makemandb.c revision 1.42 1 /* $NetBSD: makemandb.c,v 1.42 2016/07/17 15:56:14 abhinav 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 1.42 2016/07/17 15:56:14 abhinav Exp $");
21
22 #include <sys/stat.h>
23 #include <sys/types.h>
24
25 #include <assert.h>
26 #include <dirent.h>
27 #include <err.h>
28 #include <archive.h>
29 #include <libgen.h>
30 #include <md5.h>
31 #include <stdio.h>
32 #include <stdlib.h>
33 #include <string.h>
34 #include <unistd.h>
35 #include <util.h>
36
37 #include "apropos-utils.h"
38 #include "dist/man.h"
39 #include "dist/mandoc.h"
40 #include "dist/mdoc.h"
41 #include "dist/roff.h"
42
43 #define BUFLEN 1024
44 #define MDOC 0 //If the page is of mdoc(7) type
45 #define MAN 1 //If the page is of man(7) type
46
47 /*
48 * A data structure for holding section specific data.
49 */
50 typedef struct secbuff {
51 char *data;
52 size_t buflen; //Total length of buffer allocated initially
53 size_t offset; // Current offset in the buffer.
54 } secbuff;
55
56 typedef struct makemandb_flags {
57 int optimize;
58 int limit; // limit the indexing to only NAME section
59 int recreate; // Database was created from scratch
60 int verbosity; // 0: quiet, 1: default, 2: verbose
61 } makemandb_flags;
62
63 typedef struct roff_mandb_rec {
64 /* Fields for mandb table */
65 char *name; // for storing the name of the man page
66 char *name_desc; // for storing the one line description (.Nd)
67 secbuff desc; // for storing the DESCRIPTION section
68 secbuff lib; // for the LIBRARY section
69 secbuff return_vals; // RETURN VALUES
70 secbuff env; // ENVIRONMENT
71 secbuff files; // FILES
72 secbuff exit_status; // EXIT STATUS
73 secbuff diagnostics; // DIAGNOSTICS
74 secbuff errors; // ERRORS
75 char *section;
76
77 int xr_found; // To track whether a .Xr was seen when parsing a section
78
79 /* Fields for mandb_meta table */
80 char *md5_hash;
81 dev_t device;
82 ino_t inode;
83 time_t mtime;
84
85 /* Fields for mandb_links table */
86 char *machine;
87 char *links; //all the links to a page in a space separated form
88 char *file_path;
89
90 /* Non-db fields */
91 int page_type; //Indicates the type of page: mdoc or man
92 } mandb_rec;
93
94 typedef void (*proff_nf)(const struct roff_node *n, mandb_rec *);
95
96 static void append(secbuff *sbuff, const char *src);
97 static void init_secbuffs(mandb_rec *);
98 static void free_secbuffs(mandb_rec *);
99 static int check_md5(const char *, sqlite3 *, const char *, char **, void *, size_t);
100 static void cleanup(mandb_rec *);
101 static void set_section(const struct roff_man *, mandb_rec *);
102 static void set_machine(const struct roff_man *, mandb_rec *);
103 static int insert_into_db(sqlite3 *, mandb_rec *);
104 static void begin_parse(const char *, struct mparse *, mandb_rec *,
105 const void *, size_t len);
106 static void proff_node(const struct roff_node *, mandb_rec *, const proff_nf *);
107 static void pmdoc_Nm(const struct roff_node *, mandb_rec *);
108 static void pmdoc_Nd(const struct roff_node *, mandb_rec *);
109 static void pmdoc_Sh(const struct roff_node *, mandb_rec *);
110 static void pmdoc_Xr(const struct roff_node *, mandb_rec *);
