]> git.cameronkatri.com Git - mandoc.git/blob - mandocdb.c
git.ckatri.com / mandoc.git / blob
? search:


986c6a50b0d47a1eb2ac03dd93358876b5e212c5
[mandoc.git] / mandocdb.c
1 /* $Id: mandocdb.c,v 1.172 2014/12/04 17:36:00 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 #include "config.h"
19
20 #include <sys/types.h>
21 #include <sys/stat.h>
22 #include <sys/wait.h>
23
24 #include <assert.h>
25 #include <ctype.h>
26 #include <errno.h>
27 #include <fcntl.h>
28 #if HAVE_FTS
29 #include <fts.h>
30 #else
31 #include "compat_fts.h"
32 #endif
33 #include <getopt.h>
34 #include <limits.h>
35 #include <stddef.h>
36 #include <stdio.h>
37 #include <stdint.h>
38 #include <stdlib.h>
39 #include <string.h>
40 #include <unistd.h>
41
42 #if HAVE_OHASH
43 #include <ohash.h>
44 #else
45 #include "compat_ohash.h"
46 #endif
47 #include <sqlite3.h>
48
49 #include "mdoc.h"
50 #include "man.h"
51 #include "mandoc.h"
52 #include "mandoc_aux.h"
53 #include "manpath.h"
54 #include "mansearch.h"
55
56 extern int mansearch_keymax;
57 extern const char *const mansearch_keynames[];
58
59 #define SQL_EXEC(_v) \
60 if (SQLITE_OK != sqlite3_exec(db, (_v), NULL, NULL, NULL)) \
61 say("", "%s: %s", (_v), sqlite3_errmsg(db))
62 #define SQL_BIND_TEXT(_s, _i, _v) \
63 if (SQLITE_OK != sqlite3_bind_text \
64 ((_s), (_i)++, (_v), -1, SQLITE_STATIC)) \
65 say(mlink->file, "%s", sqlite3_errmsg(db))
66 #define SQL_BIND_INT(_s, _i, _v) \
67 if (SQLITE_OK != sqlite3_bind_int \
68 ((_s), (_i)++, (_v))) \
69 say(mlink->file, "%s", sqlite3_errmsg(db))
70 #define SQL_BIND_INT64(_s, _i, _v) \
71 if (SQLITE_OK != sqlite3_bind_int64 \
72 ((_s), (_i)++, (_v))) \
73 say(mlink->file, "%s", sqlite3_errmsg(db))
74 #define SQL_STEP(_s) \
75 if (SQLITE_DONE != sqlite3_step((_s))) \
76 say(mlink->file, "%s", sqlite3_errmsg(db))
77
78 enum op {
79 OP_DEFAULT = 0, /* new dbs from dir list or default config */
80 OP_CONFFILE, /* new databases from custom config file */
81 OP_UPDATE, /* delete/add entries in existing database */
82 OP_DELETE, /* delete entries from existing database */
83 OP_TEST /* change no databases, report potential problems */
84 };
85
86 struct str {
87 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_meta *,
132 const struct mdoc_node *);
133
134 struct mdoc_handler {
135 mdoc_fp fp; /* optional handler */
136 uint64_t mask; /* set unless handler returns 0 */
137 };
138
139 static void dbclose(int);
140 static void dbadd(struct mpage *, struct mchars *);
141 static void dbadd_mlink(const struct mlink *mlink);
142 static void dbadd_mlink_name(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 *, void *);
148 static void *hash_calloc(size_t, size_t, void *);
149 static void mlink_add(struct mlink *, const struct stat *);
150 static void 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_meta *,
159 const struct mdoc_node *);
160 static int parse_mdoc_body(struct mpage *, const struct mdoc_meta *,
161 const struct mdoc_node *);
162 static int parse_mdoc_head(struct mpage *, const struct mdoc_meta *,
163 const struct mdoc_node *);
164 static int parse_mdoc_Fd(struct mpage *, const struct mdoc_meta *,
165 const struct mdoc_node *);
166 static int parse_mdoc_Fn(struct mpage *, const struct mdoc_meta *,
167 const struct mdoc_node *);
168 static int parse_mdoc_Nd(struct mpage *, const struct mdoc_meta *,
169 const struct mdoc_node *);
170 static int parse_mdoc_Nm(struct mpage *, const struct mdoc_meta *,
171 const struct mdoc_node *);
172 static int parse_mdoc_Sh(struct mpage *, const struct mdoc_meta *,
173 const struct mdoc_node *);
174 static int parse_mdoc_Xr(struct mpage *, const struct mdoc_meta *,
175 const struct mdoc_node *);
176 static void putkey(const struct mpage *, char *, uint64_t);
177 static void putkeys(const struct mpage *,
178 const char *, size_t, uint64_t);
179 static void putmdockey(const struct mpage *,
180 const struct mdoc_node *, uint64_t);
181 static void render_key(struct mchars *, struct str *);
182 static void say(const char *, const char *, ...);
183 static int set_basedir(const char *, int);
184 static int treescan(void);
185 static size_t utf8(unsigned int, char [7]);
186
187 static char tempfilename[32];
188 static char *progname;
189 static int nodb; /* no database changes */
190 static int mparse_options; /* abort the parse early */
191 static int use_all; /* use all found files */
192 static int debug; /* print what we're doing */
193 static int warnings; /* warn about crap */
194 static int write_utf8; /* write UTF-8 output; else ASCII */
195 static int exitcode; /* to be returned by main */
196 static enum op op; /* operational mode */
197 static char basedir[PATH_MAX]; /* current base directory */
198 static struct ohash mpages; /* table of distinct manual pages */
199 static struct ohash mlinks; /* table of directory entries */
200 static struct ohash names; /* table of all names */
201 static struct ohash strings; /* table of all strings */
202 static sqlite3 *db = NULL; /* current database */
203 static sqlite3_stmt *stmts[STMT__MAX]; /* current statements */
204 static uint64_t name_mask;
205
206 static const struct mdoc_handler mdocs[MDOC_MAX] = {
207 { NULL, 0 }, /* Ap */
208 { NULL, 0 }, /* Dd */
209 { NULL, 0 }, /* Dt */
210 { NULL, 0 }, /* Os */
211 { parse_mdoc_Sh, TYPE_Sh }, /* Sh */
212 { parse_mdoc_head, TYPE_Ss }, /* Ss */
213 { NULL, 0 }, /* Pp */
214 { NULL, 0 }, /* D1 */
215 { NULL, 0 }, /* Dl */
216 { NULL, 0 }, /* Bd */
217 { NULL, 0 }, /* Ed */
218 { NULL, 0 }, /* Bl */
219 { NULL, 0 }, /* El */
220 { NULL, 0 }, /* It */
221 { NULL, 0 }, /* Ad */
222 { NULL, TYPE_An }, /* An */
223 { NULL, TYPE_Ar }, /* Ar */
224 { NULL, TYPE_Cd }, /* Cd */
225 { NULL, TYPE_Cm }, /* Cm */
226 { NULL, TYPE_Dv }, /* Dv */
227 { NULL, TYPE_Er }, /* Er */
228 { NULL, TYPE_Ev }, /* Ev */
229 { NULL, 0 }, /* Ex */
230 { NULL, TYPE_Fa }, /* Fa */
231 { parse_mdoc_Fd, 0 }, /* Fd */
232 { NULL, TYPE_Fl }, /* Fl */
233 { parse_mdoc_Fn, 0 }, /* Fn */
234 { NULL, TYPE_Ft }, /* Ft */
235 { NULL, TYPE_Ic }, /* Ic */
236 { NULL, TYPE_In }, /* In */
237 { NULL, TYPE_Li }, /* Li */
238 { parse_mdoc_Nd, 0 }, /* Nd */
239 { parse_mdoc_Nm, 0 }, /* Nm */
240 { NULL, 0 }, /* Op */
241 { NULL, 0 }, /* Ot */
242 { NULL, TYPE_Pa }, /* Pa */
243 { NULL, 0 }, /* Rv */
244 { NULL, TYPE_St }, /* St */
245 { NULL, TYPE_Va }, /* Va */
246 { parse_mdoc_body, TYPE_Va }, /* Vt */
247 { parse_mdoc_Xr, 0 }, /* Xr */
248 { NULL, 0 }, /* %A */
249 { NULL, 0 }, /* %B */
250 { NULL, 0 }, /* %D */
251 { NULL, 0 }, /* %I */
252 { NULL, 0 }, /* %J */
253 { NULL, 0 }, /* %N */
254 { NULL, 0 }, /* %O */
255 { NULL, 0 }, /* %P */
256 { NULL, 0 }, /* %R */
257 { NULL, 0 }, /* %T */
258 { NULL, 0 }, /* %V */
259 { NULL, 0 }, /* Ac */
260 { NULL, 0 }, /* Ao */
261 { NULL, 0 }, /* Aq */
262 { NULL, TYPE_At }, /* At */
263 { NULL, 0 }, /* Bc */
264 { NULL, 0 }, /* Bf */
265 { NULL, 0 }, /* Bo */
266 { NULL, 0 }, /* Bq */
267 { NULL, TYPE_Bsx }, /* Bsx */
268 { NULL, TYPE_Bx }, /* Bx */
269 { NULL, 0 }, /* Db */
270 { NULL, 0 }, /* Dc */
271 { NULL, 0 }, /* Do */
272 { NULL, 0 }, /* Dq */
273 { NULL, 0 }, /* Ec */
274 { NULL, 0 }, /* Ef */
275 { NULL, TYPE_Em }, /* Em */
276 { NULL, 0 }, /* Eo */
277 { NULL, TYPE_Fx }, /* Fx */
278 { NULL, TYPE_Ms }, /* Ms */
279 { NULL, 0 }, /* No */
280 { NULL, 0 }, /* Ns */
281 { NULL, TYPE_Nx }, /* Nx */
282 { NULL, TYPE_Ox }, /* Ox */
283 { NULL, 0 }, /* Pc */
284 { NULL, 0 }, /* Pf */
285 { NULL, 0 }, /* Po */
286 { NULL, 0 }, /* Pq */
287 { NULL, 0 }, /* Qc */
288 { NULL, 0 }, /* Ql */
289 { NULL, 0 }, /* Qo */
290 { NULL, 0 }, /* Qq */
291 { NULL, 0 }, /* Re */
292 { NULL, 0 }, /* Rs */
293 { NULL, 0 }, /* Sc */
294 { NULL, 0 }, /* So */
295 { NULL, 0 }, /* Sq */
296 { NULL, 0 }, /* Sm */
297 { NULL, 0 }, /* Sx */
298 { NULL, TYPE_Sy }, /* Sy */
299 { NULL, TYPE_Tn }, /* Tn */
300 { NULL, 0 }, /* Ux */
301 { NULL, 0 }, /* Xc */
302 { NULL, 0 }, /* Xo */
303 { parse_mdoc_head, 0 }, /* Fo */
304 { NULL, 0 }, /* Fc */
305 { NULL, 0 }, /* Oo */
306 { NULL, 0 }, /* Oc */
307 { NULL, 0 }, /* Bk */
308 { NULL, 0 }, /* Ek */
309 { NULL, 0 }, /* Bt */
310 { NULL, 0 }, /* Hf */
311 { NULL, 0 }, /* Fr */
312 { NULL, 0 }, /* Ud */
313 { NULL, TYPE_Lb }, /* Lb */
314 { NULL, 0 }, /* Lp */
315 { NULL, TYPE_Lk }, /* Lk */
316 { NULL, TYPE_Mt }, /* Mt */
317 { NULL, 0 }, /* Brq */
318 { NULL, 0 }, /* Bro */
319 { NULL, 0 }, /* Brc */
320 { NULL, 0 }, /* %C */
321 { NULL, 0 }, /* Es */
322 { NULL, 0 }, /* En */
323 { NULL, TYPE_Dx }, /* Dx */
324 { NULL, 0 }, /* %Q */
325 { NULL, 0 }, /* br */
326 { NULL, 0 }, /* sp */
327 { NULL, 0 }, /* %U */
328 { NULL, 0 }, /* Ta */
329 { NULL, 0 }, /* ll */
330 };
331
332
333 int
334 main(int argc, char *argv[])
335 {
336 int ch, i;
337 size_t j, sz;
338 const char *path_arg;
339 struct mchars *mc;
340 struct manpaths dirs;
341 struct mparse *mp;
342 struct ohash_info mpages_info, mlinks_info;
343
344 memset(stmts, 0, STMT__MAX * sizeof(sqlite3_stmt *));
345 memset(&dirs, 0, sizeof(struct manpaths));
346
347 mpages_info.alloc = mlinks_info.alloc = hash_alloc;
348 mpages_info.calloc = mlinks_info.calloc = hash_calloc;
349 mpages_info.free = mlinks_info.free = hash_free;
350
351 mpages_info.key_offset = offsetof(struct mpage, inodev);
352 mlinks_info.key_offset = offsetof(struct mlink, file);
353
354 progname = strrchr(argv[0], '/');
355 if (progname == NULL)
356 progname = argv[0];
357 else
358 ++progname;
359
360 /*
361 * We accept a few different invocations.
