Annotation of src/usr.bin/mandoc/read.c, Revision 1.150
1.150 ! schwarze 1: /* $OpenBSD: read.c,v 1.149 2017/06/17 22:40:27 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);
1.135 schwarze 65: static int 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,
1.137 schwarze 74: MANDOCERR_STYLE,
1.1 schwarze 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:
1.137 schwarze 85: "generic style suggestion",
1.138 schwarze 86:
1.148 schwarze 87: "Mdocdate found",
88: "Mdocdate missing",
89: "legacy man(7) date format",
1.149 schwarze 90: "RCS id missing",
1.150 ! schwarze 91: "duplicate RCS id",
1.138 schwarze 92: "useless macro",
1.139 schwarze 93: "consider using OS macro",
1.144 schwarze 94: "errnos out of order",
95: "duplicate errno",
1.140 schwarze 96: "description line ends with a full stop",
1.146 schwarze 97: "no blank before trailing delimiter",
1.147 schwarze 98: "function name without markup",
1.138 schwarze 99:
1.1 schwarze 100: "generic warning",
101:
102: /* related to the prologue */
1.57 schwarze 103: "missing manual title, using UNTITLED",
104: "missing manual title, using \"\"",
1.32 schwarze 105: "lower case character in document title",
1.57 schwarze 106: "missing manual section, using \"\"",
1.1 schwarze 107: "unknown manual section",
1.32 schwarze 108: "missing date, using today's date",
1.1 schwarze 109: "cannot parse date, using it verbatim",
1.57 schwarze 110: "missing Os macro, using \"\"",
111: "duplicate prologue macro",
112: "late prologue macro",
113: "skipping late title macro",
1.1 schwarze 114: "prologue macros out of order",
115:
116: /* related to document structure */
117: ".so is fragile, better use ln(1)",
1.28 schwarze 118: "no document body",
1.32 schwarze 119: "content before first section header",
120: "first section is not \"NAME\"",
1.129 schwarze 121: "NAME section without Nm before Nd",
1.103 schwarze 122: "NAME section without description",
123: "description not at the end of NAME",
124: "bad NAME section content",
1.129 schwarze 125: "missing comma before name",
1.96 schwarze 126: "missing description line, using \"\"",
1.134 schwarze 127: "description line outside NAME section",
1.1 schwarze 128: "sections out of conventional order",
1.32 schwarze 129: "duplicate section title",
130: "unexpected section",
1.63 schwarze 131: "unusual Xr order",
132: "unusual Xr punctuation",
1.62 schwarze 133: "AUTHORS section without An macro",
1.1 schwarze 134:
135: /* related to macros and nesting */
1.33 schwarze 136: "obsolete macro",
1.78 schwarze 137: "macro neither callable nor escaped",
1.1 schwarze 138: "skipping paragraph macro",
1.10 schwarze 139: "moving paragraph macro out of list",
1.1 schwarze 140: "skipping no-space macro",
141: "blocks badly nested",
142: "nested displays are not portable",
1.35 schwarze 143: "moving content out of list",
1.56 schwarze 144: "fill mode already enabled, skipping",
145: "fill mode already disabled, skipping",
1.1 schwarze 146: "line scope broken",
1.141 schwarze 147: "skipping blank line in line scope",
1.1 schwarze 148:
149: /* related to missing macro arguments */
1.36 schwarze 150: "skipping empty request",
151: "conditional request controls empty scope",
1.1 schwarze 152: "skipping empty macro",
1.99 schwarze 153: "empty block",
1.40 schwarze 154: "empty argument, using 0n",
1.38 schwarze 155: "missing display type, using -ragged",
156: "list type is not the first argument",
1.125 schwarze 157: "missing -width in -tag list, using 6n",
1.56 schwarze 158: "missing utility name, using \"\"",
1.98 schwarze 159: "missing function name, using \"\"",
1.38 schwarze 160: "empty head in list item",
161: "empty list item",
1.39 schwarze 162: "missing font type, using \\fR",
163: "unknown font type, using \\fR",
1.79 schwarze 164: "nothing follows prefix",
1.97 schwarze 165: "empty reference block",
1.128 schwarze 166: "missing section argument",
1.38 schwarze 167: "missing -std argument, adding it",
1.100 schwarze 168: "missing option string, using \"\"",
169: "missing resource identifier, using \"\"",
1.66 schwarze 170: "missing eqn box, using \"\"",
1.1 schwarze 171:
172: /* related to bad macro arguments */
1.42 schwarze 173: "unterminated quoted argument",
1.1 schwarze 174: "duplicate argument",
