]> git.cameronkatri.com Git - mandoc.git/blob - mandocdb.c
git.ckatri.com / mandoc.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
history | raw | HEAD
implement the roff(7) \p (break output line) escape sequence
[mandoc.git] / mandocdb.c
1 /* $Id: mandocdb.c,v 1.250 2017/05/17 22:27:12 schwarze Exp $ */
2 /*
3 * Copyright (c) 2011, 2012 Kristaps Dzonsons <kristaps@bsd.lv>
4 * Copyright (c) 2011-2017 Ingo Schwarze <schwarze@openbsd.org>
5 * Copyright (c) 2016 Ed Maste <emaste@freebsd.org>
6 *
7 * Permission to use, copy, modify, and distribute this software for any
8 * purpose with or without fee is hereby granted, provided that the above
9 * copyright notice and this permission notice appear in all copies.
10 *
11 * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHORS DISCLAIM ALL WARRANTIES
12 * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
13 * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHORS BE LIABLE FOR
14 * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
15 * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
16 * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
17 * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
18 */
19 #include "config.h"
20
21 #include <sys/types.h>
22 #include <sys/stat.h>
23 #include <sys/wait.h>
24
25 #include <assert.h>
26 #include <ctype.h>
27 #if HAVE_ERR
28 #include <err.h>
29 #endif
30 #include <errno.h>
31 #include <fcntl.h>
32 #if HAVE_FTS
33 #include <fts.h>
34 #else
35 #include "compat_fts.h"
36 #endif
37 #include <limits.h>
38 #if HAVE_SANDBOX_INIT
39 #include <sandbox.h>
40 #endif
41 #include <stdarg.h>
42 #include <stddef.h>
43 #include <stdio.h>
44 #include <stdint.h>
45 #include <stdlib.h>
46 #include <string.h>
47 #include <unistd.h>
48
49 #include "mandoc_aux.h"
50 #include "mandoc_ohash.h"
51 #include "mandoc.h"
52 #include "roff.h"
53 #include "mdoc.h"
54 #include "man.h"
55 #include "manconf.h"
56 #include "mansearch.h"
57 #include "dba_array.h"
58 #include "dba.h"
59
60 extern const char *const mansearch_keynames[];
61
62 enum op {
63 OP_DEFAULT = 0, /* new dbs from dir list or default config */
64 OP_CONFFILE, /* new databases from custom config file */
65 OP_UPDATE, /* delete/add entries in existing database */
66 OP_DELETE, /* delete entries from existing database */
67 OP_TEST /* change no databases, report potential problems */
68 };
69
70 struct str {
71 const struct mpage *mpage; /* if set, the owning parse */
72 uint64_t mask; /* bitmask in sequence */
73 char key[]; /* rendered text */
74 };
75
76 struct inodev {
77 ino_t st_ino;
78 dev_t st_dev;
79 };
80
81 struct mpage {
82 struct inodev inodev; /* used for hashing routine */
83 struct dba_array *dba;
84 char *sec; /* section from file content */
85 char *arch; /* architecture from file content */
86 char *title; /* title from file content */
87 char *desc; /* description from file content */
88 struct mpage *next; /* singly linked list */
89 struct mlink *mlinks; /* singly linked list */
90 int name_head_done;
91 enum form form; /* format from file content */
92 };
93
94 struct mlink {
95 char file[PATH_MAX]; /* filename rel. to manpath */
96 char *dsec; /* section from directory */
97 char *arch; /* architecture from directory */
98 char *name; /* name from file name (not empty) */
99 char *fsec; /* section from file name suffix */
100 struct mlink *next; /* singly linked list */
101 struct mpage *mpage; /* parent */
102 int gzip; /* filename has a .gz suffix */
103 enum form dform; /* format from directory */
104 enum form fform; /* format from file name suffix */
105 };
106
107 typedef int (*mdoc_fp)(struct mpage *, const struct roff_meta *,
108 const struct roff_node *);
109
110 struct mdoc_handler {
111 mdoc_fp fp; /* optional handler */
112 uint64_t mask; /* set unless handler returns 0 */
113 int taboo; /* node flags that must not be set */
114 };
115
116
117 int mandocdb(int, char *[]);
118
119 static void dbadd(struct dba *, struct mpage *);
120 static void dbadd_mlink(const struct mlink *mlink);
121 static void dbprune(struct dba *);
122 static void dbwrite(struct dba *);
123 static void filescan(const char *);
124 #if HAVE_FTS_COMPARE_CONST
125 static int fts_compare(const FTSENT *const *, const FTSENT *const *);
126 #else
127 static int fts_compare(const FTSENT **, const FTSENT **);
128 #endif
129 static void mlink_add(struct mlink *, const struct stat *);
130 static void mlink_check(struct mpage *, struct mlink *);
131 static void mlink_free(struct mlink *);
132 static void mlinks_undupe(struct mpage *);
133 static void mpages_free(void);
134 static void mpages_merge(struct dba *, struct mparse *);
135 static void parse_cat(struct mpage *, int);
136 static void parse_man(struct mpage *, const struct roff_meta *,
137 const struct roff_node *);
138 static void parse_mdoc(struct mpage *, const struct roff_meta *,
139 const struct roff_node *);
140 static int parse_mdoc_head(struct mpage *, const struct roff_meta *,
141 const struct roff_node *);
142 static int parse_mdoc_Fd(struct mpage *, const struct roff_meta *,
143 const struct roff_node *);
144 static void parse_mdoc_fname(struct mpage *, const struct roff_node *);
145 static int parse_mdoc_Fn(struct mpage *, const struct roff_meta *,
146 const struct roff_node *);
147 static int parse_mdoc_Fo(struct mpage *, const struct roff_meta *,
148 const struct roff_node *);
149 static int parse_mdoc_Nd(struct mpage *, const struct roff_meta *,
150 const struct roff_node *);
151 static int parse_mdoc_Nm(struct mpage *, const struct roff_meta *,
152 const struct roff_node *);
153 static int parse_mdoc_Sh(struct mpage *, const struct roff_meta *,
154 const struct roff_node *);
155 static int parse_mdoc_Va(struct mpage *, const struct roff_meta *,
156 const struct roff_node *);
157 static int parse_mdoc_Xr(struct mpage *, const struct roff_meta *,
158 const struct roff_node *);
159 static void putkey(const struct mpage *, char *, uint64_t);
160 static void putkeys(const struct mpage *, char *, size_t, uint64_t);
161 static void putmdockey(const struct mpage *,
162 const struct roff_node *, uint64_t, int);
163 static int render_string(char **, size_t *);
164 static void say(const char *, const char *, ...)
165 __attribute__((__format__ (__printf__, 2, 3)));
166 static int set_basedir(const char *, int);
167 static int treescan(void);
168 static size_t utf8(unsigned int, char [7]);
169
170 static int nodb; /* no database changes */
171 static int mparse_options; /* abort the parse early */
172 static int use_all; /* use all found files */
173 static int debug; /* print what we're doing */
174 static int warnings; /* warn about crap */
175 static int write_utf8; /* write UTF-8 output; else ASCII */
176 static int exitcode; /* to be returned by main */
177 static enum op op; /* operational mode */
178 static char basedir[PATH_MAX]; /* current base directory */
179 static struct mpage *mpage_head; /* list of distinct manual pages */
180 static struct ohash mpages; /* table of distinct manual pages */
181 static struct ohash mlinks; /* table of directory entries */
182 static struct ohash names; /* table of all names */
183 static struct ohash strings; /* table of all strings */
184 static uint64_t name_mask;
185
186 static const struct mdoc_handler __mdocs[MDOC_MAX - MDOC_Dd] = {
187 { NULL, 0, NODE_NOPRT }, /* Dd */
188 { NULL, 0, NODE_NOPRT }, /* Dt */
189 { NULL, 0, NODE_NOPRT }, /* Os */
190 { parse_mdoc_Sh, TYPE_Sh, 0 }, /* Sh */
191 { parse_mdoc_head, TYPE_Ss, 0 }, /* Ss */
