]> 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
Warn about missing mlinks.
[mandoc.git] / mandocdb.c
1 /* $Id: mandocdb.c,v 1.130 2014/04/04 16:43:16 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 fprintf(stderr, "%s\n", 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 fprintf(stderr, "%s\n", sqlite3_errmsg(db))
63 #define SQL_BIND_INT(_s, _i, _v) \
64 if (SQLITE_OK != sqlite3_bind_int \
65 ((_s), (_i)++, (_v))) \
66 fprintf(stderr, "%s\n", sqlite3_errmsg(db))
67 #define SQL_BIND_INT64(_s, _i, _v) \
68 if (SQLITE_OK != sqlite3_bind_int64 \
69 ((_s), (_i)++, (_v))) \
70 fprintf(stderr, "%s\n", sqlite3_errmsg(db))
71 #define SQL_STEP(_s) \
72 if (SQLITE_DONE != sqlite3_step((_s))) \
73 fprintf(stderr, "%s\n", 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 recno; /* id 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_KEY, /* insert parsed key */
130 STMT__MAX
131 };
132
133 typedef int (*mdoc_fp)(struct mpage *, 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 *, struct mchars *);
142 static void dbadd_mlink(const struct mlink *mlink);
143 static int dbopen(int);
144 static void dbprune(void);
145 static void filescan(const char *);
146 static void *hash_alloc(size_t, void *);
147 static void hash_free(void *, size_t, void *);
148 static void *hash_halloc(size_t, void *);
149 static void mlink_add(struct mlink *, const struct stat *);
150 static int mlink_check(struct mpage *, struct mlink *);
151 static void mlink_free(struct mlink *);
152 static void mlinks_undupe(struct mpage *);
153 static void mpages_free(void);
154 static void mpages_merge(struct mchars *, struct mparse *);
155 static void names_check(void);
156 static void parse_cat(struct mpage *, int);
157 static void parse_man(struct mpage *, const struct man_node *);
158 static void parse_mdoc(struct mpage *, const struct mdoc_node *);
159 static int parse_mdoc_body(struct mpage *, const struct mdoc_node *);
160 static int parse_mdoc_head(struct mpage *, const struct mdoc_node *);
161 static int parse_mdoc_Fd(struct mpage *, const struct mdoc_node *);
162 static int parse_mdoc_Fn(struct mpage *, const struct mdoc_node *);
163 static int parse_mdoc_Nd(struct mpage *, const struct mdoc_node *);
164 static int parse_mdoc_Nm(struct mpage *, const struct mdoc_node *);
165 static int parse_mdoc_Sh(struct mpage *, const struct mdoc_node *);
166 static int parse_mdoc_Xr(struct mpage *, const struct mdoc_node *);
167 static void putkey(const struct mpage *, char *, uint64_t);
168 static void putkeys(const struct mpage *,
169 const char *, size_t, uint64_t);
170 static void putmdockey(const struct mpage *,
171 const struct mdoc_node *, uint64_t);
172 static void render_key(struct mchars *, struct str *);
173 static void say(const char *, const char *, ...);
174 static int set_basedir(const char *);
175 static int treescan(void);
176 static size_t utf8(unsigned int, char [7]);
177
178 static char tempfilename[32];
179 static char *progname;
180 static int nodb; /* no database changes */
181 static int mparse_options; /* abort the parse early */
182 static int use_all; /* use all found files */
183 static int debug; /* print what we're doing */
184 static int warnings; /* warn about crap */
185 static int write_utf8; /* write UTF-8 output; else ASCII */
186 static int exitcode; /* to be returned by main */
187 static enum op op; /* operational mode */
188 static char basedir[PATH_MAX]; /* current base directory */
189 static struct ohash mpages; /* table of distinct manual pages */
190 static struct ohash mlinks; /* table of directory entries */
191 static struct ohash strings; /* table of all strings */
192 static sqlite3 *db = NULL; /* current database */
193 static sqlite3_stmt *stmts[STMT__MAX]; /* current statements */
194
195 static const struct mdoc_handler mdocs[MDOC_MAX] = {
196 { NULL, 0 }, /* Ap */
197 { NULL, 0 }, /* Dd */
198 { NULL, 0 }, /* Dt */
199 { NULL, 0 }, /* Os */
200 { parse_mdoc_Sh, TYPE_Sh }, /* Sh */
201 { parse_mdoc_head, TYPE_Ss }, /* Ss */
202 { NULL, 0 }, /* Pp */
203 { NULL, 0 }, /* D1 */
204 { NULL, 0 }, /* Dl */
205 { NULL, 0 }, /* Bd */
206 { NULL, 0 }, /* Ed */
207 { NULL, 0 }, /* Bl */
208 { NULL, 0 }, /* El */
209 { NULL, 0 }, /* It */
210 { NULL, 0 }, /* Ad */
211 { NULL, TYPE_An }, /* An */
212 { NULL, TYPE_Ar }, /* Ar */
213 { NULL, TYPE_Cd }, /* Cd */
214 { NULL, TYPE_Cm }, /* Cm */
215 { NULL, TYPE_Dv }, /* Dv */
216 { NULL, TYPE_Er }, /* Er */
217 { NULL, TYPE_Ev }, /* Ev */
218 { NULL, 0 }, /* Ex */
219 { NULL, TYPE_Fa }, /* Fa */
220 { parse_mdoc_Fd, 0 }, /* Fd */
221 { NULL, TYPE_Fl }, /* Fl */
222 { parse_mdoc_Fn, 0 }, /* Fn */
223 { NULL, TYPE_Ft }, /* Ft */
224 { NULL, TYPE_Ic }, /* Ic */
225 { NULL, TYPE_In }, /* In */
226 { NULL, TYPE_Li }, /* Li */
227 { parse_mdoc_Nd, TYPE_Nd }, /* Nd */
228 { parse_mdoc_Nm, 0 }, /* Nm */
229 { NULL, 0 }, /* Op */
230 { NULL, 0 }, /* Ot */
231 { NULL, TYPE_Pa }, /* Pa */
232 { NULL, 0 }, /* Rv */
233 { NULL, TYPE_St }, /* St */
234 { NULL, TYPE_Va }, /* Va */
235 { parse_mdoc_body, TYPE_Va }, /* Vt */
236 { parse_mdoc_Xr, 0 }, /* Xr */
237 { NULL, 0 }, /* %A */
238 { NULL, 0 }, /* %B */
239 { NULL, 0 }, /* %D */
240 { NULL, 0 }, /* %I */
241 { NULL, 0 }, /* %J */
242 { NULL, 0 }, /* %N */
243 { NULL, 0 }, /* %O */
244 { NULL, 0 }, /* %P */
245 { NULL, 0 }, /* %R */
246 { NULL, 0 }, /* %T */
247 { NULL, 0 }, /* %V */
248 { NULL, 0 }, /* Ac */
249 { NULL, 0 }, /* Ao */
250 { NULL, 0 }, /* Aq */
251 { NULL, TYPE_At }, /* At */
252 { NULL, 0 }, /* Bc */
253 { NULL, 0 }, /* Bf */
254 { NULL, 0 }, /* Bo */
255 { NULL, 0 }, /* Bq */
256 { NULL, TYPE_Bsx }, /* Bsx */
257 { NULL, TYPE_Bx }, /* Bx */
258 { NULL, 0 }, /* Db */
259 { NULL, 0 }, /* Dc */
260 { NULL, 0 }, /* Do */
261 { NULL, 0 }, /* Dq */
262 { NULL, 0 }, /* Ec */
263 { NULL, 0 }, /* Ef */
264 { NULL, TYPE_Em }, /* Em */
265 { NULL, 0 }, /* Eo */
266 { NULL, TYPE_Fx }, /* Fx */
267 { NULL, TYPE_Ms }, /* Ms */
268 { NULL, 0 }, /* No */
269 { NULL, 0 }, /* Ns */
270 { NULL, TYPE_Nx }, /* Nx */
271 { NULL, TYPE_Ox }, /* Ox */
272 { NULL, 0 }, /* Pc */
273 { NULL, 0 }, /* Pf */
274 { NULL, 0 }, /* Po */
275 { NULL, 0 }, /* Pq */
276 { NULL, 0 }, /* Qc */
277 { NULL, 0 }, /* Ql */
278 { NULL, 0 }, /* Qo */
279 { NULL, 0 }, /* Qq */
280 { NULL, 0 }, /* Re */
281 { NULL, 0 }, /* Rs */
282 { NULL, 0 }, /* Sc */
283 { NULL, 0 }, /* So */
284 { NULL, 0 }, /* Sq */
285 { NULL, 0 }, /* Sm */
286 { NULL, 0 }, /* Sx */
287 { NULL, TYPE_Sy }, /* Sy */
288 { NULL, TYPE_Tn }, /* Tn */
289 { NULL, 0 }, /* Ux */
290 { NULL, 0 }, /* Xc */
291 { NULL, 0 }, /* Xo */
292 { parse_mdoc_head, 0 }, /* Fo */
293 { NULL, 0 }, /* Fc */
294 { NULL, 0 }, /* Oo */
295 { NULL, 0 }, /* Oc */
296 { NULL, 0 }, /* Bk */
297 { NULL, 0 }, /* Ek */
298 { NULL, 0 }, /* Bt */
299 { NULL, 0 }, /* Hf */
300 { NULL, 0 }, /* Fr */
301 { NULL, 0 }, /* Ud */
302 { NULL, TYPE_Lb }, /* Lb */
303 { NULL, 0 }, /* Lp */
304 { NULL, TYPE_Lk }, /* Lk */
305 { NULL, TYPE_Mt }, /* Mt */
306 { NULL, 0 }, /* Brq */
307 { NULL, 0 }, /* Bro */
308 { NULL, 0 }, /* Brc */
309 { NULL, 0 }, /* %C */
310 { NULL, 0 }, /* Es */
311 { NULL, 0 }, /* En */
312 { NULL, TYPE_Dx }, /* Dx */
313 { NULL, 0 }, /* %Q */
314 { NULL, 0 }, /* br */
315 { NULL, 0 }, /* sp */
316 { NULL, 0 }, /* %U */
317 { NULL, 0 }, /* Ta */
318 };
319
320 int
321 main(int argc, char *argv[])
322 {
323 int ch, i;
324 size_t j, sz;
325 const char *path_arg;
326 struct mchars *mc;
327 struct manpaths dirs;
328 struct mparse *mp;
329 struct ohash_info mpages_info, mlinks_info;
330
331 memset(stmts, 0, STMT__MAX * sizeof(sqlite3_stmt *));
332 memset(&dirs, 0, sizeof(struct manpaths));
333
334 mpages_info.alloc = mlinks_info.alloc = hash_alloc;
335 mpages_info.halloc = mlinks_info.halloc = hash_halloc;
336 mpages_info.hfree = mlinks_info.hfree = hash_free;
337
338 mpages_info.key_offset = offsetof(struct mpage, inodev);
339 mlinks_info.key_offset = offsetof(struct mlink, file);
340
341 progname = strrchr(argv[0], '/');
342 if (progname == NULL)
343 progname = argv[0];
344 else
345 ++progname;
346
347 /*
348 * We accept a few different invocations.
