]> 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
If a manual page is installed gzip(1)ed, let makewhatis(8) take
[mandoc.git] / mandocdb.c
1 /* $OpenBSD$ */
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 #include "config.h"
19
20 #include <sys/stat.h>
21 #include <sys/types.h>
22 #include <sys/wait.h>
23
24 #include <assert.h>
25 #include <ctype.h>
26 #include <errno.h>
27 #include <fcntl.h>
28 #if HAVE_FTS
29 #include <fts.h>
30 #else
31 #include "compat_fts.h"
32 #endif
33 #include <getopt.h>
34 #include <limits.h>
35 #include <stddef.h>
36 #include <stdio.h>
37 #include <stdint.h>
38 #include <stdlib.h>
39 #include <string.h>
40 #include <unistd.h>
41
42 #if HAVE_OHASH
43 #include <ohash.h>
44 #else
45 #include "compat_ohash.h"
46 #endif
47 #include <sqlite3.h>
48
49 #include "mdoc.h"
50 #include "man.h"
51 #include "mandoc.h"
52 #include "mandoc_aux.h"
53 #include "manpath.h"
54 #include "mansearch.h"
55
56 extern int mansearch_keymax;
57 extern const char *const mansearch_keynames[];
58
59 #define SQL_EXEC(_v) \
60 if (SQLITE_OK != sqlite3_exec(db, (_v), NULL, NULL, NULL)) \
61 say("", "%s: %s", (_v), sqlite3_errmsg(db))
62 #define SQL_BIND_TEXT(_s, _i, _v) \
63 if (SQLITE_OK != sqlite3_bind_text \
64 ((_s), (_i)++, (_v), -1, SQLITE_STATIC)) \
65 say(mlink->file, "%s", sqlite3_errmsg(db))
66 #define SQL_BIND_INT(_s, _i, _v) \
67 if (SQLITE_OK != sqlite3_bind_int \
68 ((_s), (_i)++, (_v))) \
69 say(mlink->file, "%s", sqlite3_errmsg(db))
70 #define SQL_BIND_INT64(_s, _i, _v) \
71 if (SQLITE_OK != sqlite3_bind_int64 \
72 ((_s), (_i)++, (_v))) \
73 say(mlink->file, "%s", sqlite3_errmsg(db))
74 #define SQL_STEP(_s) \
75 if (SQLITE_DONE != sqlite3_step((_s))) \
76 say(mlink->file, "%s", sqlite3_errmsg(db))
77
78 enum op {
79 OP_DEFAULT = 0, /* new dbs from dir list or default config */
80 OP_CONFFILE, /* new databases from custom config file */
81 OP_UPDATE, /* delete/add entries in existing database */
82 OP_DELETE, /* delete entries from existing database */
83 OP_TEST /* change no databases, report potential problems */
84 };
85
86 struct str {
87 char *rendered; /* key in UTF-8 or ASCII form */
88 const struct mpage *mpage; /* if set, the owning parse */
89 uint64_t mask; /* bitmask in sequence */
90 char key[]; /* may contain escape sequences */
91 };
92
93 struct inodev {
94 ino_t st_ino;
95 dev_t st_dev;
96 };
97
98 struct mpage {
99 struct inodev inodev; /* used for hashing routine */
100 int64_t pageid; /* pageid in mpages SQL table */
101 char *sec; /* section from file content */
102 char *arch; /* architecture from file content */
103 char *title; /* title from file content */
104 char *desc; /* description from file content */
105 struct mlink *mlinks; /* singly linked list */
106 int form; /* format from file content */
107 };
108
109 struct mlink {
110 char file[PATH_MAX]; /* filename rel. to manpath */
111 char *dsec; /* section from directory */
112 char *arch; /* architecture from directory */
113 char *name; /* name from file name (not empty) */
114 char *fsec; /* section from file name suffix */
115 struct mlink *next; /* singly linked list */
116 struct mpage *mpage; /* parent */
117 int dform; /* format from directory */
118 int fform; /* format from file name suffix */
119 int gzip; /* filename has a .gz suffix */
120 };
121
122 enum stmt {
123 STMT_DELETE_PAGE = 0, /* delete mpage */
124 STMT_INSERT_PAGE, /* insert mpage */
125 STMT_INSERT_LINK, /* insert mlink */
126 STMT_INSERT_NAME, /* insert name */
127 STMT_INSERT_KEY, /* insert parsed key */
128 STMT__MAX
129 };
130
131 typedef int (*mdoc_fp)(struct mpage *, const struct mdoc_node *);
132
133 struct mdoc_handler {
134 mdoc_fp fp; /* optional handler */
135 uint64_t mask; /* set unless handler returns 0 */
136 };
137
138 static void dbclose(int);
139 static void dbadd(struct mpage *, struct mchars *);
140 static void dbadd_mlink(const struct mlink *mlink);
141 static int dbopen(int);
142 static void dbprune(void);
143 static void filescan(const char *);
144 static void *hash_alloc(size_t, void *);
145 static void hash_free(void *, void *);
146 static void *hash_calloc(size_t, size_t, void *);
147 static void mlink_add(struct mlink *, const struct stat *);
148 static void mlink_check(struct mpage *, struct mlink *);
149 static void mlink_free(struct mlink *);
150 static void mlinks_undupe(struct mpage *);
151 static void mpages_free(void);
152 static void mpages_merge(struct mchars *, struct mparse *);
153 static void names_check(void);
154 static void parse_cat(struct mpage *, int);
155 static void parse_man(struct mpage *, const struct man_node *);
156 static void parse_mdoc(struct mpage *, const struct mdoc_node *);
157 static int parse_mdoc_body(struct mpage *, const struct mdoc_node *);
158 static int parse_mdoc_head(struct mpage *, const struct mdoc_node *);
159 static int parse_mdoc_Fd(struct mpage *, const struct mdoc_node *);
160 static int parse_mdoc_Fn(struct mpage *, const struct mdoc_node *);
161 static int parse_mdoc_Nd(struct mpage *, const struct mdoc_node *);
162 static int parse_mdoc_Nm(struct mpage *, const struct mdoc_node *);
163 static int parse_mdoc_Sh(struct mpage *, const struct mdoc_node *);
164 static int parse_mdoc_Xr(struct mpage *, const struct mdoc_node *);
165 static void putkey(const struct mpage *, char *, uint64_t);
166 static void putkeys(const struct mpage *,
167 const char *, size_t, uint64_t);
168 static void putmdockey(const struct mpage *,
169 const struct mdoc_node *, uint64_t);
170 static void render_key(struct mchars *, struct str *);
171 static void say(const char *, const char *, ...);
172 static int set_basedir(const char *);
173 static int treescan(void);
174 static size_t utf8(unsigned int, char [7]);
175
176 static char tempfilename[32];
177 static char *progname;
178 static int nodb; /* no database changes */
179 static int mparse_options; /* abort the parse early */
180 static int use_all; /* use all found files */
181 static int debug; /* print what we're doing */
182 static int warnings; /* warn about crap */
183 static int write_utf8; /* write UTF-8 output; else ASCII */
184 static int exitcode; /* to be returned by main */
185 static enum op op; /* operational mode */
186 static char basedir[PATH_MAX]; /* current base directory */
187 static struct ohash mpages; /* table of distinct manual pages */
188 static struct ohash mlinks; /* table of directory entries */
189 static struct ohash names; /* table of all names */
190 static struct ohash strings; /* table of all strings */
191 static sqlite3 *db = NULL; /* current database */
192 static sqlite3_stmt *stmts[STMT__MAX]; /* current statements */
193 static uint64_t name_mask;
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, 0 }, /* 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
321 int
322 main(int argc, char *argv[])
323 {
324 int ch, i;
325 size_t j, sz;
326 const char *path_arg;
327 struct mchars *mc;
328 struct manpaths dirs;
329 struct mparse *mp;
330 struct ohash_info mpages_info, mlinks_info;
331
332 memset(stmts, 0, STMT__MAX * sizeof(sqlite3_stmt *));
333 memset(&dirs, 0, sizeof(struct manpaths));
334
335 mpages_info.alloc = mlinks_info.alloc = hash_alloc;
336 mpages_info.calloc = mlinks_info.calloc = hash_calloc;
337 mpages_info.free = mlinks_info.free = hash_free;
338
339 mpages_info.key_offset = offsetof(struct mpage, inodev);
340 mlinks_info.key_offset = offsetof(struct mlink, file);
341
342 progname = strrchr(argv[0], '/');
343 if (progname == NULL)
344 progname = argv[0];
345 else
346 ++progname;
347
348 /*
349 * We accept a few different invocations.
