Annotation of src/usr.bin/mandoc/read.c, Revision 1.83
1.83 ! schwarze 1: /* $OpenBSD: read.c,v 1.82 2015/01/14 17:45:25 schwarze Exp $ */
1.1 schwarze 2: /*
3: * Copyright (c) 2008, 2009, 2010, 2011 Kristaps Dzonsons <kristaps@bsd.lv>
1.82 schwarze 4: * Copyright (c) 2010-2015 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: */
1.58 schwarze 19: #include <sys/types.h>
20: #include <sys/mman.h>
1.1 schwarze 21: #include <sys/stat.h>
1.58 schwarze 22: #include <sys/wait.h>
1.1 schwarze 23:
24: #include <assert.h>
25: #include <ctype.h>
1.18 schwarze 26: #include <errno.h>
1.1 schwarze 27: #include <fcntl.h>
28: #include <stdarg.h>
29: #include <stdio.h>
30: #include <stdlib.h>
31: #include <string.h>
32: #include <unistd.h>
33:
34: #include "mandoc.h"
1.24 schwarze 35: #include "mandoc_aux.h"
1.1 schwarze 36: #include "libmandoc.h"
37: #include "mdoc.h"
38: #include "man.h"
39:
40: #define REPARSE_LIMIT 1000
41:
42: struct mparse {
43: struct man *pman; /* persistent man parser */
44: struct mdoc *pmdoc; /* persistent mdoc parser */
45: struct man *man; /* man parser */
46: struct mdoc *mdoc; /* mdoc parser */
47: struct roff *roff; /* roff parser (!NULL) */
1.69 schwarze 48: const struct mchars *mchars; /* character table */
1.23 schwarze 49: char *sodest; /* filename pointed to by .so */
1.59 schwarze 50: const char *file; /* filename of current input file */
51: struct buf *primary; /* buffer currently being parsed */
52: struct buf *secondary; /* preprocessed copy of input */
53: const char *defos; /* default operating system */
54: mandocmsg mmsg; /* warning/error message handler */
55: enum mandoclevel file_status; /* status of current parse */
56: enum mandoclevel wlevel; /* ignore messages below this */
57: int options; /* parser options */
1.70 schwarze 58: int filenc; /* encoding of the current file */
1.1 schwarze 59: int reparse_count; /* finite interp. stack */
1.59 schwarze 60: int line; /* line number in the file */
1.73 schwarze 61: pid_t child; /* the gunzip(1) process */
1.1 schwarze 62: };
63:
1.60 schwarze 64: static void choose_parser(struct mparse *);
1.1 schwarze 65: static void resize_buf(struct buf *, size_t);
1.71 schwarze 66: static void mparse_buf_r(struct mparse *, struct buf, size_t, int);
1.18 schwarze 67: static int read_whole_file(struct mparse *, const char *, int,
68: struct buf *, int *);
1.1 schwarze 69: static void mparse_end(struct mparse *);
1.15 schwarze 70: static void mparse_parse_buffer(struct mparse *, struct buf,
71: const char *);
1.1 schwarze 72:
73: static const enum mandocerr mandoclimits[MANDOCLEVEL_MAX] = {
74: MANDOCERR_OK,
75: MANDOCERR_WARNING,
76: MANDOCERR_WARNING,
77: MANDOCERR_ERROR,
78: MANDOCERR_FATAL,
79: MANDOCERR_MAX,
80: MANDOCERR_MAX
81: };
82:
83: static const char * const mandocerrs[MANDOCERR_MAX] = {
84: "ok",
85:
86: "generic warning",
87:
88: /* related to the prologue */
1.57 schwarze 89: "missing manual title, using UNTITLED",
90: "missing manual title, using \"\"",
1.32 schwarze 91: "lower case character in document title",
1.57 schwarze 92: "missing manual section, using \"\"",
1.1 schwarze 93: "unknown manual section",
1.32 schwarze 94: "missing date, using today's date",
1.1 schwarze 95: "cannot parse date, using it verbatim",
1.57 schwarze 96: "missing Os macro, using \"\"",
97: "duplicate prologue macro",
98: "late prologue macro",
99: "skipping late title macro",
1.1 schwarze 100: "prologue macros out of order",
101:
102: /* related to document structure */
103: ".so is fragile, better use ln(1)",
1.28 schwarze 104: "no document body",
1.32 schwarze 105: "content before first section header",
106: "first section is not \"NAME\"",
1.1 schwarze 107: "bad NAME section contents",
108: "sections out of conventional order",
1.32 schwarze 109: "duplicate section title",
110: "unexpected section",
1.63 schwarze 111: "unusual Xr order",
112: "unusual Xr punctuation",
1.62 schwarze 113: "AUTHORS section without An macro",
1.1 schwarze 114:
115: /* related to macros and nesting */
1.33 schwarze 116: "obsolete macro",
1.78 schwarze 117: "macro neither callable nor escaped",
1.1 schwarze 118: "skipping paragraph macro",
1.10 schwarze 119: "moving paragraph macro out of list",
1.1 schwarze 120: "skipping no-space macro",
121: "blocks badly nested",
122: "nested displays are not portable",
1.35 schwarze 123: "moving content out of list",
124: ".Vt block has child macro",
1.56 schwarze 125: "fill mode already enabled, skipping",
