]> 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
When makewhatis(8) finds an .so link after the manual being pointed to
[mandoc.git] / mandocdb.c
1 /* $Id: mandocdb.c,v 1.160 2014/09/01 23:47:59 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/stat.h>
21 #include <sys/types.h>
22 #include <sys/wait.h>
23
24 #include <assert.h>
25 #include <ctype.h>
26 #include <errno.h>
27 #include <fcntl.h>
28 #if HAVE_FTS
29 #include <fts.h>
30 #else
31 #include "compat_fts.h"
32 #endif
33 #include <getopt.h>
34 #include <limits.h>
35 #include <stddef.h>
36 #include <stdio.h>
37 #include <stdint.h>
38 #include <stdlib.h>
39 #include <string.h>
40 #include <unistd.h>
41
42 #if HAVE_OHASH
43 #include <ohash.h>
44 #else
45 #include "compat_ohash.h"
46 #endif
47 #include <sqlite3.h>
48
49 #include "mdoc.h"
50 #include "man.h"
51 #include "mandoc.h"
52 #include "mandoc_aux.h"
53 #include "manpath.h"
54 #include "mansearch.h"
55
56 extern int mansearch_keymax;
57 extern const char *const mansearch_keynames[];
58
59 #define SQL_EXEC(_v) \
60 if (SQLITE_OK != sqlite3_exec(db, (_v), NULL, NULL, NULL)) \
61 say("", "%s: %s", (_v), sqlite3_errmsg(db))
62 #define SQL_BIND_TEXT(_s, _i, _v) \
63 if (SQLITE_OK != sqlite3_bind_text \
64 ((_s), (_i)++, (_v), -1, SQLITE_STATIC)) \
65 say(mlink->file, "%s", sqlite3_errmsg(db))
66 #define SQL_BIND_INT(_s, _i, _v) \
67 if (SQLITE_OK != sqlite3_bind_int \
68 ((_s), (_i)++, (_v))) \
69 say(mlink->file, "%s", sqlite3_errmsg(db))
70 #define SQL_BIND_INT64(_s, _i, _v) \
71 if (SQLITE_OK != sqlite3_bind_int64 \
72 ((_s), (_i)++, (_v))) \
73 say(mlink->file, "%s", sqlite3_errmsg(db))
74 #define SQL_STEP(_s) \
75 if (SQLITE_DONE != sqlite3_step((_s))) \
76 say(mlink->file, "%s", sqlite3_errmsg(db))
77
78 enum op {
79 OP_DEFAULT = 0, /* new dbs from dir list or default config */
80 OP_CONFFILE, /* new databases from custom config file */
81 OP_UPDATE, /* delete/add entries in existing database */
82 OP_DELETE, /* delete entries from existing database */
83 OP_TEST /* change no databases, report potential problems */
84 };
85
86 enum form {
87 FORM_NONE, /* format is unknown */
88 FORM_SRC, /* format is -man or -mdoc */
89 FORM_CAT /* format is cat */
90 };
91
92 struct str {
93 char *rendered; /* key in UTF-8 or ASCII form */
94 const struct mpage *mpage; /* if set, the owning parse */
95 uint64_t mask; /* bitmask in sequence */
96 char key[]; /* may contain escape sequences */
97 };
98
99 struct inodev {
100 ino_t st_ino;
101 dev_t st_dev;
102 };
103
104 struct mpage {
105 struct inodev inodev; /* used for hashing routine */
106 int64_t pageid; /* pageid in mpages SQL table */
107 enum form form; /* format from file content */
108 char *sec; /* section from file content */
109 char *arch; /* architecture from file content */
110 char *title; /* title from file content */
111 char *desc; /* description from file content */
112 struct mlink *mlinks; /* singly linked list */
113 };
114
115 struct mlink {
116 char file[PATH_MAX]; /* filename rel. to manpath */
117 enum form dform; /* format from directory */
118 enum form fform; /* format from file name suffix */
119 char *dsec; /* section from directory */
120 char *arch; /* architecture from directory */
121 char *name; /* name from file name (not empty) */
122 char *fsec; /* section from file name suffix */
123 struct mlink *next; /* singly linked list */
124 struct mpage *mpage; /* parent */
125 int gzip; /* filename has a .gz suffix */
126 };
127
128 enum stmt {
129 STMT_DELETE_PAGE = 0, /* delete mpage */
130 STMT_INSERT_PAGE, /* insert mpage */
131 STMT_INSERT_LINK, /* insert mlink */
132 STMT_INSERT_NAME, /* insert name */
133 STMT_INSERT_KEY, /* insert parsed key */
134 STMT__MAX
135 };
136
137 typedef int (*mdoc_fp)(struct mpage *, const struct mdoc_node *);
138
139 struct mdoc_handler {
140 mdoc_fp fp; /* optional handler */
141 uint64_t mask; /* set unless handler returns 0 */
142 };
143
144 static void dbclose(int);
145 static void dbadd(struct mpage *, struct mchars *);
146 static void dbadd_mlink(const struct mlink *mlink);
147 static int dbopen(int);
148 static void dbprune(void);
149 static void filescan(const char *);
150 static void *hash_alloc(size_t, void *);
151 static void hash_free(void *, void *);
152 static void *hash_calloc(size_t, size_t, void *);
153 static void mlink_add(struct mlink *, const struct stat *);
154 static void mlink_check(struct mpage *, struct mlink *);
155 static void mlink_free(struct mlink *);
156 static void mlinks_undupe(struct mpage *);
157 static void mpages_free(void);
158 static void mpages_merge(struct mchars *, struct mparse *);
159 static void names_check(void);
160 static void parse_cat(struct mpage *, int);
161 static void parse_man(struct mpage *, const struct man_node *);
162 static void parse_mdoc(struct mpage *, const struct mdoc_node *);
163 static int parse_mdoc_body(struct mpage *, const struct mdoc_node *);
164 static int parse_mdoc_head(struct mpage *, const struct mdoc_node *);
165 static int parse_mdoc_Fd(struct mpage *, const struct mdoc_node *);
166 static int parse_mdoc_Fn(struct mpage *, const struct mdoc_node *);
167 static int parse_mdoc_Nd(struct mpage *, const struct mdoc_node *);
168 static int parse_mdoc_Nm(struct mpage *, const struct mdoc_node *);
169 static int parse_mdoc_Sh(struct mpage *, const struct mdoc_node *);
170 static int parse_mdoc_Xr(struct mpage *, const struct mdoc_node *);
171 static void putkey(const struct mpage *, char *, uint64_t);
172 static void putkeys(const struct mpage *,
173 const char *, size_t, uint64_t);
174 static void putmdockey(const struct mpage *,
175 const struct mdoc_node *, uint64_t);
176 static void render_key(struct mchars *, struct str *);
177 static void say(const char *, const char *, ...);
178 static int set_basedir(const char *);
179 static int treescan(void);
180 static size_t utf8(unsigned int, char [7]);
181
182 static char tempfilename[32];
183 static char *progname;
184 static int nodb; /* no database changes */
185 static int mparse_options; /* abort the parse early */
186 static int use_all; /* use all found files */
187 static int debug; /* print what we're doing */
188 static int warnings; /* warn about crap */
189 static int write_utf8; /* write UTF-8 output; else ASCII */
190 static int exitcode; /* to be returned by main */
191 static enum op op; /* operational mode */
192 static char basedir[PATH_MAX]; /* current base directory */
193 static struct ohash mpages; /* table of distinct manual pages */
194 static struct ohash mlinks; /* table of directory entries */
195 static struct ohash names; /* table of all names */
196 static struct ohash strings; /* table of all strings */
197 static sqlite3 *db = NULL; /* current database */
198 static sqlite3_stmt *stmts[STMT__MAX]; /* current statements */
199 static uint64_t name_mask;
200
201 static const struct mdoc_handler mdocs[MDOC_MAX] = {
202 { NULL, 0 }, /* Ap */
203 { NULL, 0 }, /* Dd */
204 { NULL, 0 }, /* Dt */
205 { NULL, 0 }, /* Os */
206 { parse_mdoc_Sh, TYPE_Sh }, /* Sh */
207 { parse_mdoc_head, TYPE_Ss }, /* Ss */
208 { NULL, 0 }, /* Pp */
209 { NULL, 0 }, /* D1 */
210 { NULL, 0 }, /* Dl */
211 { NULL, 0 }, /* Bd */
212 { NULL, 0 }, /* Ed */
213 { NULL, 0 }, /* Bl */
214 { NULL, 0 }, /* El */
215 { NULL, 0 }, /* It */
216 { NULL, 0 }, /* Ad */
217 { NULL, TYPE_An }, /* An */
218 { NULL, TYPE_Ar }, /* Ar */
219 { NULL, TYPE_Cd }, /* Cd */
220 { NULL, TYPE_Cm }, /* Cm */
221 { NULL, TYPE_Dv }, /* Dv */
222 { NULL, TYPE_Er }, /* Er */
223 { NULL, TYPE_Ev }, /* Ev */
224 { NULL, 0 }, /* Ex */
225 { NULL, TYPE_Fa }, /* Fa */
226 { parse_mdoc_Fd, 0 }, /* Fd */
227 { NULL, TYPE_Fl }, /* Fl */
228 { parse_mdoc_Fn, 0 }, /* Fn */
229 { NULL, TYPE_Ft }, /* Ft */
230 { NULL, TYPE_Ic }, /* Ic */
231 { NULL, TYPE_In }, /* In */
232 { NULL, TYPE_Li }, /* Li */
233 { parse_mdoc_Nd, 0 }, /* Nd */
234 { parse_mdoc_Nm, 0 }, /* Nm */
235 { NULL, 0 }, /* Op */
236 { NULL, 0 }, /* Ot */
237 { NULL, TYPE_Pa }, /* Pa */
238 { NULL, 0 }, /* Rv */
239 { NULL, TYPE_St }, /* St */
