Annotation of src/usr.bin/mandoc/read.c, Revision 1.91
1.91 ! schwarze 1: /* $OpenBSD: read.c,v 1.90 2015/01/24 01:59:40 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 */
1.91 ! schwarze 178: "non-alphabetic character in tbl options",
! 179: "skipping unknown tbl option",
! 180: "missing tbl option argument",
! 181: "wrong tbl option argument size",
1.1 schwarze 182: "no table layout cells specified",
183: "no table data cells specified",
184: "ignore data in cell",
185: "data block still open",
186: "ignoring extra data cells",
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",
1.87 schwarze 193: "skipping insecure request",
1.51 schwarze 194: "skipping item outside list",
1.46 schwarze 195: "skipping column outside column list",
196: "skipping end of block that is not open",
1.90 schwarze 197: "fewer RS blocks open, skipping",
1.46 schwarze 198: "inserting missing end of block",
199: "appending missing end of block",
200:
201: /* related to request and macro arguments */
1.1 schwarze 202: "escaped character not allowed in a name",
203: "argument count wrong",
1.75 schwarze 204: "NOT IMPLEMENTED: Bd -file",
1.49 schwarze 205: "missing list type, using -item",
1.48 schwarze 206: "missing manual name, using \"\"",
1.49 schwarze 207: "uname(3) system call failed, using UNKNOWN",
1.41 schwarze 208: "unknown standard specifier",
1.49 schwarze 209: "skipping request without numeric argument",
1.84 schwarze 210: "NOT IMPLEMENTED: .so with absolute path or \"..\"",
1.85 schwarze 211: ".so request failed",
1.39 schwarze 212: "skipping all arguments",
213: "skipping excess arguments",
1.68 schwarze 214: "divide by zero",
1.87 schwarze 215:
216: "unsupported feature",
217: "input too large",
1.89 schwarze 218: "unsupported control character",
1.87 schwarze 219: "unsupported roff request",
220: "unsupported table layout",
221: "ignoring macro in table",
1.1 schwarze 222: };
223:
224: static const char * const mandoclevels[MANDOCLEVEL_MAX] = {
225: "SUCCESS",
226: "RESERVED",
227: "WARNING",
228: "ERROR",
1.87 schwarze 229: "UNSUPP",
1.1 schwarze 230: "BADARG",
231: "SYSERR"
232: };
233:
1.25 schwarze 234:
1.1 schwarze 235: static void
236: resize_buf(struct buf *buf, size_t initial)
237: {
238:
239: buf->sz = buf->sz > initial/2 ? 2 * buf->sz : initial;
240: buf->buf = mandoc_realloc(buf->buf, buf->sz);
241: }
242:
243: static void
1.60 schwarze 244: choose_parser(struct mparse *curp)
1.1 schwarze 245: {
1.59 schwarze 246: char *cp, *ep;
247: int format;
1.1 schwarze 248:
1.59 schwarze 249: /*
250: * If neither command line arguments -mdoc or -man select
251: * a parser nor the roff parser found a .Dd or .TH macro
252: * yet, look ahead in the main input buffer.
253: */
254:
255: if ((format = roff_getformat(curp->roff)) == 0) {
256: cp = curp->primary->buf;
257: ep = cp + curp->primary->sz;
258: while (cp < ep) {
1.61 schwarze 259: if (*cp == '.' || *cp == '\'') {
1.59 schwarze 260: cp++;
261: if (cp[0] == 'D' && cp[1] == 'd') {
262: format = MPARSE_MDOC;
263: break;
264: }
265: if (cp[0] == 'T' && cp[1] == 'H') {
266: format = MPARSE_MAN;
267: break;
268: }
269: }
270: cp = memchr(cp, '\n', ep - cp);
271: if (cp == NULL)
272: break;
273: cp++;
274: }
1.1 schwarze 275: }
276:
1.59 schwarze 277: if (format == MPARSE_MDOC) {
1.25 schwarze 278: if (NULL == curp->pmdoc)
1.22 schwarze 279: curp->pmdoc = mdoc_alloc(
280: curp->roff, curp, curp->defos,
281: MPARSE_QUICK & curp->options ? 1 : 0);
1.1 schwarze 282: assert(curp->pmdoc);
283: curp->mdoc = curp->pmdoc;
284: return;
1.25 schwarze 285: }
1.1 schwarze 286:
1.59 schwarze 287: /* Fall back to man(7) as a last resort. */
288:
1.25 schwarze 289: if (NULL == curp->pman)
1.81 schwarze 290: curp->pman = man_alloc(
291: curp->roff, curp, curp->defos,
1.22 schwarze 292: MPARSE_QUICK & curp->options ? 1 : 0);
1.1 schwarze 293: assert(curp->pman);
294: curp->man = curp->pman;
295: }
296:
297: /*
1.71 schwarze 298: * Main parse routine for a buffer.