126: "fill mode already disabled, skipping",
1.1 schwarze 127: "line scope broken",
128:
129: /* related to missing macro arguments */
1.36 schwarze 130: "skipping empty request",
131: "conditional request controls empty scope",
1.1 schwarze 132: "skipping empty macro",
1.40 schwarze 133: "empty argument, using 0n",
1.1 schwarze 134: "argument count wrong",
1.38 schwarze 135: "missing display type, using -ragged",
136: "list type is not the first argument",
137: "missing -width in -tag list, using 8n",
1.56 schwarze 138: "missing utility name, using \"\"",
1.38 schwarze 139: "empty head in list item",
140: "empty list item",
1.39 schwarze 141: "missing font type, using \\fR",
142: "unknown font type, using \\fR",
1.79 schwarze 143: "nothing follows prefix",
1.38 schwarze 144: "missing -std argument, adding it",
1.66 schwarze 145: "missing eqn box, using \"\"",
1.1 schwarze 146:
147: /* related to bad macro arguments */
1.42 schwarze 148: "unterminated quoted argument",
1.1 schwarze 149: "duplicate argument",
1.54 schwarze 150: "skipping duplicate argument",
1.41 schwarze 151: "skipping duplicate display type",
152: "skipping duplicate list type",
1.54 schwarze 153: "skipping -width argument",
1.1 schwarze 154: "unknown AT&T UNIX version",
1.64 schwarze 155: "comma in function argument",
1.65 schwarze 156: "parenthesis in function name",
1.45 schwarze 157: "invalid content in Rs block",
1.41 schwarze 158: "invalid Boolean argument",
159: "unknown font, skipping request",
1.1 schwarze 160:
161: /* related to plain text */
1.42 schwarze 162: "blank line in fill mode, using .sp",
163: "tab in filled text",
164: "whitespace at end of input line",
1.1 schwarze 165: "bad comment style",
1.42 schwarze 166: "invalid escape sequence",
167: "undefined string, using \"\"",
1.3 schwarze 168:
1.1 schwarze 169: "generic error",
170:
1.3 schwarze 171: /* related to equations */
172: "unexpected equation scope closure",
173: "equation scope open on exit",
174: "overlapping equation scopes",
175: "unexpected end of equation",
176:
1.1 schwarze 177: /* related to tables */
178: "bad table syntax",
179: "bad table option",
180: "bad table layout",
181: "no table layout cells specified",
182: "no table data cells specified",
183: "ignore data in cell",
184: "data block still open",
185: "ignoring extra data cells",
1.80 schwarze 186: "ignoring macro in table",
1.1 schwarze 187:
1.46 schwarze 188: /* related to document structure and macros */
1.83 ! schwarze 189: NULL,
1.1 schwarze 190: "input stack limit exceeded, infinite loop?",
191: "skipping bad character",
1.46 schwarze 192: "skipping unknown macro",
1.51 schwarze 193: "skipping item outside list",
1.46 schwarze 194: "skipping column outside column list",
195: "skipping end of block that is not open",
196: "inserting missing end of block",
197: "appending missing end of block",
198:
199: /* related to request and macro arguments */
1.1 schwarze 200: "escaped character not allowed in a name",
201: "argument count wrong",
1.75 schwarze 202: "NOT IMPLEMENTED: Bd -file",
1.49 schwarze 203: "missing list type, using -item",
1.48 schwarze 204: "missing manual name, using \"\"",
1.49 schwarze 205: "uname(3) system call failed, using UNKNOWN",
1.41 schwarze 206: "unknown standard specifier",
1.49 schwarze 207: "skipping request without numeric argument",
1.39 schwarze 208: "skipping all arguments",
209: "skipping excess arguments",
1.68 schwarze 210: "divide by zero",
1.1 schwarze 211:
212: "generic fatal error",
213:
1.18 schwarze 214: "input too large",
1.1 schwarze 215: "NOT IMPLEMENTED: .so with absolute path or \"..\"",
1.30 schwarze 216: ".so request failed",
1.1 schwarze 217: };
218:
219: static const char * const mandoclevels[MANDOCLEVEL_MAX] = {
220: "SUCCESS",
221: "RESERVED",
222: "WARNING",
223: "ERROR",
224: "FATAL",
225: "BADARG",
226: "SYSERR"
227: };
228:
1.25 schwarze 229:
1.1 schwarze 230: static void
231: resize_buf(struct buf *buf, size_t initial)
232: {
233:
234: buf->sz = buf->sz > initial/2 ? 2 * buf->sz : initial;
235: buf->buf = mandoc_realloc(buf->buf, buf->sz);
236: }
237:
238: static void
1.60 schwarze 239: choose_parser(struct mparse *curp)
1.1 schwarze 240: {
1.59 schwarze 241: char *cp, *ep;
242: int format;
1.1 schwarze 243:
1.59 schwarze 244: /*
245: * If neither command line arguments -mdoc or -man select
246: * a parser nor the roff parser found a .Dd or .TH macro
247: * yet, look ahead in the main input buffer.
