Annotation of src/usr.bin/mandoc/read.c, Revision 1.128
1.128 ! schwarze 1: /* $OpenBSD: read.c,v 1.127 2016/12/07 22:57:35 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",
1.128 ! schwarze 148: "missing section argument",
1.38 schwarze 149: "missing -std argument, adding it",
1.100 schwarze 150: "missing option string, using \"\"",
151: "missing resource identifier, using \"\"",
1.66 schwarze 152: "missing eqn box, using \"\"",
1.1 schwarze 153:
154: /* related to bad macro arguments */
1.42 schwarze 155: "unterminated quoted argument",
1.1 schwarze 156: "duplicate argument",
1.54 schwarze 157: "skipping duplicate argument",
1.41 schwarze 158: "skipping duplicate display type",
159: "skipping duplicate list type",
1.54 schwarze 160: "skipping -width argument",
1.101 schwarze 161: "wrong number of cells",
1.1 schwarze 162: "unknown AT&T UNIX version",
1.64 schwarze 163: "comma in function argument",
1.65 schwarze 164: "parenthesis in function name",
1.45 schwarze 165: "invalid content in Rs block",
1.41 schwarze 166: "invalid Boolean argument",
167: "unknown font, skipping request",
1.101 schwarze 168: "odd number of characters in request",
1.1 schwarze 169:
170: /* related to plain text */
1.42 schwarze 171: "blank line in fill mode, using .sp",
172: "tab in filled text",
173: "whitespace at end of input line",
1.1 schwarze 174: "bad comment style",
1.42 schwarze 175: "invalid escape sequence",
176: "undefined string, using \"\"",
1.3 schwarze 177:
1.93 schwarze 178: /* related to tables */
179: "tbl line starts with span",
180: "tbl column starts with span",
181: "skipping vertical bar in tbl layout",
182:
1.1 schwarze 183: "generic error",
1.3 schwarze 184:
1.1 schwarze 185: /* related to tables */
1.91 schwarze 186: "non-alphabetic character in tbl options",
187: "skipping unknown tbl option",
188: "missing tbl option argument",
189: "wrong tbl option argument size",
1.93 schwarze 190: "empty tbl layout",
191: "invalid character in tbl layout",
192: "unmatched parenthesis in tbl layout",
1.94 schwarze 193: "tbl without any data cells",
194: "ignoring data in spanned tbl cell",
195: "ignoring extra tbl data cells",
196: "data block open at end of tbl",
1.1 schwarze 197:
1.46 schwarze 198: /* related to document structure and macros */
1.83 schwarze 199: NULL,
1.1 schwarze 200: "input stack limit exceeded, infinite loop?",
201: "skipping bad character",
1.46 schwarze 202: "skipping unknown macro",
1.87 schwarze 203: "skipping insecure request",
1.51 schwarze 204: "skipping item outside list",
1.46 schwarze 205: "skipping column outside column list",
206: "skipping end of block that is not open",
1.90 schwarze 207: "fewer RS blocks open, skipping",
1.46 schwarze 208: "inserting missing end of block",
209: "appending missing end of block",
210:
211: /* related to request and macro arguments */
1.1 schwarze 212: "escaped character not allowed in a name",
1.75 schwarze 213: "NOT IMPLEMENTED: Bd -file",
