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


fa23ad0e1e0211d825b24ab1b6eee4f43b9cd8a8
[mandoc.git] / mandocdb.c
1 /* $Id: mandocdb.c,v 1.217 2016/07/09 15:24:19 schwarze Exp $ */
2 /*
3 * Copyright (c) 2011, 2012 Kristaps Dzonsons <kristaps@bsd.lv>
4 * Copyright (c) 2011-2016 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 AUTHORS DISCLAIM ALL WARRANTIES
11 * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
12 * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHORS 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 #if HAVE_ERR
27 #include <err.h>
28 #endif
29 #include <errno.h>
30 #include <fcntl.h>
31 #if HAVE_FTS
32 #include <fts.h>
33 #else
34 #include "compat_fts.h"
35 #endif
36 #include <limits.h>
37 #include <stddef.h>
38 #include <stdio.h>
39 #include <stdint.h>
40 #include <stdlib.h>
41 #include <string.h>
42 #include <unistd.h>
43
44 #include <sqlite3.h>
45
46 #include "mandoc_aux.h"
47 #include "mandoc_ohash.h"
48 #include "mandoc.h"
49 #include "roff.h"
50 #include "mdoc.h"
51 #include "man.h"
52 #include "manconf.h"
53 #include "mansearch.h"
54
55 extern int mansearch_keymax;
56 extern const char *const mansearch_keynames[];
57
58 #define SQL_EXEC(_v) \
59 if (SQLITE_OK != sqlite3_exec(db, (_v), NULL, NULL, NULL)) \
60 say("", "%s: %s", (_v), sqlite3_errmsg(db))
61 #define SQL_BIND_TEXT(_s, _i, _v) \
62 if (SQLITE_OK != sqlite3_bind_text \
63 ((_s), (_i)++, (_v), -1, SQLITE_STATIC)) \
64 say(mlink->file, "%s", sqlite3_errmsg(db))
65 #define SQL_BIND_INT(_s, _i, _v) \
66 if (SQLITE_OK != sqlite3_bind_int \
67 ((_s), (_i)++, (_v))) \
68 say(mlink->file, "%s", sqlite3_errmsg(db))
69 #define SQL_BIND_INT64(_s, _i, _v) \
70 if (SQLITE_OK != sqlite3_bind_int64 \
71 ((_s), (_i)++, (_v))) \
72 say(mlink->file, "%s", sqlite3_errmsg(db))
73 #define SQL_STEP(_s) \
74 if (SQLITE_DONE != sqlite3_step((_s))) \
75 say(mlink->file, "%s", sqlite3_errmsg(db))
76
77 enum op {
78 OP_DEFAULT = 0, /* new dbs from dir list or default config */
79 OP_CONFFILE, /* new databases from custom config file */
80 OP_UPDATE, /* delete/add entries in existing database */
81 OP_DELETE, /* delete entries from existing database */
82 OP_TEST /* change no databases, report potential problems */
83 };
84
85 struct str {
86 const struct mpage *mpage; /* if set, the owning parse */
87 uint64_t mask; /* bitmask in sequence */
88 char key[]; /* rendered text */
89 };
90
91 struct inodev {
92 ino_t st_ino;
93 dev_t st_dev;
94 };
95
96 struct mpage {
97 struct inodev inodev; /* used for hashing routine */
98 int64_t pageid; /* pageid in mpages SQL table */
99 char *sec; /* section from file content */
100 char *arch; /* architecture from file content */
101 char *title; /* title from file content */
102 char *desc; /* description from file content */
103 struct mlink *mlinks; /* singly linked list */
104 int form; /* format from file content */
105 int name_head_done;
106 };
107
108 struct mlink {
109 char file[PATH_MAX]; /* filename rel. to manpath */
110 char *dsec; /* section from directory */
111 char *arch; /* architecture from directory */
112 char *name; /* name from file name (not empty) */
113 char *fsec; /* section from file name suffix */
114 struct mlink *next; /* singly linked list */
115 struct mpage *mpage; /* parent */
116 int dform; /* format from directory */
117 int fform; /* format from file name suffix */
118 int gzip; /* filename has a .gz suffix */
119 };
120
121 enum stmt {
122 STMT_DELETE_PAGE = 0, /* delete mpage */
123 STMT_INSERT_PAGE, /* insert mpage */
124 STMT_INSERT_LINK, /* insert mlink */
125 STMT_INSERT_NAME, /* insert name */
126 STMT_SELECT_NAME, /* retrieve existing name flags */
127 STMT_INSERT_KEY, /* insert parsed key */
128 STMT__MAX
129 };
130
131 typedef int (*mdoc_fp)(struct mpage *, const struct roff_meta *,
132 const struct roff_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 *);
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 mlink_add(struct mlink *, const struct stat *);
147 static void mlink_check(struct mpage *, struct mlink *);
148 static void mlink_free(struct mlink *);
149 static void mlinks_undupe(struct mpage *);
150 static void mpages_free(void);
151 static void mpages_merge(struct mparse *);
152 static void names_check(void);
153 static void parse_cat(struct mpage *, int);
154 static void parse_man(struct mpage *, const struct roff_meta *,
155 const struct roff_node *);
156 static void parse_mdoc(struct mpage *, const struct roff_meta *,
157 const struct roff_node *);
158 static int parse_mdoc_head(struct mpage *, const struct roff_meta *,
159 const struct roff_node *);
160 static int parse_mdoc_Fd(struct mpage *, const struct roff_meta *,
161 const struct roff_node *);
162 static void parse_mdoc_fname(struct mpage *, const struct roff_node *);
163 static int parse_mdoc_Fn(struct mpage *, const struct roff_meta *,
164 const struct roff_node *);
165 static int parse_mdoc_Fo(struct mpage *, const struct roff_meta *,
166 const struct roff_node *);
167 static int parse_mdoc_Nd(struct mpage *, const struct roff_meta *,
168 const struct roff_node *);
169 static int parse_mdoc_Nm(struct mpage *, const struct roff_meta *,
170 const struct roff_node *);
171 static int parse_mdoc_Sh(struct mpage *, const struct roff_meta *,
172 const struct roff_node *);
173 static int parse_mdoc_Va(struct mpage *, const struct roff_meta *,
174 const struct roff_node *);
175 static int parse_mdoc_Xr(struct mpage *, const struct roff_meta *,
176 const struct roff_node *);
177 static void putkey(const struct mpage *, char *, uint64_t);
178 static void putkeys(const struct mpage *, char *, size_t, uint64_t);
179 static void putmdockey(const struct mpage *,
180 const struct roff_node *, uint64_t);
181 static int render_string(char **, size_t *);
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 int nodb; /* no database changes */
189 static int mparse_options; /* abort the parse early */
190 static int use_all; /* use all found files */
191 static int debug; /* print what we're doing */
192 static int warnings; /* warn about crap */
193 static int write_utf8; /* write UTF-8 output; else ASCII */
194 static int exitcode; /* to be returned by main */
195 static enum op op; /* operational mode */
196 static char basedir[PATH_MAX]; /* current base directory */
197 static struct ohash mpages; /* table of distinct manual pages */
198 static struct ohash mlinks; /* table of directory entries */
199 static struct ohash names; /* table of all names */
200 static struct ohash strings; /* table of all strings */
201 static sqlite3 *db = NULL; /* current database */
202 static sqlite3_stmt *stmts[STMT__MAX]; /* current statements */
203 static uint64_t name_mask;
204
205 static const struct mdoc_handler mdocs[MDOC_MAX] = {
206 { NULL, 0 }, /* Ap */
207 { NULL, 0 }, /* Dd */
208 { NULL, 0 }, /* Dt */
209 { NULL, 0 }, /* Os */
210 { parse_mdoc_Sh, TYPE_Sh }, /* Sh */
211 { parse_mdoc_head, TYPE_Ss }, /* Ss */
212 { NULL, 0 }, /* Pp */
213 { NULL, 0 }, /* D1 */
214 { NULL, 0 }, /* Dl */
215 { NULL, 0 }, /* Bd */
216 { NULL, 0 }, /* Ed */
217 { NULL, 0 }, /* Bl */
218 { NULL, 0 }, /* El */
219 { NULL, 0 }, /* It */
220 { NULL, 0 }, /* Ad */
221 { NULL, TYPE_An }, /* An */
222 { NULL, TYPE_Ar }, /* Ar */
223 { NULL, TYPE_Cd }, /* Cd */
224 { NULL, TYPE_Cm }, /* Cm */
225 { NULL, TYPE_Dv }, /* Dv */
226 { NULL, TYPE_Er }, /* Er */
227 { NULL, TYPE_Ev }, /* Ev */
228 { NULL, 0 }, /* Ex */
229 { NULL, TYPE_Fa }, /* Fa */
230 { parse_mdoc_Fd, 0 }, /* Fd */
231 { NULL, TYPE_Fl }, /* Fl */
232 { parse_mdoc_Fn, 0 }, /* Fn */
233 { NULL, TYPE_Ft }, /* Ft */
234 { NULL, TYPE_Ic }, /* Ic */
235 { NULL, TYPE_In }, /* In */
236 { NULL, TYPE_Li }, /* Li */
237 { parse_mdoc_Nd, 0 }, /* Nd */
238 { parse_mdoc_Nm, 0 }, /* Nm */
239 { NULL, 0 }, /* Op */
240 { NULL, 0 }, /* Ot */
241 { NULL, TYPE_Pa }, /* Pa */
242 { NULL, 0 }, /* Rv */
243 { NULL, TYPE_St }, /* St */
244 { parse_mdoc_Va, TYPE_Va }, /* Va */