349 * The CHECKOP macro makes sure that invocation styles don't
350 * clobber each other.
351 */
352 #define CHECKOP(_op, _ch) do \
353 if (OP_DEFAULT != (_op)) { \
354 fprintf(stderr, "-%c: Conflicting option\n", (_ch)); \
355 goto usage; \
356 } while (/*CONSTCOND*/0)
357
358 path_arg = NULL;
359 op = OP_DEFAULT;
360
361 while (-1 != (ch = getopt(argc, argv, "aC:Dd:npQT:tu:v")))
362 switch (ch) {
363 case ('a'):
364 use_all = 1;
365 break;
366 case ('C'):
367 CHECKOP(op, ch);
368 path_arg = optarg;
369 op = OP_CONFFILE;
370 break;
371 case ('D'):
372 debug++;
373 break;
374 case ('d'):
375 CHECKOP(op, ch);
376 path_arg = optarg;
377 op = OP_UPDATE;
378 break;
379 case ('n'):
380 nodb = 1;
381 break;
382 case ('p'):
383 warnings = 1;
384 break;
385 case ('Q'):
386 mparse_options |= MPARSE_QUICK;
387 break;
388 case ('T'):
389 if (strcmp(optarg, "utf8")) {
390 fprintf(stderr, "-T%s: Unsupported "
391 "output format\n", optarg);
392 goto usage;
393 }
394 write_utf8 = 1;
395 break;
396 case ('t'):
397 CHECKOP(op, ch);
398 dup2(STDOUT_FILENO, STDERR_FILENO);
399 op = OP_TEST;
400 nodb = warnings = 1;
401 break;
402 case ('u'):
403 CHECKOP(op, ch);
404 path_arg = optarg;
405 op = OP_DELETE;
406 break;
407 case ('v'):
408 /* Compatibility with espie@'s makewhatis. */
409 break;
410 default:
411 goto usage;
412 }
413
414 argc -= optind;
415 argv += optind;
416
417 if (OP_CONFFILE == op && argc > 0) {
418 fprintf(stderr, "-C: Too many arguments\n");
419 goto usage;
420 }
421
422 exitcode = (int)MANDOCLEVEL_OK;
423 mp = mparse_alloc(mparse_options, MANDOCLEVEL_FATAL, NULL, NULL);
424 mc = mchars_alloc();
425
426 ohash_init(&mpages, 6, &mpages_info);
427 ohash_init(&mlinks, 6, &mlinks_info);
428
429 if (OP_UPDATE == op || OP_DELETE == op || OP_TEST == op) {
430 /*
431 * Force processing all files.
432 */
433 use_all = 1;
434
435 /*
436 * All of these deal with a specific directory.
437 * Jump into that directory then collect files specified
438 * on the command-line.
439 */
440 if (0 == set_basedir(path_arg))
441 goto out;
442 for (i = 0; i < argc; i++)
443 filescan(argv[i]);
444 if (0 == dbopen(1))
445 goto out;
446 if (OP_TEST != op)
447 dbprune();
448 if (OP_DELETE != op)
449 mpages_merge(mc, mp);
450 dbclose(1);
451 } else {
452 /*
453 * If we have arguments, use them as our manpaths.
454 * If we don't, grok from manpath(1) or however else
455 * manpath_parse() wants to do it.
456 */
457 if (argc > 0) {
458 dirs.paths = mandoc_calloc
459 (argc, sizeof(char *));
460 dirs.sz = (size_t)argc;
461 for (i = 0; i < argc; i++)
462 dirs.paths[i] = mandoc_strdup(argv[i]);
463 } else
464 manpath_parse(&dirs, path_arg, NULL, NULL);
465
466 if (0 == dirs.sz) {
467 exitcode = (int)MANDOCLEVEL_BADARG;
468 say("", "Empty manpath");
469 }
470
471 /*
472 * First scan the tree rooted at a base directory, then
473 * build a new database and finally move it into place.
474 * Ignore zero-length directories and strip trailing
475 * slashes.
476 */
477 for (j = 0; j < dirs.sz; j++) {
478 sz = strlen(dirs.paths[j]);
479 if (sz && '/' == dirs.paths[j][sz - 1])
480 dirs.paths[j][--sz] = '\0';
481 if (0 == sz)
482 continue;
483
484 if (j) {
485 ohash_init(&mpages, 6, &mpages_info);
486 ohash_init(&mlinks, 6, &mlinks_info);
487 }
488
489 if (0 == set_basedir(dirs.paths[j]))
490 goto out;
491 if (0 == treescan())
492 goto out;
493 if (0 == set_basedir(dirs.paths[j]))
494 goto out;
495 if (0 == dbopen(0))
496 goto out;
497
498 mpages_merge(mc, mp);
499 if (warnings &&
500 ! (MPARSE_QUICK & mparse_options))
501 names_check();
502 dbclose(0);
503
504 if (j + 1 < dirs.sz) {
505 mpages_free();
506 ohash_delete(&mpages);
507 ohash_delete(&mlinks);
508 }
509 }
510 }
511 out:
512 set_basedir(NULL);
513 manpath_free(&dirs);
514 mchars_free(mc);
515 mparse_free(mp);
516 mpages_free();
517 ohash_delete(&mpages);
518 ohash_delete(&mlinks);
519 return(exitcode);
520 usage:
521 fprintf(stderr, "usage: %s [-aDnpQ] [-C file] [-Tutf8]\n"
522 " %s [-aDnpQ] [-Tutf8] dir ...\n"
523 " %s [-DnpQ] [-Tutf8] -d dir [file ...]\n"
524 " %s [-Dnp] -u dir [file ...]\n"
525 " %s [-Q] -t file ...\n",
526 progname, progname, progname,
527 progname, progname);
528
529 return((int)MANDOCLEVEL_BADARG);
530 }
531
532 /*
533 * Scan a directory tree rooted at "basedir" for manpages.
534 * We use fts(), scanning directory parts along the way for clues to our
535 * section and architecture.
536 *
537 * If use_all has been specified, grok all files.
538 * If not, sanitise paths to the following:
539 *
540 * [./]man*[/<arch>]/<name>.<section>
541 * or
542 * [./]cat<section>[/<arch>]/<name>.0
543 *
544 * TODO: accomodate for multi-language directories.
545 */
546 static int
547 treescan(void)
548 {
549 FTS *f;
550 FTSENT *ff;
551 struct mlink *mlink;
552 int dform, gzip;
553 char *dsec, *arch, *fsec, *cp;
554 const char *path;
555 const char *argv[2];
556
557 argv[0] = ".";
558 argv[1] = (char *)NULL;
559
560 /*
561 * Walk through all components under the directory, using the
562 * logical descent of files.
563 */
564 f = fts_open((char * const *)argv, FTS_LOGICAL, NULL);
565 if (NULL == f) {
566 exitcode = (int)MANDOCLEVEL_SYSERR;
567 say("", "&fts_open");
568 return(0);
569 }
570
571 dsec = arch = NULL;
572 dform = FORM_NONE;
573
574 while (NULL != (ff = fts_read(f))) {
575 path = ff->fts_path + 2;
576 /*
577 * If we're a regular file, add an mlink by using the
578 * stored directory data and handling the filename.
