1 /* Id: mandocdb.c,v 1.262 2018/12/30 00:49:55 schwarze Exp */
2 /*
3 * Copyright (c) 2011, 2012 Kristaps Dzonsons <kristaps@bsd.lv>
4 * Copyright (c) 2011-2018 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 "main.h"
55 #include "man.h"
56 #include "mandoc_parse.h"
57 #include "manconf.h"
58 #include "mansearch.h"
59 #include "dba_array.h"
60 #include "dba.h"
61
62 extern const char *const mansearch_keynames[];
63
64 enum op {
65 OP_DEFAULT = 0, /* new dbs from dir list or default config */
66 OP_CONFFILE, /* new databases from custom config file */
67 OP_UPDATE, /* delete/add entries in existing database */
68 OP_DELETE, /* delete entries from existing database */
69 OP_TEST /* change no databases, report potential problems */
70 };
71
72 struct str {
73 const struct mpage *mpage; /* if set, the owning parse */
74 uint64_t mask; /* bitmask in sequence */
75 char key[]; /* rendered text */
76 };
77
78 struct inodev {
79 ino_t st_ino;
80 dev_t st_dev;
81 };
82
83 struct mpage {
84 struct inodev inodev; /* used for hashing routine */
85 struct dba_array *dba;
86 char *sec; /* section from file content */
87 char *arch; /* architecture from file content */
88 char *title; /* title from file content */
89 char *desc; /* description from file content */
90 struct mpage *next; /* singly linked list */
91 struct mlink *mlinks; /* singly linked list */
92 int name_head_done;
93 enum form form; /* format from file content */
94 };
95
96 struct mlink {
97 char file[PATH_MAX]; /* filename rel. to manpath */
98 char *dsec; /* section from directory */
99 char *arch; /* architecture from directory */
100 char *name; /* name from file name (not empty) */
101 char *fsec; /* section from file name suffix */
102 struct mlink *next; /* singly linked list */
103 struct mpage *mpage; /* parent */
104 int gzip; /* filename has a .gz suffix */
105 enum form dform; /* format from directory */
106 enum form fform; /* format from file name suffix */
107 };
108
109 typedef int (*mdoc_fp)(struct mpage *, const struct roff_meta *,
110 const struct roff_node *);
111
112 struct mdoc_handler {
113 mdoc_fp fp; /* optional handler */
114 uint64_t mask; /* set unless handler returns 0 */
115 int taboo; /* node flags that must not be set */
116 };
117
118
119 int mandocdb(int, char *[]);
120
121 static void dbadd(struct dba *, struct mpage *);
122 static void dbadd_mlink(const struct mlink *mlink);
123 static void dbprune(struct dba *);
124 static void dbwrite(struct dba *);
125 static void filescan(const char *);
126 #if HAVE_FTS_COMPARE_CONST
127 static int fts_compare(const FTSENT *const *, const FTSENT *const *);
128 #else
129 static int fts_compare(const FTSENT **, const FTSENT **);
130 #endif
131 static void mlink_add(struct mlink *, const struct stat *);
132 static void mlink_check(struct mpage *, struct mlink *);
133 static void mlink_free(struct mlink *);
134 static void mlinks_undupe(struct mpage *);
135 static void mpages_free(void);
136 static void mpages_merge(struct dba *, struct mparse *);
137 static void parse_cat(struct mpage *, int);
138 static void parse_man(struct mpage *, const struct roff_meta *,
139 const struct roff_node *);
140 static void parse_mdoc(struct mpage *, const struct roff_meta *,
141 const struct roff_node *);
142 static int parse_mdoc_head(struct mpage *, const struct roff_meta *,
143 const struct roff_node *);
144 static int parse_mdoc_Fa(struct mpage *, const struct roff_meta *,
145 const struct roff_node *);
146 static int parse_mdoc_Fd(struct mpage *, const struct roff_meta *,
147 const struct roff_node *);
148 static void parse_mdoc_fname(struct mpage *, const struct roff_node *);
149 static int parse_mdoc_Fn(struct mpage *, const struct roff_meta *,
150 const struct roff_node *);
151 static int parse_mdoc_Fo(struct mpage *, const struct roff_meta *,
152 const struct roff_node *);
153 static int parse_mdoc_Nd(struct mpage *, const struct roff_meta *,
154 const struct roff_node *);
155 static int parse_mdoc_Nm(struct mpage *, const struct roff_meta *,
156 const struct roff_node *);
157 static int parse_mdoc_Sh(struct mpage *, const struct roff_meta *,
158 const struct roff_node *);
159 static int parse_mdoc_Va(struct mpage *, const struct roff_meta *,
160 const struct roff_node *);
161 static int parse_mdoc_Xr(struct mpage *, const struct roff_meta *,
162 const struct roff_node *);
163 static void putkey(const struct mpage *, char *, uint64_t);
164 static void putkeys(const struct mpage *, char *, size_t, uint64_t);
165 static void putmdockey(const struct mpage *,
166 const struct roff_node *, uint64_t, int);
167 static int render_string(char **, size_t *);
168 static void say(const char *, const char *, ...)
169 __attribute__((__format__ (__printf__, 2, 3)));
170 static int set_basedir(const char *, int);
171 static int treescan(void);
172 static size_t utf8(unsigned int, char [7]);
173
174 static int nodb; /* no database changes */
175 static int mparse_options; /* abort the parse early */
176 static int use_all; /* use all found files */
177 static int debug; /* print what we're doing */
178 static int warnings; /* warn about crap */
179 static int write_utf8; /* write UTF-8 output; else ASCII */
180 static int exitcode; /* to be returned by main */
181 static enum op op; /* operational mode */
182 static char basedir[PATH_MAX]; /* current base directory */
183 static struct mpage *mpage_head; /* list of distinct manual pages */
184 static struct ohash mpages; /* table of distinct manual pages */
185 static struct ohash mlinks; /* table of directory entries */
186 static struct ohash names; /* table of all names */
187 static struct ohash strings; /* table of all strings */
188 static uint64_t name_mask;
189
190 static const struct mdoc_handler mdoc_handlers[MDOC_MAX - MDOC_Dd] = {
191 { NULL, 0, NODE_NOPRT }, /* Dd */
192 { NULL, 0, NODE_NOPRT }, /* Dt */
193 { NULL, 0, NODE_NOPRT }, /* Os */
194 { parse_mdoc_Sh, TYPE_Sh, 0 }, /* Sh */
195 { parse_mdoc_head, TYPE_Ss, 0 }, /* Ss */
196 { NULL, 0, 0 }, /* Pp */
197 { NULL, 0, 0 }, /* D1 */
198 { NULL, 0, 0 }, /* Dl */
199 { NULL, 0, 0 }, /* Bd */
200 { NULL, 0, 0 }, /* Ed */
201 { NULL, 0, 0 }, /* Bl */
202 { NULL, 0, 0 }, /* El */
203 { NULL, 0, 0 }, /* It */
204 { NULL, 0, 0 }, /* Ad */
205 { NULL, TYPE_An, 0 }, /* An */
206 { NULL, 0, 0 }, /* Ap */
207 { NULL, TYPE_Ar, 0 }, /* Ar */
208 { NULL, TYPE_Cd, 0 }, /* Cd */
209 { NULL, TYPE_Cm, 0 }, /* Cm */
210 { NULL, TYPE_Dv, 0 }, /* Dv */
211 { NULL, TYPE_Er, 0 }, /* Er */
212 { NULL, TYPE_Ev, 0 }, /* Ev */
213 { NULL, 0, 0 }, /* Ex */
214 { parse_mdoc_Fa, 0, 0 }, /* Fa */
215 { parse_mdoc_Fd, 0, 0 }, /* Fd */
216 { NULL, TYPE_Fl, 0 }, /* Fl */
217 { parse_mdoc_Fn, 0, 0 }, /* Fn */
218 { NULL, TYPE_Ft | TYPE_Vt, 0 }, /* Ft */
219 { NULL, TYPE_Ic, 0 }, /* Ic */
220 { NULL, TYPE_In, 0 }, /* In */
221 { NULL, TYPE_Li, 0 }, /* Li */
222 { parse_mdoc_Nd, 0, 0 }, /* Nd */
223 { parse_mdoc_Nm, 0, 0 }, /* Nm */
224 { NULL, 0, 0 }, /* Op */
225 { NULL, 0, 0 }, /* Ot */
226 { NULL, TYPE_Pa, NODE_NOSRC }, /* Pa */
227 { NULL, 0, 0 }, /* Rv */
228 { NULL, TYPE_St, 0 }, /* St */
229 { parse_mdoc_Va, TYPE_Va, 0 }, /* Va */
230 { parse_mdoc_Va, TYPE_Vt, 0 }, /* Vt */
231 { parse_mdoc_Xr, 0, 0 }, /* Xr */
232 { NULL, 0, 0 }, /* %A */
233 { NULL, 0, 0 }, /* %B */
234 { NULL, 0, 0 }, /* %D */
235 { NULL, 0, 0 }, /* %I */
236 { NULL, 0, 0 }, /* %J */
237 { NULL, 0, 0 }, /* %N */
238 { NULL, 0, 0 }, /* %O */
239 { NULL, 0, 0 }, /* %P */
240 { NULL, 0, 0 }, /* %R */
241 { NULL, 0, 0 }, /* %T */
242 { NULL, 0, 0 }, /* %V */
243 { NULL, 0, 0 }, /* Ac */
244 { NULL, 0, 0 }, /* Ao */
245 { NULL, 0, 0 }, /* Aq */
246 { NULL, TYPE_At, 0 }, /* At */