111 static void pmdoc_Pp(const struct roff_node *, mandb_rec *);
112 static void pmdoc_macro_handler(const struct roff_node *, mandb_rec *, int);
113 static void pman_parse_node(const struct roff_node *, secbuff *);
114 static void pman_parse_name(const struct roff_node *, mandb_rec *);
115 static void pman_sh(const struct roff_node *, mandb_rec *);
116 static void pman_block(const struct roff_node *, mandb_rec *);
117 static void traversedir(const char *, const char *, sqlite3 *, struct mparse *);
118 static void mdoc_parse_section(enum roff_sec, const char *, mandb_rec *);
119 static void man_parse_section(enum man_sec, const struct roff_node *, mandb_rec *);
120 static void build_file_cache(sqlite3 *, const char *, const char *,
121 struct stat *);
122 static void update_db(sqlite3 *, struct mparse *, mandb_rec *);
123 __dead static void usage(void);
124 static void optimize(sqlite3 *);
125 static char *parse_escape(const char *);
126 static void replace_hyph(char *);
127 static makemandb_flags mflags = { .verbosity = 1 };
128
129 static const proff_nf mdocs[MDOC_MAX + 1] = {
130 NULL, /* Ap */
131 NULL, /* Dd */
132 NULL, /* Dt */
133 NULL, /* Os */
134
135 pmdoc_Sh, /* Sh */
136 NULL, /* Ss */
137 pmdoc_Pp, /* Pp */
138 NULL, /* D1 */
139
140 NULL, /* Dl */
141 NULL, /* Bd */
142 NULL, /* Ed */
143 NULL, /* Bl */
144
145 NULL, /* El */
146 NULL, /* It */
147 NULL, /* Ad */
148 NULL, /* An */
149
150 NULL, /* Ar */
151 NULL, /* Cd */
152 NULL, /* Cm */
153 NULL, /* Dv */
154
155 NULL, /* Er */
156 NULL, /* Ev */
157 NULL, /* Ex */
158 NULL, /* Fa */
159
160 NULL, /* Fd */
161 NULL, /* Fl */
162 NULL, /* Fn */
163 NULL, /* Ft */
164
165 NULL, /* Ic */
166 NULL, /* In */
167 NULL, /* Li */
168 pmdoc_Nd, /* Nd */
169
170 pmdoc_Nm, /* Nm */
171 NULL, /* Op */
172 NULL, /* Ot */
173 NULL, /* Pa */
174
175 NULL, /* Rv */
176 NULL, /* St */
177 NULL, /* Va */
178 NULL, /* Vt */
179
180 pmdoc_Xr, /* Xr */
181 NULL, /* %A */
182 NULL, /* %B */
183 NULL, /* %D */
184
185 NULL, /* %I */
186 NULL, /* %J */
187 NULL, /* %N */
188 NULL, /* %O */
189
190 NULL, /* %P */
191 NULL, /* %R */
192 NULL, /* %T */
193 NULL, /* %V */
194
195 NULL, /* Ac */
196 NULL, /* Ao */
197 NULL, /* Aq */
198 NULL, /* At */
199
200 NULL, /* Bc */
201 NULL, /* Bf */
202 NULL, /* Bo */
203 NULL, /* Bq */
204
205 NULL, /* Bsx */
206 NULL, /* Bx */
207 NULL, /* Db */
208 NULL, /* Dc */
209
210 NULL, /* Do */
211 NULL, /* Dq */
212 NULL, /* Ec */
213 NULL, /* Ef */
214
215 NULL, /* Em */
216 NULL, /* Eo */
217 NULL, /* Fx */
218 NULL, /* Ms */
219
220 NULL, /* No */
221 NULL, /* Ns */
222 NULL, /* Nx */
223 NULL, /* Ox */
224
225 NULL, /* Pc */
226 NULL, /* Pf */
227 NULL, /* Po */
228 NULL, /* Pq */
229
230 NULL, /* Qc */
231 NULL, /* Ql */
232 NULL, /* Qo */
233 NULL, /* Qq */
234
235 NULL, /* Re */
236 NULL, /* Rs */
237 NULL, /* Sc */
238 NULL, /* So */
239
240 NULL, /* Sq */
241 NULL, /* Sm */
242 NULL, /* Sx */
243 NULL, /* Sy */
244
245 NULL, /* Tn */
246 NULL, /* Ux */
247 NULL, /* Xc */
248 NULL, /* Xo */
249
250 NULL, /* Fo */
251 NULL, /* Fc */
252 NULL, /* Oo */
253 NULL, /* Oc */
254
255 NULL, /* Bk */
256 NULL, /* Ek */
257 NULL, /* Bt */
258 NULL, /* Hf */