350 * The CHECKOP macro makes sure that invocation styles don't
351 * clobber each other.
352 */
353 #define CHECKOP(_op, _ch) do \
354 if (OP_DEFAULT != (_op)) { \
355 fprintf(stderr, "%s: -%c: Conflicting option\n", \
356 progname, (_ch)); \
357 goto usage; \
358 } while (/*CONSTCOND*/0)
359
360 path_arg = NULL;
361 op = OP_DEFAULT;
362
363 while (-1 != (ch = getopt(argc, argv, "aC:Dd:npQT:tu:v")))
364 switch (ch) {
365 case 'a':
366 use_all = 1;
367 break;
368 case 'C':
369 CHECKOP(op, ch);
370 path_arg = optarg;
371 op = OP_CONFFILE;
372 break;
373 case 'D':
374 debug++;
375 break;
376 case 'd':
377 CHECKOP(op, ch);
378 path_arg = optarg;
379 op = OP_UPDATE;
380 break;
381 case 'n':
382 nodb = 1;
383 break;
384 case 'p':
385 warnings = 1;
386 break;
387 case 'Q':
388 mparse_options |= MPARSE_QUICK;
389 break;
390 case 'T':
391 if (strcmp(optarg, "utf8")) {
392 fprintf(stderr, "%s: -T%s: "
393 "Unsupported output format\n",
394 progname, optarg);
395 goto usage;
396 }
397 write_utf8 = 1;
398 break;
399 case 't':
400 CHECKOP(op, ch);
401 dup2(STDOUT_FILENO, STDERR_FILENO);
402 op = OP_TEST;
403 nodb = warnings = 1;
404 break;
405 case 'u':
406 CHECKOP(op, ch);
407 path_arg = optarg;
408 op = OP_DELETE;
409 break;
410 case 'v':
411 /* Compatibility with espie@'s makewhatis. */
412 break;
413 default:
414 goto usage;
415 }
416
417 argc -= optind;
418 argv += optind;
419
420 if (OP_CONFFILE == op && argc > 0) {
421 fprintf(stderr, "%s: -C: Too many arguments\n",
422 progname);
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 mlink->dform = FORM_NONE;
836 if (strlcpy(mlink->file, start, sizeof(mlink->file)) >=
837 sizeof(mlink->file)) {
838 say(start, "Filename too long");
839 return;
840 }
841
842 /*
843 * First try to guess our directory structure.
844 * If we find a separator, try to look for man* or cat*.
845 * If we find one of these and what's underneath is a directory,
846 * assume it's an architecture.
847 */
848 if (NULL != (p = strchr(start, '/'))) {
849 *p++ = '\0';
850 if (0 == strncmp(start, "man", 3)) {
851 mlink->dform = FORM_SRC;
852 mlink->dsec = start + 3;
853 } else if (0 == strncmp(start, "cat", 3)) {
854 mlink->dform = FORM_CAT;
855 mlink->dsec = start + 3;
856 }
857
858 start = p;
859 if (NULL != mlink->dsec && NULL != (p = strchr(start, '/'))) {
860 *p++ = '\0';
861 mlink->arch = start;
862 start = p;
863 }
864 }
865
866 /*
867 * Now check the file suffix.
868 * Suffix of `.0' indicates a catpage, `.1-9' is a manpage.
869 */
870 p = strrchr(start, '\0');
871 while (p-- > start && '/' != *p && '.' != *p)
872 /* Loop. */ ;
873
874 if ('.' == *p) {
875 *p++ = '\0';
876 mlink->fsec = p;
877 }
878
879 /*
880 * Now try to parse the name.
881 * Use the filename portion of the path.
882 */
883 mlink->name = start;
884 if (NULL != (p = strrchr(start, '/'))) {
885 mlink->name = p + 1;
886 *p = '\0';
887 }
888 mlink_add(mlink, &st);
889 }
890
891 static void
892 mlink_add(struct mlink *mlink, const struct stat *st)
893 {
894 struct inodev inodev;
895 struct mpage *mpage;
896 unsigned int slot;
897
898 assert(NULL != mlink->file);
899
900 mlink->dsec = mandoc_strdup(mlink->dsec ? mlink->dsec : "");
901 mlink->arch = mandoc_strdup(mlink->arch ? mlink->arch : "");
902 mlink->name = mandoc_strdup(mlink->name ? mlink->name : "");
903 mlink->fsec = mandoc_strdup(mlink->fsec ? mlink->fsec : "");
904
905 if ('0' == *mlink->fsec) {
906 free(mlink->fsec);
907 mlink->fsec = mandoc_strdup(mlink->dsec);
908 mlink->fform = FORM_CAT;
909 } else if ('1' <= *mlink->fsec && '9' >= *mlink->fsec)
910 mlink->fform = FORM_SRC;
911 else
912 mlink->fform = FORM_NONE;
913
914 slot = ohash_qlookup(&mlinks, mlink->file);
915 assert(NULL == ohash_find(&mlinks, slot));
916 ohash_insert(&mlinks, slot, mlink);
917
918 inodev.st_ino = st->st_ino;
919 inodev.st_dev = st->st_dev;
920 slot = ohash_lookup_memory(&mpages, (char *)&inodev,
921 sizeof(struct inodev), inodev.st_ino);
922 mpage = ohash_find(&mpages, slot);
923 if (NULL == mpage) {
924 mpage = mandoc_calloc(1, sizeof(struct mpage));
925 mpage->inodev.st_ino = inodev.st_ino;
926 mpage->inodev.st_dev = inodev.st_dev;
927 ohash_insert(&mpages, slot, mpage);
928 } else
929 mlink->next = mpage->mlinks;
930 mpage->mlinks = mlink;
931 mlink->mpage = mpage;
932 }
933
934 static void
935 mlink_free(struct mlink *mlink)
936 {
937
938 free(mlink->dsec);
939 free(mlink->arch);
940 free(mlink->name);
941 free(mlink->fsec);
942 free(mlink);
943 }
944
945 static void
946 mpages_free(void)
947 {
948 struct mpage *mpage;
949 struct mlink *mlink;
950 unsigned int slot;
951
952 mpage = ohash_first(&mpages, &slot);
953 while (NULL != mpage) {
954 while (NULL != (mlink = mpage->mlinks)) {
955 mpage->mlinks = mlink->next;
956 mlink_free(mlink);
957 }
958 free(mpage->sec);
959 free(mpage->arch);
960 free(mpage->title);
961 free(mpage->desc);
962 free(mpage);
963 mpage = ohash_next(&mpages, &slot);
964 }
965 }
966
967 /*
968 * For each mlink to the mpage, check whether the path looks like
969 * it is formatted, and if it does, check whether a source manual
970 * exists by the same name, ignoring the suffix.
971 * If both conditions hold, drop the mlink.
972 */
973 static void
974 mlinks_undupe(struct mpage *mpage)
975 {
976 char buf[PATH_MAX];
977 struct mlink **prev;
978 struct mlink *mlink;
979 char *bufp;
980
981 mpage->form = FORM_CAT;
982 prev = &mpage->mlinks;
983 while (NULL != (mlink = *prev)) {
984 if (FORM_CAT != mlink->dform) {
985 mpage->form = FORM_NONE;
986 goto nextlink;
987 }
988 (void)strlcpy(buf, mlink->file, sizeof(buf));
989 bufp = strstr(buf, "cat");
990 assert(NULL != bufp);
991 memcpy(bufp, "man", 3);
992 if (NULL != (bufp = strrchr(buf, '.')))
993 *++bufp = '\0';
994 (void)strlcat(buf, mlink->dsec, sizeof(buf));
995 if (NULL == ohash_find(&mlinks,
996 ohash_qlookup(&mlinks, buf)))
997 goto nextlink;
998 if (warnings)
999 say(mlink->file, "Man source exists: %s", buf);
1000 if (use_all)
1001 goto nextlink;
1002 *prev = mlink->next;
1003 mlink_free(mlink);
1004 continue;
1005 nextlink:
1006 prev = &(*prev)->next;
1007 }
1008 }
1009
1010 static void
1011 mlink_check(struct mpage *mpage, struct mlink *mlink)
1012 {
1013 struct str *str;
1014 unsigned int slot;
1015
1016 /*
1017 * Check whether the manual section given in a file
1018 * agrees with the directory where the file is located.
1019 * Some manuals have suffixes like (3p) on their
1020 * section number either inside the file or in the
1021 * directory name, some are linked into more than one
1022 * section, like encrypt(1) = makekey(8).
