]> 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
Fix hardlink detection on platforms having padding in struct inodev,
[mandoc.git] / mandocdb.c
1 /* $Id: mandocdb.c,v 1.186 2015/03/13 00:19:41 schwarze Exp $ */
2 /*
3 * Copyright (c) 2011, 2012 Kristaps Dzonsons <kristaps@bsd.lv>
4 * Copyright (c) 2011-2015 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 #include "config.h"
19
20 #include <sys/types.h>
21 #include <sys/stat.h>
22 #include <sys/wait.h>
23
24 #include <assert.h>
25 #include <ctype.h>
26 #include <errno.h>
27 #include <fcntl.h>
28 #if HAVE_FTS
29 #include <fts.h>
30 #else
31 #include "compat_fts.h"
32 #endif
33 #include <getopt.h>
34 #include <limits.h>
35 #include <stddef.h>
36 #include <stdio.h>
37 #include <stdint.h>
38 #include <stdlib.h>
39 #include <string.h>
40 #include <unistd.h>
41
42 #if HAVE_OHASH
43 #include <ohash.h>
44 #else
45 #include "compat_ohash.h"
46 #endif
47 #include <sqlite3.h>
48
49 #include "mdoc.h"
50 #include "man.h"
51 #include "mandoc.h"
52 #include "mandoc_aux.h"
53 #include "manpath.h"
54 #include "mansearch.h"
55
56 extern int mansearch_keymax;
57 extern const char *const mansearch_keynames[];
58
59 #define SQL_EXEC(_v) \
60 if (SQLITE_OK != sqlite3_exec(db, (_v), NULL, NULL, NULL)) \
61 say("", "%s: %s", (_v), sqlite3_errmsg(db))
62 #define SQL_BIND_TEXT(_s, _i, _v) \
63 if (SQLITE_OK != sqlite3_bind_text \
64 ((_s), (_i)++, (_v), -1, SQLITE_STATIC)) \
65 say(mlink->file, "%s", sqlite3_errmsg(db))
66 #define SQL_BIND_INT(_s, _i, _v) \
67 if (SQLITE_OK != sqlite3_bind_int \
68 ((_s), (_i)++, (_v))) \
69 say(mlink->file, "%s", sqlite3_errmsg(db))
70 #define SQL_BIND_INT64(_s, _i, _v) \
71 if (SQLITE_OK != sqlite3_bind_int64 \
72 ((_s), (_i)++, (_v))) \
73 say(mlink->file, "%s", sqlite3_errmsg(db))
74 #define SQL_STEP(_s) \
75 if (SQLITE_DONE != sqlite3_step((_s))) \
76 say(mlink->file, "%s", sqlite3_errmsg(db))
77
78 enum op {
79 OP_DEFAULT = 0, /* new dbs from dir list or default config */
80 OP_CONFFILE, /* new databases from custom config file */
81 OP_UPDATE, /* delete/add entries in existing database */
82 OP_DELETE, /* delete entries from existing database */
83 OP_TEST /* change no databases, report potential problems */
84 };
85
86 struct str {
87 const struct mpage *mpage; /* if set, the owning parse */
88 uint64_t mask; /* bitmask in sequence */
89 char key[]; /* rendered text */
90 };
91
92 struct inodev {
93 ino_t st_ino;
94 dev_t st_dev;
95 };
96
97 struct mpage {
98 struct inodev inodev; /* used for hashing routine */
99 int64_t pageid; /* pageid in mpages SQL table */
100 char *sec; /* section from file content */
101 char *arch; /* architecture from file content */
102 char *title; /* title from file content */
103 char *desc; /* description from file content */
104 struct mlink *mlinks; /* singly linked list */
105 int form; /* format from file content */
106 int name_head_done;
107 };
108
109 struct mlink {
110 char file[PATH_MAX]; /* filename rel. to manpath */
111 char *dsec; /* section from directory */
112 char *arch; /* architecture from directory */
113 char *name; /* name from file name (not empty) */
114 char *fsec; /* section from file name suffix */
115 struct mlink *next; /* singly linked list */
116 struct mpage *mpage; /* parent */
117 int dform; /* format from directory */
118 int fform; /* format from file name suffix */
119 int gzip; /* filename has a .gz suffix */
120 };
121
122 enum stmt {
123 STMT_DELETE_PAGE = 0, /* delete mpage */
124 STMT_INSERT_PAGE, /* insert mpage */
125 STMT_INSERT_LINK, /* insert mlink */
126 STMT_INSERT_NAME, /* insert name */
127 STMT_SELECT_NAME, /* retrieve existing name flags */
128 STMT_INSERT_KEY, /* insert parsed key */
129 STMT__MAX
130 };
131
132 typedef int (*mdoc_fp)(struct mpage *, const struct mdoc_meta *,
133 const struct mdoc_node *);
134
135 struct mdoc_handler {
136 mdoc_fp fp; /* optional handler */
137 uint64_t mask; /* set unless handler returns 0 */
138 };
139
140 static void dbclose(int);
141 static void dbadd(struct mpage *);
142 static void dbadd_mlink(const struct mlink *mlink);
143 static void dbadd_mlink_name(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 *, void *);
149 static void *hash_calloc(size_t, 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 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_meta *,
159 const struct man_node *);
160 static void parse_mdoc(struct mpage *, const struct mdoc_meta *,
161 const struct mdoc_node *);
162 static int parse_mdoc_body(struct mpage *, const struct mdoc_meta *,
163 const struct mdoc_node *);
164 static int parse_mdoc_head(struct mpage *, const struct mdoc_meta *,
165 const struct mdoc_node *);
166 static int parse_mdoc_Fd(struct mpage *, const struct mdoc_meta *,
167 const struct mdoc_node *);
168 static void parse_mdoc_fname(struct mpage *, const struct mdoc_node *);
169 static int parse_mdoc_Fn(struct mpage *, const struct mdoc_meta *,
170 const struct mdoc_node *);
171 static int parse_mdoc_Fo(struct mpage *, const struct mdoc_meta *,
172 const struct mdoc_node *);
173 static int parse_mdoc_Nd(struct mpage *, const struct mdoc_meta *,
174 const struct mdoc_node *);
175 static int parse_mdoc_Nm(struct mpage *, const struct mdoc_meta *,
176 const struct mdoc_node *);
177 static int parse_mdoc_Sh(struct mpage *, const struct mdoc_meta *,
178 const struct mdoc_node *);
179 static int parse_mdoc_Xr(struct mpage *, const struct mdoc_meta *,
180 const struct mdoc_node *);
181 static void putkey(const struct mpage *, char *, uint64_t);
182 static void putkeys(const struct mpage *, char *, size_t, uint64_t);
183 static void putmdockey(const struct mpage *,
184 const struct mdoc_node *, uint64_t);
185 static int render_string(char **, size_t *);
186 static void say(const char *, const char *, ...);
187 static int set_basedir(const char *, int);
188 static int treescan(void);
189 static size_t utf8(unsigned int, char [7]);
190
191 static char tempfilename[32];
192 static char *progname;
193 static int nodb; /* no database changes */
194 static int mparse_options; /* abort the parse early */
195 static int use_all; /* use all found files */
196 static int debug; /* print what we're doing */
197 static int warnings; /* warn about crap */
198 static int write_utf8; /* write UTF-8 output; else ASCII */
199 static int exitcode; /* to be returned by main */
200 static enum op op; /* operational mode */
201 static char basedir[PATH_MAX]; /* current base directory */
202 static struct mchars *mchars; /* table of named characters */
203 static struct ohash mpages; /* table of distinct manual pages */
204 static struct ohash mlinks; /* table of directory entries */
205 static struct ohash names; /* table of all names */
206 static struct ohash strings; /* table of all strings */
207 static sqlite3 *db = NULL; /* current database */
208 static sqlite3_stmt *stmts[STMT__MAX]; /* current statements */
209 static uint64_t name_mask;
210
211 static const struct mdoc_handler mdocs[MDOC_MAX] = {
212 { NULL, 0 }, /* Ap */
213 { NULL, 0 }, /* Dd */
214 { NULL, 0 }, /* Dt */
215 { NULL, 0 }, /* Os */
216 { parse_mdoc_Sh, TYPE_Sh }, /* Sh */
217 { parse_mdoc_head, TYPE_Ss }, /* Ss */
218 { NULL, 0 }, /* Pp */
219 { NULL, 0 }, /* D1 */
220 { NULL, 0 }, /* Dl */
221 { NULL, 0 }, /* Bd */
222 { NULL, 0 }, /* Ed */
223 { NULL, 0 }, /* Bl */
224 { NULL, 0 }, /* El */
225 { NULL, 0 }, /* It */
226 { NULL, 0 }, /* Ad */
227 { NULL, TYPE_An }, /* An */
228 { NULL, TYPE_Ar }, /* Ar */
229 { NULL, TYPE_Cd }, /* Cd */
230 { NULL, TYPE_Cm }, /* Cm */
231 { NULL, TYPE_Dv }, /* Dv */
232 { NULL, TYPE_Er }, /* Er */
233 { NULL, TYPE_Ev }, /* Ev */
234 { NULL, 0 }, /* Ex */
235 { NULL, TYPE_Fa }, /* Fa */
236 { parse_mdoc_Fd, 0 }, /* Fd */
237 { NULL, TYPE_Fl }, /* Fl */
238 { parse_mdoc_Fn, 0 }, /* Fn */
239 { NULL, TYPE_Ft }, /* Ft */
240 { NULL, TYPE_Ic }, /* Ic */
241 { NULL, TYPE_In }, /* In */
242 { NULL, TYPE_Li }, /* Li */
243 { parse_mdoc_Nd, 0 }, /* Nd */
244 { parse_mdoc_Nm, 0 }, /* Nm */
245 { NULL, 0 }, /* Op */
246 { NULL, 0 }, /* Ot */
247 { NULL, TYPE_Pa }, /* Pa */
248 { NULL, 0 }, /* Rv */
249 { NULL, TYPE_St }, /* St */
250 { NULL, TYPE_Va }, /* Va */
251 { parse_mdoc_body, TYPE_Va }, /* Vt */
252 { parse_mdoc_Xr, 0 }, /* Xr */
