Annotation of src/usr.bin/mandoc/read.c, Revision 1.100
1.100 ! schwarze 1: /* $OpenBSD: read.c,v 1.99 2015/02/06 07:12:34 schwarze Exp $ */
1.1 schwarze 2: /*
3: * Copyright (c) 2008, 2009, 2010, 2011 Kristaps Dzonsons <kristaps@bsd.lv>
1.82 schwarze 4: * Copyright (c) 2010-2015 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: *
11: * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
12: * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
13: * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
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>
1.58 schwarze 22: #include <sys/wait.h>
1.1 schwarze 23:
24: #include <assert.h>
25: #include <ctype.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>
33:
34: #include "mandoc.h"
1.24 schwarze 35: #include "mandoc_aux.h"
1.1 schwarze 36: #include "libmandoc.h"
37: #include "mdoc.h"
38: #include "man.h"
39:
40: #define REPARSE_LIMIT 1000
41:
42: struct mparse {
43: struct man *pman; /* persistent man parser */
44: struct mdoc *pmdoc; /* persistent mdoc parser */
45: struct man *man; /* man parser */
46: struct mdoc *mdoc; /* mdoc parser */
47: struct roff *roff; /* roff parser (!NULL) */
1.69 schwarze 48: const struct mchars *mchars; /* character table */
1.23 schwarze 49: char *sodest; /* filename pointed to by .so */
1.59 schwarze 50: const char *file; /* filename of current input file */
51: struct buf *primary; /* buffer currently being parsed */
52: struct buf *secondary; /* preprocessed copy of input */
53: const char *defos; /* default operating system */
54: mandocmsg mmsg; /* warning/error message handler */
55: enum mandoclevel file_status; /* status of current parse */
56: enum mandoclevel wlevel; /* ignore messages below this */
57: int options; /* parser options */
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 */
1.73 schwarze 61: pid_t child; /* the gunzip(1) process */
1.1 schwarze 62: };
63:
1.60 schwarze 64: static void choose_parser(struct mparse *);
1.1 schwarze 65: static void resize_buf(struct buf *, size_t);
1.71 schwarze 66: static void mparse_buf_r(struct mparse *, struct buf, size_t, int);
1.18 schwarze 67: static int read_whole_file(struct mparse *, const char *, int,
68: struct buf *, int *);
1.1 schwarze 69: static void mparse_end(struct mparse *);
1.15 schwarze 70: static void mparse_parse_buffer(struct mparse *, struct buf,
71: const char *);
1.1 schwarze 72:
73: static const enum mandocerr mandoclimits[MANDOCLEVEL_MAX] = {
74: MANDOCERR_OK,
75: MANDOCERR_WARNING,
76: MANDOCERR_WARNING,
77: MANDOCERR_ERROR,
1.87 schwarze 78: MANDOCERR_UNSUPP,
1.1 schwarze 79: MANDOCERR_MAX,
80: MANDOCERR_MAX
81: };
82:
83: static const char * const mandocerrs[MANDOCERR_MAX] = {
84: "ok",
85:
86: "generic warning",
87:
88: /* related to the prologue */
1.57 schwarze 89: "missing manual title, using UNTITLED",
90: "missing manual title, using \"\"",
1.32 schwarze 91: "lower case character in document title",
1.57 schwarze 92: "missing manual section, using \"\"",
1.1 schwarze 93: "unknown manual section",
1.32 schwarze 94: "missing date, using today's date",
1.1 schwarze 95: "cannot parse date, using it verbatim",
1.57 schwarze 96: "missing Os macro, using \"\"",
97: "duplicate prologue macro",
98: "late prologue macro",
99: "skipping late title macro",
1.1 schwarze 100: "prologue macros out of order",
101:
102: /* related to document structure */
103: ".so is fragile, better use ln(1)",
1.28 schwarze 104: "no document body",
1.32 schwarze 105: "content before first section header",
106: "first section is not \"NAME\"",
1.1 schwarze 107: "bad NAME section contents",
1.96 schwarze 108: "missing description line, using \"\"",
1.1 schwarze 109: "sections out of conventional order",
1.32 schwarze 110: "duplicate section title",
111: "unexpected section",
1.63 schwarze 112: "unusual Xr order",
113: "unusual Xr punctuation",
1.62 schwarze 114: "AUTHORS section without An macro",
1.1 schwarze 115:
116: /* related to macros and nesting */
1.33 schwarze 117: "obsolete macro",
1.78 schwarze 118: "macro neither callable nor escaped",
1.1 schwarze 119: "skipping paragraph macro",
1.10 schwarze 120: "moving paragraph macro out of list",
1.1 schwarze 121: "skipping no-space macro",
122: "blocks badly nested",
123: "nested displays are not portable",
1.35 schwarze 124: "moving content out of list",
125: ".Vt block has child macro",
1.56 schwarze 126: "fill mode already enabled, skipping",
127: "fill mode already disabled, skipping",
1.1 schwarze 128: "line scope broken",
129:
130: /* related to missing macro arguments */
1.36 schwarze 131: "skipping empty request",
132: "conditional request controls empty scope",
1.1 schwarze 133: "skipping empty macro",
1.99 schwarze 134: "empty block",
1.40 schwarze 135: "empty argument, using 0n",
