makemandb.c revision 1.21 1 /* $NetBSD: makemandb.c,v 1.21 2014/01/05 19:26:44 joerg 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 1.21 2014/01/05 19:26:44 joerg 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, 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 char *dbp = get_dbpath(manconf);
354 /* No error here, it will fail in init_db in the same call */
355 if (dbp != NULL)
356 remove(dbp);
357 }
358
359 if ((db = init_db(MANDB_CREATE, manconf)) == NULL)
360 exit(EXIT_FAILURE);
361
362 sqlite3_exec(db, "PRAGMA synchronous = 0", NULL, NULL, &errmsg);
363 if (errmsg != NULL) {
364 warnx("%s", errmsg);
365 free(errmsg);
366 close_db(db);
367 exit(EXIT_FAILURE);
368 }
369
370 sqlite3_exec(db, "ATTACH DATABASE \':memory:\' AS metadb", NULL, NULL,
371 &errmsg);
372 if (errmsg != NULL) {
373 warnx("%s", errmsg);
374 free(errmsg);
375 close_db(db);
376 exit(EXIT_FAILURE);
377 }
378
379
380 /* Call man -p to get the list of man page dirs */
381 if ((file = popen(command, "r")) == NULL) {
382 close_db(db);
383 err(EXIT_FAILURE, "fopen failed");
384 }
385 free(command);
386
387 /* Begin the transaction for indexing the pages */
388 sqlite3_exec(db, "BEGIN", NULL, NULL, &errmsg);
389 if (errmsg != NULL) {
390 warnx("%s", errmsg);
391 free(errmsg);
392 exit(EXIT_FAILURE);
393 }
394
395 sqlstr = "CREATE TABLE metadb.file_cache(device, inode, mtime, parent,"
396 " file PRIMARY KEY);"
397 "CREATE UNIQUE INDEX metadb.index_file_cache_dev"
398 " ON file_cache (device, inode)";
399
400 sqlite3_exec(db, sqlstr, NULL, NULL, &errmsg);
401 if (errmsg != NULL) {
402 warnx("%s", errmsg);
403 free(errmsg);
404 close_db(db);
405 exit(EXIT_FAILURE);
406 }
407
408 if (mflags.verbosity)
409 printf("Building temporary file cache\n");
410 line = NULL;
411 linesize = 0;
412 while ((len = getline(&line, &linesize, file)) != -1) {
413 /* Replace the new line character at the end of string with '\0' */
414 line[len - 1] = '\0';
415 parent = estrdup(line);
416 char *pdir = estrdup(dirname(parent));
417 free(parent);
418 /* Traverse the man page directories and parse the pages */
419 traversedir(pdir, line, db, mp);
420 free(pdir);
421 }
422 free(line);
423
424 if (pclose(file) == -1) {
425 close_db(db);
426 cleanup(&rec);
427 free_secbuffs(&rec);
428 err(EXIT_FAILURE, "pclose error");
429 }
430
431 if (mflags.verbosity)
432 printf("Performing index update\n");
433 update_db(db, mp, &rec);
434 mparse_free(mp);
435 free_secbuffs(&rec);
436
437 /* Commit the transaction */
438 sqlite3_exec(db, "COMMIT", NULL, NULL, &errmsg);
439 if (errmsg != NULL) {
440 warnx("%s", errmsg);
441 free(errmsg);
442 exit(EXIT_FAILURE);
443 }
444
445 if (mflags.optimize)
446 optimize(db);
447
448 close_db(db);
449 return 0;
450 }
451
452 /*
453 * traversedir --
454 * Traverses the given directory recursively and passes all the man page files
455 * in the way to build_file_cache()
456 */
457 static void
458 traversedir(const char *parent, const char *file, sqlite3 *db,
459 struct mparse *mp)
460 {
461 struct stat sb;
462 struct dirent *dirp;
463 DIR *dp;
464 char *buf;
465
466 if (stat(file, &sb) < 0) {
467 if (mflags.verbosity)
468 warn("stat failed: %s", file);
469 return;
470 }
471
472 /* If it is a directory, traverse it recursively */
473 if (S_ISDIR(sb.st_mode)) {
474 if ((dp = opendir(file)) == NULL) {
475 if (mflags.verbosity)
476 warn("opendir error: %s", file);
477 return;
478 }
479
480 while ((dirp = readdir(dp)) != NULL) {
481 /* Avoid . and .. entries in a directory */
482 if (strncmp(dirp->d_name, ".", 1)) {
483 easprintf(&buf, "%s/%s", file, dirp->d_name);
484 traversedir(parent, buf, db, mp);
485 free(buf);
486 }
487 }
488 closedir(dp);
489 }
490
491 if (!S_ISREG(sb.st_mode) && !S_ISLNK(sb.st_mode))
492 return;
493
494 if (sb.st_size == 0) {
495 if (mflags.verbosity)
496 warnx("Empty file: %s", file);
497 return;
498 }
499 build_file_cache(db, parent, file, &sb);
500 }
501
502 /* build_file_cache --
503 * This function generates an md5 hash of the file passed as it's 2nd parameter
504 * and stores it in a temporary table file_cache along with the full file path.
505 * This is done to support incremental updation of the database.
506 * The temporary table file_cache is dropped thereafter in the function
507 * update_db(), once the database has been updated.
508 */
509 static void
510 build_file_cache(sqlite3 *db, const char *parent, const char *file,
511 struct stat *sb)
512 {
513 const char *sqlstr;
514 sqlite3_stmt *stmt = NULL;
515 int rc, idx;
516 assert(file != NULL);
517 dev_t device_cache = sb->st_dev;
518 ino_t inode_cache = sb->st_ino;
519 time_t mtime_cache = sb->st_mtime;
520
521 sqlstr = "INSERT INTO metadb.file_cache VALUES (:device, :inode,"
522 " :mtime, :parent, :file)";
523 rc = sqlite3_prepare_v2(db, sqlstr, -1, &stmt, NULL);
524 if (rc != SQLITE_OK) {
525 if (mflags.verbosity)
526 warnx("%s", sqlite3_errmsg(db));
527 return;
528 }
529
530 idx = sqlite3_bind_parameter_index(stmt, ":device");
531 rc = sqlite3_bind_int64(stmt, idx, device_cache);
532 if (rc != SQLITE_OK) {
533 if (mflags.verbosity)
534 warnx("%s", sqlite3_errmsg(db));
535 sqlite3_finalize(stmt);
536 return;
537 }
538
539 idx = sqlite3_bind_parameter_index(stmt, ":inode");
540 rc = sqlite3_bind_int64(stmt, idx, inode_cache);
541 if (rc != SQLITE_OK) {
542 if (mflags.verbosity)
543 warnx("%s", sqlite3_errmsg(db));
544 sqlite3_finalize(stmt);
545 return;
546 }
547
548 idx = sqlite3_bind_parameter_index(stmt, ":mtime");
549 rc = sqlite3_bind_int64(stmt, idx, mtime_cache);
550 if (rc != SQLITE_OK) {
551 if (mflags.verbosity)
552 warnx("%s", sqlite3_errmsg(db));
553 sqlite3_finalize(stmt);
554 return;
555 }
556
557 idx = sqlite3_bind_parameter_index(stmt, ":parent");
558 rc = sqlite3_bind_text(stmt, idx, parent, -1, NULL);
559 if (rc != SQLITE_OK) {
560 if (mflags.verbosity)
561 warnx("%s", sqlite3_errmsg(db));
562 sqlite3_finalize(stmt);
563 return;
564 }
565
566 idx = sqlite3_bind_parameter_index(stmt, ":file");
567 rc = sqlite3_bind_text(stmt, idx, file, -1, NULL);
568 if (rc != SQLITE_OK) {
569 if (mflags.verbosity)
570 warnx("%s", sqlite3_errmsg(db));
571 sqlite3_finalize(stmt);
572 return;
573 }
574
575 sqlite3_step(stmt);
576 sqlite3_finalize(stmt);
577 }
578
579 static void
580 update_existing_entry(sqlite3 *db, const char *file, const char *hash,
