]> 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
let the build system cope with the recent ohash changes
[mandoc.git] / mandocdb.c
1 /* $Id: mandocdb.c,v 1.152 2014/06/20 02:24:40 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 *, 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 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.calloc = mlinks_info.calloc = hash_calloc;
340 mpages_info.free = mlinks_info.free = 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 * Most of these deal with a specific directory.
437 * Jump into that directory first.
438 */
439 if (OP_TEST != op && 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_reallocarray(NULL,
475 argc, 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 == dbopen(0))
510 goto out;
511
512 mpages_merge(mc, mp);
513 if (warnings && !nodb &&
514 ! (MPARSE_QUICK & mparse_options))
515 names_check();
516 dbclose(0);
517
518 if (j + 1 < dirs.sz) {
519 mpages_free();
520 ohash_delete(&mpages);
521 ohash_delete(&mlinks);
522 }
523 }
524 }
525 out:
526 manpath_free(&dirs);
527 mchars_free(mc);
528 mparse_free(mp);
529 mpages_free();
530 ohash_delete(&mpages);
531 ohash_delete(&mlinks);
532 return(exitcode);
533 usage:
534 fprintf(stderr, "usage: %s [-aDnpQ] [-C file] [-Tutf8]\n"
535 " %s [-aDnpQ] [-Tutf8] dir ...\n"
536 " %s [-DnpQ] [-Tutf8] -d dir [file ...]\n"
537 " %s [-Dnp] -u dir [file ...]\n"
538 " %s [-Q] -t file ...\n",
539 progname, progname, progname,
540 progname, progname);
541
542 return((int)MANDOCLEVEL_BADARG);
543 }
544
545 /*
546 * Scan a directory tree rooted at "basedir" for manpages.
547 * We use fts(), scanning directory parts along the way for clues to our
548 * section and architecture.
549 *
550 * If use_all has been specified, grok all files.
551 * If not, sanitise paths to the following:
552 *
553 * [./]man*[/<arch>]/<name>.<section>
554 * or
555 * [./]cat<section>[/<arch>]/<name>.0
556 *
557 * TODO: accomodate for multi-language directories.
558 */
559 static int
560 treescan(void)
561 {
562 char buf[PATH_MAX];
563 FTS *f;
564 FTSENT *ff;
565 struct mlink *mlink;
566 int dform, gzip;
567 char *dsec, *arch, *fsec, *cp;
568 const char *path;
569 const char *argv[2];
570
571 argv[0] = ".";
572 argv[1] = (char *)NULL;
573
574 f = fts_open((char * const *)argv,
575 FTS_PHYSICAL | FTS_NOCHDIR, NULL);
576 if (NULL == f) {
577 exitcode = (int)MANDOCLEVEL_SYSERR;
578 say("", "&fts_open");
579 return(0);
580 }
581
582 dsec = arch = NULL;
583 dform = FORM_NONE;
584
585 while (NULL != (ff = fts_read(f))) {
586 path = ff->fts_path + 2;
587 switch (ff->fts_info) {
588
589 /*
590 * Symbolic links require various sanity checks,
591 * then get handled just like regular files.
592 */
593 case FTS_SL:
594 if (NULL == realpath(path, buf)) {
595 if (warnings)
596 say(path, "&realpath");
597 continue;
598 }
599 if (strstr(buf, basedir) != buf) {
600 if (warnings) say("",
601 "%s: outside base directory", buf);
602 continue;
603 }
604 /* Use logical inode to avoid mpages dupe. */
605 if (-1 == stat(path, ff->fts_statp)) {
606 if (warnings)
607 say(path, "&stat");
608 continue;
609 }
610 /* FALLTHROUGH */
611
612 /*
613 * If we're a regular file, add an mlink by using the
614 * stored directory data and handling the filename.
615 */
616 case FTS_F:
617 if (0 == strcmp(path, MANDOC_DB))
618 continue;
619 if ( ! use_all && ff->fts_level < 2) {
620 if (warnings)
621 say(path, "Extraneous file");
622 continue;
623 }
624 gzip = 0;
625 fsec = NULL;
626 while (NULL == fsec) {
627 fsec = strrchr(ff->fts_name, '.');
628 if (NULL == fsec || strcmp(fsec+1, "gz"))
629 break;
630 gzip = 1;
631 *fsec = '\0';
632 fsec = NULL;
633 }
634 if (NULL == fsec) {
635 if ( ! use_all) {
636 if (warnings)
637 say(path,
638 "No filename suffix");
639 continue;
640 }
641 } else if (0 == strcmp(++fsec, "html")) {
642 if (warnings)
643 say(path, "Skip html");
644 continue;
645 } else if (0 == strcmp(fsec, "ps")) {
646 if (warnings)
647 say(path, "Skip ps");
648 continue;
649 } else if (0 == strcmp(fsec, "pdf")) {
650 if (warnings)
651 say(path, "Skip pdf");
652 continue;
653 } else if ( ! use_all &&
654 ((FORM_SRC == dform && strcmp(fsec, dsec)) ||
655 (FORM_CAT == dform && strcmp(fsec, "0")))) {
656 if (warnings)
657 say(path, "Wrong filename suffix");
658 continue;
659 } else
660 fsec[-1] = '\0';
661
662 mlink = mandoc_calloc(1, sizeof(struct mlink));
663 if (strlcpy(mlink->file, path,
664 sizeof(mlink->file)) >=
665 sizeof(mlink->file)) {
666 say(path, "Filename too long");
667 free(mlink);
668 continue;
669 }
670 mlink->dform = dform;
671 mlink->dsec = dsec;
672 mlink->arch = arch;
673 mlink->name = ff->fts_name;
674 mlink->fsec = fsec;
675 mlink->gzip = gzip;
676 mlink_add(mlink, ff->fts_statp);
677 continue;
678
679 case FTS_D:
680 /* FALLTHROUGH */
681 case FTS_DP:
682 break;
683
684 default:
685 if (warnings)
686 say(path, "Not a regular file");
687 continue;
688 }
689
690 switch (ff->fts_level) {
691 case 0:
692 /* Ignore the root directory. */
693 break;
694 case 1:
695 /*
696 * This might contain manX/ or catX/.
697 * Try to infer this from the name.
698 * If we're not in use_all, enforce it.
699 */
700 cp = ff->fts_name;
701 if (FTS_DP == ff->fts_info)
702 break;
703
704 if (0 == strncmp(cp, "man", 3)) {
705 dform = FORM_SRC;
706 dsec = cp + 3;
707 } else if (0 == strncmp(cp, "cat", 3)) {
708 dform = FORM_CAT;
709 dsec = cp + 3;
710 } else {
711 dform = FORM_NONE;
712 dsec = NULL;
713 }
714
715 if (NULL != dsec || use_all)
716 break;
717
718 if (warnings)
719 say(path, "Unknown directory part");
720 fts_set(f, ff, FTS_SKIP);
721 break;
722 case 2:
723 /*
724 * Possibly our architecture.
725 * If we're descending, keep tabs on it.
726 */
727 if (FTS_DP != ff->fts_info && NULL != dsec)
728 arch = ff->fts_name;
729 else
730 arch = NULL;
731 break;
732 default:
733 if (FTS_DP == ff->fts_info || use_all)
734 break;
735 if (warnings)
736 say(path, "Extraneous directory part");
737 fts_set(f, ff, FTS_SKIP);
738 break;
739 }
740 }
741
742 fts_close(f);
743 return(1);
744 }
745
746 /*
747 * Add a file to the mlinks table.
748 * Do not verify that it's a "valid" looking manpage (we'll do that
749 * later).
750 *
751 * Try to infer the manual section, architecture, and page name from the
752 * path, assuming it looks like
753 *
754 * [./]man*[/<arch>]/<name>.<section>
755 * or
756 * [./]cat<section>[/<arch>]/<name>.0
757 *
758 * See treescan() for the fts(3) version of this.
759 */
760 static void
761 filescan(const char *file)
762 {
763 char buf[PATH_MAX];
764 struct stat st;
765 struct mlink *mlink;
766 char *p, *start;
767
768 assert(use_all);
769
770 if (0 == strncmp(file, "./", 2))
771 file += 2;
772
773 /*
774 * We have to do lstat(2) before realpath(3) loses
775 * the information whether this is a symbolic link.
776 * We need to know that because for symbolic links,
777 * we want to use the orginal file name, while for
778 * regular files, we want to use the real path.