1.38 schwarze 136: "missing display type, using -ragged",
137: "list type is not the first argument",
138: "missing -width in -tag list, using 8n",
1.56 schwarze 139: "missing utility name, using \"\"",
1.98 schwarze 140: "missing function name, using \"\"",
1.38 schwarze 141: "empty head in list item",
142: "empty list item",
1.39 schwarze 143: "missing font type, using \\fR",
144: "unknown font type, using \\fR",
1.79 schwarze 145: "nothing follows prefix",
1.97 schwarze 146: "empty reference block",
1.38 schwarze 147: "missing -std argument, adding it",
1.100 ! schwarze 148: "missing option string, using \"\"",
! 149: "missing resource identifier, using \"\"",
1.66 schwarze 150: "missing eqn box, using \"\"",
1.1 schwarze 151:
152: /* related to bad macro arguments */
1.42 schwarze 153: "unterminated quoted argument",
1.1 schwarze 154: "duplicate argument",
1.54 schwarze 155: "skipping duplicate argument",
1.41 schwarze 156: "skipping duplicate display type",
157: "skipping duplicate list type",
1.54 schwarze 158: "skipping -width argument",
1.1 schwarze 159: "unknown AT&T UNIX version",
1.64 schwarze 160: "comma in function argument",
1.65 schwarze 161: "parenthesis in function name",
1.45 schwarze 162: "invalid content in Rs block",
1.41 schwarze 163: "invalid Boolean argument",
164: "unknown font, skipping request",
1.1 schwarze 165:
166: /* related to plain text */
1.42 schwarze 167: "blank line in fill mode, using .sp",
168: "tab in filled text",
169: "whitespace at end of input line",
1.1 schwarze 170: "bad comment style",
1.42 schwarze 171: "invalid escape sequence",
172: "undefined string, using \"\"",
1.3 schwarze 173:
1.93 schwarze 174: /* related to tables */
175: "tbl line starts with span",
176: "tbl column starts with span",
177: "skipping vertical bar in tbl layout",
178:
1.1 schwarze 179: "generic error",
1.3 schwarze 180:
1.1 schwarze 181: /* related to tables */
1.91 schwarze 182: "non-alphabetic character in tbl options",
183: "skipping unknown tbl option",
184: "missing tbl option argument",
185: "wrong tbl option argument size",
1.93 schwarze 186: "empty tbl layout",
187: "invalid character in tbl layout",
188: "unmatched parenthesis in tbl layout",
1.94 schwarze 189: "tbl without any data cells",
190: "ignoring data in spanned tbl cell",
191: "ignoring extra tbl data cells",
192: "data block open at end of tbl",
1.1 schwarze 193:
1.46 schwarze 194: /* related to document structure and macros */
1.83 schwarze 195: NULL,
1.1 schwarze 196: "input stack limit exceeded, infinite loop?",
197: "skipping bad character",
1.46 schwarze 198: "skipping unknown macro",
1.87 schwarze 199: "skipping insecure request",
1.51 schwarze 200: "skipping item outside list",
1.46 schwarze 201: "skipping column outside column list",
202: "skipping end of block that is not open",
1.90 schwarze 203: "fewer RS blocks open, skipping",
1.46 schwarze 204: "inserting missing end of block",
205: "appending missing end of block",
206:
207: /* related to request and macro arguments */
1.1 schwarze 208: "escaped character not allowed in a name",
209: "argument count wrong",
1.75 schwarze 210: "NOT IMPLEMENTED: Bd -file",
1.49 schwarze 211: "missing list type, using -item",
1.48 schwarze 212: "missing manual name, using \"\"",
1.49 schwarze 213: "uname(3) system call failed, using UNKNOWN",
1.41 schwarze 214: "unknown standard specifier",
1.49 schwarze 215: "skipping request without numeric argument",
1.84 schwarze 216: "NOT IMPLEMENTED: .so with absolute path or \"..\"",
1.85 schwarze 217: ".so request failed",
1.39 schwarze 218: "skipping all arguments",
219: "skipping excess arguments",
1.68 schwarze 220: "divide by zero",
1.87 schwarze 221:
222: "unsupported feature",
223: "input too large",
1.89 schwarze 224: "unsupported control character",
1.87 schwarze 225: "unsupported roff request",
1.94 schwarze 226: "eqn delim option in tbl",
1.93 schwarze 227: "unsupported tbl layout modifier",
1.87 schwarze 228: "ignoring macro in table",
1.1 schwarze 229: };
230:
231: static const char * const mandoclevels[MANDOCLEVEL_MAX] = {
232: "SUCCESS",
233: "RESERVED",
234: "WARNING",
235: "ERROR",
1.87 schwarze 236: "UNSUPP",
1.1 schwarze 237: "BADARG",
238: "SYSERR"
239: };
240:
1.25 schwarze 241:
1.1 schwarze 242: static void
243: resize_buf(struct buf *buf, size_t initial)
244: {
245:
246: buf->sz = buf->sz > initial/2 ? 2 * buf->sz : initial;
247: buf->buf = mandoc_realloc(buf->buf, buf->sz);
248: }
249:
250: static void
1.60 schwarze 251: choose_parser(struct mparse *curp)
1.1 schwarze 252: {
1.59 schwarze 253: char *cp, *ep;
254: int format;
1.1 schwarze 255:
1.59 schwarze 256: /*
257: * If neither command line arguments -mdoc or -man select
258: * a parser nor the roff parser found a .Dd or .TH macro
259: * yet, look ahead in the main input buffer.
