1 /* $Id: mandocdb.c,v 1.275 2025/06/05 12:33:41 schwarze Exp $ */
2 /*
3 * Copyright (c) 2011-2021, 2024, 2025 Ingo Schwarze <schwarze@openbsd.org>
4 * Copyright (c) 2011, 2012 Kristaps Dzonsons <kristaps@bsd.lv>
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 * Implementation of the makewhatis(8) program.
20 */
21 #include "config.h"
22
23 #include <sys/types.h>
24 #include <sys/mman.h>
25 #include <sys/stat.h>
26
27 #include <assert.h>
28 #include <ctype.h>
29 #if HAVE_ERR
30 #include <err.h>
31 #endif
32 #include <errno.h>
33 #include <fcntl.h>
34 #if HAVE_FTS
35 #include <fts.h>
36 #else
37 #include "compat_fts.h"
38 #endif
39 #include <limits.h>
40 #if HAVE_SANDBOX_INIT
41 #include <sandbox.h>
42 #endif
43 #include <stdarg.h>
44 #include <stddef.h>
45 #include <stdio.h>
46 #include <stdint.h>
47 #include <stdlib.h>
48 #include <string.h>
49 #include <unistd.h>
50
51 #include "mandoc_aux.h"
52 #include "mandoc_ohash.h"
53 #include "mandoc.h"
54 #include "roff.h"
55 #include "mdoc.h"
56 #include "man.h"
57 #include "mandoc_parse.h"
58 #include "manconf.h"
59 #include "mansearch.h"
60 #include "dba_array.h"
61 #include "dba.h"
62
63 extern const char *const mansearch_keynames[];
64
65 enum op {
66 OP_DEFAULT = 0, /* new dbs from dir list or default config */
67 OP_CONFFILE, /* new databases from custom config file */
68 OP_UPDATE, /* delete/add entries in existing database */
69 OP_DELETE, /* delete entries from existing database */
70 OP_TEST /* change no databases, report potential problems */
71 };
72
73 struct str {
74 const struct mpage *mpage; /* if set, the owning parse */
75 uint64_t mask; /* bitmask in sequence */
76 char key[]; /* rendered text */
77 };
78
79 struct inodev {
80 ino_t st_ino;
81 dev_t st_dev;
82 };
83
84 struct mpage {
85 struct inodev inodev; /* used for hashing routine */
86 struct dba_array *dba;
87 char *sec; /* section from file content */
88 char *arch; /* architecture from file content */
89 char *title; /* title from file content */
90 char *desc; /* description from file content */
91 struct mpage *next; /* singly linked list */
92 struct mlink *mlinks; /* singly linked list */
93 int name_head_done;
94 enum form form; /* format from file content */
95 };
96
97 struct mlink {
98 char file[PATH_MAX]; /* filename rel. to manpath */
99 char *dsec; /* section from directory */
100 char *arch; /* architecture from directory */
101 char *name; /* name from file name (not empty) */
102 char *fsec; /* section from file name suffix */
103 struct mlink *next; /* singly linked list */
104 struct mpage *mpage; /* parent */
105 int gzip; /* filename has a .gz suffix */
106 enum form dform; /* format from directory */
107 enum form fform; /* format from file name suffix */
108 };
109
110 typedef int (*mdoc_fp)(struct mpage *, const struct roff_meta *,
111 const struct roff_node *);
112
113 struct mdoc_handler {
114 mdoc_fp fp; /* optional handler */
115 uint64_t mask; /* set unless handler returns 0 */
116 int taboo; /* node flags that must not be set */
117 };
118
119
120 int mandocdb(int, char *[]);
121
122 static void dbadd(struct dba *, struct mpage *);
123 static void dbadd_mlink(const struct mlink *);
124 static void dbprune(struct dba *);
125 static void dbwrite(struct dba *);
126 static void filescan(const char *);
127 #if HAVE_FTS_COMPARE_CONST
128 static int fts_compare(const FTSENT *const *, const FTSENT *const *);
129 #else
130 static int fts_compare(const FTSENT **, const FTSENT **);
131 #endif
132 static void mlink_add(struct mlink *, const struct stat *);
133 static void mlink_check(struct mpage *, struct mlink *);
134 static void mlink_free(struct mlink *);
135 static void mlinks_undupe(struct mpage *);
136 static void mpages_free(void);
137 static void mpages_merge(struct dba *, struct mparse *);
138 static void parse_cat(struct mpage *, int);
139 static void parse_man(struct mpage *, const struct roff_meta *,
140 const struct roff_node *);
141 static void parse_mdoc(struct mpage *, const struct roff_meta *,
142 const struct roff_node *);
143 static int parse_mdoc_head(struct mpage *, const struct roff_meta *,
144 const struct roff_node *);
145 static int parse_mdoc_Fa(struct mpage *, const struct roff_meta *,
146 const struct roff_node *);
147 static int parse_mdoc_Fd(struct mpage *, const struct roff_meta *,
148 const struct roff_node *);
149 static void parse_mdoc_fname(struct mpage *, const struct roff_node *);
150 static int parse_mdoc_Fn(struct mpage *, const struct roff_meta *,
151 const struct roff_node *);
152 static int parse_mdoc_Fo(struct mpage *, const struct roff_meta *,
153 const struct roff_node *);
154 static int parse_mdoc_Lb(struct mpage *, const struct roff_meta *,
155 const struct roff_node *);
156 static int parse_mdoc_Nd(struct mpage *, const struct roff_meta *,
157 const struct roff_node *);
158 static int parse_mdoc_Nm(struct mpage *, const struct roff_meta *,
159 const struct roff_node *);
160 static int parse_mdoc_Sh(struct mpage *, const struct roff_meta *,
161 const struct roff_node *);
162 static int parse_mdoc_Va(struct mpage *, const struct roff_meta *,
163 const struct roff_node *);
164 static int parse_mdoc_Xr(struct mpage *, const struct roff_meta *,
165 const struct roff_node *);
166 static void putkey(const struct mpage *, char *, uint64_t);
167 static void putkeys(const struct mpage *, char *, size_t, uint64_t);
168 static void putmdockey(const struct mpage *,
169 const struct roff_node *, uint64_t, int);
170 #ifdef READ_ALLOWED_PATH
171 static int read_allowed(const char *);
172 #endif
173 static int render_string(char **, size_t *);
174 static void say(const char *, const char *, ...)
175 __attribute__((__format__ (__printf__, 2, 3)));
176 static int set_basedir(const char *, int);
177 static int treescan(void);
178 static size_t utf8(unsigned int, char[5]);
179
180 static int nodb; /* no database changes */
181 static int mparse_options; /* abort the parse early */
182 static int use_all; /* use all found files */
183 static int debug; /* print what we're doing */
184 static int warnings; /* warn about crap */
185 static int write_utf8; /* write UTF-8 output; else ASCII */
186 static int exitcode; /* to be returned by main */
187 static enum op op; /* operational mode */
188 static char basedir[PATH_MAX]; /* current base directory */
189 static size_t basedir_len; /* strlen(basedir) */
190 static struct mpage *mpage_head; /* list of distinct manual pages */
191 static struct ohash mpages; /* table of distinct manual pages */
192 static struct ohash mlinks; /* table of directory entries */
193 static struct ohash names; /* table of all names */
194 static struct ohash strings; /* table of all strings */
195 static uint64_t name_mask;
196
197 static const struct mdoc_handler mdoc_handlers[MDOC_MAX - MDOC_Dd] = {
198 { NULL, 0, NODE_NOPRT }, /* Dd */
199 { NULL, 0, NODE_NOPRT }, /* Dt */
200 { NULL, 0, NODE_NOPRT }, /* Os */
201 { parse_mdoc_Sh, TYPE_Sh, 0 }, /* Sh */
202 { parse_mdoc_head, TYPE_Ss, 0 }, /* Ss */
203 { NULL, 0, 0 }, /* Pp */
204 { NULL, 0, 0 }, /* D1 */
205 { NULL, 0, 0 }, /* Dl */
206 { NULL, 0, 0 }, /* Bd */
207 { NULL, 0, 0 }, /* Ed */
208 { NULL, 0, 0 }, /* Bl */
209 { NULL, 0, 0 }, /* El */
210 { NULL, 0, 0 }, /* It */
211 { NULL, 0, 0 }, /* Ad */
212 { NULL, TYPE_An, 0 }, /* An */
213 { NULL, 0, 0 }, /* Ap */
214 { NULL, TYPE_Ar, 0 }, /* Ar */
215 { NULL, TYPE_Cd, 0 }, /* Cd */
216 { NULL, TYPE_Cm, 0 }, /* Cm */
217 { NULL, TYPE_Dv, 0 }, /* Dv */
218 { NULL, TYPE_Er, 0 }, /* Er */
219 { NULL, TYPE_Ev, 0 }, /* Ev */
220 { NULL, 0, 0 }, /* Ex */
221 { parse_mdoc_Fa, 0, 0 }, /* Fa */
222 { parse_mdoc_Fd, 0, 0 }, /* Fd */
223 { NULL, TYPE_Fl, 0 }, /* Fl */
224 { parse_mdoc_Fn, 0, 0 }, /* Fn */
225 { NULL, TYPE_Ft | TYPE_Vt, 0 }, /* Ft */
226 { NULL, TYPE_Ic, 0 }, /* Ic */
227 { NULL, TYPE_In, 0 }, /* In */
228 { NULL, TYPE_Li, 0 }, /* Li */
229 { parse_mdoc_Nd, 0, 0 }, /* Nd */
230 { parse_mdoc_Nm, 0, 0 }, /* Nm */
231 { NULL, 0, 0 }, /* Op */
232 { NULL, 0, 0 }, /* Ot */
233 { NULL, TYPE_Pa, NODE_NOSRC }, /* Pa */
234 { NULL, 0, 0 }, /* Rv */
235 { NULL, TYPE_St, 0 }, /* St */
236 { parse_mdoc_Va, TYPE_Va, 0 }, /* Va */
237 { parse_mdoc_Va, TYPE_Vt, 0 }, /* Vt */