1.54 schwarze 175: "skipping duplicate argument",
1.41 schwarze 176: "skipping duplicate display type",
177: "skipping duplicate list type",
1.54 schwarze 178: "skipping -width argument",
1.101 schwarze 179: "wrong number of cells",
1.1 schwarze 180: "unknown AT&T UNIX version",
1.64 schwarze 181: "comma in function argument",
1.65 schwarze 182: "parenthesis in function name",
1.145 schwarze 183: "unknown library name",
1.45 schwarze 184: "invalid content in Rs block",
1.41 schwarze 185: "invalid Boolean argument",
186: "unknown font, skipping request",
1.101 schwarze 187: "odd number of characters in request",
1.1 schwarze 188:
189: /* related to plain text */
1.42 schwarze 190: "blank line in fill mode, using .sp",
191: "tab in filled text",
192: "whitespace at end of input line",
1.131 schwarze 193: "new sentence, new line",
1.1 schwarze 194: "bad comment style",
1.42 schwarze 195: "invalid escape sequence",
196: "undefined string, using \"\"",
1.3 schwarze 197:
1.93 schwarze 198: /* related to tables */
199: "tbl line starts with span",
200: "tbl column starts with span",
201: "skipping vertical bar in tbl layout",
202:
1.1 schwarze 203: "generic error",
1.3 schwarze 204:
1.1 schwarze 205: /* related to tables */
1.91 schwarze 206: "non-alphabetic character in tbl options",
207: "skipping unknown tbl option",
208: "missing tbl option argument",
209: "wrong tbl option argument size",
1.93 schwarze 210: "empty tbl layout",
211: "invalid character in tbl layout",
212: "unmatched parenthesis in tbl layout",
1.94 schwarze 213: "tbl without any data cells",
214: "ignoring data in spanned tbl cell",
215: "ignoring extra tbl data cells",
216: "data block open at end of tbl",
1.1 schwarze 217:
1.46 schwarze 218: /* related to document structure and macros */
1.83 schwarze 219: NULL,
1.1 schwarze 220: "input stack limit exceeded, infinite loop?",
221: "skipping bad character",
1.46 schwarze 222: "skipping unknown macro",
1.87 schwarze 223: "skipping insecure request",
1.51 schwarze 224: "skipping item outside list",
1.46 schwarze 225: "skipping column outside column list",
226: "skipping end of block that is not open",
1.90 schwarze 227: "fewer RS blocks open, skipping",
1.46 schwarze 228: "inserting missing end of block",
229: "appending missing end of block",
230:
231: /* related to request and macro arguments */
1.1 schwarze 232: "escaped character not allowed in a name",
1.75 schwarze 233: "NOT IMPLEMENTED: Bd -file",
1.120 schwarze 234: "skipping display without arguments",
1.49 schwarze 235: "missing list type, using -item",
1.143 schwarze 236: "argument is not numeric, using 1",
1.48 schwarze 237: "missing manual name, using \"\"",
1.49 schwarze 238: "uname(3) system call failed, using UNKNOWN",
1.41 schwarze 239: "unknown standard specifier",
1.49 schwarze 240: "skipping request without numeric argument",
1.84 schwarze 241: "NOT IMPLEMENTED: .so with absolute path or \"..\"",
1.85 schwarze 242: ".so request failed",
1.39 schwarze 243: "skipping all arguments",
244: "skipping excess arguments",
1.68 schwarze 245: "divide by zero",
1.87 schwarze 246:
247: "unsupported feature",
248: "input too large",
1.89 schwarze 249: "unsupported control character",
1.87 schwarze 250: "unsupported roff request",
1.94 schwarze 251: "eqn delim option in tbl",
1.93 schwarze 252: "unsupported tbl layout modifier",
1.87 schwarze 253: "ignoring macro in table",
1.1 schwarze 254: };
255:
256: static const char * const mandoclevels[MANDOCLEVEL_MAX] = {
257: "SUCCESS",
1.137 schwarze 258: "STYLE",
1.1 schwarze 259: "WARNING",
260: "ERROR",
1.87 schwarze 261: "UNSUPP",
1.1 schwarze 262: "BADARG",
263: "SYSERR"
264: };
265:
1.25 schwarze 266:
1.1 schwarze 267: static void
268: resize_buf(struct buf *buf, size_t initial)
269: {
270:
271: buf->sz = buf->sz > initial/2 ? 2 * buf->sz : initial;
272: buf->buf = mandoc_realloc(buf->buf, buf->sz);
273: }
274:
275: static void
1.60 schwarze 276: choose_parser(struct mparse *curp)
1.1 schwarze 277: {
1.59 schwarze 278: char *cp, *ep;
279: int format;
1.1 schwarze 280:
1.59 schwarze 281: /*
282: * If neither command line arguments -mdoc or -man select
283: * a parser nor the roff parser found a .Dd or .TH macro
284: * yet, look ahead in the main input buffer.