1.120 schwarze 214: "skipping display without arguments",
1.49 schwarze 215: "missing list type, using -item",
1.48 schwarze 216: "missing manual name, using \"\"",
1.49 schwarze 217: "uname(3) system call failed, using UNKNOWN",
1.41 schwarze 218: "unknown standard specifier",
1.49 schwarze 219: "skipping request without numeric argument",
1.84 schwarze 220: "NOT IMPLEMENTED: .so with absolute path or \"..\"",
1.85 schwarze 221: ".so request failed",
1.39 schwarze 222: "skipping all arguments",
223: "skipping excess arguments",
1.68 schwarze 224: "divide by zero",
1.87 schwarze 225:
226: "unsupported feature",
227: "input too large",
1.89 schwarze 228: "unsupported control character",
1.87 schwarze 229: "unsupported roff request",
1.94 schwarze 230: "eqn delim option in tbl",
1.93 schwarze 231: "unsupported tbl layout modifier",
1.87 schwarze 232: "ignoring macro in table",
1.1 schwarze 233: };
234:
235: static const char * const mandoclevels[MANDOCLEVEL_MAX] = {
236: "SUCCESS",
237: "RESERVED",
238: "WARNING",
239: "ERROR",
1.87 schwarze 240: "UNSUPP",
1.1 schwarze 241: "BADARG",
242: "SYSERR"
243: };
244:
1.25 schwarze 245:
1.1 schwarze 246: static void
247: resize_buf(struct buf *buf, size_t initial)
248: {
249:
250: buf->sz = buf->sz > initial/2 ? 2 * buf->sz : initial;
251: buf->buf = mandoc_realloc(buf->buf, buf->sz);
252: }
253:
254: static void
1.60 schwarze 255: choose_parser(struct mparse *curp)
1.1 schwarze 256: {
1.59 schwarze 257: char *cp, *ep;
258: int format;
1.1 schwarze 259:
1.59 schwarze 260: /*
261: * If neither command line arguments -mdoc or -man select
262: * a parser nor the roff parser found a .Dd or .TH macro
263: * yet, look ahead in the main input buffer.
264: */
265:
266: if ((format = roff_getformat(curp->roff)) == 0) {
267: cp = curp->primary->buf;
268: ep = cp + curp->primary->sz;
269: while (cp < ep) {
1.61 schwarze 270: if (*cp == '.' || *cp == '\'') {
1.59 schwarze 271: cp++;
272: if (cp[0] == 'D' && cp[1] == 'd') {
273: format = MPARSE_MDOC;
274: break;
275: }
276: if (cp[0] == 'T' && cp[1] == 'H') {
277: format = MPARSE_MAN;
278: break;
279: }
280: }
281: cp = memchr(cp, '\n', ep - cp);
282: if (cp == NULL)
283: break;
284: cp++;
285: }
1.1 schwarze 286: }
287:
1.59 schwarze 288: if (format == MPARSE_MDOC) {
1.111 schwarze 289: mdoc_hash_init();
1.112 schwarze 290: curp->man->macroset = MACROSET_MDOC;
1.113 schwarze 291: curp->man->first->tok = TOKEN_NONE;
1.112 schwarze 292: } else {
293: man_hash_init();
294: curp->man->macroset = MACROSET_MAN;
1.113 schwarze 295: curp->man->first->tok = TOKEN_NONE;
1.25 schwarze 296: }
1.1 schwarze 297: }
298:
299: /*
1.71 schwarze 300: * Main parse routine for a buffer.
301: * It assumes encoding and line numbering are already set up.
302: * It can recurse directly (for invocations of user-defined
303: * macros, inline equations, and input line traps)
304: * and indirectly (for .so file inclusion).