779 */
780 if (-1 == lstat(file, &st)) {
781 exitcode = (int)MANDOCLEVEL_BADARG;
782 say(file, "&lstat");
783 return;
784 } else if (0 == ((S_IFREG | S_IFLNK) & st.st_mode)) {
785 exitcode = (int)MANDOCLEVEL_BADARG;
786 say(file, "Not a regular file");
787 return;
788 }
789
790 /*
791 * We have to resolve the file name to the real path
792 * in any case for the base directory check.
793 */
794 if (NULL == realpath(file, buf)) {
795 exitcode = (int)MANDOCLEVEL_BADARG;
796 say(file, "&realpath");
797 return;
798 }
799
800 if (OP_TEST == op)
801 start = buf;
802 else if (strstr(buf, basedir) == buf)
803 start = buf + strlen(basedir);
804 else {
805 exitcode = (int)MANDOCLEVEL_BADARG;
806 say("", "%s: outside base directory", buf);
807 return;
808 }
809
810 /*
811 * Now we are sure the file is inside our tree.
812 * If it is a symbolic link, ignore the real path
813 * and use the original name.
814 * This implies passing stuff like "cat1/../man1/foo.1"
815 * on the command line won't work. So don't do that.
816 * Note the stat(2) can still fail if the link target
817 * doesn't exist.
818 */
819 if (S_IFLNK & st.st_mode) {
820 if (-1 == stat(buf, &st)) {
821 exitcode = (int)MANDOCLEVEL_BADARG;
822 say(file, "&stat");
823 return;
824 }
825 if (strlcpy(buf, file, sizeof(buf)) >= sizeof(buf)) {
826 say(file, "Filename too long");
827 return;
828 }
829 start = buf;
830 if (OP_TEST != op && strstr(buf, basedir) == buf)
831 start += strlen(basedir);
832 }
833
834 mlink = mandoc_calloc(1, sizeof(struct mlink));
835 if (strlcpy(mlink->file, start, sizeof(mlink->file)) >=
836 sizeof(mlink->file)) {
837 say(start, "Filename too long");
838 return;
839 }
840
841 /*
842 * First try to guess our directory structure.
843 * If we find a separator, try to look for man* or cat*.
844 * If we find one of these and what's underneath is a directory,
845 * assume it's an architecture.
846 */
847 if (NULL != (p = strchr(start, '/'))) {
848 *p++ = '\0';
849 if (0 == strncmp(start, "man", 3)) {
850 mlink->dform = FORM_SRC;
851 mlink->dsec = start + 3;
852 } else if (0 == strncmp(start, "cat", 3)) {
853 mlink->dform = FORM_CAT;
854 mlink->dsec = start + 3;
855 }
856
857 start = p;
858 if (NULL != mlink->dsec && NULL != (p = strchr(start, '/'))) {
859 *p++ = '\0';
860 mlink->arch = start;
861 start = p;
862 }
863 }
864
865 /*
866 * Now check the file suffix.
867 * Suffix of `.0' indicates a catpage, `.1-9' is a manpage.
868 */
869 p = strrchr(start, '\0');
870 while (p-- > start && '/' != *p && '.' != *p)
871 /* Loop. */ ;
872
873 if ('.' == *p) {
874 *p++ = '\0';
875 mlink->fsec = p;
876 }
877
878 /*
879 * Now try to parse the name.
880 * Use the filename portion of the path.
881 */
882 mlink->name = start;
883 if (NULL != (p = strrchr(start, '/'))) {
884 mlink->name = p + 1;
885 *p = '\0';
886 }
887 mlink_add(mlink, &st);
888 }
889
890 static void
891 mlink_add(struct mlink *mlink, const struct stat *st)
892 {
893 struct inodev inodev;
894 struct mpage *mpage;
895 unsigned int slot;
896
897 assert(NULL != mlink->file);
898
899 mlink->dsec = mandoc_strdup(mlink->dsec ? mlink->dsec : "");
900 mlink->arch = mandoc_strdup(mlink->arch ? mlink->arch : "");
901 mlink->name = mandoc_strdup(mlink->name ? mlink->name : "");
902 mlink->fsec = mandoc_strdup(mlink->fsec ? mlink->fsec : "");
903
904 if ('0' == *mlink->fsec) {
905 free(mlink->fsec);
906 mlink->fsec = mandoc_strdup(mlink->dsec);
907 mlink->fform = FORM_CAT;
908 } else if ('1' <= *mlink->fsec && '9' >= *mlink->fsec)
909 mlink->fform = FORM_SRC;
910 else
911 mlink->fform = FORM_NONE;
912
913 slot = ohash_qlookup(&mlinks, mlink->file);
914 assert(NULL == ohash_find(&mlinks, slot));
915 ohash_insert(&mlinks, slot, mlink);
916
917 inodev.st_ino = st->st_ino;
918 inodev.st_dev = st->st_dev;
919 slot = ohash_lookup_memory(&mpages, (char *)&inodev,
920 sizeof(struct inodev), inodev.st_ino);
921 mpage = ohash_find(&mpages, slot);
922 if (NULL == mpage) {
923 mpage = mandoc_calloc(1, sizeof(struct mpage));
924 mpage->inodev.st_ino = inodev.st_ino;
925 mpage->inodev.st_dev = inodev.st_dev;
926 ohash_insert(&mpages, slot, mpage);
927 } else
928 mlink->next = mpage->mlinks;
929 mpage->mlinks = mlink;
930 mlink->mpage = mpage;
931 }
932
933 static void
934 mlink_free(struct mlink *mlink)
935 {
936
937 free(mlink->dsec);
938 free(mlink->arch);
939 free(mlink->name);
940 free(mlink->fsec);
941 free(mlink);
942 }
943
944 static void
945 mpages_free(void)
946 {
947 struct mpage *mpage;
948 struct mlink *mlink;
949 unsigned int slot;
950
951 mpage = ohash_first(&mpages, &slot);
952 while (NULL != mpage) {
953 while (NULL != (mlink = mpage->mlinks)) {
954 mpage->mlinks = mlink->next;
955 mlink_free(mlink);
956 }
957 free(mpage->sec);
958 free(mpage->arch);
959 free(mpage->title);
960 free(mpage->desc);
961 free(mpage);
962 mpage = ohash_next(&mpages, &slot);
963 }
964 }
965
966 /*
967 * For each mlink to the mpage, check whether the path looks like
968 * it is formatted, and if it does, check whether a source manual
969 * exists by the same name, ignoring the suffix.
970 * If both conditions hold, drop the mlink.
971 */
972 static void
973 mlinks_undupe(struct mpage *mpage)
974 {
975 char buf[PATH_MAX];
976 struct mlink **prev;
977 struct mlink *mlink;
978 char *bufp;
979
980 mpage->form = FORM_CAT;
981 prev = &mpage->mlinks;
982 while (NULL != (mlink = *prev)) {
983 if (FORM_CAT != mlink->dform) {
984 mpage->form = FORM_NONE;
985 goto nextlink;
986 }
987 (void)strlcpy(buf, mlink->file, sizeof(buf));
988 bufp = strstr(buf, "cat");
989 assert(NULL != bufp);
990 memcpy(bufp, "man", 3);
991 if (NULL != (bufp = strrchr(buf, '.')))
992 *++bufp = '\0';
993 (void)strlcat(buf, mlink->dsec, sizeof(buf));
994 if (NULL == ohash_find(&mlinks,
995 ohash_qlookup(&mlinks, buf)))
996 goto nextlink;
997 if (warnings)
998 say(mlink->file, "Man source exists: %s", buf);
999 if (use_all)
1000 goto nextlink;
1001 *prev = mlink->next;
1002 mlink_free(mlink);
1003 continue;
1004 nextlink:
1005 prev = &(*prev)->next;
1006 }
1007 }
1008
1009 static void
1010 mlink_check(struct mpage *mpage, struct mlink *mlink)
1011 {
1012 struct str *str;
1013 unsigned int slot;
1014
1015 /*
1016 * Check whether the manual section given in a file
1017 * agrees with the directory where the file is located.
1018 * Some manuals have suffixes like (3p) on their
1019 * section number either inside the file or in the
1020 * directory name, some are linked into more than one
1021 * section, like encrypt(1) = makekey(8).
1022 */
1023
1024 if (FORM_SRC == mpage->form &&
1025 strcasecmp(mpage->sec, mlink->dsec))
1026 say(mlink->file, "Section \"%s\" manual in %s directory",
1027 mpage->sec, mlink->dsec);
1028
1029 /*
1030 * Manual page directories exist for each kernel
1031 * architecture as returned by machine(1).
