]> git.cameronkatri.com Git - mandoc.git/blob - mandocdb.c
git.ckatri.com / mandoc.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
history | raw | HEAD
Integrate the makewhatis binary into the mandoc binary
[mandoc.git] / mandocdb.c
1 /* $Id: mandocdb.c,v 1.179 2014/12/09 07:29:42 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 const struct mpage *mpage; /* if set, the owning parse */
88 uint64_t mask; /* bitmask in sequence */
89 char key[]; /* rendered text */
90 };
91
92 struct inodev {
93 ino_t st_ino;
94 dev_t st_dev;
95 };
96
97 struct mpage {
98 struct inodev inodev; /* used for hashing routine */
99 int64_t pageid; /* pageid in mpages SQL table */
100 char *sec; /* section from file content */
101 char *arch; /* architecture from file content */
102 char *title; /* title from file content */
103 char *desc; /* description from file content */
104 struct mlink *mlinks; /* singly linked list */
105 int form; /* format from file content */
106 int name_head_done;
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_SELECT_NAME, /* retrieve existing name flags */
128 STMT_INSERT_KEY, /* insert parsed key */
129 STMT__MAX
130 };
131
132 typedef int (*mdoc_fp)(struct mpage *, const struct mdoc_meta *,
133 const struct mdoc_node *);
134
135 struct mdoc_handler {
136 mdoc_fp fp; /* optional handler */
137 uint64_t mask; /* set unless handler returns 0 */
138 };
139
140 static void dbclose(int);
141 static void dbadd(struct mpage *);
142 static void dbadd_mlink(const struct mlink *mlink);
143 static void dbadd_mlink_name(const struct mlink *mlink);
144 static int dbopen(int);
145 static void dbprune(void);
146 static void filescan(const char *);
147 static void *hash_alloc(size_t, void *);
148 static void hash_free(void *, void *);
149 static void *hash_calloc(size_t, size_t, void *);
150 static void mlink_add(struct mlink *, const struct stat *);
151 static void mlink_check(struct mpage *, struct mlink *);
152 static void mlink_free(struct mlink *);
153 static void mlinks_undupe(struct mpage *);
154 static void mpages_free(void);
155 static void mpages_merge(struct mparse *);
156 static void names_check(void);
157 static void parse_cat(struct mpage *, int);
158 static void parse_man(struct mpage *, const struct man_meta *,
159 const struct man_node *);
160 static void parse_mdoc(struct mpage *, const struct mdoc_meta *,
161 const struct mdoc_node *);
162 static int parse_mdoc_body(struct mpage *, const struct mdoc_meta *,
163 const struct mdoc_node *);
164 static int parse_mdoc_head(struct mpage *, const struct mdoc_meta *,
165 const struct mdoc_node *);
166 static int parse_mdoc_Fd(struct mpage *, const struct mdoc_meta *,
167 const struct mdoc_node *);
168 static void parse_mdoc_fname(struct mpage *, const struct mdoc_node *);
169 static int parse_mdoc_Fn(struct mpage *, const struct mdoc_meta *,
170 const struct mdoc_node *);
171 static int parse_mdoc_Fo(struct mpage *, const struct mdoc_meta *,
172 const struct mdoc_node *);
173 static int parse_mdoc_Nd(struct mpage *, const struct mdoc_meta *,
174 const struct mdoc_node *);
175 static int parse_mdoc_Nm(struct mpage *, const struct mdoc_meta *,
176 const struct mdoc_node *);
177 static int parse_mdoc_Sh(struct mpage *, const struct mdoc_meta *,
178 const struct mdoc_node *);
179 static int parse_mdoc_Xr(struct mpage *, const struct mdoc_meta *,
180 const struct mdoc_node *);
181 static void putkey(const struct mpage *, char *, uint64_t);
182 static void putkeys(const struct mpage *, char *, size_t, uint64_t);
183 static void putmdockey(const struct mpage *,
184 const struct mdoc_node *, uint64_t);
185 static int render_string(char **, size_t *);
186 static void say(const char *, const char *, ...);
187 static int set_basedir(const char *, int);
188 static int treescan(void);
189 static size_t utf8(unsigned int, char [7]);
190
191 static char tempfilename[32];
192 static char *progname;
193 static int nodb; /* no database changes */
194 static int mparse_options; /* abort the parse early */
195 static int use_all; /* use all found files */
196 static int debug; /* print what we're doing */
197 static int warnings; /* warn about crap */
198 static int write_utf8; /* write UTF-8 output; else ASCII */
199 static int exitcode; /* to be returned by main */
200 static enum op op; /* operational mode */
201 static char basedir[PATH_MAX]; /* current base directory */
202 static struct mchars *mchars; /* table of named characters */
203 static struct ohash mpages; /* table of distinct manual pages */
204 static struct ohash mlinks; /* table of directory entries */
205 static struct ohash names; /* table of all names */
206 static struct ohash strings; /* table of all strings */
207 static sqlite3 *db = NULL; /* current database */
208 static sqlite3_stmt *stmts[STMT__MAX]; /* current statements */
209 static uint64_t name_mask;
210
211 static const struct mdoc_handler mdocs[MDOC_MAX] = {
212 { NULL, 0 }, /* Ap */
213 { NULL, 0 }, /* Dd */
214 { NULL, 0 }, /* Dt */
215 { NULL, 0 }, /* Os */
216 { parse_mdoc_Sh, TYPE_Sh }, /* Sh */
217 { parse_mdoc_head, TYPE_Ss }, /* Ss */
218 { NULL, 0 }, /* Pp */
219 { NULL, 0 }, /* D1 */
220 { NULL, 0 }, /* Dl */
221 { NULL, 0 }, /* Bd */
222 { NULL, 0 }, /* Ed */
223 { NULL, 0 }, /* Bl */
224 { NULL, 0 }, /* El */
225 { NULL, 0 }, /* It */
226 { NULL, 0 }, /* Ad */
227 { NULL, TYPE_An }, /* An */
228 { NULL, TYPE_Ar }, /* Ar */
229 { NULL, TYPE_Cd }, /* Cd */
230 { NULL, TYPE_Cm }, /* Cm */
231 { NULL, TYPE_Dv }, /* Dv */
232 { NULL, TYPE_Er }, /* Er */
233 { NULL, TYPE_Ev }, /* Ev */
234 { NULL, 0 }, /* Ex */
235 { NULL, TYPE_Fa }, /* Fa */
236 { parse_mdoc_Fd, 0 }, /* Fd */
237 { NULL, TYPE_Fl }, /* Fl */
238 { parse_mdoc_Fn, 0 }, /* Fn */
239 { NULL, TYPE_Ft }, /* Ft */
240 { NULL, TYPE_Ic }, /* Ic */
241 { NULL, TYPE_In }, /* In */
242 { NULL, TYPE_Li }, /* Li */
243 { parse_mdoc_Nd, 0 }, /* Nd */
244 { parse_mdoc_Nm, 0 }, /* Nm */
245 { NULL, 0 }, /* Op */
246 { NULL, 0 }, /* Ot */
247 { NULL, TYPE_Pa }, /* Pa */
248 { NULL, 0 }, /* Rv */
249 { NULL, TYPE_St }, /* St */
250 { NULL, TYPE_Va }, /* Va */
251 { parse_mdoc_body, TYPE_Va }, /* Vt */
252 { parse_mdoc_Xr, 0 }, /* Xr */
253 { NULL, 0 }, /* %A */
254 { NULL, 0 }, /* %B */
255 { NULL, 0 }, /* %D */
256 { NULL, 0 }, /* %I */
257 { NULL, 0 }, /* %J */
258 { NULL, 0 }, /* %N */
259 { NULL, 0 }, /* %O */
260 { NULL, 0 }, /* %P */
261 { NULL, 0 }, /* %R */
262 { NULL, 0 }, /* %T */
263 { NULL, 0 }, /* %V */
264 { NULL, 0 }, /* Ac */
265 { NULL, 0 }, /* Ao */
266 { NULL, 0 }, /* Aq */
267 { NULL, TYPE_At }, /* At */
268 { NULL, 0 }, /* Bc */
269 { NULL, 0 }, /* Bf */
270 { NULL, 0 }, /* Bo */
271 { NULL, 0 }, /* Bq */
272 { NULL, TYPE_Bsx }, /* Bsx */
273 { NULL, TYPE_Bx }, /* Bx */
274 { NULL, 0 }, /* Db */
275 { NULL, 0 }, /* Dc */
276 { NULL, 0 }, /* Do */
277 { NULL, 0 }, /* Dq */
278 { NULL, 0 }, /* Ec */
279 { NULL, 0 }, /* Ef */
280 { NULL, TYPE_Em }, /* Em */
281 { NULL, 0 }, /* Eo */
282 { NULL, TYPE_Fx }, /* Fx */
283 { NULL, TYPE_Ms }, /* Ms */
284 { NULL, 0 }, /* No */
285 { NULL, 0 }, /* Ns */
286 { NULL, TYPE_Nx }, /* Nx */
287 { NULL, TYPE_Ox }, /* Ox */
288 { NULL, 0 }, /* Pc */
289 { NULL, 0 }, /* Pf */
290 { NULL, 0 }, /* Po */
291 { NULL, 0 }, /* Pq */
292 { NULL, 0 }, /* Qc */
293 { NULL, 0 }, /* Ql */
294 { NULL, 0 }, /* Qo */
295 { NULL, 0 }, /* Qq */
296 { NULL, 0 }, /* Re */
297 { NULL, 0 }, /* Rs */
298 { NULL, 0 }, /* Sc */
299 { NULL, 0 }, /* So */
300 { NULL, 0 }, /* Sq */
301 { NULL, 0 }, /* Sm */
302 { NULL, 0 }, /* Sx */
303 { NULL, TYPE_Sy }, /* Sy */
304 { NULL, TYPE_Tn }, /* Tn */
305 { NULL, 0 }, /* Ux */
306 { NULL, 0 }, /* Xc */
307 { NULL, 0 }, /* Xo */
308 { parse_mdoc_Fo, 0 }, /* Fo */
309 { NULL, 0 }, /* Fc */
310 { NULL, 0 }, /* Oo */
311 { NULL, 0 }, /* Oc */
312 { NULL, 0 }, /* Bk */
313 { NULL, 0 }, /* Ek */
314 { NULL, 0 }, /* Bt */
315 { NULL, 0 }, /* Hf */
316 { NULL, 0 }, /* Fr */
317 { NULL, 0 }, /* Ud */
318 { NULL, TYPE_Lb }, /* Lb */
319 { NULL, 0 }, /* Lp */
320 { NULL, TYPE_Lk }, /* Lk */
321 { NULL, TYPE_Mt }, /* Mt */
322 { NULL, 0 }, /* Brq */
323 { NULL, 0 }, /* Bro */
324 { NULL, 0 }, /* Brc */
325 { NULL, 0 }, /* %C */
326 { NULL, 0 }, /* Es */
327 { NULL, 0 }, /* En */
328 { NULL, TYPE_Dx }, /* Dx */
329 { NULL, 0 }, /* %Q */
330 { NULL, 0 }, /* br */
331 { NULL, 0 }, /* sp */
332 { NULL, 0 }, /* %U */
333 { NULL, 0 }, /* Ta */
334 { NULL, 0 }, /* ll */
335 };
336
337
338 int
339 mandocdb(int argc, char *argv[])
340 {
341 int ch, i;
342 size_t j, sz;
343 const char *path_arg;
344 struct manpaths dirs;
345 struct mparse *mp;
346 struct ohash_info mpages_info, mlinks_info;
347
348 memset(stmts, 0, STMT__MAX * sizeof(sqlite3_stmt *));
349 memset(&dirs, 0, sizeof(struct manpaths));
350
351 mpages_info.alloc = mlinks_info.alloc = hash_alloc;
352 mpages_info.calloc = mlinks_info.calloc = hash_calloc;
353 mpages_info.free = mlinks_info.free = hash_free;
354
355 mpages_info.key_offset = offsetof(struct mpage, inodev);
356 mlinks_info.key_offset = offsetof(struct mlink, file);
357
358 progname = strrchr(argv[0], '/');
359 if (progname == NULL)
360 progname = argv[0];
361 else
362 ++progname;
363
364 /*
365 * We accept a few different invocations.