238 { parse_mdoc_Xr, 0, 0 }, /* Xr */
239 { NULL, 0, 0 }, /* %A */
240 { NULL, 0, 0 }, /* %B */
241 { NULL, 0, 0 }, /* %D */
242 { NULL, 0, 0 }, /* %I */
243 { NULL, 0, 0 }, /* %J */
244 { NULL, 0, 0 }, /* %N */
245 { NULL, 0, 0 }, /* %O */
246 { NULL, 0, 0 }, /* %P */
247 { NULL, 0, 0 }, /* %R */
248 { NULL, 0, 0 }, /* %T */
249 { NULL, 0, 0 }, /* %V */
250 { NULL, 0, 0 }, /* Ac */
251 { NULL, 0, 0 }, /* Ao */
252 { NULL, 0, 0 }, /* Aq */
253 { NULL, TYPE_At, 0 }, /* At */
254 { NULL, 0, 0 }, /* Bc */
255 { NULL, 0, 0 }, /* Bf */
256 { NULL, 0, 0 }, /* Bo */
257 { NULL, 0, 0 }, /* Bq */
258 { NULL, TYPE_Bsx, NODE_NOSRC }, /* Bsx */
259 { NULL, TYPE_Bx, NODE_NOSRC }, /* Bx */
260 { NULL, 0, 0 }, /* Db */
261 { NULL, 0, 0 }, /* Dc */
262 { NULL, 0, 0 }, /* Do */
263 { NULL, 0, 0 }, /* Dq */
264 { NULL, 0, 0 }, /* Ec */
265 { NULL, 0, 0 }, /* Ef */
266 { NULL, TYPE_Em, 0 }, /* Em */
267 { NULL, 0, 0 }, /* Eo */
268 { NULL, TYPE_Fx, NODE_NOSRC }, /* Fx */
269 { NULL, TYPE_Ms, 0 }, /* Ms */
270 { NULL, 0, 0 }, /* No */
271 { NULL, 0, 0 }, /* Ns */
272 { NULL, TYPE_Nx, NODE_NOSRC }, /* Nx */
273 { NULL, TYPE_Ox, NODE_NOSRC }, /* Ox */
274 { NULL, 0, 0 }, /* Pc */
275 { NULL, 0, 0 }, /* Pf */
276 { NULL, 0, 0 }, /* Po */
277 { NULL, 0, 0 }, /* Pq */
278 { NULL, 0, 0 }, /* Qc */
279 { NULL, 0, 0 }, /* Ql */
280 { NULL, 0, 0 }, /* Qo */
281 { NULL, 0, 0 }, /* Qq */
282 { NULL, 0, 0 }, /* Re */
283 { NULL, 0, 0 }, /* Rs */
284 { NULL, 0, 0 }, /* Sc */
285 { NULL, 0, 0 }, /* So */
286 { NULL, 0, 0 }, /* Sq */
287 { NULL, 0, 0 }, /* Sm */
288 { NULL, 0, 0 }, /* Sx */
289 { NULL, TYPE_Sy, 0 }, /* Sy */
290 { NULL, TYPE_Tn, 0 }, /* Tn */
291 { NULL, 0, NODE_NOSRC }, /* Ux */
292 { NULL, 0, 0 }, /* Xc */
293 { NULL, 0, 0 }, /* Xo */
294 { parse_mdoc_Fo, 0, 0 }, /* Fo */
295 { NULL, 0, 0 }, /* Fc */
296 { NULL, 0, 0 }, /* Oo */
297 { NULL, 0, 0 }, /* Oc */
298 { NULL, 0, 0 }, /* Bk */
299 { NULL, 0, 0 }, /* Ek */
300 { NULL, 0, 0 }, /* Bt */
301 { NULL, 0, 0 }, /* Hf */
302 { NULL, 0, 0 }, /* Fr */
303 { NULL, 0, 0 }, /* Ud */
304 { parse_mdoc_Lb, 0, 0 }, /* Lb */
305 { NULL, 0, 0 }, /* Lp */
306 { NULL, TYPE_Lk, 0 }, /* Lk */
307 { NULL, TYPE_Mt, NODE_NOSRC }, /* Mt */
308 { NULL, 0, 0 }, /* Brq */
309 { NULL, 0, 0 }, /* Bro */
310 { NULL, 0, 0 }, /* Brc */
311 { NULL, 0, 0 }, /* %C */
312 { NULL, 0, 0 }, /* Es */
313 { NULL, 0, 0 }, /* En */
314 { NULL, TYPE_Dx, NODE_NOSRC }, /* Dx */
315 { NULL, 0, 0 }, /* %Q */
316 { NULL, 0, 0 }, /* %U */
317 { NULL, 0, 0 }, /* Ta */
318 };
319
320
321 int
mandocdb(int argc,char * argv[])322 mandocdb(int argc, char *argv[])
323 {
324 struct manconf conf;
325 struct mparse *mp;
326 struct dba *dba;
327 const char *path_arg, *progname;
328 size_t j, sz;
329 int ch, i;
330
331 #if HAVE_PLEDGE
332 if (pledge("stdio rpath wpath cpath", NULL) == -1) {
333 warn("pledge");
334 return (int)MANDOCLEVEL_SYSERR;
335 }
336 #endif
337
338 #if HAVE_SANDBOX_INIT
339 if (sandbox_init(kSBXProfileNoInternet, SANDBOX_NAMED, NULL) == -1) {
340 warnx("sandbox_init");
341 return (int)MANDOCLEVEL_SYSERR;
342 }
343 #endif
344
345 memset(&conf, 0, sizeof(conf));
346
347 /*
348 * We accept a few different invocations.
349 * The CHECKOP macro makes sure that invocation styles don't
350 * clobber each other.
351 */
352 #define CHECKOP(_op, _ch) do \
353 if ((_op) != OP_DEFAULT) { \
354 warnx("-%c: Conflicting option", (_ch)); \
355 goto usage; \
356 } while (/*CONSTCOND*/0)
357
358 mparse_options = MPARSE_UTF8 | MPARSE_LATIN1 | MPARSE_VALIDATE;
359 path_arg = NULL;
360 op = OP_DEFAULT;
361
362 while ((ch = getopt(argc, argv, "aC:Dd:npQT:tu:v")) != -1)
363 switch (ch) {
364 case 'a':
365 use_all = 1;
366 break;
367 case 'C':
368 CHECKOP(op, ch);
369 path_arg = optarg;
370 op = OP_CONFFILE;
371 break;
372 case 'D':
373 debug++;
374 break;
375 case 'd':
376 CHECKOP(op, ch);
377 path_arg = optarg;
378 op = OP_UPDATE;
379 break;
380 case 'n':
381 nodb = 1;
382 break;
383 case 'p':
384 warnings = 1;
385 break;
386 case 'Q':
387 mparse_options |= MPARSE_QUICK;
388 break;
389 case 'T':
390 if (strcmp(optarg, "utf8") != 0) {
391 warnx("-T%s: Unsupported output format",
392 optarg);
393 goto usage;
394 }
395 write_utf8 = 1;
396 break;
397 case 't':
398 CHECKOP(op, ch);
399 dup2(STDOUT_FILENO, STDERR_FILENO);
400 op = OP_TEST;
401 nodb = warnings = 1;
402 break;
403 case 'u':
404 CHECKOP(op, ch);
405 path_arg = optarg;
406 op = OP_DELETE;
407 break;
408 case 'v':
409 /* Compatibility with espie@'s makewhatis. */
410 break;
411 default:
412 goto usage;
413 }
414
415 argc -= optind;
416 argv += optind;
417
418 #if HAVE_PLEDGE
419 if (nodb) {
420 if (pledge("stdio rpath", NULL) == -1) {
421 warn("pledge");
422 return (int)MANDOCLEVEL_SYSERR;
423 }
424 }
425 #endif
426
427 if (op == OP_CONFFILE && argc > 0) {
428 warnx("-C: Too many arguments");
429 goto usage;
430 }
431
432 exitcode = (int)MANDOCLEVEL_OK;
433 mchars_alloc();
434 mp = mparse_alloc(mparse_options, MANDOC_OS_OTHER, NULL);
435 mandoc_ohash_init(&mpages, 6, offsetof(struct mpage, inodev));
436 mandoc_ohash_init(&mlinks, 6, offsetof(struct mlink, file));
437
438 if (op == OP_UPDATE || op == OP_DELETE || op == OP_TEST) {
439
440 /*
441 * Most of these deal with a specific directory.
442 * Jump into that directory first.
443 */
444 if (op != OP_TEST && set_basedir(path_arg, 1) == 0)
445 goto out;
446
447 dba = nodb ? dba_new(128) : dba_read(MANDOC_DB);
448 if (dba != NULL) {
449 /*
450 * The existing database is usable. Process
451 * all files specified on the command-line.
452 */
453 use_all = 1;
454 for (i = 0; i < argc; i++)
455 filescan(argv[i]);
456 if (nodb == 0)
457 dbprune(dba);
458 } else {
459 /* Database missing or corrupt. */
460 if (op != OP_UPDATE || errno != ENOENT)
461 say(MANDOC_DB, "%s: Automatically recreating"
462 " from scratch", strerror(errno));
463 exitcode = (int)MANDOCLEVEL_OK;
464 op = OP_DEFAULT;
465 if (treescan() == 0)
466 goto out;
467 dba = dba_new(128);
468 }
469 if (op != OP_DELETE)
470 mpages_merge(dba, mp);
471 if (nodb == 0)
472 dbwrite(dba);
473 dba_free(dba);
474 } else {
475 /*
476 * If we have arguments, use them as our manpaths.
477 * If we don't, use man.conf(5).
478 */
479 if (argc > 0) {
480 conf.manpath.paths = mandoc_reallocarray(NULL,
481 argc, sizeof(char *));
482 conf.manpath.sz = (size_t)argc;
483 for (i = 0; i < argc; i++)
484 conf.manpath.paths[i] = mandoc_strdup(argv[i]);
485 } else
486 manconf_parse(&conf, path_arg, NULL, NULL);
487
488 if (conf.manpath.sz == 0) {
489 exitcode = (int)MANDOCLEVEL_BADARG;
490 say("", "Empty manpath");
491 }
492
493 /*
494 * First scan the tree rooted at a base directory, then
495 * build a new database and finally move it into place.
496 * Ignore zero-length directories and strip trailing
497 * slashes.
498 */
499 for (j = 0; j < conf.manpath.sz; j++) {
500 sz = strlen(conf.manpath.paths[j]);
501 if (sz && conf.manpath.paths[j][sz - 1] == '/')
502 conf.manpath.paths[j][--sz] = '\0';
503 if (sz == 0)
504 continue;
505
506 if (j) {
507 mandoc_ohash_init(&mpages, 6,
508 offsetof(struct mpage, inodev));
509 mandoc_ohash_init(&mlinks, 6,
510 offsetof(struct mlink, file));
511 }
512
513 if (set_basedir(conf.manpath.paths[j], argc > 0) == 0)
514 continue;
515 if (treescan() == 0)
516 continue;
517 dba = dba_new(128);
518 mpages_merge(dba, mp);
519 if (nodb == 0)
520 dbwrite(dba);
521 dba_free(dba);
522
523 if (j + 1 < conf.manpath.sz) {
524 mpages_free();
525 ohash_delete(&mpages);
526 ohash_delete(&mlinks);
527 }
528 }
529 }
530 out:
531 manconf_free(&conf);
532 mparse_free(mp);
533 mchars_free();
534 mpages_free();
535 ohash_delete(&mpages);
536 ohash_delete(&mlinks);
537 #if DEBUG_MEMORY
538 mandoc_dbg_finish();
539 #endif
540 return exitcode;
541 usage:
542 progname = getprogname();
543 fprintf(stderr, "usage: %s [-aDnpQ] [-C file] [-Tutf8]\n"
544 " %s [-aDnpQ] [-Tutf8] dir ...\n"
545 " %s [-DnpQ] [-Tutf8] -d dir [file ...]\n"
546 " %s [-Dnp] -u dir [file ...]\n"
547 " %s [-Q] -t file ...\n",
548 progname, progname, progname, progname, progname);
549
550 return (int)MANDOCLEVEL_BADARG;
551 }
552
553 /*
554 * To get a singly linked list in alpha order while inserting entries
555 * at the beginning, process directory entries in reverse alpha order.
