Annotation of src/usr.bin/mandoc/read.c, Revision 1.92
1.92 ! schwarze 1: /* $OpenBSD: read.c,v 1.91 2015/01/26 00:54:09 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",
108: "sections out of conventional order",
1.32 schwarze 109: "duplicate section title",
110: "unexpected section",
1.63 schwarze 111: "unusual Xr order",
112: "unusual Xr punctuation",
1.62 schwarze 113: "AUTHORS section without An macro",
1.1 schwarze 114:
115: /* related to macros and nesting */
1.33 schwarze 116: "obsolete macro",
1.78 schwarze 117: "macro neither callable nor escaped",
1.1 schwarze 118: "skipping paragraph macro",
1.10 schwarze 119: "moving paragraph macro out of list",
1.1 schwarze 120: "skipping no-space macro",
121: "blocks badly nested",
122: "nested displays are not portable",
1.35 schwarze 123: "moving content out of list",
124: ".Vt block has child macro",
1.56 schwarze 125: "fill mode already enabled, skipping",
126: "fill mode already disabled, skipping",
1.1 schwarze 127: "line scope broken",
128:
129: /* related to missing macro arguments */
1.36 schwarze 130: "skipping empty request",
131: "conditional request controls empty scope",
1.1 schwarze 132: "skipping empty macro",
1.40 schwarze 133: "empty argument, using 0n",
1.1 schwarze 134: "argument count wrong",
1.38 schwarze 135: "missing display type, using -ragged",
136: "list type is not the first argument",
137: "missing -width in -tag list, using 8n",
1.56 schwarze 138: "missing utility name, using \"\"",
1.38 schwarze 139: "empty head in list item",
140: "empty list item",
1.39 schwarze 141: "missing font type, using \\fR",
142: "unknown font type, using \\fR",
1.79 schwarze 143: "nothing follows prefix",
1.38 schwarze 144: "missing -std argument, adding it",
1.66 schwarze 145: "missing eqn box, using \"\"",
1.1 schwarze 146:
147: /* related to bad macro arguments */
1.42 schwarze 148: "unterminated quoted argument",
1.1 schwarze 149: "duplicate argument",
1.54 schwarze 150: "skipping duplicate argument",
1.41 schwarze 151: "skipping duplicate display type",
152: "skipping duplicate list type",
1.54 schwarze 153: "skipping -width argument",
1.1 schwarze 154: "unknown AT&T UNIX version",
1.64 schwarze 155: "comma in function argument",
1.65 schwarze 156: "parenthesis in function name",
1.45 schwarze 157: "invalid content in Rs block",
1.41 schwarze 158: "invalid Boolean argument",
159: "unknown font, skipping request",
1.1 schwarze 160:
161: /* related to plain text */
1.42 schwarze 162: "blank line in fill mode, using .sp",
163: "tab in filled text",
164: "whitespace at end of input line",
1.1 schwarze 165: "bad comment style",
1.42 schwarze 166: "invalid escape sequence",
167: "undefined string, using \"\"",
1.3 schwarze 168:
1.1 schwarze 169: "generic error",
170:
1.3 schwarze 171: /* related to equations */
172: "unexpected equation scope closure",
173: "equation scope open on exit",
174: "overlapping equation scopes",
175: "unexpected end of equation",
176:
1.1 schwarze 177: /* related to tables */
1.91 schwarze 178: "non-alphabetic character in tbl options",
179: "skipping unknown tbl option",
180: "missing tbl option argument",
181: "wrong tbl option argument size",
1.1 schwarze 182: "no table layout cells specified",
183: "no table data cells specified",
184: "ignore data in cell",
185: "data block still open",
186: "ignoring extra data cells",
187:
1.46 schwarze 188: /* related to document structure and macros */
1.83 schwarze 189: NULL,
1.1 schwarze 190: "input stack limit exceeded, infinite loop?",
191: "skipping bad character",
1.46 schwarze 192: "skipping unknown macro",
