]> 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
KNF: case (FOO): -> case FOO:, remove /* LINTED */ and /* ARGSUSED */,
[mandoc.git] / mandocdb.c
1 /* $Id: mandocdb.c,v 1.141 2014/04/20 16:46:04 schwarze Exp $ */
2 /*
3 * Copyright (c) 2011, 2012 Kristaps Dzonsons <kristaps@bsd.lv>
4 * Copyright (c) 2011, 2012, 2013, 2014 Ingo Schwarze <schwarze@openbsd.org>
5 *
6 * Permission to use, copy, modify, and distribute this software for any
7 * purpose with or without fee is hereby granted, provided that the above
8 * copyright notice and this permission notice appear in all copies.
9 *
10 * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
11 * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
12 * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
13 * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
14 * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
15 * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
16 * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
17 */
18 #ifdef HAVE_CONFIG_H
19 #include "config.h"
20 #endif
21
22 #include <sys/stat.h>
23 #include <sys/wait.h>
24
25 #include <assert.h>
26 #include <ctype.h>
27 #include <errno.h>
28 #include <fcntl.h>
29 #include <fts.h>
30 #include <getopt.h>
31 #include <limits.h>
32 #include <stddef.h>
33 #include <stdio.h>
34 #include <stdint.h>
35 #include <stdlib.h>
36 #include <string.h>
37 #include <unistd.h>
38
39 #ifdef HAVE_OHASH
40 #include <ohash.h>
41 #else
42 #include "compat_ohash.h"
43 #endif
44 #include <sqlite3.h>
45
46 #include "mdoc.h"
47 #include "man.h"
48 #include "mandoc.h"
49 #include "mandoc_aux.h"
50 #include "manpath.h"
51 #include "mansearch.h"
52
53 extern int mansearch_keymax;
54 extern const char *const mansearch_keynames[];
55
56 #define SQL_EXEC(_v) \
57 if (SQLITE_OK != sqlite3_exec(db, (_v), NULL, NULL, NULL)) \
58 say("", "%s: %s", (_v), sqlite3_errmsg(db))
59 #define SQL_BIND_TEXT(_s, _i, _v) \
60 if (SQLITE_OK != sqlite3_bind_text \
61 ((_s), (_i)++, (_v), -1, SQLITE_STATIC)) \
62 say(mlink->file, "%s", sqlite3_errmsg(db))
63 #define SQL_BIND_INT(_s, _i, _v) \
64 if (SQLITE_OK != sqlite3_bind_int \
65 ((_s), (_i)++, (_v))) \
66 say(mlink->file, "%s", sqlite3_errmsg(db))
67 #define SQL_BIND_INT64(_s, _i, _v) \
68 if (SQLITE_OK != sqlite3_bind_int64 \
69 ((_s), (_i)++, (_v))) \
70 say(mlink->file, "%s", sqlite3_errmsg(db))
71 #define SQL_STEP(_s) \
72 if (SQLITE_DONE != sqlite3_step((_s))) \
73 say(mlink->file, "%s", sqlite3_errmsg(db))
74
75 enum op {
76 OP_DEFAULT = 0, /* new dbs from dir list or default config */
77 OP_CONFFILE, /* new databases from custom config file */
78 OP_UPDATE, /* delete/add entries in existing database */
79 OP_DELETE, /* delete entries from existing database */
80 OP_TEST /* change no databases, report potential problems */
81 };
82
83 enum form {
84 FORM_NONE, /* format is unknown */
85 FORM_SRC, /* format is -man or -mdoc */
86 FORM_CAT /* format is cat */
87 };
88
89 struct str {
90 char *rendered; /* key in UTF-8 or ASCII form */
91 const struct mpage *mpage; /* if set, the owning parse */
92 uint64_t mask; /* bitmask in sequence */
93 char key[]; /* may contain escape sequences */
94 };
95
96 struct inodev {
97 ino_t st_ino;
98 dev_t st_dev;
99 };
100
101 struct mpage {
102 struct inodev inodev; /* used for hashing routine */
103 int64_t pageid; /* pageid in mpages SQL table */
104 enum form form; /* format from file content */
105 char *sec; /* section from file content */
106 char *arch; /* architecture from file content */
107 char *title; /* title from file content */
108 char *desc; /* description from file content */
109 struct mlink *mlinks; /* singly linked list */
110 };
111
112 struct mlink {
113 char file[PATH_MAX]; /* filename rel. to manpath */
114 enum form dform; /* format from directory */
115 enum form fform; /* format from file name suffix */
116 char *dsec; /* section from directory */
117 char *arch; /* architecture from directory */
118 char *name; /* name from file name (not empty) */
119 char *fsec; /* section from file name suffix */
120 struct mlink *next; /* singly linked list */
121 struct mpage *mpage; /* parent */
122 int gzip; /* filename has a .gz suffix */
123 };
124
125 enum stmt {
126 STMT_DELETE_PAGE = 0, /* delete mpage */
127 STMT_INSERT_PAGE, /* insert mpage */
128 STMT_INSERT_LINK, /* insert mlink */
129 STMT_INSERT_NAME, /* insert name */
130 STMT_INSERT_KEY, /* insert parsed key */
131 STMT__MAX
132 };
133
134 typedef int (*mdoc_fp)(struct mpage *, const struct mdoc_node *);
135
136 struct mdoc_handler {
137 mdoc_fp fp; /* optional handler */
138 uint64_t mask; /* set unless handler returns 0 */
139 };
140
141 static void dbclose(int);
142 static void dbadd(struct mpage *, struct mchars *);
143 static void dbadd_mlink(const struct mlink *mlink);
144 static int dbopen(int);
145 static void dbprune(void);
146 static void filescan(const char *);
147 static void *hash_alloc(size_t, void *);
148 static void hash_free(void *, size_t, void *);
149 static void *hash_halloc(size_t, void *);
150 static void mlink_add(struct mlink *, const struct stat *);
151 static void mlink_check(struct mpage *, struct mlink *);
152 static void mlink_free(struct mlink *);
153 static void mlinks_undupe(struct mpage *);
154 static void mpages_free(void);
155 static void mpages_merge(struct mchars *, struct mparse *);
156 static void names_check(void);
157 static void parse_cat(struct mpage *, int);
158 static void parse_man(struct mpage *, const struct man_node *);
159 static void parse_mdoc(struct mpage *, const struct mdoc_node *);
160 static int parse_mdoc_body(struct mpage *, const struct mdoc_node *);
161 static int parse_mdoc_head(struct mpage *, const struct mdoc_node *);
162 static int parse_mdoc_Fd(struct mpage *, const struct mdoc_node *);
163 static int parse_mdoc_Fn(struct mpage *, const struct mdoc_node *);
164 static int parse_mdoc_Nd(struct mpage *, const struct mdoc_node *);
165 static int parse_mdoc_Nm(struct mpage *, const struct mdoc_node *);
166 static int parse_mdoc_Sh(struct mpage *, const struct mdoc_node *);
167 static int parse_mdoc_Xr(struct mpage *, const struct mdoc_node *);
168 static void putkey(const struct mpage *, char *, uint64_t);
169 static void putkeys(const struct mpage *,
170 const char *, size_t, uint64_t);
171 static void putmdockey(const struct mpage *,
172 const struct mdoc_node *, uint64_t);
173 static void render_key(struct mchars *, struct str *);
174 static void say(const char *, const char *, ...);
175 static int set_basedir(const char *);
176 static int treescan(void);
177 static size_t utf8(unsigned int, char [7]);
178
179 static char tempfilename[32];
180 static char *progname;
181 static int nodb; /* no database changes */
182 static int mparse_options; /* abort the parse early */
183 static int use_all; /* use all found files */
184 static int debug; /* print what we're doing */
185 static int warnings; /* warn about crap */
186 static int write_utf8; /* write UTF-8 output; else ASCII */
187 static int exitcode; /* to be returned by main */
188 static enum op op; /* operational mode */
189 static char basedir[PATH_MAX]; /* current base directory */
190 static struct ohash mpages; /* table of distinct manual pages */
191 static struct ohash mlinks; /* table of directory entries */
192 static struct ohash names; /* table of all names */
193 static struct ohash strings; /* table of all strings */
194 static sqlite3 *db = NULL; /* current database */
195 static sqlite3_stmt *stmts[STMT__MAX]; /* current statements */
196 static uint64_t name_mask;
197
198 static const struct mdoc_handler mdocs[MDOC_MAX] = {
199 { NULL, 0 }, /* Ap */
200 { NULL, 0 }, /* Dd */
201 { NULL, 0 }, /* Dt */
202 { NULL, 0 }, /* Os */
203 { parse_mdoc_Sh, TYPE_Sh }, /* Sh */
204 { parse_mdoc_head, TYPE_Ss }, /* Ss */
205 { NULL, 0 }, /* Pp */
206 { NULL, 0 }, /* D1 */
207 { NULL, 0 }, /* Dl */
208 { NULL, 0 }, /* Bd */
209 { NULL, 0 }, /* Ed */
210 { NULL, 0 }, /* Bl */
211 { NULL, 0 }, /* El */
212 { NULL, 0 }, /* It */
213 { NULL, 0 }, /* Ad */
214 { NULL, TYPE_An }, /* An */
215 { NULL, TYPE_Ar }, /* Ar */
216 { NULL, TYPE_Cd }, /* Cd */
217 { NULL, TYPE_Cm }, /* Cm */
218 { NULL, TYPE_Dv }, /* Dv */