1023 */
1024
1025 if (FORM_SRC == mpage->form &&
1026 strcasecmp(mpage->sec, mlink->dsec))
1027 say(mlink->file, "Section \"%s\" manual in %s directory",
1028 mpage->sec, mlink->dsec);
1029
1030 /*
1031 * Manual page directories exist for each kernel
1032 * architecture as returned by machine(1).
1033 * However, many manuals only depend on the
1034 * application architecture as returned by arch(1).
1035 * For example, some (2/ARM) manuals are shared
1036 * across the "armish" and "zaurus" kernel
1037 * architectures.
1038 * A few manuals are even shared across completely
1039 * different architectures, for example fdformat(1)
1040 * on amd64, i386, sparc, and sparc64.
1041 */
1042
1043 if (strcasecmp(mpage->arch, mlink->arch))
1044 say(mlink->file, "Architecture \"%s\" manual in "
1045 "\"%s\" directory", mpage->arch, mlink->arch);
1046
1047 /*
1048 * XXX
1049 * parse_cat() doesn't set NAME_TITLE yet.
1050 */
1051
1052 if (FORM_CAT == mpage->form)
1053 return;
1054
1055 /*
1056 * Check whether this mlink
1057 * appears as a name in the NAME section.
1058 */
1059
1060 slot = ohash_qlookup(&names, mlink->name);
1061 str = ohash_find(&names, slot);
1062 assert(NULL != str);
1063 if ( ! (NAME_TITLE & str->mask))
1064 say(mlink->file, "Name missing in NAME section");
1065 }
1066
1067 /*
1068 * Run through the files in the global vector "mpages"
1069 * and add them to the database specified in "basedir".
1070 *
1071 * This handles the parsing scheme itself, using the cues of directory
1072 * and filename to determine whether the file is parsable or not.
1073 */
1074 static void
1075 mpages_merge(struct mchars *mc, struct mparse *mp)
1076 {
1077 char any[] = "any";
1078 struct ohash_info str_info;
1079 int fd[2];
1080 struct mpage *mpage, *mpage_dest;
1081 struct mlink *mlink, *mlink_dest;
1082 struct mdoc *mdoc;
1083 struct man *man;
1084 char *sodest;
1085 char *cp;
1086 pid_t child_pid;
1087 int status;
1088 unsigned int pslot;
1089 enum mandoclevel lvl;
1090
1091 str_info.alloc = hash_alloc;
1092 str_info.calloc = hash_calloc;
1093 str_info.free = hash_free;
1094 str_info.key_offset = offsetof(struct str, key);
1095
1096 if (0 == nodb)
1097 SQL_EXEC("BEGIN TRANSACTION");
1098
1099 mpage = ohash_first(&mpages, &pslot);
1100 while (NULL != mpage) {
1101 mlinks_undupe(mpage);
1102 if (NULL == mpage->mlinks) {
1103 mpage = ohash_next(&mpages, &pslot);
1104 continue;
1105 }
1106
1107 name_mask = NAME_MASK;
1108 ohash_init(&names, 4, &str_info);
1109 ohash_init(&strings, 6, &str_info);
1110 mparse_reset(mp);
1111 mdoc = NULL;
1112 man = NULL;
1113 sodest = NULL;
1114 child_pid = 0;
1115 fd[0] = -1;
1116 fd[1] = -1;
1117
1118 if (mpage->mlinks->gzip) {
1119 if (-1 == pipe(fd)) {
1120 exitcode = (int)MANDOCLEVEL_SYSERR;
1121 say(mpage->mlinks->file, "&pipe gunzip");
1122 goto nextpage;
1123 }
1124 switch (child_pid = fork()) {
1125 case -1:
1126 exitcode = (int)MANDOCLEVEL_SYSERR;
1127 say(mpage->mlinks->file, "&fork gunzip");
1128 child_pid = 0;
1129 close(fd[1]);
1130 close(fd[0]);
1131 goto nextpage;
1132 case 0:
1133 close(fd[0]);
1134 if (-1 == dup2(fd[1], STDOUT_FILENO)) {
1135 say(mpage->mlinks->file,
1136 "&dup gunzip");
1137 exit(1);
1138 }
1139 execlp("gunzip", "gunzip", "-c",
1140 mpage->mlinks->file, NULL);
1141 say(mpage->mlinks->file, "&exec gunzip");
1142 exit(1);
1143 default:
1144 close(fd[1]);
1145 break;
1146 }
1147 }
1148
1149 /*
1150 * Try interpreting the file as mdoc(7) or man(7)
1151 * source code, unless it is already known to be
1152 * formatted. Fall back to formatted mode.
1153 */
1154 if (FORM_CAT != mpage->mlinks->dform ||
1155 FORM_CAT != mpage->mlinks->fform) {
1156 lvl = mparse_readfd(mp, fd[0], mpage->mlinks->file);
1157 if (lvl < MANDOCLEVEL_FATAL)
1158 mparse_result(mp, &mdoc, &man, &sodest);
1159 }
1160
1161 if (NULL != sodest) {
1162 mlink_dest = ohash_find(&mlinks,
1163 ohash_qlookup(&mlinks, sodest));
1164 if (NULL != mlink_dest) {
1165
1166 /* The .so target exists. */
1167
1168 mpage_dest = mlink_dest->mpage;
1169 mlink = mpage->mlinks;
1170 while (1) {
1171 mlink->mpage = mpage_dest;
1172
1173 /*
1174 * If the target was already
1175 * processed, add the links
1176 * to the database now.
1177 * Otherwise, this will
1178 * happen when we come
1179 * to the target.
1180 */
1181
1182 if (mpage_dest->pageid)
1183 dbadd_mlink(mlink);
1184
1185 if (NULL == mlink->next)
1186 break;
1187 mlink = mlink->next;
1188 }
1189
1190 /* Move all links to the target. */
1191
1192 mlink->next = mlink_dest->next;
1193 mlink_dest->next = mpage->mlinks;
1194 mpage->mlinks = NULL;
1195 }
1196 goto nextpage;
1197 } else if (NULL != mdoc) {
1198 mpage->form = FORM_SRC;
1199 mpage->sec = mdoc_meta(mdoc)->msec;
1200 mpage->sec = mandoc_strdup(
1201 NULL == mpage->sec ? "" : mpage->sec);
1202 mpage->arch = mdoc_meta(mdoc)->arch;
1203 mpage->arch = mandoc_strdup(
1204 NULL == mpage->arch ? "" : mpage->arch);
1205 mpage->title =
1206 mandoc_strdup(mdoc_meta(mdoc)->title);
1207 } else if (NULL != man) {
1208 mpage->form = FORM_SRC;
1209 mpage->sec =
1210 mandoc_strdup(man_meta(man)->msec);
1211 mpage->arch =
1212 mandoc_strdup(mpage->mlinks->arch);
1213 mpage->title =
1214 mandoc_strdup(man_meta(man)->title);
1215 } else {
1216 mpage->form = FORM_CAT;
1217 mpage->sec =
1218 mandoc_strdup(mpage->mlinks->dsec);
1219 mpage->arch =
1220 mandoc_strdup(mpage->mlinks->arch);
1221 mpage->title =
1222 mandoc_strdup(mpage->mlinks->name);
1223 }
1224 if (mpage->mlinks->gzip)
1225 mpage->form |= FORM_GZ;
1226 putkey(mpage, mpage->sec, TYPE_sec);
1227 putkey(mpage, '\0' == *mpage->arch ?
1228 any : mpage->arch, TYPE_arch);
1229
1230 for (mlink = mpage->mlinks; mlink; mlink = mlink->next) {
1231 if ('\0' != *mlink->dsec)
1232 putkey(mpage, mlink->dsec, TYPE_sec);
1233 if ('\0' != *mlink->fsec)
1234 putkey(mpage, mlink->fsec, TYPE_sec);
1235 putkey(mpage, '\0' == *mlink->arch ?
