Annotation of src/usr.bin/mandoc/read.c, Revision 1.48
1.48 ! schwarze 1: /* $Id: read.c,v 1.47 2014/07/09 11:30:07 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 */
1.32 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",
1.48 ! schwarze 131: "missing name for .Ex, using \"\"",
1.38 schwarze 132: "empty head in list item",
133: "empty list item",
1.39 schwarze 134: "missing font type, using \\fR",
135: "unknown font type, using \\fR",
1.38 schwarze 136: "missing -std argument, adding it",
1.1 schwarze 137:
138: /* related to bad macro arguments */
139: "skipping argument",
1.42 schwarze 140: "unterminated quoted argument",
1.1 schwarze 141: "duplicate argument",
1.41 schwarze 142: "skipping duplicate display type",
143: "skipping duplicate list type",
1.1 schwarze 144: "unknown AT&T UNIX version",
1.45 schwarze 145: "invalid content in Rs block",
1.41 schwarze 146: "invalid Boolean argument",
147: "unknown font, skipping request",
1.1 schwarze 148:
149: /* related to plain text */
1.42 schwarze 150: "blank line in fill mode, using .sp",
151: "tab in filled text",
152: "whitespace at end of input line",
1.1 schwarze 153: "bad comment style",
1.42 schwarze 154: "invalid escape sequence",
155: "undefined string, using \"\"",
1.3 schwarze 156:
1.1 schwarze 157: "generic error",
158:
1.3 schwarze 159: /* related to equations */
160: "unexpected equation scope closure",
161: "equation scope open on exit",
162: "overlapping equation scopes",
163: "unexpected end of equation",
164: "equation syntax error",
165:
1.1 schwarze 166: /* related to tables */
167: "bad table syntax",
168: "bad table option",
169: "bad table layout",
170: "no table layout cells specified",
171: "no table data cells specified",
172: "ignore data in cell",
173: "data block still open",
174: "ignoring extra data cells",
175:
1.46 schwarze 176: /* related to document structure and macros */
1.1 schwarze 177: "input stack limit exceeded, infinite loop?",
178: "skipping bad character",
1.46 schwarze 179: "skipping unknown macro",
180: "skipping column outside column list",
181: "skipping end of block that is not open",
182: "inserting missing end of block",
183: "appending missing end of block",
184:
185: /* related to request and macro arguments */
1.1 schwarze 186: "escaped character not allowed in a name",
187: "argument count wrong",
1.48 ! schwarze 188: "missing manual name, using \"\"",
1.41 schwarze 189: "unknown standard specifier",
1.1 schwarze 190: "uname(3) system call failed",
1.16 schwarze 191: "request requires a numeric argument",
1.38 schwarze 192: "missing list type, using -item",
1.39 schwarze 193: "skipping all arguments",
194: "skipping excess arguments",
1.1 schwarze 195:
196: "generic fatal error",
197:
1.18 schwarze 198: "input too large",
1.1 schwarze 199: "not a manual",
200: "column syntax is inconsistent",
201: "NOT IMPLEMENTED: .Bd -file",
202: "child violates parent syntax",
203: "argument count wrong, violates syntax",
204: "NOT IMPLEMENTED: .so with absolute path or \"..\"",
1.30 schwarze 205: ".so request failed",
1.1 schwarze 206: "no document prologue",
207: "static buffer exhausted",
1.18 schwarze 208:
209: /* system errors */
1.29 schwarze 210: NULL,
1.18 schwarze 211: "cannot stat file",
212: "cannot read file",
1.1 schwarze 213: };
214:
215: static const char * const mandoclevels[MANDOCLEVEL_MAX] = {
216: "SUCCESS",
217: "RESERVED",
218: "WARNING",
219: "ERROR",
220: "FATAL",
221: "BADARG",
222: "SYSERR"
223: };
224:
1.25 schwarze 225:
1.1 schwarze 226: static void
227: resize_buf(struct buf *buf, size_t initial)
228: {
229:
230: buf->sz = buf->sz > initial/2 ? 2 * buf->sz : initial;
231: buf->buf = mandoc_realloc(buf->buf, buf->sz);
232: }
233:
234: static void
235: pset(const char *buf, int pos, struct mparse *curp)
236: {
237: int i;
238:
239: /*
240: * Try to intuit which kind of manual parser should be used. If
241: * passed in by command-line (-man, -mdoc), then use that
242: * explicitly. If passed as -mandoc, then try to guess from the
243: * line: either skip dot-lines, use -mdoc when finding `.Dt', or
244: * default to -man, which is more lenient.