219 { NULL, TYPE_Er }, /* Er */
220 { NULL, TYPE_Ev }, /* Ev */
221 { NULL, 0 }, /* Ex */
222 { NULL, TYPE_Fa }, /* Fa */
223 { parse_mdoc_Fd, 0 }, /* Fd */
224 { NULL, TYPE_Fl }, /* Fl */
225 { parse_mdoc_Fn, 0 }, /* Fn */
226 { NULL, TYPE_Ft }, /* Ft */
227 { NULL, TYPE_Ic }, /* Ic */
228 { NULL, TYPE_In }, /* In */
229 { NULL, TYPE_Li }, /* Li */
230 { parse_mdoc_Nd, 0 }, /* Nd */
231 { parse_mdoc_Nm, 0 }, /* Nm */
232 { NULL, 0 }, /* Op */
233 { NULL, 0 }, /* Ot */
234 { NULL, TYPE_Pa }, /* Pa */
235 { NULL, 0 }, /* Rv */
236 { NULL, TYPE_St }, /* St */
237 { NULL, TYPE_Va }, /* Va */
238 { parse_mdoc_body, TYPE_Va }, /* Vt */
239 { parse_mdoc_Xr, 0 }, /* Xr */
240 { NULL, 0 }, /* %A */
241 { NULL, 0 }, /* %B */
242 { NULL, 0 }, /* %D */
243 { NULL, 0 }, /* %I */
244 { NULL, 0 }, /* %J */
245 { NULL, 0 }, /* %N */
246 { NULL, 0 }, /* %O */
247 { NULL, 0 }, /* %P */
248 { NULL, 0 }, /* %R */
249 { NULL, 0 }, /* %T */
250 { NULL, 0 }, /* %V */
251 { NULL, 0 }, /* Ac */
252 { NULL, 0 }, /* Ao */
253 { NULL, 0 }, /* Aq */
254 { NULL, TYPE_At }, /* At */
255 { NULL, 0 }, /* Bc */
256 { NULL, 0 }, /* Bf */
257 { NULL, 0 }, /* Bo */
258 { NULL, 0 }, /* Bq */
259 { NULL, TYPE_Bsx }, /* Bsx */
260 { NULL, TYPE_Bx }, /* Bx */
261 { NULL, 0 }, /* Db */
262 { NULL, 0 }, /* Dc */
263 { NULL, 0 }, /* Do */
264 { NULL, 0 }, /* Dq */
265 { NULL, 0 }, /* Ec */
266 { NULL, 0 }, /* Ef */
267 { NULL, TYPE_Em }, /* Em */
268 { NULL, 0 }, /* Eo */
269 { NULL, TYPE_Fx }, /* Fx */
270 { NULL, TYPE_Ms }, /* Ms */
271 { NULL, 0 }, /* No */
272 { NULL, 0 }, /* Ns */
273 { NULL, TYPE_Nx }, /* Nx */
274 { NULL, TYPE_Ox }, /* Ox */
275 { NULL, 0 }, /* Pc */
276 { NULL, 0 }, /* Pf */
277 { NULL, 0 }, /* Po */
278 { NULL, 0 }, /* Pq */
279 { NULL, 0 }, /* Qc */
280 { NULL, 0 }, /* Ql */
281 { NULL, 0 }, /* Qo */
282 { NULL, 0 }, /* Qq */
283 { NULL, 0 }, /* Re */
284 { NULL, 0 }, /* Rs */
285 { NULL, 0 }, /* Sc */
286 { NULL, 0 }, /* So */
287 { NULL, 0 }, /* Sq */
288 { NULL, 0 }, /* Sm */
289 { NULL, 0 }, /* Sx */
290 { NULL, TYPE_Sy }, /* Sy */
291 { NULL, TYPE_Tn }, /* Tn */
292 { NULL, 0 }, /* Ux */
293 { NULL, 0 }, /* Xc */
294 { NULL, 0 }, /* Xo */
295 { parse_mdoc_head, 0 }, /* Fo */
296 { NULL, 0 }, /* Fc */
297 { NULL, 0 }, /* Oo */
298 { NULL, 0 }, /* Oc */
299 { NULL, 0 }, /* Bk */
300 { NULL, 0 }, /* Ek */
301 { NULL, 0 }, /* Bt */
302 { NULL, 0 }, /* Hf */
303 { NULL, 0 }, /* Fr */
304 { NULL, 0 }, /* Ud */
305 { NULL, TYPE_Lb }, /* Lb */
306 { NULL, 0 }, /* Lp */
307 { NULL, TYPE_Lk }, /* Lk */
308 { NULL, TYPE_Mt }, /* Mt */
309 { NULL, 0 }, /* Brq */
310 { NULL, 0 }, /* Bro */
311 { NULL, 0 }, /* Brc */
312 { NULL, 0 }, /* %C */
313 { NULL, 0 }, /* Es */
314 { NULL, 0 }, /* En */
315 { NULL, TYPE_Dx }, /* Dx */
316 { NULL, 0 }, /* %Q */
317 { NULL, 0 }, /* br */
318 { NULL, 0 }, /* sp */
319 { NULL, 0 }, /* %U */
320 { NULL, 0 }, /* Ta */
321 };
322
323
324 int
325 main(int argc, char *argv[])
326 {
327 int ch, i;
328 size_t j, sz;
329 const char *path_arg;
330 struct mchars *mc;
331 struct manpaths dirs;
332 struct mparse *mp;
333 struct ohash_info mpages_info, mlinks_info;
334
335 memset(stmts, 0, STMT__MAX * sizeof(sqlite3_stmt *));
336 memset(&dirs, 0, sizeof(struct manpaths));
337
338 mpages_info.alloc = mlinks_info.alloc = hash_alloc;
339 mpages_info.halloc = mlinks_info.halloc = hash_halloc;
340 mpages_info.hfree = mlinks_info.hfree = hash_free;
341
342 mpages_info.key_offset = offsetof(struct mpage, inodev);
343 mlinks_info.key_offset = offsetof(struct mlink, file);
344
345 progname = strrchr(argv[0], '/');
346 if (progname == NULL)
347 progname = argv[0];
348 else
349 ++progname;
350
351 /*
352 * We accept a few different invocations.
353 * The CHECKOP macro makes sure that invocation styles don't
354 * clobber each other.
355 */
356 #define CHECKOP(_op, _ch) do \
357 if (OP_DEFAULT != (_op)) { \
358 fprintf(stderr, "-%c: Conflicting option\n", (_ch)); \
359 goto usage; \
360 } while (/*CONSTCOND*/0)
361
362 path_arg = NULL;
363 op = OP_DEFAULT;
364
365 while (-1 != (ch = getopt(argc, argv, "aC:Dd:npQT:tu:v")))
366 switch (ch) {
367 case 'a':
368 use_all = 1;
369 break;
370 case 'C':
371 CHECKOP(op, ch);
372 path_arg = optarg;
373 op = OP_CONFFILE;
374 break;
375 case 'D':
376 debug++;
377 break;
378 case 'd':
379 CHECKOP(op, ch);
380 path_arg = optarg;
381 op = OP_UPDATE;
382 break;
383 case 'n':
384 nodb = 1;
385 break;
386 case 'p':
387 warnings = 1;
388 break;
389 case 'Q':
390 mparse_options |= MPARSE_QUICK;
391 break;
392 case 'T':
393 if (strcmp(optarg, "utf8")) {
394 fprintf(stderr, "-T%s: Unsupported "
395 "output format\n", optarg);
396 goto usage;
397 }
398 write_utf8 = 1;
399 break;
400 case 't':
401 CHECKOP(op, ch);
402 dup2(STDOUT_FILENO, STDERR_FILENO);
403 op = OP_TEST;
404 nodb = warnings = 1;
405 break;
406 case 'u':
407 CHECKOP(op, ch);
408 path_arg = optarg;
409 op = OP_DELETE;
410 break;
411 case 'v':
412 /* Compatibility with espie@'s makewhatis. */
413 break;
414 default:
415 goto usage;
416 }
417
418 argc -= optind;
419 argv += optind;
420
421 if (OP_CONFFILE == op && argc > 0) {
422 fprintf(stderr, "-C: Too many arguments\n");
423 goto usage;
424 }
425
426 exitcode = (int)MANDOCLEVEL_OK;
427 mp = mparse_alloc(mparse_options, MANDOCLEVEL_FATAL, NULL, NULL);
428 mc = mchars_alloc();
429
430 ohash_init(&mpages, 6, &mpages_info);
431 ohash_init(&mlinks, 6, &mlinks_info);
432
433 if (OP_UPDATE == op || OP_DELETE == op || OP_TEST == op) {
434
435 /*
436 * All of these deal with a specific directory.
437 * Jump into that directory first.
438 */
439 if (0 == set_basedir(path_arg))
440 goto out;
441
442 if (dbopen(1)) {
443 /*
444 * The existing database is usable. Process
445 * all files specified on the command-line.
446 */
447 use_all = 1;
448 for (i = 0; i < argc; i++)
449 filescan(argv[i]);
450 if (OP_TEST != op)
451 dbprune();
452 } else {
453 /*
454 * Database missing or corrupt.
455 * Recreate from scratch.
456 */
457 exitcode = (int)MANDOCLEVEL_OK;
458 op = OP_DEFAULT;
459 if (0 == treescan())
460 goto out;
461 if (0 == dbopen(0))
462 goto out;
463 }
464 if (OP_DELETE != op)
465 mpages_merge(mc, mp);
466 dbclose(OP_DEFAULT == op ? 0 : 1);
467 } else {
468 /*
469 * If we have arguments, use them as our manpaths.
470 * If we don't, grok from manpath(1) or however else
471 * manpath_parse() wants to do it.
472 */
473 if (argc > 0) {
474 dirs.paths = mandoc_calloc(argc,
475 sizeof(char *));
476 dirs.sz = (size_t)argc;
477 for (i = 0; i < argc; i++)
478 dirs.paths[i] = mandoc_strdup(argv[i]);
479 } else
480 manpath_parse(&dirs, path_arg, NULL, NULL);
481
482 if (0 == dirs.sz) {
483 exitcode = (int)MANDOCLEVEL_BADARG;
484 say("", "Empty manpath");
485 }
486
487 /*
488 * First scan the tree rooted at a base directory, then
489 * build a new database and finally move it into place.
490 * Ignore zero-length directories and strip trailing
491 * slashes.
492 */
493 for (j = 0; j < dirs.sz; j++) {
494 sz = strlen(dirs.paths[j]);
495 if (sz && '/' == dirs.paths[j][sz - 1])
496 dirs.paths[j][--sz] = '\0';
497 if (0 == sz)
498 continue;
499
500 if (j) {
501 ohash_init(&mpages, 6, &mpages_info);
502 ohash_init(&mlinks, 6, &mlinks_info);
503 }
504
505 if (0 == set_basedir(dirs.paths[j]))
506 goto out;
507 if (0 == treescan())
508 goto out;
509 if (0 == set_basedir(dirs.paths[j]))
510 goto out;
511 if (0 == dbopen(0))
512 goto out;
513
514 mpages_merge(mc, mp);
515 if (warnings &&
516 ! (MPARSE_QUICK & mparse_options))
517 names_check();
518 dbclose(0);
519
520 if (j + 1 < dirs.sz) {
521 mpages_free();
522 ohash_delete(&mpages);
523 ohash_delete(&mlinks);
524 }
525 }
526 }
527 out:
528 set_basedir(NULL);
529 manpath_free(&dirs);
530 mchars_free(mc);
531 mparse_free(mp);
532 mpages_free();
533 ohash_delete(&mpages);
534 ohash_delete(&mlinks);
535 return(exitcode);
536 usage:
537 fprintf(stderr, "usage: %s [-aDnpQ] [-C file] [-Tutf8]\n"
538 " %s [-aDnpQ] [-Tutf8] dir ...\n"
539 " %s [-DnpQ] [-Tutf8] -d dir [file ...]\n"
540 " %s [-Dnp] -u dir [file ...]\n"
541 " %s [-Q] -t file ...\n",
542 progname, progname, progname,
543 progname, progname);
544
545 return((int)MANDOCLEVEL_BADARG);
546 }
547
548 /*
549 * Scan a directory tree rooted at "basedir" for manpages.
