Annotation of mandoc/read.c, Revision 1.110
1.110 ! schwarze 1: /* $Id: read.c,v 1.109 2015/01/14 22:58:16 schwarze Exp $ */
1.1 kristaps 2: /*
3: * Copyright (c) 2008, 2009, 2010, 2011 Kristaps Dzonsons <kristaps@bsd.lv>
1.107 schwarze 4: * Copyright (c) 2010-2015 Ingo Schwarze <schwarze@openbsd.org>
1.42 schwarze 5: * Copyright (c) 2010, 2012 Joerg Sonnenberger <joerg@netbsd.org>
1.1 kristaps 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: */
1.11 kristaps 19: #include "config.h"
20:
1.80 schwarze 21: #include <sys/types.h>
1.81 schwarze 22: #if HAVE_MMAP
1.82 schwarze 23: #include <sys/mman.h>
1.80 schwarze 24: #include <sys/stat.h>
1.15 kristaps 25: #endif
1.82 schwarze 26: #include <sys/wait.h>
1.1 kristaps 27:
28: #include <assert.h>
29: #include <ctype.h>
1.40 schwarze 30: #include <errno.h>
1.1 kristaps 31: #include <fcntl.h>
1.3 kristaps 32: #include <stdarg.h>
1.28 joerg 33: #include <stdint.h>
1.1 kristaps 34: #include <stdio.h>
35: #include <stdlib.h>
36: #include <string.h>
37: #include <unistd.h>
38:
39: #include "mandoc.h"
1.46 schwarze 40: #include "mandoc_aux.h"
1.3 kristaps 41: #include "libmandoc.h"
1.1 kristaps 42: #include "mdoc.h"
43: #include "man.h"
44:
45: #define REPARSE_LIMIT 1000
46:
47: struct mparse {
48: struct man *pman; /* persistent man parser */
49: struct mdoc *pmdoc; /* persistent mdoc parser */
50: struct man *man; /* man parser */
51: struct mdoc *mdoc; /* mdoc parser */
52: struct roff *roff; /* roff parser (!NULL) */
1.94 schwarze 53: const struct mchars *mchars; /* character table */
1.45 schwarze 54: char *sodest; /* filename pointed to by .so */
1.83 schwarze 55: const char *file; /* filename of current input file */
56: struct buf *primary; /* buffer currently being parsed */
57: struct buf *secondary; /* preprocessed copy of input */
58: const char *defos; /* default operating system */
59: mandocmsg mmsg; /* warning/error message handler */
60: enum mandoclevel file_status; /* status of current parse */
61: enum mandoclevel wlevel; /* ignore messages below this */
62: int options; /* parser options */
1.93 schwarze 63: int filenc; /* encoding of the current file */
1.1 kristaps 64: int reparse_count; /* finite interp. stack */
1.83 schwarze 65: int line; /* line number in the file */
1.97 schwarze 66: pid_t child; /* the gunzip(1) process */
1.1 kristaps 67: };
68:
1.84 schwarze 69: static void choose_parser(struct mparse *);
1.1 kristaps 70: static void resize_buf(struct buf *, size_t);
1.95 schwarze 71: static void mparse_buf_r(struct mparse *, struct buf, size_t, int);
1.40 schwarze 72: static int read_whole_file(struct mparse *, const char *, int,
73: struct buf *, int *);
1.1 kristaps 74: static void mparse_end(struct mparse *);
1.37 schwarze 75: static void mparse_parse_buffer(struct mparse *, struct buf,
76: const char *);
1.1 kristaps 77:
1.3 kristaps 78: static const enum mandocerr mandoclimits[MANDOCLEVEL_MAX] = {
79: MANDOCERR_OK,
80: MANDOCERR_WARNING,
81: MANDOCERR_WARNING,
82: MANDOCERR_ERROR,
83: MANDOCERR_FATAL,
84: MANDOCERR_MAX,
85: MANDOCERR_MAX
86: };
87:
1.7 kristaps 88: static const char * const mandocerrs[MANDOCERR_MAX] = {
89: "ok",
90:
91: "generic warning",
92:
93: /* related to the prologue */
1.79 schwarze 94: "missing manual title, using UNTITLED",
95: "missing manual title, using \"\"",
1.54 schwarze 96: "lower case character in document title",
1.79 schwarze 97: "missing manual section, using \"\"",
1.7 kristaps 98: "unknown manual section",
1.54 schwarze 99: "missing date, using today's date",
1.7 kristaps 100: "cannot parse date, using it verbatim",
1.79 schwarze 101: "missing Os macro, using \"\"",
102: "duplicate prologue macro",
103: "late prologue macro",
104: "skipping late title macro",
1.7 kristaps 105: "prologue macros out of order",
106:
107: /* related to document structure */
108: ".so is fragile, better use ln(1)",
1.50 schwarze 109: "no document body",
1.54 schwarze 110: "content before first section header",
111: "first section is not \"NAME\"",
1.7 kristaps 112: "bad NAME section contents",
113: "sections out of conventional order",
1.54 schwarze 114: "duplicate section title",
115: "unexpected section",
1.87 schwarze 116: "unusual Xr order",
117: "unusual Xr punctuation",
1.86 schwarze 118: "AUTHORS section without An macro",
1.7 kristaps 119:
120: /* related to macros and nesting */
1.55 schwarze 121: "obsolete macro",
1.102 schwarze 122: "macro neither callable nor escaped",
1.7 kristaps 123: "skipping paragraph macro",
1.31 schwarze 124: "moving paragraph macro out of list",
1.7 kristaps 125: "skipping no-space macro",
126: "blocks badly nested",