192 { NULL, 0, 0 }, /* Pp */
193 { NULL, 0, 0 }, /* D1 */
194 { NULL, 0, 0 }, /* Dl */
195 { NULL, 0, 0 }, /* Bd */
196 { NULL, 0, 0 }, /* Ed */
197 { NULL, 0, 0 }, /* Bl */
198 { NULL, 0, 0 }, /* El */
199 { NULL, 0, 0 }, /* It */
200 { NULL, 0, 0 }, /* Ad */
201 { NULL, TYPE_An, 0 }, /* An */
202 { NULL, 0, 0 }, /* Ap */
203 { NULL, TYPE_Ar, 0 }, /* Ar */
204 { NULL, TYPE_Cd, 0 }, /* Cd */
205 { NULL, TYPE_Cm, 0 }, /* Cm */
206 { NULL, TYPE_Dv, 0 }, /* Dv */
207 { NULL, TYPE_Er, 0 }, /* Er */
208 { NULL, TYPE_Ev, 0 }, /* Ev */
209 { NULL, 0, 0 }, /* Ex */
210 { NULL, TYPE_Fa, 0 }, /* Fa */
211 { parse_mdoc_Fd, 0, 0 }, /* Fd */
212 { NULL, TYPE_Fl, 0 }, /* Fl */
213 { parse_mdoc_Fn, 0, 0 }, /* Fn */
214 { NULL, TYPE_Ft, 0 }, /* Ft */
215 { NULL, TYPE_Ic, 0 }, /* Ic */
216 { NULL, TYPE_In, 0 }, /* In */
217 { NULL, TYPE_Li, 0 }, /* Li */
218 { parse_mdoc_Nd, 0, 0 }, /* Nd */
219 { parse_mdoc_Nm, 0, 0 }, /* Nm */
220 { NULL, 0, 0 }, /* Op */
221 { NULL, 0, 0 }, /* Ot */
222 { NULL, TYPE_Pa, NODE_NOSRC }, /* Pa */
223 { NULL, 0, 0 }, /* Rv */
224 { NULL, TYPE_St, 0 }, /* St */
225 { parse_mdoc_Va, TYPE_Va, 0 }, /* Va */
226 { parse_mdoc_Va, TYPE_Vt, 0 }, /* Vt */
227 { parse_mdoc_Xr, 0, 0 }, /* Xr */
228 { NULL, 0, 0 }, /* %A */
229 { NULL, 0, 0 }, /* %B */
230 { NULL, 0, 0 }, /* %D */
231 { NULL, 0, 0 }, /* %I */
232 { NULL, 0, 0 }, /* %J */
233 { NULL, 0, 0 }, /* %N */
234 { NULL, 0, 0 }, /* %O */
235 { NULL, 0, 0 }, /* %P */
236 { NULL, 0, 0 }, /* %R */
237 { NULL, 0, 0 }, /* %T */
238 { NULL, 0, 0 }, /* %V */
239 { NULL, 0, 0 }, /* Ac */
240 { NULL, 0, 0 }, /* Ao */
241 { NULL, 0, 0 }, /* Aq */
242 { NULL, TYPE_At, 0 }, /* At */
243 { NULL, 0, 0 }, /* Bc */
244 { NULL, 0, 0 }, /* Bf */
245 { NULL, 0, 0 }, /* Bo */
246 { NULL, 0, 0 }, /* Bq */
247 { NULL, TYPE_Bsx, NODE_NOSRC }, /* Bsx */
248 { NULL, TYPE_Bx, NODE_NOSRC }, /* Bx */
249 { NULL, 0, 0 }, /* Db */
250 { NULL, 0, 0 }, /* Dc */
251 { NULL, 0, 0 }, /* Do */
252 { NULL, 0, 0 }, /* Dq */
253 { NULL, 0, 0 }, /* Ec */
254 { NULL, 0, 0 }, /* Ef */
255 { NULL, TYPE_Em, 0 }, /* Em */
256 { NULL, 0, 0 }, /* Eo */
257 { NULL, TYPE_Fx, NODE_NOSRC }, /* Fx */
258 { NULL, TYPE_Ms, 0 }, /* Ms */
259 { NULL, 0, 0 }, /* No */
260 { NULL, 0, 0 }, /* Ns */
261 { NULL, TYPE_Nx, NODE_NOSRC }, /* Nx */
262 { NULL, TYPE_Ox, NODE_NOSRC }, /* Ox */
263 { NULL, 0, 0 }, /* Pc */
264 { NULL, 0, 0 }, /* Pf */
265 { NULL, 0, 0 }, /* Po */
266 { NULL, 0, 0 }, /* Pq */
267 { NULL, 0, 0 }, /* Qc */
268 { NULL, 0, 0 }, /* Ql */
269 { NULL, 0, 0 }, /* Qo */
270 { NULL, 0, 0 }, /* Qq */
271 { NULL, 0, 0 }, /* Re */
272 { NULL, 0, 0 }, /* Rs */
273 { NULL, 0, 0 }, /* Sc */
274 { NULL, 0, 0 }, /* So */
275 { NULL, 0, 0 }, /* Sq */
276 { NULL, 0, 0 }, /* Sm */
277 { NULL, 0, 0 }, /* Sx */
278 { NULL, TYPE_Sy, 0 }, /* Sy */
279 { NULL, TYPE_Tn, 0 }, /* Tn */
280 { NULL, 0, NODE_NOSRC }, /* Ux */
281 { NULL, 0, 0 }, /* Xc */
282 { NULL, 0, 0 }, /* Xo */
283 { parse_mdoc_Fo, 0, 0 }, /* Fo */
284 { NULL, 0, 0 }, /* Fc */
285 { NULL, 0, 0 }, /* Oo */
286 { NULL, 0, 0 }, /* Oc */
287 { NULL, 0, 0 }, /* Bk */
288 { NULL, 0, 0 }, /* Ek */
289 { NULL, 0, 0 }, /* Bt */
290 { NULL, 0, 0 }, /* Hf */
291 { NULL, 0, 0 }, /* Fr */
292 { NULL, 0, 0 }, /* Ud */
293 { NULL, TYPE_Lb, NODE_NOSRC }, /* Lb */
294 { NULL, 0, 0 }, /* Lp */
295 { NULL, TYPE_Lk, 0 }, /* Lk */
296 { NULL, TYPE_Mt, NODE_NOSRC }, /* Mt */
297 { NULL, 0, 0 }, /* Brq */
298 { NULL, 0, 0 }, /* Bro */
299 { NULL, 0, 0 }, /* Brc */
300 { NULL, 0, 0 }, /* %C */
301 { NULL, 0, 0 }, /* Es */
302 { NULL, 0, 0 }, /* En */
303 { NULL, TYPE_Dx, NODE_NOSRC }, /* Dx */
304 { NULL, 0, 0 }, /* %Q */
305 { NULL, 0, 0 }, /* %U */
306 { NULL, 0, 0 }, /* Ta */
307 };
308 static const struct mdoc_handler *const mdocs = __mdocs - MDOC_Dd;
309
310
311 int
312 mandocdb(int argc, char *argv[])
313 {
314 struct manconf conf;
315 struct mparse *mp;
316 struct dba *dba;
317 const char *path_arg, *progname;
318 size_t j, sz;
319 int ch, i;
320
321 #if HAVE_PLEDGE
322 if (pledge("stdio rpath wpath cpath fattr flock proc exec", NULL) == -1) {
323 warn("pledge");
324 return (int)MANDOCLEVEL_SYSERR;
325 }
326 #endif
327
328 #if HAVE_SANDBOX_INIT
329 if (sandbox_init(kSBXProfileNoInternet, SANDBOX_NAMED, NULL) == -1) {
330 warnx("sandbox_init");
331 return (int)MANDOCLEVEL_SYSERR;
332 }
333 #endif
334
335 memset(&conf, 0, sizeof(conf));
336
337 /*
338 * We accept a few different invocations.
339 * The CHECKOP macro makes sure that invocation styles don't
340 * clobber each other.
341 */
342 #define CHECKOP(_op, _ch) do \
343 if (OP_DEFAULT != (_op)) { \
344 warnx("-%c: Conflicting option", (_ch)); \
345 goto usage; \
346 } while (/*CONSTCOND*/0)
347
348 path_arg = NULL;
349 op = OP_DEFAULT;
350
351 while (-1 != (ch = getopt(argc, argv, "aC:Dd:npQT:tu:v")))
352 switch (ch) {
353 case 'a':
354 use_all = 1;
355 break;
356 case 'C':
357 CHECKOP(op, ch);
358 path_arg = optarg;
359 op = OP_CONFFILE;
360 break;
361 case 'D':
362 debug++;
363 break;
364 case 'd':
365 CHECKOP(op, ch);
366 path_arg = optarg;
367 op = OP_UPDATE;
368 break;
369 case 'n':
370 nodb = 1;
371 break;
372 case 'p':
373 warnings = 1;
374 break;
375 case 'Q':
376 mparse_options |= MPARSE_QUICK;
377 break;
378 case 'T':
379 if (strcmp(optarg, "utf8")) {
380 warnx("-T%s: Unsupported output format",
381 optarg);
382 goto usage;
383 }
384 write_utf8 = 1;
385 break;
386 case 't':
387 CHECKOP(op, ch);
388 dup2(STDOUT_FILENO, STDERR_FILENO);
389 op = OP_TEST;
390 nodb = warnings = 1;
391 break;
392 case 'u':
393 CHECKOP(op, ch);
394 path_arg = optarg;
395 op = OP_DELETE;
396 break;
397 case 'v':
398 /* Compatibility with espie@'s makewhatis. */
399 break;
400 default:
401 goto usage;
402 }
403
404 argc -= optind;
405 argv += optind;
406
407 #if HAVE_PLEDGE
408 if (nodb) {
409 if (pledge("stdio rpath", NULL) == -1) {
410 warn("pledge");
411 return (int)MANDOCLEVEL_SYSERR;
412 }
413 }
414 #endif
415
416 if (OP_CONFFILE == op && argc > 0) {
417 warnx("-C: Too many arguments");
418 goto usage;
419 }
420
421 exitcode = (int)MANDOCLEVEL_OK;
422 mchars_alloc();
423 mp = mparse_alloc(mparse_options, MANDOCLEVEL_BADARG, NULL, NULL);
424 mandoc_ohash_init(&mpages, 6, offsetof(struct mpage, inodev));
425 mandoc_ohash_init(&mlinks, 6, offsetof(struct mlink, file));
426
427 if (OP_UPDATE == op || OP_DELETE == op || OP_TEST == op) {
428
429 /*
430 * Most of these deal with a specific directory.
431 * Jump into that directory first.
432 */
433 if (OP_TEST != op && 0 == set_basedir(path_arg, 1))
434 goto out;
435
436 dba = nodb ? dba_new(128) : dba_read(MANDOC_DB);
437 if (dba != NULL) {
438 /*
439 * The existing database is usable. Process
440 * all files specified on the command-line.
441 */
442 #if HAVE_PLEDGE
443 if (!nodb) {
444 if (pledge("stdio rpath wpath cpath fattr flock", NULL) == -1) {
445 warn("pledge");
446 exitcode = (int)MANDOCLEVEL_SYSERR;
447 goto out;
448 }
449 }
450 #endif
451 use_all = 1;
452 for (i = 0; i < argc; i++)
453 filescan(argv[i]);
454 if (nodb == 0)
455 dbprune(dba);
456 } else {
457 /* Database missing or corrupt. */
458 if (op != OP_UPDATE || errno != ENOENT)
459 say(MANDOC_DB, "%s: Automatically recreating"
460 " from scratch", strerror(errno));
461 exitcode = (int)MANDOCLEVEL_OK;
462 op = OP_DEFAULT;
463 if (0 == treescan())
464 goto out;
465 dba = dba_new(128);
466 }
467 if (OP_DELETE != op)
468 mpages_merge(dba, mp);
469 if (nodb == 0)
470 dbwrite(dba);
471 dba_free(dba);
472 } else {
473 /*
474 * If we have arguments, use them as our manpaths.
475 * If we don't, use man.conf(5).