253 { NULL, 0 }, /* %A */
254 { NULL, 0 }, /* %B */
255 { NULL, 0 }, /* %D */
256 { NULL, 0 }, /* %I */
257 { NULL, 0 }, /* %J */
258 { NULL, 0 }, /* %N */
259 { NULL, 0 }, /* %O */
260 { NULL, 0 }, /* %P */
261 { NULL, 0 }, /* %R */
262 { NULL, 0 }, /* %T */
263 { NULL, 0 }, /* %V */
264 { NULL, 0 }, /* Ac */
265 { NULL, 0 }, /* Ao */
266 { NULL, 0 }, /* Aq */
267 { NULL, TYPE_At }, /* At */
268 { NULL, 0 }, /* Bc */
269 { NULL, 0 }, /* Bf */
270 { NULL, 0 }, /* Bo */
271 { NULL, 0 }, /* Bq */
272 { NULL, TYPE_Bsx }, /* Bsx */
273 { NULL, TYPE_Bx }, /* Bx */
274 { NULL, 0 }, /* Db */
275 { NULL, 0 }, /* Dc */
276 { NULL, 0 }, /* Do */
277 { NULL, 0 }, /* Dq */
278 { NULL, 0 }, /* Ec */
279 { NULL, 0 }, /* Ef */
280 { NULL, TYPE_Em }, /* Em */
281 { NULL, 0 }, /* Eo */
282 { NULL, TYPE_Fx }, /* Fx */
283 { NULL, TYPE_Ms }, /* Ms */
284 { NULL, 0 }, /* No */
285 { NULL, 0 }, /* Ns */
286 { NULL, TYPE_Nx }, /* Nx */
287 { NULL, TYPE_Ox }, /* Ox */
288 { NULL, 0 }, /* Pc */
289 { NULL, 0 }, /* Pf */
290 { NULL, 0 }, /* Po */
291 { NULL, 0 }, /* Pq */
292 { NULL, 0 }, /* Qc */
293 { NULL, 0 }, /* Ql */
294 { NULL, 0 }, /* Qo */
295 { NULL, 0 }, /* Qq */
296 { NULL, 0 }, /* Re */
297 { NULL, 0 }, /* Rs */
298 { NULL, 0 }, /* Sc */
299 { NULL, 0 }, /* So */
300 { NULL, 0 }, /* Sq */
301 { NULL, 0 }, /* Sm */
302 { NULL, 0 }, /* Sx */
303 { NULL, TYPE_Sy }, /* Sy */
304 { NULL, TYPE_Tn }, /* Tn */
305 { NULL, 0 }, /* Ux */
306 { NULL, 0 }, /* Xc */
307 { NULL, 0 }, /* Xo */
308 { parse_mdoc_Fo, 0 }, /* Fo */
309 { NULL, 0 }, /* Fc */
310 { NULL, 0 }, /* Oo */
311 { NULL, 0 }, /* Oc */
312 { NULL, 0 }, /* Bk */
313 { NULL, 0 }, /* Ek */
314 { NULL, 0 }, /* Bt */
315 { NULL, 0 }, /* Hf */
316 { NULL, 0 }, /* Fr */
317 { NULL, 0 }, /* Ud */
318 { NULL, TYPE_Lb }, /* Lb */
319 { NULL, 0 }, /* Lp */
320 { NULL, TYPE_Lk }, /* Lk */
321 { NULL, TYPE_Mt }, /* Mt */
322 { NULL, 0 }, /* Brq */
323 { NULL, 0 }, /* Bro */
324 { NULL, 0 }, /* Brc */
325 { NULL, 0 }, /* %C */
326 { NULL, 0 }, /* Es */
327 { NULL, 0 }, /* En */
328 { NULL, TYPE_Dx }, /* Dx */
329 { NULL, 0 }, /* %Q */
330 { NULL, 0 }, /* br */
331 { NULL, 0 }, /* sp */
332 { NULL, 0 }, /* %U */
333 { NULL, 0 }, /* Ta */
334 { NULL, 0 }, /* ll */
335 };
336
337
338 int
339 mandocdb(int argc, char *argv[])
340 {
341 int ch, i;
342 size_t j, sz;
343 const char *path_arg;
344 struct manpaths dirs;
345 struct mparse *mp;
346 struct ohash_info mpages_info, mlinks_info;
347
348 memset(stmts, 0, STMT__MAX * sizeof(sqlite3_stmt *));
349 memset(&dirs, 0, sizeof(struct manpaths));
350
351 mpages_info.alloc = mlinks_info.alloc = hash_alloc;
352 mpages_info.calloc = mlinks_info.calloc = hash_calloc;
353 mpages_info.free = mlinks_info.free = hash_free;
354 mpages_info.data = mlinks_info.data = NULL;
355
356 mpages_info.key_offset = offsetof(struct mpage, inodev);
357 mlinks_info.key_offset = offsetof(struct mlink, file);
358
359 progname = strrchr(argv[0], '/');
360 if (progname == NULL)
361 progname = argv[0];
362 else
363 ++progname;
364
365 /*
366 * We accept a few different invocations.
367 * The CHECKOP macro makes sure that invocation styles don't
368 * clobber each other.
369 */
370 #define CHECKOP(_op, _ch) do \
371 if (OP_DEFAULT != (_op)) { \
372 fprintf(stderr, "%s: -%c: Conflicting option\n", \
373 progname, (_ch)); \
374 goto usage; \
375 } while (/*CONSTCOND*/0)
376
377 path_arg = NULL;
378 op = OP_DEFAULT;
379
380 while (-1 != (ch = getopt(argc, argv, "aC:Dd:npQT:tu:v")))
381 switch (ch) {
382 case 'a':
383 use_all = 1;
384 break;
385 case 'C':
386 CHECKOP(op, ch);
387 path_arg = optarg;
388 op = OP_CONFFILE;
389 break;
390 case 'D':
391 debug++;
392 break;
393 case 'd':
394 CHECKOP(op, ch);
395 path_arg = optarg;
396 op = OP_UPDATE;
397 break;
398 case 'n':
399 nodb = 1;
400 break;
401 case 'p':
402 warnings = 1;
403 break;
404 case 'Q':
405 mparse_options |= MPARSE_QUICK;
406 break;
407 case 'T':
408 if (strcmp(optarg, "utf8")) {
409 fprintf(stderr, "%s: -T%s: "
410 "Unsupported output format\n",
411 progname, optarg);
412 goto usage;
413 }
414 write_utf8 = 1;
415 break;
416 case 't':
417 CHECKOP(op, ch);
418 dup2(STDOUT_FILENO, STDERR_FILENO);
419 op = OP_TEST;
420 nodb = warnings = 1;
421 break;
422 case 'u':
423 CHECKOP(op, ch);
424 path_arg = optarg;
425 op = OP_DELETE;
426 break;
427 case 'v':
428 /* Compatibility with espie@'s makewhatis. */
429 break;
430 default:
431 goto usage;
432 }
433
434 argc -= optind;
435 argv += optind;
436
437 if (OP_CONFFILE == op && argc > 0) {
438 fprintf(stderr, "%s: -C: Too many arguments\n",
439 progname);
440 goto usage;
441 }
442
443 exitcode = (int)MANDOCLEVEL_OK;
444 mchars = mchars_alloc();
445 mp = mparse_alloc(mparse_options, MANDOCLEVEL_BADARG, NULL,
446 mchars, NULL);
447 ohash_init(&mpages, 6, &mpages_info);
448 ohash_init(&mlinks, 6, &mlinks_info);
449
450 if (OP_UPDATE == op || OP_DELETE == op || OP_TEST == op) {
451
452 /*
453 * Most of these deal with a specific directory.
454 * Jump into that directory first.
455 */
456 if (OP_TEST != op && 0 == set_basedir(path_arg, 1))
457 goto out;
458
459 if (dbopen(1)) {
460 /*
461 * The existing database is usable. Process
462 * all files specified on the command-line.
463 */
464 use_all = 1;
465 for (i = 0; i < argc; i++)
466 filescan(argv[i]);
467 if (OP_TEST != op)
468 dbprune();
469 } else {
470 /*
471 * Database missing or corrupt.
472 * Recreate from scratch.
473 */
474 exitcode = (int)MANDOCLEVEL_OK;
475 op = OP_DEFAULT;
476 if (0 == treescan())
477 goto out;
478 if (0 == dbopen(0))
479 goto out;
480 }
481 if (OP_DELETE != op)
482 mpages_merge(mp);
483 dbclose(OP_DEFAULT == op ? 0 : 1);
484 } else {
485 /*
486 * If we have arguments, use them as our manpaths.
487 * If we don't, grok from manpath(1) or however else
488 * manpath_parse() wants to do it.
489 */
490 if (argc > 0) {
491 dirs.paths = mandoc_reallocarray(NULL,
492 argc, sizeof(char *));
493 dirs.sz = (size_t)argc;
494 for (i = 0; i < argc; i++)
495 dirs.paths[i] = mandoc_strdup(argv[i]);
496 } else
497 manpath_parse(&dirs, path_arg, NULL, NULL);
498
499 if (0 == dirs.sz) {
500 exitcode = (int)MANDOCLEVEL_BADARG;
501 say("", "Empty manpath");
502 }
503
504 /*
505 * First scan the tree rooted at a base directory, then
506 * build a new database and finally move it into place.
507 * Ignore zero-length directories and strip trailing
508 * slashes.
509 */
510 for (j = 0; j < dirs.sz; j++) {
511 sz = strlen(dirs.paths[j]);
512 if (sz && '/' == dirs.paths[j][sz - 1])
513 dirs.paths[j][--sz] = '\0';
514 if (0 == sz)
515 continue;
516
517 if (j) {
518 ohash_init(&mpages, 6, &mpages_info);
519 ohash_init(&mlinks, 6, &mlinks_info);
520 }
521
522 if (0 == set_basedir(dirs.paths[j], argc > 0))
523 continue;
524 if (0 == treescan())
525 continue;
526 if (0 == dbopen(0))
527 continue;
528
529 mpages_merge(mp);
530 if (warnings && !nodb &&
531 ! (MPARSE_QUICK & mparse_options))
532 names_check();
533 dbclose(0);
534
535 if (j + 1 < dirs.sz) {
536 mpages_free();
537 ohash_delete(&mpages);
538 ohash_delete(&mlinks);
539 }
540 }
541 }
542 out:
543 manpath_free(&dirs);
544 mparse_free(mp);
545 mchars_free(mchars);
546 mpages_free();
547 ohash_delete(&mpages);
548 ohash_delete(&mlinks);
549 return(exitcode);
550 usage:
551 fprintf(stderr, "usage: %s [-aDnpQ] [-C file] [-Tutf8]\n"
552 " %s [-aDnpQ] [-Tutf8] dir ...\n"
553 " %s [-DnpQ] [-Tutf8] -d dir [file ...]\n"
554 " %s [-Dnp] -u dir [file ...]\n"
555 " %s [-Q] -t file ...\n",
556 progname, progname, progname,
557 progname, progname);
558
559 return((int)MANDOCLEVEL_BADARG);
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: accomodate 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 dform, gzip;
584 char *dsec, *arch, *fsec, *cp;
585 const char *path;
586 const char *argv[2];
587
588 argv[0] = ".";
589 argv[1] = (char *)NULL;
590
591 f = fts_open((char * const *)argv,
592 FTS_PHYSICAL | FTS_NOCHDIR, NULL);
593 if (NULL == f) {
594 exitcode = (int)MANDOCLEVEL_SYSERR;
595 say("", "&fts_open");
596 return(0);
597 }
598
599 dsec = arch = NULL;
600 dform = FORM_NONE;
601
602 while (NULL != (ff = fts_read(f))) {
603 path = ff->fts_path + 2;
604 switch (ff->fts_info) {
605
606 /*
607 * Symbolic links require various sanity checks,
608 * then get handled just like regular files.
