Annotation of mandoc/read.c, Revision 1.91
1.91 ! schwarze 1: /* $Id: read.c,v 1.90 2014/10/12 19:31:41 schwarze Exp $ */
1.1 kristaps 2: /*
3: * Copyright (c) 2008, 2009, 2010, 2011 Kristaps Dzonsons <kristaps@bsd.lv>
1.40 schwarze 4: * Copyright (c) 2010-2014 Ingo Schwarze <schwarze@openbsd.org>
1.42 schwarze 5: * Copyright (c) 2010, 2012 Joerg Sonnenberger <joerg@netbsd.org>
1.1 kristaps 6: *
7: * Permission to use, copy, modify, and distribute this software for any
8: * purpose with or without fee is hereby granted, provided that the above
9: * copyright notice and this permission notice appear in all copies.
10: *
11: * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
12: * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
13: * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
14: * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
15: * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
16: * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
17: * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
18: */
1.11 kristaps 19: #include "config.h"
20:
1.80 schwarze 21: #include <sys/types.h>
1.81 schwarze 22: #if HAVE_MMAP
1.82 schwarze 23: #include <sys/mman.h>
1.80 schwarze 24: #include <sys/stat.h>
1.15 kristaps 25: #endif
1.82 schwarze 26: #include <sys/wait.h>
1.1 kristaps 27:
28: #include <assert.h>
29: #include <ctype.h>
1.40 schwarze 30: #include <errno.h>
1.1 kristaps 31: #include <fcntl.h>
1.3 kristaps 32: #include <stdarg.h>
1.28 joerg 33: #include <stdint.h>
1.1 kristaps 34: #include <stdio.h>
35: #include <stdlib.h>
36: #include <string.h>
37: #include <unistd.h>
38:
39: #include "mandoc.h"
1.46 schwarze 40: #include "mandoc_aux.h"
1.3 kristaps 41: #include "libmandoc.h"
1.1 kristaps 42: #include "mdoc.h"
43: #include "man.h"
1.28 joerg 44: #include "main.h"
1.1 kristaps 45:
46: #define REPARSE_LIMIT 1000
47:
48: struct buf {
1.47 schwarze 49: char *buf; /* binary input buffer */
1.1 kristaps 50: size_t sz; /* size of binary buffer */
51: };
52:
53: struct mparse {
54: struct man *pman; /* persistent man parser */
55: struct mdoc *pmdoc; /* persistent mdoc parser */
56: struct man *man; /* man parser */
57: struct mdoc *mdoc; /* mdoc parser */
58: struct roff *roff; /* roff parser (!NULL) */
1.45 schwarze 59: char *sodest; /* filename pointed to by .so */
1.83 schwarze 60: const char *file; /* filename of current input file */
61: struct buf *primary; /* buffer currently being parsed */
62: struct buf *secondary; /* preprocessed copy of input */
63: const char *defos; /* default operating system */
64: mandocmsg mmsg; /* warning/error message handler */
65: enum mandoclevel file_status; /* status of current parse */
66: enum mandoclevel wlevel; /* ignore messages below this */
67: int options; /* parser options */
1.1 kristaps 68: int reparse_count; /* finite interp. stack */
1.83 schwarze 69: int line; /* line number in the file */
1.1 kristaps 70: };
71:
1.84 schwarze 72: static void choose_parser(struct mparse *);
1.1 kristaps 73: static void resize_buf(struct buf *, size_t);
74: static void mparse_buf_r(struct mparse *, struct buf, int);
1.40 schwarze 75: static int read_whole_file(struct mparse *, const char *, int,
76: struct buf *, int *);
1.1 kristaps 77: static void mparse_end(struct mparse *);
1.37 schwarze 78: static void mparse_parse_buffer(struct mparse *, struct buf,
79: const char *);
1.1 kristaps 80:
1.3 kristaps 81: static const enum mandocerr mandoclimits[MANDOCLEVEL_MAX] = {
82: MANDOCERR_OK,
83: MANDOCERR_WARNING,
84: MANDOCERR_WARNING,
85: MANDOCERR_ERROR,
86: MANDOCERR_FATAL,
87: MANDOCERR_MAX,
88: MANDOCERR_MAX
89: };
90:
1.7 kristaps 91: static const char * const mandocerrs[MANDOCERR_MAX] = {
92: "ok",
93:
94: "generic warning",
95:
96: /* related to the prologue */
1.79 schwarze 97: "missing manual title, using UNTITLED",
98: "missing manual title, using \"\"",
1.54 schwarze 99: "lower case character in document title",
1.79 schwarze 100: "missing manual section, using \"\"",
1.7 kristaps 101: "unknown manual section",
1.32 schwarze 102: "unknown manual volume or arch",
1.54 schwarze 103: "missing date, using today's date",
1.7 kristaps 104: "cannot parse date, using it verbatim",