245 { parse_mdoc_Va, TYPE_Vt }, /* Vt */
246 { parse_mdoc_Xr, 0 }, /* Xr */
247 { NULL, 0 }, /* %A */
248 { NULL, 0 }, /* %B */
249 { NULL, 0 }, /* %D */
250 { NULL, 0 }, /* %I */
251 { NULL, 0 }, /* %J */
252 { NULL, 0 }, /* %N */
253 { NULL, 0 }, /* %O */
254 { NULL, 0 }, /* %P */
255 { NULL, 0 }, /* %R */
256 { NULL, 0 }, /* %T */
257 { NULL, 0 }, /* %V */
258 { NULL, 0 }, /* Ac */
259 { NULL, 0 }, /* Ao */
260 { NULL, 0 }, /* Aq */
261 { NULL, TYPE_At }, /* At */
262 { NULL, 0 }, /* Bc */
263 { NULL, 0 }, /* Bf */
264 { NULL, 0 }, /* Bo */
265 { NULL, 0 }, /* Bq */
266 { NULL, TYPE_Bsx }, /* Bsx */
267 { NULL, TYPE_Bx }, /* Bx */
268 { NULL, 0 }, /* Db */
269 { NULL, 0 }, /* Dc */
270 { NULL, 0 }, /* Do */
271 { NULL, 0 }, /* Dq */
272 { NULL, 0 }, /* Ec */
273 { NULL, 0 }, /* Ef */
274 { NULL, TYPE_Em }, /* Em */
275 { NULL, 0 }, /* Eo */
276 { NULL, TYPE_Fx }, /* Fx */
277 { NULL, TYPE_Ms }, /* Ms */
278 { NULL, 0 }, /* No */
279 { NULL, 0 }, /* Ns */
280 { NULL, TYPE_Nx }, /* Nx */
281 { NULL, TYPE_Ox }, /* Ox */
282 { NULL, 0 }, /* Pc */
283 { NULL, 0 }, /* Pf */
284 { NULL, 0 }, /* Po */
285 { NULL, 0 }, /* Pq */
286 { NULL, 0 }, /* Qc */
287 { NULL, 0 }, /* Ql */
288 { NULL, 0 }, /* Qo */
289 { NULL, 0 }, /* Qq */
290 { NULL, 0 }, /* Re */
291 { NULL, 0 }, /* Rs */
292 { NULL, 0 }, /* Sc */
293 { NULL, 0 }, /* So */
294 { NULL, 0 }, /* Sq */
295 { NULL, 0 }, /* Sm */
296 { NULL, 0 }, /* Sx */
297 { NULL, TYPE_Sy }, /* Sy */
298 { NULL, TYPE_Tn }, /* Tn */
299 { NULL, 0 }, /* Ux */
300 { NULL, 0 }, /* Xc */
301 { NULL, 0 }, /* Xo */
302 { parse_mdoc_Fo, 0 }, /* Fo */
303 { NULL, 0 }, /* Fc */
304 { NULL, 0 }, /* Oo */
305 { NULL, 0 }, /* Oc */
306 { NULL, 0 }, /* Bk */
307 { NULL, 0 }, /* Ek */
308 { NULL, 0 }, /* Bt */
309 { NULL, 0 }, /* Hf */
310 { NULL, 0 }, /* Fr */
311 { NULL, 0 }, /* Ud */
312 { NULL, TYPE_Lb }, /* Lb */
313 { NULL, 0 }, /* Lp */
314 { NULL, TYPE_Lk }, /* Lk */
315 { NULL, TYPE_Mt }, /* Mt */
316 { NULL, 0 }, /* Brq */
317 { NULL, 0 }, /* Bro */
318 { NULL, 0 }, /* Brc */
319 { NULL, 0 }, /* %C */
320 { NULL, 0 }, /* Es */
321 { NULL, 0 }, /* En */
322 { NULL, TYPE_Dx }, /* Dx */
323 { NULL, 0 }, /* %Q */
324 { NULL, 0 }, /* br */
325 { NULL, 0 }, /* sp */
326 { NULL, 0 }, /* %U */
327 { NULL, 0 }, /* Ta */
328 { NULL, 0 }, /* ll */
329 };
330
331
332 int
333 mandocdb(int argc, char *argv[])
334 {
335 struct manconf conf;
336 struct mparse *mp;
337 const char *path_arg, *progname;
338 size_t j, sz;
339 int ch, i;
340
341 #if HAVE_PLEDGE
342 if (pledge("stdio rpath wpath cpath fattr flock proc exec", NULL) == -1) {
343 warn("pledge");
344 return (int)MANDOCLEVEL_SYSERR;
345 }
346 #endif
347
348 memset(&conf, 0, sizeof(conf));
349 memset(stmts, 0, STMT__MAX * sizeof(sqlite3_stmt *));
350
351 /*
352 * We accept a few different invocations.
353 * The CHECKOP macro makes sure that invocation styles don't
354 * clobber each other.
355 */
356 #define CHECKOP(_op, _ch) do \
357 if (OP_DEFAULT != (_op)) { \
358 warnx("-%c: Conflicting option", (_ch)); \
359 goto usage; \
360 } while (/*CONSTCOND*/0)
361
362 path_arg = NULL;
363 op = OP_DEFAULT;
364
365 while (-1 != (ch = getopt(argc, argv, "aC:Dd:npQT:tu:v")))
366 switch (ch) {
367 case 'a':
368 use_all = 1;
369 break;
370 case 'C':
371 CHECKOP(op, ch);
372 path_arg = optarg;
373 op = OP_CONFFILE;
374 break;
375 case 'D':
376 debug++;
377 break;
378 case 'd':
379 CHECKOP(op, ch);
380 path_arg = optarg;
381 op = OP_UPDATE;
382 break;
383 case 'n':
384 nodb = 1;
385 break;
386 case 'p':
387 warnings = 1;
388 break;
389 case 'Q':
390 mparse_options |= MPARSE_QUICK;
391 break;
392 case 'T':
393 if (strcmp(optarg, "utf8")) {
394 warnx("-T%s: Unsupported output format",
395 optarg);
396 goto usage;
397 }
398 write_utf8 = 1;
399 break;
400 case 't':
401 CHECKOP(op, ch);
402 dup2(STDOUT_FILENO, STDERR_FILENO);
403 op = OP_TEST;
404 nodb = warnings = 1;
405 break;
406 case 'u':
407 CHECKOP(op, ch);
408 path_arg = optarg;
409 op = OP_DELETE;
410 break;
411 case 'v':
412 /* Compatibility with espie@'s makewhatis. */
413 break;
414 default:
415 goto usage;
416 }
417
418 argc -= optind;
419 argv += optind;
420
421 #if HAVE_PLEDGE
422 if (nodb) {
423 if (pledge("stdio rpath", NULL) == -1) {
424 warn("pledge");
425 return (int)MANDOCLEVEL_SYSERR;
426 }
427 }
428 #endif
429
430 if (OP_CONFFILE == op && argc > 0) {
431 warnx("-C: Too many arguments");
432 goto usage;
433 }
434
435 exitcode = (int)MANDOCLEVEL_OK;
436 mchars_alloc();
437 mp = mparse_alloc(mparse_options, MANDOCLEVEL_BADARG, NULL, NULL);
438 mandoc_ohash_init(&mpages, 6, offsetof(struct mpage, inodev));
439 mandoc_ohash_init(&mlinks, 6, offsetof(struct mlink, file));
440
441 if (OP_UPDATE == op || OP_DELETE == op || OP_TEST == op) {
442
443 /*
444 * Most of these deal with a specific directory.
445 * Jump into that directory first.
446 */
447 if (OP_TEST != op && 0 == set_basedir(path_arg, 1))
448 goto out;
449
450 if (dbopen(1)) {
451 /*
452 * The existing database is usable. Process
453 * all files specified on the command-line.
454 */
455 #if HAVE_PLEDGE
456 if (!nodb) {
457 if (pledge("stdio rpath wpath cpath fattr flock", NULL) == -1) {
458 warn("pledge");
459 exitcode = (int)MANDOCLEVEL_SYSERR;
460 goto out;
461 }
462 }
463 #endif
464 use_all = 1;
465 for (i = 0; i < argc; i++)
466 filescan(argv[i]);
467 if (OP_TEST != op)
468 dbprune();
469 } else {
470 /*
471 * Database missing or corrupt.
472 * Recreate from scratch.
473 */
474 exitcode = (int)MANDOCLEVEL_OK;
475 op = OP_DEFAULT;
476 if (0 == treescan())
477 goto out;
478 if (0 == dbopen(0))
479 goto out;
480 }
481 if (OP_DELETE != op)
482 mpages_merge(mp);
483 dbclose(OP_DEFAULT == op ? 0 : 1);
484 } else {
485 /*
486 * If we have arguments, use them as our manpaths.
487 * If we don't, grok from manpath(1) or however else
488 * manconf_parse() wants to do it.
489 */
490 if (argc > 0) {
491 conf.manpath.paths = mandoc_reallocarray(NULL,
492 argc, sizeof(char *));
493 conf.manpath.sz = (size_t)argc;
494 for (i = 0; i < argc; i++)
495 conf.manpath.paths[i] = mandoc_strdup(argv[i]);
496 } else
497 manconf_parse(&conf, path_arg, NULL, NULL);
498
499 if (conf.manpath.sz == 0) {
500 exitcode = (int)MANDOCLEVEL_BADARG;
501 say("", "Empty manpath");
502 }
503
504 /*
505 * First scan the tree rooted at a base directory, then
506 * build a new database and finally move it into place.
507 * Ignore zero-length directories and strip trailing
508 * slashes.
509 */
510 for (j = 0; j < conf.manpath.sz; j++) {
511 sz = strlen(conf.manpath.paths[j]);
512 if (sz && conf.manpath.paths[j][sz - 1] == '/')
513 conf.manpath.paths[j][--sz] = '\0';
514 if (0 == sz)
515 continue;
516
517 if (j) {
518 mandoc_ohash_init(&mpages, 6,
519 offsetof(struct mpage, inodev));
520 mandoc_ohash_init(&mlinks, 6,
521 offsetof(struct mlink, file));
522 }
523
524 if ( ! set_basedir(conf.manpath.paths[j], argc > 0))
525 continue;
526 if (0 == treescan())
527 continue;
528 if (0 == dbopen(0))
529 continue;
530
531 mpages_merge(mp);
532 if (warnings && !nodb &&
533 ! (MPARSE_QUICK & mparse_options))
534 names_check();
535 dbclose(0);
536
537 if (j + 1 < conf.manpath.sz) {
538 mpages_free();
539 ohash_delete(&mpages);
540 ohash_delete(&mlinks);
541 }
542 }
543 }
544 out:
545 manconf_free(&conf);
546 mparse_free(mp);
547 mchars_free();
548 mpages_free();
549 ohash_delete(&mpages);
550 ohash_delete(&mlinks);
551 return exitcode;
552 usage:
553 progname = getprogname();
554 fprintf(stderr, "usage: %s [-aDnpQ] [-C file] [-Tutf8]\n"
555 " %s [-aDnpQ] [-Tutf8] dir ...\n"
556 " %s [-DnpQ] [-Tutf8] -d dir [file ...]\n"
557 " %s [-Dnp] -u dir [file ...]\n"
558 " %s [-Q] -t file ...\n",
559 progname, progname, progname, progname, progname);
560
561 return (int)MANDOCLEVEL_BADARG;
562 }
563
564 /*
565 * Scan a directory tree rooted at "basedir" for manpages.
566 * We use fts(), scanning directory parts along the way for clues to our
567 * section and architecture.