1236 any : mlink->arch, TYPE_arch);
1237 putkey(mpage, mlink->name, NAME_FILE);
1238 }
1239
1240 assert(NULL == mpage->desc);
1241 if (NULL != mdoc) {
1242 if (NULL != (cp = mdoc_meta(mdoc)->name))
1243 putkey(mpage, cp, NAME_HEAD);
1244 parse_mdoc(mpage, mdoc_node(mdoc));
1245 } else if (NULL != man)
1246 parse_man(mpage, man_node(man));
1247 else
1248 parse_cat(mpage, fd[0]);
1249 if (NULL == mpage->desc)
1250 mpage->desc = mandoc_strdup(mpage->mlinks->name);
1251
1252 if (warnings && !use_all)
1253 for (mlink = mpage->mlinks; mlink;
1254 mlink = mlink->next)
1255 mlink_check(mpage, mlink);
1256
1257 dbadd(mpage, mc);
1258
1259 nextpage:
1260 if (child_pid) {
1261 if (-1 == waitpid(child_pid, &status, 0)) {
1262 exitcode = (int)MANDOCLEVEL_SYSERR;
1263 say(mpage->mlinks->file, "&wait gunzip");
1264 } else if (WIFSIGNALED(status)) {
1265 exitcode = (int)MANDOCLEVEL_SYSERR;
1266 say(mpage->mlinks->file,
1267 "gunzip died from signal %d",
1268 WTERMSIG(status));
1269 } else if (WEXITSTATUS(status)) {
1270 exitcode = (int)MANDOCLEVEL_SYSERR;
1271 say(mpage->mlinks->file,
1272 "gunzip failed with code %d",
1273 WEXITSTATUS(status));
1274 }
1275 }
1276 ohash_delete(&strings);
1277 ohash_delete(&names);
1278 mpage = ohash_next(&mpages, &pslot);
1279 }
1280
1281 if (0 == nodb)
1282 SQL_EXEC("END TRANSACTION");
1283 }
1284
1285 static void
1286 names_check(void)
1287 {
1288 sqlite3_stmt *stmt;
1289 const char *name, *sec, *arch, *key;
1290 int irc;
1291
1292 sqlite3_prepare_v2(db,
1293 "SELECT name, sec, arch, key FROM ("
1294 "SELECT name AS key, pageid FROM names "
1295 "WHERE bits & ? AND NOT EXISTS ("
1296 "SELECT pageid FROM mlinks "
1297 "WHERE mlinks.pageid == names.pageid "
1298 "AND mlinks.name == names.name"
1299 ")"
1300 ") JOIN ("
1301 "SELECT sec, arch, name, pageid FROM mlinks "
1302 "GROUP BY pageid"
1303 ") USING (pageid);",
1304 -1, &stmt, NULL);
1305
1306 if (SQLITE_OK != sqlite3_bind_int64(stmt, 1, NAME_TITLE))
1307 say("", "%s", sqlite3_errmsg(db));
1308
1309 while (SQLITE_ROW == (irc = sqlite3_step(stmt))) {
1310 name = (const char *)sqlite3_column_text(stmt, 0);
1311 sec = (const char *)sqlite3_column_text(stmt, 1);
1312 arch = (const char *)sqlite3_column_text(stmt, 2);
1313 key = (const char *)sqlite3_column_text(stmt, 3);
1314 say("", "%s(%s%s%s) lacks mlink \"%s\"", name, sec,
1315 '\0' == *arch ? "" : "/",
1316 '\0' == *arch ? "" : arch, key);
1317 }
1318 sqlite3_finalize(stmt);
1319 }
1320
1321 static void
1322 parse_cat(struct mpage *mpage, int fd)
1323 {
1324 FILE *stream;
1325 char *line, *p, *title;
1326 size_t len, plen, titlesz;
1327
1328 stream = (-1 == fd) ?
1329 fopen(mpage->mlinks->file, "r") :
1330 fdopen(fd, "r");
1331 if (NULL == stream) {
1332 if (-1 != fd)
1333 close(fd);
1334 if (warnings)
1335 say(mpage->mlinks->file, "&fopen");
1336 return;
1337 }
1338
1339 /* Skip to first blank line. */
1340
1341 while (NULL != (line = fgetln(stream, &len)))
1342 if ('\n' == *line)
1343 break;
1344
1345 /*
1346 * Assume the first line that is not indented
1347 * is the first section header. Skip to it.
1348 */
1349
1350 while (NULL != (line = fgetln(stream, &len)))
1351 if ('\n' != *line && ' ' != *line)
1352 break;
1353
1354 /*
1355 * Read up until the next section into a buffer.
1356 * Strip the leading and trailing newline from each read line,
1357 * appending a trailing space.
1358 * Ignore empty (whitespace-only) lines.
1359 */
1360
1361 titlesz = 0;
1362 title = NULL;
1363
1364 while (NULL != (line = fgetln(stream, &len))) {
1365 if (' ' != *line || '\n' != line[len - 1])
1366 break;
1367 while (len > 0 && isspace((unsigned char)*line)) {
1368 line++;
1369 len--;
1370 }
1371 if (1 == len)
1372 continue;
1373 title = mandoc_realloc(title, titlesz + len);
1374 memcpy(title + titlesz, line, len);
1375 titlesz += len;
1376 title[titlesz - 1] = ' ';
1377 }
1378
1379 /*
1380 * If no page content can be found, or the input line
1381 * is already the next section header, or there is no
1382 * trailing newline, reuse the page title as the page
1383 * description.
1384 */
1385
1386 if (NULL == title || '\0' == *title) {
1387 if (warnings)
1388 say(mpage->mlinks->file,
1389 "Cannot find NAME section");
1390 fclose(stream);
1391 free(title);
1392 return;
1393 }
1394
1395 title = mandoc_realloc(title, titlesz + 1);
1396 title[titlesz] = '\0';
1397
1398 /*
1399 * Skip to the first dash.
1400 * Use the remaining line as the description (no more than 70
1401 * bytes).
1402 */
1403
1404 if (NULL != (p = strstr(title, "- "))) {
1405 for (p += 2; ' ' == *p || '\b' == *p; p++)
1406 /* Skip to next word. */ ;
1407 } else {
1408 if (warnings)
1409 say(mpage->mlinks->file,
1410 "No dash in title line");
1411 p = title;
1412 }
1413
1414 plen = strlen(p);
1415
1416 /* Strip backspace-encoding from line. */
1417
1418 while (NULL != (line = memchr(p, '\b', plen))) {
1419 len = line - p;
1420 if (0 == len) {
1421 memmove(line, line + 1, plen--);
1422 continue;
1423 }
1424 memmove(line - 1, line + 1, plen - len);
1425 plen -= 2;
1426 }
1427
1428 mpage->desc = mandoc_strdup(p);
1429 fclose(stream);
1430 free(title);
1431 }
1432
1433 /*
1434 * Put a type/word pair into the word database for this particular file.
1435 */
1436 static void
1437 putkey(const struct mpage *mpage, char *value, uint64_t type)
1438 {
1439 char *cp;
1440
1441 assert(NULL != value);
1442 if (TYPE_arch == type)
1443 for (cp = value; *cp; cp++)
1444 if (isupper((unsigned char)*cp))
1445 *cp = _tolower((unsigned char)*cp);
1446 putkeys(mpage, value, strlen(value), type);
1447 }
1448
1449 /*
1450 * Grok all nodes at or below a certain mdoc node into putkey().
1451 */
1452 static void
1453 putmdockey(const struct mpage *mpage,
1454 const struct mdoc_node *n, uint64_t m)
1455 {
1456
1457 for ( ; NULL != n; n = n->next) {
1458 if (NULL != n->child)
1459 putmdockey(mpage, n->child, m);
1460 if (MDOC_TEXT == n->type)
1461 putkey(mpage, n->string, m);
1462 }
1463 }
1464
1465 static void
1466 parse_man(struct mpage *mpage, const struct man_node *n)
1467 {
1468 const struct man_node *head, *body;
1469 char *start, *title;
1470 char byte;
1471 size_t sz;
1472
1473 if (NULL == n)
1474 return;
1475
1476 /*
1477 * We're only searching for one thing: the first text child in
1478 * the BODY of a NAME section. Since we don't keep track of
1479 * sections in -man, run some hoops to find out whether we're in
1480 * the correct section or not.
1481 */
1482
1483 if (MAN_BODY == n->type && MAN_SH == n->tok) {
1484 body = n;
1485 assert(body->parent);
1486 if (NULL != (head = body->parent->head) &&
1487 1 == head->nchild &&
1488 NULL != (head = (head->child)) &&
1489 MAN_TEXT == head->type &&
1490 0 == strcmp(head->string, "NAME") &&
1491 NULL != body->child) {
1492
1493 /*
1494 * Suck the entire NAME section into memory.
1495 * Yes, we might run away.
1496 * But too many manuals have big, spread-out
1497 * NAME sections over many lines.
1498 */
1499
1500 title = NULL;
1501 man_deroff(&title, body);
1502 if (NULL == title)
1503 return;
1504
1505 /*
1506 * Go through a special heuristic dance here.
1507 * Conventionally, one or more manual names are
1508 * comma-specified prior to a whitespace, then a
1509 * dash, then a description. Try to puzzle out
1510 * the name parts here.