362 * The CHECKOP macro makes sure that invocation styles don't
363 * clobber each other.
364 */
365 #define CHECKOP(_op, _ch) do \
366 if (OP_DEFAULT != (_op)) { \
367 fprintf(stderr, "%s: -%c: Conflicting option\n", \
368 progname, (_ch)); \
369 goto usage; \
370 } while (/*CONSTCOND*/0)
371
372 path_arg = NULL;
373 op = OP_DEFAULT;
374
375 while (-1 != (ch = getopt(argc, argv, "aC:Dd:npQT:tu:v")))
376 switch (ch) {
377 case 'a':
378 use_all = 1;
379 break;
380 case 'C':
381 CHECKOP(op, ch);
382 path_arg = optarg;
383 op = OP_CONFFILE;
384 break;
385 case 'D':
386 debug++;
387 break;
388 case 'd':
389 CHECKOP(op, ch);
390 path_arg = optarg;
391 op = OP_UPDATE;
392 break;
393 case 'n':
394 nodb = 1;
395 break;
396 case 'p':
397 warnings = 1;
398 break;
399 case 'Q':
400 mparse_options |= MPARSE_QUICK;
401 break;
402 case 'T':
403 if (strcmp(optarg, "utf8")) {
404 fprintf(stderr, "%s: -T%s: "
405 "Unsupported output format\n",
406 progname, optarg);
407 goto usage;
408 }
409 write_utf8 = 1;
410 break;
411 case 't':
412 CHECKOP(op, ch);
413 dup2(STDOUT_FILENO, STDERR_FILENO);
414 op = OP_TEST;
415 nodb = warnings = 1;
416 break;
417 case 'u':
418 CHECKOP(op, ch);
419 path_arg = optarg;
420 op = OP_DELETE;
421 break;
422 case 'v':
423 /* Compatibility with espie@'s makewhatis. */
424 break;
425 default:
426 goto usage;
427 }
428
429 argc -= optind;
430 argv += optind;
431
432 if (OP_CONFFILE == op && argc > 0) {
433 fprintf(stderr, "%s: -C: Too many arguments\n",
434 progname);
435 goto usage;
436 }
437
438 exitcode = (int)MANDOCLEVEL_OK;
439 mc = mchars_alloc();
440 mp = mparse_alloc(mparse_options, MANDOCLEVEL_FATAL, NULL,
441 mc, NULL);
442 ohash_init(&mpages, 6, &mpages_info);
443 ohash_init(&mlinks, 6, &mlinks_info);
444
445 if (OP_UPDATE == op || OP_DELETE == op || OP_TEST == op) {
446
447 /*
448 * Most of these deal with a specific directory.
449 * Jump into that directory first.
450 */
451 if (OP_TEST != op && 0 == set_basedir(path_arg, 1))
452 goto out;
453
454 if (dbopen(1)) {
455 /*
456 * The existing database is usable. Process
457 * all files specified on the command-line.
458 */
459 use_all = 1;
460 for (i = 0; i < argc; i++)
461 filescan(argv[i]);
462 if (OP_TEST != op)
463 dbprune();
464 } else {
465 /*
466 * Database missing or corrupt.
467 * Recreate from scratch.
468 */
469 exitcode = (int)MANDOCLEVEL_OK;
470 op = OP_DEFAULT;
471 if (0 == treescan())
472 goto out;
473 if (0 == dbopen(0))
474 goto out;
475 }
476 if (OP_DELETE != op)
477 mpages_merge(mc, mp);
478 dbclose(OP_DEFAULT == op ? 0 : 1);
479 } else {
480 /*
481 * If we have arguments, use them as our manpaths.
482 * If we don't, grok from manpath(1) or however else
483 * manpath_parse() wants to do it.
484 */
485 if (argc > 0) {
486 dirs.paths = mandoc_reallocarray(NULL,
487 argc, sizeof(char *));
488 dirs.sz = (size_t)argc;
489 for (i = 0; i < argc; i++)
490 dirs.paths[i] = mandoc_strdup(argv[i]);
491 } else
492 manpath_parse(&dirs, path_arg, NULL, NULL);
493
494 if (0 == dirs.sz) {
495 exitcode = (int)MANDOCLEVEL_BADARG;
496 say("", "Empty manpath");
497 }
498
499 /*
500 * First scan the tree rooted at a base directory, then
501 * build a new database and finally move it into place.
502 * Ignore zero-length directories and strip trailing
503 * slashes.
504 */
505 for (j = 0; j < dirs.sz; j++) {
506 sz = strlen(dirs.paths[j]);
507 if (sz && '/' == dirs.paths[j][sz - 1])
508 dirs.paths[j][--sz] = '\0';
509 if (0 == sz)
510 continue;
511
512 if (j) {
513 ohash_init(&mpages, 6, &mpages_info);
514 ohash_init(&mlinks, 6, &mlinks_info);
515 }
516
517 if (0 == set_basedir(dirs.paths[j], argc > 0))
518 continue;
519 if (0 == treescan())
520 continue;
521 if (0 == dbopen(0))
522 continue;
523
524 mpages_merge(mc, mp);
525 if (warnings && !nodb &&
526 ! (MPARSE_QUICK & mparse_options))
527 names_check();
528 dbclose(0);
529
530 if (j + 1 < dirs.sz) {
531 mpages_free();
532 ohash_delete(&mpages);
533 ohash_delete(&mlinks);
534 }
535 }
536 }
537 out:
538 manpath_free(&dirs);
539 mparse_free(mp);
540 mchars_free(mc);
541 mpages_free();
542 ohash_delete(&mpages);
543 ohash_delete(&mlinks);
544 return(exitcode);
545 usage:
546 fprintf(stderr, "usage: %s [-aDnpQ] [-C file] [-Tutf8]\n"
547 " %s [-aDnpQ] [-Tutf8] dir ...\n"
548 " %s [-DnpQ] [-Tutf8] -d dir [file ...]\n"
549 " %s [-Dnp] -u dir [file ...]\n"
550 " %s [-Q] -t file ...\n",
551 progname, progname, progname,
552 progname, progname);
553
554 return((int)MANDOCLEVEL_BADARG);
555 }
556
557 /*
558 * Scan a directory tree rooted at "basedir" for manpages.
559 * We use fts(), scanning directory parts along the way for clues to our
560 * section and architecture.
561 *
562 * If use_all has been specified, grok all files.
563 * If not, sanitise paths to the following:
564 *
565 * [./]man*[/<arch>]/<name>.<section>
566 * or
567 * [./]cat<section>[/<arch>]/<name>.0
568 *
569 * TODO: accomodate for multi-language directories.
570 */
571 static int
572 treescan(void)
573 {
574 char buf[PATH_MAX];
575 FTS *f;
576 FTSENT *ff;
577 struct mlink *mlink;
578 int dform, gzip;
579 char *dsec, *arch, *fsec, *cp;
580 const char *path;
581 const char *argv[2];
582
583 argv[0] = ".";
584 argv[1] = (char *)NULL;
585
586 f = fts_open((char * const *)argv,
587 FTS_PHYSICAL | FTS_NOCHDIR, NULL);
588 if (NULL == f) {
589 exitcode = (int)MANDOCLEVEL_SYSERR;
590 say("", "&fts_open");
591 return(0);
592 }
593
594 dsec = arch = NULL;
595 dform = FORM_NONE;
596
597 while (NULL != (ff = fts_read(f))) {
598 path = ff->fts_path + 2;
599 switch (ff->fts_info) {
600
601 /*
602 * Symbolic links require various sanity checks,
603 * then get handled just like regular files.