568 *
569 * If use_all has been specified, grok all files.
570 * If not, sanitise paths to the following:
571 *
572 * [./]man*[/<arch>]/<name>.<section>
573 * or
574 * [./]cat<section>[/<arch>]/<name>.0
575 *
576 * TODO: accommodate for multi-language directories.
577 */
578 static int
579 treescan(void)
580 {
581 char buf[PATH_MAX];
582 FTS *f;
583 FTSENT *ff;
584 struct mlink *mlink;
585 int dform, gzip;
586 char *dsec, *arch, *fsec, *cp;
587 const char *path;
588 const char *argv[2];
589
590 argv[0] = ".";
591 argv[1] = (char *)NULL;
592
593 f = fts_open((char * const *)argv,
594 FTS_PHYSICAL | FTS_NOCHDIR, NULL);
595 if (f == NULL) {
596 exitcode = (int)MANDOCLEVEL_SYSERR;
597 say("", "&fts_open");
598 return 0;
599 }
600
601 dsec = arch = NULL;
602 dform = FORM_NONE;
603
604 while ((ff = fts_read(f)) != NULL) {
605 path = ff->fts_path + 2;
606 switch (ff->fts_info) {
607
608 /*
609 * Symbolic links require various sanity checks,
610 * then get handled just like regular files.
611 */
612 case FTS_SL:
613 if (realpath(path, buf) == NULL) {
614 if (warnings)
615 say(path, "&realpath");
616 continue;
617 }
618 if (strstr(buf, basedir) != buf
619 #ifdef HOMEBREWDIR
620 && strstr(buf, HOMEBREWDIR) != buf
621 #endif
622 ) {
623 if (warnings) say("",
624 "%s: outside base directory", buf);
625 continue;
626 }
627 /* Use logical inode to avoid mpages dupe. */
628 if (stat(path, ff->fts_statp) == -1) {
629 if (warnings)
630 say(path, "&stat");
631 continue;
632 }
633 /* FALLTHROUGH */
634
635 /*
636 * If we're a regular file, add an mlink by using the
637 * stored directory data and handling the filename.
638 */
639 case FTS_F:
640 if ( ! strcmp(path, MANDOC_DB))
641 continue;
642 if ( ! use_all && ff->fts_level < 2) {
643 if (warnings)
644 say(path, "Extraneous file");
645 continue;
646 }
647 gzip = 0;
648 fsec = NULL;
649 while (fsec == NULL) {
650 fsec = strrchr(ff->fts_name, '.');
651 if (fsec == NULL || strcmp(fsec+1, "gz"))
652 break;
653 gzip = 1;
654 *fsec = '\0';
655 fsec = NULL;
656 }
657 if (fsec == NULL) {
658 if ( ! use_all) {
659 if (warnings)
660 say(path,
661 "No filename suffix");
662 continue;
663 }
664 } else if ( ! strcmp(++fsec, "html")) {
665 if (warnings)
666 say(path, "Skip html");
667 continue;
668 } else if ( ! strcmp(fsec, "ps")) {
669 if (warnings)
670 say(path, "Skip ps");
671 continue;
672 } else if ( ! strcmp(fsec, "pdf")) {
673 if (warnings)
674 say(path, "Skip pdf");
675 continue;
676 } else if ( ! use_all &&
677 ((dform == FORM_SRC &&
678 strncmp(fsec, dsec, strlen(dsec))) ||
679 (dform == FORM_CAT && strcmp(fsec, "0")))) {
680 if (warnings)
681 say(path, "Wrong filename suffix");
682 continue;
683 } else
684 fsec[-1] = '\0';
685
686 mlink = mandoc_calloc(1, sizeof(struct mlink));
687 if (strlcpy(mlink->file, path,
688 sizeof(mlink->file)) >=
689 sizeof(mlink->file)) {
690 say(path, "Filename too long");
691 free(mlink);
692 continue;
693 }
694 mlink->dform = dform;
695 mlink->dsec = dsec;
696 mlink->arch = arch;
697 mlink->name = ff->fts_name;
698 mlink->fsec = fsec;
699 mlink->gzip = gzip;
700 mlink_add(mlink, ff->fts_statp);
701 continue;
702
703 case FTS_D:
704 case FTS_DP:
705 break;
706
707 default:
708 if (warnings)
709 say(path, "Not a regular file");
710 continue;
711 }
712
713 switch (ff->fts_level) {
714 case 0:
715 /* Ignore the root directory. */
716 break;
717 case 1:
718 /*
719 * This might contain manX/ or catX/.
720 * Try to infer this from the name.
721 * If we're not in use_all, enforce it.
722 */
723 cp = ff->fts_name;
724 if (ff->fts_info == FTS_DP) {
725 dform = FORM_NONE;
726 dsec = NULL;
727 break;
728 }
729
730 if ( ! strncmp(cp, "man", 3)) {
731 dform = FORM_SRC;
732 dsec = cp + 3;
733 } else if ( ! strncmp(cp, "cat", 3)) {
734 dform = FORM_CAT;
735 dsec = cp + 3;
736 } else {
737 dform = FORM_NONE;
738 dsec = NULL;
739 }
740
741 if (dsec != NULL || use_all)
742 break;
743
744 if (warnings)
745 say(path, "Unknown directory part");
746 fts_set(f, ff, FTS_SKIP);
747 break;
748 case 2:
749 /*
750 * Possibly our architecture.
751 * If we're descending, keep tabs on it.
752 */
753 if (ff->fts_info != FTS_DP && dsec != NULL)
754 arch = ff->fts_name;
755 else
756 arch = NULL;
757 break;
758 default:
759 if (ff->fts_info == FTS_DP || use_all)
760 break;
761 if (warnings)
762 say(path, "Extraneous directory part");
763 fts_set(f, ff, FTS_SKIP);
764 break;
765 }
766 }
767
768 fts_close(f);
769 return 1;
770 }
771
772 /*
773 * Add a file to the mlinks table.
774 * Do not verify that it's a "valid" looking manpage (we'll do that
775 * later).
776 *
777 * Try to infer the manual section, architecture, and page name from the
778 * path, assuming it looks like
779 *
780 * [./]man*[/<arch>]/<name>.<section>
781 * or
782 * [./]cat<section>[/<arch>]/<name>.0
783 *
784 * See treescan() for the fts(3) version of this.
785 */
786 static void
787 filescan(const char *file)
788 {
789 char buf[PATH_MAX];
790 struct stat st;
791 struct mlink *mlink;
792 char *p, *start;
793
794 assert(use_all);
795
796 if (0 == strncmp(file, "./", 2))
797 file += 2;
798
799 /*
800 * We have to do lstat(2) before realpath(3) loses
801 * the information whether this is a symbolic link.
802 * We need to know that because for symbolic links,
803 * we want to use the orginal file name, while for
804 * regular files, we want to use the real path.
805 */
806 if (-1 == lstat(file, &st)) {
807 exitcode = (int)MANDOCLEVEL_BADARG;
808 say(file, "&lstat");
809 return;
810 } else if (0 == ((S_IFREG | S_IFLNK) & st.st_mode)) {
811 exitcode = (int)MANDOCLEVEL_BADARG;
812 say(file, "Not a regular file");
813 return;
814 }
815
816 /*
817 * We have to resolve the file name to the real path
818 * in any case for the base directory check.
819 */
820 if (NULL == realpath(file, buf)) {
821 exitcode = (int)MANDOCLEVEL_BADARG;
822 say(file, "&realpath");
823 return;
824 }
825
826 if (OP_TEST == op)
827 start = buf;
828 else if (strstr(buf, basedir) == buf)
829 start = buf + strlen(basedir);
830 #ifdef HOMEBREWDIR
831 else if (strstr(buf, HOMEBREWDIR) == buf)
832 start = buf;
833 #endif
834 else {
835 exitcode = (int)MANDOCLEVEL_BADARG;
836 say("", "%s: outside base directory", buf);
837 return;
838 }
839
840 /*
841 * Now we are sure the file is inside our tree.
842 * If it is a symbolic link, ignore the real path
843 * and use the original name.
844 * This implies passing stuff like "cat1/../man1/foo.1"
845 * on the command line won't work. So don't do that.
846 * Note the stat(2) can still fail if the link target
847 * doesn't exist.
848 */
849 if (S_IFLNK & st.st_mode) {
850 if (-1 == stat(buf, &st)) {
851 exitcode = (int)MANDOCLEVEL_BADARG;
852 say(file, "&stat");
853 return;
854 }
855 if (strlcpy(buf, file, sizeof(buf)) >= sizeof(buf)) {
856 say(file, "Filename too long");
857 return;
858 }
859 start = buf;
860 if (OP_TEST != op && strstr(buf, basedir) == buf)
861 start += strlen(basedir);
862 }
863
864 mlink = mandoc_calloc(1, sizeof(struct mlink));
865 mlink->dform = FORM_NONE;
866 if (strlcpy(mlink->file, start, sizeof(mlink->file)) >=
867 sizeof(mlink->file)) {
868 say(start, "Filename too long");
869 free(mlink);
870 return;
871 }
872
873 /*
874 * First try to guess our directory structure.
875 * If we find a separator, try to look for man* or cat*.
876 * If we find one of these and what's underneath is a directory,
877 * assume it's an architecture.
878 */
879 if (NULL != (p = strchr(start, '/'))) {
880 *p++ = '\0';
881 if (0 == strncmp(start, "man", 3)) {
882 mlink->dform = FORM_SRC;
883 mlink->dsec = start + 3;
884 } else if (0 == strncmp(start, "cat", 3)) {
885 mlink->dform = FORM_CAT;
886 mlink->dsec = start + 3;
887 }
888
889 start = p;
890 if (NULL != mlink->dsec && NULL != (p = strchr(start, '/'))) {
891 *p++ = '\0';
892 mlink->arch = start;
893 start = p;
894 }
895 }
896
897 /*
898 * Now check the file suffix.
899 * Suffix of `.0' indicates a catpage, `.1-9' is a manpage.
900 */
901 p = strrchr(start, '\0');
902 while (p-- > start && '/' != *p && '.' != *p)
903 /* Loop. */ ;
904
905 if ('.' == *p) {
906 *p++ = '\0';
907 mlink->fsec = p;
908 }
909
910 /*
911 * Now try to parse the name.
912 * Use the filename portion of the path.