127: "nested displays are not portable",
1.57 schwarze 128: "moving content out of list",
129: ".Vt block has child macro",
1.78 schwarze 130: "fill mode already enabled, skipping",
131: "fill mode already disabled, skipping",
1.7 kristaps 132: "line scope broken",
133:
134: /* related to missing macro arguments */
1.58 schwarze 135: "skipping empty request",
136: "conditional request controls empty scope",
1.7 kristaps 137: "skipping empty macro",
1.62 schwarze 138: "empty argument, using 0n",
1.7 kristaps 139: "argument count wrong",
1.60 schwarze 140: "missing display type, using -ragged",
141: "list type is not the first argument",
142: "missing -width in -tag list, using 8n",
1.78 schwarze 143: "missing utility name, using \"\"",
1.60 schwarze 144: "empty head in list item",
145: "empty list item",
1.61 schwarze 146: "missing font type, using \\fR",
147: "unknown font type, using \\fR",
1.103 schwarze 148: "nothing follows prefix",
1.60 schwarze 149: "missing -std argument, adding it",
1.90 schwarze 150: "missing eqn box, using \"\"",
1.7 kristaps 151:
152: /* related to bad macro arguments */
1.64 schwarze 153: "unterminated quoted argument",
1.7 kristaps 154: "duplicate argument",
1.76 schwarze 155: "skipping duplicate argument",
1.63 schwarze 156: "skipping duplicate display type",
157: "skipping duplicate list type",
1.76 schwarze 158: "skipping -width argument",
1.7 kristaps 159: "unknown AT&T UNIX version",
1.88 schwarze 160: "comma in function argument",
1.89 schwarze 161: "parenthesis in function name",
1.67 schwarze 162: "invalid content in Rs block",
1.63 schwarze 163: "invalid Boolean argument",
164: "unknown font, skipping request",
1.7 kristaps 165:
166: /* related to plain text */
1.64 schwarze 167: "blank line in fill mode, using .sp",
168: "tab in filled text",
169: "whitespace at end of input line",
1.7 kristaps 170: "bad comment style",
1.64 schwarze 171: "invalid escape sequence",
172: "undefined string, using \"\"",
1.16 kristaps 173:
1.7 kristaps 174: "generic error",
1.17 kristaps 175:
176: /* related to equations */
1.20 kristaps 177: "unexpected equation scope closure",
178: "equation scope open on exit",
1.21 kristaps 179: "overlapping equation scopes",
180: "unexpected end of equation",
1.7 kristaps 181:
182: /* related to tables */
183: "bad table syntax",
184: "bad table option",
185: "bad table layout",
186: "no table layout cells specified",
187: "no table data cells specified",
188: "ignore data in cell",
189: "data block still open",
190: "ignoring extra data cells",
1.105 schwarze 191: "ignoring macro in table",
1.7 kristaps 192:
1.68 schwarze 193: /* related to document structure and macros */
1.108 schwarze 194: NULL,
1.7 kristaps 195: "input stack limit exceeded, infinite loop?",
196: "skipping bad character",
1.68 schwarze 197: "skipping unknown macro",
1.73 schwarze 198: "skipping item outside list",
1.68 schwarze 199: "skipping column outside column list",
200: "skipping end of block that is not open",
201: "inserting missing end of block",
202: "appending missing end of block",
203:
204: /* related to request and macro arguments */
1.7 kristaps 205: "escaped character not allowed in a name",
206: "argument count wrong",
1.99 schwarze 207: "NOT IMPLEMENTED: Bd -file",
1.71 schwarze 208: "missing list type, using -item",
1.70 schwarze 209: "missing manual name, using \"\"",
1.71 schwarze 210: "uname(3) system call failed, using UNKNOWN",
1.63 schwarze 211: "unknown standard specifier",
1.71 schwarze 212: "skipping request without numeric argument",
1.109 schwarze 213: "NOT IMPLEMENTED: .so with absolute path or \"..\"",
1.110 ! schwarze 214: ".so request failed",
1.61 schwarze 215: "skipping all arguments",
216: "skipping excess arguments",
1.92 kristaps 217: "divide by zero",
1.7 kristaps 218:
219: "generic fatal error",
220:
1.40 schwarze 221: "input too large",
1.7 kristaps 222: };
223:
224: static const char * const mandoclevels[MANDOCLEVEL_MAX] = {
225: "SUCCESS",
226: "RESERVED",
227: "WARNING",
228: "ERROR",
229: "FATAL",
230: "BADARG",
231: "SYSERR"
232: };
233:
1.47 schwarze 234:
1.1 kristaps 235: static void
236: resize_buf(struct buf *buf, size_t initial)
237: {
238:
239: buf->sz = buf->sz > initial/2 ? 2 * buf->sz : initial;
240: buf->buf = mandoc_realloc(buf->buf, buf->sz);
241: }
242:
243: static void
1.84 schwarze 244: choose_parser(struct mparse *curp)
1.1 kristaps 245: {
1.83 schwarze 246: char *cp, *ep;
247: int format;
1.1 kristaps 248:
1.83 schwarze 249: /*
250: * If neither command line arguments -mdoc or -man select
251: * a parser nor the roff parser found a .Dd or .TH macro
252: * yet, look ahead in the main input buffer.