550 * We use fts(), scanning directory parts along the way for clues to our
551 * section and architecture.
552 *
553 * If use_all has been specified, grok all files.
554 * If not, sanitise paths to the following:
555 *
556 * [./]man*[/<arch>]/<name>.<section>
557 * or
558 * [./]cat<section>[/<arch>]/<name>.0
559 *
560 * TODO: accomodate for multi-language directories.
561 */
562 static int
563 treescan(void)
564 {
565 char buf[PATH_MAX];
566 FTS *f;
567 FTSENT *ff;
568 struct mlink *mlink;
569 int dform, gzip;
570 char *dsec, *arch, *fsec, *cp;
571 const char *path;
572 const char *argv[2];
573
574 argv[0] = ".";
575 argv[1] = (char *)NULL;
576
577 f = fts_open((char * const *)argv,
578 FTS_PHYSICAL | FTS_NOCHDIR, NULL);
579 if (NULL == f) {
580 exitcode = (int)MANDOCLEVEL_SYSERR;
581 say("", "&fts_open");
582 return(0);
583 }
584
585 dsec = arch = NULL;
586 dform = FORM_NONE;
587
588 while (NULL != (ff = fts_read(f))) {
589 path = ff->fts_path + 2;
590 switch (ff->fts_info) {
591
592 /*
593 * Symbolic links require various sanity checks,
594 * then get handled just like regular files.
595 */
596 case FTS_SL:
597 if (NULL == realpath(path, buf)) {
598 if (warnings)
599 say(path, "&realpath");
600 continue;
601 }
602 if (strstr(buf, basedir) != buf) {
603 if (warnings) say("",
604 "%s: outside base directory", buf);
605 continue;
606 }
607 /* Use logical inode to avoid mpages dupe. */
608 if (-1 == stat(path, ff->fts_statp)) {
609 if (warnings)
610 say(path, "&stat");
611 continue;
612 }
613 /* FALLTHROUGH */
614
615 /*
616 * If we're a regular file, add an mlink by using the
617 * stored directory data and handling the filename.
618 */
619 case FTS_F:
620 if (0 == strcmp(path, MANDOC_DB))
621 continue;
622 if ( ! use_all && ff->fts_level < 2) {
623 if (warnings)
624 say(path, "Extraneous file");
625 continue;
626 }
627 gzip = 0;
628 fsec = NULL;
629 while (NULL == fsec) {
630 fsec = strrchr(ff->fts_name, '.');
631 if (NULL == fsec || strcmp(fsec+1, "gz"))
632 break;
633 gzip = 1;
634 *fsec = '\0';
635 fsec = NULL;
636 }
637 if (NULL == fsec) {
638 if ( ! use_all) {
639 if (warnings)
640 say(path,
641 "No filename suffix");
642 continue;
643 }
644 } else if (0 == strcmp(++fsec, "html")) {
645 if (warnings)
646 say(path, "Skip html");
647 continue;
648 } else if (0 == strcmp(fsec, "ps")) {
649 if (warnings)
650 say(path, "Skip ps");
651 continue;
652 } else if (0 == strcmp(fsec, "pdf")) {
653 if (warnings)
654 say(path, "Skip pdf");
655 continue;
656 } else if ( ! use_all &&
657 ((FORM_SRC == dform && strcmp(fsec, dsec)) ||
658 (FORM_CAT == dform && strcmp(fsec, "0")))) {
659 if (warnings)
660 say(path, "Wrong filename suffix");
661 continue;
662 } else
663 fsec[-1] = '\0';
664
665 mlink = mandoc_calloc(1, sizeof(struct mlink));
666 strlcpy(mlink->file, path, sizeof(mlink->file));
667 mlink->dform = dform;
668 mlink->dsec = dsec;
669 mlink->arch = arch;
670 mlink->name = ff->fts_name;
671 mlink->fsec = fsec;
672 mlink->gzip = gzip;
673 mlink_add(mlink, ff->fts_statp);
674 continue;
675
676 case FTS_D:
677 /* FALLTHROUGH */
678 case FTS_DP:
679 break;
680
681 default:
682 if (warnings)
683 say(path, "Not a regular file");
684 continue;
685 }
686
687 switch (ff->fts_level) {
688 case 0:
689 /* Ignore the root directory. */
690 break;
691 case 1:
692 /*
693 * This might contain manX/ or catX/.
694 * Try to infer this from the name.
695 * If we're not in use_all, enforce it.
696 */
697 cp = ff->fts_name;
698 if (FTS_DP == ff->fts_info)
699 break;
700
701 if (0 == strncmp(cp, "man", 3)) {
702 dform = FORM_SRC;
703 dsec = cp + 3;
704 } else if (0 == strncmp(cp, "cat", 3)) {
705 dform = FORM_CAT;
706 dsec = cp + 3;
707 } else {
708 dform = FORM_NONE;
709 dsec = NULL;
710 }
711
712 if (NULL != dsec || use_all)
713 break;
714
715 if (warnings)
716 say(path, "Unknown directory part");
717 fts_set(f, ff, FTS_SKIP);
718 break;
719 case 2:
720 /*
721 * Possibly our architecture.
722 * If we're descending, keep tabs on it.
723 */
724 if (FTS_DP != ff->fts_info && NULL != dsec)
725 arch = ff->fts_name;
726 else
727 arch = NULL;
728 break;
729 default:
730 if (FTS_DP == ff->fts_info || use_all)
731 break;
732 if (warnings)
733 say(path, "Extraneous directory part");
734 fts_set(f, ff, FTS_SKIP);
735 break;
736 }
737 }
738
739 fts_close(f);
740 return(1);
741 }
742
743 /*
744 * Add a file to the mlinks table.
745 * Do not verify that it's a "valid" looking manpage (we'll do that
746 * later).
747 *
748 * Try to infer the manual section, architecture, and page name from the
749 * path, assuming it looks like
750 *
751 * [./]man*[/<arch>]/<name>.<section>
752 * or
753 * [./]cat<section>[/<arch>]/<name>.0
754 *
755 * See treescan() for the fts(3) version of this.
756 */
757 static void
758 filescan(const char *file)
759 {
760 char buf[PATH_MAX];
761 struct stat st;
762 struct mlink *mlink;
763 char *p, *start;
764
765 assert(use_all);
766
767 if (0 == strncmp(file, "./", 2))
768 file += 2;
769
770 /*
771 * We have to do lstat(2) before realpath(3) loses
772 * the information whether this is a symbolic link.
773 * We need to know that because for symbolic links,
774 * we want to use the orginal file name, while for
775 * regular files, we want to use the real path.
776 */
777 if (-1 == lstat(file, &st)) {
778 exitcode = (int)MANDOCLEVEL_BADARG;
779 say(file, "&lstat");
780 return;
781 } else if (0 == ((S_IFREG | S_IFLNK) & st.st_mode)) {
782 exitcode = (int)MANDOCLEVEL_BADARG;
783 say(file, "Not a regular file");
784 return;
785 }
786
787 /*
788 * We have to resolve the file name to the real path
789 * in any case for the base directory check.
790 */
791 if (NULL == realpath(file, buf)) {
792 exitcode = (int)MANDOCLEVEL_BADARG;
793 say(file, "&realpath");
794 return;
795 }
796
797 if (strstr(buf, basedir) == buf)
798 start = buf + strlen(basedir) + 1;
799 else if (OP_TEST == op)
800 start = buf;
801 else {
802 exitcode = (int)MANDOCLEVEL_BADARG;
803 say("", "%s: outside base directory", buf);
804 return;
805 }
806
807 /*
808 * Now we are sure the file is inside our tree.
809 * If it is a symbolic link, ignore the real path
810 * and use the original name.
811 * This implies passing stuff like "cat1/../man1/foo.1"
812 * on the command line won't work. So don't do that.
813 * Note the stat(2) can still fail if the link target
814 * doesn't exist.
815 */
816 if (S_IFLNK & st.st_mode) {
817 if (-1 == stat(buf, &st)) {
818 exitcode = (int)MANDOCLEVEL_BADARG;
819 say(file, "&stat");
820 return;
821 }
822 strlcpy(buf, file, sizeof(buf));
823 start = strstr(buf, basedir) == buf ?
824 buf + strlen(basedir) + 1 : buf;
825 }
826
827 mlink = mandoc_calloc(1, sizeof(struct mlink));
828 strlcpy(mlink->file, start, sizeof(mlink->file));
829
830 /*
831 * First try to guess our directory structure.
832 * If we find a separator, try to look for man* or cat*.
833 * If we find one of these and what's underneath is a directory,
834 * assume it's an architecture.
835 */
836 if (NULL != (p = strchr(start, '/'))) {
837 *p++ = '\0';
838 if (0 == strncmp(start, "man", 3)) {
839 mlink->dform = FORM_SRC;
840 mlink->dsec = start + 3;
841 } else if (0 == strncmp(start, "cat", 3)) {
842 mlink->dform = FORM_CAT;
843 mlink->dsec = start + 3;
844 }
845
846 start = p;
847 if (NULL != mlink->dsec && NULL != (p = strchr(start, '/'))) {
848 *p++ = '\0';
849 mlink->arch = start;
850 start = p;
851 }
852 }
853
854 /*
855 * Now check the file suffix.
856 * Suffix of `.0' indicates a catpage, `.1-9' is a manpage.
857 */
858 p = strrchr(start, '\0');
859 while (p-- > start && '/' != *p && '.' != *p)
860 /* Loop. */ ;
861
862 if ('.' == *p) {
863 *p++ = '\0';
864 mlink->fsec = p;
865 }
866
867 /*
868 * Now try to parse the name.
869 * Use the filename portion of the path.