259
260 NULL, /* Fr */
261 NULL, /* Ud */
262 NULL, /* Lb */
263 NULL, /* Lp */
264
265 NULL, /* Lk */
266 NULL, /* Mt */
267 NULL, /* Brq */
268 NULL, /* Bro */
269
270 NULL, /* Brc */
271 NULL, /* %C */
272 NULL, /* Es */
273 NULL, /* En */
274
275 NULL, /* Dx */
276 NULL, /* %Q */
277 NULL, /* br */
278 NULL, /* sp */
279
280 NULL, /* %U */
281 NULL, /* Ta */
282 NULL, /* ll */
283 NULL, /* text */
284 };
285
286 static const proff_nf mans[MAN_MAX] = {
287 NULL, //br
288 NULL, //TH
289 pman_sh, //SH
290 NULL, //SS
291 NULL, //TP
292 NULL, //LP
293 NULL, //PP
294 NULL, //P
295 NULL, //IP
296 NULL, //HP
297 NULL, //SM
298 NULL, //SB
299 NULL, //BI
300 NULL, //IB
301 NULL, //BR
302 NULL, //RB
303 NULL, //R
304 pman_block, //B
305 NULL, //I
306 NULL, //IR
307 NULL, //RI
308 NULL, //sp
309 NULL, //nf
310 NULL, //fi
311 NULL, //RE
312 NULL, //RS
313 NULL, //DT
314 NULL, //UC
315 NULL, //PD
316 NULL, //AT
317 NULL, //in
318 NULL, //ft
319 NULL, //OP
320 NULL, //EX
321 NULL, //EE
322 NULL, //UR
323 NULL, //UE
324 NULL, //ll
325 };
326
327
328 int
329 main(int argc, char *argv[])
330 {
331 FILE *file;
332 const char *sqlstr, *manconf = NULL;
333 char *line, *command, *parent;
334 char *errmsg;
335 int ch;
336 struct mparse *mp;
337 sqlite3 *db;
338 ssize_t len;
339 size_t linesize;
340 struct roff_mandb_rec rec;
341
342 while ((ch = getopt(argc, argv, "C:floQqv")) != -1) {
343 switch (ch) {
344 case 'C':
345 manconf = optarg;
346 break;
347 case 'f':
348 mflags.recreate = 1;
349 break;
350 case 'l':
351 mflags.limit = 1;
352 break;
353 case 'o':
354 mflags.optimize = 1;
355 break;
356 case 'Q':
357 mflags.verbosity = 0;
358 break;
359 case 'q':
360 mflags.verbosity = 1;
361 break;
362 case 'v':
363 mflags.verbosity = 2;
364 break;
365 default:
366 usage();
367 }
368 }
369
370 memset(&rec, 0, sizeof(rec));
371
372 init_secbuffs(&rec);
373 mchars_alloc();
374 mp = mparse_alloc(0, MANDOCLEVEL_BADARG, NULL, NULL);
375
376 if (manconf) {
377 char *arg;
378 size_t command_len = shquote(manconf, NULL, 0) + 1;
379 arg = emalloc(command_len);
380 shquote(manconf, arg, command_len);
381 easprintf(&command, "man -p -C %s", arg);
382 free(arg);
383 } else {
384 command = estrdup("man -p");
385 manconf = MANCONF;
386 }
387
388 if (mflags.recreate) {
389 char *dbp = get_dbpath(manconf);
390 /* No error here, it will fail in init_db in the same call */
391 if (dbp != NULL)
392 remove(dbp);
393 }
394
395 if ((db = init_db(MANDB_CREATE, manconf)) == NULL)
396 exit(EXIT_FAILURE);
397
398 sqlite3_exec(db, "PRAGMA synchronous = 0", NULL, NULL, &errmsg);
399 if (errmsg != NULL) {
400 warnx("%s", errmsg);
401 free(errmsg);
402 close_db(db);
403 exit(EXIT_FAILURE);
404 }
405
406 sqlite3_exec(db, "ATTACH DATABASE \':memory:\' AS metadb", NULL, NULL,
407 &errmsg);
408 if (errmsg != NULL) {
409 warnx("%s", errmsg);
410 free(errmsg);
411 close_db(db);
412 exit(EXIT_FAILURE);
413 }
414
415
416 /* Call man -p to get the list of man page dirs */
417 if ((file = popen(command, "r")) == NULL) {
418 close_db(db);
419 err(EXIT_FAILURE, "fopen failed");
420 }
421 free(command);
422
423 /* Begin the transaction for indexing the pages */
424 sqlite3_exec(db, "BEGIN", NULL, NULL, &errmsg);