604 */
605 case FTS_SL:
606 if (NULL == realpath(path, buf)) {
607 if (warnings)
608 say(path, "&realpath");
609 continue;
610 }
611 if (strstr(buf, basedir) != buf) {
612 if (warnings) say("",
613 "%s: outside base directory", buf);
614 continue;
615 }
616 /* Use logical inode to avoid mpages dupe. */
617 if (-1 == stat(path, ff->fts_statp)) {
618 if (warnings)
619 say(path, "&stat");
620 continue;
621 }
622 /* FALLTHROUGH */
623
624 /*
625 * If we're a regular file, add an mlink by using the
626 * stored directory data and handling the filename.
627 */
628 case FTS_F:
629 if (0 == strcmp(path, MANDOC_DB))
630 continue;
631 if ( ! use_all && ff->fts_level < 2) {
632 if (warnings)
633 say(path, "Extraneous file");
634 continue;
635 }
636 gzip = 0;
637 fsec = NULL;
638 while (NULL == fsec) {
639 fsec = strrchr(ff->fts_name, '.');
640 if (NULL == fsec || strcmp(fsec+1, "gz"))
641 break;
642 gzip = 1;
643 *fsec = '\0';
644 fsec = NULL;
645 }
646 if (NULL == fsec) {
647 if ( ! use_all) {
648 if (warnings)
649 say(path,
650 "No filename suffix");
651 continue;
652 }
653 } else if (0 == strcmp(++fsec, "html")) {
654 if (warnings)
655 say(path, "Skip html");
656 continue;
657 } else if (0 == strcmp(fsec, "ps")) {
658 if (warnings)
659 say(path, "Skip ps");
660 continue;
661 } else if (0 == strcmp(fsec, "pdf")) {
662 if (warnings)
663 say(path, "Skip pdf");
664 continue;
665 } else if ( ! use_all &&
666 ((FORM_SRC == dform && strcmp(fsec, dsec)) ||
667 (FORM_CAT == dform && strcmp(fsec, "0")))) {
668 if (warnings)
669 say(path, "Wrong filename suffix");
670 continue;
671 } else
672 fsec[-1] = '\0';
673
674 mlink = mandoc_calloc(1, sizeof(struct mlink));
675 if (strlcpy(mlink->file, path,
676 sizeof(mlink->file)) >=
677 sizeof(mlink->file)) {
678 say(path, "Filename too long");
679 free(mlink);
680 continue;
681 }
682 mlink->dform = dform;
683 mlink->dsec = dsec;
684 mlink->arch = arch;
685 mlink->name = ff->fts_name;
686 mlink->fsec = fsec;
687 mlink->gzip = gzip;
688 mlink_add(mlink, ff->fts_statp);
689 continue;
690
691 case FTS_D:
692 /* FALLTHROUGH */
693 case FTS_DP:
694 break;
695
696 default:
697 if (warnings)
698 say(path, "Not a regular file");
699 continue;
700 }
701
702 switch (ff->fts_level) {
703 case 0:
704 /* Ignore the root directory. */
705 break;
706 case 1:
707 /*
708 * This might contain manX/ or catX/.
709 * Try to infer this from the name.
710 * If we're not in use_all, enforce it.
711 */
712 cp = ff->fts_name;
713 if (FTS_DP == ff->fts_info)
714 break;
715
716 if (0 == strncmp(cp, "man", 3)) {
717 dform = FORM_SRC;
718 dsec = cp + 3;
719 } else if (0 == strncmp(cp, "cat", 3)) {
720 dform = FORM_CAT;
721 dsec = cp + 3;
722 } else {
723 dform = FORM_NONE;
724 dsec = NULL;
725 }
726
727 if (NULL != dsec || use_all)
728 break;
729
730 if (warnings)
731 say(path, "Unknown directory part");
732 fts_set(f, ff, FTS_SKIP);
733 break;
734 case 2:
735 /*
736 * Possibly our architecture.
737 * If we're descending, keep tabs on it.
738 */
739 if (FTS_DP != ff->fts_info && NULL != dsec)
740 arch = ff->fts_name;
741 else
742 arch = NULL;
743 break;
744 default:
745 if (FTS_DP == ff->fts_info || use_all)
746 break;
747 if (warnings)
748 say(path, "Extraneous directory part");
749 fts_set(f, ff, FTS_SKIP);
750 break;
751 }
752 }
753
754 fts_close(f);
755 return(1);
756 }
757
758 /*
759 * Add a file to the mlinks table.
760 * Do not verify that it's a "valid" looking manpage (we'll do that
761 * later).
762 *
763 * Try to infer the manual section, architecture, and page name from the
764 * path, assuming it looks like
765 *
766 * [./]man*[/<arch>]/<name>.<section>
767 * or
768 * [./]cat<section>[/<arch>]/<name>.0
769 *
770 * See treescan() for the fts(3) version of this.
771 */
772 static void
773 filescan(const char *file)
774 {
775 char buf[PATH_MAX];
776 struct stat st;
777 struct mlink *mlink;
778 char *p, *start;
779
780 assert(use_all);
781
782 if (0 == strncmp(file, "./", 2))
783 file += 2;
784
785 /*
786 * We have to do lstat(2) before realpath(3) loses
787 * the information whether this is a symbolic link.
788 * We need to know that because for symbolic links,
789 * we want to use the orginal file name, while for
790 * regular files, we want to use the real path.
791 */
792 if (-1 == lstat(file, &st)) {
793 exitcode = (int)MANDOCLEVEL_BADARG;
794 say(file, "&lstat");
795 return;
796 } else if (0 == ((S_IFREG | S_IFLNK) & st.st_mode)) {
797 exitcode = (int)MANDOCLEVEL_BADARG;
798 say(file, "Not a regular file");
799 return;
800 }
801
802 /*
803 * We have to resolve the file name to the real path
804 * in any case for the base directory check.
805 */
806 if (NULL == realpath(file, buf)) {
807 exitcode = (int)MANDOCLEVEL_BADARG;
808 say(file, "&realpath");
809 return;
810 }
811
812 if (OP_TEST == op)
813 start = buf;
814 else if (strstr(buf, basedir) == buf)
815 start = buf + strlen(basedir);
816 else {
817 exitcode = (int)MANDOCLEVEL_BADARG;
818 say("", "%s: outside base directory", buf);
819 return;
820 }
821
822 /*
823 * Now we are sure the file is inside our tree.
824 * If it is a symbolic link, ignore the real path
825 * and use the original name.
826 * This implies passing stuff like "cat1/../man1/foo.1"
827 * on the command line won't work. So don't do that.
828 * Note the stat(2) can still fail if the link target
829 * doesn't exist.
830 */
831 if (S_IFLNK & st.st_mode) {
832 if (-1 == stat(buf, &st)) {
833 exitcode = (int)MANDOCLEVEL_BADARG;
834 say(file, "&stat");
835 return;
836 }
837 if (strlcpy(buf, file, sizeof(buf)) >= sizeof(buf)) {
838 say(file, "Filename too long");
839 return;
840 }
841 start = buf;
842 if (OP_TEST != op && strstr(buf, basedir) == buf)
843 start += strlen(basedir);
844 }
845
846 mlink = mandoc_calloc(1, sizeof(struct mlink));
847 mlink->dform = FORM_NONE;
848 if (strlcpy(mlink->file, start, sizeof(mlink->file)) >=
849 sizeof(mlink->file)) {
850 say(start, "Filename too long");
851 return;
852 }
853
854 /*
855 * First try to guess our directory structure.
856 * If we find a separator, try to look for man* or cat*.
857 * If we find one of these and what's underneath is a directory,
858 * assume it's an architecture.
859 */
860 if (NULL != (p = strchr(start, '/'))) {
861 *p++ = '\0';
862 if (0 == strncmp(start, "man", 3)) {
863 mlink->dform = FORM_SRC;
864 mlink->dsec = start + 3;
865 } else if (0 == strncmp(start, "cat", 3)) {
866 mlink->dform = FORM_CAT;
867 mlink->dsec = start + 3;
868 }
869
870 start = p;
871 if (NULL != mlink->dsec && NULL != (p = strchr(start, '/'))) {
872 *p++ = '\0';
873 mlink->arch = start;
874 start = p;
875 }
876 }
877
878 /*
879 * Now check the file suffix.
880 * Suffix of `.0' indicates a catpage, `.1-9' is a manpage.
881 */
882 p = strrchr(start, '\0');
883 while (p-- > start && '/' != *p && '.' != *p)
884 /* Loop. */ ;
885
886 if ('.' == *p) {
887 *p++ = '\0';
888 mlink->fsec = p;
889 }
890
891 /*
892 * Now try to parse the name.
893 * Use the filename portion of the path.
894 */
895 mlink->name = start;
896 if (NULL != (p = strrchr(start, '/'))) {
897 mlink->name = p + 1;
898 *p = '\0';
899 }
900 mlink_add(mlink, &st);
901 }
902
903 static void
904 mlink_add(struct mlink *mlink, const struct stat *st)
905 {
906 struct inodev inodev;
907 struct mpage *mpage;
908 unsigned int slot;
909
910 assert(NULL != mlink->file);
911
912 mlink->dsec = mandoc_strdup(mlink->dsec ? mlink->dsec : "");
913 mlink->arch = mandoc_strdup(mlink->arch ? mlink->arch : "");
914 mlink->name = mandoc_strdup(mlink->name ? mlink->name : "");
915 mlink->fsec = mandoc_strdup(mlink->fsec ? mlink->fsec : "");
916
917 if ('0' == *mlink->fsec) {
918 free(mlink->fsec);
919 mlink->fsec = mandoc_strdup(mlink->dsec);
920 mlink->fform = FORM_CAT;
921 } else if ('1' <= *mlink->fsec && '9' >= *mlink->fsec)
922 mlink->fform = FORM_SRC;
923 else
924 mlink->fform = FORM_NONE;
925
926 slot = ohash_qlookup(&mlinks, mlink->file);
927 assert(NULL == ohash_find(&mlinks, slot));
928 ohash_insert(&mlinks, slot, mlink);
929
930 inodev.st_ino = st->st_ino;
931 inodev.st_dev = st->st_dev;
932 slot = ohash_lookup_memory(&mpages, (char *)&inodev,
933 sizeof(struct inodev), inodev.st_ino);
934 mpage = ohash_find(&mpages, slot);
935 if (NULL == mpage) {
936 mpage = mandoc_calloc(1, sizeof(struct mpage));
937 mpage->inodev.st_ino = inodev.st_ino;
938 mpage->inodev.st_dev = inodev.st_dev;
939 ohash_insert(&mpages, slot, mpage);
940 } else
941 mlink->next = mpage->mlinks;
942 mpage->mlinks = mlink;
943 mlink->mpage = mpage;
944 }
945
946 static void
947 mlink_free(struct mlink *mlink)
948 {
949
950 free(mlink->dsec);
951 free(mlink->arch);
952 free(mlink->name);
953 free(mlink->fsec);
954 free(mlink);
955 }
956
957 static void
958 mpages_free(void)
959 {
960 struct mpage *mpage;
961 struct mlink *mlink;
962 unsigned int slot;
963
964 mpage = ohash_first(&mpages, &slot);
965 while (NULL != mpage) {
966 while (NULL != (mlink = mpage->mlinks)) {
967 mpage->mlinks = mlink->next;
968 mlink_free(mlink);
969 }
970 free(mpage->sec);
971 free(mpage->arch);
972 free(mpage->title);
973 free(mpage->desc);
974 free(mpage);
975 mpage = ohash_next(&mpages, &slot);
976 }
977 }
978
979 /*
980 * For each mlink to the mpage, check whether the path looks like
981 * it is formatted, and if it does, check whether a source manual
982 * exists by the same name, ignoring the suffix.