1.79 schwarze 105: "missing Os macro, using \"\"",
106: "duplicate prologue macro",
107: "late prologue macro",
108: "skipping late title macro",
1.7 kristaps 109: "prologue macros out of order",
110:
111: /* related to document structure */
112: ".so is fragile, better use ln(1)",
1.50 schwarze 113: "no document body",
1.54 schwarze 114: "content before first section header",
115: "first section is not \"NAME\"",
1.7 kristaps 116: "bad NAME section contents",
117: "sections out of conventional order",
1.54 schwarze 118: "duplicate section title",
119: "unexpected section",
1.87 schwarze 120: "unusual Xr order",
121: "unusual Xr punctuation",
1.86 schwarze 122: "AUTHORS section without An macro",
1.7 kristaps 123:
124: /* related to macros and nesting */
1.55 schwarze 125: "obsolete macro",
1.7 kristaps 126: "skipping paragraph macro",
1.31 schwarze 127: "moving paragraph macro out of list",
1.7 kristaps 128: "skipping no-space macro",
129: "blocks badly nested",
130: "nested displays are not portable",
1.57 schwarze 131: "moving content out of list",
132: ".Vt block has child macro",
1.78 schwarze 133: "fill mode already enabled, skipping",
134: "fill mode already disabled, skipping",
1.7 kristaps 135: "line scope broken",
136:
137: /* related to missing macro arguments */
1.58 schwarze 138: "skipping empty request",
139: "conditional request controls empty scope",
1.7 kristaps 140: "skipping empty macro",
1.62 schwarze 141: "empty argument, using 0n",
1.7 kristaps 142: "argument count wrong",
1.60 schwarze 143: "missing display type, using -ragged",
144: "list type is not the first argument",
145: "missing -width in -tag list, using 8n",
1.78 schwarze 146: "missing utility name, using \"\"",
1.60 schwarze 147: "empty head in list item",
148: "empty list item",
1.61 schwarze 149: "missing font type, using \\fR",
150: "unknown font type, using \\fR",
1.60 schwarze 151: "missing -std argument, adding it",
1.90 schwarze 152: "missing eqn box, using \"\"",
1.7 kristaps 153:
154: /* related to bad macro arguments */
1.64 schwarze 155: "unterminated quoted argument",
1.7 kristaps 156: "duplicate argument",
1.76 schwarze 157: "skipping duplicate argument",
1.63 schwarze 158: "skipping duplicate display type",
159: "skipping duplicate list type",
1.76 schwarze 160: "skipping -width argument",
1.7 kristaps 161: "unknown AT&T UNIX version",
1.88 schwarze 162: "comma in function argument",
1.89 schwarze 163: "parenthesis in function name",
1.67 schwarze 164: "invalid content in Rs block",
1.63 schwarze 165: "invalid Boolean argument",
166: "unknown font, skipping request",
1.7 kristaps 167:
168: /* related to plain text */
1.64 schwarze 169: "blank line in fill mode, using .sp",
170: "tab in filled text",
171: "whitespace at end of input line",
1.7 kristaps 172: "bad comment style",
1.64 schwarze 173: "invalid escape sequence",
174: "undefined string, using \"\"",
1.16 kristaps 175:
1.7 kristaps 176: "generic error",
1.17 kristaps 177:
178: /* related to equations */
1.20 kristaps 179: "unexpected equation scope closure",
180: "equation scope open on exit",
1.21 kristaps 181: "overlapping equation scopes",
182: "unexpected end of equation",
1.7 kristaps 183:
184: /* related to tables */
185: "bad table syntax",
186: "bad table option",
187: "bad table layout",
188: "no table layout cells specified",
189: "no table data cells specified",
190: "ignore data in cell",
191: "data block still open",
192: "ignoring extra data cells",
193:
1.68 schwarze 194: /* related to document structure and macros */
1.7 kristaps 195: "input stack limit exceeded, infinite loop?",
196: "skipping bad character",
1.68 schwarze 197: "skipping unknown macro",
1.73 schwarze 198: "skipping item outside list",
1.68 schwarze 199: "skipping column outside column list",
200: "skipping end of block that is not open",
201: "inserting missing end of block",
202: "appending missing end of block",
203:
204: /* related to request and macro arguments */
1.7 kristaps 205: "escaped character not allowed in a name",
206: "argument count wrong",
1.71 schwarze 207: "missing list type, using -item",
1.70 schwarze 208: "missing manual name, using \"\"",
