]> git.cameronkatri.com Git - mandoc.git/blob - mandocdb.c
git.ckatri.com / mandoc.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
history | raw | HEAD
In man(7) and cat pages, cut off excessive one line descriptions.
[mandoc.git] / mandocdb.c
1 /* $Id: mandocdb.c,v 1.257 2018/02/07 20:31:39 schwarze Exp $ */
2 /*
3 * Copyright (c) 2011, 2012 Kristaps Dzonsons <kristaps@bsd.lv>
4 * Copyright (c) 2011-2017 Ingo Schwarze <schwarze@openbsd.org>
5 * Copyright (c) 2016 Ed Maste <emaste@freebsd.org>
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 AUTHORS DISCLAIM ALL WARRANTIES
12 * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
13 * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHORS 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 */
19 #include "config.h"
20
21 #include <sys/types.h>
22 #include <sys/mman.h>
23 #include <sys/stat.h>
24
25 #include <assert.h>
26 #include <ctype.h>
27 #if HAVE_ERR
28 #include <err.h>
29 #endif
30 #include <errno.h>
31 #include <fcntl.h>
32 #if HAVE_FTS
33 #include <fts.h>
34 #else
35 #include "compat_fts.h"
36 #endif
37 #include <limits.h>
38 #if HAVE_SANDBOX_INIT
39 #include <sandbox.h>
40 #endif
41 #include <stdarg.h>
42 #include <stddef.h>
43 #include <stdio.h>
44 #include <stdint.h>
45 #include <stdlib.h>
46 #include <string.h>
47 #include <unistd.h>
48
49 #include "mandoc_aux.h"
50 #include "mandoc_ohash.h"
51 #include "mandoc.h"
52 #include "roff.h"
53 #include "mdoc.h"
54 #include "man.h"
55 #include "manconf.h"
56 #include "mansearch.h"
57 #include "dba_array.h"
58 #include "dba.h"
59
60 extern const char *const mansearch_keynames[];
61
62 enum op {
63 OP_DEFAULT = 0, /* new dbs from dir list or default config */
64 OP_CONFFILE, /* new databases from custom config file */
65 OP_UPDATE, /* delete/add entries in existing database */
66 OP_DELETE, /* delete entries from existing database */
67 OP_TEST /* change no databases, report potential problems */
68 };
69
70 struct str {
71 const struct mpage *mpage; /* if set, the owning parse */
72 uint64_t mask; /* bitmask in sequence */
73 char key[]; /* rendered text */
74 };
75
76 struct inodev {
77 ino_t st_ino;
78 dev_t st_dev;
79 };
80
81 struct mpage {
82 struct inodev inodev; /* used for hashing routine */
83 struct dba_array *dba;
84 char *sec; /* section from file content */
85 char *arch; /* architecture from file content */
86 char *title; /* title from file content */
87 char *desc; /* description from file content */
88 struct mpage *next; /* singly linked list */
89 struct mlink *mlinks; /* singly linked list */
90 int name_head_done;
91 enum form form; /* format from file content */
92 };
93
94 struct mlink {
95 char file[PATH_MAX]; /* filename rel. to manpath */
96 char *dsec; /* section from directory */
97 char *arch; /* architecture from directory */
98 char *name; /* name from file name (not empty) */
99 char *fsec; /* section from file name suffix */
100 struct mlink *next; /* singly linked list */
101 struct mpage *mpage; /* parent */
102 int gzip; /* filename has a .gz suffix */
103 enum form dform; /* format from directory */
104 enum form fform; /* format from file name suffix */
105 };
106
107 typedef int (*mdoc_fp)(struct mpage *, const struct roff_meta *,
108 const struct roff_node *);
109
110 struct mdoc_handler {
111 mdoc_fp fp; /* optional handler */
112 uint64_t mask; /* set unless handler returns 0 */
113 int taboo; /* node flags that must not be set */
114 };
115
116
117 int mandocdb(int, char *[]);
118
119 static void dbadd(struct dba *, struct mpage *);
120 static void dbadd_mlink(const struct mlink *mlink);
121 static void dbprune(struct dba *);
122 static void dbwrite(struct dba *);
123 static void filescan(const char *);
124 #if HAVE_FTS_COMPARE_CONST
125 static int fts_compare(const FTSENT *const *, const FTSENT *const *);
126 #else
127 static int fts_compare(const FTSENT **, const FTSENT **);
128 #endif
129 static void mlink_add(struct mlink *, const struct stat *);
130 static void mlink_check(struct mpage *, struct mlink *);
131 static void mlink_free(struct mlink *);
132 static void mlinks_undupe(struct mpage *);
133 static void mpages_free(void);
134 static void mpages_merge(struct dba *, struct mparse *);
135 static void parse_cat(struct mpage *, int);
136 static void parse_man(struct mpage *, const struct roff_meta *,
137 const struct roff_node *);
138 static void parse_mdoc(struct mpage *, const struct roff_meta *,
139 const struct roff_node *);
140 static int parse_mdoc_head(struct mpage *, const struct roff_meta *,
141 const struct roff_node *);
142 static int parse_mdoc_Fd(struct mpage *, const struct roff_meta *,
143 const struct roff_node *);
144 static void parse_mdoc_fname(struct mpage *, const struct roff_node *);
145 static int parse_mdoc_Fn(struct mpage *, const struct roff_meta *,
146 const struct roff_node *);
147 static int parse_mdoc_Fo(struct mpage *, const struct roff_meta *,
148 const struct roff_node *);
149 static int parse_mdoc_Nd(struct mpage *, const struct roff_meta *,
150 const struct roff_node *);
151 static int parse_mdoc_Nm(struct mpage *, const struct roff_meta *,
152 const struct roff_node *);
153 static int parse_mdoc_Sh(struct mpage *, const struct roff_meta *,
154 const struct roff_node *);
155 static int parse_mdoc_Va(struct mpage *, const struct roff_meta *,
156 const struct roff_node *);
157 static int parse_mdoc_Xr(struct mpage *, const struct roff_meta *,
158 const struct roff_node *);
159 static void putkey(const struct mpage *, char *, uint64_t);
160 static void putkeys(const struct mpage *, char *, size_t, uint64_t);
161 static void putmdockey(const struct mpage *,
162 const struct roff_node *, uint64_t, int);
163 static int render_string(char **, size_t *);
164 static void say(const char *, const char *, ...)
165 __attribute__((__format__ (__printf__, 2, 3)));
166 static int set_basedir(const char *, int);
167 static int treescan(void);
168 static size_t utf8(unsigned int, char [7]);
169
170 static int nodb; /* no database changes */
171 static int mparse_options; /* abort the parse early */
172 static int use_all; /* use all found files */
173 static int debug; /* print what we're doing */
174 static int warnings; /* warn about crap */
175 static int write_utf8; /* write UTF-8 output; else ASCII */
176 static int exitcode; /* to be returned by main */
177 static enum op op; /* operational mode */
178 static char basedir[PATH_MAX]; /* current base directory */
179 static struct mpage *mpage_head; /* list of distinct manual pages */
180 static struct ohash mpages; /* table of distinct manual pages */
181 static struct ohash mlinks; /* table of directory entries */
182 static struct ohash names; /* table of all names */
183 static struct ohash strings; /* table of all strings */
184 static uint64_t name_mask;
185
186 static const struct mdoc_handler __mdocs[MDOC_MAX - MDOC_Dd] = {
187 { NULL, 0, NODE_NOPRT }, /* Dd */
188 { NULL, 0, NODE_NOPRT }, /* Dt */
189 { NULL, 0, NODE_NOPRT }, /* Os */
190 { parse_mdoc_Sh, TYPE_Sh, 0 }, /* Sh */
191 { parse_mdoc_head, TYPE_Ss, 0 }, /* Ss */
192 { NULL, 0, 0 }, /* Pp */
193 { NULL, 0, 0 }, /* D1 */
194 { NULL, 0, 0 }, /* Dl */
195 { NULL, 0, 0 }, /* Bd */
196 { NULL, 0, 0 }, /* Ed */
197 { NULL, 0, 0 }, /* Bl */
198 { NULL, 0, 0 }, /* El */
199 { NULL, 0, 0 }, /* It */
200 { NULL, 0, 0 }, /* Ad */
201 { NULL, TYPE_An, 0 }, /* An */
202 { NULL, 0, 0 }, /* Ap */
203 { NULL, TYPE_Ar, 0 }, /* Ar */
204 { NULL, TYPE_Cd, 0 }, /* Cd */
205 { NULL, TYPE_Cm, 0 }, /* Cm */
