Annotation of src/usr.bin/mandoc/read.c, Revision 1.133
1.133 ! schwarze 1: /* $OpenBSD: read.c,v 1.132 2017/02/03 17:55:36 schwarze Exp $ */
1.1 schwarze 2: /*
3: * Copyright (c) 2008, 2009, 2010, 2011 Kristaps Dzonsons <kristaps@bsd.lv>
1.129 schwarze 4: * Copyright (c) 2010-2017 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 {
1.133 ! schwarze 46: struct roff *roff; /* roff parser (!NULL) */
1.109 schwarze 47: struct roff_man *man; /* man parser */
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\"",
1.129 schwarze 106: "NAME section without Nm before Nd",
1.103 schwarze 107: "NAME section without description",
108: "description not at the end of NAME",
109: "bad NAME section content",
1.129 schwarze 110: "missing comma before name",
1.96 schwarze 111: "missing description line, using \"\"",
1.1 schwarze 112: "sections out of conventional order",
1.32 schwarze 113: "duplicate section title",
114: "unexpected section",
1.63 schwarze 115: "unusual Xr order",
116: "unusual Xr punctuation",
1.62 schwarze 117: "AUTHORS section without An macro",
1.1 schwarze 118:
119: /* related to macros and nesting */
1.33 schwarze 120: "obsolete macro",
1.78 schwarze 121: "macro neither callable nor escaped",
1.1 schwarze 122: "skipping paragraph macro",
1.10 schwarze 123: "moving paragraph macro out of list",
1.1 schwarze 124: "skipping no-space macro",
125: "blocks badly nested",
126: "nested displays are not portable",
1.35 schwarze 127: "moving content out of list",
1.56 schwarze 128: "fill mode already enabled, skipping",
129: "fill mode already disabled, skipping",
1.1 schwarze 130: "line scope broken",
131:
132: /* related to missing macro arguments */
1.36 schwarze 133: "skipping empty request",
134: "conditional request controls empty scope",
1.1 schwarze 135: "skipping empty macro",
1.99 schwarze 136: "empty block",
1.40 schwarze 137: "empty argument, using 0n",
1.38 schwarze 138: "missing display type, using -ragged",
139: "list type is not the first argument",
1.125 schwarze 140: "missing -width in -tag list, using 6n",
1.56 schwarze 141: "missing utility name, using \"\"",
1.98 schwarze 142: "missing function name, using \"\"",
1.38 schwarze 143: "empty head in list item",
144: "empty list item",
1.39 schwarze 145: "missing font type, using \\fR",
146: "unknown font type, using \\fR",
1.79 schwarze 147: "nothing follows prefix",
1.97 schwarze 148: "empty reference block",
1.128 schwarze 149: "missing section argument",
1.38 schwarze 150: "missing -std argument, adding it",
1.100 schwarze 151: "missing option string, using \"\"",
152: "missing resource identifier, using \"\"",
1.66 schwarze 153: "missing eqn box, using \"\"",
1.1 schwarze 154:
155: /* related to bad macro arguments */
1.42 schwarze 156: "unterminated quoted argument",
1.1 schwarze 157: "duplicate argument",
1.54 schwarze 158: "skipping duplicate argument",
1.41 schwarze 159: "skipping duplicate display type",
160: "skipping duplicate list type",
1.54 schwarze 161: "skipping -width argument",
1.101 schwarze 162: "wrong number of cells",
1.1 schwarze 163: "unknown AT&T UNIX version",
1.64 schwarze 164: "comma in function argument",
1.65 schwarze 165: "parenthesis in function name",
1.45 schwarze 166: "invalid content in Rs block",
1.41 schwarze 167: "invalid Boolean argument",
168: "unknown font, skipping request",
1.101 schwarze 169: "odd number of characters in request",
1.1 schwarze 170:
171: /* related to plain text */