870 */
871 mlink->name = start;
872 if (NULL != (p = strrchr(start, '/'))) {
873 mlink->name = p + 1;
874 *p = '\0';
875 }
876 mlink_add(mlink, &st);
877 }
878
879 static void
880 mlink_add(struct mlink *mlink, const struct stat *st)
881 {
882 struct inodev inodev;
883 struct mpage *mpage;
884 unsigned int slot;
885
886 assert(NULL != mlink->file);
887
888 mlink->dsec = mandoc_strdup(mlink->dsec ? mlink->dsec : "");
889 mlink->arch = mandoc_strdup(mlink->arch ? mlink->arch : "");
890 mlink->name = mandoc_strdup(mlink->name ? mlink->name : "");
891 mlink->fsec = mandoc_strdup(mlink->fsec ? mlink->fsec : "");
892
893 if ('0' == *mlink->fsec) {
894 free(mlink->fsec);
895 mlink->fsec = mandoc_strdup(mlink->dsec);
896 mlink->fform = FORM_CAT;
897 } else if ('1' <= *mlink->fsec && '9' >= *mlink->fsec)
898 mlink->fform = FORM_SRC;
899 else
900 mlink->fform = FORM_NONE;
901
902 slot = ohash_qlookup(&mlinks, mlink->file);
903 assert(NULL == ohash_find(&mlinks, slot));
904 ohash_insert(&mlinks, slot, mlink);
905
906 inodev.st_ino = st->st_ino;
907 inodev.st_dev = st->st_dev;
908 slot = ohash_lookup_memory(&mpages, (char *)&inodev,
909 sizeof(struct inodev), inodev.st_ino);
910 mpage = ohash_find(&mpages, slot);
911 if (NULL == mpage) {
912 mpage = mandoc_calloc(1, sizeof(struct mpage));
913 mpage->inodev.st_ino = inodev.st_ino;
914 mpage->inodev.st_dev = inodev.st_dev;
915 ohash_insert(&mpages, slot, mpage);
916 } else
917 mlink->next = mpage->mlinks;
918 mpage->mlinks = mlink;
919 mlink->mpage = mpage;
920 }
921
922 static void
923 mlink_free(struct mlink *mlink)
924 {
925
926 free(mlink->dsec);
927 free(mlink->arch);
928 free(mlink->name);
929 free(mlink->fsec);
930 free(mlink);
931 }
932
933 static void
934 mpages_free(void)
935 {
936 struct mpage *mpage;
937 struct mlink *mlink;
938 unsigned int slot;
939
940 mpage = ohash_first(&mpages, &slot);
941 while (NULL != mpage) {
942 while (NULL != (mlink = mpage->mlinks)) {
943 mpage->mlinks = mlink->next;
944 mlink_free(mlink);
945 }
946 free(mpage->sec);
947 free(mpage->arch);
948 free(mpage->title);
949 free(mpage->desc);
950 free(mpage);
951 mpage = ohash_next(&mpages, &slot);
952 }
953 }
954
955 /*
956 * For each mlink to the mpage, check whether the path looks like
957 * it is formatted, and if it does, check whether a source manual
958 * exists by the same name, ignoring the suffix.
959 * If both conditions hold, drop the mlink.
960 */
961 static void
962 mlinks_undupe(struct mpage *mpage)
963 {
964 char buf[PATH_MAX];
965 struct mlink **prev;
966 struct mlink *mlink;
967 char *bufp;
968
969 mpage->form = FORM_CAT;
970 prev = &mpage->mlinks;
971 while (NULL != (mlink = *prev)) {
972 if (FORM_CAT != mlink->dform) {
973 mpage->form = FORM_NONE;
974 goto nextlink;
975 }
976 if (strlcpy(buf, mlink->file, PATH_MAX) >= PATH_MAX) {
977 if (warnings)
978 say(mlink->file, "Filename too long");
979 goto nextlink;
980 }
981 bufp = strstr(buf, "cat");
982 assert(NULL != bufp);
983 memcpy(bufp, "man", 3);
984 if (NULL != (bufp = strrchr(buf, '.')))
985 *++bufp = '\0';
986 strlcat(buf, mlink->dsec, PATH_MAX);
987 if (NULL == ohash_find(&mlinks,
988 ohash_qlookup(&mlinks, buf)))
989 goto nextlink;
990 if (warnings)
991 say(mlink->file, "Man source exists: %s", buf);
992 if (use_all)
993 goto nextlink;
994 *prev = mlink->next;
995 mlink_free(mlink);
996 continue;
997 nextlink:
998 prev = &(*prev)->next;
999 }
1000 }
1001
1002 static void
1003 mlink_check(struct mpage *mpage, struct mlink *mlink)
1004 {
1005 struct str *str;
1006 unsigned int slot;
1007
1008 /*
1009 * Check whether the manual section given in a file
1010 * agrees with the directory where the file is located.
1011 * Some manuals have suffixes like (3p) on their
1012 * section number either inside the file or in the
1013 * directory name, some are linked into more than one
1014 * section, like encrypt(1) = makekey(8).
1015 */
1016
1017 if (FORM_SRC == mpage->form &&
1018 strcasecmp(mpage->sec, mlink->dsec))
1019 say(mlink->file, "Section \"%s\" manual in %s directory",
1020 mpage->sec, mlink->dsec);
1021
1022 /*
1023 * Manual page directories exist for each kernel
1024 * architecture as returned by machine(1).
1025 * However, many manuals only depend on the
1026 * application architecture as returned by arch(1).
1027 * For example, some (2/ARM) manuals are shared
1028 * across the "armish" and "zaurus" kernel
1029 * architectures.
1030 * A few manuals are even shared across completely
1031 * different architectures, for example fdformat(1)
1032 * on amd64, i386, sparc, and sparc64.
1033 */
1034
1035 if (strcasecmp(mpage->arch, mlink->arch))
1036 say(mlink->file, "Architecture \"%s\" manual in "
1037 "\"%s\" directory", mpage->arch, mlink->arch);
1038
1039 /*
1040 * XXX
1041 * parse_cat() doesn't set NAME_TITLE yet.
1042 */
1043
1044 if (FORM_CAT == mpage->form)
1045 return;
1046
1047 /*
1048 * Check whether this mlink
1049 * appears as a name in the NAME section.
1050 */
1051
1052 slot = ohash_qlookup(&names, mlink->name);
1053 str = ohash_find(&names, slot);
1054 assert(NULL != str);
1055 if ( ! (NAME_TITLE & str->mask))
1056 say(mlink->file, "Name missing in NAME section");
1057 }
1058
1059 /*
1060 * Run through the files in the global vector "mpages"
1061 * and add them to the database specified in "basedir".
1062 *
1063 * This handles the parsing scheme itself, using the cues of directory
1064 * and filename to determine whether the file is parsable or not.
1065 */
1066 static void
1067 mpages_merge(struct mchars *mc, struct mparse *mp)
1068 {
1069 char any[] = "any";
1070 struct ohash_info str_info;
1071 int fd[2];
1072 struct mpage *mpage, *mpage_dest;
1073 struct mlink *mlink, *mlink_dest;
1074 struct mdoc *mdoc;
1075 struct man *man;
1076 char *sodest;
1077 char *cp;
1078 pid_t child_pid;
1079 int status;
1080 unsigned int pslot;
1081 enum mandoclevel lvl;
1082
1083 str_info.alloc = hash_alloc;
1084 str_info.halloc = hash_halloc;
1085 str_info.hfree = hash_free;
1086 str_info.key_offset = offsetof(struct str, key);
1087
1088 if (0 == nodb)
1089 SQL_EXEC("BEGIN TRANSACTION");
1090
1091 mpage = ohash_first(&mpages, &pslot);
1092 while (NULL != mpage) {
1093 mlinks_undupe(mpage);
1094 if (NULL == mpage->mlinks) {
1095 mpage = ohash_next(&mpages, &pslot);
1096 continue;
1097 }
1098
1099 name_mask = NAME_MASK;
1100 ohash_init(&names, 4, &str_info);
1101 ohash_init(&strings, 6, &str_info);
1102 mparse_reset(mp);
1103 mdoc = NULL;
1104 man = NULL;
1105 sodest = NULL;
1106 child_pid = 0;
1107 fd[0] = -1;
1108 fd[1] = -1;
1109
1110 if (mpage->mlinks->gzip) {
1111 if (-1 == pipe(fd)) {
1112 exitcode = (int)MANDOCLEVEL_SYSERR;
1113 say(mpage->mlinks->file, "&pipe gunzip");
1114 goto nextpage;
1115 }
1116 switch (child_pid = fork()) {
1117 case -1:
1118 exitcode = (int)MANDOCLEVEL_SYSERR;
1119 say(mpage->mlinks->file, "&fork gunzip");
1120 child_pid = 0;
1121 close(fd[1]);
1122 close(fd[0]);
1123 goto nextpage;
1124 case 0:
1125 close(fd[0]);
1126 if (-1 == dup2(fd[1], STDOUT_FILENO)) {
1127 say(mpage->mlinks->file,
1128 "&dup gunzip");
1129 exit(1);
1130 }
1131 execlp("gunzip", "gunzip", "-c",
1132 mpage->mlinks->file, NULL);
1133 say(mpage->mlinks->file, "&exec gunzip");
1134 exit(1);
1135 default:
1136 close(fd[1]);
1137 break;
1138 }
1139 }
1140
1141 /*
1142 * Try interpreting the file as mdoc(7) or man(7)
1143 * source code, unless it is already known to be
1144 * formatted. Fall back to formatted mode.
1145 */
1146 if (FORM_CAT != mpage->mlinks->dform ||
1147 FORM_CAT != mpage->mlinks->fform) {
1148 lvl = mparse_readfd(mp, fd[0], mpage->mlinks->file);
1149 if (lvl < MANDOCLEVEL_FATAL)
1150 mparse_result(mp, &mdoc, &man, &sodest);
1151 }
1152
1153 if (NULL != sodest) {
1154 mlink_dest = ohash_find(&mlinks,
1155 ohash_qlookup(&mlinks, sodest));
1156 if (NULL != mlink_dest) {
1157
1158 /* The .so target exists. */
1159
1160 mpage_dest = mlink_dest->mpage;
1161 mlink = mpage->mlinks;
1162 while (1) {
1163 mlink->mpage = mpage_dest;
1164
1165 /*
1166 * If the target was already
1167 * processed, add the links
1168 * to the database now.
1169 * Otherwise, this will
1170 * happen when we come
1171 * to the target.