1.1 schwarze 305: */
306: static void
1.71 schwarze 307: mparse_buf_r(struct mparse *curp, struct buf blk, size_t i, int start)
1.1 schwarze 308: {
309: const struct tbl_span *span;
310: struct buf ln;
1.88 schwarze 311: const char *save_file;
1.85 schwarze 312: char *cp;
1.71 schwarze 313: size_t pos; /* byte number in the ln buffer */
1.127 schwarze 314: size_t j; /* auxiliary byte number in the blk buffer */
1.1 schwarze 315: enum rofferr rr;
1.76 schwarze 316: int of;
1.1 schwarze 317: int lnn; /* line number in the real file */
1.85 schwarze 318: int fd;
1.1 schwarze 319: unsigned char c;
320:
1.71 schwarze 321: memset(&ln, 0, sizeof(ln));
1.1 schwarze 322:
1.25 schwarze 323: lnn = curp->line;
324: pos = 0;
1.1 schwarze 325:
1.71 schwarze 326: while (i < blk.sz) {
1.1 schwarze 327: if (0 == pos && '\0' == blk.buf[i])
328: break;
329:
330: if (start) {
331: curp->line = lnn;
332: curp->reparse_count = 0;
1.70 schwarze 333:
334: if (lnn < 3 &&
335: curp->filenc & MPARSE_UTF8 &&
1.71 schwarze 336: curp->filenc & MPARSE_LATIN1)
337: curp->filenc = preconv_cue(&blk, i);
1.1 schwarze 338: }
339:
1.71 schwarze 340: while (i < blk.sz && (start || blk.buf[i] != '\0')) {
1.1 schwarze 341:
342: /*
343: * When finding an unescaped newline character,
344: * leave the character loop to process the line.
345: * Skip a preceding carriage return, if any.
346: */
347:
1.71 schwarze 348: if ('\r' == blk.buf[i] && i + 1 < blk.sz &&
1.1 schwarze 349: '\n' == blk.buf[i + 1])
350: ++i;
351: if ('\n' == blk.buf[i]) {
352: ++i;
353: ++lnn;
354: break;
355: }
356:
1.13 schwarze 357: /*
1.70 schwarze 358: * Make sure we have space for the worst
359: * case of 11 bytes: "\\[u10ffff]\0"
1.13 schwarze 360: */
361:
1.71 schwarze 362: if (pos + 11 > ln.sz)
1.13 schwarze 363: resize_buf(&ln, 256);
364:
1.25 schwarze 365: /*
1.70 schwarze 366: * Encode 8-bit input.
1.1 schwarze 367: */
368:
1.70 schwarze 369: c = blk.buf[i];
370: if (c & 0x80) {
1.71 schwarze 371: if ( ! (curp->filenc && preconv_encode(
372: &blk, &i, &ln, &pos, &curp->filenc))) {
1.89 schwarze 373: mandoc_vmsg(MANDOCERR_CHAR_BAD, curp,
374: curp->line, pos, "0x%x", c);
1.70 schwarze 375: ln.buf[pos++] = '?';
376: i++;
377: }
378: continue;
379: }
380:
381: /*
382: * Exclude control characters.
383: */
1.1 schwarze 384:
1.70 schwarze 385: if (c == 0x7f || (c < 0x20 && c != 0x09)) {
1.89 schwarze 386: mandoc_vmsg(c == 0x00 || c == 0x04 ||
387: c > 0x0a ? MANDOCERR_CHAR_BAD :
388: MANDOCERR_CHAR_UNSUPP,
389: curp, curp->line, pos, "0x%x", c);
1.1 schwarze 390: i++;
1.102 schwarze 391: if (c != '\r')
392: ln.buf[pos++] = '?';
1.1 schwarze 393: continue;
394: }
395:
396: /* Trailing backslash = a plain char. */
397:
1.71 schwarze 398: if (blk.buf[i] != '\\' || i + 1 == blk.sz) {
1.1 schwarze 399: ln.buf[pos++] = blk.buf[i++];
400: continue;
401: }
402:
403: /*
404: * Found escape and at least one other character.
405: * When it's a newline character, skip it.
406: * When there is a carriage return in between,
407: * skip that one as well.