581 mandb_rec *rec, int *new_count, int *link_count, int *err_count)
582 {
583 int update_count, rc, idx;
584 const char *inner_sqlstr;
585 sqlite3_stmt *inner_stmt;
586
587 update_count = sqlite3_total_changes(db);
588 inner_sqlstr = "UPDATE mandb_meta SET device = :device,"
589 " inode = :inode, mtime = :mtime WHERE"
590 " md5_hash = :md5 AND file = :file AND"
591 " (device <> :device2 OR inode <> "
592 " :inode2 OR mtime <> :mtime2)";
593 rc = sqlite3_prepare_v2(db, inner_sqlstr, -1, &inner_stmt, NULL);
594 if (rc != SQLITE_OK) {
595 if (mflags.verbosity)
596 warnx("%s", sqlite3_errmsg(db));
597 return;
598 }
599 idx = sqlite3_bind_parameter_index(inner_stmt, ":device");
600 sqlite3_bind_int64(inner_stmt, idx, rec->device);
601 idx = sqlite3_bind_parameter_index(inner_stmt, ":inode");
602 sqlite3_bind_int64(inner_stmt, idx, rec->inode);
603 idx = sqlite3_bind_parameter_index(inner_stmt, ":mtime");
604 sqlite3_bind_int64(inner_stmt, idx, rec->mtime);
605 idx = sqlite3_bind_parameter_index(inner_stmt, ":md5");
606 sqlite3_bind_text(inner_stmt, idx, hash, -1, NULL);
607 idx = sqlite3_bind_parameter_index(inner_stmt, ":file");
608 sqlite3_bind_text(inner_stmt, idx, file, -1, NULL);
609 idx = sqlite3_bind_parameter_index(inner_stmt, ":device2");
610 sqlite3_bind_int64(inner_stmt, idx, rec->device);
611 idx = sqlite3_bind_parameter_index(inner_stmt, ":inode2");
612 sqlite3_bind_int64(inner_stmt, idx, rec->inode);
613 idx = sqlite3_bind_parameter_index(inner_stmt, ":mtime2");
614 sqlite3_bind_int64(inner_stmt, idx, rec->mtime);
615
616 rc = sqlite3_step(inner_stmt);
617 if (rc == SQLITE_DONE) {
618 /* Check if an update has been performed. */
619 if (update_count != sqlite3_total_changes(db)) {
620 if (mflags.verbosity == 2)
621 printf("Updated %s\n", file);
622 (*new_count)++;
623 } else {
624 /* Otherwise it was a hardlink. */
625 (*link_count)++;
626 }
627 } else {
628 if (mflags.verbosity == 2)
629 warnx("Could not update the meta data for %s", file);
630 (*err_count)++;
631 }
632 sqlite3_finalize(inner_stmt);
633 }
634
635 /* read_and_decompress --
636 * Reads the given file into memory. If it is compressed, decompres
637 * it before returning to the caller.
638 */
639 static int
640 read_and_decompress(const char *file, void **buf, size_t *len)
641 {
642 size_t off;
643 ssize_t r;
644 struct archive *a;
645 struct archive_entry *ae;
646
647 if ((a = archive_read_new()) == NULL)
648 errx(EXIT_FAILURE, "memory allocation failed");
649
650 if (archive_read_support_compression_all(a) != ARCHIVE_OK ||
651 archive_read_support_format_raw(a) != ARCHIVE_OK ||
652 archive_read_open_filename(a, file, 65536) != ARCHIVE_OK ||
653 archive_read_next_header(a, &ae) != ARCHIVE_OK)
654 goto archive_error;
655 *len = 65536;
656 *buf = emalloc(*len);
657 off = 0;
658 for (;;) {
659 r = archive_read_data(a, (char *)*buf + off, *len - off);
660 if (r == ARCHIVE_OK) {
661 archive_read_close(a);
662 *len = off;
663 return 0;
664 }
665 if (r <= 0) {
666 free(*buf);
667 break;
668 }
669 off += r;
670 if (off == *len) {
671 *len *= 2;
672 if (*len < off) {
673 if (mflags.verbosity)
674 warnx("File too large: %s", file);
675 free(*buf);
676 archive_read_close(a);
677 return -1;
678 }
679 *buf = erealloc(*buf, *len);
680 }
681 }
682
683 archive_error:
684 warnx("Error while reading `%s': %s", file, archive_error_string(a));
685 archive_read_close(a);
686 return -1;
687 }
688
689 /* update_db --
690 * Does an incremental updation of the database by checking the file_cache.
691 * It parses and adds the pages which are present in file_cache,
692 * but not in the database.
693 * It also removes the pages which are present in the databse,
694 * but not in the file_cache.
695 */
696 static void
697 update_db(sqlite3 *db, struct mparse *mp, mandb_rec *rec)
698 {
699 const char *sqlstr;
700 sqlite3_stmt *stmt = NULL;
701 const char *file;
702 const char *parent;
703 char *errmsg = NULL;
704 char *md5sum;
705 void *buf;
706 size_t buflen;
707 int new_count = 0; /* Counter for newly indexed/updated pages */
708 int total_count = 0; /* Counter for total number of pages */
709 int err_count = 0; /* Counter for number of failed pages */
710 int link_count = 0; /* Counter for number of hard/sym links */
711 int md5_status;
712 int rc;
713
714 sqlstr = "SELECT device, inode, mtime, parent, file"
715 " FROM metadb.file_cache fc"
716 " WHERE NOT EXISTS(SELECT 1 FROM mandb_meta WHERE"
717 " device = fc.device AND inode = fc.inode AND "
718 " mtime = fc.mtime AND file = fc.file)";
719
720 rc = sqlite3_prepare_v2(db, sqlstr, -1, &stmt, NULL);
721 if (rc != SQLITE_OK) {
722 if (mflags.verbosity)
723 warnx("%s", sqlite3_errmsg(db));
724 close_db(db);
725 errx(EXIT_FAILURE, "Could not query file cache");
726 }
727
728 buf = NULL;
729 while (sqlite3_step(stmt) == SQLITE_ROW) {
730 free(buf);
731 total_count++;
732 rec->device = sqlite3_column_int64(stmt, 0);
733 rec->inode = sqlite3_column_int64(stmt, 1);
734 rec->mtime = sqlite3_column_int64(stmt, 2);
735 parent = (const char *) sqlite3_column_text(stmt, 3);
736 file = (const char *) sqlite3_column_text(stmt, 4);
737 if (read_and_decompress(file, &buf, &buflen)) {
738 err_count++;
739 buf = NULL;
740 continue;
741 }
742 md5_status = check_md5(file, db, "mandb_meta", &md5sum, buf, buflen);
743 assert(md5sum != NULL);
744 if (md5_status == -1) {
745 if (mflags.verbosity)
746 warnx("An error occurred in checking md5 value"
747 " for file %s", file);
748 err_count++;
749 continue;
750 }
751
752 if (md5_status == 0) {
753 /*
754 * The MD5 hash is already present in the database,
755 * so simply update the metadata, ignoring symlinks.
756 */
757 struct stat sb;
758 stat(file, &sb);
759 if (S_ISLNK(sb.st_mode)) {
760 free(md5sum);
761 link_count++;
762 continue;
763 }
764 update_existing_entry(db, file, md5sum, rec,
765 &new_count, &link_count, &err_count);
766 free(md5sum);
767 continue;
768 }
769
770 if (md5_status == 1) {
771 /*
772 * The MD5 hash was not present in the database.
773 * This means is either a new file or an updated file.
774 * We should go ahead with parsing.
775 */
776 if (mflags.verbosity == 2)
777 printf("Parsing: %s\n", file);
778 rec->md5_hash = md5sum;
779 rec->file_path = estrdup(file);
780 // file_path is freed by insert_into_db itself.