206 { NULL, TYPE_Dv, 0 }, /* Dv */
207 { NULL, TYPE_Er, 0 }, /* Er */
208 { NULL, TYPE_Ev, 0 }, /* Ev */
209 { NULL, 0, 0 }, /* Ex */
210 { NULL, TYPE_Fa, 0 }, /* Fa */
211 { parse_mdoc_Fd, 0, 0 }, /* Fd */
212 { NULL, TYPE_Fl, 0 }, /* Fl */
213 { parse_mdoc_Fn, 0, 0 }, /* Fn */
214 { NULL, TYPE_Ft, 0 }, /* Ft */
215 { NULL, TYPE_Ic, 0 }, /* Ic */
216 { NULL, TYPE_In, 0 }, /* In */
217 { NULL, TYPE_Li, 0 }, /* Li */
218 { parse_mdoc_Nd, 0, 0 }, /* Nd */
219 { parse_mdoc_Nm, 0, 0 }, /* Nm */
220 { NULL, 0, 0 }, /* Op */
221 { NULL, 0, 0 }, /* Ot */
222 { NULL, TYPE_Pa, NODE_NOSRC }, /* Pa */
223 { NULL, 0, 0 }, /* Rv */
224 { NULL, TYPE_St, 0 }, /* St */
225 { parse_mdoc_Va, TYPE_Va, 0 }, /* Va */
226 { parse_mdoc_Va, TYPE_Vt, 0 }, /* Vt */
227 { parse_mdoc_Xr, 0, 0 }, /* Xr */
228 { NULL, 0, 0 }, /* %A */
229 { NULL, 0, 0 }, /* %B */
230 { NULL, 0, 0 }, /* %D */
231 { NULL, 0, 0 }, /* %I */
232 { NULL, 0, 0 }, /* %J */
233 { NULL, 0, 0 }, /* %N */
234 { NULL, 0, 0 }, /* %O */
235 { NULL, 0, 0 }, /* %P */
236 { NULL, 0, 0 }, /* %R */
237 { NULL, 0, 0 }, /* %T */
238 { NULL, 0, 0 }, /* %V */
239 { NULL, 0, 0 }, /* Ac */
240 { NULL, 0, 0 }, /* Ao */
241 { NULL, 0, 0 }, /* Aq */
242 { NULL, TYPE_At, 0 }, /* At */
243 { NULL, 0, 0 }, /* Bc */
244 { NULL, 0, 0 }, /* Bf */
245 { NULL, 0, 0 }, /* Bo */
246 { NULL, 0, 0 }, /* Bq */
247 { NULL, TYPE_Bsx, NODE_NOSRC }, /* Bsx */
248 { NULL, TYPE_Bx, NODE_NOSRC }, /* Bx */
249 { NULL, 0, 0 }, /* Db */
250 { NULL, 0, 0 }, /* Dc */
251 { NULL, 0, 0 }, /* Do */
252 { NULL, 0, 0 }, /* Dq */
253 { NULL, 0, 0 }, /* Ec */
254 { NULL, 0, 0 }, /* Ef */
255 { NULL, TYPE_Em, 0 }, /* Em */
256 { NULL, 0, 0 }, /* Eo */
257 { NULL, TYPE_Fx, NODE_NOSRC }, /* Fx */
258 { NULL, TYPE_Ms, 0 }, /* Ms */
259 { NULL, 0, 0 }, /* No */
260 { NULL, 0, 0 }, /* Ns */
261 { NULL, TYPE_Nx, NODE_NOSRC }, /* Nx */
262 { NULL, TYPE_Ox, NODE_NOSRC }, /* Ox */
263 { NULL, 0, 0 }, /* Pc */
264 { NULL, 0, 0 }, /* Pf */
265 { NULL, 0, 0 }, /* Po */
266 { NULL, 0, 0 }, /* Pq */
267 { NULL, 0, 0 }, /* Qc */
268 { NULL, 0, 0 }, /* Ql */
269 { NULL, 0, 0 }, /* Qo */
270 { NULL, 0, 0 }, /* Qq */
271 { NULL, 0, 0 }, /* Re */
272 { NULL, 0, 0 }, /* Rs */
273 { NULL, 0, 0 }, /* Sc */
274 { NULL, 0, 0 }, /* So */
275 { NULL, 0, 0 }, /* Sq */
276 { NULL, 0, 0 }, /* Sm */
277 { NULL, 0, 0 }, /* Sx */
278 { NULL, TYPE_Sy, 0 }, /* Sy */
279 { NULL, TYPE_Tn, 0 }, /* Tn */
280 { NULL, 0, NODE_NOSRC }, /* Ux */
281 { NULL, 0, 0 }, /* Xc */
282 { NULL, 0, 0 }, /* Xo */
283 { parse_mdoc_Fo, 0, 0 }, /* Fo */
284 { NULL, 0, 0 }, /* Fc */
285 { NULL, 0, 0 }, /* Oo */
286 { NULL, 0, 0 }, /* Oc */
287 { NULL, 0, 0 }, /* Bk */
288 { NULL, 0, 0 }, /* Ek */
289 { NULL, 0, 0 }, /* Bt */
290 { NULL, 0, 0 }, /* Hf */
291 { NULL, 0, 0 }, /* Fr */
292 { NULL, 0, 0 }, /* Ud */
293 { NULL, TYPE_Lb, NODE_NOSRC }, /* Lb */
294 { NULL, 0, 0 }, /* Lp */
295 { NULL, TYPE_Lk, 0 }, /* Lk */
296 { NULL, TYPE_Mt, NODE_NOSRC }, /* Mt */
297 { NULL, 0, 0 }, /* Brq */
298 { NULL, 0, 0 }, /* Bro */
299 { NULL, 0, 0 }, /* Brc */
300 { NULL, 0, 0 }, /* %C */
301 { NULL, 0, 0 }, /* Es */
302 { NULL, 0, 0 }, /* En */
303 { NULL, TYPE_Dx, NODE_NOSRC }, /* Dx */
304 { NULL, 0, 0 }, /* %Q */
305 { NULL, 0, 0 }, /* %U */
306 { NULL, 0, 0 }, /* Ta */
307 };
308 static const struct mdoc_handler *const mdocs = __mdocs - MDOC_Dd;
309
310
311 int
312 mandocdb(int argc, char *argv[])
313 {
314 struct manconf conf;
315 struct mparse *mp;
316 struct dba *dba;
317 const char *path_arg, *progname;
318 size_t j, sz;
319 int ch, i;
320
321 #if HAVE_PLEDGE
322 if (pledge("stdio rpath wpath cpath", NULL) == -1) {
323 warn("pledge");
324 return (int)MANDOCLEVEL_SYSERR;
325 }
326 #endif
327
328 #if HAVE_SANDBOX_INIT
329 if (sandbox_init(kSBXProfileNoInternet, SANDBOX_NAMED, NULL) == -1) {
330 warnx("sandbox_init");
331 return (int)MANDOCLEVEL_SYSERR;
332 }
333 #endif
334
335 memset(&conf, 0, sizeof(conf));
336
337 /*
338 * We accept a few different invocations.
339 * The CHECKOP macro makes sure that invocation styles don't
340 * clobber each other.
341 */
342 #define CHECKOP(_op, _ch) do \
343 if (OP_DEFAULT != (_op)) { \
344 warnx("-%c: Conflicting option", (_ch)); \
345 goto usage; \
346 } while (/*CONSTCOND*/0)
347
348 path_arg = NULL;
349 op = OP_DEFAULT;
350
351 while (-1 != (ch = getopt(argc, argv, "aC:Dd:npQT:tu:v")))
352 switch (ch) {
353 case 'a':
354 use_all = 1;
355 break;
356 case 'C':
357 CHECKOP(op, ch);
358 path_arg = optarg;
359 op = OP_CONFFILE;
360 break;
361 case 'D':
362 debug++;
363 break;
364 case 'd':
365 CHECKOP(op, ch);
366 path_arg = optarg;
367 op = OP_UPDATE;
368 break;
369 case 'n':
370 nodb = 1;
371 break;
372 case 'p':
373 warnings = 1;
374 break;
375 case 'Q':
376 mparse_options |= MPARSE_QUICK;
377 break;
378 case 'T':
379 if (strcmp(optarg, "utf8")) {
380 warnx("-T%s: Unsupported output format",
381 optarg);
382 goto usage;
383 }
384 write_utf8 = 1;
385 break;
386 case 't':
387 CHECKOP(op, ch);
388 dup2(STDOUT_FILENO, STDERR_FILENO);
389 op = OP_TEST;
390 nodb = warnings = 1;
391 break;
392 case 'u':
393 CHECKOP(op, ch);
394 path_arg = optarg;
395 op = OP_DELETE;
396 break;
397 case 'v':
398 /* Compatibility with espie@'s makewhatis. */
399 break;
400 default:
401 goto usage;
402 }
403
404 argc -= optind;
405 argv += optind;
406
407 #if HAVE_PLEDGE
408 if (nodb) {
409 if (pledge("stdio rpath", NULL) == -1) {
410 warn("pledge");
411 return (int)MANDOCLEVEL_SYSERR;
412 }
413 }
414 #endif
415
416 if (OP_CONFFILE == op && argc > 0) {
417 warnx("-C: Too many arguments");
418 goto usage;
419 }
420
421 exitcode = (int)MANDOCLEVEL_OK;
422 mchars_alloc();
423 mp = mparse_alloc(mparse_options, MANDOCERR_MAX, NULL,
424 MANDOC_OS_OTHER, NULL);
425 mandoc_ohash_init(&mpages, 6, offsetof(struct mpage, inodev));
426 mandoc_ohash_init(&mlinks, 6, offsetof(struct mlink, file));
427
428 if (OP_UPDATE == op || OP_DELETE == op || OP_TEST == op) {
429
430 /*
431 * Most of these deal with a specific directory.
432 * Jump into that directory first.
433 */
434 if (OP_TEST != op && 0 == set_basedir(path_arg, 1))
435 goto out;
436
437 dba = nodb ? dba_new(128) : dba_read(MANDOC_DB);
438 if (dba != NULL) {
439 /*
440 * The existing database is usable. Process
441 * all files specified on the command-line.
442 */
443 use_all = 1;
444 for (i = 0; i < argc; i++)
445 filescan(argv[i]);
446 if (nodb == 0)
447 dbprune(dba);
448 } else {
449 /* Database missing or corrupt. */
450 if (op != OP_UPDATE || errno != ENOENT)
451 say(MANDOC_DB, "%s: Automatically recreating"
452 " from scratch", strerror(errno));
453 exitcode = (int)MANDOCLEVEL_OK;
454 op = OP_DEFAULT;
455 if (0 == treescan())
456 goto out;
457 dba = dba_new(128);
458 }
459 if (OP_DELETE != op)
460 mpages_merge(dba, mp);
461 if (nodb == 0)
462 dbwrite(dba);
463 dba_free(dba);
464 } else {
465 /*
466 * If we have arguments, use them as our manpaths.
467 * If we don't, use man.conf(5).
468 */
469 if (argc > 0) {
470 conf.manpath.paths = mandoc_reallocarray(NULL,
471 argc, sizeof(char *));
472 conf.manpath.sz = (size_t)argc;
473 for (i = 0; i < argc; i++)
474 conf.manpath.paths[i] = mandoc_strdup(argv[i]);
475 } else
476 manconf_parse(&conf, path_arg, NULL, NULL);
477
478 if (conf.manpath.sz == 0) {
479 exitcode = (int)MANDOCLEVEL_BADARG;
480 say("", "Empty manpath");
481 }
482
483 /*
484 * First scan the tree rooted at a base directory, then
485 * build a new database and finally move it into place.
486 * Ignore zero-length directories and strip trailing
487 * slashes.