366 * The CHECKOP macro makes sure that invocation styles don't
367 * clobber each other.
368 */
369 #define CHECKOP(_op, _ch) do \
370 if (OP_DEFAULT != (_op)) { \
371 fprintf(stderr, "%s: -%c: Conflicting option\n", \
372 progname, (_ch)); \
373 goto usage; \
374 } while (/*CONSTCOND*/0)
375
376 path_arg = NULL;
377 op = OP_DEFAULT;
378
379 while (-1 != (ch = getopt(argc, argv, "aC:Dd:npQT:tu:v")))
380 switch (ch) {
381 case 'a':
382 use_all = 1;
383 break;
384 case 'C':
385 CHECKOP(op, ch);
386 path_arg = optarg;
387 op = OP_CONFFILE;
388 break;
389 case 'D':
390 debug++;
391 break;
392 case 'd':
393 CHECKOP(op, ch);
394 path_arg = optarg;
395 op = OP_UPDATE;
396 break;
397 case 'n':
398 nodb = 1;
399 break;
400 case 'p':
401 warnings = 1;
402 break;
403 case 'Q':
404 mparse_options |= MPARSE_QUICK;
405 break;
406 case 'T':
407 if (strcmp(optarg, "utf8")) {
408 fprintf(stderr, "%s: -T%s: "
409 "Unsupported output format\n",
410 progname, optarg);
411 goto usage;
412 }
413 write_utf8 = 1;
414 break;
415 case 't':
416 CHECKOP(op, ch);
417 dup2(STDOUT_FILENO, STDERR_FILENO);
418 op = OP_TEST;
419 nodb = warnings = 1;
420 break;
421 case 'u':
422 CHECKOP(op, ch);
423 path_arg = optarg;
424 op = OP_DELETE;
425 break;
426 case 'v':
427 /* Compatibility with espie@'s makewhatis. */
428 break;
429 default:
430 goto usage;
431 }
432
433 argc -= optind;
434 argv += optind;
435
436 if (OP_CONFFILE == op && argc > 0) {
437 fprintf(stderr, "%s: -C: Too many arguments\n",
438 progname);
439 goto usage;
440 }
441
442 exitcode = (int)MANDOCLEVEL_OK;
443 mchars = mchars_alloc();
444 mp = mparse_alloc(mparse_options, MANDOCLEVEL_FATAL, NULL,
445 mchars, NULL);
446 ohash_init(&mpages, 6, &mpages_info);
447 ohash_init(&mlinks, 6, &mlinks_info);
448
449 if (OP_UPDATE == op || OP_DELETE == op || OP_TEST == op) {
450
451 /*
452 * Most of these deal with a specific directory.
453 * Jump into that directory first.
454 */
455 if (OP_TEST != op && 0 == set_basedir(path_arg, 1))
456 goto out;
457
458 if (dbopen(1)) {
459 /*
460 * The existing database is usable. Process
461 * all files specified on the command-line.
462 */
463 use_all = 1;
464 for (i = 0; i < argc; i++)
465 filescan(argv[i]);
466 if (OP_TEST != op)
467 dbprune();
468 } else {
469 /*
470 * Database missing or corrupt.
471 * Recreate from scratch.
472 */
473 exitcode = (int)MANDOCLEVEL_OK;
474 op = OP_DEFAULT;
475 if (0 == treescan())
476 goto out;
477 if (0 == dbopen(0))
478 goto out;
479 }
480 if (OP_DELETE != op)
481 mpages_merge(mp);
482 dbclose(OP_DEFAULT == op ? 0 : 1);
483 } else {
484 /*
485 * If we have arguments, use them as our manpaths.
486 * If we don't, grok from manpath(1) or however else
487 * manpath_parse() wants to do it.
488 */
489 if (argc > 0) {
490 dirs.paths = mandoc_reallocarray(NULL,
491 argc, sizeof(char *));
492 dirs.sz = (size_t)argc;
493 for (i = 0; i < argc; i++)
494 dirs.paths[i] = mandoc_strdup(argv[i]);
495 } else
496 manpath_parse(&dirs, path_arg, NULL, NULL);
497
498 if (0 == dirs.sz) {
499 exitcode = (int)MANDOCLEVEL_BADARG;
500 say("", "Empty manpath");
501 }
502
503 /*
504 * First scan the tree rooted at a base directory, then
505 * build a new database and finally move it into place.
506 * Ignore zero-length directories and strip trailing
507 * slashes.
508 */
509 for (j = 0; j < dirs.sz; j++) {
510 sz = strlen(dirs.paths[j]);
511 if (sz && '/' == dirs.paths[j][sz - 1])
512 dirs.paths[j][--sz] = '\0';
513 if (0 == sz)
514 continue;
515
516 if (j) {
517 ohash_init(&mpages, 6, &mpages_info);
518 ohash_init(&mlinks, 6, &mlinks_info);
519 }
520
521 if (0 == set_basedir(dirs.paths[j], argc > 0))
522 continue;
523 if (0 == treescan())
524 continue;
525 if (0 == dbopen(0))
526 continue;
527
528 mpages_merge(mp);
529 if (warnings && !nodb &&
530 ! (MPARSE_QUICK & mparse_options))
531 names_check();
532 dbclose(0);
533
534 if (j + 1 < dirs.sz) {
535 mpages_free();
536 ohash_delete(&mpages);
537 ohash_delete(&mlinks);
538 }
539 }
540 }
541 out:
542 manpath_free(&dirs);
543 mparse_free(mp);
544 mchars_free(mchars);
545 mpages_free();
546 ohash_delete(&mpages);
547 ohash_delete(&mlinks);
548 return(exitcode);
549 usage:
550 fprintf(stderr, "usage: %s [-aDnpQ] [-C file] [-Tutf8]\n"
551 " %s [-aDnpQ] [-Tutf8] dir ...\n"
552 " %s [-DnpQ] [-Tutf8] -d dir [file ...]\n"
553 " %s [-Dnp] -u dir [file ...]\n"
554 " %s [-Q] -t file ...\n",
555 progname, progname, progname,
556 progname, progname);
557
558 return((int)MANDOCLEVEL_BADARG);
559 }
560
561 /*
562 * Scan a directory tree rooted at "basedir" for manpages.
563 * We use fts(), scanning directory parts along the way for clues to our
564 * section and architecture.
565 *
566 * If use_all has been specified, grok all files.
567 * If not, sanitise paths to the following:
568 *
569 * [./]man*[/<arch>]/<name>.<section>
570 * or
571 * [./]cat<section>[/<arch>]/<name>.0
572 *
573 * TODO: accomodate for multi-language directories.
574 */
575 static int
576 treescan(void)
577 {
578 char buf[PATH_MAX];
579 FTS *f;
580 FTSENT *ff;
581 struct mlink *mlink;
582 int dform, gzip;
583 char *dsec, *arch, *fsec, *cp;
584 const char *path;
585 const char *argv[2];
586
587 argv[0] = ".";
588 argv[1] = (char *)NULL;
589
590 f = fts_open((char * const *)argv,
591 FTS_PHYSICAL | FTS_NOCHDIR, NULL);
592 if (NULL == f) {
593 exitcode = (int)MANDOCLEVEL_SYSERR;
594 say("", "&fts_open");
595 return(0);
596 }
597
598 dsec = arch = NULL;
599 dform = FORM_NONE;
600
601 while (NULL != (ff = fts_read(f))) {
602 path = ff->fts_path + 2;
603 switch (ff->fts_info) {
604
605 /*
606 * Symbolic links require various sanity checks,
607 * then get handled just like regular files.
608 */
609 case FTS_SL:
610 if (NULL == realpath(path, buf)) {
611 if (warnings)
612 say(path, "&realpath");
613 continue;
614 }
615 if (strstr(buf, basedir) != buf) {
616 if (warnings) say("",
617 "%s: outside base directory", buf);
618 continue;
619 }
620 /* Use logical inode to avoid mpages dupe. */
621 if (-1 == stat(path, ff->fts_statp)) {
622 if (warnings)
623 say(path, "&stat");
624 continue;
625 }
626 /* FALLTHROUGH */
627
628 /*
629 * If we're a regular file, add an mlink by using the
630 * stored directory data and handling the filename.