253: */
254:
255: if ((format = roff_getformat(curp->roff)) == 0) {
256: cp = curp->primary->buf;
257: ep = cp + curp->primary->sz;
258: while (cp < ep) {
1.85 schwarze 259: if (*cp == '.' || *cp == '\'') {
1.83 schwarze 260: cp++;
261: if (cp[0] == 'D' && cp[1] == 'd') {
262: format = MPARSE_MDOC;
263: break;
264: }
265: if (cp[0] == 'T' && cp[1] == 'H') {
266: format = MPARSE_MAN;
267: break;
268: }
269: }
270: cp = memchr(cp, '\n', ep - cp);
271: if (cp == NULL)
272: break;
273: cp++;
274: }
1.1 kristaps 275: }
276:
1.83 schwarze 277: if (format == MPARSE_MDOC) {
1.47 schwarze 278: if (NULL == curp->pmdoc)
1.44 schwarze 279: curp->pmdoc = mdoc_alloc(
280: curp->roff, curp, curp->defos,
281: MPARSE_QUICK & curp->options ? 1 : 0);
1.1 kristaps 282: assert(curp->pmdoc);
283: curp->mdoc = curp->pmdoc;
284: return;
1.47 schwarze 285: }
1.1 kristaps 286:
1.83 schwarze 287: /* Fall back to man(7) as a last resort. */
288:
1.47 schwarze 289: if (NULL == curp->pman)
1.106 schwarze 290: curp->pman = man_alloc(
291: curp->roff, curp, curp->defos,
1.44 schwarze 292: MPARSE_QUICK & curp->options ? 1 : 0);
1.1 kristaps 293: assert(curp->pman);
294: curp->man = curp->pman;
295: }
296:
297: /*
1.95 schwarze 298: * Main parse routine for a buffer.
299: * It assumes encoding and line numbering are already set up.
300: * It can recurse directly (for invocations of user-defined
301: * macros, inline equations, and input line traps)
302: * and indirectly (for .so file inclusion).
1.1 kristaps 303: */
304: static void
1.95 schwarze 305: mparse_buf_r(struct mparse *curp, struct buf blk, size_t i, int start)
1.1 kristaps 306: {
307: const struct tbl_span *span;
308: struct buf ln;
1.110 ! schwarze 309: char *cp;
1.95 schwarze 310: size_t pos; /* byte number in the ln buffer */
1.1 kristaps 311: enum rofferr rr;
1.100 schwarze 312: int of;
1.1 kristaps 313: int lnn; /* line number in the real file */
1.110 ! schwarze 314: int fd;
! 315: pid_t save_child;
1.1 kristaps 316: unsigned char c;
317:
1.95 schwarze 318: memset(&ln, 0, sizeof(ln));
1.1 kristaps 319:
1.47 schwarze 320: lnn = curp->line;
321: pos = 0;
1.1 kristaps 322:
1.95 schwarze 323: while (i < blk.sz) {
1.1 kristaps 324: if (0 == pos && '\0' == blk.buf[i])
325: break;
326:
327: if (start) {
328: curp->line = lnn;
329: curp->reparse_count = 0;
1.93 schwarze 330:
331: if (lnn < 3 &&
332: curp->filenc & MPARSE_UTF8 &&
1.95 schwarze 333: curp->filenc & MPARSE_LATIN1)
334: curp->filenc = preconv_cue(&blk, i);
1.1 kristaps 335: }
336:
1.95 schwarze 337: while (i < blk.sz && (start || blk.buf[i] != '\0')) {
1.1 kristaps 338:
339: /*
340: * When finding an unescaped newline character,
341: * leave the character loop to process the line.
342: * Skip a preceding carriage return, if any.
343: */
344:
1.95 schwarze 345: if ('\r' == blk.buf[i] && i + 1 < blk.sz &&
1.1 kristaps 346: '\n' == blk.buf[i + 1])
347: ++i;
348: if ('\n' == blk.buf[i]) {
349: ++i;
350: ++lnn;
351: break;
352: }
353:
1.35 schwarze 354: /*
1.93 schwarze 355: * Make sure we have space for the worst
356: * case of 11 bytes: "\\[u10ffff]\0"
1.35 schwarze 357: */
358:
1.95 schwarze 359: if (pos + 11 > ln.sz)
1.35 schwarze 360: resize_buf(&ln, 256);
361:
1.47 schwarze 362: /*
1.93 schwarze 363: * Encode 8-bit input.