579 */
580 if (FTS_F == ff->fts_info) {
581 if (0 == strcmp(path, MANDOC_DB))
582 continue;
583 if ( ! use_all && ff->fts_level < 2) {
584 if (warnings)
585 say(path, "Extraneous file");
586 continue;
587 }
588 gzip = 0;
589 fsec = NULL;
590 while (NULL == fsec) {
591 fsec = strrchr(ff->fts_name, '.');
592 if (NULL == fsec || strcmp(fsec+1, "gz"))
593 break;
594 gzip = 1;
595 *fsec = '\0';
596 fsec = NULL;
597 }
598 if (NULL == fsec) {
599 if ( ! use_all) {
600 if (warnings)
601 say(path,
602 "No filename suffix");
603 continue;
604 }
605 } else if (0 == strcmp(++fsec, "html")) {
606 if (warnings)
607 say(path, "Skip html");
608 continue;
609 } else if (0 == strcmp(fsec, "ps")) {
610 if (warnings)
611 say(path, "Skip ps");
612 continue;
613 } else if (0 == strcmp(fsec, "pdf")) {
614 if (warnings)
615 say(path, "Skip pdf");
616 continue;
617 } else if ( ! use_all &&
618 ((FORM_SRC == dform && strcmp(fsec, dsec)) ||
619 (FORM_CAT == dform && strcmp(fsec, "0")))) {
620 if (warnings)
621 say(path, "Wrong filename suffix");
622 continue;
623 } else
624 fsec[-1] = '\0';
625
626 mlink = mandoc_calloc(1, sizeof(struct mlink));
627 strlcpy(mlink->file, path, sizeof(mlink->file));
628 mlink->dform = dform;
629 mlink->dsec = dsec;
630 mlink->arch = arch;
631 mlink->name = ff->fts_name;
632 mlink->fsec = fsec;
633 mlink->gzip = gzip;
634 mlink_add(mlink, ff->fts_statp);
635 continue;
636 } else if (FTS_D != ff->fts_info &&
637 FTS_DP != ff->fts_info) {
638 if (warnings)
639 say(path, "Not a regular file");
640 continue;
641 }
642
643 switch (ff->fts_level) {
644 case (0):
645 /* Ignore the root directory. */
646 break;
647 case (1):
648 /*
649 * This might contain manX/ or catX/.
650 * Try to infer this from the name.
651 * If we're not in use_all, enforce it.
652 */
653 cp = ff->fts_name;
654 if (FTS_DP == ff->fts_info)
655 break;
656
657 if (0 == strncmp(cp, "man", 3)) {
658 dform = FORM_SRC;
659 dsec = cp + 3;
660 } else if (0 == strncmp(cp, "cat", 3)) {
661 dform = FORM_CAT;
662 dsec = cp + 3;
663 } else {
664 dform = FORM_NONE;
665 dsec = NULL;
666 }
667
668 if (NULL != dsec || use_all)
669 break;
670
671 if (warnings)
672 say(path, "Unknown directory part");
673 fts_set(f, ff, FTS_SKIP);
674 break;
675 case (2):
676 /*
677 * Possibly our architecture.
678 * If we're descending, keep tabs on it.
679 */
680 if (FTS_DP != ff->fts_info && NULL != dsec)
681 arch = ff->fts_name;
682 else
683 arch = NULL;
684 break;
685 default:
686 if (FTS_DP == ff->fts_info || use_all)
687 break;
688 if (warnings)
689 say(path, "Extraneous directory part");
690 fts_set(f, ff, FTS_SKIP);
691 break;
692 }
693 }
694
695 fts_close(f);
696 return(1);
697 }
698
699 /*
700 * Add a file to the mlinks table.
701 * Do not verify that it's a "valid" looking manpage (we'll do that
702 * later).
703 *
704 * Try to infer the manual section, architecture, and page name from the
705 * path, assuming it looks like
706 *
707 * [./]man*[/<arch>]/<name>.<section>
708 * or
709 * [./]cat<section>[/<arch>]/<name>.0
710 *
711 * See treescan() for the fts(3) version of this.
712 */
713 static void
714 filescan(const char *file)
715 {
716 char buf[PATH_MAX];
717 struct stat st;
718 struct mlink *mlink;
719 char *p, *start;
720
721 assert(use_all);
722
723 if (0 == strncmp(file, "./", 2))
724 file += 2;
725
726 if (NULL == realpath(file, buf)) {
727 exitcode = (int)MANDOCLEVEL_BADARG;
728 say(file, "&realpath");
729 return;
730 }
731
732 if (strstr(buf, basedir) == buf)
733 start = buf + strlen(basedir) + 1;
734 else if (OP_TEST == op)
735 start = buf;
736 else {
737 exitcode = (int)MANDOCLEVEL_BADARG;
738 say("", "%s: outside base directory", buf);
739 return;
740 }
741
742 if (-1 == stat(buf, &st)) {
743 exitcode = (int)MANDOCLEVEL_BADARG;
744 say(file, "&stat");
745 return;
746 } else if ( ! (S_IFREG & st.st_mode)) {
747 exitcode = (int)MANDOCLEVEL_BADARG;
748 say(file, "Not a regular file");
749 return;
750 }
751
752 mlink = mandoc_calloc(1, sizeof(struct mlink));
753 strlcpy(mlink->file, start, sizeof(mlink->file));
754
755 /*
756 * First try to guess our directory structure.
757 * If we find a separator, try to look for man* or cat*.
758 * If we find one of these and what's underneath is a directory,
759 * assume it's an architecture.
760 */
761 if (NULL != (p = strchr(start, '/'))) {
762 *p++ = '\0';
763 if (0 == strncmp(start, "man", 3)) {
764 mlink->dform = FORM_SRC;
765 mlink->dsec = start + 3;
766 } else if (0 == strncmp(start, "cat", 3)) {
767 mlink->dform = FORM_CAT;
768 mlink->dsec = start + 3;
769 }
770
771 start = p;
772 if (NULL != mlink->dsec && NULL != (p = strchr(start, '/'))) {
773 *p++ = '\0';
774 mlink->arch = start;
775 start = p;
776 }
777 }
778
779 /*
780 * Now check the file suffix.
781 * Suffix of `.0' indicates a catpage, `.1-9' is a manpage.
782 */
783 p = strrchr(start, '\0');
784 while (p-- > start && '/' != *p && '.' != *p)
785 /* Loop. */ ;
786
787 if ('.' == *p) {
788 *p++ = '\0';
789 mlink->fsec = p;
790 }
791
792 /*
793 * Now try to parse the name.
794 * Use the filename portion of the path.
795 */
796 mlink->name = start;
797 if (NULL != (p = strrchr(start, '/'))) {
798 mlink->name = p + 1;
799 *p = '\0';
800 }
801 mlink_add(mlink, &st);
802 }
803
804 static void
805 mlink_add(struct mlink *mlink, const struct stat *st)
806 {
807 struct inodev inodev;
808 struct mpage *mpage;
809 unsigned int slot;
810
811 assert(NULL != mlink->file);
812
813 mlink->dsec = mandoc_strdup(mlink->dsec ? mlink->dsec : "");
814 mlink->arch = mandoc_strdup(mlink->arch ? mlink->arch : "");
815 mlink->name = mandoc_strdup(mlink->name ? mlink->name : "");
816 mlink->fsec = mandoc_strdup(mlink->fsec ? mlink->fsec : "");
817
818 if ('0' == *mlink->fsec) {
819 free(mlink->fsec);
820 mlink->fsec = mandoc_strdup(mlink->dsec);
821 mlink->fform = FORM_CAT;
822 } else if ('1' <= *mlink->fsec && '9' >= *mlink->fsec)
823 mlink->fform = FORM_SRC;
824 else
825 mlink->fform = FORM_NONE;
826
827 slot = ohash_qlookup(&mlinks, mlink->file);
828 assert(NULL == ohash_find(&mlinks, slot));
829 ohash_insert(&mlinks, slot, mlink);
830
831 inodev.st_ino = st->st_ino;
832 inodev.st_dev = st->st_dev;
833 slot = ohash_lookup_memory(&mpages, (char *)&inodev,
834 sizeof(struct inodev), inodev.st_ino);
835 mpage = ohash_find(&mpages, slot);
836 if (NULL == mpage) {
837 mpage = mandoc_calloc(1, sizeof(struct mpage));
838 mpage->inodev.st_ino = inodev.st_ino;
839 mpage->inodev.st_dev = inodev.st_dev;
840 ohash_insert(&mpages, slot, mpage);
841 } else
842 mlink->next = mpage->mlinks;
843 mpage->mlinks = mlink;
844 mlink->mpage = mpage;
845 }
846
847 static void
848 mlink_free(struct mlink *mlink)
849 {
850
851 free(mlink->dsec);
852 free(mlink->arch);
853 free(mlink->name);
854 free(mlink->fsec);
855 free(mlink);
856 }
857
858 static void
859 mpages_free(void)
860 {
861 struct mpage *mpage;
862 struct mlink *mlink;
863 unsigned int slot;
864
865 mpage = ohash_first(&mpages, &slot);
866 while (NULL != mpage) {
867 while (NULL != (mlink = mpage->mlinks)) {
868 mpage->mlinks = mlink->next;
869 mlink_free(mlink);
870 }
871 free(mpage->sec);
872 free(mpage->arch);
873 free(mpage->title);
874 free(mpage->desc);
875 free(mpage);
876 mpage = ohash_next(&mpages, &slot);
877 }
878 }
879
880 /*
881 * For each mlink to the mpage, check whether the path looks like
882 * it is formatted, and if it does, check whether a source manual
883 * exists by the same name, ignoring the suffix.
884 * If both conditions hold, drop the mlink.
885 */
886 static void
887 mlinks_undupe(struct mpage *mpage)
888 {
889 char buf[PATH_MAX];
890 struct mlink **prev;
891 struct mlink *mlink;
892 char *bufp;
893
894 mpage->form = FORM_CAT;
895 prev = &mpage->mlinks;
896 while (NULL != (mlink = *prev)) {
897 if (FORM_CAT != mlink->dform) {
898 mpage->form = FORM_NONE;
899 goto nextlink;
900 }
901 if (strlcpy(buf, mlink->file, PATH_MAX) >= PATH_MAX) {
902 if (warnings)
903 say(mlink->file, "Filename too long");
904 goto nextlink;
905 }
906 bufp = strstr(buf, "cat");
907 assert(NULL != bufp);
908 memcpy(bufp, "man", 3);
909 if (NULL != (bufp = strrchr(buf, '.')))