248: */
249:
250: if ((format = roff_getformat(curp->roff)) == 0) {
251: cp = curp->primary->buf;
252: ep = cp + curp->primary->sz;
253: while (cp < ep) {
1.61 schwarze 254: if (*cp == '.' || *cp == '\'') {
1.59 schwarze 255: cp++;
256: if (cp[0] == 'D' && cp[1] == 'd') {
257: format = MPARSE_MDOC;
258: break;
259: }
260: if (cp[0] == 'T' && cp[1] == 'H') {
261: format = MPARSE_MAN;
262: break;
263: }
264: }
265: cp = memchr(cp, '\n', ep - cp);
266: if (cp == NULL)
267: break;
268: cp++;
269: }
1.1 schwarze 270: }
271:
1.59 schwarze 272: if (format == MPARSE_MDOC) {
1.25 schwarze 273: if (NULL == curp->pmdoc)
1.22 schwarze 274: curp->pmdoc = mdoc_alloc(
275: curp->roff, curp, curp->defos,
276: MPARSE_QUICK & curp->options ? 1 : 0);
1.1 schwarze 277: assert(curp->pmdoc);
278: curp->mdoc = curp->pmdoc;
279: return;
1.25 schwarze 280: }
1.1 schwarze 281:
1.59 schwarze 282: /* Fall back to man(7) as a last resort. */
283:
1.25 schwarze 284: if (NULL == curp->pman)
1.81 schwarze 285: curp->pman = man_alloc(
286: curp->roff, curp, curp->defos,
1.22 schwarze 287: MPARSE_QUICK & curp->options ? 1 : 0);
1.1 schwarze 288: assert(curp->pman);
289: curp->man = curp->pman;
290: }
291:
292: /*
1.71 schwarze 293: * Main parse routine for a buffer.
294: * It assumes encoding and line numbering are already set up.
295: * It can recurse directly (for invocations of user-defined
296: * macros, inline equations, and input line traps)
297: * and indirectly (for .so file inclusion).
1.1 schwarze 298: */
299: static void
1.71 schwarze 300: mparse_buf_r(struct mparse *curp, struct buf blk, size_t i, int start)
1.1 schwarze 301: {
302: const struct tbl_span *span;
303: struct buf ln;
1.71 schwarze 304: size_t pos; /* byte number in the ln buffer */
1.1 schwarze 305: enum rofferr rr;
1.76 schwarze 306: int of;
1.1 schwarze 307: int lnn; /* line number in the real file */
308: unsigned char c;
309:
1.71 schwarze 310: memset(&ln, 0, sizeof(ln));
1.1 schwarze 311:
1.25 schwarze 312: lnn = curp->line;
313: pos = 0;
1.1 schwarze 314:
1.71 schwarze 315: while (i < blk.sz) {
1.1 schwarze 316: if (0 == pos && '\0' == blk.buf[i])
317: break;
318:
319: if (start) {
320: curp->line = lnn;
321: curp->reparse_count = 0;
1.70 schwarze 322:
323: if (lnn < 3 &&
324: curp->filenc & MPARSE_UTF8 &&
1.71 schwarze 325: curp->filenc & MPARSE_LATIN1)
326: curp->filenc = preconv_cue(&blk, i);
1.1 schwarze 327: }
328:
1.71 schwarze 329: while (i < blk.sz && (start || blk.buf[i] != '\0')) {
1.1 schwarze 330:
331: /*
332: * When finding an unescaped newline character,
333: * leave the character loop to process the line.
334: * Skip a preceding carriage return, if any.
