Annotation of src/usr.bin/mandoc/read.c, Revision 1.125
1.125 ! schwarze 1: /* $OpenBSD: read.c,v 1.124 2016/07/19 16:22:34 schwarze Exp $ */
1.1 schwarze 2: /*
3: * Copyright (c) 2008, 2009, 2010, 2011 Kristaps Dzonsons <kristaps@bsd.lv>
1.122 schwarze 4: * Copyright (c) 2010-2016 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: *
1.108 schwarze 11: * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHORS DISCLAIM ALL WARRANTIES
1.1 schwarze 12: * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
1.108 schwarze 13: * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHORS BE LIABLE FOR
1.1 schwarze 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>
22:
23: #include <assert.h>
24: #include <ctype.h>
1.118 schwarze 25: #include <err.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>
1.115 schwarze 33: #include <zlib.h>
1.1 schwarze 34:
1.108 schwarze 35: #include "mandoc_aux.h"
1.1 schwarze 36: #include "mandoc.h"
1.108 schwarze 37: #include "roff.h"
1.1 schwarze 38: #include "mdoc.h"
39: #include "man.h"
1.108 schwarze 40: #include "libmandoc.h"
1.114 schwarze 41: #include "roff_int.h"
1.1 schwarze 42:
43: #define REPARSE_LIMIT 1000
44:
45: struct mparse {
1.109 schwarze 46: struct roff_man *man; /* man parser */
1.1 schwarze 47: struct roff *roff; /* roff parser (!NULL) */
1.23 schwarze 48: char *sodest; /* filename pointed to by .so */
1.59 schwarze 49: const char *file; /* filename of current input file */
50: struct buf *primary; /* buffer currently being parsed */
51: struct buf *secondary; /* preprocessed copy of input */
52: const char *defos; /* default operating system */
53: mandocmsg mmsg; /* warning/error message handler */
54: enum mandoclevel file_status; /* status of current parse */
55: enum mandoclevel wlevel; /* ignore messages below this */
56: int options; /* parser options */
1.115 schwarze 57: int gzip; /* current input file is gzipped */
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 */
1.1 schwarze 61: };
62:
1.60 schwarze 63: static void choose_parser(struct mparse *);
1.1 schwarze 64: static void resize_buf(struct buf *, size_t);
1.71 schwarze 65: static void mparse_buf_r(struct mparse *, struct buf, size_t, int);
1.18 schwarze 66: static int read_whole_file(struct mparse *, const char *, int,
67: struct buf *, int *);
1.1 schwarze 68: static void mparse_end(struct mparse *);
1.15 schwarze 69: static void mparse_parse_buffer(struct mparse *, struct buf,
70: const char *);
1.1 schwarze 71:
72: static const enum mandocerr mandoclimits[MANDOCLEVEL_MAX] = {
73: MANDOCERR_OK,
74: MANDOCERR_WARNING,
75: MANDOCERR_WARNING,
76: MANDOCERR_ERROR,
1.87 schwarze 77: MANDOCERR_UNSUPP,
1.1 schwarze 78: MANDOCERR_MAX,
79: MANDOCERR_MAX
80: };
81:
82: static const char * const mandocerrs[MANDOCERR_MAX] = {
83: "ok",
84:
85: "generic warning",
86:
87: /* related to the prologue */
1.57 schwarze 88: "missing manual title, using UNTITLED",
89: "missing manual title, using \"\"",
1.32 schwarze 90: "lower case character in document title",
1.57 schwarze 91: "missing manual section, using \"\"",
1.1 schwarze 92: "unknown manual section",
1.32 schwarze 93: "missing date, using today's date",
1.1 schwarze 94: "cannot parse date, using it verbatim",
1.57 schwarze 95: "missing Os macro, using \"\"",
96: "duplicate prologue macro",
97: "late prologue macro",
98: "skipping late title macro",
1.1 schwarze 99: "prologue macros out of order",
100:
101: /* related to document structure */
102: ".so is fragile, better use ln(1)",
1.28 schwarze 103: "no document body",
1.32 schwarze 104: "content before first section header",
105: "first section is not \"NAME\"",