299: * It assumes encoding and line numbering are already set up.
300: * It can recurse directly (for invocations of user-defined
301: * macros, inline equations, and input line traps)
302: * and indirectly (for .so file inclusion).
1.1 schwarze 303: */
304: static void
1.71 schwarze 305: mparse_buf_r(struct mparse *curp, struct buf blk, size_t i, int start)
1.1 schwarze 306: {
307: const struct tbl_span *span;
308: struct buf ln;
1.88 schwarze 309: const char *save_file;
1.85 schwarze 310: char *cp;
1.71 schwarze 311: size_t pos; /* byte number in the ln buffer */
1.1 schwarze 312: enum rofferr rr;
1.76 schwarze 313: int of;
1.1 schwarze 314: int lnn; /* line number in the real file */
1.85 schwarze 315: int fd;
316: pid_t save_child;
1.1 schwarze 317: unsigned char c;
318:
1.71 schwarze 319: memset(&ln, 0, sizeof(ln));
1.1 schwarze 320:
1.25 schwarze 321: lnn = curp->line;
322: pos = 0;
1.1 schwarze 323:
1.71 schwarze 324: while (i < blk.sz) {
1.1 schwarze 325: if (0 == pos && '\0' == blk.buf[i])
326: break;
327:
328: if (start) {
329: curp->line = lnn;
330: curp->reparse_count = 0;
1.70 schwarze 331:
332: if (lnn < 3 &&
333: curp->filenc & MPARSE_UTF8 &&
1.71 schwarze 334: curp->filenc & MPARSE_LATIN1)
335: curp->filenc = preconv_cue(&blk, i);
1.1 schwarze 336: }
337:
1.71 schwarze 338: while (i < blk.sz && (start || blk.buf[i] != '\0')) {
1.1 schwarze 339:
340: /*
341: * When finding an unescaped newline character,
342: * leave the character loop to process the line.
343: * Skip a preceding carriage return, if any.
344: */
345:
1.71 schwarze 346: if ('\r' == blk.buf[i] && i + 1 < blk.sz &&
1.1 schwarze 347: '\n' == blk.buf[i + 1])
348: ++i;
349: if ('\n' == blk.buf[i]) {
350: ++i;
351: ++lnn;
352: break;
353: }
354:
1.13 schwarze 355: /*
1.70 schwarze 356: * Make sure we have space for the worst
357: * case of 11 bytes: "\\[u10ffff]\0"
1.13 schwarze 358: */
359:
1.71 schwarze 360: if (pos + 11 > ln.sz)
1.13 schwarze 361: resize_buf(&ln, 256);
362:
1.25 schwarze 363: /*
1.70 schwarze 364: * Encode 8-bit input.
1.1 schwarze 365: */
366:
1.70 schwarze 367: c = blk.buf[i];
368: if (c & 0x80) {
1.71 schwarze 369: if ( ! (curp->filenc && preconv_encode(
370: &blk, &i, &ln, &pos, &curp->filenc))) {
1.89 schwarze 371: mandoc_vmsg(MANDOCERR_CHAR_BAD, curp,
372: curp->line, pos, "0x%x", c);
1.70 schwarze 373: ln.buf[pos++] = '?';
374: i++;
375: }
376: continue;
377: }
378:
379: /*
380: * Exclude control characters.