285: */
286:
287: if ((format = roff_getformat(curp->roff)) == 0) {
288: cp = curp->primary->buf;
289: ep = cp + curp->primary->sz;
290: while (cp < ep) {
1.61 schwarze 291: if (*cp == '.' || *cp == '\'') {
1.59 schwarze 292: cp++;
293: if (cp[0] == 'D' && cp[1] == 'd') {
294: format = MPARSE_MDOC;
295: break;
296: }
297: if (cp[0] == 'T' && cp[1] == 'H') {
298: format = MPARSE_MAN;
299: break;
300: }
301: }
302: cp = memchr(cp, '\n', ep - cp);
303: if (cp == NULL)
304: break;
305: cp++;
306: }
1.1 schwarze 307: }
308:
1.59 schwarze 309: if (format == MPARSE_MDOC) {
1.112 schwarze 310: curp->man->macroset = MACROSET_MDOC;
1.136 schwarze 311: if (curp->man->mdocmac == NULL)
312: curp->man->mdocmac = roffhash_alloc(MDOC_Dd, MDOC_MAX);
1.112 schwarze 313: } else {
314: curp->man->macroset = MACROSET_MAN;
1.136 schwarze 315: if (curp->man->manmac == NULL)
316: curp->man->manmac = roffhash_alloc(MAN_TH, MAN_MAX);
1.25 schwarze 317: }
1.136 schwarze 318: curp->man->first->tok = TOKEN_NONE;
1.1 schwarze 319: }
320:
321: /*
1.71 schwarze 322: * Main parse routine for a buffer.
323: * It assumes encoding and line numbering are already set up.
324: * It can recurse directly (for invocations of user-defined
325: * macros, inline equations, and input line traps)
326: * and indirectly (for .so file inclusion).
1.1 schwarze 327: */
1.135 schwarze 328: static int
1.71 schwarze 329: mparse_buf_r(struct mparse *curp, struct buf blk, size_t i, int start)
1.1 schwarze 330: {
331: const struct tbl_span *span;
332: struct buf ln;
1.88 schwarze 333: const char *save_file;
1.85 schwarze 334: char *cp;
1.71 schwarze 335: size_t pos; /* byte number in the ln buffer */
1.1 schwarze 336: enum rofferr rr;
1.76 schwarze 337: int of;
1.1 schwarze 338: int lnn; /* line number in the real file */
1.85 schwarze 339: int fd;
1.1 schwarze 340: unsigned char c;
341:
1.71 schwarze 342: memset(&ln, 0, sizeof(ln));
1.1 schwarze 343:
1.25 schwarze 344: lnn = curp->line;
345: pos = 0;
1.1 schwarze 346:
1.71 schwarze 347: while (i < blk.sz) {
1.1 schwarze 348: if (0 == pos && '\0' == blk.buf[i])
349: break;
350:
351: if (start) {
352: curp->line = lnn;
353: curp->reparse_count = 0;
1.70 schwarze 354:
355: if (lnn < 3 &&
356: curp->filenc & MPARSE_UTF8 &&
1.71 schwarze 357: curp->filenc & MPARSE_LATIN1)
358: curp->filenc = preconv_cue(&blk, i);
1.1 schwarze 359: }
360:
1.71 schwarze 361: while (i < blk.sz && (start || blk.buf[i] != '\0')) {
1.1 schwarze 362:
363: /*
364: * When finding an unescaped newline character,
365: * leave the character loop to process the line.