1.87 schwarze 193: "skipping insecure request",
1.51 schwarze 194: "skipping item outside list",
1.46 schwarze 195: "skipping column outside column list",
196: "skipping end of block that is not open",
1.90 schwarze 197: "fewer RS blocks open, skipping",
1.46 schwarze 198: "inserting missing end of block",
199: "appending missing end of block",
200:
201: /* related to request and macro arguments */
1.1 schwarze 202: "escaped character not allowed in a name",
203: "argument count wrong",
1.75 schwarze 204: "NOT IMPLEMENTED: Bd -file",
1.49 schwarze 205: "missing list type, using -item",
1.48 schwarze 206: "missing manual name, using \"\"",
1.49 schwarze 207: "uname(3) system call failed, using UNKNOWN",
1.41 schwarze 208: "unknown standard specifier",
1.49 schwarze 209: "skipping request without numeric argument",
1.84 schwarze 210: "NOT IMPLEMENTED: .so with absolute path or \"..\"",
1.85 schwarze 211: ".so request failed",
1.39 schwarze 212: "skipping all arguments",
213: "skipping excess arguments",
1.68 schwarze 214: "divide by zero",
1.87 schwarze 215:
216: "unsupported feature",
217: "input too large",
1.89 schwarze 218: "unsupported control character",
1.87 schwarze 219: "unsupported roff request",
220: "unsupported table layout",
221: "ignoring macro in table",
1.92 ! schwarze 222: "eqn in tbl",
1.1 schwarze 223: };
224:
225: static const char * const mandoclevels[MANDOCLEVEL_MAX] = {
226: "SUCCESS",
227: "RESERVED",
228: "WARNING",
229: "ERROR",
1.87 schwarze 230: "UNSUPP",
1.1 schwarze 231: "BADARG",
232: "SYSERR"
233: };
234:
1.25 schwarze 235:
1.1 schwarze 236: static void
237: resize_buf(struct buf *buf, size_t initial)
238: {
239:
240: buf->sz = buf->sz > initial/2 ? 2 * buf->sz : initial;
241: buf->buf = mandoc_realloc(buf->buf, buf->sz);
242: }
243:
244: static void
1.60 schwarze 245: choose_parser(struct mparse *curp)
1.1 schwarze 246: {
1.59 schwarze 247: char *cp, *ep;
248: int format;
1.1 schwarze 249:
1.59 schwarze 250: /*
251: * If neither command line arguments -mdoc or -man select
252: * a parser nor the roff parser found a .Dd or .TH macro
253: * yet, look ahead in the main input buffer.
254: */
255:
256: if ((format = roff_getformat(curp->roff)) == 0) {
257: cp = curp->primary->buf;
258: ep = cp + curp->primary->sz;
259: while (cp < ep) {
1.61 schwarze 260: if (*cp == '.' || *cp == '\'') {
1.59 schwarze 261: cp++;
262: if (cp[0] == 'D' && cp[1] == 'd') {
263: format = MPARSE_MDOC;
264: break;
265: }
266: if (cp[0] == 'T' && cp[1] == 'H') {
267: format = MPARSE_MAN;
268: break;
269: }
270: }
271: cp = memchr(cp, '\n', ep - cp);
272: if (cp == NULL)
273: break;
274: cp++;
275: }
1.1 schwarze 276: }
277:
1.59 schwarze 278: if (format == MPARSE_MDOC) {
1.25 schwarze 279: if (NULL == curp->pmdoc)
1.22 schwarze 280: curp->pmdoc = mdoc_alloc(
281: curp->roff, curp, curp->defos,
282: MPARSE_QUICK & curp->options ? 1 : 0);
1.1 schwarze 283: assert(curp->pmdoc);
284: curp->mdoc = curp->pmdoc;
285: return;
1.25 schwarze 286: }
1.1 schwarze 287:
1.59 schwarze 288: /* Fall back to man(7) as a last resort. */
289:
1.25 schwarze 290: if (NULL == curp->pman)
1.81 schwarze 291: curp->pman = man_alloc(
292: curp->roff, curp, curp->defos,
1.22 schwarze 293: MPARSE_QUICK & curp->options ? 1 : 0);
1.1 schwarze 294: assert(curp->pman);
295: curp->man = curp->pman;
296: }
297:
298: /*
1.71 schwarze 299: * Main parse routine for a buffer.
300: * It assumes encoding and line numbering are already set up.
301: * It can recurse directly (for invocations of user-defined
302: * macros, inline equations, and input line traps)
303: * and indirectly (for .so file inclusion).