781 chdir(parent);
782 begin_parse(file, mp, rec, buf, buflen);
783 if (insert_into_db(db, rec) < 0) {
784 if (mflags.verbosity)
785 warnx("Error in indexing %s", file);
786 err_count++;
787 } else {
788 new_count++;
789 }
790 }
791 }
792 free(buf);
793
794 sqlite3_finalize(stmt);
795
796 if (mflags.verbosity == 2) {
797 printf("Total Number of new or updated pages encountered = %d\n"
798 "Total number of (hard or symbolic) links found = %d\n"
799 "Total number of pages that were successfully"
800 " indexed/updated = %d\n"
801 "Total number of pages that could not be indexed"
802 " due to errors = %d\n",
803 total_count - link_count, link_count, new_count, err_count);
804 }
805
806 if (mflags.recreate)
807 return;
808
809 if (mflags.verbosity == 2)
810 printf("Deleting stale index entries\n");
811
812 sqlstr = "DELETE FROM mandb_meta WHERE file NOT IN"
813 " (SELECT file FROM metadb.file_cache);"
814 "DELETE FROM mandb_links WHERE md5_hash NOT IN"
815 " (SELECT md5_hash from mandb_meta);"
816 "DROP TABLE metadb.file_cache;"
817 "DELETE FROM mandb WHERE rowid NOT IN"
818 " (SELECT id FROM mandb_meta);";
819
820 sqlite3_exec(db, sqlstr, NULL, NULL, &errmsg);
821 if (errmsg != NULL) {
822 warnx("Removing old entries failed: %s", errmsg);
823 warnx("Please rebuild database from scratch with -f.");
824 free(errmsg);
825 return;
826 }
827 }
828
829 /*
830 * begin_parse --
831 * parses the man page using libmandoc
832 */
833 static void
834 begin_parse(const char *file, struct mparse *mp, mandb_rec *rec,
835 const void *buf, size_t len)
836 {
837 struct mdoc *mdoc;
838 struct man *man;
839 mparse_reset(mp);
840
841 rec->xr_found = 0;
842
843 if (mparse_readmem(mp, buf, len, file) >= MANDOCLEVEL_FATAL) {
844 /* Printing this warning at verbosity level 2
845 * because some packages from pkgsrc might trigger several
846 * of such warnings.
847 */
848 if (mflags.verbosity == 2)
849 warnx("%s: Parse failure", file);
850 return;
851 }
852
853 mparse_result(mp, &mdoc, &man);
854 if (mdoc == NULL && man == NULL) {
855 if (mflags.verbosity == 2)
856 warnx("Not a man(7) or mdoc(7) page");
857 return;
858 }
859
860 set_machine(mdoc, rec);
861 set_section(mdoc, man, rec);
862 if (mdoc) {
863 rec->page_type = MDOC;
864 pmdoc_node(mdoc_node(mdoc), rec);
865 } else {
866 rec->page_type = MAN;
867 pman_node(man_node(man), rec);
868 }
869 }
870
871 /*
872 * set_section --
873 * Extracts the section number and normalizes it to only the numeric part
874 * (Which should be the first character of the string).
875 */
876 static void
877 set_section(const struct mdoc *md, const struct man *m, mandb_rec *rec)
878 {
879 if (md) {
880 const struct mdoc_meta *md_meta = mdoc_meta(md);
881 rec->section[0] = md_meta->msec[0];
882 } else if (m) {
883 const struct man_meta *m_meta = man_meta(m);
884 rec->section[0] = m_meta->msec[0];
885 }
886 }
887
888 /*
889 * get_machine --
890 * Extracts the machine architecture information if available.
891 */
892 static void
893 set_machine(const struct mdoc *md, mandb_rec *rec)
894 {
895 if (md == NULL)
896 return;
897 const struct mdoc_meta *md_meta = mdoc_meta(md);
898 if (md_meta->arch)
899 rec->machine = estrdup(md_meta->arch);
900 }
901
902 static void
903 pmdoc_node(const struct mdoc_node *n, mandb_rec *rec)
904 {
905
906 if (n == NULL)
907 return;
908
909 switch (n->type) {
910 case (MDOC_BODY):
911 /* FALLTHROUGH */
912 case (MDOC_TAIL):
913 /* FALLTHROUGH */
914 case (MDOC_ELEM):
915 if (mdocs[n->tok] == NULL)
916 break;
917 (*mdocs[n->tok])(n, rec);
918 break;
919 default:
920 break;
921 }
922
923 pmdoc_node(n->child, rec);
924 pmdoc_node(n->next, rec);
925 }
926
927 /*
928 * pmdoc_Nm --
929 * Extracts the Name of the manual page from the .Nm macro
930 */
931 static void
932 pmdoc_Nm(const struct mdoc_node *n, mandb_rec *rec)
933 {
934 if (n->sec != SEC_NAME)
935 return;
936
937 for (n = n->child; n; n = n->next) {
938 if (n->type == MDOC_TEXT) {
939 concat(&rec->name, n->string);
940 }
941 }
942 }
943
944 /*
945 * pmdoc_Nd --
946 * Extracts the one line description of the man page from the .Nd macro
947 */
948 static void
949 pmdoc_Nd(const struct mdoc_node *n, mandb_rec *rec)
950 {
951 /*
952 * A static variable for keeping track of whether a Xr macro was seen
953 * previously.
954 */
955 char *buf = NULL;
956 char *temp;
957
958 if (n == NULL)
959 return;
960
961 if (n->type == MDOC_TEXT) {
962 if (rec->xr_found && n->next) {
963 /*
964 * An Xr macro was seen previously, so parse this
965 * and the next node.
966 */
967 temp = estrdup(n->string);
968 n = n->next;
969 easprintf(&buf, "%s(%s)", temp, n->string);
970 concat(&rec->name_desc, buf);
971 free(buf);
972 free(temp);
973 } else {
974 concat(&rec->name_desc, n->string);
975 }
976 rec->xr_found = 0;
977 } else if (mdocs[n->tok] == pmdoc_Xr) {
978 /* Remember that we have encountered an Xr macro */
979 rec->xr_found = 1;
980 }
981
982 if (n->child)
983 pmdoc_Nd(n->child, rec);
984
985 if(n->next)
986 pmdoc_Nd(n->next, rec);
987 }
988
989 /*
990 * pmdoc_macro_handler--
991 * This function is a single point of handling all the special macros that we
992 * want to handle especially. For example the .Xr macro for properly parsing
993 * the referenced page name along with the section number, or the .Pp macro
994 * for adding a new line whenever we encounter it.
995 */
996 static void
997 pmdoc_macro_handler(const struct mdoc_node *n, mandb_rec *rec, enum mdoct doct)
998 {
999 const struct mdoc_node *sn;
1000 assert(n);
1001
1002 switch (doct) {
1003 /* Parse the man page references.
1004 * Basically the .Xr macros are used like:
1005 * .Xr ls 1
1006 * and formatted like this:
1007 * ls(1)
1008 * Prepare a buffer to format the data like the above example and call
1009 * pmdoc_parse_section to append it.
1010 */
1011 case MDOC_Xr:
1012 n = n->child;
1013 while (n->type != MDOC_TEXT && n->next)
1014 n = n->next;
1015
1016 if (n && n->type != MDOC_TEXT)
1017 return;
1018 sn = n;
1019 if (n->next)
1020 n = n->next;
1021
1022 while (n->type != MDOC_TEXT && n->next)
1023 n = n->next;
1024
1025 if (n && n->type == MDOC_TEXT) {
1026 size_t len = strlen(sn->string);
1027 char *buf = emalloc(len + 4);
1028 memcpy(buf, sn->string, len);
1029 buf[len] = '(';
1030 buf[len + 1] = n->string[0];
1031 buf[len + 2] = ')';
1032 buf[len + 3] = 0;
1033 mdoc_parse_section(n->sec, buf, rec);
1034 free(buf);
1035 }
1036
1037 break;
1038
1039 /* Parse the .Pp macro to add a new line */
1040 case MDOC_Pp:
1041 if (n->type == MDOC_TEXT)
1042 mdoc_parse_section(n->sec, "\n", rec);
1043 break;
1044 default:
1045 break;
1046 }
1047
1048 }
1049
1050 /*
1051 * pmdoc_Xr, pmdoc_Pp--
1052 * Empty stubs.
