makemandb.c revision 1.17 1 /* $NetBSD: makemandb.c,v 1.17 2013/01/14 18:01:59 christos Exp $ */
2 /*
3 * Copyright (c) 2011 Abhinav Upadhyay <er.abhinav.upadhyay (at) gmail.com>
4 * Copyright (c) 2011 Kristaps Dzonsons <kristaps (at) bsd.lv>
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
19 #include <sys/cdefs.h>
20 __RCSID("$NetBSD: makemandb.c,v 1.17 2013/01/14 18:01:59 christos Exp $");
21
22 #include <sys/stat.h>
23 #include <sys/types.h>
24
25 #include <assert.h>
26 #include <ctype.h>
27 #include <dirent.h>
28 #include <err.h>
29 #include <archive.h>
30 #include <libgen.h>
31 #include <md5.h>
32 #include <stdio.h>
33 #include <stdlib.h>
34 #include <string.h>
35 #include <unistd.h>
36 #include <util.h>
37
38 #include "apropos-utils.h"
39 #include "dist/man.h"
40 #include "dist/mandoc.h"
41 #include "dist/mdoc.h"
42 #include "sqlite3.h"
43
44 #define BUFLEN 1024
45 #define MDOC 0 //If the page is of mdoc(7) type
46 #define MAN 1 //If the page is of man(7) type
47
48 /*
49 * A data structure for holding section specific data.
50 */
51 typedef struct secbuff {
52 char *data;
53 size_t buflen; //Total length of buffer allocated initially
54 size_t offset; // Current offset in the buffer.
55 } secbuff;
56
57 typedef struct makemandb_flags {
58 int optimize;
59 int limit; // limit the indexing to only NAME section
60 int recreate; // Database was created from scratch
61 int verbosity; // 0: quiet, 1: default, 2: verbose
62 } makemandb_flags;
63
64 typedef struct mandb_rec {
65 /* Fields for mandb table */
66 char *name; // for storing the name of the man page
67 char *name_desc; // for storing the one line description (.Nd)
68 secbuff desc; // for storing the DESCRIPTION section
69 secbuff lib; // for the LIBRARY section
70 secbuff return_vals; // RETURN VALUES
71 secbuff env; // ENVIRONMENT
72 secbuff files; // FILES
73 secbuff exit_status; // EXIT STATUS
74 secbuff diagnostics; // DIAGNOSTICS
75 secbuff errors; // ERRORS
76 char section[2];
77
78 int xr_found;
79
80 /* Fields for mandb_meta table */
81 char *md5_hash;
82 dev_t device;
83 ino_t inode;
84 time_t mtime;
85
86 /* Fields for mandb_links table */
87 char *machine;
88 char *links; //all the links to a page in a space separated form
89 char *file_path;
90
91 /* Non-db fields */
92 int page_type; //Indicates the type of page: mdoc or man
93 } mandb_rec;
94
95 static void append(secbuff *sbuff, const char *src);
96 static void init_secbuffs(mandb_rec *);
97 static void free_secbuffs(mandb_rec *);
98 static int check_md5(const char *, sqlite3 *, const char *, char **, void *, size_t);
99 static void cleanup(mandb_rec *);
100 static void set_section(const struct mdoc *, const struct man *, mandb_rec *);
101 static void set_machine(const struct mdoc *, mandb_rec *);
102 static int insert_into_db(sqlite3 *, mandb_rec *);
103 static void begin_parse(const char *, struct mparse *, mandb_rec *,
104 const void *, size_t len);
105 static void pmdoc_node(const struct mdoc_node *, mandb_rec *);
106 static void pmdoc_Nm(const struct mdoc_node *, mandb_rec *);
107 static void pmdoc_Nd(const struct mdoc_node *, mandb_rec *);
108 static void pmdoc_Sh(const struct mdoc_node *, mandb_rec *);
109 static void pmdoc_Xr(const struct mdoc_node *, mandb_rec *);
110 static void pmdoc_Pp(const struct mdoc_node *, mandb_rec *);
111 static void pmdoc_macro_handler(const struct mdoc_node *, mandb_rec *,
112 enum mdoct);
113 static void pman_node(const struct man_node *n, mandb_rec *);
114 static void pman_parse_node(const struct man_node *, secbuff *);
115 static void pman_parse_name(const struct man_node *, mandb_rec *);
116 static void pman_sh(const struct man_node *, mandb_rec *);
117 static void pman_block(const struct man_node *, mandb_rec *);
118 static void traversedir(const char *, const char *, sqlite3 *, struct mparse *);
119 static void mdoc_parse_section(enum mdoc_sec, const char *, mandb_rec *);
120 static void man_parse_section(enum man_sec, const struct man_node *, mandb_rec *);
121 static void build_file_cache(sqlite3 *, const char *, const char *,
122 struct stat *);
123 static void update_db(sqlite3 *, struct mparse *, mandb_rec *);
124 __dead static void usage(void);
125 static void optimize(sqlite3 *);
126 static char *parse_escape(const char *);
127 static makemandb_flags mflags = { .verbosity = 1 };
128
129 typedef void (*pman_nf)(const struct man_node *n, mandb_rec *);
130 typedef void (*pmdoc_nf)(const struct mdoc_node *n, mandb_rec *);
131 static const pmdoc_nf mdocs[MDOC_MAX] = {
132 NULL, /* Ap */
133 NULL, /* Dd */
134 NULL, /* Dt */
135 NULL, /* Os */
136 pmdoc_Sh, /* Sh */
137 NULL, /* Ss */
138 pmdoc_Pp, /* Pp */
139 NULL, /* D1 */
140 NULL, /* Dl */
141 NULL, /* Bd */
142 NULL, /* Ed */
143 NULL, /* Bl */
144 NULL, /* El */
145 NULL, /* It */
146 NULL, /* Ad */
147 NULL, /* An */
148 NULL, /* Ar */
149 NULL, /* Cd */
150 NULL, /* Cm */
151 NULL, /* Dv */
152 NULL, /* Er */
153 NULL, /* Ev */
154 NULL, /* Ex */
155 NULL, /* Fa */
156 NULL, /* Fd */
157 NULL, /* Fl */
158 NULL, /* Fn */
159 NULL, /* Ft */
160 NULL, /* Ic */
161 NULL, /* In */
162 NULL, /* Li */
163 pmdoc_Nd, /* Nd */
164 pmdoc_Nm, /* Nm */
165 NULL, /* Op */
166 NULL, /* Ot */
167 NULL, /* Pa */
168 NULL, /* Rv */
169 NULL, /* St */
170 NULL, /* Va */
171 NULL, /* Vt */
172 pmdoc_Xr, /* Xr */
173 NULL, /* %A */
174 NULL, /* %B */
175 NULL, /* %D */
176 NULL, /* %I */
177 NULL, /* %J */
178 NULL, /* %N */
179 NULL, /* %O */
180 NULL, /* %P */
181 NULL, /* %R */
182 NULL, /* %T */
183 NULL, /* %V */
184 NULL, /* Ac */
185 NULL, /* Ao */
186 NULL, /* Aq */
187 NULL, /* At */
188 NULL, /* Bc */
189 NULL, /* Bf */
190 NULL, /* Bo */
191 NULL, /* Bq */
192 NULL, /* Bsx */
193 NULL, /* Bx */
194 NULL, /* Db */
195 NULL, /* Dc */
196 NULL, /* Do */
197 NULL, /* Dq */
198 NULL, /* Ec */
199 NULL, /* Ef */
200 NULL, /* Em */
201 NULL, /* Eo */
202 NULL, /* Fx */
203 NULL, /* Ms */
204 NULL, /* No */
205 NULL, /* Ns */
206 NULL, /* Nx */
207 NULL, /* Ox */
208 NULL, /* Pc */
209 NULL, /* Pf */
210 NULL, /* Po */
211 NULL, /* Pq */
212 NULL, /* Qc */
213 NULL, /* Ql */
214 NULL, /* Qo */
215 NULL, /* Qq */
216 NULL, /* Re */
217 NULL, /* Rs */
218 NULL, /* Sc */
219 NULL, /* So */
220 NULL, /* Sq */
221 NULL, /* Sm */