1172 */
1173
1174 if (mpage_dest->pageid)
1175 dbadd_mlink(mlink);
1176
1177 if (NULL == mlink->next)
1178 break;
1179 mlink = mlink->next;
1180 }
1181
1182 /* Move all links to the target. */
1183
1184 mlink->next = mlink_dest->next;
1185 mlink_dest->next = mpage->mlinks;
1186 mpage->mlinks = NULL;
1187 }
1188 goto nextpage;
1189 } else if (NULL != mdoc) {
1190 mpage->form = FORM_SRC;
1191 mpage->sec =
1192 mandoc_strdup(mdoc_meta(mdoc)->msec);
1193 mpage->arch = mdoc_meta(mdoc)->arch;
1194 mpage->arch = mandoc_strdup(
1195 NULL == mpage->arch ? "" : mpage->arch);
1196 mpage->title =
1197 mandoc_strdup(mdoc_meta(mdoc)->title);
1198 } else if (NULL != man) {
1199 mpage->form = FORM_SRC;
1200 mpage->sec =
1201 mandoc_strdup(man_meta(man)->msec);
1202 mpage->arch =
1203 mandoc_strdup(mpage->mlinks->arch);
1204 mpage->title =
1205 mandoc_strdup(man_meta(man)->title);
1206 } else {
1207 mpage->form = FORM_CAT;
1208 mpage->sec =
1209 mandoc_strdup(mpage->mlinks->dsec);
1210 mpage->arch =
1211 mandoc_strdup(mpage->mlinks->arch);
1212 mpage->title =
1213 mandoc_strdup(mpage->mlinks->name);
1214 }
1215 putkey(mpage, mpage->sec, TYPE_sec);
1216 putkey(mpage, '\0' == *mpage->arch ?
1217 any : mpage->arch, TYPE_arch);
1218
1219 for (mlink = mpage->mlinks; mlink; mlink = mlink->next) {
1220 if ('\0' != *mlink->dsec)
1221 putkey(mpage, mlink->dsec, TYPE_sec);
1222 if ('\0' != *mlink->fsec)
1223 putkey(mpage, mlink->fsec, TYPE_sec);
1224 putkey(mpage, '\0' == *mlink->arch ?
1225 any : mlink->arch, TYPE_arch);
1226 putkey(mpage, mlink->name, NAME_FILE);
1227 }
1228
1229 assert(NULL == mpage->desc);
1230 if (NULL != mdoc) {
1231 if (NULL != (cp = mdoc_meta(mdoc)->name))
1232 putkey(mpage, cp, NAME_HEAD);
1233 parse_mdoc(mpage, mdoc_node(mdoc));
1234 } else if (NULL != man)
1235 parse_man(mpage, man_node(man));
1236 else
1237 parse_cat(mpage, fd[0]);
1238 if (NULL == mpage->desc)
1239 mpage->desc = mandoc_strdup(mpage->mlinks->name);
1240
1241 if (warnings && !use_all)
1242 for (mlink = mpage->mlinks; mlink;
1243 mlink = mlink->next)
1244 mlink_check(mpage, mlink);
1245
1246 dbadd(mpage, mc);
1247
1248 nextpage:
1249 if (child_pid) {
1250 if (-1 == waitpid(child_pid, &status, 0)) {
1251 exitcode = (int)MANDOCLEVEL_SYSERR;
1252 say(mpage->mlinks->file, "&wait gunzip");
1253 } else if (WIFSIGNALED(status)) {
1254 exitcode = (int)MANDOCLEVEL_SYSERR;
1255 say(mpage->mlinks->file,
1256 "gunzip died from signal %d",
1257 WTERMSIG(status));
1258 } else if (WEXITSTATUS(status)) {
1259 exitcode = (int)MANDOCLEVEL_SYSERR;
1260 say(mpage->mlinks->file,
1261 "gunzip failed with code %d",
1262 WEXITSTATUS(status));
1263 }
1264 }
1265 ohash_delete(&strings);
1266 ohash_delete(&names);
1267 mpage = ohash_next(&mpages, &pslot);
1268 }
1269
1270 if (0 == nodb)
1271 SQL_EXEC("END TRANSACTION");
1272 }
1273
1274 static void
1275 names_check(void)
1276 {
1277 sqlite3_stmt *stmt;
1278 const char *name, *sec, *arch, *key;
1279 int irc;
1280
1281 sqlite3_prepare_v2(db,
1282 "SELECT name, sec, arch, key FROM ("
1283 "SELECT name AS key, pageid FROM names "
1284 "WHERE bits & ? AND NOT EXISTS ("
1285 "SELECT pageid FROM mlinks "
1286 "WHERE mlinks.pageid == names.pageid "
1287 "AND mlinks.name == names.name"
1288 ")"
1289 ") JOIN ("
1290 "SELECT * FROM mlinks GROUP BY pageid"
1291 ") USING (pageid);",
1292 -1, &stmt, NULL);
1293
1294 if (SQLITE_OK != sqlite3_bind_int64(stmt, 1, NAME_TITLE))
1295 say("", "%s", sqlite3_errmsg(db));
1296
1297 while (SQLITE_ROW == (irc = sqlite3_step(stmt))) {
1298 name = sqlite3_column_text(stmt, 0);
1299 sec = sqlite3_column_text(stmt, 1);
1300 arch = sqlite3_column_text(stmt, 2);
1301 key = sqlite3_column_text(stmt, 3);
1302 say("", "%s(%s%s%s) lacks mlink \"%s\"", name, sec,
1303 '\0' == *arch ? "" : "/",
1304 '\0' == *arch ? "" : arch, key);
1305 }
1306 sqlite3_finalize(stmt);
1307 }
1308
1309 static void
1310 parse_cat(struct mpage *mpage, int fd)
1311 {
1312 FILE *stream;
1313 char *line, *p, *title;
1314 size_t len, plen, titlesz;
1315
1316 stream = (-1 == fd) ?
1317 fopen(mpage->mlinks->file, "r") :
1318 fdopen(fd, "r");
1319 if (NULL == stream) {
1320 if (warnings)
1321 say(mpage->mlinks->file, "&fopen");
1322 return;
1323 }
1324
1325 /* Skip to first blank line. */
1326
1327 while (NULL != (line = fgetln(stream, &len)))
1328 if ('\n' == *line)
1329 break;
1330
1331 /*
1332 * Assume the first line that is not indented
1333 * is the first section header. Skip to it.
1334 */
1335
1336 while (NULL != (line = fgetln(stream, &len)))
1337 if ('\n' != *line && ' ' != *line)
1338 break;
1339
1340 /*
1341 * Read up until the next section into a buffer.
1342 * Strip the leading and trailing newline from each read line,
1343 * appending a trailing space.
1344 * Ignore empty (whitespace-only) lines.
1345 */
1346
1347 titlesz = 0;
1348 title = NULL;
1349
1350 while (NULL != (line = fgetln(stream, &len))) {
1351 if (' ' != *line || '\n' != line[len - 1])
1352 break;
1353 while (len > 0 && isspace((unsigned char)*line)) {
1354 line++;
1355 len--;
1356 }
1357 if (1 == len)
1358 continue;
1359 title = mandoc_realloc(title, titlesz + len);
1360 memcpy(title + titlesz, line, len);
1361 titlesz += len;
1362 title[titlesz - 1] = ' ';
1363 }
1364
1365 /*
1366 * If no page content can be found, or the input line
1367 * is already the next section header, or there is no
1368 * trailing newline, reuse the page title as the page
1369 * description.
1370 */
1371
1372 if (NULL == title || '\0' == *title) {
1373 if (warnings)
1374 say(mpage->mlinks->file,
1375 "Cannot find NAME section");
1376 fclose(stream);
1377 free(title);
1378 return;
1379 }
1380
1381 title = mandoc_realloc(title, titlesz + 1);
1382 title[titlesz] = '\0';
1383
1384 /*
1385 * Skip to the first dash.
1386 * Use the remaining line as the description (no more than 70
1387 * bytes).
1388 */
1389
1390 if (NULL != (p = strstr(title, "- "))) {
1391 for (p += 2; ' ' == *p || '\b' == *p; p++)
1392 /* Skip to next word. */ ;
1393 } else {
1394 if (warnings)
1395 say(mpage->mlinks->file,
1396 "No dash in title line");
1397 p = title;
1398 }
1399
1400 plen = strlen(p);
1401
1402 /* Strip backspace-encoding from line. */
1403
1404 while (NULL != (line = memchr(p, '\b', plen))) {
1405 len = line - p;
1406 if (0 == len) {
1407 memmove(line, line + 1, plen--);
1408 continue;
1409 }
1410 memmove(line - 1, line + 1, plen - len);
1411 plen -= 2;
1412 }
1413
1414 mpage->desc = mandoc_strdup(p);
1415 fclose(stream);
1416 free(title);
1417 }
1418
1419 /*
1420 * Put a type/word pair into the word database for this particular file.
1421 */
1422 static void
1423 putkey(const struct mpage *mpage, char *value, uint64_t type)
1424 {
1425 char *cp;
1426
1427 assert(NULL != value);
1428 if (TYPE_arch == type)
1429 for (cp = value; *cp; cp++)
1430 if (isupper((unsigned char)*cp))
1431 *cp = _tolower((unsigned char)*cp);
1432 putkeys(mpage, value, strlen(value), type);
1433 }
1434
1435 /*
1436 * Grok all nodes at or below a certain mdoc node into putkey().
1437 */
1438 static void
1439 putmdockey(const struct mpage *mpage,
1440 const struct mdoc_node *n, uint64_t m)
1441 {
1442
1443 for ( ; NULL != n; n = n->next) {
1444 if (NULL != n->child)
1445 putmdockey(mpage, n->child, m);
1446 if (MDOC_TEXT == n->type)
1447 putkey(mpage, n->string, m);
1448 }
1449 }
1450
1451 static void
1452 parse_man(struct mpage *mpage, const struct man_node *n)
1453 {
1454 const struct man_node *head, *body;
1455 char *start, *title;
1456 char byte;
1457 size_t sz;
1458
1459 if (NULL == n)
1460 return;
1461
1462 /*
1463 * We're only searching for one thing: the first text child in
1464 * the BODY of a NAME section. Since we don't keep track of
1465 * sections in -man, run some hoops to find out whether we're in
1466 * the correct section or not.
1467 */
1468
1469 if (MAN_BODY == n->type && MAN_SH == n->tok) {
1470 body = n;
1471 assert(body->parent);
1472 if (NULL != (head = body->parent->head) &&
1473 1 == head->nchild &&
1474 NULL != (head = (head->child)) &&
1475 MAN_TEXT == head->type &&
1476 0 == strcmp(head->string, "NAME") &&
1477 NULL != body->child) {
1478
1479 /*
1480 * Suck the entire NAME section into memory.
1481 * Yes, we might run away.
1482 * But too many manuals have big, spread-out
1483 * NAME sections over many lines.
1484 */
1485
1486 title = NULL;
1487 man_deroff(&title, body);
1488 if (NULL == title)
1489 return;
1490
1491 /*
1492 * Go through a special heuristic dance here.
1493 * Conventionally, one or more manual names are
1494 * comma-specified prior to a whitespace, then a
1495 * dash, then a description. Try to puzzle out
1496 * the name parts here.