609 */
610 case FTS_SL:
611 if (NULL == realpath(path, buf)) {
612 if (warnings)
613 say(path, "&realpath");
614 continue;
615 }
616 if (strstr(buf, basedir) != buf
617 #ifdef HOMEBREWDIR
618 && strstr(buf, HOMEBREWDIR) != buf
619 #endif
620 ) {
621 if (warnings) say("",
622 "%s: outside base directory", buf);
623 continue;
624 }
625 /* Use logical inode to avoid mpages dupe. */
626 if (-1 == stat(path, ff->fts_statp)) {
627 if (warnings)
628 say(path, "&stat");
629 continue;
630 }
631 /* FALLTHROUGH */
632
633 /*
634 * If we're a regular file, add an mlink by using the
635 * stored directory data and handling the filename.
636 */
637 case FTS_F:
638 if (0 == strcmp(path, MANDOC_DB))
639 continue;
640 if ( ! use_all && ff->fts_level < 2) {
641 if (warnings)
642 say(path, "Extraneous file");
643 continue;
644 }
645 gzip = 0;
646 fsec = NULL;
647 while (NULL == fsec) {
648 fsec = strrchr(ff->fts_name, '.');
649 if (NULL == fsec || strcmp(fsec+1, "gz"))
650 break;
651 gzip = 1;
652 *fsec = '\0';
653 fsec = NULL;
654 }
655 if (NULL == fsec) {
656 if ( ! use_all) {
657 if (warnings)
658 say(path,
659 "No filename suffix");
660 continue;
661 }
662 } else if (0 == strcmp(++fsec, "html")) {
663 if (warnings)
664 say(path, "Skip html");
665 continue;
666 } else if (0 == strcmp(fsec, "ps")) {
667 if (warnings)
668 say(path, "Skip ps");
669 continue;
670 } else if (0 == strcmp(fsec, "pdf")) {
671 if (warnings)
672 say(path, "Skip pdf");
673 continue;
674 } else if ( ! use_all &&
675 ((FORM_SRC == dform &&
676 strncmp(fsec, dsec, strlen(dsec))) ||
677 (FORM_CAT == dform && strcmp(fsec, "0")))) {
678 if (warnings)
679 say(path, "Wrong filename suffix");
680 continue;
681 } else
682 fsec[-1] = '\0';
683
684 mlink = mandoc_calloc(1, sizeof(struct mlink));
685 if (strlcpy(mlink->file, path,
686 sizeof(mlink->file)) >=
687 sizeof(mlink->file)) {
688 say(path, "Filename too long");
689 free(mlink);
690 continue;
691 }
692 mlink->dform = dform;
693 mlink->dsec = dsec;
694 mlink->arch = arch;
695 mlink->name = ff->fts_name;
696 mlink->fsec = fsec;
697 mlink->gzip = gzip;
698 mlink_add(mlink, ff->fts_statp);
699 continue;
700
701 case FTS_D:
702 /* FALLTHROUGH */
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 (FTS_DP == ff->fts_info)
724 break;
725
726 if (0 == strncmp(cp, "man", 3)) {
727 dform = FORM_SRC;
728 dsec = cp + 3;
729 } else if (0 == strncmp(cp, "cat", 3)) {
730 dform = FORM_CAT;
731 dsec = cp + 3;
732 } else {
733 dform = FORM_NONE;
734 dsec = NULL;
735 }
736
737 if (NULL != dsec || use_all)
738 break;
739
740 if (warnings)
741 say(path, "Unknown directory part");
742 fts_set(f, ff, FTS_SKIP);
743 break;
744 case 2:
745 /*
746 * Possibly our architecture.
747 * If we're descending, keep tabs on it.
748 */
749 if (FTS_DP != ff->fts_info && NULL != dsec)
750 arch = ff->fts_name;
751 else
752 arch = NULL;
753 break;
754 default:
755 if (FTS_DP == ff->fts_info || use_all)
756 break;
757 if (warnings)
758 say(path, "Extraneous directory part");
759 fts_set(f, ff, FTS_SKIP);
760 break;
761 }
762 }
763
764 fts_close(f);
765 return(1);
766 }
767
768 /*
769 * Add a file to the mlinks table.
770 * Do not verify that it's a "valid" looking manpage (we'll do that
771 * later).
772 *
773 * Try to infer the manual section, architecture, and page name from the
774 * path, assuming it looks like
775 *
776 * [./]man*[/<arch>]/<name>.<section>
777 * or
778 * [./]cat<section>[/<arch>]/<name>.0
779 *
780 * See treescan() for the fts(3) version of this.
781 */
782 static void
783 filescan(const char *file)
784 {
785 char buf[PATH_MAX];
786 struct stat st;
787 struct mlink *mlink;
788 char *p, *start;
789
790 assert(use_all);
791
792 if (0 == strncmp(file, "./", 2))
793 file += 2;
794
795 /*
796 * We have to do lstat(2) before realpath(3) loses
797 * the information whether this is a symbolic link.
798 * We need to know that because for symbolic links,
799 * we want to use the orginal file name, while for
800 * regular files, we want to use the real path.
801 */
802 if (-1 == lstat(file, &st)) {
803 exitcode = (int)MANDOCLEVEL_BADARG;
804 say(file, "&lstat");
805 return;
806 } else if (0 == ((S_IFREG | S_IFLNK) & st.st_mode)) {
807 exitcode = (int)MANDOCLEVEL_BADARG;
808 say(file, "Not a regular file");
809 return;
810 }
811
812 /*
813 * We have to resolve the file name to the real path
814 * in any case for the base directory check.
815 */
816 if (NULL == realpath(file, buf)) {
817 exitcode = (int)MANDOCLEVEL_BADARG;
818 say(file, "&realpath");
819 return;
820 }
821
822 if (OP_TEST == op)
823 start = buf;
824 else if (strstr(buf, basedir) == buf)
825 start = buf + strlen(basedir);
826 #ifdef HOMEBREWDIR
827 else if (strstr(buf, HOMEBREWDIR) == buf)
828 start = buf;
829 #endif
830 else {
831 exitcode = (int)MANDOCLEVEL_BADARG;
832 say("", "%s: outside base directory", buf);
833 return;
834 }
835
836 /*
837 * Now we are sure the file is inside our tree.
838 * If it is a symbolic link, ignore the real path
839 * and use the original name.
840 * This implies passing stuff like "cat1/../man1/foo.1"
841 * on the command line won't work. So don't do that.
842 * Note the stat(2) can still fail if the link target
843 * doesn't exist.
844 */
845 if (S_IFLNK & st.st_mode) {
846 if (-1 == stat(buf, &st)) {
847 exitcode = (int)MANDOCLEVEL_BADARG;
848 say(file, "&stat");
849 return;
850 }
851 if (strlcpy(buf, file, sizeof(buf)) >= sizeof(buf)) {
852 say(file, "Filename too long");
853 return;
854 }
855 start = buf;
856 if (OP_TEST != op && strstr(buf, basedir) == buf)
857 start += strlen(basedir);
858 }
859
860 mlink = mandoc_calloc(1, sizeof(struct mlink));
861 mlink->dform = FORM_NONE;
862 if (strlcpy(mlink->file, start, sizeof(mlink->file)) >=
863 sizeof(mlink->file)) {
864 say(start, "Filename too long");
865 free(mlink);
866 return;
867 }
868
869 /*
870 * First try to guess our directory structure.
871 * If we find a separator, try to look for man* or cat*.
872 * If we find one of these and what's underneath is a directory,
873 * assume it's an architecture.
874 */
875 if (NULL != (p = strchr(start, '/'))) {
876 *p++ = '\0';
877 if (0 == strncmp(start, "man", 3)) {
878 mlink->dform = FORM_SRC;
879 mlink->dsec = start + 3;
880 } else if (0 == strncmp(start, "cat", 3)) {
881 mlink->dform = FORM_CAT;
882 mlink->dsec = start + 3;
883 }
884
885 start = p;
886 if (NULL != mlink->dsec && NULL != (p = strchr(start, '/'))) {
887 *p++ = '\0';
888 mlink->arch = start;
889 start = p;
890 }
891 }
892
893 /*
894 * Now check the file suffix.
895 * Suffix of `.0' indicates a catpage, `.1-9' is a manpage.
896 */
897 p = strrchr(start, '\0');
898 while (p-- > start && '/' != *p && '.' != *p)
899 /* Loop. */ ;
900
901 if ('.' == *p) {
902 *p++ = '\0';
903 mlink->fsec = p;
904 }
905
906 /*
907 * Now try to parse the name.
908 * Use the filename portion of the path.
909 */
910 mlink->name = start;
911 if (NULL != (p = strrchr(start, '/'))) {
912 mlink->name = p + 1;
913 *p = '\0';
914 }
915 mlink_add(mlink, &st);
916 }
917
918 static void
919 mlink_add(struct mlink *mlink, const struct stat *st)
920 {
921 struct inodev inodev;
922 struct mpage *mpage;
923 unsigned int slot;
924
925 assert(NULL != mlink->file);
926
927 mlink->dsec = mandoc_strdup(mlink->dsec ? mlink->dsec : "");
928 mlink->arch = mandoc_strdup(mlink->arch ? mlink->arch : "");
929 mlink->name = mandoc_strdup(mlink->name ? mlink->name : "");
930 mlink->fsec = mandoc_strdup(mlink->fsec ? mlink->fsec : "");
931
932 if ('0' == *mlink->fsec) {
933 free(mlink->fsec);
934 mlink->fsec = mandoc_strdup(mlink->dsec);
935 mlink->fform = FORM_CAT;
936 } else if ('1' <= *mlink->fsec && '9' >= *mlink->fsec)
937 mlink->fform = FORM_SRC;
938 else
939 mlink->fform = FORM_NONE;
940
941 slot = ohash_qlookup(&mlinks, mlink->file);
942 assert(NULL == ohash_find(&mlinks, slot));
943 ohash_insert(&mlinks, slot, mlink);
944
945 memset(&inodev, 0, sizeof(inodev)); /* Clear padding. */
946 inodev.st_ino = st->st_ino;
947 inodev.st_dev = st->st_dev;
948 slot = ohash_lookup_memory(&mpages, (char *)&inodev,
949 sizeof(struct inodev), inodev.st_ino);
950 mpage = ohash_find(&mpages, slot);
951 if (NULL == mpage) {
952 mpage = mandoc_calloc(1, sizeof(struct mpage));
953 mpage->inodev.st_ino = inodev.st_ino;
954 mpage->inodev.st_dev = inodev.st_dev;
955 ohash_insert(&mpages, slot, mpage);
956 } else
957 mlink->next = mpage->mlinks;
958 mpage->mlinks = mlink;
959 mlink->mpage = mpage;
960 }
961
962 static void
963 mlink_free(struct mlink *mlink)
964 {
965
966 free(mlink->dsec);
967 free(mlink->arch);
968 free(mlink->name);
969 free(mlink->fsec);
970 free(mlink);
971 }
972
973 static void
974 mpages_free(void)
975 {
976 struct mpage *mpage;
977 struct mlink *mlink;
978 unsigned int slot;
979
980 mpage = ohash_first(&mpages, &slot);
981 while (NULL != mpage) {
982 while (NULL != (mlink = mpage->mlinks)) {
983 mpage->mlinks = mlink->next;
984 mlink_free(mlink);
985 }
986 free(mpage->sec);
987 free(mpage->arch);
988 free(mpage->title);
989 free(mpage->desc);
990 free(mpage);
991 mpage = ohash_next(&mpages, &slot);
992 }
993 }
994
995 /*
996 * For each mlink to the mpage, check whether the path looks like
997 * it is formatted, and if it does, check whether a source manual
998 * exists by the same name, ignoring the suffix.