1.1 kristaps 364: */
365:
1.93 schwarze 366: c = blk.buf[i];
367: if (c & 0x80) {
1.95 schwarze 368: if ( ! (curp->filenc && preconv_encode(
369: &blk, &i, &ln, &pos, &curp->filenc))) {
1.93 schwarze 370: mandoc_vmsg(MANDOCERR_BADCHAR,
371: curp, curp->line, pos,
372: "0x%x", c);
373: ln.buf[pos++] = '?';
374: i++;
375: }
376: continue;
377: }
378:
379: /*
380: * Exclude control characters.
381: */
1.1 kristaps 382:
1.93 schwarze 383: if (c == 0x7f || (c < 0x20 && c != 0x09)) {
1.78 schwarze 384: mandoc_vmsg(MANDOCERR_BADCHAR, curp,
385: curp->line, pos, "0x%x", c);
1.1 kristaps 386: i++;
1.27 joerg 387: ln.buf[pos++] = '?';
1.1 kristaps 388: continue;
389: }
390:
391: /* Trailing backslash = a plain char. */
392:
1.95 schwarze 393: if (blk.buf[i] != '\\' || i + 1 == blk.sz) {
1.1 kristaps 394: ln.buf[pos++] = blk.buf[i++];
395: continue;
396: }
397:
398: /*
399: * Found escape and at least one other character.
400: * When it's a newline character, skip it.
401: * When there is a carriage return in between,
402: * skip that one as well.
403: */
404:
1.95 schwarze 405: if ('\r' == blk.buf[i + 1] && i + 2 < blk.sz &&
1.1 kristaps 406: '\n' == blk.buf[i + 2])
407: ++i;
408: if ('\n' == blk.buf[i + 1]) {
409: i += 2;
410: ++lnn;
411: continue;
412: }
413:
1.13 kristaps 414: if ('"' == blk.buf[i + 1] || '#' == blk.buf[i + 1]) {
1.1 kristaps 415: i += 2;
416: /* Comment, skip to end of line */
1.95 schwarze 417: for (; i < blk.sz; ++i) {
1.1 kristaps 418: if ('\n' == blk.buf[i]) {
419: ++i;
420: ++lnn;
421: break;
422: }
423: }
424:
425: /* Backout trailing whitespaces */
426: for (; pos > 0; --pos) {
427: if (ln.buf[pos - 1] != ' ')
428: break;
429: if (pos > 2 && ln.buf[pos - 2] == '\\')
430: break;
431: }
432: break;
433: }
434:
1.35 schwarze 435: /* Catch escaped bogus characters. */
436:
437: c = (unsigned char) blk.buf[i+1];
438:
1.47 schwarze 439: if ( ! (isascii(c) &&
440: (isgraph(c) || isblank(c)))) {
1.78 schwarze 441: mandoc_vmsg(MANDOCERR_BADCHAR, curp,
442: curp->line, pos, "0x%x", c);
1.35 schwarze 443: i += 2;
444: ln.buf[pos++] = '?';
445: continue;
446: }
447:
1.1 kristaps 448: /* Some other escape sequence, copy & cont. */
449:
450: ln.buf[pos++] = blk.buf[i++];
451: ln.buf[pos++] = blk.buf[i++];
452: }
453:
1.95 schwarze 454: if (pos >= ln.sz)
1.1 kristaps 455: resize_buf(&ln, 256);
456:
457: ln.buf[pos] = '\0';
458:
459: /*
460: * A significant amount of complexity is contained by
461: * the roff preprocessor. It's line-oriented but can be
462: * expressed on one line, so we need at times to
463: * readjust our starting point and re-run it. The roff
464: * preprocessor can also readjust the buffers with new
465: * data, so we pass them in wholesale.
466: */
467:
468: of = 0;
469:
1.24 kristaps 470: /*
471: * Maintain a lookaside buffer of all parsed lines. We
472: * only do this if mparse_keep() has been invoked (the
473: * buffer may be accessed with mparse_getkeep()).
474: */
475:
476: if (curp->secondary) {
1.47 schwarze 477: curp->secondary->buf = mandoc_realloc(
478: curp->secondary->buf,
479: curp->secondary->sz + pos + 2);
480: memcpy(curp->secondary->buf +
481: curp->secondary->sz,
482: ln.buf, pos);
1.24 kristaps 483: curp->secondary->sz += pos;
484: curp->secondary->buf
485: [curp->secondary->sz] = '\n';
486: curp->secondary->sz++;
487: curp->secondary->buf
488: [curp->secondary->sz] = '\0';
489: }
1.1 kristaps 490: rerun:
1.96 schwarze 491: rr = roff_parseln(curp->roff, curp->line, &ln, &of);
1.1 kristaps 492:
493: switch (rr) {
1.47 schwarze 494: case ROFF_REPARSE:
1.1 kristaps 495: if (REPARSE_LIMIT >= ++curp->reparse_count)
1.95 schwarze 496: mparse_buf_r(curp, ln, of, 0);
1.1 kristaps 497: else
1.3 kristaps 498: mandoc_msg(MANDOCERR_ROFFLOOP, curp,
1.47 schwarze 499: curp->line, pos, NULL);
1.1 kristaps 500: pos = 0;
501: continue;
1.47 schwarze 502: case ROFF_APPEND:
1.95 schwarze 503: pos = strlen(ln.buf);
1.1 kristaps 504: continue;
1.47 schwarze 505: case ROFF_RERUN:
1.1 kristaps 506: goto rerun;
1.47 schwarze 507: case ROFF_IGN:
1.1 kristaps 508: pos = 0;
509: continue;
1.47 schwarze 510: case ROFF_SO:
1.95 schwarze 511: if ( ! (curp->options & MPARSE_SO) &&
512: (i >= blk.sz || blk.buf[i] == '\0')) {
1.45 schwarze 513: curp->sodest = mandoc_strdup(ln.buf + of);
514: free(ln.buf);
515: return;
516: }
1.24 kristaps 517: /*
518: * We remove `so' clauses from our lookaside
519: * buffer because we're going to descend into
520: * the file recursively.