366: * Skip a preceding carriage return, if any.
367: */
368:
1.71 schwarze 369: if ('\r' == blk.buf[i] && i + 1 < blk.sz &&
1.1 schwarze 370: '\n' == blk.buf[i + 1])
371: ++i;
372: if ('\n' == blk.buf[i]) {
373: ++i;
374: ++lnn;
375: break;
376: }
377:
1.13 schwarze 378: /*
1.70 schwarze 379: * Make sure we have space for the worst
380: * case of 11 bytes: "\\[u10ffff]\0"
1.13 schwarze 381: */
382:
1.71 schwarze 383: if (pos + 11 > ln.sz)
1.13 schwarze 384: resize_buf(&ln, 256);
385:
1.25 schwarze 386: /*
1.70 schwarze 387: * Encode 8-bit input.
1.1 schwarze 388: */
389:
1.70 schwarze 390: c = blk.buf[i];
391: if (c & 0x80) {
1.71 schwarze 392: if ( ! (curp->filenc && preconv_encode(
393: &blk, &i, &ln, &pos, &curp->filenc))) {
1.89 schwarze 394: mandoc_vmsg(MANDOCERR_CHAR_BAD, curp,
395: curp->line, pos, "0x%x", c);
1.70 schwarze 396: ln.buf[pos++] = '?';
397: i++;
398: }
399: continue;
400: }
401:
402: /*
403: * Exclude control characters.
404: */
1.1 schwarze 405:
1.70 schwarze 406: if (c == 0x7f || (c < 0x20 && c != 0x09)) {
1.89 schwarze 407: mandoc_vmsg(c == 0x00 || c == 0x04 ||
408: c > 0x0a ? MANDOCERR_CHAR_BAD :
409: MANDOCERR_CHAR_UNSUPP,
410: curp, curp->line, pos, "0x%x", c);
1.1 schwarze 411: i++;
1.102 schwarze 412: if (c != '\r')
413: ln.buf[pos++] = '?';
1.1 schwarze 414: continue;
415: }
416:
417: ln.buf[pos++] = blk.buf[i++];
418: }
419:
1.142 schwarze 420: if (pos + 1 >= ln.sz)
1.1 schwarze 421: resize_buf(&ln, 256);
422:
1.142 schwarze 423: if (i == blk.sz || blk.buf[i] == '\0')
424: ln.buf[pos++] = '\n';
1.1 schwarze 425: ln.buf[pos] = '\0';
426:
427: /*
428: * A significant amount of complexity is contained by
429: * the roff preprocessor. It's line-oriented but can be
430: * expressed on one line, so we need at times to
431: * readjust our starting point and re-run it. The roff
432: * preprocessor can also readjust the buffers with new
433: * data, so we pass them in wholesale.
434: */
435:
436: of = 0;
437:
1.4 schwarze 438: /*
439: * Maintain a lookaside buffer of all parsed lines. We
440: * only do this if mparse_keep() has been invoked (the
441: * buffer may be accessed with mparse_getkeep()).
442: */
443:
444: if (curp->secondary) {
1.25 schwarze 445: curp->secondary->buf = mandoc_realloc(
446: curp->secondary->buf,
447: curp->secondary->sz + pos + 2);
448: memcpy(curp->secondary->buf +
449: curp->secondary->sz,
450: ln.buf, pos);
1.4 schwarze 451: curp->secondary->sz += pos;
452: curp->secondary->buf
453: [curp->secondary->sz] = '\n';
454: curp->secondary->sz++;
455: curp->secondary->buf
456: [curp->secondary->sz] = '\0';
457: }
1.1 schwarze 458: rerun:
1.72 schwarze 459: rr = roff_parseln(curp->roff, curp->line, &ln, &of);
1.1 schwarze 460:
461: switch (rr) {
1.25 schwarze 462: case ROFF_REPARSE:
1.135 schwarze 463: if (++curp->reparse_count > REPARSE_LIMIT)
1.1 schwarze 464: mandoc_msg(MANDOCERR_ROFFLOOP, curp,
1.25 schwarze 465: curp->line, pos, NULL);
1.135 schwarze 466: else if (mparse_buf_r(curp, ln, of, 0) == 1 ||
467: start == 1) {
468: pos = 0;
469: continue;
470: }
471: free(ln.buf);
472: return 0;
1.25 schwarze 473: case ROFF_APPEND:
1.71 schwarze 474: pos = strlen(ln.buf);
1.1 schwarze 475: continue;
1.25 schwarze 476: case ROFF_RERUN:
1.1 schwarze 477: goto rerun;
1.25 schwarze 478: case ROFF_IGN:
1.1 schwarze 479: pos = 0;
480: continue;
1.25 schwarze 481: case ROFF_SO:
1.71 schwarze 482: if ( ! (curp->options & MPARSE_SO) &&
483: (i >= blk.sz || blk.buf[i] == '\0')) {
1.23 schwarze 484: curp->sodest = mandoc_strdup(ln.buf + of);
485: free(ln.buf);
1.135 schwarze 486: return 1;
1.23 schwarze 487: }
1.4 schwarze 488: /*
489: * We remove `so' clauses from our lookaside
490: * buffer because we're going to descend into
491: * the file recursively.