1032 * However, many manuals only depend on the
1033 * application architecture as returned by arch(1).
1034 * For example, some (2/ARM) manuals are shared
1035 * across the "armish" and "zaurus" kernel
1036 * architectures.
1037 * A few manuals are even shared across completely
1038 * different architectures, for example fdformat(1)
1039 * on amd64, i386, sparc, and sparc64.
1040 */
1041
1042 if (strcasecmp(mpage->arch, mlink->arch))
1043 say(mlink->file, "Architecture \"%s\" manual in "
1044 "\"%s\" directory", mpage->arch, mlink->arch);
1045
1046 /*
1047 * XXX
1048 * parse_cat() doesn't set NAME_TITLE yet.
1049 */
1050
1051 if (FORM_CAT == mpage->form)
1052 return;
1053
1054 /*
1055 * Check whether this mlink
1056 * appears as a name in the NAME section.
1057 */
1058
1059 slot = ohash_qlookup(&names, mlink->name);
1060 str = ohash_find(&names, slot);
1061 assert(NULL != str);
1062 if ( ! (NAME_TITLE & str->mask))
1063 say(mlink->file, "Name missing in NAME section");
1064 }
1065
1066 /*
1067 * Run through the files in the global vector "mpages"
1068 * and add them to the database specified in "basedir".
1069 *
1070 * This handles the parsing scheme itself, using the cues of directory
1071 * and filename to determine whether the file is parsable or not.
1072 */
1073 static void
1074 mpages_merge(struct mchars *mc, struct mparse *mp)
1075 {
1076 char any[] = "any";
1077 struct ohash_info str_info;
1078 int fd[2];
1079 struct mpage *mpage, *mpage_dest;
1080 struct mlink *mlink, *mlink_dest;
1081 struct mdoc *mdoc;
1082 struct man *man;
1083 char *sodest;
1084 char *cp;
1085 pid_t child_pid;
1086 int status;
1087 unsigned int pslot;
1088 enum mandoclevel lvl;
1089
1090 str_info.alloc = hash_alloc;
1091 str_info.calloc = hash_calloc;
1092 str_info.free = hash_free;
1093 str_info.key_offset = offsetof(struct str, key);
1094
1095 if (0 == nodb)
1096 SQL_EXEC("BEGIN TRANSACTION");
1097
1098 mpage = ohash_first(&mpages, &pslot);
1099 while (NULL != mpage) {
1100 mlinks_undupe(mpage);
1101 if (NULL == mpage->mlinks) {
1102 mpage = ohash_next(&mpages, &pslot);
1103 continue;
1104 }
1105
1106 name_mask = NAME_MASK;
1107 ohash_init(&names, 4, &str_info);
1108 ohash_init(&strings, 6, &str_info);
1109 mparse_reset(mp);
1110 mdoc = NULL;
1111 man = NULL;
1112 sodest = NULL;
1113 child_pid = 0;
1114 fd[0] = -1;
1115 fd[1] = -1;
1116
1117 if (mpage->mlinks->gzip) {
1118 if (-1 == pipe(fd)) {
1119 exitcode = (int)MANDOCLEVEL_SYSERR;
1120 say(mpage->mlinks->file, "&pipe gunzip");
1121 goto nextpage;
1122 }
1123 switch (child_pid = fork()) {
1124 case -1:
1125 exitcode = (int)MANDOCLEVEL_SYSERR;
1126 say(mpage->mlinks->file, "&fork gunzip");
1127 child_pid = 0;
1128 close(fd[1]);
1129 close(fd[0]);
1130 goto nextpage;
1131 case 0:
1132 close(fd[0]);
1133 if (-1 == dup2(fd[1], STDOUT_FILENO)) {
1134 say(mpage->mlinks->file,
1135 "&dup gunzip");
1136 exit(1);
1137 }
1138 execlp("gunzip", "gunzip", "-c",
1139 mpage->mlinks->file, NULL);
1140 say(mpage->mlinks->file, "&exec gunzip");
1141 exit(1);
1142 default:
1143 close(fd[1]);
1144 break;
1145 }
1146 }
1147
1148 /*
1149 * Try interpreting the file as mdoc(7) or man(7)
1150 * source code, unless it is already known to be
1151 * formatted. Fall back to formatted mode.
1152 */
1153 if (FORM_CAT != mpage->mlinks->dform ||
1154 FORM_CAT != mpage->mlinks->fform) {
1155 lvl = mparse_readfd(mp, fd[0], mpage->mlinks->file);
1156 if (lvl < MANDOCLEVEL_FATAL)
1157 mparse_result(mp, &mdoc, &man, &sodest);
1158 }
1159
1160 if (NULL != sodest) {
1161 mlink_dest = ohash_find(&mlinks,
1162 ohash_qlookup(&mlinks, sodest));
1163 if (NULL != mlink_dest) {
1164
1165 /* The .so target exists. */
1166
1167 mpage_dest = mlink_dest->mpage;
1168 mlink = mpage->mlinks;
1169 while (1) {
1170 mlink->mpage = mpage_dest;
1171
1172 /*
1173 * If the target was already
1174 * processed, add the links
1175 * to the database now.
1176 * Otherwise, this will
1177 * happen when we come
1178 * to the target.
1179 */
1180
1181 if (mpage_dest->pageid)
1182 dbadd_mlink(mlink);
1183
1184 if (NULL == mlink->next)
1185 break;
1186 mlink = mlink->next;
1187 }
1188
1189 /* Move all links to the target. */
1190
1191 mlink->next = mlink_dest->next;
1192 mlink_dest->next = mpage->mlinks;
1193 mpage->mlinks = NULL;
1194 }
1195 goto nextpage;
1196 } else if (NULL != mdoc) {
1197 mpage->form = FORM_SRC;
1198 mpage->sec =
1199 mandoc_strdup(mdoc_meta(mdoc)->msec);
1200 mpage->arch = mdoc_meta(mdoc)->arch;
1201 mpage->arch = mandoc_strdup(
1202 NULL == mpage->arch ? "" : mpage->arch);
1203 mpage->title =
1204 mandoc_strdup(mdoc_meta(mdoc)->title);
1205 } else if (NULL != man) {
1206 mpage->form = FORM_SRC;
1207 mpage->sec =
1208 mandoc_strdup(man_meta(man)->msec);
1209 mpage->arch =
1210 mandoc_strdup(mpage->mlinks->arch);
1211 mpage->title =
1212 mandoc_strdup(man_meta(man)->title);
1213 } else {
1214 mpage->form = FORM_CAT;
1215 mpage->sec =
1216 mandoc_strdup(mpage->mlinks->dsec);
1217 mpage->arch =
1218 mandoc_strdup(mpage->mlinks->arch);
1219 mpage->title =
1220 mandoc_strdup(mpage->mlinks->name);
1221 }
1222 putkey(mpage, mpage->sec, TYPE_sec);
1223 putkey(mpage, '\0' == *mpage->arch ?
1224 any : mpage->arch, TYPE_arch);
1225
1226 for (mlink = mpage->mlinks; mlink; mlink = mlink->next) {
1227 if ('\0' != *mlink->dsec)
1228 putkey(mpage, mlink->dsec, TYPE_sec);
1229 if ('\0' != *mlink->fsec)
1230 putkey(mpage, mlink->fsec, TYPE_sec);
1231 putkey(mpage, '\0' == *mlink->arch ?