335: */
336:
1.71 schwarze 337: if ('\r' == blk.buf[i] && i + 1 < blk.sz &&
1.1 schwarze 338: '\n' == blk.buf[i + 1])
339: ++i;
340: if ('\n' == blk.buf[i]) {
341: ++i;
342: ++lnn;
343: break;
344: }
345:
1.13 schwarze 346: /*
1.70 schwarze 347: * Make sure we have space for the worst
348: * case of 11 bytes: "\\[u10ffff]\0"
1.13 schwarze 349: */
350:
1.71 schwarze 351: if (pos + 11 > ln.sz)
1.13 schwarze 352: resize_buf(&ln, 256);
353:
1.25 schwarze 354: /*
1.70 schwarze 355: * Encode 8-bit input.
1.1 schwarze 356: */
357:
1.70 schwarze 358: c = blk.buf[i];
359: if (c & 0x80) {
1.71 schwarze 360: if ( ! (curp->filenc && preconv_encode(
361: &blk, &i, &ln, &pos, &curp->filenc))) {
1.70 schwarze 362: mandoc_vmsg(MANDOCERR_BADCHAR,
363: curp, curp->line, pos,
364: "0x%x", c);
365: ln.buf[pos++] = '?';
366: i++;
367: }
368: continue;
369: }
370:
371: /*
372: * Exclude control characters.
373: */
1.1 schwarze 374:
1.70 schwarze 375: if (c == 0x7f || (c < 0x20 && c != 0x09)) {
1.56 schwarze 376: mandoc_vmsg(MANDOCERR_BADCHAR, curp,
377: curp->line, pos, "0x%x", c);
1.1 schwarze 378: i++;
1.6 schwarze 379: ln.buf[pos++] = '?';
1.1 schwarze 380: continue;
381: }
382:
383: /* Trailing backslash = a plain char. */
384:
1.71 schwarze 385: if (blk.buf[i] != '\\' || i + 1 == blk.sz) {
1.1 schwarze 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:
1.71 schwarze 397: if ('\r' == blk.buf[i + 1] && i + 2 < blk.sz &&
1.1 schwarze 398: '\n' == blk.buf[i + 2])
399: ++i;
400: if ('\n' == blk.buf[i + 1]) {
401: i += 2;
402: ++lnn;
403: continue;
404: }
405:
406: if ('"' == blk.buf[i + 1] || '#' == blk.buf[i + 1]) {
407: i += 2;
408: /* Comment, skip to end of line */
1.71 schwarze 409: for (; i < blk.sz; ++i) {
1.1 schwarze 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:
1.13 schwarze 427: /* Catch escaped bogus characters. */
428:
429: c = (unsigned char) blk.buf[i+1];
430:
1.25 schwarze 431: if ( ! (isascii(c) &&
432: (isgraph(c) || isblank(c)))) {
1.56 schwarze 433: mandoc_vmsg(MANDOCERR_BADCHAR, curp,
434: curp->line, pos, "0x%x", c);
1.13 schwarze 435: i += 2;
436: ln.buf[pos++] = '?';
437: continue;
438: }
439:
1.1 schwarze 440: /* Some other escape sequence, copy & cont. */
441:
442: ln.buf[pos++] = blk.buf[i++];
443: ln.buf[pos++] = blk.buf[i++];
444: }
445:
1.71 schwarze 446: if (pos >= ln.sz)
1.1 schwarze 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:
1.4 schwarze 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) {
1.25 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);
1.4 schwarze 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 schwarze 482: rerun:
1.72 schwarze 483: rr = roff_parseln(curp->roff, curp->line, &ln, &of);
1.1 schwarze 484:
485: switch (rr) {
1.25 schwarze 486: case ROFF_REPARSE:
1.1 schwarze 487: if (REPARSE_LIMIT >= ++curp->reparse_count)
1.71 schwarze 488: mparse_buf_r(curp, ln, of, 0);
1.1 schwarze 489: else
490: mandoc_msg(MANDOCERR_ROFFLOOP, curp,
1.25 schwarze 491: curp->line, pos, NULL);
1.1 schwarze 492: pos = 0;
493: continue;
1.25 schwarze 494: case ROFF_APPEND:
1.71 schwarze 495: pos = strlen(ln.buf);
1.1 schwarze 496: continue;
1.25 schwarze 497: case ROFF_RERUN:
1.1 schwarze 498: goto rerun;
1.25 schwarze 499: case ROFF_IGN:
1.1 schwarze 500: pos = 0;
501: continue;
1.25 schwarze 502: case ROFF_ERR:
1.1 schwarze 503: assert(MANDOCLEVEL_FATAL <= curp->file_status);
504: break;
1.25 schwarze 505: case ROFF_SO:
1.71 schwarze 506: if ( ! (curp->options & MPARSE_SO) &&
507: (i >= blk.sz || blk.buf[i] == '\0')) {
1.23 schwarze 508: curp->sodest = mandoc_strdup(ln.buf + of);
509: free(ln.buf);
510: return;
511: }
1.4 schwarze 512: /*
513: * We remove `so' clauses from our lookaside
514: * buffer because we're going to descend into
515: * the file recursively.