1.42 schwarze 172: "blank line in fill mode, using .sp",
173: "tab in filled text",
174: "whitespace at end of input line",
1.131 schwarze 175: "new sentence, new line",
1.1 schwarze 176: "bad comment style",
1.42 schwarze 177: "invalid escape sequence",
178: "undefined string, using \"\"",
1.3 schwarze 179:
1.93 schwarze 180: /* related to tables */
181: "tbl line starts with span",
182: "tbl column starts with span",
183: "skipping vertical bar in tbl layout",
184:
1.1 schwarze 185: "generic error",
1.3 schwarze 186:
1.1 schwarze 187: /* related to tables */
1.91 schwarze 188: "non-alphabetic character in tbl options",
189: "skipping unknown tbl option",
190: "missing tbl option argument",
191: "wrong tbl option argument size",
1.93 schwarze 192: "empty tbl layout",
193: "invalid character in tbl layout",
194: "unmatched parenthesis in tbl layout",
1.94 schwarze 195: "tbl without any data cells",
196: "ignoring data in spanned tbl cell",
197: "ignoring extra tbl data cells",
198: "data block open at end of tbl",
1.1 schwarze 199:
1.46 schwarze 200: /* related to document structure and macros */
1.83 schwarze 201: NULL,
1.1 schwarze 202: "input stack limit exceeded, infinite loop?",
203: "skipping bad character",
1.46 schwarze 204: "skipping unknown macro",
1.87 schwarze 205: "skipping insecure request",
1.51 schwarze 206: "skipping item outside list",
1.46 schwarze 207: "skipping column outside column list",
208: "skipping end of block that is not open",
1.90 schwarze 209: "fewer RS blocks open, skipping",
1.46 schwarze 210: "inserting missing end of block",
211: "appending missing end of block",
212:
213: /* related to request and macro arguments */
1.1 schwarze 214: "escaped character not allowed in a name",
1.75 schwarze 215: "NOT IMPLEMENTED: Bd -file",
1.120 schwarze 216: "skipping display without arguments",
1.49 schwarze 217: "missing list type, using -item",
1.48 schwarze 218: "missing manual name, using \"\"",
1.49 schwarze 219: "uname(3) system call failed, using UNKNOWN",
1.41 schwarze 220: "unknown standard specifier",
1.49 schwarze 221: "skipping request without numeric argument",
1.84 schwarze 222: "NOT IMPLEMENTED: .so with absolute path or \"..\"",
1.85 schwarze 223: ".so request failed",
1.39 schwarze 224: "skipping all arguments",
225: "skipping excess arguments",
1.68 schwarze 226: "divide by zero",
1.87 schwarze 227:
228: "unsupported feature",
229: "input too large",
1.89 schwarze 230: "unsupported control character",
1.87 schwarze 231: "unsupported roff request",
1.94 schwarze 232: "eqn delim option in tbl",
1.93 schwarze 233: "unsupported tbl layout modifier",
1.87 schwarze 234: "ignoring macro in table",
1.1 schwarze 235: };
236:
237: static const char * const mandoclevels[MANDOCLEVEL_MAX] = {
238: "SUCCESS",
239: "RESERVED",
240: "WARNING",
241: "ERROR",
1.87 schwarze 242: "UNSUPP",
1.1 schwarze 243: "BADARG",
244: "SYSERR"
245: };
246:
1.25 schwarze 247:
1.1 schwarze 248: static void
249: resize_buf(struct buf *buf, size_t initial)
250: {
251:
252: buf->sz = buf->sz > initial/2 ? 2 * buf->sz : initial;
253: buf->buf = mandoc_realloc(buf->buf, buf->sz);
254: }
255:
256: static void
1.60 schwarze 257: choose_parser(struct mparse *curp)
1.1 schwarze 258: {
1.59 schwarze 259: char *cp, *ep;
260: int format;
1.1 schwarze 261:
1.59 schwarze 262: /*
263: * If neither command line arguments -mdoc or -man select
264: * a parser nor the roff parser found a .Dd or .TH macro
265: * yet, look ahead in the main input buffer.