222 NULL, /* Sx */
223 NULL, /* Sy */
224 NULL, /* Tn */
225 NULL, /* Ux */
226 NULL, /* Xc */
227 NULL, /* Xo */
228 NULL, /* Fo */
229 NULL, /* Fc */
230 NULL, /* Oo */
231 NULL, /* Oc */
232 NULL, /* Bk */
233 NULL, /* Ek */
234 NULL, /* Bt */
235 NULL, /* Hf */
236 NULL, /* Fr */
237 NULL, /* Ud */
238 NULL, /* Lb */
239 NULL, /* Lp */
240 NULL, /* Lk */
241 NULL, /* Mt */
242 NULL, /* Brq */
243 NULL, /* Bro */
244 NULL, /* Brc */
245 NULL, /* %C */
246 NULL, /* Es */
247 NULL, /* En */
248 NULL, /* Dx */
249 NULL, /* %Q */
250 NULL, /* br */
251 NULL, /* sp */
252 NULL, /* %U */
253 NULL, /* Ta */
254 };
255
256 static const pman_nf mans[MAN_MAX] = {
257 NULL, //br
258 NULL, //TH
259 pman_sh, //SH
260 NULL, //SS
261 NULL, //TP
262 NULL, //LP
263 NULL, //PP
264 NULL, //P
265 NULL, //IP
266 NULL, //HP
267 NULL, //SM
268 NULL, //SB
269 NULL, //BI
270 NULL, //IB
271 NULL, //BR
272 NULL, //RB
273 NULL, //R
274 pman_block, //B
275 NULL, //I
276 NULL, //IR
277 NULL, //RI
278 NULL, //na
279 NULL, //sp
280 NULL, //nf
281 NULL, //fi
282 NULL, //RE
283 NULL, //RS
284 NULL, //DT
285 NULL, //UC
286 NULL, //PD
287 NULL, //AT
288 NULL, //in
289 NULL, //ft
290 };
291
292
293 int
294 main(int argc, char *argv[])
295 {
296 FILE *file;
297 const char *sqlstr, *manconf = NULL;
298 char *line, *command, *parent;
299 char *errmsg;
300 int ch;
301 struct mparse *mp;
302 sqlite3 *db;
303 ssize_t len;
304 size_t linesize;
305 struct mandb_rec rec;
306
307 while ((ch = getopt(argc, argv, "C:floQqv")) != -1) {
308 switch (ch) {
309 case 'C':
310 manconf = optarg;
311 break;
312 case 'f':
313 mflags.recreate = 1;
314 break;
315 case 'l':
316 mflags.limit = 1;
317 break;
318 case 'o':
319 mflags.optimize = 1;
320 break;
321 case 'Q':
322 mflags.verbosity = 0;
323 break;
324 case 'q':
325 mflags.verbosity = 1;
326 break;
327 case 'v':
328 mflags.verbosity = 2;
329 break;
330 default:
331 usage();
332 }
333 }
334
335 memset(&rec, 0, sizeof(rec));
336
337 init_secbuffs(&rec);
338 mp = mparse_alloc(MPARSE_AUTO, MANDOCLEVEL_FATAL, NULL, NULL);
339
340 if (manconf) {
341 char *arg;
342 size_t command_len = shquote(manconf, NULL, 0) + 1;
343 arg = emalloc(command_len);
344 shquote(manconf, arg, command_len);
345 easprintf(&command, "man -p -C %s", arg);
346 free(arg);
347 } else {
348 command = estrdup("man -p");
349 manconf = MANCONF;
350 }
351
352 if (mflags.recreate)
353 remove(get_dbpath(manconf));
354
355 if ((db = init_db(MANDB_CREATE, manconf)) == NULL)
356 exit(EXIT_FAILURE);
357
358 sqlite3_exec(db, "PRAGMA synchronous = 0", NULL, NULL, &errmsg);
359 if (errmsg != NULL) {
360 warnx("%s", errmsg);
361 free(errmsg);
362 close_db(db);
363 exit(EXIT_FAILURE);
364 }
365
366 sqlite3_exec(db, "ATTACH DATABASE \':memory:\' AS metadb", NULL, NULL,
367 &errmsg);
368 if (errmsg != NULL) {
369 warnx("%s", errmsg);
370 free(errmsg);
371 close_db(db);
372 exit(EXIT_FAILURE);
373 }
374
375
376 /* Call man -p to get the list of man page dirs */
377 if ((file = popen(command, "r")) == NULL) {
378 close_db(db);
379 err(EXIT_FAILURE, "fopen failed");
380 }
381 free(command);
382
383 /* Begin the transaction for indexing the pages */
384 sqlite3_exec(db, "BEGIN", NULL, NULL, &errmsg);
385 if (errmsg != NULL) {
386 warnx("%s", errmsg);
387 free(errmsg);
388 exit(EXIT_FAILURE);
389 }
390
391 sqlstr = "CREATE TABLE metadb.file_cache(device, inode, mtime, parent,"
392 " file PRIMARY KEY);"
393 "CREATE UNIQUE INDEX metadb.index_file_cache_dev"
394 " ON file_cache (device, inode)";
395
396 sqlite3_exec(db, sqlstr, NULL, NULL, &errmsg);
397 if (errmsg != NULL) {
398 warnx("%s", errmsg);
399 free(errmsg);
400 close_db(db);
401 exit(EXIT_FAILURE);
402 }
403
404 if (mflags.verbosity)
405 printf("Building temporary file cache\n");
406 line = NULL;
407 linesize = 0;
408 while ((len = getline(&line, &linesize, file)) != -1) {
409 /* Replace the new line character at the end of string with '\0' */
410 line[len - 1] = '\0';
411 parent = estrdup(line);
412 char *pdir = estrdup(dirname(parent));
413 free(parent);
414 /* Traverse the man page directories and parse the pages */
415 traversedir(pdir, line, db, mp);
416 free(pdir);
417 }
418 free(line);
419
420 if (pclose(file) == -1) {
421 close_db(db);
422 cleanup(&rec);
423 free_secbuffs(&rec);
424 err(EXIT_FAILURE, "pclose error");
425 }
426
427 if (mflags.verbosity)
428 printf("Performing index update\n");
429 update_db(db, mp, &rec);
430 mparse_free(mp);
431 free_secbuffs(&rec);
432
433 /* Commit the transaction */
434 sqlite3_exec(db, "COMMIT", NULL, NULL, &errmsg);
435 if (errmsg != NULL) {
436 warnx("%s", errmsg);
437 free(errmsg);
438 exit(EXIT_FAILURE);
439 }
440
441 if (mflags.optimize)
442 optimize(db);
443
444 close_db(db);
445 return 0;
446 }
447
448 /*
449 * traversedir --
450 * Traverses the given directory recursively and passes all the man page files
451 * in the way to build_file_cache()
452 */
453 static void
454 traversedir(const char *parent, const char *file, sqlite3 *db,
455 struct mparse *mp)
456 {
457 struct stat sb;
458 struct dirent *dirp;
459 DIR *dp;
460 char *buf;
461
462 if (stat(file, &sb) < 0) {
463 if (mflags.verbosity)
464 warn("stat failed: %s", file);
465 return;
466 }
467
468 /* If it is a regular file or a symlink, pass it to build_cache() */
469 if (S_ISREG(sb.st_mode) || S_ISLNK(sb.st_mode)) {
470 build_file_cache(db, parent, file, &sb);
471 return;
472 }
473
474 /* If it is a directory, traverse it recursively */
475 if (S_ISDIR(sb.st_mode)) {
476 if ((dp = opendir(file)) == NULL) {
477 if (mflags.verbosity)
478 warn("opendir error: %s", file);
479 return;
480 }
481
482 while ((dirp = readdir(dp)) != NULL) {
483 /* Avoid . and .. entries in a directory */
484 if (strncmp(dirp->d_name, ".", 1)) {
485 easprintf(&buf, "%s/%s", file, dirp->d_name);
486 traversedir(parent, buf, db, mp);
487 free(buf);
488 }
489 }
490 closedir(dp);
491 }
492 }
493
494 /* build_file_cache --
495 * This function generates an md5 hash of the file passed as it's 2nd parameter
496 * and stores it in a temporary table file_cache along with the full file path.
497 * This is done to support incremental updation of the database.
498 * The temporary table file_cache is dropped thereafter in the function
499 * update_db(), once the database has been updated.