245: *
246: * Separate out pmdoc/pman from mdoc/man: the first persists
247: * through all parsers, while the latter is used per-parse.
248: */
249:
250: if ('.' == buf[0] || '\'' == buf[0]) {
251: for (i = 1; buf[i]; i++)
252: if (' ' != buf[i] && '\t' != buf[i])
253: break;
254: if ('\0' == buf[i])
255: return;
256: }
257:
1.22 schwarze 258: if (MPARSE_MDOC & curp->options) {
1.25 schwarze 259: if (NULL == curp->pmdoc)
1.22 schwarze 260: curp->pmdoc = mdoc_alloc(
261: curp->roff, curp, curp->defos,
262: MPARSE_QUICK & curp->options ? 1 : 0);
1.1 schwarze 263: assert(curp->pmdoc);
264: curp->mdoc = curp->pmdoc;
265: return;
1.22 schwarze 266: } else if (MPARSE_MAN & curp->options) {
1.25 schwarze 267: if (NULL == curp->pman)
1.19 schwarze 268: curp->pman = man_alloc(curp->roff, curp,
1.22 schwarze 269: MPARSE_QUICK & curp->options ? 1 : 0);
1.1 schwarze 270: assert(curp->pman);
271: curp->man = curp->pman;
272: return;
273: }
274:
275: if (pos >= 3 && 0 == memcmp(buf, ".Dd", 3)) {
1.25 schwarze 276: if (NULL == curp->pmdoc)
1.22 schwarze 277: curp->pmdoc = mdoc_alloc(
278: curp->roff, curp, curp->defos,
279: MPARSE_QUICK & curp->options ? 1 : 0);
1.1 schwarze 280: assert(curp->pmdoc);
281: curp->mdoc = curp->pmdoc;
282: return;
1.25 schwarze 283: }
1.1 schwarze 284:
1.25 schwarze 285: if (NULL == curp->pman)
1.22 schwarze 286: curp->pman = man_alloc(curp->roff, curp,
287: MPARSE_QUICK & curp->options ? 1 : 0);
1.1 schwarze 288: assert(curp->pman);
289: curp->man = curp->pman;
290: }
291:
292: /*
293: * Main parse routine for an opened file. This is called for each
294: * opened file and simply loops around the full input file, possibly
295: * nesting (i.e., with `so').
296: */
297: static void
298: mparse_buf_r(struct mparse *curp, struct buf blk, int start)
299: {
300: const struct tbl_span *span;
301: struct buf ln;
302: enum rofferr rr;
303: int i, of, rc;
304: int pos; /* byte number in the ln buffer */
305: int lnn; /* line number in the real file */
306: unsigned char c;
307:
308: memset(&ln, 0, sizeof(struct buf));
309:
1.25 schwarze 310: lnn = curp->line;
311: pos = 0;
1.1 schwarze 312:
313: for (i = 0; i < (int)blk.sz; ) {
314: if (0 == pos && '\0' == blk.buf[i])
315: break;
316:
317: if (start) {
318: curp->line = lnn;
319: curp->reparse_count = 0;
320: }
321:
322: while (i < (int)blk.sz && (start || '\0' != blk.buf[i])) {
323:
324: /*
325: * When finding an unescaped newline character,
326: * leave the character loop to process the line.
327: * Skip a preceding carriage return, if any.
328: */
329:
330: if ('\r' == blk.buf[i] && i + 1 < (int)blk.sz &&
331: '\n' == blk.buf[i + 1])
332: ++i;
333: if ('\n' == blk.buf[i]) {
334: ++i;
335: ++lnn;
336: break;
337: }
338:
1.13 schwarze 339: /*
340: * Make sure we have space for at least
341: * one backslash and one other character
342: * and the trailing NUL byte.
343: */
344:
345: if (pos + 2 >= (int)ln.sz)
346: resize_buf(&ln, 256);
347:
1.25 schwarze 348: /*
1.1 schwarze 349: * Warn about bogus characters. If you're using
350: * non-ASCII encoding, you're screwing your
351: * readers. Since I'd rather this not happen,
1.6 schwarze 352: * I'll be helpful and replace these characters
353: * with "?", so we don't display gibberish.
354: * Note to manual writers: use special characters.