1053 * The parser calls these functions each time it encounters
1054 * a .Xr or .Pp macro. We are parsing all the data from
1055 * the pmdoc_Sh function, so don't do anything here.
1056 * (See if else blocks in pmdoc_Sh.)
1057 */
1058 static void
1059 pmdoc_Xr(const struct mdoc_node *n, mandb_rec *rec)
1060 {
1061 }
1062
1063 static void
1064 pmdoc_Pp(const struct mdoc_node *n, mandb_rec *rec)
1065 {
1066 }
1067
1068 /*
1069 * pmdoc_Sh --
1070 * Called when a .Sh macro is encountered and loops through its body, calling
1071 * mdoc_parse_section to append the data to the section specific buffer.
1072 * Two special macros which may occur inside the body of Sh are .Nm and .Xr and
1073 * they need special handling, thus the separate if branches for them.
1074 */
1075 static void
1076 pmdoc_Sh(const struct mdoc_node *n, mandb_rec *rec)
1077 {
1078 if (n == NULL)
1079 return;
1080 int xr_found = 0;
1081
1082 if (n->type == MDOC_TEXT) {
1083 mdoc_parse_section(n->sec, n->string, rec);
1084 } else if (mdocs[n->tok] == pmdoc_Nm && rec->name != NULL) {
1085 /*
1086 * When encountering a .Nm macro, substitute it
1087 * with its previously cached value of the argument.
1088 */
1089 mdoc_parse_section(n->sec, rec->name, rec);
1090 } else if (mdocs[n->tok] == pmdoc_Xr) {
1091 /*
1092 * When encountering other inline macros,
1093 * call pmdoc_macro_handler.
1094 */
1095 pmdoc_macro_handler(n, rec, MDOC_Xr);
1096 xr_found = 1;
1097 } else if (mdocs[n->tok] == pmdoc_Pp) {
1098 pmdoc_macro_handler(n, rec, MDOC_Pp);
1099 }
1100
1101 /*
1102 * If an Xr macro was encountered then the child node has
1103 * already been explored by pmdoc_macro_handler.
1104 */
1105 if (xr_found == 0)
1106 pmdoc_Sh(n->child, rec);
1107 pmdoc_Sh(n->next, rec);
1108 }
1109
1110 /*
1111 * mdoc_parse_section--
1112 * Utility function for parsing sections of the mdoc type pages.
1113 * Takes two params:
1114 * 1. sec is an enum which indicates the section in which we are present
1115 * 2. string is the string which we need to append to the secbuff for this
1116 * particular section.
1117 * The function appends string to the global section buffer and returns.
1118 */
1119 static void
1120 mdoc_parse_section(enum mdoc_sec sec, const char *string, mandb_rec *rec)
1121 {
1122 /*
1123 * If the user specified the 'l' flag, then parse and store only the
1124 * NAME section. Ignore the rest.
1125 */
1126 if (mflags.limit)
1127 return;
1128
1129 switch (sec) {
1130 case SEC_LIBRARY:
1131 append(&rec->lib, string);
1132 break;
1133 case SEC_RETURN_VALUES:
1134 append(&rec->return_vals, string);
1135 break;
1136 case SEC_ENVIRONMENT:
1137 append(&rec->env, string);
1138 break;
1139 case SEC_FILES:
1140 append(&rec->files, string);
1141 break;
1142 case SEC_EXIT_STATUS:
1143 append(&rec->exit_status, string);
1144 break;
1145 case SEC_DIAGNOSTICS:
1146 append(&rec->diagnostics, string);
1147 break;
1148 case SEC_ERRORS:
1149 append(&rec->errors, string);
1150 break;
1151 case SEC_NAME:
1152 case SEC_SYNOPSIS:
1153 case SEC_EXAMPLES:
1154 case SEC_STANDARDS:
1155 case SEC_HISTORY:
1156 case SEC_AUTHORS:
1157 case SEC_BUGS:
1158 break;
1159 default:
1160 append(&rec->desc, string);
1161 break;
1162 }
1163 }
1164
1165 static void
1166 pman_node(const struct man_node *n, mandb_rec *rec)
1167 {
1168 if (n == NULL)
1169 return;
1170
1171 switch (n->type) {
1172 case (MAN_BODY):
1173 /* FALLTHROUGH */
1174 case (MAN_TAIL):
1175 /* FALLTHROUGH */
1176 case (MAN_BLOCK):
1177 /* FALLTHROUGH */
1178 case (MAN_ELEM):
1179 if (mans[n->tok] != NULL)
1180 (*mans[n->tok])(n, rec);
1181 break;
1182 default:
1183 break;
1184 }
1185
1186 pman_node(n->child, rec);
1187 pman_node(n->next, rec);
1188 }
1189
1190 /*
1191 * pman_parse_name --
1192 * Parses the NAME section and puts the complete content in the name_desc
1193 * variable.
1194 */
1195 static void
1196 pman_parse_name(const struct man_node *n, mandb_rec *rec)
1197 {
1198 if (n == NULL)
1199 return;
1200
1201 if (n->type == MAN_TEXT) {
1202 char *tmp = parse_escape(n->string);
1203 concat(&rec->name_desc, tmp);
1204 free(tmp);
1205 }
1206
1207 if (n->child)
1208 pman_parse_name(n->child, rec);
1209
1210 if(n->next)
1211 pman_parse_name(n->next, rec);
1212 }
1213
1214 /*
1215 * A stub function to be able to parse the macros like .B embedded inside
1216 * a section.
1217 */
1218 static void
1219 pman_block(const struct man_node *n, mandb_rec *rec)
1220 {
1221 }
1222
1223 /*
1224 * pman_sh --
1225 * This function does one of the two things:
1226 * 1. If the present section is NAME, then it will:
1227 * (a) Extract the name of the page (in case of multiple comma separated
1228 * names, it will pick up the first one).
1229 * (b) Build a space spearated list of all the symlinks/hardlinks to
1230 * this page and store in the buffer 'links'. These are extracted from
1231 * the comma separated list of names in the NAME section as well.
1232 * (c) Move on to the one line description section, which is after the list
1233 * of names in the NAME section.
1234 * 2. Otherwise, it will check the section name and call the man_parse_section
1235 * function, passing the enum corresponding that section.
1236 */
1237 static void
1238 pman_sh(const struct man_node *n, mandb_rec *rec)
1239 {
1240 static const struct {
1241 enum man_sec section;
1242 const char *header;
1243 } mapping[] = {
1244 { MANSEC_DESCRIPTION, "DESCRIPTION" },
1245 { MANSEC_SYNOPSIS, "SYNOPSIS" },
1246 { MANSEC_LIBRARY, "LIBRARY" },
1247 { MANSEC_ERRORS, "ERRORS" },
1248 { MANSEC_FILES, "FILES" },
1249 { MANSEC_RETURN_VALUES, "RETURN VALUE" },
1250 { MANSEC_RETURN_VALUES, "RETURN VALUES" },
1251 { MANSEC_EXIT_STATUS, "EXIT STATUS" },
1252 { MANSEC_EXAMPLES, "EXAMPLES" },
1253 { MANSEC_EXAMPLES, "EXAMPLE" },
1254 { MANSEC_STANDARDS, "STANDARDS" },
1255 { MANSEC_HISTORY, "HISTORY" },
1256 { MANSEC_BUGS, "BUGS" },
1257 { MANSEC_AUTHORS, "AUTHORS" },
1258 { MANSEC_COPYRIGHT, "COPYRIGHT" },
1259 };
1260 const struct man_node *head;
1261 char *name_desc;
1262 int sz;
1263 size_t i;
1264
1265 if ((head = n->parent->head) == NULL || (head = head->child) == NULL ||
1266 head->type != MAN_TEXT)
1267 return;
1268
1269 /*
1270 * Check if this section should be extracted and
1271 * where it should be stored. Handled the trival cases first.