913 */
914 mlink->name = start;
915 if (NULL != (p = strrchr(start, '/'))) {
916 mlink->name = p + 1;
917 *p = '\0';
918 }
919 mlink_add(mlink, &st);
920 }
921
922 static void
923 mlink_add(struct mlink *mlink, const struct stat *st)
924 {
925 struct inodev inodev;
926 struct mpage *mpage;
927 unsigned int slot;
928
929 assert(NULL != mlink->file);
930
931 mlink->dsec = mandoc_strdup(mlink->dsec ? mlink->dsec : "");
932 mlink->arch = mandoc_strdup(mlink->arch ? mlink->arch : "");
933 mlink->name = mandoc_strdup(mlink->name ? mlink->name : "");
934 mlink->fsec = mandoc_strdup(mlink->fsec ? mlink->fsec : "");
935
936 if ('0' == *mlink->fsec) {
937 free(mlink->fsec);
938 mlink->fsec = mandoc_strdup(mlink->dsec);
939 mlink->fform = FORM_CAT;
940 } else if ('1' <= *mlink->fsec && '9' >= *mlink->fsec)
941 mlink->fform = FORM_SRC;
942 else
943 mlink->fform = FORM_NONE;
944
945 slot = ohash_qlookup(&mlinks, mlink->file);
946 assert(NULL == ohash_find(&mlinks, slot));
947 ohash_insert(&mlinks, slot, mlink);
948
949 memset(&inodev, 0, sizeof(inodev)); /* Clear padding. */
950 inodev.st_ino = st->st_ino;
951 inodev.st_dev = st->st_dev;
952 slot = ohash_lookup_memory(&mpages, (char *)&inodev,
953 sizeof(struct inodev), inodev.st_ino);
954 mpage = ohash_find(&mpages, slot);
955 if (NULL == mpage) {
956 mpage = mandoc_calloc(1, sizeof(struct mpage));
957 mpage->inodev.st_ino = inodev.st_ino;
958 mpage->inodev.st_dev = inodev.st_dev;
959 ohash_insert(&mpages, slot, mpage);
960 } else
961 mlink->next = mpage->mlinks;
962 mpage->mlinks = mlink;
963 mlink->mpage = mpage;
964 }
965
966 static void
967 mlink_free(struct mlink *mlink)
968 {
969
970 free(mlink->dsec);
971 free(mlink->arch);
972 free(mlink->name);
973 free(mlink->fsec);
974 free(mlink);
975 }
976
977 static void
978 mpages_free(void)
979 {
980 struct mpage *mpage;
981 struct mlink *mlink;
982 unsigned int slot;
983
984 mpage = ohash_first(&mpages, &slot);
985 while (NULL != mpage) {
986 while (NULL != (mlink = mpage->mlinks)) {
987 mpage->mlinks = mlink->next;
988 mlink_free(mlink);
989 }
990 free(mpage->sec);
991 free(mpage->arch);
992 free(mpage->title);
993 free(mpage->desc);
994 free(mpage);
995 mpage = ohash_next(&mpages, &slot);
996 }
997 }
998
999 /*
1000 * For each mlink to the mpage, check whether the path looks like
1001 * it is formatted, and if it does, check whether a source manual
1002 * exists by the same name, ignoring the suffix.
1003 * If both conditions hold, drop the mlink.
1004 */
1005 static void
1006 mlinks_undupe(struct mpage *mpage)
1007 {
1008 char buf[PATH_MAX];
1009 struct mlink **prev;
1010 struct mlink *mlink;
1011 char *bufp;
1012
1013 mpage->form = FORM_CAT;
1014 prev = &mpage->mlinks;
1015 while (NULL != (mlink = *prev)) {
1016 if (FORM_CAT != mlink->dform) {
1017 mpage->form = FORM_NONE;
1018 goto nextlink;
1019 }
1020 (void)strlcpy(buf, mlink->file, sizeof(buf));
1021 bufp = strstr(buf, "cat");
1022 assert(NULL != bufp);
1023 memcpy(bufp, "man", 3);
1024 if (NULL != (bufp = strrchr(buf, '.')))
1025 *++bufp = '\0';
1026 (void)strlcat(buf, mlink->dsec, sizeof(buf));
1027 if (NULL == ohash_find(&mlinks,
1028 ohash_qlookup(&mlinks, buf)))
1029 goto nextlink;
1030 if (warnings)
1031 say(mlink->file, "Man source exists: %s", buf);
1032 if (use_all)
1033 goto nextlink;
1034 *prev = mlink->next;
1035 mlink_free(mlink);
1036 continue;
1037 nextlink:
1038 prev = &(*prev)->next;
1039 }
1040 }
1041
1042 static void
1043 mlink_check(struct mpage *mpage, struct mlink *mlink)
1044 {
1045 struct str *str;
1046 unsigned int slot;
1047
1048 /*
1049 * Check whether the manual section given in a file
1050 * agrees with the directory where the file is located.
1051 * Some manuals have suffixes like (3p) on their
1052 * section number either inside the file or in the
1053 * directory name, some are linked into more than one
1054 * section, like encrypt(1) = makekey(8).
1055 */
1056
1057 if (FORM_SRC == mpage->form &&
1058 strcasecmp(mpage->sec, mlink->dsec))
1059 say(mlink->file, "Section \"%s\" manual in %s directory",
1060 mpage->sec, mlink->dsec);
1061
1062 /*
1063 * Manual page directories exist for each kernel
1064 * architecture as returned by machine(1).
1065 * However, many manuals only depend on the
1066 * application architecture as returned by arch(1).
1067 * For example, some (2/ARM) manuals are shared
1068 * across the "armish" and "zaurus" kernel
1069 * architectures.
1070 * A few manuals are even shared across completely
1071 * different architectures, for example fdformat(1)
1072 * on amd64, i386, sparc, and sparc64.
1073 */
1074
1075 if (strcasecmp(mpage->arch, mlink->arch))
1076 say(mlink->file, "Architecture \"%s\" manual in "
1077 "\"%s\" directory", mpage->arch, mlink->arch);
1078
1079 /*
1080 * XXX
1081 * parse_cat() doesn't set NAME_TITLE yet.
1082 */
1083
1084 if (FORM_CAT == mpage->form)
1085 return;
1086
1087 /*
1088 * Check whether this mlink
1089 * appears as a name in the NAME section.
1090 */
1091
1092 slot = ohash_qlookup(&names, mlink->name);
1093 str = ohash_find(&names, slot);
1094 assert(NULL != str);
1095 if ( ! (NAME_TITLE & str->mask))
1096 say(mlink->file, "Name missing in NAME section");
1097 }
1098
1099 /*
1100 * Run through the files in the global vector "mpages"
1101 * and add them to the database specified in "basedir".
1102 *
1103 * This handles the parsing scheme itself, using the cues of directory
1104 * and filename to determine whether the file is parsable or not.
1105 */
1106 static void
1107 mpages_merge(struct mparse *mp)
1108 {
1109 char any[] = "any";
1110 struct mpage *mpage, *mpage_dest;
1111 struct mlink *mlink, *mlink_dest;
1112 struct roff_man *man;
1113 char *sodest;
1114 char *cp;
1115 int fd;
1116 unsigned int pslot;
1117
1118 if ( ! nodb)
1119 SQL_EXEC("BEGIN TRANSACTION");
1120
1121 mpage = ohash_first(&mpages, &pslot);
1122 while (mpage != NULL) {
1123 mlinks_undupe(mpage);
1124 if ((mlink = mpage->mlinks) == NULL) {
1125 mpage = ohash_next(&mpages, &pslot);
1126 continue;
1127 }
1128
1129 name_mask = NAME_MASK;
1130 mandoc_ohash_init(&names, 4, offsetof(struct str, key));
1131 mandoc_ohash_init(&strings, 6, offsetof(struct str, key));
1132 mparse_reset(mp);
1133 man = NULL;
1134 sodest = NULL;
1135
1136 if ((fd = mparse_open(mp, mlink->file)) == -1) {
1137 say(mlink->file, "&open");
1138 goto nextpage;
1139 }
1140
1141 /*
1142 * Interpret the file as mdoc(7) or man(7) source
1143 * code, unless it is known to be formatted.
1144 */
1145 if (mlink->dform != FORM_CAT || mlink->fform != FORM_CAT) {
1146 mparse_readfd(mp, fd, mlink->file);
1147 close(fd);
1148 mparse_result(mp, &man, &sodest);
1149 }
1150
1151 if (sodest != NULL) {
1152 mlink_dest = ohash_find(&mlinks,
1153 ohash_qlookup(&mlinks, sodest));
1154 if (mlink_dest == NULL) {
1155 mandoc_asprintf(&cp, "%s.gz", sodest);
1156 mlink_dest = ohash_find(&mlinks,
1157 ohash_qlookup(&mlinks, cp));
1158 free(cp);
1159 }
1160 if (mlink_dest != NULL) {
1161
1162 /* The .so target exists. */
1163
1164 mpage_dest = mlink_dest->mpage;
1165 while (1) {
1166 mlink->mpage = mpage_dest;
1167
1168 /*
1169 * If the target was already
1170 * processed, add the links
1171 * to the database now.
1172 * Otherwise, this will
1173 * happen when we come
1174 * to the target.
1175 */
1176
1177 if (mpage_dest->pageid)
1178 dbadd_mlink_name(mlink);
1179
1180 if (mlink->next == NULL)
1181 break;
1182 mlink = mlink->next;
1183 }
1184
1185 /* Move all links to the target. */
1186
1187 mlink->next = mlink_dest->next;
1188 mlink_dest->next = mpage->mlinks;
1189 mpage->mlinks = NULL;
1190 }
1191 goto nextpage;
1192 } else if (man != NULL && man->macroset == MACROSET_MDOC) {
1193 mdoc_validate(man);
1194 mpage->form = FORM_SRC;
1195 mpage->sec = man->meta.msec;
1196 mpage->sec = mandoc_strdup(
1197 mpage->sec == NULL ? "" : mpage->sec);
1198 mpage->arch = man->meta.arch;
1199 mpage->arch = mandoc_strdup(
1200 mpage->arch == NULL ? "" : mpage->arch);
1201 mpage->title = mandoc_strdup(man->meta.title);
1202 } else if (man != NULL && man->macroset == MACROSET_MAN) {
1203 man_validate(man);
1204 mpage->form = FORM_SRC;
1205 mpage->sec = mandoc_strdup(man->meta.msec);
1206 mpage->arch = mandoc_strdup(mlink->arch);
1207 mpage->title = mandoc_strdup(man->meta.title);
1208 } else {
1209 mpage->form = FORM_CAT;
1210 mpage->sec = mandoc_strdup(mlink->dsec);
1211 mpage->arch = mandoc_strdup(mlink->arch);
1212 mpage->title = mandoc_strdup(mlink->name);
1213 }
1214 putkey(mpage, mpage->sec, TYPE_sec);
1215 if (*mpage->arch != '\0')
1216 putkey(mpage, mpage->arch, TYPE_arch);
1217
1218 for ( ; mlink != NULL; mlink = mlink->next) {
1219 if ('\0' != *mlink->dsec)
1220 putkey(mpage, mlink->dsec, TYPE_sec);
1221 if ('\0' != *mlink->fsec)
1222 putkey(mpage, mlink->fsec, TYPE_sec);
1223 putkey(mpage, '\0' == *mlink->arch ?