910 *++bufp = '\0';
911 strlcat(buf, mlink->dsec, PATH_MAX);
912 if (NULL == ohash_find(&mlinks,
913 ohash_qlookup(&mlinks, buf)))
914 goto nextlink;
915 if (warnings)
916 say(mlink->file, "Man source exists: %s", buf);
917 if (use_all)
918 goto nextlink;
919 *prev = mlink->next;
920 mlink_free(mlink);
921 continue;
922 nextlink:
923 prev = &(*prev)->next;
924 }
925 }
926
927 static int
928 mlink_check(struct mpage *mpage, struct mlink *mlink)
929 {
930 int match;
931
932 match = 1;
933
934 /*
935 * Check whether the manual section given in a file
936 * agrees with the directory where the file is located.
937 * Some manuals have suffixes like (3p) on their
938 * section number either inside the file or in the
939 * directory name, some are linked into more than one
940 * section, like encrypt(1) = makekey(8).
941 */
942
943 if (FORM_SRC == mpage->form &&
944 strcasecmp(mpage->sec, mlink->dsec)) {
945 match = 0;
946 say(mlink->file, "Section \"%s\" manual in %s directory",
947 mpage->sec, mlink->dsec);
948 }
949
950 /*
951 * Manual page directories exist for each kernel
952 * architecture as returned by machine(1).
953 * However, many manuals only depend on the
954 * application architecture as returned by arch(1).
955 * For example, some (2/ARM) manuals are shared
956 * across the "armish" and "zaurus" kernel
957 * architectures.
958 * A few manuals are even shared across completely
959 * different architectures, for example fdformat(1)
960 * on amd64, i386, sparc, and sparc64.
961 */
962
963 if (strcasecmp(mpage->arch, mlink->arch)) {
964 match = 0;
965 say(mlink->file, "Architecture \"%s\" manual in "
966 "\"%s\" directory", mpage->arch, mlink->arch);
967 }
968
969 if (strcasecmp(mpage->title, mlink->name))
970 match = 0;
971
972 return(match);
973 }
974
975 /*
976 * Run through the files in the global vector "mpages"
977 * and add them to the database specified in "basedir".
978 *
979 * This handles the parsing scheme itself, using the cues of directory
980 * and filename to determine whether the file is parsable or not.
981 */
982 static void
983 mpages_merge(struct mchars *mc, struct mparse *mp)
984 {
985 char any[] = "any";
986 struct ohash_info str_info;
987 int fd[2];
988 struct mpage *mpage, *mpage_dest;
989 struct mlink *mlink, *mlink_dest;
990 struct mdoc *mdoc;
991 struct man *man;
992 char *sodest;
993 char *cp;
994 pid_t child_pid;
995 int match, status;
996 unsigned int pslot;
997 enum mandoclevel lvl;
998
999 str_info.alloc = hash_alloc;
1000 str_info.halloc = hash_halloc;
1001 str_info.hfree = hash_free;
1002 str_info.key_offset = offsetof(struct str, key);
1003
1004 if (0 == nodb)
1005 SQL_EXEC("BEGIN TRANSACTION");
1006
1007 mpage = ohash_first(&mpages, &pslot);
1008 while (NULL != mpage) {
1009 mlinks_undupe(mpage);
1010 if (NULL == mpage->mlinks) {
1011 mpage = ohash_next(&mpages, &pslot);
1012 continue;
1013 }
1014
1015 ohash_init(&strings, 6, &str_info);
1016 mparse_reset(mp);
1017 mdoc = NULL;
1018 man = NULL;
1019 sodest = NULL;
1020 child_pid = 0;
1021 fd[0] = -1;
1022 fd[1] = -1;
1023
1024 if (mpage->mlinks->gzip) {
1025 if (-1 == pipe(fd)) {
1026 exitcode = (int)MANDOCLEVEL_SYSERR;
1027 say(mpage->mlinks->file, "&pipe gunzip");
1028 goto nextpage;
1029 }
1030 switch (child_pid = fork()) {
1031 case (-1):
1032 exitcode = (int)MANDOCLEVEL_SYSERR;
1033 say(mpage->mlinks->file, "&fork gunzip");
1034 child_pid = 0;
1035 close(fd[1]);
1036 close(fd[0]);
1037 goto nextpage;
1038 case (0):
1039 close(fd[0]);
1040 if (-1 == dup2(fd[1], STDOUT_FILENO)) {
1041 say(mpage->mlinks->file,
1042 "&dup gunzip");
1043 exit(1);
1044 }
1045 execlp("gunzip", "gunzip", "-c",
1046 mpage->mlinks->file, NULL);
1047 say(mpage->mlinks->file, "&exec gunzip");
1048 exit(1);
1049 default:
1050 close(fd[1]);
1051 break;
1052 }
1053 }
1054
1055 /*
1056 * Try interpreting the file as mdoc(7) or man(7)
1057 * source code, unless it is already known to be
1058 * formatted. Fall back to formatted mode.
1059 */
1060 if (FORM_CAT != mpage->mlinks->dform ||
1061 FORM_CAT != mpage->mlinks->fform) {
1062 lvl = mparse_readfd(mp, fd[0], mpage->mlinks->file);
1063 if (lvl < MANDOCLEVEL_FATAL)
1064 mparse_result(mp, &mdoc, &man, &sodest);
1065 }
1066
1067 if (NULL != sodest) {
1068 mlink_dest = ohash_find(&mlinks,
1069 ohash_qlookup(&mlinks, sodest));
1070 if (NULL != mlink_dest) {
1071
1072 /* The .so target exists. */
1073
1074 mpage_dest = mlink_dest->mpage;
1075 mlink = mpage->mlinks;
1076 while (1) {
1077 mlink->mpage = mpage_dest;
1078
1079 /*
1080 * If the target was already
1081 * processed, add the links
1082 * to the database now.
1083 * Otherwise, this will
1084 * happen when we come
1085 * to the target.
1086 */
1087
1088 if (mpage_dest->recno)
1089 dbadd_mlink(mlink);
1090
1091 if (NULL == mlink->next)
1092 break;
1093 mlink = mlink->next;
1094 }
1095
1096 /* Move all links to the target. */
1097
1098 mlink->next = mlink_dest->next;
1099 mlink_dest->next = mpage->mlinks;
1100 mpage->mlinks = NULL;
1101 }
1102 goto nextpage;
1103 } else if (NULL != mdoc) {
1104 mpage->form = FORM_SRC;
1105 mpage->sec =
1106 mandoc_strdup(mdoc_meta(mdoc)->msec);
1107 mpage->arch = mdoc_meta(mdoc)->arch;
1108 mpage->arch = mandoc_strdup(
1109 NULL == mpage->arch ? "" : mpage->arch);
1110 mpage->title =
1111 mandoc_strdup(mdoc_meta(mdoc)->title);
1112 } else if (NULL != man) {
1113 mpage->form = FORM_SRC;
1114 mpage->sec =
1115 mandoc_strdup(man_meta(man)->msec);
1116 mpage->arch =
1117 mandoc_strdup(mpage->mlinks->arch);
1118 mpage->title =
1119 mandoc_strdup(man_meta(man)->title);
1120 } else {
1121 mpage->form = FORM_CAT;
1122 mpage->sec =
1123 mandoc_strdup(mpage->mlinks->dsec);
1124 mpage->arch =
1125 mandoc_strdup(mpage->mlinks->arch);
1126 mpage->title =
1127 mandoc_strdup(mpage->mlinks->name);
1128 }
1129 putkey(mpage, mpage->sec, TYPE_sec);
1130 putkey(mpage, '\0' == *mpage->arch ?
1131 any : mpage->arch, TYPE_arch);
1132
1133 for (mlink = mpage->mlinks; mlink; mlink = mlink->next) {
1134 if ('\0' != *mlink->dsec)
1135 putkey(mpage, mlink->dsec, TYPE_sec);
1136 if ('\0' != *mlink->fsec)
1137 putkey(mpage, mlink->fsec, TYPE_sec);
1138 putkey(mpage, '\0' == *mlink->arch ?
1139 any : mlink->arch, TYPE_arch);
1140 putkey(mpage, mlink->name, TYPE_Nm);
1141 }
1142
1143 if (warnings && !use_all) {
1144 match = 0;
1145 for (mlink = mpage->mlinks; mlink;
1146 mlink = mlink->next)
1147 if (mlink_check(mpage, mlink))
1148 match = 1;
1149 } else
1150 match = 1;
1151
1152 if (NULL != mdoc) {
1153 if (NULL != (cp = mdoc_meta(mdoc)->name))
1154 putkey(mpage, cp, TYPE_Nm);
1155 assert(NULL == mpage->desc);
1156 parse_mdoc(mpage, mdoc_node(mdoc));
1157 putkey(mpage, NULL != mpage->desc ?
