Annotation of src/usr.bin/mandoc/read.c, Revision 1.156
1.156 ! schwarze 1: /* $OpenBSD: read.c,v 1.155 2017/06/29 15:21:46 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 */
1.151 schwarze 52: const char *os_s; /* default operating system */
1.59 schwarze 53: mandocmsg mmsg; /* warning/error message handler */
54: enum mandoclevel file_status; /* status of current parse */
1.151 schwarze 55: enum mandocerr mmin; /* ignore messages below this */
1.59 schwarze 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:
1.151 schwarze 85: "base system convention",
1.138 schwarze 86:
1.148 schwarze 87: "Mdocdate found",
88: "Mdocdate missing",
1.153 schwarze 89: "unknown architecture",
1.152 schwarze 90: "operating system explicitly specified",
1.151 schwarze 91: "RCS id missing",
1.156 ! schwarze 92: "referenced manual not found",
1.151 schwarze 93:
94: "generic style suggestion",
95:
1.148 schwarze 96: "legacy man(7) date format",
1.150 schwarze 97: "duplicate RCS id",
1.154 schwarze 98: "typo in section name",
1.138 schwarze 99: "useless macro",
1.139 schwarze 100: "consider using OS macro",
1.144 schwarze 101: "errnos out of order",
102: "duplicate errno",
1.140 schwarze 103: "description line ends with a full stop",
1.146 schwarze 104: "no blank before trailing delimiter",
1.147 schwarze 105: "function name without markup",
1.138 schwarze 106:
1.1 schwarze 107: "generic warning",
108:
109: /* related to the prologue */
1.57 schwarze 110: "missing manual title, using UNTITLED",
111: "missing manual title, using \"\"",
1.32 schwarze 112: "lower case character in document title",
1.57 schwarze 113: "missing manual section, using \"\"",
1.1 schwarze 114: "unknown manual section",
1.32 schwarze 115: "missing date, using today's date",
1.1 schwarze 116: "cannot parse date, using it verbatim",
1.57 schwarze 117: "missing Os macro, using \"\"",
118: "duplicate prologue macro",
119: "late prologue macro",
120: "skipping late title macro",
1.1 schwarze 121: "prologue macros out of order",
122:
123: /* related to document structure */
124: ".so is fragile, better use ln(1)",
1.28 schwarze 125: "no document body",
1.32 schwarze 126: "content before first section header",
127: "first section is not \"NAME\"",
1.129 schwarze 128: "NAME section without Nm before Nd",
1.103 schwarze 129: "NAME section without description",
130: "description not at the end of NAME",
131: "bad NAME section content",
1.129 schwarze 132: "missing comma before name",
1.96 schwarze 133: "missing description line, using \"\"",
1.134 schwarze 134: "description line outside NAME section",
1.1 schwarze 135: "sections out of conventional order",
1.32 schwarze 136: "duplicate section title",
137: "unexpected section",
1.63 schwarze 138: "unusual Xr order",
139: "unusual Xr punctuation",
1.62 schwarze 140: "AUTHORS section without An macro",
1.1 schwarze 141:
142: /* related to macros and nesting */
1.33 schwarze 143: "obsolete macro",
1.78 schwarze 144: "macro neither callable nor escaped",
1.1 schwarze 145: "skipping paragraph macro",
1.10 schwarze 146: "moving paragraph macro out of list",
1.1 schwarze 147: "skipping no-space macro",
148: "blocks badly nested",
149: "nested displays are not portable",
1.35 schwarze 150: "moving content out of list",
1.155 schwarze 151: "first macro on line",
1.56 schwarze 152: "fill mode already enabled, skipping",
153: "fill mode already disabled, skipping",
1.1 schwarze 154: "line scope broken",
1.141 schwarze 155: "skipping blank line in line scope",
1.1 schwarze 156:
157: /* related to missing macro arguments */
1.36 schwarze 158: "skipping empty request",
159: "conditional request controls empty scope",
1.1 schwarze 160: "skipping empty macro",
1.99 schwarze 161: "empty block",
1.40 schwarze 162: "empty argument, using 0n",
1.38 schwarze 163: "missing display type, using -ragged",
