Annotation of src/usr.bin/mandoc/read.c, Revision 1.89
1.89 ! schwarze 1: /* $OpenBSD: read.c,v 1.88 2015/01/22 19:26:16 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,
1.87 schwarze 78: MANDOCERR_UNSUPP,
1.1 schwarze 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: "no table layout cells specified",
179: "no table data cells specified",
180: "ignore data in cell",
181: "data block still open",
182: "ignoring extra data cells",
183:
1.46 schwarze 184: /* related to document structure and macros */
1.83 schwarze 185: NULL,
1.1 schwarze 186: "input stack limit exceeded, infinite loop?",
187: "skipping bad character",
1.46 schwarze 188: "skipping unknown macro",
1.87 schwarze 189: "skipping insecure request",
1.51 schwarze 190: "skipping item outside list",
1.46 schwarze 191: "skipping column outside column list",
192: "skipping end of block that is not open",
193: "inserting missing end of block",
194: "appending missing end of block",
195:
196: /* related to request and macro arguments */
1.1 schwarze 197: "escaped character not allowed in a name",
198: "argument count wrong",
1.75 schwarze 199: "NOT IMPLEMENTED: Bd -file",
1.49 schwarze 200: "missing list type, using -item",
1.48 schwarze 201: "missing manual name, using \"\"",
1.49 schwarze 202: "uname(3) system call failed, using UNKNOWN",
1.41 schwarze 203: "unknown standard specifier",
1.49 schwarze 204: "skipping request without numeric argument",
1.84 schwarze 205: "NOT IMPLEMENTED: .so with absolute path or \"..\"",
1.85 schwarze 206: ".so request failed",
1.39 schwarze 207: "skipping all arguments",
208: "skipping excess arguments",
1.68 schwarze 209: "divide by zero",
1.87 schwarze 210:
211: "unsupported feature",
212: "input too large",
1.89 ! schwarze 213: "unsupported control character",
1.87 schwarze 214: "unsupported roff request",
215: "unsupported table syntax",
216: "unsupported table option",
217: "unsupported table layout",
218: "ignoring macro in table",
1.1 schwarze 219: };
220:
221: static const char * const mandoclevels[MANDOCLEVEL_MAX] = {
222: "SUCCESS",
223: "RESERVED",
224: "WARNING",
225: "ERROR",
1.87 schwarze 226: "UNSUPP",
1.1 schwarze 227: "BADARG",
228: "SYSERR"
229: };
230:
1.25 schwarze 231:
1.1 schwarze 232: static void
233: resize_buf(struct buf *buf, size_t initial)
234: {
235:
236: buf->sz = buf->sz > initial/2 ? 2 * buf->sz : initial;
237: buf->buf = mandoc_realloc(buf->buf, buf->sz);
238: }
239:
240: static void
1.60 schwarze 241: choose_parser(struct mparse *curp)
1.1 schwarze 242: {
1.59 schwarze 243: char *cp, *ep;
244: int format;
1.1 schwarze 245:
1.59 schwarze 246: /*
247: * If neither command line arguments -mdoc or -man select
248: * a parser nor the roff parser found a .Dd or .TH macro
249: * yet, look ahead in the main input buffer.
250: */
251:
252: if ((format = roff_getformat(curp->roff)) == 0) {
253: cp = curp->primary->buf;
254: ep = cp + curp->primary->sz;
255: while (cp < ep) {
1.61 schwarze 256: if (*cp == '.' || *cp == '\'') {
1.59 schwarze 257: cp++;
258: if (cp[0] == 'D' && cp[1] == 'd') {
259: format = MPARSE_MDOC;
260: break;
261: }
262: if (cp[0] == 'T' && cp[1] == 'H') {
263: format = MPARSE_MAN;
264: break;
265: }
266: }
267: cp = memchr(cp, '\n', ep - cp);
268: if (cp == NULL)
269: break;
270: cp++;
271: }
1.1 schwarze 272: }
273:
1.59 schwarze 274: if (format == MPARSE_MDOC) {
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:
1.59 schwarze 284: /* Fall back to man(7) as a last resort. */
285:
1.25 schwarze 286: if (NULL == curp->pman)
1.81 schwarze 287: curp->pman = man_alloc(
288: curp->roff, curp, curp->defos,
1.22 schwarze 289: MPARSE_QUICK & curp->options ? 1 : 0);
1.1 schwarze 290: assert(curp->pman);
291: curp->man = curp->pman;
292: }
293:
294: /*
1.71 schwarze 295: * Main parse routine for a buffer.