1272 */
1273 for (i = 0; i < sizeof(mapping) / sizeof(mapping[0]); ++i) {
1274 if (strcmp(head->string, mapping[i].header) == 0) {
1275 man_parse_section(mapping[i].section, n, rec);
1276 return;
1277 }
1278 }
1279
1280 if (strcmp(head->string, "NAME") == 0) {
1281 /*
1282 * We are in the NAME section.
1283 * pman_parse_name will put the complete content in name_desc.
1284 */
1285 pman_parse_name(n, rec);
1286
1287 name_desc = rec->name_desc;
1288 if (name_desc == NULL)
1289 return;
1290
1291 /* Remove any leading spaces. */
1292 while (name_desc[0] == ' ')
1293 name_desc++;
1294
1295 /* If the line begins with a "\&", avoid those */
1296 if (name_desc[0] == '\\' && name_desc[1] == '&')
1297 name_desc += 2;
1298
1299 /* Now name_desc should be left with a comma-space
1300 * separated list of names and the one line description
1301 * of the page:
1302 * "a, b, c \- sample description"
1303 * Take out the first name, before the first comma
1304 * (or space) and store it in rec->name.
1305 * If the page has aliases then they should be
1306 * in the form of a comma separated list.
1307 * Keep looping while there is a comma in name_desc,
1308 * extract the alias name and store in rec->links.
1309 * When there are no more commas left, break out.
1310 */
1311 int has_alias = 0; // Any more aliases left?
1312 while (*name_desc) {
1313 /* Remove any leading spaces or hyphens. */
1314 if (name_desc[0] == ' ' || name_desc[0] =='-') {
1315 name_desc++;
1316 continue;
1317 }
1318 sz = strcspn(name_desc, ", ");
1319
1320 /* Extract the first term and store it in rec->name. */
1321 if (rec->name == NULL) {
1322 if (name_desc[sz] == ',')
1323 has_alias = 1;
1324 name_desc[sz] = 0;
1325 rec->name = emalloc(sz + 1);
1326 memcpy(rec->name, name_desc, sz + 1);
1327 name_desc += sz + 1;
1328 continue;
1329 }
1330
1331 /*
1332 * Once rec->name is set, rest of the names
1333 * are to be treated as links or aliases.
1334 */
1335 if (rec->name && has_alias) {
1336 if (name_desc[sz] != ',') {
1337 /* No more commas left -->
1338 * no more aliases to take out
1339 */
1340 has_alias = 0;
1341 }
1342 name_desc[sz] = 0;
1343 concat2(&rec->links, name_desc, sz);
1344 name_desc += sz + 1;
1345 continue;
1346 }
1347 break;
1348 }
1349
1350 /* Parse any escape sequences that might be there */
1351 char *temp = parse_escape(name_desc);
1352 free(rec->name_desc);
1353 rec->name_desc = temp;
1354 temp = parse_escape(rec->name);
1355 free(rec->name);
1356 rec->name = temp;
1357 return;
1358 }
1359
1360 /* The RETURN VALUE section might be specified in multiple ways */
1361 if (strcmp(head->string, "RETURN") == 0 &&
1362 head->next != NULL && head->next->type == MAN_TEXT &&
1363 (strcmp(head->next->string, "VALUE") == 0 ||
1364 strcmp(head->next->string, "VALUES") == 0)) {
1365 man_parse_section(MANSEC_RETURN_VALUES, n, rec);
1366 return;
1367 }
1368
1369 /*
1370 * EXIT STATUS section can also be specified all on one line or on two
1371 * separate lines.
1372 */
1373 if (strcmp(head->string, "EXIT") == 0 &&
1374 head->next != NULL && head->next->type == MAN_TEXT &&
1375 strcmp(head->next->string, "STATUS") == 0) {
1376 man_parse_section(MANSEC_EXIT_STATUS, n, rec);
1377 return;
1378 }
1379
1380 /* Store the rest of the content in desc. */
1381 man_parse_section(MANSEC_NONE, n, rec);
1382 }
1383
1384 /*
1385 * pman_parse_node --
1386 * Generic function to iterate through a node. Usually called from
1387 * man_parse_section to parse a particular section of the man page.
1388 */
1389 static void
1390 pman_parse_node(const struct man_node *n, secbuff *s)
1391 {
1392 if (n == NULL)
1393 return;
1394
1395 if (n->type == MAN_TEXT)
1396 append(s, n->string);
1397
1398 pman_parse_node(n->child, s);
1399 pman_parse_node(n->next, s);
1400 }
1401
1402 /*
1403 * man_parse_section --
1404 * Takes two parameters:
1405 * sec: Tells which section we are present in
1406 * n: Is the present node of the AST.
1407 * Depending on the section, we call pman_parse_node to parse that section and
1408 * concatenate the content from that section into the buffer for that section.
1409 */
1410 static void
1411 man_parse_section(enum man_sec sec, const struct man_node *n, mandb_rec *rec)
1412 {
1413 /*
1414 * If the user sepecified the 'l' flag then just parse
1415 * the NAME section, ignore the rest.
1416 */
1417 if (mflags.limit)
1418 return;
1419
1420 switch (sec) {
1421 case MANSEC_LIBRARY:
1422 pman_parse_node(n, &rec->lib);
1423 break;
1424 case MANSEC_RETURN_VALUES:
1425 pman_parse_node(n, &rec->return_vals);
1426 break;
1427 case MANSEC_ENVIRONMENT:
1428 pman_parse_node(n, &rec->env);
1429 break;
1430 case MANSEC_FILES:
1431 pman_parse_node(n, &rec->files);
1432 break;
1433 case MANSEC_EXIT_STATUS:
1434 pman_parse_node(n, &rec->exit_status);
1435 break;
1436 case MANSEC_DIAGNOSTICS:
1437 pman_parse_node(n, &rec->diagnostics);
1438 break;
1439 case MANSEC_ERRORS:
1440 pman_parse_node(n, &rec->errors);
1441 break;
1442 case MANSEC_NAME:
1443 case MANSEC_SYNOPSIS:
1444 case MANSEC_EXAMPLES:
1445 case MANSEC_STANDARDS:
1446 case MANSEC_HISTORY:
1447 case MANSEC_BUGS:
1448 case MANSEC_AUTHORS:
1449 case MANSEC_COPYRIGHT:
1450 break;
1451 default:
1452 pman_parse_node(n, &rec->desc);
1453 break;
1454 }
1455
1456 }
1457
1458 /*
1459 * insert_into_db --
1460 * Inserts the parsed data of the man page in the Sqlite databse.
1461 * If any of the values is NULL, then we cleanup and return -1 indicating
1462 * an error.
1463 * Otherwise, store the data in the database and return 0.
1464 */
1465 static int
1466 insert_into_db(sqlite3 *db, mandb_rec *rec)
1467 {
1468 int rc = 0;
1469 int idx = -1;
1470 const char *sqlstr = NULL;
1471 sqlite3_stmt *stmt = NULL;
1472 char *ln = NULL;
1473 char *errmsg = NULL;
1474 long int mandb_rowid;
1475
1476 /*
1477 * At the very minimum we want to make sure that we store
1478 * the following data:
1479 * Name, one line description, and the MD5 hash
1480 */
1481 if (rec->name == NULL || rec->name_desc == NULL ||
1482 rec->md5_hash == NULL) {
1483 cleanup(rec);
1484 return -1;
1485 }
1486
1487 /* Write null byte at the end of all the sec_buffs */
1488 rec->desc.data[rec->desc.offset] = 0;
1489 rec->lib.data[rec->lib.offset] = 0;
1490 rec->env.data[rec->env.offset] = 0;
1491 rec->return_vals.data[rec->return_vals.offset] = 0;
1492 rec->exit_status.data[rec->exit_status.offset] = 0;
1493 rec->files.data[rec->files.offset] = 0;
1494 rec->diagnostics.data[rec->diagnostics.offset] = 0;
1495 rec->errors.data[rec->errors.offset] = 0;
1496
1497 /*
1498 * In case of a mdoc page: (sorry, no better place to put this code)
1499 * parse the comma separated list of names of man pages,
1500 * the first name will be stored in the mandb table, rest will be
1501 * treated as links and put in the mandb_links table.