476 */
477 if (argc > 0) {
478 conf.manpath.paths = mandoc_reallocarray(NULL,
479 argc, sizeof(char *));
480 conf.manpath.sz = (size_t)argc;
481 for (i = 0; i < argc; i++)
482 conf.manpath.paths[i] = mandoc_strdup(argv[i]);
483 } else
484 manconf_parse(&conf, path_arg, NULL, NULL);
485
486 if (conf.manpath.sz == 0) {
487 exitcode = (int)MANDOCLEVEL_BADARG;
488 say("", "Empty manpath");
489 }
490
491 /*
492 * First scan the tree rooted at a base directory, then
493 * build a new database and finally move it into place.
494 * Ignore zero-length directories and strip trailing
495 * slashes.
496 */
497 for (j = 0; j < conf.manpath.sz; j++) {
498 sz = strlen(conf.manpath.paths[j]);
499 if (sz && conf.manpath.paths[j][sz - 1] == '/')
500 conf.manpath.paths[j][--sz] = '\0';
501 if (0 == sz)
502 continue;
503
504 if (j) {
505 mandoc_ohash_init(&mpages, 6,
506 offsetof(struct mpage, inodev));
507 mandoc_ohash_init(&mlinks, 6,
508 offsetof(struct mlink, file));
509 }
510
511 if ( ! set_basedir(conf.manpath.paths[j], argc > 0))
512 continue;
513 if (0 == treescan())
514 continue;
515 dba = dba_new(128);
516 mpages_merge(dba, mp);
517 if (nodb == 0)
518 dbwrite(dba);
519 dba_free(dba);
520
521 if (j + 1 < conf.manpath.sz) {
522 mpages_free();
523 ohash_delete(&mpages);
524 ohash_delete(&mlinks);
525 }
526 }
527 }
528 out:
529 manconf_free(&conf);
530 mparse_free(mp);
531 mchars_free();
532 mpages_free();
533 ohash_delete(&mpages);
534 ohash_delete(&mlinks);
535 return exitcode;
536 usage:
537 progname = getprogname();
538 fprintf(stderr, "usage: %s [-aDnpQ] [-C file] [-Tutf8]\n"
539 " %s [-aDnpQ] [-Tutf8] dir ...\n"
540 " %s [-DnpQ] [-Tutf8] -d dir [file ...]\n"
541 " %s [-Dnp] -u dir [file ...]\n"
542 " %s [-Q] -t file ...\n",
543 progname, progname, progname, progname, progname);
544
545 return (int)MANDOCLEVEL_BADARG;
546 }
547
548 /*
549 * To get a singly linked list in alpha order while inserting entries
550 * at the beginning, process directory entries in reverse alpha order.
551 */
552 static int
553 #if HAVE_FTS_COMPARE_CONST
554 fts_compare(const FTSENT *const *a, const FTSENT *const *b)
555 #else
556 fts_compare(const FTSENT **a, const FTSENT **b)
557 #endif
558 {
559 return -strcmp((*a)->fts_name, (*b)->fts_name);
560 }
561
562 /*
563 * Scan a directory tree rooted at "basedir" for manpages.
564 * We use fts(), scanning directory parts along the way for clues to our
565 * section and architecture.
566 *
567 * If use_all has been specified, grok all files.
568 * If not, sanitise paths to the following:
569 *
570 * [./]man*[/<arch>]/<name>.<section>
571 * or
572 * [./]cat<section>[/<arch>]/<name>.0
573 *
574 * TODO: accommodate for multi-language directories.
575 */
576 static int
577 treescan(void)
578 {
579 char buf[PATH_MAX];
580 FTS *f;
581 FTSENT *ff;
582 struct mlink *mlink;
583 int gzip;
584 enum form dform;
585 char *dsec, *arch, *fsec, *cp;
586 const char *path;
587 const char *argv[2];
588
589 argv[0] = ".";
590 argv[1] = NULL;
591
592 f = fts_open((char * const *)argv, FTS_PHYSICAL | FTS_NOCHDIR,
593 fts_compare);
594 if (f == NULL) {
595 exitcode = (int)MANDOCLEVEL_SYSERR;
596 say("", "&fts_open");
597 return 0;
598 }
599
600 dsec = arch = NULL;
601 dform = FORM_NONE;
602
603 while ((ff = fts_read(f)) != NULL) {
604 path = ff->fts_path + 2;
605 switch (ff->fts_info) {
606
607 /*
608 * Symbolic links require various sanity checks,
609 * then get handled just like regular files.
610 */
611 case FTS_SL:
612 if (realpath(path, buf) == NULL) {
613 if (warnings)
614 say(path, "&realpath");
615 continue;
616 }
617 if (strstr(buf, basedir) != buf
618 #ifdef HOMEBREWDIR
619 && strstr(buf, HOMEBREWDIR) != buf
620 #endif
621 ) {
622 if (warnings) say("",
623 "%s: outside base directory", buf);
624 continue;
625 }
626 /* Use logical inode to avoid mpages dupe. */
627 if (stat(path, ff->fts_statp) == -1) {
628 if (warnings)
629 say(path, "&stat");
630 continue;
631 }
632 /* FALLTHROUGH */
633
634 /*
635 * If we're a regular file, add an mlink by using the
636 * stored directory data and handling the filename.
637 */
638 case FTS_F:
639 if ( ! strcmp(path, MANDOC_DB))
640 continue;
641 if ( ! use_all && ff->fts_level < 2) {
642 if (warnings)
643 say(path, "Extraneous file");
644 continue;
645 }
646 gzip = 0;
647 fsec = NULL;
648 while (fsec == NULL) {
649 fsec = strrchr(ff->fts_name, '.');
650 if (fsec == NULL || strcmp(fsec+1, "gz"))
651 break;
652 gzip = 1;
653 *fsec = '\0';
654 fsec = NULL;
655 }
656 if (fsec == NULL) {
657 if ( ! use_all) {
658 if (warnings)
659 say(path,
660 "No filename suffix");
661 continue;
662 }
663 } else if ( ! strcmp(++fsec, "html")) {
664 if (warnings)
665 say(path, "Skip html");
666 continue;
667 } else if ( ! strcmp(fsec, "ps")) {
668 if (warnings)
669 say(path, "Skip ps");
670 continue;
671 } else if ( ! strcmp(fsec, "pdf")) {
672 if (warnings)
673 say(path, "Skip pdf");
674 continue;
675 } else if ( ! use_all &&
676 ((dform == FORM_SRC &&
677 strncmp(fsec, dsec, strlen(dsec))) ||
678 (dform == FORM_CAT && strcmp(fsec, "0")))) {
679 if (warnings)
680 say(path, "Wrong filename suffix");
681 continue;
682 } else
683 fsec[-1] = '\0';
684
685 mlink = mandoc_calloc(1, sizeof(struct mlink));
686 if (strlcpy(mlink->file, path,
687 sizeof(mlink->file)) >=
688 sizeof(mlink->file)) {
689 say(path, "Filename too long");
690 free(mlink);
691 continue;
692 }
693 mlink->dform = dform;
694 mlink->dsec = dsec;
695 mlink->arch = arch;
696 mlink->name = ff->fts_name;
697 mlink->fsec = fsec;
698 mlink->gzip = gzip;
699 mlink_add(mlink, ff->fts_statp);
700 continue;
701
702 case FTS_D:
703 case FTS_DP:
704 break;
705
706 default:
707 if (warnings)
708 say(path, "Not a regular file");
709 continue;
710 }
711
712 switch (ff->fts_level) {
713 case 0:
714 /* Ignore the root directory. */
715 break;
716 case 1:
717 /*
718 * This might contain manX/ or catX/.
719 * Try to infer this from the name.
720 * If we're not in use_all, enforce it.
721 */
722 cp = ff->fts_name;
723 if (ff->fts_info == FTS_DP) {
724 dform = FORM_NONE;
725 dsec = NULL;
726 break;
727 }
728
729 if ( ! strncmp(cp, "man", 3)) {
730 dform = FORM_SRC;
731 dsec = cp + 3;
732 } else if ( ! strncmp(cp, "cat", 3)) {
733 dform = FORM_CAT;
734 dsec = cp + 3;
735 } else {
736 dform = FORM_NONE;
737 dsec = NULL;
738 }
739
740 if (dsec != NULL || use_all)
741 break;
742
743 if (warnings)
744 say(path, "Unknown directory part");
745 fts_set(f, ff, FTS_SKIP);
746 break;
747 case 2:
748 /*
749 * Possibly our architecture.
750 * If we're descending, keep tabs on it.
751 */
752 if (ff->fts_info != FTS_DP && dsec != NULL)
753 arch = ff->fts_name;
754 else
755 arch = NULL;
756 break;
757 default:
758 if (ff->fts_info == FTS_DP || use_all)
759 break;
760 if (warnings)
761 say(path, "Extraneous directory part");
762 fts_set(f, ff, FTS_SKIP);
763 break;
764 }
765 }
766
767 fts_close(f);
768 return 1;
769 }
770
771 /*
772 * Add a file to the mlinks table.
773 * Do not verify that it's a "valid" looking manpage (we'll do that
774 * later).
775 *
776 * Try to infer the manual section, architecture, and page name from the
777 * path, assuming it looks like
778 *
779 * [./]man*[/<arch>]/<name>.<section>
780 * or
781 * [./]cat<section>[/<arch>]/<name>.0
782 *
783 * See treescan() for the fts(3) version of this.
784 */
785 static void
786 filescan(const char *file)
787 {
788 char buf[PATH_MAX];
789 struct stat st;
790 struct mlink *mlink;
791 char *p, *start;
792
793 assert(use_all);
794
795 if (0 == strncmp(file, "./", 2))
796 file += 2;
797
798 /*
799 * We have to do lstat(2) before realpath(3) loses
800 * the information whether this is a symbolic link.
801 * We need to know that because for symbolic links,
802 * we want to use the orginal file name, while for
803 * regular files, we want to use the real path.
804 */
805 if (-1 == lstat(file, &st)) {
806 exitcode = (int)MANDOCLEVEL_BADARG;
807 say(file, "&lstat");
808 return;
809 } else if (0 == ((S_IFREG | S_IFLNK) & st.st_mode)) {
810 exitcode = (int)MANDOCLEVEL_BADARG;
811 say(file, "Not a regular file");
812 return;
813 }
814
815 /*
816 * We have to resolve the file name to the real path
817 * in any case for the base directory check.