556 */
557 static int
558 #if HAVE_FTS_COMPARE_CONST
fts_compare(const FTSENT * const * a,const FTSENT * const * b)559 fts_compare(const FTSENT *const *a, const FTSENT *const *b)
560 #else
561 fts_compare(const FTSENT **a, const FTSENT **b)
562 #endif
563 {
564 return -strcmp((*a)->fts_name, (*b)->fts_name);
565 }
566
567 /*
568 * Scan a directory tree rooted at "basedir" for manpages.
569 * We use fts(), scanning directory parts along the way for clues to our
570 * section and architecture.
571 *
572 * If use_all has been specified, grok all files.
573 * If not, sanitise paths to the following:
574 *
575 * [./]man*[/<arch>]/<name>.<section>
576 * or
577 * [./]cat<section>[/<arch>]/<name>.0
578 *
579 * TODO: accommodate for multi-language directories.
580 */
581 static int
treescan(void)582 treescan(void)
583 {
584 char buf[PATH_MAX];
585 FTS *f;
586 FTSENT *ff;
587 struct mlink *mlink;
588 int gzip;
589 enum form dform;
590 char *dsec, *arch, *fsec, *cp;
591 const char *path;
592 const char *argv[2];
593
594 argv[0] = ".";
595 argv[1] = NULL;
596
597 f = fts_open((char * const *)argv, FTS_PHYSICAL | FTS_NOCHDIR,
598 fts_compare);
599 if (f == NULL) {
600 exitcode = (int)MANDOCLEVEL_SYSERR;
601 say("", "&fts_open");
602 return 0;
603 }
604
605 dsec = arch = NULL;
606 dform = FORM_NONE;
607
608 while ((ff = fts_read(f)) != NULL) {
609 path = ff->fts_path + 2;
610 switch (ff->fts_info) {
611
612 /*
613 * Symbolic links require various sanity checks,
614 * then get handled just like regular files.
615 */
616 case FTS_SL:
617 if (realpath(path, buf) == NULL) {
618 if (warnings)
619 say(path, "&realpath");
620 continue;
621 }
622 if (strncmp(buf, basedir, basedir_len) != 0
623 #ifdef READ_ALLOWED_PATH
624 && !read_allowed(buf)
625 #endif
626 ) {
627 if (warnings) say("",
628 "%s: outside base directory", buf);
629 continue;
630 }
631 /* Use logical inode to avoid mpages dupe. */
632 if (stat(path, ff->fts_statp) == -1) {
633 if (warnings)
634 say(path, "&stat");
635 continue;
636 }
637 if ((ff->fts_statp->st_mode & S_IFMT) != S_IFREG)
638 continue;
639 /* FALLTHROUGH */
640
641 /*
642 * If we're a regular file, add an mlink by using the
643 * stored directory data and handling the filename.
644 */
645 case FTS_F:
646 if ( ! strcmp(path, MANDOC_DB))
647 continue;
648 if ( ! use_all && ff->fts_level < 2) {
649 if (warnings)
650 say(path, "Extraneous file");
651 continue;
652 }
653 gzip = 0;
654 fsec = NULL;
655 while (fsec == NULL) {
656 fsec = strrchr(ff->fts_name, '.');
657 if (fsec == NULL || strcmp(fsec+1, "gz"))
658 break;
659 gzip = 1;
660 *fsec = '\0';
661 fsec = NULL;
662 }
663 if (fsec == NULL) {
664 if ( ! use_all) {
665 if (warnings)
666 say(path,
667 "No filename suffix");
668 continue;
669 }
670 } else if ( ! strcmp(++fsec, "html")) {
671 if (warnings)
672 say(path, "Skip html");
673 continue;
674 } else if ( ! strcmp(fsec, "ps")) {
675 if (warnings)
676 say(path, "Skip ps");
677 continue;
678 } else if ( ! strcmp(fsec, "pdf")) {
679 if (warnings)
680 say(path, "Skip pdf");
681 continue;
682 } else if ( ! use_all &&
683 ((dform == FORM_SRC &&
684 strncmp(fsec, dsec, strlen(dsec))) ||
685 (dform == FORM_CAT && strcmp(fsec, "0")))) {
686 if (warnings)
687 say(path, "Wrong filename suffix");
688 continue;
689 } else
690 fsec[-1] = '\0';
691
692 mlink = mandoc_calloc(1, sizeof(struct mlink));
693 if (strlcpy(mlink->file, path,
694 sizeof(mlink->file)) >=
695 sizeof(mlink->file)) {
696 say(path, "Filename too long");
697 free(mlink);
698 continue;
699 }
700 mlink->dform = dform;
701 mlink->dsec = dsec;
702 mlink->arch = arch;
703 mlink->name = ff->fts_name;
704 mlink->fsec = fsec;
705 mlink->gzip = gzip;
706 mlink_add(mlink, ff->fts_statp);
707 continue;
708
709 case FTS_D:
710 case FTS_DP:
711 break;
712
713 default:
714 if (warnings)
715 say(path, "Not a regular file");
716 continue;
717 }
718
719 switch (ff->fts_level) {
720 case 0:
721 /* Ignore the root directory. */
722 break;
723 case 1:
724 /*
725 * This might contain manX/ or catX/.
726 * Try to infer this from the name.
727 * If we're not in use_all, enforce it.
728 */
729 cp = ff->fts_name;
730 if (ff->fts_info == FTS_DP) {
731 dform = FORM_NONE;
732 dsec = NULL;
733 break;
734 }
735
736 if ( ! strncmp(cp, "man", 3)) {
737 dform = FORM_SRC;
738 dsec = cp + 3;
739 } else if ( ! strncmp(cp, "cat", 3)) {
740 dform = FORM_CAT;
741 dsec = cp + 3;
742 } else {
743 dform = FORM_NONE;
744 dsec = NULL;
745 }
746
747 if (dsec != NULL || use_all)
748 break;
749
750 if (warnings)
751 say(path, "Unknown directory part");
752 fts_set(f, ff, FTS_SKIP);
753 break;
754 case 2:
755 /*
756 * Possibly our architecture.
757 * If we're descending, keep tabs on it.
758 */
759 if (ff->fts_info != FTS_DP && dsec != NULL)
760 arch = ff->fts_name;
761 else
762 arch = NULL;
763 break;
764 default:
765 if (ff->fts_info == FTS_DP || use_all)
766 break;
767 if (warnings)
768 say(path, "Extraneous directory part");
769 fts_set(f, ff, FTS_SKIP);
770 break;
771 }
772 }
773
774 fts_close(f);
775 return 1;
776 }
777
778 /*
779 * Add a file to the mlinks table.
780 * Do not verify that it's a "valid" looking manpage (we'll do that
781 * later).
782 *
783 * Try to infer the manual section, architecture, and page name from the
784 * path, assuming it looks like
785 *
786 * [./]man*[/<arch>]/<name>.<section>
787 * or
788 * [./]cat<section>[/<arch>]/<name>.0
789 *
790 * See treescan() for the fts(3) version of this.
791 */
792 static void
filescan(const char * infile)793 filescan(const char *infile)
794 {
795 struct stat st;
796 struct mlink *mlink;
797 char *linkfile, *p, *realdir, *start, *usefile;
798 size_t realdir_len;
799
800 assert(use_all);
801
802 if (strncmp(infile, "./", 2) == 0)
803 infile += 2;
804
805 /*
806 * We have to do lstat(2) before realpath(3) loses
807 * the information whether this is a symbolic link.
808 * We need to know that because for symbolic links,
809 * we want to use the original file name, while for
810 * regular files, we want to use the real path.
811 */
812 if (lstat(infile, &st) == -1) {
813 exitcode = (int)MANDOCLEVEL_BADARG;
814 say(infile, "&lstat");
815 return;
816 } else if (S_ISREG(st.st_mode) == 0 && S_ISLNK(st.st_mode) == 0) {
817 exitcode = (int)MANDOCLEVEL_BADARG;
818 say(infile, "Not a regular file");
819 return;
820 }
821
822 /*
823 * We have to resolve the file name to the real path
824 * in any case for the base directory check.
825 */
826 if ((usefile = realpath(infile, NULL)) == NULL) {
827 exitcode = (int)MANDOCLEVEL_BADARG;
828 say(infile, "&realpath");
829 return;
830 }
831
832 if (op == OP_TEST)
833 start = usefile;
834 else if (strncmp(usefile, basedir, basedir_len) == 0)
835 start = usefile + basedir_len;
836 #ifdef READ_ALLOWED_PATH
837 else if (read_allowed(usefile))
838 start = usefile;
839 #endif
840 else {
841 exitcode = (int)MANDOCLEVEL_BADARG;
842 say("", "%s: outside base directory", infile);
843 free(usefile);
844 return;
845 }
846
847 /*
848 * Now we are sure the file is inside our tree.
849 * If it is a symbolic link, ignore the real path
850 * and use the original name.
851 */
852 do {
853 if (S_ISLNK(st.st_mode) == 0)
854 break;
855
856 /*
857 * Some implementations of realpath(3) may succeed
858 * even if the target of the link does not exist,
859 * so check again for extra safety.
860 */
861 if (stat(usefile, &st) == -1) {
862 exitcode = (int)MANDOCLEVEL_BADARG;
863 say(infile, "&stat");
864 free(usefile);
865 return;
866 }
867 linkfile = mandoc_strdup(infile);
868 if (op == OP_TEST) {
869 free(usefile);
870 start = usefile = linkfile;
871 break;
872 }
873 if (strncmp(infile, basedir, basedir_len) == 0) {
874 free(usefile);
875 usefile = linkfile;
876 start = usefile + basedir_len;
877 break;
878 }
879
880 /*
881 * This symbolic link points into the basedir
882 * from the outside. Let's see whether any of
883 * the parent directories resolve to the basedir.
884 */
885 p = strchr(linkfile, '\0');
886 do {
887 while (*--p != '/')
888 continue;
889 *p = '\0';
890 if ((realdir = realpath(linkfile, NULL)) == NULL) {
891 exitcode = (int)MANDOCLEVEL_BADARG;
892 say(infile, "&realpath");
893 free(linkfile);
894 free(usefile);
895 return;
896 }
897 realdir_len = strlen(realdir) + 1;
898 free(realdir);
899 *p = '/';
900 } while (realdir_len > basedir_len);
901
902 /*
903 * If one of the directories resolves to the basedir,
904 * use the rest of the original name.