266: */
267:
268: if ((format = roff_getformat(curp->roff)) == 0) {
269: cp = curp->primary->buf;
270: ep = cp + curp->primary->sz;
271: while (cp < ep) {
1.61 schwarze 272: if (*cp == '.' || *cp == '\'') {
1.59 schwarze 273: cp++;
274: if (cp[0] == 'D' && cp[1] == 'd') {
275: format = MPARSE_MDOC;
276: break;
277: }
278: if (cp[0] == 'T' && cp[1] == 'H') {
279: format = MPARSE_MAN;
280: break;
281: }
282: }
283: cp = memchr(cp, '\n', ep - cp);
284: if (cp == NULL)
285: break;
286: cp++;
287: }
1.1 schwarze 288: }
289:
1.59 schwarze 290: if (format == MPARSE_MDOC) {
1.111 schwarze 291: mdoc_hash_init();
1.112 schwarze 292: curp->man->macroset = MACROSET_MDOC;
1.113 schwarze 293: curp->man->first->tok = TOKEN_NONE;
1.112 schwarze 294: } else {
295: man_hash_init();
296: curp->man->macroset = MACROSET_MAN;
1.113 schwarze 297: curp->man->first->tok = TOKEN_NONE;
1.25 schwarze 298: }
1.1 schwarze 299: }
300:
301: /*
1.71 schwarze 302: * Main parse routine for a buffer.
303: * It assumes encoding and line numbering are already set up.
304: * It can recurse directly (for invocations of user-defined
305: * macros, inline equations, and input line traps)
306: * and indirectly (for .so file inclusion).
1.1 schwarze 307: */
308: static void
1.71 schwarze 309: mparse_buf_r(struct mparse *curp, struct buf blk, size_t i, int start)
1.1 schwarze 310: {
311: const struct tbl_span *span;
312: struct buf ln;
1.88 schwarze 313: const char *save_file;
1.85 schwarze 314: char *cp;
1.71 schwarze 315: size_t pos; /* byte number in the ln buffer */
1.127 schwarze 316: size_t j; /* auxiliary byte number in the blk buffer */
1.1 schwarze 317: enum rofferr rr;
1.76 schwarze 318: int of;
1.1 schwarze 319: int lnn; /* line number in the real file */
1.85 schwarze 320: int fd;
1.1 schwarze 321: unsigned char c;
322:
1.71 schwarze 323: memset(&ln, 0, sizeof(ln));
1.1 schwarze 324:
1.25 schwarze 325: lnn = curp->line;
326: pos = 0;
1.1 schwarze 327:
1.71 schwarze 328: while (i < blk.sz) {
1.1 schwarze 329: if (0 == pos && '\0' == blk.buf[i])
330: break;
331:
332: if (start) {
333: curp->line = lnn;
334: curp->reparse_count = 0;
1.70 schwarze 335:
336: if (lnn < 3 &&
337: curp->filenc & MPARSE_UTF8 &&
1.71 schwarze 338: curp->filenc & MPARSE_LATIN1)
339: curp->filenc = preconv_cue(&blk, i);
1.1 schwarze 340: }
341:
1.71 schwarze 342: while (i < blk.sz && (start || blk.buf[i] != '\0')) {
1.1 schwarze 343:
344: /*
345: * When finding an unescaped newline character,
346: * leave the character loop to process the line.
347: * Skip a preceding carriage return, if any.
348: */
349:
1.71 schwarze 350: if ('\r' == blk.buf[i] && i + 1 < blk.sz &&
1.1 schwarze 351: '\n' == blk.buf[i + 1])
352: ++i;
353: if ('\n' == blk.buf[i]) {
354: ++i;
355: ++lnn;
356: break;
357: }
358:
1.13 schwarze 359: /*
1.70 schwarze 360: * Make sure we have space for the worst
361: * case of 11 bytes: "\\[u10ffff]\0"
1.13 schwarze 362: */
363:
1.71 schwarze 364: if (pos + 11 > ln.sz)
1.13 schwarze 365: resize_buf(&ln, 256);
366:
1.25 schwarze 367: /*
1.70 schwarze 368: * Encode 8-bit input.
1.1 schwarze 369: */
370:
1.70 schwarze 371: c = blk.buf[i];
372: if (c & 0x80) {
1.71 schwarze 373: if ( ! (curp->filenc && preconv_encode(
374: &blk, &i, &ln, &pos, &curp->filenc))) {
1.89 schwarze 375: mandoc_vmsg(MANDOCERR_CHAR_BAD, curp,
376: curp->line, pos, "0x%x", c);
1.70 schwarze 377: ln.buf[pos++] = '?';
378: i++;
379: }
380: continue;
381: }
382:
383: /*
384: * Exclude control characters.