1497 */
1498
1499 start = title;
1500 for ( ;; ) {
1501 sz = strcspn(start, " ,");
1502 if ('\0' == start[sz])
1503 break;
1504
1505 byte = start[sz];
1506 start[sz] = '\0';
1507
1508 /*
1509 * Assume a stray trailing comma in the
1510 * name list if a name begins with a dash.
1511 */
1512
1513 if ('-' == start[0] ||
1514 ('\\' == start[0] && '-' == start[1]))
1515 break;
1516
1517 putkey(mpage, start, NAME_TITLE);
1518
1519 if (' ' == byte) {
1520 start += sz + 1;
1521 break;
1522 }
1523
1524 assert(',' == byte);
1525 start += sz + 1;
1526 while (' ' == *start)
1527 start++;
1528 }
1529
1530 if (start == title) {
1531 putkey(mpage, start, NAME_TITLE);
1532 free(title);
1533 return;
1534 }
1535
1536 while (isspace((unsigned char)*start))
1537 start++;
1538
1539 if (0 == strncmp(start, "-", 1))
1540 start += 1;
1541 else if (0 == strncmp(start, "\\-\\-", 4))
1542 start += 4;
1543 else if (0 == strncmp(start, "\\-", 2))
1544 start += 2;
1545 else if (0 == strncmp(start, "\\(en", 4))
1546 start += 4;
1547 else if (0 == strncmp(start, "\\(em", 4))
1548 start += 4;
1549
1550 while (' ' == *start)
1551 start++;
1552
1553 mpage->desc = mandoc_strdup(start);
1554 free(title);
1555 return;
1556 }
1557 }
1558
1559 for (n = n->child; n; n = n->next) {
1560 if (NULL != mpage->desc)
1561 break;
1562 parse_man(mpage, n);
1563 }
1564 }
1565
1566 static void
1567 parse_mdoc(struct mpage *mpage, const struct mdoc_node *n)
1568 {
1569
1570 assert(NULL != n);
1571 for (n = n->child; NULL != n; n = n->next) {
1572 switch (n->type) {
1573 case MDOC_ELEM:
1574 /* FALLTHROUGH */
1575 case MDOC_BLOCK:
1576 /* FALLTHROUGH */
1577 case MDOC_HEAD:
1578 /* FALLTHROUGH */
1579 case MDOC_BODY:
1580 /* FALLTHROUGH */
1581 case MDOC_TAIL:
1582 if (NULL != mdocs[n->tok].fp)
1583 if (0 == (*mdocs[n->tok].fp)(mpage, n))
1584 break;
1585 if (mdocs[n->tok].mask)
1586 putmdockey(mpage, n->child,
1587 mdocs[n->tok].mask);
1588 break;
1589 default:
1590 assert(MDOC_ROOT != n->type);
1591 continue;
1592 }
1593 if (NULL != n->child)
1594 parse_mdoc(mpage, n);
1595 }
1596 }
1597
1598 static int
1599 parse_mdoc_Fd(struct mpage *mpage, const struct mdoc_node *n)
1600 {
1601 const char *start, *end;
1602 size_t sz;
1603
1604 if (SEC_SYNOPSIS != n->sec ||
1605 NULL == (n = n->child) ||
1606 MDOC_TEXT != n->type)
1607 return(0);
1608
1609 /*
1610 * Only consider those `Fd' macro fields that begin with an
1611 * "inclusion" token (versus, e.g., #define).
1612 */
1613
1614 if (strcmp("#include", n->string))
1615 return(0);
1616
1617 if (NULL == (n = n->next) || MDOC_TEXT != n->type)
1618 return(0);
1619
1620 /*
1621 * Strip away the enclosing angle brackets and make sure we're
1622 * not zero-length.
1623 */
1624
1625 start = n->string;
1626 if ('<' == *start || '"' == *start)
1627 start++;
1628
1629 if (0 == (sz = strlen(start)))
1630 return(0);
1631
1632 end = &start[(int)sz - 1];
1633 if ('>' == *end || '"' == *end)
1634 end--;
1635
1636 if (end > start)
1637 putkeys(mpage, start, end - start + 1, TYPE_In);
1638 return(0);
1639 }
1640
1641 static int
1642 parse_mdoc_Fn(struct mpage *mpage, const struct mdoc_node *n)
1643 {
1644 char *cp;
1645
1646 if (NULL == (n = n->child) || MDOC_TEXT != n->type)
1647 return(0);
1648
1649 /*
1650 * Parse: .Fn "struct type *name" "char *arg".
1651 * First strip away pointer symbol.
1652 * Then store the function name, then type.
1653 * Finally, store the arguments.
1654 */
1655
1656 if (NULL == (cp = strrchr(n->string, ' ')))
1657 cp = n->string;
1658
1659 while ('*' == *cp)
1660 cp++;
1661
1662 putkey(mpage, cp, TYPE_Fn);
1663
1664 if (n->string < cp)
1665 putkeys(mpage, n->string, cp - n->string, TYPE_Ft);
1666
1667 for (n = n->next; NULL != n; n = n->next)
1668 if (MDOC_TEXT == n->type)
1669 putkey(mpage, n->string, TYPE_Fa);
1670
1671 return(0);
1672 }
1673
1674 static int
1675 parse_mdoc_Xr(struct mpage *mpage, const struct mdoc_node *n)
1676 {
1677 char *cp;
1678
1679 if (NULL == (n = n->child))
1680 return(0);
1681
1682 if (NULL == n->next) {
1683 putkey(mpage, n->string, TYPE_Xr);
1684 return(0);
1685 }
1686
1687 mandoc_asprintf(&cp, "%s(%s)", n->string, n->next->string);
1688 putkey(mpage, cp, TYPE_Xr);
1689 free(cp);
1690 return(0);
1691 }
1692
1693 static int
1694 parse_mdoc_Nd(struct mpage *mpage, const struct mdoc_node *n)
1695 {
1696
1697 if (MDOC_BODY == n->type)
1698 mdoc_deroff(&mpage->desc, n);
1699 return(0);
1700 }
1701
1702 static int
1703 parse_mdoc_Nm(struct mpage *mpage, const struct mdoc_node *n)
1704 {
1705
1706 if (SEC_NAME == n->sec)
1707 putmdockey(mpage, n->child, NAME_TITLE);
1708 else if (SEC_SYNOPSIS == n->sec && MDOC_HEAD == n->type)
1709 putmdockey(mpage, n->child, NAME_SYN);
1710 return(0);
1711 }
1712
1713 static int
1714 parse_mdoc_Sh(struct mpage *mpage, const struct mdoc_node *n)
1715 {
1716
1717 return(SEC_CUSTOM == n->sec && MDOC_HEAD == n->type);
1718 }
1719
1720 static int
1721 parse_mdoc_head(struct mpage *mpage, const struct mdoc_node *n)
1722 {
1723
1724 return(MDOC_HEAD == n->type);
1725 }
1726
1727 static int
1728 parse_mdoc_body(struct mpage *mpage, const struct mdoc_node *n)
1729 {
1730
1731 return(MDOC_BODY == n->type);
1732 }
1733
1734 /*
1735 * Add a string to the hash table for the current manual.
1736 * Each string has a bitmask telling which macros it belongs to.
1737 * When we finish the manual, we'll dump the table.
1738 */
1739 static void
1740 putkeys(const struct mpage *mpage,
1741 const char *cp, size_t sz, uint64_t v)
1742 {
1743 struct ohash *htab;
1744 struct str *s;
1745 const char *end;
1746 unsigned int slot;
1747 int i;
1748
1749 if (0 == sz)
1750 return;
1751
1752 if (TYPE_Nm & v) {
1753 htab = &names;
1754 v &= name_mask;
1755 name_mask &= ~NAME_FIRST;
1756 if (debug > 1)
1757 say(mpage->mlinks->file,
1758 "Adding name %*s", sz, cp);
1759 } else {
1760 htab = &strings;
1761 if (debug > 1)
1762 for (i = 0; i < mansearch_keymax; i++)
1763 if (1 << i & v)
1764 say(mpage->mlinks->file,
1765 "Adding key %s=%*s",
1766 mansearch_keynames[i], sz, cp);
1767 }
1768
1769 end = cp + sz;
1770 slot = ohash_qlookupi(htab, cp, &end);
1771 s = ohash_find(htab, slot);
1772
1773 if (NULL != s && mpage == s->mpage) {
1774 s->mask |= v;
1775 return;
1776 } else if (NULL == s) {
1777 s = mandoc_calloc(sizeof(struct str) + sz + 1, 1);
1778 memcpy(s->key, cp, sz);
1779 ohash_insert(htab, slot, s);
1780 }
1781 s->mpage = mpage;
1782 s->mask = v;
1783 }
1784
1785 /*
1786 * Take a Unicode codepoint and produce its UTF-8 encoding.
1787 * This isn't the best way to do this, but it works.
1788 * The magic numbers are from the UTF-8 packaging.
1789 * They're not as scary as they seem: read the UTF-8 spec for details.
1790 */
1791 static size_t
1792 utf8(unsigned int cp, char out[7])
1793 {
1794 size_t rc;
1795
1796 rc = 0;
1797 if (cp <= 0x0000007F) {
1798 rc = 1;
1799 out[0] = (char)cp;
1800 } else if (cp <= 0x000007FF) {
1801 rc = 2;
1802 out[0] = (cp >> 6 & 31) | 192;
1803 out[1] = (cp & 63) | 128;
1804 } else if (cp <= 0x0000FFFF) {
1805 rc = 3;
1806 out[0] = (cp >> 12 & 15) | 224;
1807 out[1] = (cp >> 6 & 63) | 128;
1808 out[2] = (cp & 63) | 128;
1809 } else if (cp <= 0x001FFFFF) {
1810 rc = 4;
1811 out[0] = (cp >> 18 & 7) | 240;
1812 out[1] = (cp >> 12 & 63) | 128;
1813 out[2] = (cp >> 6 & 63) | 128;
1814 out[3] = (cp & 63) | 128;
1815 } else if (cp <= 0x03FFFFFF) {
1816 rc = 5;
1817 out[0] = (cp >> 24 & 3) | 248;
1818 out[1] = (cp >> 18 & 63) | 128;
1819 out[2] = (cp >> 12 & 63) | 128;
1820 out[3] = (cp >> 6 & 63) | 128;
1821 out[4] = (cp & 63) | 128;
1822 } else if (cp <= 0x7FFFFFFF) {
1823 rc = 6;
1824 out[0] = (cp >> 30 & 1) | 252;
1825 out[1] = (cp >> 24 & 63) | 128;
1826 out[2] = (cp >> 18 & 63) | 128;
1827 out[3] = (cp >> 12 & 63) | 128;
1828 out[4] = (cp >> 6 & 63) | 128;
1829 out[5] = (cp & 63) | 128;
1830 } else
1831 return(0);
1832
1833 out[rc] = '\0';
1834 return(rc);
1835 }
1836
1837 /*
1838 * Store the rendered version of a key, or alias the pointer
1839 * if the key contains no escape sequences.