1511 */
1512
1513 start = title;
1514 for ( ;; ) {
1515 sz = strcspn(start, " ,");
1516 if ('\0' == start[sz])
1517 break;
1518
1519 byte = start[sz];
1520 start[sz] = '\0';
1521
1522 /*
1523 * Assume a stray trailing comma in the
1524 * name list if a name begins with a dash.
1525 */
1526
1527 if ('-' == start[0] ||
1528 ('\\' == start[0] && '-' == start[1]))
1529 break;
1530
1531 putkey(mpage, start, NAME_TITLE);
1532
1533 if (' ' == byte) {
1534 start += sz + 1;
1535 break;
1536 }
1537
1538 assert(',' == byte);
1539 start += sz + 1;
1540 while (' ' == *start)
1541 start++;
1542 }
1543
1544 if (start == title) {
1545 putkey(mpage, start, NAME_TITLE);
1546 free(title);
1547 return;
1548 }
1549
1550 while (isspace((unsigned char)*start))
1551 start++;
1552
1553 if (0 == strncmp(start, "-", 1))
1554 start += 1;
1555 else if (0 == strncmp(start, "\\-\\-", 4))
1556 start += 4;
1557 else if (0 == strncmp(start, "\\-", 2))
1558 start += 2;
1559 else if (0 == strncmp(start, "\\(en", 4))
1560 start += 4;
1561 else if (0 == strncmp(start, "\\(em", 4))
1562 start += 4;
1563
1564 while (' ' == *start)
1565 start++;
1566
1567 mpage->desc = mandoc_strdup(start);
1568 free(title);
1569 return;
1570 }
1571 }
1572
1573 for (n = n->child; n; n = n->next) {
1574 if (NULL != mpage->desc)
1575 break;
1576 parse_man(mpage, n);
1577 }
1578 }
1579
1580 static void
1581 parse_mdoc(struct mpage *mpage, const struct mdoc_node *n)
1582 {
1583
1584 assert(NULL != n);
1585 for (n = n->child; NULL != n; n = n->next) {
1586 switch (n->type) {
1587 case MDOC_ELEM:
1588 /* FALLTHROUGH */
1589 case MDOC_BLOCK:
1590 /* FALLTHROUGH */
1591 case MDOC_HEAD:
1592 /* FALLTHROUGH */
1593 case MDOC_BODY:
1594 /* FALLTHROUGH */
1595 case MDOC_TAIL:
1596 if (NULL != mdocs[n->tok].fp)
1597 if (0 == (*mdocs[n->tok].fp)(mpage, n))
1598 break;
1599 if (mdocs[n->tok].mask)
1600 putmdockey(mpage, n->child,
1601 mdocs[n->tok].mask);
1602 break;
1603 default:
1604 assert(MDOC_ROOT != n->type);
1605 continue;
1606 }
1607 if (NULL != n->child)
1608 parse_mdoc(mpage, n);
1609 }
1610 }
1611
1612 static int
1613 parse_mdoc_Fd(struct mpage *mpage, const struct mdoc_node *n)
1614 {
1615 const char *start, *end;
1616 size_t sz;
1617
1618 if (SEC_SYNOPSIS != n->sec ||
1619 NULL == (n = n->child) ||
1620 MDOC_TEXT != n->type)
1621 return(0);
1622
1623 /*
1624 * Only consider those `Fd' macro fields that begin with an
1625 * "inclusion" token (versus, e.g., #define).
1626 */
1627
1628 if (strcmp("#include", n->string))
1629 return(0);
1630
1631 if (NULL == (n = n->next) || MDOC_TEXT != n->type)
1632 return(0);
1633
1634 /*
1635 * Strip away the enclosing angle brackets and make sure we're
1636 * not zero-length.
1637 */
1638
1639 start = n->string;
1640 if ('<' == *start || '"' == *start)
1641 start++;
1642
1643 if (0 == (sz = strlen(start)))
1644 return(0);
1645
1646 end = &start[(int)sz - 1];
1647 if ('>' == *end || '"' == *end)
1648 end--;
1649
1650 if (end > start)
1651 putkeys(mpage, start, end - start + 1, TYPE_In);
1652 return(0);
1653 }
1654
1655 static int
1656 parse_mdoc_Fn(struct mpage *mpage, const struct mdoc_node *n)
1657 {
1658 char *cp;
1659
1660 if (NULL == (n = n->child) || MDOC_TEXT != n->type)
1661 return(0);
1662
1663 /*
1664 * Parse: .Fn "struct type *name" "char *arg".
1665 * First strip away pointer symbol.
1666 * Then store the function name, then type.
1667 * Finally, store the arguments.
1668 */
1669
1670 if (NULL == (cp = strrchr(n->string, ' ')))
1671 cp = n->string;
1672
1673 while ('*' == *cp)
1674 cp++;
1675
1676 putkey(mpage, cp, TYPE_Fn);
1677
1678 if (n->string < cp)
1679 putkeys(mpage, n->string, cp - n->string, TYPE_Ft);
1680
1681 for (n = n->next; NULL != n; n = n->next)
1682 if (MDOC_TEXT == n->type)
1683 putkey(mpage, n->string, TYPE_Fa);
1684
1685 return(0);
1686 }
1687
1688 static int
1689 parse_mdoc_Xr(struct mpage *mpage, const struct mdoc_node *n)
1690 {
1691 char *cp;
1692
1693 if (NULL == (n = n->child))
1694 return(0);
1695
1696 if (NULL == n->next) {
1697 putkey(mpage, n->string, TYPE_Xr);
1698 return(0);
1699 }
1700
1701 mandoc_asprintf(&cp, "%s(%s)", n->string, n->next->string);
1702 putkey(mpage, cp, TYPE_Xr);
1703 free(cp);
1704 return(0);
1705 }
1706
1707 static int
1708 parse_mdoc_Nd(struct mpage *mpage, const struct mdoc_node *n)
1709 {
1710
1711 if (MDOC_BODY == n->type)
1712 mdoc_deroff(&mpage->desc, n);
1713 return(0);
1714 }
1715
1716 static int
1717 parse_mdoc_Nm(struct mpage *mpage, const struct mdoc_node *n)
1718 {
1719
1720 if (SEC_NAME == n->sec)
1721 putmdockey(mpage, n->child, NAME_TITLE);
1722 else if (SEC_SYNOPSIS == n->sec && MDOC_HEAD == n->type)
1723 putmdockey(mpage, n->child, NAME_SYN);
1724 return(0);
1725 }
1726
1727 static int
1728 parse_mdoc_Sh(struct mpage *mpage, const struct mdoc_node *n)
1729 {
1730
1731 return(SEC_CUSTOM == n->sec && MDOC_HEAD == n->type);
1732 }
1733
1734 static int
1735 parse_mdoc_head(struct mpage *mpage, const struct mdoc_node *n)
1736 {
1737
1738 return(MDOC_HEAD == n->type);
1739 }
1740
1741 static int
1742 parse_mdoc_body(struct mpage *mpage, const struct mdoc_node *n)
1743 {
1744
1745 return(MDOC_BODY == n->type);
1746 }
1747
1748 /*
1749 * Add a string to the hash table for the current manual.
1750 * Each string has a bitmask telling which macros it belongs to.
1751 * When we finish the manual, we'll dump the table.
1752 */
1753 static void
1754 putkeys(const struct mpage *mpage,
1755 const char *cp, size_t sz, uint64_t v)
1756 {
1757 struct ohash *htab;
1758 struct str *s;
1759 const char *end;
1760 unsigned int slot;
1761 int i;
1762
1763 if (0 == sz)
1764 return;
1765
1766 if (TYPE_Nm & v) {
1767 htab = &names;
1768 v &= name_mask;
1769 name_mask &= ~NAME_FIRST;
1770 if (debug > 1)
1771 say(mpage->mlinks->file,
1772 "Adding name %*s", sz, cp);
1773 } else {
1774 htab = &strings;
1775 if (debug > 1)
1776 for (i = 0; i < mansearch_keymax; i++)
1777 if (1 << i & v)
1778 say(mpage->mlinks->file,
1779 "Adding key %s=%*s",
1780 mansearch_keynames[i], sz, cp);
1781 }
1782
1783 end = cp + sz;
1784 slot = ohash_qlookupi(htab, cp, &end);
1785 s = ohash_find(htab, slot);
1786
1787 if (NULL != s && mpage == s->mpage) {
1788 s->mask |= v;
1789 return;
1790 } else if (NULL == s) {
1791 s = mandoc_calloc(1, sizeof(struct str) + sz + 1);
1792 memcpy(s->key, cp, sz);
1793 ohash_insert(htab, slot, s);
1794 }
1795 s->mpage = mpage;
1796 s->mask = v;
1797 }
1798
1799 /*
1800 * Take a Unicode codepoint and produce its UTF-8 encoding.