1.103 schwarze 106: "NAME section without name",
107: "NAME section without description",
108: "description not at the end of NAME",
109: "bad NAME section content",
1.96 schwarze 110: "missing description line, using \"\"",
1.1 schwarze 111: "sections out of conventional order",
1.32 schwarze 112: "duplicate section title",
113: "unexpected section",
1.63 schwarze 114: "unusual Xr order",
115: "unusual Xr punctuation",
1.62 schwarze 116: "AUTHORS section without An macro",
1.1 schwarze 117:
118: /* related to macros and nesting */
1.33 schwarze 119: "obsolete macro",
1.78 schwarze 120: "macro neither callable nor escaped",
1.1 schwarze 121: "skipping paragraph macro",
1.10 schwarze 122: "moving paragraph macro out of list",
1.1 schwarze 123: "skipping no-space macro",
124: "blocks badly nested",
125: "nested displays are not portable",
1.35 schwarze 126: "moving content out of list",
1.56 schwarze 127: "fill mode already enabled, skipping",
128: "fill mode already disabled, skipping",
1.1 schwarze 129: "line scope broken",
130:
131: /* related to missing macro arguments */
1.36 schwarze 132: "skipping empty request",
133: "conditional request controls empty scope",
1.1 schwarze 134: "skipping empty macro",
1.99 schwarze 135: "empty block",
1.40 schwarze 136: "empty argument, using 0n",
1.38 schwarze 137: "missing display type, using -ragged",
138: "list type is not the first argument",
1.125 ! schwarze 139: "missing -width in -tag list, using 6n",
1.56 schwarze 140: "missing utility name, using \"\"",
1.98 schwarze 141: "missing function name, using \"\"",
1.38 schwarze 142: "empty head in list item",
143: "empty list item",
1.39 schwarze 144: "missing font type, using \\fR",
145: "unknown font type, using \\fR",
1.79 schwarze 146: "nothing follows prefix",
1.97 schwarze 147: "empty reference block",
1.38 schwarze 148: "missing -std argument, adding it",
1.100 schwarze 149: "missing option string, using \"\"",
150: "missing resource identifier, using \"\"",
1.66 schwarze 151: "missing eqn box, using \"\"",
1.1 schwarze 152:
153: /* related to bad macro arguments */
1.42 schwarze 154: "unterminated quoted argument",
1.1 schwarze 155: "duplicate argument",
1.54 schwarze 156: "skipping duplicate argument",
1.41 schwarze 157: "skipping duplicate display type",
158: "skipping duplicate list type",
1.54 schwarze 159: "skipping -width argument",
1.101 schwarze 160: "wrong number of cells",
1.1 schwarze 161: "unknown AT&T UNIX version",
1.64 schwarze 162: "comma in function argument",
1.65 schwarze 163: "parenthesis in function name",
1.45 schwarze 164: "invalid content in Rs block",
1.41 schwarze 165: "invalid Boolean argument",
166: "unknown font, skipping request",
1.101 schwarze 167: "odd number of characters in request",
1.1 schwarze 168:
169: /* related to plain text */
1.42 schwarze 170: "blank line in fill mode, using .sp",
171: "tab in filled text",
172: "whitespace at end of input line",
1.1 schwarze 173: "bad comment style",
1.42 schwarze 174: "invalid escape sequence",
175: "undefined string, using \"\"",
1.3 schwarze 176:
1.93 schwarze 177: /* related to tables */
178: "tbl line starts with span",
179: "tbl column starts with span",
180: "skipping vertical bar in tbl layout",
181:
1.1 schwarze 182: "generic error",
1.3 schwarze 183:
1.1 schwarze 184: /* related to tables */
1.91 schwarze 185: "non-alphabetic character in tbl options",
186: "skipping unknown tbl option",
187: "missing tbl option argument",
188: "wrong tbl option argument size",
1.93 schwarze 189: "empty tbl layout",
190: "invalid character in tbl layout",
191: "unmatched parenthesis in tbl layout",