905 * Otherwise, the best we can do
906 * is to use the filename pointed to.
907 */
908 if (realdir_len == basedir_len) {
909 free(usefile);
910 usefile = linkfile;
911 start = p + 1;
912 } else {
913 free(linkfile);
914 start = usefile + basedir_len;
915 }
916 } while (/* CONSTCOND */ 0);
917
918 mlink = mandoc_calloc(1, sizeof(struct mlink));
919 mlink->dform = FORM_NONE;
920 if (strlcpy(mlink->file, start, sizeof(mlink->file)) >=
921 sizeof(mlink->file)) {
922 say(start, "Filename too long");
923 free(mlink);
924 free(usefile);
925 return;
926 }
927
928 /*
929 * In test mode or when the original name is absolute
930 * but outside our tree, guess the base directory.
931 */
932
933 if (op == OP_TEST || (start == usefile && *start == '/')) {
934 if (strncmp(usefile, "man/", 4) == 0)
935 start = usefile + 4;
936 else if ((start = strstr(usefile, "/man/")) != NULL)
937 start += 5;
938 else
939 start = usefile;
940 }
941
942 /*
943 * First try to guess our directory structure.
944 * If we find a separator, try to look for man* or cat*.
945 * If we find one of these and what's underneath is a directory,
946 * assume it's an architecture.
947 */
948 if ((p = strchr(start, '/')) != NULL) {
949 *p++ = '\0';
950 if (strncmp(start, "man", 3) == 0) {
951 mlink->dform = FORM_SRC;
952 mlink->dsec = start + 3;
953 } else if (strncmp(start, "cat", 3) == 0) {
954 mlink->dform = FORM_CAT;
955 mlink->dsec = start + 3;
956 }
957
958 start = p;
959 if (mlink->dsec != NULL && (p = strchr(start, '/')) != NULL) {
960 *p++ = '\0';
961 mlink->arch = start;
962 start = p;
963 }
964 }
965
966 /*
967 * Now check the file suffix.
968 * Suffix of `.0' indicates a catpage, `.1-9' is a manpage.
969 */
970 p = strrchr(start, '\0');
971 while (p-- > start && *p != '/' && *p != '.')
972 continue;
973
974 if (*p == '.') {
975 *p++ = '\0';
976 mlink->fsec = p;
977 }
978
979 /*
980 * Now try to parse the name.
981 * Use the filename portion of the path.
982 */
983 mlink->name = start;
984 if ((p = strrchr(start, '/')) != NULL) {
985 mlink->name = p + 1;
986 *p = '\0';
987 }
988 mlink_add(mlink, &st);
989 free(usefile);
990 }
991
992 static void
mlink_add(struct mlink * mlink,const struct stat * st)993 mlink_add(struct mlink *mlink, const struct stat *st)
994 {
995 struct inodev inodev;
996 struct mpage *mpage;
997 unsigned int slot;
998
999 assert(NULL != mlink->file);
1000
1001 mlink->dsec = mandoc_strdup(mlink->dsec ? mlink->dsec : "");
1002 mlink->arch = mandoc_strdup(mlink->arch ? mlink->arch : "");
1003 mlink->name = mandoc_strdup(mlink->name ? mlink->name : "");
1004 mlink->fsec = mandoc_strdup(mlink->fsec ? mlink->fsec : "");
1005
1006 if ('0' == *mlink->fsec) {
1007 free(mlink->fsec);
1008 mlink->fsec = mandoc_strdup(mlink->dsec);
1009 mlink->fform = FORM_CAT;
1010 } else if ('1' <= *mlink->fsec && '9' >= *mlink->fsec)
1011 mlink->fform = FORM_SRC;
1012 else
1013 mlink->fform = FORM_NONE;
1014
1015 slot = ohash_qlookup(&mlinks, mlink->file);
1016 assert(NULL == ohash_find(&mlinks, slot));
1017 ohash_insert(&mlinks, slot, mlink);
1018
1019 memset(&inodev, 0, sizeof(inodev)); /* Clear padding. */
1020 inodev.st_ino = st->st_ino;
1021 inodev.st_dev = st->st_dev;
1022 slot = ohash_lookup_memory(&mpages, (char *)&inodev,
1023 sizeof(struct inodev), inodev.st_ino);
1024 mpage = ohash_find(&mpages, slot);
1025 if (NULL == mpage) {
1026 mpage = mandoc_calloc(1, sizeof(struct mpage));
1027 mpage->inodev.st_ino = inodev.st_ino;
1028 mpage->inodev.st_dev = inodev.st_dev;
1029 mpage->form = FORM_NONE;
1030 mpage->next = mpage_head;
1031 mpage_head = mpage;
1032 ohash_insert(&mpages, slot, mpage);
1033 } else
1034 mlink->next = mpage->mlinks;
1035 mpage->mlinks = mlink;
1036 mlink->mpage = mpage;
1037 }
1038
1039 static void
mlink_free(struct mlink * mlink)1040 mlink_free(struct mlink *mlink)
1041 {
1042
1043 free(mlink->dsec);
1044 free(mlink->arch);
1045 free(mlink->name);
1046 free(mlink->fsec);
1047 free(mlink);
1048 }
1049
1050 static void
mpages_free(void)1051 mpages_free(void)
1052 {
1053 struct mpage *mpage;
1054 struct mlink *mlink;
1055
1056 while ((mpage = mpage_head) != NULL) {
1057 while ((mlink = mpage->mlinks) != NULL) {
1058 mpage->mlinks = mlink->next;
1059 mlink_free(mlink);
1060 }
1061 mpage_head = mpage->next;
1062 free(mpage->sec);
1063 free(mpage->arch);
1064 free(mpage->title);
1065 free(mpage->desc);
1066 free(mpage);
1067 }
1068 }
1069
1070 /*
1071 * For each mlink to the mpage, check whether the path looks like
1072 * it is formatted, and if it does, check whether a source manual
1073 * exists by the same name, ignoring the suffix.
1074 * If both conditions hold, drop the mlink.
1075 */
1076 static void
mlinks_undupe(struct mpage * mpage)1077 mlinks_undupe(struct mpage *mpage)
1078 {
1079 char buf[PATH_MAX];
1080 struct mlink **prev;
1081 struct mlink *mlink;
1082 char *bufp;
1083
1084 mpage->form = FORM_CAT;
1085 prev = &mpage->mlinks;
1086 while (NULL != (mlink = *prev)) {
1087 if (FORM_CAT != mlink->dform) {
1088 mpage->form = FORM_NONE;
1089 goto nextlink;
1090 }
1091 (void)strlcpy(buf, mlink->file, sizeof(buf));
1092 bufp = strstr(buf, "cat");
1093 assert(NULL != bufp);
1094 memcpy(bufp, "man", 3);
1095 if (NULL != (bufp = strrchr(buf, '.')))
1096 *++bufp = '\0';
1097 (void)strlcat(buf, mlink->dsec, sizeof(buf));
1098 if (NULL == ohash_find(&mlinks,
1099 ohash_qlookup(&mlinks, buf)))
1100 goto nextlink;
1101 if (warnings)
1102 say(mlink->file, "Man source exists: %s", buf);
1103 if (use_all)
1104 goto nextlink;
1105 *prev = mlink->next;
1106 mlink_free(mlink);
1107 continue;
1108 nextlink:
1109 prev = &(*prev)->next;
1110 }
1111 }
1112
1113 static void
mlink_check(struct mpage * mpage,struct mlink * mlink)1114 mlink_check(struct mpage *mpage, struct mlink *mlink)
1115 {
1116 struct str *str;
1117 unsigned int slot;
1118
1119 /*
1120 * Check whether the manual section given in a file
1121 * agrees with the directory where the file is located.
1122 * Some manuals have suffixes like (3p) on their
1123 * section number either inside the file or in the
1124 * directory name, some are linked into more than one
1125 * section, like encrypt(1) = makekey(8).
1126 */
1127
1128 if (FORM_SRC == mpage->form &&
1129 strcasecmp(mpage->sec, mlink->dsec))
1130 say(mlink->file, "Section \"%s\" manual in %s directory",
1131 mpage->sec, mlink->dsec);
1132
1133 /*
1134 * Manual page directories exist for each kernel
1135 * architecture as returned by machine(1).
1136 * However, many manuals only depend on the
1137 * application architecture as returned by arch(1).
1138 * For example, some (2/ARM) manuals are shared
1139 * across the "armish" and "zaurus" kernel
1140 * architectures.
1141 * A few manuals are even shared across completely
1142 * different architectures, for example fdformat(1)
1143 * on amd64, i386, and sparc64.
1144 */
1145
1146 if (strcasecmp(mpage->arch, mlink->arch))
1147 say(mlink->file, "Architecture \"%s\" manual in "
1148 "\"%s\" directory", mpage->arch, mlink->arch);
1149
1150 /*
1151 * XXX
1152 * parse_cat() doesn't set NAME_TITLE yet.
1153 */
1154
1155 if (FORM_CAT == mpage->form)
1156 return;
1157
1158 /*
1159 * Check whether this mlink
1160 * appears as a name in the NAME section.
1161 */
1162
1163 slot = ohash_qlookup(&names, mlink->name);
1164 str = ohash_find(&names, slot);
1165 assert(NULL != str);
1166 if ( ! (NAME_TITLE & str->mask))
1167 say(mlink->file, "Name missing in NAME section");
1168 }
1169
1170 /*
1171 * Run through the files in the global vector "mpages"
1172 * and add them to the database specified in "basedir".
1173 *
1174 * This handles the parsing scheme itself, using the cues of directory
1175 * and filename to determine whether the file is parsable or not.
1176 */
1177 static void
mpages_merge(struct dba * dba,struct mparse * mp)1178 mpages_merge(struct dba *dba, struct mparse *mp)
1179 {
1180 struct mpage *mpage, *mpage_dest;
1181 struct mlink *mlink, *mlink_dest;
1182 struct roff_meta *meta;
1183 char *cp;
1184 int fd;
1185
1186 for (mpage = mpage_head; mpage != NULL; mpage = mpage->next) {
1187 mlinks_undupe(mpage);
1188 if ((mlink = mpage->mlinks) == NULL)
1189 continue;
1190
1191 name_mask = NAME_MASK;
1192 mandoc_ohash_init(&names, 4, offsetof(struct str, key));
1193 mandoc_ohash_init(&strings, 6, offsetof(struct str, key));
1194 mparse_reset(mp);
1195 meta = NULL;
1196
1197 if ((fd = mparse_open(mp, mlink->file)) == -1) {
1198 say(mlink->file, "&open");
1199 goto nextpage;
1200 }
1201
1202 /*
1203 * Interpret the file as mdoc(7) or man(7) source
1204 * code, unless it is known to be formatted.