1.71 schwarze 209: "uname(3) system call failed, using UNKNOWN",
1.63 schwarze 210: "unknown standard specifier",
1.71 schwarze 211: "skipping request without numeric argument",
1.61 schwarze 212: "skipping all arguments",
213: "skipping excess arguments",
1.7 kristaps 214:
215: "generic fatal error",
216:
1.40 schwarze 217: "input too large",
1.78 schwarze 218: "NOT IMPLEMENTED: Bd -file",
1.7 kristaps 219: "NOT IMPLEMENTED: .so with absolute path or \"..\"",
1.52 schwarze 220: ".so request failed",
1.40 schwarze 221:
222: /* system errors */
1.82 schwarze 223: "cannot dup file descriptor",
224: "cannot exec",
225: "gunzip failed with code",
226: "cannot fork",
1.51 schwarze 227: NULL,
1.82 schwarze 228: "cannot open pipe",
229: "cannot read file",
230: "gunzip died from signal",
1.40 schwarze 231: "cannot stat file",
1.82 schwarze 232: "wait failed",
1.7 kristaps 233: };
234:
235: static const char * const mandoclevels[MANDOCLEVEL_MAX] = {
236: "SUCCESS",
237: "RESERVED",
238: "WARNING",
239: "ERROR",
240: "FATAL",
241: "BADARG",
242: "SYSERR"
243: };
244:
1.47 schwarze 245:
1.1 kristaps 246: static void
247: resize_buf(struct buf *buf, size_t initial)
248: {
249:
250: buf->sz = buf->sz > initial/2 ? 2 * buf->sz : initial;
251: buf->buf = mandoc_realloc(buf->buf, buf->sz);
252: }
253:
254: static void
1.84 schwarze 255: choose_parser(struct mparse *curp)
1.1 kristaps 256: {
1.83 schwarze 257: char *cp, *ep;
258: int format;
1.1 kristaps 259:
1.83 schwarze 260: /*
261: * If neither command line arguments -mdoc or -man select
262: * a parser nor the roff parser found a .Dd or .TH macro
263: * yet, look ahead in the main input buffer.
264: */
265:
266: if ((format = roff_getformat(curp->roff)) == 0) {
267: cp = curp->primary->buf;
268: ep = cp + curp->primary->sz;
269: while (cp < ep) {
1.85 schwarze 270: if (*cp == '.' || *cp == '\'') {
1.83 schwarze 271: cp++;
272: if (cp[0] == 'D' && cp[1] == 'd') {
273: format = MPARSE_MDOC;
274: break;
275: }
276: if (cp[0] == 'T' && cp[1] == 'H') {
277: format = MPARSE_MAN;
278: break;
279: }
280: }
281: cp = memchr(cp, '\n', ep - cp);
282: if (cp == NULL)
283: break;
284: cp++;
285: }
1.1 kristaps 286: }
287:
1.83 schwarze 288: if (format == MPARSE_MDOC) {
1.47 schwarze 289: if (NULL == curp->pmdoc)
1.44 schwarze 290: curp->pmdoc = mdoc_alloc(
291: curp->roff, curp, curp->defos,
292: MPARSE_QUICK & curp->options ? 1 : 0);
1.1 kristaps 293: assert(curp->pmdoc);
294: curp->mdoc = curp->pmdoc;
295: return;
1.47 schwarze 296: }
1.1 kristaps 297:
1.83 schwarze 298: /* Fall back to man(7) as a last resort. */
299:
1.47 schwarze 300: if (NULL == curp->pman)
1.44 schwarze 301: curp->pman = man_alloc(curp->roff, curp,
302: MPARSE_QUICK & curp->options ? 1 : 0);
1.1 kristaps 303: assert(curp->pman);
304: curp->man = curp->pman;
305: }
306:
307: /*
308: * Main parse routine for an opened file. This is called for each
309: * opened file and simply loops around the full input file, possibly
310: * nesting (i.e., with `so').
311: */
312: static void
313: mparse_buf_r(struct mparse *curp, struct buf blk, int start)
314: {
315: const struct tbl_span *span;
316: struct buf ln;
317: enum rofferr rr;
318: int i, of, rc;
319: int pos; /* byte number in the ln buffer */
320: int lnn; /* line number in the real file */
321: unsigned char c;
322:
323: memset(&ln, 0, sizeof(struct buf));
324:
1.47 schwarze 325: lnn = curp->line;
326: pos = 0;
1.1 kristaps 327:
328: for (i = 0; i < (int)blk.sz; ) {
329: if (0 == pos && '\0' == blk.buf[i])
330: break;
331:
332: if (start) {
333: curp->line = lnn;
334: curp->reparse_count = 0;
335: }
336:
337: while (i < (int)blk.sz && (start || '\0' != blk.buf[i])) {
338:
339: /*
340: * When finding an unescaped newline character,
341: * leave the character loop to process the line.
342: * Skip a preceding carriage return, if any.
343: */
344:
345: if ('\r' == blk.buf[i] && i + 1 < (int)blk.sz &&
346: '\n' == blk.buf[i + 1])
347: ++i;
348: if ('\n' == blk.buf[i]) {
349: ++i;
350: ++lnn;
351: break;
352: }
353:
1.35 schwarze 354: /*
355: * Make sure we have space for at least
356: * one backslash and one other character
357: * and the trailing NUL byte.