425 if (errmsg != NULL) {
426 warnx("%s", errmsg);
427 free(errmsg);
428 close_db(db);
429 exit(EXIT_FAILURE);
430 }
431
432 sqlstr = "CREATE TABLE metadb.file_cache(device, inode, mtime, parent,"
433 " file PRIMARY KEY);"
434 "CREATE UNIQUE INDEX metadb.index_file_cache_dev"
435 " ON file_cache (device, inode)";
436
437 sqlite3_exec(db, sqlstr, NULL, NULL, &errmsg);
438 if (errmsg != NULL) {
439 warnx("%s", errmsg);
440 free(errmsg);
441 close_db(db);
442 exit(EXIT_FAILURE);
443 }
444
445 if (mflags.verbosity)
446 printf("Building temporary file cache\n");
447 line = NULL;
448 linesize = 0;
449 while ((len = getline(&line, &linesize, file)) != -1) {
450 /* Replace the new line character at the end of string with '\0' */
451 line[len - 1] = '\0';
452 parent = estrdup(line);
453 char *pdir = estrdup(dirname(parent));
454 free(parent);
455 /* Traverse the man page directories and parse the pages */
456 traversedir(pdir, line, db, mp);
457 free(pdir);
458 }
459 free(line);
460
461 if (pclose(file) == -1) {
462 close_db(db);
463 cleanup(&rec);
464 free_secbuffs(&rec);
465 err(EXIT_FAILURE, "pclose error");
466 }
467
468 if (mflags.verbosity)
469 printf("Performing index update\n");
470 update_db(db, mp, &rec);
471 mparse_free(mp);
472 mchars_free();
473 free_secbuffs(&rec);
474
475 /* Commit the transaction */
476 sqlite3_exec(db, "COMMIT", NULL, NULL, &errmsg);
477 if (errmsg != NULL) {
478 warnx("%s", errmsg);
479 free(errmsg);
480 close_db(db);
481 exit(EXIT_FAILURE);
482 }
483
484 if (mflags.optimize)
485 optimize(db);
486
487 close_db(db);
488 return 0;
489 }
490
491 /*
492 * traversedir --
493 * Traverses the given directory recursively and passes all the man page files
494 * in the way to build_file_cache()
495 */
496 static void
497 traversedir(const char *parent, const char *file, sqlite3 *db,
498 struct mparse *mp)
499 {
500 struct stat sb;
501 struct dirent *dirp;
502 DIR *dp;
503 char *buf;
504
505 if (stat(file, &sb) < 0) {
506 if (mflags.verbosity)
507 warn("stat failed: %s", file);
508 return;
509 }
510
511 /* If it is a directory, traverse it recursively */
512 if (S_ISDIR(sb.st_mode)) {
513 if ((dp = opendir(file)) == NULL) {
514 if (mflags.verbosity)
515 warn("opendir error: %s", file);
516 return;
517 }
518
519 while ((dirp = readdir(dp)) != NULL) {
520 /* Avoid . and .. entries in a directory */
521 if (strncmp(dirp->d_name, ".", 1)) {
522 easprintf(&buf, "%s/%s", file, dirp->d_name);
523 traversedir(parent, buf, db, mp);
524 free(buf);
525 }
526 }
527 closedir(dp);
528 }
529
530 if (!S_ISREG(sb.st_mode) && !S_ISLNK(sb.st_mode))
531 return;
532
533 if (sb.st_size == 0) {
534 if (mflags.verbosity)
535 warnx("Empty file: %s", file);
536 return;
537 }
538 build_file_cache(db, parent, file, &sb);
539 }
540
541 /* build_file_cache --
542 * This function generates an md5 hash of the file passed as its 2nd parameter
543 * and stores it in a temporary table file_cache along with the full file path.
544 * This is done to support incremental updation of the database.
545 * The temporary table file_cache is dropped thereafter in the function
546 * update_db(), once the database has been updated.