240 { NULL, TYPE_Va }, /* Va */
241 { parse_mdoc_body, TYPE_Va }, /* Vt */
242 { parse_mdoc_Xr, 0 }, /* Xr */
243 { NULL, 0 }, /* %A */
244 { NULL, 0 }, /* %B */
245 { NULL, 0 }, /* %D */
246 { NULL, 0 }, /* %I */
247 { NULL, 0 }, /* %J */
248 { NULL, 0 }, /* %N */
249 { NULL, 0 }, /* %O */
250 { NULL, 0 }, /* %P */
251 { NULL, 0 }, /* %R */
252 { NULL, 0 }, /* %T */
253 { NULL, 0 }, /* %V */
254 { NULL, 0 }, /* Ac */
255 { NULL, 0 }, /* Ao */
256 { NULL, 0 }, /* Aq */
257 { NULL, TYPE_At }, /* At */
258 { NULL, 0 }, /* Bc */
259 { NULL, 0 }, /* Bf */
260 { NULL, 0 }, /* Bo */
261 { NULL, 0 }, /* Bq */
262 { NULL, TYPE_Bsx }, /* Bsx */
263 { NULL, TYPE_Bx }, /* Bx */
264 { NULL, 0 }, /* Db */
265 { NULL, 0 }, /* Dc */
266 { NULL, 0 }, /* Do */
267 { NULL, 0 }, /* Dq */
268 { NULL, 0 }, /* Ec */
269 { NULL, 0 }, /* Ef */
270 { NULL, TYPE_Em }, /* Em */
271 { NULL, 0 }, /* Eo */
272 { NULL, TYPE_Fx }, /* Fx */
273 { NULL, TYPE_Ms }, /* Ms */
274 { NULL, 0 }, /* No */
275 { NULL, 0 }, /* Ns */
276 { NULL, TYPE_Nx }, /* Nx */
277 { NULL, TYPE_Ox }, /* Ox */
278 { NULL, 0 }, /* Pc */
279 { NULL, 0 }, /* Pf */
280 { NULL, 0 }, /* Po */
281 { NULL, 0 }, /* Pq */
282 { NULL, 0 }, /* Qc */
283 { NULL, 0 }, /* Ql */
284 { NULL, 0 }, /* Qo */
285 { NULL, 0 }, /* Qq */
286 { NULL, 0 }, /* Re */
287 { NULL, 0 }, /* Rs */
288 { NULL, 0 }, /* Sc */
289 { NULL, 0 }, /* So */
290 { NULL, 0 }, /* Sq */
291 { NULL, 0 }, /* Sm */
292 { NULL, 0 }, /* Sx */
293 { NULL, TYPE_Sy }, /* Sy */
294 { NULL, TYPE_Tn }, /* Tn */
295 { NULL, 0 }, /* Ux */
296 { NULL, 0 }, /* Xc */
297 { NULL, 0 }, /* Xo */
298 { parse_mdoc_head, 0 }, /* Fo */
299 { NULL, 0 }, /* Fc */
300 { NULL, 0 }, /* Oo */
301 { NULL, 0 }, /* Oc */
302 { NULL, 0 }, /* Bk */
303 { NULL, 0 }, /* Ek */
304 { NULL, 0 }, /* Bt */
305 { NULL, 0 }, /* Hf */
306 { NULL, 0 }, /* Fr */
307 { NULL, 0 }, /* Ud */
308 { NULL, TYPE_Lb }, /* Lb */
309 { NULL, 0 }, /* Lp */
310 { NULL, TYPE_Lk }, /* Lk */
311 { NULL, TYPE_Mt }, /* Mt */
312 { NULL, 0 }, /* Brq */
313 { NULL, 0 }, /* Bro */
314 { NULL, 0 }, /* Brc */
315 { NULL, 0 }, /* %C */
316 { NULL, 0 }, /* Es */
317 { NULL, 0 }, /* En */
318 { NULL, TYPE_Dx }, /* Dx */
319 { NULL, 0 }, /* %Q */
320 { NULL, 0 }, /* br */
321 { NULL, 0 }, /* sp */
322 { NULL, 0 }, /* %U */
323 { NULL, 0 }, /* Ta */
324 };
325
326
327 int
328 main(int argc, char *argv[])
329 {
330 int ch, i;
331 size_t j, sz;
332 const char *path_arg;
333 struct mchars *mc;
334 struct manpaths dirs;
335 struct mparse *mp;
336 struct ohash_info mpages_info, mlinks_info;
337
338 memset(stmts, 0, STMT__MAX * sizeof(sqlite3_stmt *));
339 memset(&dirs, 0, sizeof(struct manpaths));
340
341 mpages_info.alloc = mlinks_info.alloc = hash_alloc;
342 mpages_info.calloc = mlinks_info.calloc = hash_calloc;
343 mpages_info.free = mlinks_info.free = hash_free;
344
345 mpages_info.key_offset = offsetof(struct mpage, inodev);
346 mlinks_info.key_offset = offsetof(struct mlink, file);
347
348 progname = strrchr(argv[0], '/');
349 if (progname == NULL)
350 progname = argv[0];
351 else
352 ++progname;
353
354 /*
355 * We accept a few different invocations.
356 * The CHECKOP macro makes sure that invocation styles don't
357 * clobber each other.
358 */
359 #define CHECKOP(_op, _ch) do \
360 if (OP_DEFAULT != (_op)) { \
361 fprintf(stderr, "%s: -%c: Conflicting option\n", \
362 progname, (_ch)); \
363 goto usage; \
364 } while (/*CONSTCOND*/0)
365
366 path_arg = NULL;
367 op = OP_DEFAULT;
368
369 while (-1 != (ch = getopt(argc, argv, "aC:Dd:npQT:tu:v")))
370 switch (ch) {
371 case 'a':
372 use_all = 1;
373 break;
374 case 'C':
375 CHECKOP(op, ch);
376 path_arg = optarg;
377 op = OP_CONFFILE;
378 break;
379 case 'D':
380 debug++;
381 break;
382 case 'd':
383 CHECKOP(op, ch);
384 path_arg = optarg;
385 op = OP_UPDATE;
386 break;
387 case 'n':
388 nodb = 1;
389 break;
390 case 'p':
391 warnings = 1;
392 break;
393 case 'Q':
394 mparse_options |= MPARSE_QUICK;
395 break;
396 case 'T':
397 if (strcmp(optarg, "utf8")) {
398 fprintf(stderr, "%s: -T%s: "
399 "Unsupported output format\n",
400 progname, optarg);
401 goto usage;
402 }
403 write_utf8 = 1;
404 break;
405 case 't':
406 CHECKOP(op, ch);
407 dup2(STDOUT_FILENO, STDERR_FILENO);
408 op = OP_TEST;
409 nodb = warnings = 1;
410 break;
411 case 'u':
412 CHECKOP(op, ch);
413 path_arg = optarg;
414 op = OP_DELETE;
415 break;
416 case 'v':
417 /* Compatibility with espie@'s makewhatis. */
418 break;
419 default:
420 goto usage;
421 }
422
423 argc -= optind;
424 argv += optind;
425
426 if (OP_CONFFILE == op && argc > 0) {
427 fprintf(stderr, "%s: -C: Too many arguments\n",
428 progname);
429 goto usage;
430 }
431
432 exitcode = (int)MANDOCLEVEL_OK;
433 mp = mparse_alloc(mparse_options, MANDOCLEVEL_FATAL, NULL, NULL);
434 mc = mchars_alloc();
435
436 ohash_init(&mpages, 6, &mpages_info);
437 ohash_init(&mlinks, 6, &mlinks_info);
438
439 if (OP_UPDATE == op || OP_DELETE == op || OP_TEST == op) {
440
441 /*
442 * Most of these deal with a specific directory.
443 * Jump into that directory first.
444 */
445 if (OP_TEST != op && 0 == set_basedir(path_arg))
446 goto out;
447
448 if (dbopen(1)) {
449 /*
450 * The existing database is usable. Process
451 * all files specified on the command-line.
452 */
453 use_all = 1;
454 for (i = 0; i < argc; i++)
455 filescan(argv[i]);
456 if (OP_TEST != op)
457 dbprune();
458 } else {
459 /*
460 * Database missing or corrupt.
461 * Recreate from scratch.
462 */
463 exitcode = (int)MANDOCLEVEL_OK;
464 op = OP_DEFAULT;
465 if (0 == treescan())
466 goto out;
467 if (0 == dbopen(0))
468 goto out;
469 }
470 if (OP_DELETE != op)
471 mpages_merge(mc, mp);
472 dbclose(OP_DEFAULT == op ? 0 : 1);
473 } else {
474 /*
475 * If we have arguments, use them as our manpaths.
476 * If we don't, grok from manpath(1) or however else
477 * manpath_parse() wants to do it.
478 */
479 if (argc > 0) {
480 dirs.paths = mandoc_reallocarray(NULL,
481 argc, sizeof(char *));
482 dirs.sz = (size_t)argc;
483 for (i = 0; i < argc; i++)
484 dirs.paths[i] = mandoc_strdup(argv[i]);
485 } else
486 manpath_parse(&dirs, path_arg, NULL, NULL);
487
488 if (0 == dirs.sz) {
489 exitcode = (int)MANDOCLEVEL_BADARG;
490 say("", "Empty manpath");
491 }
492
493 /*
494 * First scan the tree rooted at a base directory, then
495 * build a new database and finally move it into place.
496 * Ignore zero-length directories and strip trailing
497 * slashes.
498 */
499 for (j = 0; j < dirs.sz; j++) {
500 sz = strlen(dirs.paths[j]);
501 if (sz && '/' == dirs.paths[j][sz - 1])
502 dirs.paths[j][--sz] = '\0';
503 if (0 == sz)
504 continue;
505
506 if (j) {
507 ohash_init(&mpages, 6, &mpages_info);
508 ohash_init(&mlinks, 6, &mlinks_info);
509 }
510
511 if (0 == set_basedir(dirs.paths[j]))
512 goto out;
513 if (0 == treescan())
514 goto out;
515 if (0 == dbopen(0))
516 goto out;
517
518 mpages_merge(mc, mp);
519 if (warnings && !nodb &&
520 ! (MPARSE_QUICK & mparse_options))
521 names_check();
522 dbclose(0);
523
524 if (j + 1 < dirs.sz) {
525 mpages_free();
526 ohash_delete(&mpages);
527 ohash_delete(&mlinks);
528 }
529 }
530 }
531 out:
532 manpath_free(&dirs);
533 mchars_free(mc);
534 mparse_free(mp);
535 mpages_free();
536 ohash_delete(&mpages);
537 ohash_delete(&mlinks);
538 return(exitcode);
539 usage:
540 fprintf(stderr, "usage: %s [-aDnpQ] [-C file] [-Tutf8]\n"
541 " %s [-aDnpQ] [-Tutf8] dir ...\n"
542 " %s [-DnpQ] [-Tutf8] -d dir [file ...]\n"
543 " %s [-Dnp] -u dir [file ...]\n"
544 " %s [-Q] -t file ...\n",
545 progname, progname, progname,
546 progname, progname);
547
548 return((int)MANDOCLEVEL_BADARG);
549 }
550
551 /*
552 * Scan a directory tree rooted at "basedir" for manpages.
553 * We use fts(), scanning directory parts along the way for clues to our
554 * section and architecture.
555 *
556 * If use_all has been specified, grok all files.
557 * If not, sanitise paths to the following:
558 *
559 * [./]man*[/<arch>]/<name>.<section>
560 * or
561 * [./]cat<section>[/<arch>]/<name>.0
562 *
563 * TODO: accomodate for multi-language directories.