1.1 schwarze 355: */
356:
357: c = (unsigned char) blk.buf[i];
358:
1.25 schwarze 359: if ( ! (isascii(c) &&
360: (isgraph(c) || isblank(c)))) {
1.1 schwarze 361: mandoc_msg(MANDOCERR_BADCHAR, curp,
1.25 schwarze 362: curp->line, pos, NULL);
1.1 schwarze 363: i++;
1.6 schwarze 364: ln.buf[pos++] = '?';
1.1 schwarze 365: continue;
366: }
367:
368: /* Trailing backslash = a plain char. */
369:
370: if ('\\' != blk.buf[i] || i + 1 == (int)blk.sz) {
371: ln.buf[pos++] = blk.buf[i++];
372: continue;
373: }
374:
375: /*
376: * Found escape and at least one other character.
377: * When it's a newline character, skip it.
378: * When there is a carriage return in between,
379: * skip that one as well.
380: */
381:
382: if ('\r' == blk.buf[i + 1] && i + 2 < (int)blk.sz &&
383: '\n' == blk.buf[i + 2])
384: ++i;
385: if ('\n' == blk.buf[i + 1]) {
386: i += 2;
387: ++lnn;
388: continue;
389: }
390:
391: if ('"' == blk.buf[i + 1] || '#' == blk.buf[i + 1]) {
392: i += 2;
393: /* Comment, skip to end of line */
394: for (; i < (int)blk.sz; ++i) {
395: if ('\n' == blk.buf[i]) {
396: ++i;
397: ++lnn;
398: break;
399: }
400: }
401:
402: /* Backout trailing whitespaces */
403: for (; pos > 0; --pos) {
404: if (ln.buf[pos - 1] != ' ')
405: break;
406: if (pos > 2 && ln.buf[pos - 2] == '\\')
407: break;
408: }
409: break;
410: }
411:
1.13 schwarze 412: /* Catch escaped bogus characters. */
413:
414: c = (unsigned char) blk.buf[i+1];
415:
1.25 schwarze 416: if ( ! (isascii(c) &&
417: (isgraph(c) || isblank(c)))) {
1.13 schwarze 418: mandoc_msg(MANDOCERR_BADCHAR, curp,
1.25 schwarze 419: curp->line, pos, NULL);
1.13 schwarze 420: i += 2;
421: ln.buf[pos++] = '?';
422: continue;
423: }
424:
1.1 schwarze 425: /* Some other escape sequence, copy & cont. */
426:
427: ln.buf[pos++] = blk.buf[i++];
428: ln.buf[pos++] = blk.buf[i++];
429: }
430:
1.25 schwarze 431: if (pos >= (int)ln.sz)
1.1 schwarze 432: resize_buf(&ln, 256);
433:
434: ln.buf[pos] = '\0';
435:
436: /*
437: * A significant amount of complexity is contained by
438: * the roff preprocessor. It's line-oriented but can be
439: * expressed on one line, so we need at times to
440: * readjust our starting point and re-run it. The roff
441: * preprocessor can also readjust the buffers with new
442: * data, so we pass them in wholesale.
443: */
444:
445: of = 0;
446:
1.4 schwarze 447: /*
448: * Maintain a lookaside buffer of all parsed lines. We
449: * only do this if mparse_keep() has been invoked (the
450: * buffer may be accessed with mparse_getkeep()).
451: */
452:
453: if (curp->secondary) {
1.25 schwarze 454: curp->secondary->buf = mandoc_realloc(
455: curp->secondary->buf,
456: curp->secondary->sz + pos + 2);
457: memcpy(curp->secondary->buf +
458: curp->secondary->sz,
459: ln.buf, pos);
1.4 schwarze 460: curp->secondary->sz += pos;
461: curp->secondary->buf
462: [curp->secondary->sz] = '\n';
463: curp->secondary->sz++;
464: curp->secondary->buf
465: [curp->secondary->sz] = '\0';
466: }
1.1 schwarze 467: rerun:
1.25 schwarze 468: rr = roff_parseln(curp->roff, curp->line,
469: &ln.buf, &ln.sz, of, &of);
1.1 schwarze 470:
471: switch (rr) {
1.25 schwarze 472: case ROFF_REPARSE:
1.1 schwarze 473: if (REPARSE_LIMIT >= ++curp->reparse_count)
474: mparse_buf_r(curp, ln, 0);
475: else
476: mandoc_msg(MANDOCERR_ROFFLOOP, curp,
1.25 schwarze 477: curp->line, pos, NULL);
1.1 schwarze 478: pos = 0;
479: continue;
1.25 schwarze 480: case ROFF_APPEND:
1.1 schwarze 481: pos = (int)strlen(ln.buf);
482: continue;
1.25 schwarze 483: case ROFF_RERUN:
1.1 schwarze 484: goto rerun;
1.25 schwarze 485: case ROFF_IGN:
1.1 schwarze 486: pos = 0;
487: continue;
1.25 schwarze 488: case ROFF_ERR:
1.1 schwarze 489: assert(MANDOCLEVEL_FATAL <= curp->file_status);
490: break;
1.25 schwarze 491: case ROFF_SO:
1.23 schwarze 492: if (0 == (MPARSE_SO & curp->options) &&
493: (i >= (int)blk.sz || '\0' == blk.buf[i])) {
494: curp->sodest = mandoc_strdup(ln.buf + of);
495: free(ln.buf);
496: return;
497: }
1.4 schwarze 498: /*
499: * We remove `so' clauses from our lookaside
500: * buffer because we're going to descend into
501: * the file recursively.