385: */
1.1 schwarze 386:
1.70 schwarze 387: if (c == 0x7f || (c < 0x20 && c != 0x09)) {
1.89 schwarze 388: mandoc_vmsg(c == 0x00 || c == 0x04 ||
389: c > 0x0a ? MANDOCERR_CHAR_BAD :
390: MANDOCERR_CHAR_UNSUPP,
391: curp, curp->line, pos, "0x%x", c);
1.1 schwarze 392: i++;
1.102 schwarze 393: if (c != '\r')
394: ln.buf[pos++] = '?';
1.1 schwarze 395: continue;
396: }
397:
398: /* Trailing backslash = a plain char. */
399:
1.71 schwarze 400: if (blk.buf[i] != '\\' || i + 1 == blk.sz) {
1.1 schwarze 401: ln.buf[pos++] = blk.buf[i++];
402: continue;
403: }
404:
405: /*
406: * Found escape and at least one other character.
407: * When it's a newline character, skip it.
408: * When there is a carriage return in between,
409: * skip that one as well.
410: */
411:
1.71 schwarze 412: if ('\r' == blk.buf[i + 1] && i + 2 < blk.sz &&
1.1 schwarze 413: '\n' == blk.buf[i + 2])
414: ++i;
415: if ('\n' == blk.buf[i + 1]) {
416: i += 2;
417: ++lnn;
418: continue;
419: }
420:
421: if ('"' == blk.buf[i + 1] || '#' == blk.buf[i + 1]) {
1.127 schwarze 422: j = i;
1.1 schwarze 423: i += 2;
424: /* Comment, skip to end of line */
1.71 schwarze 425: for (; i < blk.sz; ++i) {
1.126 schwarze 426: if (blk.buf[i] != '\n')
427: continue;
428: if (blk.buf[i - 1] == ' ' ||
429: blk.buf[i - 1] == '\t')
430: mandoc_msg(
431: MANDOCERR_SPACE_EOL,
432: curp, curp->line,
1.127 schwarze 433: pos + i-1 - j, NULL);
1.126 schwarze 434: ++i;
435: ++lnn;
436: break;
1.1 schwarze 437: }
438:
439: /* Backout trailing whitespaces */
440: for (; pos > 0; --pos) {
441: if (ln.buf[pos - 1] != ' ')
442: break;
443: if (pos > 2 && ln.buf[pos - 2] == '\\')
444: break;
445: }
446: break;
447: }
448:
1.13 schwarze 449: /* Catch escaped bogus characters. */
450:
451: c = (unsigned char) blk.buf[i+1];
452:
1.25 schwarze 453: if ( ! (isascii(c) &&
454: (isgraph(c) || isblank(c)))) {
1.89 schwarze 455: mandoc_vmsg(MANDOCERR_CHAR_BAD, curp,
1.56 schwarze 456: curp->line, pos, "0x%x", c);
1.13 schwarze 457: i += 2;
458: ln.buf[pos++] = '?';
459: continue;
460: }
461:
1.1 schwarze 462: /* Some other escape sequence, copy & cont. */
463:
464: ln.buf[pos++] = blk.buf[i++];
465: ln.buf[pos++] = blk.buf[i++];
466: }
467:
1.71 schwarze 468: if (pos >= ln.sz)
1.1 schwarze 469: resize_buf(&ln, 256);
470:
471: ln.buf[pos] = '\0';
472:
473: /*
474: * A significant amount of complexity is contained by
475: * the roff preprocessor. It's line-oriented but can be
476: * expressed on one line, so we need at times to
477: * readjust our starting point and re-run it. The roff
478: * preprocessor can also readjust the buffers with new
479: * data, so we pass them in wholesale.
480: */
481:
482: of = 0;
483:
1.4 schwarze 484: /*
485: * Maintain a lookaside buffer of all parsed lines. We
486: * only do this if mparse_keep() has been invoked (the
487: * buffer may be accessed with mparse_getkeep()).