296: * It assumes encoding and line numbering are already set up.
297: * It can recurse directly (for invocations of user-defined
298: * macros, inline equations, and input line traps)
299: * and indirectly (for .so file inclusion).
1.1 schwarze 300: */
301: static void
1.71 schwarze 302: mparse_buf_r(struct mparse *curp, struct buf blk, size_t i, int start)
1.1 schwarze 303: {
304: const struct tbl_span *span;
305: struct buf ln;
1.88 schwarze 306: const char *save_file;
1.85 schwarze 307: char *cp;
1.71 schwarze 308: size_t pos; /* byte number in the ln buffer */
1.1 schwarze 309: enum rofferr rr;
1.76 schwarze 310: int of;
1.1 schwarze 311: int lnn; /* line number in the real file */
1.85 schwarze 312: int fd;
313: pid_t save_child;
1.1 schwarze 314: unsigned char c;
315:
1.71 schwarze 316: memset(&ln, 0, sizeof(ln));
1.1 schwarze 317:
1.25 schwarze 318: lnn = curp->line;
319: pos = 0;
1.1 schwarze 320:
1.71 schwarze 321: while (i < blk.sz) {
1.1 schwarze 322: if (0 == pos && '\0' == blk.buf[i])
323: break;
324:
325: if (start) {
326: curp->line = lnn;
327: curp->reparse_count = 0;
1.70 schwarze 328:
329: if (lnn < 3 &&
330: curp->filenc & MPARSE_UTF8 &&
1.71 schwarze 331: curp->filenc & MPARSE_LATIN1)
332: curp->filenc = preconv_cue(&blk, i);
1.1 schwarze 333: }
334:
1.71 schwarze 335: while (i < blk.sz && (start || blk.buf[i] != '\0')) {
1.1 schwarze 336:
337: /*
338: * When finding an unescaped newline character,
339: * leave the character loop to process the line.
340: * Skip a preceding carriage return, if any.
341: */
342:
1.71 schwarze 343: if ('\r' == blk.buf[i] && i + 1 < blk.sz &&
1.1 schwarze 344: '\n' == blk.buf[i + 1])
345: ++i;
346: if ('\n' == blk.buf[i]) {
347: ++i;
348: ++lnn;
349: break;
350: }
351:
1.13 schwarze 352: /*
1.70 schwarze 353: * Make sure we have space for the worst
354: * case of 11 bytes: "\\[u10ffff]\0"
1.13 schwarze 355: */
356:
1.71 schwarze 357: if (pos + 11 > ln.sz)
1.13 schwarze 358: resize_buf(&ln, 256);
359:
1.25 schwarze 360: /*
1.70 schwarze 361: * Encode 8-bit input.
1.1 schwarze 362: */
363:
1.70 schwarze 364: c = blk.buf[i];
365: if (c & 0x80) {
1.71 schwarze 366: if ( ! (curp->filenc && preconv_encode(
367: &blk, &i, &ln, &pos, &curp->filenc))) {
1.89 ! schwarze 368: mandoc_vmsg(MANDOCERR_CHAR_BAD, curp,
! 369: curp->line, pos, "0x%x", c);
1.70 schwarze 370: ln.buf[pos++] = '?';
371: i++;
372: }
373: continue;
374: }
375:
376: /*
377: * Exclude control characters.