1801 * This isn't the best way to do this, but it works.
1802 * The magic numbers are from the UTF-8 packaging.
1803 * They're not as scary as they seem: read the UTF-8 spec for details.
1804 */
1805 static size_t
1806 utf8(unsigned int cp, char out[7])
1807 {
1808 size_t rc;
1809
1810 rc = 0;
1811 if (cp <= 0x0000007F) {
1812 rc = 1;
1813 out[0] = (char)cp;
1814 } else if (cp <= 0x000007FF) {
1815 rc = 2;
1816 out[0] = (cp >> 6 & 31) | 192;
1817 out[1] = (cp & 63) | 128;
1818 } else if (cp <= 0x0000FFFF) {
1819 rc = 3;
1820 out[0] = (cp >> 12 & 15) | 224;
1821 out[1] = (cp >> 6 & 63) | 128;
1822 out[2] = (cp & 63) | 128;
1823 } else if (cp <= 0x001FFFFF) {
1824 rc = 4;
1825 out[0] = (cp >> 18 & 7) | 240;
1826 out[1] = (cp >> 12 & 63) | 128;
1827 out[2] = (cp >> 6 & 63) | 128;
1828 out[3] = (cp & 63) | 128;
1829 } else if (cp <= 0x03FFFFFF) {
1830 rc = 5;
1831 out[0] = (cp >> 24 & 3) | 248;
1832 out[1] = (cp >> 18 & 63) | 128;
1833 out[2] = (cp >> 12 & 63) | 128;
1834 out[3] = (cp >> 6 & 63) | 128;
1835 out[4] = (cp & 63) | 128;
1836 } else if (cp <= 0x7FFFFFFF) {
1837 rc = 6;
1838 out[0] = (cp >> 30 & 1) | 252;
1839 out[1] = (cp >> 24 & 63) | 128;
1840 out[2] = (cp >> 18 & 63) | 128;
1841 out[3] = (cp >> 12 & 63) | 128;
1842 out[4] = (cp >> 6 & 63) | 128;
1843 out[5] = (cp & 63) | 128;
1844 } else
1845 return(0);
1846
1847 out[rc] = '\0';
1848 return(rc);
1849 }
1850
1851 /*
1852 * Store the rendered version of a key, or alias the pointer
1853 * if the key contains no escape sequences.
1854 */
1855 static void
1856 render_key(struct mchars *mc, struct str *key)
1857 {
1858 size_t sz, bsz, pos;
1859 char utfbuf[7], res[6];
1860 char *buf;
1861 const char *seq, *cpp, *val;
1862 int len, u;
1863 enum mandoc_esc esc;
1864
1865 assert(NULL == key->rendered);
1866
1867 res[0] = '\\';
1868 res[1] = '\t';
1869 res[2] = ASCII_NBRSP;
1870 res[3] = ASCII_HYPH;
1871 res[4] = ASCII_BREAK;
1872 res[5] = '\0';
1873
1874 val = key->key;
1875 bsz = strlen(val);
1876
1877 /*
1878 * Pre-check: if we have no stop-characters, then set the
1879 * pointer as ourselvse and get out of here.
1880 */
1881 if (strcspn(val, res) == bsz) {
1882 key->rendered = key->key;
1883 return;
1884 }
1885
1886 /* Pre-allocate by the length of the input */
1887
1888 buf = mandoc_malloc(++bsz);
1889 pos = 0;
1890
1891 while ('\0' != *val) {
1892 /*
1893 * Halt on the first escape sequence.
1894 * This also halts on the end of string, in which case
1895 * we just copy, fallthrough, and exit the loop.
1896 */
1897 if ((sz = strcspn(val, res)) > 0) {
1898 memcpy(&buf[pos], val, sz);
1899 pos += sz;
1900 val += sz;
1901 }
1902
1903 switch (*val) {
1904 case ASCII_HYPH:
1905 buf[pos++] = '-';
1906 val++;
1907 continue;
1908 case '\t':
1909 /* FALLTHROUGH */
1910 case ASCII_NBRSP:
1911 buf[pos++] = ' ';
1912 val++;
1913 /* FALLTHROUGH */
1914 case ASCII_BREAK:
1915 continue;
1916 default:
1917 break;
1918 }
1919 if ('\\' != *val)
1920 break;
1921
1922 /* Read past the slash. */
1923
1924 val++;
1925
1926 /*
1927 * Parse the escape sequence and see if it's a
1928 * predefined character or special character.
1929 */
1930
1931 esc = mandoc_escape((const char **)&val,
1932 &seq, &len);
1933 if (ESCAPE_ERROR == esc)
1934 break;
1935 if (ESCAPE_SPECIAL != esc)
1936 continue;
1937
1938 /*
1939 * Render the special character
1940 * as either UTF-8 or ASCII.
1941 */
1942
1943 if (write_utf8) {
1944 if (0 == (u = mchars_spec2cp(mc, seq, len)))
1945 continue;
1946 cpp = utfbuf;
1947 if (0 == (sz = utf8(u, utfbuf)))
1948 continue;
1949 sz = strlen(cpp);
1950 } else {
1951 cpp = mchars_spec2str(mc, seq, len, &sz);
1952 if (NULL == cpp)
1953 continue;
1954 if (ASCII_NBRSP == *cpp) {
1955 cpp = " ";
1956 sz = 1;
1957 }
1958 }
1959
1960 /* Copy the rendered glyph into the stream. */
1961
1962 bsz += sz;
1963 buf = mandoc_realloc(buf, bsz);
1964 memcpy(&buf[pos], cpp, sz);
1965 pos += sz;
1966 }
1967
1968 buf[pos] = '\0';
1969 key->rendered = buf;
1970 }
1971
1972 static void
1973 dbadd_mlink(const struct mlink *mlink)
1974 {
1975 size_t i;
1976
1977 i = 1;
1978 SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->dsec);
1979 SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->arch);
1980 SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->name);
1981 SQL_BIND_INT64(stmts[STMT_INSERT_LINK], i, mlink->mpage->pageid);
1982 SQL_STEP(stmts[STMT_INSERT_LINK]);
1983 sqlite3_reset(stmts[STMT_INSERT_LINK]);
1984
1985 i = 1;
1986 SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, NAME_FILE);
1987 SQL_BIND_TEXT(stmts[STMT_INSERT_NAME], i, mlink->name);
1988 SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, mlink->mpage->pageid);
1989 SQL_STEP(stmts[STMT_INSERT_NAME]);
1990 sqlite3_reset(stmts[STMT_INSERT_NAME]);
1991 }
1992
1993 /*
1994 * Flush the current page's terms (and their bits) into the database.
1995 * Wrap the entire set of additions in a transaction to make sqlite be a
1996 * little faster.
1997 * Also, handle escape sequences at the last possible moment.