488 */
489 for (j = 0; j < conf.manpath.sz; j++) {
490 sz = strlen(conf.manpath.paths[j]);
491 if (sz && conf.manpath.paths[j][sz - 1] == '/')
492 conf.manpath.paths[j][--sz] = '\0';
493 if (0 == sz)
494 continue;
495
496 if (j) {
497 mandoc_ohash_init(&mpages, 6,
498 offsetof(struct mpage, inodev));
499 mandoc_ohash_init(&mlinks, 6,
500 offsetof(struct mlink, file));
501 }
502
503 if ( ! set_basedir(conf.manpath.paths[j], argc > 0))
504 continue;
505 if (0 == treescan())
506 continue;
507 dba = dba_new(128);
508 mpages_merge(dba, mp);
509 if (nodb == 0)
510 dbwrite(dba);
511 dba_free(dba);
512
513 if (j + 1 < conf.manpath.sz) {
514 mpages_free();
515 ohash_delete(&mpages);
516 ohash_delete(&mlinks);
517 }
518 }
519 }
520 out:
521 manconf_free(&conf);
522 mparse_free(mp);
523 mchars_free();
524 mpages_free();
525 ohash_delete(&mpages);
526 ohash_delete(&mlinks);
527 return exitcode;
528 usage:
529 progname = getprogname();
530 fprintf(stderr, "usage: %s [-aDnpQ] [-C file] [-Tutf8]\n"
531 " %s [-aDnpQ] [-Tutf8] dir ...\n"
532 " %s [-DnpQ] [-Tutf8] -d dir [file ...]\n"
533 " %s [-Dnp] -u dir [file ...]\n"
534 " %s [-Q] -t file ...\n",
535 progname, progname, progname, progname, progname);
536
537 return (int)MANDOCLEVEL_BADARG;
538 }
539
540 /*
541 * To get a singly linked list in alpha order while inserting entries
542 * at the beginning, process directory entries in reverse alpha order.
543 */
544 static int
545 #if HAVE_FTS_COMPARE_CONST
546 fts_compare(const FTSENT *const *a, const FTSENT *const *b)
547 #else
548 fts_compare(const FTSENT **a, const FTSENT **b)
549 #endif
550 {
551 return -strcmp((*a)->fts_name, (*b)->fts_name);
552 }
553
554 /*
555 * Scan a directory tree rooted at "basedir" for manpages.
556 * We use fts(), scanning directory parts along the way for clues to our
557 * section and architecture.
558 *
559 * If use_all has been specified, grok all files.
560 * If not, sanitise paths to the following:
561 *
562 * [./]man*[/<arch>]/<name>.<section>
563 * or
564 * [./]cat<section>[/<arch>]/<name>.0
565 *
566 * TODO: accommodate for multi-language directories.
567 */
568 static int
569 treescan(void)
570 {
571 char buf[PATH_MAX];
572 FTS *f;
573 FTSENT *ff;
574 struct mlink *mlink;
575 int gzip;
576 enum form dform;
577 char *dsec, *arch, *fsec, *cp;
578 const char *path;
579 const char *argv[2];
580
581 argv[0] = ".";
582 argv[1] = NULL;
583
584 f = fts_open((char * const *)argv, FTS_PHYSICAL | FTS_NOCHDIR,
585 fts_compare);
586 if (f == NULL) {
587 exitcode = (int)MANDOCLEVEL_SYSERR;
588 say("", "&fts_open");
589 return 0;
590 }
591
592 dsec = arch = NULL;
593 dform = FORM_NONE;
594
595 while ((ff = fts_read(f)) != NULL) {
596 path = ff->fts_path + 2;
597 switch (ff->fts_info) {
598
599 /*
600 * Symbolic links require various sanity checks,
601 * then get handled just like regular files.
602 */
603 case FTS_SL:
604 if (realpath(path, buf) == NULL) {
605 if (warnings)
606 say(path, "&realpath");
607 continue;
608 }
609 if (strstr(buf, basedir) != buf
610 #ifdef HOMEBREWDIR
611 && strstr(buf, HOMEBREWDIR) != buf
612 #endif
613 ) {
614 if (warnings) say("",
615 "%s: outside base directory", buf);
616 continue;
617 }
618 /* Use logical inode to avoid mpages dupe. */
619 if (stat(path, ff->fts_statp) == -1) {
620 if (warnings)
621 say(path, "&stat");
622 continue;
623 }
624 /* FALLTHROUGH */
625
626 /*
627 * If we're a regular file, add an mlink by using the
628 * stored directory data and handling the filename.
629 */
630 case FTS_F:
631 if ( ! strcmp(path, MANDOC_DB))
632 continue;
633 if ( ! use_all && ff->fts_level < 2) {
634 if (warnings)
635 say(path, "Extraneous file");
636 continue;
637 }
638 gzip = 0;
639 fsec = NULL;
640 while (fsec == NULL) {
641 fsec = strrchr(ff->fts_name, '.');
642 if (fsec == NULL || strcmp(fsec+1, "gz"))
643 break;
644 gzip = 1;
645 *fsec = '\0';
646 fsec = NULL;
647 }
648 if (fsec == NULL) {
649 if ( ! use_all) {
650 if (warnings)
651 say(path,
652 "No filename suffix");
653 continue;
654 }
655 } else if ( ! strcmp(++fsec, "html")) {
656 if (warnings)
657 say(path, "Skip html");
658 continue;
659 } else if ( ! strcmp(fsec, "ps")) {
660 if (warnings)
661 say(path, "Skip ps");
662 continue;
663 } else if ( ! strcmp(fsec, "pdf")) {
664 if (warnings)
665 say(path, "Skip pdf");
666 continue;
667 } else if ( ! use_all &&
668 ((dform == FORM_SRC &&
669 strncmp(fsec, dsec, strlen(dsec))) ||
670 (dform == FORM_CAT && strcmp(fsec, "0")))) {
671 if (warnings)
672 say(path, "Wrong filename suffix");
673 continue;
674 } else
675 fsec[-1] = '\0';
676
677 mlink = mandoc_calloc(1, sizeof(struct mlink));
678 if (strlcpy(mlink->file, path,
679 sizeof(mlink->file)) >=
680 sizeof(mlink->file)) {
681 say(path, "Filename too long");
682 free(mlink);
683 continue;
684 }
685 mlink->dform = dform;
686 mlink->dsec = dsec;
687 mlink->arch = arch;
688 mlink->name = ff->fts_name;
689 mlink->fsec = fsec;
690 mlink->gzip = gzip;
691 mlink_add(mlink, ff->fts_statp);
692 continue;
693
694 case FTS_D:
695 case FTS_DP:
696 break;
697
698 default:
699 if (warnings)
700 say(path, "Not a regular file");
701 continue;
702 }
703
704 switch (ff->fts_level) {
705 case 0:
706 /* Ignore the root directory. */
707 break;
708 case 1:
709 /*
710 * This might contain manX/ or catX/.
711 * Try to infer this from the name.
712 * If we're not in use_all, enforce it.
713 */
714 cp = ff->fts_name;
715 if (ff->fts_info == FTS_DP) {
716 dform = FORM_NONE;
717 dsec = NULL;
718 break;
719 }
720
721 if ( ! strncmp(cp, "man", 3)) {
722 dform = FORM_SRC;
723 dsec = cp + 3;
724 } else if ( ! strncmp(cp, "cat", 3)) {
725 dform = FORM_CAT;
726 dsec = cp + 3;
727 } else {
728 dform = FORM_NONE;
729 dsec = NULL;
730 }
731
732 if (dsec != NULL || use_all)
733 break;
734
735 if (warnings)
736 say(path, "Unknown directory part");
737 fts_set(f, ff, FTS_SKIP);
738 break;
739 case 2:
740 /*
741 * Possibly our architecture.
742 * If we're descending, keep tabs on it.
743 */
744 if (ff->fts_info != FTS_DP && dsec != NULL)
745 arch = ff->fts_name;
746 else
747 arch = NULL;
748 break;
749 default:
750 if (ff->fts_info == FTS_DP || use_all)
751 break;
752 if (warnings)
753 say(path, "Extraneous directory part");
754 fts_set(f, ff, FTS_SKIP);
755 break;
756 }
757 }
758
759 fts_close(f);
760 return 1;
761 }
762
763 /*
764 * Add a file to the mlinks table.
765 * Do not verify that it's a "valid" looking manpage (we'll do that
766 * later).
767 *
768 * Try to infer the manual section, architecture, and page name from the
769 * path, assuming it looks like
770 *
771 * [./]man*[/<arch>]/<name>.<section>
772 * or
773 * [./]cat<section>[/<arch>]/<name>.0
774 *
775 * See treescan() for the fts(3) version of this.
776 */
777 static void
778 filescan(const char *file)
779 {
780 char buf[PATH_MAX];
781 struct stat st;
782 struct mlink *mlink;
783 char *p, *start;
784
785 assert(use_all);
786
787 if (0 == strncmp(file, "./", 2))
788 file += 2;
789
790 /*
791 * We have to do lstat(2) before realpath(3) loses
792 * the information whether this is a symbolic link.
793 * We need to know that because for symbolic links,
794 * we want to use the orginal file name, while for
795 * regular files, we want to use the real path.
796 */
797 if (-1 == lstat(file, &st)) {
798 exitcode = (int)MANDOCLEVEL_BADARG;
799 say(file, "&lstat");
800 return;
801 } else if (0 == ((S_IFREG | S_IFLNK) & st.st_mode)) {
802 exitcode = (int)MANDOCLEVEL_BADARG;
803 say(file, "Not a regular file");
804 return;
805 }
806
807 /*
808 * We have to resolve the file name to the real path
809 * in any case for the base directory check.
810 */
811 if (NULL == realpath(file, buf)) {
812 exitcode = (int)MANDOCLEVEL_BADARG;
813 say(file, "&realpath");
814 return;
815 }
816
817 if (OP_TEST == op)
818 start = buf;
819 else if (strstr(buf, basedir) == buf)
820 start = buf + strlen(basedir);
821 #ifdef HOMEBREWDIR
822 else if (strstr(buf, HOMEBREWDIR) == buf)
823 start = buf;
824 #endif
825 else {
826 exitcode = (int)MANDOCLEVEL_BADARG;
827 say("", "%s: outside base directory", buf);
828 return;
829 }
830
831 /*
832 * Now we are sure the file is inside our tree.