247 { NULL, 0, 0 }, /* Bc */
248 { NULL, 0, 0 }, /* Bf */
249 { NULL, 0, 0 }, /* Bo */
250 { NULL, 0, 0 }, /* Bq */
251 { NULL, TYPE_Bsx, NODE_NOSRC }, /* Bsx */
252 { NULL, TYPE_Bx, NODE_NOSRC }, /* Bx */
253 { NULL, 0, 0 }, /* Db */
254 { NULL, 0, 0 }, /* Dc */
255 { NULL, 0, 0 }, /* Do */
256 { NULL, 0, 0 }, /* Dq */
257 { NULL, 0, 0 }, /* Ec */
258 { NULL, 0, 0 }, /* Ef */
259 { NULL, TYPE_Em, 0 }, /* Em */
260 { NULL, 0, 0 }, /* Eo */
261 { NULL, TYPE_Fx, NODE_NOSRC }, /* Fx */
262 { NULL, TYPE_Ms, 0 }, /* Ms */
263 { NULL, 0, 0 }, /* No */
264 { NULL, 0, 0 }, /* Ns */
265 { NULL, TYPE_Nx, NODE_NOSRC }, /* Nx */
266 { NULL, TYPE_Ox, NODE_NOSRC }, /* Ox */
267 { NULL, 0, 0 }, /* Pc */
268 { NULL, 0, 0 }, /* Pf */
269 { NULL, 0, 0 }, /* Po */
270 { NULL, 0, 0 }, /* Pq */
271 { NULL, 0, 0 }, /* Qc */
272 { NULL, 0, 0 }, /* Ql */
273 { NULL, 0, 0 }, /* Qo */
274 { NULL, 0, 0 }, /* Qq */
275 { NULL, 0, 0 }, /* Re */
276 { NULL, 0, 0 }, /* Rs */
277 { NULL, 0, 0 }, /* Sc */
278 { NULL, 0, 0 }, /* So */
279 { NULL, 0, 0 }, /* Sq */
280 { NULL, 0, 0 }, /* Sm */
281 { NULL, 0, 0 }, /* Sx */
282 { NULL, TYPE_Sy, 0 }, /* Sy */
283 { NULL, TYPE_Tn, 0 }, /* Tn */
284 { NULL, 0, NODE_NOSRC }, /* Ux */
285 { NULL, 0, 0 }, /* Xc */
286 { NULL, 0, 0 }, /* Xo */
287 { parse_mdoc_Fo, 0, 0 }, /* Fo */
288 { NULL, 0, 0 }, /* Fc */
289 { NULL, 0, 0 }, /* Oo */
290 { NULL, 0, 0 }, /* Oc */
291 { NULL, 0, 0 }, /* Bk */
292 { NULL, 0, 0 }, /* Ek */
293 { NULL, 0, 0 }, /* Bt */
294 { NULL, 0, 0 }, /* Hf */
295 { NULL, 0, 0 }, /* Fr */
296 { NULL, 0, 0 }, /* Ud */
297 { NULL, TYPE_Lb, NODE_NOSRC }, /* Lb */
298 { NULL, 0, 0 }, /* Lp */
299 { NULL, TYPE_Lk, 0 }, /* Lk */
300 { NULL, TYPE_Mt, NODE_NOSRC }, /* Mt */
301 { NULL, 0, 0 }, /* Brq */
302 { NULL, 0, 0 }, /* Bro */
303 { NULL, 0, 0 }, /* Brc */
304 { NULL, 0, 0 }, /* %C */
305 { NULL, 0, 0 }, /* Es */
306 { NULL, 0, 0 }, /* En */
307 { NULL, TYPE_Dx, NODE_NOSRC }, /* Dx */
308 { NULL, 0, 0 }, /* %Q */
309 { NULL, 0, 0 }, /* %U */
310 { NULL, 0, 0 }, /* Ta */
311 };
312
313
314 int
mandocdb(int argc,char * argv[])315 mandocdb(int argc, char *argv[])
316 {
317 struct manconf conf;
318 struct mparse *mp;
319 struct dba *dba;
320 const char *path_arg, *progname;
321 size_t j, sz;
322 int ch, i;
323
324 #if HAVE_PLEDGE
325 if (pledge("stdio rpath wpath cpath", NULL) == -1) {
326 warn("pledge");
327 return (int)MANDOCLEVEL_SYSERR;
328 }
329 #endif
330
331 #if HAVE_SANDBOX_INIT
332 if (sandbox_init(kSBXProfileNoInternet, SANDBOX_NAMED, NULL) == -1) {
333 warnx("sandbox_init");
334 return (int)MANDOCLEVEL_SYSERR;
335 }
336 #endif
337
338 memset(&conf, 0, sizeof(conf));
339
340 /*
341 * We accept a few different invocations.
342 * The CHECKOP macro makes sure that invocation styles don't
343 * clobber each other.
344 */
345 #define CHECKOP(_op, _ch) do \
346 if (OP_DEFAULT != (_op)) { \
347 warnx("-%c: Conflicting option", (_ch)); \
348 goto usage; \
349 } while (/*CONSTCOND*/0)
350
351 mparse_options = MPARSE_VALIDATE;
352 path_arg = NULL;
353 op = OP_DEFAULT;
354
355 while (-1 != (ch = getopt(argc, argv, "aC:Dd:npQT:tu:v")))
356 switch (ch) {
357 case 'a':
358 use_all = 1;
359 break;
360 case 'C':
361 CHECKOP(op, ch);
362 path_arg = optarg;
363 op = OP_CONFFILE;
364 break;
365 case 'D':
366 debug++;
367 break;
368 case 'd':
369 CHECKOP(op, ch);
370 path_arg = optarg;
371 op = OP_UPDATE;
372 break;
373 case 'n':
374 nodb = 1;
375 break;
376 case 'p':
377 warnings = 1;
378 break;
379 case 'Q':
380 mparse_options |= MPARSE_QUICK;
381 break;
382 case 'T':
383 if (strcmp(optarg, "utf8")) {
384 warnx("-T%s: Unsupported output format",
385 optarg);
386 goto usage;
387 }
388 write_utf8 = 1;
389 break;
390 case 't':
391 CHECKOP(op, ch);
392 dup2(STDOUT_FILENO, STDERR_FILENO);
393 op = OP_TEST;
394 nodb = warnings = 1;
395 break;
396 case 'u':
397 CHECKOP(op, ch);
398 path_arg = optarg;
399 op = OP_DELETE;
400 break;
401 case 'v':
402 /* Compatibility with espie@'s makewhatis. */
403 break;
404 default:
405 goto usage;
406 }
407
408 argc -= optind;
409 argv += optind;
410
411 #if HAVE_PLEDGE
412 if (nodb) {
413 if (pledge("stdio rpath", NULL) == -1) {
414 warn("pledge");
415 return (int)MANDOCLEVEL_SYSERR;
416 }
417 }
418 #endif
419
420 if (OP_CONFFILE == op && argc > 0) {
421 warnx("-C: Too many arguments");
422 goto usage;
423 }
424
425 exitcode = (int)MANDOCLEVEL_OK;
426 mchars_alloc();
427 mp = mparse_alloc(mparse_options, MANDOC_OS_OTHER, NULL);
428 mandoc_ohash_init(&mpages, 6, offsetof(struct mpage, inodev));
429 mandoc_ohash_init(&mlinks, 6, offsetof(struct mlink, file));
430
431 if (OP_UPDATE == op || OP_DELETE == op || OP_TEST == op) {
432
433 /*
434 * Most of these deal with a specific directory.
435 * Jump into that directory first.
436 */
437 if (OP_TEST != op && 0 == set_basedir(path_arg, 1))
438 goto out;
439
440 dba = nodb ? dba_new(128) : dba_read(MANDOC_DB);
441 if (dba != NULL) {
442 /*
443 * The existing database is usable. Process
444 * all files specified on the command-line.
445 */
446 use_all = 1;
447 for (i = 0; i < argc; i++)
448 filescan(argv[i]);
449 if (nodb == 0)
450 dbprune(dba);
451 } else {
452 /* Database missing or corrupt. */
453 if (op != OP_UPDATE || errno != ENOENT)
454 say(MANDOC_DB, "%s: Automatically recreating"
455 " from scratch", strerror(errno));
456 exitcode = (int)MANDOCLEVEL_OK;
457 op = OP_DEFAULT;
458 if (0 == treescan())
459 goto out;
460 dba = dba_new(128);
461 }
462 if (OP_DELETE != op)
463 mpages_merge(dba, mp);
464 if (nodb == 0)
465 dbwrite(dba);
466 dba_free(dba);
467 } else {
468 /*
469 * If we have arguments, use them as our manpaths.
470 * If we don't, use man.conf(5).
471 */
472 if (argc > 0) {
473 conf.manpath.paths = mandoc_reallocarray(NULL,
474 argc, sizeof(char *));
475 conf.manpath.sz = (size_t)argc;
476 for (i = 0; i < argc; i++)
477 conf.manpath.paths[i] = mandoc_strdup(argv[i]);
478 } else
479 manconf_parse(&conf, path_arg, NULL, NULL);
480
481 if (conf.manpath.sz == 0) {
482 exitcode = (int)MANDOCLEVEL_BADARG;
483 say("", "Empty manpath");
484 }
485
486 /*
487 * First scan the tree rooted at a base directory, then
488 * build a new database and finally move it into place.
489 * Ignore zero-length directories and strip trailing
490 * slashes.
491 */
492 for (j = 0; j < conf.manpath.sz; j++) {
493 sz = strlen(conf.manpath.paths[j]);
494 if (sz && conf.manpath.paths[j][sz - 1] == '/')
495 conf.manpath.paths[j][--sz] = '\0';
496 if (0 == sz)
497 continue;
498
499 if (j) {
500 mandoc_ohash_init(&mpages, 6,
501 offsetof(struct mpage, inodev));
502 mandoc_ohash_init(&mlinks, 6,
503 offsetof(struct mlink, file));
504 }
505
506 if ( ! set_basedir(conf.manpath.paths[j], argc > 0))
507 continue;
508 if (0 == treescan())
509 continue;
510 dba = dba_new(128);
511 mpages_merge(dba, mp);
512 if (nodb == 0)
513 dbwrite(dba);
514 dba_free(dba);
515
516 if (j + 1 < conf.manpath.sz) {
517 mpages_free();
518 ohash_delete(&mpages);
519 ohash_delete(&mlinks);
520 }
521 }
522 }
523 out:
524 manconf_free(&conf);
525 mparse_free(mp);
526 mchars_free();
527 mpages_free();
528 ohash_delete(&mpages);
529 ohash_delete(&mlinks);
530 return exitcode;
531 usage:
532 progname = getprogname();
533 fprintf(stderr, "usage: %s [-aDnpQ] [-C file] [-Tutf8]\n"
534 " %s [-aDnpQ] [-Tutf8] dir ...\n"
535 " %s [-DnpQ] [-Tutf8] -d dir [file ...]\n"
536 " %s [-Dnp] -u dir [file ...]\n"
537 " %s [-Q] -t file ...\n",
538 progname, progname, progname, progname, progname);
539
540 return (int)MANDOCLEVEL_BADARG;
541 }
542
543 /*
544 * To get a singly linked list in alpha order while inserting entries
545 * at the beginning, process directory entries in reverse alpha order.