1.1 schwarze 304: */
305: static void
1.71 schwarze 306: mparse_buf_r(struct mparse *curp, struct buf blk, size_t i, int start)
1.1 schwarze 307: {
308: const struct tbl_span *span;
309: struct buf ln;
1.88 schwarze 310: const char *save_file;
1.85 schwarze 311: char *cp;
1.71 schwarze 312: size_t pos; /* byte number in the ln buffer */
1.1 schwarze 313: enum rofferr rr;
1.76 schwarze 314: int of;
1.1 schwarze 315: int lnn; /* line number in the real file */
1.85 schwarze 316: int fd;
317: pid_t save_child;
1.1 schwarze 318: unsigned char c;
319:
1.71 schwarze 320: memset(&ln, 0, sizeof(ln));
1.1 schwarze 321:
1.25 schwarze 322: lnn = curp->line;
323: pos = 0;
1.1 schwarze 324:
1.71 schwarze 325: while (i < blk.sz) {
1.1 schwarze 326: if (0 == pos && '\0' == blk.buf[i])
327: break;
328:
329: if (start) {
330: curp->line = lnn;
331: curp->reparse_count = 0;
1.70 schwarze 332:
333: if (lnn < 3 &&
334: curp->filenc & MPARSE_UTF8 &&
1.71 schwarze 335: curp->filenc & MPARSE_LATIN1)
336: curp->filenc = preconv_cue(&blk, i);
1.1 schwarze 337: }
338:
1.71 schwarze 339: while (i < blk.sz && (start || blk.buf[i] != '\0')) {
1.1 schwarze 340:
341: /*
342: * When finding an unescaped newline character,
343: * leave the character loop to process the line.
344: * Skip a preceding carriage return, if any.
345: */
346:
1.71 schwarze 347: if ('\r' == blk.buf[i] && i + 1 < blk.sz &&
1.1 schwarze 348: '\n' == blk.buf[i + 1])
349: ++i;
350: if ('\n' == blk.buf[i]) {
351: ++i;
352: ++lnn;
353: break;
354: }
355:
1.13 schwarze 356: /*
1.70 schwarze 357: * Make sure we have space for the worst
358: * case of 11 bytes: "\\[u10ffff]\0"
1.13 schwarze 359: */
360:
1.71 schwarze 361: if (pos + 11 > ln.sz)
1.13 schwarze 362: resize_buf(&ln, 256);
363:
1.25 schwarze 364: /*
1.70 schwarze 365: * Encode 8-bit input.
1.1 schwarze 366: */
367:
1.70 schwarze 368: c = blk.buf[i];
369: if (c & 0x80) {
1.71 schwarze 370: if ( ! (curp->filenc && preconv_encode(
371: &blk, &i, &ln, &pos, &curp->filenc))) {
1.89 schwarze 372: mandoc_vmsg(MANDOCERR_CHAR_BAD, curp,
373: curp->line, pos, "0x%x", c);
1.70 schwarze 374: ln.buf[pos++] = '?';
375: i++;
376: }
377: continue;
378: }
379:
380: /*
381: * Exclude control characters.
382: */
1.1 schwarze 383:
1.70 schwarze 384: if (c == 0x7f || (c < 0x20 && c != 0x09)) {
1.89 schwarze 385: mandoc_vmsg(c == 0x00 || c == 0x04 ||
386: c > 0x0a ? MANDOCERR_CHAR_BAD :
387: MANDOCERR_CHAR_UNSUPP,
388: curp, curp->line, pos, "0x%x", c);
1.1 schwarze 389: i++;
1.6 schwarze 390: ln.buf[pos++] = '?';
1.1 schwarze 391: continue;
392: }
393:
394: /* Trailing backslash = a plain char. */
395:
1.71 schwarze 396: if (blk.buf[i] != '\\' || i + 1 == blk.sz) {
1.1 schwarze 397: ln.buf[pos++] = blk.buf[i++];
398: continue;
399: }
400:
401: /*
402: * Found escape and at least one other character.
403: * When it's a newline character, skip it.
404: * When there is a carriage return in between,
405: * skip that one as well.