833 * If it is a symbolic link, ignore the real path
834 * and use the original name.
835 * This implies passing stuff like "cat1/../man1/foo.1"
836 * on the command line won't work. So don't do that.
837 * Note the stat(2) can still fail if the link target
838 * doesn't exist.
839 */
840 if (S_IFLNK & st.st_mode) {
841 if (-1 == stat(buf, &st)) {
842 exitcode = (int)MANDOCLEVEL_BADARG;
843 say(file, "&stat");
844 return;
845 }
846 if (strlcpy(buf, file, sizeof(buf)) >= sizeof(buf)) {
847 say(file, "Filename too long");
848 return;
849 }
850 start = buf;
851 if (OP_TEST != op && strstr(buf, basedir) == buf)
852 start += strlen(basedir);
853 }
854
855 mlink = mandoc_calloc(1, sizeof(struct mlink));
856 mlink->dform = FORM_NONE;
857 if (strlcpy(mlink->file, start, sizeof(mlink->file)) >=
858 sizeof(mlink->file)) {
859 say(start, "Filename too long");
860 free(mlink);
861 return;
862 }
863
864 /*
865 * In test mode or when the original name is absolute
866 * but outside our tree, guess the base directory.
867 */
868
869 if (op == OP_TEST || (start == buf && *start == '/')) {
870 if (strncmp(buf, "man/", 4) == 0)
871 start = buf + 4;
872 else if ((start = strstr(buf, "/man/")) != NULL)
873 start += 5;
874 else
875 start = buf;
876 }
877
878 /*
879 * First try to guess our directory structure.
880 * If we find a separator, try to look for man* or cat*.
881 * If we find one of these and what's underneath is a directory,
882 * assume it's an architecture.
883 */
884 if (NULL != (p = strchr(start, '/'))) {
885 *p++ = '\0';
886 if (0 == strncmp(start, "man", 3)) {
887 mlink->dform = FORM_SRC;
888 mlink->dsec = start + 3;
889 } else if (0 == strncmp(start, "cat", 3)) {
890 mlink->dform = FORM_CAT;
891 mlink->dsec = start + 3;
892 }
893
894 start = p;
895 if (NULL != mlink->dsec && NULL != (p = strchr(start, '/'))) {
896 *p++ = '\0';
897 mlink->arch = start;
898 start = p;
899 }
900 }
901
902 /*
903 * Now check the file suffix.
904 * Suffix of `.0' indicates a catpage, `.1-9' is a manpage.
905 */
906 p = strrchr(start, '\0');
907 while (p-- > start && '/' != *p && '.' != *p)
908 /* Loop. */ ;
909
910 if ('.' == *p) {
911 *p++ = '\0';
912 mlink->fsec = p;
913 }
914
915 /*
916 * Now try to parse the name.
917 * Use the filename portion of the path.
918 */
919 mlink->name = start;
920 if (NULL != (p = strrchr(start, '/'))) {
921 mlink->name = p + 1;
922 *p = '\0';
923 }
924 mlink_add(mlink, &st);
925 }
926
927 static void
928 mlink_add(struct mlink *mlink, const struct stat *st)
929 {
930 struct inodev inodev;
931 struct mpage *mpage;
932 unsigned int slot;
933
934 assert(NULL != mlink->file);
935
936 mlink->dsec = mandoc_strdup(mlink->dsec ? mlink->dsec : "");
937 mlink->arch = mandoc_strdup(mlink->arch ? mlink->arch : "");
938 mlink->name = mandoc_strdup(mlink->name ? mlink->name : "");
939 mlink->fsec = mandoc_strdup(mlink->fsec ? mlink->fsec : "");
940
941 if ('0' == *mlink->fsec) {
942 free(mlink->fsec);
943 mlink->fsec = mandoc_strdup(mlink->dsec);
944 mlink->fform = FORM_CAT;
945 } else if ('1' <= *mlink->fsec && '9' >= *mlink->fsec)
946 mlink->fform = FORM_SRC;
947 else
948 mlink->fform = FORM_NONE;
949
950 slot = ohash_qlookup(&mlinks, mlink->file);
951 assert(NULL == ohash_find(&mlinks, slot));
952 ohash_insert(&mlinks, slot, mlink);
953
954 memset(&inodev, 0, sizeof(inodev)); /* Clear padding. */
955 inodev.st_ino = st->st_ino;
956 inodev.st_dev = st->st_dev;
957 slot = ohash_lookup_memory(&mpages, (char *)&inodev,
958 sizeof(struct inodev), inodev.st_ino);
959 mpage = ohash_find(&mpages, slot);
960 if (NULL == mpage) {
961 mpage = mandoc_calloc(1, sizeof(struct mpage));
962 mpage->inodev.st_ino = inodev.st_ino;
963 mpage->inodev.st_dev = inodev.st_dev;
964 mpage->form = FORM_NONE;
965 mpage->next = mpage_head;
966 mpage_head = mpage;
967 ohash_insert(&mpages, slot, mpage);
968 } else
969 mlink->next = mpage->mlinks;
970 mpage->mlinks = mlink;
971 mlink->mpage = mpage;
972 }
973
974 static void
975 mlink_free(struct mlink *mlink)
976 {
977
978 free(mlink->dsec);
979 free(mlink->arch);
980 free(mlink->name);
981 free(mlink->fsec);
982 free(mlink);
983 }
984
985 static void
986 mpages_free(void)
987 {
988 struct mpage *mpage;
989 struct mlink *mlink;
990
991 while ((mpage = mpage_head) != NULL) {
992 while ((mlink = mpage->mlinks) != NULL) {
993 mpage->mlinks = mlink->next;
994 mlink_free(mlink);
995 }
996 mpage_head = mpage->next;
997 free(mpage->sec);
998 free(mpage->arch);
999 free(mpage->title);
1000 free(mpage->desc);
1001 free(mpage);
1002 }
1003 }
1004
1005 /*
1006 * For each mlink to the mpage, check whether the path looks like
1007 * it is formatted, and if it does, check whether a source manual
1008 * exists by the same name, ignoring the suffix.
1009 * If both conditions hold, drop the mlink.
1010 */
1011 static void
1012 mlinks_undupe(struct mpage *mpage)
1013 {
1014 char buf[PATH_MAX];
1015 struct mlink **prev;
1016 struct mlink *mlink;
1017 char *bufp;
1018
1019 mpage->form = FORM_CAT;
1020 prev = &mpage->mlinks;
1021 while (NULL != (mlink = *prev)) {
1022 if (FORM_CAT != mlink->dform) {
1023 mpage->form = FORM_NONE;
1024 goto nextlink;
1025 }
1026 (void)strlcpy(buf, mlink->file, sizeof(buf));
1027 bufp = strstr(buf, "cat");
1028 assert(NULL != bufp);
1029 memcpy(bufp, "man", 3);
1030 if (NULL != (bufp = strrchr(buf, '.')))
1031 *++bufp = '\0';
1032 (void)strlcat(buf, mlink->dsec, sizeof(buf));
1033 if (NULL == ohash_find(&mlinks,
1034 ohash_qlookup(&mlinks, buf)))
1035 goto nextlink;
1036 if (warnings)
1037 say(mlink->file, "Man source exists: %s", buf);
1038 if (use_all)
1039 goto nextlink;
1040 *prev = mlink->next;
1041 mlink_free(mlink);
1042 continue;
1043 nextlink:
1044 prev = &(*prev)->next;
1045 }
1046 }
1047
1048 static void
1049 mlink_check(struct mpage *mpage, struct mlink *mlink)
1050 {
1051 struct str *str;
1052 unsigned int slot;
1053
1054 /*
1055 * Check whether the manual section given in a file
1056 * agrees with the directory where the file is located.
1057 * Some manuals have suffixes like (3p) on their
1058 * section number either inside the file or in the
1059 * directory name, some are linked into more than one
1060 * section, like encrypt(1) = makekey(8).
1061 */
1062
1063 if (FORM_SRC == mpage->form &&
1064 strcasecmp(mpage->sec, mlink->dsec))
1065 say(mlink->file, "Section \"%s\" manual in %s directory",
1066 mpage->sec, mlink->dsec);
1067
1068 /*
1069 * Manual page directories exist for each kernel
1070 * architecture as returned by machine(1).
1071 * However, many manuals only depend on the
1072 * application architecture as returned by arch(1).
1073 * For example, some (2/ARM) manuals are shared
1074 * across the "armish" and "zaurus" kernel
1075 * architectures.
1076 * A few manuals are even shared across completely
1077 * different architectures, for example fdformat(1)
1078 * on amd64, i386, and sparc64.
1079 */
1080
1081 if (strcasecmp(mpage->arch, mlink->arch))
1082 say(mlink->file, "Architecture \"%s\" manual in "
1083 "\"%s\" directory", mpage->arch, mlink->arch);
1084
1085 /*
1086 * XXX
1087 * parse_cat() doesn't set NAME_TITLE yet.
1088 */
1089
1090 if (FORM_CAT == mpage->form)
1091 return;
1092
1093 /*
1094 * Check whether this mlink
1095 * appears as a name in the NAME section.
1096 */
1097
1098 slot = ohash_qlookup(&names, mlink->name);
1099 str = ohash_find(&names, slot);
1100 assert(NULL != str);
1101 if ( ! (NAME_TITLE & str->mask))
1102 say(mlink->file, "Name missing in NAME section");
1103 }
1104
1105 /*
1106 * Run through the files in the global vector "mpages"
1107 * and add them to the database specified in "basedir".
1108 *
1109 * This handles the parsing scheme itself, using the cues of directory
1110 * and filename to determine whether the file is parsable or not.