546 */
547 static int
548 #if HAVE_FTS_COMPARE_CONST
fts_compare(const FTSENT * const * a,const FTSENT * const * b)549 fts_compare(const FTSENT *const *a, const FTSENT *const *b)
550 #else
551 fts_compare(const FTSENT **a, const FTSENT **b)
552 #endif
553 {
554 return -strcmp((*a)->fts_name, (*b)->fts_name);
555 }
556
557 /*
558 * Scan a directory tree rooted at "basedir" for manpages.
559 * We use fts(), scanning directory parts along the way for clues to our
560 * section and architecture.
561 *
562 * If use_all has been specified, grok all files.
563 * If not, sanitise paths to the following:
564 *
565 * [./]man*[/<arch>]/<name>.<section>
566 * or
567 * [./]cat<section>[/<arch>]/<name>.0
568 *
569 * TODO: accommodate for multi-language directories.
570 */
571 static int
treescan(void)572 treescan(void)
573 {
574 char buf[PATH_MAX];
575 FTS *f;
576 FTSENT *ff;
577 struct mlink *mlink;
578 int gzip;
579 enum form dform;
580 char *dsec, *arch, *fsec, *cp;
581 const char *path;
582 const char *argv[2];
583
584 argv[0] = ".";
585 argv[1] = NULL;
586
587 f = fts_open((char * const *)__UNCONST(*argv),
588 FTS_PHYSICAL | FTS_NOCHDIR, fts_compare);
589 if (f == NULL) {
590 exitcode = (int)MANDOCLEVEL_SYSERR;
591 say("", "&fts_open");
592 return 0;
593 }
594
595 dsec = arch = NULL;
596 dform = FORM_NONE;
597
598 while ((ff = fts_read(f)) != NULL) {
599 path = ff->fts_path + 2;
600 switch (ff->fts_info) {
601
602 /*
603 * Symbolic links require various sanity checks,
604 * then get handled just like regular files.
605 */
606 case FTS_SL:
607 if (realpath(path, buf) == NULL) {
608 if (warnings)
609 say(path, "&realpath");
610 continue;
611 }
612 if (strstr(buf, basedir) != buf
613 #ifdef HOMEBREWDIR
614 && strstr(buf, HOMEBREWDIR) != buf
615 #endif
616 ) {
617 if (warnings) say("",
618 "%s: outside base directory", buf);
619 continue;
620 }
621 /* Use logical inode to avoid mpages dupe. */
622 if (stat(path, ff->fts_statp) == -1) {
623 if (warnings)
624 say(path, "&stat");
625 continue;
626 }
627 /* FALLTHROUGH */
628
629 /*
630 * If we're a regular file, add an mlink by using the
631 * stored directory data and handling the filename.
632 */
633 case FTS_F:
634 if ( ! strcmp(path, MANDOC_DB))
635 continue;
636 if ( ! use_all && ff->fts_level < 2) {
637 if (warnings)
638 say(path, "Extraneous file");
639 continue;
640 }
641 gzip = 0;
642 fsec = NULL;
643 while (fsec == NULL) {
644 fsec = strrchr(ff->fts_name, '.');
645 if (fsec == NULL || strcmp(fsec+1, "gz"))
646 break;
647 gzip = 1;
648 *fsec = '\0';
649 fsec = NULL;
650 }
651 if (fsec == NULL) {
652 if ( ! use_all) {
653 if (warnings)
654 say(path,
655 "No filename suffix");
656 continue;
657 }
658 } else if ( ! strcmp(++fsec, "html")) {
659 if (warnings)
660 say(path, "Skip html");
661 continue;
662 } else if ( ! strcmp(fsec, "ps")) {
663 if (warnings)
664 say(path, "Skip ps");
665 continue;
666 } else if ( ! strcmp(fsec, "pdf")) {
667 if (warnings)
668 say(path, "Skip pdf");
669 continue;
670 } else if ( ! use_all &&
671 ((dform == FORM_SRC &&
672 strncmp(fsec, dsec, strlen(dsec))) ||
673 (dform == FORM_CAT && strcmp(fsec, "0")))) {
674 if (warnings)
675 say(path, "Wrong filename suffix");
676 continue;
677 } else
678 fsec[-1] = '\0';
679
680 mlink = mandoc_calloc(1, sizeof(struct mlink));
681 if (strlcpy(mlink->file, path,
682 sizeof(mlink->file)) >=
683 sizeof(mlink->file)) {
684 say(path, "Filename too long");
685 free(mlink);
686 continue;
687 }
688 mlink->dform = dform;
689 mlink->dsec = dsec;
690 mlink->arch = arch;
691 mlink->name = ff->fts_name;
692 mlink->fsec = fsec;
693 mlink->gzip = gzip;
694 mlink_add(mlink, ff->fts_statp);
695 continue;
696
697 case FTS_D:
698 case FTS_DP:
699 break;
700
701 default:
702 if (warnings)
703 say(path, "Not a regular file");
704 continue;
705 }
706
707 switch (ff->fts_level) {
708 case 0:
709 /* Ignore the root directory. */
710 break;
711 case 1:
712 /*
713 * This might contain manX/ or catX/.
714 * Try to infer this from the name.
715 * If we're not in use_all, enforce it.
716 */
717 cp = ff->fts_name;
718 if (ff->fts_info == FTS_DP) {
719 dform = FORM_NONE;
720 dsec = NULL;
721 break;
722 }
723
724 if ( ! strncmp(cp, "man", 3)) {
725 dform = FORM_SRC;
726 dsec = cp + 3;
727 } else if ( ! strncmp(cp, "cat", 3)) {
728 dform = FORM_CAT;
729 dsec = cp + 3;
730 } else {
731 dform = FORM_NONE;
732 dsec = NULL;
733 }
734
735 if (dsec != NULL || use_all)
736 break;
737
738 if (warnings)
739 say(path, "Unknown directory part");
740 fts_set(f, ff, FTS_SKIP);
741 break;
742 case 2:
743 /*
744 * Possibly our architecture.
745 * If we're descending, keep tabs on it.
746 */
747 if (ff->fts_info != FTS_DP && dsec != NULL)
748 arch = ff->fts_name;
749 else
750 arch = NULL;
751 break;
752 default:
753 if (ff->fts_info == FTS_DP || use_all)
754 break;
755 if (warnings)
756 say(path, "Extraneous directory part");
757 fts_set(f, ff, FTS_SKIP);
758 break;
759 }
760 }
761
762 fts_close(f);
763 return 1;
764 }
765
766 /*
767 * Add a file to the mlinks table.
768 * Do not verify that it's a "valid" looking manpage (we'll do that
769 * later).
770 *
771 * Try to infer the manual section, architecture, and page name from the
772 * path, assuming it looks like
773 *
774 * [./]man*[/<arch>]/<name>.<section>
775 * or
776 * [./]cat<section>[/<arch>]/<name>.0
777 *
778 * See treescan() for the fts(3) version of this.
779 */
780 static void
filescan(const char * file)781 filescan(const char *file)
782 {
783 char buf[PATH_MAX];
784 struct stat st;
785 struct mlink *mlink;
786 char *p, *start;
787
788 assert(use_all);
789
790 if (0 == strncmp(file, "./", 2))
791 file += 2;
792
793 /*
794 * We have to do lstat(2) before realpath(3) loses
795 * the information whether this is a symbolic link.
796 * We need to know that because for symbolic links,
797 * we want to use the orginal file name, while for
798 * regular files, we want to use the real path.
799 */
800 if (-1 == lstat(file, &st)) {
801 exitcode = (int)MANDOCLEVEL_BADARG;
802 say(file, "&lstat");
803 return;
804 } else if (0 == ((S_IFREG | S_IFLNK) & st.st_mode)) {
805 exitcode = (int)MANDOCLEVEL_BADARG;
806 say(file, "Not a regular file");
807 return;
808 }
809
810 /*
811 * We have to resolve the file name to the real path
812 * in any case for the base directory check.
813 */
814 if (NULL == realpath(file, buf)) {
815 exitcode = (int)MANDOCLEVEL_BADARG;
816 say(file, "&realpath");
817 return;
818 }
819
820 if (OP_TEST == op)
821 start = buf;
822 else if (strstr(buf, basedir) == buf)
823 start = buf + strlen(basedir);
824 #ifdef HOMEBREWDIR
825 else if (strstr(buf, HOMEBREWDIR) == buf)
826 start = buf;
827 #endif
828 else {
829 exitcode = (int)MANDOCLEVEL_BADARG;
830 say("", "%s: outside base directory", buf);
831 return;
832 }
833
834 /*
835 * Now we are sure the file is inside our tree.
836 * If it is a symbolic link, ignore the real path
837 * and use the original name.
838 * This implies passing stuff like "cat1/../man1/foo.1"
839 * on the command line won't work. So don't do that.
840 * Note the stat(2) can still fail if the link target
841 * doesn't exist.
842 */
843 if (S_IFLNK & st.st_mode) {
844 if (-1 == stat(buf, &st)) {
845 exitcode = (int)MANDOCLEVEL_BADARG;
846 say(file, "&stat");
847 return;
848 }
849 if (strlcpy(buf, file, sizeof(buf)) >= sizeof(buf)) {
850 say(file, "Filename too long");
851 return;
852 }
853 start = buf;
854 if (OP_TEST != op && strstr(buf, basedir) == buf)
855 start += strlen(basedir);
856 }
857
858 mlink = mandoc_calloc(1, sizeof(struct mlink));
859 mlink->dform = FORM_NONE;
860 if (strlcpy(mlink->file, start, sizeof(mlink->file)) >=
861 sizeof(mlink->file)) {
862 say(start, "Filename too long");
863 free(mlink);
864 return;
865 }
866
867 /*
868 * In test mode or when the original name is absolute
869 * but outside our tree, guess the base directory.
870 */
871
872 if (op == OP_TEST || (start == buf && *start == '/')) {
873 if (strncmp(buf, "man/", 4) == 0)
874 start = buf + 4;
875 else if ((start = strstr(buf, "/man/")) != NULL)
876 start += 5;
877 else
878 start = buf;
879 }
880
881 /*
882 * First try to guess our directory structure.