358: */
359:
360: if (pos + 2 >= (int)ln.sz)
361: resize_buf(&ln, 256);
362:
1.47 schwarze 363: /*
1.1 kristaps 364: * Warn about bogus characters. If you're using
365: * non-ASCII encoding, you're screwing your
366: * readers. Since I'd rather this not happen,
1.27 joerg 367: * I'll be helpful and replace these characters
368: * with "?", so we don't display gibberish.
369: * Note to manual writers: use special characters.
1.1 kristaps 370: */
371:
372: c = (unsigned char) blk.buf[i];
373:
1.47 schwarze 374: if ( ! (isascii(c) &&
375: (isgraph(c) || isblank(c)))) {
1.78 schwarze 376: mandoc_vmsg(MANDOCERR_BADCHAR, curp,
377: curp->line, pos, "0x%x", c);
1.1 kristaps 378: i++;
1.27 joerg 379: ln.buf[pos++] = '?';
1.1 kristaps 380: continue;
381: }
382:
383: /* Trailing backslash = a plain char. */
384:
385: if ('\\' != blk.buf[i] || i + 1 == (int)blk.sz) {
386: ln.buf[pos++] = blk.buf[i++];
387: continue;
388: }
389:
390: /*
391: * Found escape and at least one other character.
392: * When it's a newline character, skip it.
393: * When there is a carriage return in between,
394: * skip that one as well.
395: */
396:
397: if ('\r' == blk.buf[i + 1] && i + 2 < (int)blk.sz &&
398: '\n' == blk.buf[i + 2])
399: ++i;
400: if ('\n' == blk.buf[i + 1]) {
401: i += 2;
402: ++lnn;
403: continue;
404: }
405:
1.13 kristaps 406: if ('"' == blk.buf[i + 1] || '#' == blk.buf[i + 1]) {
1.1 kristaps 407: i += 2;
408: /* Comment, skip to end of line */
409: for (; i < (int)blk.sz; ++i) {
410: if ('\n' == blk.buf[i]) {
411: ++i;
412: ++lnn;
413: break;
414: }
415: }
416:
417: /* Backout trailing whitespaces */
418: for (; pos > 0; --pos) {
419: if (ln.buf[pos - 1] != ' ')
420: break;
421: if (pos > 2 && ln.buf[pos - 2] == '\\')
422: break;
423: }
424: break;
425: }
426:
1.35 schwarze 427: /* Catch escaped bogus characters. */
428:
429: c = (unsigned char) blk.buf[i+1];
430:
1.47 schwarze 431: if ( ! (isascii(c) &&
432: (isgraph(c) || isblank(c)))) {
1.78 schwarze 433: mandoc_vmsg(MANDOCERR_BADCHAR, curp,
434: curp->line, pos, "0x%x", c);
1.35 schwarze 435: i += 2;
436: ln.buf[pos++] = '?';
437: continue;
438: }
439:
1.1 kristaps 440: /* Some other escape sequence, copy & cont. */
441:
442: ln.buf[pos++] = blk.buf[i++];
443: ln.buf[pos++] = blk.buf[i++];
444: }
445:
1.47 schwarze 446: if (pos >= (int)ln.sz)
1.1 kristaps 447: resize_buf(&ln, 256);
448:
449: ln.buf[pos] = '\0';
450:
451: /*
452: * A significant amount of complexity is contained by
453: * the roff preprocessor. It's line-oriented but can be
454: * expressed on one line, so we need at times to
455: * readjust our starting point and re-run it. The roff
456: * preprocessor can also readjust the buffers with new
457: * data, so we pass them in wholesale.
458: */
459:
460: of = 0;
461:
1.24 kristaps 462: /*
463: * Maintain a lookaside buffer of all parsed lines. We
464: * only do this if mparse_keep() has been invoked (the
465: * buffer may be accessed with mparse_getkeep()).
466: */
467:
468: if (curp->secondary) {
1.47 schwarze 469: curp->secondary->buf = mandoc_realloc(
470: curp->secondary->buf,
471: curp->secondary->sz + pos + 2);
472: memcpy(curp->secondary->buf +
473: curp->secondary->sz,
474: ln.buf, pos);
1.24 kristaps 475: curp->secondary->sz += pos;
476: curp->secondary->buf
477: [curp->secondary->sz] = '\n';
478: curp->secondary->sz++;
479: curp->secondary->buf
480: [curp->secondary->sz] = '\0';
481: }
1.1 kristaps 482: rerun:
1.47 schwarze 483: rr = roff_parseln(curp->roff, curp->line,
484: &ln.buf, &ln.sz, of, &of);
1.1 kristaps 485:
486: switch (rr) {
1.47 schwarze 487: case ROFF_REPARSE:
1.1 kristaps 488: if (REPARSE_LIMIT >= ++curp->reparse_count)
489: mparse_buf_r(curp, ln, 0);
490: else
1.3 kristaps 491: mandoc_msg(MANDOCERR_ROFFLOOP, curp,
1.47 schwarze 492: curp->line, pos, NULL);
1.1 kristaps 493: pos = 0;
494: continue;
1.47 schwarze 495: case ROFF_APPEND:
1.1 kristaps 496: pos = (int)strlen(ln.buf);
497: continue;
1.47 schwarze 498: case ROFF_RERUN:
1.1 kristaps 499: goto rerun;
1.47 schwarze 500: case ROFF_IGN:
1.1 kristaps 501: pos = 0;
502: continue;
1.47 schwarze 503: case ROFF_ERR:
1.1 kristaps 504: assert(MANDOCLEVEL_FATAL <= curp->file_status);
505: break;
1.47 schwarze 506: case ROFF_SO:
1.45 schwarze 507: if (0 == (MPARSE_SO & curp->options) &&
508: (i >= (int)blk.sz || '\0' == blk.buf[i])) {
509: curp->sodest = mandoc_strdup(ln.buf + of);
510: free(ln.buf);
511: return;
512: }
1.24 kristaps 513: /*
514: * We remove `so' clauses from our lookaside
515: * buffer because we're going to descend into
516: * the file recursively.