631 */
632 case FTS_F:
633 if (0 == strcmp(path, MANDOC_DB))
634 continue;
635 if ( ! use_all && ff->fts_level < 2) {
636 if (warnings)
637 say(path, "Extraneous file");
638 continue;
639 }
640 gzip = 0;
641 fsec = NULL;
642 while (NULL == fsec) {
643 fsec = strrchr(ff->fts_name, '.');
644 if (NULL == fsec || strcmp(fsec+1, "gz"))
645 break;
646 gzip = 1;
647 *fsec = '\0';
648 fsec = NULL;
649 }
650 if (NULL == fsec) {
651 if ( ! use_all) {
652 if (warnings)
653 say(path,
654 "No filename suffix");
655 continue;
656 }
657 } else if (0 == strcmp(++fsec, "html")) {
658 if (warnings)
659 say(path, "Skip html");
660 continue;
661 } else if (0 == strcmp(fsec, "ps")) {
662 if (warnings)
663 say(path, "Skip ps");
664 continue;
665 } else if (0 == strcmp(fsec, "pdf")) {
666 if (warnings)
667 say(path, "Skip pdf");
668 continue;
669 } else if ( ! use_all &&
670 ((FORM_SRC == dform && strcmp(fsec, dsec)) ||
671 (FORM_CAT == dform && strcmp(fsec, "0")))) {
672 if (warnings)
673 say(path, "Wrong filename suffix");
674 continue;
675 } else
676 fsec[-1] = '\0';
677
678 mlink = mandoc_calloc(1, sizeof(struct mlink));
679 if (strlcpy(mlink->file, path,
680 sizeof(mlink->file)) >=
681 sizeof(mlink->file)) {
682 say(path, "Filename too long");
683 free(mlink);
684 continue;
685 }
686 mlink->dform = dform;
687 mlink->dsec = dsec;
688 mlink->arch = arch;
689 mlink->name = ff->fts_name;
690 mlink->fsec = fsec;
691 mlink->gzip = gzip;
692 mlink_add(mlink, ff->fts_statp);
693 continue;
694
695 case FTS_D:
696 /* FALLTHROUGH */
697 case FTS_DP:
698 break;
699
700 default:
701 if (warnings)
702 say(path, "Not a regular file");
703 continue;
704 }
705
706 switch (ff->fts_level) {
707 case 0:
708 /* Ignore the root directory. */
709 break;
710 case 1:
711 /*
712 * This might contain manX/ or catX/.
713 * Try to infer this from the name.
714 * If we're not in use_all, enforce it.
715 */
716 cp = ff->fts_name;
717 if (FTS_DP == ff->fts_info)
718 break;
719
720 if (0 == strncmp(cp, "man", 3)) {
721 dform = FORM_SRC;
722 dsec = cp + 3;
723 } else if (0 == strncmp(cp, "cat", 3)) {
724 dform = FORM_CAT;
725 dsec = cp + 3;
726 } else {
727 dform = FORM_NONE;
728 dsec = NULL;
729 }
730
731 if (NULL != dsec || use_all)
732 break;
733
734 if (warnings)
735 say(path, "Unknown directory part");
736 fts_set(f, ff, FTS_SKIP);
737 break;
738 case 2:
739 /*
740 * Possibly our architecture.
741 * If we're descending, keep tabs on it.
742 */
743 if (FTS_DP != ff->fts_info && NULL != dsec)
744 arch = ff->fts_name;
745 else
746 arch = NULL;
747 break;
748 default:
749 if (FTS_DP == ff->fts_info || use_all)
750 break;
751 if (warnings)
752 say(path, "Extraneous directory part");
753 fts_set(f, ff, FTS_SKIP);
754 break;
755 }
756 }
757
758 fts_close(f);
759 return(1);
760 }
761
762 /*
763 * Add a file to the mlinks table.
764 * Do not verify that it's a "valid" looking manpage (we'll do that
765 * later).
766 *
767 * Try to infer the manual section, architecture, and page name from the
768 * path, assuming it looks like
769 *
770 * [./]man*[/<arch>]/<name>.<section>
771 * or
772 * [./]cat<section>[/<arch>]/<name>.0
773 *
774 * See treescan() for the fts(3) version of this.
775 */
776 static void
777 filescan(const char *file)
778 {
779 char buf[PATH_MAX];
780 struct stat st;
781 struct mlink *mlink;
782 char *p, *start;
783
784 assert(use_all);
785
786 if (0 == strncmp(file, "./", 2))
787 file += 2;
788
789 /*
790 * We have to do lstat(2) before realpath(3) loses
791 * the information whether this is a symbolic link.
792 * We need to know that because for symbolic links,
793 * we want to use the orginal file name, while for
794 * regular files, we want to use the real path.
795 */
796 if (-1 == lstat(file, &st)) {
797 exitcode = (int)MANDOCLEVEL_BADARG;
798 say(file, "&lstat");
799 return;
800 } else if (0 == ((S_IFREG | S_IFLNK) & st.st_mode)) {
801 exitcode = (int)MANDOCLEVEL_BADARG;
802 say(file, "Not a regular file");
803 return;
804 }
805
806 /*
807 * We have to resolve the file name to the real path
808 * in any case for the base directory check.
809 */
810 if (NULL == realpath(file, buf)) {
811 exitcode = (int)MANDOCLEVEL_BADARG;
812 say(file, "&realpath");
813 return;
814 }
815
816 if (OP_TEST == op)
817 start = buf;
818 else if (strstr(buf, basedir) == buf)
819 start = buf + strlen(basedir);
820 else {
821 exitcode = (int)MANDOCLEVEL_BADARG;
822 say("", "%s: outside base directory", buf);
823 return;
824 }
825
826 /*
827 * Now we are sure the file is inside our tree.
828 * If it is a symbolic link, ignore the real path
829 * and use the original name.
830 * This implies passing stuff like "cat1/../man1/foo.1"
831 * on the command line won't work. So don't do that.
832 * Note the stat(2) can still fail if the link target
833 * doesn't exist.
834 */
835 if (S_IFLNK & st.st_mode) {
836 if (-1 == stat(buf, &st)) {
837 exitcode = (int)MANDOCLEVEL_BADARG;
838 say(file, "&stat");
839 return;
840 }
841 if (strlcpy(buf, file, sizeof(buf)) >= sizeof(buf)) {
842 say(file, "Filename too long");
843 return;
844 }
845 start = buf;
846 if (OP_TEST != op && strstr(buf, basedir) == buf)
847 start += strlen(basedir);
848 }
849
850 mlink = mandoc_calloc(1, sizeof(struct mlink));
851 mlink->dform = FORM_NONE;
852 if (strlcpy(mlink->file, start, sizeof(mlink->file)) >=
853 sizeof(mlink->file)) {
854 say(start, "Filename too long");
855 return;
856 }
857
858 /*
859 * First try to guess our directory structure.
860 * If we find a separator, try to look for man* or cat*.
861 * If we find one of these and what's underneath is a directory,
862 * assume it's an architecture.
863 */
864 if (NULL != (p = strchr(start, '/'))) {
865 *p++ = '\0';
866 if (0 == strncmp(start, "man", 3)) {
867 mlink->dform = FORM_SRC;
868 mlink->dsec = start + 3;
869 } else if (0 == strncmp(start, "cat", 3)) {
870 mlink->dform = FORM_CAT;
871 mlink->dsec = start + 3;
872 }
873
874 start = p;
875 if (NULL != mlink->dsec && NULL != (p = strchr(start, '/'))) {
876 *p++ = '\0';
877 mlink->arch = start;
878 start = p;
879 }
880 }
881
882 /*
883 * Now check the file suffix.
884 * Suffix of `.0' indicates a catpage, `.1-9' is a manpage.
885 */
886 p = strrchr(start, '\0');
887 while (p-- > start && '/' != *p && '.' != *p)
888 /* Loop. */ ;
889
890 if ('.' == *p) {
891 *p++ = '\0';
892 mlink->fsec = p;
893 }
894
895 /*
896 * Now try to parse the name.
897 * Use the filename portion of the path.
898 */
899 mlink->name = start;
900 if (NULL != (p = strrchr(start, '/'))) {
901 mlink->name = p + 1;
902 *p = '\0';
903 }
904 mlink_add(mlink, &st);
905 }
906
907 static void
908 mlink_add(struct mlink *mlink, const struct stat *st)
909 {
910 struct inodev inodev;
911 struct mpage *mpage;
912 unsigned int slot;
913
914 assert(NULL != mlink->file);
915
916 mlink->dsec = mandoc_strdup(mlink->dsec ? mlink->dsec : "");
917 mlink->arch = mandoc_strdup(mlink->arch ? mlink->arch : "");
918 mlink->name = mandoc_strdup(mlink->name ? mlink->name : "");
919 mlink->fsec = mandoc_strdup(mlink->fsec ? mlink->fsec : "");
920
921 if ('0' == *mlink->fsec) {
922 free(mlink->fsec);
923 mlink->fsec = mandoc_strdup(mlink->dsec);
924 mlink->fform = FORM_CAT;
925 } else if ('1' <= *mlink->fsec && '9' >= *mlink->fsec)
926 mlink->fform = FORM_SRC;
927 else
928 mlink->fform = FORM_NONE;
929
930 slot = ohash_qlookup(&mlinks, mlink->file);
931 assert(NULL == ohash_find(&mlinks, slot));
932 ohash_insert(&mlinks, slot, mlink);
933
934 inodev.st_ino = st->st_ino;
935 inodev.st_dev = st->st_dev;
936 slot = ohash_lookup_memory(&mpages, (char *)&inodev,
937 sizeof(struct inodev), inodev.st_ino);
938 mpage = ohash_find(&mpages, slot);
939 if (NULL == mpage) {
940 mpage = mandoc_calloc(1, sizeof(struct mpage));
941 mpage->inodev.st_ino = inodev.st_ino;
942 mpage->inodev.st_dev = inodev.st_dev;
943 ohash_insert(&mpages, slot, mpage);
944 } else
945 mlink->next = mpage->mlinks;
946 mpage->mlinks = mlink;
947 mlink->mpage = mpage;
948 }
949
950 static void
951 mlink_free(struct mlink *mlink)
952 {
953
954 free(mlink->dsec);
955 free(mlink->arch);
956 free(mlink->name);
957 free(mlink->fsec);
958 free(mlink);
959 }
960
961 static void
962 mpages_free(void)
963 {
964 struct mpage *mpage;
965 struct mlink *mlink;
966 unsigned int slot;
967
968 mpage = ohash_first(&mpages, &slot);
969 while (NULL != mpage) {
970 while (NULL != (mlink = mpage->mlinks)) {
971 mpage->mlinks = mlink->next;
972 mlink_free(mlink);
973 }
974 free(mpage->sec);
975 free(mpage->arch);
976 free(mpage->title);
977 free(mpage->desc);
978 free(mpage);
979 mpage = ohash_next(&mpages, &slot);
980 }
981 }
982
983 /*
984 * For each mlink to the mpage, check whether the path looks like
985 * it is formatted, and if it does, check whether a source manual
986 * exists by the same name, ignoring the suffix.
987 * If both conditions hold, drop the mlink.
988 */
989 static void
990 mlinks_undupe(struct mpage *mpage)
991 {
992 char buf[PATH_MAX];
993 struct mlink **prev;
994 struct mlink *mlink;
995 char *bufp;
996
997 mpage->form = FORM_CAT;
998 prev = &mpage->mlinks;
999 while (NULL != (mlink = *prev)) {
1000 if (FORM_CAT != mlink->dform) {
1001 mpage->form = FORM_NONE;
1002 goto nextlink;
1003 }
1004 (void)strlcpy(buf, mlink->file, sizeof(buf));
1005 bufp = strstr(buf, "cat");
1006 assert(NULL != bufp);
1007 memcpy(bufp, "man", 3);
1008 if (NULL != (bufp = strrchr(buf, '.')))