564 */
565 static int
566 treescan(void)
567 {
568 char buf[PATH_MAX];
569 FTS *f;
570 FTSENT *ff;
571 struct mlink *mlink;
572 int dform, gzip;
573 char *dsec, *arch, *fsec, *cp;
574 const char *path;
575 const char *argv[2];
576
577 argv[0] = ".";
578 argv[1] = (char *)NULL;
579
580 f = fts_open((char * const *)argv,
581 FTS_PHYSICAL | FTS_NOCHDIR, NULL);
582 if (NULL == f) {
583 exitcode = (int)MANDOCLEVEL_SYSERR;
584 say("", "&fts_open");
585 return(0);
586 }
587
588 dsec = arch = NULL;
589 dform = FORM_NONE;
590
591 while (NULL != (ff = fts_read(f))) {
592 path = ff->fts_path + 2;
593 switch (ff->fts_info) {
594
595 /*
596 * Symbolic links require various sanity checks,
597 * then get handled just like regular files.
598 */
599 case FTS_SL:
600 if (NULL == realpath(path, buf)) {
601 if (warnings)
602 say(path, "&realpath");
603 continue;
604 }
605 if (strstr(buf, basedir) != buf) {
606 if (warnings) say("",
607 "%s: outside base directory", buf);
608 continue;
609 }
610 /* Use logical inode to avoid mpages dupe. */
611 if (-1 == stat(path, ff->fts_statp)) {
612 if (warnings)
613 say(path, "&stat");
614 continue;
615 }
616 /* FALLTHROUGH */
617
618 /*
619 * If we're a regular file, add an mlink by using the
620 * stored directory data and handling the filename.
621 */
622 case FTS_F:
623 if (0 == strcmp(path, MANDOC_DB))
624 continue;
625 if ( ! use_all && ff->fts_level < 2) {
626 if (warnings)
627 say(path, "Extraneous file");
628 continue;
629 }
630 gzip = 0;
631 fsec = NULL;
632 while (NULL == fsec) {
633 fsec = strrchr(ff->fts_name, '.');
634 if (NULL == fsec || strcmp(fsec+1, "gz"))
635 break;
636 gzip = 1;
637 *fsec = '\0';
638 fsec = NULL;
639 }
640 if (NULL == fsec) {
641 if ( ! use_all) {
642 if (warnings)
643 say(path,
644 "No filename suffix");
645 continue;
646 }
647 } else if (0 == strcmp(++fsec, "html")) {
648 if (warnings)
649 say(path, "Skip html");
650 continue;
651 } else if (0 == strcmp(fsec, "ps")) {
652 if (warnings)
653 say(path, "Skip ps");
654 continue;
655 } else if (0 == strcmp(fsec, "pdf")) {
656 if (warnings)
657 say(path, "Skip pdf");
658 continue;
659 } else if ( ! use_all &&
660 ((FORM_SRC == dform && strcmp(fsec, dsec)) ||
661 (FORM_CAT == dform && strcmp(fsec, "0")))) {
662 if (warnings)
663 say(path, "Wrong filename suffix");
664 continue;
665 } else
666 fsec[-1] = '\0';
667
668 mlink = mandoc_calloc(1, sizeof(struct mlink));
669 if (strlcpy(mlink->file, path,
670 sizeof(mlink->file)) >=
671 sizeof(mlink->file)) {
672 say(path, "Filename too long");
673 free(mlink);
674 continue;
675 }
676 mlink->dform = dform;
677 mlink->dsec = dsec;
678 mlink->arch = arch;
679 mlink->name = ff->fts_name;
680 mlink->fsec = fsec;
681 mlink->gzip = gzip;
682 mlink_add(mlink, ff->fts_statp);
683 continue;
684
685 case FTS_D:
686 /* FALLTHROUGH */
687 case FTS_DP:
688 break;
689
690 default:
691 if (warnings)
692 say(path, "Not a regular file");
693 continue;
694 }
695
696 switch (ff->fts_level) {
697 case 0:
698 /* Ignore the root directory. */
699 break;
700 case 1:
701 /*
702 * This might contain manX/ or catX/.
703 * Try to infer this from the name.
704 * If we're not in use_all, enforce it.
705 */
706 cp = ff->fts_name;
707 if (FTS_DP == ff->fts_info)
708 break;
709
710 if (0 == strncmp(cp, "man", 3)) {
711 dform = FORM_SRC;
712 dsec = cp + 3;
713 } else if (0 == strncmp(cp, "cat", 3)) {
714 dform = FORM_CAT;
715 dsec = cp + 3;
716 } else {
717 dform = FORM_NONE;
718 dsec = NULL;
719 }
720
721 if (NULL != dsec || use_all)
722 break;
723
724 if (warnings)
725 say(path, "Unknown directory part");
726 fts_set(f, ff, FTS_SKIP);
727 break;
728 case 2:
729 /*
730 * Possibly our architecture.
731 * If we're descending, keep tabs on it.
732 */
733 if (FTS_DP != ff->fts_info && NULL != dsec)
734 arch = ff->fts_name;
735 else
736 arch = NULL;
737 break;
738 default:
739 if (FTS_DP == ff->fts_info || use_all)
740 break;
741 if (warnings)
742 say(path, "Extraneous directory part");
743 fts_set(f, ff, FTS_SKIP);
744 break;
745 }
746 }
747
748 fts_close(f);
749 return(1);
750 }
751
752 /*
753 * Add a file to the mlinks table.
754 * Do not verify that it's a "valid" looking manpage (we'll do that
755 * later).
756 *
757 * Try to infer the manual section, architecture, and page name from the
758 * path, assuming it looks like
759 *
760 * [./]man*[/<arch>]/<name>.<section>
761 * or
762 * [./]cat<section>[/<arch>]/<name>.0
763 *
764 * See treescan() for the fts(3) version of this.
765 */
766 static void
767 filescan(const char *file)
768 {
769 char buf[PATH_MAX];
770 struct stat st;
771 struct mlink *mlink;
772 char *p, *start;
773
774 assert(use_all);
775
776 if (0 == strncmp(file, "./", 2))
777 file += 2;
778
779 /*
780 * We have to do lstat(2) before realpath(3) loses
781 * the information whether this is a symbolic link.
782 * We need to know that because for symbolic links,
783 * we want to use the orginal file name, while for
784 * regular files, we want to use the real path.
785 */
786 if (-1 == lstat(file, &st)) {
787 exitcode = (int)MANDOCLEVEL_BADARG;
788 say(file, "&lstat");
789 return;
790 } else if (0 == ((S_IFREG | S_IFLNK) & st.st_mode)) {
791 exitcode = (int)MANDOCLEVEL_BADARG;
792 say(file, "Not a regular file");
793 return;
794 }
795
796 /*
797 * We have to resolve the file name to the real path
798 * in any case for the base directory check.
799 */
800 if (NULL == realpath(file, buf)) {
801 exitcode = (int)MANDOCLEVEL_BADARG;
802 say(file, "&realpath");
803 return;
804 }
805
806 if (OP_TEST == op)
807 start = buf;
808 else if (strstr(buf, basedir) == buf)
809 start = buf + strlen(basedir);
810 else {
811 exitcode = (int)MANDOCLEVEL_BADARG;
812 say("", "%s: outside base directory", buf);
813 return;
814 }
815
816 /*
817 * Now we are sure the file is inside our tree.
818 * If it is a symbolic link, ignore the real path
819 * and use the original name.
820 * This implies passing stuff like "cat1/../man1/foo.1"
821 * on the command line won't work. So don't do that.
822 * Note the stat(2) can still fail if the link target
823 * doesn't exist.
824 */
825 if (S_IFLNK & st.st_mode) {
826 if (-1 == stat(buf, &st)) {
827 exitcode = (int)MANDOCLEVEL_BADARG;
828 say(file, "&stat");
829 return;
830 }
831 if (strlcpy(buf, file, sizeof(buf)) >= sizeof(buf)) {
832 say(file, "Filename too long");
833 return;
834 }
835 start = buf;
836 if (OP_TEST != op && strstr(buf, basedir) == buf)
837 start += strlen(basedir);
838 }
839
840 mlink = mandoc_calloc(1, sizeof(struct mlink));
841 if (strlcpy(mlink->file, start, sizeof(mlink->file)) >=
842 sizeof(mlink->file)) {
843 say(start, "Filename too long");
844 return;
845 }
846
847 /*
848 * First try to guess our directory structure.
849 * If we find a separator, try to look for man* or cat*.
850 * If we find one of these and what's underneath is a directory,
851 * assume it's an architecture.
852 */
853 if (NULL != (p = strchr(start, '/'))) {
854 *p++ = '\0';
855 if (0 == strncmp(start, "man", 3)) {
856 mlink->dform = FORM_SRC;
857 mlink->dsec = start + 3;
858 } else if (0 == strncmp(start, "cat", 3)) {
859 mlink->dform = FORM_CAT;
860 mlink->dsec = start + 3;
861 }
862
863 start = p;
864 if (NULL != mlink->dsec && NULL != (p = strchr(start, '/'))) {
865 *p++ = '\0';
866 mlink->arch = start;
867 start = p;
868 }
869 }
870
871 /*
872 * Now check the file suffix.
873 * Suffix of `.0' indicates a catpage, `.1-9' is a manpage.
874 */
875 p = strrchr(start, '\0');
876 while (p-- > start && '/' != *p && '.' != *p)
877 /* Loop. */ ;
878
879 if ('.' == *p) {
880 *p++ = '\0';
881 mlink->fsec = p;
882 }
883
884 /*
885 * Now try to parse the name.
886 * Use the filename portion of the path.
887 */
888 mlink->name = start;
889 if (NULL != (p = strrchr(start, '/'))) {
890 mlink->name = p + 1;
891 *p = '\0';
892 }
893 mlink_add(mlink, &st);
894 }
895
896 static void
897 mlink_add(struct mlink *mlink, const struct stat *st)
898 {
899 struct inodev inodev;
900 struct mpage *mpage;
901 unsigned int slot;
902
903 assert(NULL != mlink->file);
904
905 mlink->dsec = mandoc_strdup(mlink->dsec ? mlink->dsec : "");
906 mlink->arch = mandoc_strdup(mlink->arch ? mlink->arch : "");
907 mlink->name = mandoc_strdup(mlink->name ? mlink->name : "");
908 mlink->fsec = mandoc_strdup(mlink->fsec ? mlink->fsec : "");
909
910 if ('0' == *mlink->fsec) {
911 free(mlink->fsec);
912 mlink->fsec = mandoc_strdup(mlink->dsec);
913 mlink->fform = FORM_CAT;
914 } else if ('1' <= *mlink->fsec && '9' >= *mlink->fsec)
915 mlink->fform = FORM_SRC;
916 else
917 mlink->fform = FORM_NONE;
918
919 slot = ohash_qlookup(&mlinks, mlink->file);
920 assert(NULL == ohash_find(&mlinks, slot));
921 ohash_insert(&mlinks, slot, mlink);
922
923 inodev.st_ino = st->st_ino;
924 inodev.st_dev = st->st_dev;
925 slot = ohash_lookup_memory(&mpages, (char *)&inodev,
926 sizeof(struct inodev), inodev.st_ino);
927 mpage = ohash_find(&mpages, slot);
928 if (NULL == mpage) {
929 mpage = mandoc_calloc(1, sizeof(struct mpage));
930 mpage->inodev.st_ino = inodev.st_ino;
931 mpage->inodev.st_dev = inodev.st_dev;
932 ohash_insert(&mpages, slot, mpage);
933 } else
934 mlink->next = mpage->mlinks;
935 mpage->mlinks = mlink;
936 mlink->mpage = mpage;
937 }
938
939 static void
940 mlink_free(struct mlink *mlink)
941 {
942
943 free(mlink->dsec);
944 free(mlink->arch);
945 free(mlink->name);
946 free(mlink->fsec);
947 free(mlink);
948 }
949
950 static void
951 mpages_free(void)
952 {
953 struct mpage *mpage;
954 struct mlink *mlink;
955 unsigned int slot;
956
957 mpage = ohash_first(&mpages, &slot);
958 while (NULL != mpage) {
959 while (NULL != (mlink = mpage->mlinks)) {
960 mpage->mlinks = mlink->next;
961 mlink_free(mlink);
962 }
963 free(mpage->sec);
964 free(mpage->arch);
965 free(mpage->title);
966 free(mpage->desc);
967 free(mpage);
968 mpage = ohash_next(&mpages, &slot);
969 }
970 }
971
972 /*
973 * For each mlink to the mpage, check whether the path looks like
974 * it is formatted, and if it does, check whether a source manual
975 * exists by the same name, ignoring the suffix.