378: */
1.1 schwarze 379:
1.70 schwarze 380: if (c == 0x7f || (c < 0x20 && c != 0x09)) {
1.89 ! schwarze 381: mandoc_vmsg(c == 0x00 || c == 0x04 ||
! 382: c > 0x0a ? MANDOCERR_CHAR_BAD :
! 383: MANDOCERR_CHAR_UNSUPP,
! 384: curp, curp->line, pos, "0x%x", c);
1.1 schwarze 385: i++;
1.6 schwarze 386: ln.buf[pos++] = '?';
1.1 schwarze 387: continue;
388: }
389:
390: /* Trailing backslash = a plain char. */
391:
1.71 schwarze 392: if (blk.buf[i] != '\\' || i + 1 == blk.sz) {
1.1 schwarze 393: ln.buf[pos++] = blk.buf[i++];
394: continue;
395: }
396:
397: /*
398: * Found escape and at least one other character.
399: * When it's a newline character, skip it.
400: * When there is a carriage return in between,
401: * skip that one as well.
402: */
403:
1.71 schwarze 404: if ('\r' == blk.buf[i + 1] && i + 2 < blk.sz &&
1.1 schwarze 405: '\n' == blk.buf[i + 2])
406: ++i;
407: if ('\n' == blk.buf[i + 1]) {
408: i += 2;
409: ++lnn;
410: continue;
411: }
412:
413: if ('"' == blk.buf[i + 1] || '#' == blk.buf[i + 1]) {
414: i += 2;
415: /* Comment, skip to end of line */
1.71 schwarze 416: for (; i < blk.sz; ++i) {
1.1 schwarze 417: if ('\n' == blk.buf[i]) {
418: ++i;
419: ++lnn;
420: break;
421: }
422: }
423:
424: /* Backout trailing whitespaces */
425: for (; pos > 0; --pos) {
426: if (ln.buf[pos - 1] != ' ')
427: break;
428: if (pos > 2 && ln.buf[pos - 2] == '\\')
429: break;
430: }
431: break;
432: }
433:
1.13 schwarze 434: /* Catch escaped bogus characters. */
435:
436: c = (unsigned char) blk.buf[i+1];
437:
1.25 schwarze 438: if ( ! (isascii(c) &&
439: (isgraph(c) || isblank(c)))) {
1.89 ! schwarze 440: mandoc_vmsg(MANDOCERR_CHAR_BAD, curp,
1.56 schwarze 441: curp->line, pos, "0x%x", c);
1.13 schwarze 442: i += 2;
443: ln.buf[pos++] = '?';
444: continue;
445: }
446:
1.1 schwarze 447: /* Some other escape sequence, copy & cont. */
448:
449: ln.buf[pos++] = blk.buf[i++];
450: ln.buf[pos++] = blk.buf[i++];
451: }
452:
1.71 schwarze 453: if (pos >= ln.sz)
1.1 schwarze 454: resize_buf(&ln, 256);
455:
456: ln.buf[pos] = '\0';
457:
458: /*
459: * A significant amount of complexity is contained by
460: * the roff preprocessor. It's line-oriented but can be
461: * expressed on one line, so we need at times to
462: * readjust our starting point and re-run it. The roff
463: * preprocessor can also readjust the buffers with new
464: * data, so we pass them in wholesale.
465: */
466:
467: of = 0;
468:
1.4 schwarze 469: /*
470: * Maintain a lookaside buffer of all parsed lines. We
471: * only do this if mparse_keep() has been invoked (the
472: * buffer may be accessed with mparse_getkeep()).