492: */
1.25 schwarze 493: if (curp->secondary)
1.4 schwarze 494: curp->secondary->sz -= pos + 1;
1.88 schwarze 495: save_file = curp->file;
1.122 schwarze 496: if ((fd = mparse_open(curp, ln.buf + of)) != -1) {
1.85 schwarze 497: mparse_readfd(curp, fd, ln.buf + of);
1.121 schwarze 498: close(fd);
1.88 schwarze 499: curp->file = save_file;
500: } else {
501: curp->file = save_file;
1.30 schwarze 502: mandoc_vmsg(MANDOCERR_SO_FAIL,
503: curp, curp->line, pos,
504: ".so %s", ln.buf + of);
1.85 schwarze 505: ln.sz = mandoc_asprintf(&cp,
506: ".sp\nSee the file %s.\n.sp",
507: ln.buf + of);
508: free(ln.buf);
509: ln.buf = cp;
510: of = 0;
511: mparse_buf_r(curp, ln, of, 0);
1.30 schwarze 512: }
1.1 schwarze 513: pos = 0;
514: continue;
515: default:
516: break;
517: }
518:
1.124 schwarze 519: if (curp->man->macroset == MACROSET_NONE)
1.60 schwarze 520: choose_parser(curp);
1.1 schwarze 521:
1.25 schwarze 522: /*
1.60 schwarze 523: * Lastly, push down into the parsers themselves.
1.1 schwarze 524: * If libroff returns ROFF_TBL, then add it to the
525: * currently open parse. Since we only get here if
526: * there does exist data (see tbl_data.c), we're
527: * guaranteed that something's been allocated.
528: * Do the same for ROFF_EQN.
529: */
530:
1.114 schwarze 531: if (rr == ROFF_TBL)
1.76 schwarze 532: while ((span = roff_span(curp->roff)) != NULL)
1.114 schwarze 533: roff_addtbl(curp->man, span);
534: else if (rr == ROFF_EQN)
535: roff_addeqn(curp->man, roff_eqn(curp->roff));
536: else if ((curp->man->macroset == MACROSET_MDOC ?
1.110 schwarze 537: mdoc_parseln(curp->man, curp->line, ln.buf, of) :
1.76 schwarze 538: man_parseln(curp->man, curp->line, ln.buf, of)) == 2)
539: break;
1.1 schwarze 540:
541: /* Temporary buffers typically are not full. */
542:
543: if (0 == start && '\0' == blk.buf[i])
544: break;
545:
546: /* Start the next input line. */
547:
548: pos = 0;
549: }
550:
551: free(ln.buf);
1.135 schwarze 552: return 1;
1.1 schwarze 553: }
554:
555: static int
1.18 schwarze 556: read_whole_file(struct mparse *curp, const char *file, int fd,
557: struct buf *fb, int *with_mmap)
1.1 schwarze 558: {
559: struct stat st;
1.115 schwarze 560: gzFile gz;
1.1 schwarze 561: size_t off;
562: ssize_t ssz;
563:
1.118 schwarze 564: if (fstat(fd, &st) == -1)
565: err((int)MANDOCLEVEL_SYSERR, "%s", file);
1.1 schwarze 566:
567: /*
568: * If we're a regular file, try just reading in the whole entry
569: * via mmap(). This is faster than reading it into blocks, and
570: * since each file is only a few bytes to begin with, I'm not
571: * concerned that this is going to tank any machines.