381: */
1.1 schwarze 382:
1.70 schwarze 383: if (c == 0x7f || (c < 0x20 && c != 0x09)) {
1.89 schwarze 384: mandoc_vmsg(c == 0x00 || c == 0x04 ||
385: c > 0x0a ? MANDOCERR_CHAR_BAD :
386: MANDOCERR_CHAR_UNSUPP,
387: curp, curp->line, pos, "0x%x", c);
1.1 schwarze 388: i++;
1.6 schwarze 389: ln.buf[pos++] = '?';
1.1 schwarze 390: continue;
391: }
392:
393: /* Trailing backslash = a plain char. */
394:
1.71 schwarze 395: if (blk.buf[i] != '\\' || i + 1 == blk.sz) {
1.1 schwarze 396: ln.buf[pos++] = blk.buf[i++];
397: continue;
398: }
399:
400: /*
401: * Found escape and at least one other character.
402: * When it's a newline character, skip it.
403: * When there is a carriage return in between,
404: * skip that one as well.
405: */
406:
1.71 schwarze 407: if ('\r' == blk.buf[i + 1] && i + 2 < blk.sz &&
1.1 schwarze 408: '\n' == blk.buf[i + 2])
409: ++i;
410: if ('\n' == blk.buf[i + 1]) {
411: i += 2;
412: ++lnn;
413: continue;
414: }
415:
416: if ('"' == blk.buf[i + 1] || '#' == blk.buf[i + 1]) {
417: i += 2;
418: /* Comment, skip to end of line */
1.71 schwarze 419: for (; i < blk.sz; ++i) {
1.1 schwarze 420: if ('\n' == blk.buf[i]) {
421: ++i;
422: ++lnn;
423: break;
424: }
425: }
426:
427: /* Backout trailing whitespaces */
428: for (; pos > 0; --pos) {
429: if (ln.buf[pos - 1] != ' ')
430: break;
431: if (pos > 2 && ln.buf[pos - 2] == '\\')
432: break;
433: }
434: break;
435: }
436:
1.13 schwarze 437: /* Catch escaped bogus characters. */
438:
439: c = (unsigned char) blk.buf[i+1];
440:
1.25 schwarze 441: if ( ! (isascii(c) &&
442: (isgraph(c) || isblank(c)))) {
1.89 schwarze 443: mandoc_vmsg(MANDOCERR_CHAR_BAD, curp,
1.56 schwarze 444: curp->line, pos, "0x%x", c);
1.13 schwarze 445: i += 2;
446: ln.buf[pos++] = '?';
447: continue;
448: }
449:
1.1 schwarze 450: /* Some other escape sequence, copy & cont. */
451:
452: ln.buf[pos++] = blk.buf[i++];
453: ln.buf[pos++] = blk.buf[i++];
454: }
455:
1.71 schwarze 456: if (pos >= ln.sz)
1.1 schwarze 457: resize_buf(&ln, 256);
458:
459: ln.buf[pos] = '\0';
460:
461: /*
462: * A significant amount of complexity is contained by
463: * the roff preprocessor. It's line-oriented but can be
464: * expressed on one line, so we need at times to
465: * readjust our starting point and re-run it. The roff
466: * preprocessor can also readjust the buffers with new
467: * data, so we pass them in wholesale.
468: */
469:
470: of = 0;
471:
1.4 schwarze 472: /*
473: * Maintain a lookaside buffer of all parsed lines. We
474: * only do this if mparse_keep() has been invoked (the
475: * buffer may be accessed with mparse_getkeep()).