1998 */
1999 static void
2000 dbadd(struct mpage *mpage, struct mchars *mc)
2001 {
2002 struct mlink *mlink;
2003 struct str *key;
2004 size_t i;
2005 unsigned int slot;
2006
2007 mlink = mpage->mlinks;
2008
2009 if (nodb) {
2010 for (key = ohash_first(&names, &slot); NULL != key;
2011 key = ohash_next(&names, &slot)) {
2012 if (key->rendered != key->key)
2013 free(key->rendered);
2014 free(key);
2015 }
2016 for (key = ohash_first(&strings, &slot); NULL != key;
2017 key = ohash_next(&strings, &slot)) {
2018 if (key->rendered != key->key)
2019 free(key->rendered);
2020 free(key);
2021 }
2022 if (0 == debug)
2023 return;
2024 while (NULL != mlink) {
2025 fputs(mlink->name, stdout);
2026 if (NULL == mlink->next ||
2027 strcmp(mlink->dsec, mlink->next->dsec) ||
2028 strcmp(mlink->fsec, mlink->next->fsec) ||
2029 strcmp(mlink->arch, mlink->next->arch)) {
2030 putchar('(');
2031 if ('\0' == *mlink->dsec)
2032 fputs(mlink->fsec, stdout);
2033 else
2034 fputs(mlink->dsec, stdout);
2035 if ('\0' != *mlink->arch)
2036 printf("/%s", mlink->arch);
2037 putchar(')');
2038 }
2039 mlink = mlink->next;
2040 if (NULL != mlink)
2041 fputs(", ", stdout);
2042 }
2043 printf(" - %s\n", mpage->desc);
2044 return;
2045 }
2046
2047 if (debug)
2048 say(mlink->file, "Adding to database");
2049
2050 i = strlen(mpage->desc) + 1;
2051 key = mandoc_calloc(1, sizeof(struct str) + i);
2052 memcpy(key->key, mpage->desc, i);
2053 render_key(mc, key);
2054
2055 i = 1;
2056 SQL_BIND_TEXT(stmts[STMT_INSERT_PAGE], i, key->rendered);
2057 SQL_BIND_INT(stmts[STMT_INSERT_PAGE], i, mpage->form);
2058 SQL_STEP(stmts[STMT_INSERT_PAGE]);
2059 mpage->pageid = sqlite3_last_insert_rowid(db);
2060 sqlite3_reset(stmts[STMT_INSERT_PAGE]);
2061
2062 if (key->rendered != key->key)
2063 free(key->rendered);
2064 free(key);
2065
2066 while (NULL != mlink) {
2067 dbadd_mlink(mlink);
2068 mlink = mlink->next;
2069 }
2070 mlink = mpage->mlinks;
2071
2072 for (key = ohash_first(&names, &slot); NULL != key;
2073 key = ohash_next(&names, &slot)) {
2074 assert(key->mpage == mpage);
2075 if (NULL == key->rendered)
2076 render_key(mc, key);
2077 i = 1;
2078 SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, key->mask);
2079 SQL_BIND_TEXT(stmts[STMT_INSERT_NAME], i, key->rendered);
2080 SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, mpage->pageid);
2081 SQL_STEP(stmts[STMT_INSERT_NAME]);
2082 sqlite3_reset(stmts[STMT_INSERT_NAME]);
2083 if (key->rendered != key->key)
2084 free(key->rendered);
2085 free(key);
2086 }
2087 for (key = ohash_first(&strings, &slot); NULL != key;
2088 key = ohash_next(&strings, &slot)) {
2089 assert(key->mpage == mpage);
2090 if (NULL == key->rendered)
2091 render_key(mc, key);
2092 i = 1;
2093 SQL_BIND_INT64(stmts[STMT_INSERT_KEY], i, key->mask);
2094 SQL_BIND_TEXT(stmts[STMT_INSERT_KEY], i, key->rendered);
2095 SQL_BIND_INT64(stmts[STMT_INSERT_KEY], i, mpage->pageid);
2096 SQL_STEP(stmts[STMT_INSERT_KEY]);
2097 sqlite3_reset(stmts[STMT_INSERT_KEY]);
2098 if (key->rendered != key->key)
2099 free(key->rendered);
2100 free(key);
2101 }
2102 }
2103
2104 static void
2105 dbprune(void)
2106 {
2107 struct mpage *mpage;
2108 struct mlink *mlink;
2109 size_t i;
2110 unsigned int slot;
2111
2112 if (0 == nodb)
2113 SQL_EXEC("BEGIN TRANSACTION");
2114
2115 for (mpage = ohash_first(&mpages, &slot); NULL != mpage;
2116 mpage = ohash_next(&mpages, &slot)) {
2117 mlink = mpage->mlinks;
2118 if (debug)
2119 say(mlink->file, "Deleting from database");
2120 if (nodb)
2121 continue;
2122 for ( ; NULL != mlink; mlink = mlink->next) {
2123 i = 1;
2124 SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE],
2125 i, mlink->dsec);
2126 SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE],
2127 i, mlink->arch);
2128 SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE],
2129 i, mlink->name);
2130 SQL_STEP(stmts[STMT_DELETE_PAGE]);
2131 sqlite3_reset(stmts[STMT_DELETE_PAGE]);
2132 }
2133 }
2134
2135 if (0 == nodb)
2136 SQL_EXEC("END TRANSACTION");
2137 }
2138
2139 /*
2140 * Close an existing database and its prepared statements.
2141 * If "real" is not set, rename the temporary file into the real one.
2142 */
2143 static void
2144 dbclose(int real)
2145 {
2146 size_t i;
2147 int status;
2148 pid_t child;
2149
2150 if (nodb)
2151 return;
2152
2153 for (i = 0; i < STMT__MAX; i++) {
2154 sqlite3_finalize(stmts[i]);
2155 stmts[i] = NULL;
2156 }
2157
2158 sqlite3_close(db);
2159 db = NULL;
2160
2161 if (real)
2162 return;
2163
2164 if ('\0' == *tempfilename) {
2165 if (-1 == rename(MANDOC_DB "~", MANDOC_DB)) {
2166 exitcode = (int)MANDOCLEVEL_SYSERR;
2167 say(MANDOC_DB, "&rename");
2168 }
2169 return;
2170 }
2171
2172 switch (child = fork()) {
2173 case -1:
2174 exitcode = (int)MANDOCLEVEL_SYSERR;
2175 say("", "&fork cmp");
2176 return;
2177 case 0:
2178 execlp("cmp", "cmp", "-s",
2179 tempfilename, MANDOC_DB, NULL);
2180 say("", "&exec cmp");
2181 exit(0);
2182 default:
2183 break;
2184 }
2185 if (-1 == waitpid(child, &status, 0)) {
2186 exitcode = (int)MANDOCLEVEL_SYSERR;
2187 say("", "&wait cmp");
2188 } else if (WIFSIGNALED(status)) {
2189 exitcode = (int)MANDOCLEVEL_SYSERR;
2190 say("", "cmp died from signal %d", WTERMSIG(status));
2191 } else if (WEXITSTATUS(status)) {
2192 exitcode = (int)MANDOCLEVEL_SYSERR;
2193 say(MANDOC_DB,
2194 "Data changed, but cannot replace database");
2195 }
2196
2197 *strrchr(tempfilename, '/') = '\0';
2198 switch (child = fork()) {
2199 case -1:
2200 exitcode = (int)MANDOCLEVEL_SYSERR;
2201 say("", "&fork rm");
2202 return;
2203 case 0:
2204 execlp("rm", "rm", "-rf", tempfilename, NULL);
2205 say("", "&exec rm");
2206 exit((int)MANDOCLEVEL_SYSERR);
2207 default:
2208 break;
2209 }
2210 if (-1 == waitpid(child, &status, 0)) {
2211 exitcode = (int)MANDOCLEVEL_SYSERR;
2212 say("", "&wait rm");
2213 } else if (WIFSIGNALED(status) || WEXITSTATUS(status)) {
2214 exitcode = (int)MANDOCLEVEL_SYSERR;
2215 say("", "%s: Cannot remove temporary directory",
2216 tempfilename);
2217 }
2218 }
2219
2220 /*
2221 * This is straightforward stuff.
2222 * Open a database connection to a "temporary" database, then open a set
2223 * of prepared statements we'll use over and over again.
2224 * If "real" is set, we use the existing database; if not, we truncate a
2225 * temporary one.
2226 * Must be matched by dbclose().