572: */
573:
1.115 schwarze 574: if (curp->gzip == 0 && S_ISREG(st.st_mode)) {
1.106 schwarze 575: if (st.st_size > 0x7fffffff) {
1.86 schwarze 576: mandoc_msg(MANDOCERR_TOOLARGE, curp, 0, 0, NULL);
1.117 schwarze 577: return 0;
1.1 schwarze 578: }
579: *with_mmap = 1;
580: fb->sz = (size_t)st.st_size;
1.15 schwarze 581: fb->buf = mmap(NULL, fb->sz, PROT_READ, MAP_SHARED, fd, 0);
1.1 schwarze 582: if (fb->buf != MAP_FAILED)
1.117 schwarze 583: return 1;
1.1 schwarze 584: }
585:
1.115 schwarze 586: if (curp->gzip) {
1.118 schwarze 587: if ((gz = gzdopen(fd, "rb")) == NULL)
588: err((int)MANDOCLEVEL_SYSERR, "%s", file);
1.115 schwarze 589: } else
590: gz = NULL;
591:
1.1 schwarze 592: /*
593: * If this isn't a regular file (like, say, stdin), then we must
594: * go the old way and just read things in bit by bit.
595: */
596:
597: *with_mmap = 0;
598: off = 0;
599: fb->sz = 0;
600: fb->buf = NULL;
601: for (;;) {
602: if (off == fb->sz) {
603: if (fb->sz == (1U << 31)) {
1.86 schwarze 604: mandoc_msg(MANDOCERR_TOOLARGE, curp,
605: 0, 0, NULL);
1.1 schwarze 606: break;
607: }
608: resize_buf(fb, 65536);
609: }
1.115 schwarze 610: ssz = curp->gzip ?
611: gzread(gz, fb->buf + (int)off, fb->sz - off) :
612: read(fd, fb->buf + (int)off, fb->sz - off);
1.1 schwarze 613: if (ssz == 0) {
614: fb->sz = off;
1.117 schwarze 615: return 1;
1.1 schwarze 616: }
1.118 schwarze 617: if (ssz == -1)
618: err((int)MANDOCLEVEL_SYSERR, "%s", file);
1.1 schwarze 619: off += (size_t)ssz;
620: }
621:
622: free(fb->buf);
623: fb->buf = NULL;
1.117 schwarze 624: return 0;
1.1 schwarze 625: }
626:
627: static void
628: mparse_end(struct mparse *curp)
629: {
1.110 schwarze 630: if (curp->man->macroset == MACROSET_NONE)
631: curp->man->macroset = MACROSET_MAN;
632: if (curp->man->macroset == MACROSET_MDOC)
633: mdoc_endparse(curp->man);
634: else
1.86 schwarze 635: man_endparse(curp->man);
1.1 schwarze 636: roff_endparse(curp->roff);
637: }
638:
1.15 schwarze 639: static void
640: mparse_parse_buffer(struct mparse *curp, struct buf blk, const char *file)
1.1 schwarze 641: {
1.61 schwarze 642: struct buf *svprimary;
1.1 schwarze 643: const char *svfile;
1.71 schwarze 644: size_t offset;
1.14 schwarze 645: static int recursion_depth;
646:
647: if (64 < recursion_depth) {
648: mandoc_msg(MANDOCERR_ROFFLOOP, curp, curp->line, 0, NULL);
1.15 schwarze 649: return;
1.14 schwarze 650: }
1.1 schwarze 651:
1.15 schwarze 652: /* Line number is per-file. */
653: svfile = curp->file;
654: curp->file = file;
1.61 schwarze 655: svprimary = curp->primary;
1.59 schwarze 656: curp->primary = &blk;
1.15 schwarze 657: curp->line = 1;
658: recursion_depth++;
659:
1.70 schwarze 660: /* Skip an UTF-8 byte order mark. */
661: if (curp->filenc & MPARSE_UTF8 && blk.sz > 2 &&
662: (unsigned char)blk.buf[0] == 0xef &&
663: (unsigned char)blk.buf[1] == 0xbb &&
664: (unsigned char)blk.buf[2] == 0xbf) {
1.71 schwarze 665: offset = 3;
1.70 schwarze 666: curp->filenc &= ~MPARSE_LATIN1;
1.71 schwarze 667: } else
668: offset = 0;
1.70 schwarze 669:
1.71 schwarze 670: mparse_buf_r(curp, blk, offset, 1);
1.15 schwarze 671:
1.86 schwarze 672: if (--recursion_depth == 0)
1.15 schwarze 673: mparse_end(curp);
674:
1.61 schwarze 675: curp->primary = svprimary;
1.15 schwarze 676: curp->file = svfile;
677: }
678:
1.74 schwarze 679: /*
680: * Read the whole file into memory and call the parsers.