1.94 schwarze 192: "tbl without any data cells",
193: "ignoring data in spanned tbl cell",
194: "ignoring extra tbl data cells",
195: "data block open at end of tbl",
1.1 schwarze 196:
1.46 schwarze 197: /* related to document structure and macros */
1.83 schwarze 198: NULL,
1.1 schwarze 199: "input stack limit exceeded, infinite loop?",
200: "skipping bad character",
1.46 schwarze 201: "skipping unknown macro",
1.87 schwarze 202: "skipping insecure request",
1.51 schwarze 203: "skipping item outside list",
1.46 schwarze 204: "skipping column outside column list",
205: "skipping end of block that is not open",
1.90 schwarze 206: "fewer RS blocks open, skipping",
1.46 schwarze 207: "inserting missing end of block",
208: "appending missing end of block",
209:
210: /* related to request and macro arguments */
1.1 schwarze 211: "escaped character not allowed in a name",
1.75 schwarze 212: "NOT IMPLEMENTED: Bd -file",
1.120 schwarze 213: "skipping display without arguments",
1.49 schwarze 214: "missing list type, using -item",
1.48 schwarze 215: "missing manual name, using \"\"",
1.49 schwarze 216: "uname(3) system call failed, using UNKNOWN",
1.41 schwarze 217: "unknown standard specifier",
1.49 schwarze 218: "skipping request without numeric argument",
1.84 schwarze 219: "NOT IMPLEMENTED: .so with absolute path or \"..\"",
1.85 schwarze 220: ".so request failed",
1.39 schwarze 221: "skipping all arguments",
222: "skipping excess arguments",
1.68 schwarze 223: "divide by zero",
1.87 schwarze 224:
225: "unsupported feature",
226: "input too large",
1.89 schwarze 227: "unsupported control character",
1.87 schwarze 228: "unsupported roff request",
1.94 schwarze 229: "eqn delim option in tbl",
1.93 schwarze 230: "unsupported tbl layout modifier",
1.87 schwarze 231: "ignoring macro in table",
1.1 schwarze 232: };
233:
234: static const char * const mandoclevels[MANDOCLEVEL_MAX] = {
235: "SUCCESS",
236: "RESERVED",
237: "WARNING",
238: "ERROR",
1.87 schwarze 239: "UNSUPP",
1.1 schwarze 240: "BADARG",
241: "SYSERR"
242: };
243:
1.25 schwarze 244:
1.1 schwarze 245: static void
246: resize_buf(struct buf *buf, size_t initial)
247: {
248:
249: buf->sz = buf->sz > initial/2 ? 2 * buf->sz : initial;
250: buf->buf = mandoc_realloc(buf->buf, buf->sz);
251: }
252:
253: static void
1.60 schwarze 254: choose_parser(struct mparse *curp)
1.1 schwarze 255: {
1.59 schwarze 256: char *cp, *ep;
257: int format;
1.1 schwarze 258:
1.59 schwarze 259: /*
260: * If neither command line arguments -mdoc or -man select
261: * a parser nor the roff parser found a .Dd or .TH macro
262: * yet, look ahead in the main input buffer.
263: */
264:
265: if ((format = roff_getformat(curp->roff)) == 0) {
266: cp = curp->primary->buf;
267: ep = cp + curp->primary->sz;
268: while (cp < ep) {
1.61 schwarze 269: if (*cp == '.' || *cp == '\'') {
1.59 schwarze 270: cp++;
271: if (cp[0] == 'D' && cp[1] == 'd') {
272: format = MPARSE_MDOC;
273: break;
274: }
275: if (cp[0] == 'T' && cp[1] == 'H') {
276: format = MPARSE_MAN;
277: break;
278: }
279: }
280: cp = memchr(cp, '\n', ep - cp);
281: if (cp == NULL)
282: break;
283: cp++;
284: }
1.1 schwarze 285: }
286:
1.59 schwarze 287: if (format == MPARSE_MDOC) {
1.111 schwarze 288: mdoc_hash_init();
1.112 schwarze 289: curp->man->macroset = MACROSET_MDOC;
1.113 schwarze 290: curp->man->first->tok = TOKEN_NONE;
1.112 schwarze 291: } else {
292: man_hash_init();
293: curp->man->macroset = MACROSET_MAN;
1.113 schwarze 294: curp->man->first->tok = TOKEN_NONE;
1.25 schwarze 295: }
1.1 schwarze 296: }
297:
298: /*
1.71 schwarze 299: * Main parse routine for a buffer.