406: */
407:
1.71 schwarze 408: if ('\r' == blk.buf[i + 1] && i + 2 < blk.sz &&
1.1 schwarze 409: '\n' == blk.buf[i + 2])
410: ++i;
411: if ('\n' == blk.buf[i + 1]) {
412: i += 2;
413: ++lnn;
414: continue;
415: }
416:
417: if ('"' == blk.buf[i + 1] || '#' == blk.buf[i + 1]) {
418: i += 2;
419: /* Comment, skip to end of line */
1.71 schwarze 420: for (; i < blk.sz; ++i) {
1.1 schwarze 421: if ('\n' == blk.buf[i]) {
422: ++i;
423: ++lnn;
424: break;
425: }
426: }
427:
428: /* Backout trailing whitespaces */
429: for (; pos > 0; --pos) {
430: if (ln.buf[pos - 1] != ' ')
431: break;
432: if (pos > 2 && ln.buf[pos - 2] == '\\')
433: break;
434: }
435: break;
436: }
437:
1.13 schwarze 438: /* Catch escaped bogus characters. */
439:
440: c = (unsigned char) blk.buf[i+1];
441:
1.25 schwarze 442: if ( ! (isascii(c) &&
443: (isgraph(c) || isblank(c)))) {
1.89 schwarze 444: mandoc_vmsg(MANDOCERR_CHAR_BAD, curp,
1.56 schwarze 445: curp->line, pos, "0x%x", c);
1.13 schwarze 446: i += 2;
447: ln.buf[pos++] = '?';
448: continue;
449: }
450:
1.1 schwarze 451: /* Some other escape sequence, copy & cont. */
452:
453: ln.buf[pos++] = blk.buf[i++];
454: ln.buf[pos++] = blk.buf[i++];
455: }
456:
1.71 schwarze 457: if (pos >= ln.sz)
1.1 schwarze 458: resize_buf(&ln, 256);
459:
460: ln.buf[pos] = '\0';
461:
462: /*
463: * A significant amount of complexity is contained by
464: * the roff preprocessor. It's line-oriented but can be
465: * expressed on one line, so we need at times to
466: * readjust our starting point and re-run it. The roff
467: * preprocessor can also readjust the buffers with new
468: * data, so we pass them in wholesale.
469: */
470:
471: of = 0;
472:
1.4 schwarze 473: /*
474: * Maintain a lookaside buffer of all parsed lines. We
475: * only do this if mparse_keep() has been invoked (the
476: * buffer may be accessed with mparse_getkeep()).
477: */
478:
479: if (curp->secondary) {
1.25 schwarze 480: curp->secondary->buf = mandoc_realloc(
481: curp->secondary->buf,
482: curp->secondary->sz + pos + 2);
483: memcpy(curp->secondary->buf +
484: curp->secondary->sz,
485: ln.buf, pos);
1.4 schwarze 486: curp->secondary->sz += pos;
487: curp->secondary->buf
488: [curp->secondary->sz] = '\n';
489: curp->secondary->sz++;
490: curp->secondary->buf
491: [curp->secondary->sz] = '\0';
492: }
1.1 schwarze 493: rerun:
1.72 schwarze 494: rr = roff_parseln(curp->roff, curp->line, &ln, &of);
1.1 schwarze 495:
496: switch (rr) {
1.25 schwarze 497: case ROFF_REPARSE:
1.1 schwarze 498: if (REPARSE_LIMIT >= ++curp->reparse_count)
1.71 schwarze 499: mparse_buf_r(curp, ln, of, 0);
1.1 schwarze 500: else
501: mandoc_msg(MANDOCERR_ROFFLOOP, curp,
1.25 schwarze 502: curp->line, pos, NULL);
1.1 schwarze 503: pos = 0;
504: continue;
1.25 schwarze 505: case ROFF_APPEND:
1.71 schwarze 506: pos = strlen(ln.buf);
1.1 schwarze 507: continue;
1.25 schwarze 508: case ROFF_RERUN:
1.1 schwarze 509: goto rerun;
1.25 schwarze 510: case ROFF_IGN:
1.1 schwarze 511: pos = 0;
512: continue;
1.25 schwarze 513: case ROFF_SO:
1.71 schwarze 514: if ( ! (curp->options & MPARSE_SO) &&
515: (i >= blk.sz || blk.buf[i] == '\0')) {
1.23 schwarze 516: curp->sodest = mandoc_strdup(ln.buf + of);
517: free(ln.buf);
518: return;
519: }
1.4 schwarze 520: /*
521: * We remove `so' clauses from our lookaside
522: * buffer because we're going to descend into
523: * the file recursively.