516: */
1.25 schwarze 517: if (curp->secondary)
1.4 schwarze 518: curp->secondary->sz -= pos + 1;
1.14 schwarze 519: mparse_readfd(curp, -1, ln.buf + of);
1.30 schwarze 520: if (MANDOCLEVEL_FATAL <= curp->file_status) {
521: mandoc_vmsg(MANDOCERR_SO_FAIL,
522: curp, curp->line, pos,
523: ".so %s", ln.buf + of);
1.1 schwarze 524: break;
1.30 schwarze 525: }
1.1 schwarze 526: pos = 0;
527: continue;
528: default:
529: break;
530: }
531:
532: /*
533: * If we encounter errors in the recursive parse, make
534: * sure we don't continue parsing.
535: */
536:
537: if (MANDOCLEVEL_FATAL <= curp->file_status)
538: break;
539:
540: /*
541: * If input parsers have not been allocated, do so now.
1.2 schwarze 542: * We keep these instanced between parsers, but set them
1.1 schwarze 543: * locally per parse routine since we can use different
544: * parsers with each one.
545: */
546:
547: if ( ! (curp->man || curp->mdoc))
1.60 schwarze 548: choose_parser(curp);
1.1 schwarze 549:
1.25 schwarze 550: /*
1.60 schwarze 551: * Lastly, push down into the parsers themselves.
1.1 schwarze 552: * If libroff returns ROFF_TBL, then add it to the
553: * currently open parse. Since we only get here if
554: * there does exist data (see tbl_data.c), we're
555: * guaranteed that something's been allocated.
556: * Do the same for ROFF_EQN.
557: */
558:
1.76 schwarze 559: if (rr == ROFF_TBL) {
560: while ((span = roff_span(curp->roff)) != NULL)
561: if (curp->man == NULL)
562: mdoc_addspan(curp->mdoc, span);
563: else
564: man_addspan(curp->man, span);
565: } else if (rr == ROFF_EQN) {
566: if (curp->man == NULL)
567: mdoc_addeqn(curp->mdoc, roff_eqn(curp->roff));
568: else
569: man_addeqn(curp->man, roff_eqn(curp->roff));
570: } else if ((curp->man == NULL ?
571: mdoc_parseln(curp->mdoc, curp->line, ln.buf, of) :
572: man_parseln(curp->man, curp->line, ln.buf, of)) == 2)
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)) {
1.82 schwarze 597: perror(file);
598: exit((int)MANDOCLEVEL_SYSERR);
1.1 schwarze 599: }
600:
601: /*
602: * If we're a regular file, try just reading in the whole entry
603: * via mmap(). This is faster than reading it into blocks, and
604: * since each file is only a few bytes to begin with, I'm not
605: * concerned that this is going to tank any machines.
606: */
607:
608: if (S_ISREG(st.st_mode)) {
609: if (st.st_size >= (1U << 31)) {
1.18 schwarze 610: curp->file_status = MANDOCLEVEL_FATAL;
611: if (curp->mmsg)
612: (*curp->mmsg)(MANDOCERR_TOOLARGE,
613: curp->file_status, file, 0, 0, NULL);
1.1 schwarze 614: return(0);
615: }
616: *with_mmap = 1;
617: fb->sz = (size_t)st.st_size;
1.15 schwarze 618: fb->buf = mmap(NULL, fb->sz, PROT_READ, MAP_SHARED, fd, 0);
1.1 schwarze 619: if (fb->buf != MAP_FAILED)
620: return(1);
621: }
622:
623: /*
624: * If this isn't a regular file (like, say, stdin), then we must
625: * go the old way and just read things in bit by bit.