500 */
501 static void
502 build_file_cache(sqlite3 *db, const char *parent, const char *file,
503 struct stat *sb)
504 {
505 const char *sqlstr;
506 sqlite3_stmt *stmt = NULL;
507 int rc, idx;
508 assert(file != NULL);
509 dev_t device_cache = sb->st_dev;
510 ino_t inode_cache = sb->st_ino;
511 time_t mtime_cache = sb->st_mtime;
512
513 sqlstr = "INSERT INTO metadb.file_cache VALUES (:device, :inode,"
514 " :mtime, :parent, :file)";
515 rc = sqlite3_prepare_v2(db, sqlstr, -1, &stmt, NULL);
516 if (rc != SQLITE_OK) {
517 if (mflags.verbosity)
518 warnx("%s", sqlite3_errmsg(db));
519 return;
520 }
521
522 idx = sqlite3_bind_parameter_index(stmt, ":device");
523 rc = sqlite3_bind_int64(stmt, idx, device_cache);
524 if (rc != SQLITE_OK) {
525 if (mflags.verbosity)
526 warnx("%s", sqlite3_errmsg(db));
527 sqlite3_finalize(stmt);
528 return;
529 }
530
531 idx = sqlite3_bind_parameter_index(stmt, ":inode");
532 rc = sqlite3_bind_int64(stmt, idx, inode_cache);
533 if (rc != SQLITE_OK) {
534 if (mflags.verbosity)
535 warnx("%s", sqlite3_errmsg(db));
536 sqlite3_finalize(stmt);
537 return;
538 }
539
540 idx = sqlite3_bind_parameter_index(stmt, ":mtime");
541 rc = sqlite3_bind_int64(stmt, idx, mtime_cache);
542 if (rc != SQLITE_OK) {
543 if (mflags.verbosity)
544 warnx("%s", sqlite3_errmsg(db));
545 sqlite3_finalize(stmt);
546 return;
547 }
548
549 idx = sqlite3_bind_parameter_index(stmt, ":parent");
550 rc = sqlite3_bind_text(stmt, idx, parent, -1, NULL);
551 if (rc != SQLITE_OK) {
552 if (mflags.verbosity)
553 warnx("%s", sqlite3_errmsg(db));
554 sqlite3_finalize(stmt);
555 return;
556 }
557
558 idx = sqlite3_bind_parameter_index(stmt, ":file");
559 rc = sqlite3_bind_text(stmt, idx, file, -1, NULL);
560 if (rc != SQLITE_OK) {
561 if (mflags.verbosity)
562 warnx("%s", sqlite3_errmsg(db));
563 sqlite3_finalize(stmt);
564 return;
565 }
566
567 sqlite3_step(stmt);
568 sqlite3_finalize(stmt);
569 }
570
571 static void
572 update_existing_entry(sqlite3 *db, const char *file, const char *hash,
573 mandb_rec *rec, int *new_count, int *link_count, int *err_count)
574 {
575 int update_count, rc, idx;
576 const char *inner_sqlstr;
577 sqlite3_stmt *inner_stmt;
578
579 update_count = sqlite3_total_changes(db);
580 inner_sqlstr = "UPDATE mandb_meta SET device = :device,"
581 " inode = :inode, mtime = :mtime WHERE"
582 " md5_hash = :md5 AND file = :file AND"
583 " (device <> :device2 OR inode <> "
584 " :inode2 OR mtime <> :mtime2)";
585 rc = sqlite3_prepare_v2(db, inner_sqlstr, -1, &inner_stmt, NULL);
586 if (rc != SQLITE_OK) {
587 if (mflags.verbosity)
588 warnx("%s", sqlite3_errmsg(db));
589 return;
590 }
591 idx = sqlite3_bind_parameter_index(inner_stmt, ":device");
592 sqlite3_bind_int64(inner_stmt, idx, rec->device);
593 idx = sqlite3_bind_parameter_index(inner_stmt, ":inode");
594 sqlite3_bind_int64(inner_stmt, idx, rec->inode);
595 idx = sqlite3_bind_parameter_index(inner_stmt, ":mtime");
596 sqlite3_bind_int64(inner_stmt, idx, rec->mtime);
597 idx = sqlite3_bind_parameter_index(inner_stmt, ":md5");
598 sqlite3_bind_text(inner_stmt, idx, hash, -1, NULL);
599 idx = sqlite3_bind_parameter_index(inner_stmt, ":file");
600 sqlite3_bind_text(inner_stmt, idx, file, -1, NULL);
601 idx = sqlite3_bind_parameter_index(inner_stmt, ":device2");
602 sqlite3_bind_int64(inner_stmt, idx, rec->device);
603 idx = sqlite3_bind_parameter_index(inner_stmt, ":inode2");
604 sqlite3_bind_int64(inner_stmt, idx, rec->inode);
605 idx = sqlite3_bind_parameter_index(inner_stmt, ":mtime2");
606 sqlite3_bind_int64(inner_stmt, idx, rec->mtime);
607
608 rc = sqlite3_step(inner_stmt);
609 if (rc == SQLITE_DONE) {
610 /* Check if an update has been performed. */
611 if (update_count != sqlite3_total_changes(db)) {
612 if (mflags.verbosity == 2)
613 printf("Updated %s\n", file);
614 (*new_count)++;
615 } else {
616 /* Otherwise it was a hardlink. */
617 (*link_count)++;
618 }
619 } else {
620 if (mflags.verbosity == 2)
621 warnx("Could not update the meta data for %s", file);
622 (*err_count)++;
623 }
624 sqlite3_finalize(inner_stmt);
625 }
626
627 /* read_and_decompress --
628 * Reads the given file into memory. If it is compressed, decompres
629 * it before returning to the caller.
630 */
631 static int
632 read_and_decompress(const char *file, void **buf, size_t *len)
633 {
634 size_t off;
635 ssize_t r;
636 struct archive *a;
637 struct archive_entry *ae;
638
639 if ((a = archive_read_new()) == NULL)
640 errx(EXIT_FAILURE, "memory allocation failed");
641
642 if (archive_read_support_compression_all(a) != ARCHIVE_OK ||
643 archive_read_support_format_raw(a) != ARCHIVE_OK ||
644 archive_read_open_filename(a, file, 65536) != ARCHIVE_OK ||
645 archive_read_next_header(a, &ae) != ARCHIVE_OK)
646 goto archive_error;
647 *len = 65536;
648 *buf = emalloc(*len);
649 off = 0;
650 for (;;) {
651 r = archive_read_data(a, (char *)*buf + off, *len - off);
652 if (r == ARCHIVE_OK) {
653 archive_read_close(a);
654 *len = off;
655 return 0;
656 }
657 if (r <= 0) {
658 free(*buf);
659 break;
660 }
661 off += r;
662 if (off == *len) {
663 *len *= 2;
664 if (*len < off) {
665 if (mflags.verbosity)
666 warnx("File too large: %s", file);
667 free(*buf);
668 archive_read_close(a);
669 return -1;
670 }
671 *buf = erealloc(*buf, *len);
672 }
673 }
674
675 archive_error:
676 warnx("Error while reading `%s': %s", file, archive_error_string(a));
677 archive_read_close(a);
678 return -1;
679 }
680
681 /* update_db --
682 * Does an incremental updation of the database by checking the file_cache.
683 * It parses and adds the pages which are present in file_cache,
684 * but not in the database.
685 * It also removes the pages which are present in the databse,
686 * but not in the file_cache.
687 */
688 static void
689 update_db(sqlite3 *db, struct mparse *mp, mandb_rec *rec)
690 {
691 const char *sqlstr;
692 sqlite3_stmt *stmt = NULL;
693 const char *file;
694 const char *parent;
695 char *errmsg = NULL;
696 char *md5sum;
697 void *buf;
698 size_t buflen;
699 int new_count = 0; /* Counter for newly indexed/updated pages */
700 int total_count = 0; /* Counter for total number of pages */
701 int err_count = 0; /* Counter for number of failed pages */
702 int link_count = 0; /* Counter for number of hard/sym links */
703 int md5_status;
704 int rc;
705
706 sqlstr = "SELECT device, inode, mtime, parent, file"
707 " FROM metadb.file_cache fc"
708 " WHERE NOT EXISTS(SELECT 1 FROM mandb_meta WHERE"
709 " device = fc.device AND inode = fc.inode AND "
710 " mtime = fc.mtime AND file = fc.file)";
711
712 rc = sqlite3_prepare_v2(db, sqlstr, -1, &stmt, NULL);
713 if (rc != SQLITE_OK) {
714 if (mflags.verbosity)
715 warnx("%s", sqlite3_errmsg(db));
716 close_db(db);
717 errx(EXIT_FAILURE, "Could not query file cache");
718 }
719
720 buf = NULL;
721 while (sqlite3_step(stmt) == SQLITE_ROW) {
722 free(buf);
723 total_count++;
724 rec->device = sqlite3_column_int64(stmt, 0);
725 rec->inode = sqlite3_column_int64(stmt, 1);
726 rec->mtime = sqlite3_column_int64(stmt, 2);
727 parent = (const char *) sqlite3_column_text(stmt, 3);
728 file = (const char *) sqlite3_column_text(stmt, 4);
729 if (read_and_decompress(file, &buf, &buflen)) {
730 err_count++;
731 buf = NULL;
732 continue;
733 }
734 md5_status = check_md5(file, db, "mandb_meta", &md5sum, buf, buflen);
735 assert(md5sum != NULL);
736 if (md5_status == -1) {
737 if (mflags.verbosity)
738 warnx("An error occurred in checking md5 value"
739 " for file %s", file);
740 err_count++;
741 continue;
742 }
743
744 if (md5_status == 0) {
745 /*
746 * The MD5 hash is already present in the database,
747 * so simply update the metadata, ignoring symlinks.
748 */
749 struct stat sb;
750 stat(file, &sb);
751 if (S_ISLNK(sb.st_mode)) {
752 free(md5sum);
753 link_count++;
754 continue;
755 }
756 update_existing_entry(db, file, md5sum, rec,
757 &new_count, &link_count, &err_count);
758 free(md5sum);
759 continue;
760 }
761
762 if (md5_status == 1) {
763 /*
764 * The MD5 hash was not present in the database.
765 * This means is either a new file or an updated file.
766 * We should go ahead with parsing.
767 */
768 if (mflags.verbosity == 2)
769 printf("Parsing: %s\n", file);
770 rec->md5_hash = md5sum;
771 rec->file_path = estrdup(file);
772 // file_path is freed by insert_into_db itself.