488: */
489:
490: if (curp->secondary) {
1.25 schwarze 491: curp->secondary->buf = mandoc_realloc(
492: curp->secondary->buf,
493: curp->secondary->sz + pos + 2);
494: memcpy(curp->secondary->buf +
495: curp->secondary->sz,
496: ln.buf, pos);
1.4 schwarze 497: curp->secondary->sz += pos;
498: curp->secondary->buf
499: [curp->secondary->sz] = '\n';
500: curp->secondary->sz++;
501: curp->secondary->buf
502: [curp->secondary->sz] = '\0';
503: }
1.1 schwarze 504: rerun:
1.72 schwarze 505: rr = roff_parseln(curp->roff, curp->line, &ln, &of);
1.1 schwarze 506:
507: switch (rr) {
1.25 schwarze 508: case ROFF_REPARSE:
1.1 schwarze 509: if (REPARSE_LIMIT >= ++curp->reparse_count)
1.71 schwarze 510: mparse_buf_r(curp, ln, of, 0);
1.1 schwarze 511: else
512: mandoc_msg(MANDOCERR_ROFFLOOP, curp,
1.25 schwarze 513: curp->line, pos, NULL);
1.1 schwarze 514: pos = 0;
515: continue;
1.25 schwarze 516: case ROFF_APPEND:
1.71 schwarze 517: pos = strlen(ln.buf);
1.1 schwarze 518: continue;
1.25 schwarze 519: case ROFF_RERUN:
1.1 schwarze 520: goto rerun;
1.25 schwarze 521: case ROFF_IGN:
1.1 schwarze 522: pos = 0;
523: continue;
1.25 schwarze 524: case ROFF_SO:
1.71 schwarze 525: if ( ! (curp->options & MPARSE_SO) &&
526: (i >= blk.sz || blk.buf[i] == '\0')) {
1.23 schwarze 527: curp->sodest = mandoc_strdup(ln.buf + of);
528: free(ln.buf);
529: return;
530: }
1.4 schwarze 531: /*
532: * We remove `so' clauses from our lookaside
533: * buffer because we're going to descend into
534: * the file recursively.
535: */
1.25 schwarze 536: if (curp->secondary)
1.4 schwarze 537: curp->secondary->sz -= pos + 1;
1.88 schwarze 538: save_file = curp->file;
1.122 schwarze 539: if ((fd = mparse_open(curp, ln.buf + of)) != -1) {
1.85 schwarze 540: mparse_readfd(curp, fd, ln.buf + of);
1.121 schwarze 541: close(fd);
1.88 schwarze 542: curp->file = save_file;
543: } else {
544: curp->file = save_file;
1.30 schwarze 545: mandoc_vmsg(MANDOCERR_SO_FAIL,
546: curp, curp->line, pos,
547: ".so %s", ln.buf + of);
1.85 schwarze 548: ln.sz = mandoc_asprintf(&cp,
549: ".sp\nSee the file %s.\n.sp",
550: ln.buf + of);
551: free(ln.buf);
552: ln.buf = cp;
553: of = 0;
554: mparse_buf_r(curp, ln, of, 0);
1.30 schwarze 555: }
1.1 schwarze 556: pos = 0;
557: continue;
558: default:
559: break;
560: }
561:
1.124 schwarze 562: if (curp->man->macroset == MACROSET_NONE)
1.60 schwarze 563: choose_parser(curp);
1.1 schwarze 564:
1.25 schwarze 565: /*
1.60 schwarze 566: * Lastly, push down into the parsers themselves.
1.1 schwarze 567: * If libroff returns ROFF_TBL, then add it to the
568: * currently open parse. Since we only get here if
569: * there does exist data (see tbl_data.c), we're
570: * guaranteed that something's been allocated.
571: * Do the same for ROFF_EQN.
572: */
573:
1.114 schwarze 574: if (rr == ROFF_TBL)
1.76 schwarze 575: while ((span = roff_span(curp->roff)) != NULL)
1.114 schwarze 576: roff_addtbl(curp->man, span);
577: else if (rr == ROFF_EQN)
578: roff_addeqn(curp->man, roff_eqn(curp->roff));
579: else if ((curp->man->macroset == MACROSET_MDOC ?
1.110 schwarze 580: mdoc_parseln(curp->man, curp->line, ln.buf, of) :
1.76 schwarze 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;
1.115 schwarze 602: gzFile gz;
1.1 schwarze 603: size_t off;
604: ssize_t ssz;
605:
1.118 schwarze 606: if (fstat(fd, &st) == -1)
607: err((int)MANDOCLEVEL_SYSERR, "%s", file);
1.1 schwarze 608:
609: /*
610: * If we're a regular file, try just reading in the whole entry
611: * via mmap(). This is faster than reading it into blocks, and
612: * since each file is only a few bytes to begin with, I'm not
613: * concerned that this is going to tank any machines.