300: * It assumes encoding and line numbering are already set up.
301: * It can recurse directly (for invocations of user-defined
302: * macros, inline equations, and input line traps)
303: * and indirectly (for .so file inclusion).
1.1 schwarze 304: */
305: static void
1.71 schwarze 306: mparse_buf_r(struct mparse *curp, struct buf blk, size_t i, int start)
1.1 schwarze 307: {
308: const struct tbl_span *span;
309: struct buf ln;
1.88 schwarze 310: const char *save_file;
1.85 schwarze 311: char *cp;
1.71 schwarze 312: size_t pos; /* byte number in the ln buffer */
1.1 schwarze 313: enum rofferr rr;
1.76 schwarze 314: int of;
1.1 schwarze 315: int lnn; /* line number in the real file */
1.85 schwarze 316: int fd;
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++;
1.102 schwarze 389: if (c != '\r')
390: ln.buf[pos++] = '?';
1.1 schwarze 391: continue;
392: }
393:
394: /* Trailing backslash = a plain char. */
395:
1.71 schwarze 396: if (blk.buf[i] != '\\' || i + 1 == blk.sz) {
1.1 schwarze 397: ln.buf[pos++] = blk.buf[i++];
398: continue;
399: }
400:
401: /*
402: * Found escape and at least one other character.
403: * When it's a newline character, skip it.
404: * When there is a carriage return in between,
405: * skip that one as well.
406: */
407:
1.71 schwarze 408: if ('\r' == blk.buf[i + 1] && i + 2 < blk.sz &&
1.1 schwarze 409: '\n' == blk.buf[i + 2])
410: ++i;
411: if ('\n' == blk.buf[i + 1]) {
412: i += 2;
413: ++lnn;
414: continue;
415: }
416:
417: if ('"' == blk.buf[i + 1] || '#' == blk.buf[i + 1]) {
418: i += 2;
419: /* Comment, skip to end of line */
1.71 schwarze 420: for (; i < blk.sz; ++i) {
1.1 schwarze 421: if ('\n' == blk.buf[i]) {
422: ++i;
423: ++lnn;
424: break;
425: }
426: }
427:
428: /* Backout trailing whitespaces */
429: for (; pos > 0; --pos) {
430: if (ln.buf[pos - 1] != ' ')
431: break;
432: if (pos > 2 && ln.buf[pos - 2] == '\\')
433: break;
434: }
435: break;
436: }
437:
1.13 schwarze 438: /* Catch escaped bogus characters. */
439:
440: c = (unsigned char) blk.buf[i+1];
441:
1.25 schwarze 442: if ( ! (isascii(c) &&
443: (isgraph(c) || isblank(c)))) {
1.89 schwarze 444: mandoc_vmsg(MANDOCERR_CHAR_BAD, curp,
1.56 schwarze 445: curp->line, pos, "0x%x", c);
1.13 schwarze 446: i += 2;
447: ln.buf[pos++] = '?';
448: continue;
449: }
450:
1.1 schwarze 451: /* Some other escape sequence, copy & cont. */
452:
453: ln.buf[pos++] = blk.buf[i++];
454: ln.buf[pos++] = blk.buf[i++];
455: }
456:
1.71 schwarze 457: if (pos >= ln.sz)
1.1 schwarze 458: resize_buf(&ln, 256);
459:
460: ln.buf[pos] = '\0';
461:
462: /*
463: * A significant amount of complexity is contained by
464: * the roff preprocessor. It's line-oriented but can be
465: * expressed on one line, so we need at times to
466: * readjust our starting point and re-run it. The roff
467: * preprocessor can also readjust the buffers with new
468: * data, so we pass them in wholesale.
469: */
470:
471: of = 0;
472:
1.4 schwarze 473: /*
474: * Maintain a lookaside buffer of all parsed lines. We
475: * only do this if mparse_keep() has been invoked (the
476: * buffer may be accessed with mparse_getkeep()).