1232 any : mlink->arch, TYPE_arch);
1233 putkey(mpage, mlink->name, NAME_FILE);
1234 }
1235
1236 assert(NULL == mpage->desc);
1237 if (NULL != mdoc) {
1238 if (NULL != (cp = mdoc_meta(mdoc)->name))
1239 putkey(mpage, cp, NAME_HEAD);
1240 parse_mdoc(mpage, mdoc_node(mdoc));
1241 } else if (NULL != man)
1242 parse_man(mpage, man_node(man));
1243 else
1244 parse_cat(mpage, fd[0]);
1245 if (NULL == mpage->desc)
1246 mpage->desc = mandoc_strdup(mpage->mlinks->name);
1247
1248 if (warnings && !use_all)
1249 for (mlink = mpage->mlinks; mlink;
1250 mlink = mlink->next)
1251 mlink_check(mpage, mlink);
1252
1253 dbadd(mpage, mc);
1254
1255 nextpage:
1256 if (child_pid) {
1257 if (-1 == waitpid(child_pid, &status, 0)) {
1258 exitcode = (int)MANDOCLEVEL_SYSERR;
1259 say(mpage->mlinks->file, "&wait gunzip");
1260 } else if (WIFSIGNALED(status)) {
1261 exitcode = (int)MANDOCLEVEL_SYSERR;
1262 say(mpage->mlinks->file,
1263 "gunzip died from signal %d",
1264 WTERMSIG(status));
1265 } else if (WEXITSTATUS(status)) {
1266 exitcode = (int)MANDOCLEVEL_SYSERR;
1267 say(mpage->mlinks->file,
1268 "gunzip failed with code %d",
1269 WEXITSTATUS(status));
1270 }
1271 }
1272 ohash_delete(&strings);
1273 ohash_delete(&names);
1274 mpage = ohash_next(&mpages, &pslot);
1275 }
1276
1277 if (0 == nodb)
1278 SQL_EXEC("END TRANSACTION");
1279 }
1280
1281 static void
1282 names_check(void)
1283 {
1284 sqlite3_stmt *stmt;
1285 const char *name, *sec, *arch, *key;
1286 int irc;
1287
1288 sqlite3_prepare_v2(db,
1289 "SELECT name, sec, arch, key FROM ("
1290 "SELECT name AS key, pageid FROM names "
1291 "WHERE bits & ? AND NOT EXISTS ("
1292 "SELECT pageid FROM mlinks "
1293 "WHERE mlinks.pageid == names.pageid "
1294 "AND mlinks.name == names.name"
1295 ")"
1296 ") JOIN ("
1297 "SELECT sec, arch, name, pageid FROM mlinks "
1298 "GROUP BY pageid"
1299 ") USING (pageid);",
1300 -1, &stmt, NULL);
1301
1302 if (SQLITE_OK != sqlite3_bind_int64(stmt, 1, NAME_TITLE))
1303 say("", "%s", sqlite3_errmsg(db));
1304
1305 while (SQLITE_ROW == (irc = sqlite3_step(stmt))) {
1306 name = sqlite3_column_text(stmt, 0);
1307 sec = sqlite3_column_text(stmt, 1);
1308 arch = sqlite3_column_text(stmt, 2);
1309 key = sqlite3_column_text(stmt, 3);
1310 say("", "%s(%s%s%s) lacks mlink \"%s\"", name, sec,
1311 '\0' == *arch ? "" : "/",
1312 '\0' == *arch ? "" : arch, key);
1313 }
1314 sqlite3_finalize(stmt);
1315 }
1316
1317 static void
1318 parse_cat(struct mpage *mpage, int fd)
1319 {
1320 FILE *stream;
1321 char *line, *p, *title;
1322 size_t len, plen, titlesz;
1323
1324 stream = (-1 == fd) ?
1325 fopen(mpage->mlinks->file, "r") :
1326 fdopen(fd, "r");
1327 if (NULL == stream) {
1328 if (warnings)
1329 say(mpage->mlinks->file, "&fopen");
1330 return;
1331 }
1332
1333 /* Skip to first blank line. */
1334
1335 while (NULL != (line = fgetln(stream, &len)))
1336 if ('\n' == *line)
1337 break;
1338
1339 /*
1340 * Assume the first line that is not indented
1341 * is the first section header. Skip to it.
1342 */
1343
1344 while (NULL != (line = fgetln(stream, &len)))
1345 if ('\n' != *line && ' ' != *line)
1346 break;
1347
1348 /*
1349 * Read up until the next section into a buffer.
1350 * Strip the leading and trailing newline from each read line,
1351 * appending a trailing space.
1352 * Ignore empty (whitespace-only) lines.
1353 */
1354
1355 titlesz = 0;
1356 title = NULL;
1357
1358 while (NULL != (line = fgetln(stream, &len))) {
1359 if (' ' != *line || '\n' != line[len - 1])
1360 break;
1361 while (len > 0 && isspace((unsigned char)*line)) {
1362 line++;
1363 len--;
1364 }
1365 if (1 == len)
1366 continue;
1367 title = mandoc_realloc(title, titlesz + len);
1368 memcpy(title + titlesz, line, len);
1369 titlesz += len;
1370 title[titlesz - 1] = ' ';
1371 }
1372
1373 /*
1374 * If no page content can be found, or the input line
1375 * is already the next section header, or there is no
1376 * trailing newline, reuse the page title as the page
1377 * description.
1378 */
1379
1380 if (NULL == title || '\0' == *title) {
1381 if (warnings)
1382 say(mpage->mlinks->file,
1383 "Cannot find NAME section");
1384 fclose(stream);
1385 free(title);
1386 return;
1387 }
1388
1389 title = mandoc_realloc(title, titlesz + 1);
1390 title[titlesz] = '\0';
1391
1392 /*
1393 * Skip to the first dash.
1394 * Use the remaining line as the description (no more than 70
1395 * bytes).
1396 */
1397
1398 if (NULL != (p = strstr(title, "- "))) {
1399 for (p += 2; ' ' == *p || '\b' == *p; p++)
1400 /* Skip to next word. */ ;
1401 } else {
1402 if (warnings)
1403 say(mpage->mlinks->file,
1404 "No dash in title line");
1405 p = title;
1406 }
1407
1408 plen = strlen(p);
1409
1410 /* Strip backspace-encoding from line. */
1411
1412 while (NULL != (line = memchr(p, '\b', plen))) {
1413 len = line - p;
1414 if (0 == len) {
1415 memmove(line, line + 1, plen--);
1416 continue;
1417 }
1418 memmove(line - 1, line + 1, plen - len);
1419 plen -= 2;
1420 }
1421
1422 mpage->desc = mandoc_strdup(p);
1423 fclose(stream);
1424 free(title);
1425 }
1426
1427 /*
1428 * Put a type/word pair into the word database for this particular file.
1429 */
1430 static void
1431 putkey(const struct mpage *mpage, char *value, uint64_t type)
1432 {
1433 char *cp;
1434
1435 assert(NULL != value);
1436 if (TYPE_arch == type)
1437 for (cp = value; *cp; cp++)
1438 if (isupper((unsigned char)*cp))
1439 *cp = _tolower((unsigned char)*cp);
1440 putkeys(mpage, value, strlen(value), type);
1441 }
1442
1443 /*
1444 * Grok all nodes at or below a certain mdoc node into putkey().
1445 */
1446 static void
1447 putmdockey(const struct mpage *mpage,
1448 const struct mdoc_node *n, uint64_t m)
1449 {
1450
1451 for ( ; NULL != n; n = n->next) {
1452 if (NULL != n->child)
1453 putmdockey(mpage, n->child, m);
1454 if (MDOC_TEXT == n->type)
1455 putkey(mpage, n->string, m);
1456 }
1457 }
1458
1459 static void
1460 parse_man(struct mpage *mpage, const struct man_node *n)
1461 {
1462 const struct man_node *head, *body;
1463 char *start, *title;
1464 char byte;
1465 size_t sz;
1466
1467 if (NULL == n)
1468 return;
1469
1470 /*
1471 * We're only searching for one thing: the first text child in
1472 * the BODY of a NAME section. Since we don't keep track of
1473 * sections in -man, run some hoops to find out whether we're in
1474 * the correct section or not.
1475 */
1476
1477 if (MAN_BODY == n->type && MAN_SH == n->tok) {
1478 body = n;
1479 assert(body->parent);
1480 if (NULL != (head = body->parent->head) &&
1481 1 == head->nchild &&
1482 NULL != (head = (head->child)) &&
1483 MAN_TEXT == head->type &&
1484 0 == strcmp(head->string, "NAME") &&
1485 NULL != body->child) {
1486
1487 /*
1488 * Suck the entire NAME section into memory.
1489 * Yes, we might run away.
1490 * But too many manuals have big, spread-out
1491 * NAME sections over many lines.
1492 */
1493
1494 title = NULL;
1495 man_deroff(&title, body);
1496 if (NULL == title)
1497 return;
1498
1499 /*
1500 * Go through a special heuristic dance here.
1501 * Conventionally, one or more manual names are
1502 * comma-specified prior to a whitespace, then a
1503 * dash, then a description. Try to puzzle out
1504 * the name parts here.
1505 */
1506
1507 start = title;
1508 for ( ;; ) {
1509 sz = strcspn(start, " ,");
1510 if ('\0' == start[sz])
1511 break;
1512
1513 byte = start[sz];
1514 start[sz] = '\0';
1515
1516 /*
1517 * Assume a stray trailing comma in the
1518 * name list if a name begins with a dash.