517: */
1.47 schwarze 518: if (curp->secondary)
1.25 kristaps 519: curp->secondary->sz -= pos + 1;
1.36 schwarze 520: mparse_readfd(curp, -1, ln.buf + of);
1.52 schwarze 521: if (MANDOCLEVEL_FATAL <= curp->file_status) {
522: mandoc_vmsg(MANDOCERR_SO_FAIL,
523: curp, curp->line, pos,
524: ".so %s", ln.buf + of);
1.1 kristaps 525: break;
1.52 schwarze 526: }
1.1 kristaps 527: pos = 0;
528: continue;
529: default:
530: break;
531: }
532:
533: /*
534: * If we encounter errors in the recursive parse, make
535: * sure we don't continue parsing.
536: */
537:
538: if (MANDOCLEVEL_FATAL <= curp->file_status)
539: break;
540:
541: /*
542: * If input parsers have not been allocated, do so now.
1.14 kristaps 543: * We keep these instanced between parsers, but set them
1.1 kristaps 544: * locally per parse routine since we can use different
545: * parsers with each one.
546: */
547:
548: if ( ! (curp->man || curp->mdoc))
1.84 schwarze 549: choose_parser(curp);
1.1 kristaps 550:
1.47 schwarze 551: /*
1.84 schwarze 552: * Lastly, push down into the parsers themselves.
1.1 kristaps 553: * If libroff returns ROFF_TBL, then add it to the
554: * currently open parse. Since we only get here if
555: * there does exist data (see tbl_data.c), we're
556: * guaranteed that something's been allocated.
557: * Do the same for ROFF_EQN.
558: */
559:
560: rc = -1;
561:
562: if (ROFF_TBL == rr)
563: while (NULL != (span = roff_span(curp->roff))) {
564: rc = curp->man ?
1.47 schwarze 565: man_addspan(curp->man, span) :
566: mdoc_addspan(curp->mdoc, span);
1.1 kristaps 567: if (0 == rc)
568: break;
569: }
570: else if (ROFF_EQN == rr)
1.47 schwarze 571: rc = curp->mdoc ?
572: mdoc_addeqn(curp->mdoc,
573: roff_eqn(curp->roff)) :
574: man_addeqn(curp->man,
575: roff_eqn(curp->roff));
1.1 kristaps 576: else if (curp->man || curp->mdoc)
577: rc = curp->man ?
1.47 schwarze 578: man_parseln(curp->man,
579: curp->line, ln.buf, of) :
580: mdoc_parseln(curp->mdoc,
581: curp->line, ln.buf, of);
1.1 kristaps 582:
583: if (0 == rc) {
584: assert(MANDOCLEVEL_FATAL <= curp->file_status);
585: break;
1.41 schwarze 586: } else if (2 == rc)
587: break;
1.1 kristaps 588:
589: /* Temporary buffers typically are not full. */
590:
591: if (0 == start && '\0' == blk.buf[i])
592: break;
593:
594: /* Start the next input line. */
595:
596: pos = 0;
597: }
598:
599: free(ln.buf);
600: }
601:
602: static int
1.40 schwarze 603: read_whole_file(struct mparse *curp, const char *file, int fd,
604: struct buf *fb, int *with_mmap)
1.1 kristaps 605: {
606: size_t off;
607: ssize_t ssz;
608:
1.81 schwarze 609: #if HAVE_MMAP
1.15 kristaps 610: struct stat st;
1.1 kristaps 611: if (-1 == fstat(fd, &st)) {
1.40 schwarze 612: curp->file_status = MANDOCLEVEL_SYSERR;
613: if (curp->mmsg)
614: (*curp->mmsg)(MANDOCERR_SYSSTAT, curp->file_status,
615: file, 0, 0, strerror(errno));
1.1 kristaps 616: return(0);
617: }
618:
619: /*
620: * If we're a regular file, try just reading in the whole entry
621: * via mmap(). This is faster than reading it into blocks, and
622: * since each file is only a few bytes to begin with, I'm not
623: * concerned that this is going to tank any machines.