477: */
478:
479: if (curp->secondary) {
1.25 schwarze 480: curp->secondary->buf = mandoc_realloc(
481: curp->secondary->buf,
482: curp->secondary->sz + pos + 2);
483: memcpy(curp->secondary->buf +
484: curp->secondary->sz,
485: ln.buf, pos);
1.4 schwarze 486: curp->secondary->sz += pos;
487: curp->secondary->buf
488: [curp->secondary->sz] = '\n';
489: curp->secondary->sz++;
490: curp->secondary->buf
491: [curp->secondary->sz] = '\0';
492: }
1.1 schwarze 493: rerun:
1.72 schwarze 494: rr = roff_parseln(curp->roff, curp->line, &ln, &of);
1.1 schwarze 495:
496: switch (rr) {
1.25 schwarze 497: case ROFF_REPARSE:
1.1 schwarze 498: if (REPARSE_LIMIT >= ++curp->reparse_count)
1.71 schwarze 499: mparse_buf_r(curp, ln, of, 0);
1.1 schwarze 500: else
501: mandoc_msg(MANDOCERR_ROFFLOOP, curp,
1.25 schwarze 502: curp->line, pos, NULL);
1.1 schwarze 503: pos = 0;
504: continue;
1.25 schwarze 505: case ROFF_APPEND:
1.71 schwarze 506: pos = strlen(ln.buf);
1.1 schwarze 507: continue;
1.25 schwarze 508: case ROFF_RERUN:
1.1 schwarze 509: goto rerun;
1.25 schwarze 510: case ROFF_IGN:
1.1 schwarze 511: pos = 0;
512: continue;
1.25 schwarze 513: case ROFF_SO:
1.71 schwarze 514: if ( ! (curp->options & MPARSE_SO) &&
515: (i >= blk.sz || blk.buf[i] == '\0')) {
1.23 schwarze 516: curp->sodest = mandoc_strdup(ln.buf + of);
517: free(ln.buf);
518: return;
519: }
1.4 schwarze 520: /*
521: * We remove `so' clauses from our lookaside
522: * buffer because we're going to descend into
523: * the file recursively.
524: */
1.25 schwarze 525: if (curp->secondary)
1.4 schwarze 526: curp->secondary->sz -= pos + 1;
1.88 schwarze 527: save_file = curp->file;
1.122 schwarze 528: if ((fd = mparse_open(curp, ln.buf + of)) != -1) {
1.85 schwarze 529: mparse_readfd(curp, fd, ln.buf + of);
1.121 schwarze 530: close(fd);
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: }
1.1 schwarze 545: pos = 0;
546: continue;
547: default:
548: break;
549: }
550:
1.124 schwarze 551: if (curp->man->macroset == MACROSET_NONE)
1.60 schwarze 552: choose_parser(curp);
1.1 schwarze 553:
1.25 schwarze 554: /*
1.60 schwarze 555: * Lastly, push down into the parsers themselves.
1.1 schwarze 556: * If libroff returns ROFF_TBL, then add it to the
557: * currently open parse. Since we only get here if
558: * there does exist data (see tbl_data.c), we're
559: * guaranteed that something's been allocated.
560: * Do the same for ROFF_EQN.
561: */
562:
1.114 schwarze 563: if (rr == ROFF_TBL)
1.76 schwarze 564: while ((span = roff_span(curp->roff)) != NULL)
1.114 schwarze 565: roff_addtbl(curp->man, span);
566: else if (rr == ROFF_EQN)
567: roff_addeqn(curp->man, roff_eqn(curp->roff));
568: else if ((curp->man->macroset == MACROSET_MDOC ?