260: */
261:
262: if ((format = roff_getformat(curp->roff)) == 0) {
263: cp = curp->primary->buf;
264: ep = cp + curp->primary->sz;
265: while (cp < ep) {
1.61 schwarze 266: if (*cp == '.' || *cp == '\'') {
1.59 schwarze 267: cp++;
268: if (cp[0] == 'D' && cp[1] == 'd') {
269: format = MPARSE_MDOC;
270: break;
271: }
272: if (cp[0] == 'T' && cp[1] == 'H') {
273: format = MPARSE_MAN;
274: break;
275: }
276: }
277: cp = memchr(cp, '\n', ep - cp);
278: if (cp == NULL)
279: break;
280: cp++;
281: }
1.1 schwarze 282: }
283:
1.59 schwarze 284: if (format == MPARSE_MDOC) {
1.25 schwarze 285: if (NULL == curp->pmdoc)
1.22 schwarze 286: curp->pmdoc = mdoc_alloc(
287: curp->roff, curp, curp->defos,
288: MPARSE_QUICK & curp->options ? 1 : 0);
1.1 schwarze 289: assert(curp->pmdoc);
290: curp->mdoc = curp->pmdoc;
291: return;
1.25 schwarze 292: }
1.1 schwarze 293:
1.59 schwarze 294: /* Fall back to man(7) as a last resort. */
295:
1.25 schwarze 296: if (NULL == curp->pman)
1.81 schwarze 297: curp->pman = man_alloc(
298: curp->roff, curp, curp->defos,
1.22 schwarze 299: MPARSE_QUICK & curp->options ? 1 : 0);
1.1 schwarze 300: assert(curp->pman);
301: curp->man = curp->pman;
302: }
303:
304: /*
1.71 schwarze 305: * Main parse routine for a buffer.
306: * It assumes encoding and line numbering are already set up.
307: * It can recurse directly (for invocations of user-defined
308: * macros, inline equations, and input line traps)
309: * and indirectly (for .so file inclusion).
1.1 schwarze 310: */
311: static void
1.71 schwarze 312: mparse_buf_r(struct mparse *curp, struct buf blk, size_t i, int start)
1.1 schwarze 313: {
314: const struct tbl_span *span;
315: struct buf ln;
1.88 schwarze 316: const char *save_file;
1.85 schwarze 317: char *cp;
1.71 schwarze 318: size_t pos; /* byte number in the ln buffer */
1.1 schwarze 319: enum rofferr rr;
1.76 schwarze 320: int of;
1.1 schwarze 321: int lnn; /* line number in the real file */
1.85 schwarze 322: int fd;
323: pid_t save_child;
1.1 schwarze 324: unsigned char c;
325:
1.71 schwarze 326: memset(&ln, 0, sizeof(ln));
1.1 schwarze 327:
1.25 schwarze 328: lnn = curp->line;
329: pos = 0;
1.1 schwarze 330:
1.71 schwarze 331: while (i < blk.sz) {
1.1 schwarze 332: if (0 == pos && '\0' == blk.buf[i])
333: break;
334:
335: if (start) {
336: curp->line = lnn;
337: curp->reparse_count = 0;
1.70 schwarze 338:
339: if (lnn < 3 &&
340: curp->filenc & MPARSE_UTF8 &&
1.71 schwarze 341: curp->filenc & MPARSE_LATIN1)
342: curp->filenc = preconv_cue(&blk, i);
1.1 schwarze 343: }
344:
1.71 schwarze 345: while (i < blk.sz && (start || blk.buf[i] != '\0')) {
1.1 schwarze 346:
347: /*
348: * When finding an unescaped newline character,
349: * leave the character loop to process the line.
350: * Skip a preceding carriage return, if any.
351: */
352:
1.71 schwarze 353: if ('\r' == blk.buf[i] && i + 1 < blk.sz &&
1.1 schwarze 354: '\n' == blk.buf[i + 1])
355: ++i;
356: if ('\n' == blk.buf[i]) {
357: ++i;
358: ++lnn;
359: break;
360: }
361:
1.13 schwarze 362: /*
1.70 schwarze 363: * Make sure we have space for the worst
364: * case of 11 bytes: "\\[u10ffff]\0"
1.13 schwarze 365: */
366:
1.71 schwarze 367: if (pos + 11 > ln.sz)
1.13 schwarze 368: resize_buf(&ln, 256);
369:
1.25 schwarze 370: /*
1.70 schwarze 371: * Encode 8-bit input.