983 * If both conditions hold, drop the mlink.
984 */
985 static void
986 mlinks_undupe(struct mpage *mpage)
987 {
988 char buf[PATH_MAX];
989 struct mlink **prev;
990 struct mlink *mlink;
991 char *bufp;
992
993 mpage->form = FORM_CAT;
994 prev = &mpage->mlinks;
995 while (NULL != (mlink = *prev)) {
996 if (FORM_CAT != mlink->dform) {
997 mpage->form = FORM_NONE;
998 goto nextlink;
999 }
1000 (void)strlcpy(buf, mlink->file, sizeof(buf));
1001 bufp = strstr(buf, "cat");
1002 assert(NULL != bufp);
1003 memcpy(bufp, "man", 3);
1004 if (NULL != (bufp = strrchr(buf, '.')))
1005 *++bufp = '\0';
1006 (void)strlcat(buf, mlink->dsec, sizeof(buf));
1007 if (NULL == ohash_find(&mlinks,
1008 ohash_qlookup(&mlinks, buf)))
1009 goto nextlink;
1010 if (warnings)
1011 say(mlink->file, "Man source exists: %s", buf);
1012 if (use_all)
1013 goto nextlink;
1014 *prev = mlink->next;
1015 mlink_free(mlink);
1016 continue;
1017 nextlink:
1018 prev = &(*prev)->next;
1019 }
1020 }
1021
1022 static void
1023 mlink_check(struct mpage *mpage, struct mlink *mlink)
1024 {
1025 struct str *str;
1026 unsigned int slot;
1027
1028 /*
1029 * Check whether the manual section given in a file
1030 * agrees with the directory where the file is located.
1031 * Some manuals have suffixes like (3p) on their
1032 * section number either inside the file or in the
1033 * directory name, some are linked into more than one
1034 * section, like encrypt(1) = makekey(8).
1035 */
1036
1037 if (FORM_SRC == mpage->form &&
1038 strcasecmp(mpage->sec, mlink->dsec))
1039 say(mlink->file, "Section \"%s\" manual in %s directory",
1040 mpage->sec, mlink->dsec);
1041
1042 /*
1043 * Manual page directories exist for each kernel
1044 * architecture as returned by machine(1).
1045 * However, many manuals only depend on the
1046 * application architecture as returned by arch(1).
1047 * For example, some (2/ARM) manuals are shared
1048 * across the "armish" and "zaurus" kernel
1049 * architectures.
1050 * A few manuals are even shared across completely
1051 * different architectures, for example fdformat(1)
1052 * on amd64, i386, sparc, and sparc64.
1053 */
1054
1055 if (strcasecmp(mpage->arch, mlink->arch))
1056 say(mlink->file, "Architecture \"%s\" manual in "
1057 "\"%s\" directory", mpage->arch, mlink->arch);
1058
1059 /*
1060 * XXX
1061 * parse_cat() doesn't set NAME_TITLE yet.
1062 */
1063
1064 if (FORM_CAT == mpage->form)
1065 return;
1066
1067 /*
1068 * Check whether this mlink
1069 * appears as a name in the NAME section.
1070 */
1071
1072 slot = ohash_qlookup(&names, mlink->name);
1073 str = ohash_find(&names, slot);
1074 assert(NULL != str);
1075 if ( ! (NAME_TITLE & str->mask))
1076 say(mlink->file, "Name missing in NAME section");
1077 }
1078
1079 /*
1080 * Run through the files in the global vector "mpages"
1081 * and add them to the database specified in "basedir".
1082 *
1083 * This handles the parsing scheme itself, using the cues of directory
1084 * and filename to determine whether the file is parsable or not.
1085 */
1086 static void
1087 mpages_merge(struct mchars *mc, struct mparse *mp)
1088 {
1089 char any[] = "any";
1090 struct ohash_info str_info;
1091 struct mpage *mpage, *mpage_dest;
1092 struct mlink *mlink, *mlink_dest;
1093 struct mdoc *mdoc;
1094 struct man *man;
1095 char *sodest;
1096 char *cp;
1097 int fd;
1098 unsigned int pslot;
1099 enum mandoclevel lvl;
1100
1101 str_info.alloc = hash_alloc;
1102 str_info.calloc = hash_calloc;
1103 str_info.free = hash_free;
1104 str_info.key_offset = offsetof(struct str, key);
1105
1106 if ( ! nodb)
1107 SQL_EXEC("BEGIN TRANSACTION");
1108
1109 mpage = ohash_first(&mpages, &pslot);
1110 while (mpage != NULL) {
1111 mlinks_undupe(mpage);
1112 if (mpage->mlinks == NULL) {
1113 mpage = ohash_next(&mpages, &pslot);
1114 continue;
1115 }
1116
1117 name_mask = NAME_MASK;
1118 ohash_init(&names, 4, &str_info);
1119 ohash_init(&strings, 6, &str_info);
1120 mparse_reset(mp);
1121 mdoc = NULL;
1122 man = NULL;
1123 sodest = NULL;
1124
1125 mparse_open(mp, &fd, mpage->mlinks->file);
1126 if (fd == -1) {
1127 say(mpage->mlinks->file, "&open");
1128 goto nextpage;
1129 }
1130
1131 /*
1132 * Try interpreting the file as mdoc(7) or man(7)
1133 * source code, unless it is already known to be
1134 * formatted. Fall back to formatted mode.
1135 */
1136 if (mpage->mlinks->dform != FORM_CAT ||
1137 mpage->mlinks->fform != FORM_CAT) {
1138 lvl = mparse_readfd(mp, fd, mpage->mlinks->file);
1139 if (lvl < MANDOCLEVEL_FATAL)
1140 mparse_result(mp, &mdoc, &man, &sodest);
1141 }
1142
1143 if (sodest != NULL) {
1144 mlink_dest = ohash_find(&mlinks,
1145 ohash_qlookup(&mlinks, sodest));
1146 if (mlink_dest == NULL) {
1147 mandoc_asprintf(&cp, "%s.gz", sodest);
1148 mlink_dest = ohash_find(&mlinks,
1149 ohash_qlookup(&mlinks, cp));
1150 free(cp);
1151 }
1152 if (mlink_dest != NULL) {
1153
1154 /* The .so target exists. */
1155
1156 mpage_dest = mlink_dest->mpage;
1157 mlink = mpage->mlinks;
1158 while (1) {
1159 mlink->mpage = mpage_dest;
1160
1161 /*
1162 * If the target was already
1163 * processed, add the links
1164 * to the database now.
1165 * Otherwise, this will
1166 * happen when we come
1167 * to the target.
1168 */
1169
1170 if (mpage_dest->pageid)
1171 dbadd_mlink_name(mlink);
1172
1173 if (mlink->next == NULL)
1174 break;
1175 mlink = mlink->next;
1176 }
1177
1178 /* Move all links to the target. */
1179
1180 mlink->next = mlink_dest->next;
1181 mlink_dest->next = mpage->mlinks;
1182 mpage->mlinks = NULL;
1183 }
1184 goto nextpage;
1185 } else if (mdoc != NULL) {
1186 mpage->form = FORM_SRC;
1187 mpage->sec = mdoc_meta(mdoc)->msec;
1188 mpage->sec = mandoc_strdup(
1189 mpage->sec == NULL ? "" : mpage->sec);
1190 mpage->arch = mdoc_meta(mdoc)->arch;
1191 mpage->arch = mandoc_strdup(
1192 mpage->arch == NULL ? "" : mpage->arch);
1193 mpage->title =
1194 mandoc_strdup(mdoc_meta(mdoc)->title);
1195 } else if (man != NULL) {
1196 mpage->form = FORM_SRC;
1197 mpage->sec =
1198 mandoc_strdup(man_meta(man)->msec);
1199 mpage->arch =
1200 mandoc_strdup(mpage->mlinks->arch);
1201 mpage->title =
1202 mandoc_strdup(man_meta(man)->title);
1203 } else {
1204 mpage->form = FORM_CAT;
1205 mpage->sec =
1206 mandoc_strdup(mpage->mlinks->dsec);
1207 mpage->arch =
1208 mandoc_strdup(mpage->mlinks->arch);
1209 mpage->title =
1210 mandoc_strdup(mpage->mlinks->name);
1211 }
1212 putkey(mpage, mpage->sec, TYPE_sec);
1213 if (*mpage->arch != '\0')
1214 putkey(mpage, mpage->arch, TYPE_arch);
1215
1216 for (mlink = mpage->mlinks; mlink; mlink = mlink->next) {
1217 if ('\0' != *mlink->dsec)
1218 putkey(mpage, mlink->dsec, TYPE_sec);
1219 if ('\0' != *mlink->fsec)
1220 putkey(mpage, mlink->fsec, TYPE_sec);
1221 putkey(mpage, '\0' == *mlink->arch ?