1.110 schwarze 569: mdoc_parseln(curp->man, curp->line, ln.buf, of) :
1.76 schwarze 570: man_parseln(curp->man, curp->line, ln.buf, of)) == 2)
571: break;
1.1 schwarze 572:
573: /* Temporary buffers typically are not full. */
574:
575: if (0 == start && '\0' == blk.buf[i])
576: break;
577:
578: /* Start the next input line. */
579:
580: pos = 0;
581: }
582:
583: free(ln.buf);
584: }
585:
586: static int
1.18 schwarze 587: read_whole_file(struct mparse *curp, const char *file, int fd,
588: struct buf *fb, int *with_mmap)
1.1 schwarze 589: {
590: struct stat st;
1.115 schwarze 591: gzFile gz;
1.1 schwarze 592: size_t off;
593: ssize_t ssz;
594:
1.118 schwarze 595: if (fstat(fd, &st) == -1)
596: err((int)MANDOCLEVEL_SYSERR, "%s", file);
1.1 schwarze 597:
598: /*
599: * If we're a regular file, try just reading in the whole entry
600: * via mmap(). This is faster than reading it into blocks, and
601: * since each file is only a few bytes to begin with, I'm not
602: * concerned that this is going to tank any machines.
603: */
604:
1.115 schwarze 605: if (curp->gzip == 0 && S_ISREG(st.st_mode)) {
1.106 schwarze 606: if (st.st_size > 0x7fffffff) {
1.86 schwarze 607: mandoc_msg(MANDOCERR_TOOLARGE, curp, 0, 0, NULL);
1.117 schwarze 608: return 0;
1.1 schwarze 609: }
610: *with_mmap = 1;
611: fb->sz = (size_t)st.st_size;
1.15 schwarze 612: fb->buf = mmap(NULL, fb->sz, PROT_READ, MAP_SHARED, fd, 0);
1.1 schwarze 613: if (fb->buf != MAP_FAILED)
1.117 schwarze 614: return 1;
1.1 schwarze 615: }
616:
1.115 schwarze 617: if (curp->gzip) {
1.118 schwarze 618: if ((gz = gzdopen(fd, "rb")) == NULL)
619: err((int)MANDOCLEVEL_SYSERR, "%s", file);
1.115 schwarze 620: } else
621: gz = NULL;
622:
1.1 schwarze 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)) {
1.86 schwarze 635: mandoc_msg(MANDOCERR_TOOLARGE, curp,
636: 0, 0, NULL);
1.1 schwarze 637: break;
638: }
639: resize_buf(fb, 65536);
640: }
1.115 schwarze 641: ssz = curp->gzip ?
642: gzread(gz, fb->buf + (int)off, fb->sz - off) :
643: read(fd, fb->buf + (int)off, fb->sz - off);
1.1 schwarze 644: if (ssz == 0) {
645: fb->sz = off;
1.117 schwarze 646: return 1;
1.1 schwarze 647: }
1.118 schwarze 648: if (ssz == -1)
649: err((int)MANDOCLEVEL_SYSERR, "%s", file);
1.1 schwarze 650: off += (size_t)ssz;
651: }
652:
653: free(fb->buf);
654: fb->buf = NULL;
1.117 schwarze 655: return 0;
1.1 schwarze 656: }
657:
658: static void
659: mparse_end(struct mparse *curp)
660: {
1.110 schwarze 661: if (curp->man->macroset == MACROSET_NONE)
662: curp->man->macroset = MACROSET_MAN;
663: if (curp->man->macroset == MACROSET_MDOC)
664: mdoc_endparse(curp->man);
665: else
1.86 schwarze 666: man_endparse(curp->man);
1.1 schwarze 667: roff_endparse(curp->roff);
668: }
669:
1.15 schwarze 670: static void
671: mparse_parse_buffer(struct mparse *curp, struct buf blk, const char *file)
1.1 schwarze 672: {
1.61 schwarze 673: struct buf *svprimary;
1.1 schwarze 674: const char *svfile;
1.71 schwarze 675: size_t offset;
1.14 schwarze 676: static int recursion_depth;
677:
678: if (64 < recursion_depth) {
679: mandoc_msg(MANDOCERR_ROFFLOOP, curp, curp->line, 0, NULL);
1.15 schwarze 680: return;
1.14 schwarze 681: }
1.1 schwarze 682:
1.15 schwarze 683: /* Line number is per-file. */
684: svfile = curp->file;
685: curp->file = file;
1.61 schwarze 686: svprimary = curp->primary;
1.59 schwarze 687: curp->primary = &blk;
1.15 schwarze 688: curp->line = 1;
689: recursion_depth++;
690:
1.70 schwarze 691: /* Skip an UTF-8 byte order mark. */
692: if (curp->filenc & MPARSE_UTF8 && blk.sz > 2 &&
693: (unsigned char)blk.buf[0] == 0xef &&
694: (unsigned char)blk.buf[1] == 0xbb &&
695: (unsigned char)blk.buf[2] == 0xbf) {
1.71 schwarze 696: offset = 3;
1.70 schwarze 697: curp->filenc &= ~MPARSE_LATIN1;
1.71 schwarze 698: } else
699: offset = 0;
1.70 schwarze 700:
1.71 schwarze 701: mparse_buf_r(curp, blk, offset, 1);
1.15 schwarze 702:
1.86 schwarze 703: if (--recursion_depth == 0)
1.15 schwarze 704: mparse_end(curp);
705:
1.61 schwarze 706: curp->primary = svprimary;
1.15 schwarze 707: curp->file = svfile;
708: }
709:
1.74 schwarze 710: /*
711: * Read the whole file into memory and call the parsers.