1224 any : mlink->arch, TYPE_arch);
1225 putkey(mpage, mlink->name, NAME_FILE);
1226 }
1227
1228 assert(mpage->desc == NULL);
1229 if (man != NULL && man->macroset == MACROSET_MDOC)
1230 parse_mdoc(mpage, &man->meta, man->first);
1231 else if (man != NULL)
1232 parse_man(mpage, &man->meta, man->first);
1233 else
1234 parse_cat(mpage, fd);
1235 if (mpage->desc == NULL)
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);
1244 mlink = mpage->mlinks;
1245
1246 nextpage:
1247 ohash_delete(&strings);
1248 ohash_delete(&names);
1249 mpage = ohash_next(&mpages, &pslot);
1250 }
1251
1252 if (0 == nodb)
1253 SQL_EXEC("END TRANSACTION");
1254 }
1255
1256 static void
1257 names_check(void)
1258 {
1259 sqlite3_stmt *stmt;
1260 const char *name, *sec, *arch, *key;
1261
1262 sqlite3_prepare_v2(db,
1263 "SELECT name, sec, arch, key FROM ("
1264 "SELECT name AS key, pageid FROM names "
1265 "WHERE bits & ? AND NOT EXISTS ("
1266 "SELECT pageid FROM mlinks "
1267 "WHERE mlinks.pageid == names.pageid "
1268 "AND mlinks.name == names.name"
1269 ")"
1270 ") JOIN ("
1271 "SELECT sec, arch, name, pageid FROM mlinks "
1272 "GROUP BY pageid"
1273 ") USING (pageid);",
1274 -1, &stmt, NULL);
1275
1276 if (sqlite3_bind_int64(stmt, 1, NAME_TITLE) != SQLITE_OK)
1277 say("", "%s", sqlite3_errmsg(db));
1278
1279 while (sqlite3_step(stmt) == SQLITE_ROW) {
1280 name = (const char *)sqlite3_column_text(stmt, 0);
1281 sec = (const char *)sqlite3_column_text(stmt, 1);
1282 arch = (const char *)sqlite3_column_text(stmt, 2);
1283 key = (const char *)sqlite3_column_text(stmt, 3);
1284 say("", "%s(%s%s%s) lacks mlink \"%s\"", name, sec,
1285 '\0' == *arch ? "" : "/",
1286 '\0' == *arch ? "" : arch, key);
1287 }
1288 sqlite3_finalize(stmt);
1289 }
1290
1291 static void
1292 parse_cat(struct mpage *mpage, int fd)
1293 {
1294 FILE *stream;
1295 char *line, *p, *title;
1296 size_t linesz, plen, titlesz;
1297 ssize_t len;
1298 int offs;
1299
1300 stream = (-1 == fd) ?
1301 fopen(mpage->mlinks->file, "r") :
1302 fdopen(fd, "r");
1303 if (NULL == stream) {
1304 if (-1 != fd)
1305 close(fd);
1306 if (warnings)
1307 say(mpage->mlinks->file, "&fopen");
1308 return;
1309 }
1310
1311 line = NULL;
1312 linesz = 0;
1313
1314 /* Skip to first blank line. */
1315
1316 while (getline(&line, &linesz, stream) != -1)
1317 if (*line == '\n')
1318 break;
1319
1320 /*
1321 * Assume the first line that is not indented
1322 * is the first section header. Skip to it.
1323 */
1324
1325 while (getline(&line, &linesz, stream) != -1)
1326 if (*line != '\n' && *line != ' ')
1327 break;
1328
1329 /*
1330 * Read up until the next section into a buffer.
1331 * Strip the leading and trailing newline from each read line,
1332 * appending a trailing space.
1333 * Ignore empty (whitespace-only) lines.
1334 */
1335
1336 titlesz = 0;
1337 title = NULL;
1338
1339 while ((len = getline(&line, &linesz, stream)) != -1) {
1340 if (*line != ' ')
1341 break;
1342 offs = 0;
1343 while (isspace((unsigned char)line[offs]))
1344 offs++;
1345 if (line[offs] == '\0')
1346 continue;
1347 title = mandoc_realloc(title, titlesz + len - offs);
1348 memcpy(title + titlesz, line + offs, len - offs);
1349 titlesz += len - offs;
1350 title[titlesz - 1] = ' ';
1351 }
1352 free(line);
1353
1354 /*
1355 * If no page content can be found, or the input line
1356 * is already the next section header, or there is no
1357 * trailing newline, reuse the page title as the page
1358 * description.
1359 */
1360
1361 if (NULL == title || '\0' == *title) {
1362 if (warnings)
1363 say(mpage->mlinks->file,
1364 "Cannot find NAME section");
1365 fclose(stream);
1366 free(title);
1367 return;
1368 }
1369
1370 title[titlesz - 1] = '\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 roff_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 (n->type == ROFFT_TEXT)
1435 putkey(mpage, n->string, m);
1436 }
1437 }
1438
1439 static void
1440 parse_man(struct mpage *mpage, const struct roff_meta *meta,
1441 const struct roff_node *n)
1442 {
1443 const struct roff_node *head, *body;
1444 char *start, *title;
1445 char byte;
1446 size_t sz;
1447
1448 if (n == NULL)
1449 return;
1450
1451 /*
1452 * We're only searching for one thing: the first text child in
1453 * the BODY of a NAME section. Since we don't keep track of
1454 * sections in -man, run some hoops to find out whether we're in
1455 * the correct section or not.
1456 */
1457
1458 if (n->type == ROFFT_BODY && n->tok == MAN_SH) {
1459 body = n;
1460 if ((head = body->parent->head) != NULL &&
1461 (head = head->child) != NULL &&
1462 head->next == NULL &&
1463 head->type == ROFFT_TEXT &&
1464 strcmp(head->string, "NAME") == 0 &&
1465 body->child != NULL) {
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 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 if ( ! (mpage->name_head_done ||
1507 strcasecmp(start, meta->title))) {
1508 putkey(mpage, start, NAME_HEAD);
1509 mpage->name_head_done = 1;
1510 }
1511
1512 if (' ' == byte) {
1513 start += sz + 1;
1514 break;
1515 }
1516
1517 assert(',' == byte);
1518 start += sz + 1;
1519 while (' ' == *start)
1520 start++;
1521 }
1522
1523 if (start == title) {
1524 putkey(mpage, start, NAME_TITLE);
1525 if ( ! (mpage->name_head_done ||
1526 strcasecmp(start, meta->title))) {
1527 putkey(mpage, start, NAME_HEAD);
1528 mpage->name_head_done = 1;
1529 }
1530 free(title);
1531 return;
1532 }
1533
1534 while (isspace((unsigned char)*start))
1535 start++;
1536
1537 if (0 == strncmp(start, "-", 1))
1538 start += 1;
1539 else if (0 == strncmp(start, "\\-\\-", 4))
1540 start += 4;
1541 else if (0 == strncmp(start, "\\-", 2))
1542 start += 2;
1543 else if (0 == strncmp(start, "\\(en", 4))
1544 start += 4;
1545 else if (0 == strncmp(start, "\\(em", 4))
1546 start += 4;
1547
1548 while (' ' == *start)
1549 start++;
1550
1551 mpage->desc = mandoc_strdup(start);
1552 free(title);
1553 return;
1554 }
1555 }
1556
1557 for (n = n->child; n; n = n->next) {
1558 if (NULL != mpage->desc)
1559 break;
1560 parse_man(mpage, meta, n);
1561 }
1562 }
1563
1564 static void
1565 parse_mdoc(struct mpage *mpage, const struct roff_meta *meta,
1566 const struct roff_node *n)
1567 {
1568
1569 assert(NULL != n);
1570 for (n = n->child; NULL != n; n = n->next) {
1571 switch (n->type) {
1572 case ROFFT_ELEM:
1573 case ROFFT_BLOCK:
1574 case ROFFT_HEAD:
1575 case ROFFT_BODY:
1576 case ROFFT_TAIL:
1577 if (NULL != mdocs[n->tok].fp)
1578 if (0 == (*mdocs[n->tok].fp)(mpage, meta, n))
1579 break;
1580 if (mdocs[n->tok].mask)
1581 putmdockey(mpage, n->child,
1582 mdocs[n->tok].mask);
1583 break;
1584 default:
1585 assert(n->type != ROFFT_ROOT);
1586 continue;
1587 }
1588 if (NULL != n->child)
1589 parse_mdoc(mpage, meta, n);
1590 }
1591 }
1592
1593 static int
1594 parse_mdoc_Fd(struct mpage *mpage, const struct roff_meta *meta,
1595 const struct roff_node *n)
1596 {
1597 char *start, *end;
1598 size_t sz;
1599
1600 if (SEC_SYNOPSIS != n->sec ||
1601 NULL == (n = n->child) ||
1602 n->type != ROFFT_TEXT)
1603 return 0;
1604
1605 /*
1606 * Only consider those `Fd' macro fields that begin with an
1607 * "inclusion" token (versus, e.g., #define).
1608 */
1609
1610 if (strcmp("#include", n->string))
1611 return 0;
1612
1613 if ((n = n->next) == NULL || n->type != ROFFT_TEXT)
1614 return 0;
1615
1616 /*
1617 * Strip away the enclosing angle brackets and make sure we're
1618 * not zero-length.