626: */
627:
628: *with_mmap = 0;
629: off = 0;
630: fb->sz = 0;
631: fb->buf = NULL;
632: for (;;) {
633: if (off == fb->sz) {
634: if (fb->sz == (1U << 31)) {
1.18 schwarze 635: curp->file_status = MANDOCLEVEL_FATAL;
636: if (curp->mmsg)
637: (*curp->mmsg)(MANDOCERR_TOOLARGE,
638: curp->file_status,
639: file, 0, 0, NULL);
1.1 schwarze 640: break;
641: }
642: resize_buf(fb, 65536);
643: }
644: ssz = read(fd, fb->buf + (int)off, fb->sz - off);
645: if (ssz == 0) {
646: fb->sz = off;
647: return(1);
648: }
649: if (ssz == -1) {
1.82 schwarze 650: perror(file);
651: exit((int)MANDOCLEVEL_SYSERR);
1.1 schwarze 652: }
653: off += (size_t)ssz;
654: }
655:
656: free(fb->buf);
657: fb->buf = NULL;
658: return(0);
659: }
660:
661: static void
662: mparse_end(struct mparse *curp)
663: {
664:
665: if (MANDOCLEVEL_FATAL <= curp->file_status)
666: return;
667:
1.50 schwarze 668: if (curp->mdoc == NULL &&
669: curp->man == NULL &&
670: curp->sodest == NULL) {
671: if (curp->options & MPARSE_MDOC)
672: curp->mdoc = curp->pmdoc;
673: else {
674: if (curp->pman == NULL)
1.81 schwarze 675: curp->pman = man_alloc(
676: curp->roff, curp, curp->defos,
1.50 schwarze 677: curp->options & MPARSE_QUICK ? 1 : 0);
678: curp->man = curp->pman;
679: }
680: }
681:
1.1 schwarze 682: if (curp->mdoc && ! mdoc_endparse(curp->mdoc)) {
683: assert(MANDOCLEVEL_FATAL <= curp->file_status);
684: return;
685: }
686:
687: if (curp->man && ! man_endparse(curp->man)) {
688: assert(MANDOCLEVEL_FATAL <= curp->file_status);
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: {
1.61 schwarze 698: struct buf *svprimary;
1.1 schwarze 699: const char *svfile;
1.71 schwarze 700: size_t offset;
1.14 schwarze 701: static int recursion_depth;
702:
703: if (64 < recursion_depth) {
704: mandoc_msg(MANDOCERR_ROFFLOOP, curp, curp->line, 0, NULL);
1.15 schwarze 705: return;
1.14 schwarze 706: }
1.1 schwarze 707:
1.15 schwarze 708: /* Line number is per-file. */
709: svfile = curp->file;
710: curp->file = file;
1.61 schwarze 711: svprimary = curp->primary;
1.59 schwarze 712: curp->primary = &blk;
1.15 schwarze 713: curp->line = 1;
714: recursion_depth++;
715:
1.70 schwarze 716: /* Skip an UTF-8 byte order mark. */
717: if (curp->filenc & MPARSE_UTF8 && blk.sz > 2 &&
718: (unsigned char)blk.buf[0] == 0xef &&
719: (unsigned char)blk.buf[1] == 0xbb &&
720: (unsigned char)blk.buf[2] == 0xbf) {
1.71 schwarze 721: offset = 3;
1.70 schwarze 722: curp->filenc &= ~MPARSE_LATIN1;
1.71 schwarze 723: } else
724: offset = 0;
1.70 schwarze 725:
1.71 schwarze 726: mparse_buf_r(curp, blk, offset, 1);
1.15 schwarze 727:
728: if (0 == --recursion_depth && MANDOCLEVEL_FATAL > curp->file_status)
729: mparse_end(curp);
730:
1.61 schwarze 731: curp->primary = svprimary;
1.15 schwarze 732: curp->file = svfile;
733: }
734:
1.74 schwarze 735: /*
736: * If a file descriptor is given, use it and assume it points
737: * to the named file. Otherwise, open the named file.
738: * Read the whole file into memory and call the parsers.
739: * Called recursively when an .so request is encountered.