476: */
477:
478: if (curp->secondary) {
1.25 schwarze 479: curp->secondary->buf = mandoc_realloc(
480: curp->secondary->buf,
481: curp->secondary->sz + pos + 2);
482: memcpy(curp->secondary->buf +
483: curp->secondary->sz,
484: ln.buf, pos);
1.4 schwarze 485: curp->secondary->sz += pos;
486: curp->secondary->buf
487: [curp->secondary->sz] = '\n';
488: curp->secondary->sz++;
489: curp->secondary->buf
490: [curp->secondary->sz] = '\0';
491: }
1.1 schwarze 492: rerun:
1.72 schwarze 493: rr = roff_parseln(curp->roff, curp->line, &ln, &of);
1.1 schwarze 494:
495: switch (rr) {
1.25 schwarze 496: case ROFF_REPARSE:
1.1 schwarze 497: if (REPARSE_LIMIT >= ++curp->reparse_count)
1.71 schwarze 498: mparse_buf_r(curp, ln, of, 0);
1.1 schwarze 499: else
500: mandoc_msg(MANDOCERR_ROFFLOOP, curp,
1.25 schwarze 501: curp->line, pos, NULL);
1.1 schwarze 502: pos = 0;
503: continue;
1.25 schwarze 504: case ROFF_APPEND:
1.71 schwarze 505: pos = strlen(ln.buf);
1.1 schwarze 506: continue;
1.25 schwarze 507: case ROFF_RERUN:
1.1 schwarze 508: goto rerun;
1.25 schwarze 509: case ROFF_IGN:
1.1 schwarze 510: pos = 0;
511: continue;
1.25 schwarze 512: case ROFF_SO:
1.71 schwarze 513: if ( ! (curp->options & MPARSE_SO) &&
514: (i >= blk.sz || blk.buf[i] == '\0')) {
1.23 schwarze 515: curp->sodest = mandoc_strdup(ln.buf + of);
516: free(ln.buf);
517: return;
518: }
1.4 schwarze 519: /*
520: * We remove `so' clauses from our lookaside
521: * buffer because we're going to descend into
522: * the file recursively.
523: */
1.25 schwarze 524: if (curp->secondary)
1.4 schwarze 525: curp->secondary->sz -= pos + 1;
1.88 schwarze 526: save_file = curp->file;
1.85 schwarze 527: save_child = curp->child;
528: if (mparse_open(curp, &fd, ln.buf + of) ==
1.88 schwarze 529: MANDOCLEVEL_OK) {
1.85 schwarze 530: mparse_readfd(curp, fd, ln.buf + of);
1.88 schwarze 531: curp->file = save_file;
532: } else {
533: curp->file = save_file;
1.30 schwarze 534: mandoc_vmsg(MANDOCERR_SO_FAIL,
535: curp, curp->line, pos,
536: ".so %s", ln.buf + of);
1.85 schwarze 537: ln.sz = mandoc_asprintf(&cp,
538: ".sp\nSee the file %s.\n.sp",
539: ln.buf + of);
540: free(ln.buf);
541: ln.buf = cp;
542: of = 0;
543: mparse_buf_r(curp, ln, of, 0);
1.30 schwarze 544: }
1.85 schwarze 545: curp->child = save_child;
1.1 schwarze 546: pos = 0;
547: continue;
548: default:
549: break;
550: }
551:
552: /*
553: * If input parsers have not been allocated, do so now.
1.2 schwarze 554: * We keep these instanced between parsers, but set them
1.1 schwarze 555: * locally per parse routine since we can use different
556: * parsers with each one.
557: */
558:
559: if ( ! (curp->man || curp->mdoc))
1.60 schwarze 560: choose_parser(curp);
1.1 schwarze 561:
1.25 schwarze 562: /*
1.60 schwarze 563: * Lastly, push down into the parsers themselves.
1.1 schwarze 564: * If libroff returns ROFF_TBL, then add it to the
565: * currently open parse. Since we only get here if
566: * there does exist data (see tbl_data.c), we're
567: * guaranteed that something's been allocated.
568: * Do the same for ROFF_EQN.