818 */
819 if (NULL == realpath(file, buf)) {
820 exitcode = (int)MANDOCLEVEL_BADARG;
821 say(file, "&realpath");
822 return;
823 }
824
825 if (OP_TEST == op)
826 start = buf;
827 else if (strstr(buf, basedir) == buf)
828 start = buf + strlen(basedir);
829 #ifdef HOMEBREWDIR
830 else if (strstr(buf, HOMEBREWDIR) == buf)
831 start = buf;
832 #endif
833 else {
834 exitcode = (int)MANDOCLEVEL_BADARG;
835 say("", "%s: outside base directory", buf);
836 return;
837 }
838
839 /*
840 * Now we are sure the file is inside our tree.
841 * If it is a symbolic link, ignore the real path
842 * and use the original name.
843 * This implies passing stuff like "cat1/../man1/foo.1"
844 * on the command line won't work. So don't do that.
845 * Note the stat(2) can still fail if the link target
846 * doesn't exist.
847 */
848 if (S_IFLNK & st.st_mode) {
849 if (-1 == stat(buf, &st)) {
850 exitcode = (int)MANDOCLEVEL_BADARG;
851 say(file, "&stat");
852 return;
853 }
854 if (strlcpy(buf, file, sizeof(buf)) >= sizeof(buf)) {
855 say(file, "Filename too long");
856 return;
857 }
858 start = buf;
859 if (OP_TEST != op && strstr(buf, basedir) == buf)
860 start += strlen(basedir);
861 }
862
863 mlink = mandoc_calloc(1, sizeof(struct mlink));
864 mlink->dform = FORM_NONE;
865 if (strlcpy(mlink->file, start, sizeof(mlink->file)) >=
866 sizeof(mlink->file)) {
867 say(start, "Filename too long");
868 free(mlink);
869 return;
870 }
871
872 /*
873 * In test mode or when the original name is absolute
874 * but outside our tree, guess the base directory.
875 */
876
877 if (op == OP_TEST || (start == buf && *start == '/')) {
878 if (strncmp(buf, "man/", 4) == 0)
879 start = buf + 4;
880 else if ((start = strstr(buf, "/man/")) != NULL)
881 start += 5;
882 else
883 start = buf;
884 }
885
886 /*
887 * First try to guess our directory structure.
888 * If we find a separator, try to look for man* or cat*.
889 * If we find one of these and what's underneath is a directory,
890 * assume it's an architecture.
891 */
892 if (NULL != (p = strchr(start, '/'))) {
893 *p++ = '\0';
894 if (0 == strncmp(start, "man", 3)) {
895 mlink->dform = FORM_SRC;
896 mlink->dsec = start + 3;
897 } else if (0 == strncmp(start, "cat", 3)) {
898 mlink->dform = FORM_CAT;
899 mlink->dsec = start + 3;
900 }
901
902 start = p;
903 if (NULL != mlink->dsec && NULL != (p = strchr(start, '/'))) {
904 *p++ = '\0';
905 mlink->arch = start;
906 start = p;
907 }
908 }
909
910 /*
911 * Now check the file suffix.
912 * Suffix of `.0' indicates a catpage, `.1-9' is a manpage.
913 */
914 p = strrchr(start, '\0');
915 while (p-- > start && '/' != *p && '.' != *p)
916 /* Loop. */ ;
917
918 if ('.' == *p) {
919 *p++ = '\0';
920 mlink->fsec = p;
921 }
922
923 /*
924 * Now try to parse the name.
925 * Use the filename portion of the path.
926 */
927 mlink->name = start;
928 if (NULL != (p = strrchr(start, '/'))) {
929 mlink->name = p + 1;
930 *p = '\0';
931 }
932 mlink_add(mlink, &st);
933 }
934
935 static void
936 mlink_add(struct mlink *mlink, const struct stat *st)
937 {
938 struct inodev inodev;
939 struct mpage *mpage;
940 unsigned int slot;
941
942 assert(NULL != mlink->file);
943
944 mlink->dsec = mandoc_strdup(mlink->dsec ? mlink->dsec : "");
945 mlink->arch = mandoc_strdup(mlink->arch ? mlink->arch : "");
946 mlink->name = mandoc_strdup(mlink->name ? mlink->name : "");
947 mlink->fsec = mandoc_strdup(mlink->fsec ? mlink->fsec : "");
948
949 if ('0' == *mlink->fsec) {
950 free(mlink->fsec);
951 mlink->fsec = mandoc_strdup(mlink->dsec);
952 mlink->fform = FORM_CAT;
953 } else if ('1' <= *mlink->fsec && '9' >= *mlink->fsec)
954 mlink->fform = FORM_SRC;
955 else
956 mlink->fform = FORM_NONE;
957
958 slot = ohash_qlookup(&mlinks, mlink->file);
959 assert(NULL == ohash_find(&mlinks, slot));
960 ohash_insert(&mlinks, slot, mlink);
961
962 memset(&inodev, 0, sizeof(inodev)); /* Clear padding. */
963 inodev.st_ino = st->st_ino;
964 inodev.st_dev = st->st_dev;
965 slot = ohash_lookup_memory(&mpages, (char *)&inodev,
966 sizeof(struct inodev), inodev.st_ino);
967 mpage = ohash_find(&mpages, slot);
968 if (NULL == mpage) {
969 mpage = mandoc_calloc(1, sizeof(struct mpage));
970 mpage->inodev.st_ino = inodev.st_ino;
971 mpage->inodev.st_dev = inodev.st_dev;
972 mpage->form = FORM_NONE;
973 mpage->next = mpage_head;
974 mpage_head = mpage;
975 ohash_insert(&mpages, slot, mpage);
976 } else
977 mlink->next = mpage->mlinks;
978 mpage->mlinks = mlink;
979 mlink->mpage = mpage;
980 }
981
982 static void
983 mlink_free(struct mlink *mlink)
984 {
985
986 free(mlink->dsec);
987 free(mlink->arch);
988 free(mlink->name);
989 free(mlink->fsec);
990 free(mlink);
991 }
992
993 static void
994 mpages_free(void)
995 {
996 struct mpage *mpage;
997 struct mlink *mlink;
998
999 while ((mpage = mpage_head) != NULL) {
1000 while ((mlink = mpage->mlinks) != NULL) {
1001 mpage->mlinks = mlink->next;
1002 mlink_free(mlink);
1003 }
1004 mpage_head = mpage->next;
1005 free(mpage->sec);
1006 free(mpage->arch);
1007 free(mpage->title);
1008 free(mpage->desc);
1009 free(mpage);
1010 }
1011 }
1012
1013 /*
1014 * For each mlink to the mpage, check whether the path looks like
1015 * it is formatted, and if it does, check whether a source manual
1016 * exists by the same name, ignoring the suffix.
1017 * If both conditions hold, drop the mlink.
1018 */
1019 static void
1020 mlinks_undupe(struct mpage *mpage)
1021 {
1022 char buf[PATH_MAX];
1023 struct mlink **prev;
1024 struct mlink *mlink;
1025 char *bufp;
1026
1027 mpage->form = FORM_CAT;
1028 prev = &mpage->mlinks;
1029 while (NULL != (mlink = *prev)) {
1030 if (FORM_CAT != mlink->dform) {
1031 mpage->form = FORM_NONE;
1032 goto nextlink;
1033 }
1034 (void)strlcpy(buf, mlink->file, sizeof(buf));
1035 bufp = strstr(buf, "cat");
1036 assert(NULL != bufp);
1037 memcpy(bufp, "man", 3);
1038 if (NULL != (bufp = strrchr(buf, '.')))
1039 *++bufp = '\0';
1040 (void)strlcat(buf, mlink->dsec, sizeof(buf));
1041 if (NULL == ohash_find(&mlinks,
1042 ohash_qlookup(&mlinks, buf)))
1043 goto nextlink;
1044 if (warnings)
1045 say(mlink->file, "Man source exists: %s", buf);
1046 if (use_all)
1047 goto nextlink;
1048 *prev = mlink->next;
1049 mlink_free(mlink);
1050 continue;
1051 nextlink:
1052 prev = &(*prev)->next;
1053 }
1054 }
1055
1056 static void
1057 mlink_check(struct mpage *mpage, struct mlink *mlink)
1058 {
1059 struct str *str;
1060 unsigned int slot;
1061
1062 /*
1063 * Check whether the manual section given in a file
1064 * agrees with the directory where the file is located.
1065 * Some manuals have suffixes like (3p) on their
1066 * section number either inside the file or in the
1067 * directory name, some are linked into more than one
1068 * section, like encrypt(1) = makekey(8).
1069 */
1070
1071 if (FORM_SRC == mpage->form &&
1072 strcasecmp(mpage->sec, mlink->dsec))
1073 say(mlink->file, "Section \"%s\" manual in %s directory",
1074 mpage->sec, mlink->dsec);
1075
1076 /*
1077 * Manual page directories exist for each kernel
1078 * architecture as returned by machine(1).
1079 * However, many manuals only depend on the
1080 * application architecture as returned by arch(1).
1081 * For example, some (2/ARM) manuals are shared
1082 * across the "armish" and "zaurus" kernel
1083 * architectures.
1084 * A few manuals are even shared across completely
1085 * different architectures, for example fdformat(1)
1086 * on amd64, i386, and sparc64.
1087 */
1088
1089 if (strcasecmp(mpage->arch, mlink->arch))
1090 say(mlink->file, "Architecture \"%s\" manual in "
1091 "\"%s\" directory", mpage->arch, mlink->arch);
1092
1093 /*
1094 * XXX
1095 * parse_cat() doesn't set NAME_TITLE yet.
1096 */
1097
1098 if (FORM_CAT == mpage->form)
1099 return;
1100
1101 /*
1102 * Check whether this mlink
1103 * appears as a name in the NAME section.
1104 */
1105
1106 slot = ohash_qlookup(&names, mlink->name);
1107 str = ohash_find(&names, slot);
1108 assert(NULL != str);
1109 if ( ! (NAME_TITLE & str->mask))
1110 say(mlink->file, "Name missing in NAME section");
1111 }
1112
1113 /*
1114 * Run through the files in the global vector "mpages"
1115 * and add them to the database specified in "basedir".