1009 *++bufp = '\0';
1010 (void)strlcat(buf, mlink->dsec, sizeof(buf));
1011 if (NULL == ohash_find(&mlinks,
1012 ohash_qlookup(&mlinks, buf)))
1013 goto nextlink;
1014 if (warnings)
1015 say(mlink->file, "Man source exists: %s", buf);
1016 if (use_all)
1017 goto nextlink;
1018 *prev = mlink->next;
1019 mlink_free(mlink);
1020 continue;
1021 nextlink:
1022 prev = &(*prev)->next;
1023 }
1024 }
1025
1026 static void
1027 mlink_check(struct mpage *mpage, struct mlink *mlink)
1028 {
1029 struct str *str;
1030 unsigned int slot;
1031
1032 /*
1033 * Check whether the manual section given in a file
1034 * agrees with the directory where the file is located.
1035 * Some manuals have suffixes like (3p) on their
1036 * section number either inside the file or in the
1037 * directory name, some are linked into more than one
1038 * section, like encrypt(1) = makekey(8).
1039 */
1040
1041 if (FORM_SRC == mpage->form &&
1042 strcasecmp(mpage->sec, mlink->dsec))
1043 say(mlink->file, "Section \"%s\" manual in %s directory",
1044 mpage->sec, mlink->dsec);
1045
1046 /*
1047 * Manual page directories exist for each kernel
1048 * architecture as returned by machine(1).
1049 * However, many manuals only depend on the
1050 * application architecture as returned by arch(1).
1051 * For example, some (2/ARM) manuals are shared
1052 * across the "armish" and "zaurus" kernel
1053 * architectures.
1054 * A few manuals are even shared across completely
1055 * different architectures, for example fdformat(1)
1056 * on amd64, i386, sparc, and sparc64.
1057 */
1058
1059 if (strcasecmp(mpage->arch, mlink->arch))
1060 say(mlink->file, "Architecture \"%s\" manual in "
1061 "\"%s\" directory", mpage->arch, mlink->arch);
1062
1063 /*
1064 * XXX
1065 * parse_cat() doesn't set NAME_TITLE yet.
1066 */
1067
1068 if (FORM_CAT == mpage->form)
1069 return;
1070
1071 /*
1072 * Check whether this mlink
1073 * appears as a name in the NAME section.
1074 */
1075
1076 slot = ohash_qlookup(&names, mlink->name);
1077 str = ohash_find(&names, slot);
1078 assert(NULL != str);
1079 if ( ! (NAME_TITLE & str->mask))
1080 say(mlink->file, "Name missing in NAME section");
1081 }
1082
1083 /*
1084 * Run through the files in the global vector "mpages"
1085 * and add them to the database specified in "basedir".
1086 *
1087 * This handles the parsing scheme itself, using the cues of directory
1088 * and filename to determine whether the file is parsable or not.
1089 */
1090 static void
1091 mpages_merge(struct mparse *mp)
1092 {
1093 char any[] = "any";
1094 struct ohash_info str_info;
1095 struct mpage *mpage, *mpage_dest;
1096 struct mlink *mlink, *mlink_dest;
1097 struct mdoc *mdoc;
1098 struct man *man;
1099 char *sodest;
1100 char *cp;
1101 int fd;
1102 unsigned int pslot;
1103 enum mandoclevel lvl;
1104
1105 str_info.alloc = hash_alloc;
1106 str_info.calloc = hash_calloc;
1107 str_info.free = hash_free;
1108 str_info.key_offset = offsetof(struct str, key);
1109
1110 if ( ! nodb)
1111 SQL_EXEC("BEGIN TRANSACTION");
1112
1113 mpage = ohash_first(&mpages, &pslot);
1114 while (mpage != NULL) {
1115 mlinks_undupe(mpage);
1116 if (mpage->mlinks == NULL) {
1117 mpage = ohash_next(&mpages, &pslot);
1118 continue;
1119 }
1120
1121 name_mask = NAME_MASK;
1122 ohash_init(&names, 4, &str_info);
1123 ohash_init(&strings, 6, &str_info);
1124 mparse_reset(mp);
1125 mdoc = NULL;
1126 man = NULL;
1127 sodest = NULL;
1128
1129 mparse_open(mp, &fd, mpage->mlinks->file);
1130 if (fd == -1) {
1131 say(mpage->mlinks->file, "&open");
1132 goto nextpage;
1133 }
1134
1135 /*
1136 * Try interpreting the file as mdoc(7) or man(7)
1137 * source code, unless it is already known to be
1138 * formatted. Fall back to formatted mode.
1139 */
1140 if (mpage->mlinks->dform != FORM_CAT ||
1141 mpage->mlinks->fform != FORM_CAT) {
1142 lvl = mparse_readfd(mp, fd, mpage->mlinks->file);
1143 if (lvl < MANDOCLEVEL_FATAL)
1144 mparse_result(mp, &mdoc, &man, &sodest);
1145 }
1146
1147 if (sodest != NULL) {
1148 mlink_dest = ohash_find(&mlinks,
1149 ohash_qlookup(&mlinks, sodest));
1150 if (mlink_dest == NULL) {
1151 mandoc_asprintf(&cp, "%s.gz", sodest);
1152 mlink_dest = ohash_find(&mlinks,
1153 ohash_qlookup(&mlinks, cp));
1154 free(cp);
1155 }
1156 if (mlink_dest != NULL) {
1157
1158 /* The .so target exists. */
1159
1160 mpage_dest = mlink_dest->mpage;
1161 mlink = mpage->mlinks;
1162 while (1) {
1163 mlink->mpage = mpage_dest;
1164
1165 /*
1166 * If the target was already
1167 * processed, add the links
1168 * to the database now.
1169 * Otherwise, this will
1170 * happen when we come
1171 * to the target.
1172 */
1173
1174 if (mpage_dest->pageid)
1175 dbadd_mlink_name(mlink);
1176
1177 if (mlink->next == NULL)
1178 break;
1179 mlink = mlink->next;
1180 }
1181
1182 /* Move all links to the target. */
1183
1184 mlink->next = mlink_dest->next;
1185 mlink_dest->next = mpage->mlinks;
1186 mpage->mlinks = NULL;
1187 }
1188 goto nextpage;
1189 } else if (mdoc != NULL) {
1190 mpage->form = FORM_SRC;
1191 mpage->sec = mdoc_meta(mdoc)->msec;
1192 mpage->sec = mandoc_strdup(
1193 mpage->sec == NULL ? "" : mpage->sec);
1194 mpage->arch = mdoc_meta(mdoc)->arch;
1195 mpage->arch = mandoc_strdup(
1196 mpage->arch == NULL ? "" : mpage->arch);
1197 mpage->title =
1198 mandoc_strdup(mdoc_meta(mdoc)->title);
1199 } else if (man != NULL) {
1200 mpage->form = FORM_SRC;
1201 mpage->sec =
1202 mandoc_strdup(man_meta(man)->msec);
1203 mpage->arch =
1204 mandoc_strdup(mpage->mlinks->arch);
1205 mpage->title =
1206 mandoc_strdup(man_meta(man)->title);
1207 } else {
1208 mpage->form = FORM_CAT;
1209 mpage->sec =
1210 mandoc_strdup(mpage->mlinks->dsec);
1211 mpage->arch =
1212 mandoc_strdup(mpage->mlinks->arch);
1213 mpage->title =
1214 mandoc_strdup(mpage->mlinks->name);
1215 }
1216 putkey(mpage, mpage->sec, TYPE_sec);
1217 if (*mpage->arch != '\0')
1218 putkey(mpage, mpage->arch, TYPE_arch);
1219
1220 for (mlink = mpage->mlinks; mlink; mlink = mlink->next) {
1221 if ('\0' != *mlink->dsec)
1222 putkey(mpage, mlink->dsec, TYPE_sec);
1223 if ('\0' != *mlink->fsec)
1224 putkey(mpage, mlink->fsec, TYPE_sec);
1225 putkey(mpage, '\0' == *mlink->arch ?
1226 any : mlink->arch, TYPE_arch);
1227 putkey(mpage, mlink->name, NAME_FILE);
1228 }
1229
1230 assert(mpage->desc == NULL);
1231 if (mdoc != NULL)
1232 parse_mdoc(mpage, mdoc_meta(mdoc), mdoc_node(mdoc));
1233 else if (man != NULL)
1234 parse_man(mpage, man_meta(man), man_node(man));
1235 else
1236 parse_cat(mpage, fd);
1237 if (mpage->desc == NULL)
1238 mpage->desc = mandoc_strdup(mpage->mlinks->name);
1239
1240 if (warnings && !use_all)
1241 for (mlink = mpage->mlinks; mlink;
1242 mlink = mlink->next)
1243 mlink_check(mpage, mlink);
1244
1245 dbadd(mpage);
1246
1247 nextpage:
1248 if (mparse_wait(mp) != MANDOCLEVEL_OK) {
1249 exitcode = (int)MANDOCLEVEL_SYSERR;
1250 say(mpage->mlinks->file, "&wait gunzip");
1251 }
1252 ohash_delete(&strings);
1253 ohash_delete(&names);
1254 mpage = ohash_next(&mpages, &pslot);
1255 }
1256
1257 if (0 == nodb)
1258 SQL_EXEC("END TRANSACTION");
1259 }
1260
1261 static void
1262 names_check(void)
1263 {
1264 sqlite3_stmt *stmt;
1265 const char *name, *sec, *arch, *key;
1266 int irc;
1267
1268 sqlite3_prepare_v2(db,
1269 "SELECT name, sec, arch, key FROM ("
1270 "SELECT name AS key, pageid FROM names "
1271 "WHERE bits & ? AND NOT EXISTS ("
1272 "SELECT pageid FROM mlinks "
1273 "WHERE mlinks.pageid == names.pageid "
1274 "AND mlinks.name == names.name"
1275 ")"
1276 ") JOIN ("
1277 "SELECT sec, arch, name, pageid FROM mlinks "
1278 "GROUP BY pageid"
1279 ") USING (pageid);",
1280 -1, &stmt, NULL);
1281
1282 if (SQLITE_OK != sqlite3_bind_int64(stmt, 1, NAME_TITLE))
1283 say("", "%s", sqlite3_errmsg(db));
1284
1285 while (SQLITE_ROW == (irc = sqlite3_step(stmt))) {
1286 name = (const char *)sqlite3_column_text(stmt, 0);
1287 sec = (const char *)sqlite3_column_text(stmt, 1);
1288 arch = (const char *)sqlite3_column_text(stmt, 2);
1289 key = (const char *)sqlite3_column_text(stmt, 3);
1290 say("", "%s(%s%s%s) lacks mlink \"%s\"", name, sec,
1291 '\0' == *arch ? "" : "/",
1292 '\0' == *arch ? "" : arch, key);
1293 }
1294 sqlite3_finalize(stmt);
1295 }
1296
1297 static void
1298 parse_cat(struct mpage *mpage, int fd)
1299 {
1300 FILE *stream;
1301 char *line, *p, *title;
1302 size_t len, plen, titlesz;
1303
1304 stream = (-1 == fd) ?