164: "list type is not the first argument",
1.125 schwarze 165: "missing -width in -tag list, using 6n",
1.56 schwarze 166: "missing utility name, using \"\"",
1.98 schwarze 167: "missing function name, using \"\"",
1.38 schwarze 168: "empty head in list item",
169: "empty list item",
1.155 schwarze 170: "missing argument, using next line",
1.39 schwarze 171: "missing font type, using \\fR",
172: "unknown font type, using \\fR",
1.79 schwarze 173: "nothing follows prefix",
1.97 schwarze 174: "empty reference block",
1.128 schwarze 175: "missing section argument",
1.38 schwarze 176: "missing -std argument, adding it",
1.100 schwarze 177: "missing option string, using \"\"",
178: "missing resource identifier, using \"\"",
1.66 schwarze 179: "missing eqn box, using \"\"",
1.1 schwarze 180:
181: /* related to bad macro arguments */
1.42 schwarze 182: "unterminated quoted argument",
1.1 schwarze 183: "duplicate argument",
1.54 schwarze 184: "skipping duplicate argument",
1.41 schwarze 185: "skipping duplicate display type",
186: "skipping duplicate list type",
1.54 schwarze 187: "skipping -width argument",
1.101 schwarze 188: "wrong number of cells",
1.1 schwarze 189: "unknown AT&T UNIX version",
1.64 schwarze 190: "comma in function argument",
1.65 schwarze 191: "parenthesis in function name",
1.145 schwarze 192: "unknown library name",
1.45 schwarze 193: "invalid content in Rs block",
1.41 schwarze 194: "invalid Boolean argument",
195: "unknown font, skipping request",
1.101 schwarze 196: "odd number of characters in request",
1.1 schwarze 197:
198: /* related to plain text */
1.42 schwarze 199: "blank line in fill mode, using .sp",
200: "tab in filled text",
201: "whitespace at end of input line",
1.131 schwarze 202: "new sentence, new line",
1.1 schwarze 203: "bad comment style",
1.42 schwarze 204: "invalid escape sequence",
205: "undefined string, using \"\"",
1.3 schwarze 206:
1.93 schwarze 207: /* related to tables */
208: "tbl line starts with span",
209: "tbl column starts with span",
210: "skipping vertical bar in tbl layout",
211:
1.1 schwarze 212: "generic error",
1.3 schwarze 213:
1.1 schwarze 214: /* related to tables */
1.91 schwarze 215: "non-alphabetic character in tbl options",
216: "skipping unknown tbl option",
217: "missing tbl option argument",
218: "wrong tbl option argument size",
1.93 schwarze 219: "empty tbl layout",
220: "invalid character in tbl layout",
221: "unmatched parenthesis in tbl layout",
1.94 schwarze 222: "tbl without any data cells",
223: "ignoring data in spanned tbl cell",
224: "ignoring extra tbl data cells",
225: "data block open at end of tbl",
1.1 schwarze 226:
1.46 schwarze 227: /* related to document structure and macros */
1.83 schwarze 228: NULL,
1.1 schwarze 229: "input stack limit exceeded, infinite loop?",
230: "skipping bad character",
1.46 schwarze 231: "skipping unknown macro",
1.87 schwarze 232: "skipping insecure request",
1.51 schwarze 233: "skipping item outside list",
1.46 schwarze 234: "skipping column outside column list",
235: "skipping end of block that is not open",
1.90 schwarze 236: "fewer RS blocks open, skipping",
1.46 schwarze 237: "inserting missing end of block",
238: "appending missing end of block",
239:
240: /* related to request and macro arguments */
1.1 schwarze 241: "escaped character not allowed in a name",
1.75 schwarze 242: "NOT IMPLEMENTED: Bd -file",
1.120 schwarze 243: "skipping display without arguments",
1.49 schwarze 244: "missing list type, using -item",
1.143 schwarze 245: "argument is not numeric, using 1",
1.48 schwarze 246: "missing manual name, using \"\"",
1.49 schwarze 247: "uname(3) system call failed, using UNKNOWN",
1.41 schwarze 248: "unknown standard specifier",
1.49 schwarze 249: "skipping request without numeric argument",
1.84 schwarze 250: "NOT IMPLEMENTED: .so with absolute path or \"..\"",
1.85 schwarze 251: ".so request failed",
