert(MANDOCLEVEL_FATAL <= curp->file_status);
571: break;
1.19 schwarze 572: } else if (2 == rc)
573: break;
1.1 schwarze 574:
575: /* Temporary buffers typically are not full. */
576:
577: if (0 == start && '\0' == blk.buf[i])
578: break;
579:
580: /* Start the next input line. */
581:
582: pos = 0;
583: }
584:
585: free(ln.buf);
586: }
587:
588: static int
1.18 schwarze 589: read_whole_file(struct mparse *curp, const char *file, int fd,
590: struct buf *fb, int *with_mmap)
1.1 schwarze 591: {
592: struct stat st;
593: size_t off;
594: ssize_t ssz;
595:
596: if (-1 == fstat(fd, &st)) {
1.18 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 schwarze 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)) {
1.18 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 schwarze 617: return(0);
618: }
619: *with_mmap = 1;
620: fb->sz = (size_t)st.st_size;
1.15 schwarze 621: fb->buf = mmap(NULL, fb->sz, PROT_READ, MAP_SHARED, fd, 0);
1.1 schwarze 622: if (fb->buf != MAP_FAILED)
623: return(1);
624: }
625:
626: /*
627: * If this isn't a regular file (like, say, stdin), then we must
628: * go the old way and just read things in bit by bit.
629: */
630:
631: *with_mmap = 0;
632: off = 0;
633: fb->sz = 0;
634: fb->buf = NULL;
635: for (;;) {
636: if (off == fb->sz) {
637: if (fb->sz == (1U << 31)) {
1.18 schwarze 638: curp->file_status = MANDOCLEVEL_FATAL;
639: if (curp->mmsg)
640: (*curp->mmsg)(MANDOCERR_TOOLARGE,
641: curp->file_status,
642: file, 0, 0, NULL);
1.1 schwarze 643: break;
644: }
645: resize_buf(fb, 65536);
646: }
647: ssz = read(fd, fb->buf + (int)off, fb->sz - off);
648: if (ssz == 0) {
649: fb->sz = off;
650: return(1);
651: }
652: if (ssz == -1) {
1.18 schwarze 653: curp->file_status = MANDOCLEVEL_SYSERR;
654: if (curp->mmsg)
655: (*curp->mmsg)(MANDOCERR_SYSREAD,
656: curp->file_status, file, 0, 0,
657: strerror(errno));
1.1 schwarze 658: break;
659: }
660: off += (size_t)ssz;
661: }
662:
663: free(fb->buf);
664: fb->buf = NULL;
665: return(0);
666: }
667:
668: static void
669: mparse_end(struct mparse *curp)
670: {
671:
672: if (MANDOCLEVEL_FATAL <= curp->file_status)
673: return;
674:
675: if (curp->mdoc && ! mdoc_endparse(curp->mdoc)) {
676: assert(MANDOCLEVEL_FATAL <= curp->file_status);
677: return;
678: }
679:
680: if (curp->man && ! man_endparse(curp->man)) {
681: assert(MANDOCLEVEL_FATAL <= curp->file_status);
682: return;
683: }
684:
1.23 schwarze 685: if ( ! (curp->mdoc || curp->man || curp->sodest)) {
1.28 schwarze 686: mandoc_msg(MANDOCERR_NOTMANUAL, curp, 0, 0, NULL);
1.1 schwarze 687: curp->file_status = MANDOCLEVEL_FATAL;
688: return;
689: }
690:
691: roff_endparse(curp->roff);
692: }
693:
1.15 schwarze 694: static void
695: mparse_parse_buffer(struct mparse *curp, struct buf blk, const char *file)
1.1 schwarze 696: {
697: const char *svfile;
1.14 schwarze 698: static int recursion_depth;
699:
700: if (64 < recursion_depth) {
701: mandoc_msg(MANDOCERR_ROFFLOOP, curp, curp->line, 0, NULL);
1.15 schwarze 702: return;
1.14 schwarze 703: }
1.1 schwarze 704:
1.15 schwarze 705: /* Line number is per-file. */
706: svfile = curp->file;
707: curp->file = file;
708: curp->line = 1;
709: recursion_depth++;
710:
711: mparse_buf_r(curp, blk, 1);
712:
713: if (0 == --recursion_depth && MANDOCLEVEL_FATAL > curp->file_status)
714: mparse_end(curp);
715:
716: curp->file = svfile;
717: }
718:
719: enum mandoclevel
720: mparse_readfd(struct mparse *curp, int fd, const char *file)
721: {
722: struct buf blk;
723: int with_mmap;
724:
1.18 schwarze 725: if (-1 == fd && -1 == (fd = open(file, O_RDONLY, 0))) {
726: curp->file_status = MANDOCLEVEL_SYSERR;
727: if (curp->mmsg)
728: (*curp->mmsg)(MANDOCERR_SYSOPEN,
729: curp->file_status,
730: file, 0, 0, strerror(errno));
731: goto out;
732: }
733:
1.15 schwarze 734: /*
735: * Run for each opened file; may be called more than once for
736: * each full parse sequence if the opened file is nested (i.e.,
737: * from `so'). Simply sucks in the whole file and moves into
738: * the parse phase for the file.