976 * If both conditions hold, drop the mlink.
977 */
978 static void
979 mlinks_undupe(struct mpage *mpage)
980 {
981 char buf[PATH_MAX];
982 struct mlink **prev;
983 struct mlink *mlink;
984 char *bufp;
985
986 mpage->form = FORM_CAT;
987 prev = &mpage->mlinks;
988 while (NULL != (mlink = *prev)) {
989 if (FORM_CAT != mlink->dform) {
990 mpage->form = FORM_NONE;
991 goto nextlink;
992 }
993 (void)strlcpy(buf, mlink->file, sizeof(buf));
994 bufp = strstr(buf, "cat");
995 assert(NULL != bufp);
996 memcpy(bufp, "man", 3);
997 if (NULL != (bufp = strrchr(buf, '.')))
998 *++bufp = '\0';
999 (void)strlcat(buf, mlink->dsec, sizeof(buf));
1000 if (NULL == ohash_find(&mlinks,
1001 ohash_qlookup(&mlinks, buf)))
1002 goto nextlink;
1003 if (warnings)
1004 say(mlink->file, "Man source exists: %s", buf);
1005 if (use_all)
1006 goto nextlink;
1007 *prev = mlink->next;
1008 mlink_free(mlink);
1009 continue;
1010 nextlink:
1011 prev = &(*prev)->next;
1012 }
1013 }
1014
1015 static void
1016 mlink_check(struct mpage *mpage, struct mlink *mlink)
1017 {
1018 struct str *str;
1019 unsigned int slot;
1020
1021 /*
1022 * Check whether the manual section given in a file
1023 * agrees with the directory where the file is located.
1024 * Some manuals have suffixes like (3p) on their
1025 * section number either inside the file or in the
1026 * directory name, some are linked into more than one
1027 * section, like encrypt(1) = makekey(8).
1028 */
1029
1030 if (FORM_SRC == mpage->form &&
1031 strcasecmp(mpage->sec, mlink->dsec))
1032 say(mlink->file, "Section \"%s\" manual in %s directory",
1033 mpage->sec, mlink->dsec);
1034
1035 /*
1036 * Manual page directories exist for each kernel
1037 * architecture as returned by machine(1).
1038 * However, many manuals only depend on the
1039 * application architecture as returned by arch(1).
1040 * For example, some (2/ARM) manuals are shared
1041 * across the "armish" and "zaurus" kernel
1042 * architectures.
1043 * A few manuals are even shared across completely
1044 * different architectures, for example fdformat(1)
1045 * on amd64, i386, sparc, and sparc64.
1046 */
1047
1048 if (strcasecmp(mpage->arch, mlink->arch))
1049 say(mlink->file, "Architecture \"%s\" manual in "
1050 "\"%s\" directory", mpage->arch, mlink->arch);
1051
1052 /*
1053 * XXX
1054 * parse_cat() doesn't set NAME_TITLE yet.
1055 */
1056
1057 if (FORM_CAT == mpage->form)
1058 return;
1059
1060 /*
1061 * Check whether this mlink
1062 * appears as a name in the NAME section.
1063 */
1064
1065 slot = ohash_qlookup(&names, mlink->name);
1066 str = ohash_find(&names, slot);
1067 assert(NULL != str);
1068 if ( ! (NAME_TITLE & str->mask))
1069 say(mlink->file, "Name missing in NAME section");
1070 }
1071
1072 /*
1073 * Run through the files in the global vector "mpages"
1074 * and add them to the database specified in "basedir".
1075 *
1076 * This handles the parsing scheme itself, using the cues of directory
1077 * and filename to determine whether the file is parsable or not.
1078 */
1079 static void
1080 mpages_merge(struct mchars *mc, struct mparse *mp)
1081 {
1082 char any[] = "any";
1083 struct ohash_info str_info;
1084 int fd[2];
1085 struct mpage *mpage, *mpage_dest;
1086 struct mlink *mlink, *mlink_dest;
1087 struct mdoc *mdoc;
1088 struct man *man;
1089 char *sodest;
1090 char *cp;
1091 pid_t child_pid;
1092 int status;
1093 unsigned int pslot;
1094 enum mandoclevel lvl;
1095
1096 str_info.alloc = hash_alloc;
1097 str_info.calloc = hash_calloc;
1098 str_info.free = hash_free;
1099 str_info.key_offset = offsetof(struct str, key);
1100
1101 if (0 == nodb)
1102 SQL_EXEC("BEGIN TRANSACTION");
1103
1104 mpage = ohash_first(&mpages, &pslot);
1105 while (NULL != mpage) {
1106 mlinks_undupe(mpage);
1107 if (NULL == mpage->mlinks) {
1108 mpage = ohash_next(&mpages, &pslot);
1109 continue;
1110 }
1111
1112 name_mask = NAME_MASK;
1113 ohash_init(&names, 4, &str_info);
1114 ohash_init(&strings, 6, &str_info);
1115 mparse_reset(mp);
1116 mdoc = NULL;
1117 man = NULL;
1118 sodest = NULL;
1119 child_pid = 0;
1120 fd[0] = -1;
1121 fd[1] = -1;
1122
1123 if (mpage->mlinks->gzip) {
1124 if (-1 == pipe(fd)) {
1125 exitcode = (int)MANDOCLEVEL_SYSERR;
1126 say(mpage->mlinks->file, "&pipe gunzip");
1127 goto nextpage;
1128 }
1129 switch (child_pid = fork()) {
1130 case -1:
1131 exitcode = (int)MANDOCLEVEL_SYSERR;
1132 say(mpage->mlinks->file, "&fork gunzip");
1133 child_pid = 0;
1134 close(fd[1]);
1135 close(fd[0]);
1136 goto nextpage;
1137 case 0:
1138 close(fd[0]);
1139 if (-1 == dup2(fd[1], STDOUT_FILENO)) {
1140 say(mpage->mlinks->file,
1141 "&dup gunzip");
1142 exit(1);
1143 }
1144 execlp("gunzip", "gunzip", "-c",
1145 mpage->mlinks->file, NULL);
1146 say(mpage->mlinks->file, "&exec gunzip");
1147 exit(1);
1148 default:
1149 close(fd[1]);
1150 break;
1151 }
1152 }
1153
1154 /*
1155 * Try interpreting the file as mdoc(7) or man(7)
1156 * source code, unless it is already known to be
1157 * formatted. Fall back to formatted mode.
1158 */
1159 if (FORM_CAT != mpage->mlinks->dform ||
1160 FORM_CAT != mpage->mlinks->fform) {
1161 lvl = mparse_readfd(mp, fd[0], mpage->mlinks->file);
1162 if (lvl < MANDOCLEVEL_FATAL)
1163 mparse_result(mp, &mdoc, &man, &sodest);
1164 }
1165
1166 if (NULL != sodest) {
1167 mlink_dest = ohash_find(&mlinks,
1168 ohash_qlookup(&mlinks, sodest));
1169 if (NULL != mlink_dest) {
1170
1171 /* The .so target exists. */
1172
1173 mpage_dest = mlink_dest->mpage;
1174 mlink = mpage->mlinks;
1175 while (1) {
1176 mlink->mpage = mpage_dest;
1177
1178 /*
1179 * If the target was already
1180 * processed, add the links
1181 * to the database now.
1182 * Otherwise, this will
1183 * happen when we come
1184 * to the target.
1185 */
1186
1187 if (mpage_dest->pageid)
1188 dbadd_mlink(mlink);
1189
1190 if (NULL == mlink->next)
1191 break;
1192 mlink = mlink->next;
1193 }
1194
1195 /* Move all links to the target. */
1196
1197 mlink->next = mlink_dest->next;
1198 mlink_dest->next = mpage->mlinks;
1199 mpage->mlinks = NULL;
1200 }
1201 goto nextpage;
1202 } else if (NULL != mdoc) {
1203 mpage->form = FORM_SRC;
1204 mpage->sec = mdoc_meta(mdoc)->msec;
1205 mpage->sec = mandoc_strdup(
1206 NULL == mpage->sec ? "" : mpage->sec);
1207 mpage->arch = mdoc_meta(mdoc)->arch;
1208 mpage->arch = mandoc_strdup(
1209 NULL == mpage->arch ? "" : mpage->arch);
1210 mpage->title =
1211 mandoc_strdup(mdoc_meta(mdoc)->title);
1212 } else if (NULL != man) {
1213 mpage->form = FORM_SRC;
1214 mpage->sec =
1215 mandoc_strdup(man_meta(man)->msec);
1216 mpage->arch =
1217 mandoc_strdup(mpage->mlinks->arch);
1218 mpage->title =
1219 mandoc_strdup(man_meta(man)->title);
1220 } else {
1221 mpage->form = FORM_CAT;
1222 mpage->sec =
1223 mandoc_strdup(mpage->mlinks->dsec);
1224 mpage->arch =
1225 mandoc_strdup(mpage->mlinks->arch);
1226 mpage->title =
1227 mandoc_strdup(mpage->mlinks->name);
1228 }
1229 putkey(mpage, mpage->sec, TYPE_sec);
1230 putkey(mpage, '\0' == *mpage->arch ?
1231 any : mpage->arch, TYPE_arch);
1232
1233 for (mlink = mpage->mlinks; mlink; mlink = mlink->next) {
1234 if ('\0' != *mlink->dsec)
1235 putkey(mpage, mlink->dsec, TYPE_sec);
1236 if ('\0' != *mlink->fsec)
1237 putkey(mpage, mlink->fsec, TYPE_sec);
1238 putkey(mpage, '\0' == *mlink->arch ?