883 * If we find a separator, try to look for man* or cat*.
884 * If we find one of these and what's underneath is a directory,
885 * assume it's an architecture.
886 */
887 if (NULL != (p = strchr(start, '/'))) {
888 *p++ = '\0';
889 if (0 == strncmp(start, "man", 3)) {
890 mlink->dform = FORM_SRC;
891 mlink->dsec = start + 3;
892 } else if (0 == strncmp(start, "cat", 3)) {
893 mlink->dform = FORM_CAT;
894 mlink->dsec = start + 3;
895 }
896
897 start = p;
898 if (NULL != mlink->dsec && NULL != (p = strchr(start, '/'))) {
899 *p++ = '\0';
900 mlink->arch = start;
901 start = p;
902 }
903 }
904
905 /*
906 * Now check the file suffix.
907 * Suffix of `.0' indicates a catpage, `.1-9' is a manpage.
908 */
909 p = strrchr(start, '\0');
910 while (p-- > start && '/' != *p && '.' != *p)
911 /* Loop. */ ;
912
913 if ('.' == *p) {
914 *p++ = '\0';
915 mlink->fsec = p;
916 }
917
918 /*
919 * Now try to parse the name.
920 * Use the filename portion of the path.
921 */
922 mlink->name = start;
923 if (NULL != (p = strrchr(start, '/'))) {
924 mlink->name = p + 1;
925 *p = '\0';
926 }
927 mlink_add(mlink, &st);
928 }
929
930 static void
mlink_add(struct mlink * mlink,const struct stat * st)931 mlink_add(struct mlink *mlink, const struct stat *st)
932 {
933 struct inodev inodev;
934 struct mpage *mpage;
935 unsigned int slot;
936
937 assert(NULL != mlink->file);
938
939 mlink->dsec = mandoc_strdup(mlink->dsec ? mlink->dsec : "");
940 mlink->arch = mandoc_strdup(mlink->arch ? mlink->arch : "");
941 mlink->name = mandoc_strdup(mlink->name ? mlink->name : "");
942 mlink->fsec = mandoc_strdup(mlink->fsec ? mlink->fsec : "");
943
944 if ('0' == *mlink->fsec) {
945 free(mlink->fsec);
946 mlink->fsec = mandoc_strdup(mlink->dsec);
947 mlink->fform = FORM_CAT;
948 } else if ('1' <= *mlink->fsec && '9' >= *mlink->fsec)
949 mlink->fform = FORM_SRC;
950 else
951 mlink->fform = FORM_NONE;
952
953 slot = ohash_qlookup(&mlinks, mlink->file);
954 assert(NULL == ohash_find(&mlinks, slot));
955 ohash_insert(&mlinks, slot, mlink);
956
957 memset(&inodev, 0, sizeof(inodev)); /* Clear padding. */
958 inodev.st_ino = st->st_ino;
959 inodev.st_dev = st->st_dev;
960 slot = ohash_lookup_memory(&mpages, (char *)&inodev,
961 sizeof(struct inodev), inodev.st_ino);
962 mpage = ohash_find(&mpages, slot);
963 if (NULL == mpage) {
964 mpage = mandoc_calloc(1, sizeof(struct mpage));
965 mpage->inodev.st_ino = inodev.st_ino;
966 mpage->inodev.st_dev = inodev.st_dev;
967 mpage->form = FORM_NONE;
968 mpage->next = mpage_head;
969 mpage_head = mpage;
970 ohash_insert(&mpages, slot, mpage);
971 } else
972 mlink->next = mpage->mlinks;
973 mpage->mlinks = mlink;
974 mlink->mpage = mpage;
975 }
976
977 static void
mlink_free(struct mlink * mlink)978 mlink_free(struct mlink *mlink)
979 {
980
981 free(mlink->dsec);
982 free(mlink->arch);
983 free(mlink->name);
984 free(mlink->fsec);
985 free(mlink);
986 }
987
988 static void
mpages_free(void)989 mpages_free(void)
990 {
991 struct mpage *mpage;
992 struct mlink *mlink;
993
994 while ((mpage = mpage_head) != NULL) {
995 while ((mlink = mpage->mlinks) != NULL) {
996 mpage->mlinks = mlink->next;
997 mlink_free(mlink);
998 }
999 mpage_head = mpage->next;
1000 free(mpage->sec);
1001 free(mpage->arch);
1002 free(mpage->title);
1003 free(mpage->desc);
1004 free(mpage);
1005 }
1006 }
1007
1008 /*
1009 * For each mlink to the mpage, check whether the path looks like
1010 * it is formatted, and if it does, check whether a source manual
1011 * exists by the same name, ignoring the suffix.
1012 * If both conditions hold, drop the mlink.
1013 */
1014 static void
mlinks_undupe(struct mpage * mpage)1015 mlinks_undupe(struct mpage *mpage)
1016 {
1017 char buf[PATH_MAX];
1018 struct mlink **prev;
1019 struct mlink *mlink;
1020 char *bufp;
1021
1022 mpage->form = FORM_CAT;
1023 prev = &mpage->mlinks;
1024 while (NULL != (mlink = *prev)) {
1025 if (FORM_CAT != mlink->dform) {
1026 mpage->form = FORM_NONE;
1027 goto nextlink;
1028 }
1029 (void)strlcpy(buf, mlink->file, sizeof(buf));
1030 bufp = strstr(buf, "cat");
1031 assert(NULL != bufp);
1032 memcpy(bufp, "man", 3);
1033 if (NULL != (bufp = strrchr(buf, '.')))
1034 *++bufp = '\0';
1035 (void)strlcat(buf, mlink->dsec, sizeof(buf));
1036 if (NULL == ohash_find(&mlinks,
1037 ohash_qlookup(&mlinks, buf)))
1038 goto nextlink;
1039 if (warnings)
1040 say(mlink->file, "Man source exists: %s", buf);
1041 if (use_all)
1042 goto nextlink;
1043 *prev = mlink->next;
1044 mlink_free(mlink);
1045 continue;
1046 nextlink:
1047 prev = &(*prev)->next;
1048 }
1049 }
1050
1051 static void
mlink_check(struct mpage * mpage,struct mlink * mlink)1052 mlink_check(struct mpage *mpage, struct mlink *mlink)
1053 {
1054 struct str *str;
1055 unsigned int slot;
1056
1057 /*
1058 * Check whether the manual section given in a file
1059 * agrees with the directory where the file is located.
1060 * Some manuals have suffixes like (3p) on their
1061 * section number either inside the file or in the
1062 * directory name, some are linked into more than one
1063 * section, like encrypt(1) = makekey(8).
1064 */
1065
1066 if (FORM_SRC == mpage->form &&
1067 strcasecmp(mpage->sec, mlink->dsec))
1068 say(mlink->file, "Section \"%s\" manual in %s directory",
1069 mpage->sec, mlink->dsec);
1070
1071 /*
1072 * Manual page directories exist for each kernel
1073 * architecture as returned by machine(1).
1074 * However, many manuals only depend on the
1075 * application architecture as returned by arch(1).
1076 * For example, some (2/ARM) manuals are shared
1077 * across the "armish" and "zaurus" kernel
1078 * architectures.
1079 * A few manuals are even shared across completely
1080 * different architectures, for example fdformat(1)
1081 * on amd64, i386, and sparc64.
1082 */
1083
1084 if (strcasecmp(mpage->arch, mlink->arch))
1085 say(mlink->file, "Architecture \"%s\" manual in "
1086 "\"%s\" directory", mpage->arch, mlink->arch);
1087
1088 /*
1089 * XXX
1090 * parse_cat() doesn't set NAME_TITLE yet.
1091 */
1092
1093 if (FORM_CAT == mpage->form)
1094 return;
1095
1096 /*
1097 * Check whether this mlink
1098 * appears as a name in the NAME section.
1099 */
1100
1101 slot = ohash_qlookup(&names, mlink->name);
1102 str = ohash_find(&names, slot);
1103 assert(NULL != str);
1104 if ( ! (NAME_TITLE & str->mask))
1105 say(mlink->file, "Name missing in NAME section");
1106 }
1107
1108 /*
1109 * Run through the files in the global vector "mpages"
1110 * and add them to the database specified in "basedir".
1111 *
1112 * This handles the parsing scheme itself, using the cues of directory
1113 * and filename to determine whether the file is parsable or not.
1114 */
1115 static void
mpages_merge(struct dba * dba,struct mparse * mp)1116 mpages_merge(struct dba *dba, struct mparse *mp)
1117 {
1118 struct mpage *mpage, *mpage_dest;
1119 struct mlink *mlink, *mlink_dest;
1120 struct roff_meta *meta;
1121 char *cp;
1122 int fd;
1123
1124 for (mpage = mpage_head; mpage != NULL; mpage = mpage->next) {
1125 mlinks_undupe(mpage);
1126 if ((mlink = mpage->mlinks) == NULL)
1127 continue;
1128
1129 name_mask = NAME_MASK;
1130 mandoc_ohash_init(&names, 4, offsetof(struct str, key));
1131 mandoc_ohash_init(&strings, 6, offsetof(struct str, key));
1132 mparse_reset(mp);
1133 meta = NULL;
1134
1135 if ((fd = mparse_open(mp, mlink->file)) == -1) {
1136 say(mlink->file, "&open");
1137 goto nextpage;
1138 }
1139
1140 /*
1141 * Interpret the file as mdoc(7) or man(7) source
1142 * code, unless it is known to be formatted.
1143 */
1144 if (mlink->dform != FORM_CAT || mlink->fform != FORM_CAT) {
1145 mparse_readfd(mp, fd, mlink->file);
1146 close(fd);
1147 fd = -1;
1148 meta = mparse_result(mp);
1149 }
1150
1151 if (meta != NULL && meta->sodest != NULL) {
1152 mlink_dest = ohash_find(&mlinks,
1153 ohash_qlookup(&mlinks, meta->sodest));
1154 if (mlink_dest == NULL) {
1155 mandoc_asprintf(&cp, "%s.gz", meta->sodest);
1156 mlink_dest = ohash_find(&mlinks,
1157 ohash_qlookup(&mlinks, cp));
1158 free(cp);
1159 }
1160 if (mlink_dest != NULL) {
1161
1162 /* The .so target exists. */
1163
1164 mpage_dest = mlink_dest->mpage;
1165 while (1) {
1166 mlink->mpage = mpage_dest;
1167
1168 /*
1169 * If the target was already
1170 * processed, add the links
1171 * to the database now.