408: */
409:
1.71 schwarze 410: if ('\r' == blk.buf[i + 1] && i + 2 < blk.sz &&
1.1 schwarze 411: '\n' == blk.buf[i + 2])
412: ++i;
413: if ('\n' == blk.buf[i + 1]) {
414: i += 2;
415: ++lnn;
416: continue;
417: }
418:
419: if ('"' == blk.buf[i + 1] || '#' == blk.buf[i + 1]) {
1.127 schwarze 420: j = i;
1.1 schwarze 421: i += 2;
422: /* Comment, skip to end of line */
1.71 schwarze 423: for (; i < blk.sz; ++i) {
1.126 schwarze 424: if (blk.buf[i] != '\n')
425: continue;
426: if (blk.buf[i - 1] == ' ' ||
427: blk.buf[i - 1] == '\t')
428: mandoc_msg(
429: MANDOCERR_SPACE_EOL,
430: curp, curp->line,
1.127 schwarze 431: pos + i-1 - j, NULL);
1.126 schwarze 432: ++i;
433: ++lnn;
434: break;
1.1 schwarze 435: }
436:
437: /* Backout trailing whitespaces */
438: for (; pos > 0; --pos) {
439: if (ln.buf[pos - 1] != ' ')
440: break;
441: if (pos > 2 && ln.buf[pos - 2] == '\\')
442: break;
443: }
444: break;
445: }
446:
1.13 schwarze 447: /* Catch escaped bogus characters. */
448:
449: c = (unsigned char) blk.buf[i+1];
450:
1.25 schwarze 451: if ( ! (isascii(c) &&
452: (isgraph(c) || isblank(c)))) {
1.89 schwarze 453: mandoc_vmsg(MANDOCERR_CHAR_BAD, curp,
1.56 schwarze 454: curp->line, pos, "0x%x", c);
1.13 schwarze 455: i += 2;
456: ln.buf[pos++] = '?';
457: continue;
458: }
459:
1.1 schwarze 460: /* Some other escape sequence, copy & cont. */
461:
462: ln.buf[pos++] = blk.buf[i++];
463: ln.buf[pos++] = blk.buf[i++];
464: }
465:
1.71 schwarze 466: if (pos >= ln.sz)
1.1 schwarze 467: resize_buf(&ln, 256);
468:
469: ln.buf[pos] = '\0';
470:
471: /*
472: * A significant amount of complexity is contained by
473: * the roff preprocessor. It's line-oriented but can be
474: * expressed on one line, so we need at times to
475: * readjust our starting point and re-run it. The roff
476: * preprocessor can also readjust the buffers with new
477: * data, so we pass them in wholesale.
478: */
479:
480: of = 0;
481:
1.4 schwarze 482: /*
483: * Maintain a lookaside buffer of all parsed lines. We
484: * only do this if mparse_keep() has been invoked (the
485: * buffer may be accessed with mparse_getkeep()).
486: */
487:
488: if (curp->secondary) {
1.25 schwarze 489: curp->secondary->buf = mandoc_realloc(
490: curp->secondary->buf,
491: curp->secondary->sz + pos + 2);
492: memcpy(curp->secondary->buf +
493: curp->secondary->sz,
494: ln.buf, pos);
1.4 schwarze 495: curp->secondary->sz += pos;
496: curp->secondary->buf
497: [curp->secondary->sz] = '\n';
498: curp->secondary->sz++;
499: curp->secondary->buf
500: [curp->secondary->sz] = '\0';
501: }
1.1 schwarze 502: rerun:
1.72 schwarze 503: rr = roff_parseln(curp->roff, curp->line, &ln, &of);
1.1 schwarze 504:
505: switch (rr) {
1.25 schwarze 506: case ROFF_REPARSE:
1.1 schwarze 507: if (REPARSE_LIMIT >= ++curp->reparse_count)
1.71 schwarze 508: mparse_buf_r(curp, ln, of, 0);
1.1 schwarze 509: else
510: mandoc_msg(MANDOCERR_ROFFLOOP, curp,
1.25 schwarze 511: curp->line, pos, NULL);
1.1 schwarze 512: pos = 0;
513: continue;
1.25 schwarze 514: case ROFF_APPEND:
1.71 schwarze 515: pos = strlen(ln.buf);
1.1 schwarze 516: continue;
1.25 schwarze 517: case ROFF_RERUN:
1.1 schwarze 518: goto rerun;
1.25 schwarze 519: case ROFF_IGN:
1.1 schwarze 520: pos = 0;
521: continue;
1.25 schwarze 522: case ROFF_SO:
1.71 schwarze 523: if ( ! (curp->options & MPARSE_SO) &&
524: (i >= blk.sz || blk.buf[i] == '\0')) {
1.23 schwarze 525: curp->sodest = mandoc_strdup(ln.buf + of);
526: free(ln.buf);
527: return;
528: }
1.4 schwarze 529: /*
530: * We remove `so' clauses from our lookaside
531: * buffer because we're going to descend into
532: * the file recursively.