1111 */
1112 static void
1113 mpages_merge(struct dba *dba, struct mparse *mp)
1114 {
1115 struct mpage *mpage, *mpage_dest;
1116 struct mlink *mlink, *mlink_dest;
1117 struct roff_man *man;
1118 char *sodest;
1119 char *cp;
1120 int fd;
1121
1122 for (mpage = mpage_head; mpage != NULL; mpage = mpage->next) {
1123 mlinks_undupe(mpage);
1124 if ((mlink = mpage->mlinks) == NULL)
1125 continue;
1126
1127 name_mask = NAME_MASK;
1128 mandoc_ohash_init(&names, 4, offsetof(struct str, key));
1129 mandoc_ohash_init(&strings, 6, offsetof(struct str, key));
1130 mparse_reset(mp);
1131 man = NULL;
1132 sodest = NULL;
1133
1134 if ((fd = mparse_open(mp, mlink->file)) == -1) {
1135 say(mlink->file, "&open");
1136 goto nextpage;
1137 }
1138
1139 /*
1140 * Interpret the file as mdoc(7) or man(7) source
1141 * code, unless it is known to be formatted.
1142 */
1143 if (mlink->dform != FORM_CAT || mlink->fform != FORM_CAT) {
1144 mparse_readfd(mp, fd, mlink->file);
1145 close(fd);
1146 fd = -1;
1147 mparse_result(mp, &man, &sodest);
1148 }
1149
1150 if (sodest != NULL) {
1151 mlink_dest = ohash_find(&mlinks,
1152 ohash_qlookup(&mlinks, sodest));
1153 if (mlink_dest == NULL) {
1154 mandoc_asprintf(&cp, "%s.gz", sodest);
1155 mlink_dest = ohash_find(&mlinks,
1156 ohash_qlookup(&mlinks, cp));
1157 free(cp);
1158 }
1159 if (mlink_dest != NULL) {
1160
1161 /* The .so target exists. */
1162
1163 mpage_dest = mlink_dest->mpage;
1164 while (1) {
1165 mlink->mpage = mpage_dest;
1166
1167 /*
1168 * If the target was already
1169 * processed, add the links
1170 * to the database now.
1171 * Otherwise, this will
1172 * happen when we come
1173 * to the target.
1174 */
1175
1176 if (mpage_dest->dba != NULL)
1177 dbadd_mlink(mlink);
1178
1179 if (mlink->next == NULL)
1180 break;
1181 mlink = mlink->next;
1182 }
1183
1184 /* Move all links to the target. */
1185
1186 mlink->next = mlink_dest->next;
1187 mlink_dest->next = mpage->mlinks;
1188 mpage->mlinks = NULL;
1189 }
1190 goto nextpage;
1191 } else if (man != NULL && man->macroset == MACROSET_MDOC) {
1192 mdoc_validate(man);
1193 mpage->form = FORM_SRC;
1194 mpage->sec = man->meta.msec;
1195 mpage->sec = mandoc_strdup(
1196 mpage->sec == NULL ? "" : mpage->sec);
1197 mpage->arch = man->meta.arch;
1198 mpage->arch = mandoc_strdup(
1199 mpage->arch == NULL ? "" : mpage->arch);
1200 mpage->title = mandoc_strdup(man->meta.title);
1201 } else if (man != NULL && man->macroset == MACROSET_MAN) {
1202 man_validate(man);
1203 if (*man->meta.msec != '\0' ||
1204 *man->meta.title != '\0') {
1205 mpage->form = FORM_SRC;
1206 mpage->sec = mandoc_strdup(man->meta.msec);
1207 mpage->arch = mandoc_strdup(mlink->arch);
1208 mpage->title = mandoc_strdup(man->meta.title);
1209 } else
1210 man = NULL;
1211 }
1212
1213 assert(mpage->desc == NULL);
1214 if (man == NULL) {
1215 mpage->form = FORM_CAT;
1216 mpage->sec = mandoc_strdup(mlink->dsec);
1217 mpage->arch = mandoc_strdup(mlink->arch);
1218 mpage->title = mandoc_strdup(mlink->name);
1219 parse_cat(mpage, fd);
1220 } else if (man->macroset == MACROSET_MDOC)
1221 parse_mdoc(mpage, &man->meta, man->first);
1222 else
1223 parse_man(mpage, &man->meta, man->first);
1224 if (mpage->desc == NULL) {
1225 mpage->desc = mandoc_strdup(mlink->name);
1226 if (warnings)
1227 say(mlink->file, "No one-line description, "
1228 "using filename \"%s\"", mlink->name);
1229 }
1230
1231 for (mlink = mpage->mlinks;
1232 mlink != NULL;
1233 mlink = mlink->next) {
1234 putkey(mpage, mlink->name, NAME_FILE);
1235 if (warnings && !use_all)
1236 mlink_check(mpage, mlink);
1237 }
1238
1239 dbadd(dba, mpage);
1240
1241 nextpage:
1242 ohash_delete(&strings);
1243 ohash_delete(&names);
1244 }
1245 }
1246
1247 static void
1248 parse_cat(struct mpage *mpage, int fd)
1249 {
1250 FILE *stream;
1251 struct mlink *mlink;
1252 char *line, *p, *title, *sec;
1253 size_t linesz, plen, titlesz;
1254 ssize_t len;
1255 int offs;
1256
1257 mlink = mpage->mlinks;
1258 stream = fd == -1 ? fopen(mlink->file, "r") : fdopen(fd, "r");
1259 if (stream == NULL) {
1260 if (fd != -1)
1261 close(fd);
1262 if (warnings)
1263 say(mlink->file, "&fopen");
1264 return;
1265 }
1266
1267 line = NULL;
1268 linesz = 0;
1269
1270 /* Parse the section number from the header line. */
1271
1272 while (getline(&line, &linesz, stream) != -1) {
1273 if (*line == '\n')
1274 continue;
1275 if ((sec = strchr(line, '(')) == NULL)
1276 break;
1277 if ((p = strchr(++sec, ')')) == NULL)
1278 break;
1279 free(mpage->sec);
1280 mpage->sec = mandoc_strndup(sec, p - sec);
1281 if (warnings && *mlink->dsec != '\0' &&
1282 strcasecmp(mpage->sec, mlink->dsec))
1283 say(mlink->file,
1284 "Section \"%s\" manual in %s directory",
1285 mpage->sec, mlink->dsec);
1286 break;
1287 }
1288
1289 /* Skip to first blank line. */
1290
1291 while (line == NULL || *line != '\n')
1292 if (getline(&line, &linesz, stream) == -1)
1293 break;
1294
1295 /*
1296 * Assume the first line that is not indented
1297 * is the first section header. Skip to it.
1298 */
1299
1300 while (getline(&line, &linesz, stream) != -1)
1301 if (*line != '\n' && *line != ' ')
1302 break;
1303
1304 /*
1305 * Read up until the next section into a buffer.
1306 * Strip the leading and trailing newline from each read line,
1307 * appending a trailing space.
1308 * Ignore empty (whitespace-only) lines.
1309 */
1310
1311 titlesz = 0;
1312 title = NULL;
1313
1314 while ((len = getline(&line, &linesz, stream)) != -1) {
1315 if (*line != ' ')
1316 break;
1317 offs = 0;
1318 while (isspace((unsigned char)line[offs]))
1319 offs++;
1320 if (line[offs] == '\0')
1321 continue;
1322 title = mandoc_realloc(title, titlesz + len - offs);
1323 memcpy(title + titlesz, line + offs, len - offs);
1324 titlesz += len - offs;
1325 title[titlesz - 1] = ' ';
1326 }
1327 free(line);
1328
1329 /*
1330 * If no page content can be found, or the input line
1331 * is already the next section header, or there is no
1332 * trailing newline, reuse the page title as the page
1333 * description.
1334 */
1335
1336 if (NULL == title || '\0' == *title) {
1337 if (warnings)
1338 say(mlink->file, "Cannot find NAME section");
1339 fclose(stream);
1340 free(title);
1341 return;
1342 }
1343
1344 title[titlesz - 1] = '\0';
1345
1346 /*
1347 * Skip to the first dash.
1348 * Use the remaining line as the description (no more than 70
1349 * bytes).
1350 */
1351
1352 if (NULL != (p = strstr(title, "- "))) {
1353 for (p += 2; ' ' == *p || '\b' == *p; p++)
1354 /* Skip to next word. */ ;
1355 } else {
1356 if (warnings)
1357 say(mlink->file, "No dash in title line, "
1358 "reusing \"%s\" as one-line description", title);
1359 p = title;
1360 }
1361
1362 plen = strlen(p);
1363
1364 /* Strip backspace-encoding from line. */
1365
1366 while (NULL != (line = memchr(p, '\b', plen))) {
1367 len = line - p;
1368 if (0 == len) {
1369 memmove(line, line + 1, plen--);
1370 continue;
1371 }
1372 memmove(line - 1, line + 1, plen - len);
1373 plen -= 2;
1374 }
1375
1376 /*
1377 * Cut off excessive one-line descriptions.
1378 * Bad pages are not worth better heuristics.
1379 */
1380
1381 mpage->desc = mandoc_strndup(p, 150);
1382 fclose(stream);
1383 free(title);
1384 }
1385
1386 /*
1387 * Put a type/word pair into the word database for this particular file.
1388 */
1389 static void
1390 putkey(const struct mpage *mpage, char *value, uint64_t type)
1391 {
1392 putkeys(mpage, value, strlen(value), type);
1393 }
1394
1395 /*
1396 * Grok all nodes at or below a certain mdoc node into putkey().
1397 */
1398 static void
1399 putmdockey(const struct mpage *mpage,
1400 const struct roff_node *n, uint64_t m, int taboo)
1401 {
1402
1403 for ( ; NULL != n; n = n->next) {
1404 if (n->flags & taboo)
1405 continue;
1406 if (NULL != n->child)
1407 putmdockey(mpage, n->child, m, taboo);
1408 if (n->type == ROFFT_TEXT)
1409 putkey(mpage, n->string, m);
1410 }
1411 }
1412
1413 static void
1414 parse_man(struct mpage *mpage, const struct roff_meta *meta,
1415 const struct roff_node *n)
1416 {
1417 const struct roff_node *head, *body;
1418 char *start, *title;
1419 char byte;
1420 size_t sz;
1421
1422 if (n == NULL)
1423 return;
1424
1425 /*
1426 * We're only searching for one thing: the first text child in
1427 * the BODY of a NAME section. Since we don't keep track of
1428 * sections in -man, run some hoops to find out whether we're in
1429 * the correct section or not.