1305 fopen(mpage->mlinks->file, "r") :
1306 fdopen(fd, "r");
1307 if (NULL == stream) {
1308 if (-1 != fd)
1309 close(fd);
1310 if (warnings)
1311 say(mpage->mlinks->file, "&fopen");
1312 return;
1313 }
1314
1315 /* Skip to first blank line. */
1316
1317 while (NULL != (line = fgetln(stream, &len)))
1318 if ('\n' == *line)
1319 break;
1320
1321 /*
1322 * Assume the first line that is not indented
1323 * is the first section header. Skip to it.
1324 */
1325
1326 while (NULL != (line = fgetln(stream, &len)))
1327 if ('\n' != *line && ' ' != *line)
1328 break;
1329
1330 /*
1331 * Read up until the next section into a buffer.
1332 * Strip the leading and trailing newline from each read line,
1333 * appending a trailing space.
1334 * Ignore empty (whitespace-only) lines.
1335 */
1336
1337 titlesz = 0;
1338 title = NULL;
1339
1340 while (NULL != (line = fgetln(stream, &len))) {
1341 if (' ' != *line || '\n' != line[len - 1])
1342 break;
1343 while (len > 0 && isspace((unsigned char)*line)) {
1344 line++;
1345 len--;
1346 }
1347 if (1 == len)
1348 continue;
1349 title = mandoc_realloc(title, titlesz + len);
1350 memcpy(title + titlesz, line, len);
1351 titlesz += len;
1352 title[titlesz - 1] = ' ';
1353 }
1354
1355 /*
1356 * If no page content can be found, or the input line
1357 * is already the next section header, or there is no
1358 * trailing newline, reuse the page title as the page
1359 * description.
1360 */
1361
1362 if (NULL == title || '\0' == *title) {
1363 if (warnings)
1364 say(mpage->mlinks->file,
1365 "Cannot find NAME section");
1366 fclose(stream);
1367 free(title);
1368 return;
1369 }
1370
1371 title = mandoc_realloc(title, titlesz + 1);
1372 title[titlesz] = '\0';
1373
1374 /*
1375 * Skip to the first dash.
1376 * Use the remaining line as the description (no more than 70
1377 * bytes).
1378 */
1379
1380 if (NULL != (p = strstr(title, "- "))) {
1381 for (p += 2; ' ' == *p || '\b' == *p; p++)
1382 /* Skip to next word. */ ;
1383 } else {
1384 if (warnings)
1385 say(mpage->mlinks->file,
1386 "No dash in title line");
1387 p = title;
1388 }
1389
1390 plen = strlen(p);
1391
1392 /* Strip backspace-encoding from line. */
1393
1394 while (NULL != (line = memchr(p, '\b', plen))) {
1395 len = line - p;
1396 if (0 == len) {
1397 memmove(line, line + 1, plen--);
1398 continue;
1399 }
1400 memmove(line - 1, line + 1, plen - len);
1401 plen -= 2;
1402 }
1403
1404 mpage->desc = mandoc_strdup(p);
1405 fclose(stream);
1406 free(title);
1407 }
1408
1409 /*
1410 * Put a type/word pair into the word database for this particular file.
1411 */
1412 static void
1413 putkey(const struct mpage *mpage, char *value, uint64_t type)
1414 {
1415 char *cp;
1416
1417 assert(NULL != value);
1418 if (TYPE_arch == type)
1419 for (cp = value; *cp; cp++)
1420 if (isupper((unsigned char)*cp))
1421 *cp = _tolower((unsigned char)*cp);
1422 putkeys(mpage, value, strlen(value), type);
1423 }
1424
1425 /*
1426 * Grok all nodes at or below a certain mdoc node into putkey().
1427 */
1428 static void
1429 putmdockey(const struct mpage *mpage,
1430 const struct mdoc_node *n, uint64_t m)
1431 {
1432
1433 for ( ; NULL != n; n = n->next) {
1434 if (NULL != n->child)
1435 putmdockey(mpage, n->child, m);
1436 if (MDOC_TEXT == n->type)
1437 putkey(mpage, n->string, m);
1438 }
1439 }
1440
1441 static void
1442 parse_man(struct mpage *mpage, const struct man_meta *meta,
1443 const struct man_node *n)
1444 {
1445 const struct man_node *head, *body;
1446 char *start, *title;
1447 char byte;
1448 size_t sz;
1449
1450 if (NULL == n)
1451 return;
1452
1453 /*
1454 * We're only searching for one thing: the first text child in
1455 * the BODY of a NAME section. Since we don't keep track of
1456 * sections in -man, run some hoops to find out whether we're in
1457 * the correct section or not.
1458 */
1459
1460 if (MAN_BODY == n->type && MAN_SH == n->tok) {
1461 body = n;
1462 assert(body->parent);
1463 if (NULL != (head = body->parent->head) &&
1464 1 == head->nchild &&
1465 NULL != (head = (head->child)) &&
1466 MAN_TEXT == head->type &&
1467 0 == strcmp(head->string, "NAME") &&
1468 NULL != body->child) {
1469
1470 /*
1471 * Suck the entire NAME section into memory.
1472 * Yes, we might run away.
1473 * But too many manuals have big, spread-out
1474 * NAME sections over many lines.
1475 */
1476
1477 title = NULL;
1478 man_deroff(&title, body);
1479 if (NULL == title)
1480 return;
1481
1482 /*
1483 * Go through a special heuristic dance here.
1484 * Conventionally, one or more manual names are
1485 * comma-specified prior to a whitespace, then a
1486 * dash, then a description. Try to puzzle out
1487 * the name parts here.
1488 */
1489
1490 start = title;
1491 for ( ;; ) {
1492 sz = strcspn(start, " ,");
1493 if ('\0' == start[sz])
1494 break;
1495
1496 byte = start[sz];
1497 start[sz] = '\0';
1498
1499 /*
1500 * Assume a stray trailing comma in the
1501 * name list if a name begins with a dash.
1502 */
1503
1504 if ('-' == start[0] ||
1505 ('\\' == start[0] && '-' == start[1]))
1506 break;
1507
1508 putkey(mpage, start, NAME_TITLE);
1509 if ( ! (mpage->name_head_done ||
1510 strcasecmp(start, meta->title))) {
1511 putkey(mpage, start, NAME_HEAD);
1512 mpage->name_head_done = 1;
1513 }
1514
1515 if (' ' == byte) {
1516 start += sz + 1;
1517 break;
1518 }
1519
1520 assert(',' == byte);
1521 start += sz + 1;
1522 while (' ' == *start)
1523 start++;
1524 }
1525
1526 if (start == title) {
1527 putkey(mpage, start, NAME_TITLE);
1528 if ( ! (mpage->name_head_done ||
1529 strcasecmp(start, meta->title))) {
1530 putkey(mpage, start, NAME_HEAD);
1531 mpage->name_head_done = 1;
1532 }
1533 free(title);
1534 return;
1535 }
1536
1537 while (isspace((unsigned char)*start))
1538 start++;
1539
1540 if (0 == strncmp(start, "-", 1))
1541 start += 1;
1542 else if (0 == strncmp(start, "\\-\\-", 4))
1543 start += 4;
1544 else if (0 == strncmp(start, "\\-", 2))
1545 start += 2;
1546 else if (0 == strncmp(start, "\\(en", 4))
1547 start += 4;
1548 else if (0 == strncmp(start, "\\(em", 4))
1549 start += 4;
1550
1551 while (' ' == *start)
1552 start++;
1553
1554 mpage->desc = mandoc_strdup(start);
1555 free(title);
1556 return;
1557 }
1558 }
1559
1560 for (n = n->child; n; n = n->next) {
1561 if (NULL != mpage->desc)
1562 break;
1563 parse_man(mpage, meta, n);
1564 }
1565 }
1566
1567 static void
1568 parse_mdoc(struct mpage *mpage, const struct mdoc_meta *meta,
1569 const struct mdoc_node *n)
1570 {
1571
1572 assert(NULL != n);
1573 for (n = n->child; NULL != n; n = n->next) {
1574 switch (n->type) {
1575 case MDOC_ELEM:
1576 /* FALLTHROUGH */
1577 case MDOC_BLOCK:
1578 /* FALLTHROUGH */
1579 case MDOC_HEAD:
1580 /* FALLTHROUGH */
1581 case MDOC_BODY:
1582 /* FALLTHROUGH */
1583 case MDOC_TAIL:
1584 if (NULL != mdocs[n->tok].fp)
1585 if (0 == (*mdocs[n->tok].fp)(mpage, meta, n))
1586 break;
1587 if (mdocs[n->tok].mask)
1588 putmdockey(mpage, n->child,
1589 mdocs[n->tok].mask);
1590 break;
1591 default:
1592 assert(MDOC_ROOT != n->type);
1593 continue;
1594 }
1595 if (NULL != n->child)
1596 parse_mdoc(mpage, meta, n);
1597 }
1598 }
1599
1600 static int
1601 parse_mdoc_Fd(struct mpage *mpage, const struct mdoc_meta *meta,
1602 const struct mdoc_node *n)
1603 {
1604 char *start, *end;
1605 size_t sz;
1606
1607 if (SEC_SYNOPSIS != n->sec ||
1608 NULL == (n = n->child) ||
1609 MDOC_TEXT != n->type)
1610 return(0);
1611
1612 /*
1613 * Only consider those `Fd' macro fields that begin with an
1614 * "inclusion" token (versus, e.g., #define).
1615 */
1616
1617 if (strcmp("#include", n->string))
1618 return(0);
1619
1620 if (NULL == (n = n->next) || MDOC_TEXT != n->type)
1621 return(0);
1622
1623 /*
1624 * Strip away the enclosing angle brackets and make sure we're
1625 * not zero-length.