533: */
1.25 schwarze 534: if (curp->secondary)
1.4 schwarze 535: curp->secondary->sz -= pos + 1;
1.88 schwarze 536: save_file = curp->file;
1.122 schwarze 537: if ((fd = mparse_open(curp, ln.buf + of)) != -1) {
1.85 schwarze 538: mparse_readfd(curp, fd, ln.buf + of);
1.121 schwarze 539: close(fd);
1.88 schwarze 540: curp->file = save_file;
541: } else {
542: curp->file = save_file;
1.30 schwarze 543: mandoc_vmsg(MANDOCERR_SO_FAIL,
544: curp, curp->line, pos,
545: ".so %s", ln.buf + of);
1.85 schwarze 546: ln.sz = mandoc_asprintf(&cp,
547: ".sp\nSee the file %s.\n.sp",
548: ln.buf + of);
549: free(ln.buf);
550: ln.buf = cp;
551: of = 0;
552: mparse_buf_r(curp, ln, of, 0);
1.30 schwarze 553: }
1.1 schwarze 554: pos = 0;
555: continue;
556: default:
557: break;
558: }
559:
1.124 schwarze 560: if (curp->man->macroset == MACROSET_NONE)
1.60 schwarze 561: choose_parser(curp);
1.1 schwarze 562:
1.25 schwarze 563: /*
1.60 schwarze 564: * Lastly, push down into the parsers themselves.
1.1 schwarze 565: * If libroff returns ROFF_TBL, then add it to the
566: * currently open parse. Since we only get here if
567: * there does exist data (see tbl_data.c), we're
568: * guaranteed that something's been allocated.
569: * Do the same for ROFF_EQN.
570: */
571:
1.114 schwarze 572: if (rr == ROFF_TBL)
1.76 schwarze 573: while ((span = roff_span(curp->roff)) != NULL)
1.114 schwarze 574: roff_addtbl(curp->man, span);
575: else if (rr == ROFF_EQN)
576: roff_addeqn(curp->man, roff_eqn(curp->roff));
577: else if ((curp->man->macroset == MACROSET_MDOC ?
1.110 schwarze 578: mdoc_parseln(curp->man, curp->line, ln.buf, of) :
1.76 schwarze 579: man_parseln(curp->man, curp->line, ln.buf, of)) == 2)
580: break;
1.1 schwarze 581:
582: /* Temporary buffers typically are not full. */
583:
584: if (0 == start && '\0' == blk.buf[i])
585: break;
586:
587: /* Start the next input line. */
588:
589: pos = 0;
590: }
591:
592: free(ln.buf);
593: }
594:
595: static int
1.18 schwarze 596: read_whole_file(struct mparse *curp, const char *file, int fd,
597: struct buf *fb, int *with_mmap)
1.1 schwarze 598: {
599: struct stat st;
1.115 schwarze 600: gzFile gz;
1.1 schwarze 601: size_t off;
602: ssize_t ssz;
603:
1.118 schwarze 604: if (fstat(fd, &st) == -1)
605: err((int)MANDOCLEVEL_SYSERR, "%s", file);
1.1 schwarze 606:
607: /*
608: * If we're a regular file, try just reading in the whole entry
609: * via mmap(). This is faster than reading it into blocks, and
610: * since each file is only a few bytes to begin with, I'm not
611: * concerned that this is going to tank any machines.