569: */
570:
1.76 schwarze 571: if (rr == ROFF_TBL) {
572: while ((span = roff_span(curp->roff)) != NULL)
573: if (curp->man == NULL)
574: mdoc_addspan(curp->mdoc, span);
575: else
576: man_addspan(curp->man, span);
577: } else if (rr == ROFF_EQN) {
578: if (curp->man == NULL)
579: mdoc_addeqn(curp->mdoc, roff_eqn(curp->roff));
580: else
581: man_addeqn(curp->man, roff_eqn(curp->roff));
582: } else if ((curp->man == NULL ?
583: mdoc_parseln(curp->mdoc, curp->line, ln.buf, of) :
584: man_parseln(curp->man, curp->line, ln.buf, of)) == 2)
585: break;
1.1 schwarze 586:
587: /* Temporary buffers typically are not full. */
588:
589: if (0 == start && '\0' == blk.buf[i])
590: break;
591:
592: /* Start the next input line. */
593:
594: pos = 0;
595: }
596:
597: free(ln.buf);
598: }
599:
600: static int
1.18 schwarze 601: read_whole_file(struct mparse *curp, const char *file, int fd,
602: struct buf *fb, int *with_mmap)
1.1 schwarze 603: {
604: struct stat st;
605: size_t off;
606: ssize_t ssz;
607:
608: if (-1 == fstat(fd, &st)) {
1.82 schwarze 609: perror(file);
610: exit((int)MANDOCLEVEL_SYSERR);
1.1 schwarze 611: }
612:
613: /*
614: * If we're a regular file, try just reading in the whole entry
615: * via mmap(). This is faster than reading it into blocks, and
616: * since each file is only a few bytes to begin with, I'm not
617: * concerned that this is going to tank any machines.
618: */
619:
620: if (S_ISREG(st.st_mode)) {
621: if (st.st_size >= (1U << 31)) {
1.86 schwarze 622: mandoc_msg(MANDOCERR_TOOLARGE, curp, 0, 0, NULL);
1.1 schwarze 623: return(0);
624: }
625: *with_mmap = 1;
626: fb->sz = (size_t)st.st_size;
1.15 schwarze 627: fb->buf = mmap(NULL, fb->sz, PROT_READ, MAP_SHARED, fd, 0);
1.1 schwarze 628: if (fb->buf != MAP_FAILED)
629: return(1);
630: }
631:
632: /*
633: * If this isn't a regular file (like, say, stdin), then we must
634: * go the old way and just read things in bit by bit.
635: */
636:
637: *with_mmap = 0;
638: off = 0;
639: fb->sz = 0;
640: fb->buf = NULL;
641: for (;;) {
642: if (off == fb->sz) {
643: if (fb->sz == (1U << 31)) {
1.86 schwarze 644: mandoc_msg(MANDOCERR_TOOLARGE, curp,
645: 0, 0, NULL);
1.1 schwarze 646: break;
647: }
648: resize_buf(fb, 65536);
649: }
650: ssz = read(fd, fb->buf + (int)off, fb->sz - off);
651: if (ssz == 0) {
652: fb->sz = off;
653: return(1);
654: }
655: if (ssz == -1) {
1.82 schwarze 656: perror(file);
657: exit((int)MANDOCLEVEL_SYSERR);
1.1 schwarze 658: }
659: off += (size_t)ssz;
660: }
661:
662: free(fb->buf);
663: fb->buf = NULL;
664: return(0);
665: }
666:
667: static void
668: mparse_end(struct mparse *curp)
669: {
670:
1.50 schwarze 671: if (curp->mdoc == NULL &&
672: curp->man == NULL &&
673: curp->sodest == NULL) {
674: if (curp->options & MPARSE_MDOC)
675: curp->mdoc = curp->pmdoc;
676: else {
677: if (curp->pman == NULL)
1.81 schwarze 678: curp->pman = man_alloc(
679: curp->roff, curp, curp->defos,