1840 */
1841 static void
1842 render_key(struct mchars *mc, struct str *key)
1843 {
1844 size_t sz, bsz, pos;
1845 char utfbuf[7], res[6];
1846 char *buf;
1847 const char *seq, *cpp, *val;
1848 int len, u;
1849 enum mandoc_esc esc;
1850
1851 assert(NULL == key->rendered);
1852
1853 res[0] = '\\';
1854 res[1] = '\t';
1855 res[2] = ASCII_NBRSP;
1856 res[3] = ASCII_HYPH;
1857 res[4] = ASCII_BREAK;
1858 res[5] = '\0';
1859
1860 val = key->key;
1861 bsz = strlen(val);
1862
1863 /*
1864 * Pre-check: if we have no stop-characters, then set the
1865 * pointer as ourselvse and get out of here.
1866 */
1867 if (strcspn(val, res) == bsz) {
1868 key->rendered = key->key;
1869 return;
1870 }
1871
1872 /* Pre-allocate by the length of the input */
1873
1874 buf = mandoc_malloc(++bsz);
1875 pos = 0;
1876
1877 while ('\0' != *val) {
1878 /*
1879 * Halt on the first escape sequence.
1880 * This also halts on the end of string, in which case
1881 * we just copy, fallthrough, and exit the loop.
1882 */
1883 if ((sz = strcspn(val, res)) > 0) {
1884 memcpy(&buf[pos], val, sz);
1885 pos += sz;
1886 val += sz;
1887 }
1888
1889 switch (*val) {
1890 case ASCII_HYPH:
1891 buf[pos++] = '-';
1892 val++;
1893 continue;
1894 case '\t':
1895 /* FALLTHROUGH */
1896 case ASCII_NBRSP:
1897 buf[pos++] = ' ';
1898 val++;
1899 /* FALLTHROUGH */
1900 case ASCII_BREAK:
1901 continue;
1902 default:
1903 break;
1904 }
1905 if ('\\' != *val)
1906 break;
1907
1908 /* Read past the slash. */
1909
1910 val++;
1911
1912 /*
1913 * Parse the escape sequence and see if it's a
1914 * predefined character or special character.
1915 */
1916
1917 esc = mandoc_escape((const char **)&val,
1918 &seq, &len);
1919 if (ESCAPE_ERROR == esc)
1920 break;
1921 if (ESCAPE_SPECIAL != esc)
1922 continue;
1923
1924 /*
1925 * Render the special character
1926 * as either UTF-8 or ASCII.
1927 */
1928
1929 if (write_utf8) {
1930 if (0 == (u = mchars_spec2cp(mc, seq, len)))
1931 continue;
1932 cpp = utfbuf;
1933 if (0 == (sz = utf8(u, utfbuf)))
1934 continue;
1935 sz = strlen(cpp);
1936 } else {
1937 cpp = mchars_spec2str(mc, seq, len, &sz);
1938 if (NULL == cpp)
1939 continue;
1940 if (ASCII_NBRSP == *cpp) {
1941 cpp = " ";
1942 sz = 1;
1943 }
1944 }
1945
1946 /* Copy the rendered glyph into the stream. */
1947
1948 bsz += sz;
1949 buf = mandoc_realloc(buf, bsz);
1950 memcpy(&buf[pos], cpp, sz);
1951 pos += sz;
1952 }
1953
1954 buf[pos] = '\0';
1955 key->rendered = buf;
1956 }
1957
1958 static void
1959 dbadd_mlink(const struct mlink *mlink)
1960 {
1961 size_t i;
1962
1963 i = 1;
1964 SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->dsec);
1965 SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->arch);
1966 SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->name);
1967 SQL_BIND_INT64(stmts[STMT_INSERT_LINK], i, mlink->mpage->pageid);
1968 SQL_STEP(stmts[STMT_INSERT_LINK]);
1969 sqlite3_reset(stmts[STMT_INSERT_LINK]);
1970 }
1971
1972 /*
1973 * Flush the current page's terms (and their bits) into the database.
1974 * Wrap the entire set of additions in a transaction to make sqlite be a
1975 * little faster.
1976 * Also, handle escape sequences at the last possible moment.
1977 */
1978 static void
1979 dbadd(struct mpage *mpage, struct mchars *mc)
1980 {
1981 struct mlink *mlink;
1982 struct str *key;
1983 size_t i;
1984 unsigned int slot;
1985
1986 mlink = mpage->mlinks;
1987
1988 if (nodb) {
1989 while (NULL != mlink) {
1990 fputs(mlink->name, stdout);
1991 if (NULL == mlink->next ||
1992 strcmp(mlink->dsec, mlink->next->dsec) ||
1993 strcmp(mlink->fsec, mlink->next->fsec) ||
1994 strcmp(mlink->arch, mlink->next->arch)) {
1995 putchar('(');
1996 if ('\0' == *mlink->dsec)
1997 fputs(mlink->fsec, stdout);
1998 else
1999 fputs(mlink->dsec, stdout);
2000 if ('\0' != *mlink->arch)
2001 printf("/%s", mlink->arch);
2002 putchar(')');
2003 }
2004 mlink = mlink->next;
2005 if (NULL != mlink)
2006 fputs(", ", stdout);
2007 }
2008 printf(" - %s\n", mpage->desc);
2009 return;
2010 }
2011
2012 if (debug)
2013 say(mlink->file, "Adding to database");
2014
2015 i = 1;
2016 SQL_BIND_TEXT(stmts[STMT_INSERT_PAGE], i, mpage->desc);
2017 SQL_BIND_INT(stmts[STMT_INSERT_PAGE], i, FORM_SRC == mpage->form);
2018 SQL_STEP(stmts[STMT_INSERT_PAGE]);
2019 mpage->pageid = sqlite3_last_insert_rowid(db);
2020 sqlite3_reset(stmts[STMT_INSERT_PAGE]);
2021
2022 while (NULL != mlink) {
2023 dbadd_mlink(mlink);
2024 mlink = mlink->next;
2025 }
2026 mlink = mpage->mlinks;
2027
2028 for (key = ohash_first(&names, &slot); NULL != key;
2029 key = ohash_next(&names, &slot)) {
2030 assert(key->mpage == mpage);
2031 if (NULL == key->rendered)
2032 render_key(mc, key);
2033 i = 1;
2034 SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, key->mask);
2035 SQL_BIND_TEXT(stmts[STMT_INSERT_NAME], i, key->rendered);
2036 SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, mpage->pageid);
2037 SQL_STEP(stmts[STMT_INSERT_NAME]);
2038 sqlite3_reset(stmts[STMT_INSERT_NAME]);
2039 if (key->rendered != key->key)
2040 free(key->rendered);
2041 free(key);
2042 }
2043 for (key = ohash_first(&strings, &slot); NULL != key;
2044 key = ohash_next(&strings, &slot)) {
2045 assert(key->mpage == mpage);
2046 if (NULL == key->rendered)
2047 render_key(mc, key);
2048 i = 1;
2049 SQL_BIND_INT64(stmts[STMT_INSERT_KEY], i, key->mask);
2050 SQL_BIND_TEXT(stmts[STMT_INSERT_KEY], i, key->rendered);
2051 SQL_BIND_INT64(stmts[STMT_INSERT_KEY], i, mpage->pageid);
2052 SQL_STEP(stmts[STMT_INSERT_KEY]);
2053 sqlite3_reset(stmts[STMT_INSERT_KEY]);
2054 if (key->rendered != key->key)
2055 free(key->rendered);
2056 free(key);
2057 }
2058 }
2059
2060 static void
2061 dbprune(void)
2062 {
2063 struct mpage *mpage;
2064 struct mlink *mlink;
2065 size_t i;
2066 unsigned int slot;
2067
2068 if (0 == nodb)
2069 SQL_EXEC("BEGIN TRANSACTION");
2070
2071 for (mpage = ohash_first(&mpages, &slot); NULL != mpage;
2072 mpage = ohash_next(&mpages, &slot)) {
2073 mlink = mpage->mlinks;
2074 if (debug)
2075 say(mlink->file, "Deleting from database");
2076 if (nodb)
2077 continue;
2078 for ( ; NULL != mlink; mlink = mlink->next) {
2079 i = 1;
2080 SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE],
2081 i, mlink->dsec);
2082 SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE],
2083 i, mlink->arch);
2084 SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE],
2085 i, mlink->name);
2086 SQL_STEP(stmts[STMT_DELETE_PAGE]);
2087 sqlite3_reset(stmts[STMT_DELETE_PAGE]);
2088 }
2089 }
2090
2091 if (0 == nodb)
2092 SQL_EXEC("END TRANSACTION");
2093 }
2094
2095 /*
2096 * Close an existing database and its prepared statements.
2097 * If "real" is not set, rename the temporary file into the real one.
2098 */
2099 static void
2100 dbclose(int real)
2101 {
2102 size_t i;
2103 int status;
2104 pid_t child;
2105
2106 if (nodb)
2107 return;
2108
2109 for (i = 0; i < STMT__MAX; i++) {
2110 sqlite3_finalize(stmts[i]);
2111 stmts[i] = NULL;
2112 }
2113
2114 sqlite3_close(db);
2115 db = NULL;
2116
2117 if (real)
2118 return;
2119
2120 if ('\0' == *tempfilename) {
2121 if (-1 == rename(MANDOC_DB "~", MANDOC_DB)) {
2122 exitcode = (int)MANDOCLEVEL_SYSERR;
2123 say(MANDOC_DB, "&rename");
2124 }
2125 return;
2126 }
2127
2128 switch (child = fork()) {
2129 case -1:
2130 exitcode = (int)MANDOCLEVEL_SYSERR;
2131 say("", "&fork cmp");
2132 return;
2133 case 0:
2134 execlp("cmp", "cmp", "-s",
2135 tempfilename, MANDOC_DB, NULL);
2136 say("", "&exec cmp");
2137 exit(0);
2138 default:
2139 break;
2140 }
2141 if (-1 == waitpid(child, &status, 0)) {
2142 exitcode = (int)MANDOCLEVEL_SYSERR;
2143 say("", "&wait cmp");
2144 } else if (WIFSIGNALED(status)) {
2145 exitcode = (int)MANDOCLEVEL_SYSERR;
2146 say("", "cmp died from signal %d", WTERMSIG(status));
2147 } else if (WEXITSTATUS(status)) {
2148 exitcode = (int)MANDOCLEVEL_SYSERR;
2149 say(MANDOC_DB,
2150 "Data changed, but cannot replace database");
2151 }
2152
2153 *strrchr(tempfilename, '/') = '\0';
2154 switch (child = fork()) {
2155 case -1:
2156 exitcode = (int)MANDOCLEVEL_SYSERR;
2157 say("", "&fork rm");
2158 return;
2159 case 0:
2160 execlp("rm", "rm", "-rf", tempfilename, NULL);
2161 say("", "&exec rm");
2162 exit((int)MANDOCLEVEL_SYSERR);
2163 default:
2164 break;
2165 }
2166 if (-1 == waitpid(child, &status, 0)) {
2167 exitcode = (int)MANDOCLEVEL_SYSERR;
2168 say("", "&wait rm");
2169 } else if (WIFSIGNALED(status) || WEXITSTATUS(status)) {
2170 exitcode = (int)MANDOCLEVEL_SYSERR;
2171 say("", "%s: Cannot remove temporary directory",
2172 tempfilename);
2173 }
2174 }
2175
2176 /*
2177 * This is straightforward stuff.