473: */
474:
475: if (curp->secondary) {
1.25 schwarze 476: curp->secondary->buf = mandoc_realloc(
477: curp->secondary->buf,
478: curp->secondary->sz + pos + 2);
479: memcpy(curp->secondary->buf +
480: curp->secondary->sz,
481: ln.buf, pos);
1.4 schwarze 482: curp->secondary->sz += pos;
483: curp->secondary->buf
484: [curp->secondary->sz] = '\n';
485: curp->secondary->sz++;
486: curp->secondary->buf
487: [curp->secondary->sz] = '\0';
488: }
1.1 schwarze 489: rerun:
1.72 schwarze 490: rr = roff_parseln(curp->roff, curp->line, &ln, &of);
1.1 schwarze 491:
492: switch (rr) {
1.25 schwarze 493: case ROFF_REPARSE:
1.1 schwarze 494: if (REPARSE_LIMIT >= ++curp->reparse_count)
1.71 schwarze 495: mparse_buf_r(curp, ln, of, 0);
1.1 schwarze 496: else
497: mandoc_msg(MANDOCERR_ROFFLOOP, curp,
1.25 schwarze 498: curp->line, pos, NULL);
1.1 schwarze 499: pos = 0;
500: continue;
1.25 schwarze 501: case ROFF_APPEND:
1.71 schwarze 502: pos = strlen(ln.buf);
1.1 schwarze 503: continue;
1.25 schwarze 504: case ROFF_RERUN:
1.1 schwarze 505: goto rerun;
1.25 schwarze 506: case ROFF_IGN:
1.1 schwarze 507: pos = 0;
508: continue;
1.25 schwarze 509: case ROFF_SO:
1.71 schwarze 510: if ( ! (curp->options & MPARSE_SO) &&
511: (i >= blk.sz || blk.buf[i] == '\0')) {
1.23 schwarze 512: curp->sodest = mandoc_strdup(ln.buf + of);
513: free(ln.buf);
514: return;
515: }
1.4 schwarze 516: /*
517: * We remove `so' clauses from our lookaside
518: * buffer because we're going to descend into
519: * the file recursively.
520: */
1.25 schwarze 521: if (curp->secondary)
1.4 schwarze 522: curp->secondary->sz -= pos + 1;
1.88 schwarze 523: save_file = curp->file;
1.85 schwarze 524: save_child = curp->child;
525: if (mparse_open(curp, &fd, ln.buf + of) ==
1.88 schwarze 526: MANDOCLEVEL_OK) {
1.85 schwarze 527: mparse_readfd(curp, fd, ln.buf + of);
1.88 schwarze 528: curp->file = save_file;
529: } else {
530: curp->file = save_file;
1.30 schwarze 531: mandoc_vmsg(MANDOCERR_SO_FAIL,
532: curp, curp->line, pos,
533: ".so %s", ln.buf + of);
1.85 schwarze 534: ln.sz = mandoc_asprintf(&cp,
535: ".sp\nSee the file %s.\n.sp",
536: ln.buf + of);
537: free(ln.buf);
538: ln.buf = cp;
539: of = 0;
540: mparse_buf_r(curp, ln, of, 0);
1.30 schwarze 541: }
1.85 schwarze 542: curp->child = save_child;
1.1 schwarze 543: pos = 0;
544: continue;
545: default:
546: break;
547: }
548:
549: /*
550: * If input parsers have not been allocated, do so now.
1.2 schwarze 551: * We keep these instanced between parsers, but set them
1.1 schwarze 552: * locally per parse routine since we can use different
553: * parsers with each one.
554: */
555:
556: if ( ! (curp->man || curp->mdoc))
1.60 schwarze 557: choose_parser(curp);
1.1 schwarze 558:
1.25 schwarze 559: /*
1.60 schwarze 560: * Lastly, push down into the parsers themselves.
1.1 schwarze 561: * If libroff returns ROFF_TBL, then add it to the
562: * currently open parse. Since we only get here if
563: * there does exist data (see tbl_data.c), we're
564: * guaranteed that something's been allocated.
565: * Do the same for ROFF_EQN.