502: */
1.25 schwarze 503: if (curp->secondary)
1.4 schwarze 504: curp->secondary->sz -= pos + 1;
1.14 schwarze 505: mparse_readfd(curp, -1, ln.buf + of);
1.30 schwarze 506: if (MANDOCLEVEL_FATAL <= curp->file_status) {
507: mandoc_vmsg(MANDOCERR_SO_FAIL,
508: curp, curp->line, pos,
509: ".so %s", ln.buf + of);
1.1 schwarze 510: break;
1.30 schwarze 511: }
1.1 schwarze 512: pos = 0;
513: continue;
514: default:
515: break;
516: }
517:
518: /*
519: * If we encounter errors in the recursive parse, make
520: * sure we don't continue parsing.
521: */
522:
523: if (MANDOCLEVEL_FATAL <= curp->file_status)
524: break;
525:
526: /*
527: * If input parsers have not been allocated, do so now.
1.2 schwarze 528: * We keep these instanced between parsers, but set them
1.1 schwarze 529: * locally per parse routine since we can use different
530: * parsers with each one.
531: */
532:
533: if ( ! (curp->man || curp->mdoc))
534: pset(ln.buf + of, pos - of, curp);
535:
1.25 schwarze 536: /*
1.1 schwarze 537: * Lastly, push down into the parsers themselves. One
538: * of these will have already been set in the pset()
539: * routine.
540: * If libroff returns ROFF_TBL, then add it to the
541: * currently open parse. Since we only get here if
542: * there does exist data (see tbl_data.c), we're
543: * guaranteed that something's been allocated.
544: * Do the same for ROFF_EQN.
545: */
546:
547: rc = -1;
548:
549: if (ROFF_TBL == rr)
550: while (NULL != (span = roff_span(curp->roff))) {
551: rc = curp->man ?
1.25 schwarze 552: man_addspan(curp->man, span) :
553: mdoc_addspan(curp->mdoc, span);
1.1 schwarze 554: if (0 == rc)
555: break;
556: }
557: else if (ROFF_EQN == rr)
1.25 schwarze 558: rc = curp->mdoc ?
559: mdoc_addeqn(curp->mdoc,
560: roff_eqn(curp->roff)) :
561: man_addeqn(curp->man,
562: roff_eqn(curp->roff));
1.1 schwarze 563: else if (curp->man || curp->mdoc)
564: rc = curp->man ?
1.25 schwarze 565: man_parseln(curp->man,
566: curp->line, ln.buf, of) :
567: mdoc_parseln(curp->mdoc,
568: curp->line, ln.buf, of);
1.1 schwarze 569:
570: if (0 == rc) {
571: assert(MANDOCLEVEL_FATAL <= curp->file_status);
572: break;
1.19 schwarze 573: } else if (2 == rc)
574: break;
1.1 schwarze 575:
576: /* Temporary buffers typically are not full. */
577:
578: if (0 == start && '\0' == blk.buf[i])
579: break;
580:
581: /* Start the next input line. */
582:
583: pos = 0;
584: }
585:
586: free(ln.buf);
587: }
588:
589: static int
1.18 schwarze 590: read_whole_file(struct mparse *curp, const char *file, int fd,
591: struct buf *fb, int *with_mmap)
1.1 schwarze 592: {
593: struct stat st;
594: size_t off;
595: ssize_t ssz;
596:
597: if (-1 == fstat(fd, &st)) {
1.18 schwarze 598: curp->file_status = MANDOCLEVEL_SYSERR;
599: if (curp->mmsg)
600: (*curp->mmsg)(MANDOCERR_SYSSTAT, curp->file_status,
601: file, 0, 0, strerror(errno));
1.1 schwarze 602: return(0);
603: }
604:
605: /*
606: * If we're a regular file, try just reading in the whole entry
607: * via mmap(). This is faster than reading it into blocks, and
608: * since each file is only a few bytes to begin with, I'm not
609: * concerned that this is going to tank any machines.