1.1 schwarze 372: */
373:
1.70 schwarze 374: c = blk.buf[i];
375: if (c & 0x80) {
1.71 schwarze 376: if ( ! (curp->filenc && preconv_encode(
377: &blk, &i, &ln, &pos, &curp->filenc))) {
1.89 schwarze 378: mandoc_vmsg(MANDOCERR_CHAR_BAD, curp,
379: curp->line, pos, "0x%x", c);
1.70 schwarze 380: ln.buf[pos++] = '?';
381: i++;
382: }
383: continue;
384: }
385:
386: /*
387: * Exclude control characters.
388: */
1.1 schwarze 389:
1.70 schwarze 390: if (c == 0x7f || (c < 0x20 && c != 0x09)) {
1.89 schwarze 391: mandoc_vmsg(c == 0x00 || c == 0x04 ||
392: c > 0x0a ? MANDOCERR_CHAR_BAD :
393: MANDOCERR_CHAR_UNSUPP,
394: curp, curp->line, pos, "0x%x", c);
1.1 schwarze 395: i++;
1.6 schwarze 396: ln.buf[pos++] = '?';
1.1 schwarze 397: continue;
398: }
399:
400: /* Trailing backslash = a plain char. */
401:
1.71 schwarze 402: if (blk.buf[i] != '\\' || i + 1 == blk.sz) {
1.1 schwarze 403: ln.buf[pos++] = blk.buf[i++];
404: continue;
405: }
406:
407: /*
408: * Found escape and at least one other character.
409: * When it's a newline character, skip it.
410: * When there is a carriage return in between,
411: * skip that one as well.
412: */
413:
1.71 schwarze 414: if ('\r' == blk.buf[i + 1] && i + 2 < blk.sz &&
1.1 schwarze 415: '\n' == blk.buf[i + 2])
416: ++i;
417: if ('\n' == blk.buf[i + 1]) {
418: i += 2;
419: ++lnn;
420: continue;
421: }
422:
423: if ('"' == blk.buf[i + 1] || '#' == blk.buf[i + 1]) {
424: i += 2;
425: /* Comment, skip to end of line */
1.71 schwarze 426: for (; i < blk.sz; ++i) {
1.1 schwarze 427: if ('\n' == blk.buf[i]) {
428: ++i;
429: ++lnn;
430: break;
431: }
432: }
433:
434: /* Backout trailing whitespaces */
435: for (; pos > 0; --pos) {
436: if (ln.buf[pos - 1] != ' ')
437: break;
438: if (pos > 2 && ln.buf[pos - 2] == '\\')
439: break;
440: }
441: break;
442: }
443:
1.13 schwarze 444: /* Catch escaped bogus characters. */
445:
446: c = (unsigned char) blk.buf[i+1];
447:
1.25 schwarze 448: if ( ! (isascii(c) &&
449: (isgraph(c) || isblank(c)))) {
1.89 schwarze 450: mandoc_vmsg(MANDOCERR_CHAR_BAD, curp,
1.56 schwarze 451: curp->line, pos, "0x%x", c);
1.13 schwarze 452: i += 2;
453: ln.buf[pos++] = '?';
454: continue;
455: }
456:
1.1 schwarze 457: /* Some other escape sequence, copy & cont. */
458:
459: ln.buf[pos++] = blk.buf[i++];
460: ln.buf[pos++] = blk.buf[i++];
461: }
462:
1.71 schwarze 463: if (pos >= ln.sz)
1.1 schwarze 464: resize_buf(&ln, 256);
465:
466: ln.buf[pos] = '\0';
467:
468: /*
469: * A significant amount of complexity is contained by
470: * the roff preprocessor. It's line-oriented but can be
471: * expressed on one line, so we need at times to
472: * readjust our starting point and re-run it. The roff
473: * preprocessor can also readjust the buffers with new
474: * data, so we pass them in wholesale.
475: */
476:
477: of = 0;
478:
1.4 schwarze 479: /*
480: * Maintain a lookaside buffer of all parsed lines. We
481: * only do this if mparse_keep() has been invoked (the
482: * buffer may be accessed with mparse_getkeep()).