773 chdir(parent);
774 begin_parse(file, mp, rec, buf, buflen);
775 if (insert_into_db(db, rec) < 0) {
776 if (mflags.verbosity)
777 warnx("Error in indexing %s", file);
778 err_count++;
779 } else {
780 new_count++;
781 }
782 }
783 }
784 free(buf);
785
786 sqlite3_finalize(stmt);
787
788 if (mflags.verbosity == 2) {
789 printf("Total Number of new or updated pages encountered = %d\n"
790 "Total number of (hard or symbolic) links found = %d\n"
791 "Total number of pages that were successfully"
792 " indexed/updated = %d\n"
793 "Total number of pages that could not be indexed"
794 " due to errors = %d\n",
795 total_count - link_count, link_count, new_count, err_count);
796 }
797
798 if (mflags.recreate)
799 return;
800
801 if (mflags.verbosity == 2)
802 printf("Deleting stale index entries\n");
803
804 sqlstr = "DELETE FROM mandb_meta WHERE file NOT IN"
805 " (SELECT file FROM metadb.file_cache);"
806 "DELETE FROM mandb_links WHERE md5_hash NOT IN"
807 " (SELECT md5_hash from mandb_meta);"
808 "DROP TABLE metadb.file_cache;"
809 "DELETE FROM mandb WHERE rowid NOT IN"
810 " (SELECT id FROM mandb_meta);";
811
812 sqlite3_exec(db, sqlstr, NULL, NULL, &errmsg);
813 if (errmsg != NULL) {
814 warnx("Removing old entries failed: %s", errmsg);
815 warnx("Please rebuild database from scratch with -f.");
816 free(errmsg);
817 return;
818 }
819 }
820
821 /*
822 * begin_parse --
823 * parses the man page using libmandoc
824 */
825 static void
826 begin_parse(const char *file, struct mparse *mp, mandb_rec *rec,
827 const void *buf, size_t len)
828 {
829 struct mdoc *mdoc;
830 struct man *man;
831 mparse_reset(mp);
832
833 rec->xr_found = 0;
834
835 if (mparse_readmem(mp, buf, len, file) >= MANDOCLEVEL_FATAL) {
836 /* Printing this warning at verbosity level 2
837 * because some packages from pkgsrc might trigger several
838 * of such warnings.
839 */
840 if (mflags.verbosity == 2)
841 warnx("%s: Parse failure", file);
842 return;
843 }
844
845 mparse_result(mp, &mdoc, &man);
846 if (mdoc == NULL && man == NULL) {
847 if (mflags.verbosity == 2)
848 warnx("Not a man(7) or mdoc(7) page");
849 return;
850 }
851
852 set_machine(mdoc, rec);
853 set_section(mdoc, man, rec);
854 if (mdoc) {
855 rec->page_type = MDOC;
856 pmdoc_node(mdoc_node(mdoc), rec);
857 } else {
858 rec->page_type = MAN;
859 pman_node(man_node(man), rec);
860 }
861 }
862
863 /*
864 * set_section --
865 * Extracts the section number and normalizes it to only the numeric part
866 * (Which should be the first character of the string).
867 */
868 static void
869 set_section(const struct mdoc *md, const struct man *m, mandb_rec *rec)
870 {
871 if (md) {
872 const struct mdoc_meta *md_meta = mdoc_meta(md);
873 rec->section[0] = md_meta->msec[0];
874 } else if (m) {
875 const struct man_meta *m_meta = man_meta(m);
876 rec->section[0] = m_meta->msec[0];
877 }
878 }
879
880 /*
881 * get_machine --
882 * Extracts the machine architecture information if available.
883 */
884 static void
885 set_machine(const struct mdoc *md, mandb_rec *rec)
886 {
887 if (md == NULL)
888 return;
889 const struct mdoc_meta *md_meta = mdoc_meta(md);
890 if (md_meta->arch)
891 rec->machine = estrdup(md_meta->arch);
892 }
893
894 static void
895 pmdoc_node(const struct mdoc_node *n, mandb_rec *rec)
896 {
897
898 if (n == NULL)
899 return;
900
901 switch (n->type) {
902 case (MDOC_BODY):
903 /* FALLTHROUGH */
904 case (MDOC_TAIL):
905 /* FALLTHROUGH */
906 case (MDOC_ELEM):
907 if (mdocs[n->tok] == NULL)
908 break;
909 (*mdocs[n->tok])(n, rec);
910 break;
911 default:
912 break;
913 }
914
915 pmdoc_node(n->child, rec);
916 pmdoc_node(n->next, rec);
917 }
918
919 /*
920 * pmdoc_Nm --
921 * Extracts the Name of the manual page from the .Nm macro
922 */
923 static void
924 pmdoc_Nm(const struct mdoc_node *n, mandb_rec *rec)
925 {
926 if (n->sec != SEC_NAME)
927 return;
928
929 for (n = n->child; n; n = n->next) {
930 if (n->type == MDOC_TEXT) {
931 concat(&rec->name, n->string);
932 }
933 }
934 }
935
936 /*
937 * pmdoc_Nd --
938 * Extracts the one line description of the man page from the .Nd macro
939 */
940 static void
941 pmdoc_Nd(const struct mdoc_node *n, mandb_rec *rec)
942 {
943 /*
944 * A static variable for keeping track of whether a Xr macro was seen
945 * previously.
946 */
947 char *buf = NULL;
948 char *temp;
949
950 if (n == NULL)
951 return;
952
953 if (n->type == MDOC_TEXT) {
954 if (rec->xr_found && n->next) {
955 /*
956 * An Xr macro was seen previously, so parse this
957 * and the next node.
958 */
959 temp = estrdup(n->string);
960 n = n->next;
961 easprintf(&buf, "%s(%s)", temp, n->string);
962 concat(&rec->name_desc, buf);
963 free(buf);
964 free(temp);
965 } else {
966 concat(&rec->name_desc, n->string);
967 }
968 rec->xr_found = 0;
969 } else if (mdocs[n->tok] == pmdoc_Xr) {
970 /* Remember that we have encountered an Xr macro */
971 rec->xr_found = 1;
972 }
973
974 if (n->child)
975 pmdoc_Nd(n->child, rec);
976
977 if(n->next)
978 pmdoc_Nd(n->next, rec);
979 }
980
981 /*
982 * pmdoc_macro_handler--
983 * This function is a single point of handling all the special macros that we
984 * want to handle especially. For example the .Xr macro for properly parsing
985 * the referenced page name along with the section number, or the .Pp macro
986 * for adding a new line whenever we encounter it.
987 */
988 static void
989 pmdoc_macro_handler(const struct mdoc_node *n, mandb_rec *rec, enum mdoct doct)
990 {
991 const struct mdoc_node *sn;
992 assert(n);
993
994 switch (doct) {
995 /* Parse the man page references.
996 * Basically the .Xr macros are used like:
997 * .Xr ls 1
998 * and formatted like this:
999 * ls(1)
1000 * Prepare a buffer to format the data like the above example and call
1001 * pmdoc_parse_section to append it.
1002 */
1003 case MDOC_Xr:
1004 n = n->child;
1005 while (n->type != MDOC_TEXT && n->next)
1006 n = n->next;
1007
1008 if (n && n->type != MDOC_TEXT)
1009 return;
1010 sn = n;
1011 if (n->next)
1012 n = n->next;
1013
1014 while (n->type != MDOC_TEXT && n->next)
1015 n = n->next;
1016
1017 if (n && n->type == MDOC_TEXT) {
1018 size_t len = strlen(sn->string);
1019 char *buf = emalloc(len + 4);
1020 memcpy(buf, sn->string, len);
1021 buf[len] = '(';
1022 buf[len + 1] = n->string[0];
1023 buf[len + 2] = ')';
1024 buf[len + 3] = 0;
1025 mdoc_parse_section(n->sec, buf, rec);
1026 free(buf);
1027 }
1028
1029 break;
1030
1031 /* Parse the .Pp macro to add a new line */
1032 case MDOC_Pp:
1033 if (n->type == MDOC_TEXT)
1034 mdoc_parse_section(n->sec, "\n", rec);
1035 break;
1036 default:
1037 break;
1038 }
1039
1040 }
1041
1042 /*
1043 * pmdoc_Xr, pmdoc_Pp--
1044 * Empty stubs.
1045 * The parser calls these functions each time it encounters
1046 * a .Xr or .Pp macro. We are parsing all the data from
1047 * the pmdoc_Sh function, so don't do anything here.
1048 * (See if else blocks in pmdoc_Sh.)
1049 */
1050 static void
1051 pmdoc_Xr(const struct mdoc_node *n, mandb_rec *rec)
1052 {
1053 }
1054
1055 static void
1056 pmdoc_Pp(const struct mdoc_node *n, mandb_rec *rec)
1057 {
1058 }
1059
1060 /*
1061 * pmdoc_Sh --
1062 * Called when a .Sh macro is encountered and loops through its body, calling
1063 * mdoc_parse_section to append the data to the section specific buffer.
1064 * Two special macros which may occur inside the body of Sh are .Nm and .Xr and
1065 * they need special handling, thus the separate if branches for them.
1066 */
1067 static void
1068 pmdoc_Sh(const struct mdoc_node *n, mandb_rec *rec)
1069 {
1070 if (n == NULL)
1071 return;
1072 int xr_found = 0;
1073
1074 if (n->type == MDOC_TEXT) {
1075 mdoc_parse_section(n->sec, n->string, rec);
1076 } else if (mdocs[n->tok] == pmdoc_Nm && rec->name != NULL) {
1077 /*
1078 * When encountering a .Nm macro, substitute it
1079 * with its previously cached value of the argument.
1080 */
1081 mdoc_parse_section(n->sec, rec->name, rec);
1082 } else if (mdocs[n->tok] == pmdoc_Xr) {
1083 /*
1084 * When encountering other inline macros,
1085 * call pmdoc_macro_handler.