624: */
625:
626: if (S_ISREG(st.st_mode)) {
627: if (st.st_size >= (1U << 31)) {
1.40 schwarze 628: curp->file_status = MANDOCLEVEL_FATAL;
629: if (curp->mmsg)
630: (*curp->mmsg)(MANDOCERR_TOOLARGE,
631: curp->file_status, file, 0, 0, NULL);
1.1 kristaps 632: return(0);
633: }
634: *with_mmap = 1;
635: fb->sz = (size_t)st.st_size;
1.37 schwarze 636: fb->buf = mmap(NULL, fb->sz, PROT_READ, MAP_SHARED, fd, 0);
1.1 kristaps 637: if (fb->buf != MAP_FAILED)
638: return(1);
639: }
1.15 kristaps 640: #endif
1.1 kristaps 641:
642: /*
643: * If this isn't a regular file (like, say, stdin), then we must
644: * go the old way and just read things in bit by bit.
645: */
646:
647: *with_mmap = 0;
648: off = 0;
649: fb->sz = 0;
650: fb->buf = NULL;
651: for (;;) {
652: if (off == fb->sz) {
653: if (fb->sz == (1U << 31)) {
1.40 schwarze 654: curp->file_status = MANDOCLEVEL_FATAL;
655: if (curp->mmsg)
656: (*curp->mmsg)(MANDOCERR_TOOLARGE,
657: curp->file_status,
658: file, 0, 0, NULL);
1.1 kristaps 659: break;
660: }
661: resize_buf(fb, 65536);
662: }
663: ssz = read(fd, fb->buf + (int)off, fb->sz - off);
664: if (ssz == 0) {
665: fb->sz = off;
666: return(1);
667: }
668: if (ssz == -1) {
1.40 schwarze 669: curp->file_status = MANDOCLEVEL_SYSERR;
670: if (curp->mmsg)
671: (*curp->mmsg)(MANDOCERR_SYSREAD,
672: curp->file_status, file, 0, 0,
673: strerror(errno));
1.1 kristaps 674: break;
675: }
676: off += (size_t)ssz;
677: }
678:
679: free(fb->buf);
680: fb->buf = NULL;
681: return(0);
682: }
683:
684: static void
685: mparse_end(struct mparse *curp)
686: {
687:
688: if (MANDOCLEVEL_FATAL <= curp->file_status)
689: return;
690:
1.72 schwarze 691: if (curp->mdoc == NULL &&
692: curp->man == NULL &&
693: curp->sodest == NULL) {
694: if (curp->options & MPARSE_MDOC)
695: curp->mdoc = curp->pmdoc;
696: else {
697: if (curp->pman == NULL)
698: curp->pman = man_alloc(curp->roff, curp,
699: curp->options & MPARSE_QUICK ? 1 : 0);
700: curp->man = curp->pman;
701: }
702: }
703:
1.1 kristaps 704: if (curp->mdoc && ! mdoc_endparse(curp->mdoc)) {
705: assert(MANDOCLEVEL_FATAL <= curp->file_status);
706: return;
707: }
708:
709: if (curp->man && ! man_endparse(curp->man)) {
710: assert(MANDOCLEVEL_FATAL <= curp->file_status);
711: return;
712: }
713:
714: roff_endparse(curp->roff);
715: }
716:
717: static void
1.36 schwarze 718: mparse_parse_buffer(struct mparse *curp, struct buf blk, const char *file)
1.28 joerg 719: {
1.85 schwarze 720: struct buf *svprimary;
1.28 joerg 721: const char *svfile;
1.36 schwarze 722: static int recursion_depth;
723:
724: if (64 < recursion_depth) {
725: mandoc_msg(MANDOCERR_ROFFLOOP, curp, curp->line, 0, NULL);
726: return;
727: }
1.28 joerg 728:
729: /* Line number is per-file. */
730: svfile = curp->file;
731: curp->file = file;
1.85 schwarze 732: svprimary = curp->primary;
1.83 schwarze 733: curp->primary = &blk;
1.28 joerg 734: curp->line = 1;
1.36 schwarze 735: recursion_depth++;
1.28 joerg 736:
737: mparse_buf_r(curp, blk, 1);
738:
1.36 schwarze 739: if (0 == --recursion_depth && MANDOCLEVEL_FATAL > curp->file_status)
1.28 joerg 740: mparse_end(curp);
741:
1.85 schwarze 742: curp->primary = svprimary;
1.28 joerg 743: curp->file = svfile;
744: }
745:
746: enum mandoclevel
747: mparse_readmem(struct mparse *curp, const void *buf, size_t len,
748: const char *file)
749: {
750: struct buf blk;
751:
752: blk.buf = UNCONST(buf);
753: blk.sz = len;
754:
1.36 schwarze 755: mparse_parse_buffer(curp, blk, file);
1.28 joerg 756: return(curp->file_status);
757: }
758:
1.36 schwarze 759: enum mandoclevel
760: mparse_readfd(struct mparse *curp, int fd, const char *file)
1.1 kristaps 761: {
1.28 joerg 762: struct buf blk;
763: int with_mmap;
1.1 kristaps 764:
1.40 schwarze 765: if (-1 == fd && -1 == (fd = open(file, O_RDONLY, 0))) {
766: curp->file_status = MANDOCLEVEL_SYSERR;
767: if (curp->mmsg)
768: (*curp->mmsg)(MANDOCERR_SYSOPEN,
769: curp->file_status,
770: file, 0, 0, strerror(errno));
1.91 ! schwarze 771: return(curp->file_status);
1.40 schwarze 772: }
773:
1.28 joerg 774: /*
775: * Run for each opened file; may be called more than once for
776: * each full parse sequence if the opened file is nested (i.e.,
777: * from `so'). Simply sucks in the whole file and moves into
778: * the parse phase for the file.