1.50 schwarze 680: curp->options & MPARSE_QUICK ? 1 : 0);
681: curp->man = curp->pman;
682: }
683: }
1.86 schwarze 684: if (curp->mdoc)
685: mdoc_endparse(curp->mdoc);
686: if (curp->man)
687: man_endparse(curp->man);
1.1 schwarze 688: roff_endparse(curp->roff);
689: }
690:
1.15 schwarze 691: static void
692: mparse_parse_buffer(struct mparse *curp, struct buf blk, const char *file)
1.1 schwarze 693: {
1.61 schwarze 694: struct buf *svprimary;
1.1 schwarze 695: const char *svfile;
1.71 schwarze 696: size_t offset;
1.14 schwarze 697: static int recursion_depth;
698:
699: if (64 < recursion_depth) {
700: mandoc_msg(MANDOCERR_ROFFLOOP, curp, curp->line, 0, NULL);
1.15 schwarze 701: return;
1.14 schwarze 702: }
1.1 schwarze 703:
1.15 schwarze 704: /* Line number is per-file. */
705: svfile = curp->file;
706: curp->file = file;
1.61 schwarze 707: svprimary = curp->primary;
1.59 schwarze 708: curp->primary = &blk;
1.15 schwarze 709: curp->line = 1;
710: recursion_depth++;
711:
1.70 schwarze 712: /* Skip an UTF-8 byte order mark. */
713: if (curp->filenc & MPARSE_UTF8 && blk.sz > 2 &&
714: (unsigned char)blk.buf[0] == 0xef &&
715: (unsigned char)blk.buf[1] == 0xbb &&
716: (unsigned char)blk.buf[2] == 0xbf) {
1.71 schwarze 717: offset = 3;
1.70 schwarze 718: curp->filenc &= ~MPARSE_LATIN1;
1.71 schwarze 719: } else
720: offset = 0;
1.70 schwarze 721:
1.71 schwarze 722: mparse_buf_r(curp, blk, offset, 1);
1.15 schwarze 723:
1.86 schwarze 724: if (--recursion_depth == 0)
1.15 schwarze 725: mparse_end(curp);
726:
1.61 schwarze 727: curp->primary = svprimary;
1.15 schwarze 728: curp->file = svfile;
729: }
730:
1.74 schwarze 731: /*
732: * Read the whole file into memory and call the parsers.
733: * Called recursively when an .so request is encountered.
734: */
1.15 schwarze 735: enum mandoclevel
736: mparse_readfd(struct mparse *curp, int fd, const char *file)
737: {
738: struct buf blk;
739: int with_mmap;
1.70 schwarze 740: int save_filenc;
1.1 schwarze 741:
1.67 schwarze 742: if (read_whole_file(curp, file, fd, &blk, &with_mmap)) {
1.70 schwarze 743: save_filenc = curp->filenc;
744: curp->filenc = curp->options &
745: (MPARSE_UTF8 | MPARSE_LATIN1);
1.67 schwarze 746: mparse_parse_buffer(curp, blk, file);
1.70 schwarze 747: curp->filenc = save_filenc;
1.67 schwarze 748: if (with_mmap)
749: munmap(blk.buf, blk.sz);
750: else
751: free(blk.buf);
752: }
1.1 schwarze 753:
1.74 schwarze 754: if (fd != STDIN_FILENO && close(fd) == -1)
1.1 schwarze 755: perror(file);
1.67 schwarze 756:
1.74 schwarze 757: mparse_wait(curp);
1.1 schwarze 758: return(curp->file_status);
1.58 schwarze 759: }
760:
761: enum mandoclevel
1.73 schwarze 762: mparse_open(struct mparse *curp, int *fd, const char *file)
1.58 schwarze 763: {
764: int pfd[2];
1.74 schwarze 765: int save_errno;
1.58 schwarze 766: char *cp;
767:
768: curp->file = file;
1.74 schwarze 769:
770: /* Unless zipped, try to just open the file. */
771:
1.58 schwarze 772: if ((cp = strrchr(file, '.')) == NULL ||