614: */
615:
1.115 schwarze 616: if (curp->gzip == 0 && S_ISREG(st.st_mode)) {
1.106 schwarze 617: if (st.st_size > 0x7fffffff) {
1.86 schwarze 618: mandoc_msg(MANDOCERR_TOOLARGE, curp, 0, 0, NULL);
1.117 schwarze 619: return 0;
1.1 schwarze 620: }
621: *with_mmap = 1;
622: fb->sz = (size_t)st.st_size;
1.15 schwarze 623: fb->buf = mmap(NULL, fb->sz, PROT_READ, MAP_SHARED, fd, 0);
1.1 schwarze 624: if (fb->buf != MAP_FAILED)
1.117 schwarze 625: return 1;
1.1 schwarze 626: }
627:
1.115 schwarze 628: if (curp->gzip) {
1.118 schwarze 629: if ((gz = gzdopen(fd, "rb")) == NULL)
630: err((int)MANDOCLEVEL_SYSERR, "%s", file);
1.115 schwarze 631: } else
632: gz = NULL;
633:
1.1 schwarze 634: /*
635: * If this isn't a regular file (like, say, stdin), then we must
636: * go the old way and just read things in bit by bit.
637: */
638:
639: *with_mmap = 0;
640: off = 0;
641: fb->sz = 0;
642: fb->buf = NULL;
643: for (;;) {
644: if (off == fb->sz) {
645: if (fb->sz == (1U << 31)) {
1.86 schwarze 646: mandoc_msg(MANDOCERR_TOOLARGE, curp,
647: 0, 0, NULL);
1.1 schwarze 648: break;
649: }
650: resize_buf(fb, 65536);
651: }
1.115 schwarze 652: ssz = curp->gzip ?
653: gzread(gz, fb->buf + (int)off, fb->sz - off) :
654: read(fd, fb->buf + (int)off, fb->sz - off);
1.1 schwarze 655: if (ssz == 0) {
656: fb->sz = off;
1.117 schwarze 657: return 1;
1.1 schwarze 658: }
1.118 schwarze 659: if (ssz == -1)
660: err((int)MANDOCLEVEL_SYSERR, "%s", file);
1.1 schwarze 661: off += (size_t)ssz;
662: }
663:
664: free(fb->buf);
665: fb->buf = NULL;
1.117 schwarze 666: return 0;
1.1 schwarze 667: }
668:
669: static void
670: mparse_end(struct mparse *curp)
671: {
1.110 schwarze 672: if (curp->man->macroset == MACROSET_NONE)
673: curp->man->macroset = MACROSET_MAN;
674: if (curp->man->macroset == MACROSET_MDOC)
675: mdoc_endparse(curp->man);
676: else
1.86 schwarze 677: man_endparse(curp->man);
1.1 schwarze 678: roff_endparse(curp->roff);
679: }
680:
1.15 schwarze 681: static void
682: mparse_parse_buffer(struct mparse *curp, struct buf blk, const char *file)
1.1 schwarze 683: {
1.61 schwarze 684: struct buf *svprimary;
1.1 schwarze 685: const char *svfile;
1.71 schwarze 686: size_t offset;
1.14 schwarze 687: static int recursion_depth;
688:
689: if (64 < recursion_depth) {
690: mandoc_msg(MANDOCERR_ROFFLOOP, curp, curp->line, 0, NULL);
1.15 schwarze 691: return;
1.14 schwarze 692: }
1.1 schwarze 693:
1.15 schwarze 694: /* Line number is per-file. */
695: svfile = curp->file;
696: curp->file = file;
1.61 schwarze 697: svprimary = curp->primary;
1.59 schwarze 698: curp->primary = &blk;
1.15 schwarze 699: curp->line = 1;
700: recursion_depth++;
701:
1.70 schwarze 702: /* Skip an UTF-8 byte order mark. */
703: if (curp->filenc & MPARSE_UTF8 && blk.sz > 2 &&
704: (unsigned char)blk.buf[0] == 0xef &&
705: (unsigned char)blk.buf[1] == 0xbb &&
706: (unsigned char)blk.buf[2] == 0xbf) {
1.71 schwarze 707: offset = 3;
1.70 schwarze 708: curp->filenc &= ~MPARSE_LATIN1;
1.71 schwarze 709: } else
710: offset = 0;
1.70 schwarze 711:
1.71 schwarze 712: mparse_buf_r(curp, blk, offset, 1);
1.15 schwarze 713:
1.86 schwarze 714: if (--recursion_depth == 0)
1.15 schwarze 715: mparse_end(curp);
716:
1.61 schwarze 717: curp->primary = svprimary;
1.15 schwarze 718: curp->file = svfile;
719: }
720:
1.74 schwarze 721: /*
722: * Read the whole file into memory and call the parsers.