2227 */
2228 static int
2229 dbopen(int real)
2230 {
2231 const char *sql;
2232 int rc, ofl;
2233
2234 if (nodb)
2235 return(1);
2236
2237 *tempfilename = '\0';
2238 ofl = SQLITE_OPEN_READWRITE;
2239
2240 if (real) {
2241 rc = sqlite3_open_v2(MANDOC_DB, &db, ofl, NULL);
2242 if (SQLITE_OK != rc) {
2243 exitcode = (int)MANDOCLEVEL_SYSERR;
2244 if (SQLITE_CANTOPEN != rc)
2245 say(MANDOC_DB, "%s", sqlite3_errstr(rc));
2246 return(0);
2247 }
2248 goto prepare_statements;
2249 }
2250
2251 ofl |= SQLITE_OPEN_CREATE | SQLITE_OPEN_EXCLUSIVE;
2252
2253 remove(MANDOC_DB "~");
2254 rc = sqlite3_open_v2(MANDOC_DB "~", &db, ofl, NULL);
2255 if (SQLITE_OK == rc)
2256 goto create_tables;
2257 if (MPARSE_QUICK & mparse_options) {
2258 exitcode = (int)MANDOCLEVEL_SYSERR;
2259 say(MANDOC_DB "~", "%s", sqlite3_errstr(rc));
2260 return(0);
2261 }
2262
2263 (void)strlcpy(tempfilename, "/tmp/mandocdb.XXXXXX",
2264 sizeof(tempfilename));
2265 if (NULL == mkdtemp(tempfilename)) {
2266 exitcode = (int)MANDOCLEVEL_SYSERR;
2267 say("", "&%s", tempfilename);
2268 return(0);
2269 }
2270 (void)strlcat(tempfilename, "/" MANDOC_DB,
2271 sizeof(tempfilename));
2272 rc = sqlite3_open_v2(tempfilename, &db, ofl, NULL);
2273 if (SQLITE_OK != rc) {
2274 exitcode = (int)MANDOCLEVEL_SYSERR;
2275 say("", "%s: %s", tempfilename, sqlite3_errstr(rc));
2276 return(0);
2277 }
2278
2279 create_tables:
2280 sql = "CREATE TABLE \"mpages\" (\n"
2281 " \"desc\" TEXT NOT NULL,\n"
2282 " \"form\" INTEGER NOT NULL,\n"
2283 " \"pageid\" INTEGER PRIMARY KEY AUTOINCREMENT NOT NULL\n"
2284 ");\n"
2285 "\n"
2286 "CREATE TABLE \"mlinks\" (\n"
2287 " \"sec\" TEXT NOT NULL,\n"
2288 " \"arch\" TEXT NOT NULL,\n"
2289 " \"name\" TEXT NOT NULL,\n"
2290 " \"pageid\" INTEGER NOT NULL REFERENCES mpages(pageid) "
2291 "ON DELETE CASCADE\n"
2292 ");\n"
2293 "CREATE INDEX mlinks_pageid_idx ON mlinks (pageid);\n"
2294 "\n"
2295 "CREATE TABLE \"names\" (\n"
2296 " \"bits\" INTEGER NOT NULL,\n"
2297 " \"name\" TEXT NOT NULL,\n"
2298 " \"pageid\" INTEGER NOT NULL REFERENCES mpages(pageid) "
2299 "ON DELETE CASCADE\n"
2300 ");\n"
2301 "\n"
2302 "CREATE TABLE \"keys\" (\n"
2303 " \"bits\" INTEGER NOT NULL,\n"
2304 " \"key\" TEXT NOT NULL,\n"
2305 " \"pageid\" INTEGER NOT NULL REFERENCES mpages(pageid) "
2306 "ON DELETE CASCADE\n"
2307 ");\n"
2308 "CREATE INDEX keys_pageid_idx ON keys (pageid);\n";
2309
2310 if (SQLITE_OK != sqlite3_exec(db, sql, NULL, NULL, NULL)) {
2311 exitcode = (int)MANDOCLEVEL_SYSERR;
2312 say(MANDOC_DB, "%s", sqlite3_errmsg(db));
2313 sqlite3_close(db);
2314 return(0);
2315 }
2316
2317 prepare_statements:
2318 if (SQLITE_OK != sqlite3_exec(db,
2319 "PRAGMA foreign_keys = ON", NULL, NULL, NULL)) {
2320 exitcode = (int)MANDOCLEVEL_SYSERR;
2321 say(MANDOC_DB, "PRAGMA foreign_keys: %s",
2322 sqlite3_errmsg(db));
2323 sqlite3_close(db);
2324 return(0);
2325 }
2326
2327 sql = "DELETE FROM mpages WHERE pageid IN "
2328 "(SELECT pageid FROM mlinks WHERE "
2329 "sec=? AND arch=? AND name=?)";
2330 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_DELETE_PAGE], NULL);
2331 sql = "INSERT INTO mpages "
2332 "(desc,form) VALUES (?,?)";
2333 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_PAGE], NULL);
2334 sql = "INSERT INTO mlinks "
2335 "(sec,arch,name,pageid) VALUES (?,?,?,?)";
2336 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_LINK], NULL);
2337 sql = "INSERT INTO names "
2338 "(bits,name,pageid) VALUES (?,?,?)";
2339 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_NAME], NULL);
2340 sql = "INSERT INTO keys "
2341 "(bits,key,pageid) VALUES (?,?,?)";
2342 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_KEY], NULL);
2343
2344 #ifndef __APPLE__
2345 /*
2346 * When opening a new database, we can turn off
2347 * synchronous mode for much better performance.
2348 */
2349
2350 if (real && SQLITE_OK != sqlite3_exec(db,
2351 "PRAGMA synchronous = OFF", NULL, NULL, NULL)) {
2352 exitcode = (int)MANDOCLEVEL_SYSERR;
2353 say(MANDOC_DB, "PRAGMA synchronous: %s",
2354 sqlite3_errmsg(db));
2355 sqlite3_close(db);
2356 return(0);
2357 }
2358 #endif
2359
2360 return(1);
2361 }
2362
2363 static void *
2364 hash_calloc(size_t n, size_t sz, void *arg)
2365 {
2366
2367 return(mandoc_calloc(n, sz));
2368 }
2369
2370 static void *
2371 hash_alloc(size_t sz, void *arg)
2372 {
2373
2374 return(mandoc_malloc(sz));
2375 }
2376
2377 static void
2378 hash_free(void *p, void *arg)
2379 {
2380
2381 free(p);
2382 }
2383
2384 static int
2385 set_basedir(const char *targetdir)
2386 {
2387 static char startdir[PATH_MAX];
2388 static int getcwd_status; /* 1 = ok, 2 = failure */
2389 static int chdir_status; /* 1 = changed directory */
2390 char *cp;
2391
2392 /*
2393 * Remember the original working directory, if possible.
2394 * This will be needed if the second or a later directory
2395 * on the command line is given as a relative path.
2396 * Do not error out if the current directory is not
2397 * searchable: Maybe it won't be needed after all.
2398 */
2399 if (0 == getcwd_status) {
2400 if (NULL == getcwd(startdir, sizeof(startdir))) {
2401 getcwd_status = 2;
2402 (void)strlcpy(startdir, strerror(errno),
2403 sizeof(startdir));
2404 } else
2405 getcwd_status = 1;
2406 }
2407
2408 /*
2409 * We are leaving the old base directory.
2410 * Do not use it any longer, not even for messages.
2411 */
2412 *basedir = '\0';
2413
2414 /*
2415 * If and only if the directory was changed earlier and
2416 * the next directory to process is given as a relative path,
2417 * first go back, or bail out if that is impossible.
2418 */
2419 if (chdir_status && '/' != *targetdir) {
2420 if (2 == getcwd_status) {
2421 exitcode = (int)MANDOCLEVEL_SYSERR;
2422 say("", "getcwd: %s", startdir);
2423 return(0);
2424 }
2425 if (-1 == chdir(startdir)) {
2426 exitcode = (int)MANDOCLEVEL_SYSERR;
2427 say("", "&chdir %s", startdir);
2428 return(0);
2429 }
2430 }
2431
2432 /*
2433 * Always resolve basedir to the canonicalized absolute
2434 * pathname and append a trailing slash, such that
2435 * we can reliably check whether files are inside.
2436 */
2437 if (NULL == realpath(targetdir, basedir)) {
2438 exitcode = (int)MANDOCLEVEL_BADARG;
2439 say("", "&%s: realpath", targetdir);
2440 return(0);
2441 } else if (-1 == chdir(basedir)) {
2442 exitcode = (int)MANDOCLEVEL_BADARG;
2443 say("", "&chdir");
2444 return(0);
2445 }
2446 chdir_status = 1;
2447 cp = strchr(basedir, '\0');
2448 if ('/' != cp[-1]) {
2449 if (cp - basedir >= PATH_MAX - 1) {
2450 exitcode = (int)MANDOCLEVEL_SYSERR;
2451 say("", "Filename too long");
2452 return(0);
2453 }
2454 *cp++ = '/';
2455 *cp = '\0';
2456 }
2457 return(1);
2458 }
2459
2460 static void
2461 say(const char *file, const char *format, ...)
2462 {
2463 va_list ap;
2464 int use_errno;
2465
2466 if ('\0' != *basedir)
2467 fprintf(stderr, "%s", basedir);
2468 if ('\0' != *basedir && '\0' != *file)
2469 fputc('/', stderr);
2470 if ('\0' != *file)
2471 fprintf(stderr, "%s", file);
2472
2473 use_errno = 1;
2474 if (NULL != format) {
2475 switch (*format) {
2476 case '&':
2477 format++;
2478 break;
2479 case '\0':
2480 format = NULL;
2481 break;
2482 default:
2483 use_errno = 0;
2484 break;
2485 }
2486 }
2487 if (NULL != format) {
2488 if ('\0' != *basedir || '\0' != *file)
2489 fputs(": ", stderr);
2490 va_start(ap, format);
2491 vfprintf(stderr, format, ap);
2492 va_end(ap);
2493 }
2494 if (use_errno) {
2495 if ('\0' != *basedir || '\0' != *file || NULL != format)
2496 fputs(": ", stderr);
2497 perror(NULL);
2498 } else
2499 fputc('\n', stderr);
2500 }
A git conversion mirror of mandoc.bsd.lv