681: * Called recursively when an .so request is encountered.
682: */
1.15 schwarze 683: enum mandoclevel
684: mparse_readfd(struct mparse *curp, int fd, const char *file)
685: {
686: struct buf blk;
687: int with_mmap;
1.70 schwarze 688: int save_filenc;
1.1 schwarze 689:
1.67 schwarze 690: if (read_whole_file(curp, file, fd, &blk, &with_mmap)) {
1.70 schwarze 691: save_filenc = curp->filenc;
692: curp->filenc = curp->options &
693: (MPARSE_UTF8 | MPARSE_LATIN1);
1.67 schwarze 694: mparse_parse_buffer(curp, blk, file);
1.70 schwarze 695: curp->filenc = save_filenc;
1.67 schwarze 696: if (with_mmap)
697: munmap(blk.buf, blk.sz);
698: else
699: free(blk.buf);
700: }
1.117 schwarze 701: return curp->file_status;
1.58 schwarze 702: }
703:
1.122 schwarze 704: int
705: mparse_open(struct mparse *curp, const char *file)
1.58 schwarze 706: {
707: char *cp;
1.122 schwarze 708: int fd;
1.58 schwarze 709:
710: curp->file = file;
1.115 schwarze 711: cp = strrchr(file, '.');
712: curp->gzip = (cp != NULL && ! strcmp(cp + 1, "gz"));
1.74 schwarze 713:
1.115 schwarze 714: /* First try to use the filename as it is. */
1.74 schwarze 715:
1.122 schwarze 716: if ((fd = open(file, O_RDONLY)) != -1)
717: return fd;
1.74 schwarze 718:
1.115 schwarze 719: /*
720: * If that doesn't work and the filename doesn't
721: * already end in .gz, try appending .gz.
722: */
1.74 schwarze 723:
1.115 schwarze 724: if ( ! curp->gzip) {
1.74 schwarze 725: mandoc_asprintf(&cp, "%s.gz", file);
1.123 schwarze 726: fd = open(cp, O_RDONLY);
1.83 schwarze 727: free(cp);
1.122 schwarze 728: if (fd != -1) {
1.115 schwarze 729: curp->gzip = 1;
1.122 schwarze 730: return fd;
1.58 schwarze 731: }
732: }
733:
1.115 schwarze 734: /* Neither worked, give up. */
1.73 schwarze 735:
1.115 schwarze 736: mandoc_msg(MANDOCERR_FILE, curp, 0, 0, strerror(errno));
1.122 schwarze 737: return -1;
1.1 schwarze 738: }
739:
740: struct mparse *
1.69 schwarze 741: mparse_alloc(int options, enum mandoclevel wlevel, mandocmsg mmsg,
1.119 schwarze 742: const char *defos)
1.1 schwarze 743: {
744: struct mparse *curp;
745:
746: curp = mandoc_calloc(1, sizeof(struct mparse));
747:
1.22 schwarze 748: curp->options = options;
1.1 schwarze 749: curp->wlevel = wlevel;
750: curp->mmsg = mmsg;
1.7 schwarze 751: curp->defos = defos;
1.1 schwarze 752:
1.119 schwarze 753: curp->roff = roff_alloc(curp, options);
1.112 schwarze 754: curp->man = roff_man_alloc( curp->roff, curp, curp->defos,
755: curp->options & MPARSE_QUICK ? 1 : 0);
1.111 schwarze 756: if (curp->options & MPARSE_MDOC) {
1.112 schwarze 757: curp->man->macroset = MACROSET_MDOC;
1.136 schwarze 758: if (curp->man->mdocmac == NULL)
759: curp->man->mdocmac = roffhash_alloc(MDOC_Dd, MDOC_MAX);
1.112 schwarze 760: } else if (curp->options & MPARSE_MAN) {
761: curp->man->macroset = MACROSET_MAN;
1.136 schwarze 762: if (curp->man->manmac == NULL)
763: curp->man->manmac = roffhash_alloc(MAN_TH, MAN_MAX);
1.111 schwarze 764: }