1.39 schwarze 252: "skipping all arguments",
253: "skipping excess arguments",
1.68 schwarze 254: "divide by zero",
1.87 schwarze 255:
256: "unsupported feature",
257: "input too large",
1.89 schwarze 258: "unsupported control character",
1.87 schwarze 259: "unsupported roff request",
1.94 schwarze 260: "eqn delim option in tbl",
1.93 schwarze 261: "unsupported tbl layout modifier",
1.87 schwarze 262: "ignoring macro in table",
1.1 schwarze 263: };
264:
265: static const char * const mandoclevels[MANDOCLEVEL_MAX] = {
266: "SUCCESS",
1.137 schwarze 267: "STYLE",
1.1 schwarze 268: "WARNING",
269: "ERROR",
1.87 schwarze 270: "UNSUPP",
1.1 schwarze 271: "BADARG",
272: "SYSERR"
273: };
274:
1.25 schwarze 275:
1.1 schwarze 276: static void
277: resize_buf(struct buf *buf, size_t initial)
278: {
279:
280: buf->sz = buf->sz > initial/2 ? 2 * buf->sz : initial;
281: buf->buf = mandoc_realloc(buf->buf, buf->sz);
282: }
283:
284: static void
1.60 schwarze 285: choose_parser(struct mparse *curp)
1.1 schwarze 286: {
1.59 schwarze 287: char *cp, *ep;
288: int format;
1.1 schwarze 289:
1.59 schwarze 290: /*
291: * If neither command line arguments -mdoc or -man select
292: * a parser nor the roff parser found a .Dd or .TH macro
293: * yet, look ahead in the main input buffer.
294: */
295:
296: if ((format = roff_getformat(curp->roff)) == 0) {
297: cp = curp->primary->buf;
298: ep = cp + curp->primary->sz;
299: while (cp < ep) {
1.61 schwarze 300: if (*cp == '.' || *cp == '\'') {
1.59 schwarze 301: cp++;
302: if (cp[0] == 'D' && cp[1] == 'd') {
303: format = MPARSE_MDOC;
304: break;
305: }
306: if (cp[0] == 'T' && cp[1] == 'H') {
307: format = MPARSE_MAN;
308: break;
309: }
310: }
311: cp = memchr(cp, '\n', ep - cp);
312: if (cp == NULL)
313: break;
314: cp++;
315: }
1.1 schwarze 316: }
317:
1.59 schwarze 318: if (format == MPARSE_MDOC) {
1.112 schwarze 319: curp->man->macroset = MACROSET_MDOC;
1.136 schwarze 320: if (curp->man->mdocmac == NULL)
321: curp->man->mdocmac = roffhash_alloc(MDOC_Dd, MDOC_MAX);
1.112 schwarze 322: } else {
323: curp->man->macroset = MACROSET_MAN;
1.136 schwarze 324: if (curp->man->manmac == NULL)
325: curp->man->manmac = roffhash_alloc(MAN_TH, MAN_MAX);
1.25 schwarze 326: }
1.136 schwarze 327: curp->man->first->tok = TOKEN_NONE;
1.1 schwarze 328: }
329:
330: /*
1.71 schwarze 331: * Main parse routine for a buffer.
332: * It assumes encoding and line numbering are already set up.
333: * It can recurse directly (for invocations of user-defined
334: * macros, inline equations, and input line traps)
335: * and indirectly (for .so file inclusion).
1.1 schwarze 336: */
1.135 schwarze 337: static int
1.71 schwarze 338: mparse_buf_r(struct mparse *curp, struct buf blk, size_t i, int start)
1.1 schwarze 339: {
340: const struct tbl_span *span;
341: struct buf ln;
1.88 schwarze 342: const char *save_file;
1.85 schwarze 343: char *cp;
1.71 schwarze 344: size_t pos; /* byte number in the ln buffer */
1.1 schwarze 345: enum rofferr rr;
1.76 schwarze 346: int of;
1.1 schwarze 347: int lnn; /* line number in the real file */
1.85 schwarze 348: int fd;
1.1 schwarze 349: unsigned char c;
350:
1.71 schwarze 351: memset(&ln, 0, sizeof(ln));
1.1 schwarze 352:
1.25 schwarze 353: lnn = curp->line;
354: pos = 0;
1.1 schwarze 355:
1.71 schwarze 356: while (i < blk.sz) {
1.1 schwarze 357: if (0 == pos && '\0' == blk.buf[i])
358: break;
359:
360: if (start) {
361: curp->line = lnn;
362: curp->reparse_count = 0;
1.70 schwarze 363:
364: if (lnn < 3 &&
365: curp->filenc & MPARSE_UTF8 &&
1.71 schwarze 366: curp->filenc & MPARSE_LATIN1)
367: curp->filenc = preconv_cue(&blk, i);
1.1 schwarze 368: }
369:
1.71 schwarze 370: while (i < blk.sz && (start || blk.buf[i] != '\0')) {
1.1 schwarze 371:
372: /*
373: * When finding an unescaped newline character,
374: * leave the character loop to process the line.