521: */
1.47 schwarze 522: if (curp->secondary)
1.25 kristaps 523: curp->secondary->sz -= pos + 1;
1.110 ! schwarze 524: save_child = curp->child;
! 525: if (mparse_open(curp, &fd, ln.buf + of) ==
! 526: MANDOCLEVEL_OK)
! 527: mparse_readfd(curp, fd, ln.buf + of);
! 528: else {
1.52 schwarze 529: mandoc_vmsg(MANDOCERR_SO_FAIL,
530: curp, curp->line, pos,
531: ".so %s", ln.buf + of);
1.110 ! schwarze 532: ln.sz = mandoc_asprintf(&cp,
! 533: ".sp\nSee the file %s.\n.sp",
! 534: ln.buf + of);
! 535: free(ln.buf);
! 536: ln.buf = cp;
! 537: of = 0;
! 538: mparse_buf_r(curp, ln, of, 0);
1.52 schwarze 539: }
1.110 ! schwarze 540: curp->child = save_child;
1.1 kristaps 541: pos = 0;
542: continue;
543: default:
544: break;
545: }
546:
547: /*
548: * If we encounter errors in the recursive parse, make
549: * sure we don't continue parsing.
550: */
551:
552: if (MANDOCLEVEL_FATAL <= curp->file_status)
553: break;
554:
555: /*
556: * If input parsers have not been allocated, do so now.
1.14 kristaps 557: * We keep these instanced between parsers, but set them
1.1 kristaps 558: * locally per parse routine since we can use different
559: * parsers with each one.
560: */
561:
562: if ( ! (curp->man || curp->mdoc))
1.84 schwarze 563: choose_parser(curp);
1.1 kristaps 564:
1.47 schwarze 565: /*
1.84 schwarze 566: * Lastly, push down into the parsers themselves.
1.1 kristaps 567: * If libroff returns ROFF_TBL, then add it to the
568: * currently open parse. Since we only get here if
569: * there does exist data (see tbl_data.c), we're
570: * guaranteed that something's been allocated.
571: * Do the same for ROFF_EQN.
572: */
573:
1.100 schwarze 574: if (rr == ROFF_TBL) {
575: while ((span = roff_span(curp->roff)) != NULL)
576: if (curp->man == NULL)
577: mdoc_addspan(curp->mdoc, span);
578: else
579: man_addspan(curp->man, span);
580: } else if (rr == ROFF_EQN) {
581: if (curp->man == NULL)
582: mdoc_addeqn(curp->mdoc, roff_eqn(curp->roff));
583: else
584: man_addeqn(curp->man, roff_eqn(curp->roff));
585: } else if ((curp->man == NULL ?
586: mdoc_parseln(curp->mdoc, curp->line, ln.buf, of) :
587: man_parseln(curp->man, curp->line, ln.buf, of)) == 2)
588: break;
1.1 kristaps 589:
590: /* Temporary buffers typically are not full. */
591:
592: if (0 == start && '\0' == blk.buf[i])
593: break;
594:
595: /* Start the next input line. */
596:
597: pos = 0;
598: }
599:
600: free(ln.buf);
601: }
602:
603: static int
1.40 schwarze 604: read_whole_file(struct mparse *curp, const char *file, int fd,
605: struct buf *fb, int *with_mmap)
1.1 kristaps 606: {
607: size_t off;
608: ssize_t ssz;
609:
1.81 schwarze 610: #if HAVE_MMAP
1.15 kristaps 611: struct stat st;
1.1 kristaps 612: if (-1 == fstat(fd, &st)) {
1.107 schwarze 613: perror(file);
614: exit((int)MANDOCLEVEL_SYSERR);
1.1 kristaps 615: }
616:
617: /*
618: * If we're a regular file, try just reading in the whole entry
619: * via mmap(). This is faster than reading it into blocks, and
620: * since each file is only a few bytes to begin with, I'm not
621: * concerned that this is going to tank any machines.