999 * If both conditions hold, drop the mlink.
1000 */
1001 static void
1002 mlinks_undupe(struct mpage *mpage)
1003 {
1004 char buf[PATH_MAX];
1005 struct mlink **prev;
1006 struct mlink *mlink;
1007 char *bufp;
1008
1009 mpage->form = FORM_CAT;
1010 prev = &mpage->mlinks;
1011 while (NULL != (mlink = *prev)) {
1012 if (FORM_CAT != mlink->dform) {
1013 mpage->form = FORM_NONE;
1014 goto nextlink;
1015 }
1016 (void)strlcpy(buf, mlink->file, sizeof(buf));
1017 bufp = strstr(buf, "cat");
1018 assert(NULL != bufp);
1019 memcpy(bufp, "man", 3);
1020 if (NULL != (bufp = strrchr(buf, '.')))
1021 *++bufp = '\0';
1022 (void)strlcat(buf, mlink->dsec, sizeof(buf));
1023 if (NULL == ohash_find(&mlinks,
1024 ohash_qlookup(&mlinks, buf)))
1025 goto nextlink;
1026 if (warnings)
1027 say(mlink->file, "Man source exists: %s", buf);
1028 if (use_all)
1029 goto nextlink;
1030 *prev = mlink->next;
1031 mlink_free(mlink);
1032 continue;
1033 nextlink:
1034 prev = &(*prev)->next;
1035 }
1036 }
1037
1038 static void
1039 mlink_check(struct mpage *mpage, struct mlink *mlink)
1040 {
1041 struct str *str;
1042 unsigned int slot;
1043
1044 /*
1045 * Check whether the manual section given in a file
1046 * agrees with the directory where the file is located.
1047 * Some manuals have suffixes like (3p) on their
1048 * section number either inside the file or in the
1049 * directory name, some are linked into more than one
1050 * section, like encrypt(1) = makekey(8).
1051 */
1052
1053 if (FORM_SRC == mpage->form &&
1054 strcasecmp(mpage->sec, mlink->dsec))
1055 say(mlink->file, "Section \"%s\" manual in %s directory",
1056 mpage->sec, mlink->dsec);
1057
1058 /*
1059 * Manual page directories exist for each kernel
1060 * architecture as returned by machine(1).
1061 * However, many manuals only depend on the
1062 * application architecture as returned by arch(1).
1063 * For example, some (2/ARM) manuals are shared
1064 * across the "armish" and "zaurus" kernel
1065 * architectures.
1066 * A few manuals are even shared across completely
1067 * different architectures, for example fdformat(1)
1068 * on amd64, i386, sparc, and sparc64.
1069 */
1070
1071 if (strcasecmp(mpage->arch, mlink->arch))
1072 say(mlink->file, "Architecture \"%s\" manual in "
1073 "\"%s\" directory", mpage->arch, mlink->arch);
1074
1075 /*
1076 * XXX
1077 * parse_cat() doesn't set NAME_TITLE yet.
1078 */
1079
1080 if (FORM_CAT == mpage->form)
1081 return;
1082
1083 /*
1084 * Check whether this mlink
1085 * appears as a name in the NAME section.
1086 */
1087
1088 slot = ohash_qlookup(&names, mlink->name);
1089 str = ohash_find(&names, slot);
1090 assert(NULL != str);
1091 if ( ! (NAME_TITLE & str->mask))
1092 say(mlink->file, "Name missing in NAME section");
1093 }
1094
1095 /*
1096 * Run through the files in the global vector "mpages"
1097 * and add them to the database specified in "basedir".
1098 *
1099 * This handles the parsing scheme itself, using the cues of directory
1100 * and filename to determine whether the file is parsable or not.
1101 */
1102 static void
1103 mpages_merge(struct mparse *mp)
1104 {
1105 char any[] = "any";
1106 struct ohash_info str_info;
1107 struct mpage *mpage, *mpage_dest;
1108 struct mlink *mlink, *mlink_dest;
1109 struct mdoc *mdoc;
1110 struct man *man;
1111 char *sodest;
1112 char *cp;
1113 int fd;
1114 unsigned int pslot;
1115
1116 str_info.alloc = hash_alloc;
1117 str_info.calloc = hash_calloc;
1118 str_info.free = hash_free;
1119 str_info.data = NULL;
1120 str_info.key_offset = offsetof(struct str, key);
1121
1122 if ( ! nodb)
1123 SQL_EXEC("BEGIN TRANSACTION");
1124
1125 mpage = ohash_first(&mpages, &pslot);
1126 while (mpage != NULL) {
1127 mlinks_undupe(mpage);
1128 if ((mlink = mpage->mlinks) == NULL) {
1129 mpage = ohash_next(&mpages, &pslot);
1130 continue;
1131 }
1132
1133 name_mask = NAME_MASK;
1134 ohash_init(&names, 4, &str_info);
1135 ohash_init(&strings, 6, &str_info);
1136 mparse_reset(mp);
1137 mdoc = NULL;
1138 man = NULL;
1139 sodest = NULL;
1140
1141 mparse_open(mp, &fd, mlink->file);
1142 if (fd == -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 mparse_result(mp, &mdoc, &man, &sodest);
1154 }
1155
1156 if (sodest != NULL) {
1157 mlink_dest = ohash_find(&mlinks,
1158 ohash_qlookup(&mlinks, sodest));
1159 if (mlink_dest == NULL) {
1160 mandoc_asprintf(&cp, "%s.gz", sodest);
1161 mlink_dest = ohash_find(&mlinks,
1162 ohash_qlookup(&mlinks, cp));
1163 free(cp);
1164 }
1165 if (mlink_dest != NULL) {
1166
1167 /* The .so target exists. */
1168
1169 mpage_dest = mlink_dest->mpage;
1170 while (1) {
1171 mlink->mpage = mpage_dest;
1172
1173 /*
1174 * If the target was already
1175 * processed, add the links
1176 * to the database now.
1177 * Otherwise, this will
1178 * happen when we come
1179 * to the target.
1180 */
1181
1182 if (mpage_dest->pageid)
1183 dbadd_mlink_name(mlink);
1184
1185 if (mlink->next == NULL)
1186 break;
1187 mlink = mlink->next;
1188 }
1189
1190 /* Move all links to the target. */
1191
1192 mlink->next = mlink_dest->next;
1193 mlink_dest->next = mpage->mlinks;
1194 mpage->mlinks = NULL;
1195 }
1196 goto nextpage;
1197 } else if (mdoc != NULL) {
1198 mpage->form = FORM_SRC;
1199 mpage->sec = mdoc_meta(mdoc)->msec;
1200 mpage->sec = mandoc_strdup(
1201 mpage->sec == NULL ? "" : mpage->sec);
1202 mpage->arch = mdoc_meta(mdoc)->arch;
1203 mpage->arch = mandoc_strdup(
1204 mpage->arch == NULL ? "" : mpage->arch);
1205 mpage->title =
1206 mandoc_strdup(mdoc_meta(mdoc)->title);
1207 } else if (man != NULL) {
1208 mpage->form = FORM_SRC;
1209 mpage->sec = mandoc_strdup(man_meta(man)->msec);
1210 mpage->arch = mandoc_strdup(mlink->arch);
1211 mpage->title = mandoc_strdup(man_meta(man)->title);
1212 } else {
1213 mpage->form = FORM_CAT;
1214 mpage->sec = mandoc_strdup(mlink->dsec);
1215 mpage->arch = mandoc_strdup(mlink->arch);
1216 mpage->title = mandoc_strdup(mlink->name);
1217 }
1218 putkey(mpage, mpage->sec, TYPE_sec);
1219 if (*mpage->arch != '\0')
1220 putkey(mpage, mpage->arch, TYPE_arch);
1221
1222 for ( ; mlink != NULL; mlink = mlink->next) {
1223 if ('\0' != *mlink->dsec)
1224 putkey(mpage, mlink->dsec, TYPE_sec);
1225 if ('\0' != *mlink->fsec)
1226 putkey(mpage, mlink->fsec, TYPE_sec);
1227 putkey(mpage, '\0' == *mlink->arch ?
1228 any : mlink->arch, TYPE_arch);
1229 putkey(mpage, mlink->name, NAME_FILE);
1230 }
1231
1232 assert(mpage->desc == NULL);
1233 if (mdoc != NULL)
1234 parse_mdoc(mpage, mdoc_meta(mdoc), mdoc_node(mdoc));
1235 else if (man != NULL)
1236 parse_man(mpage, man_meta(man), man_node(man));
1237 else
1238 parse_cat(mpage, fd);
1239 if (mpage->desc == NULL)
1240 mpage->desc = mandoc_strdup(mpage->mlinks->name);
1241
1242 if (warnings && !use_all)
1243 for (mlink = mpage->mlinks; mlink;
1244 mlink = mlink->next)
1245 mlink_check(mpage, mlink);
1246
1247 dbadd(mpage);
1248 mlink = mpage->mlinks;
1249
1250 nextpage:
1251 if (mparse_wait(mp) != MANDOCLEVEL_OK) {
1252 exitcode = (int)MANDOCLEVEL_SYSERR;
1253 say(mlink->file, "&wait gunzip");
1254 }
1255 ohash_delete(&strings);
1256 ohash_delete(&names);
1257 mpage = ohash_next(&mpages, &pslot);
1258 }
1259
1260 if (0 == nodb)
1261 SQL_EXEC("END TRANSACTION");
1262 }
1263
1264 static void
1265 names_check(void)
1266 {
1267 sqlite3_stmt *stmt;
1268 const char *name, *sec, *arch, *key;
1269 int irc;
1270
1271 sqlite3_prepare_v2(db,
1272 "SELECT name, sec, arch, key FROM ("
1273 "SELECT name AS key, pageid FROM names "
1274 "WHERE bits & ? AND NOT EXISTS ("
1275 "SELECT pageid FROM mlinks "
1276 "WHERE mlinks.pageid == names.pageid "
1277 "AND mlinks.name == names.name"
1278 ")"
1279 ") JOIN ("
1280 "SELECT sec, arch, name, pageid FROM mlinks "
1281 "GROUP BY pageid"
1282 ") USING (pageid);",
1283 -1, &stmt, NULL);
1284
1285 if (SQLITE_OK != sqlite3_bind_int64(stmt, 1, NAME_TITLE))
1286 say("", "%s", sqlite3_errmsg(db));
1287
1288 while (SQLITE_ROW == (irc = sqlite3_step(stmt))) {
1289 name = (const char *)sqlite3_column_text(stmt, 0);
1290 sec = (const char *)sqlite3_column_text(stmt, 1);
1291 arch = (const char *)sqlite3_column_text(stmt, 2);
1292 key = (const char *)sqlite3_column_text(stmt, 3);
1293 say("", "%s(%s%s%s) lacks mlink \"%s\"", name, sec,
1294 '\0' == *arch ? "" : "/",
1295 '\0' == *arch ? "" : arch, key);
1296 }
1297 sqlite3_finalize(stmt);
1298 }
1299
1300 static void
1301 parse_cat(struct mpage *mpage, int fd)
1302 {
1303 FILE *stream;
1304 char *line, *p, *title;
1305 size_t len, plen, titlesz;
1306
1307 stream = (-1 == fd) ?