375: * Skip a preceding carriage return, if any.
376: */
377:
1.71 schwarze 378: if ('\r' == blk.buf[i] && i + 1 < blk.sz &&
1.1 schwarze 379: '\n' == blk.buf[i + 1])
380: ++i;
381: if ('\n' == blk.buf[i]) {
382: ++i;
383: ++lnn;
384: break;
385: }
386:
1.13 schwarze 387: /*
1.70 schwarze 388: * Make sure we have space for the worst
389: * case of 11 bytes: "\\[u10ffff]\0"
1.13 schwarze 390: */
391:
1.71 schwarze 392: if (pos + 11 > ln.sz)
1.13 schwarze 393: resize_buf(&ln, 256);
394:
1.25 schwarze 395: /*
1.70 schwarze 396: * Encode 8-bit input.
1.1 schwarze 397: */
398:
1.70 schwarze 399: c = blk.buf[i];
400: if (c & 0x80) {
1.71 schwarze 401: if ( ! (curp->filenc && preconv_encode(
402: &blk, &i, &ln, &pos, &curp->filenc))) {
1.89 schwarze 403: mandoc_vmsg(MANDOCERR_CHAR_BAD, curp,
404: curp->line, pos, "0x%x", c);
1.70 schwarze 405: ln.buf[pos++] = '?';
406: i++;
407: }
408: continue;
409: }
410:
411: /*
412: * Exclude control characters.
413: */
1.1 schwarze 414:
1.70 schwarze 415: if (c == 0x7f || (c < 0x20 && c != 0x09)) {
1.89 schwarze 416: mandoc_vmsg(c == 0x00 || c == 0x04 ||
417: c > 0x0a ? MANDOCERR_CHAR_BAD :
418: MANDOCERR_CHAR_UNSUPP,
419: curp, curp->line, pos, "0x%x", c);
1.1 schwarze 420: i++;
1.102 schwarze 421: if (c != '\r')
422: ln.buf[pos++] = '?';
1.1 schwarze 423: continue;
424: }
425:
426: ln.buf[pos++] = blk.buf[i++];
427: }
428:
1.142 schwarze 429: if (pos + 1 >= ln.sz)
1.1 schwarze 430: resize_buf(&ln, 256);
431:
1.142 schwarze 432: if (i == blk.sz || blk.buf[i] == '\0')
433: ln.buf[pos++] = '\n';
1.1 schwarze 434: ln.buf[pos] = '\0';
435:
436: /*
437: * A significant amount of complexity is contained by
438: * the roff preprocessor. It's line-oriented but can be
439: * expressed on one line, so we need at times to
440: * readjust our starting point and re-run it. The roff
441: * preprocessor can also readjust the buffers with new
442: * data, so we pass them in wholesale.
443: */
444:
445: of = 0;
446:
1.4 schwarze 447: /*
448: * Maintain a lookaside buffer of all parsed lines. We
449: * only do this if mparse_keep() has been invoked (the
450: * buffer may be accessed with mparse_getkeep()).