1205 */
1206 if (mlink->dform != FORM_CAT || mlink->fform != FORM_CAT) {
1207 mparse_readfd(mp, fd, mlink->file);
1208 close(fd);
1209 fd = -1;
1210 meta = mparse_result(mp);
1211 }
1212
1213 if (meta != NULL && meta->sodest != NULL) {
1214 mlink_dest = ohash_find(&mlinks,
1215 ohash_qlookup(&mlinks, meta->sodest));
1216 if (mlink_dest == NULL) {
1217 mandoc_asprintf(&cp, "%s.gz", meta->sodest);
1218 mlink_dest = ohash_find(&mlinks,
1219 ohash_qlookup(&mlinks, cp));
1220 free(cp);
1221 }
1222 if (mlink_dest != NULL) {
1223
1224 /* The .so target exists. */
1225
1226 mpage_dest = mlink_dest->mpage;
1227 while (1) {
1228 mlink->mpage = mpage_dest;
1229
1230 /*
1231 * If the target was already
1232 * processed, add the links
1233 * to the database now.
1234 * Otherwise, this will
1235 * happen when we come
1236 * to the target.
1237 */
1238
1239 if (mpage_dest->dba != NULL)
1240 dbadd_mlink(mlink);
1241
1242 if (mlink->next == NULL)
1243 break;
1244 mlink = mlink->next;
1245 }
1246
1247 /* Move all links to the target. */
1248
1249 mlink->next = mlink_dest->next;
1250 mlink_dest->next = mpage->mlinks;
1251 mpage->mlinks = NULL;
1252 goto nextpage;
1253 }
1254 meta->macroset = MACROSET_NONE;
1255 }
1256 if (meta != NULL && meta->macroset == MACROSET_MDOC) {
1257 mpage->form = FORM_SRC;
1258 mpage->sec = meta->msec;
1259 mpage->sec = mandoc_strdup(
1260 mpage->sec == NULL ? "" : mpage->sec);
1261 mpage->arch = meta->arch;
1262 mpage->arch = mandoc_strdup(
1263 mpage->arch == NULL ? "" : mpage->arch);
1264 mpage->title = mandoc_strdup(meta->title);
1265 } else if (meta != NULL && meta->macroset == MACROSET_MAN) {
1266 if (*meta->msec != '\0' || *meta->title != '\0') {
1267 mpage->form = FORM_SRC;
1268 mpage->sec = mandoc_strdup(meta->msec);
1269 mpage->arch = mandoc_strdup(mlink->arch);
1270 mpage->title = mandoc_strdup(meta->title);
1271 } else
1272 meta = NULL;
1273 }
1274
1275 assert(mpage->desc == NULL);
1276 if (meta == NULL || meta->sodest != NULL) {
1277 mpage->sec = mandoc_strdup(mlink->dsec);
1278 mpage->arch = mandoc_strdup(mlink->arch);
1279 mpage->title = mandoc_strdup(mlink->name);
1280 if (meta == NULL) {
1281 mpage->form = FORM_CAT;
1282 parse_cat(mpage, fd);
1283 } else
1284 mpage->form = FORM_SRC;
1285 } else if (meta->macroset == MACROSET_MDOC)
1286 parse_mdoc(mpage, meta, meta->first);
1287 else
1288 parse_man(mpage, meta, meta->first);
1289 if (mpage->desc == NULL) {
1290 mpage->desc = mandoc_strdup(mlink->name);
1291 if (warnings)
1292 say(mlink->file, "No one-line description, "
1293 "using filename \"%s\"", mlink->name);
1294 }
1295
1296 for (mlink = mpage->mlinks;
1297 mlink != NULL;
1298 mlink = mlink->next) {
1299 putkey(mpage, mlink->name, NAME_FILE);
1300 if (warnings && !use_all)
1301 mlink_check(mpage, mlink);
1302 }
1303
1304 dbadd(dba, mpage);
1305
1306 nextpage:
1307 ohash_delete(&strings);
1308 ohash_delete(&names);
1309 }
1310 }
1311
1312 static void
parse_cat(struct mpage * mpage,int fd)1313 parse_cat(struct mpage *mpage, int fd)
1314 {
1315 FILE *stream;
1316 struct mlink *mlink;
1317 char *line, *p, *title, *sec;
1318 size_t linesz, plen, titlesz;
1319 ssize_t len;
1320 int offs;
1321
1322 mlink = mpage->mlinks;
1323 stream = fd == -1 ? fopen(mlink->file, "r") : fdopen(fd, "r");
1324 if (stream == NULL) {
1325 if (fd != -1)
1326 close(fd);
1327 if (warnings)
1328 say(mlink->file, "&fopen");
1329 return;
1330 }
1331
1332 line = NULL;
1333 linesz = 0;
1334
1335 /* Parse the section number from the header line. */
1336
1337 while (getline(&line, &linesz, stream) != -1) {
1338 if (*line == '\n')
1339 continue;
1340 if ((sec = strchr(line, '(')) == NULL)
1341 break;
1342 if ((p = strchr(++sec, ')')) == NULL)
1343 break;
1344 free(mpage->sec);
1345 mpage->sec = mandoc_strndup(sec, p - sec);
1346 if (warnings && *mlink->dsec != '\0' &&
1347 strcasecmp(mpage->sec, mlink->dsec))
1348 say(mlink->file,
1349 "Section \"%s\" manual in %s directory",
1350 mpage->sec, mlink->dsec);
1351 break;
1352 }
1353
1354 /* Skip to first blank line. */
1355
1356 while (line == NULL || *line != '\n')
1357 if (getline(&line, &linesz, stream) == -1)
1358 break;
1359
1360 /*
1361 * Assume the first line that is not indented
1362 * is the first section header. Skip to it.
1363 */
1364
1365 while (getline(&line, &linesz, stream) != -1)
1366 if (*line != '\n' && *line != ' ')
1367 break;
1368
1369 /*
1370 * Read up until the next section into a buffer.
1371 * Strip the leading and trailing newline from each read line,
1372 * appending a trailing space.
1373 * Ignore empty (whitespace-only) lines.
1374 */
1375
1376 titlesz = 0;
1377 title = NULL;
1378
1379 while ((len = getline(&line, &linesz, stream)) != -1) {
1380 if (*line != ' ')
1381 break;
1382 offs = 0;
1383 while (isspace((unsigned char)line[offs]))
1384 offs++;
1385 if (line[offs] == '\0')
1386 continue;
1387 title = mandoc_realloc(title, titlesz + len - offs);
1388 memcpy(title + titlesz, line + offs, len - offs);
1389 titlesz += len - offs;
1390 title[titlesz - 1] = ' ';
1391 }
1392 free(line);
1393
1394 /*
1395 * If no page content can be found, or the input line
1396 * is already the next section header, or there is no
1397 * trailing newline, reuse the page title as the page
1398 * description.
1399 */
1400
1401 if (NULL == title || '\0' == *title) {
1402 if (warnings)
1403 say(mlink->file, "Cannot find NAME section");
1404 fclose(stream);
1405 free(title);
1406 return;
1407 }
1408
1409 title[titlesz - 1] = '\0';
1410
1411 /*
1412 * Skip to the first dash.
1413 * Use the remaining line as the description (no more than 70
1414 * bytes).
1415 */
1416
1417 if (NULL != (p = strstr(title, "- "))) {
1418 for (p += 2; ' ' == *p || '\b' == *p; p++)
1419 /* Skip to next word. */ ;
1420 } else {
1421 if (warnings)
1422 say(mlink->file, "No dash in title line, "
1423 "reusing \"%s\" as one-line description", title);
1424 p = title;
1425 }
1426
1427 plen = strlen(p);
1428
1429 /* Strip backspace-encoding from line. */
1430
1431 while (NULL != (line = memchr(p, '\b', plen))) {
1432 len = line - p;
1433 if (0 == len) {
1434 memmove(line, line + 1, plen--);
1435 continue;
1436 }
1437 memmove(line - 1, line + 1, plen - len);
1438 plen -= 2;
1439 }
1440
1441 /*
1442 * Cut off excessive one-line descriptions.
1443 * Bad pages are not worth better heuristics.
1444 */
1445
1446 mpage->desc = mandoc_strndup(p, 150);
1447 fclose(stream);
1448 free(title);
1449 }
1450
1451 /*
1452 * Put a type/word pair into the word database for this particular file.
1453 */
1454 static void
putkey(const struct mpage * mpage,char * value,uint64_t type)1455 putkey(const struct mpage *mpage, char *value, uint64_t type)
1456 {
1457 putkeys(mpage, value, strlen(value), type);
1458 }
1459
1460 /*
1461 * Grok all nodes at or below a certain mdoc node into putkey().
1462 */
1463 static void
putmdockey(const struct mpage * mpage,const struct roff_node * n,uint64_t m,int taboo)1464 putmdockey(const struct mpage *mpage,
1465 const struct roff_node *n, uint64_t m, int taboo)
1466 {
1467
1468 for ( ; NULL != n; n = n->next) {
1469 if (n->flags & taboo)
1470 continue;
1471 if (NULL != n->child)
1472 putmdockey(mpage, n->child, m, taboo);
1473 if (n->type == ROFFT_TEXT)
1474 putkey(mpage, n->string, m);
1475 }
1476 }
1477
1478 static void
parse_man(struct mpage * mpage,const struct roff_meta * meta,const struct roff_node * n)1479 parse_man(struct mpage *mpage, const struct roff_meta *meta,
1480 const struct roff_node *n)
1481 {
1482 const struct roff_node *head, *body;
1483 char *start, *title;
1484 char byte;
1485 size_t sz;
1486
1487 if (n == NULL)
1488 return;
1489
1490 /*
1491 * We're only searching for one thing: the first text child in
1492 * the BODY of a NAME section. Since we don't keep track of
1493 * sections in -man, run some hoops to find out whether we're in
1494 * the correct section or not.
1495 */
1496
1497 if (n->type == ROFFT_BODY && n->tok == MAN_SH) {
1498 body = n;
1499 if ((head = body->parent->head) != NULL &&
1500 (head = head->child) != NULL &&
1501 head->next == NULL &&
1502 head->type == ROFFT_TEXT &&
1503 strcmp(head->string, "NAME") == 0 &&
1504 body->child != NULL) {
1505
1506 /*
1507 * Suck the entire NAME section into memory.
1508 * Yes, we might run away.
1509 * But too many manuals have big, spread-out
1510 * NAME sections over many lines.
1511 */
1512
1513 title = NULL;
1514 deroff(&title, body);
1515 if (NULL == title)
1516 return;
1517
1518 /*
1519 * Go through a special heuristic dance here.
1520 * Conventionally, one or more manual names are
1521 * comma-specified prior to a whitespace, then a
1522 * dash, then a description. Try to puzzle out
1523 * the name parts here.