622: */
623:
624: if (S_ISREG(st.st_mode)) {
625: if (st.st_size >= (1U << 31)) {
1.40 schwarze 626: curp->file_status = MANDOCLEVEL_FATAL;
627: if (curp->mmsg)
628: (*curp->mmsg)(MANDOCERR_TOOLARGE,
629: curp->file_status, file, 0, 0, NULL);
1.1 kristaps 630: return(0);
631: }
632: *with_mmap = 1;
633: fb->sz = (size_t)st.st_size;
1.37 schwarze 634: fb->buf = mmap(NULL, fb->sz, PROT_READ, MAP_SHARED, fd, 0);
1.1 kristaps 635: if (fb->buf != MAP_FAILED)
636: return(1);
637: }
1.15 kristaps 638: #endif
1.1 kristaps 639:
640: /*
641: * If this isn't a regular file (like, say, stdin), then we must
642: * go the old way and just read things in bit by bit.
643: */
644:
645: *with_mmap = 0;
646: off = 0;
647: fb->sz = 0;
648: fb->buf = NULL;
649: for (;;) {
650: if (off == fb->sz) {
651: if (fb->sz == (1U << 31)) {
1.40 schwarze 652: curp->file_status = MANDOCLEVEL_FATAL;
653: if (curp->mmsg)
654: (*curp->mmsg)(MANDOCERR_TOOLARGE,
655: curp->file_status,
656: file, 0, 0, NULL);
1.1 kristaps 657: break;
658: }
659: resize_buf(fb, 65536);
660: }
661: ssz = read(fd, fb->buf + (int)off, fb->sz - off);
662: if (ssz == 0) {
663: fb->sz = off;
664: return(1);
665: }
666: if (ssz == -1) {
1.107 schwarze 667: perror(file);
668: exit((int)MANDOCLEVEL_SYSERR);
1.1 kristaps 669: }
670: off += (size_t)ssz;
671: }
672:
673: free(fb->buf);
674: fb->buf = NULL;
675: return(0);
676: }
677:
678: static void
679: mparse_end(struct mparse *curp)
680: {
681:
682: if (MANDOCLEVEL_FATAL <= curp->file_status)
683: return;
684:
1.72 schwarze 685: if (curp->mdoc == NULL &&
686: curp->man == NULL &&
687: curp->sodest == NULL) {
688: if (curp->options & MPARSE_MDOC)
689: curp->mdoc = curp->pmdoc;
690: else {
691: if (curp->pman == NULL)
1.106 schwarze 692: curp->pman = man_alloc(
693: curp->roff, curp, curp->defos,
1.72 schwarze 694: curp->options & MPARSE_QUICK ? 1 : 0);
695: curp->man = curp->pman;
696: }
697: }
698:
1.1 kristaps 699: if (curp->mdoc && ! mdoc_endparse(curp->mdoc)) {
700: assert(MANDOCLEVEL_FATAL <= curp->file_status);
701: return;
702: }
703:
704: if (curp->man && ! man_endparse(curp->man)) {
705: assert(MANDOCLEVEL_FATAL <= curp->file_status);
706: return;
707: }
708:
709: roff_endparse(curp->roff);
710: }
711:
712: static void
1.36 schwarze 713: mparse_parse_buffer(struct mparse *curp, struct buf blk, const char *file)
1.28 joerg 714: {
1.85 schwarze 715: struct buf *svprimary;
1.28 joerg 716: const char *svfile;
1.95 schwarze 717: size_t offset;
1.36 schwarze 718: static int recursion_depth;
719:
720: if (64 < recursion_depth) {
721: mandoc_msg(MANDOCERR_ROFFLOOP, curp, curp->line, 0, NULL);
722: return;
723: }
1.28 joerg 724:
725: /* Line number is per-file. */
726: svfile = curp->file;
727: curp->file = file;
1.85 schwarze 728: svprimary = curp->primary;
1.83 schwarze 729: curp->primary = &blk;
1.28 joerg 730: curp->line = 1;
1.36 schwarze 731: recursion_depth++;
1.28 joerg 732:
1.93 schwarze 733: /* Skip an UTF-8 byte order mark. */
734: if (curp->filenc & MPARSE_UTF8 && blk.sz > 2 &&
735: (unsigned char)blk.buf[0] == 0xef &&
736: (unsigned char)blk.buf[1] == 0xbb &&
737: (unsigned char)blk.buf[2] == 0xbf) {
1.95 schwarze 738: offset = 3;
1.93 schwarze 739: curp->filenc &= ~MPARSE_LATIN1;
1.95 schwarze 740: } else
741: offset = 0;
1.93 schwarze 742:
1.95 schwarze 743: mparse_buf_r(curp, blk, offset, 1);
1.28 joerg 744:
1.36 schwarze 745: if (0 == --recursion_depth && MANDOCLEVEL_FATAL > curp->file_status)
1.28 joerg 746: mparse_end(curp);
747:
1.85 schwarze 748: curp->primary = svprimary;
1.28 joerg 749: curp->file = svfile;
750: }
751:
752: enum mandoclevel
1.104 schwarze 753: mparse_readmem(struct mparse *curp, void *buf, size_t len,
1.28 joerg 754: const char *file)
755: {
756: struct buf blk;
757:
1.104 schwarze 758: blk.buf = buf;
1.28 joerg 759: blk.sz = len;
760:
1.36 schwarze 761: mparse_parse_buffer(curp, blk, file);
1.28 joerg 762: return(curp->file_status);
763: }
764:
1.98 schwarze 765: /*
766: * Read the whole file into memory and call the parsers.