1308 fopen(mpage->mlinks->file, "r") :
1309 fdopen(fd, "r");
1310 if (NULL == stream) {
1311 if (-1 != fd)
1312 close(fd);
1313 if (warnings)
1314 say(mpage->mlinks->file, "&fopen");
1315 return;
1316 }
1317
1318 /* Skip to first blank line. */
1319
1320 while (NULL != (line = fgetln(stream, &len)))
1321 if ('\n' == *line)
1322 break;
1323
1324 /*
1325 * Assume the first line that is not indented
1326 * is the first section header. Skip to it.
1327 */
1328
1329 while (NULL != (line = fgetln(stream, &len)))
1330 if ('\n' != *line && ' ' != *line)
1331 break;
1332
1333 /*
1334 * Read up until the next section into a buffer.
1335 * Strip the leading and trailing newline from each read line,
1336 * appending a trailing space.
1337 * Ignore empty (whitespace-only) lines.
1338 */
1339
1340 titlesz = 0;
1341 title = NULL;
1342
1343 while (NULL != (line = fgetln(stream, &len))) {
1344 if (' ' != *line || '\n' != line[len - 1])
1345 break;
1346 while (len > 0 && isspace((unsigned char)*line)) {
1347 line++;
1348 len--;
1349 }
1350 if (1 == len)
1351 continue;
1352 title = mandoc_realloc(title, titlesz + len);
1353 memcpy(title + titlesz, line, len);
1354 titlesz += len;
1355 title[titlesz - 1] = ' ';
1356 }
1357
1358 /*
1359 * If no page content can be found, or the input line
1360 * is already the next section header, or there is no
1361 * trailing newline, reuse the page title as the page
1362 * description.
1363 */
1364
1365 if (NULL == title || '\0' == *title) {
1366 if (warnings)
1367 say(mpage->mlinks->file,
1368 "Cannot find NAME section");
1369 fclose(stream);
1370 free(title);
1371 return;
1372 }
1373
1374 title = mandoc_realloc(title, titlesz + 1);
1375 title[titlesz] = '\0';
1376
1377 /*
1378 * Skip to the first dash.
1379 * Use the remaining line as the description (no more than 70
1380 * bytes).
1381 */
1382
1383 if (NULL != (p = strstr(title, "- "))) {
1384 for (p += 2; ' ' == *p || '\b' == *p; p++)
1385 /* Skip to next word. */ ;
1386 } else {
1387 if (warnings)
1388 say(mpage->mlinks->file,
1389 "No dash in title line");
1390 p = title;
1391 }
1392
1393 plen = strlen(p);
1394
1395 /* Strip backspace-encoding from line. */
1396
1397 while (NULL != (line = memchr(p, '\b', plen))) {
1398 len = line - p;
1399 if (0 == len) {
1400 memmove(line, line + 1, plen--);
1401 continue;
1402 }
1403 memmove(line - 1, line + 1, plen - len);
1404 plen -= 2;
1405 }
1406
1407 mpage->desc = mandoc_strdup(p);
1408 fclose(stream);
1409 free(title);
1410 }
1411
1412 /*
1413 * Put a type/word pair into the word database for this particular file.
1414 */
1415 static void
1416 putkey(const struct mpage *mpage, char *value, uint64_t type)
1417 {
1418 char *cp;
1419
1420 assert(NULL != value);
1421 if (TYPE_arch == type)
1422 for (cp = value; *cp; cp++)
1423 if (isupper((unsigned char)*cp))
1424 *cp = _tolower((unsigned char)*cp);
1425 putkeys(mpage, value, strlen(value), type);
1426 }
1427
1428 /*
1429 * Grok all nodes at or below a certain mdoc node into putkey().
1430 */
1431 static void
1432 putmdockey(const struct mpage *mpage,
1433 const struct mdoc_node *n, uint64_t m)
1434 {
1435
1436 for ( ; NULL != n; n = n->next) {
1437 if (NULL != n->child)
1438 putmdockey(mpage, n->child, m);
1439 if (MDOC_TEXT == n->type)
1440 putkey(mpage, n->string, m);
1441 }
1442 }
1443
1444 static void
1445 parse_man(struct mpage *mpage, const struct man_meta *meta,
1446 const struct man_node *n)
1447 {
1448 const struct man_node *head, *body;
1449 char *start, *title;
1450 char byte;
1451 size_t sz;
1452
1453 if (NULL == n)
1454 return;
1455
1456 /*
1457 * We're only searching for one thing: the first text child in
1458 * the BODY of a NAME section. Since we don't keep track of
1459 * sections in -man, run some hoops to find out whether we're in
1460 * the correct section or not.
1461 */
1462
1463 if (MAN_BODY == n->type && MAN_SH == n->tok) {
1464 body = n;
1465 assert(body->parent);
1466 if (NULL != (head = body->parent->head) &&
1467 1 == head->nchild &&
1468 NULL != (head = (head->child)) &&
1469 MAN_TEXT == head->type &&
1470 0 == strcmp(head->string, "NAME") &&
1471 NULL != body->child) {
1472
1473 /*
1474 * Suck the entire NAME section into memory.
1475 * Yes, we might run away.
1476 * But too many manuals have big, spread-out
1477 * NAME sections over many lines.
1478 */
1479
1480 title = NULL;
1481 man_deroff(&title, body);
1482 if (NULL == title)
1483 return;
1484
1485 /*
1486 * Go through a special heuristic dance here.
1487 * Conventionally, one or more manual names are
1488 * comma-specified prior to a whitespace, then a
1489 * dash, then a description. Try to puzzle out
1490 * the name parts here.
1491 */
1492
1493 start = title;
1494 for ( ;; ) {
1495 sz = strcspn(start, " ,");
1496 if ('\0' == start[sz])
1497 break;
1498
1499 byte = start[sz];
1500 start[sz] = '\0';
1501
1502 /*
1503 * Assume a stray trailing comma in the
1504 * name list if a name begins with a dash.
1505 */
1506
1507 if ('-' == start[0] ||
1508 ('\\' == start[0] && '-' == start[1]))
1509 break;
1510
1511 putkey(mpage, start, NAME_TITLE);
1512 if ( ! (mpage->name_head_done ||
1513 strcasecmp(start, meta->title))) {
1514 putkey(mpage, start, NAME_HEAD);
1515 mpage->name_head_done = 1;
1516 }
1517
1518 if (' ' == byte) {
1519 start += sz + 1;
1520 break;
1521 }
1522
1523 assert(',' == byte);
1524 start += sz + 1;
1525 while (' ' == *start)
1526 start++;
1527 }
1528
1529 if (start == title) {
1530 putkey(mpage, start, NAME_TITLE);
1531 if ( ! (mpage->name_head_done ||
1532 strcasecmp(start, meta->title))) {
1533 putkey(mpage, start, NAME_HEAD);
1534 mpage->name_head_done = 1;
1535 }
1536 free(title);
1537 return;
1538 }
1539
1540 while (isspace((unsigned char)*start))
1541 start++;
1542
1543 if (0 == strncmp(start, "-", 1))
1544 start += 1;
1545 else if (0 == strncmp(start, "\\-\\-", 4))
1546 start += 4;
1547 else if (0 == strncmp(start, "\\-", 2))
1548 start += 2;
1549 else if (0 == strncmp(start, "\\(en", 4))
1550 start += 4;
1551 else if (0 == strncmp(start, "\\(em", 4))
1552 start += 4;
1553
1554 while (' ' == *start)
1555 start++;
1556
1557 mpage->desc = mandoc_strdup(start);
1558 free(title);
1559 return;
1560 }
1561 }
1562
1563 for (n = n->child; n; n = n->next) {
1564 if (NULL != mpage->desc)
1565 break;
1566 parse_man(mpage, meta, n);
1567 }
1568 }
1569
1570 static void
1571 parse_mdoc(struct mpage *mpage, const struct mdoc_meta *meta,
1572 const struct mdoc_node *n)
1573 {
1574
1575 assert(NULL != n);
1576 for (n = n->child; NULL != n; n = n->next) {
1577 switch (n->type) {
1578 case MDOC_ELEM:
1579 /* FALLTHROUGH */
1580 case MDOC_BLOCK:
1581 /* FALLTHROUGH */
1582 case MDOC_HEAD:
1583 /* FALLTHROUGH */
1584 case MDOC_BODY:
1585 /* FALLTHROUGH */
1586 case MDOC_TAIL:
1587 if (NULL != mdocs[n->tok].fp)
1588 if (0 == (*mdocs[n->tok].fp)(mpage, meta, n))
1589 break;
1590 if (mdocs[n->tok].mask)
1591 putmdockey(mpage, n->child,
1592 mdocs[n->tok].mask);
1593 break;
1594 default:
1595 assert(MDOC_ROOT != n->type);
1596 continue;
1597 }
1598 if (NULL != n->child)
1599 parse_mdoc(mpage, meta, n);
1600 }
1601 }
1602
1603 static int
1604 parse_mdoc_Fd(struct mpage *mpage, const struct mdoc_meta *meta,
1605 const struct mdoc_node *n)
1606 {
1607 char *start, *end;
1608 size_t sz;
1609
1610 if (SEC_SYNOPSIS != n->sec ||
1611 NULL == (n = n->child) ||
1612 MDOC_TEXT != n->type)
1613 return(0);
1614
1615 /*
1616 * Only consider those `Fd' macro fields that begin with an
1617 * "inclusion" token (versus, e.g., #define).
1618 */
1619
1620 if (strcmp("#include", n->string))
1621 return(0);
1622
1623 if (NULL == (n = n->next) || MDOC_TEXT != n->type)
1624 return(0);
1625
1626 /*
1627 * Strip away the enclosing angle brackets and make sure we're
1628 * not zero-length.