1222 any : mlink->arch, TYPE_arch);
1223 putkey(mpage, mlink->name, NAME_FILE);
1224 }
1225
1226 assert(NULL == mpage->desc);
1227 if (NULL != mdoc) {
1228 if (NULL != (cp = mdoc_meta(mdoc)->name))
1229 putkey(mpage, cp, NAME_HEAD);
1230 parse_mdoc(mpage, mdoc_meta(mdoc), mdoc_node(mdoc));
1231 } else if (NULL != man)
1232 parse_man(mpage, man_node(man));
1233 else
1234 parse_cat(mpage, fd);
1235 if (NULL == mpage->desc)
1236 mpage->desc = mandoc_strdup(mpage->mlinks->name);
1237
1238 if (warnings && !use_all)
1239 for (mlink = mpage->mlinks; mlink;
1240 mlink = mlink->next)
1241 mlink_check(mpage, mlink);
1242
1243 dbadd(mpage, mc);
1244
1245 nextpage:
1246 if (mparse_wait(mp) != MANDOCLEVEL_OK) {
1247 exitcode = (int)MANDOCLEVEL_SYSERR;
1248 say(mpage->mlinks->file, "&wait gunzip");
1249 }
1250 ohash_delete(&strings);
1251 ohash_delete(&names);
1252 mpage = ohash_next(&mpages, &pslot);
1253 }
1254
1255 if (0 == nodb)
1256 SQL_EXEC("END TRANSACTION");
1257 }
1258
1259 static void
1260 names_check(void)
1261 {
1262 sqlite3_stmt *stmt;
1263 const char *name, *sec, *arch, *key;
1264 int irc;
1265
1266 sqlite3_prepare_v2(db,
1267 "SELECT name, sec, arch, key FROM ("
1268 "SELECT name AS key, pageid FROM names "
1269 "WHERE bits & ? AND NOT EXISTS ("
1270 "SELECT pageid FROM mlinks "
1271 "WHERE mlinks.pageid == names.pageid "
1272 "AND mlinks.name == names.name"
1273 ")"
1274 ") JOIN ("
1275 "SELECT sec, arch, name, pageid FROM mlinks "
1276 "GROUP BY pageid"
1277 ") USING (pageid);",
1278 -1, &stmt, NULL);
1279
1280 if (SQLITE_OK != sqlite3_bind_int64(stmt, 1, NAME_TITLE))
1281 say("", "%s", sqlite3_errmsg(db));
1282
1283 while (SQLITE_ROW == (irc = sqlite3_step(stmt))) {
1284 name = (const char *)sqlite3_column_text(stmt, 0);
1285 sec = (const char *)sqlite3_column_text(stmt, 1);
1286 arch = (const char *)sqlite3_column_text(stmt, 2);
1287 key = (const char *)sqlite3_column_text(stmt, 3);
1288 say("", "%s(%s%s%s) lacks mlink \"%s\"", name, sec,
1289 '\0' == *arch ? "" : "/",
1290 '\0' == *arch ? "" : arch, key);
1291 }
1292 sqlite3_finalize(stmt);
1293 }
1294
1295 static void
1296 parse_cat(struct mpage *mpage, int fd)
1297 {
1298 FILE *stream;
1299 char *line, *p, *title;
1300 size_t len, plen, titlesz;
1301
1302 stream = (-1 == fd) ?
1303 fopen(mpage->mlinks->file, "r") :
1304 fdopen(fd, "r");
1305 if (NULL == stream) {
1306 if (-1 != fd)
1307 close(fd);
1308 if (warnings)
1309 say(mpage->mlinks->file, "&fopen");
1310 return;
1311 }
1312
1313 /* Skip to first blank line. */
1314
1315 while (NULL != (line = fgetln(stream, &len)))
1316 if ('\n' == *line)
1317 break;
1318
1319 /*
1320 * Assume the first line that is not indented
1321 * is the first section header. Skip to it.
1322 */
1323
1324 while (NULL != (line = fgetln(stream, &len)))
1325 if ('\n' != *line && ' ' != *line)
1326 break;
1327
1328 /*
1329 * Read up until the next section into a buffer.
1330 * Strip the leading and trailing newline from each read line,
1331 * appending a trailing space.
1332 * Ignore empty (whitespace-only) lines.
1333 */
1334
1335 titlesz = 0;
1336 title = NULL;
1337
1338 while (NULL != (line = fgetln(stream, &len))) {
1339 if (' ' != *line || '\n' != line[len - 1])
1340 break;
1341 while (len > 0 && isspace((unsigned char)*line)) {
1342 line++;
1343 len--;
1344 }
1345 if (1 == len)
1346 continue;
1347 title = mandoc_realloc(title, titlesz + len);
1348 memcpy(title + titlesz, line, len);
1349 titlesz += len;
1350 title[titlesz - 1] = ' ';
1351 }
1352
1353 /*
1354 * If no page content can be found, or the input line
1355 * is already the next section header, or there is no
1356 * trailing newline, reuse the page title as the page
1357 * description.
1358 */
1359
1360 if (NULL == title || '\0' == *title) {
1361 if (warnings)
1362 say(mpage->mlinks->file,
1363 "Cannot find NAME section");
1364 fclose(stream);
1365 free(title);
1366 return;
1367 }
1368
1369 title = mandoc_realloc(title, titlesz + 1);
1370 title[titlesz] = '\0';
1371
1372 /*
1373 * Skip to the first dash.
1374 * Use the remaining line as the description (no more than 70
1375 * bytes).
1376 */
1377
1378 if (NULL != (p = strstr(title, "- "))) {
1379 for (p += 2; ' ' == *p || '\b' == *p; p++)
1380 /* Skip to next word. */ ;
1381 } else {
1382 if (warnings)
1383 say(mpage->mlinks->file,
1384 "No dash in title line");
1385 p = title;
1386 }
1387
1388 plen = strlen(p);
1389
1390 /* Strip backspace-encoding from line. */
1391
1392 while (NULL != (line = memchr(p, '\b', plen))) {
1393 len = line - p;
1394 if (0 == len) {
1395 memmove(line, line + 1, plen--);
1396 continue;
1397 }
1398 memmove(line - 1, line + 1, plen - len);
1399 plen -= 2;
1400 }
1401
1402 mpage->desc = mandoc_strdup(p);
1403 fclose(stream);
1404 free(title);
1405 }
1406
1407 /*
1408 * Put a type/word pair into the word database for this particular file.
1409 */
1410 static void
1411 putkey(const struct mpage *mpage, char *value, uint64_t type)
1412 {
1413 char *cp;
1414
1415 assert(NULL != value);
1416 if (TYPE_arch == type)
1417 for (cp = value; *cp; cp++)
1418 if (isupper((unsigned char)*cp))
1419 *cp = _tolower((unsigned char)*cp);
1420 putkeys(mpage, value, strlen(value), type);
1421 }
1422
1423 /*
1424 * Grok all nodes at or below a certain mdoc node into putkey().
1425 */
1426 static void
1427 putmdockey(const struct mpage *mpage,
1428 const struct mdoc_node *n, uint64_t m)
1429 {
1430
1431 for ( ; NULL != n; n = n->next) {
1432 if (NULL != n->child)
1433 putmdockey(mpage, n->child, m);
1434 if (MDOC_TEXT == n->type)
1435 putkey(mpage, n->string, m);
1436 }
1437 }
1438
1439 static void
1440 parse_man(struct mpage *mpage, const struct man_node *n)
1441 {
1442 const struct man_node *head, *body;
1443 char *start, *title;
1444 char byte;
1445 size_t sz;
1446
1447 if (NULL == n)
1448 return;
1449
1450 /*
1451 * We're only searching for one thing: the first text child in
1452 * the BODY of a NAME section. Since we don't keep track of
1453 * sections in -man, run some hoops to find out whether we're in
1454 * the correct section or not.
1455 */
1456
1457 if (MAN_BODY == n->type && MAN_SH == n->tok) {
1458 body = n;
1459 assert(body->parent);
1460 if (NULL != (head = body->parent->head) &&
1461 1 == head->nchild &&
1462 NULL != (head = (head->child)) &&
1463 MAN_TEXT == head->type &&
1464 0 == strcmp(head->string, "NAME") &&
1465 NULL != body->child) {
1466
1467 /*
1468 * Suck the entire NAME section into memory.
1469 * Yes, we might run away.
1470 * But too many manuals have big, spread-out
1471 * NAME sections over many lines.
1472 */
1473
1474 title = NULL;
1475 man_deroff(&title, body);
1476 if (NULL == title)
1477 return;
1478
1479 /*
1480 * Go through a special heuristic dance here.
1481 * Conventionally, one or more manual names are
1482 * comma-specified prior to a whitespace, then a
1483 * dash, then a description. Try to puzzle out
1484 * the name parts here.
1485 */
1486
1487 start = title;
1488 for ( ;; ) {
1489 sz = strcspn(start, " ,");
1490 if ('\0' == start[sz])
1491 break;
1492
1493 byte = start[sz];
1494 start[sz] = '\0';
1495
1496 /*
1497 * Assume a stray trailing comma in the
1498 * name list if a name begins with a dash.
1499 */
1500
1501 if ('-' == start[0] ||
1502 ('\\' == start[0] && '-' == start[1]))
1503 break;
1504
1505 putkey(mpage, start, NAME_TITLE);
1506
1507 if (' ' == byte) {
1508 start += sz + 1;
1509 break;
1510 }
1511
1512 assert(',' == byte);
1513 start += sz + 1;
1514 while (' ' == *start)
1515 start++;
1516 }
1517
1518 if (start == title) {
1519 putkey(mpage, start, NAME_TITLE);
1520 free(title);
1521 return;
1522 }
1523
1524 while (isspace((unsigned char)*start))
1525 start++;
1526
1527 if (0 == strncmp(start, "-", 1))
1528 start += 1;
1529 else if (0 == strncmp(start, "\\-\\-", 4))
1530 start += 4;
1531 else if (0 == strncmp(start, "\\-", 2))
1532 start += 2;
1533 else if (0 == strncmp(start, "\\(en", 4))
1534 start += 4;
1535 else if (0 == strncmp(start, "\\(em", 4))
1536 start += 4;
1537
1538 while (' ' == *start)
1539 start++;
1540
1541 mpage->desc = mandoc_strdup(start);
1542 free(title);
1543 return;
1544 }
1545 }
1546
1547 for (n = n->child; n; n = n->next) {
1548 if (NULL != mpage->desc)
1549 break;
1550 parse_man(mpage, n);
1551 }
1552 }
1553
1554 static void
1555 parse_mdoc(struct mpage *mpage, const struct mdoc_meta *meta,
1556 const struct mdoc_node *n)
1557 {
1558
1559 assert(NULL != n);
1560 for (n = n->child; NULL != n; n = n->next) {
1561 switch (n->type) {
1562 case MDOC_ELEM:
1563 /* FALLTHROUGH */
1564 case MDOC_BLOCK:
1565 /* FALLTHROUGH */
1566 case MDOC_HEAD:
1567 /* FALLTHROUGH */
1568 case MDOC_BODY:
1569 /* FALLTHROUGH */
1570 case MDOC_TAIL:
1571 if (NULL != mdocs[n->tok].fp)
1572 if (0 == (*mdocs[n->tok].fp)(mpage, meta, n))
1573 break;
1574 if (mdocs[n->tok].mask)
1575 putmdockey(mpage, n->child,
1576 mdocs[n->tok].mask);
1577 break;
1578 default:
1579 assert(MDOC_ROOT != n->type);
1580 continue;
1581 }
1582 if (NULL != n->child)
1583 parse_mdoc(mpage, meta, n);
1584 }
1585 }
1586
1587 static int
1588 parse_mdoc_Fd(struct mpage *mpage, const struct mdoc_meta *meta,
1589 const struct mdoc_node *n)
1590 {
1591 const char *start, *end;
1592 size_t sz;
1593
1594 if (SEC_SYNOPSIS != n->sec ||
1595 NULL == (n = n->child) ||
1596 MDOC_TEXT != n->type)
1597 return(0);
1598
1599 /*
1600 * Only consider those `Fd' macro fields that begin with an
1601 * "inclusion" token (versus, e.g., #define).