1626 */
1627
1628 start = n->string;
1629 if ('<' == *start || '"' == *start)
1630 start++;
1631
1632 if (0 == (sz = strlen(start)))
1633 return(0);
1634
1635 end = &start[(int)sz - 1];
1636 if ('>' == *end || '"' == *end)
1637 end--;
1638
1639 if (end > start)
1640 putkeys(mpage, start, end - start + 1, TYPE_In);
1641 return(0);
1642 }
1643
1644 static void
1645 parse_mdoc_fname(struct mpage *mpage, const struct mdoc_node *n)
1646 {
1647 char *cp;
1648 size_t sz;
1649
1650 if (n->type != MDOC_TEXT)
1651 return;
1652
1653 /* Skip function pointer punctuation. */
1654
1655 cp = n->string;
1656 while (*cp == '(' || *cp == '*')
1657 cp++;
1658 sz = strcspn(cp, "()");
1659
1660 putkeys(mpage, cp, sz, TYPE_Fn);
1661 if (n->sec == SEC_SYNOPSIS)
1662 putkeys(mpage, cp, sz, NAME_SYN);
1663 }
1664
1665 static int
1666 parse_mdoc_Fn(struct mpage *mpage, const struct mdoc_meta *meta,
1667 const struct mdoc_node *n)
1668 {
1669
1670 if (n->child == NULL)
1671 return(0);
1672
1673 parse_mdoc_fname(mpage, n->child);
1674
1675 for (n = n->child->next; n != NULL; n = n->next)
1676 if (n->type == MDOC_TEXT)
1677 putkey(mpage, n->string, TYPE_Fa);
1678
1679 return(0);
1680 }
1681
1682 static int
1683 parse_mdoc_Fo(struct mpage *mpage, const struct mdoc_meta *meta,
1684 const struct mdoc_node *n)
1685 {
1686
1687 if (n->type != MDOC_HEAD)
1688 return(1);
1689
1690 if (n->child != NULL)
1691 parse_mdoc_fname(mpage, n->child);
1692
1693 return(0);
1694 }
1695
1696 static int
1697 parse_mdoc_Xr(struct mpage *mpage, const struct mdoc_meta *meta,
1698 const struct mdoc_node *n)
1699 {
1700 char *cp;
1701
1702 if (NULL == (n = n->child))
1703 return(0);
1704
1705 if (NULL == n->next) {
1706 putkey(mpage, n->string, TYPE_Xr);
1707 return(0);
1708 }
1709
1710 mandoc_asprintf(&cp, "%s(%s)", n->string, n->next->string);
1711 putkey(mpage, cp, TYPE_Xr);
1712 free(cp);
1713 return(0);
1714 }
1715
1716 static int
1717 parse_mdoc_Nd(struct mpage *mpage, const struct mdoc_meta *meta,
1718 const struct mdoc_node *n)
1719 {
1720
1721 if (MDOC_BODY == n->type)
1722 mdoc_deroff(&mpage->desc, n);
1723 return(0);
1724 }
1725
1726 static int
1727 parse_mdoc_Nm(struct mpage *mpage, const struct mdoc_meta *meta,
1728 const struct mdoc_node *n)
1729 {
1730
1731 if (SEC_NAME == n->sec)
1732 putmdockey(mpage, n->child, NAME_TITLE);
1733 else if (SEC_SYNOPSIS == n->sec && MDOC_HEAD == n->type) {
1734 if (n->child == NULL)
1735 putkey(mpage, meta->name, NAME_SYN);
1736 else
1737 putmdockey(mpage, n->child, NAME_SYN);
1738 }
1739 if ( ! (mpage->name_head_done ||
1740 n->child == NULL || n->child->string == NULL ||
1741 strcasecmp(n->child->string, meta->title))) {
1742 putkey(mpage, n->child->string, NAME_HEAD);
1743 mpage->name_head_done = 1;
1744 }
1745 return(0);
1746 }
1747
1748 static int
1749 parse_mdoc_Sh(struct mpage *mpage, const struct mdoc_meta *meta,
1750 const struct mdoc_node *n)
1751 {
1752
1753 return(SEC_CUSTOM == n->sec && MDOC_HEAD == n->type);
1754 }
1755
1756 static int
1757 parse_mdoc_head(struct mpage *mpage, const struct mdoc_meta *meta,
1758 const struct mdoc_node *n)
1759 {
1760
1761 return(MDOC_HEAD == n->type);
1762 }
1763
1764 static int
1765 parse_mdoc_body(struct mpage *mpage, const struct mdoc_meta *meta,
1766 const struct mdoc_node *n)
1767 {
1768
1769 return(MDOC_BODY == n->type);
1770 }
1771
1772 /*
1773 * Add a string to the hash table for the current manual.
1774 * Each string has a bitmask telling which macros it belongs to.
1775 * When we finish the manual, we'll dump the table.
1776 */
1777 static void
1778 putkeys(const struct mpage *mpage, char *cp, size_t sz, uint64_t v)
1779 {
1780 struct ohash *htab;
1781 struct str *s;
1782 const char *end;
1783 unsigned int slot;
1784 int i, mustfree;
1785
1786 if (0 == sz)
1787 return;
1788
1789 mustfree = render_string(&cp, &sz);
1790
1791 if (TYPE_Nm & v) {
1792 htab = &names;
1793 v &= name_mask;
1794 if (v & NAME_FIRST)
1795 name_mask &= ~NAME_FIRST;
1796 if (debug > 1)
1797 say(mpage->mlinks->file,
1798 "Adding name %*s, bits=%d", sz, cp, v);
1799 } else {
1800 htab = &strings;
1801 if (debug > 1)
1802 for (i = 0; i < mansearch_keymax; i++)
1803 if ((uint64_t)1 << i & v)
1804 say(mpage->mlinks->file,
1805 "Adding key %s=%*s",
1806 mansearch_keynames[i], sz, cp);
1807 }
1808
1809 end = cp + sz;
1810 slot = ohash_qlookupi(htab, cp, &end);
1811 s = ohash_find(htab, slot);
1812
1813 if (NULL != s && mpage == s->mpage) {
1814 s->mask |= v;
1815 return;
1816 } else if (NULL == s) {
1817 s = mandoc_calloc(1, sizeof(struct str) + sz + 1);
1818 memcpy(s->key, cp, sz);
1819 ohash_insert(htab, slot, s);
1820 }
1821 s->mpage = mpage;
1822 s->mask = v;
1823
1824 if (mustfree)
1825 free(cp);
1826 }
1827
1828 /*
1829 * Take a Unicode codepoint and produce its UTF-8 encoding.
1830 * This isn't the best way to do this, but it works.
1831 * The magic numbers are from the UTF-8 packaging.
1832 * They're not as scary as they seem: read the UTF-8 spec for details.
1833 */
1834 static size_t
1835 utf8(unsigned int cp, char out[7])
1836 {
1837 size_t rc;
1838
1839 rc = 0;
1840 if (cp <= 0x0000007F) {
1841 rc = 1;
1842 out[0] = (char)cp;
1843 } else if (cp <= 0x000007FF) {
1844 rc = 2;
1845 out[0] = (cp >> 6 & 31) | 192;
1846 out[1] = (cp & 63) | 128;
1847 } else if (cp <= 0x0000FFFF) {
1848 rc = 3;
1849 out[0] = (cp >> 12 & 15) | 224;
1850 out[1] = (cp >> 6 & 63) | 128;
1851 out[2] = (cp & 63) | 128;
1852 } else if (cp <= 0x001FFFFF) {
1853 rc = 4;
1854 out[0] = (cp >> 18 & 7) | 240;
1855 out[1] = (cp >> 12 & 63) | 128;
1856 out[2] = (cp >> 6 & 63) | 128;
1857 out[3] = (cp & 63) | 128;
1858 } else if (cp <= 0x03FFFFFF) {
1859 rc = 5;
1860 out[0] = (cp >> 24 & 3) | 248;
1861 out[1] = (cp >> 18 & 63) | 128;
1862 out[2] = (cp >> 12 & 63) | 128;
1863 out[3] = (cp >> 6 & 63) | 128;
1864 out[4] = (cp & 63) | 128;
1865 } else if (cp <= 0x7FFFFFFF) {
1866 rc = 6;
1867 out[0] = (cp >> 30 & 1) | 252;
1868 out[1] = (cp >> 24 & 63) | 128;
1869 out[2] = (cp >> 18 & 63) | 128;
1870 out[3] = (cp >> 12 & 63) | 128;
1871 out[4] = (cp >> 6 & 63) | 128;
1872 out[5] = (cp & 63) | 128;
1873 } else
1874 return(0);
1875
1876 out[rc] = '\0';
1877 return(rc);
1878 }
1879
1880 /*
1881 * If the string contains escape sequences,
1882 * replace it with an allocated rendering and return 1,
1883 * such that the caller can free it after use.
1884 * Otherwise, do nothing and return 0.
1885 */
1886 static int
1887 render_string(char **public, size_t *psz)
1888 {
1889 const char *src, *scp, *addcp, *seq;
1890 char *dst;
1891 size_t ssz, dsz, addsz;
1892 char utfbuf[7], res[6];
1893 int seqlen, unicode;
1894
1895 res[0] = '\\';
1896 res[1] = '\t';
1897 res[2] = ASCII_NBRSP;
1898 res[3] = ASCII_HYPH;
1899 res[4] = ASCII_BREAK;
1900 res[5] = '\0';
1901
1902 src = scp = *public;
1903 ssz = *psz;
1904 dst = NULL;
1905 dsz = 0;
1906
1907 while (scp < src + *psz) {
1908
1909 /* Leave normal characters unchanged. */
1910
1911 if (strchr(res, *scp) == NULL) {
1912 if (dst != NULL)
1913 dst[dsz++] = *scp;
1914 scp++;
1915 continue;
1916 }
1917
1918 /*
1919 * Found something that requires replacing,
1920 * make sure we have a destination buffer.
1921 */
1922
1923 if (dst == NULL) {
1924 dst = mandoc_malloc(ssz + 1);
1925 dsz = scp - src;
1926 memcpy(dst, src, dsz);
1927 }
1928
1929 /* Handle single-char special characters. */
1930
1931 switch (*scp) {
1932 case '\\':
1933 break;
1934 case '\t':
1935 /* FALLTHROUGH */
1936 case ASCII_NBRSP:
1937 dst[dsz++] = ' ';
1938 scp++;
1939 continue;
1940 case ASCII_HYPH:
1941 dst[dsz++] = '-';
1942 /* FALLTHROUGH */
1943 case ASCII_BREAK:
1944 scp++;
1945 continue;
1946 default:
1947 abort();
1948 }
1949
1950 /*
1951 * Found an escape sequence.
1952 * Read past the slash, then parse it.
1953 * Ignore everything except characters.
1954 */
1955
1956 scp++;
1957 if (mandoc_escape(&scp, &seq, &seqlen) != ESCAPE_SPECIAL)
1958 continue;
1959
1960 /*
1961 * Render the special character
1962 * as either UTF-8 or ASCII.