451: */
452:
453: if (curp->secondary) {
1.25 schwarze 454: curp->secondary->buf = mandoc_realloc(
455: curp->secondary->buf,
456: curp->secondary->sz + pos + 2);
457: memcpy(curp->secondary->buf +
458: curp->secondary->sz,
459: ln.buf, pos);
1.4 schwarze 460: curp->secondary->sz += pos;
461: curp->secondary->buf
462: [curp->secondary->sz] = '\n';
463: curp->secondary->sz++;
464: curp->secondary->buf
465: [curp->secondary->sz] = '\0';
466: }
1.1 schwarze 467: rerun:
1.72 schwarze 468: rr = roff_parseln(curp->roff, curp->line, &ln, &of);
1.1 schwarze 469:
470: switch (rr) {
1.25 schwarze 471: case ROFF_REPARSE:
1.135 schwarze 472: if (++curp->reparse_count > REPARSE_LIMIT)
1.1 schwarze 473: mandoc_msg(MANDOCERR_ROFFLOOP, curp,
1.25 schwarze 474: curp->line, pos, NULL);
1.135 schwarze 475: else if (mparse_buf_r(curp, ln, of, 0) == 1 ||
476: start == 1) {
477: pos = 0;
478: continue;
479: }
480: free(ln.buf);
481: return 0;
1.25 schwarze 482: case ROFF_APPEND:
1.71 schwarze 483: pos = strlen(ln.buf);
1.1 schwarze 484: continue;
1.25 schwarze 485: case ROFF_RERUN:
1.1 schwarze 486: goto rerun;
1.25 schwarze 487: case ROFF_IGN:
1.1 schwarze 488: pos = 0;
489: continue;
1.25 schwarze 490: case ROFF_SO:
1.71 schwarze 491: if ( ! (curp->options & MPARSE_SO) &&
492: (i >= blk.sz || blk.buf[i] == '\0')) {
1.23 schwarze 493: curp->sodest = mandoc_strdup(ln.buf + of);
494: free(ln.buf);
1.135 schwarze 495: return 1;
1.23 schwarze 496: }
1.4 schwarze 497: /*
498: * We remove `so' clauses from our lookaside
499: * buffer because we're going to descend into
500: * the file recursively.
501: */
1.25 schwarze 502: if (curp->secondary)
1.4 schwarze 503: curp->secondary->sz -= pos + 1;
1.88 schwarze 504: save_file = curp->file;
1.122 schwarze 505: if ((fd = mparse_open(curp, ln.buf + of)) != -1) {
1.85 schwarze 506: mparse_readfd(curp, fd, ln.buf + of);
1.121 schwarze 507: close(fd);
1.88 schwarze 508: curp->file = save_file;
509: } else {
510: curp->file = save_file;
1.30 schwarze 511: mandoc_vmsg(MANDOCERR_SO_FAIL,
512: curp, curp->line, pos,
513: ".so %s", ln.buf + of);
1.85 schwarze 514: ln.sz = mandoc_asprintf(&cp,
515: ".sp\nSee the file %s.\n.sp",
516: ln.buf + of);
517: free(ln.buf);
518: ln.buf = cp;
519: of = 0;
520: mparse_buf_r(curp, ln, of, 0);
1.30 schwarze 521: }
1.1 schwarze 522: pos = 0;
523: continue;
524: default:
525: break;
526: }
527:
1.124 schwarze 528: if (curp->man->macroset == MACROSET_NONE)
1.60 schwarze 529: choose_parser(curp);
1.1 schwarze 530:
1.25 schwarze 531: /*
1.60 schwarze 532: * Lastly, push down into the parsers themselves.
1.1 schwarze 533: * If libroff returns ROFF_TBL, then add it to the
534: * currently open parse. Since we only get here if
535: * there does exist data (see tbl_data.c), we're
536: * guaranteed that something's been allocated.
537: * Do the same for ROFF_EQN.
538: */
539:
1.114 schwarze 540: if (rr == ROFF_TBL)
1.76 schwarze 541: while ((span = roff_span(curp->roff)) != NULL)
1.114 schwarze 542: roff_addtbl(curp->man, span);
543: else if (rr == ROFF_EQN)
544: roff_addeqn(curp->man, roff_eqn(curp->roff));
545: else if ((curp->man->macroset == MACROSET_MDOC ?
1.110 schwarze 546: mdoc_parseln(curp->man, curp->line, ln.buf, of) :
1.76 schwarze 547: man_parseln(curp->man, curp->line, ln.buf, of)) == 2)
548: break;
1.1 schwarze 549:
550: /* Temporary buffers typically are not full. */
551:
552: if (0 == start && '\0' == blk.buf[i])
553: break;
554:
555: /* Start the next input line. */
556:
557: pos = 0;
558: }
559:
560: free(ln.buf);
1.135 schwarze 561: return 1;
1.1 schwarze 562: }
563:
564: static int
1.18 schwarze 565: read_whole_file(struct mparse *curp, const char *file, int fd,
566: struct buf *fb, int *with_mmap)
1.1 schwarze 567: {
568: struct stat st;
1.115 schwarze 569: gzFile gz;
1.1 schwarze 570: size_t off;
571: ssize_t ssz;
572:
1.118 schwarze 573: if (fstat(fd, &st) == -1)
574: err((int)MANDOCLEVEL_SYSERR, "%s", file);
1.1 schwarze 575:
576: /*
577: * If we're a regular file, try just reading in the whole entry
578: * via mmap(). This is faster than reading it into blocks, and
579: * since each file is only a few bytes to begin with, I'm not
580: * concerned that this is going to tank any machines.