1519 */
1520
1521 if ('-' == start[0] ||
1522 ('\\' == start[0] && '-' == start[1]))
1523 break;
1524
1525 putkey(mpage, start, NAME_TITLE);
1526
1527 if (' ' == byte) {
1528 start += sz + 1;
1529 break;
1530 }
1531
1532 assert(',' == byte);
1533 start += sz + 1;
1534 while (' ' == *start)
1535 start++;
1536 }
1537
1538 if (start == title) {
1539 putkey(mpage, start, NAME_TITLE);
1540 free(title);
1541 return;
1542 }
1543
1544 while (isspace((unsigned char)*start))
1545 start++;
1546
1547 if (0 == strncmp(start, "-", 1))
1548 start += 1;
1549 else if (0 == strncmp(start, "\\-\\-", 4))
1550 start += 4;
1551 else if (0 == strncmp(start, "\\-", 2))
1552 start += 2;
1553 else if (0 == strncmp(start, "\\(en", 4))
1554 start += 4;
1555 else if (0 == strncmp(start, "\\(em", 4))
1556 start += 4;
1557
1558 while (' ' == *start)
1559 start++;
1560
1561 mpage->desc = mandoc_strdup(start);
1562 free(title);
1563 return;
1564 }
1565 }
1566
1567 for (n = n->child; n; n = n->next) {
1568 if (NULL != mpage->desc)
1569 break;
1570 parse_man(mpage, n);
1571 }
1572 }
1573
1574 static void
1575 parse_mdoc(struct mpage *mpage, const struct mdoc_node *n)
1576 {
1577
1578 assert(NULL != n);
1579 for (n = n->child; NULL != n; n = n->next) {
1580 switch (n->type) {
1581 case MDOC_ELEM:
1582 /* FALLTHROUGH */
1583 case MDOC_BLOCK:
1584 /* FALLTHROUGH */
1585 case MDOC_HEAD:
1586 /* FALLTHROUGH */
1587 case MDOC_BODY:
1588 /* FALLTHROUGH */
1589 case MDOC_TAIL:
1590 if (NULL != mdocs[n->tok].fp)
1591 if (0 == (*mdocs[n->tok].fp)(mpage, n))
1592 break;
1593 if (mdocs[n->tok].mask)
1594 putmdockey(mpage, n->child,
1595 mdocs[n->tok].mask);
1596 break;
1597 default:
1598 assert(MDOC_ROOT != n->type);
1599 continue;
1600 }
1601 if (NULL != n->child)
1602 parse_mdoc(mpage, n);
1603 }
1604 }
1605
1606 static int
1607 parse_mdoc_Fd(struct mpage *mpage, const struct mdoc_node *n)
1608 {
1609 const char *start, *end;
1610 size_t sz;
1611
1612 if (SEC_SYNOPSIS != n->sec ||
1613 NULL == (n = n->child) ||
1614 MDOC_TEXT != n->type)
1615 return(0);
1616
1617 /*
1618 * Only consider those `Fd' macro fields that begin with an
1619 * "inclusion" token (versus, e.g., #define).
1620 */
1621
1622 if (strcmp("#include", n->string))
1623 return(0);
1624
1625 if (NULL == (n = n->next) || MDOC_TEXT != n->type)
1626 return(0);
1627
1628 /*
1629 * Strip away the enclosing angle brackets and make sure we're
1630 * not zero-length.
1631 */
1632
1633 start = n->string;
1634 if ('<' == *start || '"' == *start)
1635 start++;
1636
1637 if (0 == (sz = strlen(start)))
1638 return(0);
1639
1640 end = &start[(int)sz - 1];
1641 if ('>' == *end || '"' == *end)
1642 end--;
1643
1644 if (end > start)
1645 putkeys(mpage, start, end - start + 1, TYPE_In);
1646 return(0);
1647 }
1648
1649 static int
1650 parse_mdoc_Fn(struct mpage *mpage, const struct mdoc_node *n)
1651 {
1652 char *cp;
1653
1654 if (NULL == (n = n->child) || MDOC_TEXT != n->type)
1655 return(0);
1656
1657 /*
1658 * Parse: .Fn "struct type *name" "char *arg".
1659 * First strip away pointer symbol.
1660 * Then store the function name, then type.
1661 * Finally, store the arguments.
1662 */
1663
1664 if (NULL == (cp = strrchr(n->string, ' ')))
1665 cp = n->string;
1666
1667 while ('*' == *cp)
1668 cp++;
1669
1670 putkey(mpage, cp, TYPE_Fn);
1671
1672 if (n->string < cp)
1673 putkeys(mpage, n->string, cp - n->string, TYPE_Ft);
1674
1675 for (n = n->next; NULL != n; n = n->next)
1676 if (MDOC_TEXT == n->type)
1677 putkey(mpage, n->string, TYPE_Fa);
1678
1679 return(0);
1680 }
1681
1682 static int
1683 parse_mdoc_Xr(struct mpage *mpage, const struct mdoc_node *n)
1684 {
1685 char *cp;
1686
1687 if (NULL == (n = n->child))
1688 return(0);
1689
1690 if (NULL == n->next) {
1691 putkey(mpage, n->string, TYPE_Xr);
1692 return(0);
1693 }
1694
1695 mandoc_asprintf(&cp, "%s(%s)", n->string, n->next->string);
1696 putkey(mpage, cp, TYPE_Xr);
1697 free(cp);
1698 return(0);
1699 }
1700
1701 static int
1702 parse_mdoc_Nd(struct mpage *mpage, const struct mdoc_node *n)
1703 {
1704
1705 if (MDOC_BODY == n->type)
1706 mdoc_deroff(&mpage->desc, n);
1707 return(0);
1708 }
1709
1710 static int
1711 parse_mdoc_Nm(struct mpage *mpage, const struct mdoc_node *n)
1712 {
1713
1714 if (SEC_NAME == n->sec)
1715 putmdockey(mpage, n->child, NAME_TITLE);
1716 else if (SEC_SYNOPSIS == n->sec && MDOC_HEAD == n->type)
1717 putmdockey(mpage, n->child, NAME_SYN);
1718 return(0);
1719 }
1720
1721 static int
1722 parse_mdoc_Sh(struct mpage *mpage, const struct mdoc_node *n)
1723 {
1724
1725 return(SEC_CUSTOM == n->sec && MDOC_HEAD == n->type);
1726 }
1727
1728 static int
1729 parse_mdoc_head(struct mpage *mpage, const struct mdoc_node *n)
1730 {
1731
1732 return(MDOC_HEAD == n->type);
1733 }
1734
1735 static int
1736 parse_mdoc_body(struct mpage *mpage, const struct mdoc_node *n)
1737 {
1738
1739 return(MDOC_BODY == n->type);
1740 }
1741
1742 /*
1743 * Add a string to the hash table for the current manual.
1744 * Each string has a bitmask telling which macros it belongs to.
1745 * When we finish the manual, we'll dump the table.
1746 */
1747 static void
1748 putkeys(const struct mpage *mpage,
1749 const char *cp, size_t sz, uint64_t v)
1750 {
1751 struct ohash *htab;
1752 struct str *s;
1753 const char *end;
1754 unsigned int slot;
1755 int i;
1756
1757 if (0 == sz)
1758 return;
1759
1760 if (TYPE_Nm & v) {
1761 htab = &names;
1762 v &= name_mask;
1763 name_mask &= ~NAME_FIRST;
1764 if (debug > 1)
1765 say(mpage->mlinks->file,
1766 "Adding name %*s", sz, cp);
1767 } else {
1768 htab = &strings;
1769 if (debug > 1)
1770 for (i = 0; i < mansearch_keymax; i++)
1771 if (1 << i & v)
1772 say(mpage->mlinks->file,
1773 "Adding key %s=%*s",
1774 mansearch_keynames[i], sz, cp);
1775 }
1776
1777 end = cp + sz;
1778 slot = ohash_qlookupi(htab, cp, &end);
1779 s = ohash_find(htab, slot);
1780
1781 if (NULL != s && mpage == s->mpage) {
1782 s->mask |= v;
1783 return;
1784 } else if (NULL == s) {
1785 s = mandoc_calloc(1, sizeof(struct str) + sz + 1);
1786 memcpy(s->key, cp, sz);
1787 ohash_insert(htab, slot, s);
1788 }
1789 s->mpage = mpage;
1790 s->mask = v;
1791 }
1792
1793 /*
1794 * Take a Unicode codepoint and produce its UTF-8 encoding.