1502 */
1503 if (rec->page_type == MDOC) {
1504 char *tmp;
1505 rec->links = estrdup(rec->name);
1506 free(rec->name);
1507 int sz = strcspn(rec->links, " \0");
1508 rec->name = emalloc(sz + 1);
1509 memcpy(rec->name, rec->links, sz);
1510 if(rec->name[sz - 1] == ',')
1511 rec->name[sz - 1] = 0;
1512 else
1513 rec->name[sz] = 0;
1514 while (rec->links[sz] == ' ')
1515 ++sz;
1516 tmp = estrdup(rec->links + sz);
1517 free(rec->links);
1518 rec->links = tmp;
1519 }
1520
1521 /*------------------------ Populate the mandb table---------------------------*/
1522 sqlstr = "INSERT INTO mandb VALUES (:section, :name, :name_desc, :desc,"
1523 " :lib, :return_vals, :env, :files, :exit_status,"
1524 " :diagnostics, :errors, :md5_hash, :machine)";
1525
1526 rc = sqlite3_prepare_v2(db, sqlstr, -1, &stmt, NULL);
1527 if (rc != SQLITE_OK)
1528 goto Out;
1529
1530 idx = sqlite3_bind_parameter_index(stmt, ":name");
1531 rc = sqlite3_bind_text(stmt, idx, rec->name, -1, NULL);
1532 if (rc != SQLITE_OK) {
1533 sqlite3_finalize(stmt);
1534 goto Out;
1535 }
1536
1537 idx = sqlite3_bind_parameter_index(stmt, ":section");
1538 rc = sqlite3_bind_text(stmt, idx, rec->section, -1, NULL);
1539 if (rc != SQLITE_OK) {
1540 sqlite3_finalize(stmt);
1541 goto Out;
1542 }
1543
1544 idx = sqlite3_bind_parameter_index(stmt, ":name_desc");
1545 rc = sqlite3_bind_text(stmt, idx, rec->name_desc, -1, NULL);
1546 if (rc != SQLITE_OK) {
1547 sqlite3_finalize(stmt);
1548 goto Out;
1549 }
1550
1551 idx = sqlite3_bind_parameter_index(stmt, ":desc");
1552 rc = sqlite3_bind_text(stmt, idx, rec->desc.data,
1553 rec->desc.offset + 1, NULL);
1554 if (rc != SQLITE_OK) {
1555 sqlite3_finalize(stmt);
1556 goto Out;
1557 }
1558
1559 idx = sqlite3_bind_parameter_index(stmt, ":lib");
1560 rc = sqlite3_bind_text(stmt, idx, rec->lib.data, rec->lib.offset + 1, NULL);
1561 if (rc != SQLITE_OK) {
1562 sqlite3_finalize(stmt);
1563 goto Out;
1564 }
1565
1566 idx = sqlite3_bind_parameter_index(stmt, ":return_vals");
1567 rc = sqlite3_bind_text(stmt, idx, rec->return_vals.data,
1568 rec->return_vals.offset + 1, NULL);
1569 if (rc != SQLITE_OK) {
1570 sqlite3_finalize(stmt);
1571 goto Out;
1572 }
1573
1574 idx = sqlite3_bind_parameter_index(stmt, ":env");
1575 rc = sqlite3_bind_text(stmt, idx, rec->env.data, rec->env.offset + 1, NULL);
1576 if (rc != SQLITE_OK) {
1577 sqlite3_finalize(stmt);
1578 goto Out;
1579 }
1580
1581 idx = sqlite3_bind_parameter_index(stmt, ":files");
1582 rc = sqlite3_bind_text(stmt, idx, rec->files.data,
1583 rec->files.offset + 1, NULL);
1584 if (rc != SQLITE_OK) {
1585 sqlite3_finalize(stmt);
1586 goto Out;
1587 }
1588
1589 idx = sqlite3_bind_parameter_index(stmt, ":exit_status");
1590 rc = sqlite3_bind_text(stmt, idx, rec->exit_status.data,
1591 rec->exit_status.offset + 1, NULL);
1592 if (rc != SQLITE_OK) {
1593 sqlite3_finalize(stmt);
1594 goto Out;
1595 }
1596
1597 idx = sqlite3_bind_parameter_index(stmt, ":diagnostics");
1598 rc = sqlite3_bind_text(stmt, idx, rec->diagnostics.data,
1599 rec->diagnostics.offset + 1, NULL);
1600 if (rc != SQLITE_OK) {
1601 sqlite3_finalize(stmt);
1602 goto Out;
1603 }
1604
1605 idx = sqlite3_bind_parameter_index(stmt, ":errors");
1606 rc = sqlite3_bind_text(stmt, idx, rec->errors.data,
1607 rec->errors.offset + 1, NULL);
1608 if (rc != SQLITE_OK) {
1609 sqlite3_finalize(stmt);
1610 goto Out;
1611 }
1612
1613 idx = sqlite3_bind_parameter_index(stmt, ":md5_hash");
1614 rc = sqlite3_bind_text(stmt, idx, rec->md5_hash, -1, NULL);
1615 if (rc != SQLITE_OK) {
1616 sqlite3_finalize(stmt);
1617 goto Out;
1618 }
1619
1620 idx = sqlite3_bind_parameter_index(stmt, ":machine");
1621 if (rec->machine)
1622 rc = sqlite3_bind_text(stmt, idx, rec->machine, -1, NULL);
1623 else
1624 rc = sqlite3_bind_null(stmt, idx);
1625 if (rc != SQLITE_OK) {
1626 sqlite3_finalize(stmt);
1627 goto Out;
1628 }
1629
1630 rc = sqlite3_step(stmt);
1631 if (rc != SQLITE_DONE) {
1632 sqlite3_finalize(stmt);
1633 goto Out;
1634 }
1635
1636 sqlite3_finalize(stmt);
1637
1638 /* Get the row id of the last inserted row */
1639 mandb_rowid = sqlite3_last_insert_rowid(db);
1640
1641 /*------------------------Populate the mandb_meta table-----------------------*/
1642 sqlstr = "INSERT INTO mandb_meta VALUES (:device, :inode, :mtime,"
1643 " :file, :md5_hash, :id)";
1644 rc = sqlite3_prepare_v2(db, sqlstr, -1, &stmt, NULL);
1645 if (rc != SQLITE_OK)
1646 goto Out;
1647
1648 idx = sqlite3_bind_parameter_index(stmt, ":device");
1649 rc = sqlite3_bind_int64(stmt, idx, rec->device);
1650 if (rc != SQLITE_OK) {
1651 sqlite3_finalize(stmt);
1652 goto Out;
1653 }
1654
1655 idx = sqlite3_bind_parameter_index(stmt, ":inode");
1656 rc = sqlite3_bind_int64(stmt, idx, rec->inode);
1657 if (rc != SQLITE_OK) {
1658 sqlite3_finalize(stmt);
1659 goto Out;
1660 }
1661
1662 idx = sqlite3_bind_parameter_index(stmt, ":mtime");
1663 rc = sqlite3_bind_int64(stmt, idx, rec->mtime);
1664 if (rc != SQLITE_OK) {
1665 sqlite3_finalize(stmt);
1666 goto Out;
1667 }
1668
1669 idx = sqlite3_bind_parameter_index(stmt, ":file");
1670 rc = sqlite3_bind_text(stmt, idx, rec->file_path, -1, NULL);
1671 if (rc != SQLITE_OK) {
1672 sqlite3_finalize(stmt);
1673 goto Out;
1674 }
1675
1676 idx = sqlite3_bind_parameter_index(stmt, ":md5_hash");
1677 rc = sqlite3_bind_text(stmt, idx, rec->md5_hash, -1, NULL);
1678 if (rc != SQLITE_OK) {
1679 sqlite3_finalize(stmt);
1680 goto Out;
1681 }
1682
1683 idx = sqlite3_bind_parameter_index(stmt, ":id");
1684 rc = sqlite3_bind_int64(stmt, idx, mandb_rowid);
1685 if (rc != SQLITE_OK) {
1686 sqlite3_finalize(stmt);
1687 goto Out;
1688 }
1689
1690 rc = sqlite3_step(stmt);
1691 sqlite3_finalize(stmt);
1692 if (rc == SQLITE_CONSTRAINT) {
1693 /* The *most* probable reason for reaching here is that
1694 * the UNIQUE contraint on the file column of the mandb_meta
1695 * table was violated.