524: */
1.25 schwarze 525: if (curp->secondary)
1.4 schwarze 526: curp->secondary->sz -= pos + 1;
1.88 schwarze 527: save_file = curp->file;
1.85 schwarze 528: save_child = curp->child;
529: if (mparse_open(curp, &fd, ln.buf + of) ==
1.88 schwarze 530: MANDOCLEVEL_OK) {
1.85 schwarze 531: mparse_readfd(curp, fd, ln.buf + of);
1.88 schwarze 532: curp->file = save_file;
533: } else {
534: curp->file = save_file;
1.30 schwarze 535: mandoc_vmsg(MANDOCERR_SO_FAIL,
536: curp, curp->line, pos,
537: ".so %s", ln.buf + of);
1.85 schwarze 538: ln.sz = mandoc_asprintf(&cp,
539: ".sp\nSee the file %s.\n.sp",
540: ln.buf + of);
541: free(ln.buf);
542: ln.buf = cp;
543: of = 0;
544: mparse_buf_r(curp, ln, of, 0);
1.30 schwarze 545: }
1.85 schwarze 546: curp->child = save_child;
1.1 schwarze 547: pos = 0;
548: continue;
549: default:
550: break;
551: }
552:
553: /*
554: * If input parsers have not been allocated, do so now.
1.2 schwarze 555: * We keep these instanced between parsers, but set them
1.1 schwarze 556: * locally per parse routine since we can use different
557: * parsers with each one.
558: */
559:
560: if ( ! (curp->man || curp->mdoc))
1.60 schwarze 561: choose_parser(curp);
1.1 schwarze 562:
1.25 schwarze 563: /*
1.60 schwarze 564: * Lastly, push down into the parsers themselves.
1.1 schwarze 565: * If libroff returns ROFF_TBL, then add it to the
566: * currently open parse. Since we only get here if
567: * there does exist data (see tbl_data.c), we're
568: * guaranteed that something's been allocated.
569: * Do the same for ROFF_EQN.
570: */
571:
1.76 schwarze 572: if (rr == ROFF_TBL) {
573: while ((span = roff_span(curp->roff)) != NULL)
574: if (curp->man == NULL)
575: mdoc_addspan(curp->mdoc, span);
576: else
577: man_addspan(curp->man, span);
578: } else if (rr == ROFF_EQN) {
579: if (curp->man == NULL)
580: mdoc_addeqn(curp->mdoc, roff_eqn(curp->roff));
581: else
582: man_addeqn(curp->man, roff_eqn(curp->roff));
583: } else if ((curp->man == NULL ?
584: mdoc_parseln(curp->mdoc, curp->line, ln.buf, of) :
585: man_parseln(curp->man, curp->line, ln.buf, of)) == 2)
586: break;
1.1 schwarze 587:
588: /* Temporary buffers typically are not full. */
589:
590: if (0 == start && '\0' == blk.buf[i])
591: break;
592:
593: /* Start the next input line. */
594:
595: pos = 0;
596: }
597:
598: free(ln.buf);
599: }
600:
601: static int
1.18 schwarze 602: read_whole_file(struct mparse *curp, const char *file, int fd,
603: struct buf *fb, int *with_mmap)
1.1 schwarze 604: {
605: struct stat st;
606: size_t off;
607: ssize_t ssz;
608:
609: if (-1 == fstat(fd, &st)) {
1.82 schwarze 610: perror(file);
611: exit((int)MANDOCLEVEL_SYSERR);
1.1 schwarze 612: }
613:
614: /*
615: * If we're a regular file, try just reading in the whole entry
616: * via mmap(). This is faster than reading it into blocks, and
617: * since each file is only a few bytes to begin with, I'm not
618: * concerned that this is going to tank any machines.
619: */
620:
621: if (S_ISREG(st.st_mode)) {
622: if (st.st_size >= (1U << 31)) {
1.86 schwarze 623: mandoc_msg(MANDOCERR_TOOLARGE, curp, 0, 0, NULL);
1.1 schwarze 624: return(0);
625: }
626: *with_mmap = 1;
627: fb->sz = (size_t)st.st_size;
1.15 schwarze 628: fb->buf = mmap(NULL, fb->sz, PROT_READ, MAP_SHARED, fd, 0);
1.1 schwarze 629: if (fb->buf != MAP_FAILED)
630: return(1);
631: }
632:
633: /*
634: * If this isn't a regular file (like, say, stdin), then we must
635: * go the old way and just read things in bit by bit.