566: */
567:
1.76 schwarze 568: if (rr == ROFF_TBL) {
569: while ((span = roff_span(curp->roff)) != NULL)
570: if (curp->man == NULL)
571: mdoc_addspan(curp->mdoc, span);
572: else
573: man_addspan(curp->man, span);
574: } else if (rr == ROFF_EQN) {
575: if (curp->man == NULL)
576: mdoc_addeqn(curp->mdoc, roff_eqn(curp->roff));
577: else
578: man_addeqn(curp->man, roff_eqn(curp->roff));
579: } else if ((curp->man == NULL ?
580: mdoc_parseln(curp->mdoc, curp->line, ln.buf, of) :
581: man_parseln(curp->man, curp->line, ln.buf, of)) == 2)
582: break;
1.1 schwarze 583:
584: /* Temporary buffers typically are not full. */
585:
586: if (0 == start && '\0' == blk.buf[i])
587: break;
588:
589: /* Start the next input line. */
590:
591: pos = 0;
592: }
593:
594: free(ln.buf);
595: }
596:
597: static int
1.18 schwarze 598: read_whole_file(struct mparse *curp, const char *file, int fd,
599: struct buf *fb, int *with_mmap)
1.1 schwarze 600: {
601: struct stat st;
602: size_t off;
603: ssize_t ssz;
604:
605: if (-1 == fstat(fd, &st)) {
1.82 schwarze 606: perror(file);
607: exit((int)MANDOCLEVEL_SYSERR);
1.1 schwarze 608: }
609:
610: /*
611: * If we're a regular file, try just reading in the whole entry
612: * via mmap(). This is faster than reading it into blocks, and
613: * since each file is only a few bytes to begin with, I'm not
614: * concerned that this is going to tank any machines.
615: */
616:
617: if (S_ISREG(st.st_mode)) {
618: if (st.st_size >= (1U << 31)) {
1.86 schwarze 619: mandoc_msg(MANDOCERR_TOOLARGE, curp, 0, 0, NULL);
1.1 schwarze 620: return(0);
621: }
622: *with_mmap = 1;
623: fb->sz = (size_t)st.st_size;
1.15 schwarze 624: fb->buf = mmap(NULL, fb->sz, PROT_READ, MAP_SHARED, fd, 0);
1.1 schwarze 625: if (fb->buf != MAP_FAILED)
626: return(1);
627: }
628:
629: /*
630: * If this isn't a regular file (like, say, stdin), then we must
631: * go the old way and just read things in bit by bit.
632: */
633:
634: *with_mmap = 0;
635: off = 0;
636: fb->sz = 0;
637: fb->buf = NULL;
638: for (;;) {
639: if (off == fb->sz) {
640: if (fb->sz == (1U << 31)) {
1.86 schwarze 641: mandoc_msg(MANDOCERR_TOOLARGE, curp,
642: 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) {
1.82 schwarze 653: perror(file);
654: exit((int)MANDOCLEVEL_SYSERR);
1.1 schwarze 655: }
656: off += (size_t)ssz;
657: }
658:
659: free(fb->buf);
660: fb->buf = NULL;
661: return(0);
662: }
663:
664: static void
665: mparse_end(struct mparse *curp)
666: {
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: }
1.86 schwarze 681: if (curp->mdoc)
682: mdoc_endparse(curp->mdoc);
683: if (curp->man)
684: man_endparse(curp->man);
1.1 schwarze 685: roff_endparse(curp->roff);
686: }
687:
1.15 schwarze 688: static void
689: mparse_parse_buffer(struct mparse *curp, struct buf blk, const char *file)
1.1 schwarze 690: {
1.61 schwarze 691: struct buf *svprimary;