1619 */
1620
1621 start = n->string;
1622 if ('<' == *start || '"' == *start)
1623 start++;
1624
1625 if (0 == (sz = strlen(start)))
1626 return 0;
1627
1628 end = &start[(int)sz - 1];
1629 if ('>' == *end || '"' == *end)
1630 end--;
1631
1632 if (end > start)
1633 putkeys(mpage, start, end - start + 1, TYPE_In);
1634 return 0;
1635 }
1636
1637 static void
1638 parse_mdoc_fname(struct mpage *mpage, const struct roff_node *n)
1639 {
1640 char *cp;
1641 size_t sz;
1642
1643 if (n->type != ROFFT_TEXT)
1644 return;
1645
1646 /* Skip function pointer punctuation. */
1647
1648 cp = n->string;
1649 while (*cp == '(' || *cp == '*')
1650 cp++;
1651 sz = strcspn(cp, "()");
1652
1653 putkeys(mpage, cp, sz, TYPE_Fn);
1654 if (n->sec == SEC_SYNOPSIS)
1655 putkeys(mpage, cp, sz, NAME_SYN);
1656 }
1657
1658 static int
1659 parse_mdoc_Fn(struct mpage *mpage, const struct roff_meta *meta,
1660 const struct roff_node *n)
1661 {
1662
1663 if (n->child == NULL)
1664 return 0;
1665
1666 parse_mdoc_fname(mpage, n->child);
1667
1668 for (n = n->child->next; n != NULL; n = n->next)
1669 if (n->type == ROFFT_TEXT)
1670 putkey(mpage, n->string, TYPE_Fa);
1671
1672 return 0;
1673 }
1674
1675 static int
1676 parse_mdoc_Fo(struct mpage *mpage, const struct roff_meta *meta,
1677 const struct roff_node *n)
1678 {
1679
1680 if (n->type != ROFFT_HEAD)
1681 return 1;
1682
1683 if (n->child != NULL)
1684 parse_mdoc_fname(mpage, n->child);
1685
1686 return 0;
1687 }
1688
1689 static int
1690 parse_mdoc_Va(struct mpage *mpage, const struct roff_meta *meta,
1691 const struct roff_node *n)
1692 {
1693 char *cp;
1694
1695 if (n->type != ROFFT_ELEM && n->type != ROFFT_BODY)
1696 return 0;
1697
1698 if (n->child != NULL &&
1699 n->child->next == NULL &&
1700 n->child->type == ROFFT_TEXT)
1701 return 1;
1702
1703 cp = NULL;
1704 deroff(&cp, n);
1705 if (cp != NULL) {
1706 putkey(mpage, cp, TYPE_Vt | (n->tok == MDOC_Va ||
1707 n->type == ROFFT_BODY ? TYPE_Va : 0));
1708 free(cp);
1709 }
1710
1711 return 0;
1712 }
1713
1714 static int
1715 parse_mdoc_Xr(struct mpage *mpage, const struct roff_meta *meta,
1716 const struct roff_node *n)
1717 {
1718 char *cp;
1719
1720 if (NULL == (n = n->child))
1721 return 0;
1722
1723 if (NULL == n->next) {
1724 putkey(mpage, n->string, TYPE_Xr);
1725 return 0;
1726 }
1727
1728 mandoc_asprintf(&cp, "%s(%s)", n->string, n->next->string);
1729 putkey(mpage, cp, TYPE_Xr);
1730 free(cp);
1731 return 0;
1732 }
1733
1734 static int
1735 parse_mdoc_Nd(struct mpage *mpage, const struct roff_meta *meta,
1736 const struct roff_node *n)
1737 {
1738
1739 if (n->type == ROFFT_BODY)
1740 deroff(&mpage->desc, n);
1741 return 0;
1742 }
1743
1744 static int
1745 parse_mdoc_Nm(struct mpage *mpage, const struct roff_meta *meta,
1746 const struct roff_node *n)
1747 {
1748
1749 if (SEC_NAME == n->sec)
1750 putmdockey(mpage, n->child, NAME_TITLE);
1751 else if (n->sec == SEC_SYNOPSIS && n->type == ROFFT_HEAD) {
1752 if (n->child == NULL)
1753 putkey(mpage, meta->name, NAME_SYN);
1754 else
1755 putmdockey(mpage, n->child, NAME_SYN);
1756 }
1757 if ( ! (mpage->name_head_done ||
1758 n->child == NULL || n->child->string == NULL ||
1759 strcasecmp(n->child->string, meta->title))) {
1760 putkey(mpage, n->child->string, ROFFT_HEAD);
1761 mpage->name_head_done = 1;
1762 }
1763 return 0;
1764 }
1765
1766 static int
1767 parse_mdoc_Sh(struct mpage *mpage, const struct roff_meta *meta,
1768 const struct roff_node *n)
1769 {
1770
1771 return n->sec == SEC_CUSTOM && n->type == ROFFT_HEAD;
1772 }
1773
1774 static int
1775 parse_mdoc_head(struct mpage *mpage, const struct roff_meta *meta,
1776 const struct roff_node *n)
1777 {
1778
1779 return n->type == ROFFT_HEAD;
1780 }
1781
1782 /*
1783 * Add a string to the hash table for the current manual.
1784 * Each string has a bitmask telling which macros it belongs to.
1785 * When we finish the manual, we'll dump the table.
1786 */
1787 static void
1788 putkeys(const struct mpage *mpage, char *cp, size_t sz, uint64_t v)
1789 {
1790 struct ohash *htab;
1791 struct str *s;
1792 const char *end;
1793 unsigned int slot;
1794 int i, mustfree;
1795
1796 if (0 == sz)
1797 return;
1798
1799 mustfree = render_string(&cp, &sz);
1800
1801 if (TYPE_Nm & v) {
1802 htab = &names;
1803 v &= name_mask;
1804 if (v & NAME_FIRST)
1805 name_mask &= ~NAME_FIRST;
1806 if (debug > 1)
1807 say(mpage->mlinks->file,
1808 "Adding name %*s, bits=%d", sz, cp, v);
1809 } else {
1810 htab = &strings;
1811 if (debug > 1)
1812 for (i = 0; i < mansearch_keymax; i++)
1813 if ((uint64_t)1 << i & v)
1814 say(mpage->mlinks->file,
1815 "Adding key %s=%*s",
1816 mansearch_keynames[i], sz, cp);
1817 }
1818
1819 end = cp + sz;
1820 slot = ohash_qlookupi(htab, cp, &end);
1821 s = ohash_find(htab, slot);
1822
1823 if (NULL != s && mpage == s->mpage) {
1824 s->mask |= v;
1825 return;
1826 } else if (NULL == s) {
1827 s = mandoc_calloc(1, sizeof(struct str) + sz + 1);
1828 memcpy(s->key, cp, sz);
1829 ohash_insert(htab, slot, s);
1830 }
1831 s->mpage = mpage;
1832 s->mask = v;
1833
1834 if (mustfree)
1835 free(cp);
1836 }
1837
1838 /*
1839 * Take a Unicode codepoint and produce its UTF-8 encoding.
1840 * This isn't the best way to do this, but it works.
1841 * The magic numbers are from the UTF-8 packaging.
1842 * They're not as scary as they seem: read the UTF-8 spec for details.
1843 */
1844 static size_t
1845 utf8(unsigned int cp, char out[7])
1846 {
1847 size_t rc;
1848
1849 rc = 0;
1850 if (cp <= 0x0000007F) {
1851 rc = 1;
1852 out[0] = (char)cp;
1853 } else if (cp <= 0x000007FF) {
1854 rc = 2;
1855 out[0] = (cp >> 6 & 31) | 192;
1856 out[1] = (cp & 63) | 128;
1857 } else if (cp <= 0x0000FFFF) {
1858 rc = 3;
1859 out[0] = (cp >> 12 & 15) | 224;
1860 out[1] = (cp >> 6 & 63) | 128;
1861 out[2] = (cp & 63) | 128;
1862 } else if (cp <= 0x001FFFFF) {
1863 rc = 4;
1864 out[0] = (cp >> 18 & 7) | 240;
1865 out[1] = (cp >> 12 & 63) | 128;
1866 out[2] = (cp >> 6 & 63) | 128;
1867 out[3] = (cp & 63) | 128;
1868 } else if (cp <= 0x03FFFFFF) {
1869 rc = 5;
1870 out[0] = (cp >> 24 & 3) | 248;
1871 out[1] = (cp >> 18 & 63) | 128;
1872 out[2] = (cp >> 12 & 63) | 128;
1873 out[3] = (cp >> 6 & 63) | 128;
1874 out[4] = (cp & 63) | 128;
1875 } else if (cp <= 0x7FFFFFFF) {
1876 rc = 6;
1877 out[0] = (cp >> 30 & 1) | 252;
1878 out[1] = (cp >> 24 & 63) | 128;
1879 out[2] = (cp >> 18 & 63) | 128;
1880 out[3] = (cp >> 12 & 63) | 128;
1881 out[4] = (cp >> 6 & 63) | 128;
1882 out[5] = (cp & 63) | 128;
1883 } else
1884 return 0;
1885
1886 out[rc] = '\0';
1887 return rc;
1888 }
1889
1890 /*
1891 * If the string contains escape sequences,
1892 * replace it with an allocated rendering and return 1,
1893 * such that the caller can free it after use.
1894 * Otherwise, do nothing and return 0.
1895 */
1896 static int
1897 render_string(char **public, size_t *psz)
1898 {
1899 const char *src, *scp, *addcp, *seq;
1900 char *dst;
1901 size_t ssz, dsz, addsz;
1902 char utfbuf[7], res[6];
1903 int seqlen, unicode;
1904
1905 res[0] = '\\';
1906 res[1] = '\t';
1907 res[2] = ASCII_NBRSP;
1908 res[3] = ASCII_HYPH;
1909 res[4] = ASCII_BREAK;
1910 res[5] = '\0';
1911
1912 src = scp = *public;
1913 ssz = *psz;
1914 dst = NULL;
1915 dsz = 0;
1916
1917 while (scp < src + *psz) {
1918
1919 /* Leave normal characters unchanged. */
1920
1921 if (strchr(res, *scp) == NULL) {
1922 if (dst != NULL)
1923 dst[dsz++] = *scp;
1924 scp++;
1925 continue;
1926 }
1927
1928 /*
1929 * Found something that requires replacing,
1930 * make sure we have a destination buffer.
1931 */
1932
1933 if (dst == NULL) {
1934 dst = mandoc_malloc(ssz + 1);
1935 dsz = scp - src;
1936 memcpy(dst, src, dsz);
1937 }
1938
1939 /* Handle single-char special characters. */
1940
1941 switch (*scp) {
1942 case '\\':
1943 break;
1944 case '\t':
1945 case ASCII_NBRSP:
1946 dst[dsz++] = ' ';
1947 scp++;
1948 continue;
1949 case ASCII_HYPH:
1950 dst[dsz++] = '-';
1951 /* FALLTHROUGH */
1952 case ASCII_BREAK:
1953 scp++;
1954 continue;
1955 default:
1956 abort();
1957 }
1958
1959 /*
1960 * Found an escape sequence.
1961 * Read past the slash, then parse it.
1962 * Ignore everything except characters.
1963 */
1964
1965 scp++;
1966 if (mandoc_escape(&scp, &seq, &seqlen) != ESCAPE_SPECIAL)
1967 continue;
1968
1969 /*
1970 * Render the special character
1971 * as either UTF-8 or ASCII.