1.113 schwarze 765: curp->man->first->tok = TOKEN_NONE;
1.117 schwarze 766: return curp;
1.1 schwarze 767: }
768:
769: void
770: mparse_reset(struct mparse *curp)
771: {
772: roff_reset(curp->roff);
1.124 schwarze 773: roff_man_reset(curp->man);
1.133 schwarze 774:
775: free(curp->sodest);
776: curp->sodest = NULL;
777:
1.4 schwarze 778: if (curp->secondary)
779: curp->secondary->sz = 0;
1.1 schwarze 780:
781: curp->file_status = MANDOCLEVEL_OK;
1.132 schwarze 782: curp->gzip = 0;
1.1 schwarze 783: }
784:
785: void
786: mparse_free(struct mparse *curp)
787: {
788:
1.136 schwarze 789: roffhash_free(curp->man->mdocmac);
790: roffhash_free(curp->man->manmac);
1.112 schwarze 791: roff_man_free(curp->man);
1.133 schwarze 792: roff_free(curp->roff);
1.4 schwarze 793: if (curp->secondary)
794: free(curp->secondary->buf);
1.1 schwarze 795:
1.4 schwarze 796: free(curp->secondary);
1.23 schwarze 797: free(curp->sodest);
1.1 schwarze 798: free(curp);
799: }
800:
801: void
1.110 schwarze 802: mparse_result(struct mparse *curp, struct roff_man **man,
803: char **sodest)
1.1 schwarze 804: {
805:
1.23 schwarze 806: if (sodest && NULL != (*sodest = curp->sodest)) {
807: *man = NULL;
808: return;
809: }
1.1 schwarze 810: if (man)
811: *man = curp->man;
1.130 schwarze 812: }
813:
814: void
815: mparse_updaterc(struct mparse *curp, enum mandoclevel *rc)
816: {
817: if (curp->file_status > *rc)
818: *rc = curp->file_status;
1.1 schwarze 819: }
820:
821: void
822: mandoc_vmsg(enum mandocerr t, struct mparse *m,
823: int ln, int pos, const char *fmt, ...)
824: {
825: char buf[256];
826: va_list ap;
827:
828: va_start(ap, fmt);
1.26 schwarze 829: (void)vsnprintf(buf, sizeof(buf), fmt, ap);
1.1 schwarze 830: va_end(ap);
831:
832: mandoc_msg(t, m, ln, pos, buf);
833: }
834:
835: void
1.25 schwarze 836: mandoc_msg(enum mandocerr er, struct mparse *m,
1.1 schwarze 837: int ln, int col, const char *msg)
838: {
839: enum mandoclevel level;
840:
1.87 schwarze 841: level = MANDOCLEVEL_UNSUPP;
1.1 schwarze 842: while (er < mandoclimits[level])
843: level--;
844:
1.83 schwarze 845: if (level < m->wlevel && er != MANDOCERR_FILE)
1.1 schwarze 846: return;
847:
848: if (m->mmsg)
849: (*m->mmsg)(er, level, m->file, ln, col, msg);
850:
851: if (m->file_status < level)
852: m->file_status = level;
853: }
854:
855: const char *
856: mparse_strerror(enum mandocerr er)
857: {
858:
1.117 schwarze 859: return mandocerrs[er];
1.1 schwarze 860: }
861:
862: const char *
863: mparse_strlevel(enum mandoclevel lvl)
864: {
1.117 schwarze 865: return mandoclevels[lvl];
1.4 schwarze 866: }
867:
868: void
869: mparse_keep(struct mparse *p)
870: {
871:
872: assert(NULL == p->secondary);
873: p->secondary = mandoc_calloc(1, sizeof(struct buf));
874: }
875:
876: const char *
877: mparse_getkeep(const struct mparse *p)
878: {
879:
880: assert(p->secondary);
1.117 schwarze 881: return p->secondary->sz ? p->secondary->buf : NULL;
1.1 schwarze 882: }