1602 */
1603
1604 if (strcmp("#include", n->string))
1605 return(0);
1606
1607 if (NULL == (n = n->next) || MDOC_TEXT != n->type)
1608 return(0);
1609
1610 /*
1611 * Strip away the enclosing angle brackets and make sure we're
1612 * not zero-length.
1613 */
1614
1615 start = n->string;
1616 if ('<' == *start || '"' == *start)
1617 start++;
1618
1619 if (0 == (sz = strlen(start)))
1620 return(0);
1621
1622 end = &start[(int)sz - 1];
1623 if ('>' == *end || '"' == *end)
1624 end--;
1625
1626 if (end > start)
1627 putkeys(mpage, start, end - start + 1, TYPE_In);
1628 return(0);
1629 }
1630
1631 static int
1632 parse_mdoc_Fn(struct mpage *mpage, const struct mdoc_meta *meta,
1633 const struct mdoc_node *n)
1634 {
1635 char *cp;
1636
1637 if (NULL == (n = n->child) || MDOC_TEXT != n->type)
1638 return(0);
1639
1640 /*
1641 * Parse: .Fn "struct type *name" "char *arg".
1642 * First strip away pointer symbol.
1643 * Then store the function name, then type.
1644 * Finally, store the arguments.
1645 */
1646
1647 if (NULL == (cp = strrchr(n->string, ' ')))
1648 cp = n->string;
1649
1650 while ('*' == *cp)
1651 cp++;
1652
1653 putkey(mpage, cp, TYPE_Fn);
1654
1655 if (n->string < cp)
1656 putkeys(mpage, n->string, cp - n->string, TYPE_Ft);
1657
1658 for (n = n->next; NULL != n; n = n->next)
1659 if (MDOC_TEXT == n->type)
1660 putkey(mpage, n->string, TYPE_Fa);
1661
1662 return(0);
1663 }
1664
1665 static int
1666 parse_mdoc_Xr(struct mpage *mpage, const struct mdoc_meta *meta,
1667 const struct mdoc_node *n)
1668 {
1669 char *cp;
1670
1671 if (NULL == (n = n->child))
1672 return(0);
1673
1674 if (NULL == n->next) {
1675 putkey(mpage, n->string, TYPE_Xr);
1676 return(0);
1677 }
1678
1679 mandoc_asprintf(&cp, "%s(%s)", n->string, n->next->string);
1680 putkey(mpage, cp, TYPE_Xr);
1681 free(cp);
1682 return(0);
1683 }
1684
1685 static int
1686 parse_mdoc_Nd(struct mpage *mpage, const struct mdoc_meta *meta,
1687 const struct mdoc_node *n)
1688 {
1689
1690 if (MDOC_BODY == n->type)
1691 mdoc_deroff(&mpage->desc, n);
1692 return(0);
1693 }
1694
1695 static int
1696 parse_mdoc_Nm(struct mpage *mpage, const struct mdoc_meta *meta,
1697 const struct mdoc_node *n)
1698 {
1699
1700 if (SEC_NAME == n->sec)
1701 putmdockey(mpage, n->child, NAME_TITLE);
1702 else if (SEC_SYNOPSIS == n->sec && MDOC_HEAD == n->type) {
1703 if (n->child == NULL)
1704 putkey(mpage, meta->name, NAME_SYN);
1705 else
1706 putmdockey(mpage, n->child, NAME_SYN);
1707 }
1708 return(0);
1709 }
1710
1711 static int
1712 parse_mdoc_Sh(struct mpage *mpage, const struct mdoc_meta *meta,
1713 const struct mdoc_node *n)
1714 {
1715
1716 return(SEC_CUSTOM == n->sec && MDOC_HEAD == n->type);
1717 }
1718
1719 static int
1720 parse_mdoc_head(struct mpage *mpage, const struct mdoc_meta *meta,
1721 const struct mdoc_node *n)
1722 {
1723
1724 return(MDOC_HEAD == n->type);
1725 }
1726
1727 static int
1728 parse_mdoc_body(struct mpage *mpage, const struct mdoc_meta *meta,
1729 const struct mdoc_node *n)
1730 {
1731
1732 return(MDOC_BODY == n->type);
1733 }
1734
1735 /*
1736 * Add a string to the hash table for the current manual.
1737 * Each string has a bitmask telling which macros it belongs to.
1738 * When we finish the manual, we'll dump the table.
1739 */
1740 static void
1741 putkeys(const struct mpage *mpage,
1742 const char *cp, size_t sz, uint64_t v)
1743 {
1744 struct ohash *htab;
1745 struct str *s;
1746 const char *end;
1747 unsigned int slot;
1748 int i;
1749
1750 if (0 == sz)
1751 return;
1752
1753 if (TYPE_Nm & v) {
1754 htab = &names;
1755 v &= name_mask;
1756 if (v & NAME_FIRST)
1757 name_mask &= ~NAME_FIRST;
1758 if (debug > 1)
1759 say(mpage->mlinks->file,
1760 "Adding name %*s", sz, cp);
1761 } else {
1762 htab = &strings;
1763 if (debug > 1)
1764 for (i = 0; i < mansearch_keymax; i++)
1765 if ((uint64_t)1 << i & v)
1766 say(mpage->mlinks->file,
1767 "Adding key %s=%*s",
1768 mansearch_keynames[i], sz, cp);
1769 }
1770
1771 end = cp + sz;
1772 slot = ohash_qlookupi(htab, cp, &end);
1773 s = ohash_find(htab, slot);
1774
1775 if (NULL != s && mpage == s->mpage) {
1776 s->mask |= v;
1777 return;
1778 } else if (NULL == s) {
1779 s = mandoc_calloc(1, sizeof(struct str) + sz + 1);
1780 memcpy(s->key, cp, sz);
1781 ohash_insert(htab, slot, s);
1782 }
1783 s->mpage = mpage;
1784 s->mask = v;
1785 }
1786
1787 /*
1788 * Take a Unicode codepoint and produce its UTF-8 encoding.
1789 * This isn't the best way to do this, but it works.
1790 * The magic numbers are from the UTF-8 packaging.
1791 * They're not as scary as they seem: read the UTF-8 spec for details.
1792 */
1793 static size_t
1794 utf8(unsigned int cp, char out[7])
1795 {
1796 size_t rc;
1797
1798 rc = 0;
1799 if (cp <= 0x0000007F) {
1800 rc = 1;
1801 out[0] = (char)cp;
1802 } else if (cp <= 0x000007FF) {
1803 rc = 2;
1804 out[0] = (cp >> 6 & 31) | 192;
1805 out[1] = (cp & 63) | 128;
1806 } else if (cp <= 0x0000FFFF) {
1807 rc = 3;
1808 out[0] = (cp >> 12 & 15) | 224;
1809 out[1] = (cp >> 6 & 63) | 128;
1810 out[2] = (cp & 63) | 128;
1811 } else if (cp <= 0x001FFFFF) {
1812 rc = 4;
1813 out[0] = (cp >> 18 & 7) | 240;
1814 out[1] = (cp >> 12 & 63) | 128;
1815 out[2] = (cp >> 6 & 63) | 128;
1816 out[3] = (cp & 63) | 128;
1817 } else if (cp <= 0x03FFFFFF) {
1818 rc = 5;
1819 out[0] = (cp >> 24 & 3) | 248;
1820 out[1] = (cp >> 18 & 63) | 128;
1821 out[2] = (cp >> 12 & 63) | 128;
1822 out[3] = (cp >> 6 & 63) | 128;
1823 out[4] = (cp & 63) | 128;
1824 } else if (cp <= 0x7FFFFFFF) {
1825 rc = 6;
1826 out[0] = (cp >> 30 & 1) | 252;
1827 out[1] = (cp >> 24 & 63) | 128;
1828 out[2] = (cp >> 18 & 63) | 128;
1829 out[3] = (cp >> 12 & 63) | 128;
1830 out[4] = (cp >> 6 & 63) | 128;
1831 out[5] = (cp & 63) | 128;
1832 } else
1833 return(0);
1834
1835 out[rc] = '\0';
1836 return(rc);
1837 }
1838
1839 /*
1840 * Store the rendered version of a key, or alias the pointer
1841 * if the key contains no escape sequences.
1842 */
1843 static void
1844 render_key(struct mchars *mc, struct str *key)
1845 {
1846 size_t sz, bsz, pos;
1847 char utfbuf[7], res[6];
1848 char *buf;
1849 const char *seq, *cpp, *val;
1850 int len, u;
1851 enum mandoc_esc esc;
1852
1853 assert(NULL == key->rendered);
1854
1855 res[0] = '\\';
1856 res[1] = '\t';
1857 res[2] = ASCII_NBRSP;
1858 res[3] = ASCII_HYPH;
1859 res[4] = ASCII_BREAK;
1860 res[5] = '\0';
1861
1862 val = key->key;
1863 bsz = strlen(val);
1864
1865 /*
1866 * Pre-check: if we have no stop-characters, then set the
1867 * pointer as ourselvse and get out of here.