1963 */
1964
1965 if (write_utf8) {
1966 unicode = mchars_spec2cp(mchars, seq, seqlen);
1967 if (unicode <= 0)
1968 continue;
1969 addsz = utf8(unicode, utfbuf);
1970 if (addsz == 0)
1971 continue;
1972 addcp = utfbuf;
1973 } else {
1974 addcp = mchars_spec2str(mchars, seq, seqlen, &addsz);
1975 if (addcp == NULL)
1976 continue;
1977 if (*addcp == ASCII_NBRSP) {
1978 addcp = " ";
1979 addsz = 1;
1980 }
1981 }
1982
1983 /* Copy the rendered glyph into the stream. */
1984
1985 ssz += addsz;
1986 dst = mandoc_realloc(dst, ssz + 1);
1987 memcpy(dst + dsz, addcp, addsz);
1988 dsz += addsz;
1989 }
1990 if (dst != NULL) {
1991 *public = dst;
1992 *psz = dsz;
1993 }
1994
1995 /* Trim trailing whitespace and NUL-terminate. */
1996
1997 while (*psz > 0 && (*public)[*psz - 1] == ' ')
1998 --*psz;
1999 if (dst != NULL) {
2000 (*public)[*psz] = '\0';
2001 return(1);
2002 } else
2003 return(0);
2004 }
2005
2006 static void
2007 dbadd_mlink(const struct mlink *mlink)
2008 {
2009 size_t i;
2010
2011 i = 1;
2012 SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->dsec);
2013 SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->arch);
2014 SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->name);
2015 SQL_BIND_INT64(stmts[STMT_INSERT_LINK], i, mlink->mpage->pageid);
2016 SQL_STEP(stmts[STMT_INSERT_LINK]);
2017 sqlite3_reset(stmts[STMT_INSERT_LINK]);
2018 }
2019
2020 static void
2021 dbadd_mlink_name(const struct mlink *mlink)
2022 {
2023 uint64_t bits;
2024 size_t i;
2025
2026 dbadd_mlink(mlink);
2027
2028 i = 1;
2029 SQL_BIND_INT64(stmts[STMT_SELECT_NAME], i, mlink->mpage->pageid);
2030 bits = NAME_FILE & NAME_MASK;
2031 if (sqlite3_step(stmts[STMT_SELECT_NAME]) == SQLITE_ROW) {
2032 bits |= sqlite3_column_int64(stmts[STMT_SELECT_NAME], 0);
2033 sqlite3_reset(stmts[STMT_SELECT_NAME]);
2034 }
2035
2036 i = 1;
2037 SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, bits);
2038 SQL_BIND_TEXT(stmts[STMT_INSERT_NAME], i, mlink->name);
2039 SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, mlink->mpage->pageid);
2040 SQL_STEP(stmts[STMT_INSERT_NAME]);
2041 sqlite3_reset(stmts[STMT_INSERT_NAME]);
2042 }
2043
2044 /*
2045 * Flush the current page's terms (and their bits) into the database.
2046 * Wrap the entire set of additions in a transaction to make sqlite be a
2047 * little faster.
2048 * Also, handle escape sequences at the last possible moment.
2049 */
2050 static void
2051 dbadd(struct mpage *mpage)
2052 {
2053 struct mlink *mlink;
2054 struct str *key;
2055 char *cp;
2056 size_t i;
2057 unsigned int slot;
2058 int mustfree;
2059
2060 mlink = mpage->mlinks;
2061
2062 if (nodb) {
2063 for (key = ohash_first(&names, &slot); NULL != key;
2064 key = ohash_next(&names, &slot))
2065 free(key);
2066 for (key = ohash_first(&strings, &slot); NULL != key;
2067 key = ohash_next(&strings, &slot))
2068 free(key);
2069 if (0 == debug)
2070 return;
2071 while (NULL != mlink) {
2072 fputs(mlink->name, stdout);
2073 if (NULL == mlink->next ||
2074 strcmp(mlink->dsec, mlink->next->dsec) ||
2075 strcmp(mlink->fsec, mlink->next->fsec) ||
2076 strcmp(mlink->arch, mlink->next->arch)) {
2077 putchar('(');
2078 if ('\0' == *mlink->dsec)
2079 fputs(mlink->fsec, stdout);
2080 else
2081 fputs(mlink->dsec, stdout);
2082 if ('\0' != *mlink->arch)
2083 printf("/%s", mlink->arch);
2084 putchar(')');
2085 }
2086 mlink = mlink->next;
2087 if (NULL != mlink)
2088 fputs(", ", stdout);
2089 }
2090 printf(" - %s\n", mpage->desc);
2091 return;
2092 }
2093
2094 if (debug)
2095 say(mlink->file, "Adding to database");
2096
2097 cp = mpage->desc;
2098 i = strlen(cp);
2099 mustfree = render_string(&cp, &i);
2100 i = 1;
2101 SQL_BIND_TEXT(stmts[STMT_INSERT_PAGE], i, cp);
2102 SQL_BIND_INT(stmts[STMT_INSERT_PAGE], i, mpage->form);
2103 SQL_STEP(stmts[STMT_INSERT_PAGE]);
2104 mpage->pageid = sqlite3_last_insert_rowid(db);
2105 sqlite3_reset(stmts[STMT_INSERT_PAGE]);
2106 if (mustfree)
2107 free(cp);
2108
2109 while (NULL != mlink) {
2110 dbadd_mlink(mlink);
2111 mlink = mlink->next;
2112 }
2113 mlink = mpage->mlinks;
2114
2115 for (key = ohash_first(&names, &slot); NULL != key;
2116 key = ohash_next(&names, &slot)) {
2117 assert(key->mpage == mpage);
2118 i = 1;
2119 SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, key->mask);
2120 SQL_BIND_TEXT(stmts[STMT_INSERT_NAME], i, key->key);
2121 SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, mpage->pageid);
2122 SQL_STEP(stmts[STMT_INSERT_NAME]);
2123 sqlite3_reset(stmts[STMT_INSERT_NAME]);
2124 free(key);
2125 }
2126 for (key = ohash_first(&strings, &slot); NULL != key;
2127 key = ohash_next(&strings, &slot)) {
2128 assert(key->mpage == mpage);
2129 i = 1;
2130 SQL_BIND_INT64(stmts[STMT_INSERT_KEY], i, key->mask);
2131 SQL_BIND_TEXT(stmts[STMT_INSERT_KEY], i, key->key);
2132 SQL_BIND_INT64(stmts[STMT_INSERT_KEY], i, mpage->pageid);
2133 SQL_STEP(stmts[STMT_INSERT_KEY]);
2134 sqlite3_reset(stmts[STMT_INSERT_KEY]);
2135 free(key);
2136 }
2137 }
2138
2139 static void
2140 dbprune(void)
2141 {
2142 struct mpage *mpage;
2143 struct mlink *mlink;
2144 size_t i;
2145 unsigned int slot;
2146
2147 if (0 == nodb)
2148 SQL_EXEC("BEGIN TRANSACTION");
2149
2150 for (mpage = ohash_first(&mpages, &slot); NULL != mpage;
2151 mpage = ohash_next(&mpages, &slot)) {
2152 mlink = mpage->mlinks;
2153 if (debug)
2154 say(mlink->file, "Deleting from database");
2155 if (nodb)
2156 continue;
2157 for ( ; NULL != mlink; mlink = mlink->next) {
2158 i = 1;
2159 SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE],
2160 i, mlink->dsec);
2161 SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE],
2162 i, mlink->arch);
2163 SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE],
2164 i, mlink->name);
2165 SQL_STEP(stmts[STMT_DELETE_PAGE]);
2166 sqlite3_reset(stmts[STMT_DELETE_PAGE]);
2167 }
2168 }
2169
2170 if (0 == nodb)
2171 SQL_EXEC("END TRANSACTION");
2172 }
2173
2174 /*
2175 * Close an existing database and its prepared statements.
2176 * If "real" is not set, rename the temporary file into the real one.
2177 */
2178 static void
2179 dbclose(int real)
2180 {
2181 size_t i;
2182 int status;
2183 pid_t child;
2184
2185 if (nodb)
2186 return;
2187
2188 for (i = 0; i < STMT__MAX; i++) {
2189 sqlite3_finalize(stmts[i]);
2190 stmts[i] = NULL;
2191 }
2192
2193 sqlite3_close(db);
2194 db = NULL;
2195
2196 if (real)
2197 return;
2198
2199 if ('\0' == *tempfilename) {
2200 if (-1 == rename(MANDOC_DB "~", MANDOC_DB)) {
2201 exitcode = (int)MANDOCLEVEL_SYSERR;
2202 say(MANDOC_DB, "&rename");
2203 }
2204 return;
2205 }
2206
2207 switch (child = fork()) {
2208 case -1:
2209 exitcode = (int)MANDOCLEVEL_SYSERR;
2210 say("", "&fork cmp");
2211 return;
2212 case 0:
2213 execlp("cmp", "cmp", "-s",
2214 tempfilename, MANDOC_DB, NULL);
2215 say("", "&exec cmp");
2216 exit(0);
2217 default:
2218 break;
2219 }
2220 if (-1 == waitpid(child, &status, 0)) {
2221 exitcode = (int)MANDOCLEVEL_SYSERR;
2222 say("", "&wait cmp");
2223 } else if (WIFSIGNALED(status)) {
2224 exitcode = (int)MANDOCLEVEL_SYSERR;
2225 say("", "cmp died from signal %d", WTERMSIG(status));
2226 } else if (WEXITSTATUS(status)) {
2227 exitcode = (int)MANDOCLEVEL_SYSERR;
2228 say(MANDOC_DB,
2229 "Data changed, but cannot replace database");
2230 }
2231
2232 *strrchr(tempfilename, '/') = '\0';
2233 switch (child = fork()) {
2234 case -1:
2235 exitcode = (int)MANDOCLEVEL_SYSERR;
2236 say("", "&fork rm");
2237 return;
2238 case 0:
2239 execlp("rm", "rm", "-rf", tempfilename, NULL);
2240 say("", "&exec rm");
2241 exit((int)MANDOCLEVEL_SYSERR);
2242 default:
2243 break;
2244 }
2245 if (-1 == waitpid(child, &status, 0)) {
2246 exitcode = (int)MANDOCLEVEL_SYSERR;
2247 say("", "&wait rm");
2248 } else if (WIFSIGNALED(status) || WEXITSTATUS(status)) {
2249 exitcode = (int)MANDOCLEVEL_SYSERR;
2250 say("", "%s: Cannot remove temporary directory",
2251 tempfilename);
2252 }
2253 }
2254
2255 /*
2256 * This is straightforward stuff.
2257 * Open a database connection to a "temporary" database, then open a set
2258 * of prepared statements we'll use over and over again.
2259 * If "real" is set, we use the existing database; if not, we truncate a
2260 * temporary one.