2178 * Open a database connection to a "temporary" database, then open a set
2179 * of prepared statements we'll use over and over again.
2180 * If "real" is set, we use the existing database; if not, we truncate a
2181 * temporary one.
2182 * Must be matched by dbclose().
2183 */
2184 static int
2185 dbopen(int real)
2186 {
2187 const char *sql;
2188 int rc, ofl;
2189
2190 if (nodb)
2191 return(1);
2192
2193 *tempfilename = '\0';
2194 ofl = SQLITE_OPEN_READWRITE;
2195
2196 if (real) {
2197 rc = sqlite3_open_v2(MANDOC_DB, &db, ofl, NULL);
2198 if (SQLITE_OK != rc) {
2199 exitcode = (int)MANDOCLEVEL_SYSERR;
2200 say(MANDOC_DB, "%s", sqlite3_errmsg(db));
2201 return(0);
2202 }
2203 goto prepare_statements;
2204 }
2205
2206 ofl |= SQLITE_OPEN_CREATE | SQLITE_OPEN_EXCLUSIVE;
2207
2208 remove(MANDOC_DB "~");
2209 rc = sqlite3_open_v2(MANDOC_DB "~", &db, ofl, NULL);
2210 if (SQLITE_OK == rc)
2211 goto create_tables;
2212 if (MPARSE_QUICK & mparse_options) {
2213 exitcode = (int)MANDOCLEVEL_SYSERR;
2214 say(MANDOC_DB "~", "%s", sqlite3_errmsg(db));
2215 return(0);
2216 }
2217
2218 if (strlcpy(tempfilename, "/tmp/mandocdb.XXXXXX",
2219 sizeof(tempfilename)) >= sizeof(tempfilename)) {
2220 exitcode = (int)MANDOCLEVEL_SYSERR;
2221 say("", "/tmp/mandocdb.XXXXXX: Filename too long");
2222 return(0);
2223 }
2224 if (NULL == mkdtemp(tempfilename)) {
2225 exitcode = (int)MANDOCLEVEL_SYSERR;
2226 say("", "&%s", tempfilename);
2227 return(0);
2228 }
2229 if (strlcat(tempfilename, "/" MANDOC_DB,
2230 sizeof(tempfilename)) >= sizeof(tempfilename)) {
2231 exitcode = (int)MANDOCLEVEL_SYSERR;
2232 say("", "%s/" MANDOC_DB ": Filename too long",
2233 tempfilename);
2234 return(0);
2235 }
2236 rc = sqlite3_open_v2(tempfilename, &db, ofl, NULL);
2237 if (SQLITE_OK != rc) {
2238 exitcode = (int)MANDOCLEVEL_SYSERR;
2239 say("", "%s: %s", tempfilename, sqlite3_errmsg(db));
2240 return(0);
2241 }
2242
2243 create_tables:
2244 sql = "CREATE TABLE \"mpages\" (\n"
2245 " \"desc\" TEXT NOT NULL,\n"
2246 " \"form\" INTEGER NOT NULL,\n"
2247 " \"pageid\" INTEGER PRIMARY KEY AUTOINCREMENT NOT NULL\n"
2248 ");\n"
2249 "\n"
2250 "CREATE TABLE \"mlinks\" (\n"
2251 " \"sec\" TEXT NOT NULL,\n"
2252 " \"arch\" TEXT NOT NULL,\n"
2253 " \"name\" TEXT NOT NULL,\n"
2254 " \"pageid\" INTEGER NOT NULL REFERENCES mpages(pageid) "
2255 "ON DELETE CASCADE\n"
2256 ");\n"
2257 "CREATE INDEX mlinks_pageid_idx ON mlinks (pageid);\n"
2258 "\n"
2259 "CREATE TABLE \"names\" (\n"
2260 " \"bits\" INTEGER NOT NULL,\n"
2261 " \"name\" TEXT NOT NULL,\n"
2262 " \"pageid\" INTEGER NOT NULL REFERENCES mpages(pageid) "
2263 "ON DELETE CASCADE\n"
2264 ");\n"
2265 "\n"
2266 "CREATE TABLE \"keys\" (\n"
2267 " \"bits\" INTEGER NOT NULL,\n"
2268 " \"key\" TEXT NOT NULL,\n"
2269 " \"pageid\" INTEGER NOT NULL REFERENCES mpages(pageid) "
2270 "ON DELETE CASCADE\n"
2271 ");\n"
2272 "CREATE INDEX keys_pageid_idx ON keys (pageid);\n";
2273
2274 if (SQLITE_OK != sqlite3_exec(db, sql, NULL, NULL, NULL)) {
2275 exitcode = (int)MANDOCLEVEL_SYSERR;
2276 say(MANDOC_DB, "%s", sqlite3_errmsg(db));
2277 return(0);
2278 }
2279
2280 prepare_statements:
2281 SQL_EXEC("PRAGMA foreign_keys = ON");
2282 sql = "DELETE FROM mpages WHERE pageid IN "
2283 "(SELECT pageid FROM mlinks WHERE "
2284 "sec=? AND arch=? AND name=?)";
2285 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_DELETE_PAGE], NULL);
2286 sql = "INSERT INTO mpages "
2287 "(desc,form) VALUES (?,?)";
2288 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_PAGE], NULL);
2289 sql = "INSERT INTO mlinks "
2290 "(sec,arch,name,pageid) VALUES (?,?,?,?)";
2291 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_LINK], NULL);
2292 sql = "INSERT INTO names "
2293 "(bits,name,pageid) VALUES (?,?,?)";
2294 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_NAME], NULL);
2295 sql = "INSERT INTO keys "
2296 "(bits,key,pageid) VALUES (?,?,?)";
2297 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_KEY], NULL);
2298
2299 #ifndef __APPLE__
2300 /*
2301 * When opening a new database, we can turn off
2302 * synchronous mode for much better performance.
2303 */
2304
2305 if (real)
2306 SQL_EXEC("PRAGMA synchronous = OFF");
2307 #endif
2308
2309 return(1);
2310 }
2311
2312 static void *
2313 hash_halloc(size_t sz, void *arg)
2314 {
2315
2316 return(mandoc_calloc(sz, 1));
2317 }
2318
2319 static void *
2320 hash_alloc(size_t sz, void *arg)
2321 {
2322
2323 return(mandoc_malloc(sz));
2324 }
2325
2326 static void
2327 hash_free(void *p, size_t sz, void *arg)
2328 {
2329
2330 free(p);
2331 }
2332
2333 static int
2334 set_basedir(const char *targetdir)
2335 {
2336 static char startdir[PATH_MAX];
2337 static int fd;
2338
2339 /*
2340 * Remember where we started by keeping a fd open to the origin
2341 * path component: throughout this utility, we chdir() a lot to
2342 * handle relative paths, and by doing this, we can return to
2343 * the starting point.
2344 */
2345 if ('\0' == *startdir) {
2346 if (NULL == getcwd(startdir, PATH_MAX)) {
2347 exitcode = (int)MANDOCLEVEL_SYSERR;
2348 if (NULL != targetdir)
2349 say("", "&getcwd");
2350 return(0);
2351 }
2352 if (-1 == (fd = open(startdir, O_RDONLY, 0))) {
2353 exitcode = (int)MANDOCLEVEL_SYSERR;
2354 say("", "&open %s", startdir);
2355 return(0);
2356 }
2357 if (NULL == targetdir)
2358 targetdir = startdir;
2359 } else {
2360 if (-1 == fd)
2361 return(0);
2362 if (-1 == fchdir(fd)) {
2363 close(fd);
2364 basedir[0] = '\0';
2365 exitcode = (int)MANDOCLEVEL_SYSERR;
2366 say("", "&chdir %s", startdir);
2367 return(0);
2368 }
2369 if (NULL == targetdir) {
2370 close(fd);
2371 return(1);
2372 }
2373 }
2374 if (NULL == realpath(targetdir, basedir)) {
2375 basedir[0] = '\0';
2376 exitcode = (int)MANDOCLEVEL_BADARG;
2377 say("", "&%s: realpath", targetdir);
2378 return(0);
2379 } else if (-1 == chdir(basedir)) {
2380 exitcode = (int)MANDOCLEVEL_BADARG;
2381 say("", "&chdir");
2382 return(0);
2383 }
2384 return(1);
2385 }
2386
2387 static void
2388 say(const char *file, const char *format, ...)
2389 {
2390 va_list ap;
2391 int use_errno;
2392
2393 if ('\0' != *basedir)
2394 fprintf(stderr, "%s", basedir);
2395 if ('\0' != *basedir && '\0' != *file)
2396 fputs("//", stderr);
2397 if ('\0' != *file)
2398 fprintf(stderr, "%s", file);
2399
2400 use_errno = 1;
2401 if (NULL != format) {
2402 switch (*format) {
2403 case '&':
2404 format++;
2405 break;
2406 case '\0':
2407 format = NULL;
2408 break;
2409 default:
2410 use_errno = 0;
2411 break;
2412 }
2413 }
2414 if (NULL != format) {
2415 if ('\0' != *basedir || '\0' != *file)
2416 fputs(": ", stderr);
2417 va_start(ap, format);
2418 vfprintf(stderr, format, ap);
2419 va_end(ap);
2420 }
2421 if (use_errno) {
2422 if ('\0' != *basedir || '\0' != *file || NULL != format)
2423 fputs(": ", stderr);
2424 perror(NULL);
2425 } else
2426 fputc('\n', stderr);
2427 }
A git conversion mirror of mandoc.bsd.lv