1172 * Otherwise, this will
1173 * happen when we come
1174 * to the target.
1175 */
1176
1177 if (mpage_dest->dba != NULL)
1178 dbadd_mlink(mlink);
1179
1180 if (mlink->next == NULL)
1181 break;
1182 mlink = mlink->next;
1183 }
1184
1185 /* Move all links to the target. */
1186
1187 mlink->next = mlink_dest->next;
1188 mlink_dest->next = mpage->mlinks;
1189 mpage->mlinks = NULL;
1190 }
1191 goto nextpage;
1192 } else if (meta != NULL && meta->macroset == MACROSET_MDOC) {
1193 mpage->form = FORM_SRC;
1194 mpage->sec = meta->msec;
1195 mpage->sec = mandoc_strdup(
1196 mpage->sec == NULL ? "" : mpage->sec);
1197 mpage->arch = meta->arch;
1198 mpage->arch = mandoc_strdup(
1199 mpage->arch == NULL ? "" : mpage->arch);
1200 mpage->title = mandoc_strdup(meta->title);
1201 } else if (meta != NULL && meta->macroset == MACROSET_MAN) {
1202 if (*meta->msec != '\0' || *meta->title != '\0') {
1203 mpage->form = FORM_SRC;
1204 mpage->sec = mandoc_strdup(meta->msec);
1205 mpage->arch = mandoc_strdup(mlink->arch);
1206 mpage->title = mandoc_strdup(meta->title);
1207 } else
1208 meta = NULL;
1209 }
1210
1211 assert(mpage->desc == NULL);
1212 if (meta == NULL) {
1213 mpage->form = FORM_CAT;
1214 mpage->sec = mandoc_strdup(mlink->dsec);
1215 mpage->arch = mandoc_strdup(mlink->arch);
1216 mpage->title = mandoc_strdup(mlink->name);
1217 parse_cat(mpage, fd);
1218 } else if (meta->macroset == MACROSET_MDOC)
1219 parse_mdoc(mpage, meta, meta->first);
1220 else
1221 parse_man(mpage, meta, meta->first);
1222 if (mpage->desc == NULL) {
1223 mpage->desc = mandoc_strdup(mlink->name);
1224 if (warnings)
1225 say(mlink->file, "No one-line description, "
1226 "using filename \"%s\"", mlink->name);
1227 }
1228
1229 for (mlink = mpage->mlinks;
1230 mlink != NULL;
1231 mlink = mlink->next) {
1232 putkey(mpage, mlink->name, NAME_FILE);
1233 if (warnings && !use_all)
1234 mlink_check(mpage, mlink);
1235 }
1236
1237 dbadd(dba, mpage);
1238
1239 nextpage:
1240 ohash_delete(&strings);
1241 ohash_delete(&names);
1242 }
1243 }
1244
1245 static void
parse_cat(struct mpage * mpage,int fd)1246 parse_cat(struct mpage *mpage, int fd)
1247 {
1248 FILE *stream;
1249 struct mlink *mlink;
1250 char *line, *p, *title, *sec;
1251 size_t linesz, plen, titlesz;
1252 ssize_t len;
1253 int offs;
1254
1255 mlink = mpage->mlinks;
1256 stream = fd == -1 ? fopen(mlink->file, "r") : fdopen(fd, "r");
1257 if (stream == NULL) {
1258 if (fd != -1)
1259 close(fd);
1260 if (warnings)
1261 say(mlink->file, "&fopen");
1262 return;
1263 }
1264
1265 line = NULL;
1266 linesz = 0;
1267
1268 /* Parse the section number from the header line. */
1269
1270 while (getline(&line, &linesz, stream) != -1) {
1271 if (*line == '\n')
1272 continue;
1273 if ((sec = strchr(line, '(')) == NULL)
1274 break;
1275 if ((p = strchr(++sec, ')')) == NULL)
1276 break;
1277 free(mpage->sec);
1278 mpage->sec = mandoc_strndup(sec, p - sec);
1279 if (warnings && *mlink->dsec != '\0' &&
1280 strcasecmp(mpage->sec, mlink->dsec))
1281 say(mlink->file,
1282 "Section \"%s\" manual in %s directory",
1283 mpage->sec, mlink->dsec);
1284 break;
1285 }
1286
1287 /* Skip to first blank line. */
1288
1289 while (line == NULL || *line != '\n')
1290 if (getline(&line, &linesz, stream) == -1)
1291 break;
1292
1293 /*
1294 * Assume the first line that is not indented
1295 * is the first section header. Skip to it.
1296 */
1297
1298 while (getline(&line, &linesz, stream) != -1)
1299 if (*line != '\n' && *line != ' ')
1300 break;
1301
1302 /*
1303 * Read up until the next section into a buffer.
1304 * Strip the leading and trailing newline from each read line,
1305 * appending a trailing space.
1306 * Ignore empty (whitespace-only) lines.
1307 */
1308
1309 titlesz = 0;
1310 title = NULL;
1311
1312 while ((len = getline(&line, &linesz, stream)) != -1) {
1313 if (*line != ' ')
1314 break;
1315 offs = 0;
1316 while (isspace((unsigned char)line[offs]))
1317 offs++;
1318 if (line[offs] == '\0')
1319 continue;
1320 title = mandoc_realloc(title, titlesz + len - offs);
1321 memcpy(title + titlesz, line + offs, len - offs);
1322 titlesz += len - offs;
1323 title[titlesz - 1] = ' ';
1324 }
1325 free(line);
1326
1327 /*
1328 * If no page content can be found, or the input line
1329 * is already the next section header, or there is no
1330 * trailing newline, reuse the page title as the page
1331 * description.
1332 */
1333
1334 if (NULL == title || '\0' == *title) {
1335 if (warnings)
1336 say(mlink->file, "Cannot find NAME section");
1337 fclose(stream);
1338 free(title);
1339 return;
1340 }
1341
1342 title[titlesz - 1] = '\0';
1343
1344 /*
1345 * Skip to the first dash.
1346 * Use the remaining line as the description (no more than 70
1347 * bytes).
1348 */
1349
1350 if (NULL != (p = strstr(title, "- "))) {
1351 for (p += 2; ' ' == *p || '\b' == *p; p++)
1352 /* Skip to next word. */ ;
1353 } else {
1354 if (warnings)
1355 say(mlink->file, "No dash in title line, "
1356 "reusing \"%s\" as one-line description", title);
1357 p = title;
1358 }
1359
1360 plen = strlen(p);
1361
1362 /* Strip backspace-encoding from line. */
1363
1364 while (NULL != (line = memchr(p, '\b', plen))) {
1365 len = line - p;
1366 if (0 == len) {
1367 memmove(line, line + 1, plen--);
1368 continue;
1369 }
1370 memmove(line - 1, line + 1, plen - len);
1371 plen -= 2;
1372 }
1373
1374 /*
1375 * Cut off excessive one-line descriptions.
1376 * Bad pages are not worth better heuristics.
1377 */
1378
1379 mpage->desc = mandoc_strndup(p, 150);
1380 fclose(stream);
1381 free(title);
1382 }
1383
1384 /*
1385 * Put a type/word pair into the word database for this particular file.
1386 */
1387 static void
putkey(const struct mpage * mpage,char * value,uint64_t type)1388 putkey(const struct mpage *mpage, char *value, uint64_t type)
1389 {
1390 putkeys(mpage, value, strlen(value), type);
1391 }
1392
1393 /*
1394 * Grok all nodes at or below a certain mdoc node into putkey().
1395 */
1396 static void
putmdockey(const struct mpage * mpage,const struct roff_node * n,uint64_t m,int taboo)1397 putmdockey(const struct mpage *mpage,
1398 const struct roff_node *n, uint64_t m, int taboo)
1399 {
1400
1401 for ( ; NULL != n; n = n->next) {
1402 if (n->flags & taboo)
1403 continue;
1404 if (NULL != n->child)
1405 putmdockey(mpage, n->child, m, taboo);
1406 if (n->type == ROFFT_TEXT)
1407 putkey(mpage, n->string, m);
1408 }
1409 }
1410
1411 static void
parse_man(struct mpage * mpage,const struct roff_meta * meta,const struct roff_node * n)1412 parse_man(struct mpage *mpage, const struct roff_meta *meta,
1413 const struct roff_node *n)
1414 {
1415 const struct roff_node *head, *body;
1416 char *start, *title;
1417 char byte;
1418 size_t sz;
1419
1420 if (n == NULL)
1421 return;
1422
1423 /*
1424 * We're only searching for one thing: the first text child in
1425 * the BODY of a NAME section. Since we don't keep track of
1426 * sections in -man, run some hoops to find out whether we're in
1427 * the correct section or not.
1428 */
1429
1430 if (n->type == ROFFT_BODY && n->tok == MAN_SH) {
1431 body = n;
1432 if ((head = body->parent->head) != NULL &&
1433 (head = head->child) != NULL &&
1434 head->next == NULL &&
1435 head->type == ROFFT_TEXT &&
1436 strcmp(head->string, "NAME") == 0 &&
1437 body->child != NULL) {
1438
1439 /*
1440 * Suck the entire NAME section into memory.
1441 * Yes, we might run away.
1442 * But too many manuals have big, spread-out
1443 * NAME sections over many lines.
1444 */
1445
1446 title = NULL;
1447 deroff(&title, body);
1448 if (NULL == title)
1449 return;
1450
1451 /*
1452 * Go through a special heuristic dance here.
1453 * Conventionally, one or more manual names are
1454 * comma-specified prior to a whitespace, then a
1455 * dash, then a description. Try to puzzle out
1456 * the name parts here.
1457 */
1458
1459 start = title;
1460 for ( ;; ) {
1461 sz = strcspn(start, " ,");
1462 if ('\0' == start[sz])
1463 break;
1464
1465 byte = start[sz];
1466 start[sz] = '\0';
1467
1468 /*
1469 * Assume a stray trailing comma in the
1470 * name list if a name begins with a dash.