483: */
484:
485: if (curp->secondary) {
1.25 schwarze 486: curp->secondary->buf = mandoc_realloc(
487: curp->secondary->buf,
488: curp->secondary->sz + pos + 2);
489: memcpy(curp->secondary->buf +
490: curp->secondary->sz,
491: ln.buf, pos);
1.4 schwarze 492: curp->secondary->sz += pos;
493: curp->secondary->buf
494: [curp->secondary->sz] = '\n';
495: curp->secondary->sz++;
496: curp->secondary->buf
497: [curp->secondary->sz] = '\0';
498: }
1.1 schwarze 499: rerun:
1.72 schwarze 500: rr = roff_parseln(curp->roff, curp->line, &ln, &of);
1.1 schwarze 501:
502: switch (rr) {
1.25 schwarze 503: case ROFF_REPARSE:
1.1 schwarze 504: if (REPARSE_LIMIT >= ++curp->reparse_count)
1.71 schwarze 505: mparse_buf_r(curp, ln, of, 0);
1.1 schwarze 506: else
507: mandoc_msg(MANDOCERR_ROFFLOOP, curp,
1.25 schwarze 508: curp->line, pos, NULL);
1.1 schwarze 509: pos = 0;
510: continue;
1.25 schwarze 511: case ROFF_APPEND:
1.71 schwarze 512: pos = strlen(ln.buf);
1.1 schwarze 513: continue;
1.25 schwarze 514: case ROFF_RERUN:
1.1 schwarze 515: goto rerun;
1.25 schwarze 516: case ROFF_IGN:
1.1 schwarze 517: pos = 0;
518: continue;
1.25 schwarze 519: case ROFF_SO:
1.71 schwarze 520: if ( ! (curp->options & MPARSE_SO) &&
521: (i >= blk.sz || blk.buf[i] == '\0')) {
1.23 schwarze 522: curp->sodest = mandoc_strdup(ln.buf + of);
523: free(ln.buf);
524: return;
525: }
1.4 schwarze 526: /*
527: * We remove `so' clauses from our lookaside
528: * buffer because we're going to descend into
529: * the file recursively.
530: */
1.25 schwarze 531: if (curp->secondary)
1.4 schwarze 532: curp->secondary->sz -= pos + 1;
1.88 schwarze 533: save_file = curp->file;
1.85 schwarze 534: save_child = curp->child;
535: if (mparse_open(curp, &fd, ln.buf + of) ==
1.88 schwarze 536: MANDOCLEVEL_OK) {
1.85 schwarze 537: mparse_readfd(curp, fd, ln.buf + of);
1.88 schwarze 538: curp->file = save_file;
539: } else {
540: curp->file = save_file;
1.30 schwarze 541: mandoc_vmsg(MANDOCERR_SO_FAIL,
542: curp, curp->line, pos,
543: ".so %s", ln.buf + of);
1.85 schwarze 544: ln.sz = mandoc_asprintf(&cp,
545: ".sp\nSee the file %s.\n.sp",
546: ln.buf + of);
547: free(ln.buf);
548: ln.buf = cp;
549: of = 0;
550: mparse_buf_r(curp, ln, of, 0);
1.30 schwarze 551: }
1.85 schwarze 552: curp->child = save_child;
1.1 schwarze 553: pos = 0;
554: continue;
555: default:
556: break;
557: }
558:
559: /*
560: * If input parsers have not been allocated, do so now.
1.2 schwarze 561: * We keep these instanced between parsers, but set them
1.1 schwarze 562: * locally per parse routine since we can use different
563: * parsers with each one.
564: */
565:
566: if ( ! (curp->man || curp->mdoc))
1.60 schwarze 567: choose_parser(curp);
1.1 schwarze 568:
1.25 schwarze 569: /*
1.60 schwarze 570: * Lastly, push down into the parsers themselves.
1.1 schwarze 571: * If libroff returns ROFF_TBL, then add it to the
572: * currently open parse. Since we only get here if
573: * there does exist data (see tbl_data.c), we're
574: * guaranteed that something's been allocated.
575: * Do the same for ROFF_EQN.
576: */
577:
1.76 schwarze 578: if (rr == ROFF_TBL) {
579: while ((span = roff_span(curp->roff)) != NULL)
580: if (curp->man == NULL)
581: mdoc_addspan(curp->mdoc, span);
582: else
583: man_addspan(curp->man, span);
584: } else if (rr == ROFF_EQN) {
585: if (curp->man == NULL)
586: mdoc_addeqn(curp->mdoc, roff_eqn(curp->roff));
587: else
588: man_addeqn(curp->man, roff_eqn(curp->roff));
589: } else if ((curp->man == NULL ?
590: mdoc_parseln(curp->mdoc, curp->line, ln.buf, of) :
591: man_parseln(curp->man, curp->line, ln.buf, of)) == 2)
592: break;
1.1 schwarze 593:
594: /* Temporary buffers typically are not full. */
595:
596: if (0 == start && '\0' == blk.buf[i])
597: break;
598:
599: /* Start the next input line. */
600:
601: pos = 0;
602: }
603:
604: free(ln.buf);
605: }
606:
607: static int
1.18 schwarze 608: read_whole_file(struct mparse *curp, const char *file, int fd,
609: struct buf *fb, int *with_mmap)
1.1 schwarze 610: {
611: struct stat st;
612: size_t off;
613: ssize_t ssz;
614:
615: if (-1 == fstat(fd, &st)) {
1.82 schwarze 616: perror(file);
617: exit((int)MANDOCLEVEL_SYSERR);
1.1 schwarze 618: }
619:
620: /*
621: * If we're a regular file, try just reading in the whole entry
622: * via mmap(). This is faster than reading it into blocks, and
623: * since each file is only a few bytes to begin with, I'm not
624: * concerned that this is going to tank any machines.