2261 * Must be matched by dbclose().
2262 */
2263 static int
2264 dbopen(int real)
2265 {
2266 const char *sql;
2267 int rc, ofl;
2268
2269 if (nodb)
2270 return(1);
2271
2272 *tempfilename = '\0';
2273 ofl = SQLITE_OPEN_READWRITE;
2274
2275 if (real) {
2276 rc = sqlite3_open_v2(MANDOC_DB, &db, ofl, NULL);
2277 if (SQLITE_OK != rc) {
2278 exitcode = (int)MANDOCLEVEL_SYSERR;
2279 if (SQLITE_CANTOPEN != rc)
2280 say(MANDOC_DB, "%s", sqlite3_errstr(rc));
2281 return(0);
2282 }
2283 goto prepare_statements;
2284 }
2285
2286 ofl |= SQLITE_OPEN_CREATE | SQLITE_OPEN_EXCLUSIVE;
2287
2288 remove(MANDOC_DB "~");
2289 rc = sqlite3_open_v2(MANDOC_DB "~", &db, ofl, NULL);
2290 if (SQLITE_OK == rc)
2291 goto create_tables;
2292 if (MPARSE_QUICK & mparse_options) {
2293 exitcode = (int)MANDOCLEVEL_SYSERR;
2294 say(MANDOC_DB "~", "%s", sqlite3_errstr(rc));
2295 return(0);
2296 }
2297
2298 (void)strlcpy(tempfilename, "/tmp/mandocdb.XXXXXX",
2299 sizeof(tempfilename));
2300 if (NULL == mkdtemp(tempfilename)) {
2301 exitcode = (int)MANDOCLEVEL_SYSERR;
2302 say("", "&%s", tempfilename);
2303 return(0);
2304 }
2305 (void)strlcat(tempfilename, "/" MANDOC_DB,
2306 sizeof(tempfilename));
2307 rc = sqlite3_open_v2(tempfilename, &db, ofl, NULL);
2308 if (SQLITE_OK != rc) {
2309 exitcode = (int)MANDOCLEVEL_SYSERR;
2310 say("", "%s: %s", tempfilename, sqlite3_errstr(rc));
2311 return(0);
2312 }
2313
2314 create_tables:
2315 sql = "CREATE TABLE \"mpages\" (\n"
2316 " \"desc\" TEXT NOT NULL,\n"
2317 " \"form\" INTEGER NOT NULL,\n"
2318 " \"pageid\" INTEGER PRIMARY KEY AUTOINCREMENT NOT NULL\n"
2319 ");\n"
2320 "\n"
2321 "CREATE TABLE \"mlinks\" (\n"
2322 " \"sec\" TEXT NOT NULL,\n"
2323 " \"arch\" TEXT NOT NULL,\n"
2324 " \"name\" TEXT NOT NULL,\n"
2325 " \"pageid\" INTEGER NOT NULL REFERENCES mpages(pageid) "
2326 "ON DELETE CASCADE\n"
2327 ");\n"
2328 "CREATE INDEX mlinks_pageid_idx ON mlinks (pageid);\n"
2329 "\n"
2330 "CREATE TABLE \"names\" (\n"
2331 " \"bits\" INTEGER NOT NULL,\n"
2332 " \"name\" TEXT NOT NULL,\n"
2333 " \"pageid\" INTEGER NOT NULL REFERENCES mpages(pageid) "
2334 "ON DELETE CASCADE,\n"
2335 " UNIQUE (\"name\", \"pageid\") ON CONFLICT REPLACE\n"
2336 ");\n"
2337 "\n"
2338 "CREATE TABLE \"keys\" (\n"
2339 " \"bits\" INTEGER NOT NULL,\n"
2340 " \"key\" TEXT NOT NULL,\n"
2341 " \"pageid\" INTEGER NOT NULL REFERENCES mpages(pageid) "
2342 "ON DELETE CASCADE\n"
2343 ");\n"
2344 "CREATE INDEX keys_pageid_idx ON keys (pageid);\n";
2345
2346 if (SQLITE_OK != sqlite3_exec(db, sql, NULL, NULL, NULL)) {
2347 exitcode = (int)MANDOCLEVEL_SYSERR;
2348 say(MANDOC_DB, "%s", sqlite3_errmsg(db));
2349 sqlite3_close(db);
2350 return(0);
2351 }
2352
2353 prepare_statements:
2354 if (SQLITE_OK != sqlite3_exec(db,
2355 "PRAGMA foreign_keys = ON", NULL, NULL, NULL)) {
2356 exitcode = (int)MANDOCLEVEL_SYSERR;
2357 say(MANDOC_DB, "PRAGMA foreign_keys: %s",
2358 sqlite3_errmsg(db));
2359 sqlite3_close(db);
2360 return(0);
2361 }
2362
2363 sql = "DELETE FROM mpages WHERE pageid IN "
2364 "(SELECT pageid FROM mlinks WHERE "
2365 "sec=? AND arch=? AND name=?)";
2366 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_DELETE_PAGE], NULL);
2367 sql = "INSERT INTO mpages "
2368 "(desc,form) VALUES (?,?)";
2369 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_PAGE], NULL);
2370 sql = "INSERT INTO mlinks "
2371 "(sec,arch,name,pageid) VALUES (?,?,?,?)";
2372 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_LINK], NULL);
2373 sql = "SELECT bits FROM names where pageid = ?";
2374 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_SELECT_NAME], NULL);
2375 sql = "INSERT INTO names "
2376 "(bits,name,pageid) VALUES (?,?,?)";
2377 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_NAME], NULL);
2378 sql = "INSERT INTO keys "
2379 "(bits,key,pageid) VALUES (?,?,?)";
2380 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_KEY], NULL);
2381
2382 #ifndef __APPLE__
2383 /*
2384 * When opening a new database, we can turn off
2385 * synchronous mode for much better performance.
2386 */
2387
2388 if (real && SQLITE_OK != sqlite3_exec(db,
2389 "PRAGMA synchronous = OFF", NULL, NULL, NULL)) {
2390 exitcode = (int)MANDOCLEVEL_SYSERR;
2391 say(MANDOC_DB, "PRAGMA synchronous: %s",
2392 sqlite3_errmsg(db));
2393 sqlite3_close(db);
2394 return(0);
2395 }
2396 #endif
2397
2398 return(1);
2399 }
2400
2401 static void *
2402 hash_calloc(size_t n, size_t sz, void *arg)
2403 {
2404
2405 return(mandoc_calloc(n, sz));
2406 }
2407
2408 static void *
2409 hash_alloc(size_t sz, void *arg)
2410 {
2411
2412 return(mandoc_malloc(sz));
2413 }
2414
2415 static void
2416 hash_free(void *p, void *arg)
2417 {
2418
2419 free(p);
2420 }
2421
2422 static int
2423 set_basedir(const char *targetdir, int report_baddir)
2424 {
2425 static char startdir[PATH_MAX];
2426 static int getcwd_status; /* 1 = ok, 2 = failure */
2427 static int chdir_status; /* 1 = changed directory */
2428 char *cp;
2429
2430 /*
2431 * Remember the original working directory, if possible.
2432 * This will be needed if the second or a later directory
2433 * on the command line is given as a relative path.
2434 * Do not error out if the current directory is not
2435 * searchable: Maybe it won't be needed after all.
2436 */
2437 if (0 == getcwd_status) {
2438 if (NULL == getcwd(startdir, sizeof(startdir))) {
2439 getcwd_status = 2;
2440 (void)strlcpy(startdir, strerror(errno),
2441 sizeof(startdir));
2442 } else
2443 getcwd_status = 1;
2444 }
2445
2446 /*
2447 * We are leaving the old base directory.
2448 * Do not use it any longer, not even for messages.
2449 */
2450 *basedir = '\0';
2451
2452 /*
2453 * If and only if the directory was changed earlier and
2454 * the next directory to process is given as a relative path,
2455 * first go back, or bail out if that is impossible.
2456 */
2457 if (chdir_status && '/' != *targetdir) {
2458 if (2 == getcwd_status) {
2459 exitcode = (int)MANDOCLEVEL_SYSERR;
2460 say("", "getcwd: %s", startdir);
2461 return(0);
2462 }
2463 if (-1 == chdir(startdir)) {
2464 exitcode = (int)MANDOCLEVEL_SYSERR;
2465 say("", "&chdir %s", startdir);
2466 return(0);
2467 }
2468 }
2469
2470 /*
2471 * Always resolve basedir to the canonicalized absolute
2472 * pathname and append a trailing slash, such that
2473 * we can reliably check whether files are inside.
2474 */
2475 if (NULL == realpath(targetdir, basedir)) {
2476 if (report_baddir || errno != ENOENT) {
2477 exitcode = (int)MANDOCLEVEL_BADARG;
2478 say("", "&%s: realpath", targetdir);
2479 }
2480 return(0);
2481 } else if (-1 == chdir(basedir)) {
2482 if (report_baddir || errno != ENOENT) {
2483 exitcode = (int)MANDOCLEVEL_BADARG;
2484 say("", "&chdir");
2485 }
2486 return(0);
2487 }
2488 chdir_status = 1;
2489 cp = strchr(basedir, '\0');
2490 if ('/' != cp[-1]) {
2491 if (cp - basedir >= PATH_MAX - 1) {
2492 exitcode = (int)MANDOCLEVEL_SYSERR;
2493 say("", "Filename too long");
2494 return(0);
2495 }
2496 *cp++ = '/';
2497 *cp = '\0';
2498 }
2499 return(1);
2500 }
2501
2502 static void
2503 say(const char *file, const char *format, ...)
2504 {
2505 va_list ap;
2506 int use_errno;
2507
2508 if ('\0' != *basedir)
2509 fprintf(stderr, "%s", basedir);
2510 if ('\0' != *basedir && '\0' != *file)
2511 fputc('/', stderr);
2512 if ('\0' != *file)
2513 fprintf(stderr, "%s", file);
2514
2515 use_errno = 1;
2516 if (NULL != format) {
2517 switch (*format) {
2518 case '&':
2519 format++;
2520 break;
2521 case '\0':
2522 format = NULL;
2523 break;
2524 default:
2525 use_errno = 0;
2526 break;
2527 }
2528 }
2529 if (NULL != format) {
2530 if ('\0' != *basedir || '\0' != *file)
2531 fputs(": ", stderr);
2532 va_start(ap, format);
2533 vfprintf(stderr, format, ap);
2534 va_end(ap);
2535 }
2536 if (use_errno) {
2537 if ('\0' != *basedir || '\0' != *file || NULL != format)
2538 fputs(": ", stderr);
2539 perror(NULL);
2540 } else
2541 fputc('\n', stderr);
2542 }
A git conversion mirror of mandoc.bsd.lv