1524 */
1525
1526 start = title;
1527 for ( ;; ) {
1528 sz = strcspn(start, " ,");
1529 if ('\0' == start[sz])
1530 break;
1531
1532 byte = start[sz];
1533 start[sz] = '\0';
1534
1535 /*
1536 * Assume a stray trailing comma in the
1537 * name list if a name begins with a dash.
1538 */
1539
1540 if ('-' == start[0] ||
1541 ('\\' == start[0] && '-' == start[1]))
1542 break;
1543
1544 putkey(mpage, start, NAME_TITLE);
1545 if ( ! (mpage->name_head_done ||
1546 strcasecmp(start, meta->title))) {
1547 putkey(mpage, start, NAME_HEAD);
1548 mpage->name_head_done = 1;
1549 }
1550
1551 if (' ' == byte) {
1552 start += sz + 1;
1553 break;
1554 }
1555
1556 assert(',' == byte);
1557 start += sz + 1;
1558 while (' ' == *start)
1559 start++;
1560 }
1561
1562 if (start == title) {
1563 putkey(mpage, start, NAME_TITLE);
1564 if ( ! (mpage->name_head_done ||
1565 strcasecmp(start, meta->title))) {
1566 putkey(mpage, start, NAME_HEAD);
1567 mpage->name_head_done = 1;
1568 }
1569 free(title);
1570 return;
1571 }
1572
1573 while (isspace((unsigned char)*start))
1574 start++;
1575
1576 if (0 == strncmp(start, "-", 1))
1577 start += 1;
1578 else if (0 == strncmp(start, "\\-\\-", 4))
1579 start += 4;
1580 else if (0 == strncmp(start, "\\-", 2))
1581 start += 2;
1582 else if (0 == strncmp(start, "\\(en", 4))
1583 start += 4;
1584 else if (0 == strncmp(start, "\\(em", 4))
1585 start += 4;
1586
1587 while (' ' == *start)
1588 start++;
1589
1590 /*
1591 * Cut off excessive one-line descriptions.
1592 * Bad pages are not worth better heuristics.
1593 */
1594
1595 mpage->desc = mandoc_strndup(start, 150);
1596 free(title);
1597 return;
1598 }
1599 }
1600
1601 for (n = n->child; n; n = n->next) {
1602 if (NULL != mpage->desc)
1603 break;
1604 parse_man(mpage, meta, n);
1605 }
1606 }
1607
1608 static void
parse_mdoc(struct mpage * mpage,const struct roff_meta * meta,const struct roff_node * n)1609 parse_mdoc(struct mpage *mpage, const struct roff_meta *meta,
1610 const struct roff_node *n)
1611 {
1612 const struct mdoc_handler *handler;
1613
1614 for (n = n->child; n != NULL; n = n->next) {
1615 if (n->tok == TOKEN_NONE || n->tok < ROFF_MAX)
1616 continue;
1617 assert(n->tok >= MDOC_Dd && n->tok < MDOC_MAX);
1618 handler = mdoc_handlers + (n->tok - MDOC_Dd);
1619 if (n->flags & handler->taboo)
1620 continue;
1621
1622 switch (n->type) {
1623 case ROFFT_ELEM:
1624 case ROFFT_BLOCK:
1625 case ROFFT_HEAD:
1626 case ROFFT_BODY:
1627 case ROFFT_TAIL:
1628 if (handler->fp != NULL &&
1629 (*handler->fp)(mpage, meta, n) == 0)
1630 break;
1631 if (handler->mask)
1632 putmdockey(mpage, n->child,
1633 handler->mask, handler->taboo);
1634 break;
1635 default:
1636 continue;
1637 }
1638 if (NULL != n->child)
1639 parse_mdoc(mpage, meta, n);
1640 }
1641 }
1642
1643 static int
parse_mdoc_Fa(struct mpage * mpage,const struct roff_meta * meta,const struct roff_node * n)1644 parse_mdoc_Fa(struct mpage *mpage, const struct roff_meta *meta,
1645 const struct roff_node *n)
1646 {
1647 uint64_t mask;
1648
1649 mask = TYPE_Fa;
1650 if (n->sec == SEC_SYNOPSIS)
1651 mask |= TYPE_Vt;
1652
1653 putmdockey(mpage, n->child, mask, 0);
1654 return 0;
1655 }
1656
1657 static int
parse_mdoc_Fd(struct mpage * mpage,const struct roff_meta * meta,const struct roff_node * n)1658 parse_mdoc_Fd(struct mpage *mpage, const struct roff_meta *meta,
1659 const struct roff_node *n)
1660 {
1661 char *start, *end;
1662 size_t sz;
1663
1664 if (SEC_SYNOPSIS != n->sec ||
1665 NULL == (n = n->child) ||
1666 n->type != ROFFT_TEXT)
1667 return 0;
1668
1669 /*
1670 * Only consider those `Fd' macro fields that begin with an
1671 * "inclusion" token (versus, e.g., #define).
1672 */
1673
1674 if (strcmp("#include", n->string))
1675 return 0;
1676
1677 if ((n = n->next) == NULL || n->type != ROFFT_TEXT)
1678 return 0;
1679
1680 /*
1681 * Strip away the enclosing angle brackets and make sure we're
1682 * not zero-length.
1683 */
1684
1685 start = n->string;
1686 if ('<' == *start || '"' == *start)
1687 start++;
1688
1689 if (0 == (sz = strlen(start)))
1690 return 0;
1691
1692 end = &start[(int)sz - 1];
1693 if ('>' == *end || '"' == *end)
1694 end--;
1695
1696 if (end > start)
1697 putkeys(mpage, start, end - start + 1, TYPE_In);
1698 return 0;
1699 }
1700
1701 static void
parse_mdoc_fname(struct mpage * mpage,const struct roff_node * n)1702 parse_mdoc_fname(struct mpage *mpage, const struct roff_node *n)
1703 {
1704 char *cp;
1705 size_t sz;
1706
1707 if (n->type != ROFFT_TEXT)
1708 return;
1709
1710 /* Skip function pointer punctuation. */
1711
1712 cp = n->string;
1713 while (*cp == '(' || *cp == '*')
1714 cp++;
1715 sz = strcspn(cp, "()");
1716
1717 putkeys(mpage, cp, sz, TYPE_Fn);
1718 if (n->sec == SEC_SYNOPSIS)
1719 putkeys(mpage, cp, sz, NAME_SYN);
1720 }
1721
1722 static int
parse_mdoc_Fn(struct mpage * mpage,const struct roff_meta * meta,const struct roff_node * n)1723 parse_mdoc_Fn(struct mpage *mpage, const struct roff_meta *meta,
1724 const struct roff_node *n)
1725 {
1726 uint64_t mask;
1727
1728 if (n->child == NULL)
1729 return 0;
1730
1731 parse_mdoc_fname(mpage, n->child);
1732
1733 n = n->child->next;
1734 if (n != NULL && n->type == ROFFT_TEXT) {
1735 mask = TYPE_Fa;
1736 if (n->sec == SEC_SYNOPSIS)
1737 mask |= TYPE_Vt;
1738 putmdockey(mpage, n, mask, 0);
1739 }
1740
1741 return 0;
1742 }
1743
1744 static int
parse_mdoc_Fo(struct mpage * mpage,const struct roff_meta * meta,const struct roff_node * n)1745 parse_mdoc_Fo(struct mpage *mpage, const struct roff_meta *meta,
1746 const struct roff_node *n)
1747 {
1748
1749 if (n->type != ROFFT_HEAD)
1750 return 1;
1751
1752 if (n->child != NULL)
1753 parse_mdoc_fname(mpage, n->child);
1754
1755 return 0;
1756 }
1757
1758 static int
parse_mdoc_Lb(struct mpage * mpage,const struct roff_meta * meta,const struct roff_node * n)1759 parse_mdoc_Lb(struct mpage *mpage, const struct roff_meta *meta,
1760 const struct roff_node *n)
1761 {
1762 char *cp;
1763
1764 for (n = n->child; n != NULL; n = n->next) {
1765 if (n->flags & NODE_NOSRC)
1766 continue;
1767 cp = n->string;
1768 if (n->sec == SEC_SYNOPSIS)
1769 mandoc_asprintf(&cp, "lib%s", cp);
1770 putkey(mpage, cp, TYPE_Lb);
1771 if (n->sec == SEC_SYNOPSIS)
1772 free(cp);
1773 }
1774 return 0;
1775 }
1776
1777 static int
parse_mdoc_Va(struct mpage * mpage,const struct roff_meta * meta,const struct roff_node * n)1778 parse_mdoc_Va(struct mpage *mpage, const struct roff_meta *meta,
1779 const struct roff_node *n)
1780 {
1781 char *cp;
1782
1783 if (n->type != ROFFT_ELEM && n->type != ROFFT_BODY)
1784 return 0;
1785
1786 if (n->child != NULL &&
1787 n->child->next == NULL &&
1788 n->child->type == ROFFT_TEXT)
1789 return 1;
1790
1791 cp = NULL;
1792 deroff(&cp, n);
1793 if (cp != NULL) {
1794 putkey(mpage, cp, TYPE_Vt | (n->tok == MDOC_Va ||
1795 n->type == ROFFT_BODY ? TYPE_Va : 0));
1796 free(cp);
1797 }
1798
1799 return 0;
1800 }
1801
1802 static int
parse_mdoc_Xr(struct mpage * mpage,const struct roff_meta * meta,const struct roff_node * n)1803 parse_mdoc_Xr(struct mpage *mpage, const struct roff_meta *meta,
1804 const struct roff_node *n)
1805 {
1806 char *cp;
1807
1808 if (NULL == (n = n->child))
1809 return 0;
1810
1811 if (NULL == n->next) {
1812 putkey(mpage, n->string, TYPE_Xr);
1813 return 0;
1814 }
1815
1816 mandoc_asprintf(&cp, "%s(%s)", n->string, n->next->string);
1817 putkey(mpage, cp, TYPE_Xr);
1818 free(cp);
1819 return 0;
1820 }
1821
1822 static int
parse_mdoc_Nd(struct mpage * mpage,const struct roff_meta * meta,const struct roff_node * n)1823 parse_mdoc_Nd(struct mpage *mpage, const struct roff_meta *meta,
1824 const struct roff_node *n)
1825 {
1826
1827 if (n->type == ROFFT_BODY)
1828 deroff(&mpage->desc, n);
1829 return 0;
1830 }
1831
1832 static int
parse_mdoc_Nm(struct mpage * mpage,const struct roff_meta * meta,const struct roff_node * n)1833 parse_mdoc_Nm(struct mpage *mpage, const struct roff_meta *meta,
1834 const struct roff_node *n)
1835 {
1836
1837 if (SEC_NAME == n->sec)
1838 putmdockey(mpage, n->child, NAME_TITLE, 0);
1839 else if (n->sec == SEC_SYNOPSIS && n->type == ROFFT_HEAD) {
1840 if (n->child == NULL)
1841 putkey(mpage, meta->name, NAME_SYN);
1842 else
1843 putmdockey(mpage, n->child, NAME_SYN, 0);
1844 }
1845 if ( ! (mpage->name_head_done ||
1846 n->child == NULL || n->child->string == NULL ||
1847 strcasecmp(n->child->string, meta->title))) {
1848 putkey(mpage, n->child->string, NAME_HEAD);
1849 mpage->name_head_done = 1;
1850 }
1851 return 0;
1852 }
1853
1854 static int
parse_mdoc_Sh(struct mpage * mpage,const struct roff_meta * meta,const struct roff_node * n)1855 parse_mdoc_Sh(struct mpage *mpage, const struct roff_meta *meta,
1856 const struct roff_node *n)
1857 {
1858
1859 return n->sec == SEC_CUSTOM && n->type == ROFFT_HEAD;
1860 }
1861
1862 static int
parse_mdoc_head(struct mpage * mpage,const struct roff_meta * meta,const struct roff_node * n)1863 parse_mdoc_head(struct mpage *mpage, const struct roff_meta *meta,
1864 const struct roff_node *n)
1865 {
1866
1867 return n->type == ROFFT_HEAD;
1868 }
1869
1870 /*
1871 * Add a string to the hash table for the current manual.