1086 */
1087 pmdoc_macro_handler(n, rec, MDOC_Xr);
1088 xr_found = 1;
1089 } else if (mdocs[n->tok] == pmdoc_Pp) {
1090 pmdoc_macro_handler(n, rec, MDOC_Pp);
1091 }
1092
1093 /*
1094 * If an Xr macro was encountered then the child node has
1095 * already been explored by pmdoc_macro_handler.
1096 */
1097 if (xr_found == 0)
1098 pmdoc_Sh(n->child, rec);
1099 pmdoc_Sh(n->next, rec);
1100 }
1101
1102 /*
1103 * mdoc_parse_section--
1104 * Utility function for parsing sections of the mdoc type pages.
1105 * Takes two params:
1106 * 1. sec is an enum which indicates the section in which we are present
1107 * 2. string is the string which we need to append to the secbuff for this
1108 * particular section.
1109 * The function appends string to the global section buffer and returns.
1110 */
1111 static void
1112 mdoc_parse_section(enum mdoc_sec sec, const char *string, mandb_rec *rec)
1113 {
1114 /*
1115 * If the user specified the 'l' flag, then parse and store only the
1116 * NAME section. Ignore the rest.
1117 */
1118 if (mflags.limit)
1119 return;
1120
1121 switch (sec) {
1122 case SEC_LIBRARY:
1123 append(&rec->lib, string);
1124 break;
1125 case SEC_RETURN_VALUES:
1126 append(&rec->return_vals, string);
1127 break;
1128 case SEC_ENVIRONMENT:
1129 append(&rec->env, string);
1130 break;
1131 case SEC_FILES:
1132 append(&rec->files, string);
1133 break;
1134 case SEC_EXIT_STATUS:
1135 append(&rec->exit_status, string);
1136 break;
1137 case SEC_DIAGNOSTICS:
1138 append(&rec->diagnostics, string);
1139 break;
1140 case SEC_ERRORS:
1141 append(&rec->errors, string);
1142 break;
1143 case SEC_NAME:
1144 case SEC_SYNOPSIS:
1145 case SEC_EXAMPLES:
1146 case SEC_STANDARDS:
1147 case SEC_HISTORY:
1148 case SEC_AUTHORS:
1149 case SEC_BUGS:
1150 break;
1151 default:
1152 append(&rec->desc, string);
1153 break;
1154 }
1155 }
1156
1157 static void
1158 pman_node(const struct man_node *n, mandb_rec *rec)
1159 {
1160 if (n == NULL)
1161 return;
1162
1163 switch (n->type) {
1164 case (MAN_BODY):
1165 /* FALLTHROUGH */
1166 case (MAN_TAIL):
1167 /* FALLTHROUGH */
1168 case (MAN_BLOCK):
1169 /* FALLTHROUGH */
1170 case (MAN_ELEM):
1171 if (mans[n->tok] != NULL)
1172 (*mans[n->tok])(n, rec);
1173 break;
1174 default:
1175 break;
1176 }
1177
1178 pman_node(n->child, rec);
1179 pman_node(n->next, rec);
1180 }
1181
1182 /*
1183 * pman_parse_name --
1184 * Parses the NAME section and puts the complete content in the name_desc
1185 * variable.
1186 */
1187 static void
1188 pman_parse_name(const struct man_node *n, mandb_rec *rec)
1189 {
1190 if (n == NULL)
1191 return;
1192
1193 if (n->type == MAN_TEXT) {
1194 char *tmp = parse_escape(n->string);
1195 concat(&rec->name_desc, tmp);
1196 free(tmp);
1197 }
1198
1199 if (n->child)
1200 pman_parse_name(n->child, rec);
1201
1202 if(n->next)
1203 pman_parse_name(n->next, rec);
1204 }
1205
1206 /*
1207 * A stub function to be able to parse the macros like .B embedded inside
1208 * a section.
1209 */
1210 static void
1211 pman_block(const struct man_node *n, mandb_rec *rec)
1212 {
1213 }
1214
1215 /*
1216 * pman_sh --
1217 * This function does one of the two things:
1218 * 1. If the present section is NAME, then it will:
1219 * (a) Extract the name of the page (in case of multiple comma separated
1220 * names, it will pick up the first one).
1221 * (b) Build a space spearated list of all the symlinks/hardlinks to
1222 * this page and store in the buffer 'links'. These are extracted from
1223 * the comma separated list of names in the NAME section as well.
1224 * (c) Move on to the one line description section, which is after the list
1225 * of names in the NAME section.
1226 * 2. Otherwise, it will check the section name and call the man_parse_section
1227 * function, passing the enum corresponding that section.
1228 */
1229 static void
1230 pman_sh(const struct man_node *n, mandb_rec *rec)
1231 {
1232 static const struct {
1233 enum man_sec section;
1234 const char *header;
1235 } mapping[] = {
1236 { MANSEC_DESCRIPTION, "DESCRIPTION" },
1237 { MANSEC_SYNOPSIS, "SYNOPSIS" },
1238 { MANSEC_LIBRARY, "LIBRARY" },
1239 { MANSEC_ERRORS, "ERRORS" },
1240 { MANSEC_FILES, "FILES" },
1241 { MANSEC_RETURN_VALUES, "RETURN VALUE" },
1242 { MANSEC_RETURN_VALUES, "RETURN VALUES" },
1243 { MANSEC_EXIT_STATUS, "EXIT STATUS" },
1244 { MANSEC_EXAMPLES, "EXAMPLES" },
1245 { MANSEC_EXAMPLES, "EXAMPLE" },
1246 { MANSEC_STANDARDS, "STANDARDS" },
1247 { MANSEC_HISTORY, "HISTORY" },
1248 { MANSEC_BUGS, "BUGS" },
1249 { MANSEC_AUTHORS, "AUTHORS" },
1250 { MANSEC_COPYRIGHT, "COPYRIGHT" },
1251 };
1252 const struct man_node *head;
1253 char *name_desc;
1254 int sz;
1255 size_t i;
1256
1257 if ((head = n->parent->head) == NULL || (head = head->child) == NULL ||
1258 head->type != MAN_TEXT)
1259 return;
1260
1261 /*
1262 * Check if this section should be extracted and
1263 * where it should be stored. Handled the trival cases first.
1264 */
1265 for (i = 0; i < sizeof(mapping) / sizeof(mapping[0]); ++i) {
1266 if (strcmp(head->string, mapping[i].header) == 0) {
1267 man_parse_section(mapping[i].section, n, rec);
1268 return;
1269 }
1270 }
1271
1272 if (strcmp(head->string, "NAME") == 0) {
1273 /*
1274 * We are in the NAME section.
1275 * pman_parse_name will put the complete content in name_desc.
1276 */
1277 pman_parse_name(n, rec);
1278
1279 name_desc = rec->name_desc;
1280 if (name_desc == NULL)
1281 return;
1282
1283 /* Remove any leading spaces. */
1284 while (name_desc[0] == ' ')
1285 name_desc++;
1286
1287 /* If the line begins with a "\&", avoid those */
1288 if (name_desc[0] == '\\' && name_desc[1] == '&')
1289 name_desc += 2;
1290
1291 /* Now name_desc should be left with a comma-space
1292 * separated list of names and the one line description
1293 * of the page:
1294 * "a, b, c \- sample description"
1295 * Take out the first name, before the first comma
1296 * (or space) and store it in rec->name.
1297 * If the page has aliases then they should be
1298 * in the form of a comma separated list.
1299 * Keep looping while there is a comma in name_desc,
1300 * extract the alias name and store in rec->links.
1301 * When there are no more commas left, break out.
1302 */
1303 int has_alias = 0; // Any more aliases left?
1304 while (*name_desc) {
1305 /* Remove any leading spaces or hyphens. */
1306 if (name_desc[0] == ' ' || name_desc[0] =='-') {
1307 name_desc++;
1308 continue;
1309 }
1310 sz = strcspn(name_desc, ", ");
1311
1312 /* Extract the first term and store it in rec->name. */
1313 if (rec->name == NULL) {
1314 if (name_desc[sz] == ',')
1315 has_alias = 1;
1316 name_desc[sz] = 0;
1317 rec->name = emalloc(sz + 1);
1318 memcpy(rec->name, name_desc, sz + 1);
1319 name_desc += sz + 1;
1320 continue;
1321 }
1322
1323 /*
1324 * Once rec->name is set, rest of the names
1325 * are to be treated as links or aliases.
1326 */
1327 if (rec->name && has_alias) {
1328 if (name_desc[sz] != ',') {
1329 /* No more commas left -->
1330 * no more aliases to take out
1331 */
1332 has_alias = 0;
1333 }
1334 name_desc[sz] = 0;
1335 concat2(&rec->links, name_desc, sz);
1336 name_desc += sz + 1;
1337 continue;
1338 }
1339 break;
1340 }
1341
1342 /* Parse any escape sequences that might be there */
1343 char *temp = parse_escape(name_desc);
1344 free(rec->name_desc);
1345 rec->name_desc = temp;
1346 temp = parse_escape(rec->name);
1347 free(rec->name);
1348 rec->name = temp;
1349 return;
1350 }
1351
1352 /* The RETURN VALUE section might be specified in multiple ways */
1353 if (strcmp(head->string, "RETURN") == 0 &&
1354 head->next != NULL && head->next->type == MAN_TEXT &&
1355 (strcmp(head->next->string, "VALUE") == 0 ||
1356 strcmp(head->next->string, "VALUES") == 0)) {
1357 man_parse_section(MANSEC_RETURN_VALUES, n, rec);
1358 return;
1359 }
1360
1361 /*
1362 * EXIT STATUS section can also be specified all on one line or on two
1363 * separate lines.