1629 */
1630
1631 start = n->string;
1632 if ('<' == *start || '"' == *start)
1633 start++;
1634
1635 if (0 == (sz = strlen(start)))
1636 return(0);
1637
1638 end = &start[(int)sz - 1];
1639 if ('>' == *end || '"' == *end)
1640 end--;
1641
1642 if (end > start)
1643 putkeys(mpage, start, end - start + 1, TYPE_In);
1644 return(0);
1645 }
1646
1647 static void
1648 parse_mdoc_fname(struct mpage *mpage, const struct mdoc_node *n)
1649 {
1650 char *cp;
1651 size_t sz;
1652
1653 if (n->type != MDOC_TEXT)
1654 return;
1655
1656 /* Skip function pointer punctuation. */
1657
1658 cp = n->string;
1659 while (*cp == '(' || *cp == '*')
1660 cp++;
1661 sz = strcspn(cp, "()");
1662
1663 putkeys(mpage, cp, sz, TYPE_Fn);
1664 if (n->sec == SEC_SYNOPSIS)
1665 putkeys(mpage, cp, sz, NAME_SYN);
1666 }
1667
1668 static int
1669 parse_mdoc_Fn(struct mpage *mpage, const struct mdoc_meta *meta,
1670 const struct mdoc_node *n)
1671 {
1672
1673 if (n->child == NULL)
1674 return(0);
1675
1676 parse_mdoc_fname(mpage, n->child);
1677
1678 for (n = n->child->next; n != NULL; n = n->next)
1679 if (n->type == MDOC_TEXT)
1680 putkey(mpage, n->string, TYPE_Fa);
1681
1682 return(0);
1683 }
1684
1685 static int
1686 parse_mdoc_Fo(struct mpage *mpage, const struct mdoc_meta *meta,
1687 const struct mdoc_node *n)
1688 {
1689
1690 if (n->type != MDOC_HEAD)
1691 return(1);
1692
1693 if (n->child != NULL)
1694 parse_mdoc_fname(mpage, n->child);
1695
1696 return(0);
1697 }
1698
1699 static int
1700 parse_mdoc_Xr(struct mpage *mpage, const struct mdoc_meta *meta,
1701 const struct mdoc_node *n)
1702 {
1703 char *cp;
1704
1705 if (NULL == (n = n->child))
1706 return(0);
1707
1708 if (NULL == n->next) {
1709 putkey(mpage, n->string, TYPE_Xr);
1710 return(0);
1711 }
1712
1713 mandoc_asprintf(&cp, "%s(%s)", n->string, n->next->string);
1714 putkey(mpage, cp, TYPE_Xr);
1715 free(cp);
1716 return(0);
1717 }
1718
1719 static int
1720 parse_mdoc_Nd(struct mpage *mpage, const struct mdoc_meta *meta,
1721 const struct mdoc_node *n)
1722 {
1723
1724 if (MDOC_BODY == n->type)
1725 mdoc_deroff(&mpage->desc, n);
1726 return(0);
1727 }
1728
1729 static int
1730 parse_mdoc_Nm(struct mpage *mpage, const struct mdoc_meta *meta,
1731 const struct mdoc_node *n)
1732 {
1733
1734 if (SEC_NAME == n->sec)
1735 putmdockey(mpage, n->child, NAME_TITLE);
1736 else if (SEC_SYNOPSIS == n->sec && MDOC_HEAD == n->type) {
1737 if (n->child == NULL)
1738 putkey(mpage, meta->name, NAME_SYN);
1739 else
1740 putmdockey(mpage, n->child, NAME_SYN);
1741 }
1742 if ( ! (mpage->name_head_done ||
1743 n->child == NULL || n->child->string == NULL ||
1744 strcasecmp(n->child->string, meta->title))) {
1745 putkey(mpage, n->child->string, NAME_HEAD);
1746 mpage->name_head_done = 1;
1747 }
1748 return(0);
1749 }
1750
1751 static int
1752 parse_mdoc_Sh(struct mpage *mpage, const struct mdoc_meta *meta,
1753 const struct mdoc_node *n)
1754 {
1755
1756 return(SEC_CUSTOM == n->sec && MDOC_HEAD == n->type);
1757 }
1758
1759 static int
1760 parse_mdoc_head(struct mpage *mpage, const struct mdoc_meta *meta,
1761 const struct mdoc_node *n)
1762 {
1763
1764 return(MDOC_HEAD == n->type);
1765 }
1766
1767 static int
1768 parse_mdoc_body(struct mpage *mpage, const struct mdoc_meta *meta,
1769 const struct mdoc_node *n)
1770 {
1771
1772 return(MDOC_BODY == n->type);
1773 }
1774
1775 /*
1776 * Add a string to the hash table for the current manual.
1777 * Each string has a bitmask telling which macros it belongs to.
1778 * When we finish the manual, we'll dump the table.
1779 */
1780 static void
1781 putkeys(const struct mpage *mpage, char *cp, size_t sz, uint64_t v)
1782 {
1783 struct ohash *htab;
1784 struct str *s;
1785 const char *end;
1786 unsigned int slot;
1787 int i, mustfree;
1788
1789 if (0 == sz)
1790 return;
1791
1792 mustfree = render_string(&cp, &sz);
1793
1794 if (TYPE_Nm & v) {
1795 htab = &names;
1796 v &= name_mask;
1797 if (v & NAME_FIRST)
1798 name_mask &= ~NAME_FIRST;
1799 if (debug > 1)
1800 say(mpage->mlinks->file,
1801 "Adding name %*s, bits=%d", sz, cp, v);
1802 } else {
1803 htab = &strings;
1804 if (debug > 1)
1805 for (i = 0; i < mansearch_keymax; i++)
1806 if ((uint64_t)1 << i & v)
1807 say(mpage->mlinks->file,
1808 "Adding key %s=%*s",
1809 mansearch_keynames[i], sz, cp);
1810 }
1811
1812 end = cp + sz;
1813 slot = ohash_qlookupi(htab, cp, &end);
1814 s = ohash_find(htab, slot);
1815
1816 if (NULL != s && mpage == s->mpage) {
1817 s->mask |= v;
1818 return;
1819 } else if (NULL == s) {
1820 s = mandoc_calloc(1, sizeof(struct str) + sz + 1);
1821 memcpy(s->key, cp, sz);
1822 ohash_insert(htab, slot, s);
1823 }
1824 s->mpage = mpage;
1825 s->mask = v;
1826
1827 if (mustfree)
1828 free(cp);
1829 }
1830
1831 /*
1832 * Take a Unicode codepoint and produce its UTF-8 encoding.
1833 * This isn't the best way to do this, but it works.
1834 * The magic numbers are from the UTF-8 packaging.
1835 * They're not as scary as they seem: read the UTF-8 spec for details.
1836 */
1837 static size_t
1838 utf8(unsigned int cp, char out[7])
1839 {
1840 size_t rc;
1841
1842 rc = 0;
1843 if (cp <= 0x0000007F) {
1844 rc = 1;
1845 out[0] = (char)cp;
1846 } else if (cp <= 0x000007FF) {
1847 rc = 2;
1848 out[0] = (cp >> 6 & 31) | 192;
1849 out[1] = (cp & 63) | 128;
1850 } else if (cp <= 0x0000FFFF) {
1851 rc = 3;
1852 out[0] = (cp >> 12 & 15) | 224;
1853 out[1] = (cp >> 6 & 63) | 128;
1854 out[2] = (cp & 63) | 128;
1855 } else if (cp <= 0x001FFFFF) {
1856 rc = 4;
1857 out[0] = (cp >> 18 & 7) | 240;
1858 out[1] = (cp >> 12 & 63) | 128;
1859 out[2] = (cp >> 6 & 63) | 128;
1860 out[3] = (cp & 63) | 128;
1861 } else if (cp <= 0x03FFFFFF) {
1862 rc = 5;
1863 out[0] = (cp >> 24 & 3) | 248;
1864 out[1] = (cp >> 18 & 63) | 128;
1865 out[2] = (cp >> 12 & 63) | 128;
1866 out[3] = (cp >> 6 & 63) | 128;
1867 out[4] = (cp & 63) | 128;
1868 } else if (cp <= 0x7FFFFFFF) {
1869 rc = 6;
1870 out[0] = (cp >> 30 & 1) | 252;
1871 out[1] = (cp >> 24 & 63) | 128;
1872 out[2] = (cp >> 18 & 63) | 128;
1873 out[3] = (cp >> 12 & 63) | 128;
1874 out[4] = (cp >> 6 & 63) | 128;
1875 out[5] = (cp & 63) | 128;
1876 } else
1877 return(0);
1878
1879 out[rc] = '\0';
1880 return(rc);
1881 }
1882
1883 /*
1884 * If the string contains escape sequences,
1885 * replace it with an allocated rendering and return 1,
1886 * such that the caller can free it after use.
1887 * Otherwise, do nothing and return 0.
1888 */
1889 static int
1890 render_string(char **public, size_t *psz)
1891 {
1892 const char *src, *scp, *addcp, *seq;
1893 char *dst;
1894 size_t ssz, dsz, addsz;
1895 char utfbuf[7], res[6];
1896 int seqlen, unicode;
1897
1898 res[0] = '\\';
1899 res[1] = '\t';
1900 res[2] = ASCII_NBRSP;
1901 res[3] = ASCII_HYPH;
1902 res[4] = ASCII_BREAK;
1903 res[5] = '\0';
1904
1905 src = scp = *public;
1906 ssz = *psz;
1907 dst = NULL;
1908 dsz = 0;
1909
1910 while (scp < src + *psz) {
1911
1912 /* Leave normal characters unchanged. */
1913
1914 if (strchr(res, *scp) == NULL) {
1915 if (dst != NULL)
1916 dst[dsz++] = *scp;
1917 scp++;
1918 continue;
1919 }
1920
1921 /*
1922 * Found something that requires replacing,
1923 * make sure we have a destination buffer.
1924 */
1925
1926 if (dst == NULL) {
1927 dst = mandoc_malloc(ssz + 1);
1928 dsz = scp - src;
1929 memcpy(dst, src, dsz);
1930 }
1931
1932 /* Handle single-char special characters. */
1933
1934 switch (*scp) {
1935 case '\\':
1936 break;
1937 case '\t':
1938 /* FALLTHROUGH */
1939 case ASCII_NBRSP:
1940 dst[dsz++] = ' ';
1941 scp++;
1942 continue;
1943 case ASCII_HYPH:
1944 dst[dsz++] = '-';
1945 /* FALLTHROUGH */
1946 case ASCII_BREAK:
1947 scp++;
1948 continue;
1949 default:
1950 abort();
1951 }
1952
1953 /*
1954 * Found an escape sequence.
1955 * Read past the slash, then parse it.
1956 * Ignore everything except characters.