1872 * Each string has a bitmask telling which macros it belongs to.
1873 * When we finish the manual, we'll dump the table.
1874 */
1875 static void
putkeys(const struct mpage * mpage,char * cp,size_t sz,uint64_t v)1876 putkeys(const struct mpage *mpage, char *cp, size_t sz, uint64_t v)
1877 {
1878 struct ohash *htab;
1879 struct str *s;
1880 const char *end;
1881 unsigned int slot;
1882 int i, mustfree;
1883
1884 if (0 == sz)
1885 return;
1886
1887 mustfree = render_string(&cp, &sz);
1888
1889 if (TYPE_Nm & v) {
1890 htab = &names;
1891 v &= name_mask;
1892 if (v & NAME_FIRST)
1893 name_mask &= ~NAME_FIRST;
1894 if (debug > 1)
1895 say(mpage->mlinks->file,
1896 "Adding name %*s, bits=0x%llx", (int)sz, cp,
1897 (unsigned long long)v);
1898 } else {
1899 htab = &strings;
1900 if (debug > 1)
1901 for (i = 0; i < KEY_MAX; i++)
1902 if ((uint64_t)1 << i & v)
1903 say(mpage->mlinks->file,
1904 "Adding key %s=%*s",
1905 mansearch_keynames[i], (int)sz, cp);
1906 }
1907
1908 end = cp + sz;
1909 slot = ohash_qlookupi(htab, cp, &end);
1910 s = ohash_find(htab, slot);
1911
1912 if (NULL != s && mpage == s->mpage) {
1913 s->mask |= v;
1914 return;
1915 } else if (NULL == s) {
1916 s = mandoc_calloc(1, sizeof(struct str) + sz + 1);
1917 memcpy(s->key, cp, sz);
1918 ohash_insert(htab, slot, s);
1919 }
1920 s->mpage = mpage;
1921 s->mask = v;
1922
1923 if (mustfree)
1924 free(cp);
1925 }
1926
1927 /*
1928 * Take a Unicode codepoint and produce its UTF-8 encoding.
1929 * This isn't the best way to do this, but it works.
1930 * The magic numbers are from the UTF-8 packaging.
1931 * Read the UTF-8 spec or the utf8(7) manual page for details.
1932 */
1933 static size_t
utf8(unsigned int cp,char out[5])1934 utf8(unsigned int cp, char out[5])
1935 {
1936 size_t rc;
1937
1938 if (cp <= 0x7f) {
1939 rc = 1;
1940 out[0] = (char)cp;
1941 } else if (cp <= 0x7ff) {
1942 rc = 2;
1943 out[0] = (cp >> 6 & 31) | 192;
1944 out[1] = (cp & 63) | 128;
1945 } else if (cp >= 0xd800 && cp <= 0xdfff) {
1946 rc = 0; /* reject UTF-16 surrogate */
1947 } else if (cp <= 0xffff) {
1948 rc = 3;
1949 out[0] = (cp >> 12 & 15) | 224;
1950 out[1] = (cp >> 6 & 63) | 128;
1951 out[2] = (cp & 63) | 128;
1952 } else if (cp <= 0x10ffff) {
1953 rc = 4;
1954 out[0] = (cp >> 18 & 7) | 240;
1955 out[1] = (cp >> 12 & 63) | 128;
1956 out[2] = (cp >> 6 & 63) | 128;
1957 out[3] = (cp & 63) | 128;
1958 } else
1959 rc = 0;
1960
1961 out[rc] = '\0';
1962 return rc;
1963 }
1964
1965 /*
1966 * If the string contains escape sequences,
1967 * replace it with an allocated rendering and return 1,
1968 * such that the caller can free it after use.
1969 * Otherwise, do nothing and return 0.
1970 */
1971 static int
render_string(char ** public,size_t * psz)1972 render_string(char **public, size_t *psz)
1973 {
1974 const char *src, *scp, *addcp, *seq;
1975 char *dst;
1976 size_t ssz, dsz, addsz;
1977 char utfbuf[7], res[6];
1978 int seqlen, unicode;
1979
1980 res[0] = '\\';
1981 res[1] = '\t';
1982 res[2] = ASCII_NBRSP;
1983 res[3] = ASCII_HYPH;
1984 res[4] = ASCII_BREAK;
1985 res[5] = '\0';
1986
1987 src = scp = *public;
1988 ssz = *psz;
1989 dst = NULL;
1990 dsz = 0;
1991
1992 while (scp < src + *psz) {
1993
1994 /* Leave normal characters unchanged. */
1995
1996 if (strchr(res, *scp) == NULL) {
1997 if (dst != NULL)
1998 dst[dsz++] = *scp;
1999 scp++;
2000 continue;
2001 }
2002
2003 /*
2004 * Found something that requires replacing,
2005 * make sure we have a destination buffer.
2006 */
2007
2008 if (dst == NULL) {
2009 dst = mandoc_malloc(ssz + 1);
2010 dsz = scp - src;
2011 memcpy(dst, src, dsz);
2012 }
2013
2014 /* Handle single-char special characters. */
2015
2016 switch (*scp) {
2017 case '\\':
2018 break;
2019 case '\t':
2020 case ASCII_NBRSP:
2021 dst[dsz++] = ' ';
2022 scp++;
2023 continue;
2024 case ASCII_HYPH:
2025 dst[dsz++] = '-';
2026 /* FALLTHROUGH */
2027 case ASCII_BREAK:
2028 scp++;
2029 continue;
2030 default:
2031 abort();
2032 }
2033
2034 /*
2035 * Found an escape sequence.
2036 * Read past the slash, then parse it.
2037 * Ignore everything except characters.
2038 */
2039
2040 scp++;
2041 switch (mandoc_escape(&scp, &seq, &seqlen)) {
2042 case ESCAPE_UNICODE:
2043 unicode = mchars_num2uc(seq + 1, seqlen - 1);
2044 break;
2045 case ESCAPE_NUMBERED:
2046 unicode = mchars_num2char(seq, seqlen);
2047 break;
2048 case ESCAPE_SPECIAL:
2049 unicode = mchars_spec2cp(seq, seqlen);
2050 break;
2051 default:
2052 unicode = -1;
2053 break;
2054 }
2055 if (unicode <= 0)
2056 continue;
2057
2058 /*
2059 * Render the special character
2060 * as either UTF-8 or ASCII.
2061 */
2062
2063 if (write_utf8) {
2064 addsz = utf8(unicode, utfbuf);
2065 if (addsz == 0)
2066 continue;
2067 addcp = utfbuf;
2068 } else {
2069 addcp = mchars_uc2str(unicode);
2070 if (addcp == NULL)
2071 continue;
2072 if (*addcp == ASCII_NBRSP)
2073 addcp = " ";
2074 addsz = strlen(addcp);
2075 }
2076
2077 /* Copy the rendered glyph into the stream. */
2078
2079 ssz += addsz;
2080 dst = mandoc_realloc(dst, ssz + 1);
2081 memcpy(dst + dsz, addcp, addsz);
2082 dsz += addsz;
2083 }
2084 if (dst != NULL) {
2085 *public = dst;
2086 *psz = dsz;
2087 }
2088
2089 /* Trim trailing whitespace and NUL-terminate. */
2090
2091 while (*psz > 0 && (*public)[*psz - 1] == ' ')
2092 --*psz;
2093 if (dst != NULL) {
2094 (*public)[*psz] = '\0';
2095 return 1;
2096 } else
2097 return 0;
2098 }
2099
2100 static void
dbadd_mlink(const struct mlink * mlink)2101 dbadd_mlink(const struct mlink *mlink)
2102 {
2103 dba_page_alias(mlink->mpage->dba, mlink->name, NAME_FILE);
2104 dba_page_add(mlink->mpage->dba, DBP_SECT, mlink->dsec);
2105 dba_page_add(mlink->mpage->dba, DBP_SECT, mlink->fsec);
2106 dba_page_add(mlink->mpage->dba, DBP_ARCH, mlink->arch);
2107 dba_page_add(mlink->mpage->dba, DBP_FILE, mlink->file);
2108 }
2109
2110 /*
2111 * Flush the current page's terms (and their bits) into the database.
2112 * Also, handle escape sequences at the last possible moment.