1239 any : mlink->arch, TYPE_arch);
1240 putkey(mpage, mlink->name, NAME_FILE);
1241 }
1242
1243 assert(NULL == mpage->desc);
1244 if (NULL != mdoc) {
1245 if (NULL != (cp = mdoc_meta(mdoc)->name))
1246 putkey(mpage, cp, NAME_HEAD);
1247 parse_mdoc(mpage, mdoc_node(mdoc));
1248 } else if (NULL != man)
1249 parse_man(mpage, man_node(man));
1250 else
1251 parse_cat(mpage, fd[0]);
1252 if (NULL == mpage->desc)
1253 mpage->desc = mandoc_strdup(mpage->mlinks->name);
1254
1255 if (warnings && !use_all)
1256 for (mlink = mpage->mlinks; mlink;
1257 mlink = mlink->next)
1258 mlink_check(mpage, mlink);
1259
1260 dbadd(mpage, mc);
1261
1262 nextpage:
1263 if (child_pid) {
1264 if (-1 == waitpid(child_pid, &status, 0)) {
1265 exitcode = (int)MANDOCLEVEL_SYSERR;
1266 say(mpage->mlinks->file, "&wait gunzip");
1267 } else if (WIFSIGNALED(status)) {
1268 exitcode = (int)MANDOCLEVEL_SYSERR;
1269 say(mpage->mlinks->file,
1270 "gunzip died from signal %d",
1271 WTERMSIG(status));
1272 } else if (WEXITSTATUS(status)) {
1273 exitcode = (int)MANDOCLEVEL_SYSERR;
1274 say(mpage->mlinks->file,
1275 "gunzip failed with code %d",
1276 WEXITSTATUS(status));
1277 }
1278 }
1279 ohash_delete(&strings);
1280 ohash_delete(&names);
1281 mpage = ohash_next(&mpages, &pslot);
1282 }
1283
1284 if (0 == nodb)
1285 SQL_EXEC("END TRANSACTION");
1286 }
1287
1288 static void
1289 names_check(void)
1290 {
1291 sqlite3_stmt *stmt;
1292 const char *name, *sec, *arch, *key;
1293 int irc;
1294
1295 sqlite3_prepare_v2(db,
1296 "SELECT name, sec, arch, key FROM ("
1297 "SELECT name AS key, pageid FROM names "
1298 "WHERE bits & ? AND NOT EXISTS ("
1299 "SELECT pageid FROM mlinks "
1300 "WHERE mlinks.pageid == names.pageid "
1301 "AND mlinks.name == names.name"
1302 ")"
1303 ") JOIN ("
1304 "SELECT sec, arch, name, pageid FROM mlinks "
1305 "GROUP BY pageid"
1306 ") USING (pageid);",
1307 -1, &stmt, NULL);
1308
1309 if (SQLITE_OK != sqlite3_bind_int64(stmt, 1, NAME_TITLE))
1310 say("", "%s", sqlite3_errmsg(db));
1311
1312 while (SQLITE_ROW == (irc = sqlite3_step(stmt))) {
1313 name = (const char *)sqlite3_column_text(stmt, 0);
1314 sec = (const char *)sqlite3_column_text(stmt, 1);
1315 arch = (const char *)sqlite3_column_text(stmt, 2);
1316 key = (const char *)sqlite3_column_text(stmt, 3);
1317 say("", "%s(%s%s%s) lacks mlink \"%s\"", name, sec,
1318 '\0' == *arch ? "" : "/",
1319 '\0' == *arch ? "" : arch, key);
1320 }
1321 sqlite3_finalize(stmt);
1322 }
1323
1324 static void
1325 parse_cat(struct mpage *mpage, int fd)
1326 {
1327 FILE *stream;
1328 char *line, *p, *title;
1329 size_t len, plen, titlesz;
1330
1331 stream = (-1 == fd) ?
1332 fopen(mpage->mlinks->file, "r") :
1333 fdopen(fd, "r");
1334 if (NULL == stream) {
1335 if (-1 != fd)
1336 close(fd);
1337 if (warnings)
1338 say(mpage->mlinks->file, "&fopen");
1339 return;
1340 }
1341
1342 /* Skip to first blank line. */
1343
1344 while (NULL != (line = fgetln(stream, &len)))
1345 if ('\n' == *line)
1346 break;
1347
1348 /*
1349 * Assume the first line that is not indented
1350 * is the first section header. Skip to it.
1351 */
1352
1353 while (NULL != (line = fgetln(stream, &len)))
1354 if ('\n' != *line && ' ' != *line)
1355 break;
1356
1357 /*
1358 * Read up until the next section into a buffer.
1359 * Strip the leading and trailing newline from each read line,
1360 * appending a trailing space.
1361 * Ignore empty (whitespace-only) lines.
1362 */
1363
1364 titlesz = 0;
1365 title = NULL;
1366
1367 while (NULL != (line = fgetln(stream, &len))) {
1368 if (' ' != *line || '\n' != line[len - 1])
1369 break;
1370 while (len > 0 && isspace((unsigned char)*line)) {
1371 line++;
1372 len--;
1373 }
1374 if (1 == len)
1375 continue;
1376 title = mandoc_realloc(title, titlesz + len);
1377 memcpy(title + titlesz, line, len);
1378 titlesz += len;
1379 title[titlesz - 1] = ' ';
1380 }
1381
1382 /*
1383 * If no page content can be found, or the input line
1384 * is already the next section header, or there is no
1385 * trailing newline, reuse the page title as the page
1386 * description.
1387 */
1388
1389 if (NULL == title || '\0' == *title) {
1390 if (warnings)
1391 say(mpage->mlinks->file,
1392 "Cannot find NAME section");
1393 fclose(stream);
1394 free(title);
1395 return;
1396 }
1397
1398 title = mandoc_realloc(title, titlesz + 1);
1399 title[titlesz] = '\0';
1400
1401 /*
1402 * Skip to the first dash.
1403 * Use the remaining line as the description (no more than 70
1404 * bytes).
1405 */
1406
1407 if (NULL != (p = strstr(title, "- "))) {
1408 for (p += 2; ' ' == *p || '\b' == *p; p++)
1409 /* Skip to next word. */ ;
1410 } else {
1411 if (warnings)
1412 say(mpage->mlinks->file,
1413 "No dash in title line");
1414 p = title;
1415 }
1416
1417 plen = strlen(p);
1418
1419 /* Strip backspace-encoding from line. */
1420
1421 while (NULL != (line = memchr(p, '\b', plen))) {
1422 len = line - p;
1423 if (0 == len) {
1424 memmove(line, line + 1, plen--);
1425 continue;
1426 }
1427 memmove(line - 1, line + 1, plen - len);
1428 plen -= 2;
1429 }
1430
1431 mpage->desc = mandoc_strdup(p);
1432 fclose(stream);
1433 free(title);
1434 }
1435
1436 /*
1437 * Put a type/word pair into the word database for this particular file.
1438 */
1439 static void
1440 putkey(const struct mpage *mpage, char *value, uint64_t type)
1441 {
1442 char *cp;
1443
1444 assert(NULL != value);
1445 if (TYPE_arch == type)
1446 for (cp = value; *cp; cp++)
1447 if (isupper((unsigned char)*cp))
1448 *cp = _tolower((unsigned char)*cp);
1449 putkeys(mpage, value, strlen(value), type);
1450 }
1451
1452 /*
1453 * Grok all nodes at or below a certain mdoc node into putkey().
1454 */
1455 static void
1456 putmdockey(const struct mpage *mpage,
1457 const struct mdoc_node *n, uint64_t m)
1458 {
1459
1460 for ( ; NULL != n; n = n->next) {
1461 if (NULL != n->child)
1462 putmdockey(mpage, n->child, m);
1463 if (MDOC_TEXT == n->type)
1464 putkey(mpage, n->string, m);
1465 }
1466 }
1467
1468 static void
1469 parse_man(struct mpage *mpage, const struct man_node *n)
1470 {
1471 const struct man_node *head, *body;
1472 char *start, *title;
1473 char byte;
1474 size_t sz;
1475
1476 if (NULL == n)
1477 return;
1478
1479 /*
1480 * We're only searching for one thing: the first text child in
1481 * the BODY of a NAME section. Since we don't keep track of
1482 * sections in -man, run some hoops to find out whether we're in
1483 * the correct section or not.
1484 */
1485
1486 if (MAN_BODY == n->type && MAN_SH == n->tok) {
1487 body = n;
1488 assert(body->parent);
1489 if (NULL != (head = body->parent->head) &&
1490 1 == head->nchild &&
1491 NULL != (head = (head->child)) &&
1492 MAN_TEXT == head->type &&
1493 0 == strcmp(head->string, "NAME") &&
1494 NULL != body->child) {
1495
1496 /*
1497 * Suck the entire NAME section into memory.
1498 * Yes, we might run away.
1499 * But too many manuals have big, spread-out
1500 * NAME sections over many lines.
1501 */
1502
1503 title = NULL;
1504 man_deroff(&title, body);
1505 if (NULL == title)
1506 return;
1507
1508 /*
1509 * Go through a special heuristic dance here.
1510 * Conventionally, one or more manual names are
1511 * comma-specified prior to a whitespace, then a
1512 * dash, then a description. Try to puzzle out
1513 * the name parts here.
1514 */
1515
1516 start = title;
1517 for ( ;; ) {
1518 sz = strcspn(start, " ,");
1519 if ('\0' == start[sz])
1520 break;
1521
1522 byte = start[sz];
1523 start[sz] = '\0';
1524
1525 /*
1526 * Assume a stray trailing comma in the
1527 * name list if a name begins with a dash.