625: */
626:
627: if (S_ISREG(st.st_mode)) {
628: if (st.st_size >= (1U << 31)) {
1.86 schwarze 629: mandoc_msg(MANDOCERR_TOOLARGE, curp, 0, 0, NULL);
1.1 schwarze 630: return(0);
631: }
632: *with_mmap = 1;
633: fb->sz = (size_t)st.st_size;
1.15 schwarze 634: fb->buf = mmap(NULL, fb->sz, PROT_READ, MAP_SHARED, fd, 0);
1.1 schwarze 635: if (fb->buf != MAP_FAILED)
636: return(1);
637: }
638:
639: /*
640: * If this isn't a regular file (like, say, stdin), then we must
641: * go the old way and just read things in bit by bit.
642: */
643:
644: *with_mmap = 0;
645: off = 0;
646: fb->sz = 0;
647: fb->buf = NULL;
648: for (;;) {
649: if (off == fb->sz) {
650: if (fb->sz == (1U << 31)) {
1.86 schwarze 651: mandoc_msg(MANDOCERR_TOOLARGE, curp,
652: 0, 0, NULL);
1.1 schwarze 653: break;
654: }
655: resize_buf(fb, 65536);
656: }
657: ssz = read(fd, fb->buf + (int)off, fb->sz - off);
658: if (ssz == 0) {
659: fb->sz = off;
660: return(1);
661: }
662: if (ssz == -1) {
1.82 schwarze 663: perror(file);
664: exit((int)MANDOCLEVEL_SYSERR);
1.1 schwarze 665: }
666: off += (size_t)ssz;
667: }
668:
669: free(fb->buf);
670: fb->buf = NULL;
671: return(0);
672: }
673:
674: static void
675: mparse_end(struct mparse *curp)
676: {
677:
1.50 schwarze 678: if (curp->mdoc == NULL &&
679: curp->man == NULL &&
680: curp->sodest == NULL) {
681: if (curp->options & MPARSE_MDOC)
682: curp->mdoc = curp->pmdoc;
683: else {
684: if (curp->pman == NULL)
1.81 schwarze 685: curp->pman = man_alloc(
686: curp->roff, curp, curp->defos,
1.50 schwarze 687: curp->options & MPARSE_QUICK ? 1 : 0);
688: curp->man = curp->pman;
689: }
690: }
1.86 schwarze 691: if (curp->mdoc)
692: mdoc_endparse(curp->mdoc);
693: if (curp->man)
694: man_endparse(curp->man);
1.1 schwarze 695: roff_endparse(curp->roff);
696: }
697:
1.15 schwarze 698: static void
699: mparse_parse_buffer(struct mparse *curp, struct buf blk, const char *file)
1.1 schwarze 700: {
1.61 schwarze 701: struct buf *svprimary;
1.1 schwarze 702: const char *svfile;
1.71 schwarze 703: size_t offset;
1.14 schwarze 704: static int recursion_depth;
705:
706: if (64 < recursion_depth) {
707: mandoc_msg(MANDOCERR_ROFFLOOP, curp, curp->line, 0, NULL);
1.15 schwarze 708: return;
1.14 schwarze 709: }
1.1 schwarze 710:
1.15 schwarze 711: /* Line number is per-file. */
712: svfile = curp->file;
713: curp->file = file;
1.61 schwarze 714: svprimary = curp->primary;
1.59 schwarze 715: curp->primary = &blk;
1.15 schwarze 716: curp->line = 1;
717: recursion_depth++;
718:
1.70 schwarze 719: /* Skip an UTF-8 byte order mark. */
720: if (curp->filenc & MPARSE_UTF8 && blk.sz > 2 &&
721: (unsigned char)blk.buf[0] == 0xef &&
722: (unsigned char)blk.buf[1] == 0xbb &&
723: (unsigned char)blk.buf[2] == 0xbf) {
1.71 schwarze 724: offset = 3;
1.70 schwarze 725: curp->filenc &= ~MPARSE_LATIN1;
1.71 schwarze 726: } else
727: offset = 0;
1.70 schwarze 728:
1.71 schwarze 729: mparse_buf_r(curp, blk, offset, 1);
1.15 schwarze 730:
1.86 schwarze 731: if (--recursion_depth == 0)
1.15 schwarze 732: mparse_end(curp);
733:
1.61 schwarze 734: curp->primary = svprimary;
1.15 schwarze 735: curp->file = svfile;
736: }
737:
1.74 schwarze 738: /*
739: * Read the whole file into memory and call the parsers.
740: * Called recursively when an .so request is encountered.