1116 *
1117 * This handles the parsing scheme itself, using the cues of directory
1118 * and filename to determine whether the file is parsable or not.
1119 */
1120 static void
1121 mpages_merge(struct dba *dba, struct mparse *mp)
1122 {
1123 struct mpage *mpage, *mpage_dest;
1124 struct mlink *mlink, *mlink_dest;
1125 struct roff_man *man;
1126 char *sodest;
1127 char *cp;
1128 int fd;
1129
1130 for (mpage = mpage_head; mpage != NULL; mpage = mpage->next) {
1131 mlinks_undupe(mpage);
1132 if ((mlink = mpage->mlinks) == NULL)
1133 continue;
1134
1135 name_mask = NAME_MASK;
1136 mandoc_ohash_init(&names, 4, offsetof(struct str, key));
1137 mandoc_ohash_init(&strings, 6, offsetof(struct str, key));
1138 mparse_reset(mp);
1139 man = NULL;
1140 sodest = NULL;
1141
1142 if ((fd = mparse_open(mp, mlink->file)) == -1) {
1143 say(mlink->file, "&open");
1144 goto nextpage;
1145 }
1146
1147 /*
1148 * Interpret the file as mdoc(7) or man(7) source
1149 * code, unless it is known to be formatted.
1150 */
1151 if (mlink->dform != FORM_CAT || mlink->fform != FORM_CAT) {
1152 mparse_readfd(mp, fd, mlink->file);
1153 close(fd);
1154 fd = -1;
1155 mparse_result(mp, &man, &sodest);
1156 }
1157
1158 if (sodest != NULL) {
1159 mlink_dest = ohash_find(&mlinks,
1160 ohash_qlookup(&mlinks, sodest));
1161 if (mlink_dest == NULL) {
1162 mandoc_asprintf(&cp, "%s.gz", sodest);
1163 mlink_dest = ohash_find(&mlinks,
1164 ohash_qlookup(&mlinks, cp));
1165 free(cp);
1166 }
1167 if (mlink_dest != NULL) {
1168
1169 /* The .so target exists. */
1170
1171 mpage_dest = mlink_dest->mpage;
1172 while (1) {
1173 mlink->mpage = mpage_dest;
1174
1175 /*
1176 * If the target was already
1177 * processed, add the links
1178 * to the database now.
1179 * Otherwise, this will
1180 * happen when we come
1181 * to the target.
1182 */
1183
1184 if (mpage_dest->dba != NULL)
1185 dbadd_mlink(mlink);
1186
1187 if (mlink->next == NULL)
1188 break;
1189 mlink = mlink->next;
1190 }
1191
1192 /* Move all links to the target. */
1193
1194 mlink->next = mlink_dest->next;
1195 mlink_dest->next = mpage->mlinks;
1196 mpage->mlinks = NULL;
1197 }
1198 goto nextpage;
1199 } else if (man != NULL && man->macroset == MACROSET_MDOC) {
1200 mdoc_validate(man);
1201 mpage->form = FORM_SRC;
1202 mpage->sec = man->meta.msec;
1203 mpage->sec = mandoc_strdup(
1204 mpage->sec == NULL ? "" : mpage->sec);
1205 mpage->arch = man->meta.arch;
1206 mpage->arch = mandoc_strdup(
1207 mpage->arch == NULL ? "" : mpage->arch);
1208 mpage->title = mandoc_strdup(man->meta.title);
1209 } else if (man != NULL && man->macroset == MACROSET_MAN) {
1210 man_validate(man);
1211 if (*man->meta.msec != '\0' ||
1212 *man->meta.title != '\0') {
1213 mpage->form = FORM_SRC;
1214 mpage->sec = mandoc_strdup(man->meta.msec);
1215 mpage->arch = mandoc_strdup(mlink->arch);
1216 mpage->title = mandoc_strdup(man->meta.title);
1217 } else
1218 man = NULL;
1219 }
1220
1221 assert(mpage->desc == NULL);
1222 if (man == NULL) {
1223 mpage->form = FORM_CAT;
1224 mpage->sec = mandoc_strdup(mlink->dsec);
1225 mpage->arch = mandoc_strdup(mlink->arch);
1226 mpage->title = mandoc_strdup(mlink->name);
1227 parse_cat(mpage, fd);
1228 } else if (man->macroset == MACROSET_MDOC)
1229 parse_mdoc(mpage, &man->meta, man->first);
1230 else
1231 parse_man(mpage, &man->meta, man->first);
1232 if (mpage->desc == NULL) {
1233 mpage->desc = mandoc_strdup(mlink->name);
1234 if (warnings)
1235 say(mlink->file, "No one-line description, "
1236 "using filename \"%s\"", mlink->name);
1237 }
1238
1239 for (mlink = mpage->mlinks;
1240 mlink != NULL;
1241 mlink = mlink->next) {
1242 putkey(mpage, mlink->name, NAME_FILE);
1243 if (warnings && !use_all)
1244 mlink_check(mpage, mlink);
1245 }
1246
1247 dbadd(dba, mpage);
1248
1249 nextpage:
1250 ohash_delete(&strings);
1251 ohash_delete(&names);
1252 }
1253 }
1254
1255 static void
1256 parse_cat(struct mpage *mpage, int fd)
1257 {
1258 FILE *stream;
1259 struct mlink *mlink;
1260 char *line, *p, *title, *sec;
1261 size_t linesz, plen, titlesz;
1262 ssize_t len;
1263 int offs;
1264
1265 mlink = mpage->mlinks;
1266 stream = fd == -1 ? fopen(mlink->file, "r") : fdopen(fd, "r");
1267 if (stream == NULL) {
1268 if (fd != -1)
1269 close(fd);
1270 if (warnings)
1271 say(mlink->file, "&fopen");
1272 return;
1273 }
1274
1275 line = NULL;
1276 linesz = 0;
1277
1278 /* Parse the section number from the header line. */
1279
1280 while (getline(&line, &linesz, stream) != -1) {
1281 if (*line == '\n')
1282 continue;
1283 if ((sec = strchr(line, '(')) == NULL)
1284 break;
1285 if ((p = strchr(++sec, ')')) == NULL)
1286 break;
1287 free(mpage->sec);
1288 mpage->sec = mandoc_strndup(sec, p - sec);
1289 if (warnings && *mlink->dsec != '\0' &&
1290 strcasecmp(mpage->sec, mlink->dsec))
1291 say(mlink->file,
1292 "Section \"%s\" manual in %s directory",
1293 mpage->sec, mlink->dsec);
1294 break;
1295 }
1296
1297 /* Skip to first blank line. */
1298
1299 while (line == NULL || *line != '\n')
1300 if (getline(&line, &linesz, stream) == -1)
1301 break;
1302
1303 /*
1304 * Assume the first line that is not indented
1305 * is the first section header. Skip to it.
1306 */
1307
1308 while (getline(&line, &linesz, stream) != -1)
1309 if (*line != '\n' && *line != ' ')
1310 break;
1311
1312 /*
1313 * Read up until the next section into a buffer.
1314 * Strip the leading and trailing newline from each read line,
1315 * appending a trailing space.
1316 * Ignore empty (whitespace-only) lines.
1317 */
1318
1319 titlesz = 0;
1320 title = NULL;
1321
1322 while ((len = getline(&line, &linesz, stream)) != -1) {
1323 if (*line != ' ')
1324 break;
1325 offs = 0;
1326 while (isspace((unsigned char)line[offs]))
1327 offs++;
1328 if (line[offs] == '\0')
1329 continue;
1330 title = mandoc_realloc(title, titlesz + len - offs);
1331 memcpy(title + titlesz, line + offs, len - offs);
1332 titlesz += len - offs;
1333 title[titlesz - 1] = ' ';
1334 }
1335 free(line);
1336
1337 /*
1338 * If no page content can be found, or the input line
1339 * is already the next section header, or there is no
1340 * trailing newline, reuse the page title as the page
1341 * description.
1342 */
1343
1344 if (NULL == title || '\0' == *title) {
1345 if (warnings)
1346 say(mlink->file, "Cannot find NAME section");
1347 fclose(stream);
1348 free(title);
1349 return;
1350 }
1351
1352 title[titlesz - 1] = '\0';
1353
1354 /*
1355 * Skip to the first dash.
1356 * Use the remaining line as the description (no more than 70
1357 * bytes).
1358 */
1359
1360 if (NULL != (p = strstr(title, "- "))) {
1361 for (p += 2; ' ' == *p || '\b' == *p; p++)
1362 /* Skip to next word. */ ;
1363 } else {
1364 if (warnings)
1365 say(mlink->file, "No dash in title line, "
1366 "reusing \"%s\" as one-line description", title);
1367 p = title;
1368 }
1369
1370 plen = strlen(p);
1371
1372 /* Strip backspace-encoding from line. */
1373
1374 while (NULL != (line = memchr(p, '\b', plen))) {
1375 len = line - p;
1376 if (0 == len) {
1377 memmove(line, line + 1, plen--);
1378 continue;
1379 }
1380 memmove(line - 1, line + 1, plen - len);
1381 plen -= 2;
1382 }
1383
1384 mpage->desc = mandoc_strdup(p);
1385 fclose(stream);
1386 free(title);
1387 }
1388
1389 /*
1390 * Put a type/word pair into the word database for this particular file.
1391 */
1392 static void
1393 putkey(const struct mpage *mpage, char *value, uint64_t type)
1394 {
1395 putkeys(mpage, value, strlen(value), type);
1396 }
1397
1398 /*
1399 * Grok all nodes at or below a certain mdoc node into putkey().