740: */
1.15 schwarze 741: enum mandoclevel
742: mparse_readfd(struct mparse *curp, int fd, const char *file)
743: {
744: struct buf blk;
745: int with_mmap;
1.70 schwarze 746: int save_filenc;
1.74 schwarze 747: pid_t save_child;
1.15 schwarze 748:
1.74 schwarze 749: save_child = curp->child;
750: if (fd != -1)
751: curp->child = 0;
1.83 ! schwarze 752: else if (mparse_open(curp, &fd, file) != MANDOCLEVEL_OK)
1.74 schwarze 753: goto out;
1.1 schwarze 754:
1.67 schwarze 755: if (read_whole_file(curp, file, fd, &blk, &with_mmap)) {
1.70 schwarze 756: save_filenc = curp->filenc;
757: curp->filenc = curp->options &
758: (MPARSE_UTF8 | MPARSE_LATIN1);
1.67 schwarze 759: mparse_parse_buffer(curp, blk, file);
1.70 schwarze 760: curp->filenc = save_filenc;
1.67 schwarze 761: if (with_mmap)
762: munmap(blk.buf, blk.sz);
763: else
764: free(blk.buf);
765: }
1.1 schwarze 766:
1.74 schwarze 767: if (fd != STDIN_FILENO && close(fd) == -1)
1.1 schwarze 768: perror(file);
1.67 schwarze 769:
1.74 schwarze 770: mparse_wait(curp);
771: out:
772: curp->child = save_child;
1.1 schwarze 773: return(curp->file_status);
1.58 schwarze 774: }
775:
776: enum mandoclevel
1.73 schwarze 777: mparse_open(struct mparse *curp, int *fd, const char *file)
1.58 schwarze 778: {
779: int pfd[2];
1.74 schwarze 780: int save_errno;
1.58 schwarze 781: char *cp;
782:
783: curp->file = file;
1.74 schwarze 784:
785: /* Unless zipped, try to just open the file. */
786:
1.58 schwarze 787: if ((cp = strrchr(file, '.')) == NULL ||
788: strcmp(cp + 1, "gz")) {
1.73 schwarze 789: curp->child = 0;
1.74 schwarze 790: if ((*fd = open(file, O_RDONLY)) != -1)
791: return(MANDOCLEVEL_OK);
792:
793: /* Open failed; try to append ".gz". */
794:
795: mandoc_asprintf(&cp, "%s.gz", file);
796: file = cp;
797: } else
798: cp = NULL;
799:
800: /* Before forking, make sure the file can be read. */
801:
802: save_errno = errno;
803: if (access(file, R_OK) == -1) {
804: if (cp != NULL)
805: errno = save_errno;
1.83 ! schwarze 806: free(cp);
! 807: *fd = -1;
! 808: curp->child = 0;
! 809: mandoc_msg(MANDOCERR_FILE, curp, 0, 0, strerror(errno));
! 810: return(MANDOCLEVEL_ERROR);
1.58 schwarze 811: }
812:
1.74 schwarze 813: /* Run gunzip(1). */
814:
1.58 schwarze 815: if (pipe(pfd) == -1) {
1.82 schwarze 816: perror("pipe");
817: exit((int)MANDOCLEVEL_SYSERR);
1.58 schwarze 818: }
819:
1.73 schwarze 820: switch (curp->child = fork()) {
1.58 schwarze 821: case -1:
1.82 schwarze 822: perror("fork");
823: exit((int)MANDOCLEVEL_SYSERR);
1.58 schwarze 824: case 0:
825: close(pfd[0]);
826: if (dup2(pfd[1], STDOUT_FILENO) == -1) {
1.82 schwarze 827: perror("dup");
828: exit((int)MANDOCLEVEL_SYSERR);
1.58 schwarze 829: }
830: execlp("gunzip", "gunzip", "-c", file, NULL);
1.82 schwarze 831: perror("exec");
832: exit((int)MANDOCLEVEL_SYSERR);
1.58 schwarze 833: default:
834: close(pfd[1]);
835: *fd = pfd[0];
836: return(MANDOCLEVEL_OK);
837: }
838: }
839:
840: enum mandoclevel
1.73 schwarze 841: mparse_wait(struct mparse *curp)
1.58 schwarze 842: {
843: int status;
844:
1.73 schwarze 845: if (curp->child == 0)
846: return(MANDOCLEVEL_OK);
847:
848: if (waitpid(curp->child, &status, 0) == -1) {
1.82 schwarze 849: perror("wait");
850: exit((int)MANDOCLEVEL_SYSERR);
1.58 schwarze 851: }
852: if (WIFSIGNALED(status)) {
1.83 ! schwarze 853: mandoc_vmsg(MANDOCERR_FILE, curp, 0, 0,
! 854: "gunzip died from signal %d", WTERMSIG(status));
! 855: return(MANDOCLEVEL_ERROR);
1.58 schwarze 856: }
857: if (WEXITSTATUS(status)) {
1.83 ! schwarze 858: mandoc_vmsg(MANDOCERR_FILE, curp, 0, 0,
! 859: "gunzip failed with code %d", WEXITSTATUS(status));
! 860: return(MANDOCLEVEL_ERROR);
1.58 schwarze 861: }