712: * Called recursively when an .so request is encountered.
713: */
1.15 schwarze 714: enum mandoclevel
715: mparse_readfd(struct mparse *curp, int fd, const char *file)
716: {
717: struct buf blk;
718: int with_mmap;
1.70 schwarze 719: int save_filenc;
1.1 schwarze 720:
1.67 schwarze 721: if (read_whole_file(curp, file, fd, &blk, &with_mmap)) {
1.70 schwarze 722: save_filenc = curp->filenc;
723: curp->filenc = curp->options &
724: (MPARSE_UTF8 | MPARSE_LATIN1);
1.67 schwarze 725: mparse_parse_buffer(curp, blk, file);
1.70 schwarze 726: curp->filenc = save_filenc;
1.67 schwarze 727: if (with_mmap)
728: munmap(blk.buf, blk.sz);
729: else
730: free(blk.buf);
731: }
1.117 schwarze 732: return curp->file_status;
1.58 schwarze 733: }
734:
1.122 schwarze 735: int
736: mparse_open(struct mparse *curp, const char *file)
1.58 schwarze 737: {
738: char *cp;
1.122 schwarze 739: int fd;
1.58 schwarze 740:
741: curp->file = file;
1.115 schwarze 742: cp = strrchr(file, '.');
743: curp->gzip = (cp != NULL && ! strcmp(cp + 1, "gz"));
1.74 schwarze 744:
1.115 schwarze 745: /* First try to use the filename as it is. */
1.74 schwarze 746:
1.122 schwarze 747: if ((fd = open(file, O_RDONLY)) != -1)
748: return fd;
1.74 schwarze 749:
1.115 schwarze 750: /*
751: * If that doesn't work and the filename doesn't
752: * already end in .gz, try appending .gz.
753: */
1.74 schwarze 754:
1.115 schwarze 755: if ( ! curp->gzip) {
1.74 schwarze 756: mandoc_asprintf(&cp, "%s.gz", file);
1.123 schwarze 757: fd = open(cp, O_RDONLY);
1.83 schwarze 758: free(cp);
1.122 schwarze 759: if (fd != -1) {
1.115 schwarze 760: curp->gzip = 1;
1.122 schwarze 761: return fd;
1.58 schwarze 762: }
763: }
764:
1.115 schwarze 765: /* Neither worked, give up. */
1.73 schwarze 766:
1.115 schwarze 767: mandoc_msg(MANDOCERR_FILE, curp, 0, 0, strerror(errno));
1.122 schwarze 768: return -1;
1.1 schwarze 769: }
770:
771: struct mparse *
1.69 schwarze 772: mparse_alloc(int options, enum mandoclevel wlevel, mandocmsg mmsg,
1.119 schwarze 773: const char *defos)
1.1 schwarze 774: {
775: struct mparse *curp;
776:
777: curp = mandoc_calloc(1, sizeof(struct mparse));
778:
1.22 schwarze 779: curp->options = options;
1.1 schwarze 780: curp->wlevel = wlevel;
781: curp->mmsg = mmsg;
1.7 schwarze 782: curp->defos = defos;
1.1 schwarze 783:
1.119 schwarze 784: curp->roff = roff_alloc(curp, options);
1.112 schwarze 785: curp->man = roff_man_alloc( curp->roff, curp, curp->defos,
786: curp->options & MPARSE_QUICK ? 1 : 0);
1.111 schwarze 787: if (curp->options & MPARSE_MDOC) {
788: mdoc_hash_init();
1.112 schwarze 789: curp->man->macroset = MACROSET_MDOC;
790: } else if (curp->options & MPARSE_MAN) {
1.111 schwarze 791: man_hash_init();