1795 * This isn't the best way to do this, but it works.
1796 * The magic numbers are from the UTF-8 packaging.
1797 * They're not as scary as they seem: read the UTF-8 spec for details.
1798 */
1799 static size_t
1800 utf8(unsigned int cp, char out[7])
1801 {
1802 size_t rc;
1803
1804 rc = 0;
1805 if (cp <= 0x0000007F) {
1806 rc = 1;
1807 out[0] = (char)cp;
1808 } else if (cp <= 0x000007FF) {
1809 rc = 2;
1810 out[0] = (cp >> 6 & 31) | 192;
1811 out[1] = (cp & 63) | 128;
1812 } else if (cp <= 0x0000FFFF) {
1813 rc = 3;
1814 out[0] = (cp >> 12 & 15) | 224;
1815 out[1] = (cp >> 6 & 63) | 128;
1816 out[2] = (cp & 63) | 128;
1817 } else if (cp <= 0x001FFFFF) {
1818 rc = 4;
1819 out[0] = (cp >> 18 & 7) | 240;
1820 out[1] = (cp >> 12 & 63) | 128;
1821 out[2] = (cp >> 6 & 63) | 128;
1822 out[3] = (cp & 63) | 128;
1823 } else if (cp <= 0x03FFFFFF) {
1824 rc = 5;
1825 out[0] = (cp >> 24 & 3) | 248;
1826 out[1] = (cp >> 18 & 63) | 128;
1827 out[2] = (cp >> 12 & 63) | 128;
1828 out[3] = (cp >> 6 & 63) | 128;
1829 out[4] = (cp & 63) | 128;
1830 } else if (cp <= 0x7FFFFFFF) {
1831 rc = 6;
1832 out[0] = (cp >> 30 & 1) | 252;
1833 out[1] = (cp >> 24 & 63) | 128;
1834 out[2] = (cp >> 18 & 63) | 128;
1835 out[3] = (cp >> 12 & 63) | 128;
1836 out[4] = (cp >> 6 & 63) | 128;
1837 out[5] = (cp & 63) | 128;
1838 } else
1839 return(0);
1840
1841 out[rc] = '\0';
1842 return(rc);
1843 }
1844
1845 /*
1846 * Store the rendered version of a key, or alias the pointer
1847 * if the key contains no escape sequences.
1848 */
1849 static void
1850 render_key(struct mchars *mc, struct str *key)
1851 {
1852 size_t sz, bsz, pos;
1853 char utfbuf[7], res[6];
1854 char *buf;
1855 const char *seq, *cpp, *val;
1856 int len, u;
1857 enum mandoc_esc esc;
1858
1859 assert(NULL == key->rendered);
1860
1861 res[0] = '\\';
1862 res[1] = '\t';
1863 res[2] = ASCII_NBRSP;
1864 res[3] = ASCII_HYPH;
1865 res[4] = ASCII_BREAK;
1866 res[5] = '\0';
1867
1868 val = key->key;
1869 bsz = strlen(val);
1870
1871 /*
1872 * Pre-check: if we have no stop-characters, then set the
1873 * pointer as ourselvse and get out of here.
1874 */
1875 if (strcspn(val, res) == bsz) {
1876 key->rendered = key->key;
1877 return;
1878 }
1879
1880 /* Pre-allocate by the length of the input */
1881
1882 buf = mandoc_malloc(++bsz);
1883 pos = 0;
1884
1885 while ('\0' != *val) {
1886 /*
1887 * Halt on the first escape sequence.
1888 * This also halts on the end of string, in which case
1889 * we just copy, fallthrough, and exit the loop.
1890 */
1891 if ((sz = strcspn(val, res)) > 0) {
1892 memcpy(&buf[pos], val, sz);
1893 pos += sz;
1894 val += sz;
1895 }
1896
1897 switch (*val) {
1898 case ASCII_HYPH:
1899 buf[pos++] = '-';
1900 val++;
1901 continue;
1902 case '\t':
1903 /* FALLTHROUGH */
1904 case ASCII_NBRSP:
1905 buf[pos++] = ' ';
1906 val++;
1907 /* FALLTHROUGH */
1908 case ASCII_BREAK:
1909 continue;
1910 default:
1911 break;
1912 }
1913 if ('\\' != *val)
1914 break;
1915
1916 /* Read past the slash. */
1917
1918 val++;
1919
1920 /*
1921 * Parse the escape sequence and see if it's a
1922 * predefined character or special character.
1923 */
1924
1925 esc = mandoc_escape((const char **)&val,
1926 &seq, &len);
1927 if (ESCAPE_ERROR == esc)
1928 break;
1929 if (ESCAPE_SPECIAL != esc)
1930 continue;
1931
1932 /*
1933 * Render the special character
1934 * as either UTF-8 or ASCII.
1935 */
1936
1937 if (write_utf8) {
1938 if (0 == (u = mchars_spec2cp(mc, seq, len)))
1939 continue;
1940 cpp = utfbuf;
1941 if (0 == (sz = utf8(u, utfbuf)))
1942 continue;
1943 sz = strlen(cpp);
1944 } else {
1945 cpp = mchars_spec2str(mc, seq, len, &sz);
1946 if (NULL == cpp)
1947 continue;
1948 if (ASCII_NBRSP == *cpp) {
1949 cpp = " ";
1950 sz = 1;
1951 }
1952 }
1953
1954 /* Copy the rendered glyph into the stream. */
1955
1956 bsz += sz;
1957 buf = mandoc_realloc(buf, bsz);
1958 memcpy(&buf[pos], cpp, sz);
1959 pos += sz;
1960 }
1961
1962 buf[pos] = '\0';
1963 key->rendered = buf;
1964 }
1965
1966 static void
1967 dbadd_mlink(const struct mlink *mlink)
1968 {
1969 size_t i;
1970
1971 i = 1;
1972 SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->dsec);
1973 SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->arch);
1974 SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->name);
1975 SQL_BIND_INT64(stmts[STMT_INSERT_LINK], i, mlink->mpage->pageid);
1976 SQL_STEP(stmts[STMT_INSERT_LINK]);
1977 sqlite3_reset(stmts[STMT_INSERT_LINK]);
1978 }
1979
1980 /*
1981 * Flush the current page's terms (and their bits) into the database.
1982 * Wrap the entire set of additions in a transaction to make sqlite be a
1983 * little faster.
1984 * Also, handle escape sequences at the last possible moment.