1972 */
1973
1974 if (write_utf8) {
1975 unicode = mchars_spec2cp(seq, seqlen);
1976 if (unicode <= 0)
1977 continue;
1978 addsz = utf8(unicode, utfbuf);
1979 if (addsz == 0)
1980 continue;
1981 addcp = utfbuf;
1982 } else {
1983 addcp = mchars_spec2str(seq, seqlen, &addsz);
1984 if (addcp == NULL)
1985 continue;
1986 if (*addcp == ASCII_NBRSP) {
1987 addcp = " ";
1988 addsz = 1;
1989 }
1990 }
1991
1992 /* Copy the rendered glyph into the stream. */
1993
1994 ssz += addsz;
1995 dst = mandoc_realloc(dst, ssz + 1);
1996 memcpy(dst + dsz, addcp, addsz);
1997 dsz += addsz;
1998 }
1999 if (dst != NULL) {
2000 *public = dst;
2001 *psz = dsz;
2002 }
2003
2004 /* Trim trailing whitespace and NUL-terminate. */
2005
2006 while (*psz > 0 && (*public)[*psz - 1] == ' ')
2007 --*psz;
2008 if (dst != NULL) {
2009 (*public)[*psz] = '\0';
2010 return 1;
2011 } else
2012 return 0;
2013 }
2014
2015 static void
2016 dbadd_mlink(const struct mlink *mlink)
2017 {
2018 size_t i;
2019
2020 i = 1;
2021 SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->dsec);
2022 SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->arch);
2023 SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->name);
2024 SQL_BIND_INT64(stmts[STMT_INSERT_LINK], i, mlink->mpage->pageid);
2025 SQL_STEP(stmts[STMT_INSERT_LINK]);
2026 sqlite3_reset(stmts[STMT_INSERT_LINK]);
2027 }
2028
2029 static void
2030 dbadd_mlink_name(const struct mlink *mlink)
2031 {
2032 uint64_t bits;
2033 size_t i;
2034
2035 dbadd_mlink(mlink);
2036
2037 i = 1;
2038 SQL_BIND_INT64(stmts[STMT_SELECT_NAME], i, mlink->mpage->pageid);
2039 bits = NAME_FILE & NAME_MASK;
2040 if (sqlite3_step(stmts[STMT_SELECT_NAME]) == SQLITE_ROW) {
2041 bits |= sqlite3_column_int64(stmts[STMT_SELECT_NAME], 0);
2042 sqlite3_reset(stmts[STMT_SELECT_NAME]);
2043 }
2044
2045 i = 1;
2046 SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, bits);
2047 SQL_BIND_TEXT(stmts[STMT_INSERT_NAME], i, mlink->name);
2048 SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, mlink->mpage->pageid);
2049 SQL_STEP(stmts[STMT_INSERT_NAME]);
2050 sqlite3_reset(stmts[STMT_INSERT_NAME]);
2051 }
2052
2053 /*
2054 * Flush the current page's terms (and their bits) into the database.
2055 * Wrap the entire set of additions in a transaction to make sqlite be a
2056 * little faster.
2057 * Also, handle escape sequences at the last possible moment.
2058 */
2059 static void
2060 dbadd(struct mpage *mpage)
2061 {
2062 struct mlink *mlink;
2063 struct str *key;
2064 char *cp;
2065 size_t i;
2066 unsigned int slot;
2067 int mustfree;
2068
2069 mlink = mpage->mlinks;
2070
2071 if (nodb) {
2072 for (key = ohash_first(&names, &slot); NULL != key;
2073 key = ohash_next(&names, &slot))
2074 free(key);
2075 for (key = ohash_first(&strings, &slot); NULL != key;
2076 key = ohash_next(&strings, &slot))
2077 free(key);
2078 if (0 == debug)
2079 return;
2080 while (NULL != mlink) {
2081 fputs(mlink->name, stdout);
2082 if (NULL == mlink->next ||
2083 strcmp(mlink->dsec, mlink->next->dsec) ||
2084 strcmp(mlink->fsec, mlink->next->fsec) ||
2085 strcmp(mlink->arch, mlink->next->arch)) {
2086 putchar('(');
2087 if ('\0' == *mlink->dsec)
2088 fputs(mlink->fsec, stdout);
2089 else
2090 fputs(mlink->dsec, stdout);
2091 if ('\0' != *mlink->arch)
2092 printf("/%s", mlink->arch);
2093 putchar(')');
2094 }
2095 mlink = mlink->next;
2096 if (NULL != mlink)
2097 fputs(", ", stdout);
2098 }
2099 printf(" - %s\n", mpage->desc);
2100 return;
2101 }
2102
2103 if (debug)
2104 say(mlink->file, "Adding to database");
2105
2106 cp = mpage->desc;
2107 i = strlen(cp);
2108 mustfree = render_string(&cp, &i);
2109 i = 1;
2110 SQL_BIND_TEXT(stmts[STMT_INSERT_PAGE], i, cp);
2111 SQL_BIND_INT(stmts[STMT_INSERT_PAGE], i, mpage->form);
2112 SQL_STEP(stmts[STMT_INSERT_PAGE]);
2113 mpage->pageid = sqlite3_last_insert_rowid(db);
2114 sqlite3_reset(stmts[STMT_INSERT_PAGE]);
2115 if (mustfree)
2116 free(cp);
2117
2118 while (NULL != mlink) {
2119 dbadd_mlink(mlink);
2120 mlink = mlink->next;
2121 }
2122 mlink = mpage->mlinks;
2123
2124 for (key = ohash_first(&names, &slot); NULL != key;
2125 key = ohash_next(&names, &slot)) {
2126 assert(key->mpage == mpage);
2127 i = 1;
2128 SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, key->mask);
2129 SQL_BIND_TEXT(stmts[STMT_INSERT_NAME], i, key->key);
2130 SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, mpage->pageid);
2131 SQL_STEP(stmts[STMT_INSERT_NAME]);
2132 sqlite3_reset(stmts[STMT_INSERT_NAME]);
2133 free(key);
2134 }
2135 for (key = ohash_first(&strings, &slot); NULL != key;
2136 key = ohash_next(&strings, &slot)) {
2137 assert(key->mpage == mpage);
2138 i = 1;
2139 SQL_BIND_INT64(stmts[STMT_INSERT_KEY], i, key->mask);
2140 SQL_BIND_TEXT(stmts[STMT_INSERT_KEY], i, key->key);
2141 SQL_BIND_INT64(stmts[STMT_INSERT_KEY], i, mpage->pageid);
2142 SQL_STEP(stmts[STMT_INSERT_KEY]);
2143 sqlite3_reset(stmts[STMT_INSERT_KEY]);
2144 free(key);
2145 }
2146 }
2147
2148 static void
2149 dbprune(void)
2150 {
2151 struct mpage *mpage;
2152 struct mlink *mlink;
2153 size_t i;
2154 unsigned int slot;
2155
2156 if (0 == nodb)
2157 SQL_EXEC("BEGIN TRANSACTION");
2158
2159 for (mpage = ohash_first(&mpages, &slot); NULL != mpage;
2160 mpage = ohash_next(&mpages, &slot)) {
2161 mlink = mpage->mlinks;
2162 if (debug)
2163 say(mlink->file, "Deleting from database");
2164 if (nodb)
2165 continue;
2166 for ( ; NULL != mlink; mlink = mlink->next) {
2167 i = 1;
2168 SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE],
2169 i, mlink->dsec);
2170 SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE],
2171 i, mlink->arch);
2172 SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE],
2173 i, mlink->name);
2174 SQL_STEP(stmts[STMT_DELETE_PAGE]);
2175 sqlite3_reset(stmts[STMT_DELETE_PAGE]);
2176 }
2177 }
2178
2179 if (0 == nodb)
2180 SQL_EXEC("END TRANSACTION");
2181 }
2182
2183 /*
2184 * Close an existing database and its prepared statements.
2185 * If "real" is not set, rename the temporary file into the real one.
2186 */
2187 static void
2188 dbclose(int real)
2189 {
2190 size_t i;
2191 int status;
2192 pid_t child;
2193
2194 if (nodb)
2195 return;
2196
2197 for (i = 0; i < STMT__MAX; i++) {
2198 sqlite3_finalize(stmts[i]);
2199 stmts[i] = NULL;
2200 }
2201
2202 sqlite3_close(db);
2203 db = NULL;
2204
2205 if (real)
2206 return;
2207
2208 if ('\0' == *tempfilename) {
2209 if (-1 == rename(MANDOC_DB "~", MANDOC_DB)) {
2210 exitcode = (int)MANDOCLEVEL_SYSERR;
2211 say(MANDOC_DB, "&rename");
2212 }
2213 return;
2214 }
2215
2216 switch (child = fork()) {
2217 case -1:
2218 exitcode = (int)MANDOCLEVEL_SYSERR;
2219 say("", "&fork cmp");
2220 return;
2221 case 0:
2222 execlp("cmp", "cmp", "-s",
2223 tempfilename, MANDOC_DB, (char *)NULL);
2224 say("", "&exec cmp");
2225 exit(0);
2226 default:
2227 break;
2228 }
2229 if (-1 == waitpid(child, &status, 0)) {
2230 exitcode = (int)MANDOCLEVEL_SYSERR;
2231 say("", "&wait cmp");
2232 } else if (WIFSIGNALED(status)) {
2233 exitcode = (int)MANDOCLEVEL_SYSERR;
2234 say("", "cmp died from signal %d", WTERMSIG(status));
2235 } else if (WEXITSTATUS(status)) {
2236 exitcode = (int)MANDOCLEVEL_SYSERR;
2237 say(MANDOC_DB,
2238 "Data changed, but cannot replace database");
2239 }
2240
2241 *strrchr(tempfilename, '/') = '\0';
2242 switch (child = fork()) {
2243 case -1:
2244 exitcode = (int)MANDOCLEVEL_SYSERR;
2245 say("", "&fork rm");
2246 return;
2247 case 0:
2248 execlp("rm", "rm", "-rf", tempfilename, (char *)NULL);
2249 say("", "&exec rm");
2250 exit((int)MANDOCLEVEL_SYSERR);
2251 default:
2252 break;
2253 }
2254 if (-1 == waitpid(child, &status, 0)) {
2255 exitcode = (int)MANDOCLEVEL_SYSERR;
2256 say("", "&wait rm");
2257 } else if (WIFSIGNALED(status) || WEXITSTATUS(status)) {
2258 exitcode = (int)MANDOCLEVEL_SYSERR;
2259 say("", "%s: Cannot remove temporary directory",
2260 tempfilename);
2261 }
2262 }
2263
2264 /*
2265 * This is straightforward stuff.