1.112 schwarze 792: curp->man->macroset = MACROSET_MAN;
1.111 schwarze 793: }
1.113 schwarze 794: curp->man->first->tok = TOKEN_NONE;
1.117 schwarze 795: return curp;
1.1 schwarze 796: }
797:
798: void
799: mparse_reset(struct mparse *curp)
800: {
801: roff_reset(curp->roff);
1.124 schwarze 802: roff_man_reset(curp->man);
1.4 schwarze 803: if (curp->secondary)
804: curp->secondary->sz = 0;
1.1 schwarze 805:
806: curp->file_status = MANDOCLEVEL_OK;
1.23 schwarze 807:
808: free(curp->sodest);
809: curp->sodest = NULL;
1.1 schwarze 810: }
811:
812: void
813: mparse_free(struct mparse *curp)
814: {
815:
1.112 schwarze 816: roff_man_free(curp->man);
1.1 schwarze 817: if (curp->roff)
818: roff_free(curp->roff);
1.4 schwarze 819: if (curp->secondary)
820: free(curp->secondary->buf);
1.1 schwarze 821:
1.4 schwarze 822: free(curp->secondary);
1.23 schwarze 823: free(curp->sodest);
1.1 schwarze 824: free(curp);
825: }
826:
827: void
1.110 schwarze 828: mparse_result(struct mparse *curp, struct roff_man **man,
829: char **sodest)
1.1 schwarze 830: {
831:
1.23 schwarze 832: if (sodest && NULL != (*sodest = curp->sodest)) {
833: *man = NULL;
834: return;
835: }
1.1 schwarze 836: if (man)
837: *man = curp->man;
838: }
839:
840: void
841: mandoc_vmsg(enum mandocerr t, struct mparse *m,
842: int ln, int pos, const char *fmt, ...)
843: {
844: char buf[256];
845: va_list ap;
846:
847: va_start(ap, fmt);
1.26 schwarze 848: (void)vsnprintf(buf, sizeof(buf), fmt, ap);
1.1 schwarze 849: va_end(ap);
850:
851: mandoc_msg(t, m, ln, pos, buf);
852: }
853:
854: void
1.25 schwarze 855: mandoc_msg(enum mandocerr er, struct mparse *m,
1.1 schwarze 856: int ln, int col, const char *msg)
857: {
858: enum mandoclevel level;
859:
1.87 schwarze 860: level = MANDOCLEVEL_UNSUPP;
1.1 schwarze 861: while (er < mandoclimits[level])
862: level--;
863:
1.83 schwarze 864: if (level < m->wlevel && er != MANDOCERR_FILE)
1.1 schwarze 865: return;
866:
867: if (m->mmsg)
868: (*m->mmsg)(er, level, m->file, ln, col, msg);
869:
870: if (m->file_status < level)
871: m->file_status = level;
872: }
873:
874: const char *
875: mparse_strerror(enum mandocerr er)
876: {
877:
1.117 schwarze 878: return mandocerrs[er];
1.1 schwarze 879: }
880:
881: const char *
882: mparse_strlevel(enum mandoclevel lvl)
883: {
1.117 schwarze 884: return mandoclevels[lvl];
1.4 schwarze 885: }
886:
887: void
888: mparse_keep(struct mparse *p)
889: {
890:
891: assert(NULL == p->secondary);
892: p->secondary = mandoc_calloc(1, sizeof(struct buf));
893: }
894:
895: const char *
896: mparse_getkeep(const struct mparse *p)
897: {
898:
899: assert(p->secondary);
1.117 schwarze 900: return p->secondary->sz ? p->secondary->buf : NULL;
1.1 schwarze 901: }