1471 */
1472
1473 if ('-' == start[0] ||
1474 ('\\' == start[0] && '-' == start[1]))
1475 break;
1476
1477 putkey(mpage, start, NAME_TITLE);
1478 if ( ! (mpage->name_head_done ||
1479 strcasecmp(start, meta->title))) {
1480 putkey(mpage, start, NAME_HEAD);
1481 mpage->name_head_done = 1;
1482 }
1483
1484 if (' ' == byte) {
1485 start += sz + 1;
1486 break;
1487 }
1488
1489 assert(',' == byte);
1490 start += sz + 1;
1491 while (' ' == *start)
1492 start++;
1493 }
1494
1495 if (start == title) {
1496 putkey(mpage, start, NAME_TITLE);
1497 if ( ! (mpage->name_head_done ||
1498 strcasecmp(start, meta->title))) {
1499 putkey(mpage, start, NAME_HEAD);
1500 mpage->name_head_done = 1;
1501 }
1502 free(title);
1503 return;
1504 }
1505
1506 while (isspace((unsigned char)*start))
1507 start++;
1508
1509 if (0 == strncmp(start, "-", 1))
1510 start += 1;
1511 else if (0 == strncmp(start, "\\-\\-", 4))
1512 start += 4;
1513 else if (0 == strncmp(start, "\\-", 2))
1514 start += 2;
1515 else if (0 == strncmp(start, "\\(en", 4))
1516 start += 4;
1517 else if (0 == strncmp(start, "\\(em", 4))
1518 start += 4;
1519
1520 while (' ' == *start)
1521 start++;
1522
1523 /*
1524 * Cut off excessive one-line descriptions.
1525 * Bad pages are not worth better heuristics.
1526 */
1527
1528 mpage->desc = mandoc_strndup(start, 150);
1529 free(title);
1530 return;
1531 }
1532 }
1533
1534 for (n = n->child; n; n = n->next) {
1535 if (NULL != mpage->desc)
1536 break;
1537 parse_man(mpage, meta, n);
1538 }
1539 }
1540
1541 static void
parse_mdoc(struct mpage * mpage,const struct roff_meta * meta,const struct roff_node * n)1542 parse_mdoc(struct mpage *mpage, const struct roff_meta *meta,
1543 const struct roff_node *n)
1544 {
1545 const struct mdoc_handler *handler;
1546
1547 for (n = n->child; n != NULL; n = n->next) {
1548 if (n->tok == TOKEN_NONE || n->tok < ROFF_MAX)
1549 continue;
1550 assert(n->tok >= MDOC_Dd && n->tok < MDOC_MAX);
1551 handler = mdoc_handlers + (n->tok - MDOC_Dd);
1552 if (n->flags & handler->taboo)
1553 continue;
1554
1555 switch (n->type) {
1556 case ROFFT_ELEM:
1557 case ROFFT_BLOCK:
1558 case ROFFT_HEAD:
1559 case ROFFT_BODY:
1560 case ROFFT_TAIL:
1561 if (handler->fp != NULL &&
1562 (*handler->fp)(mpage, meta, n) == 0)
1563 break;
1564 if (handler->mask)
1565 putmdockey(mpage, n->child,
1566 handler->mask, handler->taboo);
1567 break;
1568 default:
1569 continue;
1570 }
1571 if (NULL != n->child)
1572 parse_mdoc(mpage, meta, n);
1573 }
1574 }
1575
1576 static int
parse_mdoc_Fa(struct mpage * mpage,const struct roff_meta * meta,const struct roff_node * n)1577 parse_mdoc_Fa(struct mpage *mpage, const struct roff_meta *meta,
1578 const struct roff_node *n)
1579 {
1580 uint64_t mask;
1581
1582 mask = TYPE_Fa;
1583 if (n->sec == SEC_SYNOPSIS)
1584 mask |= TYPE_Vt;
1585
1586 putmdockey(mpage, n->child, mask, 0);
1587 return 0;
1588 }
1589
1590 static int
parse_mdoc_Fd(struct mpage * mpage,const struct roff_meta * meta,const struct roff_node * n)1591 parse_mdoc_Fd(struct mpage *mpage, const struct roff_meta *meta,
1592 const struct roff_node *n)
1593 {
1594 char *start, *end;
1595 size_t sz;
1596
1597 if (SEC_SYNOPSIS != n->sec ||
1598 NULL == (n = n->child) ||
1599 n->type != ROFFT_TEXT)
1600 return 0;
1601
1602 /*
1603 * Only consider those `Fd' macro fields that begin with an
1604 * "inclusion" token (versus, e.g., #define).
1605 */
1606
1607 if (strcmp("#include", n->string))
1608 return 0;
1609
1610 if ((n = n->next) == NULL || n->type != ROFFT_TEXT)
1611 return 0;
1612
1613 /*
1614 * Strip away the enclosing angle brackets and make sure we're
1615 * not zero-length.
1616 */
1617
1618 start = n->string;
1619 if ('<' == *start || '"' == *start)
1620 start++;
1621
1622 if (0 == (sz = strlen(start)))
1623 return 0;
1624
1625 end = &start[(int)sz - 1];
1626 if ('>' == *end || '"' == *end)
1627 end--;
1628
1629 if (end > start)
1630 putkeys(mpage, start, end - start + 1, TYPE_In);
1631 return 0;
1632 }
1633
1634 static void
parse_mdoc_fname(struct mpage * mpage,const struct roff_node * n)1635 parse_mdoc_fname(struct mpage *mpage, const struct roff_node *n)
1636 {
1637 char *cp;
1638 size_t sz;
1639
1640 if (n->type != ROFFT_TEXT)
1641 return;
1642
1643 /* Skip function pointer punctuation. */
1644
1645 cp = n->string;
1646 while (*cp == '(' || *cp == '*')
1647 cp++;
1648 sz = strcspn(cp, "()");
1649
1650 putkeys(mpage, cp, sz, TYPE_Fn);
1651 if (n->sec == SEC_SYNOPSIS)
1652 putkeys(mpage, cp, sz, NAME_SYN);
1653 }
1654
1655 static int
parse_mdoc_Fn(struct mpage * mpage,const struct roff_meta * meta,const struct roff_node * n)1656 parse_mdoc_Fn(struct mpage *mpage, const struct roff_meta *meta,
1657 const struct roff_node *n)
1658 {
1659 uint64_t mask;
1660
1661 if (n->child == NULL)
1662 return 0;
1663
1664 parse_mdoc_fname(mpage, n->child);
1665
1666 n = n->child->next;
1667 if (n != NULL && n->type == ROFFT_TEXT) {
1668 mask = TYPE_Fa;
1669 if (n->sec == SEC_SYNOPSIS)
1670 mask |= TYPE_Vt;
1671 putmdockey(mpage, n, mask, 0);
1672 }
1673
1674 return 0;
1675 }
1676
1677 static int
parse_mdoc_Fo(struct mpage * mpage,const struct roff_meta * meta,const struct roff_node * n)1678 parse_mdoc_Fo(struct mpage *mpage, const struct roff_meta *meta,
1679 const struct roff_node *n)
1680 {
1681
1682 if (n->type != ROFFT_HEAD)
1683 return 1;
1684
1685 if (n->child != NULL)
1686 parse_mdoc_fname(mpage, n->child);
1687
1688 return 0;
1689 }
1690
1691 static int
parse_mdoc_Va(struct mpage * mpage,const struct roff_meta * meta,const struct roff_node * n)1692 parse_mdoc_Va(struct mpage *mpage, const struct roff_meta *meta,
1693 const struct roff_node *n)
1694 {
1695 char *cp;
1696
1697 if (n->type != ROFFT_ELEM && n->type != ROFFT_BODY)
1698 return 0;
1699
1700 if (n->child != NULL &&
1701 n->child->next == NULL &&
1702 n->child->type == ROFFT_TEXT)
1703 return 1;
1704
1705 cp = NULL;
1706 deroff(&cp, n);
1707 if (cp != NULL) {
1708 putkey(mpage, cp, TYPE_Vt | (n->tok == MDOC_Va ||
1709 n->type == ROFFT_BODY ? TYPE_Va : 0));
1710 free(cp);
1711 }
1712
1713 return 0;
1714 }
1715
1716 static int
parse_mdoc_Xr(struct mpage * mpage,const struct roff_meta * meta,const struct roff_node * n)1717 parse_mdoc_Xr(struct mpage *mpage, const struct roff_meta *meta,
1718 const struct roff_node *n)
1719 {
1720 char *cp;
1721
1722 if (NULL == (n = n->child))
1723 return 0;
1724
1725 if (NULL == n->next) {
1726 putkey(mpage, n->string, TYPE_Xr);
1727 return 0;
1728 }
1729
1730 mandoc_asprintf(&cp, "%s(%s)", n->string, n->next->string);
1731 putkey(mpage, cp, TYPE_Xr);
1732 free(cp);
1733 return 0;
1734 }
1735
1736 static int
parse_mdoc_Nd(struct mpage * mpage,const struct roff_meta * meta,const struct roff_node * n)1737 parse_mdoc_Nd(struct mpage *mpage, const struct roff_meta *meta,
1738 const struct roff_node *n)
1739 {
1740
1741 if (n->type == ROFFT_BODY)
1742 deroff(&mpage->desc, n);
1743 return 0;
1744 }
1745
1746 static int
parse_mdoc_Nm(struct mpage * mpage,const struct roff_meta * meta,const struct roff_node * n)1747 parse_mdoc_Nm(struct mpage *mpage, const struct roff_meta *meta,
1748 const struct roff_node *n)
1749 {
1750
1751 if (SEC_NAME == n->sec)
1752 putmdockey(mpage, n->child, NAME_TITLE, 0);
1753 else if (n->sec == SEC_SYNOPSIS && n->type == ROFFT_HEAD) {
1754 if (n->child == NULL)
1755 putkey(mpage, meta->name, NAME_SYN);
1756 else
1757 putmdockey(mpage, n->child, NAME_SYN, 0);
1758 }
1759 if ( ! (mpage->name_head_done ||
1760 n->child == NULL || n->child->string == NULL ||
1761 strcasecmp(n->child->string, meta->title))) {
1762 putkey(mpage, n->child->string, NAME_HEAD);
1763 mpage->name_head_done = 1;
1764 }
1765 return 0;
1766 }
1767
1768 static int
parse_mdoc_Sh(struct mpage * mpage,const struct roff_meta * meta,const struct roff_node * n)1769 parse_mdoc_Sh(struct mpage *mpage, const struct roff_meta *meta,
1770 const struct roff_node *n)
1771 {
1772
1773 return n->sec == SEC_CUSTOM && n->type == ROFFT_HEAD;
1774 }
1775
1776 static int
parse_mdoc_head(struct mpage * mpage,const struct roff_meta * meta,const struct roff_node * n)1777 parse_mdoc_head(struct mpage *mpage, const struct roff_meta *meta,
1778 const struct roff_node *n)
1779 {
1780
1781 return n->type == ROFFT_HEAD;
1782 }
1783
1784 /*
1785 * Add a string to the hash table for the current manual.