1.1 schwarze 692: const char *svfile;
1.71 schwarze 693: size_t offset;
1.14 schwarze 694: static int recursion_depth;
695:
696: if (64 < recursion_depth) {
697: mandoc_msg(MANDOCERR_ROFFLOOP, curp, curp->line, 0, NULL);
1.15 schwarze 698: return;
1.14 schwarze 699: }
1.1 schwarze 700:
1.15 schwarze 701: /* Line number is per-file. */
702: svfile = curp->file;
703: curp->file = file;
1.61 schwarze 704: svprimary = curp->primary;
1.59 schwarze 705: curp->primary = &blk;
1.15 schwarze 706: curp->line = 1;
707: recursion_depth++;
708:
1.70 schwarze 709: /* Skip an UTF-8 byte order mark. */
710: if (curp->filenc & MPARSE_UTF8 && blk.sz > 2 &&
711: (unsigned char)blk.buf[0] == 0xef &&
712: (unsigned char)blk.buf[1] == 0xbb &&
713: (unsigned char)blk.buf[2] == 0xbf) {
1.71 schwarze 714: offset = 3;
1.70 schwarze 715: curp->filenc &= ~MPARSE_LATIN1;
1.71 schwarze 716: } else
717: offset = 0;
1.70 schwarze 718:
1.71 schwarze 719: mparse_buf_r(curp, blk, offset, 1);
1.15 schwarze 720:
1.86 schwarze 721: if (--recursion_depth == 0)
1.15 schwarze 722: mparse_end(curp);
723:
1.61 schwarze 724: curp->primary = svprimary;
1.15 schwarze 725: curp->file = svfile;
726: }
727:
1.74 schwarze 728: /*
729: * Read the whole file into memory and call the parsers.
730: * Called recursively when an .so request is encountered.
731: */
1.15 schwarze 732: enum mandoclevel
733: mparse_readfd(struct mparse *curp, int fd, const char *file)
734: {
735: struct buf blk;
736: int with_mmap;
1.70 schwarze 737: int save_filenc;
1.1 schwarze 738:
1.67 schwarze 739: if (read_whole_file(curp, file, fd, &blk, &with_mmap)) {
1.70 schwarze 740: save_filenc = curp->filenc;
741: curp->filenc = curp->options &
742: (MPARSE_UTF8 | MPARSE_LATIN1);
1.67 schwarze 743: mparse_parse_buffer(curp, blk, file);
1.70 schwarze 744: curp->filenc = save_filenc;
1.67 schwarze 745: if (with_mmap)
746: munmap(blk.buf, blk.sz);
747: else
748: free(blk.buf);
749: }
1.1 schwarze 750:
1.74 schwarze 751: if (fd != STDIN_FILENO && close(fd) == -1)
1.1 schwarze 752: perror(file);
1.67 schwarze 753:
1.74 schwarze 754: mparse_wait(curp);
1.1 schwarze 755: return(curp->file_status);
1.58 schwarze 756: }
757:
758: enum mandoclevel
1.73 schwarze 759: mparse_open(struct mparse *curp, int *fd, const char *file)
1.58 schwarze 760: {
761: int pfd[2];
1.74 schwarze 762: int save_errno;
1.58 schwarze 763: char *cp;
764:
765: curp->file = file;
1.74 schwarze 766:
767: /* Unless zipped, try to just open the file. */
768:
1.58 schwarze 769: if ((cp = strrchr(file, '.')) == NULL ||
770: strcmp(cp + 1, "gz")) {
1.73 schwarze 771: curp->child = 0;
1.74 schwarze 772: if ((*fd = open(file, O_RDONLY)) != -1)
773: return(MANDOCLEVEL_OK);
774:
775: /* Open failed; try to append ".gz". */
776:
777: mandoc_asprintf(&cp, "%s.gz", file);
778: file = cp;
779: } else
780: cp = NULL;