1985 */
1986 static void
1987 dbadd(struct mpage *mpage, struct mchars *mc)
1988 {
1989 struct mlink *mlink;
1990 struct str *key;
1991 size_t i;
1992 unsigned int slot;
1993
1994 mlink = mpage->mlinks;
1995
1996 if (nodb) {
1997 for (key = ohash_first(&names, &slot); NULL != key;
1998 key = ohash_next(&names, &slot)) {
1999 if (key->rendered != key->key)
2000 free(key->rendered);
2001 free(key);
2002 }
2003 for (key = ohash_first(&strings, &slot); NULL != key;
2004 key = ohash_next(&strings, &slot)) {
2005 if (key->rendered != key->key)
2006 free(key->rendered);
2007 free(key);
2008 }
2009 if (0 == debug)
2010 return;
2011 while (NULL != mlink) {
2012 fputs(mlink->name, stdout);
2013 if (NULL == mlink->next ||
2014 strcmp(mlink->dsec, mlink->next->dsec) ||
2015 strcmp(mlink->fsec, mlink->next->fsec) ||
2016 strcmp(mlink->arch, mlink->next->arch)) {
2017 putchar('(');
2018 if ('\0' == *mlink->dsec)
2019 fputs(mlink->fsec, stdout);
2020 else
2021 fputs(mlink->dsec, stdout);
2022 if ('\0' != *mlink->arch)
2023 printf("/%s", mlink->arch);
2024 putchar(')');
2025 }
2026 mlink = mlink->next;
2027 if (NULL != mlink)
2028 fputs(", ", stdout);
2029 }
2030 printf(" - %s\n", mpage->desc);
2031 return;
2032 }
2033
2034 if (debug)
2035 say(mlink->file, "Adding to database");
2036
2037 i = strlen(mpage->desc) + 1;
2038 key = mandoc_calloc(1, sizeof(struct str) + i);
2039 memcpy(key->key, mpage->desc, i);
2040 render_key(mc, key);
2041
2042 i = 1;
2043 SQL_BIND_TEXT(stmts[STMT_INSERT_PAGE], i, key->rendered);
2044 SQL_BIND_INT(stmts[STMT_INSERT_PAGE], i, FORM_SRC == mpage->form);
2045 SQL_STEP(stmts[STMT_INSERT_PAGE]);
2046 mpage->pageid = sqlite3_last_insert_rowid(db);
2047 sqlite3_reset(stmts[STMT_INSERT_PAGE]);
2048
2049 if (key->rendered != key->key)
2050 free(key->rendered);
2051 free(key);
2052
2053 while (NULL != mlink) {
2054 dbadd_mlink(mlink);
2055 mlink = mlink->next;
2056 }
2057 mlink = mpage->mlinks;
2058
2059 for (key = ohash_first(&names, &slot); NULL != key;
2060 key = ohash_next(&names, &slot)) {
2061 assert(key->mpage == mpage);
2062 if (NULL == key->rendered)
2063 render_key(mc, key);
2064 i = 1;
2065 SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, key->mask);
2066 SQL_BIND_TEXT(stmts[STMT_INSERT_NAME], i, key->rendered);
2067 SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, mpage->pageid);
2068 SQL_STEP(stmts[STMT_INSERT_NAME]);
2069 sqlite3_reset(stmts[STMT_INSERT_NAME]);
2070 if (key->rendered != key->key)
2071 free(key->rendered);
2072 free(key);
2073 }
2074 for (key = ohash_first(&strings, &slot); NULL != key;
2075 key = ohash_next(&strings, &slot)) {
2076 assert(key->mpage == mpage);
2077 if (NULL == key->rendered)
2078 render_key(mc, key);
2079 i = 1;
2080 SQL_BIND_INT64(stmts[STMT_INSERT_KEY], i, key->mask);
2081 SQL_BIND_TEXT(stmts[STMT_INSERT_KEY], i, key->rendered);
2082 SQL_BIND_INT64(stmts[STMT_INSERT_KEY], i, mpage->pageid);
2083 SQL_STEP(stmts[STMT_INSERT_KEY]);
2084 sqlite3_reset(stmts[STMT_INSERT_KEY]);
2085 if (key->rendered != key->key)
2086 free(key->rendered);
2087 free(key);
2088 }
2089 }
2090
2091 static void
2092 dbprune(void)
2093 {
2094 struct mpage *mpage;
2095 struct mlink *mlink;
2096 size_t i;
2097 unsigned int slot;
2098
2099 if (0 == nodb)
2100 SQL_EXEC("BEGIN TRANSACTION");
2101
2102 for (mpage = ohash_first(&mpages, &slot); NULL != mpage;
2103 mpage = ohash_next(&mpages, &slot)) {
2104 mlink = mpage->mlinks;
2105 if (debug)
2106 say(mlink->file, "Deleting from database");
2107 if (nodb)
2108 continue;
2109 for ( ; NULL != mlink; mlink = mlink->next) {
2110 i = 1;
2111 SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE],
2112 i, mlink->dsec);
2113 SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE],
2114 i, mlink->arch);
2115 SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE],
2116 i, mlink->name);
2117 SQL_STEP(stmts[STMT_DELETE_PAGE]);
2118 sqlite3_reset(stmts[STMT_DELETE_PAGE]);
2119 }
2120 }
2121
2122 if (0 == nodb)
2123 SQL_EXEC("END TRANSACTION");
2124 }
2125
2126 /*
2127 * Close an existing database and its prepared statements.
2128 * If "real" is not set, rename the temporary file into the real one.
2129 */
2130 static void
2131 dbclose(int real)
2132 {
2133 size_t i;
2134 int status;
2135 pid_t child;
2136
2137 if (nodb)
2138 return;
2139
2140 for (i = 0; i < STMT__MAX; i++) {
2141 sqlite3_finalize(stmts[i]);
2142 stmts[i] = NULL;
2143 }
2144
2145 sqlite3_close(db);
2146 db = NULL;
2147
2148 if (real)
2149 return;
2150
2151 if ('\0' == *tempfilename) {
2152 if (-1 == rename(MANDOC_DB "~", MANDOC_DB)) {
2153 exitcode = (int)MANDOCLEVEL_SYSERR;
2154 say(MANDOC_DB, "&rename");
2155 }
2156 return;
2157 }
2158
2159 switch (child = fork()) {
2160 case -1:
2161 exitcode = (int)MANDOCLEVEL_SYSERR;
2162 say("", "&fork cmp");
2163 return;
2164 case 0:
2165 execlp("cmp", "cmp", "-s",
2166 tempfilename, MANDOC_DB, NULL);
2167 say("", "&exec cmp");
2168 exit(0);
2169 default:
2170 break;
2171 }
2172 if (-1 == waitpid(child, &status, 0)) {
2173 exitcode = (int)MANDOCLEVEL_SYSERR;
2174 say("", "&wait cmp");
2175 } else if (WIFSIGNALED(status)) {
2176 exitcode = (int)MANDOCLEVEL_SYSERR;
2177 say("", "cmp died from signal %d", WTERMSIG(status));
2178 } else if (WEXITSTATUS(status)) {
2179 exitcode = (int)MANDOCLEVEL_SYSERR;
2180 say(MANDOC_DB,
2181 "Data changed, but cannot replace database");
2182 }
2183
2184 *strrchr(tempfilename, '/') = '\0';
2185 switch (child = fork()) {
2186 case -1:
2187 exitcode = (int)MANDOCLEVEL_SYSERR;
2188 say("", "&fork rm");
2189 return;
2190 case 0:
2191 execlp("rm", "rm", "-rf", tempfilename, NULL);
2192 say("", "&exec rm");
2193 exit((int)MANDOCLEVEL_SYSERR);
2194 default:
2195 break;
2196 }
2197 if (-1 == waitpid(child, &status, 0)) {
2198 exitcode = (int)MANDOCLEVEL_SYSERR;
2199 say("", "&wait rm");
2200 } else if (WIFSIGNALED(status) || WEXITSTATUS(status)) {
2201 exitcode = (int)MANDOCLEVEL_SYSERR;
2202 say("", "%s: Cannot remove temporary directory",
2203 tempfilename);
2204 }
2205 }
2206
2207 /*
2208 * This is straightforward stuff.
2209 * Open a database connection to a "temporary" database, then open a set
2210 * of prepared statements we'll use over and over again.
2211 * If "real" is set, we use the existing database; if not, we truncate a
2212 * temporary one.
2213 * Must be matched by dbclose().