1957 */
1958
1959 scp++;
1960 if (mandoc_escape(&scp, &seq, &seqlen) != ESCAPE_SPECIAL)
1961 continue;
1962
1963 /*
1964 * Render the special character
1965 * as either UTF-8 or ASCII.
1966 */
1967
1968 if (write_utf8) {
1969 unicode = mchars_spec2cp(mchars, seq, seqlen);
1970 if (unicode <= 0)
1971 continue;
1972 addsz = utf8(unicode, utfbuf);
1973 if (addsz == 0)
1974 continue;
1975 addcp = utfbuf;
1976 } else {
1977 addcp = mchars_spec2str(mchars, seq, seqlen, &addsz);
1978 if (addcp == NULL)
1979 continue;
1980 if (*addcp == ASCII_NBRSP) {
1981 addcp = " ";
1982 addsz = 1;
1983 }
1984 }
1985
1986 /* Copy the rendered glyph into the stream. */
1987
1988 ssz += addsz;
1989 dst = mandoc_realloc(dst, ssz + 1);
1990 memcpy(dst + dsz, addcp, addsz);
1991 dsz += addsz;
1992 }
1993 if (dst != NULL) {
1994 *public = dst;
1995 *psz = dsz;
1996 }
1997
1998 /* Trim trailing whitespace and NUL-terminate. */
1999
2000 while (*psz > 0 && (*public)[*psz - 1] == ' ')
2001 --*psz;
2002 if (dst != NULL) {
2003 (*public)[*psz] = '\0';
2004 return(1);
2005 } else
2006 return(0);
2007 }
2008
2009 static void
2010 dbadd_mlink(const struct mlink *mlink)
2011 {
2012 size_t i;
2013
2014 i = 1;
2015 SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->dsec);
2016 SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->arch);
2017 SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->name);
2018 SQL_BIND_INT64(stmts[STMT_INSERT_LINK], i, mlink->mpage->pageid);
2019 SQL_STEP(stmts[STMT_INSERT_LINK]);
2020 sqlite3_reset(stmts[STMT_INSERT_LINK]);
2021 }
2022
2023 static void
2024 dbadd_mlink_name(const struct mlink *mlink)
2025 {
2026 uint64_t bits;
2027 size_t i;
2028
2029 dbadd_mlink(mlink);
2030
2031 i = 1;
2032 SQL_BIND_INT64(stmts[STMT_SELECT_NAME], i, mlink->mpage->pageid);
2033 bits = NAME_FILE & NAME_MASK;
2034 if (sqlite3_step(stmts[STMT_SELECT_NAME]) == SQLITE_ROW) {
2035 bits |= sqlite3_column_int64(stmts[STMT_SELECT_NAME], 0);
2036 sqlite3_reset(stmts[STMT_SELECT_NAME]);
2037 }
2038
2039 i = 1;
2040 SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, bits);
2041 SQL_BIND_TEXT(stmts[STMT_INSERT_NAME], i, mlink->name);
2042 SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, mlink->mpage->pageid);
2043 SQL_STEP(stmts[STMT_INSERT_NAME]);
2044 sqlite3_reset(stmts[STMT_INSERT_NAME]);
2045 }
2046
2047 /*
2048 * Flush the current page's terms (and their bits) into the database.
2049 * Wrap the entire set of additions in a transaction to make sqlite be a
2050 * little faster.
2051 * Also, handle escape sequences at the last possible moment.
2052 */
2053 static void
2054 dbadd(struct mpage *mpage)
2055 {
2056 struct mlink *mlink;
2057 struct str *key;
2058 char *cp;
2059 size_t i;
2060 unsigned int slot;
2061 int mustfree;
2062
2063 mlink = mpage->mlinks;
2064
2065 if (nodb) {
2066 for (key = ohash_first(&names, &slot); NULL != key;
2067 key = ohash_next(&names, &slot))
2068 free(key);
2069 for (key = ohash_first(&strings, &slot); NULL != key;
2070 key = ohash_next(&strings, &slot))
2071 free(key);
2072 if (0 == debug)
2073 return;
2074 while (NULL != mlink) {
2075 fputs(mlink->name, stdout);
2076 if (NULL == mlink->next ||
2077 strcmp(mlink->dsec, mlink->next->dsec) ||
2078 strcmp(mlink->fsec, mlink->next->fsec) ||
2079 strcmp(mlink->arch, mlink->next->arch)) {
2080 putchar('(');
2081 if ('\0' == *mlink->dsec)
2082 fputs(mlink->fsec, stdout);
2083 else
2084 fputs(mlink->dsec, stdout);
2085 if ('\0' != *mlink->arch)
2086 printf("/%s", mlink->arch);
2087 putchar(')');
2088 }
2089 mlink = mlink->next;
2090 if (NULL != mlink)
2091 fputs(", ", stdout);
2092 }
2093 printf(" - %s\n", mpage->desc);
2094 return;
2095 }
2096
2097 if (debug)
2098 say(mlink->file, "Adding to database");
2099
2100 cp = mpage->desc;
2101 i = strlen(cp);
2102 mustfree = render_string(&cp, &i);
2103 i = 1;
2104 SQL_BIND_TEXT(stmts[STMT_INSERT_PAGE], i, cp);
2105 SQL_BIND_INT(stmts[STMT_INSERT_PAGE], i, mpage->form);
2106 SQL_STEP(stmts[STMT_INSERT_PAGE]);
2107 mpage->pageid = sqlite3_last_insert_rowid(db);
2108 sqlite3_reset(stmts[STMT_INSERT_PAGE]);
2109 if (mustfree)
2110 free(cp);
2111
2112 while (NULL != mlink) {
2113 dbadd_mlink(mlink);
2114 mlink = mlink->next;
2115 }
2116 mlink = mpage->mlinks;
2117
2118 for (key = ohash_first(&names, &slot); NULL != key;
2119 key = ohash_next(&names, &slot)) {
2120 assert(key->mpage == mpage);
2121 i = 1;
2122 SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, key->mask);
2123 SQL_BIND_TEXT(stmts[STMT_INSERT_NAME], i, key->key);
2124 SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, mpage->pageid);
2125 SQL_STEP(stmts[STMT_INSERT_NAME]);
2126 sqlite3_reset(stmts[STMT_INSERT_NAME]);
2127 free(key);
2128 }
2129 for (key = ohash_first(&strings, &slot); NULL != key;
2130 key = ohash_next(&strings, &slot)) {
2131 assert(key->mpage == mpage);
2132 i = 1;
2133 SQL_BIND_INT64(stmts[STMT_INSERT_KEY], i, key->mask);
2134 SQL_BIND_TEXT(stmts[STMT_INSERT_KEY], i, key->key);
2135 SQL_BIND_INT64(stmts[STMT_INSERT_KEY], i, mpage->pageid);
2136 SQL_STEP(stmts[STMT_INSERT_KEY]);
2137 sqlite3_reset(stmts[STMT_INSERT_KEY]);
2138 free(key);
2139 }
2140 }
2141
2142 static void
2143 dbprune(void)
2144 {
2145 struct mpage *mpage;
2146 struct mlink *mlink;
2147 size_t i;
2148 unsigned int slot;
2149
2150 if (0 == nodb)
2151 SQL_EXEC("BEGIN TRANSACTION");
2152
2153 for (mpage = ohash_first(&mpages, &slot); NULL != mpage;
2154 mpage = ohash_next(&mpages, &slot)) {
2155 mlink = mpage->mlinks;
2156 if (debug)
2157 say(mlink->file, "Deleting from database");
2158 if (nodb)
2159 continue;
2160 for ( ; NULL != mlink; mlink = mlink->next) {
2161 i = 1;
2162 SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE],
2163 i, mlink->dsec);
2164 SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE],
2165 i, mlink->arch);
2166 SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE],
2167 i, mlink->name);
2168 SQL_STEP(stmts[STMT_DELETE_PAGE]);
2169 sqlite3_reset(stmts[STMT_DELETE_PAGE]);
2170 }
2171 }
2172
2173 if (0 == nodb)
2174 SQL_EXEC("END TRANSACTION");
2175 }
2176
2177 /*
2178 * Close an existing database and its prepared statements.
2179 * If "real" is not set, rename the temporary file into the real one.
2180 */
2181 static void
2182 dbclose(int real)
2183 {
2184 size_t i;
2185 int status;
2186 pid_t child;
2187
2188 if (nodb)
2189 return;
2190
2191 for (i = 0; i < STMT__MAX; i++) {
2192 sqlite3_finalize(stmts[i]);
2193 stmts[i] = NULL;
2194 }
2195
2196 sqlite3_close(db);
2197 db = NULL;
2198
2199 if (real)
2200 return;
2201
2202 if ('\0' == *tempfilename) {
2203 if (-1 == rename(MANDOC_DB "~", MANDOC_DB)) {
2204 exitcode = (int)MANDOCLEVEL_SYSERR;
2205 say(MANDOC_DB, "&rename");
2206 }
2207 return;
2208 }
2209
2210 switch (child = fork()) {
2211 case -1:
2212 exitcode = (int)MANDOCLEVEL_SYSERR;
2213 say("", "&fork cmp");
2214 return;
2215 case 0:
2216 execlp("cmp", "cmp", "-s",
2217 tempfilename, MANDOC_DB, NULL);
2218 say("", "&exec cmp");
2219 exit(0);
2220 default:
2221 break;
2222 }
2223 if (-1 == waitpid(child, &status, 0)) {
2224 exitcode = (int)MANDOCLEVEL_SYSERR;
2225 say("", "&wait cmp");
2226 } else if (WIFSIGNALED(status)) {
2227 exitcode = (int)MANDOCLEVEL_SYSERR;
2228 say("", "cmp died from signal %d", WTERMSIG(status));
2229 } else if (WEXITSTATUS(status)) {
2230 exitcode = (int)MANDOCLEVEL_SYSERR;
2231 say(MANDOC_DB,
2232 "Data changed, but cannot replace database");
2233 }
2234
2235 *strrchr(tempfilename, '/') = '\0';
2236 switch (child = fork()) {
2237 case -1:
2238 exitcode = (int)MANDOCLEVEL_SYSERR;
2239 say("", "&fork rm");
2240 return;
2241 case 0:
2242 execlp("rm", "rm", "-rf", tempfilename, NULL);
2243 say("", "&exec rm");
2244 exit((int)MANDOCLEVEL_SYSERR);
2245 default:
2246 break;
2247 }
2248 if (-1 == waitpid(child, &status, 0)) {
2249 exitcode = (int)MANDOCLEVEL_SYSERR;
2250 say("", "&wait rm");
2251 } else if (WIFSIGNALED(status) || WEXITSTATUS(status)) {
2252 exitcode = (int)MANDOCLEVEL_SYSERR;
2253 say("", "%s: Cannot remove temporary directory",
2254 tempfilename);
2255 }
2256 }
2257
2258 /*
2259 * This is straightforward stuff.