612: */
613:
1.115 schwarze 614: if (curp->gzip == 0 && S_ISREG(st.st_mode)) {
1.106 schwarze 615: if (st.st_size > 0x7fffffff) {
1.86 schwarze 616: mandoc_msg(MANDOCERR_TOOLARGE, curp, 0, 0, NULL);
1.117 schwarze 617: return 0;
1.1 schwarze 618: }
619: *with_mmap = 1;
620: fb->sz = (size_t)st.st_size;
1.15 schwarze 621: fb->buf = mmap(NULL, fb->sz, PROT_READ, MAP_SHARED, fd, 0);
1.1 schwarze 622: if (fb->buf != MAP_FAILED)
1.117 schwarze 623: return 1;
1.1 schwarze 624: }
625:
1.115 schwarze 626: if (curp->gzip) {
1.118 schwarze 627: if ((gz = gzdopen(fd, "rb")) == NULL)
628: err((int)MANDOCLEVEL_SYSERR, "%s", file);
1.115 schwarze 629: } else
630: gz = NULL;
631:
1.1 schwarze 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: }
1.115 schwarze 650: ssz = curp->gzip ?
651: gzread(gz, fb->buf + (int)off, fb->sz - off) :
652: read(fd, fb->buf + (int)off, fb->sz - off);
1.1 schwarze 653: if (ssz == 0) {
654: fb->sz = off;
1.117 schwarze 655: return 1;
1.1 schwarze 656: }
1.118 schwarze 657: if (ssz == -1)
658: err((int)MANDOCLEVEL_SYSERR, "%s", file);
1.1 schwarze 659: off += (size_t)ssz;
660: }
661:
662: free(fb->buf);
663: fb->buf = NULL;
1.117 schwarze 664: return 0;
1.1 schwarze 665: }
666:
667: static void
668: mparse_end(struct mparse *curp)
669: {
1.110 schwarze 670: if (curp->man->macroset == MACROSET_NONE)
671: curp->man->macroset = MACROSET_MAN;
672: if (curp->man->macroset == MACROSET_MDOC)
673: mdoc_endparse(curp->man);
674: else
1.86 schwarze 675: man_endparse(curp->man);
1.1 schwarze 676: roff_endparse(curp->roff);
677: }
678:
1.15 schwarze 679: static void
680: mparse_parse_buffer(struct mparse *curp, struct buf blk, const char *file)
1.1 schwarze 681: {
1.61 schwarze 682: struct buf *svprimary;
1.1 schwarze 683: const char *svfile;
1.71 schwarze 684: size_t offset;
1.14 schwarze 685: static int recursion_depth;
686:
687: if (64 < recursion_depth) {
688: mandoc_msg(MANDOCERR_ROFFLOOP, curp, curp->line, 0, NULL);
1.15 schwarze 689: return;
1.14 schwarze 690: }
1.1 schwarze 691:
1.15 schwarze 692: /* Line number is per-file. */
693: svfile = curp->file;
694: curp->file = file;
1.61 schwarze 695: svprimary = curp->primary;
1.59 schwarze 696: curp->primary = &blk;
1.15 schwarze 697: curp->line = 1;
698: recursion_depth++;
699:
1.70 schwarze 700: /* Skip an UTF-8 byte order mark. */
701: if (curp->filenc & MPARSE_UTF8 && blk.sz > 2 &&
702: (unsigned char)blk.buf[0] == 0xef &&
703: (unsigned char)blk.buf[1] == 0xbb &&
704: (unsigned char)blk.buf[2] == 0xbf) {
1.71 schwarze 705: offset = 3;
1.70 schwarze 706: curp->filenc &= ~MPARSE_LATIN1;
1.71 schwarze 707: } else
708: offset = 0;
1.70 schwarze 709:
1.71 schwarze 710: mparse_buf_r(curp, blk, offset, 1);
1.15 schwarze 711:
1.86 schwarze 712: if (--recursion_depth == 0)
1.15 schwarze 713: mparse_end(curp);
714:
1.61 schwarze 715: curp->primary = svprimary;
1.15 schwarze 716: curp->file = svfile;
717: }
718:
1.74 schwarze 719: /*
720: * Read the whole file into memory and call the parsers.