773: strcmp(cp + 1, "gz")) {
1.73 schwarze 774: curp->child = 0;
1.74 schwarze 775: if ((*fd = open(file, O_RDONLY)) != -1)
776: return(MANDOCLEVEL_OK);
777:
778: /* Open failed; try to append ".gz". */
779:
780: mandoc_asprintf(&cp, "%s.gz", file);
781: file = cp;
782: } else
783: cp = NULL;
784:
785: /* Before forking, make sure the file can be read. */
786:
787: save_errno = errno;
788: if (access(file, R_OK) == -1) {
789: if (cp != NULL)
790: errno = save_errno;
1.83 schwarze 791: free(cp);
792: *fd = -1;
793: curp->child = 0;
794: mandoc_msg(MANDOCERR_FILE, curp, 0, 0, strerror(errno));
795: return(MANDOCLEVEL_ERROR);
1.58 schwarze 796: }
797:
1.74 schwarze 798: /* Run gunzip(1). */
799:
1.58 schwarze 800: if (pipe(pfd) == -1) {
1.82 schwarze 801: perror("pipe");
802: exit((int)MANDOCLEVEL_SYSERR);
1.58 schwarze 803: }
804:
1.73 schwarze 805: switch (curp->child = fork()) {
1.58 schwarze 806: case -1:
1.82 schwarze 807: perror("fork");
808: exit((int)MANDOCLEVEL_SYSERR);
1.58 schwarze 809: case 0:
810: close(pfd[0]);
811: if (dup2(pfd[1], STDOUT_FILENO) == -1) {
1.82 schwarze 812: perror("dup");
813: exit((int)MANDOCLEVEL_SYSERR);
1.58 schwarze 814: }
815: execlp("gunzip", "gunzip", "-c", file, NULL);
1.82 schwarze 816: perror("exec");
817: exit((int)MANDOCLEVEL_SYSERR);
1.58 schwarze 818: default:
819: close(pfd[1]);
820: *fd = pfd[0];
821: return(MANDOCLEVEL_OK);
822: }
823: }
824:
825: enum mandoclevel
1.73 schwarze 826: mparse_wait(struct mparse *curp)
1.58 schwarze 827: {
828: int status;
829:
1.73 schwarze 830: if (curp->child == 0)
831: return(MANDOCLEVEL_OK);
832:
833: if (waitpid(curp->child, &status, 0) == -1) {
1.82 schwarze 834: perror("wait");
835: exit((int)MANDOCLEVEL_SYSERR);
1.58 schwarze 836: }
837: if (WIFSIGNALED(status)) {
1.83 schwarze 838: mandoc_vmsg(MANDOCERR_FILE, curp, 0, 0,
839: "gunzip died from signal %d", WTERMSIG(status));
840: return(MANDOCLEVEL_ERROR);
1.58 schwarze 841: }
842: if (WEXITSTATUS(status)) {
1.83 schwarze 843: mandoc_vmsg(MANDOCERR_FILE, curp, 0, 0,
844: "gunzip failed with code %d", WEXITSTATUS(status));
845: return(MANDOCLEVEL_ERROR);
1.58 schwarze 846: }
847: return(MANDOCLEVEL_OK);
1.1 schwarze 848: }
849:
850: struct mparse *
1.69 schwarze 851: mparse_alloc(int options, enum mandoclevel wlevel, mandocmsg mmsg,
852: const struct mchars *mchars, const char *defos)
1.1 schwarze 853: {
854: struct mparse *curp;
855:
856: curp = mandoc_calloc(1, sizeof(struct mparse));
857:
1.22 schwarze 858: curp->options = options;
1.1 schwarze 859: curp->wlevel = wlevel;
860: curp->mmsg = mmsg;
1.7 schwarze 861: curp->defos = defos;
1.1 schwarze 862:
1.69 schwarze 863: curp->mchars = mchars;
864: curp->roff = roff_alloc(curp, curp->mchars, options);
1.50 schwarze 865: if (curp->options & MPARSE_MDOC)
866: curp->pmdoc = mdoc_alloc(
867: curp->roff, curp, curp->defos,
868: curp->options & MPARSE_QUICK ? 1 : 0);
869: if (curp->options & MPARSE_MAN)