2266 * Open a database connection to a "temporary" database, then open a set
2267 * of prepared statements we'll use over and over again.
2268 * If "real" is set, we use the existing database; if not, we truncate a
2269 * temporary one.
2270 * Must be matched by dbclose().
2271 */
2272 static int
2273 dbopen(int real)
2274 {
2275 const char *sql;
2276 int rc, ofl;
2277
2278 if (nodb)
2279 return 1;
2280
2281 *tempfilename = '\0';
2282 ofl = SQLITE_OPEN_READWRITE;
2283
2284 if (real) {
2285 rc = sqlite3_open_v2(MANDOC_DB, &db, ofl, NULL);
2286 if (SQLITE_OK != rc) {
2287 exitcode = (int)MANDOCLEVEL_SYSERR;
2288 if (SQLITE_CANTOPEN != rc)
2289 say(MANDOC_DB, "%s", sqlite3_errstr(rc));
2290 return 0;
2291 }
2292 goto prepare_statements;
2293 }
2294
2295 ofl |= SQLITE_OPEN_CREATE | SQLITE_OPEN_EXCLUSIVE;
2296
2297 remove(MANDOC_DB "~");
2298 rc = sqlite3_open_v2(MANDOC_DB "~", &db, ofl, NULL);
2299 if (SQLITE_OK == rc)
2300 goto create_tables;
2301 if (MPARSE_QUICK & mparse_options) {
2302 exitcode = (int)MANDOCLEVEL_SYSERR;
2303 say(MANDOC_DB "~", "%s", sqlite3_errstr(rc));
2304 return 0;
2305 }
2306
2307 (void)strlcpy(tempfilename, "/tmp/mandocdb.XXXXXX",
2308 sizeof(tempfilename));
2309 if (NULL == mkdtemp(tempfilename)) {
2310 exitcode = (int)MANDOCLEVEL_SYSERR;
2311 say("", "&%s", tempfilename);
2312 return 0;
2313 }
2314 (void)strlcat(tempfilename, "/" MANDOC_DB,
2315 sizeof(tempfilename));
2316 rc = sqlite3_open_v2(tempfilename, &db, ofl, NULL);
2317 if (SQLITE_OK != rc) {
2318 exitcode = (int)MANDOCLEVEL_SYSERR;
2319 say("", "%s: %s", tempfilename, sqlite3_errstr(rc));
2320 return 0;
2321 }
2322
2323 create_tables:
2324 sql = "CREATE TABLE \"mpages\" (\n"
2325 " \"desc\" TEXT NOT NULL,\n"
2326 " \"form\" INTEGER NOT NULL,\n"
2327 " \"pageid\" INTEGER PRIMARY KEY AUTOINCREMENT NOT NULL\n"
2328 ");\n"
2329 "\n"
2330 "CREATE TABLE \"mlinks\" (\n"
2331 " \"sec\" TEXT NOT NULL,\n"
2332 " \"arch\" TEXT NOT NULL,\n"
2333 " \"name\" TEXT NOT NULL,\n"
2334 " \"pageid\" INTEGER NOT NULL REFERENCES mpages(pageid) "
2335 "ON DELETE CASCADE\n"
2336 ");\n"
2337 "CREATE INDEX mlinks_pageid_idx ON mlinks (pageid);\n"
2338 "\n"
2339 "CREATE TABLE \"names\" (\n"
2340 " \"bits\" INTEGER NOT NULL,\n"
2341 " \"name\" TEXT NOT NULL,\n"
2342 " \"pageid\" INTEGER NOT NULL REFERENCES mpages(pageid) "
2343 "ON DELETE CASCADE,\n"
2344 " UNIQUE (\"name\", \"pageid\") ON CONFLICT REPLACE\n"
2345 ");\n"
2346 "\n"
2347 "CREATE TABLE \"keys\" (\n"
2348 " \"bits\" INTEGER NOT NULL,\n"
2349 " \"key\" TEXT NOT NULL,\n"
2350 " \"pageid\" INTEGER NOT NULL REFERENCES mpages(pageid) "
2351 "ON DELETE CASCADE\n"
2352 ");\n"
2353 "CREATE INDEX keys_pageid_idx ON keys (pageid);\n";
2354
2355 if (SQLITE_OK != sqlite3_exec(db, sql, NULL, NULL, NULL)) {
2356 exitcode = (int)MANDOCLEVEL_SYSERR;
2357 say(MANDOC_DB, "%s", sqlite3_errmsg(db));
2358 sqlite3_close(db);
2359 return 0;
2360 }
2361
2362 prepare_statements:
2363 if (SQLITE_OK != sqlite3_exec(db,
2364 "PRAGMA foreign_keys = ON", NULL, NULL, NULL)) {
2365 exitcode = (int)MANDOCLEVEL_SYSERR;
2366 say(MANDOC_DB, "PRAGMA foreign_keys: %s",
2367 sqlite3_errmsg(db));
2368 sqlite3_close(db);
2369 return 0;
2370 }
2371
2372 sql = "DELETE FROM mpages WHERE pageid IN "
2373 "(SELECT pageid FROM mlinks WHERE "
2374 "sec=? AND arch=? AND name=?)";
2375 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_DELETE_PAGE], NULL);
2376 sql = "INSERT INTO mpages "
2377 "(desc,form) VALUES (?,?)";
2378 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_PAGE], NULL);
2379 sql = "INSERT INTO mlinks "
2380 "(sec,arch,name,pageid) VALUES (?,?,?,?)";
2381 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_LINK], NULL);
2382 sql = "SELECT bits FROM names where pageid = ?";
2383 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_SELECT_NAME], NULL);
2384 sql = "INSERT INTO names "
2385 "(bits,name,pageid) VALUES (?,?,?)";
2386 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_NAME], NULL);
2387 sql = "INSERT INTO keys "
2388 "(bits,key,pageid) VALUES (?,?,?)";
2389 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_KEY], NULL);
2390
2391 #ifndef __APPLE__
2392 /*
2393 * When opening a new database, we can turn off
2394 * synchronous mode for much better performance.
2395 */
2396
2397 if (real && SQLITE_OK != sqlite3_exec(db,
2398 "PRAGMA synchronous = OFF", NULL, NULL, NULL)) {
2399 exitcode = (int)MANDOCLEVEL_SYSERR;
2400 say(MANDOC_DB, "PRAGMA synchronous: %s",
2401 sqlite3_errmsg(db));
2402 sqlite3_close(db);
2403 return 0;
2404 }
2405 #endif
2406
2407 return 1;
2408 }
2409
2410 static int
2411 set_basedir(const char *targetdir, int report_baddir)
2412 {
2413 static char startdir[PATH_MAX];
2414 static int getcwd_status; /* 1 = ok, 2 = failure */
2415 static int chdir_status; /* 1 = changed directory */
2416 char *cp;
2417
2418 /*
2419 * Remember the original working directory, if possible.
2420 * This will be needed if the second or a later directory
2421 * on the command line is given as a relative path.
2422 * Do not error out if the current directory is not
2423 * searchable: Maybe it won't be needed after all.
2424 */
2425 if (0 == getcwd_status) {
2426 if (NULL == getcwd(startdir, sizeof(startdir))) {
2427 getcwd_status = 2;
2428 (void)strlcpy(startdir, strerror(errno),
2429 sizeof(startdir));
2430 } else
2431 getcwd_status = 1;
2432 }
2433
2434 /*
2435 * We are leaving the old base directory.
2436 * Do not use it any longer, not even for messages.
2437 */
2438 *basedir = '\0';
2439
2440 /*
2441 * If and only if the directory was changed earlier and
2442 * the next directory to process is given as a relative path,
2443 * first go back, or bail out if that is impossible.
2444 */
2445 if (chdir_status && '/' != *targetdir) {
2446 if (2 == getcwd_status) {
2447 exitcode = (int)MANDOCLEVEL_SYSERR;
2448 say("", "getcwd: %s", startdir);
2449 return 0;
2450 }
2451 if (-1 == chdir(startdir)) {
2452 exitcode = (int)MANDOCLEVEL_SYSERR;
2453 say("", "&chdir %s", startdir);
2454 return 0;
2455 }
2456 }
2457
2458 /*
2459 * Always resolve basedir to the canonicalized absolute
2460 * pathname and append a trailing slash, such that
2461 * we can reliably check whether files are inside.
2462 */
2463 if (NULL == realpath(targetdir, basedir)) {
2464 if (report_baddir || errno != ENOENT) {
2465 exitcode = (int)MANDOCLEVEL_BADARG;
2466 say("", "&%s: realpath", targetdir);
2467 }
2468 return 0;
2469 } else if (-1 == chdir(basedir)) {
2470 if (report_baddir || errno != ENOENT) {
2471 exitcode = (int)MANDOCLEVEL_BADARG;
2472 say("", "&chdir");
2473 }
2474 return 0;
2475 }
2476 chdir_status = 1;
2477 cp = strchr(basedir, '\0');
2478 if ('/' != cp[-1]) {
2479 if (cp - basedir >= PATH_MAX - 1) {
2480 exitcode = (int)MANDOCLEVEL_SYSERR;
2481 say("", "Filename too long");
2482 return 0;
2483 }
2484 *cp++ = '/';
2485 *cp = '\0';
2486 }
2487 return 1;
2488 }
2489
2490 static void
2491 say(const char *file, const char *format, ...)
2492 {
2493 va_list ap;
2494 int use_errno;
2495
2496 if ('\0' != *basedir)
2497 fprintf(stderr, "%s", basedir);
2498 if ('\0' != *basedir && '\0' != *file)
2499 fputc('/', stderr);
2500 if ('\0' != *file)
2501 fprintf(stderr, "%s", file);
2502
2503 use_errno = 1;
2504 if (NULL != format) {
2505 switch (*format) {
2506 case '&':
2507 format++;
2508 break;
2509 case '\0':
2510 format = NULL;
2511 break;
2512 default:
2513 use_errno = 0;
2514 break;
2515 }
2516 }
2517 if (NULL != format) {
2518 if ('\0' != *basedir || '\0' != *file)
2519 fputs(": ", stderr);
2520 va_start(ap, format);
2521 vfprintf(stderr, format, ap);
2522 va_end(ap);
2523 }
2524 if (use_errno) {
2525 if ('\0' != *basedir || '\0' != *file || NULL != format)
2526 fputs(": ", stderr);
2527 perror(NULL);
2528 } else
2529 fputc('\n', stderr);
2530 }
A git conversion mirror of mandoc.bsd.lv