1786 * Each string has a bitmask telling which macros it belongs to.
1787 * When we finish the manual, we'll dump the table.
1788 */
1789 static void
putkeys(const struct mpage * mpage,char * cp,size_t sz,uint64_t v)1790 putkeys(const struct mpage *mpage, char *cp, size_t sz, uint64_t v)
1791 {
1792 struct ohash *htab;
1793 struct str *s;
1794 const char *end;
1795 unsigned int slot;
1796 int i, mustfree;
1797
1798 if (0 == sz)
1799 return;
1800
1801 mustfree = render_string(&cp, &sz);
1802
1803 if (TYPE_Nm & v) {
1804 htab = &names;
1805 v &= name_mask;
1806 if (v & NAME_FIRST)
1807 name_mask &= ~NAME_FIRST;
1808 if (debug > 1)
1809 say(mpage->mlinks->file,
1810 "Adding name %*s, bits=0x%llx", (int)sz, cp,
1811 (unsigned long long)v);
1812 } else {
1813 htab = &strings;
1814 if (debug > 1)
1815 for (i = 0; i < KEY_MAX; i++)
1816 if ((uint64_t)1 << i & v)
1817 say(mpage->mlinks->file,
1818 "Adding key %s=%*s",
1819 mansearch_keynames[i], (int)sz, cp);
1820 }
1821
1822 end = cp + sz;
1823 slot = ohash_qlookupi(htab, cp, &end);
1824 s = ohash_find(htab, slot);
1825
1826 if (NULL != s && mpage == s->mpage) {
1827 s->mask |= v;
1828 return;
1829 } else if (NULL == s) {
1830 s = mandoc_calloc(1, sizeof(struct str) + sz + 1);
1831 memcpy(s->key, cp, sz);
1832 ohash_insert(htab, slot, s);
1833 }
1834 s->mpage = mpage;
1835 s->mask = v;
1836
1837 if (mustfree)
1838 free(cp);
1839 }
1840
1841 /*
1842 * Take a Unicode codepoint and produce its UTF-8 encoding.
1843 * This isn't the best way to do this, but it works.
1844 * The magic numbers are from the UTF-8 packaging.
1845 * They're not as scary as they seem: read the UTF-8 spec for details.
1846 */
1847 static size_t
utf8(unsigned int cp,char out[7])1848 utf8(unsigned int cp, char out[7])
1849 {
1850 size_t rc;
1851
1852 rc = 0;
1853 if (cp <= 0x0000007F) {
1854 rc = 1;
1855 out[0] = (char)cp;
1856 } else if (cp <= 0x000007FF) {
1857 rc = 2;
1858 out[0] = (cp >> 6 & 31) | 192;
1859 out[1] = (cp & 63) | 128;
1860 } else if (cp <= 0x0000FFFF) {
1861 rc = 3;
1862 out[0] = (cp >> 12 & 15) | 224;
1863 out[1] = (cp >> 6 & 63) | 128;
1864 out[2] = (cp & 63) | 128;
1865 } else if (cp <= 0x001FFFFF) {
1866 rc = 4;
1867 out[0] = (cp >> 18 & 7) | 240;
1868 out[1] = (cp >> 12 & 63) | 128;
1869 out[2] = (cp >> 6 & 63) | 128;
1870 out[3] = (cp & 63) | 128;
1871 } else if (cp <= 0x03FFFFFF) {
1872 rc = 5;
1873 out[0] = (cp >> 24 & 3) | 248;
1874 out[1] = (cp >> 18 & 63) | 128;
1875 out[2] = (cp >> 12 & 63) | 128;
1876 out[3] = (cp >> 6 & 63) | 128;
1877 out[4] = (cp & 63) | 128;
1878 } else if (cp <= 0x7FFFFFFF) {
1879 rc = 6;
1880 out[0] = (cp >> 30 & 1) | 252;
1881 out[1] = (cp >> 24 & 63) | 128;
1882 out[2] = (cp >> 18 & 63) | 128;
1883 out[3] = (cp >> 12 & 63) | 128;
1884 out[4] = (cp >> 6 & 63) | 128;
1885 out[5] = (cp & 63) | 128;
1886 } else
1887 return 0;
1888
1889 out[rc] = '\0';
1890 return rc;
1891 }
1892
1893 /*
1894 * If the string contains escape sequences,
1895 * replace it with an allocated rendering and return 1,
1896 * such that the caller can free it after use.
1897 * Otherwise, do nothing and return 0.
1898 */
1899 static int
render_string(char ** public,size_t * psz)1900 render_string(char **public, size_t *psz)
1901 {
1902 const char *src, *scp, *addcp, *seq;
1903 char *dst;
1904 size_t ssz, dsz, addsz;
1905 char utfbuf[7], res[6];
1906 int seqlen, unicode;
1907
1908 res[0] = '\\';
1909 res[1] = '\t';
1910 res[2] = ASCII_NBRSP;
1911 res[3] = ASCII_HYPH;
1912 res[4] = ASCII_BREAK;
1913 res[5] = '\0';
1914
1915 src = scp = *public;
1916 ssz = *psz;
1917 dst = NULL;
1918 dsz = 0;
1919
1920 while (scp < src + *psz) {
1921
1922 /* Leave normal characters unchanged. */
1923
1924 if (strchr(res, *scp) == NULL) {
1925 if (dst != NULL)
1926 dst[dsz++] = *scp;
1927 scp++;
1928 continue;
1929 }
1930
1931 /*
1932 * Found something that requires replacing,
1933 * make sure we have a destination buffer.
1934 */
1935
1936 if (dst == NULL) {
1937 dst = mandoc_malloc(ssz + 1);
1938 dsz = scp - src;
1939 memcpy(dst, src, dsz);
1940 }
1941
1942 /* Handle single-char special characters. */
1943
1944 switch (*scp) {
1945 case '\\':
1946 break;
1947 case '\t':
1948 case ASCII_NBRSP:
1949 dst[dsz++] = ' ';
1950 scp++;
1951 continue;
1952 case ASCII_HYPH:
1953 dst[dsz++] = '-';
1954 /* FALLTHROUGH */
1955 case ASCII_BREAK:
1956 scp++;
1957 continue;
1958 default:
1959 abort();
1960 }
1961
1962 /*
1963 * Found an escape sequence.
1964 * Read past the slash, then parse it.
1965 * Ignore everything except characters.
1966 */
1967
1968 scp++;
1969 if (mandoc_escape(&scp, &seq, &seqlen) != ESCAPE_SPECIAL)
1970 continue;
1971
1972 /*
1973 * Render the special character
1974 * as either UTF-8 or ASCII.
1975 */
1976
1977 if (write_utf8) {
1978 unicode = mchars_spec2cp(seq, seqlen);
1979 if (unicode <= 0)
1980 continue;
1981 addsz = utf8(unicode, utfbuf);
1982 if (addsz == 0)
1983 continue;
1984 addcp = utfbuf;
1985 } else {
1986 addcp = mchars_spec2str(seq, seqlen, &addsz);
1987 if (addcp == NULL)
1988 continue;
1989 if (*addcp == ASCII_NBRSP) {
1990 addcp = " ";
1991 addsz = 1;
1992 }
1993 }
1994
1995 /* Copy the rendered glyph into the stream. */
1996
1997 ssz += addsz;
1998 dst = mandoc_realloc(dst, ssz + 1);
1999 memcpy(dst + dsz, addcp, addsz);
2000 dsz += addsz;
2001 }
2002 if (dst != NULL) {
2003 *public = dst;
2004 *psz = dsz;
2005 }
2006
2007 /* Trim trailing whitespace and NUL-terminate. */
2008
2009 while (*psz > 0 && (*public)[*psz - 1] == ' ')
2010 --*psz;
2011 if (dst != NULL) {
2012 (*public)[*psz] = '\0';
2013 return 1;
2014 } else
2015 return 0;
2016 }
2017
2018 static void
dbadd_mlink(const struct mlink * mlink)2019 dbadd_mlink(const struct mlink *mlink)
2020 {
2021 dba_page_alias(mlink->mpage->dba, mlink->name, NAME_FILE);
2022 dba_page_add(mlink->mpage->dba, DBP_SECT, mlink->dsec);
2023 dba_page_add(mlink->mpage->dba, DBP_SECT, mlink->fsec);
2024 dba_page_add(mlink->mpage->dba, DBP_ARCH, mlink->arch);
2025 dba_page_add(mlink->mpage->dba, DBP_FILE, mlink->file);
2026 }
2027
2028 /*
2029 * Flush the current page's terms (and their bits) into the database.
2030 * Also, handle escape sequences at the last possible moment.