723: * Called recursively when an .so request is encountered.
724: */
1.15 schwarze 725: enum mandoclevel
726: mparse_readfd(struct mparse *curp, int fd, const char *file)
727: {
728: struct buf blk;
729: int with_mmap;
1.70 schwarze 730: int save_filenc;
1.1 schwarze 731:
1.67 schwarze 732: if (read_whole_file(curp, file, fd, &blk, &with_mmap)) {
1.70 schwarze 733: save_filenc = curp->filenc;
734: curp->filenc = curp->options &
735: (MPARSE_UTF8 | MPARSE_LATIN1);
1.67 schwarze 736: mparse_parse_buffer(curp, blk, file);
1.70 schwarze 737: curp->filenc = save_filenc;
1.67 schwarze 738: if (with_mmap)
739: munmap(blk.buf, blk.sz);
740: else
741: free(blk.buf);
742: }
1.117 schwarze 743: return curp->file_status;
1.58 schwarze 744: }
745:
1.122 schwarze 746: int
747: mparse_open(struct mparse *curp, const char *file)
1.58 schwarze 748: {
749: char *cp;
1.122 schwarze 750: int fd;
1.58 schwarze 751:
752: curp->file = file;
1.115 schwarze 753: cp = strrchr(file, '.');
754: curp->gzip = (cp != NULL && ! strcmp(cp + 1, "gz"));
1.74 schwarze 755:
1.115 schwarze 756: /* First try to use the filename as it is. */
1.74 schwarze 757:
1.122 schwarze 758: if ((fd = open(file, O_RDONLY)) != -1)
759: return fd;
1.74 schwarze 760:
1.115 schwarze 761: /*
762: * If that doesn't work and the filename doesn't
763: * already end in .gz, try appending .gz.
764: */
1.74 schwarze 765:
1.115 schwarze 766: if ( ! curp->gzip) {
1.74 schwarze 767: mandoc_asprintf(&cp, "%s.gz", file);
1.123 schwarze 768: fd = open(cp, O_RDONLY);
1.83 schwarze 769: free(cp);
1.122 schwarze 770: if (fd != -1) {
1.115 schwarze 771: curp->gzip = 1;
1.122 schwarze 772: return fd;
1.58 schwarze 773: }
774: }
775:
1.115 schwarze 776: /* Neither worked, give up. */
1.73 schwarze 777:
1.115 schwarze 778: mandoc_msg(MANDOCERR_FILE, curp, 0, 0, strerror(errno));
1.122 schwarze 779: return -1;
1.1 schwarze 780: }
781:
782: struct mparse *
1.69 schwarze 783: mparse_alloc(int options, enum mandoclevel wlevel, mandocmsg mmsg,
1.119 schwarze 784: const char *defos)
1.1 schwarze 785: {
786: struct mparse *curp;
787:
788: curp = mandoc_calloc(1, sizeof(struct mparse));
789:
1.22 schwarze 790: curp->options = options;
1.1 schwarze 791: curp->wlevel = wlevel;
792: curp->mmsg = mmsg;
1.7 schwarze 793: curp->defos = defos;
1.1 schwarze 794:
1.119 schwarze 795: curp->roff = roff_alloc(curp, options);
1.112 schwarze 796: curp->man = roff_man_alloc( curp->roff, curp, curp->defos,
797: curp->options & MPARSE_QUICK ? 1 : 0);
1.111 schwarze 798: if (curp->options & MPARSE_MDOC) {
799: mdoc_hash_init();
1.112 schwarze 800: curp->man->macroset = MACROSET_MDOC;
801: } else if (curp->options & MPARSE_MAN) {
1.111 schwarze 802: man_hash_init();
1.112 schwarze 803: curp->man->macroset = MACROSET_MAN;
1.111 schwarze 804: }
1.113 schwarze 805: curp->man->first->tok = TOKEN_NONE;