547 */
548 static void
549 build_file_cache(sqlite3 *db, const char *parent, const char *file,
550 struct stat *sb)
551 {
552 const char *sqlstr;
553 sqlite3_stmt *stmt = NULL;
554 int rc, idx;
555 assert(file != NULL);
556 dev_t device_cache = sb->st_dev;
557 ino_t inode_cache = sb->st_ino;
558 time_t mtime_cache = sb->st_mtime;
559
560 sqlstr = "INSERT INTO metadb.file_cache VALUES (:device, :inode,"
561 " :mtime, :parent, :file)";
562 rc = sqlite3_prepare_v2(db, sqlstr, -1, &stmt, NULL);
563 if (rc != SQLITE_OK) {
564 if (mflags.verbosity)
565 warnx("%s", sqlite3_errmsg(db));
566 return;
567 }
568
569 idx = sqlite3_bind_parameter_index(stmt, ":device");
570 rc = sqlite3_bind_int64(stmt, idx, device_cache);
571 if (rc != SQLITE_OK) {
572 if (mflags.verbosity)
573 warnx("%s", sqlite3_errmsg(db));
574 sqlite3_finalize(stmt);
575 return;
576 }
577
578 idx = sqlite3_bind_parameter_index(stmt, ":inode");
579 rc = sqlite3_bind_int64(stmt, idx, inode_cache);
580 if (rc != SQLITE_OK) {
581 if (mflags.verbosity)
582 warnx("%s", sqlite3_errmsg(db));
583 sqlite3_finalize(stmt);
584 return;
585 }
586
587 idx = sqlite3_bind_parameter_index(stmt, ":mtime");
588 rc = sqlite3_bind_int64(stmt, idx, mtime_cache);
589 if (rc != SQLITE_OK) {
590 if (mflags.verbosity)
591 warnx("%s", sqlite3_errmsg(db));
592 sqlite3_finalize(stmt);
593 return;
594 }
595
596 idx = sqlite3_bind_parameter_index(stmt, ":parent");
597 rc = sqlite3_bind_text(stmt, idx, parent, -1, NULL);
598 if (rc != SQLITE_OK) {
599 if (mflags.verbosity)
600 warnx("%s", sqlite3_errmsg(db));
601 sqlite3_finalize(stmt);
602 return;
603 }
604
605 idx = sqlite3_bind_parameter_index(stmt, ":file");
606 rc = sqlite3_bind_text(stmt, idx, file, -1, NULL);
607 if (rc != SQLITE_OK) {
608 if (mflags.verbosity)
609 warnx("%s", sqlite3_errmsg(db));
610 sqlite3_finalize(stmt);
611 return;
612 }
613
614 sqlite3_step(stmt);
615 sqlite3_finalize(stmt);
616 }
617
618 static void
619 update_existing_entry(sqlite3 *db, const char *file, const char *hash,
620 mandb_rec *rec, int *new_count, int *link_count, int *err_count)
621 {
622 int update_count, rc, idx;
623 const char *inner_sqlstr;
624 sqlite3_stmt *inner_stmt;
625
626 update_count = sqlite3_total_changes(db);
627 inner_sqlstr = "UPDATE mandb_meta SET device = :device,"
628 " inode = :inode, mtime = :mtime WHERE"
629 " md5_hash = :md5 AND file = :file AND"
630 " (device <> :device2 OR inode <> "
631 " :inode2 OR mtime <> :mtime2)";
632 rc = sqlite3_prepare_v2(db, inner_sqlstr, -1, &inner_stmt, NULL);
633 if (rc != SQLITE_OK) {
634 if (mflags.verbosity)
635 warnx("%s", sqlite3_errmsg(db));
636 return;
637 }
638 idx = sqlite3_bind_parameter_index(inner_stmt, ":device");
639 sqlite3_bind_int64(inner_stmt, idx, rec->device);
640 idx = sqlite3_bind_parameter_index(inner_stmt, ":inode");
641 sqlite3_bind_int64(inner_stmt, idx, rec->inode);
642 idx = sqlite3_bind_parameter_index(inner_stmt, ":mtime");
643 sqlite3_bind_int64(inner_stmt, idx, rec->mtime);
644 idx = sqlite3_bind_parameter_index(inner_stmt, ":md5");
645 sqlite3_bind_text(inner_stmt, idx, hash, -1, NULL);
646 idx = sqlite3_bind_parameter_index(inner_stmt, ":file");
647 sqlite3_bind_text(inner_stmt, idx, file, -1, NULL);
648 idx = sqlite3_bind_parameter_index(inner_stmt, ":device2");
649 sqlite3_bind_int64(inner_stmt, idx, rec->device);
650 idx = sqlite3_bind_parameter_index(inner_stmt, ":inode2");
651 sqlite3_bind_int64(inner_stmt, idx, rec->inode);
652 idx = sqlite3_bind_parameter_index(inner_stmt, ":mtime2");
653 sqlite3_bind_int64(inner_stmt, idx, rec->mtime);
654
655 rc = sqlite3_step(inner_stmt);
656 if (rc == SQLITE_DONE) {
657 /* Check if an update has been performed. */
658 if (update_count != sqlite3_total_changes(db)) {
659 if (mflags.verbosity == 2)
660 printf("Updated %s\n", file);
661 (*new_count)++;
662 } else {
663 /* Otherwise it was a hardlink. */
664 (*link_count)++;
665 }
666 } else {
667 if (mflags.verbosity == 2)
668 warnx("Could not update the meta data for %s", file);
669 (*err_count)++;
670 }
671 sqlite3_finalize(inner_stmt);
672 }
673
674 /* read_and_decompress --
675 * Reads the given file into memory. If it is compressed, decompress
676 * it before returning to the caller.