610: */
611:
612: if (S_ISREG(st.st_mode)) {
613: if (st.st_size >= (1U << 31)) {
1.18 schwarze 614: curp->file_status = MANDOCLEVEL_FATAL;
615: if (curp->mmsg)
616: (*curp->mmsg)(MANDOCERR_TOOLARGE,
617: curp->file_status, file, 0, 0, NULL);
1.1 schwarze 618: return(0);
619: }
620: *with_mmap = 1;
621: fb->sz = (size_t)st.st_size;
1.15 schwarze 622: fb->buf = mmap(NULL, fb->sz, PROT_READ, MAP_SHARED, fd, 0);
1.1 schwarze 623: if (fb->buf != MAP_FAILED)
624: return(1);
625: }
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)) {
1.18 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 schwarze 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) {
1.18 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 schwarze 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:
676: if (curp->mdoc && ! mdoc_endparse(curp->mdoc)) {
677: assert(MANDOCLEVEL_FATAL <= curp->file_status);
678: return;
679: }
680:
681: if (curp->man && ! man_endparse(curp->man)) {
682: assert(MANDOCLEVEL_FATAL <= curp->file_status);
683: return;
684: }
685:
1.23 schwarze 686: if ( ! (curp->mdoc || curp->man || curp->sodest)) {
1.28 schwarze 687: mandoc_msg(MANDOCERR_NOTMANUAL, curp, 0, 0, NULL);
1.1 schwarze 688: curp->file_status = MANDOCLEVEL_FATAL;
689: return;
690: }
691:
692: roff_endparse(curp->roff);
693: }
694:
1.15 schwarze 695: static void
696: mparse_parse_buffer(struct mparse *curp, struct buf blk, const char *file)
1.1 schwarze 697: {
698: const char *svfile;
1.14 schwarze 699: static int recursion_depth;
700:
701: if (64 < recursion_depth) {
702: mandoc_msg(MANDOCERR_ROFFLOOP, curp, curp->line, 0, NULL);
1.15 schwarze 703: return;
1.14 schwarze 704: }
1.1 schwarze 705:
1.15 schwarze 706: /* Line number is per-file. */
707: svfile = curp->file;
708: curp->file = file;
709: curp->line = 1;
710: recursion_depth++;
711:
712: mparse_buf_r(curp, blk, 1);
713:
714: if (0 == --recursion_depth && MANDOCLEVEL_FATAL > curp->file_status)
715: mparse_end(curp);
716:
717: curp->file = svfile;
718: }
719:
720: enum mandoclevel
721: mparse_readfd(struct mparse *curp, int fd, const char *file)
722: {
723: struct buf blk;
724: int with_mmap;
725:
1.18 schwarze 726: if (-1 == fd && -1 == (fd = open(file, O_RDONLY, 0))) {
727: curp->file_status = MANDOCLEVEL_SYSERR;
728: if (curp->mmsg)
729: (*curp->mmsg)(MANDOCERR_SYSOPEN,
730: curp->file_status,
731: file, 0, 0, strerror(errno));
732: goto out;
733: }
734:
1.15 schwarze 735: /*
736: * Run for each opened file; may be called more than once for
737: * each full parse sequence if the opened file is nested (i.e.,
738: * from `so'). Simply sucks in the whole file and moves into
739: * the parse phase for the file.