636: */
637:
638: *with_mmap = 0;
639: off = 0;
640: fb->sz = 0;
641: fb->buf = NULL;
642: for (;;) {
643: if (off == fb->sz) {
644: if (fb->sz == (1U << 31)) {
1.86 schwarze 645: mandoc_msg(MANDOCERR_TOOLARGE, curp,
646: 0, 0, NULL);
1.1 schwarze 647: break;
648: }
649: resize_buf(fb, 65536);
650: }
651: ssz = read(fd, fb->buf + (int)off, fb->sz - off);
652: if (ssz == 0) {
653: fb->sz = off;
654: return(1);
655: }
656: if (ssz == -1) {
1.82 schwarze 657: perror(file);
658: exit((int)MANDOCLEVEL_SYSERR);
1.1 schwarze 659: }
660: off += (size_t)ssz;
661: }
662:
663: free(fb->buf);
664: fb->buf = NULL;
665: return(0);
666: }
667:
668: static void
669: mparse_end(struct mparse *curp)
670: {
671:
1.50 schwarze 672: if (curp->mdoc == NULL &&
673: curp->man == NULL &&
674: curp->sodest == NULL) {
675: if (curp->options & MPARSE_MDOC)
676: curp->mdoc = curp->pmdoc;
677: else {
678: if (curp->pman == NULL)
1.81 schwarze 679: curp->pman = man_alloc(
680: curp->roff, curp, curp->defos,
1.50 schwarze 681: curp->options & MPARSE_QUICK ? 1 : 0);
682: curp->man = curp->pman;
683: }
684: }
1.86 schwarze 685: if (curp->mdoc)
686: mdoc_endparse(curp->mdoc);
687: if (curp->man)
688: man_endparse(curp->man);
1.1 schwarze 689: roff_endparse(curp->roff);
690: }
691:
1.15 schwarze 692: static void
693: mparse_parse_buffer(struct mparse *curp, struct buf blk, const char *file)
1.1 schwarze 694: {
1.61 schwarze 695: struct buf *svprimary;
1.1 schwarze 696: const char *svfile;
1.71 schwarze 697: size_t offset;
1.14 schwarze 698: static int recursion_depth;
699:
700: if (64 < recursion_depth) {
701: mandoc_msg(MANDOCERR_ROFFLOOP, curp, curp->line, 0, NULL);
1.15 schwarze 702: return;
1.14 schwarze 703: }
1.1 schwarze 704:
1.15 schwarze 705: /* Line number is per-file. */
706: svfile = curp->file;
707: curp->file = file;
1.61 schwarze 708: svprimary = curp->primary;
1.59 schwarze 709: curp->primary = &blk;
1.15 schwarze 710: curp->line = 1;
711: recursion_depth++;
712:
1.70 schwarze 713: /* Skip an UTF-8 byte order mark. */
714: if (curp->filenc & MPARSE_UTF8 && blk.sz > 2 &&
715: (unsigned char)blk.buf[0] == 0xef &&
716: (unsigned char)blk.buf[1] == 0xbb &&
717: (unsigned char)blk.buf[2] == 0xbf) {
1.71 schwarze 718: offset = 3;
1.70 schwarze 719: curp->filenc &= ~MPARSE_LATIN1;
1.71 schwarze 720: } else
721: offset = 0;
1.70 schwarze 722:
1.71 schwarze 723: mparse_buf_r(curp, blk, offset, 1);
1.15 schwarze 724:
1.86 schwarze 725: if (--recursion_depth == 0)
1.15 schwarze 726: mparse_end(curp);
727:
1.61 schwarze 728: curp->primary = svprimary;
1.15 schwarze 729: curp->file = svfile;
730: }
731:
1.74 schwarze 732: /*
733: * Read the whole file into memory and call the parsers.
734: * Called recursively when an .so request is encountered.