741: */
1.15 schwarze 742: enum mandoclevel
743: mparse_readfd(struct mparse *curp, int fd, const char *file)
744: {
745: struct buf blk;
746: int with_mmap;
1.70 schwarze 747: int save_filenc;
1.1 schwarze 748:
1.67 schwarze 749: if (read_whole_file(curp, file, fd, &blk, &with_mmap)) {
1.70 schwarze 750: save_filenc = curp->filenc;
751: curp->filenc = curp->options &
752: (MPARSE_UTF8 | MPARSE_LATIN1);
1.67 schwarze 753: mparse_parse_buffer(curp, blk, file);
1.70 schwarze 754: curp->filenc = save_filenc;
1.67 schwarze 755: if (with_mmap)
756: munmap(blk.buf, blk.sz);
757: else
758: free(blk.buf);
759: }
1.1 schwarze 760:
1.74 schwarze 761: if (fd != STDIN_FILENO && close(fd) == -1)
1.1 schwarze 762: perror(file);
1.67 schwarze 763:
1.74 schwarze 764: mparse_wait(curp);
1.1 schwarze 765: return(curp->file_status);
1.58 schwarze 766: }
767:
768: enum mandoclevel
1.73 schwarze 769: mparse_open(struct mparse *curp, int *fd, const char *file)
1.58 schwarze 770: {
771: int pfd[2];
1.74 schwarze 772: int save_errno;
1.58 schwarze 773: char *cp;
774:
775: curp->file = file;
1.74 schwarze 776:
777: /* Unless zipped, try to just open the file. */
778:
1.58 schwarze 779: if ((cp = strrchr(file, '.')) == NULL ||
780: strcmp(cp + 1, "gz")) {
1.73 schwarze 781: curp->child = 0;
1.74 schwarze 782: if ((*fd = open(file, O_RDONLY)) != -1)
783: return(MANDOCLEVEL_OK);
784:
785: /* Open failed; try to append ".gz". */
786:
787: mandoc_asprintf(&cp, "%s.gz", file);
788: file = cp;
789: } else
790: cp = NULL;
791:
792: /* Before forking, make sure the file can be read. */
793:
794: save_errno = errno;
795: if (access(file, R_OK) == -1) {
796: if (cp != NULL)
797: errno = save_errno;
1.83 schwarze 798: free(cp);
799: *fd = -1;
800: curp->child = 0;
801: mandoc_msg(MANDOCERR_FILE, curp, 0, 0, strerror(errno));
802: return(MANDOCLEVEL_ERROR);
1.58 schwarze 803: }
804:
1.74 schwarze 805: /* Run gunzip(1). */
806:
1.58 schwarze 807: if (pipe(pfd) == -1) {
1.82 schwarze 808: perror("pipe");
809: exit((int)MANDOCLEVEL_SYSERR);
1.58 schwarze 810: }
811:
1.73 schwarze 812: switch (curp->child = fork()) {
1.58 schwarze 813: case -1:
1.82 schwarze 814: perror("fork");
815: exit((int)MANDOCLEVEL_SYSERR);
1.58 schwarze 816: case 0:
817: close(pfd[0]);
818: if (dup2(pfd[1], STDOUT_FILENO) == -1) {
1.82 schwarze 819: perror("dup");
820: exit((int)MANDOCLEVEL_SYSERR);
1.58 schwarze 821: }
822: execlp("gunzip", "gunzip", "-c", file, NULL);
1.82 schwarze 823: perror("exec");
824: exit((int)MANDOCLEVEL_SYSERR);
1.58 schwarze 825: default:
826: close(pfd[1]);
827: *fd = pfd[0];
828: return(MANDOCLEVEL_OK);
829: }
830: }
831:
832: enum mandoclevel
1.73 schwarze 833: mparse_wait(struct mparse *curp)
1.58 schwarze 834: {
835: int status;
836:
1.73 schwarze 837: if (curp->child == 0)
838: return(MANDOCLEVEL_OK);
839:
840: if (waitpid(curp->child, &status, 0) == -1) {
1.82 schwarze 841: perror("wait");
842: exit((int)MANDOCLEVEL_SYSERR);
1.58 schwarze 843: }
844: if (WIFSIGNALED(status)) {
1.83 schwarze 845: mandoc_vmsg(MANDOCERR_FILE, curp, 0, 0,
846: "gunzip died from signal %d", WTERMSIG(status));
847: return(MANDOCLEVEL_ERROR);
1.58 schwarze 848: }
849: if (WEXITSTATUS(status)) {
1.83 schwarze 850: mandoc_vmsg(MANDOCERR_FILE, curp, 0, 0,
851: "gunzip failed with code %d", WEXITSTATUS(status));
852: return(MANDOCLEVEL_ERROR);
1.58 schwarze 853: }
854: return(MANDOCLEVEL_OK);
1.1 schwarze 855: }
856:
857: struct mparse *
1.69 schwarze 858: mparse_alloc(int options, enum mandoclevel wlevel, mandocmsg mmsg,
859: const struct mchars *mchars, const char *defos)
1.1 schwarze 860: {