2113 */
2114 static void
dbadd(struct dba * dba,struct mpage * mpage)2115 dbadd(struct dba *dba, struct mpage *mpage)
2116 {
2117 struct mlink *mlink;
2118 struct str *key;
2119 char *cp;
2120 uint64_t mask;
2121 size_t i;
2122 unsigned int slot;
2123 int mustfree;
2124
2125 mlink = mpage->mlinks;
2126
2127 if (nodb) {
2128 for (key = ohash_first(&names, &slot); NULL != key;
2129 key = ohash_next(&names, &slot))
2130 free(key);
2131 for (key = ohash_first(&strings, &slot); NULL != key;
2132 key = ohash_next(&strings, &slot))
2133 free(key);
2134 if (0 == debug)
2135 return;
2136 while (NULL != mlink) {
2137 fputs(mlink->name, stdout);
2138 if (NULL == mlink->next ||
2139 strcmp(mlink->dsec, mlink->next->dsec) ||
2140 strcmp(mlink->fsec, mlink->next->fsec) ||
2141 strcmp(mlink->arch, mlink->next->arch)) {
2142 putchar('(');
2143 if ('\0' == *mlink->dsec)
2144 fputs(mlink->fsec, stdout);
2145 else
2146 fputs(mlink->dsec, stdout);
2147 if ('\0' != *mlink->arch)
2148 printf("/%s", mlink->arch);
2149 putchar(')');
2150 }
2151 mlink = mlink->next;
2152 if (NULL != mlink)
2153 fputs(", ", stdout);
2154 }
2155 printf(" - %s\n", mpage->desc);
2156 return;
2157 }
2158
2159 if (debug)
2160 say(mlink->file, "Adding to database");
2161
2162 cp = mpage->desc;
2163 i = strlen(cp);
2164 mustfree = render_string(&cp, &i);
2165 mpage->dba = dba_page_new(dba->pages,
2166 *mpage->arch == '\0' ? mlink->arch : mpage->arch,
2167 cp, mlink->file, mpage->form);
2168 if (mustfree)
2169 free(cp);
2170 dba_page_add(mpage->dba, DBP_SECT, mpage->sec);
2171
2172 while (mlink != NULL) {
2173 dbadd_mlink(mlink);
2174 mlink = mlink->next;
2175 }
2176
2177 for (key = ohash_first(&names, &slot); NULL != key;
2178 key = ohash_next(&names, &slot)) {
2179 assert(key->mpage == mpage);
2180 dba_page_alias(mpage->dba, key->key, key->mask);
2181 free(key);
2182 }
2183 for (key = ohash_first(&strings, &slot); NULL != key;
2184 key = ohash_next(&strings, &slot)) {
2185 assert(key->mpage == mpage);
2186 i = 0;
2187 for (mask = TYPE_Xr; mask <= TYPE_Lb; mask *= 2) {
2188 if (key->mask & mask)
2189 dba_macro_add(dba->macros, i,
2190 key->key, mpage->dba);
2191 i++;
2192 }
2193 free(key);
2194 }
2195 }
2196
2197 static void
dbprune(struct dba * dba)2198 dbprune(struct dba *dba)
2199 {
2200 struct dba_array *page, *files;
2201 char *file;
2202
2203 dba_array_FOREACH(dba->pages, page) {
2204 files = dba_array_get(page, DBP_FILE);
2205 dba_array_FOREACH(files, file) {
2206 if (*file < ' ')
2207 file++;
2208 if (ohash_find(&mlinks, ohash_qlookup(&mlinks,
2209 file)) != NULL) {
2210 if (debug)
2211 say(file, "Deleting from database");
2212 dba_array_del(dba->pages);
2213 break;
2214 }
2215 }
2216 }
2217 }
2218
2219 /*
2220 * Write the database from memory to disk.
2221 */
2222 static void
dbwrite(struct dba * dba)2223 dbwrite(struct dba *dba)
2224 {
2225 struct stat sb1, sb2;
2226 char tfn[33], *cp1, *cp2;
2227 off_t i;
2228 int fd1, fd2;
2229
2230 /*
2231 * Do not write empty databases, and delete existing ones
2232 * when makewhatis -u causes them to become empty.
2233 */
2234
2235 dba_array_start(dba->pages);
2236 if (dba_array_next(dba->pages) == NULL) {
2237 if (unlink(MANDOC_DB) == -1 && errno != ENOENT)
2238 say(MANDOC_DB, "&unlink");
2239 return;
2240 }
2241
2242 /*
2243 * Build the database in a temporary file,
2244 * then atomically move it into place.
2245 */
2246
2247 if (dba_write(MANDOC_DB "~", dba) != -1) {
2248 if (rename(MANDOC_DB "~", MANDOC_DB) == -1) {
2249 exitcode = (int)MANDOCLEVEL_SYSERR;
2250 say(MANDOC_DB, "&rename");
2251 unlink(MANDOC_DB "~");
2252 }
2253 return;
2254 }
2255
2256 /*
2257 * We lack write permission and cannot replace the database
2258 * file, but let's at least check whether the data changed.
2259 */
2260
2261 (void)strlcpy(tfn, "/tmp/mandocdb.XXXXXXXX", sizeof(tfn));
2262 if (mkdtemp(tfn) == NULL) {
2263 exitcode = (int)MANDOCLEVEL_SYSERR;
2264 say("", "&%s", tfn);
2265 return;
2266 }
2267 cp1 = cp2 = MAP_FAILED;
2268 fd1 = fd2 = -1;
2269 (void)strlcat(tfn, "/" MANDOC_DB, sizeof(tfn));
2270 if (dba_write(tfn, dba) == -1) {
2271 say(tfn, "&dba_write");
2272 goto err;
2273 }
2274 if ((fd1 = open(MANDOC_DB, O_RDONLY)) == -1) {
2275 say(MANDOC_DB, "&open");
2276 goto err;
2277 }
2278 if ((fd2 = open(tfn, O_RDONLY)) == -1) {
2279 say(tfn, "&open");
2280 goto err;
2281 }
2282 if (fstat(fd1, &sb1) == -1) {
2283 say(MANDOC_DB, "&fstat");
2284 goto err;
2285 }
2286 if (fstat(fd2, &sb2) == -1) {
2287 say(tfn, "&fstat");
2288 goto err;
2289 }
2290 if (sb1.st_size != sb2.st_size)
2291 goto err;
2292 if ((cp1 = mmap(NULL, sb1.st_size, PROT_READ, MAP_PRIVATE,
2293 fd1, 0)) == MAP_FAILED) {
2294 say(MANDOC_DB, "&mmap");
2295 goto err;
2296 }
2297 if ((cp2 = mmap(NULL, sb2.st_size, PROT_READ, MAP_PRIVATE,
2298 fd2, 0)) == MAP_FAILED) {
2299 say(tfn, "&mmap");
2300 goto err;
2301 }
2302 for (i = 0; i < sb1.st_size; i++)
2303 if (cp1[i] != cp2[i])
2304 goto err;
2305 goto out;
2306
2307 err:
2308 exitcode = (int)MANDOCLEVEL_SYSERR;
2309 say(MANDOC_DB, "Data changed, but cannot replace database");
2310
2311 out:
2312 if (cp1 != MAP_FAILED)
2313 munmap(cp1, sb1.st_size);
2314 if (cp2 != MAP_FAILED)
2315 munmap(cp2, sb2.st_size);
2316 if (fd1 != -1)
2317 close(fd1);
2318 if (fd2 != -1)
2319 close(fd2);
2320 unlink(tfn);
2321 *strrchr(tfn, '/') = '\0';
2322 rmdir(tfn);
2323 }
2324
2325 static int
set_basedir(const char * targetdir,int report_baddir)2326 set_basedir(const char *targetdir, int report_baddir)
2327 {
2328 static char startdir[PATH_MAX];
2329 static int getcwd_status; /* 1 = ok, 2 = failure */
2330 static int chdir_status; /* 1 = changed directory */
2331
2332 /*
2333 * Remember the original working directory, if possible.
2334 * This will be needed if the second or a later directory
2335 * on the command line is given as a relative path.
2336 * Do not error out if the current directory is not
2337 * searchable: Maybe it won't be needed after all.
2338 */
2339 if (getcwd_status == 0) {
2340 if (getcwd(startdir, sizeof(startdir)) == NULL) {
2341 getcwd_status = 2;
2342 (void)strlcpy(startdir, strerror(errno),
2343 sizeof(startdir));
2344 } else
2345 getcwd_status = 1;
2346 }
2347
2348 /*
2349 * We are leaving the old base directory.
2350 * Do not use it any longer, not even for messages.
2351 */
2352 *basedir = '\0';
2353 basedir_len = 0;
2354
2355 /*
2356 * If and only if the directory was changed earlier and
2357 * the next directory to process is given as a relative path,
2358 * first go back, or bail out if that is impossible.
2359 */
2360 if (chdir_status && *targetdir != '/') {
2361 if (getcwd_status == 2) {
2362 exitcode = (int)MANDOCLEVEL_SYSERR;
2363 say("", "getcwd: %s", startdir);
2364 return 0;
2365 }
2366 if (chdir(startdir) == -1) {
2367 exitcode = (int)MANDOCLEVEL_SYSERR;
2368 say("", "&chdir %s", startdir);
2369 return 0;
2370 }
2371 }
2372
2373 /*
2374 * Always resolve basedir to the canonicalized absolute
2375 * pathname and append a trailing slash, such that
2376 * we can reliably check whether files are inside.
2377 */
2378 if (realpath(targetdir, basedir) == NULL) {
2379 if (report_baddir || errno != ENOENT) {
2380 exitcode = (int)MANDOCLEVEL_BADARG;
2381 say("", "&%s: realpath", targetdir);
2382 }
2383 *basedir = '\0';
2384 return 0;
2385 } else if (chdir(basedir) == -1) {
2386 if (report_baddir || errno != ENOENT) {
2387 exitcode = (int)MANDOCLEVEL_BADARG;
2388 say("", "&chdir");
2389 }
2390 *basedir = '\0';
2391 return 0;
2392 }
2393 chdir_status = 1;
2394 basedir_len = strlen(basedir);
2395 if (basedir[basedir_len - 1] != '/') {
2396 if (basedir_len >= PATH_MAX - 1) {
2397 exitcode = (int)MANDOCLEVEL_SYSERR;
2398 say("", "Filename too long");
2399 *basedir = '\0';
2400 basedir_len = 0;
2401 return 0;
2402 }
2403 basedir[basedir_len++] = '/';
2404 basedir[basedir_len] = '\0';
2405 }
2406 return 1;
2407 }
2408
2409 #ifdef READ_ALLOWED_PATH
2410 static int
read_allowed(const char * candidate)2411 read_allowed(const char *candidate)
2412 {
2413 const char *cp;
2414 size_t len;
2415
2416 for (cp = READ_ALLOWED_PATH;; cp += len) {
2417 while (*cp == ':')
2418 cp++;
2419 if (*cp == '\0')
2420 return 0;
2421 len = strcspn(cp, ":");
2422 if (strncmp(candidate, cp, len) == 0)
2423 return 1;
2424 }
2425 }
2426 #endif
2427
2428 static void
say(const char * file,const char * format,...)2429 say(const char *file, const char *format, ...)
2430 {
2431 va_list ap;
2432 int use_errno;
2433
2434 if (*basedir != '\0')
2435 fprintf(stderr, "%s", basedir);
2436 if (*basedir != '\0' && *file != '\0')
2437 fputc('/', stderr);
2438 if (*file != '\0')
2439 fprintf(stderr, "%s", file);
2440
2441 use_errno = 1;
2442 if (format != NULL) {
2443 switch (*format) {
2444 case '&':
2445 format++;
2446 break;
2447 case '\0':
2448 format = NULL;
2449 break;
2450 default:
2451 use_errno = 0;
2452 break;
2453 }
2454 }
2455 if (format != NULL) {
2456 if (*basedir != '\0' || *file != '\0')
2457 fputs(": ", stderr);
2458 va_start(ap, format);
2459 vfprintf(stderr, format, ap);
2460 va_end(ap);
2461 }
2462 if (use_errno) {
2463 if (*basedir != '\0' || *file != '\0' || format != NULL)
2464 fputs(": ", stderr);
2465 perror(NULL);
2466 } else
2467 fputc('\n', stderr);
2468 }
2469