2214 */
2215 static int
2216 dbopen(int real)
2217 {
2218 const char *sql;
2219 int rc, ofl;
2220
2221 if (nodb)
2222 return(1);
2223
2224 *tempfilename = '\0';
2225 ofl = SQLITE_OPEN_READWRITE;
2226
2227 if (real) {
2228 rc = sqlite3_open_v2(MANDOC_DB, &db, ofl, NULL);
2229 if (SQLITE_OK != rc) {
2230 exitcode = (int)MANDOCLEVEL_SYSERR;
2231 if (SQLITE_CANTOPEN != rc)
2232 say(MANDOC_DB, "%s", sqlite3_errstr(rc));
2233 return(0);
2234 }
2235 goto prepare_statements;
2236 }
2237
2238 ofl |= SQLITE_OPEN_CREATE | SQLITE_OPEN_EXCLUSIVE;
2239
2240 remove(MANDOC_DB "~");
2241 rc = sqlite3_open_v2(MANDOC_DB "~", &db, ofl, NULL);
2242 if (SQLITE_OK == rc)
2243 goto create_tables;
2244 if (MPARSE_QUICK & mparse_options) {
2245 exitcode = (int)MANDOCLEVEL_SYSERR;
2246 say(MANDOC_DB "~", "%s", sqlite3_errstr(rc));
2247 return(0);
2248 }
2249
2250 (void)strlcpy(tempfilename, "/tmp/mandocdb.XXXXXX",
2251 sizeof(tempfilename));
2252 if (NULL == mkdtemp(tempfilename)) {
2253 exitcode = (int)MANDOCLEVEL_SYSERR;
2254 say("", "&%s", tempfilename);
2255 return(0);
2256 }
2257 (void)strlcat(tempfilename, "/" MANDOC_DB,
2258 sizeof(tempfilename));
2259 rc = sqlite3_open_v2(tempfilename, &db, ofl, NULL);
2260 if (SQLITE_OK != rc) {
2261 exitcode = (int)MANDOCLEVEL_SYSERR;
2262 say("", "%s: %s", tempfilename, sqlite3_errstr(rc));
2263 return(0);
2264 }
2265
2266 create_tables:
2267 sql = "CREATE TABLE \"mpages\" (\n"
2268 " \"desc\" TEXT NOT NULL,\n"
2269 " \"form\" INTEGER NOT NULL,\n"
2270 " \"pageid\" INTEGER PRIMARY KEY AUTOINCREMENT NOT NULL\n"
2271 ");\n"
2272 "\n"
2273 "CREATE TABLE \"mlinks\" (\n"
2274 " \"sec\" TEXT NOT NULL,\n"
2275 " \"arch\" TEXT NOT NULL,\n"
2276 " \"name\" TEXT NOT NULL,\n"
2277 " \"pageid\" INTEGER NOT NULL REFERENCES mpages(pageid) "
2278 "ON DELETE CASCADE\n"
2279 ");\n"
2280 "CREATE INDEX mlinks_pageid_idx ON mlinks (pageid);\n"
2281 "\n"
2282 "CREATE TABLE \"names\" (\n"
2283 " \"bits\" INTEGER NOT NULL,\n"
2284 " \"name\" TEXT NOT NULL,\n"
2285 " \"pageid\" INTEGER NOT NULL REFERENCES mpages(pageid) "
2286 "ON DELETE CASCADE\n"
2287 ");\n"
2288 "\n"
2289 "CREATE TABLE \"keys\" (\n"
2290 " \"bits\" INTEGER NOT NULL,\n"
2291 " \"key\" TEXT NOT NULL,\n"
2292 " \"pageid\" INTEGER NOT NULL REFERENCES mpages(pageid) "
2293 "ON DELETE CASCADE\n"
2294 ");\n"
2295 "CREATE INDEX keys_pageid_idx ON keys (pageid);\n";
2296
2297 if (SQLITE_OK != sqlite3_exec(db, sql, NULL, NULL, NULL)) {
2298 exitcode = (int)MANDOCLEVEL_SYSERR;
2299 say(MANDOC_DB, "%s", sqlite3_errmsg(db));
2300 sqlite3_close(db);
2301 return(0);
2302 }
2303
2304 prepare_statements:
2305 if (SQLITE_OK != sqlite3_exec(db,
2306 "PRAGMA foreign_keys = ON", NULL, NULL, NULL)) {
2307 exitcode = (int)MANDOCLEVEL_SYSERR;
2308 say(MANDOC_DB, "PRAGMA foreign_keys: %s",
2309 sqlite3_errmsg(db));
2310 sqlite3_close(db);
2311 return(0);
2312 }
2313
2314 sql = "DELETE FROM mpages WHERE pageid IN "
2315 "(SELECT pageid FROM mlinks WHERE "
2316 "sec=? AND arch=? AND name=?)";
2317 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_DELETE_PAGE], NULL);
2318 sql = "INSERT INTO mpages "
2319 "(desc,form) VALUES (?,?)";
2320 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_PAGE], NULL);
2321 sql = "INSERT INTO mlinks "
2322 "(sec,arch,name,pageid) VALUES (?,?,?,?)";
2323 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_LINK], NULL);
2324 sql = "INSERT INTO names "
2325 "(bits,name,pageid) VALUES (?,?,?)";
2326 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_NAME], NULL);
2327 sql = "INSERT INTO keys "
2328 "(bits,key,pageid) VALUES (?,?,?)";
2329 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_KEY], NULL);
2330
2331 #ifndef __APPLE__
2332 /*
2333 * When opening a new database, we can turn off
2334 * synchronous mode for much better performance.
2335 */
2336
2337 if (real && SQLITE_OK != sqlite3_exec(db,
2338 "PRAGMA synchronous = OFF", NULL, NULL, NULL)) {
2339 exitcode = (int)MANDOCLEVEL_SYSERR;
2340 say(MANDOC_DB, "PRAGMA synchronous: %s",
2341 sqlite3_errmsg(db));
2342 sqlite3_close(db);
2343 return(0);
2344 }
2345 #endif
2346
2347 return(1);
2348 }
2349
2350 static void *
2351 hash_calloc(size_t n, size_t sz, void *arg)
2352 {
2353
2354 return(mandoc_calloc(n, sz));
2355 }
2356
2357 static void *
2358 hash_alloc(size_t sz, void *arg)
2359 {
2360
2361 return(mandoc_malloc(sz));
2362 }
2363
2364 static void
2365 hash_free(void *p, void *arg)
2366 {
2367
2368 free(p);
2369 }
2370
2371 static int
2372 set_basedir(const char *targetdir)
2373 {
2374 static char startdir[PATH_MAX];
2375 static int getcwd_status; /* 1 = ok, 2 = failure */
2376 static int chdir_status; /* 1 = changed directory */
2377 char *cp;
2378
2379 /*
2380 * Remember the original working directory, if possible.
2381 * This will be needed if the second or a later directory
2382 * on the command line is given as a relative path.
2383 * Do not error out if the current directory is not
2384 * searchable: Maybe it won't be needed after all.
2385 */
2386 if (0 == getcwd_status) {
2387 if (NULL == getcwd(startdir, sizeof(startdir))) {
2388 getcwd_status = 2;
2389 (void)strlcpy(startdir, strerror(errno),
2390 sizeof(startdir));
2391 } else
2392 getcwd_status = 1;
2393 }
2394
2395 /*
2396 * We are leaving the old base directory.
2397 * Do not use it any longer, not even for messages.
2398 */
2399 *basedir = '\0';
2400
2401 /*
2402 * If and only if the directory was changed earlier and
2403 * the next directory to process is given as a relative path,
2404 * first go back, or bail out if that is impossible.
2405 */
2406 if (chdir_status && '/' != *targetdir) {
2407 if (2 == getcwd_status) {
2408 exitcode = (int)MANDOCLEVEL_SYSERR;
2409 say("", "getcwd: %s", startdir);
2410 return(0);
2411 }
2412 if (-1 == chdir(startdir)) {
2413 exitcode = (int)MANDOCLEVEL_SYSERR;
2414 say("", "&chdir %s", startdir);
2415 return(0);
2416 }
2417 }
2418
2419 /*
2420 * Always resolve basedir to the canonicalized absolute
2421 * pathname and append a trailing slash, such that
2422 * we can reliably check whether files are inside.
2423 */
2424 if (NULL == realpath(targetdir, basedir)) {
2425 exitcode = (int)MANDOCLEVEL_BADARG;
2426 say("", "&%s: realpath", targetdir);
2427 return(0);
2428 } else if (-1 == chdir(basedir)) {
2429 exitcode = (int)MANDOCLEVEL_BADARG;
2430 say("", "&chdir");
2431 return(0);
2432 }
2433 chdir_status = 1;
2434 cp = strchr(basedir, '\0');
2435 if ('/' != cp[-1]) {
2436 if (cp - basedir >= PATH_MAX - 1) {
2437 exitcode = (int)MANDOCLEVEL_SYSERR;
2438 say("", "Filename too long");
2439 return(0);
2440 }
2441 *cp++ = '/';
2442 *cp = '\0';
2443 }
2444 return(1);
2445 }
2446
2447 static void
2448 say(const char *file, const char *format, ...)
2449 {
2450 va_list ap;
2451 int use_errno;
2452
2453 if ('\0' != *basedir)
2454 fprintf(stderr, "%s", basedir);
2455 if ('\0' != *basedir && '\0' != *file)
2456 fputc('/', stderr);
2457 if ('\0' != *file)
2458 fprintf(stderr, "%s", file);
2459
2460 use_errno = 1;
2461 if (NULL != format) {
2462 switch (*format) {
2463 case '&':
2464 format++;
2465 break;
2466 case '\0':
2467 format = NULL;
2468 break;
2469 default:
2470 use_errno = 0;
2471 break;
2472 }
2473 }
2474 if (NULL != format) {
2475 if ('\0' != *basedir || '\0' != *file)
2476 fputs(": ", stderr);
2477 va_start(ap, format);
2478 vfprintf(stderr, format, ap);
2479 va_end(ap);
2480 }
2481 if (use_errno) {
2482 if ('\0' != *basedir || '\0' != *file || NULL != format)
2483 fputs(": ", stderr);
2484 perror(NULL);
2485 } else
2486 fputc('\n', stderr);
2487 }
A git conversion mirror of mandoc.bsd.lv