1528 */
1529
1530 if ('-' == start[0] ||
1531 ('\\' == start[0] && '-' == start[1]))
1532 break;
1533
1534 putkey(mpage, start, NAME_TITLE);
1535
1536 if (' ' == byte) {
1537 start += sz + 1;
1538 break;
1539 }
1540
1541 assert(',' == byte);
1542 start += sz + 1;
1543 while (' ' == *start)
1544 start++;
1545 }
1546
1547 if (start == title) {
1548 putkey(mpage, start, NAME_TITLE);
1549 free(title);
1550 return;
1551 }
1552
1553 while (isspace((unsigned char)*start))
1554 start++;
1555
1556 if (0 == strncmp(start, "-", 1))
1557 start += 1;
1558 else if (0 == strncmp(start, "\\-\\-", 4))
1559 start += 4;
1560 else if (0 == strncmp(start, "\\-", 2))
1561 start += 2;
1562 else if (0 == strncmp(start, "\\(en", 4))
1563 start += 4;
1564 else if (0 == strncmp(start, "\\(em", 4))
1565 start += 4;
1566
1567 while (' ' == *start)
1568 start++;
1569
1570 mpage->desc = mandoc_strdup(start);
1571 free(title);
1572 return;
1573 }
1574 }
1575
1576 for (n = n->child; n; n = n->next) {
1577 if (NULL != mpage->desc)
1578 break;
1579 parse_man(mpage, n);
1580 }
1581 }
1582
1583 static void
1584 parse_mdoc(struct mpage *mpage, const struct mdoc_node *n)
1585 {
1586
1587 assert(NULL != n);
1588 for (n = n->child; NULL != n; n = n->next) {
1589 switch (n->type) {
1590 case MDOC_ELEM:
1591 /* FALLTHROUGH */
1592 case MDOC_BLOCK:
1593 /* FALLTHROUGH */
1594 case MDOC_HEAD:
1595 /* FALLTHROUGH */
1596 case MDOC_BODY:
1597 /* FALLTHROUGH */
1598 case MDOC_TAIL:
1599 if (NULL != mdocs[n->tok].fp)
1600 if (0 == (*mdocs[n->tok].fp)(mpage, n))
1601 break;
1602 if (mdocs[n->tok].mask)
1603 putmdockey(mpage, n->child,
1604 mdocs[n->tok].mask);
1605 break;
1606 default:
1607 assert(MDOC_ROOT != n->type);
1608 continue;
1609 }
1610 if (NULL != n->child)
1611 parse_mdoc(mpage, n);
1612 }
1613 }
1614
1615 static int
1616 parse_mdoc_Fd(struct mpage *mpage, const struct mdoc_node *n)
1617 {
1618 const char *start, *end;
1619 size_t sz;
1620
1621 if (SEC_SYNOPSIS != n->sec ||
1622 NULL == (n = n->child) ||
1623 MDOC_TEXT != n->type)
1624 return(0);
1625
1626 /*
1627 * Only consider those `Fd' macro fields that begin with an
1628 * "inclusion" token (versus, e.g., #define).
1629 */
1630
1631 if (strcmp("#include", n->string))
1632 return(0);
1633
1634 if (NULL == (n = n->next) || MDOC_TEXT != n->type)
1635 return(0);
1636
1637 /*
1638 * Strip away the enclosing angle brackets and make sure we're
1639 * not zero-length.
1640 */
1641
1642 start = n->string;
1643 if ('<' == *start || '"' == *start)
1644 start++;
1645
1646 if (0 == (sz = strlen(start)))
1647 return(0);
1648
1649 end = &start[(int)sz - 1];
1650 if ('>' == *end || '"' == *end)
1651 end--;
1652
1653 if (end > start)
1654 putkeys(mpage, start, end - start + 1, TYPE_In);
1655 return(0);
1656 }
1657
1658 static int
1659 parse_mdoc_Fn(struct mpage *mpage, const struct mdoc_node *n)
1660 {
1661 char *cp;
1662
1663 if (NULL == (n = n->child) || MDOC_TEXT != n->type)
1664 return(0);
1665
1666 /*
1667 * Parse: .Fn "struct type *name" "char *arg".
1668 * First strip away pointer symbol.
1669 * Then store the function name, then type.
1670 * Finally, store the arguments.
1671 */
1672
1673 if (NULL == (cp = strrchr(n->string, ' ')))
1674 cp = n->string;
1675
1676 while ('*' == *cp)
1677 cp++;
1678
1679 putkey(mpage, cp, TYPE_Fn);
1680
1681 if (n->string < cp)
1682 putkeys(mpage, n->string, cp - n->string, TYPE_Ft);
1683
1684 for (n = n->next; NULL != n; n = n->next)
1685 if (MDOC_TEXT == n->type)
1686 putkey(mpage, n->string, TYPE_Fa);
1687
1688 return(0);
1689 }
1690
1691 static int
1692 parse_mdoc_Xr(struct mpage *mpage, const struct mdoc_node *n)
1693 {
1694 char *cp;
1695
1696 if (NULL == (n = n->child))
1697 return(0);
1698
1699 if (NULL == n->next) {
1700 putkey(mpage, n->string, TYPE_Xr);
1701 return(0);
1702 }
1703
1704 mandoc_asprintf(&cp, "%s(%s)", n->string, n->next->string);
1705 putkey(mpage, cp, TYPE_Xr);
1706 free(cp);
1707 return(0);
1708 }
1709
1710 static int
1711 parse_mdoc_Nd(struct mpage *mpage, const struct mdoc_node *n)
1712 {
1713
1714 if (MDOC_BODY == n->type)
1715 mdoc_deroff(&mpage->desc, n);
1716 return(0);
1717 }
1718
1719 static int
1720 parse_mdoc_Nm(struct mpage *mpage, const struct mdoc_node *n)
1721 {
1722
1723 if (SEC_NAME == n->sec)
1724 putmdockey(mpage, n->child, NAME_TITLE);
1725 else if (SEC_SYNOPSIS == n->sec && MDOC_HEAD == n->type)
1726 putmdockey(mpage, n->child, NAME_SYN);
1727 return(0);
1728 }
1729
1730 static int
1731 parse_mdoc_Sh(struct mpage *mpage, const struct mdoc_node *n)
1732 {
1733
1734 return(SEC_CUSTOM == n->sec && MDOC_HEAD == n->type);
1735 }
1736
1737 static int
1738 parse_mdoc_head(struct mpage *mpage, const struct mdoc_node *n)
1739 {
1740
1741 return(MDOC_HEAD == n->type);
1742 }
1743
1744 static int
1745 parse_mdoc_body(struct mpage *mpage, const struct mdoc_node *n)
1746 {
1747
1748 return(MDOC_BODY == n->type);
1749 }
1750
1751 /*
1752 * Add a string to the hash table for the current manual.
1753 * Each string has a bitmask telling which macros it belongs to.
1754 * When we finish the manual, we'll dump the table.
1755 */
1756 static void
1757 putkeys(const struct mpage *mpage,
1758 const char *cp, size_t sz, uint64_t v)
1759 {
1760 struct ohash *htab;
1761 struct str *s;
1762 const char *end;
1763 unsigned int slot;
1764 int i;
1765
1766 if (0 == sz)
1767 return;
1768
1769 if (TYPE_Nm & v) {
1770 htab = &names;
1771 v &= name_mask;
1772 name_mask &= ~NAME_FIRST;
1773 if (debug > 1)
1774 say(mpage->mlinks->file,
1775 "Adding name %*s", sz, cp);
1776 } else {
1777 htab = &strings;
1778 if (debug > 1)
1779 for (i = 0; i < mansearch_keymax; i++)
1780 if (1 << i & v)
1781 say(mpage->mlinks->file,
1782 "Adding key %s=%*s",
1783 mansearch_keynames[i], sz, cp);
1784 }
1785
1786 end = cp + sz;
1787 slot = ohash_qlookupi(htab, cp, &end);
1788 s = ohash_find(htab, slot);
1789
1790 if (NULL != s && mpage == s->mpage) {
1791 s->mask |= v;
1792 return;
1793 } else if (NULL == s) {
1794 s = mandoc_calloc(1, sizeof(struct str) + sz + 1);
1795 memcpy(s->key, cp, sz);
1796 ohash_insert(htab, slot, s);
1797 }
1798 s->mpage = mpage;
1799 s->mask = v;
1800 }
1801
1802 /*
1803 * Take a Unicode codepoint and produce its UTF-8 encoding.
1804 * This isn't the best way to do this, but it works.
1805 * The magic numbers are from the UTF-8 packaging.
1806 * They're not as scary as they seem: read the UTF-8 spec for details.
1807 */
1808 static size_t
1809 utf8(unsigned int cp, char out[7])
1810 {
1811 size_t rc;
1812
1813 rc = 0;
1814 if (cp <= 0x0000007F) {
1815 rc = 1;
1816 out[0] = (char)cp;
1817 } else if (cp <= 0x000007FF) {
1818 rc = 2;
1819 out[0] = (cp >> 6 & 31) | 192;
1820 out[1] = (cp & 63) | 128;
1821 } else if (cp <= 0x0000FFFF) {
1822 rc = 3;
1823 out[0] = (cp >> 12 & 15) | 224;
1824 out[1] = (cp >> 6 & 63) | 128;
1825 out[2] = (cp & 63) | 128;
1826 } else if (cp <= 0x001FFFFF) {
1827 rc = 4;
1828 out[0] = (cp >> 18 & 7) | 240;
1829 out[1] = (cp >> 12 & 63) | 128;
1830 out[2] = (cp >> 6 & 63) | 128;
1831 out[3] = (cp & 63) | 128;
1832 } else if (cp <= 0x03FFFFFF) {
1833 rc = 5;
1834 out[0] = (cp >> 24 & 3) | 248;
1835 out[1] = (cp >> 18 & 63) | 128;
1836 out[2] = (cp >> 12 & 63) | 128;
1837 out[3] = (cp >> 6 & 63) | 128;
1838 out[4] = (cp & 63) | 128;
1839 } else if (cp <= 0x7FFFFFFF) {
1840 rc = 6;
1841 out[0] = (cp >> 30 & 1) | 252;
1842 out[1] = (cp >> 24 & 63) | 128;
1843 out[2] = (cp >> 18 & 63) | 128;
1844 out[3] = (cp >> 12 & 63) | 128;
1845 out[4] = (cp >> 6 & 63) | 128;
1846 out[5] = (cp & 63) | 128;
1847 } else
1848 return(0);
1849
1850 out[rc] = '\0';
1851 return(rc);
1852 }
1853
1854 /*
1855 * Store the rendered version of a key, or alias the pointer
1856 * if the key contains no escape sequences.
1857 */
1858 static void
1859 render_key(struct mchars *mc, struct str *key)
1860 {
1861 size_t sz, bsz, pos;
1862 char utfbuf[7], res[6];
1863 char *buf;
1864 const char *seq, *cpp, *val;
1865 int len, u;
1866 enum mandoc_esc esc;
1867
1868 assert(NULL == key->rendered);
1869
1870 res[0] = '\\';
1871 res[1] = '\t';
1872 res[2] = ASCII_NBRSP;
1873 res[3] = ASCII_HYPH;
1874 res[4] = ASCII_BREAK;
1875 res[5] = '\0';
1876
1877 val = key->key;
1878 bsz = strlen(val);
1879
1880 /*
1881 * Pre-check: if we have no stop-characters, then set the
1882 * pointer as ourselvse and get out of here.
1883 */
1884 if (strcspn(val, res) == bsz) {
1885 key->rendered = key->key;
1886 return;
1887 }
1888
1889 /* Pre-allocate by the length of the input */
1890
1891 buf = mandoc_malloc(++bsz);
1892 pos = 0;
1893
1894 while ('\0' != *val) {
1895 /*
1896 * Halt on the first escape sequence.
1897 * This also halts on the end of string, in which case
1898 * we just copy, fallthrough, and exit the loop.