581: */
582:
1.115 schwarze 583: if (curp->gzip == 0 && S_ISREG(st.st_mode)) {
1.106 schwarze 584: if (st.st_size > 0x7fffffff) {
1.86 schwarze 585: mandoc_msg(MANDOCERR_TOOLARGE, curp, 0, 0, NULL);
1.117 schwarze 586: return 0;
1.1 schwarze 587: }
588: *with_mmap = 1;
589: fb->sz = (size_t)st.st_size;
1.15 schwarze 590: fb->buf = mmap(NULL, fb->sz, PROT_READ, MAP_SHARED, fd, 0);
1.1 schwarze 591: if (fb->buf != MAP_FAILED)
1.117 schwarze 592: return 1;
1.1 schwarze 593: }
594:
1.115 schwarze 595: if (curp->gzip) {
1.118 schwarze 596: if ((gz = gzdopen(fd, "rb")) == NULL)
597: err((int)MANDOCLEVEL_SYSERR, "%s", file);
1.115 schwarze 598: } else
599: gz = NULL;
600:
1.1 schwarze 601: /*
602: * If this isn't a regular file (like, say, stdin), then we must
603: * go the old way and just read things in bit by bit.
604: */
605:
606: *with_mmap = 0;
607: off = 0;
608: fb->sz = 0;
609: fb->buf = NULL;
610: for (;;) {
611: if (off == fb->sz) {
612: if (fb->sz == (1U << 31)) {
1.86 schwarze 613: mandoc_msg(MANDOCERR_TOOLARGE, curp,
614: 0, 0, NULL);
1.1 schwarze 615: break;
616: }
617: resize_buf(fb, 65536);
618: }
1.115 schwarze 619: ssz = curp->gzip ?
620: gzread(gz, fb->buf + (int)off, fb->sz - off) :
621: read(fd, fb->buf + (int)off, fb->sz - off);
1.1 schwarze 622: if (ssz == 0) {
623: fb->sz = off;
1.117 schwarze 624: return 1;
1.1 schwarze 625: }
1.118 schwarze 626: if (ssz == -1)
627: err((int)MANDOCLEVEL_SYSERR, "%s", file);
1.1 schwarze 628: off += (size_t)ssz;
629: }
630:
631: free(fb->buf);
632: fb->buf = NULL;
1.117 schwarze 633: return 0;
1.1 schwarze 634: }
635:
636: static void
637: mparse_end(struct mparse *curp)
638: {
1.110 schwarze 639: if (curp->man->macroset == MACROSET_NONE)
640: curp->man->macroset = MACROSET_MAN;
641: if (curp->man->macroset == MACROSET_MDOC)
642: mdoc_endparse(curp->man);
643: else
1.86 schwarze 644: man_endparse(curp->man);
1.1 schwarze 645: roff_endparse(curp->roff);
646: }
647:
1.15 schwarze 648: static void
649: mparse_parse_buffer(struct mparse *curp, struct buf blk, const char *file)
1.1 schwarze 650: {
1.61 schwarze 651: struct buf *svprimary;
1.1 schwarze 652: const char *svfile;
1.71 schwarze 653: size_t offset;
1.14 schwarze 654: static int recursion_depth;
655:
656: if (64 < recursion_depth) {
657: mandoc_msg(MANDOCERR_ROFFLOOP, curp, curp->line, 0, NULL);
1.15 schwarze 658: return;
1.14 schwarze 659: }
1.1 schwarze 660:
1.15 schwarze 661: /* Line number is per-file. */
662: svfile = curp->file;
663: curp->file = file;
1.61 schwarze 664: svprimary = curp->primary;
1.59 schwarze 665: curp->primary = &blk;
1.15 schwarze 666: curp->line = 1;
667: recursion_depth++;
668:
1.70 schwarze 669: /* Skip an UTF-8 byte order mark. */
670: if (curp->filenc & MPARSE_UTF8 && blk.sz > 2 &&
671: (unsigned char)blk.buf[0] == 0xef &&
672: (unsigned char)blk.buf[1] == 0xbb &&
673: (unsigned char)blk.buf[2] == 0xbf) {
1.71 schwarze 674: offset = 3;
1.70 schwarze 675: curp->filenc &= ~MPARSE_LATIN1;
1.71 schwarze 676: } else
677: offset = 0;
1.70 schwarze 678:
1.71 schwarze 679: mparse_buf_r(curp, blk, offset, 1);
1.15 schwarze 680:
1.86 schwarze 681: if (--recursion_depth == 0)
1.15 schwarze 682: mparse_end(curp);
683:
1.61 schwarze 684: curp->primary = svprimary;
1.15 schwarze 685: curp->file = svfile;
686: }
687:
1.74 schwarze 688: /*
689: * Read the whole file into memory and call the parsers.