1.81 schwarze 870: curp->pman = man_alloc(
871: curp->roff, curp, curp->defos,
1.50 schwarze 872: curp->options & MPARSE_QUICK ? 1 : 0);
873:
1.1 schwarze 874: return(curp);
875: }
876:
877: void
878: mparse_reset(struct mparse *curp)
879: {
880:
881: roff_reset(curp->roff);
882:
883: if (curp->mdoc)
884: mdoc_reset(curp->mdoc);
885: if (curp->man)
886: man_reset(curp->man);
1.4 schwarze 887: if (curp->secondary)
888: curp->secondary->sz = 0;
1.1 schwarze 889:
890: curp->file_status = MANDOCLEVEL_OK;
891: curp->mdoc = NULL;
892: curp->man = NULL;
1.23 schwarze 893:
894: free(curp->sodest);
895: curp->sodest = NULL;
1.1 schwarze 896: }
897:
898: void
899: mparse_free(struct mparse *curp)
900: {
901:
902: if (curp->pmdoc)
903: mdoc_free(curp->pmdoc);
904: if (curp->pman)
905: man_free(curp->pman);
906: if (curp->roff)
907: roff_free(curp->roff);
1.4 schwarze 908: if (curp->secondary)
909: free(curp->secondary->buf);
1.1 schwarze 910:
1.4 schwarze 911: free(curp->secondary);
1.23 schwarze 912: free(curp->sodest);
1.1 schwarze 913: free(curp);
914: }
915:
916: void
1.23 schwarze 917: mparse_result(struct mparse *curp,
918: struct mdoc **mdoc, struct man **man, char **sodest)
1.1 schwarze 919: {
920:
1.23 schwarze 921: if (sodest && NULL != (*sodest = curp->sodest)) {
922: *mdoc = NULL;
923: *man = NULL;
924: return;
925: }
1.1 schwarze 926: if (mdoc)
927: *mdoc = curp->mdoc;
928: if (man)
929: *man = curp->man;
930: }
931:
932: void
933: mandoc_vmsg(enum mandocerr t, struct mparse *m,
934: int ln, int pos, const char *fmt, ...)
935: {
936: char buf[256];
937: va_list ap;
938:
939: va_start(ap, fmt);
1.26 schwarze 940: (void)vsnprintf(buf, sizeof(buf), fmt, ap);
1.1 schwarze 941: va_end(ap);
942:
943: mandoc_msg(t, m, ln, pos, buf);
944: }
945:
946: void
1.25 schwarze 947: mandoc_msg(enum mandocerr er, struct mparse *m,
1.1 schwarze 948: int ln, int col, const char *msg)
949: {
950: enum mandoclevel level;
951:
1.87 schwarze 952: level = MANDOCLEVEL_UNSUPP;
1.1 schwarze 953: while (er < mandoclimits[level])
954: level--;
955:
1.83 schwarze 956: if (level < m->wlevel && er != MANDOCERR_FILE)
1.1 schwarze 957: return;
958:
959: if (m->mmsg)
960: (*m->mmsg)(er, level, m->file, ln, col, msg);
961:
962: if (m->file_status < level)
963: m->file_status = level;
964: }
965:
966: const char *
967: mparse_strerror(enum mandocerr er)
968: {
969:
970: return(mandocerrs[er]);
971: }
972:
973: const char *
974: mparse_strlevel(enum mandoclevel lvl)
975: {
976: return(mandoclevels[lvl]);
1.4 schwarze 977: }
978:
979: void
980: mparse_keep(struct mparse *p)
981: {
982:
983: assert(NULL == p->secondary);
984: p->secondary = mandoc_calloc(1, sizeof(struct buf));
985: }
986:
987: const char *
988: mparse_getkeep(const struct mparse *p)
989: {
990:
991: assert(p->secondary);
992: return(p->secondary->sz ? p->secondary->buf : NULL);
1.1 schwarze 993: }
![[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