1899 */
1900 if ((sz = strcspn(val, res)) > 0) {
1901 memcpy(&buf[pos], val, sz);
1902 pos += sz;
1903 val += sz;
1904 }
1905
1906 switch (*val) {
1907 case ASCII_HYPH:
1908 buf[pos++] = '-';
1909 val++;
1910 continue;
1911 case '\t':
1912 /* FALLTHROUGH */
1913 case ASCII_NBRSP:
1914 buf[pos++] = ' ';
1915 val++;
1916 /* FALLTHROUGH */
1917 case ASCII_BREAK:
1918 continue;
1919 default:
1920 break;
1921 }
1922 if ('\\' != *val)
1923 break;
1924
1925 /* Read past the slash. */
1926
1927 val++;
1928
1929 /*
1930 * Parse the escape sequence and see if it's a
1931 * predefined character or special character.
1932 */
1933
1934 esc = mandoc_escape((const char **)&val,
1935 &seq, &len);
1936 if (ESCAPE_ERROR == esc)
1937 break;
1938 if (ESCAPE_SPECIAL != esc)
1939 continue;
1940
1941 /*
1942 * Render the special character
1943 * as either UTF-8 or ASCII.
1944 */
1945
1946 if (write_utf8) {
1947 if (0 == (u = mchars_spec2cp(mc, seq, len)))
1948 continue;
1949 cpp = utfbuf;
1950 if (0 == (sz = utf8(u, utfbuf)))
1951 continue;
1952 sz = strlen(cpp);
1953 } else {
1954 cpp = mchars_spec2str(mc, seq, len, &sz);
1955 if (NULL == cpp)
1956 continue;
1957 if (ASCII_NBRSP == *cpp) {
1958 cpp = " ";
1959 sz = 1;
1960 }
1961 }
1962
1963 /* Copy the rendered glyph into the stream. */
1964
1965 bsz += sz;
1966 buf = mandoc_realloc(buf, bsz);
1967 memcpy(&buf[pos], cpp, sz);
1968 pos += sz;
1969 }
1970
1971 buf[pos] = '\0';
1972 key->rendered = buf;
1973 }
1974
1975 static void
1976 dbadd_mlink(const struct mlink *mlink)
1977 {
1978 size_t i;
1979
1980 i = 1;
1981 SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->dsec);
1982 SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->arch);
1983 SQL_BIND_TEXT(stmts[STMT_INSERT_LINK], i, mlink->name);
1984 SQL_BIND_INT64(stmts[STMT_INSERT_LINK], i, mlink->mpage->pageid);
1985 SQL_STEP(stmts[STMT_INSERT_LINK]);
1986 sqlite3_reset(stmts[STMT_INSERT_LINK]);
1987
1988 i = 1;
1989 SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, NAME_FILE);
1990 SQL_BIND_TEXT(stmts[STMT_INSERT_NAME], i, mlink->name);
1991 SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, mlink->mpage->pageid);
1992 SQL_STEP(stmts[STMT_INSERT_NAME]);
1993 sqlite3_reset(stmts[STMT_INSERT_NAME]);
1994 }
1995
1996 /*
1997 * Flush the current page's terms (and their bits) into the database.
1998 * Wrap the entire set of additions in a transaction to make sqlite be a
1999 * little faster.
2000 * Also, handle escape sequences at the last possible moment.
2001 */
2002 static void
2003 dbadd(struct mpage *mpage, struct mchars *mc)
2004 {
2005 struct mlink *mlink;
2006 struct str *key;
2007 size_t i;
2008 unsigned int slot;
2009
2010 mlink = mpage->mlinks;
2011
2012 if (nodb) {
2013 for (key = ohash_first(&names, &slot); NULL != key;
2014 key = ohash_next(&names, &slot)) {
2015 if (key->rendered != key->key)
2016 free(key->rendered);
2017 free(key);
2018 }
2019 for (key = ohash_first(&strings, &slot); NULL != key;
2020 key = ohash_next(&strings, &slot)) {
2021 if (key->rendered != key->key)
2022 free(key->rendered);
2023 free(key);
2024 }
2025 if (0 == debug)
2026 return;
2027 while (NULL != mlink) {
2028 fputs(mlink->name, stdout);
2029 if (NULL == mlink->next ||
2030 strcmp(mlink->dsec, mlink->next->dsec) ||
2031 strcmp(mlink->fsec, mlink->next->fsec) ||
2032 strcmp(mlink->arch, mlink->next->arch)) {
2033 putchar('(');
2034 if ('\0' == *mlink->dsec)
2035 fputs(mlink->fsec, stdout);
2036 else
2037 fputs(mlink->dsec, stdout);
2038 if ('\0' != *mlink->arch)
2039 printf("/%s", mlink->arch);
2040 putchar(')');
2041 }
2042 mlink = mlink->next;
2043 if (NULL != mlink)
2044 fputs(", ", stdout);
2045 }
2046 printf(" - %s\n", mpage->desc);
2047 return;
2048 }
2049
2050 if (debug)
2051 say(mlink->file, "Adding to database");
2052
2053 i = strlen(mpage->desc) + 1;
2054 key = mandoc_calloc(1, sizeof(struct str) + i);
2055 memcpy(key->key, mpage->desc, i);
2056 render_key(mc, key);
2057
2058 i = 1;
2059 SQL_BIND_TEXT(stmts[STMT_INSERT_PAGE], i, key->rendered);
2060 SQL_BIND_INT(stmts[STMT_INSERT_PAGE], i, FORM_SRC == mpage->form);
2061 SQL_STEP(stmts[STMT_INSERT_PAGE]);
2062 mpage->pageid = sqlite3_last_insert_rowid(db);
2063 sqlite3_reset(stmts[STMT_INSERT_PAGE]);
2064
2065 if (key->rendered != key->key)
2066 free(key->rendered);
2067 free(key);
2068
2069 while (NULL != mlink) {
2070 dbadd_mlink(mlink);
2071 mlink = mlink->next;
2072 }
2073 mlink = mpage->mlinks;
2074
2075 for (key = ohash_first(&names, &slot); NULL != key;
2076 key = ohash_next(&names, &slot)) {
2077 assert(key->mpage == mpage);
2078 if (NULL == key->rendered)
2079 render_key(mc, key);
2080 i = 1;
2081 SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, key->mask);
2082 SQL_BIND_TEXT(stmts[STMT_INSERT_NAME], i, key->rendered);
2083 SQL_BIND_INT64(stmts[STMT_INSERT_NAME], i, mpage->pageid);
2084 SQL_STEP(stmts[STMT_INSERT_NAME]);
2085 sqlite3_reset(stmts[STMT_INSERT_NAME]);
2086 if (key->rendered != key->key)
2087 free(key->rendered);
2088 free(key);
2089 }
2090 for (key = ohash_first(&strings, &slot); NULL != key;
2091 key = ohash_next(&strings, &slot)) {
2092 assert(key->mpage == mpage);
2093 if (NULL == key->rendered)
2094 render_key(mc, key);
2095 i = 1;
2096 SQL_BIND_INT64(stmts[STMT_INSERT_KEY], i, key->mask);
2097 SQL_BIND_TEXT(stmts[STMT_INSERT_KEY], i, key->rendered);
2098 SQL_BIND_INT64(stmts[STMT_INSERT_KEY], i, mpage->pageid);
2099 SQL_STEP(stmts[STMT_INSERT_KEY]);
2100 sqlite3_reset(stmts[STMT_INSERT_KEY]);
2101 if (key->rendered != key->key)
2102 free(key->rendered);
2103 free(key);
2104 }
2105 }
2106
2107 static void
2108 dbprune(void)
2109 {
2110 struct mpage *mpage;
2111 struct mlink *mlink;
2112 size_t i;
2113 unsigned int slot;
2114
2115 if (0 == nodb)
2116 SQL_EXEC("BEGIN TRANSACTION");
2117
2118 for (mpage = ohash_first(&mpages, &slot); NULL != mpage;
2119 mpage = ohash_next(&mpages, &slot)) {
2120 mlink = mpage->mlinks;
2121 if (debug)
2122 say(mlink->file, "Deleting from database");
2123 if (nodb)
2124 continue;
2125 for ( ; NULL != mlink; mlink = mlink->next) {
2126 i = 1;
2127 SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE],
2128 i, mlink->dsec);
2129 SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE],
2130 i, mlink->arch);
2131 SQL_BIND_TEXT(stmts[STMT_DELETE_PAGE],
2132 i, mlink->name);
2133 SQL_STEP(stmts[STMT_DELETE_PAGE]);
2134 sqlite3_reset(stmts[STMT_DELETE_PAGE]);
2135 }
2136 }
2137
2138 if (0 == nodb)
2139 SQL_EXEC("END TRANSACTION");
2140 }
2141
2142 /*
2143 * Close an existing database and its prepared statements.
2144 * If "real" is not set, rename the temporary file into the real one.
2145 */
2146 static void
2147 dbclose(int real)
2148 {
2149 size_t i;
2150 int status;
2151 pid_t child;
2152
2153 if (nodb)
2154 return;
2155
2156 for (i = 0; i < STMT__MAX; i++) {
2157 sqlite3_finalize(stmts[i]);
2158 stmts[i] = NULL;
2159 }
2160
2161 sqlite3_close(db);
2162 db = NULL;
2163
2164 if (real)
2165 return;
2166
2167 if ('\0' == *tempfilename) {
2168 if (-1 == rename(MANDOC_DB "~", MANDOC_DB)) {
2169 exitcode = (int)MANDOCLEVEL_SYSERR;
2170 say(MANDOC_DB, "&rename");
2171 }
2172 return;
2173 }
2174
2175 switch (child = fork()) {
2176 case -1:
2177 exitcode = (int)MANDOCLEVEL_SYSERR;
2178 say("", "&fork cmp");
2179 return;
2180 case 0:
2181 execlp("cmp", "cmp", "-s",
2182 tempfilename, MANDOC_DB, NULL);
2183 say("", "&exec cmp");
2184 exit(0);
2185 default:
2186 break;
2187 }
2188 if (-1 == waitpid(child, &status, 0)) {
2189 exitcode = (int)MANDOCLEVEL_SYSERR;
2190 say("", "&wait cmp");
2191 } else if (WIFSIGNALED(status)) {
2192 exitcode = (int)MANDOCLEVEL_SYSERR;
2193 say("", "cmp died from signal %d", WTERMSIG(status));
2194 } else if (WEXITSTATUS(status)) {
2195 exitcode = (int)MANDOCLEVEL_SYSERR;
2196 say(MANDOC_DB,
2197 "Data changed, but cannot replace database");
2198 }
2199
2200 *strrchr(tempfilename, '/') = '\0';
2201 switch (child = fork()) {
2202 case -1:
2203 exitcode = (int)MANDOCLEVEL_SYSERR;
2204 say("", "&fork rm");
2205 return;
2206 case 0:
2207 execlp("rm", "rm", "-rf", tempfilename, NULL);
2208 say("", "&exec rm");
2209 exit((int)MANDOCLEVEL_SYSERR);
2210 default:
2211 break;
2212 }
2213 if (-1 == waitpid(child, &status, 0)) {
2214 exitcode = (int)MANDOCLEVEL_SYSERR;
2215 say("", "&wait rm");
2216 } else if (WIFSIGNALED(status) || WEXITSTATUS(status)) {
2217 exitcode = (int)MANDOCLEVEL_SYSERR;
2218 say("", "%s: Cannot remove temporary directory",
2219 tempfilename);
2220 }
2221 }
2222
2223 /*
2224 * This is straightforward stuff.