1696 * This can happen when a file was updated/modified.
1697 * To fix this we need to do two things:
1698 * 1. Delete the row for the older version of this file
1699 * from mandb table.
1700 * 2. Run an UPDATE query to update the row for this file
1701 * in the mandb_meta table.
1702 */
1703 warnx("Trying to update index for %s", rec->file_path);
1704 char *sql = sqlite3_mprintf("DELETE FROM mandb "
1705 "WHERE rowid = (SELECT id"
1706 " FROM mandb_meta"
1707 " WHERE file = %Q)",
1708 rec->file_path);
1709 sqlite3_exec(db, sql, NULL, NULL, &errmsg);
1710 sqlite3_free(sql);
1711 if (errmsg != NULL) {
1712 if (mflags.verbosity)
1713 warnx("%s", errmsg);
1714 free(errmsg);
1715 }
1716 sqlstr = "UPDATE mandb_meta SET device = :device,"
1717 " inode = :inode, mtime = :mtime, id = :id,"
1718 " md5_hash = :md5 WHERE file = :file";
1719 rc = sqlite3_prepare_v2(db, sqlstr, -1, &stmt, NULL);
1720 if (rc != SQLITE_OK) {
1721 if (mflags.verbosity)
1722 warnx("Update failed with error: %s",
1723 sqlite3_errmsg(db));
1724 close_db(db);
1725 cleanup(rec);
1726 errx(EXIT_FAILURE,
1727 "Consider running makemandb with -f option");
1728 }
1729
1730 idx = sqlite3_bind_parameter_index(stmt, ":device");
1731 sqlite3_bind_int64(stmt, idx, rec->device);
1732 idx = sqlite3_bind_parameter_index(stmt, ":inode");
1733 sqlite3_bind_int64(stmt, idx, rec->inode);
1734 idx = sqlite3_bind_parameter_index(stmt, ":mtime");
1735 sqlite3_bind_int64(stmt, idx, rec->mtime);
1736 idx = sqlite3_bind_parameter_index(stmt, ":id");
1737 sqlite3_bind_int64(stmt, idx, mandb_rowid);
1738 idx = sqlite3_bind_parameter_index(stmt, ":md5");
1739 sqlite3_bind_text(stmt, idx, rec->md5_hash, -1, NULL);
1740 idx = sqlite3_bind_parameter_index(stmt, ":file");
1741 sqlite3_bind_text(stmt, idx, rec->file_path, -1, NULL);
1742 rc = sqlite3_step(stmt);
1743 sqlite3_finalize(stmt);
1744
1745 if (rc != SQLITE_DONE) {
1746 if (mflags.verbosity)
1747 warnx("%s", sqlite3_errmsg(db));
1748 close_db(db);
1749 cleanup(rec);
1750 errx(EXIT_FAILURE,
1751 "Consider running makemandb with -f option");
1752 }
1753 } else if (rc != SQLITE_DONE) {
1754 /* Otherwise make this error fatal */
1755 warnx("Failed at %s\n%s", rec->file_path, sqlite3_errmsg(db));
1756 cleanup(rec);
1757 close_db(db);
1758 exit(EXIT_FAILURE);
1759 }
1760
1761 /*------------------------ Populate the mandb_links table---------------------*/
1762 char *str = NULL;
1763 char *links;
1764 if (rec->links && strlen(rec->links)) {
1765 links = rec->links;
1766 for(ln = strtok(links, " "); ln; ln = strtok(NULL, " ")) {
1767 if (ln[0] == ',')
1768 ln++;
1769 if(ln[strlen(ln) - 1] == ',')
1770 ln[strlen(ln) - 1] = 0;
1771
1772 str = sqlite3_mprintf("INSERT INTO mandb_links"
1773 " VALUES (%Q, %Q, %Q, %Q, %Q)",
1774 ln, rec->name, rec->section,
1775 rec->machine, rec->md5_hash);
1776 sqlite3_exec(db, str, NULL, NULL, &errmsg);
1777 sqlite3_free(str);
1778 if (errmsg != NULL) {
1779 warnx("%s", errmsg);
1780 cleanup(rec);
1781 free(errmsg);
1782 return -1;
1783 }
1784 }
1785 }
1786
1787 cleanup(rec);
1788 return 0;
1789
1790 Out:
1791 if (mflags.verbosity)
1792 warnx("%s", sqlite3_errmsg(db));
1793 cleanup(rec);
1794 return -1;
1795 }
1796
1797 /*
1798 * check_md5--
1799 * Generates the md5 hash of the file and checks if it already doesn't exist
1800 * in the table (passed as the 3rd parameter).
1801 * This function is being used to avoid hardlinks.
1802 * On successful completion it will also set the value of the fourth parameter
1803 * to the md5 hash of the file (computed previously). It is the responsibility
1804 * of the caller to free this buffer.
1805 * Return values:
1806 * -1: If an error occurs somewhere and sets the md5 return buffer to NULL.
1807 * 0: If the md5 hash does not exist in the table.
1808 * 1: If the hash exists in the database.