690: * Called recursively when an .so request is encountered.
691: */
1.15 schwarze 692: enum mandoclevel
693: mparse_readfd(struct mparse *curp, int fd, const char *file)
694: {
695: struct buf blk;
696: int with_mmap;
1.70 schwarze 697: int save_filenc;
1.1 schwarze 698:
1.67 schwarze 699: if (read_whole_file(curp, file, fd, &blk, &with_mmap)) {
1.70 schwarze 700: save_filenc = curp->filenc;
701: curp->filenc = curp->options &
702: (MPARSE_UTF8 | MPARSE_LATIN1);
1.67 schwarze 703: mparse_parse_buffer(curp, blk, file);
1.70 schwarze 704: curp->filenc = save_filenc;
1.67 schwarze 705: if (with_mmap)
706: munmap(blk.buf, blk.sz);
707: else
708: free(blk.buf);
709: }
1.117 schwarze 710: return curp->file_status;
1.58 schwarze 711: }
712:
1.122 schwarze 713: int
714: mparse_open(struct mparse *curp, const char *file)
1.58 schwarze 715: {
716: char *cp;
1.122 schwarze 717: int fd;
1.58 schwarze 718:
719: curp->file = file;
1.115 schwarze 720: cp = strrchr(file, '.');
721: curp->gzip = (cp != NULL && ! strcmp(cp + 1, "gz"));
1.74 schwarze 722:
1.115 schwarze 723: /* First try to use the filename as it is. */
1.74 schwarze 724:
1.122 schwarze 725: if ((fd = open(file, O_RDONLY)) != -1)
726: return fd;
1.74 schwarze 727:
1.115 schwarze 728: /*
729: * If that doesn't work and the filename doesn't
730: * already end in .gz, try appending .gz.
731: */
1.74 schwarze 732:
1.115 schwarze 733: if ( ! curp->gzip) {
1.74 schwarze 734: mandoc_asprintf(&cp, "%s.gz", file);
1.123 schwarze 735: fd = open(cp, O_RDONLY);
1.83 schwarze 736: free(cp);
1.122 schwarze 737: if (fd != -1) {
1.115 schwarze 738: curp->gzip = 1;
1.122 schwarze 739: return fd;
1.58 schwarze 740: }
741: }
742:
1.115 schwarze 743: /* Neither worked, give up. */
1.73 schwarze 744:
1.115 schwarze 745: mandoc_msg(MANDOCERR_FILE, curp, 0, 0, strerror(errno));
1.122 schwarze 746: return -1;
1.1 schwarze 747: }
748:
749: struct mparse *
1.151 schwarze 750: mparse_alloc(int options, enum mandocerr mmin, mandocmsg mmsg,
751: enum mandoc_os os_e, const char *os_s)
1.1 schwarze 752: {
753: struct mparse *curp;
754:
755: curp = mandoc_calloc(1, sizeof(struct mparse));
756:
1.22 schwarze 757: curp->options = options;
1.151 schwarze 758: curp->mmin = mmin;
1.1 schwarze 759: curp->mmsg = mmsg;
1.151 schwarze 760: curp->os_s = os_s;
1.1 schwarze 761:
1.119 schwarze 762: curp->roff = roff_alloc(curp, options);
1.151 schwarze 763: curp->man = roff_man_alloc(curp->roff, curp, curp->os_s,
1.112 schwarze 764: curp->options & MPARSE_QUICK ? 1 : 0);
1.111 schwarze 765: if (curp->options & MPARSE_MDOC) {
1.112 schwarze 766: curp->man->macroset = MACROSET_MDOC;
1.136 schwarze 767: if (curp->man->mdocmac == NULL)
768: curp->man->mdocmac = roffhash_alloc(MDOC_Dd, MDOC_MAX);
1.112 schwarze 769: } else if (curp->options & MPARSE_MAN) {
770: curp->man->macroset = MACROSET_MAN;
1.136 schwarze 771: if (curp->man->manmac == NULL)
772: curp->man->manmac = roffhash_alloc(MAN_TH, MAN_MAX);
1.111 schwarze 773: }