1400 */
1401 static void
1402 putmdockey(const struct mpage *mpage,
1403 const struct roff_node *n, uint64_t m, int taboo)
1404 {
1405
1406 for ( ; NULL != n; n = n->next) {
1407 if (n->flags & taboo)
1408 continue;
1409 if (NULL != n->child)
1410 putmdockey(mpage, n->child, m, taboo);
1411 if (n->type == ROFFT_TEXT)
1412 putkey(mpage, n->string, m);
1413 }
1414 }
1415
1416 static void
1417 parse_man(struct mpage *mpage, const struct roff_meta *meta,
1418 const struct roff_node *n)
1419 {
1420 const struct roff_node *head, *body;
1421 char *start, *title;
1422 char byte;
1423 size_t sz;
1424
1425 if (n == NULL)
1426 return;
1427
1428 /*
1429 * We're only searching for one thing: the first text child in
1430 * the BODY of a NAME section. Since we don't keep track of
1431 * sections in -man, run some hoops to find out whether we're in
1432 * the correct section or not.
1433 */
1434
1435 if (n->type == ROFFT_BODY && n->tok == MAN_SH) {
1436 body = n;
1437 if ((head = body->parent->head) != NULL &&
1438 (head = head->child) != NULL &&
1439 head->next == NULL &&
1440 head->type == ROFFT_TEXT &&
1441 strcmp(head->string, "NAME") == 0 &&
1442 body->child != NULL) {
1443
1444 /*
1445 * Suck the entire NAME section into memory.
1446 * Yes, we might run away.
1447 * But too many manuals have big, spread-out
1448 * NAME sections over many lines.
1449 */
1450
1451 title = NULL;
1452 deroff(&title, body);
1453 if (NULL == title)
1454 return;
1455
1456 /*
1457 * Go through a special heuristic dance here.
1458 * Conventionally, one or more manual names are
1459 * comma-specified prior to a whitespace, then a
1460 * dash, then a description. Try to puzzle out
1461 * the name parts here.
1462 */
1463
1464 start = title;
1465 for ( ;; ) {
1466 sz = strcspn(start, " ,");
1467 if ('\0' == start[sz])
1468 break;
1469
1470 byte = start[sz];
1471 start[sz] = '\0';
1472
1473 /*
1474 * Assume a stray trailing comma in the
1475 * name list if a name begins with a dash.
1476 */
1477
1478 if ('-' == start[0] ||
1479 ('\\' == start[0] && '-' == start[1]))
1480 break;
1481
1482 putkey(mpage, start, NAME_TITLE);
1483 if ( ! (mpage->name_head_done ||
1484 strcasecmp(start, meta->title))) {
1485 putkey(mpage, start, NAME_HEAD);
1486 mpage->name_head_done = 1;
1487 }
1488
1489 if (' ' == byte) {
1490 start += sz + 1;
1491 break;
1492 }
1493
1494 assert(',' == byte);
1495 start += sz + 1;
1496 while (' ' == *start)
1497 start++;
1498 }
1499
1500 if (start == title) {
1501 putkey(mpage, start, NAME_TITLE);
1502 if ( ! (mpage->name_head_done ||
1503 strcasecmp(start, meta->title))) {
1504 putkey(mpage, start, NAME_HEAD);
1505 mpage->name_head_done = 1;
1506 }
1507 free(title);
1508 return;
1509 }
1510
1511 while (isspace((unsigned char)*start))
1512 start++;
1513
1514 if (0 == strncmp(start, "-", 1))
1515 start += 1;
1516 else if (0 == strncmp(start, "\\-\\-", 4))
1517 start += 4;
1518 else if (0 == strncmp(start, "\\-", 2))
1519 start += 2;
1520 else if (0 == strncmp(start, "\\(en", 4))
1521 start += 4;
1522 else if (0 == strncmp(start, "\\(em", 4))
1523 start += 4;
1524
1525 while (' ' == *start)
1526 start++;
1527
1528 mpage->desc = mandoc_strdup(start);
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
1542 parse_mdoc(struct mpage *mpage, const struct roff_meta *meta,
1543 const struct roff_node *n)
1544 {
1545
1546 for (n = n->child; n != NULL; n = n->next) {
1547 if (n->tok == TOKEN_NONE ||
1548 n->tok < ROFF_MAX ||
1549 n->flags & mdocs[n->tok].taboo)
1550 continue;
1551 assert(n->tok >= MDOC_Dd && n->tok < MDOC_MAX);
1552 switch (n->type) {
1553 case ROFFT_ELEM:
1554 case ROFFT_BLOCK:
1555 case ROFFT_HEAD:
1556 case ROFFT_BODY:
1557 case ROFFT_TAIL:
1558 if (mdocs[n->tok].fp != NULL &&
1559 (*mdocs[n->tok].fp)(mpage, meta, n) == 0)
1560 break;
1561 if (mdocs[n->tok].mask)
1562 putmdockey(mpage, n->child,
1563 mdocs[n->tok].mask, mdocs[n->tok].taboo);
1564 break;
1565 default:
1566 continue;
1567 }
1568 if (NULL != n->child)
1569 parse_mdoc(mpage, meta, n);
1570 }
1571 }
1572
1573 static int
1574 parse_mdoc_Fd(struct mpage *mpage, const struct roff_meta *meta,
1575 const struct roff_node *n)
1576 {
1577 char *start, *end;
1578 size_t sz;
1579
1580 if (SEC_SYNOPSIS != n->sec ||
1581 NULL == (n = n->child) ||
1582 n->type != ROFFT_TEXT)
1583 return 0;
1584
1585 /*
1586 * Only consider those `Fd' macro fields that begin with an
1587 * "inclusion" token (versus, e.g., #define).
1588 */
1589
1590 if (strcmp("#include", n->string))
1591 return 0;
1592
1593 if ((n = n->next) == NULL || n->type != ROFFT_TEXT)
1594 return 0;
1595
1596 /*
1597 * Strip away the enclosing angle brackets and make sure we're
1598 * not zero-length.
1599 */
1600
1601 start = n->string;
1602 if ('<' == *start || '"' == *start)
1603 start++;
1604
1605 if (0 == (sz = strlen(start)))
1606 return 0;
1607
1608 end = &start[(int)sz - 1];
1609 if ('>' == *end || '"' == *end)
1610 end--;
1611
1612 if (end > start)
1613 putkeys(mpage, start, end - start + 1, TYPE_In);
1614 return 0;
1615 }
1616
1617 static void
1618 parse_mdoc_fname(struct mpage *mpage, const struct roff_node *n)
1619 {
1620 char *cp;
1621 size_t sz;
1622
1623 if (n->type != ROFFT_TEXT)
1624 return;
1625
1626 /* Skip function pointer punctuation. */
1627
1628 cp = n->string;
1629 while (*cp == '(' || *cp == '*')
1630 cp++;
1631 sz = strcspn(cp, "()");
1632
1633 putkeys(mpage, cp, sz, TYPE_Fn);
1634 if (n->sec == SEC_SYNOPSIS)
1635 putkeys(mpage, cp, sz, NAME_SYN);
1636 }
1637
1638 static int
1639 parse_mdoc_Fn(struct mpage *mpage, const struct roff_meta *meta,
1640 const struct roff_node *n)
1641 {
1642
1643 if (n->child == NULL)
1644 return 0;
1645
1646 parse_mdoc_fname(mpage, n->child);
1647
1648 for (n = n->child->next; n != NULL; n = n->next)
1649 if (n->type == ROFFT_TEXT)
1650 putkey(mpage, n->string, TYPE_Fa);
1651
1652 return 0;
1653 }
1654
1655 static int
1656 parse_mdoc_Fo(struct mpage *mpage, const struct roff_meta *meta,
1657 const struct roff_node *n)
1658 {
1659
1660 if (n->type != ROFFT_HEAD)
1661 return 1;
1662
1663 if (n->child != NULL)
1664 parse_mdoc_fname(mpage, n->child);
1665
1666 return 0;
1667 }
1668
1669 static int
1670 parse_mdoc_Va(struct mpage *mpage, const struct roff_meta *meta,
1671 const struct roff_node *n)
1672 {
1673 char *cp;
1674
1675 if (n->type != ROFFT_ELEM && n->type != ROFFT_BODY)
1676 return 0;
1677
1678 if (n->child != NULL &&
1679 n->child->next == NULL &&
1680 n->child->type == ROFFT_TEXT)
1681 return 1;
1682
1683 cp = NULL;
1684 deroff(&cp, n);
1685 if (cp != NULL) {
1686 putkey(mpage, cp, TYPE_Vt | (n->tok == MDOC_Va ||
1687 n->type == ROFFT_BODY ? TYPE_Va : 0));
1688 free(cp);
1689 }
1690
1691 return 0;
1692 }
1693
1694 static int
1695 parse_mdoc_Xr(struct mpage *mpage, const struct roff_meta *meta,
1696 const struct roff_node *n)
1697 {
1698 char *cp;
1699
1700 if (NULL == (n = n->child))
1701 return 0;
1702
1703 if (NULL == n->next) {
1704 putkey(mpage, n->string, TYPE_Xr);
1705 return 0;
1706 }
1707
1708 mandoc_asprintf(&cp, "%s(%s)", n->string, n->next->string);
1709 putkey(mpage, cp, TYPE_Xr);
1710 free(cp);
1711 return 0;
1712 }
1713
1714 static int
1715 parse_mdoc_Nd(struct mpage *mpage, const struct roff_meta *meta,
1716 const struct roff_node *n)
1717 {
1718
1719 if (n->type == ROFFT_BODY)
1720 deroff(&mpage->desc, n);
1721 return 0;
1722 }
1723
1724 static int
1725 parse_mdoc_Nm(struct mpage *mpage, const struct roff_meta *meta,
1726 const struct roff_node *n)
1727 {
1728
1729 if (SEC_NAME == n->sec)
1730 putmdockey(mpage, n->child, NAME_TITLE, 0);
1731 else if (n->sec == SEC_SYNOPSIS && n->type == ROFFT_HEAD) {
1732 if (n->child == NULL)
1733 putkey(mpage, meta->name, NAME_SYN);
1734 else
1735 putmdockey(mpage, n->child, NAME_SYN, 0);
1736 }
1737 if ( ! (mpage->name_head_done ||
1738 n->child == NULL || n->child->string == NULL ||
1739 strcasecmp(n->child->string, meta->title))) {
1740 putkey(mpage, n->child->string, NAME_HEAD);
1741 mpage->name_head_done = 1;
1742 }
1743 return 0;
1744 }
1745
1746 static int
1747 parse_mdoc_Sh(struct mpage *mpage, const struct roff_meta *meta,
1748 const struct roff_node *n)
1749 {
1750
1751 return n->sec == SEC_CUSTOM && n->type == ROFFT_HEAD;
1752 }
1753
1754 static int
1755 parse_mdoc_head(struct mpage *mpage, const struct roff_meta *meta,
1756 const struct roff_node *n)
1757 {
1758
1759 return n->type == ROFFT_HEAD;
1760 }
1761
1762 /*
1763 * Add a string to the hash table for the current manual.