767: * Called recursively when an .so request is encountered.
768: */
1.36 schwarze 769: enum mandoclevel
770: mparse_readfd(struct mparse *curp, int fd, const char *file)
1.1 kristaps 771: {
1.28 joerg 772: struct buf blk;
773: int with_mmap;
1.93 schwarze 774: int save_filenc;
1.1 kristaps 775:
1.91 schwarze 776: if (read_whole_file(curp, file, fd, &blk, &with_mmap)) {
1.93 schwarze 777: save_filenc = curp->filenc;
778: curp->filenc = curp->options &
779: (MPARSE_UTF8 | MPARSE_LATIN1);
1.91 schwarze 780: mparse_parse_buffer(curp, blk, file);
1.93 schwarze 781: curp->filenc = save_filenc;
1.81 schwarze 782: #if HAVE_MMAP
1.91 schwarze 783: if (with_mmap)
784: munmap(blk.buf, blk.sz);
785: else
1.28 joerg 786: #endif
1.91 schwarze 787: free(blk.buf);
788: }
1.1 kristaps 789:
1.98 schwarze 790: if (fd != STDIN_FILENO && close(fd) == -1)
1.1 kristaps 791: perror(file);
1.91 schwarze 792:
1.98 schwarze 793: mparse_wait(curp);
1.1 kristaps 794: return(curp->file_status);
1.82 schwarze 795: }
796:
797: enum mandoclevel
1.97 schwarze 798: mparse_open(struct mparse *curp, int *fd, const char *file)
1.82 schwarze 799: {
800: int pfd[2];
1.98 schwarze 801: int save_errno;
1.82 schwarze 802: char *cp;
803:
804: curp->file = file;
1.98 schwarze 805:
806: /* Unless zipped, try to just open the file. */
807:
1.82 schwarze 808: if ((cp = strrchr(file, '.')) == NULL ||
809: strcmp(cp + 1, "gz")) {
1.97 schwarze 810: curp->child = 0;
1.98 schwarze 811: if ((*fd = open(file, O_RDONLY)) != -1)
812: return(MANDOCLEVEL_OK);
813:
814: /* Open failed; try to append ".gz". */
815:
816: mandoc_asprintf(&cp, "%s.gz", file);
817: file = cp;
818: } else
819: cp = NULL;
820:
821: /* Before forking, make sure the file can be read. */
822:
823: save_errno = errno;
824: if (access(file, R_OK) == -1) {
825: if (cp != NULL)
826: errno = save_errno;
1.108 schwarze 827: free(cp);
828: *fd = -1;
829: curp->child = 0;
830: mandoc_msg(MANDOCERR_FILE, curp, 0, 0, strerror(errno));
831: return(MANDOCLEVEL_ERROR);
1.82 schwarze 832: }
833:
1.98 schwarze 834: /* Run gunzip(1). */
835:
1.82 schwarze 836: if (pipe(pfd) == -1) {
1.107 schwarze 837: perror("pipe");
838: exit((int)MANDOCLEVEL_SYSERR);
1.82 schwarze 839: }
840:
1.97 schwarze 841: switch (curp->child = fork()) {
1.82 schwarze 842: case -1:
1.107 schwarze 843: perror("fork");
844: exit((int)MANDOCLEVEL_SYSERR);
1.82 schwarze 845: case 0:
846: close(pfd[0]);
847: if (dup2(pfd[1], STDOUT_FILENO) == -1) {
1.107 schwarze 848: perror("dup");
849: exit((int)MANDOCLEVEL_SYSERR);
1.82 schwarze 850: }
851: execlp("gunzip", "gunzip", "-c", file, NULL);
1.107 schwarze 852: perror("exec");
853: exit((int)MANDOCLEVEL_SYSERR);
1.82 schwarze 854: default:
855: close(pfd[1]);
856: *fd = pfd[0];
857: return(MANDOCLEVEL_OK);
858: }
859: }
860:
861: enum mandoclevel
1.97 schwarze 862: mparse_wait(struct mparse *curp)
1.82 schwarze 863: {
864: int status;
865:
1.97 schwarze 866: if (curp->child == 0)
867: return(MANDOCLEVEL_OK);
868:
869: if (waitpid(curp->child, &status, 0) == -1) {
1.107 schwarze 870: perror("wait");
871: exit((int)MANDOCLEVEL_SYSERR);
1.82 schwarze 872: }
873: if (WIFSIGNALED(status)) {
1.108 schwarze 874: mandoc_vmsg(MANDOCERR_FILE, curp, 0, 0,
875: "gunzip died from signal %d", WTERMSIG(status));
876: return(MANDOCLEVEL_ERROR);
1.82 schwarze 877: }
878: if (WEXITSTATUS(status)) {
1.108 schwarze 879: mandoc_vmsg(MANDOCERR_FILE, curp, 0, 0,
880: "gunzip failed with code %d", WEXITSTATUS(status));
881: return(MANDOCLEVEL_ERROR);
1.82 schwarze 882: }
883: return(MANDOCLEVEL_OK);
1.1 kristaps 884: }
885:
886: struct mparse *
1.94 schwarze 887: mparse_alloc(int options, enum mandoclevel wlevel, mandocmsg mmsg,