1.113 schwarze 774: curp->man->first->tok = TOKEN_NONE;
1.151 schwarze 775: curp->man->meta.os_e = os_e;
1.117 schwarze 776: return curp;
1.1 schwarze 777: }
778:
779: void
780: mparse_reset(struct mparse *curp)
781: {
782: roff_reset(curp->roff);
1.124 schwarze 783: roff_man_reset(curp->man);
1.133 schwarze 784:
785: free(curp->sodest);
786: curp->sodest = NULL;
787:
1.4 schwarze 788: if (curp->secondary)
789: curp->secondary->sz = 0;
1.1 schwarze 790:
791: curp->file_status = MANDOCLEVEL_OK;
1.132 schwarze 792: curp->gzip = 0;
1.1 schwarze 793: }
794:
795: void
796: mparse_free(struct mparse *curp)
797: {
798:
1.136 schwarze 799: roffhash_free(curp->man->mdocmac);
800: roffhash_free(curp->man->manmac);
1.112 schwarze 801: roff_man_free(curp->man);
1.133 schwarze 802: roff_free(curp->roff);
1.4 schwarze 803: if (curp->secondary)
804: free(curp->secondary->buf);
1.1 schwarze 805:
1.4 schwarze 806: free(curp->secondary);
1.23 schwarze 807: free(curp->sodest);
1.1 schwarze 808: free(curp);
809: }
810:
811: void
1.110 schwarze 812: mparse_result(struct mparse *curp, struct roff_man **man,
813: char **sodest)
1.1 schwarze 814: {
815:
1.23 schwarze 816: if (sodest && NULL != (*sodest = curp->sodest)) {
817: *man = NULL;
818: return;
819: }
1.1 schwarze 820: if (man)
821: *man = curp->man;
1.130 schwarze 822: }
823:
824: void
825: mparse_updaterc(struct mparse *curp, enum mandoclevel *rc)
826: {
827: if (curp->file_status > *rc)
828: *rc = curp->file_status;
1.1 schwarze 829: }
830:
831: void
832: mandoc_vmsg(enum mandocerr t, struct mparse *m,
833: int ln, int pos, const char *fmt, ...)
834: {
835: char buf[256];
836: va_list ap;
837:
838: va_start(ap, fmt);
1.26 schwarze 839: (void)vsnprintf(buf, sizeof(buf), fmt, ap);
1.1 schwarze 840: va_end(ap);
841:
842: mandoc_msg(t, m, ln, pos, buf);
843: }
844:
845: void
1.25 schwarze 846: mandoc_msg(enum mandocerr er, struct mparse *m,
1.1 schwarze 847: int ln, int col, const char *msg)
848: {
849: enum mandoclevel level;
850:
1.151 schwarze 851: if (er < m->mmin && er != MANDOCERR_FILE)
852: return;
853:
1.87 schwarze 854: level = MANDOCLEVEL_UNSUPP;
1.1 schwarze 855: while (er < mandoclimits[level])
856: level--;
857:
858: if (m->mmsg)
859: (*m->mmsg)(er, level, m->file, ln, col, msg);
860:
861: if (m->file_status < level)
862: m->file_status = level;
863: }
864:
865: const char *
866: mparse_strerror(enum mandocerr er)
867: {
868:
1.117 schwarze 869: return mandocerrs[er];
1.1 schwarze 870: }
871:
872: const char *
873: mparse_strlevel(enum mandoclevel lvl)
874: {
1.117 schwarze 875: return mandoclevels[lvl];
1.4 schwarze 876: }
877:
878: void
879: mparse_keep(struct mparse *p)
880: {
881:
882: assert(NULL == p->secondary);
883: p->secondary = mandoc_calloc(1, sizeof(struct buf));
884: }
885:
886: const char *
887: mparse_getkeep(const struct mparse *p)
888: {
889:
890: assert(p->secondary);
1.117 schwarze 891: return p->secondary->sz ? p->secondary->buf : NULL;
1.1 schwarze 892: }