1430 */
1431
1432 if (n->type == ROFFT_BODY && n->tok == MAN_SH) {
1433 body = n;
1434 if ((head = body->parent->head) != NULL &&
1435 (head = head->child) != NULL &&
1436 head->next == NULL &&
1437 head->type == ROFFT_TEXT &&
1438 strcmp(head->string, "NAME") == 0 &&
1439 body->child != NULL) {
1440
1441 /*
1442 * Suck the entire NAME section into memory.
1443 * Yes, we might run away.
1444 * But too many manuals have big, spread-out
1445 * NAME sections over many lines.
1446 */
1447
1448 title = NULL;
1449 deroff(&title, body);
1450 if (NULL == title)
1451 return;
1452
1453 /*
1454 * Go through a special heuristic dance here.
1455 * Conventionally, one or more manual names are
1456 * comma-specified prior to a whitespace, then a
1457 * dash, then a description. Try to puzzle out
1458 * the name parts here.
1459 */
1460
1461 start = title;
1462 for ( ;; ) {
1463 sz = strcspn(start, " ,");
1464 if ('\0' == start[sz])
1465 break;
1466
1467 byte = start[sz];
1468 start[sz] = '\0';
1469
1470 /*
1471 * Assume a stray trailing comma in the
1472 * name list if a name begins with a dash.
1473 */
1474
1475 if ('-' == start[0] ||
1476 ('\\' == start[0] && '-' == start[1]))
1477 break;
1478
1479 putkey(mpage, start, NAME_TITLE);
1480 if ( ! (mpage->name_head_done ||
1481 strcasecmp(start, meta->title))) {
1482 putkey(mpage, start, NAME_HEAD);
1483 mpage->name_head_done = 1;
1484 }
1485
1486 if (' ' == byte) {
1487 start += sz + 1;
1488 break;
1489 }
1490
1491 assert(',' == byte);
1492 start += sz + 1;
1493 while (' ' == *start)
1494 start++;
1495 }
1496
1497 if (start == title) {
1498 putkey(mpage, start, NAME_TITLE);
1499 if ( ! (mpage->name_head_done ||
1500 strcasecmp(start, meta->title))) {
1501 putkey(mpage, start, NAME_HEAD);
1502 mpage->name_head_done = 1;
1503 }
1504 free(title);
1505 return;
1506 }
1507
1508 while (isspace((unsigned char)*start))
1509 start++;
1510
1511 if (0 == strncmp(start, "-", 1))
1512 start += 1;
1513 else if (0 == strncmp(start, "\\-\\-", 4))
1514 start += 4;
1515 else if (0 == strncmp(start, "\\-", 2))
1516 start += 2;
1517 else if (0 == strncmp(start, "\\(en", 4))
1518 start += 4;
1519 else if (0 == strncmp(start, "\\(em", 4))
1520 start += 4;
1521
1522 while (' ' == *start)
1523 start++;
1524
1525 /*
1526 * Cut off excessive one-line descriptions.
1527 * Bad pages are not worth better heuristics.
1528 */
1529
1530 mpage->desc = mandoc_strndup(start, 150);
1531 free(title);
1532 return;
1533 }
1534 }
1535
1536 for (n = n->child; n; n = n->next) {
1537 if (NULL != mpage->desc)
1538 break;
1539 parse_man(mpage, meta, n);
1540 }
1541 }
1542
1543 static void
1544 parse_mdoc(struct mpage *mpage, const struct roff_meta *meta,
1545 const struct roff_node *n)
1546 {
1547
1548 for (n = n->child; n != NULL; n = n->next) {
1549 if (n->tok == TOKEN_NONE ||
1550 n->tok < ROFF_MAX ||
1551 n->flags & mdocs[n->tok].taboo)
1552 continue;
1553 assert(n->tok >= MDOC_Dd && n->tok < MDOC_MAX);
1554 switch (n->type) {
1555 case ROFFT_ELEM:
1556 case ROFFT_BLOCK:
1557 case ROFFT_HEAD:
1558 case ROFFT_BODY:
1559 case ROFFT_TAIL:
1560 if (mdocs[n->tok].fp != NULL &&
1561 (*mdocs[n->tok].fp)(mpage, meta, n) == 0)
1562 break;
1563 if (mdocs[n->tok].mask)
1564 putmdockey(mpage, n->child,
1565 mdocs[n->tok].mask, mdocs[n->tok].taboo);
1566 break;
1567 default:
1568 continue;
1569 }
1570 if (NULL != n->child)
1571 parse_mdoc(mpage, meta, n);
1572 }
1573 }
1574
1575 static int
1576 parse_mdoc_Fd(struct mpage *mpage, const struct roff_meta *meta,
1577 const struct roff_node *n)
1578 {
1579 char *start, *end;
1580 size_t sz;
1581
1582 if (SEC_SYNOPSIS != n->sec ||
1583 NULL == (n = n->child) ||
1584 n->type != ROFFT_TEXT)
1585 return 0;
1586
1587 /*
1588 * Only consider those `Fd' macro fields that begin with an
1589 * "inclusion" token (versus, e.g., #define).
1590 */
1591
1592 if (strcmp("#include", n->string))
1593 return 0;
1594
1595 if ((n = n->next) == NULL || n->type != ROFFT_TEXT)
1596 return 0;
1597
1598 /*
1599 * Strip away the enclosing angle brackets and make sure we're
1600 * not zero-length.
1601 */
1602
1603 start = n->string;
1604 if ('<' == *start || '"' == *start)
1605 start++;
1606
1607 if (0 == (sz = strlen(start)))
1608 return 0;
1609
1610 end = &start[(int)sz - 1];
1611 if ('>' == *end || '"' == *end)
1612 end--;
1613
1614 if (end > start)
1615 putkeys(mpage, start, end - start + 1, TYPE_In);
1616 return 0;
1617 }
1618
1619 static void
1620 parse_mdoc_fname(struct mpage *mpage, const struct roff_node *n)
1621 {
1622 char *cp;
1623 size_t sz;
1624
1625 if (n->type != ROFFT_TEXT)
1626 return;
1627
1628 /* Skip function pointer punctuation. */
1629
1630 cp = n->string;
1631 while (*cp == '(' || *cp == '*')
1632 cp++;
1633 sz = strcspn(cp, "()");
1634
1635 putkeys(mpage, cp, sz, TYPE_Fn);
1636 if (n->sec == SEC_SYNOPSIS)
1637 putkeys(mpage, cp, sz, NAME_SYN);
1638 }
1639
1640 static int
1641 parse_mdoc_Fn(struct mpage *mpage, const struct roff_meta *meta,
1642 const struct roff_node *n)
1643 {
1644
1645 if (n->child == NULL)
1646 return 0;
1647
1648 parse_mdoc_fname(mpage, n->child);
1649
1650 for (n = n->child->next; n != NULL; n = n->next)
1651 if (n->type == ROFFT_TEXT)
1652 putkey(mpage, n->string, TYPE_Fa);
1653
1654 return 0;
1655 }
1656
1657 static int
1658 parse_mdoc_Fo(struct mpage *mpage, const struct roff_meta *meta,
1659 const struct roff_node *n)
1660 {
1661
1662 if (n->type != ROFFT_HEAD)
1663 return 1;
1664
1665 if (n->child != NULL)
1666 parse_mdoc_fname(mpage, n->child);
1667
1668 return 0;
1669 }
1670
1671 static int
1672 parse_mdoc_Va(struct mpage *mpage, const struct roff_meta *meta,
1673 const struct roff_node *n)
1674 {
1675 char *cp;
1676
1677 if (n->type != ROFFT_ELEM && n->type != ROFFT_BODY)
1678 return 0;
1679
1680 if (n->child != NULL &&
1681 n->child->next == NULL &&
1682 n->child->type == ROFFT_TEXT)
1683 return 1;
1684
1685 cp = NULL;
1686 deroff(&cp, n);
1687 if (cp != NULL) {
1688 putkey(mpage, cp, TYPE_Vt | (n->tok == MDOC_Va ||
1689 n->type == ROFFT_BODY ? TYPE_Va : 0));
1690 free(cp);
1691 }
1692
1693 return 0;
1694 }
1695
1696 static int
1697 parse_mdoc_Xr(struct mpage *mpage, const struct roff_meta *meta,
1698 const struct roff_node *n)
1699 {
1700 char *cp;
1701
1702 if (NULL == (n = n->child))
1703 return 0;
1704
1705 if (NULL == n->next) {
1706 putkey(mpage, n->string, TYPE_Xr);
1707 return 0;
1708 }
1709
1710 mandoc_asprintf(&cp, "%s(%s)", n->string, n->next->string);
1711 putkey(mpage, cp, TYPE_Xr);
1712 free(cp);
1713 return 0;
1714 }
1715
1716 static int
1717 parse_mdoc_Nd(struct mpage *mpage, const struct roff_meta *meta,
1718 const struct roff_node *n)
1719 {
1720
1721 if (n->type == ROFFT_BODY)
1722 deroff(&mpage->desc, n);
1723 return 0;
1724 }
1725
1726 static int
1727 parse_mdoc_Nm(struct mpage *mpage, const struct roff_meta *meta,
1728 const struct roff_node *n)
1729 {
1730
1731 if (SEC_NAME == n->sec)
1732 putmdockey(mpage, n->child, NAME_TITLE, 0);
1733 else if (n->sec == SEC_SYNOPSIS && n->type == ROFFT_HEAD) {
1734 if (n->child == NULL)
1735 putkey(mpage, meta->name, NAME_SYN);
1736 else
1737 putmdockey(mpage, n->child, NAME_SYN, 0);
1738 }
1739 if ( ! (mpage->name_head_done ||
1740 n->child == NULL || n->child->string == NULL ||
1741 strcasecmp(n->child->string, meta->title))) {
1742 putkey(mpage, n->child->string, NAME_HEAD);
1743 mpage->name_head_done = 1;
1744 }
1745 return 0;
1746 }
1747
1748 static int
1749 parse_mdoc_Sh(struct mpage *mpage, const struct roff_meta *meta,
1750 const struct roff_node *n)
1751 {
1752
1753 return n->sec == SEC_CUSTOM && n->type == ROFFT_HEAD;
1754 }
1755
1756 static int
1757 parse_mdoc_head(struct mpage *mpage, const struct roff_meta *meta,
1758 const struct roff_node *n)
1759 {
1760
1761 return n->type == ROFFT_HEAD;
1762 }
1763
1764 /*
1765 * Add a string to the hash table for the current manual.