2225 * Open a database connection to a "temporary" database, then open a set
2226 * of prepared statements we'll use over and over again.
2227 * If "real" is set, we use the existing database; if not, we truncate a
2228 * temporary one.
2229 * Must be matched by dbclose().
2230 */
2231 static int
2232 dbopen(int real)
2233 {
2234 const char *sql;
2235 int rc, ofl;
2236
2237 if (nodb)
2238 return(1);
2239
2240 *tempfilename = '\0';
2241 ofl = SQLITE_OPEN_READWRITE;
2242
2243 if (real) {
2244 rc = sqlite3_open_v2(MANDOC_DB, &db, ofl, NULL);
2245 if (SQLITE_OK != rc) {
2246 exitcode = (int)MANDOCLEVEL_SYSERR;
2247 if (SQLITE_CANTOPEN != rc)
2248 say(MANDOC_DB, "%s", sqlite3_errstr(rc));
2249 return(0);
2250 }
2251 goto prepare_statements;
2252 }
2253
2254 ofl |= SQLITE_OPEN_CREATE | SQLITE_OPEN_EXCLUSIVE;
2255
2256 remove(MANDOC_DB "~");
2257 rc = sqlite3_open_v2(MANDOC_DB "~", &db, ofl, NULL);
2258 if (SQLITE_OK == rc)
2259 goto create_tables;
2260 if (MPARSE_QUICK & mparse_options) {
2261 exitcode = (int)MANDOCLEVEL_SYSERR;
2262 say(MANDOC_DB "~", "%s", sqlite3_errstr(rc));
2263 return(0);
2264 }
2265
2266 (void)strlcpy(tempfilename, "/tmp/mandocdb.XXXXXX",
2267 sizeof(tempfilename));
2268 if (NULL == mkdtemp(tempfilename)) {
2269 exitcode = (int)MANDOCLEVEL_SYSERR;
2270 say("", "&%s", tempfilename);
2271 return(0);
2272 }
2273 (void)strlcat(tempfilename, "/" MANDOC_DB,
2274 sizeof(tempfilename));
2275 rc = sqlite3_open_v2(tempfilename, &db, ofl, NULL);
2276 if (SQLITE_OK != rc) {
2277 exitcode = (int)MANDOCLEVEL_SYSERR;
2278 say("", "%s: %s", tempfilename, sqlite3_errstr(rc));
2279 return(0);
2280 }
2281
2282 create_tables:
2283 sql = "CREATE TABLE \"mpages\" (\n"
2284 " \"desc\" TEXT NOT NULL,\n"
2285 " \"form\" INTEGER NOT NULL,\n"
2286 " \"pageid\" INTEGER PRIMARY KEY AUTOINCREMENT NOT NULL\n"
2287 ");\n"
2288 "\n"
2289 "CREATE TABLE \"mlinks\" (\n"
2290 " \"sec\" TEXT NOT NULL,\n"
2291 " \"arch\" TEXT NOT NULL,\n"
2292 " \"name\" TEXT NOT NULL,\n"
2293 " \"pageid\" INTEGER NOT NULL REFERENCES mpages(pageid) "
2294 "ON DELETE CASCADE\n"
2295 ");\n"
2296 "CREATE INDEX mlinks_pageid_idx ON mlinks (pageid);\n"
2297 "\n"
2298 "CREATE TABLE \"names\" (\n"
2299 " \"bits\" INTEGER NOT NULL,\n"
2300 " \"name\" TEXT NOT NULL,\n"
2301 " \"pageid\" INTEGER NOT NULL REFERENCES mpages(pageid) "
2302 "ON DELETE CASCADE\n"
2303 ");\n"
2304 "\n"
2305 "CREATE TABLE \"keys\" (\n"
2306 " \"bits\" INTEGER NOT NULL,\n"
2307 " \"key\" TEXT NOT NULL,\n"
2308 " \"pageid\" INTEGER NOT NULL REFERENCES mpages(pageid) "
2309 "ON DELETE CASCADE\n"
2310 ");\n"
2311 "CREATE INDEX keys_pageid_idx ON keys (pageid);\n";
2312
2313 if (SQLITE_OK != sqlite3_exec(db, sql, NULL, NULL, NULL)) {
2314 exitcode = (int)MANDOCLEVEL_SYSERR;
2315 say(MANDOC_DB, "%s", sqlite3_errmsg(db));
2316 sqlite3_close(db);
2317 return(0);
2318 }
2319
2320 prepare_statements:
2321 if (SQLITE_OK != sqlite3_exec(db,
2322 "PRAGMA foreign_keys = ON", NULL, NULL, NULL)) {
2323 exitcode = (int)MANDOCLEVEL_SYSERR;
2324 say(MANDOC_DB, "PRAGMA foreign_keys: %s",
2325 sqlite3_errmsg(db));
2326 sqlite3_close(db);
2327 return(0);
2328 }
2329
2330 sql = "DELETE FROM mpages WHERE pageid IN "
2331 "(SELECT pageid FROM mlinks WHERE "
2332 "sec=? AND arch=? AND name=?)";
2333 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_DELETE_PAGE], NULL);
2334 sql = "INSERT INTO mpages "
2335 "(desc,form) VALUES (?,?)";
2336 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_PAGE], NULL);
2337 sql = "INSERT INTO mlinks "
2338 "(sec,arch,name,pageid) VALUES (?,?,?,?)";
2339 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_LINK], NULL);
2340 sql = "INSERT INTO names "
2341 "(bits,name,pageid) VALUES (?,?,?)";
2342 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_NAME], NULL);
2343 sql = "INSERT INTO keys "
2344 "(bits,key,pageid) VALUES (?,?,?)";
2345 sqlite3_prepare_v2(db, sql, -1, &stmts[STMT_INSERT_KEY], NULL);
2346
2347 #ifndef __APPLE__
2348 /*
2349 * When opening a new database, we can turn off
2350 * synchronous mode for much better performance.
2351 */
2352
2353 if (real && SQLITE_OK != sqlite3_exec(db,
2354 "PRAGMA synchronous = OFF", NULL, NULL, NULL)) {
2355 exitcode = (int)MANDOCLEVEL_SYSERR;
2356 say(MANDOC_DB, "PRAGMA synchronous: %s",
2357 sqlite3_errmsg(db));
2358 sqlite3_close(db);
2359 return(0);
2360 }
2361 #endif
2362
2363 return(1);
2364 }
2365
2366 static void *
2367 hash_calloc(size_t n, size_t sz, void *arg)
2368 {
2369
2370 return(mandoc_calloc(n, sz));
2371 }
2372
2373 static void *
2374 hash_alloc(size_t sz, void *arg)
2375 {
2376
2377 return(mandoc_malloc(sz));
2378 }
2379
2380 static void
2381 hash_free(void *p, void *arg)
2382 {
2383
2384 free(p);
2385 }
2386
2387 static int
2388 set_basedir(const char *targetdir)
2389 {
2390 static char startdir[PATH_MAX];
2391 static int getcwd_status; /* 1 = ok, 2 = failure */
2392 static int chdir_status; /* 1 = changed directory */
2393 char *cp;
2394
2395 /*
2396 * Remember the original working directory, if possible.
2397 * This will be needed if the second or a later directory
2398 * on the command line is given as a relative path.
2399 * Do not error out if the current directory is not
2400 * searchable: Maybe it won't be needed after all.
2401 */
2402 if (0 == getcwd_status) {
2403 if (NULL == getcwd(startdir, sizeof(startdir))) {
2404 getcwd_status = 2;
2405 (void)strlcpy(startdir, strerror(errno),
2406 sizeof(startdir));
2407 } else
2408 getcwd_status = 1;
2409 }
2410
2411 /*
2412 * We are leaving the old base directory.
2413 * Do not use it any longer, not even for messages.
2414 */
2415 *basedir = '\0';
2416
2417 /*
2418 * If and only if the directory was changed earlier and
2419 * the next directory to process is given as a relative path,
2420 * first go back, or bail out if that is impossible.
2421 */
2422 if (chdir_status && '/' != *targetdir) {
2423 if (2 == getcwd_status) {
2424 exitcode = (int)MANDOCLEVEL_SYSERR;
2425 say("", "getcwd: %s", startdir);
2426 return(0);
2427 }
2428 if (-1 == chdir(startdir)) {
2429 exitcode = (int)MANDOCLEVEL_SYSERR;
2430 say("", "&chdir %s", startdir);
2431 return(0);
2432 }
2433 }
2434
2435 /*
2436 * Always resolve basedir to the canonicalized absolute
2437 * pathname and append a trailing slash, such that
2438 * we can reliably check whether files are inside.
2439 */
2440 if (NULL == realpath(targetdir, basedir)) {
2441 exitcode = (int)MANDOCLEVEL_BADARG;
2442 say("", "&%s: realpath", targetdir);
2443 return(0);
2444 } else if (-1 == chdir(basedir)) {
2445 exitcode = (int)MANDOCLEVEL_BADARG;
2446 say("", "&chdir");
2447 return(0);
2448 }
2449 chdir_status = 1;
2450 cp = strchr(basedir, '\0');
2451 if ('/' != cp[-1]) {
2452 if (cp - basedir >= PATH_MAX - 1) {
2453 exitcode = (int)MANDOCLEVEL_SYSERR;
2454 say("", "Filename too long");
2455 return(0);
2456 }
2457 *cp++ = '/';
2458 *cp = '\0';
2459 }
2460 return(1);
2461 }
2462
2463 static void
2464 say(const char *file, const char *format, ...)
2465 {
2466 va_list ap;
2467 int use_errno;
2468
2469 if ('\0' != *basedir)
2470 fprintf(stderr, "%s", basedir);
2471 if ('\0' != *basedir && '\0' != *file)
2472 fputc('/', stderr);
2473 if ('\0' != *file)
2474 fprintf(stderr, "%s", file);
2475
2476 use_errno = 1;
2477 if (NULL != format) {
2478 switch (*format) {
2479 case '&':
2480 format++;
2481 break;
2482 case '\0':
2483 format = NULL;
2484 break;
2485 default:
2486 use_errno = 0;
2487 break;
2488 }
2489 }
2490 if (NULL != format) {
2491 if ('\0' != *basedir || '\0' != *file)
2492 fputs(": ", stderr);
2493 va_start(ap, format);
2494 vfprintf(stderr, format, ap);
2495 va_end(ap);
2496 }
2497 if (use_errno) {
2498 if ('\0' != *basedir || '\0' != *file || NULL != format)
2499 fputs(": ", stderr);
2500 perror(NULL);
2501 } else
2502 fputc('\n', stderr);
2503 }
A git conversion mirror of mandoc.bsd.lv