1868 */
1869 if (strcspn(val, res) == bsz) {
1870 key->rendered = key->key;
1871 return;
1872 }
1873
1874 /* Pre-allocate by the length of the input */
1875
1876 buf = mandoc_malloc(++bsz);
1877 pos = 0;
1878
1879 while ('\0' != *val) {
1880 /*
1881 * Halt on the first escape sequence.
1882 * This also halts on the end of string, in which case
1883 * we just copy, fallthrough, and exit the loop.
1884 */
1885 if ((sz = strcspn(val, res)) > 0) {
1886 memcpy(&buf[pos], val, sz);
1887 pos += sz;
1888 val += sz;
1889 }
1890
1891 switch (*val) {
1892 case ASCII_HYPH:
1893 buf[pos++] = '-';
1894 val++;
1895 continue;
1896 case '\t':
1897 /* FALLTHROUGH */
1898 case ASCII_NBRSP:
1899 buf[pos++] = ' ';
1900 val++;
1901 /* FALLTHROUGH */
1902 case ASCII_BREAK:
1903 continue;
1904 default:
1905 break;
1906 }
1907 if ('\\' != *val)
1908 break;
1909
1910 /* Read past the slash. */
1911
1912 val++;
1913
1914 /*
1915 * Parse the escape sequence and see if it's a
1916 * predefined character or special character.
1917 */
1918
1919 esc = mandoc_escape((const char **)&val,
1920 &seq, &len);
1921 if (ESCAPE_ERROR == esc)
1922 break;
1923 if (ESCAPE_SPECIAL != esc)
1924 continue;
1925
1926 /*
1927 * Render the special character
1928 * as either UTF-8 or ASCII.
1929 */
1930
1931 if (write_utf8) {
1932 if ((u = mchars_spec2cp(mc, seq, len)) <= 0)
1933 continue;
1934 cpp = utfbuf;
1935 if (0 == (sz = utf8(u, utfbuf)))
1936 continue;
1937 sz = strlen(cpp);
1938 } else {
1939 cpp = mchars_spec2str(mc, seq, len, &sz);
1940 if (NULL == cpp)
1941 continue;
1942 if (ASCII_NBRSP == *cpp) {
1943 cpp = " ";
1944 sz = 1;
1945 }
1946 }
1947
1948 /* Copy the rendered glyph into the stream. */
1949
1950 bsz += sz;
1951 buf = mandoc_realloc(buf, bsz);
1952 memcpy(&buf[pos], cpp, sz);
1953 pos += sz;
1954 }
1955
1956 buf[pos] = '\0';
1957 key->rendered = buf;
1958 }
1959
1960 static void
1961 dbadd_mlink(const struct mlink *mlink)
1962 {
1963 size_t i;
1964
1965 i = 1;
1966 SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->dsec);
1967 SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->arch);
1968 SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->name);
1969 SQL_BIND_INT64(stmts[STMT_INSERT_LINK], i, mlink->mpage->pageid);
1970 SQL_STEP(stmts[STMT_INSERT_LINK]);
1971 sqlite3_reset(stmts[STMT_INSERT_LINK]);
1972 }
1973
1974 static void
1975 dbadd_mlink_name(const struct mlink *mlink)
1976 {
1977 size_t i;
1978
1979 dbadd_mlink(mlink);
1980
1981 i = 1;
1982 SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, NAME_FILE & NAME_MASK);
1983 SQL_BIND_TEXT(stmts[STMT_INSERT_NAME], i, mlink->name);
1984 SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, mlink->mpage->pageid);
1985 SQL_STEP(stmts[STMT_INSERT_NAME]);
1986 sqlite3_reset(stmts[STMT_INSERT_NAME]);
1987 }
1988
1989 /*
1990 * Flush the current page's terms (and their bits) into the database.
1991 * Wrap the entire set of additions in a transaction to make sqlite be a
1992 * little faster.
1993 * Also, handle escape sequences at the last possible moment.
1994 */
1995 static void
1996 dbadd(struct mpage *mpage, struct mchars *mc)
1997 {
1998 struct mlink *mlink;
1999 struct str *key;
2000 size_t i;
2001 unsigned int slot;
2002
2003 mlink = mpage->mlinks;
2004
2005 if (nodb) {
2006 for (key = ohash_first(&names, &slot); NULL != key;
2007 key = ohash_next(&names, &slot)) {
2008 if (key->rendered != key->key)
2009 free(key->rendered);
2010 free(key);
2011 }
2012 for (key = ohash_first(&strings, &slot); NULL != key;
2013 key = ohash_next(&strings, &slot)) {
2014 if (key->rendered != key->key)
2015 free(key->rendered);
2016 free(key);
2017 }
2018 if (0 == debug)
2019 return;
2020 while (NULL != mlink) {
2021 fputs(mlink->name, stdout);
2022 if (NULL == mlink->next ||
2023 strcmp(mlink->dsec, mlink->next->dsec) ||
2024 strcmp(mlink->fsec, mlink->next->fsec) ||
2025 strcmp(mlink->arch, mlink->next->arch)) {
2026 putchar('(');
2027 if ('\0' == *mlink->dsec)
2028 fputs(mlink->fsec, stdout);
2029 else
2030 fputs(mlink->dsec, stdout);
2031 if ('\0' != *mlink->arch)
2032 printf("/%s", mlink->arch);
2033 putchar(')');
2034 }
2035 mlink = mlink->next;
2036 if (NULL != mlink)
2037 fputs(", ", stdout);
2038 }
2039 printf(" - %s\n", mpage->desc);
2040 return;
2041 }
2042
2043 if (debug)
2044 say(mlink->file, "Adding to database");
2045
2046 i = strlen(mpage->desc) + 1;
2047 key = mandoc_calloc(1, sizeof(struct str) + i);
2048 memcpy(key->key, mpage->desc, i);
2049 render_key(mc, key);
2050
2051 i = 1;
2052 SQL_BIND_TEXT(stmts[STMT_INSERT_PAGE], i, key->rendered);
2053 SQL_BIND_INT(stmts[STMT_INSERT_PAGE], i, mpage->form);
2054 SQL_STEP(stmts[STMT_INSERT_PAGE]);
2055 mpage->pageid = sqlite3_last_insert_rowid(db);
2056 sqlite3_reset(stmts[STMT_INSERT_PAGE]);
2057
2058 if (key->rendered != key->key)
2059 free(key->rendered);
2060 free(key);
2061
2062 while (NULL != mlink) {
2063 dbadd_mlink(mlink);
2064 mlink = mlink->next;
2065 }
2066 mlink = mpage->mlinks;
2067
2068 for (key = ohash_first(&names, &slot); NULL != key;
2069 key = ohash_next(&names, &slot)) {
2070 assert(key->mpage == mpage);
2071 if (NULL == key->rendered)
2072 render_key(mc, key);
2073 i = 1;
2074 SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, key->mask);
2075 SQL_BIND_TEXT(stmts[STMT_INSERT_NAME], i, key->rendered);
2076 SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, mpage->pageid);
2077 SQL_STEP(stmts[STMT_INSERT_NAME]);
2078 sqlite3_reset(stmts[STMT_INSERT_NAME]);
2079 if (key->rendered != key->key)
2080 free(key->rendered);
2081 free(key);
2082 }
2083 for (key = ohash_first(&strings, &slot); NULL != key;
2084 key = ohash_next(&strings, &slot)) {
2085 assert(key->mpage == mpage);
2086 if (NULL == key->rendered)
2087 render_key(mc, key);
2088 i = 1;
2089 SQL_BIND_INT64(stmts[STMT_INSERT_KEY], i, key->mask);
2090 SQL_BIND_TEXT(stmts[STMT_INSERT_KEY], i, key->rendered);
2091 SQL_BIND_INT64(stmts[STMT_INSERT_KEY], i, mpage->pageid);
2092 SQL_STEP(stmts[STMT_INSERT_KEY]);
2093 sqlite3_reset(stmts[STMT_INSERT_KEY]);
2094 if (key->rendered != key->key)
2095 free(key->rendered);
2096 free(key);
2097 }
2098 }
2099
2100 static void
2101 dbprune(void)
2102 {
2103 struct mpage *mpage;
2104 struct mlink *mlink;
2105 size_t i;
2106 unsigned int slot;
2107
2108 if (0 == nodb)
2109 SQL_EXEC("BEGIN TRANSACTION");
2110
2111 for (mpage = ohash_first(&mpages, &slot); NULL != mpage;
2112 mpage = ohash_next(&mpages, &slot)) {
2113 mlink = mpage->mlinks;
2114 if (debug)
2115 say(mlink->file, "Deleting from database");
2116 if (nodb)
2117 continue;
2118 for ( ; NULL != mlink; mlink = mlink->next) {
2119 i = 1;
2120 SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE],
2121 i, mlink->dsec);
2122 SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE],
2123 i, mlink->arch);
2124 SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE],
2125 i, mlink->name);
2126 SQL_STEP(stmts[STMT_DELETE_PAGE]);
2127 sqlite3_reset(stmts[STMT_DELETE_PAGE]);
2128 }
2129 }
2130
2131 if (0 == nodb)
2132 SQL_EXEC("END TRANSACTION");
2133 }
2134
2135 /*
2136 * Close an existing database and its prepared statements.
2137 * If "real" is not set, rename the temporary file into the real one.