677 */
678 static int
679 read_and_decompress(const char *file, void **bufp, size_t *len)
680 {
681 size_t off;
682 ssize_t r;
683 struct archive *a;
684 struct archive_entry *ae;
685 char *buf;
686
687 if ((a = archive_read_new()) == NULL)
688 errx(EXIT_FAILURE, "memory allocation failed");
689
690 *bufp = NULL;
691 if (archive_read_support_compression_all(a) != ARCHIVE_OK ||
692 archive_read_support_format_raw(a) != ARCHIVE_OK ||
693 archive_read_open_filename(a, file, 65536) != ARCHIVE_OK ||
694 archive_read_next_header(a, &ae) != ARCHIVE_OK)
695 goto archive_error;
696 *len = 65536;
697 buf = emalloc(*len);
698 off = 0;
699 for (;;) {
700 r = archive_read_data(a, buf + off, *len - off);
701 if (r == ARCHIVE_OK) {
702 archive_read_finish(a);
703 *bufp = buf;
704 *len = off;
705 return 0;
706 }
707 if (r <= 0) {
708 free(buf);
709 break;
710 }
711 off += r;
712 if (off == *len) {
713 *len *= 2;
714 if (*len < off) {
715 if (mflags.verbosity)
716 warnx("File too large: %s", file);
717 free(buf);
718 archive_read_finish(a);
719 return -1;
720 }
721 buf = erealloc(buf, *len);
722 }
723 }
724
725 archive_error:
726 warnx("Error while reading `%s': %s", file, archive_error_string(a));
727 archive_read_finish(a);
728 return -1;
729 }
730
731 /* update_db --
732 * Does an incremental updation of the database by checking the file_cache.
733 * It parses and adds the pages which are present in file_cache,
734 * but not in the database.
735 * It also removes the pages which are present in the databse,
736 * but not in the file_cache.
737 */
738 static void
739 update_db(sqlite3 *db, struct mparse *mp, mandb_rec *rec)
740 {
741 const char *sqlstr;
742 sqlite3_stmt *stmt = NULL;
743 char *file;
744 char *parent;
745 char *errmsg = NULL;
746 char *md5sum;
747 void *buf;
748 size_t buflen;
749 struct sql_row {
750 struct sql_row *next;
751 dev_t device;
752 ino_t inode;
753 time_t mtime;
754 char *parent;
755 char *file;
756 } *rows, *row;
757 int new_count = 0; /* Counter for newly indexed/updated pages */
758 int total_count = 0; /* Counter for total number of pages */
759 int err_count = 0; /* Counter for number of failed pages */
760 int link_count = 0; /* Counter for number of hard/sym links */
761 int md5_status;
762 int rc;
763
764 sqlstr = "SELECT device, inode, mtime, parent, file"
765 " FROM metadb.file_cache fc"
766 " WHERE NOT EXISTS(SELECT 1 FROM mandb_meta WHERE"
767 " device = fc.device AND inode = fc.inode AND "
768 " mtime = fc.mtime AND file = fc.file)";
769
770 rc = sqlite3_prepare_v2(db, sqlstr, -1, &stmt, NULL);
771 if (rc != SQLITE_OK) {
772 if (mflags.verbosity)
773 warnx("%s", sqlite3_errmsg(db));
774 close_db(db);
775 errx(EXIT_FAILURE, "Could not query file cache");
776 }
777
778 buf = NULL;
779 rows = NULL;
780 while (sqlite3_step(stmt) == SQLITE_ROW) {
781 row = emalloc(sizeof(struct sql_row));
782 row->device = sqlite3_column_int64(stmt, 0);
783 row->inode = sqlite3_column_int64(stmt, 1);
784 row->mtime = sqlite3_column_int64(stmt, 2);
785 row->parent = estrdup((const char *) sqlite3_column_text(stmt, 3));