1158 mpage->desc : mpage->mlinks->name, TYPE_Nd);
1159 } else if (NULL != man)
1160 parse_man(mpage, man_node(man));
1161 else
1162 parse_cat(mpage, fd[0]);
1163
1164 dbadd(mpage, mc);
1165
1166 nextpage:
1167 if (child_pid) {
1168 if (-1 == waitpid(child_pid, &status, 0)) {
1169 exitcode = (int)MANDOCLEVEL_SYSERR;
1170 say(mpage->mlinks->file, "&wait gunzip");
1171 } else if (WIFSIGNALED(status)) {
1172 exitcode = (int)MANDOCLEVEL_SYSERR;
1173 say(mpage->mlinks->file,
1174 "gunzip died from signal %d",
1175 WTERMSIG(status));
1176 } else if (WEXITSTATUS(status)) {
1177 exitcode = (int)MANDOCLEVEL_SYSERR;
1178 say(mpage->mlinks->file,
1179 "gunzip failed with code %d",
1180 WEXITSTATUS(status));
1181 }
1182 }
1183 ohash_delete(&strings);
1184 mpage = ohash_next(&mpages, &pslot);
1185 }
1186
1187 if (0 == nodb)
1188 SQL_EXEC("END TRANSACTION");
1189 }
1190
1191 static void
1192 names_check(void)
1193 {
1194 sqlite3_stmt *stmt;
1195 const char *name, *sec, *arch, *key;
1196 size_t i;
1197 int irc;
1198
1199 sqlite3_prepare_v2(db,
1200 "SELECT name, sec, arch, key FROM ("
1201 "SELECT key, pageid FROM keys "
1202 "WHERE bits & ? AND NOT EXISTS ("
1203 "SELECT pageid FROM mlinks "
1204 "WHERE mlinks.pageid == keys.pageid "
1205 "AND mlinks.name == keys.key"
1206 ")"
1207 ") JOIN ("
1208 "SELECT * FROM mlinks GROUP BY pageid"
1209 ") USING (pageid);",
1210 -1, &stmt, NULL);
1211
1212 i = 1;
1213 SQL_BIND_INT64(stmt, i, TYPE_NAME);
1214
1215 while (SQLITE_ROW == (irc = sqlite3_step(stmt))) {
1216 name = sqlite3_column_text(stmt, 0);
1217 sec = sqlite3_column_text(stmt, 1);
1218 arch = sqlite3_column_text(stmt, 2);
1219 key = sqlite3_column_text(stmt, 3);
1220 say("", "%s(%s%s%s) lacks mlink \"%s\"", name, sec,
1221 '\0' == *arch ? "" : "/",
1222 '\0' == *arch ? "" : arch, key);
1223 }
1224 sqlite3_finalize(stmt);
1225 }
1226
1227 static void
1228 parse_cat(struct mpage *mpage, int fd)
1229 {
1230 FILE *stream;
1231 char *line, *p, *title;
1232 size_t len, plen, titlesz;
1233
1234 stream = (-1 == fd) ?
1235 fopen(mpage->mlinks->file, "r") :
1236 fdopen(fd, "r");
1237 if (NULL == stream) {
1238 if (warnings)
1239 say(mpage->mlinks->file, "&fopen");
1240 return;
1241 }
1242
1243 /* Skip to first blank line. */
1244
1245 while (NULL != (line = fgetln(stream, &len)))
1246 if ('\n' == *line)
1247 break;
1248
1249 /*
1250 * Assume the first line that is not indented
1251 * is the first section header. Skip to it.
1252 */
1253
1254 while (NULL != (line = fgetln(stream, &len)))
1255 if ('\n' != *line && ' ' != *line)
1256 break;
1257
1258 /*
1259 * Read up until the next section into a buffer.
1260 * Strip the leading and trailing newline from each read line,
1261 * appending a trailing space.
1262 * Ignore empty (whitespace-only) lines.
1263 */
1264
1265 titlesz = 0;
1266 title = NULL;
1267
1268 while (NULL != (line = fgetln(stream, &len))) {
1269 if (' ' != *line || '\n' != line[len - 1])
1270 break;
1271 while (len > 0 && isspace((unsigned char)*line)) {
1272 line++;
1273 len--;
1274 }
1275 if (1 == len)
1276 continue;
1277 title = mandoc_realloc(title, titlesz + len);
1278 memcpy(title + titlesz, line, len);
1279 titlesz += len;
1280 title[titlesz - 1] = ' ';
1281 }
1282
1283 /*
1284 * If no page content can be found, or the input line
1285 * is already the next section header, or there is no
1286 * trailing newline, reuse the page title as the page
1287 * description.
1288 */
1289
1290 if (NULL == title || '\0' == *title) {
1291 if (warnings)
1292 say(mpage->mlinks->file,
1293 "Cannot find NAME section");
1294 assert(NULL == mpage->desc);
1295 mpage->desc = mandoc_strdup(mpage->mlinks->name);
1296 putkey(mpage, mpage->mlinks->name, TYPE_Nd);
1297 fclose(stream);
1298 free(title);
1299 return;
1300 }
1301
1302 title = mandoc_realloc(title, titlesz + 1);
1303 title[titlesz] = '\0';
1304
1305 /*
1306 * Skip to the first dash.
1307 * Use the remaining line as the description (no more than 70
1308 * bytes).
1309 */
1310
1311 if (NULL != (p = strstr(title, "- "))) {
1312 for (p += 2; ' ' == *p || '\b' == *p; p++)
1313 /* Skip to next word. */ ;
1314 } else {
1315 if (warnings)
1316 say(mpage->mlinks->file,
1317 "No dash in title line");
1318 p = title;
1319 }
1320
1321 plen = strlen(p);
1322
1323 /* Strip backspace-encoding from line. */
1324
1325 while (NULL != (line = memchr(p, '\b', plen))) {
1326 len = line - p;
1327 if (0 == len) {
1328 memmove(line, line + 1, plen--);
1329 continue;
1330 }
1331 memmove(line - 1, line + 1, plen - len);
1332 plen -= 2;
1333 }
1334
1335 assert(NULL == mpage->desc);
1336 mpage->desc = mandoc_strdup(p);
1337 putkey(mpage, mpage->desc, TYPE_Nd);
1338 fclose(stream);
1339 free(title);
1340 }
1341
1342 /*
1343 * Put a type/word pair into the word database for this particular file.
1344 */
1345 static void
1346 putkey(const struct mpage *mpage, char *value, uint64_t type)
1347 {
1348 char *cp;
1349
1350 assert(NULL != value);
1351 if (TYPE_arch == type)
1352 for (cp = value; *cp; cp++)
1353 if (isupper((unsigned char)*cp))
1354 *cp = _tolower((unsigned char)*cp);
1355 putkeys(mpage, value, strlen(value), type);
1356 }
1357
1358 /*
1359 * Grok all nodes at or below a certain mdoc node into putkey().
1360 */
1361 static void
1362 putmdockey(const struct mpage *mpage,
1363 const struct mdoc_node *n, uint64_t m)
1364 {
1365
1366 for ( ; NULL != n; n = n->next) {
1367 if (NULL != n->child)
1368 putmdockey(mpage, n->child, m);
1369 if (MDOC_TEXT == n->type)
1370 putkey(mpage, n->string, m);
1371 }
1372 }
1373
1374 static void
1375 parse_man(struct mpage *mpage, const struct man_node *n)
1376 {
1377 const struct man_node *head, *body;
1378 char *start, *title;
1379 char byte;
1380 size_t sz;
1381
1382 if (NULL == n)
1383 return;
1384
1385 /*
1386 * We're only searching for one thing: the first text child in
1387 * the BODY of a NAME section. Since we don't keep track of
1388 * sections in -man, run some hoops to find out whether we're in
1389 * the correct section or not.
1390 */
1391
1392 if (MAN_BODY == n->type && MAN_SH == n->tok) {
1393 body = n;
1394 assert(body->parent);
1395 if (NULL != (head = body->parent->head) &&
1396 1 == head->nchild &&
1397 NULL != (head = (head->child)) &&
1398 MAN_TEXT == head->type &&
1399 0 == strcmp(head->string, "NAME") &&
1400 NULL != body->child) {
1401
1402 /*
1403 * Suck the entire NAME section into memory.
1404 * Yes, we might run away.
1405 * But too many manuals have big, spread-out
1406 * NAME sections over many lines.
1407 */
1408
1409 title = NULL;
1410 man_deroff(&title, body);
1411 if (NULL == title)
1412 return;
1413
1414 /*
1415 * Go through a special heuristic dance here.
1416 * Conventionally, one or more manual names are
1417 * comma-specified prior to a whitespace, then a
1418 * dash, then a description. Try to puzzle out
1419 * the name parts here.
1420 */
1421
1422 start = title;
1423 for ( ;; ) {
1424 sz = strcspn(start, " ,");
1425 if ('\0' == start[sz])
1426 break;
1427
1428 byte = start[sz];
1429 start[sz] = '\0';
1430
1431 /*
1432 * Assume a stray trailing comma in the
1433 * name list if a name begins with a dash.