779: */
1.1 kristaps 780:
1.91 ! schwarze 781: if (read_whole_file(curp, file, fd, &blk, &with_mmap)) {
! 782: mparse_parse_buffer(curp, blk, file);
1.81 schwarze 783: #if HAVE_MMAP
1.91 ! schwarze 784: if (with_mmap)
! 785: munmap(blk.buf, blk.sz);
! 786: else
1.28 joerg 787: #endif
1.91 ! schwarze 788: free(blk.buf);
! 789: }
1.1 kristaps 790:
791: if (STDIN_FILENO != fd && -1 == close(fd))
792: perror(file);
1.91 ! schwarze 793:
1.1 kristaps 794: return(curp->file_status);
1.82 schwarze 795: }
796:
797: enum mandoclevel
798: mparse_open(struct mparse *curp, int *fd, const char *file,
799: pid_t *child_pid)
800: {
801: int pfd[2];
802: char *cp;
803: enum mandocerr err;
804:
805: pfd[1] = -1;
806: curp->file = file;
807: if ((cp = strrchr(file, '.')) == NULL ||
808: strcmp(cp + 1, "gz")) {
809: *child_pid = 0;
810: if ((*fd = open(file, O_RDONLY)) == -1) {
811: err = MANDOCERR_SYSOPEN;
812: goto out;
813: }
814: return(MANDOCLEVEL_OK);
815: }
816:
817: if (pipe(pfd) == -1) {
818: err = MANDOCERR_SYSPIPE;
819: goto out;
820: }
821:
822: switch (*child_pid = fork()) {
823: case -1:
824: err = MANDOCERR_SYSFORK;
825: close(pfd[0]);
826: close(pfd[1]);
827: pfd[1] = -1;
828: break;
829: case 0:
830: close(pfd[0]);
831: if (dup2(pfd[1], STDOUT_FILENO) == -1) {
832: err = MANDOCERR_SYSDUP;
833: break;
834: }
835: execlp("gunzip", "gunzip", "-c", file, NULL);
836: err = MANDOCERR_SYSEXEC;
837: break;
838: default:
839: close(pfd[1]);
840: *fd = pfd[0];
841: return(MANDOCLEVEL_OK);
842: }
843:
844: out:
845: *fd = -1;
846: *child_pid = 0;
847: curp->file_status = MANDOCLEVEL_SYSERR;
848: if (curp->mmsg)
849: (*curp->mmsg)(err, curp->file_status, file,
850: 0, 0, strerror(errno));
851: if (pfd[1] != -1)
852: exit(1);
853: return(curp->file_status);
854: }
855:
856: enum mandoclevel
857: mparse_wait(struct mparse *curp, pid_t child_pid)
858: {
859: int status;
860:
861: if (waitpid(child_pid, &status, 0) == -1) {
862: mandoc_msg(MANDOCERR_SYSWAIT, curp, 0, 0,
863: strerror(errno));
864: curp->file_status = MANDOCLEVEL_SYSERR;
865: return(curp->file_status);
866: }
867: if (WIFSIGNALED(status)) {
868: mandoc_vmsg(MANDOCERR_SYSSIG, curp, 0, 0,
869: "%d", WTERMSIG(status));
870: curp->file_status = MANDOCLEVEL_SYSERR;
871: return(curp->file_status);
872: }
873: if (WEXITSTATUS(status)) {
874: mandoc_vmsg(MANDOCERR_SYSEXIT, curp, 0, 0,
875: "%d", WEXITSTATUS(status));
876: curp->file_status = MANDOCLEVEL_SYSERR;
877: return(curp->file_status);
878: }
879: return(MANDOCLEVEL_OK);
1.1 kristaps 880: }
881:
882: struct mparse *
1.44 schwarze 883: mparse_alloc(int options, enum mandoclevel wlevel,
1.69 schwarze 884: mandocmsg mmsg, const char *defos)
1.1 kristaps 885: {
886: struct mparse *curp;
1.10 kristaps 887:
888: assert(wlevel <= MANDOCLEVEL_FATAL);
1.1 kristaps 889:
890: curp = mandoc_calloc(1, sizeof(struct mparse));
891:
1.44 schwarze 892: curp->options = options;
1.3 kristaps 893: curp->wlevel = wlevel;
1.1 kristaps 894: curp->mmsg = mmsg;
1.29 schwarze 895: curp->defos = defos;
1.1 kristaps 896:
1.44 schwarze 897: curp->roff = roff_alloc(curp, options);
1.72 schwarze 898: if (curp->options & MPARSE_MDOC)
899: curp->pmdoc = mdoc_alloc(
900: curp->roff, curp, curp->defos,
901: curp->options & MPARSE_QUICK ? 1 : 0);
902: if (curp->options & MPARSE_MAN)
903: curp->pman = man_alloc(curp->roff, curp,
904: curp->options & MPARSE_QUICK ? 1 : 0);
905:
1.1 kristaps 906: return(curp);
907: }
908:
909: void
910: mparse_reset(struct mparse *curp)
911: {
912:
913: roff_reset(curp->roff);
914:
915: if (curp->mdoc)
916: mdoc_reset(curp->mdoc);
917: if (curp->man)
918: man_reset(curp->man);
1.24 kristaps 919: if (curp->secondary)
920: curp->secondary->sz = 0;
1.1 kristaps 921:
922: curp->file_status = MANDOCLEVEL_OK;
923: curp->mdoc = NULL;
924: curp->man = NULL;
1.45 schwarze 925:
926: free(curp->sodest);
927: curp->sodest = NULL;
1.1 kristaps 928: }
929:
930: void
931: mparse_free(struct mparse *curp)
932: {
933:
934: if (curp->pmdoc)
935: mdoc_free(curp->pmdoc);
936: if (curp->pman)
937: man_free(curp->pman);
938: if (curp->roff)
939: roff_free(curp->roff);
1.24 kristaps 940: if (curp->secondary)
941: free(curp->secondary->buf);
1.1 kristaps 942:
1.24 kristaps 943: free(curp->secondary);
1.45 schwarze 944: free(curp->sodest);
1.1 kristaps 945: free(curp);
946: }
947:
948: void
1.45 schwarze 949: mparse_result(struct mparse *curp,
950: struct mdoc **mdoc, struct man **man, char **sodest)
1.1 kristaps 951: {
952:
1.45 schwarze 953: if (sodest && NULL != (*sodest = curp->sodest)) {
954: *mdoc = NULL;
955: *man = NULL;
956: return;
957: }
1.9 kristaps 958: if (mdoc)
959: *mdoc = curp->mdoc;
960: if (man)
961: *man = curp->man;
1.3 kristaps 962: }
963:
964: void
965: mandoc_vmsg(enum mandocerr t, struct mparse *m,
966: int ln, int pos, const char *fmt, ...)
967: {
968: char buf[256];
969: va_list ap;
970:
971: va_start(ap, fmt);
1.48 schwarze 972: (void)vsnprintf(buf, sizeof(buf), fmt, ap);
1.3 kristaps 973: va_end(ap);
974:
975: mandoc_msg(t, m, ln, pos, buf);
976: }
977:
978: void
1.47 schwarze 979: mandoc_msg(enum mandocerr er, struct mparse *m,
1.3 kristaps 980: int ln, int col, const char *msg)
981: {
982: enum mandoclevel level;
983:
984: level = MANDOCLEVEL_FATAL;
985: while (er < mandoclimits[level])
986: level--;
987:
988: if (level < m->wlevel)
989: return;
990:
1.8 kristaps 991: if (m->mmsg)
992: (*m->mmsg)(er, level, m->file, ln, col, msg);
1.3 kristaps 993:
994: if (m->file_status < level)
995: m->file_status = level;
1.7 kristaps 996: }
997:
998: const char *
999: mparse_strerror(enum mandocerr er)
1000: {
1001:
1002: return(mandocerrs[er]);
1003: }
1004:
1005: const char *
1006: mparse_strlevel(enum mandoclevel lvl)
1007: {
1008: return(mandoclevels[lvl]);
1.24 kristaps 1009: }
1010:
1011: void
1012: mparse_keep(struct mparse *p)
1013: {
1014:
1015: assert(NULL == p->secondary);
1016: p->secondary = mandoc_calloc(1, sizeof(struct buf));
1017: }
1018:
1019: const char *
1020: mparse_getkeep(const struct mparse *p)
1021: {
1022:
1023: assert(p->secondary);
1024: return(p->secondary->sz ? p->secondary->buf : NULL);
1.1 kristaps 1025: }
CVSweb