735: */
1.15 schwarze 736: enum mandoclevel
737: mparse_readfd(struct mparse *curp, int fd, const char *file)
738: {
739: struct buf blk;
740: int with_mmap;
1.70 schwarze 741: int save_filenc;
1.1 schwarze 742:
1.67 schwarze 743: if (read_whole_file(curp, file, fd, &blk, &with_mmap)) {
1.70 schwarze 744: save_filenc = curp->filenc;
745: curp->filenc = curp->options &
746: (MPARSE_UTF8 | MPARSE_LATIN1);
1.67 schwarze 747: mparse_parse_buffer(curp, blk, file);
1.70 schwarze 748: curp->filenc = save_filenc;
1.67 schwarze 749: if (with_mmap)
750: munmap(blk.buf, blk.sz);
751: else
752: free(blk.buf);
753: }
1.1 schwarze 754:
1.74 schwarze 755: if (fd != STDIN_FILENO && close(fd) == -1)
1.1 schwarze 756: perror(file);
1.67 schwarze 757:
1.74 schwarze 758: mparse_wait(curp);
1.1 schwarze 759: return(curp->file_status);
1.58 schwarze 760: }
761:
762: enum mandoclevel
1.73 schwarze 763: mparse_open(struct mparse *curp, int *fd, const char *file)
1.58 schwarze 764: {
765: int pfd[2];
1.74 schwarze 766: int save_errno;
1.58 schwarze 767: char *cp;
768:
769: curp->file = file;
1.74 schwarze 770:
771: /* Unless zipped, try to just open the file. */
772:
1.58 schwarze 773: if ((cp = strrchr(file, '.')) == NULL ||
774: strcmp(cp + 1, "gz")) {
1.73 schwarze 775: curp->child = 0;
1.74 schwarze 776: if ((*fd = open(file, O_RDONLY)) != -1)
777: return(MANDOCLEVEL_OK);
778:
779: /* Open failed; try to append ".gz". */
780:
781: mandoc_asprintf(&cp, "%s.gz", file);
782: file = cp;
783: } else
784: cp = NULL;
785:
786: /* Before forking, make sure the file can be read. */
787:
788: save_errno = errno;
789: if (access(file, R_OK) == -1) {
790: if (cp != NULL)
791: errno = save_errno;
1.83 schwarze 792: free(cp);
793: *fd = -1;
794: curp->child = 0;
795: mandoc_msg(MANDOCERR_FILE, curp, 0, 0, strerror(errno));
796: return(MANDOCLEVEL_ERROR);
1.58 schwarze 797: }
798:
1.74 schwarze 799: /* Run gunzip(1). */
800:
1.58 schwarze 801: if (pipe(pfd) == -1) {
1.82 schwarze 802: perror("pipe");
803: exit((int)MANDOCLEVEL_SYSERR);
1.58 schwarze 804: }
805:
1.73 schwarze 806: switch (curp->child = fork()) {
1.58 schwarze 807: case -1:
1.82 schwarze 808: perror("fork");
809: exit((int)MANDOCLEVEL_SYSERR);
1.58 schwarze 810: case 0:
811: close(pfd[0]);
812: if (dup2(pfd[1], STDOUT_FILENO) == -1) {
1.82 schwarze 813: perror("dup");
814: exit((int)MANDOCLEVEL_SYSERR);
1.58 schwarze 815: }
816: execlp("gunzip", "gunzip", "-c", file, NULL);
1.82 schwarze 817: perror("exec");
818: exit((int)MANDOCLEVEL_SYSERR);
1.58 schwarze 819: default:
820: close(pfd[1]);
821: *fd = pfd[0];
822: return(MANDOCLEVEL_OK);
823: }
824: }
825:
826: enum mandoclevel
1.73 schwarze 827: mparse_wait(struct mparse *curp)
1.58 schwarze 828: {
829: int status;
830:
1.73 schwarze 831: if (curp->child == 0)
832: return(MANDOCLEVEL_OK);
833:
834: if (waitpid(curp->child, &status, 0) == -1) {
1.82 schwarze 835: perror("wait");
836: exit((int)MANDOCLEVEL_SYSERR);
1.58 schwarze 837: }
838: if (WIFSIGNALED(status)) {
1.83 schwarze 839: mandoc_vmsg(MANDOCERR_FILE, curp, 0, 0,
840: "gunzip died from signal %d", WTERMSIG(status));
841: return(MANDOCLEVEL_ERROR);
1.58 schwarze 842: }
843: if (WEXITSTATUS(status)) {
1.83 schwarze 844: mandoc_vmsg(MANDOCERR_FILE, curp, 0, 0,
845: "gunzip failed with code %d", WEXITSTATUS(status));
846: return(MANDOCLEVEL_ERROR);
1.58 schwarze 847: }
848: return(MANDOCLEVEL_OK);
1.1 schwarze 849: }
850:
851: struct mparse *
1.69 schwarze 852: mparse_alloc(int options, enum mandoclevel wlevel, mandocmsg mmsg,