1434 */
1435
1436 if ('-' == start[0] ||
1437 ('\\' == start[0] && '-' == start[1]))
1438 break;
1439
1440 putkey(mpage, start, TYPE_NAME | TYPE_Nm);
1441
1442 if (' ' == byte) {
1443 start += sz + 1;
1444 break;
1445 }
1446
1447 assert(',' == byte);
1448 start += sz + 1;
1449 while (' ' == *start)
1450 start++;
1451 }
1452
1453 if (start == title) {
1454 putkey(mpage, start, TYPE_NAME | TYPE_Nm);
1455 free(title);
1456 return;
1457 }
1458
1459 while (isspace((unsigned char)*start))
1460 start++;
1461
1462 if (0 == strncmp(start, "-", 1))
1463 start += 1;
1464 else if (0 == strncmp(start, "\\-\\-", 4))
1465 start += 4;
1466 else if (0 == strncmp(start, "\\-", 2))
1467 start += 2;
1468 else if (0 == strncmp(start, "\\(en", 4))
1469 start += 4;
1470 else if (0 == strncmp(start, "\\(em", 4))
1471 start += 4;
1472
1473 while (' ' == *start)
1474 start++;
1475
1476 assert(NULL == mpage->desc);
1477 mpage->desc = mandoc_strdup(start);
1478 putkey(mpage, mpage->desc, TYPE_Nd);
1479 free(title);
1480 return;
1481 }
1482 }
1483
1484 for (n = n->child; n; n = n->next) {
1485 if (NULL != mpage->desc)
1486 break;
1487 parse_man(mpage, n);
1488 }
1489 }
1490
1491 static void
1492 parse_mdoc(struct mpage *mpage, const struct mdoc_node *n)
1493 {
1494
1495 assert(NULL != n);
1496 for (n = n->child; NULL != n; n = n->next) {
1497 switch (n->type) {
1498 case (MDOC_ELEM):
1499 /* FALLTHROUGH */
1500 case (MDOC_BLOCK):
1501 /* FALLTHROUGH */
1502 case (MDOC_HEAD):
1503 /* FALLTHROUGH */
1504 case (MDOC_BODY):
1505 /* FALLTHROUGH */
1506 case (MDOC_TAIL):
1507 if (NULL != mdocs[n->tok].fp)
1508 if (0 == (*mdocs[n->tok].fp)(mpage, n))
1509 break;
1510 if (mdocs[n->tok].mask)
1511 putmdockey(mpage, n->child,
1512 mdocs[n->tok].mask);
1513 break;
1514 default:
1515 assert(MDOC_ROOT != n->type);
1516 continue;
1517 }
1518 if (NULL != n->child)
1519 parse_mdoc(mpage, n);
1520 }
1521 }
1522
1523 static int
1524 parse_mdoc_Fd(struct mpage *mpage, const struct mdoc_node *n)
1525 {
1526 const char *start, *end;
1527 size_t sz;
1528
1529 if (SEC_SYNOPSIS != n->sec ||
1530 NULL == (n = n->child) ||
1531 MDOC_TEXT != n->type)
1532 return(0);
1533
1534 /*
1535 * Only consider those `Fd' macro fields that begin with an
1536 * "inclusion" token (versus, e.g., #define).
1537 */
1538
1539 if (strcmp("#include", n->string))
1540 return(0);
1541
1542 if (NULL == (n = n->next) || MDOC_TEXT != n->type)
1543 return(0);
1544
1545 /*
1546 * Strip away the enclosing angle brackets and make sure we're
1547 * not zero-length.
1548 */
1549
1550 start = n->string;
1551 if ('<' == *start || '"' == *start)
1552 start++;
1553
1554 if (0 == (sz = strlen(start)))
1555 return(0);
1556
1557 end = &start[(int)sz - 1];
1558 if ('>' == *end || '"' == *end)
1559 end--;
1560
1561 if (end > start)
1562 putkeys(mpage, start, end - start + 1, TYPE_In);
1563 return(0);
1564 }
1565
1566 static int
1567 parse_mdoc_Fn(struct mpage *mpage, const struct mdoc_node *n)
1568 {
1569 char *cp;
1570
1571 if (NULL == (n = n->child) || MDOC_TEXT != n->type)
1572 return(0);
1573
1574 /*
1575 * Parse: .Fn "struct type *name" "char *arg".
1576 * First strip away pointer symbol.
1577 * Then store the function name, then type.
1578 * Finally, store the arguments.
1579 */
1580
1581 if (NULL == (cp = strrchr(n->string, ' ')))
1582 cp = n->string;
1583
1584 while ('*' == *cp)
1585 cp++;
1586
1587 putkey(mpage, cp, TYPE_Fn);
1588
1589 if (n->string < cp)
1590 putkeys(mpage, n->string, cp - n->string, TYPE_Ft);
1591
1592 for (n = n->next; NULL != n; n = n->next)
1593 if (MDOC_TEXT == n->type)
1594 putkey(mpage, n->string, TYPE_Fa);
1595
1596 return(0);
1597 }
1598
1599 static int
1600 parse_mdoc_Xr(struct mpage *mpage, const struct mdoc_node *n)
1601 {
1602 char *cp;
1603
1604 if (NULL == (n = n->child))
1605 return(0);
1606
1607 if (NULL == n->next) {
1608 putkey(mpage, n->string, TYPE_Xr);
1609 return(0);
1610 }
1611
1612 mandoc_asprintf(&cp, "%s(%s)", n->string, n->next->string);
1613 putkey(mpage, cp, TYPE_Xr);
1614 free(cp);
1615 return(0);
1616 }
1617
1618 static int
1619 parse_mdoc_Nd(struct mpage *mpage, const struct mdoc_node *n)
1620 {
1621
1622 if (MDOC_BODY == n->type)
1623 mdoc_deroff(&mpage->desc, n);
1624 return(0);
1625 }
1626
1627 static int
1628 parse_mdoc_Nm(struct mpage *mpage, const struct mdoc_node *n)
1629 {
1630
1631 if (SEC_NAME == n->sec)
1632 putmdockey(mpage, n->child, TYPE_NAME | TYPE_Nm);
1633 else if (SEC_SYNOPSIS == n->sec && MDOC_HEAD == n->type)
1634 putmdockey(mpage, n->child, TYPE_Nm);
1635 return(0);
1636 }
1637
1638 static int
1639 parse_mdoc_Sh(struct mpage *mpage, const struct mdoc_node *n)
1640 {
1641
1642 return(SEC_CUSTOM == n->sec && MDOC_HEAD == n->type);
1643 }
1644
1645 static int
1646 parse_mdoc_head(struct mpage *mpage, const struct mdoc_node *n)
1647 {
1648
1649 return(MDOC_HEAD == n->type);
1650 }
1651
1652 static int
1653 parse_mdoc_body(struct mpage *mpage, const struct mdoc_node *n)
1654 {
1655
1656 return(MDOC_BODY == n->type);
1657 }
1658
1659 /*
1660 * Add a string to the hash table for the current manual.
1661 * Each string has a bitmask telling which macros it belongs to.
1662 * When we finish the manual, we'll dump the table.
1663 */
1664 static void
1665 putkeys(const struct mpage *mpage,
1666 const char *cp, size_t sz, uint64_t v)
1667 {
1668 struct str *s;
1669 const char *end;
1670 uint64_t mask;
1671 unsigned int slot;
1672 int i;
1673
1674 if (0 == sz)
1675 return;
1676
1677 if (debug > 1) {
1678 for (i = 0, mask = 1;
1679 i < mansearch_keymax;
1680 i++, mask <<= 1)
1681 if (mask & v)
1682 break;
1683 say(mpage->mlinks->file, "Adding key %s=%*s",
1684 mansearch_keynames[i], sz, cp);
1685 }
1686
1687 end = cp + sz;
1688 slot = ohash_qlookupi(&strings, cp, &end);
1689 s = ohash_find(&strings, slot);
1690
1691 if (NULL != s && mpage == s->mpage) {
1692 s->mask |= v;
1693 return;
1694 } else if (NULL == s) {
1695 s = mandoc_calloc(sizeof(struct str) + sz + 1, 1);
1696 memcpy(s->key, cp, sz);
1697 ohash_insert(&strings, slot, s);
1698 }
1699 s->mpage = mpage;
1700 s->mask = v;
1701 }
1702
1703 /*
1704 * Take a Unicode codepoint and produce its UTF-8 encoding.
1705 * This isn't the best way to do this, but it works.
1706 * The magic numbers are from the UTF-8 packaging.
1707 * They're not as scary as they seem: read the UTF-8 spec for details.
1708 */
1709 static size_t
1710 utf8(unsigned int cp, char out[7])
1711 {
1712 size_t rc;
1713
1714 rc = 0;
1715 if (cp <= 0x0000007F) {
1716 rc = 1;
1717 out[0] = (char)cp;
1718 } else if (cp <= 0x000007FF) {
1719 rc = 2;
1720 out[0] = (cp >> 6 & 31) | 192;
1721 out[1] = (cp & 63) | 128;
1722 } else if (cp <= 0x0000FFFF) {
1723 rc = 3;
1724 out[0] = (cp >> 12 & 15) | 224;
1725 out[1] = (cp >> 6 & 63) | 128;
1726 out[2] = (cp & 63) | 128;
1727 } else if (cp <= 0x001FFFFF) {
1728 rc = 4;
1729 out[0] = (cp >> 18 & 7) | 240;
1730 out[1] = (cp >> 12 & 63) | 128;
1731 out[2] = (cp >> 6 & 63) | 128;
1732 out[3] = (cp & 63) | 128;
1733 } else if (cp <= 0x03FFFFFF) {
1734 rc = 5;
1735 out[0] = (cp >> 24 & 3) | 248;
1736 out[1] = (cp >> 18 & 63) | 128;
1737 out[2] = (cp >> 12 & 63) | 128;
1738 out[3] = (cp >> 6 & 63) | 128;
1739 out[4] = (cp & 63) | 128;
1740 } else if (cp <= 0x7FFFFFFF) {
1741 rc = 6;
1742 out[0] = (cp >> 30 & 1) | 252;
1743 out[1] = (cp >> 24 & 63) | 128;
1744 out[2] = (cp >> 18 & 63) | 128;
1745 out[3] = (cp >> 12 & 63) | 128;
1746 out[4] = (cp >> 6 & 63) | 128;
1747 out[5] = (cp & 63) | 128;
1748 } else
1749 return(0);
1750
1751 out[rc] = '\0';
1752 return(rc);
1753 }
1754
1755 /*
1756 * Store the rendered version of a key, or alias the pointer
1757 * if the key contains no escape sequences.
1758 */
1759 static void
1760 render_key(struct mchars *mc, struct str *key)
1761 {
1762 size_t sz, bsz, pos;
1763 char utfbuf[7], res[6];
1764 char *buf;
1765 const char *seq, *cpp, *val;
1766 int len, u;
1767 enum mandoc_esc esc;
1768
1769 assert(NULL == key->rendered);
1770
1771 res[0] = '\\';
1772 res[1] = '\t';
1773 res[2] = ASCII_NBRSP;
1774 res[3] = ASCII_HYPH;
1775 res[4] = ASCII_BREAK;
1776 res[5] = '\0';
1777
1778 val = key->key;
1779 bsz = strlen(val);
1780
1781 /*
1782 * Pre-check: if we have no stop-characters, then set the
1783 * pointer as ourselvse and get out of here.