2138 */
2139 static void
2140 dbclose(int real)
2141 {
2142 size_t i;
2143 int status;
2144 pid_t child;
2145
2146 if (nodb)
2147 return;
2148
2149 for (i = 0; i < STMT__MAX; i++) {
2150 sqlite3_finalize(stmts[i]);
2151 stmts[i] = NULL;
2152 }
2153
2154 sqlite3_close(db);
2155 db = NULL;
2156
2157 if (real)
2158 return;
2159
2160 if ('\0' == *tempfilename) {
2161 if (-1 == rename(MANDOC_DB "~", MANDOC_DB)) {
2162 exitcode = (int)MANDOCLEVEL_SYSERR;
2163 say(MANDOC_DB, "&rename");
2164 }
2165 return;
2166 }
2167
2168 switch (child = fork()) {
2169 case -1:
2170 exitcode = (int)MANDOCLEVEL_SYSERR;
2171 say("", "&fork cmp");
2172 return;
2173 case 0:
2174 execlp("cmp", "cmp", "-s",
2175 tempfilename, MANDOC_DB, NULL);
2176 say("", "&exec cmp");
2177 exit(0);
2178 default:
2179 break;
2180 }
2181 if (-1 == waitpid(child, &status, 0)) {
2182 exitcode = (int)MANDOCLEVEL_SYSERR;
2183 say("", "&wait cmp");
2184 } else if (WIFSIGNALED(status)) {
2185 exitcode = (int)MANDOCLEVEL_SYSERR;
2186 say("", "cmp died from signal %d", WTERMSIG(status));
2187 } else if (WEXITSTATUS(status)) {
2188 exitcode = (int)MANDOCLEVEL_SYSERR;
2189 say(MANDOC_DB,
2190 "Data changed, but cannot replace database");
2191 }
2192
2193 *strrchr(tempfilename, '/') = '\0';
2194 switch (child = fork()) {
2195 case -1:
2196 exitcode = (int)MANDOCLEVEL_SYSERR;
2197 say("", "&fork rm");
2198 return;
2199 case 0:
2200 execlp("rm", "rm", "-rf", tempfilename, NULL);
2201 say("", "&exec rm");
2202 exit((int)MANDOCLEVEL_SYSERR);
2203 default:
2204 break;
2205 }
2206 if (-1 == waitpid(child, &status, 0)) {
2207 exitcode = (int)MANDOCLEVEL_SYSERR;
2208 say("", "&wait rm");
2209 } else if (WIFSIGNALED(status) || WEXITSTATUS(status)) {
2210 exitcode = (int)MANDOCLEVEL_SYSERR;
2211 say("", "%s: Cannot remove temporary directory",
2212 tempfilename);
2213 }
2214 }
2215
2216 /*
2217 * This is straightforward stuff.
2218 * Open a database connection to a "temporary" database, then open a set
2219 * of prepared statements we'll use over and over again.
2220 * If "real" is set, we use the existing database; if not, we truncate a
2221 * temporary one.
2222 * Must be matched by dbclose().
2223 */
2224 static int
2225 dbopen(int real)
2226 {
2227 const char *sql;
2228 int rc, ofl;
2229
2230 if (nodb)
2231 return(1);
2232
2233 *tempfilename = '\0';
2234 ofl = SQLITE_OPEN_READWRITE;
2235
2236 if (real) {
2237 rc = sqlite3_open_v2(MANDOC_DB, &db, ofl, NULL);
2238 if (SQLITE_OK != rc) {
2239 exitcode = (int)MANDOCLEVEL_SYSERR;
2240 if (SQLITE_CANTOPEN != rc)
2241 say(MANDOC_DB, "%s", sqlite3_errstr(rc));
2242 return(0);
2243 }
2244 goto prepare_statements;
2245 }
2246
2247 ofl |= SQLITE_OPEN_CREATE | SQLITE_OPEN_EXCLUSIVE;
2248
2249 remove(MANDOC_DB "~");
2250 rc = sqlite3_open_v2(MANDOC_DB "~", &db, ofl, NULL);
2251 if (SQLITE_OK == rc)
2252 goto create_tables;
2253 if (MPARSE_QUICK & mparse_options) {
2254 exitcode = (int)MANDOCLEVEL_SYSERR;
2255 say(MANDOC_DB "~", "%s", sqlite3_errstr(rc));
2256 return(0);
2257 }
2258
2259 (void)strlcpy(tempfilename, "/tmp/mandocdb.XXXXXX",
2260 sizeof(tempfilename));
2261 if (NULL == mkdtemp(tempfilename)) {
2262 exitcode = (int)MANDOCLEVEL_SYSERR;
2263 say("", "&%s", tempfilename);
2264 return(0);
2265 }
2266 (void)strlcat(tempfilename, "/" MANDOC_DB,
2267 sizeof(tempfilename));
2268 rc = sqlite3_open_v2(tempfilename, &db, ofl, NULL);
2269 if (SQLITE_OK != rc) {
2270 exitcode = (int)MANDOCLEVEL_SYSERR;
2271 say("", "%s: %s", tempfilename, sqlite3_errstr(rc));
2272 return(0);
2273 }
2274
2275 create_tables:
2276 sql = "CREATE TABLE \"mpages\" (\n"
2277 " \"desc\" TEXT NOT NULL,\n"
2278 " \"form\" INTEGER NOT NULL,\n"
2279 " \"pageid\" INTEGER PRIMARY KEY AUTOINCREMENT NOT NULL\n"
2280 ");\n"
2281 "\n"
2282 "CREATE TABLE \"mlinks\" (\n"
2283 " \"sec\" TEXT NOT NULL,\n"
2284 " \"arch\" TEXT NOT NULL,\n"
2285 " \"name\" TEXT NOT NULL,\n"
2286 " \"pageid\" INTEGER NOT NULL REFERENCES mpages(pageid) "
2287 "ON DELETE CASCADE\n"
2288 ");\n"
2289 "CREATE INDEX mlinks_pageid_idx ON mlinks (pageid);\n"
2290 "\n"
2291 "CREATE TABLE \"names\" (\n"
2292 " \"bits\" INTEGER NOT NULL,\n"
2293 " \"name\" TEXT NOT NULL,\n"
2294 " \"pageid\" INTEGER NOT NULL REFERENCES mpages(pageid) "
2295 "ON DELETE CASCADE\n"
2296 ");\n"
2297 "\n"
2298 "CREATE TABLE \"keys\" (\n"
2299 " \"bits\" INTEGER NOT NULL,\n"
2300 " \"key\" TEXT NOT NULL,\n"
2301 " \"pageid\" INTEGER NOT NULL REFERENCES mpages(pageid) "
2302 "ON DELETE CASCADE\n"
2303 ");\n"
2304 "CREATE INDEX keys_pageid_idx ON keys (pageid);\n";
2305
2306 if (SQLITE_OK != sqlite3_exec(db, sql, NULL, NULL, NULL)) {
2307 exitcode = (int)MANDOCLEVEL_SYSERR;
2308 say(MANDOC_DB, "%s", sqlite3_errmsg(db));
2309 sqlite3_close(db);
2310 return(0);
2311 }
2312
2313 prepare_statements:
2314 if (SQLITE_OK != sqlite3_exec(db,
2315 "PRAGMA foreign_keys = ON", NULL, NULL, NULL)) {
2316 exitcode = (int)MANDOCLEVEL_SYSERR;
2317 say(MANDOC_DB, "PRAGMA foreign_keys: %s",
2318 sqlite3_errmsg(db));
2319 sqlite3_close(db);
2320 return(0);
2321 }
2322
2323 sql = "DELETE FROM mpages WHERE pageid IN "
2324 "(SELECT pageid FROM mlinks WHERE "
2325 "sec=? AND arch=? AND name=?)";
2326 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_DELETE_PAGE], NULL);
2327 sql = "INSERT INTO mpages "
2328 "(desc,form) VALUES (?,?)";
2329 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_PAGE], NULL);
2330 sql = "INSERT INTO mlinks "
2331 "(sec,arch,name,pageid) VALUES (?,?,?,?)";
2332 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_LINK], NULL);
2333 sql = "INSERT INTO names "
2334 "(bits,name,pageid) VALUES (?,?,?)";
2335 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_NAME], NULL);
2336 sql = "INSERT INTO keys "
2337 "(bits,key,pageid) VALUES (?,?,?)";
2338 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_KEY], NULL);
2339
2340 #ifndef __APPLE__
2341 /*
2342 * When opening a new database, we can turn off
2343 * synchronous mode for much better performance.
2344 */
2345
2346 if (real && SQLITE_OK != sqlite3_exec(db,
2347 "PRAGMA synchronous = OFF", NULL, NULL, NULL)) {
2348 exitcode = (int)MANDOCLEVEL_SYSERR;
2349 say(MANDOC_DB, "PRAGMA synchronous: %s",
2350 sqlite3_errmsg(db));
2351 sqlite3_close(db);
2352 return(0);
2353 }
2354 #endif
2355
2356 return(1);
2357 }
2358
2359 static void *
2360 hash_calloc(size_t n, size_t sz, void *arg)
2361 {
2362
2363 return(mandoc_calloc(n, sz));
2364 }
2365
2366 static void *
2367 hash_alloc(size_t sz, void *arg)
2368 {
2369
2370 return(mandoc_malloc(sz));
2371 }
2372
2373 static void
2374 hash_free(void *p, void *arg)
2375 {
2376
2377 free(p);
2378 }
2379
2380 static int
2381 set_basedir(const char *targetdir, int report_baddir)
2382 {
2383 static char startdir[PATH_MAX];
2384 static int getcwd_status; /* 1 = ok, 2 = failure */
2385 static int chdir_status; /* 1 = changed directory */
2386 char *cp;
2387
2388 /*
2389 * Remember the original working directory, if possible.
2390 * This will be needed if the second or a later directory
2391 * on the command line is given as a relative path.
2392 * Do not error out if the current directory is not
2393 * searchable: Maybe it won't be needed after all.
2394 */
2395 if (0 == getcwd_status) {
2396 if (NULL == getcwd(startdir, sizeof(startdir))) {
2397 getcwd_status = 2;
2398 (void)strlcpy(startdir, strerror(errno),
2399 sizeof(startdir));
2400 } else
2401 getcwd_status = 1;
2402 }
2403
2404 /*
2405 * We are leaving the old base directory.
2406 * Do not use it any longer, not even for messages.
2407 */
2408 *basedir = '\0';
2409
2410 /*
2411 * If and only if the directory was changed earlier and
2412 * the next directory to process is given as a relative path,
2413 * first go back, or bail out if that is impossible.
2414 */
2415 if (chdir_status && '/' != *targetdir) {
2416 if (2 == getcwd_status) {
2417 exitcode = (int)MANDOCLEVEL_SYSERR;
2418 say("", "getcwd: %s", startdir);
2419 return(0);
2420 }
2421 if (-1 == chdir(startdir)) {
2422 exitcode = (int)MANDOCLEVEL_SYSERR;
2423 say("", "&chdir %s", startdir);
2424 return(0);
2425 }
2426 }
2427
2428 /*
2429 * Always resolve basedir to the canonicalized absolute
2430 * pathname and append a trailing slash, such that
2431 * we can reliably check whether files are inside.
2432 */
2433 if (NULL == realpath(targetdir, basedir)) {
2434 if (report_baddir || errno != ENOENT) {
2435 exitcode = (int)MANDOCLEVEL_BADARG;
2436 say("", "&%s: realpath", targetdir);
2437 }
2438 return(0);
2439 } else if (-1 == chdir(basedir)) {
2440 if (report_baddir || errno != ENOENT) {
2441 exitcode = (int)MANDOCLEVEL_BADARG;
2442 say("", "&chdir");
2443 }
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