861: struct mparse *curp;
862:
863: curp = mandoc_calloc(1, sizeof(struct mparse));
864:
1.22 schwarze 865: curp->options = options;
1.1 schwarze 866: curp->wlevel = wlevel;
867: curp->mmsg = mmsg;
1.7 schwarze 868: curp->defos = defos;
1.1 schwarze 869:
1.69 schwarze 870: curp->mchars = mchars;
871: curp->roff = roff_alloc(curp, curp->mchars, options);
1.50 schwarze 872: if (curp->options & MPARSE_MDOC)
873: curp->pmdoc = mdoc_alloc(
874: curp->roff, curp, curp->defos,
875: curp->options & MPARSE_QUICK ? 1 : 0);
876: if (curp->options & MPARSE_MAN)
1.81 schwarze 877: curp->pman = man_alloc(
878: curp->roff, curp, curp->defos,
1.50 schwarze 879: curp->options & MPARSE_QUICK ? 1 : 0);
880:
1.1 schwarze 881: return(curp);
882: }
883:
884: void
885: mparse_reset(struct mparse *curp)
886: {
887:
888: roff_reset(curp->roff);
889:
890: if (curp->mdoc)
891: mdoc_reset(curp->mdoc);
892: if (curp->man)
893: man_reset(curp->man);
1.4 schwarze 894: if (curp->secondary)
895: curp->secondary->sz = 0;
1.1 schwarze 896:
897: curp->file_status = MANDOCLEVEL_OK;
898: curp->mdoc = NULL;
899: curp->man = NULL;
1.23 schwarze 900:
901: free(curp->sodest);
902: curp->sodest = NULL;
1.1 schwarze 903: }
904:
905: void
906: mparse_free(struct mparse *curp)
907: {
908:
909: if (curp->pmdoc)
910: mdoc_free(curp->pmdoc);
911: if (curp->pman)
912: man_free(curp->pman);
913: if (curp->roff)
914: roff_free(curp->roff);
1.4 schwarze 915: if (curp->secondary)
916: free(curp->secondary->buf);
1.1 schwarze 917:
1.4 schwarze 918: free(curp->secondary);
1.23 schwarze 919: free(curp->sodest);
1.1 schwarze 920: free(curp);
921: }
922:
923: void
1.23 schwarze 924: mparse_result(struct mparse *curp,
925: struct mdoc **mdoc, struct man **man, char **sodest)
1.1 schwarze 926: {
927:
1.23 schwarze 928: if (sodest && NULL != (*sodest = curp->sodest)) {
929: *mdoc = NULL;
930: *man = NULL;
931: return;
932: }
1.1 schwarze 933: if (mdoc)
934: *mdoc = curp->mdoc;
935: if (man)
936: *man = curp->man;
937: }
938:
939: void
940: mandoc_vmsg(enum mandocerr t, struct mparse *m,
941: int ln, int pos, const char *fmt, ...)
942: {
943: char buf[256];
944: va_list ap;
945:
946: va_start(ap, fmt);
1.26 schwarze 947: (void)vsnprintf(buf, sizeof(buf), fmt, ap);
1.1 schwarze 948: va_end(ap);
949:
950: mandoc_msg(t, m, ln, pos, buf);
951: }
952:
953: void
1.25 schwarze 954: mandoc_msg(enum mandocerr er, struct mparse *m,
1.1 schwarze 955: int ln, int col, const char *msg)
956: {
957: enum mandoclevel level;
958:
1.87 schwarze 959: level = MANDOCLEVEL_UNSUPP;
1.1 schwarze 960: while (er < mandoclimits[level])
961: level--;
962:
1.83 schwarze 963: if (level < m->wlevel && er != MANDOCERR_FILE)
1.1 schwarze 964: return;
965:
966: if (m->mmsg)
967: (*m->mmsg)(er, level, m->file, ln, col, msg);
968:
969: if (m->file_status < level)
970: m->file_status = level;
971: }
972:
973: const char *
974: mparse_strerror(enum mandocerr er)
975: {
976:
977: return(mandocerrs[er]);
978: }
979:
980: const char *
981: mparse_strlevel(enum mandoclevel lvl)
982: {
983: return(mandoclevels[lvl]);
1.4 schwarze 984: }
985:
986: void
987: mparse_keep(struct mparse *p)
988: {
989:
990: assert(NULL == p->secondary);
991: p->secondary = mandoc_calloc(1, sizeof(struct buf));
992: }
993:
994: const char *
995: mparse_getkeep(const struct mparse *p)
996: {
997:
998: assert(p->secondary);
999: return(p->secondary->sz ? p->secondary->buf : NULL);
1.1 schwarze 1000: }
![[BACK]](/icons/back.gif){kind=icon}
Return to read.c CVS log ![[TXT]](/icons/text.gif){kind=icon}
![[DIR]](/icons/dir.gif){kind=icon}
Up to [local] / src / usr.bin / mandoc