721: * Called recursively when an .so request is encountered.
722: */
1.15 schwarze 723: enum mandoclevel
724: mparse_readfd(struct mparse *curp, int fd, const char *file)
725: {
726: struct buf blk;
727: int with_mmap;
1.70 schwarze 728: int save_filenc;
1.1 schwarze 729:
1.67 schwarze 730: if (read_whole_file(curp, file, fd, &blk, &with_mmap)) {
1.70 schwarze 731: save_filenc = curp->filenc;
732: curp->filenc = curp->options &
733: (MPARSE_UTF8 | MPARSE_LATIN1);
1.67 schwarze 734: mparse_parse_buffer(curp, blk, file);
1.70 schwarze 735: curp->filenc = save_filenc;
1.67 schwarze 736: if (with_mmap)
737: munmap(blk.buf, blk.sz);
738: else
739: free(blk.buf);
740: }
1.117 schwarze 741: return curp->file_status;
1.58 schwarze 742: }
743:
1.122 schwarze 744: int
745: mparse_open(struct mparse *curp, const char *file)
1.58 schwarze 746: {
747: char *cp;
1.122 schwarze 748: int fd;
1.58 schwarze 749:
750: curp->file = file;
1.115 schwarze 751: cp = strrchr(file, '.');
752: curp->gzip = (cp != NULL && ! strcmp(cp + 1, "gz"));
1.74 schwarze 753:
1.115 schwarze 754: /* First try to use the filename as it is. */
1.74 schwarze 755:
1.122 schwarze 756: if ((fd = open(file, O_RDONLY)) != -1)
757: return fd;
1.74 schwarze 758:
1.115 schwarze 759: /*
760: * If that doesn't work and the filename doesn't
761: * already end in .gz, try appending .gz.
762: */
1.74 schwarze 763:
1.115 schwarze 764: if ( ! curp->gzip) {
1.74 schwarze 765: mandoc_asprintf(&cp, "%s.gz", file);
1.123 schwarze 766: fd = open(cp, O_RDONLY);
1.83 schwarze 767: free(cp);
1.122 schwarze 768: if (fd != -1) {
1.115 schwarze 769: curp->gzip = 1;
1.122 schwarze 770: return fd;
1.58 schwarze 771: }
772: }
773:
1.115 schwarze 774: /* Neither worked, give up. */
1.73 schwarze 775:
1.115 schwarze 776: mandoc_msg(MANDOCERR_FILE, curp, 0, 0, strerror(errno));
1.122 schwarze 777: return -1;
1.1 schwarze 778: }
779:
780: struct mparse *
1.69 schwarze 781: mparse_alloc(int options, enum mandoclevel wlevel, mandocmsg mmsg,
1.119 schwarze 782: const char *defos)
1.1 schwarze 783: {
784: struct mparse *curp;
785:
786: curp = mandoc_calloc(1, sizeof(struct mparse));
787:
1.22 schwarze 788: curp->options = options;
1.1 schwarze 789: curp->wlevel = wlevel;
790: curp->mmsg = mmsg;
1.7 schwarze 791: curp->defos = defos;
1.1 schwarze 792:
1.119 schwarze 793: curp->roff = roff_alloc(curp, options);
1.112 schwarze 794: curp->man = roff_man_alloc( curp->roff, curp, curp->defos,
795: curp->options & MPARSE_QUICK ? 1 : 0);
1.111 schwarze 796: if (curp->options & MPARSE_MDOC) {
797: mdoc_hash_init();
1.112 schwarze 798: curp->man->macroset = MACROSET_MDOC;
799: } else if (curp->options & MPARSE_MAN) {
1.111 schwarze 800: man_hash_init();