1809 */
1810 static int
1811 check_md5(const char *file, sqlite3 *db, const char *table, char **md5sum,
1812 void *buf, size_t buflen)
1813 {
1814 int rc = 0;
1815 int idx = -1;
1816 char *sqlstr = NULL;
1817 sqlite3_stmt *stmt = NULL;
1818
1819 assert(file != NULL);
1820 *md5sum = MD5Data(buf, buflen, NULL);
1821 if (*md5sum == NULL) {
1822 if (mflags.verbosity)
1823 warn("md5 failed: %s", file);
1824 return -1;
1825 }
1826
1827 easprintf(&sqlstr, "SELECT * FROM %s WHERE md5_hash = :md5_hash",
1828 table);
1829 rc = sqlite3_prepare_v2(db, sqlstr, -1, &stmt, NULL);
1830 if (rc != SQLITE_OK) {
1831 free(sqlstr);
1832 free(*md5sum);
1833 *md5sum = NULL;
1834 return -1;
1835 }
1836
1837 idx = sqlite3_bind_parameter_index(stmt, ":md5_hash");
1838 rc = sqlite3_bind_text(stmt, idx, *md5sum, -1, NULL);
1839 if (rc != SQLITE_OK) {
1840 if (mflags.verbosity)
1841 warnx("%s", sqlite3_errmsg(db));
1842 sqlite3_finalize(stmt);
1843 free(sqlstr);
1844 free(*md5sum);
1845 *md5sum = NULL;
1846 return -1;
1847 }
1848
1849 if (sqlite3_step(stmt) == SQLITE_ROW) {
1850 sqlite3_finalize(stmt);
1851 free(sqlstr);
1852 return 0;
1853 }
1854
1855 sqlite3_finalize(stmt);
1856 free(sqlstr);
1857 return 1;
1858 }
1859
1860 /* Optimize the index for faster search */
1861 static void
1862 optimize(sqlite3 *db)
1863 {
1864 const char *sqlstr;
1865 char *errmsg = NULL;
1866
1867 if (mflags.verbosity == 2)
1868 printf("Optimizing the database index\n");
1869 sqlstr = "INSERT INTO mandb(mandb) VALUES (\'optimize\');"
1870 "VACUUM";
1871 sqlite3_exec(db, sqlstr, NULL, NULL, &errmsg);
1872 if (errmsg != NULL) {
1873 if (mflags.verbosity)
1874 warnx("%s", errmsg);
1875 free(errmsg);
1876 return;
1877 }
1878 }
1879
1880 /*
1881 * cleanup --
1882 * cleans up the global buffers
1883 */
1884 static void
1885 cleanup(mandb_rec *rec)
1886 {
1887 rec->desc.offset = 0;
1888 rec->lib.offset = 0;
1889 rec->return_vals.offset = 0;
1890 rec->env.offset = 0;
1891 rec->exit_status.offset = 0;
1892 rec->diagnostics.offset = 0;
1893 rec->errors.offset = 0;
1894 rec->files.offset = 0;
1895
1896 free(rec->machine);
1897 rec->machine = NULL;
1898
1899 free(rec->links);
1900 rec->links = NULL;
1901
1902 free(rec->file_path);
1903 rec->file_path = NULL;
1904
1905 free(rec->name);
1906 rec->name = NULL;
1907
1908 free(rec->name_desc);
1909 rec->name_desc = NULL;
1910
1911 free(rec->md5_hash);
1912 rec->md5_hash = NULL;
1913 }
1914
1915 /*
1916 * init_secbuffs--
1917 * Sets the value of buflen for all the sec_buff field of rec. And then
1918 * allocate memory to each sec_buff member of rec.
1919 */
1920 static void
1921 init_secbuffs(mandb_rec *rec)
1922 {
1923 /*
1924 * Some sec_buff might need more memory, for example desc,
1925 * which stores the data of the DESCRIPTION section,
1926 * while some might need very small amount of memory.
1927 * Therefore explicitly setting the value of buflen field for
1928 * each sec_buff.
1929 */
1930 rec->desc.buflen = 10 * BUFLEN;
1931 rec->desc.data = emalloc(rec->desc.buflen);
1932 rec->desc.offset = 0;
1933
1934 rec->lib.buflen = BUFLEN / 2;
1935 rec->lib.data = emalloc(rec->lib.buflen);
1936 rec->lib.offset = 0;
1937
1938 rec->return_vals.buflen = BUFLEN;
1939 rec->return_vals.data = emalloc(rec->return_vals.buflen);
1940 rec->return_vals.offset = 0;
1941
1942 rec->exit_status.buflen = BUFLEN;
1943 rec->exit_status.data = emalloc(rec->exit_status.buflen);
1944 rec->exit_status.offset = 0;
1945
1946 rec->env.buflen = BUFLEN;
1947 rec->env.data = emalloc(rec->env.buflen);
1948 rec->env.offset = 0;
1949
1950 rec->files.buflen = BUFLEN;
1951 rec->files.data = emalloc(rec->files.buflen);
1952 rec->files.offset = 0;
1953
1954 rec->diagnostics.buflen = BUFLEN;
1955 rec->diagnostics.data = emalloc(rec->diagnostics.buflen);
1956 rec->diagnostics.offset = 0;
1957
1958 rec->errors.buflen = BUFLEN;
1959 rec->errors.data = emalloc(rec->errors.buflen);
1960 rec->errors.offset = 0;
1961 }
1962
1963 /*
1964 * free_secbuffs--
1965 * This function should be called at the end, when all the pages have been
1966 * parsed.
1967 * It frees the memory allocated to sec_buffs by init_secbuffs in the starting.
1968 */
1969 static void
1970 free_secbuffs(mandb_rec *rec)
1971 {
1972 free(rec->desc.data);
1973 free(rec->lib.data);
1974 free(rec->return_vals.data);
1975 free(rec->exit_status.data);
1976 free(rec->env.data);
1977 free(rec->files.data);
1978 free(rec->diagnostics.data);
1979 free(rec->errors.data);
1980 }
1981
1982 static void
1983 replace_hyph(char *str)
1984 {
1985 char *iter = str;
1986 while ((iter = strchr(iter, ASCII_HYPH)) != NULL)
1987 *iter = '-';
1988 }
1989
1990 static char *
1991 parse_escape(const char *str)
1992 {
1993 const char *backslash, *last_backslash;
1994 char *result, *iter;
1995 size_t len;
1996
1997 assert(str);
1998
1999 last_backslash = str;
2000 backslash = strchr(str, '\\');
2001 if (backslash == NULL) {
2002 result = estrdup(str);
2003 replace_hyph(result);
2004 return result;
2005 }
2006
2007 result = emalloc(strlen(str) + 1);
2008 iter = result;
2009
2010 do {
2011 len = backslash - last_backslash;
2012 memcpy(iter, last_backslash, len);
2013 iter += len;
2014 if (backslash[1] == '-' || backslash[1] == ' ') {
2015 *iter++ = backslash[1];
2016 last_backslash = backslash + 2;
2017 backslash = strchr(backslash + 2, '\\');
2018 } else {
2019 ++backslash;
2020 mandoc_escape(&backslash, NULL, NULL);
2021 last_backslash = backslash;
2022 if (backslash == NULL)
2023 break;
2024 backslash = strchr(last_backslash, '\\');
2025 }
2026 } while (backslash != NULL);
2027 if (last_backslash != NULL)
2028 strcpy(iter, last_backslash);
2029
2030 replace_hyph(result);
2031 return result;
2032 }
2033
2034 /*
2035 * append--
2036 * Concatenates a space and src at the end of sbuff->data (much like concat in
2037 * apropos-utils.c).
2038 * Rather than reallocating space for writing data, it uses the value of the
2039 * offset field of sec_buff to write new data at the free space left in the
2040 * buffer.
2041 * In case the size of the data to be appended exceeds the number of bytes left
2042 * in the buffer, it reallocates buflen number of bytes and then continues.
2043 * Value of offset field should be adjusted as new data is written.
2044 *
2045 * NOTE: This function does not write the null byte at the end of the buffers,
2046 * write a null byte at the position pointed to by offset before inserting data
2047 * in the db.
2048 */
2049 static void
2050 append(secbuff *sbuff, const char *src)
2051 {
2052 short flag = 0;
2053 size_t srclen, newlen;
2054 char *temp;
2055
2056 assert(src != NULL);
2057 temp = parse_escape(src);
2058 srclen = strlen(temp);
2059
2060 if (sbuff->data == NULL) {
2061 sbuff->data = emalloc(sbuff->buflen);
2062 sbuff->offset = 0;
2063 }
2064
2065 newlen = sbuff->offset + srclen + 2;
2066 if (newlen >= sbuff->buflen) {
2067 while (sbuff->buflen < newlen)
2068 sbuff->buflen += sbuff->buflen;
2069 sbuff->data = erealloc(sbuff->data, sbuff->buflen);
2070 flag = 1;
2071 }
2072
2073 /* Append a space at the end of the buffer. */
2074 if (sbuff->offset || flag)
2075 sbuff->data[sbuff->offset++] = ' ';
2076 /* Now, copy src at the end of the buffer. */
2077 memcpy(sbuff->data + sbuff->offset, temp, srclen);
2078 sbuff->offset += srclen;
2079 free(temp);
2080 }
2081
2082 static void
2083 usage(void)
2084 {
2085 fprintf(stderr, "Usage: %s [-floQqv] [-C path]\n", getprogname());
2086 exit(1);
2087 }
2088