781:
782: /* Before forking, make sure the file can be read. */
783:
784: save_errno = errno;
785: if (access(file, R_OK) == -1) {
786: if (cp != NULL)
787: errno = save_errno;
1.83 schwarze 788: free(cp);
789: *fd = -1;
790: curp->child = 0;
791: mandoc_msg(MANDOCERR_FILE, curp, 0, 0, strerror(errno));
792: return(MANDOCLEVEL_ERROR);
1.58 schwarze 793: }
794:
1.74 schwarze 795: /* Run gunzip(1). */
796:
1.58 schwarze 797: if (pipe(pfd) == -1) {
1.82 schwarze 798: perror("pipe");
799: exit((int)MANDOCLEVEL_SYSERR);
1.58 schwarze 800: }
801:
1.73 schwarze 802: switch (curp->child = fork()) {
1.58 schwarze 803: case -1:
1.82 schwarze 804: perror("fork");
805: exit((int)MANDOCLEVEL_SYSERR);
1.58 schwarze 806: case 0:
807: close(pfd[0]);
808: if (dup2(pfd[1], STDOUT_FILENO) == -1) {
1.82 schwarze 809: perror("dup");
810: exit((int)MANDOCLEVEL_SYSERR);
1.58 schwarze 811: }
812: execlp("gunzip", "gunzip", "-c", file, NULL);
1.82 schwarze 813: perror("exec");
814: exit((int)MANDOCLEVEL_SYSERR);
1.58 schwarze 815: default:
816: close(pfd[1]);
817: *fd = pfd[0];
818: return(MANDOCLEVEL_OK);
819: }
820: }
821:
822: enum mandoclevel
1.73 schwarze 823: mparse_wait(struct mparse *curp)
1.58 schwarze 824: {
825: int status;
826:
1.73 schwarze 827: if (curp->child == 0)
828: return(MANDOCLEVEL_OK);
829:
830: if (waitpid(curp->child, &status, 0) == -1) {
1.82 schwarze 831: perror("wait");
832: exit((int)MANDOCLEVEL_SYSERR);
1.58 schwarze 833: }
834: if (WIFSIGNALED(status)) {
1.83 schwarze 835: mandoc_vmsg(MANDOCERR_FILE, curp, 0, 0,
836: "gunzip died from signal %d", WTERMSIG(status));
837: return(MANDOCLEVEL_ERROR);
1.58 schwarze 838: }
839: if (WEXITSTATUS(status)) {
1.83 schwarze 840: mandoc_vmsg(MANDOCERR_FILE, curp, 0, 0,
841: "gunzip failed with code %d", WEXITSTATUS(status));
842: return(MANDOCLEVEL_ERROR);
1.58 schwarze 843: }
844: return(MANDOCLEVEL_OK);
1.1 schwarze 845: }
846:
847: struct mparse *
1.69 schwarze 848: mparse_alloc(int options, enum mandoclevel wlevel, mandocmsg mmsg,
849: const struct mchars *mchars, const char *defos)
1.1 schwarze 850: {
851: struct mparse *curp;
852:
853: curp = mandoc_calloc(1, sizeof(struct mparse));
854:
1.22 schwarze 855: curp->options = options;
1.1 schwarze 856: curp->wlevel = wlevel;
857: curp->mmsg = mmsg;
1.7 schwarze 858: curp->defos = defos;
1.1 schwarze 859:
1.69 schwarze 860: curp->mchars = mchars;
861: curp->roff = roff_alloc(curp, curp->mchars, options);
1.50 schwarze 862: if (curp->options & MPARSE_MDOC)
863: curp->pmdoc = mdoc_alloc(
864: curp->roff, curp, curp->defos,
865: curp->options & MPARSE_QUICK ? 1 : 0);
866: if (curp->options & MPARSE_MAN)
1.81 schwarze 867: curp->pman = man_alloc(
868: curp->roff, curp, curp->defos,
1.50 schwarze 869: curp->options & MPARSE_QUICK ? 1 : 0);
870:
1.1 schwarze 871: return(curp);
872: }
873:
874: void