1.112 schwarze 801: curp->man->macroset = MACROSET_MAN;
1.111 schwarze 802: }
1.113 schwarze 803: curp->man->first->tok = TOKEN_NONE;
1.117 schwarze 804: return curp;
1.1 schwarze 805: }
806:
807: void
808: mparse_reset(struct mparse *curp)
809: {
810: roff_reset(curp->roff);
1.124 schwarze 811: roff_man_reset(curp->man);
1.4 schwarze 812: if (curp->secondary)
813: curp->secondary->sz = 0;
1.1 schwarze 814:
815: curp->file_status = MANDOCLEVEL_OK;
1.23 schwarze 816:
817: free(curp->sodest);
818: curp->sodest = NULL;
1.1 schwarze 819: }
820:
821: void
822: mparse_free(struct mparse *curp)
823: {
824:
1.112 schwarze 825: roff_man_free(curp->man);
1.1 schwarze 826: if (curp->roff)
827: roff_free(curp->roff);
1.4 schwarze 828: if (curp->secondary)
829: free(curp->secondary->buf);
1.1 schwarze 830:
1.4 schwarze 831: free(curp->secondary);
1.23 schwarze 832: free(curp->sodest);
1.1 schwarze 833: free(curp);
834: }
835:
836: void
1.110 schwarze 837: mparse_result(struct mparse *curp, struct roff_man **man,
838: char **sodest)
1.1 schwarze 839: {
840:
1.23 schwarze 841: if (sodest && NULL != (*sodest = curp->sodest)) {
842: *man = NULL;
843: return;
844: }
1.1 schwarze 845: if (man)
846: *man = curp->man;
847: }
848:
849: void
850: mandoc_vmsg(enum mandocerr t, struct mparse *m,
851: int ln, int pos, const char *fmt, ...)
852: {
853: char buf[256];
854: va_list ap;
855:
856: va_start(ap, fmt);
1.26 schwarze 857: (void)vsnprintf(buf, sizeof(buf), fmt, ap);
1.1 schwarze 858: va_end(ap);
859:
860: mandoc_msg(t, m, ln, pos, buf);
861: }
862:
863: void
1.25 schwarze 864: mandoc_msg(enum mandocerr er, struct mparse *m,
1.1 schwarze 865: int ln, int col, const char *msg)
866: {
867: enum mandoclevel level;
868:
1.87 schwarze 869: level = MANDOCLEVEL_UNSUPP;
1.1 schwarze 870: while (er < mandoclimits[level])
871: level--;
872:
1.83 schwarze 873: if (level < m->wlevel && er != MANDOCERR_FILE)
1.1 schwarze 874: return;
875:
876: if (m->mmsg)
877: (*m->mmsg)(er, level, m->file, ln, col, msg);
878:
879: if (m->file_status < level)
880: m->file_status = level;
881: }
882:
883: const char *
884: mparse_strerror(enum mandocerr er)
885: {
886:
1.117 schwarze 887: return mandocerrs[er];
1.1 schwarze 888: }
889:
890: const char *
891: mparse_strlevel(enum mandoclevel lvl)
892: {
1.117 schwarze 893: return mandoclevels[lvl];
1.4 schwarze 894: }
895:
896: void
897: mparse_keep(struct mparse *p)
898: {
899:
900: assert(NULL == p->secondary);
901: p->secondary = mandoc_calloc(1, sizeof(struct buf));
902: }
903:
904: const char *
905: mparse_getkeep(const struct mparse *p)
906: {
907:
908: assert(p->secondary);
1.117 schwarze 909: return p->secondary->sz ? p->secondary->buf : NULL;
1.1 schwarze 910: }