739: */
1.1 schwarze 740:
1.18 schwarze 741: if ( ! read_whole_file(curp, file, fd, &blk, &with_mmap))
1.15 schwarze 742: goto out;
1.1 schwarze 743:
1.15 schwarze 744: mparse_parse_buffer(curp, blk, file);
1.1 schwarze 745:
1.15 schwarze 746: if (with_mmap)
747: munmap(blk.buf, blk.sz);
748: else
749: free(blk.buf);
1.1 schwarze 750:
751: if (STDIN_FILENO != fd && -1 == close(fd))
752: perror(file);
1.14 schwarze 753: out:
1.1 schwarze 754: return(curp->file_status);
755: }
756:
757: struct mparse *
1.22 schwarze 758: mparse_alloc(int options, enum mandoclevel wlevel,
759: mandocmsg mmsg, char *defos)
1.1 schwarze 760: {
761: struct mparse *curp;
762:
763: assert(wlevel <= MANDOCLEVEL_FATAL);
764:
765: curp = mandoc_calloc(1, sizeof(struct mparse));
766:
1.22 schwarze 767: curp->options = options;
1.1 schwarze 768: curp->wlevel = wlevel;
769: curp->mmsg = mmsg;
1.7 schwarze 770: curp->defos = defos;
1.1 schwarze 771:
1.22 schwarze 772: curp->roff = roff_alloc(curp, options);
1.1 schwarze 773: return(curp);
774: }
775:
776: void
777: mparse_reset(struct mparse *curp)
778: {
779:
780: roff_reset(curp->roff);
781:
782: if (curp->mdoc)
783: mdoc_reset(curp->mdoc);
784: if (curp->man)
785: man_reset(curp->man);
1.4 schwarze 786: if (curp->secondary)
787: curp->secondary->sz = 0;
1.1 schwarze 788:
789: curp->file_status = MANDOCLEVEL_OK;
790: curp->mdoc = NULL;
791: curp->man = NULL;
1.23 schwarze 792:
793: free(curp->sodest);
794: curp->sodest = NULL;
1.1 schwarze 795: }
796:
797: void
798: mparse_free(struct mparse *curp)
799: {
800:
801: if (curp->pmdoc)
802: mdoc_free(curp->pmdoc);
803: if (curp->pman)
804: man_free(curp->pman);
805: if (curp->roff)
806: roff_free(curp->roff);
1.4 schwarze 807: if (curp->secondary)
808: free(curp->secondary->buf);
1.1 schwarze 809:
1.4 schwarze 810: free(curp->secondary);
1.23 schwarze 811: free(curp->sodest);
1.1 schwarze 812: free(curp);
813: }
814:
815: void
1.23 schwarze 816: mparse_result(struct mparse *curp,
817: struct mdoc **mdoc, struct man **man, char **sodest)
1.1 schwarze 818: {
819:
1.23 schwarze 820: if (sodest && NULL != (*sodest = curp->sodest)) {
821: *mdoc = NULL;
822: *man = NULL;
823: return;
824: }
1.1 schwarze 825: if (mdoc)
826: *mdoc = curp->mdoc;
827: if (man)
828: *man = curp->man;
829: }
830:
831: void
832: mandoc_vmsg(enum mandocerr t, struct mparse *m,
833: int ln, int pos, const char *fmt, ...)
834: {
835: char buf[256];
836: va_list ap;
837:
838: va_start(ap, fmt);
1.26 schwarze 839: (void)vsnprintf(buf, sizeof(buf), fmt, ap);
1.1 schwarze 840: va_end(ap);
841:
842: mandoc_msg(t, m, ln, pos, buf);
843: }
844:
845: void
1.25 schwarze 846: mandoc_msg(enum mandocerr er, struct mparse *m,
1.1 schwarze 847: int ln, int col, const char *msg)
848: {
849: enum mandoclevel level;
850:
851: level = MANDOCLEVEL_FATAL;
852: while (er < mandoclimits[level])
853: level--;
854:
855: if (level < m->wlevel)
856: return;
857:
858: if (m->mmsg)
859: (*m->mmsg)(er, level, m->file, ln, col, msg);
860:
861: if (m->file_status < level)
862: m->file_status = level;
863: }
864:
865: const char *
866: mparse_strerror(enum mandocerr er)
867: {
868:
869: return(mandocerrs[er]);
870: }
871:
872: const char *
873: mparse_strlevel(enum mandoclevel lvl)
874: {
875: return(mandoclevels[lvl]);
1.4 schwarze 876: }
877:
878: void
879: mparse_keep(struct mparse *p)
880: {
881:
882: assert(NULL == p->secondary);
883: p->secondary = mandoc_calloc(1, sizeof(struct buf));
884: }
885:
886: const char *
887: mparse_getkeep(const struct mparse *p)
888: {
889:
890: assert(p->secondary);
891: return(p->secondary->sz ? p->secondary->buf : NULL);
1.1 schwarze 892: }