740: */
1.1 schwarze 741:
1.18 schwarze 742: if ( ! read_whole_file(curp, file, fd, &blk, &with_mmap))
1.15 schwarze 743: goto out;
1.1 schwarze 744:
1.15 schwarze 745: mparse_parse_buffer(curp, blk, file);
1.1 schwarze 746:
1.15 schwarze 747: if (with_mmap)
748: munmap(blk.buf, blk.sz);
749: else
750: free(blk.buf);
1.1 schwarze 751:
752: if (STDIN_FILENO != fd && -1 == close(fd))
753: perror(file);
1.14 schwarze 754: out:
1.1 schwarze 755: return(curp->file_status);
756: }
757:
758: struct mparse *
1.22 schwarze 759: mparse_alloc(int options, enum mandoclevel wlevel,
1.47 schwarze 760: mandocmsg mmsg, const char *defos)
1.1 schwarze 761: {
762: struct mparse *curp;
763:
764: assert(wlevel <= MANDOCLEVEL_FATAL);
765:
766: curp = mandoc_calloc(1, sizeof(struct mparse));
767:
1.22 schwarze 768: curp->options = options;
1.1 schwarze 769: curp->wlevel = wlevel;
770: curp->mmsg = mmsg;
1.7 schwarze 771: curp->defos = defos;
1.1 schwarze 772:
1.22 schwarze 773: curp->roff = roff_alloc(curp, options);
1.1 schwarze 774: return(curp);
775: }
776:
777: void
778: mparse_reset(struct mparse *curp)
779: {
780:
781: roff_reset(curp->roff);
782:
783: if (curp->mdoc)
784: mdoc_reset(curp->mdoc);
785: if (curp->man)
786: man_reset(curp->man);
1.4 schwarze 787: if (curp->secondary)
788: curp->secondary->sz = 0;
1.1 schwarze 789:
790: curp->file_status = MANDOCLEVEL_OK;
791: curp->mdoc = NULL;
792: curp->man = NULL;
1.23 schwarze 793:
794: free(curp->sodest);
795: curp->sodest = NULL;
1.1 schwarze 796: }
797:
798: void
799: mparse_free(struct mparse *curp)
800: {
801:
802: if (curp->pmdoc)
803: mdoc_free(curp->pmdoc);
804: if (curp->pman)
805: man_free(curp->pman);
806: if (curp->roff)
807: roff_free(curp->roff);
1.4 schwarze 808: if (curp->secondary)
809: free(curp->secondary->buf);
1.1 schwarze 810:
1.4 schwarze 811: free(curp->secondary);
1.23 schwarze 812: free(curp->sodest);
1.1 schwarze 813: free(curp);
814: }
815:
816: void
1.23 schwarze 817: mparse_result(struct mparse *curp,
818: struct mdoc **mdoc, struct man **man, char **sodest)
1.1 schwarze 819: {
820:
1.23 schwarze 821: if (sodest && NULL != (*sodest = curp->sodest)) {
822: *mdoc = NULL;
823: *man = NULL;
824: return;
825: }
1.1 schwarze 826: if (mdoc)
827: *mdoc = curp->mdoc;
828: if (man)
829: *man = curp->man;
830: }
831:
832: void
833: mandoc_vmsg(enum mandocerr t, struct mparse *m,
834: int ln, int pos, const char *fmt, ...)
835: {
836: char buf[256];
837: va_list ap;
838:
839: va_start(ap, fmt);
1.26 schwarze 840: (void)vsnprintf(buf, sizeof(buf), fmt, ap);
1.1 schwarze 841: va_end(ap);
842:
843: mandoc_msg(t, m, ln, pos, buf);
844: }
845:
846: void
1.25 schwarze 847: mandoc_msg(enum mandocerr er, struct mparse *m,
1.1 schwarze 848: int ln, int col, const char *msg)
849: {
850: enum mandoclevel level;
851:
852: level = MANDOCLEVEL_FATAL;
853: while (er < mandoclimits[level])
854: level--;
855:
856: if (level < m->wlevel)
857: return;
858:
859: if (m->mmsg)
860: (*m->mmsg)(er, level, m->file, ln, col, msg);
861:
862: if (m->file_status < level)
863: m->file_status = level;
864: }
865:
866: const char *
867: mparse_strerror(enum mandocerr er)
868: {
869:
870: return(mandocerrs[er]);
871: }
872:
873: const char *
874: mparse_strlevel(enum mandoclevel lvl)
875: {
876: return(mandoclevels[lvl]);
1.4 schwarze 877: }
878:
879: void
880: mparse_keep(struct mparse *p)
881: {
882:
883: assert(NULL == p->secondary);
884: p->secondary = mandoc_calloc(1, sizeof(struct buf));
885: }
886:
887: const char *
888: mparse_getkeep(const struct mparse *p)
889: {
890:
891: assert(p->secondary);
892: return(p->secondary->sz ? p->secondary->buf : NULL);
1.1 schwarze 893: }