1364 */
1365 if (strcmp(head->string, "EXIT") == 0 &&
1366 head->next != NULL && head->next->type == MAN_TEXT &&
1367 strcmp(head->next->string, "STATUS") == 0) {
1368 man_parse_section(MANSEC_EXIT_STATUS, n, rec);
1369 return;
1370 }
1371
1372 /* Store the rest of the content in desc. */
1373 man_parse_section(MANSEC_NONE, n, rec);
1374 }
1375
1376 /*
1377 * pman_parse_node --
1378 * Generic function to iterate through a node. Usually called from
1379 * man_parse_section to parse a particular section of the man page.
1380 */
1381 static void
1382 pman_parse_node(const struct man_node *n, secbuff *s)
1383 {
1384 if (n == NULL)
1385 return;
1386
1387 if (n->type == MAN_TEXT)
1388 append(s, n->string);
1389
1390 pman_parse_node(n->child, s);
1391 pman_parse_node(n->next, s);
1392 }
1393
1394 /*
1395 * man_parse_section --
1396 * Takes two parameters:
1397 * sec: Tells which section we are present in
1398 * n: Is the present node of the AST.
1399 * Depending on the section, we call pman_parse_node to parse that section and
1400 * concatenate the content from that section into the buffer for that section.
1401 */
1402 static void
1403 man_parse_section(enum man_sec sec, const struct man_node *n, mandb_rec *rec)
1404 {
1405 /*
1406 * If the user sepecified the 'l' flag then just parse
1407 * the NAME section, ignore the rest.
1408 */
1409 if (mflags.limit)
1410 return;
1411
1412 switch (sec) {
1413 case MANSEC_LIBRARY:
1414 pman_parse_node(n, &rec->lib);
1415 break;
1416 case MANSEC_RETURN_VALUES:
1417 pman_parse_node(n, &rec->return_vals);
1418 break;
1419 case MANSEC_ENVIRONMENT:
1420 pman_parse_node(n, &rec->env);
1421 break;
1422 case MANSEC_FILES:
1423 pman_parse_node(n, &rec->files);
1424 break;
1425 case MANSEC_EXIT_STATUS:
1426 pman_parse_node(n, &rec->exit_status);
1427 break;
1428 case MANSEC_DIAGNOSTICS:
1429 pman_parse_node(n, &rec->diagnostics);
1430 break;
1431 case MANSEC_ERRORS:
1432 pman_parse_node(n, &rec->errors);
1433 break;
1434 case MANSEC_NAME:
1435 case MANSEC_SYNOPSIS:
1436 case MANSEC_EXAMPLES:
1437 case MANSEC_STANDARDS:
1438 case MANSEC_HISTORY:
1439 case MANSEC_BUGS:
1440 case MANSEC_AUTHORS:
1441 case MANSEC_COPYRIGHT:
1442 break;
1443 default:
1444 pman_parse_node(n, &rec->desc);
1445 break;
1446 }
1447
1448 }
1449
1450 /*
1451 * insert_into_db --
1452 * Inserts the parsed data of the man page in the Sqlite databse.
1453 * If any of the values is NULL, then we cleanup and return -1 indicating
1454 * an error.
1455 * Otherwise, store the data in the database and return 0.
1456 */
1457 static int
1458 insert_into_db(sqlite3 *db, mandb_rec *rec)
1459 {
1460 int rc = 0;
1461 int idx = -1;
1462 const char *sqlstr = NULL;
1463 sqlite3_stmt *stmt = NULL;
1464 char *ln = NULL;
1465 char *errmsg = NULL;
1466 long int mandb_rowid;
1467
1468 /*
1469 * At the very minimum we want to make sure that we store
1470 * the following data:
1471 * Name, one line description, and the MD5 hash
1472 */
1473 if (rec->name == NULL || rec->name_desc == NULL ||
1474 rec->md5_hash == NULL) {
1475 cleanup(rec);
1476 return -1;
1477 }
1478
1479 /* Write null byte at the end of all the sec_buffs */
1480 rec->desc.data[rec->desc.offset] = 0;
1481 rec->lib.data[rec->lib.offset] = 0;
1482 rec->env.data[rec->env.offset] = 0;
1483 rec->return_vals.data[rec->return_vals.offset] = 0;
1484 rec->exit_status.data[rec->exit_status.offset] = 0;
1485 rec->files.data[rec->files.offset] = 0;
1486 rec->diagnostics.data[rec->diagnostics.offset] = 0;
1487 rec->errors.data[rec->errors.offset] = 0;
1488
1489 /*
1490 * In case of a mdoc page: (sorry, no better place to put this code)
1491 * parse the comma separated list of names of man pages,
1492 * the first name will be stored in the mandb table, rest will be
1493 * treated as links and put in the mandb_links table.
1494 */
1495 if (rec->page_type == MDOC) {
1496 char *tmp;
1497 rec->links = estrdup(rec->name);
1498 free(rec->name);
1499 int sz = strcspn(rec->links, " \0");
1500 rec->name = emalloc(sz + 1);
1501 memcpy(rec->name, rec->links, sz);
1502 if(rec->name[sz - 1] == ',')
1503 rec->name[sz - 1] = 0;
1504 else
1505 rec->name[sz] = 0;
1506 while (rec->links[sz] == ' ')
1507 ++sz;
1508 tmp = estrdup(rec->links + sz);
1509 free(rec->links);
1510 rec->links = tmp;
1511 }
1512
1513 /*------------------------ Populate the mandb table---------------------------*/
1514 sqlstr = "INSERT INTO mandb VALUES (:section, :name, :name_desc, :desc,"
1515 " :lib, :return_vals, :env, :files, :exit_status,"
1516 " :diagnostics, :errors, :md5_hash, :machine)";
1517
1518 rc = sqlite3_prepare_v2(db, sqlstr, -1, &stmt, NULL);
1519 if (rc != SQLITE_OK)
1520 goto Out;
1521
1522 idx = sqlite3_bind_parameter_index(stmt, ":name");
1523 rc = sqlite3_bind_text(stmt, idx, rec->name, -1, NULL);
1524 if (rc != SQLITE_OK) {
1525 sqlite3_finalize(stmt);
1526 goto Out;
1527 }
1528
1529 idx = sqlite3_bind_parameter_index(stmt, ":section");
1530 rc = sqlite3_bind_text(stmt, idx, rec->section, -1, NULL);
1531 if (rc != SQLITE_OK) {
1532 sqlite3_finalize(stmt);
1533 goto Out;
1534 }
1535
1536 idx = sqlite3_bind_parameter_index(stmt, ":name_desc");
1537 rc = sqlite3_bind_text(stmt, idx, rec->name_desc, -1, NULL);
1538 if (rc != SQLITE_OK) {
1539 sqlite3_finalize(stmt);
1540 goto Out;
1541 }
1542
1543 idx = sqlite3_bind_parameter_index(stmt, ":desc");
1544 rc = sqlite3_bind_text(stmt, idx, rec->desc.data,
1545 rec->desc.offset + 1, NULL);
1546 if (rc != SQLITE_OK) {
1547 sqlite3_finalize(stmt);
1548 goto Out;
1549 }
1550
1551 idx = sqlite3_bind_parameter_index(stmt, ":lib");
1552 rc = sqlite3_bind_text(stmt, idx, rec->lib.data, rec->lib.offset + 1, NULL);
1553 if (rc != SQLITE_OK) {
1554 sqlite3_finalize(stmt);
1555 goto Out;
1556 }
1557
1558 idx = sqlite3_bind_parameter_index(stmt, ":return_vals");
1559 rc = sqlite3_bind_text(stmt, idx, rec->return_vals.data,
1560 rec->return_vals.offset + 1, NULL);
1561 if (rc != SQLITE_OK) {
1562 sqlite3_finalize(stmt);
1563 goto Out;
1564 }
1565
1566 idx = sqlite3_bind_parameter_index(stmt, ":env");
1567 rc = sqlite3_bind_text(stmt, idx, rec->env.data, rec->env.offset + 1, NULL);
1568 if (rc != SQLITE_OK) {
1569 sqlite3_finalize(stmt);
1570 goto Out;
1571 }
1572
1573 idx = sqlite3_bind_parameter_index(stmt, ":files");
1574 rc = sqlite3_bind_text(stmt, idx, rec->files.data,
1575 rec->files.offset + 1, NULL);
1576 if (rc != SQLITE_OK) {
1577 sqlite3_finalize(stmt);
1578 goto Out;
1579 }
1580
1581 idx = sqlite3_bind_parameter_index(stmt, ":exit_status");
1582 rc = sqlite3_bind_text(stmt, idx, rec->exit_status.data,
1583 rec->exit_status.offset + 1, NULL);
1584 if (rc != SQLITE_OK) {
1585 sqlite3_finalize(stmt);
1586 goto Out;
1587 }
1588
1589 idx = sqlite3_bind_parameter_index(stmt, ":diagnostics");