875: mparse_reset(struct mparse *curp)
876: {
877:
878: roff_reset(curp->roff);
879:
880: if (curp->mdoc)
881: mdoc_reset(curp->mdoc);
882: if (curp->man)
883: man_reset(curp->man);
1.4 schwarze 884: if (curp->secondary)
885: curp->secondary->sz = 0;
1.1 schwarze 886:
887: curp->file_status = MANDOCLEVEL_OK;
888: curp->mdoc = NULL;
889: curp->man = NULL;
1.23 schwarze 890:
891: free(curp->sodest);
892: curp->sodest = NULL;
1.1 schwarze 893: }
894:
895: void
896: mparse_free(struct mparse *curp)
897: {
898:
899: if (curp->pmdoc)
900: mdoc_free(curp->pmdoc);
901: if (curp->pman)
902: man_free(curp->pman);
903: if (curp->roff)
904: roff_free(curp->roff);
1.4 schwarze 905: if (curp->secondary)
906: free(curp->secondary->buf);
1.1 schwarze 907:
1.4 schwarze 908: free(curp->secondary);
1.23 schwarze 909: free(curp->sodest);
1.1 schwarze 910: free(curp);
911: }
912:
913: void
1.23 schwarze 914: mparse_result(struct mparse *curp,
915: struct mdoc **mdoc, struct man **man, char **sodest)
1.1 schwarze 916: {
917:
1.23 schwarze 918: if (sodest && NULL != (*sodest = curp->sodest)) {
919: *mdoc = NULL;
920: *man = NULL;
921: return;
922: }
1.1 schwarze 923: if (mdoc)
924: *mdoc = curp->mdoc;
925: if (man)
926: *man = curp->man;
927: }
928:
929: void
930: mandoc_vmsg(enum mandocerr t, struct mparse *m,
931: int ln, int pos, const char *fmt, ...)
932: {
933: char buf[256];
934: va_list ap;
935:
936: va_start(ap, fmt);
1.26 schwarze 937: (void)vsnprintf(buf, sizeof(buf), fmt, ap);
1.1 schwarze 938: va_end(ap);
939:
940: mandoc_msg(t, m, ln, pos, buf);
941: }
942:
943: void
1.25 schwarze 944: mandoc_msg(enum mandocerr er, struct mparse *m,
1.1 schwarze 945: int ln, int col, const char *msg)
946: {
947: enum mandoclevel level;
948:
1.87 schwarze 949: level = MANDOCLEVEL_UNSUPP;
1.1 schwarze 950: while (er < mandoclimits[level])
951: level--;
952:
1.83 schwarze 953: if (level < m->wlevel && er != MANDOCERR_FILE)
1.1 schwarze 954: return;
955:
956: if (m->mmsg)
957: (*m->mmsg)(er, level, m->file, ln, col, msg);
958:
959: if (m->file_status < level)
960: m->file_status = level;
961: }
962:
963: const char *
964: mparse_strerror(enum mandocerr er)
965: {
966:
967: return(mandocerrs[er]);
968: }
969:
970: const char *
971: mparse_strlevel(enum mandoclevel lvl)
972: {
973: return(mandoclevels[lvl]);
1.4 schwarze 974: }
975:
976: void
977: mparse_keep(struct mparse *p)
978: {
979:
980: assert(NULL == p->secondary);
981: p->secondary = mandoc_calloc(1, sizeof(struct buf));
982: }
983:
984: const char *
985: mparse_getkeep(const struct mparse *p)
986: {
987:
988: assert(p->secondary);
989: return(p->secondary->sz ? p->secondary->buf : NULL);
1.1 schwarze 990: }
![[BACK]](/icons/back.gif){kind=icon}
Return to read.c CVS log ![[TXT]](/icons/text.gif){kind=icon}
![[DIR]](/icons/dir.gif){kind=icon}
Up to [local] / src / usr.bin / mandoc