862: return(MANDOCLEVEL_OK);
1.1 schwarze 863: }
864:
865: struct mparse *
1.69 schwarze 866: mparse_alloc(int options, enum mandoclevel wlevel, mandocmsg mmsg,
867: const struct mchars *mchars, const char *defos)
1.1 schwarze 868: {
869: struct mparse *curp;
870:
871: assert(wlevel <= MANDOCLEVEL_FATAL);
872:
873: curp = mandoc_calloc(1, sizeof(struct mparse));
874:
1.22 schwarze 875: curp->options = options;
1.1 schwarze 876: curp->wlevel = wlevel;
877: curp->mmsg = mmsg;
1.7 schwarze 878: curp->defos = defos;
1.1 schwarze 879:
1.69 schwarze 880: curp->mchars = mchars;
881: curp->roff = roff_alloc(curp, curp->mchars, options);
1.50 schwarze 882: if (curp->options & MPARSE_MDOC)
883: curp->pmdoc = mdoc_alloc(
884: curp->roff, curp, curp->defos,
885: curp->options & MPARSE_QUICK ? 1 : 0);
886: if (curp->options & MPARSE_MAN)
1.81 schwarze 887: curp->pman = man_alloc(
888: curp->roff, curp, curp->defos,
1.50 schwarze 889: curp->options & MPARSE_QUICK ? 1 : 0);
890:
1.1 schwarze 891: return(curp);
892: }
893:
894: void
895: mparse_reset(struct mparse *curp)
896: {
897:
898: roff_reset(curp->roff);
899:
900: if (curp->mdoc)
901: mdoc_reset(curp->mdoc);
902: if (curp->man)
903: man_reset(curp->man);
1.4 schwarze 904: if (curp->secondary)
905: curp->secondary->sz = 0;
1.1 schwarze 906:
907: curp->file_status = MANDOCLEVEL_OK;
908: curp->mdoc = NULL;
909: curp->man = NULL;
1.23 schwarze 910:
911: free(curp->sodest);
912: curp->sodest = NULL;
1.1 schwarze 913: }
914:
915: void
916: mparse_free(struct mparse *curp)
917: {
918:
919: if (curp->pmdoc)
920: mdoc_free(curp->pmdoc);
921: if (curp->pman)
922: man_free(curp->pman);
923: if (curp->roff)
924: roff_free(curp->roff);
1.4 schwarze 925: if (curp->secondary)
926: free(curp->secondary->buf);
1.1 schwarze 927:
1.4 schwarze 928: free(curp->secondary);
1.23 schwarze 929: free(curp->sodest);
1.1 schwarze 930: free(curp);
931: }
932:
933: void
1.23 schwarze 934: mparse_result(struct mparse *curp,
935: struct mdoc **mdoc, struct man **man, char **sodest)
1.1 schwarze 936: {
937:
1.23 schwarze 938: if (sodest && NULL != (*sodest = curp->sodest)) {
939: *mdoc = NULL;
940: *man = NULL;
941: return;
942: }
1.1 schwarze 943: if (mdoc)
944: *mdoc = curp->mdoc;
945: if (man)
946: *man = curp->man;
947: }
948:
949: void
950: mandoc_vmsg(enum mandocerr t, struct mparse *m,
951: int ln, int pos, const char *fmt, ...)
952: {
953: char buf[256];
954: va_list ap;
955:
956: va_start(ap, fmt);
1.26 schwarze 957: (void)vsnprintf(buf, sizeof(buf), fmt, ap);
1.1 schwarze 958: va_end(ap);
959:
960: mandoc_msg(t, m, ln, pos, buf);
961: }
962:
963: void
1.25 schwarze 964: mandoc_msg(enum mandocerr er, struct mparse *m,
1.1 schwarze 965: int ln, int col, const char *msg)
966: {
967: enum mandoclevel level;
968:
969: level = MANDOCLEVEL_FATAL;
970: while (er < mandoclimits[level])
971: level--;
972:
1.83 ! schwarze 973: if (level < m->wlevel && er != MANDOCERR_FILE)
1.1 schwarze 974: return;
975:
976: if (m->mmsg)
977: (*m->mmsg)(er, level, m->file, ln, col, msg);
978:
979: if (m->file_status < level)
980: m->file_status = level;
981: }
982:
983: const char *
984: mparse_strerror(enum mandocerr er)
985: {
986:
987: return(mandocerrs[er]);
988: }
989:
990: const char *
991: mparse_strlevel(enum mandoclevel lvl)
992: {
993: return(mandoclevels[lvl]);
1.4 schwarze 994: }
995:
996: void
997: mparse_keep(struct mparse *p)
998: {
999:
1000: assert(NULL == p->secondary);
1001: p->secondary = mandoc_calloc(1, sizeof(struct buf));
1002: }
1003:
1004: const char *
1005: mparse_getkeep(const struct mparse *p)
1006: {
1007:
1008: assert(p->secondary);
1009: return(p->secondary->sz ? p->secondary->buf : NULL);
1.1 schwarze 1010: }