1766 * Each string has a bitmask telling which macros it belongs to.
1767 * When we finish the manual, we'll dump the table.
1768 */
1769 static void
1770 putkeys(const struct mpage *mpage, char *cp, size_t sz, uint64_t v)
1771 {
1772 struct ohash *htab;
1773 struct str *s;
1774 const char *end;
1775 unsigned int slot;
1776 int i, mustfree;
1777
1778 if (0 == sz)
1779 return;
1780
1781 mustfree = render_string(&cp, &sz);
1782
1783 if (TYPE_Nm & v) {
1784 htab = &names;
1785 v &= name_mask;
1786 if (v & NAME_FIRST)
1787 name_mask &= ~NAME_FIRST;
1788 if (debug > 1)
1789 say(mpage->mlinks->file,
1790 "Adding name %*s, bits=0x%llx", (int)sz, cp,
1791 (unsigned long long)v);
1792 } else {
1793 htab = &strings;
1794 if (debug > 1)
1795 for (i = 0; i < KEY_MAX; i++)
1796 if ((uint64_t)1 << i & v)
1797 say(mpage->mlinks->file,
1798 "Adding key %s=%*s",
1799 mansearch_keynames[i], (int)sz, cp);
1800 }
1801
1802 end = cp + sz;
1803 slot = ohash_qlookupi(htab, cp, &end);
1804 s = ohash_find(htab, slot);
1805
1806 if (NULL != s && mpage == s->mpage) {
1807 s->mask |= v;
1808 return;
1809 } else if (NULL == s) {
1810 s = mandoc_calloc(1, sizeof(struct str) + sz + 1);
1811 memcpy(s->key, cp, sz);
1812 ohash_insert(htab, slot, s);
1813 }
1814 s->mpage = mpage;
1815 s->mask = v;
1816
1817 if (mustfree)
1818 free(cp);
1819 }
1820
1821 /*
1822 * Take a Unicode codepoint and produce its UTF-8 encoding.
1823 * This isn't the best way to do this, but it works.
1824 * The magic numbers are from the UTF-8 packaging.
1825 * They're not as scary as they seem: read the UTF-8 spec for details.
1826 */
1827 static size_t
1828 utf8(unsigned int cp, char out[7])
1829 {
1830 size_t rc;
1831
1832 rc = 0;
1833 if (cp <= 0x0000007F) {
1834 rc = 1;
1835 out[0] = (char)cp;
1836 } else if (cp <= 0x000007FF) {
1837 rc = 2;
1838 out[0] = (cp >> 6 & 31) | 192;
1839 out[1] = (cp & 63) | 128;
1840 } else if (cp <= 0x0000FFFF) {
1841 rc = 3;
1842 out[0] = (cp >> 12 & 15) | 224;
1843 out[1] = (cp >> 6 & 63) | 128;
1844 out[2] = (cp & 63) | 128;
1845 } else if (cp <= 0x001FFFFF) {
1846 rc = 4;
1847 out[0] = (cp >> 18 & 7) | 240;
1848 out[1] = (cp >> 12 & 63) | 128;
1849 out[2] = (cp >> 6 & 63) | 128;
1850 out[3] = (cp & 63) | 128;
1851 } else if (cp <= 0x03FFFFFF) {
1852 rc = 5;
1853 out[0] = (cp >> 24 & 3) | 248;
1854 out[1] = (cp >> 18 & 63) | 128;
1855 out[2] = (cp >> 12 & 63) | 128;
1856 out[3] = (cp >> 6 & 63) | 128;
1857 out[4] = (cp & 63) | 128;
1858 } else if (cp <= 0x7FFFFFFF) {
1859 rc = 6;
1860 out[0] = (cp >> 30 & 1) | 252;
1861 out[1] = (cp >> 24 & 63) | 128;
1862 out[2] = (cp >> 18 & 63) | 128;
1863 out[3] = (cp >> 12 & 63) | 128;
1864 out[4] = (cp >> 6 & 63) | 128;
1865 out[5] = (cp & 63) | 128;
1866 } else
1867 return 0;
1868
1869 out[rc] = '\0';
1870 return rc;
1871 }
1872
1873 /*
1874 * If the string contains escape sequences,
1875 * replace it with an allocated rendering and return 1,
1876 * such that the caller can free it after use.
1877 * Otherwise, do nothing and return 0.
1878 */
1879 static int
1880 render_string(char **public, size_t *psz)
1881 {
1882 const char *src, *scp, *addcp, *seq;
1883 char *dst;
1884 size_t ssz, dsz, addsz;
1885 char utfbuf[7], res[6];
1886 int seqlen, unicode;
1887
1888 res[0] = '\\';
1889 res[1] = '\t';
1890 res[2] = ASCII_NBRSP;
1891 res[3] = ASCII_HYPH;
1892 res[4] = ASCII_BREAK;
1893 res[5] = '\0';
1894
1895 src = scp = *public;
1896 ssz = *psz;
1897 dst = NULL;
1898 dsz = 0;
1899
1900 while (scp < src + *psz) {
1901
1902 /* Leave normal characters unchanged. */
1903
1904 if (strchr(res, *scp) == NULL) {
1905 if (dst != NULL)
1906 dst[dsz++] = *scp;
1907 scp++;
1908 continue;
1909 }
1910
1911 /*
1912 * Found something that requires replacing,
1913 * make sure we have a destination buffer.
1914 */
1915
1916 if (dst == NULL) {
1917 dst = mandoc_malloc(ssz + 1);
1918 dsz = scp - src;
1919 memcpy(dst, src, dsz);
1920 }
1921
1922 /* Handle single-char special characters. */
1923
1924 switch (*scp) {
1925 case '\\':
1926 break;
1927 case '\t':
1928 case ASCII_NBRSP:
1929 dst[dsz++] = ' ';
1930 scp++;
1931 continue;
1932 case ASCII_HYPH:
1933 dst[dsz++] = '-';
1934 /* FALLTHROUGH */
1935 case ASCII_BREAK:
1936 scp++;
1937 continue;
1938 default:
1939 abort();
1940 }
1941
1942 /*
1943 * Found an escape sequence.
1944 * Read past the slash, then parse it.
1945 * Ignore everything except characters.
1946 */
1947
1948 scp++;
1949 if (mandoc_escape(&scp, &seq, &seqlen) != ESCAPE_SPECIAL)
1950 continue;
1951
1952 /*
1953 * Render the special character
1954 * as either UTF-8 or ASCII.
1955 */
1956
1957 if (write_utf8) {
1958 unicode = mchars_spec2cp(seq, seqlen);
1959 if (unicode <= 0)
1960 continue;
1961 addsz = utf8(unicode, utfbuf);
1962 if (addsz == 0)
1963 continue;
1964 addcp = utfbuf;
1965 } else {
1966 addcp = mchars_spec2str(seq, seqlen, &addsz);
1967 if (addcp == NULL)
1968 continue;
1969 if (*addcp == ASCII_NBRSP) {
1970 addcp = " ";
1971 addsz = 1;
1972 }
1973 }
1974
1975 /* Copy the rendered glyph into the stream. */
1976
1977 ssz += addsz;
1978 dst = mandoc_realloc(dst, ssz + 1);
1979 memcpy(dst + dsz, addcp, addsz);
1980 dsz += addsz;
1981 }
1982 if (dst != NULL) {
1983 *public = dst;
1984 *psz = dsz;
1985 }
1986
1987 /* Trim trailing whitespace and NUL-terminate. */
1988
1989 while (*psz > 0 && (*public)[*psz - 1] == ' ')
1990 --*psz;
1991 if (dst != NULL) {
1992 (*public)[*psz] = '\0';
1993 return 1;
1994 } else
1995 return 0;
1996 }
1997
1998 static void
1999 dbadd_mlink(const struct mlink *mlink)
2000 {
2001 dba_page_alias(mlink->mpage->dba, mlink->name, NAME_FILE);
2002 dba_page_add(mlink->mpage->dba, DBP_SECT, mlink->dsec);
2003 dba_page_add(mlink->mpage->dba, DBP_SECT, mlink->fsec);
2004 dba_page_add(mlink->mpage->dba, DBP_ARCH, mlink->arch);
2005 dba_page_add(mlink->mpage->dba, DBP_FILE, mlink->file);
2006 }
2007
2008 /*
2009 * Flush the current page's terms (and their bits) into the database.
2010 * Also, handle escape sequences at the last possible moment.