853: const struct mchars *mchars, const char *defos)
1.1 schwarze 854: {
855: struct mparse *curp;
856:
857: curp = mandoc_calloc(1, sizeof(struct mparse));
858:
1.22 schwarze 859: curp->options = options;
1.1 schwarze 860: curp->wlevel = wlevel;
861: curp->mmsg = mmsg;
1.7 schwarze 862: curp->defos = defos;
1.1 schwarze 863:
1.69 schwarze 864: curp->mchars = mchars;
865: curp->roff = roff_alloc(curp, curp->mchars, options);
1.50 schwarze 866: if (curp->options & MPARSE_MDOC)
867: curp->pmdoc = mdoc_alloc(
868: curp->roff, curp, curp->defos,
869: curp->options & MPARSE_QUICK ? 1 : 0);
870: if (curp->options & MPARSE_MAN)
1.81 schwarze 871: curp->pman = man_alloc(
872: curp->roff, curp, curp->defos,
1.50 schwarze 873: curp->options & MPARSE_QUICK ? 1 : 0);
874:
1.1 schwarze 875: return(curp);
876: }
877:
878: void
879: mparse_reset(struct mparse *curp)
880: {
881:
882: roff_reset(curp->roff);
883:
884: if (curp->mdoc)
885: mdoc_reset(curp->mdoc);
886: if (curp->man)
887: man_reset(curp->man);
1.4 schwarze 888: if (curp->secondary)
889: curp->secondary->sz = 0;
1.1 schwarze 890:
891: curp->file_status = MANDOCLEVEL_OK;
892: curp->mdoc = NULL;
893: curp->man = NULL;
1.23 schwarze 894:
895: free(curp->sodest);
896: curp->sodest = NULL;
1.1 schwarze 897: }
898:
899: void
900: mparse_free(struct mparse *curp)
901: {
902:
903: if (curp->pmdoc)
904: mdoc_free(curp->pmdoc);
905: if (curp->pman)
906: man_free(curp->pman);
907: if (curp->roff)
908: roff_free(curp->roff);
1.4 schwarze 909: if (curp->secondary)
910: free(curp->secondary->buf);
1.1 schwarze 911:
1.4 schwarze 912: free(curp->secondary);
1.23 schwarze 913: free(curp->sodest);
1.1 schwarze 914: free(curp);
915: }
916:
917: void
1.23 schwarze 918: mparse_result(struct mparse *curp,
919: struct mdoc **mdoc, struct man **man, char **sodest)
1.1 schwarze 920: {
921:
1.23 schwarze 922: if (sodest && NULL != (*sodest = curp->sodest)) {
923: *mdoc = NULL;
924: *man = NULL;
925: return;
926: }
1.1 schwarze 927: if (mdoc)
928: *mdoc = curp->mdoc;
929: if (man)
930: *man = curp->man;
931: }
932:
933: void
934: mandoc_vmsg(enum mandocerr t, struct mparse *m,
935: int ln, int pos, const char *fmt, ...)
936: {
937: char buf[256];
938: va_list ap;
939:
940: va_start(ap, fmt);
1.26 schwarze 941: (void)vsnprintf(buf, sizeof(buf), fmt, ap);
1.1 schwarze 942: va_end(ap);
943:
944: mandoc_msg(t, m, ln, pos, buf);
945: }
946:
947: void
1.25 schwarze 948: mandoc_msg(enum mandocerr er, struct mparse *m,
1.1 schwarze 949: int ln, int col, const char *msg)
950: {
951: enum mandoclevel level;
952:
1.87 schwarze 953: level = MANDOCLEVEL_UNSUPP;
1.1 schwarze 954: while (er < mandoclimits[level])
955: level--;
956:
1.83 schwarze 957: if (level < m->wlevel && er != MANDOCERR_FILE)
1.1 schwarze 958: return;
959:
960: if (m->mmsg)
961: (*m->mmsg)(er, level, m->file, ln, col, msg);
962:
963: if (m->file_status < level)
964: m->file_status = level;
965: }
966:
967: const char *
968: mparse_strerror(enum mandocerr er)
969: {
970:
971: return(mandocerrs[er]);
972: }
973:
974: const char *
975: mparse_strlevel(enum mandoclevel lvl)
976: {
977: return(mandoclevels[lvl]);
1.4 schwarze 978: }
979:
980: void
981: mparse_keep(struct mparse *p)
982: {
983:
984: assert(NULL == p->secondary);
985: p->secondary = mandoc_calloc(1, sizeof(struct buf));
986: }
987:
988: const char *
989: mparse_getkeep(const struct mparse *p)
990: {
991:
992: assert(p->secondary);
993: return(p->secondary->sz ? p->secondary->buf : NULL);
1.1 schwarze 994: }
![[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