2260 * Open a database connection to a "temporary" database, then open a set
2261 * of prepared statements we'll use over and over again.
2262 * If "real" is set, we use the existing database; if not, we truncate a
2263 * temporary one.
2264 * Must be matched by dbclose().
2265 */
2266 static int
2267 dbopen(int real)
2268 {
2269 const char *sql;
2270 int rc, ofl;
2271
2272 if (nodb)
2273 return(1);
2274
2275 *tempfilename = '\0';
2276 ofl = SQLITE_OPEN_READWRITE;
2277
2278 if (real) {
2279 rc = sqlite3_open_v2(MANDOC_DB, &db, ofl, NULL);
2280 if (SQLITE_OK != rc) {
2281 exitcode = (int)MANDOCLEVEL_SYSERR;
2282 if (SQLITE_CANTOPEN != rc)
2283 say(MANDOC_DB, "%s", sqlite3_errstr(rc));
2284 return(0);
2285 }
2286 goto prepare_statements;
2287 }
2288
2289 ofl |= SQLITE_OPEN_CREATE | SQLITE_OPEN_EXCLUSIVE;
2290
2291 remove(MANDOC_DB "~");
2292 rc = sqlite3_open_v2(MANDOC_DB "~", &db, ofl, NULL);
2293 if (SQLITE_OK == rc)
2294 goto create_tables;
2295 if (MPARSE_QUICK & mparse_options) {
2296 exitcode = (int)MANDOCLEVEL_SYSERR;
2297 say(MANDOC_DB "~", "%s", sqlite3_errstr(rc));
2298 return(0);
2299 }
2300
2301 (void)strlcpy(tempfilename, "/tmp/mandocdb.XXXXXX",
2302 sizeof(tempfilename));
2303 if (NULL == mkdtemp(tempfilename)) {
2304 exitcode = (int)MANDOCLEVEL_SYSERR;
2305 say("", "&%s", tempfilename);
2306 return(0);
2307 }
2308 (void)strlcat(tempfilename, "/" MANDOC_DB,
2309 sizeof(tempfilename));
2310 rc = sqlite3_open_v2(tempfilename, &db, ofl, NULL);
2311 if (SQLITE_OK != rc) {
2312 exitcode = (int)MANDOCLEVEL_SYSERR;
2313 say("", "%s: %s", tempfilename, sqlite3_errstr(rc));
2314 return(0);
2315 }
2316
2317 create_tables:
2318 sql = "CREATE TABLE \"mpages\" (\n"
2319 " \"desc\" TEXT NOT NULL,\n"
2320 " \"form\" INTEGER NOT NULL,\n"
2321 " \"pageid\" INTEGER PRIMARY KEY AUTOINCREMENT NOT NULL\n"
2322 ");\n"
2323 "\n"
2324 "CREATE TABLE \"mlinks\" (\n"
2325 " \"sec\" TEXT NOT NULL,\n"
2326 " \"arch\" TEXT NOT NULL,\n"
2327 " \"name\" TEXT NOT NULL,\n"
2328 " \"pageid\" INTEGER NOT NULL REFERENCES mpages(pageid) "
2329 "ON DELETE CASCADE\n"
2330 ");\n"
2331 "CREATE INDEX mlinks_pageid_idx ON mlinks (pageid);\n"
2332 "\n"
2333 "CREATE TABLE \"names\" (\n"
2334 " \"bits\" INTEGER NOT NULL,\n"
2335 " \"name\" TEXT NOT NULL,\n"
2336 " \"pageid\" INTEGER NOT NULL REFERENCES mpages(pageid) "
2337 "ON DELETE CASCADE,\n"
2338 " UNIQUE (\"name\", \"pageid\") ON CONFLICT REPLACE\n"
2339 ");\n"
2340 "\n"
2341 "CREATE TABLE \"keys\" (\n"
2342 " \"bits\" INTEGER NOT NULL,\n"
2343 " \"key\" TEXT NOT NULL,\n"
2344 " \"pageid\" INTEGER NOT NULL REFERENCES mpages(pageid) "
2345 "ON DELETE CASCADE\n"
2346 ");\n"
2347 "CREATE INDEX keys_pageid_idx ON keys (pageid);\n";
2348
2349 if (SQLITE_OK != sqlite3_exec(db, sql, NULL, NULL, NULL)) {
2350 exitcode = (int)MANDOCLEVEL_SYSERR;
2351 say(MANDOC_DB, "%s", sqlite3_errmsg(db));
2352 sqlite3_close(db);
2353 return(0);
2354 }
2355
2356 prepare_statements:
2357 if (SQLITE_OK != sqlite3_exec(db,
2358 "PRAGMA foreign_keys = ON", NULL, NULL, NULL)) {
2359 exitcode = (int)MANDOCLEVEL_SYSERR;
2360 say(MANDOC_DB, "PRAGMA foreign_keys: %s",
2361 sqlite3_errmsg(db));
2362 sqlite3_close(db);
2363 return(0);
2364 }
2365
2366 sql = "DELETE FROM mpages WHERE pageid IN "
2367 "(SELECT pageid FROM mlinks WHERE "
2368 "sec=? AND arch=? AND name=?)";
2369 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_DELETE_PAGE], NULL);
2370 sql = "INSERT INTO mpages "
2371 "(desc,form) VALUES (?,?)";
2372 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_PAGE], NULL);
2373 sql = "INSERT INTO mlinks "
2374 "(sec,arch,name,pageid) VALUES (?,?,?,?)";
2375 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_LINK], NULL);
2376 sql = "SELECT bits FROM names where pageid = ?";
2377 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_SELECT_NAME], NULL);
2378 sql = "INSERT INTO names "
2379 "(bits,name,pageid) VALUES (?,?,?)";
2380 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_NAME], NULL);
2381 sql = "INSERT INTO keys "
2382 "(bits,key,pageid) VALUES (?,?,?)";
2383 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_KEY], NULL);
2384
2385 #ifndef __APPLE__
2386 /*
2387 * When opening a new database, we can turn off
2388 * synchronous mode for much better performance.
2389 */
2390
2391 if (real && SQLITE_OK != sqlite3_exec(db,
2392 "PRAGMA synchronous = OFF", NULL, NULL, NULL)) {
2393 exitcode = (int)MANDOCLEVEL_SYSERR;
2394 say(MANDOC_DB, "PRAGMA synchronous: %s",
2395 sqlite3_errmsg(db));
2396 sqlite3_close(db);
2397 return(0);
2398 }
2399 #endif
2400
2401 return(1);
2402 }
2403
2404 static void *
2405 hash_calloc(size_t n, size_t sz, void *arg)
2406 {
2407
2408 return(mandoc_calloc(n, sz));
2409 }
2410
2411 static void *
2412 hash_alloc(size_t sz, void *arg)
2413 {
2414
2415 return(mandoc_malloc(sz));
2416 }
2417
2418 static void
2419 hash_free(void *p, void *arg)
2420 {
2421
2422 free(p);
2423 }
2424
2425 static int
2426 set_basedir(const char *targetdir, int report_baddir)
2427 {
2428 static char startdir[PATH_MAX];
2429 static int getcwd_status; /* 1 = ok, 2 = failure */
2430 static int chdir_status; /* 1 = changed directory */
2431 char *cp;
2432
2433 /*
2434 * Remember the original working directory, if possible.
2435 * This will be needed if the second or a later directory
2436 * on the command line is given as a relative path.
2437 * Do not error out if the current directory is not
2438 * searchable: Maybe it won't be needed after all.
2439 */
2440 if (0 == getcwd_status) {
2441 if (NULL == getcwd(startdir, sizeof(startdir))) {
2442 getcwd_status = 2;
2443 (void)strlcpy(startdir, strerror(errno),
2444 sizeof(startdir));
2445 } else
2446 getcwd_status = 1;
2447 }
2448
2449 /*
2450 * We are leaving the old base directory.
2451 * Do not use it any longer, not even for messages.
2452 */
2453 *basedir = '\0';
2454
2455 /*
2456 * If and only if the directory was changed earlier and
2457 * the next directory to process is given as a relative path,
2458 * first go back, or bail out if that is impossible.
2459 */
2460 if (chdir_status && '/' != *targetdir) {
2461 if (2 == getcwd_status) {
2462 exitcode = (int)MANDOCLEVEL_SYSERR;
2463 say("", "getcwd: %s", startdir);
2464 return(0);
2465 }
2466 if (-1 == chdir(startdir)) {
2467 exitcode = (int)MANDOCLEVEL_SYSERR;
2468 say("", "&chdir %s", startdir);
2469 return(0);
2470 }
2471 }
2472
2473 /*
2474 * Always resolve basedir to the canonicalized absolute
2475 * pathname and append a trailing slash, such that
2476 * we can reliably check whether files are inside.
2477 */
2478 if (NULL == realpath(targetdir, basedir)) {
2479 if (report_baddir || errno != ENOENT) {
2480 exitcode = (int)MANDOCLEVEL_BADARG;
2481 say("", "&%s: realpath", targetdir);
2482 }
2483 return(0);
2484 } else if (-1 == chdir(basedir)) {
2485 if (report_baddir || errno != ENOENT) {
2486 exitcode = (int)MANDOCLEVEL_BADARG;
2487 say("", "&chdir");
2488 }
2489 return(0);
2490 }
2491 chdir_status = 1;
2492 cp = strchr(basedir, '\0');
2493 if ('/' != cp[-1]) {
2494 if (cp - basedir >= PATH_MAX - 1) {
2495 exitcode = (int)MANDOCLEVEL_SYSERR;
2496 say("", "Filename too long");
2497 return(0);
2498 }
2499 *cp++ = '/';
2500 *cp = '\0';
2501 }
2502 return(1);
2503 }
2504
2505 static void
2506 say(const char *file, const char *format, ...)
2507 {
2508 va_list ap;
2509 int use_errno;
2510
2511 if ('\0' != *basedir)
2512 fprintf(stderr, "%s", basedir);
2513 if ('\0' != *basedir && '\0' != *file)
2514 fputc('/', stderr);
2515 if ('\0' != *file)
2516 fprintf(stderr, "%s", file);
2517
2518 use_errno = 1;
2519 if (NULL != format) {
2520 switch (*format) {
2521 case '&':
2522 format++;
2523 break;
2524 case '\0':
2525 format = NULL;
2526 break;
2527 default:
2528 use_errno = 0;
2529 break;
2530 }
2531 }
2532 if (NULL != format) {
2533 if ('\0' != *basedir || '\0' != *file)
2534 fputs(": ", stderr);
2535 va_start(ap, format);
2536 vfprintf(stderr, format, ap);
2537 va_end(ap);
2538 }
2539 if (use_errno) {
2540 if ('\0' != *basedir || '\0' != *file || NULL != format)
2541 fputs(": ", stderr);
2542 perror(NULL);
2543 } else
2544 fputc('\n', stderr);
2545 }
A git conversion mirror of mandoc.bsd.lv