1784 */
1785 if (strcspn(val, res) == bsz) {
1786 key->rendered = key->key;
1787 return;
1788 }
1789
1790 /* Pre-allocate by the length of the input */
1791
1792 buf = mandoc_malloc(++bsz);
1793 pos = 0;
1794
1795 while ('\0' != *val) {
1796 /*
1797 * Halt on the first escape sequence.
1798 * This also halts on the end of string, in which case
1799 * we just copy, fallthrough, and exit the loop.
1800 */
1801 if ((sz = strcspn(val, res)) > 0) {
1802 memcpy(&buf[pos], val, sz);
1803 pos += sz;
1804 val += sz;
1805 }
1806
1807 switch (*val) {
1808 case (ASCII_HYPH):
1809 buf[pos++] = '-';
1810 val++;
1811 continue;
1812 case ('\t'):
1813 /* FALLTHROUGH */
1814 case (ASCII_NBRSP):
1815 buf[pos++] = ' ';
1816 val++;
1817 /* FALLTHROUGH */
1818 case (ASCII_BREAK):
1819 continue;
1820 default:
1821 break;
1822 }
1823 if ('\\' != *val)
1824 break;
1825
1826 /* Read past the slash. */
1827
1828 val++;
1829
1830 /*
1831 * Parse the escape sequence and see if it's a
1832 * predefined character or special character.
1833 */
1834
1835 esc = mandoc_escape
1836 ((const char **)&val, &seq, &len);
1837 if (ESCAPE_ERROR == esc)
1838 break;
1839 if (ESCAPE_SPECIAL != esc)
1840 continue;
1841
1842 /*
1843 * Render the special character
1844 * as either UTF-8 or ASCII.
1845 */
1846
1847 if (write_utf8) {
1848 if (0 == (u = mchars_spec2cp(mc, seq, len)))
1849 continue;
1850 cpp = utfbuf;
1851 if (0 == (sz = utf8(u, utfbuf)))
1852 continue;
1853 sz = strlen(cpp);
1854 } else {
1855 cpp = mchars_spec2str(mc, seq, len, &sz);
1856 if (NULL == cpp)
1857 continue;
1858 if (ASCII_NBRSP == *cpp) {
1859 cpp = " ";
1860 sz = 1;
1861 }
1862 }
1863
1864 /* Copy the rendered glyph into the stream. */
1865
1866 bsz += sz;
1867 buf = mandoc_realloc(buf, bsz);
1868 memcpy(&buf[pos], cpp, sz);
1869 pos += sz;
1870 }
1871
1872 buf[pos] = '\0';
1873 key->rendered = buf;
1874 }
1875
1876 static void
1877 dbadd_mlink(const struct mlink *mlink)
1878 {
1879 size_t i;
1880
1881 i = 1;
1882 SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->dsec);
1883 SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->arch);
1884 SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->name);
1885 SQL_BIND_INT64(stmts[STMT_INSERT_LINK], i, mlink->mpage->recno);
1886 SQL_STEP(stmts[STMT_INSERT_LINK]);
1887 sqlite3_reset(stmts[STMT_INSERT_LINK]);
1888 }
1889
1890 /*
1891 * Flush the current page's terms (and their bits) into the database.
1892 * Wrap the entire set of additions in a transaction to make sqlite be a
1893 * little faster.
1894 * Also, handle escape sequences at the last possible moment.
1895 */
1896 static void
1897 dbadd(struct mpage *mpage, struct mchars *mc)
1898 {
1899 struct mlink *mlink;
1900 struct str *key;
1901 size_t i;
1902 unsigned int slot;
1903
1904 mlink = mpage->mlinks;
1905
1906 if (nodb) {
1907 while (NULL != mlink) {
1908 fputs(mlink->name, stdout);
1909 if (NULL == mlink->next ||
1910 strcmp(mlink->dsec, mlink->next->dsec) ||
1911 strcmp(mlink->fsec, mlink->next->fsec) ||
1912 strcmp(mlink->arch, mlink->next->arch)) {
1913 putchar('(');
1914 if ('\0' == *mlink->dsec)
1915 fputs(mlink->fsec, stdout);
1916 else
1917 fputs(mlink->dsec, stdout);
1918 if ('\0' != *mlink->arch)
1919 printf("/%s", mlink->arch);
1920 putchar(')');
1921 }
1922 mlink = mlink->next;
1923 if (NULL != mlink)
1924 fputs(", ", stdout);
1925 }
1926 for (key = ohash_first(&strings, &slot); NULL != key;
1927 key = ohash_next(&strings, &slot)) {
1928 if (TYPE_Nd & key->mask) {
1929 if (NULL == key->rendered)
1930 render_key(mc, key);
1931 printf(" - %s", key->rendered);
1932 break;
1933 }
1934 }
1935 putchar('\n');
1936 return;
1937 }
1938
1939 if (debug)
1940 say(mlink->file, "Adding to database");
1941
1942 i = 1;
1943 SQL_BIND_INT(stmts[STMT_INSERT_PAGE], i, FORM_SRC == mpage->form);
1944 SQL_STEP(stmts[STMT_INSERT_PAGE]);
1945 mpage->recno = sqlite3_last_insert_rowid(db);
1946 sqlite3_reset(stmts[STMT_INSERT_PAGE]);
1947
1948 while (NULL != mlink) {
1949 dbadd_mlink(mlink);
1950 mlink = mlink->next;
1951 }
1952
1953 for (key = ohash_first(&strings, &slot); NULL != key;
1954 key = ohash_next(&strings, &slot)) {
1955 assert(key->mpage == mpage);
1956 if (NULL == key->rendered)
1957 render_key(mc, key);
1958 i = 1;
1959 SQL_BIND_INT64(stmts[STMT_INSERT_KEY], i, key->mask);
1960 SQL_BIND_TEXT(stmts[STMT_INSERT_KEY], i, key->rendered);
1961 SQL_BIND_INT64(stmts[STMT_INSERT_KEY], i, mpage->recno);
1962 SQL_STEP(stmts[STMT_INSERT_KEY]);
1963 sqlite3_reset(stmts[STMT_INSERT_KEY]);
1964 if (key->rendered != key->key)
1965 free(key->rendered);
1966 free(key);
1967 }
1968 }
1969
1970 static void
1971 dbprune(void)
1972 {
1973 struct mpage *mpage;
1974 struct mlink *mlink;
1975 size_t i;
1976 unsigned int slot;
1977
1978 if (0 == nodb)
1979 SQL_EXEC("BEGIN TRANSACTION");
1980
1981 for (mpage = ohash_first(&mpages, &slot); NULL != mpage;
1982 mpage = ohash_next(&mpages, &slot)) {
1983 mlink = mpage->mlinks;
1984 if (debug)
1985 say(mlink->file, "Deleting from database");
1986 if (nodb)
1987 continue;
1988 for ( ; NULL != mlink; mlink = mlink->next) {
1989 i = 1;
1990 SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE],
1991 i, mlink->dsec);
1992 SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE],
1993 i, mlink->arch);
1994 SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE],
1995 i, mlink->name);
1996 SQL_STEP(stmts[STMT_DELETE_PAGE]);
1997 sqlite3_reset(stmts[STMT_DELETE_PAGE]);
1998 }
1999 }
2000
2001 if (0 == nodb)
2002 SQL_EXEC("END TRANSACTION");
2003 }
2004
2005 /*
2006 * Close an existing database and its prepared statements.
2007 * If "real" is not set, rename the temporary file into the real one.
2008 */
2009 static void
2010 dbclose(int real)
2011 {
2012 size_t i;
2013 int status;
2014 pid_t child;
2015
2016 if (nodb)
2017 return;
2018
2019 for (i = 0; i < STMT__MAX; i++) {
2020 sqlite3_finalize(stmts[i]);
2021 stmts[i] = NULL;
2022 }
2023
2024 sqlite3_close(db);
2025 db = NULL;
2026
2027 if (real)
2028 return;
2029
2030 if ('\0' == *tempfilename) {
2031 if (-1 == rename(MANDOC_DB "~", MANDOC_DB)) {
2032 exitcode = (int)MANDOCLEVEL_SYSERR;
2033 say(MANDOC_DB, "&rename");
2034 }
2035 return;
2036 }
2037
2038 switch (child = fork()) {
2039 case (-1):
2040 exitcode = (int)MANDOCLEVEL_SYSERR;
2041 say("", "&fork cmp");
2042 return;
2043 case (0):
2044 execlp("cmp", "cmp", "-s",
2045 tempfilename, MANDOC_DB, NULL);
2046 say("", "&exec cmp");
2047 exit(0);
2048 default:
2049 break;
2050 }
2051 if (-1 == waitpid(child, &status, 0)) {
2052 exitcode = (int)MANDOCLEVEL_SYSERR;
2053 say("", "&wait cmp");
2054 } else if (WIFSIGNALED(status)) {
2055 exitcode = (int)MANDOCLEVEL_SYSERR;
2056 say("", "cmp died from signal %d", WTERMSIG(status));
2057 } else if (WEXITSTATUS(status)) {
2058 exitcode = (int)MANDOCLEVEL_SYSERR;
2059 say(MANDOC_DB,
2060 "Data changed, but cannot replace database");
2061 }
2062
2063 *strrchr(tempfilename, '/') = '\0';
2064 switch (child = fork()) {
2065 case (-1):
2066 exitcode = (int)MANDOCLEVEL_SYSERR;
2067 say("", "&fork rm");
2068 return;
2069 case (0):
2070 execlp("rm", "rm", "-rf", tempfilename, NULL);
2071 say("", "&exec rm");
2072 exit((int)MANDOCLEVEL_SYSERR);
2073 default:
2074 break;
2075 }
2076 if (-1 == waitpid(child, &status, 0)) {
2077 exitcode = (int)MANDOCLEVEL_SYSERR;
2078 say("", "&wait rm");
2079 } else if (WIFSIGNALED(status) || WEXITSTATUS(status)) {
2080 exitcode = (int)MANDOCLEVEL_SYSERR;
2081 say("", "%s: Cannot remove temporary directory",
2082 tempfilename);
2083 }
2084 }
2085
2086 /*
2087 * This is straightforward stuff.