1.117 schwarze 806: return curp;
1.1 schwarze 807: }
808:
809: void
810: mparse_reset(struct mparse *curp)
811: {
812: roff_reset(curp->roff);
1.124 schwarze 813: roff_man_reset(curp->man);
1.133 ! schwarze 814:
! 815: free(curp->sodest);
! 816: curp->sodest = NULL;
! 817:
1.4 schwarze 818: if (curp->secondary)
819: curp->secondary->sz = 0;
1.1 schwarze 820:
821: curp->file_status = MANDOCLEVEL_OK;
1.132 schwarze 822: curp->gzip = 0;
1.1 schwarze 823: }
824:
825: void
826: mparse_free(struct mparse *curp)
827: {
828:
1.112 schwarze 829: roff_man_free(curp->man);
1.133 ! schwarze 830: roff_free(curp->roff);
1.4 schwarze 831: if (curp->secondary)
832: free(curp->secondary->buf);
1.1 schwarze 833:
1.4 schwarze 834: free(curp->secondary);
1.23 schwarze 835: free(curp->sodest);
1.1 schwarze 836: free(curp);
837: }
838:
839: void
1.110 schwarze 840: mparse_result(struct mparse *curp, struct roff_man **man,
841: char **sodest)
1.1 schwarze 842: {
843:
1.23 schwarze 844: if (sodest && NULL != (*sodest = curp->sodest)) {
845: *man = NULL;
846: return;
847: }
1.1 schwarze 848: if (man)
849: *man = curp->man;
1.130 schwarze 850: }
851:
852: void
853: mparse_updaterc(struct mparse *curp, enum mandoclevel *rc)
854: {
855: if (curp->file_status > *rc)
856: *rc = curp->file_status;
1.1 schwarze 857: }
858:
859: void
860: mandoc_vmsg(enum mandocerr t, struct mparse *m,
861: int ln, int pos, const char *fmt, ...)
862: {
863: char buf[256];
864: va_list ap;
865:
866: va_start(ap, fmt);
1.26 schwarze 867: (void)vsnprintf(buf, sizeof(buf), fmt, ap);
1.1 schwarze 868: va_end(ap);
869:
870: mandoc_msg(t, m, ln, pos, buf);
871: }
872:
873: void
1.25 schwarze 874: mandoc_msg(enum mandocerr er, struct mparse *m,
1.1 schwarze 875: int ln, int col, const char *msg)
876: {
877: enum mandoclevel level;
878:
1.87 schwarze 879: level = MANDOCLEVEL_UNSUPP;
1.1 schwarze 880: while (er < mandoclimits[level])
881: level--;
882:
1.83 schwarze 883: if (level < m->wlevel && er != MANDOCERR_FILE)
1.1 schwarze 884: return;
885:
886: if (m->mmsg)
887: (*m->mmsg)(er, level, m->file, ln, col, msg);
888:
889: if (m->file_status < level)
890: m->file_status = level;
891: }
892:
893: const char *
894: mparse_strerror(enum mandocerr er)
895: {
896:
1.117 schwarze 897: return mandocerrs[er];
1.1 schwarze 898: }
899:
900: const char *
901: mparse_strlevel(enum mandoclevel lvl)
902: {
1.117 schwarze 903: return mandoclevels[lvl];
1.4 schwarze 904: }
905:
906: void
907: mparse_keep(struct mparse *p)
908: {
909:
910: assert(NULL == p->secondary);
911: p->secondary = mandoc_calloc(1, sizeof(struct buf));
912: }
913:
914: const char *
915: mparse_getkeep(const struct mparse *p)
916: {
917:
918: assert(p->secondary);
1.117 schwarze 919: return p->secondary->sz ? p->secondary->buf : NULL;
1.1 schwarze 920: }