1764 * Each string has a bitmask telling which macros it belongs to.
1765 * When we finish the manual, we'll dump the table.
1766 */
1767 static void
1768 putkeys(const struct mpage *mpage, char *cp, size_t sz, uint64_t v)
1769 {
1770 struct ohash *htab;
1771 struct str *s;
1772 const char *end;
1773 unsigned int slot;
1774 int i, mustfree;
1775
1776 if (0 == sz)
1777 return;
1778
1779 mustfree = render_string(&cp, &sz);
1780
1781 if (TYPE_Nm & v) {
1782 htab = &names;
1783 v &= name_mask;
1784 if (v & NAME_FIRST)
1785 name_mask &= ~NAME_FIRST;
1786 if (debug > 1)
1787 say(mpage->mlinks->file,
1788 "Adding name %*s, bits=0x%llx", (int)sz, cp,
1789 (unsigned long long)v);
1790 } else {
1791 htab = &strings;
1792 if (debug > 1)
1793 for (i = 0; i < KEY_MAX; i++)
1794 if ((uint64_t)1 << i & v)
1795 say(mpage->mlinks->file,
1796 "Adding key %s=%*s",
1797 mansearch_keynames[i], (int)sz, cp);
1798 }
1799
1800 end = cp + sz;
1801 slot = ohash_qlookupi(htab, cp, &end);
1802 s = ohash_find(htab, slot);
1803
1804 if (NULL != s && mpage == s->mpage) {
1805 s->mask |= v;
1806 return;
1807 } else if (NULL == s) {
1808 s = mandoc_calloc(1, sizeof(struct str) + sz + 1);
1809 memcpy(s->key, cp, sz);
1810 ohash_insert(htab, slot, s);
1811 }
1812 s->mpage = mpage;
1813 s->mask = v;
1814
1815 if (mustfree)
1816 free(cp);
1817 }
1818
1819 /*
1820 * Take a Unicode codepoint and produce its UTF-8 encoding.
1821 * This isn't the best way to do this, but it works.
1822 * The magic numbers are from the UTF-8 packaging.
1823 * They're not as scary as they seem: read the UTF-8 spec for details.
1824 */
1825 static size_t
1826 utf8(unsigned int cp, char out[7])
1827 {
1828 size_t rc;
1829
1830 rc = 0;
1831 if (cp <= 0x0000007F) {
1832 rc = 1;
1833 out[0] = (char)cp;
1834 } else if (cp <= 0x000007FF) {
1835 rc = 2;
1836 out[0] = (cp >> 6 & 31) | 192;
1837 out[1] = (cp & 63) | 128;
1838 } else if (cp <= 0x0000FFFF) {
1839 rc = 3;
1840 out[0] = (cp >> 12 & 15) | 224;
1841 out[1] = (cp >> 6 & 63) | 128;
1842 out[2] = (cp & 63) | 128;
1843 } else if (cp <= 0x001FFFFF) {
1844 rc = 4;
1845 out[0] = (cp >> 18 & 7) | 240;
1846 out[1] = (cp >> 12 & 63) | 128;
1847 out[2] = (cp >> 6 & 63) | 128;
1848 out[3] = (cp & 63) | 128;
1849 } else if (cp <= 0x03FFFFFF) {
1850 rc = 5;
1851 out[0] = (cp >> 24 & 3) | 248;
1852 out[1] = (cp >> 18 & 63) | 128;
1853 out[2] = (cp >> 12 & 63) | 128;
1854 out[3] = (cp >> 6 & 63) | 128;
1855 out[4] = (cp & 63) | 128;
1856 } else if (cp <= 0x7FFFFFFF) {
1857 rc = 6;
1858 out[0] = (cp >> 30 & 1) | 252;
1859 out[1] = (cp >> 24 & 63) | 128;
1860 out[2] = (cp >> 18 & 63) | 128;
1861 out[3] = (cp >> 12 & 63) | 128;
1862 out[4] = (cp >> 6 & 63) | 128;
1863 out[5] = (cp & 63) | 128;
1864 } else
1865 return 0;
1866
1867 out[rc] = '\0';
1868 return rc;
1869 }
1870
1871 /*
1872 * If the string contains escape sequences,
1873 * replace it with an allocated rendering and return 1,
1874 * such that the caller can free it after use.
1875 * Otherwise, do nothing and return 0.
1876 */
1877 static int
1878 render_string(char **public, size_t *psz)
1879 {
1880 const char *src, *scp, *addcp, *seq;
1881 char *dst;
1882 size_t ssz, dsz, addsz;
1883 char utfbuf[7], res[6];
1884 int seqlen, unicode;
1885
1886 res[0] = '\\';
1887 res[1] = '\t';
1888 res[2] = ASCII_NBRSP;
1889 res[3] = ASCII_HYPH;
1890 res[4] = ASCII_BREAK;
1891 res[5] = '\0';
1892
1893 src = scp = *public;
1894 ssz = *psz;
1895 dst = NULL;
1896 dsz = 0;
1897
1898 while (scp < src + *psz) {
1899
1900 /* Leave normal characters unchanged. */
1901
1902 if (strchr(res, *scp) == NULL) {
1903 if (dst != NULL)
1904 dst[dsz++] = *scp;
1905 scp++;
1906 continue;
1907 }
1908
1909 /*
1910 * Found something that requires replacing,
1911 * make sure we have a destination buffer.
1912 */
1913
1914 if (dst == NULL) {
1915 dst = mandoc_malloc(ssz + 1);
1916 dsz = scp - src;
1917 memcpy(dst, src, dsz);
1918 }
1919
1920 /* Handle single-char special characters. */
1921
1922 switch (*scp) {
1923 case '\\':
1924 break;
1925 case '\t':
1926 case ASCII_NBRSP:
1927 dst[dsz++] = ' ';
1928 scp++;
1929 continue;
1930 case ASCII_HYPH:
1931 dst[dsz++] = '-';
1932 /* FALLTHROUGH */
1933 case ASCII_BREAK:
1934 scp++;
1935 continue;
1936 default:
1937 abort();
1938 }
1939
1940 /*
1941 * Found an escape sequence.
1942 * Read past the slash, then parse it.
1943 * Ignore everything except characters.
1944 */
1945
1946 scp++;
1947 if (mandoc_escape(&scp, &seq, &seqlen) != ESCAPE_SPECIAL)
1948 continue;
1949
1950 /*
1951 * Render the special character
1952 * as either UTF-8 or ASCII.
1953 */
1954
1955 if (write_utf8) {
1956 unicode = mchars_spec2cp(seq, seqlen);
1957 if (unicode <= 0)
1958 continue;
1959 addsz = utf8(unicode, utfbuf);
1960 if (addsz == 0)
1961 continue;
1962 addcp = utfbuf;
1963 } else {
1964 addcp = mchars_spec2str(seq, seqlen, &addsz);
1965 if (addcp == NULL)
1966 continue;
1967 if (*addcp == ASCII_NBRSP) {
1968 addcp = " ";
1969 addsz = 1;
1970 }
1971 }
1972
1973 /* Copy the rendered glyph into the stream. */
1974
1975 ssz += addsz;
1976 dst = mandoc_realloc(dst, ssz + 1);
1977 memcpy(dst + dsz, addcp, addsz);
1978 dsz += addsz;
1979 }
1980 if (dst != NULL) {
1981 *public = dst;
1982 *psz = dsz;
1983 }
1984
1985 /* Trim trailing whitespace and NUL-terminate. */
1986
1987 while (*psz > 0 && (*public)[*psz - 1] == ' ')
1988 --*psz;
1989 if (dst != NULL) {
1990 (*public)[*psz] = '\0';
1991 return 1;
1992 } else
1993 return 0;
1994 }
1995
1996 static void
1997 dbadd_mlink(const struct mlink *mlink)
1998 {
1999 dba_page_alias(mlink->mpage->dba, mlink->name, NAME_FILE);
2000 dba_page_add(mlink->mpage->dba, DBP_SECT, mlink->dsec);
2001 dba_page_add(mlink->mpage->dba, DBP_SECT, mlink->fsec);
2002 dba_page_add(mlink->mpage->dba, DBP_ARCH, mlink->arch);
2003 dba_page_add(mlink->mpage->dba, DBP_FILE, mlink->file);
2004 }
2005
2006 /*
2007 * Flush the current page's terms (and their bits) into the database.
2008 * Also, handle escape sequences at the last possible moment.