888: const struct mchars *mchars, const char *defos)
1.1 kristaps 889: {
890: struct mparse *curp;
1.10 kristaps 891:
892: assert(wlevel <= MANDOCLEVEL_FATAL);
1.1 kristaps 893:
894: curp = mandoc_calloc(1, sizeof(struct mparse));
895:
1.44 schwarze 896: curp->options = options;
1.3 kristaps 897: curp->wlevel = wlevel;
1.1 kristaps 898: curp->mmsg = mmsg;
1.29 schwarze 899: curp->defos = defos;
1.1 kristaps 900:
1.94 schwarze 901: curp->mchars = mchars;
902: curp->roff = roff_alloc(curp, curp->mchars, options);
1.72 schwarze 903: if (curp->options & MPARSE_MDOC)
904: curp->pmdoc = mdoc_alloc(
905: curp->roff, curp, curp->defos,
906: curp->options & MPARSE_QUICK ? 1 : 0);
907: if (curp->options & MPARSE_MAN)
1.106 schwarze 908: curp->pman = man_alloc(
909: curp->roff, curp, curp->defos,
1.72 schwarze 910: curp->options & MPARSE_QUICK ? 1 : 0);
911:
1.1 kristaps 912: return(curp);
913: }
914:
915: void
916: mparse_reset(struct mparse *curp)
917: {
918:
919: roff_reset(curp->roff);
920:
921: if (curp->mdoc)
922: mdoc_reset(curp->mdoc);
923: if (curp->man)
924: man_reset(curp->man);
1.24 kristaps 925: if (curp->secondary)
926: curp->secondary->sz = 0;
1.1 kristaps 927:
928: curp->file_status = MANDOCLEVEL_OK;
929: curp->mdoc = NULL;
930: curp->man = NULL;
1.45 schwarze 931:
932: free(curp->sodest);
933: curp->sodest = NULL;
1.1 kristaps 934: }
935:
936: void
937: mparse_free(struct mparse *curp)
938: {
939:
940: if (curp->pmdoc)
941: mdoc_free(curp->pmdoc);
942: if (curp->pman)
943: man_free(curp->pman);
944: if (curp->roff)
945: roff_free(curp->roff);
1.24 kristaps 946: if (curp->secondary)
947: free(curp->secondary->buf);
1.1 kristaps 948:
1.24 kristaps 949: free(curp->secondary);
1.45 schwarze 950: free(curp->sodest);
1.1 kristaps 951: free(curp);
952: }
953:
954: void
1.45 schwarze 955: mparse_result(struct mparse *curp,
956: struct mdoc **mdoc, struct man **man, char **sodest)
1.1 kristaps 957: {
958:
1.45 schwarze 959: if (sodest && NULL != (*sodest = curp->sodest)) {
960: *mdoc = NULL;
961: *man = NULL;
962: return;
963: }
1.9 kristaps 964: if (mdoc)
965: *mdoc = curp->mdoc;
966: if (man)
967: *man = curp->man;
1.3 kristaps 968: }
969:
970: void
971: mandoc_vmsg(enum mandocerr t, struct mparse *m,
972: int ln, int pos, const char *fmt, ...)
973: {
974: char buf[256];
975: va_list ap;
976:
977: va_start(ap, fmt);
1.48 schwarze 978: (void)vsnprintf(buf, sizeof(buf), fmt, ap);
1.3 kristaps 979: va_end(ap);
980:
981: mandoc_msg(t, m, ln, pos, buf);
982: }
983:
984: void
1.47 schwarze 985: mandoc_msg(enum mandocerr er, struct mparse *m,
1.3 kristaps 986: int ln, int col, const char *msg)
987: {
988: enum mandoclevel level;
989:
990: level = MANDOCLEVEL_FATAL;
991: while (er < mandoclimits[level])
992: level--;
993:
1.108 schwarze 994: if (level < m->wlevel && er != MANDOCERR_FILE)
1.3 kristaps 995: return;
996:
1.8 kristaps 997: if (m->mmsg)
998: (*m->mmsg)(er, level, m->file, ln, col, msg);
1.3 kristaps 999:
1000: if (m->file_status < level)
1001: m->file_status = level;
1.7 kristaps 1002: }
1003:
1004: const char *
1005: mparse_strerror(enum mandocerr er)
1006: {
1007:
1008: return(mandocerrs[er]);
1009: }
1010:
1011: const char *
1012: mparse_strlevel(enum mandoclevel lvl)
1013: {
1014: return(mandoclevels[lvl]);
1.24 kristaps 1015: }
1016:
1017: void
1018: mparse_keep(struct mparse *p)
1019: {
1020:
1021: assert(NULL == p->secondary);
1022: p->secondary = mandoc_calloc(1, sizeof(struct buf));
1023: }
1024:
1025: const char *
1026: mparse_getkeep(const struct mparse *p)
1027: {
1028:
1029: assert(p->secondary);
1030: return(p->secondary->sz ? p->secondary->buf : NULL);
1.1 kristaps 1031: }
CVSweb