1590 rc = sqlite3_bind_text(stmt, idx, rec->diagnostics.data,
1591 rec->diagnostics.offset + 1, NULL);
1592 if (rc != SQLITE_OK) {
1593 sqlite3_finalize(stmt);
1594 goto Out;
1595 }
1596
1597 idx = sqlite3_bind_parameter_index(stmt, ":errors");
1598 rc = sqlite3_bind_text(stmt, idx, rec->errors.data,
1599 rec->errors.offset + 1, NULL);
1600 if (rc != SQLITE_OK) {
1601 sqlite3_finalize(stmt);
1602 goto Out;
1603 }
1604
1605 idx = sqlite3_bind_parameter_index(stmt, ":md5_hash");
1606 rc = sqlite3_bind_text(stmt, idx, rec->md5_hash, -1, NULL);
1607 if (rc != SQLITE_OK) {
1608 sqlite3_finalize(stmt);
1609 goto Out;
1610 }
1611
1612 idx = sqlite3_bind_parameter_index(stmt, ":machine");
1613 if (rec->machine)
1614 rc = sqlite3_bind_text(stmt, idx, rec->machine, -1, NULL);
1615 else
1616 rc = sqlite3_bind_null(stmt, idx);
1617 if (rc != SQLITE_OK) {
1618 sqlite3_finalize(stmt);
1619 goto Out;
1620 }
1621
1622 rc = sqlite3_step(stmt);
1623 if (rc != SQLITE_DONE) {
1624 sqlite3_finalize(stmt);
1625 goto Out;
1626 }
1627
1628 sqlite3_finalize(stmt);
1629
1630 /* Get the row id of the last inserted row */
1631 mandb_rowid = sqlite3_last_insert_rowid(db);
1632
1633 /*------------------------Populate the mandb_meta table-----------------------*/
1634 sqlstr = "INSERT INTO mandb_meta VALUES (:device, :inode, :mtime,"
1635 " :file, :md5_hash, :id)";
1636 rc = sqlite3_prepare_v2(db, sqlstr, -1, &stmt, NULL);
1637 if (rc != SQLITE_OK)
1638 goto Out;
1639
1640 idx = sqlite3_bind_parameter_index(stmt, ":device");
1641 rc = sqlite3_bind_int64(stmt, idx, rec->device);
1642 if (rc != SQLITE_OK) {
1643 sqlite3_finalize(stmt);
1644 goto Out;
1645 }
1646
1647 idx = sqlite3_bind_parameter_index(stmt, ":inode");
1648 rc = sqlite3_bind_int64(stmt, idx, rec->inode);
1649 if (rc != SQLITE_OK) {
1650 sqlite3_finalize(stmt);
1651 goto Out;
1652 }
1653
1654 idx = sqlite3_bind_parameter_index(stmt, ":mtime");
1655 rc = sqlite3_bind_int64(stmt, idx, rec->mtime);
1656 if (rc != SQLITE_OK) {
1657 sqlite3_finalize(stmt);
1658 goto Out;
1659 }
1660
1661 idx = sqlite3_bind_parameter_index(stmt, ":file");
1662 rc = sqlite3_bind_text(stmt, idx, rec->file_path, -1, NULL);
1663 if (rc != SQLITE_OK) {
1664 sqlite3_finalize(stmt);
1665 goto Out;
1666 }
1667
1668 idx = sqlite3_bind_parameter_index(stmt, ":md5_hash");
1669 rc = sqlite3_bind_text(stmt, idx, rec->md5_hash, -1, NULL);
1670 if (rc != SQLITE_OK) {
1671 sqlite3_finalize(stmt);
1672 goto Out;
1673 }
1674
1675 idx = sqlite3_bind_parameter_index(stmt, ":id");
1676 rc = sqlite3_bind_int64(stmt, idx, mandb_rowid);
1677 if (rc != SQLITE_OK) {
1678 sqlite3_finalize(stmt);
1679 goto Out;
1680 }
1681
1682 rc = sqlite3_step(stmt);
1683 sqlite3_finalize(stmt);
1684 if (rc == SQLITE_CONSTRAINT) {
1685 /* The *most* probable reason for reaching here is that
1686 * the UNIQUE contraint on the file column of the mandb_meta
1687 * table was violated.
1688 * This can happen when a file was updated/modified.
1689 * To fix this we need to do two things:
1690 * 1. Delete the row for the older version of this file
1691 * from mandb table.
1692 * 2. Run an UPDATE query to update the row for this file
1693 * in the mandb_meta table.
1694 */
1695 warnx("Trying to update index for %s", rec->file_path);
1696 char *sql = sqlite3_mprintf("DELETE FROM mandb "
1697 "WHERE rowid = (SELECT id"
1698 " FROM mandb_meta"
1699 " WHERE file = %Q)",
1700 rec->file_path);
1701 sqlite3_exec(db, sql, NULL, NULL, &errmsg);
1702 sqlite3_free(sql);
1703 if (errmsg != NULL) {
1704 if (mflags.verbosity)
1705 warnx("%s", errmsg);
1706 free(errmsg);
1707 }
1708 sqlstr = "UPDATE mandb_meta SET device = :device,"
1709 " inode = :inode, mtime = :mtime, id = :id,"
1710 " md5_hash = :md5 WHERE file = :file";
1711 rc = sqlite3_prepare_v2(db, sqlstr, -1, &stmt, NULL);
1712 if (rc != SQLITE_OK) {
1713 if (mflags.verbosity)
1714 warnx("Update failed with error: %s",
1715 sqlite3_errmsg(db));
1716 close_db(db);
1717 cleanup(rec);
1718 errx(EXIT_FAILURE,
1719 "Consider running makemandb with -f option");
1720 }
1721
1722 idx = sqlite3_bind_parameter_index(stmt, ":device");
1723 sqlite3_bind_int64(stmt, idx, rec->device);
1724 idx = sqlite3_bind_parameter_index(stmt, ":inode");
1725 sqlite3_bind_int64(stmt, idx, rec->inode);
1726 idx = sqlite3_bind_parameter_index(stmt, ":mtime");
1727 sqlite3_bind_int64(stmt, idx, rec->mtime);
1728 idx = sqlite3_bind_parameter_index(stmt, ":id");
1729 sqlite3_bind_int64(stmt, idx, mandb_rowid);
1730 idx = sqlite3_bind_parameter_index(stmt, ":md5");
1731 sqlite3_bind_text(stmt, idx, rec->md5_hash, -1, NULL);
1732 idx = sqlite3_bind_parameter_index(stmt, ":file");
1733 sqlite3_bind_text(stmt, idx, rec->file_path, -1, NULL);
1734 rc = sqlite3_step(stmt);
1735 sqlite3_finalize(stmt);
1736
1737 if (rc != SQLITE_DONE) {
1738 if (mflags.verbosity)
1739 warnx("%s", sqlite3_errmsg(db));
1740 close_db(db);
1741 cleanup(rec);
1742 errx(EXIT_FAILURE,
1743 "Consider running makemandb with -f option");
1744 }
1745 } else if (rc != SQLITE_DONE) {
1746 /* Otherwise make this error fatal */
1747 warnx("Failed at %s\n%s", rec->file_path, sqlite3_errmsg(db));
1748 cleanup(rec);
1749 close_db(db);
1750 exit(EXIT_FAILURE);
1751 }
1752
1753 /*------------------------ Populate the mandb_links table---------------------*/
1754 char *str = NULL;
1755 char *links;
1756 if (rec->links && strlen(rec->links)) {
1757 links = rec->links;
1758 for(ln = strtok(links, " "); ln; ln = strtok(NULL, " ")) {
1759 if (ln[0] == ',')
1760 ln++;
1761 if(ln[strlen(ln) - 1] == ',')
1762 ln[strlen(ln) - 1] = 0;
1763
1764 str = sqlite3_mprintf("INSERT INTO mandb_links"
1765 " VALUES (%Q, %Q, %Q, %Q, %Q)",
1766 ln, rec->name, rec->section,
1767 rec->machine, rec->md5_hash);
1768 sqlite3_exec(db, str, NULL, NULL, &errmsg);
1769 sqlite3_free(str);
1770 if (errmsg != NULL) {
1771 warnx("%s", errmsg);
1772 cleanup(rec);
1773 free(errmsg);
1774 return -1;
1775 }
1776 }
1777 }
1778
1779 cleanup(rec);
1780 return 0;
1781
1782 Out:
1783 if (mflags.verbosity)
1784 warnx("%s", sqlite3_errmsg(db));
1785 cleanup(rec);
1786 return -1;
1787 }
1788
1789 /*
1790 * check_md5--
1791 * Generates the md5 hash of the file and checks if it already doesn't exist
1792 * in the table (passed as the 3rd parameter).
1793 * This function is being used to avoid hardlinks.
1794 * On successful completion it will also set the value of the fourth parameter
1795 * to the md5 hash of the file (computed previously). It is the responsibility
1796 * of the caller to free this buffer.
1797 * Return values:
1798 * -1: If an error occurs somewhere and sets the md5 return buffer to NULL.
1799 * 0: If the md5 hash does not exist in the table.
1800 * 1: If the hash exists in the database.