2088 * Open a database connection to a "temporary" database, then open a set
2089 * of prepared statements we'll use over and over again.
2090 * If "real" is set, we use the existing database; if not, we truncate a
2091 * temporary one.
2092 * Must be matched by dbclose().
2093 */
2094 static int
2095 dbopen(int real)
2096 {
2097 const char *sql;
2098 int rc, ofl;
2099
2100 if (nodb)
2101 return(1);
2102
2103 *tempfilename = '\0';
2104 ofl = SQLITE_OPEN_READWRITE;
2105
2106 if (real) {
2107 rc = sqlite3_open_v2(MANDOC_DB, &db, ofl, NULL);
2108 if (SQLITE_OK != rc) {
2109 exitcode = (int)MANDOCLEVEL_SYSERR;
2110 say(MANDOC_DB, "%s", sqlite3_errmsg(db));
2111 return(0);
2112 }
2113 goto prepare_statements;
2114 }
2115
2116 ofl |= SQLITE_OPEN_CREATE | SQLITE_OPEN_EXCLUSIVE;
2117
2118 remove(MANDOC_DB "~");
2119 rc = sqlite3_open_v2(MANDOC_DB "~", &db, ofl, NULL);
2120 if (SQLITE_OK == rc)
2121 goto create_tables;
2122 if (MPARSE_QUICK & mparse_options) {
2123 exitcode = (int)MANDOCLEVEL_SYSERR;
2124 say(MANDOC_DB "~", "%s", sqlite3_errmsg(db));
2125 return(0);
2126 }
2127
2128 if (strlcpy(tempfilename, "/tmp/mandocdb.XXXXXX",
2129 sizeof(tempfilename)) >= sizeof(tempfilename)) {
2130 exitcode = (int)MANDOCLEVEL_SYSERR;
2131 say("", "/tmp/mandocdb.XXXXXX: Filename too long");
2132 return(0);
2133 }
2134 if (NULL == mkdtemp(tempfilename)) {
2135 exitcode = (int)MANDOCLEVEL_SYSERR;
2136 say("", "&%s", tempfilename);
2137 return(0);
2138 }
2139 if (strlcat(tempfilename, "/" MANDOC_DB,
2140 sizeof(tempfilename)) >= sizeof(tempfilename)) {
2141 exitcode = (int)MANDOCLEVEL_SYSERR;
2142 say("", "%s/" MANDOC_DB ": Filename too long",
2143 tempfilename);
2144 return(0);
2145 }
2146 rc = sqlite3_open_v2(tempfilename, &db, ofl, NULL);
2147 if (SQLITE_OK != rc) {
2148 exitcode = (int)MANDOCLEVEL_SYSERR;
2149 say("", "%s: %s", tempfilename, sqlite3_errmsg(db));
2150 return(0);
2151 }
2152
2153 create_tables:
2154 sql = "CREATE TABLE \"mpages\" (\n"
2155 " \"form\" INTEGER NOT NULL,\n"
2156 " \"id\" INTEGER PRIMARY KEY AUTOINCREMENT NOT NULL\n"
2157 ");\n"
2158 "\n"
2159 "CREATE TABLE \"mlinks\" (\n"
2160 " \"sec\" TEXT NOT NULL,\n"
2161 " \"arch\" TEXT NOT NULL,\n"
2162 " \"name\" TEXT NOT NULL,\n"
2163 " \"pageid\" INTEGER NOT NULL REFERENCES mpages(id) "
2164 "ON DELETE CASCADE\n"
2165 ");\n"
2166 "\n"
2167 "CREATE TABLE \"keys\" (\n"
2168 " \"bits\" INTEGER NOT NULL,\n"
2169 " \"key\" TEXT NOT NULL,\n"
2170 " \"pageid\" INTEGER NOT NULL REFERENCES mpages(id) "
2171 "ON DELETE CASCADE\n"
2172 ");\n";
2173
2174 if (SQLITE_OK != sqlite3_exec(db, sql, NULL, NULL, NULL)) {
2175 exitcode = (int)MANDOCLEVEL_SYSERR;
2176 say(MANDOC_DB, "%s", sqlite3_errmsg(db));
2177 return(0);
2178 }
2179
2180 prepare_statements:
2181 SQL_EXEC("PRAGMA foreign_keys = ON");
2182 sql = "DELETE FROM mpages WHERE id IN "
2183 "(SELECT pageid FROM mlinks WHERE "
2184 "sec=? AND arch=? AND name=?)";
2185 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_DELETE_PAGE], NULL);
2186 sql = "INSERT INTO mpages "
2187 "(form) VALUES (?)";
2188 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_PAGE], NULL);
2189 sql = "INSERT INTO mlinks "
2190 "(sec,arch,name,pageid) VALUES (?,?,?,?)";
2191 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_LINK], NULL);
2192 sql = "INSERT INTO keys "
2193 "(bits,key,pageid) VALUES (?,?,?)";
2194 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_KEY], NULL);
2195
2196 #ifndef __APPLE__
2197 /*
2198 * When opening a new database, we can turn off
2199 * synchronous mode for much better performance.
2200 */
2201
2202 if (real)
2203 SQL_EXEC("PRAGMA synchronous = OFF");
2204 #endif
2205
2206 return(1);
2207 }
2208
2209 static void *
2210 hash_halloc(size_t sz, void *arg)
2211 {
2212
2213 return(mandoc_calloc(sz, 1));
2214 }
2215
2216 static void *
2217 hash_alloc(size_t sz, void *arg)
2218 {
2219
2220 return(mandoc_malloc(sz));
2221 }
2222
2223 static void
2224 hash_free(void *p, size_t sz, void *arg)
2225 {
2226
2227 free(p);
2228 }
2229
2230 static int
2231 set_basedir(const char *targetdir)
2232 {
2233 static char startdir[PATH_MAX];
2234 static int fd;
2235
2236 /*
2237 * Remember where we started by keeping a fd open to the origin
2238 * path component: throughout this utility, we chdir() a lot to
2239 * handle relative paths, and by doing this, we can return to
2240 * the starting point.
2241 */
2242 if ('\0' == *startdir) {
2243 if (NULL == getcwd(startdir, PATH_MAX)) {
2244 exitcode = (int)MANDOCLEVEL_SYSERR;
2245 if (NULL != targetdir)
2246 say("", "&getcwd");
2247 return(0);
2248 }
2249 if (-1 == (fd = open(startdir, O_RDONLY, 0))) {
2250 exitcode = (int)MANDOCLEVEL_SYSERR;
2251 say("", "&open %s", startdir);
2252 return(0);
2253 }
2254 if (NULL == targetdir)
2255 targetdir = startdir;
2256 } else {
2257 if (-1 == fd)
2258 return(0);
2259 if (-1 == fchdir(fd)) {
2260 close(fd);
2261 basedir[0] = '\0';
2262 exitcode = (int)MANDOCLEVEL_SYSERR;
2263 say("", "&chdir %s", startdir);
2264 return(0);
2265 }
2266 if (NULL == targetdir) {
2267 close(fd);
2268 return(1);
2269 }
2270 }
2271 if (NULL == realpath(targetdir, basedir)) {
2272 basedir[0] = '\0';
2273 exitcode = (int)MANDOCLEVEL_BADARG;
2274 say("", "&%s: realpath", targetdir);
2275 return(0);
2276 } else if (-1 == chdir(basedir)) {
2277 exitcode = (int)MANDOCLEVEL_BADARG;
2278 say("", "&chdir");
2279 return(0);
2280 }
2281 return(1);
2282 }
2283
2284 static void
2285 say(const char *file, const char *format, ...)
2286 {
2287 va_list ap;
2288 int use_errno;
2289
2290 if ('\0' != *basedir)
2291 fprintf(stderr, "%s", basedir);
2292 if ('\0' != *basedir && '\0' != *file)
2293 fputs("//", stderr);
2294 if ('\0' != *file)
2295 fprintf(stderr, "%s", file);
2296
2297 use_errno = 1;
2298 if (NULL != format) {
2299 switch (*format) {
2300 case ('&'):
2301 format++;
2302 break;
2303 case ('\0'):
2304 format = NULL;
2305 break;
2306 default:
2307 use_errno = 0;
2308 break;
2309 }
2310 }
2311 if (NULL != format) {
2312 if ('\0' != *basedir || '\0' != *file)
2313 fputs(": ", stderr);
2314 va_start(ap, format);
2315 vfprintf(stderr, format, ap);
2316 va_end(ap);
2317 }
2318 if (use_errno) {
2319 if ('\0' != *basedir || '\0' != *file || NULL != format)
2320 fputs(": ", stderr);
2321 perror(NULL);
2322 } else
2323 fputc('\n', stderr);
2324 }
A git conversion mirror of mandoc.bsd.lv