2031 */
2032 static void
dbadd(struct dba * dba,struct mpage * mpage)2033 dbadd(struct dba *dba, struct mpage *mpage)
2034 {
2035 struct mlink *mlink;
2036 struct str *key;
2037 char *cp;
2038 uint64_t mask;
2039 size_t i;
2040 unsigned int slot;
2041 int mustfree;
2042
2043 mlink = mpage->mlinks;
2044
2045 if (nodb) {
2046 for (key = ohash_first(&names, &slot); NULL != key;
2047 key = ohash_next(&names, &slot))
2048 free(key);
2049 for (key = ohash_first(&strings, &slot); NULL != key;
2050 key = ohash_next(&strings, &slot))
2051 free(key);
2052 if (0 == debug)
2053 return;
2054 while (NULL != mlink) {
2055 fputs(mlink->name, stdout);
2056 if (NULL == mlink->next ||
2057 strcmp(mlink->dsec, mlink->next->dsec) ||
2058 strcmp(mlink->fsec, mlink->next->fsec) ||
2059 strcmp(mlink->arch, mlink->next->arch)) {
2060 putchar('(');
2061 if ('\0' == *mlink->dsec)
2062 fputs(mlink->fsec, stdout);
2063 else
2064 fputs(mlink->dsec, stdout);
2065 if ('\0' != *mlink->arch)
2066 printf("/%s", mlink->arch);
2067 putchar(')');
2068 }
2069 mlink = mlink->next;
2070 if (NULL != mlink)
2071 fputs(", ", stdout);
2072 }
2073 printf(" - %s\n", mpage->desc);
2074 return;
2075 }
2076
2077 if (debug)
2078 say(mlink->file, "Adding to database");
2079
2080 cp = mpage->desc;
2081 i = strlen(cp);
2082 mustfree = render_string(&cp, &i);
2083 mpage->dba = dba_page_new(dba->pages,
2084 *mpage->arch == '\0' ? mlink->arch : mpage->arch,
2085 cp, mlink->file, mpage->form);
2086 if (mustfree)
2087 free(cp);
2088 dba_page_add(mpage->dba, DBP_SECT, mpage->sec);
2089
2090 while (mlink != NULL) {
2091 dbadd_mlink(mlink);
2092 mlink = mlink->next;
2093 }
2094
2095 for (key = ohash_first(&names, &slot); NULL != key;
2096 key = ohash_next(&names, &slot)) {
2097 assert(key->mpage == mpage);
2098 dba_page_alias(mpage->dba, key->key, key->mask);
2099 free(key);
2100 }
2101 for (key = ohash_first(&strings, &slot); NULL != key;
2102 key = ohash_next(&strings, &slot)) {
2103 assert(key->mpage == mpage);
2104 i = 0;
2105 for (mask = TYPE_Xr; mask <= TYPE_Lb; mask *= 2) {
2106 if (key->mask & mask)
2107 dba_macro_add(dba->macros, i,
2108 key->key, mpage->dba);
2109 i++;
2110 }
2111 free(key);
2112 }
2113 }
2114
2115 static void
dbprune(struct dba * dba)2116 dbprune(struct dba *dba)
2117 {
2118 struct dba_array *page, *files;
2119 char *file;
2120
2121 dba_array_FOREACH(dba->pages, page) {
2122 files = dba_array_get(page, DBP_FILE);
2123 dba_array_FOREACH(files, file) {
2124 if (*file < ' ')
2125 file++;
2126 if (ohash_find(&mlinks, ohash_qlookup(&mlinks,
2127 file)) != NULL) {
2128 if (debug)
2129 say(file, "Deleting from database");
2130 dba_array_del(dba->pages);
2131 break;
2132 }
2133 }
2134 }
2135 }
2136
2137 /*
2138 * Write the database from memory to disk.
2139 */
2140 static void
dbwrite(struct dba * dba)2141 dbwrite(struct dba *dba)
2142 {
2143 struct stat sb1, sb2;
2144 char tfn[33], *cp1, *cp2;
2145 off_t i;
2146 int fd1, fd2;
2147
2148 /*
2149 * Do not write empty databases, and delete existing ones
2150 * when makewhatis -u causes them to become empty.
2151 */
2152
2153 dba_array_start(dba->pages);
2154 if (dba_array_next(dba->pages) == NULL) {
2155 if (unlink(MANDOC_DB) == -1 && errno != ENOENT)
2156 say(MANDOC_DB, "&unlink");
2157 return;
2158 }
2159
2160 /*
2161 * Build the database in a temporary file,
2162 * then atomically move it into place.
2163 */
2164
2165 if (dba_write(MANDOC_DB "~", dba) != -1) {
2166 if (rename(MANDOC_DB "~", MANDOC_DB) == -1) {
2167 exitcode = (int)MANDOCLEVEL_SYSERR;
2168 say(MANDOC_DB, "&rename");
2169 unlink(MANDOC_DB "~");
2170 }
2171 return;
2172 }
2173
2174 /*
2175 * We lack write permission and cannot replace the database
2176 * file, but let's at least check whether the data changed.
2177 */
2178
2179 (void)strlcpy(tfn, "/tmp/mandocdb.XXXXXXXX", sizeof(tfn));
2180 if (mkdtemp(tfn) == NULL) {
2181 exitcode = (int)MANDOCLEVEL_SYSERR;
2182 say("", "&%s", tfn);
2183 return;
2184 }
2185 cp1 = cp2 = MAP_FAILED;
2186 fd1 = fd2 = -1;
2187 (void)strlcat(tfn, "/" MANDOC_DB, sizeof(tfn));
2188 if (dba_write(tfn, dba) == -1) {
2189 say(tfn, "&dba_write");
2190 goto err;
2191 }
2192 if ((fd1 = open(MANDOC_DB, O_RDONLY, 0)) == -1) {
2193 say(MANDOC_DB, "&open");
2194 goto err;
2195 }
2196 if ((fd2 = open(tfn, O_RDONLY, 0)) == -1) {
2197 say(tfn, "&open");
2198 goto err;
2199 }
2200 if (fstat(fd1, &sb1) == -1) {
2201 say(MANDOC_DB, "&fstat");
2202 goto err;
2203 }
2204 if (fstat(fd2, &sb2) == -1) {
2205 say(tfn, "&fstat");
2206 goto err;
2207 }
2208 if (sb1.st_size != sb2.st_size)
2209 goto err;
2210 if ((cp1 = mmap(NULL, sb1.st_size, PROT_READ, MAP_PRIVATE,
2211 fd1, 0)) == MAP_FAILED) {
2212 say(MANDOC_DB, "&mmap");
2213 goto err;
2214 }
2215 if ((cp2 = mmap(NULL, sb2.st_size, PROT_READ, MAP_PRIVATE,
2216 fd2, 0)) == MAP_FAILED) {
2217 say(tfn, "&mmap");
2218 goto err;
2219 }
2220 for (i = 0; i < sb1.st_size; i++)
2221 if (cp1[i] != cp2[i])
2222 goto err;
2223 goto out;
2224
2225 err:
2226 exitcode = (int)MANDOCLEVEL_SYSERR;
2227 say(MANDOC_DB, "Data changed, but cannot replace database");
2228
2229 out:
2230 if (cp1 != MAP_FAILED)
2231 munmap(cp1, sb1.st_size);
2232 if (cp2 != MAP_FAILED)
2233 munmap(cp2, sb2.st_size);
2234 if (fd1 != -1)
2235 close(fd1);
2236 if (fd2 != -1)
2237 close(fd2);
2238 unlink(tfn);
2239 *strrchr(tfn, '/') = '\0';
2240 rmdir(tfn);
2241 }
2242
2243 static int
set_basedir(const char * targetdir,int report_baddir)2244 set_basedir(const char *targetdir, int report_baddir)
2245 {
2246 static char startdir[PATH_MAX];
2247 static int getcwd_status; /* 1 = ok, 2 = failure */
2248 static int chdir_status; /* 1 = changed directory */
2249 char *cp;
2250
2251 /*
2252 * Remember the original working directory, if possible.
2253 * This will be needed if the second or a later directory
2254 * on the command line is given as a relative path.
2255 * Do not error out if the current directory is not
2256 * searchable: Maybe it won't be needed after all.
2257 */
2258 if (0 == getcwd_status) {
2259 if (NULL == getcwd(startdir, sizeof(startdir))) {
2260 getcwd_status = 2;
2261 (void)strlcpy(startdir, strerror(errno),
2262 sizeof(startdir));
2263 } else
2264 getcwd_status = 1;
2265 }
2266
2267 /*
2268 * We are leaving the old base directory.
2269 * Do not use it any longer, not even for messages.
2270 */
2271 *basedir = '\0';
2272
2273 /*
2274 * If and only if the directory was changed earlier and
2275 * the next directory to process is given as a relative path,
2276 * first go back, or bail out if that is impossible.
2277 */
2278 if (chdir_status && '/' != *targetdir) {
2279 if (2 == getcwd_status) {
2280 exitcode = (int)MANDOCLEVEL_SYSERR;
2281 say("", "getcwd: %s", startdir);
2282 return 0;
2283 }
2284 if (-1 == chdir(startdir)) {
2285 exitcode = (int)MANDOCLEVEL_SYSERR;
2286 say("", "&chdir %s", startdir);
2287 return 0;
2288 }
2289 }
2290
2291 /*
2292 * Always resolve basedir to the canonicalized absolute
2293 * pathname and append a trailing slash, such that
2294 * we can reliably check whether files are inside.
2295 */
2296 if (NULL == realpath(targetdir, basedir)) {
2297 if (report_baddir || errno != ENOENT) {
2298 exitcode = (int)MANDOCLEVEL_BADARG;
2299 say("", "&%s: realpath", targetdir);
2300 }
2301 return 0;
2302 } else if (-1 == chdir(basedir)) {
2303 if (report_baddir || errno != ENOENT) {
2304 exitcode = (int)MANDOCLEVEL_BADARG;
2305 say("", "&chdir");
2306 }
2307 return 0;
2308 }
2309 chdir_status = 1;
2310 cp = strchr(basedir, '\0');
2311 if ('/' != cp[-1]) {
2312 if (cp - basedir >= PATH_MAX - 1) {
2313 exitcode = (int)MANDOCLEVEL_SYSERR;
2314 say("", "Filename too long");
2315 return 0;
2316 }
2317 *cp++ = '/';
2318 *cp = '\0';
2319 }
2320 return 1;
2321 }
2322
2323 static void
say(const char * file,const char * format,...)2324 say(const char *file, const char *format, ...)
2325 {
2326 va_list ap;
2327 int use_errno;
2328
2329 if ('\0' != *basedir)
2330 fprintf(stderr, "%s", basedir);
2331 if ('\0' != *basedir && '\0' != *file)
2332 fputc('/', stderr);
2333 if ('\0' != *file)
2334 fprintf(stderr, "%s", file);
2335
2336 use_errno = 1;
2337 if (NULL != format) {
2338 switch (*format) {
2339 case '&':
2340 format++;
2341 break;
2342 case '\0':
2343 format = NULL;
2344 break;
2345 default:
2346 use_errno = 0;
2347 break;
2348 }
2349 }
2350 if (NULL != format) {
2351 if ('\0' != *basedir || '\0' != *file)
2352 fputs(": ", stderr);
2353 va_start(ap, format);
2354 vfprintf(stderr, format, ap);
2355 va_end(ap);
2356 }
2357 if (use_errno) {
2358 if ('\0' != *basedir || '\0' != *file || NULL != format)
2359 fputs(": ", stderr);
2360 perror(NULL);
2361 } else
2362 fputc('\n', stderr);
2363 }
2364