1801 */
1802 static int
1803 check_md5(const char *file, sqlite3 *db, const char *table, char **md5sum,
1804 void *buf, size_t buflen)
1805 {
1806 int rc = 0;
1807 int idx = -1;
1808 char *sqlstr = NULL;
1809 sqlite3_stmt *stmt = NULL;
1810
1811 assert(file != NULL);
1812 *md5sum = MD5Data(buf, buflen, NULL);
1813 if (*md5sum == NULL) {
1814 if (mflags.verbosity)
1815 warn("md5 failed: %s", file);
1816 return -1;
1817 }
1818
1819 easprintf(&sqlstr, "SELECT * FROM %s WHERE md5_hash = :md5_hash",
1820 table);
1821 rc = sqlite3_prepare_v2(db, sqlstr, -1, &stmt, NULL);
1822 if (rc != SQLITE_OK) {
1823 free(sqlstr);
1824 free(*md5sum);
1825 *md5sum = NULL;
1826 return -1;
1827 }
1828
1829 idx = sqlite3_bind_parameter_index(stmt, ":md5_hash");
1830 rc = sqlite3_bind_text(stmt, idx, *md5sum, -1, NULL);
1831 if (rc != SQLITE_OK) {
1832 if (mflags.verbosity)
1833 warnx("%s", sqlite3_errmsg(db));
1834 sqlite3_finalize(stmt);
1835 free(sqlstr);
1836 free(*md5sum);
1837 *md5sum = NULL;
1838 return -1;
1839 }
1840
1841 if (sqlite3_step(stmt) == SQLITE_ROW) {
1842 sqlite3_finalize(stmt);
1843 free(sqlstr);
1844 return 0;
1845 }
1846
1847 sqlite3_finalize(stmt);
1848 free(sqlstr);
1849 return 1;
1850 }
1851
1852 /* Optimize the index for faster search */
1853 static void
1854 optimize(sqlite3 *db)
1855 {
1856 const char *sqlstr;
1857 char *errmsg = NULL;
1858
1859 if (mflags.verbosity == 2)
1860 printf("Optimizing the database index\n");
1861 sqlstr = "INSERT INTO mandb(mandb) VALUES (\'optimize\');"
1862 "VACUUM";
1863 sqlite3_exec(db, sqlstr, NULL, NULL, &errmsg);
1864 if (errmsg != NULL) {
1865 if (mflags.verbosity)
1866 warnx("%s", errmsg);
1867 free(errmsg);
1868 return;
1869 }
1870 }
1871
1872 /*
1873 * cleanup --
1874 * cleans up the global buffers
1875 */
1876 static void
1877 cleanup(mandb_rec *rec)
1878 {
1879 rec->desc.offset = 0;
1880 rec->lib.offset = 0;
1881 rec->return_vals.offset = 0;
1882 rec->env.offset = 0;
1883 rec->exit_status.offset = 0;
1884 rec->diagnostics.offset = 0;
1885 rec->errors.offset = 0;
1886 rec->files.offset = 0;
1887
1888 free(rec->machine);
1889 rec->machine = NULL;
1890
1891 free(rec->links);
1892 rec->links = NULL;
1893
1894 free(rec->file_path);
1895 rec->file_path = NULL;
1896
1897 free(rec->name);
1898 rec->name = NULL;
1899
1900 free(rec->name_desc);
1901 rec->name_desc = NULL;
1902
1903 free(rec->md5_hash);
1904 rec->md5_hash = NULL;
1905 }
1906
1907 /*
1908 * init_secbuffs--
1909 * Sets the value of buflen for all the sec_buff field of rec. And then
1910 * allocate memory to each sec_buff member of rec.
1911 */
1912 static void
1913 init_secbuffs(mandb_rec *rec)
1914 {
1915 /*
1916 * Some sec_buff might need more memory, for example desc,
1917 * which stores the data of the DESCRIPTION section,
1918 * while some might need very small amount of memory.
1919 * Therefore explicitly setting the value of buflen field for
1920 * each sec_buff.
1921 */
1922 rec->desc.buflen = 10 * BUFLEN;
1923 rec->desc.data = emalloc(rec->desc.buflen);
1924 rec->desc.offset = 0;
1925
1926 rec->lib.buflen = BUFLEN / 2;
1927 rec->lib.data = emalloc(rec->lib.buflen);
1928 rec->lib.offset = 0;
1929
1930 rec->return_vals.buflen = BUFLEN;
1931 rec->return_vals.data = emalloc(rec->return_vals.buflen);
1932 rec->return_vals.offset = 0;
1933
1934 rec->exit_status.buflen = BUFLEN;
1935 rec->exit_status.data = emalloc(rec->exit_status.buflen);
1936 rec->exit_status.offset = 0;
1937
1938 rec->env.buflen = BUFLEN;
1939 rec->env.data = emalloc(rec->env.buflen);
1940 rec->env.offset = 0;
1941
1942 rec->files.buflen = BUFLEN;
1943 rec->files.data = emalloc(rec->files.buflen);
1944 rec->files.offset = 0;
1945
1946 rec->diagnostics.buflen = BUFLEN;
1947 rec->diagnostics.data = emalloc(rec->diagnostics.buflen);
1948 rec->diagnostics.offset = 0;
1949
1950 rec->errors.buflen = BUFLEN;
1951 rec->errors.data = emalloc(rec->errors.buflen);
1952 rec->errors.offset = 0;
1953 }
1954
1955 /*
1956 * free_secbuffs--
1957 * This function should be called at the end, when all the pages have been
1958 * parsed.
1959 * It frees the memory allocated to sec_buffs by init_secbuffs in the starting.
1960 */
1961 static void
1962 free_secbuffs(mandb_rec *rec)
1963 {
1964 free(rec->desc.data);
1965 free(rec->lib.data);
1966 free(rec->return_vals.data);
1967 free(rec->exit_status.data);
1968 free(rec->env.data);
1969 free(rec->files.data);
1970 free(rec->diagnostics.data);
1971 free(rec->errors.data);
1972 }
1973
1974 static void
1975 replace_hyph(char *str)
1976 {
1977 char *iter = str;
1978 while ((iter = strchr(iter, ASCII_HYPH)) != NULL)
1979 *iter = '-';
1980 }
1981
1982 static char *
1983 parse_escape(const char *str)
1984 {
1985 const char *backslash, *last_backslash;
1986 char *result, *iter;
1987 size_t len;
1988
1989 assert(str);
1990
1991 last_backslash = str;
1992 backslash = strchr(str, '\\');
1993 if (backslash == NULL) {
1994 result = estrdup(str);
1995 replace_hyph(result);
1996 return result;
1997 }
1998
1999 result = emalloc(strlen(str) + 1);
2000 iter = result;
2001
2002 do {
2003 len = backslash - last_backslash;
2004 memcpy(iter, last_backslash, len);
2005 iter += len;
2006 if (backslash[1] == '-' || backslash[1] == ' ') {
2007 *iter++ = backslash[1];
2008 last_backslash = backslash + 2;
2009 backslash = strchr(backslash + 2, '\\');
2010 } else {
2011 ++backslash;
2012 mandoc_escape(&backslash, NULL, NULL);
2013 last_backslash = backslash;
2014 if (backslash == NULL)
2015 break;
2016 backslash = strchr(last_backslash, '\\');
2017 }
2018 } while (backslash != NULL);
2019 if (last_backslash != NULL)
2020 strcpy(iter, last_backslash);
2021
2022 replace_hyph(result);
2023 return result;
2024 }
2025
2026 /*
2027 * append--
2028 * Concatenates a space and src at the end of sbuff->data (much like concat in
2029 * apropos-utils.c).
2030 * Rather than reallocating space for writing data, it uses the value of the
2031 * offset field of sec_buff to write new data at the free space left in the
2032 * buffer.
2033 * In case the size of the data to be appended exceeds the number of bytes left
2034 * in the buffer, it reallocates buflen number of bytes and then continues.
2035 * Value of offset field should be adjusted as new data is written.
2036 *
2037 * NOTE: This function does not write the null byte at the end of the buffers,
2038 * write a null byte at the position pointed to by offset before inserting data
2039 * in the db.
2040 */
2041 static void
2042 append(secbuff *sbuff, const char *src)
2043 {
2044 short flag = 0;
2045 size_t srclen, newlen;
2046 char *temp;
2047
2048 assert(src != NULL);
2049 temp = parse_escape(src);
2050 srclen = strlen(temp);
2051
2052 if (sbuff->data == NULL) {
2053 sbuff->data = emalloc(sbuff->buflen);
2054 sbuff->offset = 0;
2055 }
2056
2057 newlen = sbuff->offset + srclen + 2;
2058 if (newlen >= sbuff->buflen) {
2059 while (sbuff->buflen < newlen)
2060 sbuff->buflen += sbuff->buflen;
2061 sbuff->data = erealloc(sbuff->data, sbuff->buflen);
2062 flag = 1;
2063 }
2064
2065 /* Append a space at the end of the buffer. */
2066 if (sbuff->offset || flag)
2067 sbuff->data[sbuff->offset++] = ' ';
2068 /* Now, copy src at the end of the buffer. */
2069 memcpy(sbuff->data + sbuff->offset, temp, srclen);
2070 sbuff->offset += srclen;
2071 free(temp);
2072 }
2073
2074 static void
2075 usage(void)
2076 {
2077 fprintf(stderr, "Usage: %s [-floQqv] [-C path]\n", getprogname());
2078 exit(1);
2079 }
2080