2011 */
2012 static void
2013 dbadd(struct dba *dba, struct mpage *mpage)
2014 {
2015 struct mlink *mlink;
2016 struct str *key;
2017 char *cp;
2018 uint64_t mask;
2019 size_t i;
2020 unsigned int slot;
2021 int mustfree;
2022
2023 mlink = mpage->mlinks;
2024
2025 if (nodb) {
2026 for (key = ohash_first(&names, &slot); NULL != key;
2027 key = ohash_next(&names, &slot))
2028 free(key);
2029 for (key = ohash_first(&strings, &slot); NULL != key;
2030 key = ohash_next(&strings, &slot))
2031 free(key);
2032 if (0 == debug)
2033 return;
2034 while (NULL != mlink) {
2035 fputs(mlink->name, stdout);
2036 if (NULL == mlink->next ||
2037 strcmp(mlink->dsec, mlink->next->dsec) ||
2038 strcmp(mlink->fsec, mlink->next->fsec) ||
2039 strcmp(mlink->arch, mlink->next->arch)) {
2040 putchar('(');
2041 if ('\0' == *mlink->dsec)
2042 fputs(mlink->fsec, stdout);
2043 else
2044 fputs(mlink->dsec, stdout);
2045 if ('\0' != *mlink->arch)
2046 printf("/%s", mlink->arch);
2047 putchar(')');
2048 }
2049 mlink = mlink->next;
2050 if (NULL != mlink)
2051 fputs(", ", stdout);
2052 }
2053 printf(" - %s\n", mpage->desc);
2054 return;
2055 }
2056
2057 if (debug)
2058 say(mlink->file, "Adding to database");
2059
2060 cp = mpage->desc;
2061 i = strlen(cp);
2062 mustfree = render_string(&cp, &i);
2063 mpage->dba = dba_page_new(dba->pages,
2064 *mpage->arch == '\0' ? mlink->arch : mpage->arch,
2065 cp, mlink->file, mpage->form);
2066 if (mustfree)
2067 free(cp);
2068 dba_page_add(mpage->dba, DBP_SECT, mpage->sec);
2069
2070 while (mlink != NULL) {
2071 dbadd_mlink(mlink);
2072 mlink = mlink->next;
2073 }
2074
2075 for (key = ohash_first(&names, &slot); NULL != key;
2076 key = ohash_next(&names, &slot)) {
2077 assert(key->mpage == mpage);
2078 dba_page_alias(mpage->dba, key->key, key->mask);
2079 free(key);
2080 }
2081 for (key = ohash_first(&strings, &slot); NULL != key;
2082 key = ohash_next(&strings, &slot)) {
2083 assert(key->mpage == mpage);
2084 i = 0;
2085 for (mask = TYPE_Xr; mask <= TYPE_Lb; mask *= 2) {
2086 if (key->mask & mask)
2087 dba_macro_add(dba->macros, i,
2088 key->key, mpage->dba);
2089 i++;
2090 }
2091 free(key);
2092 }
2093 }
2094
2095 static void
2096 dbprune(struct dba *dba)
2097 {
2098 struct dba_array *page, *files;
2099 char *file;
2100
2101 dba_array_FOREACH(dba->pages, page) {
2102 files = dba_array_get(page, DBP_FILE);
2103 dba_array_FOREACH(files, file) {
2104 if (*file < ' ')
2105 file++;
2106 if (ohash_find(&mlinks, ohash_qlookup(&mlinks,
2107 file)) != NULL) {
2108 if (debug)
2109 say(file, "Deleting from database");
2110 dba_array_del(dba->pages);
2111 break;
2112 }
2113 }
2114 }
2115 }
2116
2117 /*
2118 * Write the database from memory to disk.
2119 */
2120 static void
2121 dbwrite(struct dba *dba)
2122 {
2123 struct stat sb1, sb2;
2124 char tfn[33], *cp1, *cp2;
2125 off_t i;
2126 int fd1, fd2;
2127
2128 /*
2129 * Do not write empty databases, and delete existing ones
2130 * when makewhatis -u causes them to become empty.
2131 */
2132
2133 dba_array_start(dba->pages);
2134 if (dba_array_next(dba->pages) == NULL) {
2135 if (unlink(MANDOC_DB) == -1 && errno != ENOENT)
2136 say(MANDOC_DB, "&unlink");
2137 return;
2138 }
2139
2140 /*
2141 * Build the database in a temporary file,
2142 * then atomically move it into place.
2143 */
2144
2145 if (dba_write(MANDOC_DB "~", dba) != -1) {
2146 if (rename(MANDOC_DB "~", MANDOC_DB) == -1) {
2147 exitcode = (int)MANDOCLEVEL_SYSERR;
2148 say(MANDOC_DB, "&rename");
2149 unlink(MANDOC_DB "~");
2150 }
2151 return;
2152 }
2153
2154 /*
2155 * We lack write permission and cannot replace the database
2156 * file, but let's at least check whether the data changed.
2157 */
2158
2159 (void)strlcpy(tfn, "/tmp/mandocdb.XXXXXXXX", sizeof(tfn));
2160 if (mkdtemp(tfn) == NULL) {
2161 exitcode = (int)MANDOCLEVEL_SYSERR;
2162 say("", "&%s", tfn);
2163 return;
2164 }
2165 cp1 = cp2 = MAP_FAILED;
2166 fd1 = fd2 = -1;
2167 (void)strlcat(tfn, "/" MANDOC_DB, sizeof(tfn));
2168 if (dba_write(tfn, dba) == -1) {
2169 say(tfn, "&dba_write");
2170 goto err;
2171 }
2172 if ((fd1 = open(MANDOC_DB, O_RDONLY, 0)) == -1) {
2173 say(MANDOC_DB, "&open");
2174 goto err;
2175 }
2176 if ((fd2 = open(tfn, O_RDONLY, 0)) == -1) {
2177 say(tfn, "&open");
2178 goto err;
2179 }
2180 if (fstat(fd1, &sb1) == -1) {
2181 say(MANDOC_DB, "&fstat");
2182 goto err;
2183 }
2184 if (fstat(fd2, &sb2) == -1) {
2185 say(tfn, "&fstat");
2186 goto err;
2187 }
2188 if (sb1.st_size != sb2.st_size)
2189 goto err;
2190 if ((cp1 = mmap(NULL, sb1.st_size, PROT_READ, MAP_PRIVATE,
2191 fd1, 0)) == MAP_FAILED) {
2192 say(MANDOC_DB, "&mmap");
2193 goto err;
2194 }
2195 if ((cp2 = mmap(NULL, sb2.st_size, PROT_READ, MAP_PRIVATE,
2196 fd2, 0)) == MAP_FAILED) {
2197 say(tfn, "&mmap");
2198 goto err;
2199 }
2200 for (i = 0; i < sb1.st_size; i++)
2201 if (cp1[i] != cp2[i])
2202 goto err;
2203 goto out;
2204
2205 err:
2206 exitcode = (int)MANDOCLEVEL_SYSERR;
2207 say(MANDOC_DB, "Data changed, but cannot replace database");
2208
2209 out:
2210 if (cp1 != MAP_FAILED)
2211 munmap(cp1, sb1.st_size);
2212 if (cp2 != MAP_FAILED)
2213 munmap(cp2, sb2.st_size);
2214 if (fd1 != -1)
2215 close(fd1);
2216 if (fd2 != -1)
2217 close(fd2);
2218 unlink(tfn);
2219 *strrchr(tfn, '/') = '\0';
2220 rmdir(tfn);
2221 }
2222
2223 static int
2224 set_basedir(const char *targetdir, int report_baddir)
2225 {
2226 static char startdir[PATH_MAX];
2227 static int getcwd_status; /* 1 = ok, 2 = failure */
2228 static int chdir_status; /* 1 = changed directory */
2229 char *cp;
2230
2231 /*
2232 * Remember the original working directory, if possible.
2233 * This will be needed if the second or a later directory
2234 * on the command line is given as a relative path.
2235 * Do not error out if the current directory is not
2236 * searchable: Maybe it won't be needed after all.
2237 */
2238 if (0 == getcwd_status) {
2239 if (NULL == getcwd(startdir, sizeof(startdir))) {
2240 getcwd_status = 2;
2241 (void)strlcpy(startdir, strerror(errno),
2242 sizeof(startdir));
2243 } else
2244 getcwd_status = 1;
2245 }
2246
2247 /*
2248 * We are leaving the old base directory.
2249 * Do not use it any longer, not even for messages.
2250 */
2251 *basedir = '\0';
2252
2253 /*
2254 * If and only if the directory was changed earlier and
2255 * the next directory to process is given as a relative path,
2256 * first go back, or bail out if that is impossible.
2257 */
2258 if (chdir_status && '/' != *targetdir) {
2259 if (2 == getcwd_status) {
2260 exitcode = (int)MANDOCLEVEL_SYSERR;
2261 say("", "getcwd: %s", startdir);
2262 return 0;
2263 }
2264 if (-1 == chdir(startdir)) {
2265 exitcode = (int)MANDOCLEVEL_SYSERR;
2266 say("", "&chdir %s", startdir);
2267 return 0;
2268 }
2269 }
2270
2271 /*
2272 * Always resolve basedir to the canonicalized absolute
2273 * pathname and append a trailing slash, such that
2274 * we can reliably check whether files are inside.
2275 */
2276 if (NULL == realpath(targetdir, basedir)) {
2277 if (report_baddir || errno != ENOENT) {
2278 exitcode = (int)MANDOCLEVEL_BADARG;
2279 say("", "&%s: realpath", targetdir);
2280 }
2281 return 0;
2282 } else if (-1 == chdir(basedir)) {
2283 if (report_baddir || errno != ENOENT) {
2284 exitcode = (int)MANDOCLEVEL_BADARG;
2285 say("", "&chdir");
2286 }
2287 return 0;
2288 }
2289 chdir_status = 1;
2290 cp = strchr(basedir, '\0');
2291 if ('/' != cp[-1]) {
2292 if (cp - basedir >= PATH_MAX - 1) {
2293 exitcode = (int)MANDOCLEVEL_SYSERR;
2294 say("", "Filename too long");
2295 return 0;
2296 }
2297 *cp++ = '/';
2298 *cp = '\0';
2299 }
2300 return 1;
2301 }
2302
2303 static void
2304 say(const char *file, const char *format, ...)
2305 {
2306 va_list ap;
2307 int use_errno;
2308
2309 if ('\0' != *basedir)
2310 fprintf(stderr, "%s", basedir);
2311 if ('\0' != *basedir && '\0' != *file)
2312 fputc('/', stderr);
2313 if ('\0' != *file)
2314 fprintf(stderr, "%s", file);
2315
2316 use_errno = 1;
2317 if (NULL != format) {
2318 switch (*format) {
2319 case '&':
2320 format++;
2321 break;
2322 case '\0':
2323 format = NULL;
2324 break;
2325 default:
2326 use_errno = 0;
2327 break;
2328 }
2329 }
2330 if (NULL != format) {
2331 if ('\0' != *basedir || '\0' != *file)
2332 fputs(": ", stderr);
2333 va_start(ap, format);
2334 vfprintf(stderr, format, ap);
2335 va_end(ap);
2336 }
2337 if (use_errno) {
2338 if ('\0' != *basedir || '\0' != *file || NULL != format)
2339 fputs(": ", stderr);
2340 perror(NULL);
2341 } else
2342 fputc('\n', stderr);
2343 }
A git conversion mirror of mandoc.bsd.lv