2009 */
2010 static void
2011 dbadd(struct dba *dba, struct mpage *mpage)
2012 {
2013 struct mlink *mlink;
2014 struct str *key;
2015 char *cp;
2016 uint64_t mask;
2017 size_t i;
2018 unsigned int slot;
2019 int mustfree;
2020
2021 mlink = mpage->mlinks;
2022
2023 if (nodb) {
2024 for (key = ohash_first(&names, &slot); NULL != key;
2025 key = ohash_next(&names, &slot))
2026 free(key);
2027 for (key = ohash_first(&strings, &slot); NULL != key;
2028 key = ohash_next(&strings, &slot))
2029 free(key);
2030 if (0 == debug)
2031 return;
2032 while (NULL != mlink) {
2033 fputs(mlink->name, stdout);
2034 if (NULL == mlink->next ||
2035 strcmp(mlink->dsec, mlink->next->dsec) ||
2036 strcmp(mlink->fsec, mlink->next->fsec) ||
2037 strcmp(mlink->arch, mlink->next->arch)) {
2038 putchar('(');
2039 if ('\0' == *mlink->dsec)
2040 fputs(mlink->fsec, stdout);
2041 else
2042 fputs(mlink->dsec, stdout);
2043 if ('\0' != *mlink->arch)
2044 printf("/%s", mlink->arch);
2045 putchar(')');
2046 }
2047 mlink = mlink->next;
2048 if (NULL != mlink)
2049 fputs(", ", stdout);
2050 }
2051 printf(" - %s\n", mpage->desc);
2052 return;
2053 }
2054
2055 if (debug)
2056 say(mlink->file, "Adding to database");
2057
2058 cp = mpage->desc;
2059 i = strlen(cp);
2060 mustfree = render_string(&cp, &i);
2061 mpage->dba = dba_page_new(dba->pages,
2062 *mpage->arch == '\0' ? mlink->arch : mpage->arch,
2063 cp, mlink->file, mpage->form);
2064 if (mustfree)
2065 free(cp);
2066 dba_page_add(mpage->dba, DBP_SECT, mpage->sec);
2067
2068 while (mlink != NULL) {
2069 dbadd_mlink(mlink);
2070 mlink = mlink->next;
2071 }
2072
2073 for (key = ohash_first(&names, &slot); NULL != key;
2074 key = ohash_next(&names, &slot)) {
2075 assert(key->mpage == mpage);
2076 dba_page_alias(mpage->dba, key->key, key->mask);
2077 free(key);
2078 }
2079 for (key = ohash_first(&strings, &slot); NULL != key;
2080 key = ohash_next(&strings, &slot)) {
2081 assert(key->mpage == mpage);
2082 i = 0;
2083 for (mask = TYPE_Xr; mask <= TYPE_Lb; mask *= 2) {
2084 if (key->mask & mask)
2085 dba_macro_add(dba->macros, i,
2086 key->key, mpage->dba);
2087 i++;
2088 }
2089 free(key);
2090 }
2091 }
2092
2093 static void
2094 dbprune(struct dba *dba)
2095 {
2096 struct dba_array *page, *files;
2097 char *file;
2098
2099 dba_array_FOREACH(dba->pages, page) {
2100 files = dba_array_get(page, DBP_FILE);
2101 dba_array_FOREACH(files, file) {
2102 if (*file < ' ')
2103 file++;
2104 if (ohash_find(&mlinks, ohash_qlookup(&mlinks,
2105 file)) != NULL) {
2106 if (debug)
2107 say(file, "Deleting from database");
2108 dba_array_del(dba->pages);
2109 break;
2110 }
2111 }
2112 }
2113 }
2114
2115 /*
2116 * Write the database from memory to disk.
2117 */
2118 static void
2119 dbwrite(struct dba *dba)
2120 {
2121 char tfn[32];
2122 int status;
2123 pid_t child;
2124
2125 /*
2126 * Do not write empty databases, and delete existing ones
2127 * when makewhatis -u causes them to become empty.
2128 */
2129
2130 dba_array_start(dba->pages);
2131 if (dba_array_next(dba->pages) == NULL) {
2132 if (unlink(MANDOC_DB) == -1)
2133 say(MANDOC_DB, "&unlink");
2134 return;
2135 }
2136
2137 /*
2138 * Build the database in a temporary file,
2139 * then atomically move it into place.
2140 */
2141
2142 if (dba_write(MANDOC_DB "~", dba) != -1) {
2143 if (rename(MANDOC_DB "~", MANDOC_DB) == -1) {
2144 exitcode = (int)MANDOCLEVEL_SYSERR;
2145 say(MANDOC_DB, "&rename");
2146 unlink(MANDOC_DB "~");
2147 }
2148 return;
2149 }
2150
2151 /*
2152 * We lack write permission and cannot replace the database
2153 * file, but let's at least check whether the data changed.
2154 */
2155
2156 (void)strlcpy(tfn, "/tmp/mandocdb.XXXXXXXX", sizeof(tfn));
2157 if (mkdtemp(tfn) == NULL) {
2158 exitcode = (int)MANDOCLEVEL_SYSERR;
2159 say("", "&%s", tfn);
2160 return;
2161 }
2162
2163 (void)strlcat(tfn, "/" MANDOC_DB, sizeof(tfn));
2164 if (dba_write(tfn, dba) == -1) {
2165 exitcode = (int)MANDOCLEVEL_SYSERR;
2166 say(tfn, "&dba_write");
2167 goto out;
2168 }
2169
2170 switch (child = fork()) {
2171 case -1:
2172 exitcode = (int)MANDOCLEVEL_SYSERR;
2173 say("", "&fork cmp");
2174 return;
2175 case 0:
2176 execlp("cmp", "cmp", "-s", tfn, MANDOC_DB, (char *)NULL);
2177 say("", "&exec cmp");
2178 exit(0);
2179 default:
2180 break;
2181 }
2182 if (waitpid(child, &status, 0) == -1) {
2183 exitcode = (int)MANDOCLEVEL_SYSERR;
2184 say("", "&wait cmp");
2185 } else if (WIFSIGNALED(status)) {
2186 exitcode = (int)MANDOCLEVEL_SYSERR;
2187 say("", "cmp died from signal %d", WTERMSIG(status));
2188 } else if (WEXITSTATUS(status)) {
2189 exitcode = (int)MANDOCLEVEL_SYSERR;
2190 say(MANDOC_DB,
2191 "Data changed, but cannot replace database");
2192 }
2193
2194 out:
2195 *strrchr(tfn, '/') = '\0';
2196 switch (child = fork()) {
2197 case -1:
2198 exitcode = (int)MANDOCLEVEL_SYSERR;
2199 say("", "&fork rm");
2200 return;
2201 case 0:
2202 execlp("rm", "rm", "-rf", tfn, (char *)NULL);
2203 say("", "&exec rm");
2204 exit((int)MANDOCLEVEL_SYSERR);
2205 default:
2206 break;
2207 }
2208 if (waitpid(child, &status, 0) == -1) {
2209 exitcode = (int)MANDOCLEVEL_SYSERR;
2210 say("", "&wait rm");
2211 } else if (WIFSIGNALED(status) || WEXITSTATUS(status)) {
2212 exitcode = (int)MANDOCLEVEL_SYSERR;
2213 say("", "%s: Cannot remove temporary directory", tfn);
2214 }
2215 }
2216
2217 static int
2218 set_basedir(const char *targetdir, int report_baddir)
2219 {
2220 static char startdir[PATH_MAX];
2221 static int getcwd_status; /* 1 = ok, 2 = failure */
2222 static int chdir_status; /* 1 = changed directory */
2223 char *cp;
2224
2225 /*
2226 * Remember the original working directory, if possible.
2227 * This will be needed if the second or a later directory
2228 * on the command line is given as a relative path.
2229 * Do not error out if the current directory is not
2230 * searchable: Maybe it won't be needed after all.
2231 */
2232 if (0 == getcwd_status) {
2233 if (NULL == getcwd(startdir, sizeof(startdir))) {
2234 getcwd_status = 2;
2235 (void)strlcpy(startdir, strerror(errno),
2236 sizeof(startdir));
2237 } else
2238 getcwd_status = 1;
2239 }
2240
2241 /*
2242 * We are leaving the old base directory.
2243 * Do not use it any longer, not even for messages.
2244 */
2245 *basedir = '\0';
2246
2247 /*
2248 * If and only if the directory was changed earlier and
2249 * the next directory to process is given as a relative path,
2250 * first go back, or bail out if that is impossible.
2251 */
2252 if (chdir_status && '/' != *targetdir) {
2253 if (2 == getcwd_status) {
2254 exitcode = (int)MANDOCLEVEL_SYSERR;
2255 say("", "getcwd: %s", startdir);
2256 return 0;
2257 }
2258 if (-1 == chdir(startdir)) {
2259 exitcode = (int)MANDOCLEVEL_SYSERR;
2260 say("", "&chdir %s", startdir);
2261 return 0;
2262 }
2263 }
2264
2265 /*
2266 * Always resolve basedir to the canonicalized absolute
2267 * pathname and append a trailing slash, such that
2268 * we can reliably check whether files are inside.
2269 */
2270 if (NULL == realpath(targetdir, basedir)) {
2271 if (report_baddir || errno != ENOENT) {
2272 exitcode = (int)MANDOCLEVEL_BADARG;
2273 say("", "&%s: realpath", targetdir);
2274 }
2275 return 0;
2276 } else if (-1 == chdir(basedir)) {
2277 if (report_baddir || errno != ENOENT) {
2278 exitcode = (int)MANDOCLEVEL_BADARG;
2279 say("", "&chdir");
2280 }
2281 return 0;
2282 }
2283 chdir_status = 1;
2284 cp = strchr(basedir, '\0');
2285 if ('/' != cp[-1]) {
2286 if (cp - basedir >= PATH_MAX - 1) {
2287 exitcode = (int)MANDOCLEVEL_SYSERR;
2288 say("", "Filename too long");
2289 return 0;
2290 }
2291 *cp++ = '/';
2292 *cp = '\0';
2293 }
2294 return 1;
2295 }
2296
2297 static void
2298 say(const char *file, const char *format, ...)
2299 {
2300 va_list ap;
2301 int use_errno;
2302
2303 if ('\0' != *basedir)
2304 fprintf(stderr, "%s", basedir);
2305 if ('\0' != *basedir && '\0' != *file)
2306 fputc('/', stderr);
2307 if ('\0' != *file)
2308 fprintf(stderr, "%s", file);
2309
2310 use_errno = 1;
2311 if (NULL != format) {
2312 switch (*format) {
2313 case '&':
2314 format++;
2315 break;
2316 case '\0':
2317 format = NULL;
2318 break;
2319 default:
2320 use_errno = 0;
2321 break;
2322 }
2323 }
2324 if (NULL != format) {
2325 if ('\0' != *basedir || '\0' != *file)
2326 fputs(": ", stderr);
2327 va_start(ap, format);
2328 vfprintf(stderr, format, ap);
2329 va_end(ap);
2330 }
2331 if (use_errno) {
2332 if ('\0' != *basedir || '\0' != *file || NULL != format)
2333 fputs(": ", stderr);
2334 perror(NULL);
2335 } else
2336 fputc('\n', stderr);
2337 }
A git conversion mirror of mandoc.bsd.lv