786 row->file = estrdup((const char *) sqlite3_column_text(stmt, 4));
787 row->next = rows;
788 rows = row;
789 total_count++;
790 }
791 sqlite3_finalize(stmt);
792
793 for ( ; rows != NULL; free(parent), free(file), free(buf)) {
794 row = rows;
795 rows = rows->next;
796
797 rec->device = row->device;
798 rec->inode = row->inode;
799 rec->mtime = row->mtime;
800 parent = row->parent;
801 file = row->file;
802 free(row);
803
804 if (read_and_decompress(file, &buf, &buflen)) {
805 err_count++;
806 continue;
807 }
808 md5_status = check_md5(file, db, "mandb_meta", &md5sum, buf,
809 buflen);
810 assert(md5sum != NULL);
811 if (md5_status == -1) {
812 if (mflags.verbosity)
813 warnx("An error occurred in checking md5 value"
814 " for file %s", file);
815 err_count++;
816 continue;
817 }
818
819 if (md5_status == 0) {
820 /*
821 * The MD5 hash is already present in the database,
822 * so simply update the metadata, ignoring symlinks.
823 */
824 struct stat sb;
825 stat(file, &sb);
826 if (S_ISLNK(sb.st_mode)) {
827 free(md5sum);
828 link_count++;
829 continue;
830 }
831 update_existing_entry(db, file, md5sum, rec,
832 &new_count, &link_count, &err_count);
833 free(md5sum);
834 continue;
835 }
836
837 if (md5_status == 1) {
838 /*
839 * The MD5 hash was not present in the database.
840 * This means is either a new file or an updated file.
841 * We should go ahead with parsing.
842 */
843 if (chdir(parent) == -1) {
844 if (mflags.verbosity)
845 warn("chdir failed for `%s', could "
846 "not index `%s'", parent, file);
847 err_count++;
848 free(md5sum);
849 continue;
850 }
851
852 if (mflags.verbosity == 2)
853 printf("Parsing: %s\n", file);
854 rec->md5_hash = md5sum;
855 rec->file_path = estrdup(file);
856 // file_path is freed by insert_into_db itself.
857 begin_parse(file, mp, rec, buf, buflen);
858 if (insert_into_db(db, rec) < 0) {
859 if (mflags.verbosity)
860 warnx("Error in indexing `%s'", file);
861 err_count++;
862 } else {
863 new_count++;
864 }
865 }
866 }
867
868 if (mflags.verbosity == 2) {
869 printf("Total Number of new or updated pages encountered = %d\n"
870 "Total number of (hard or symbolic) links found = %d\n"
871 "Total number of pages that were successfully"
872 " indexed/updated = %d\n"
873 "Total number of pages that could not be indexed"
874 " due to errors = %d\n",
875 total_count - link_count, link_count, new_count, err_count);
876 }
877
878 if (mflags.recreate)
879 return;
880
881 if (mflags.verbosity == 2)
882 printf("Deleting stale index entries\n");
883
884 sqlstr = "DELETE FROM mandb_meta WHERE file NOT IN"
885 " (SELECT file FROM metadb.file_cache);"
886 "DELETE FROM mandb_links WHERE md5_hash NOT IN"
887 " (SELECT md5_hash from mandb_meta);"
888 "DROP TABLE metadb.file_cache;"
889 "DELETE FROM mandb WHERE rowid NOT IN"
890 " (SELECT id FROM mandb_meta);";
891
892 sqlite3_exec(db, sqlstr, NULL, NULL, &errmsg);
893 if (errmsg != NULL) {
894 warnx("Removing old entries failed: %s", errmsg);
895 warnx("Please rebuild database from scratch with -f.");
896 free(errmsg);
897 return;
898 }
899 }
900
901 /*
902 * begin_parse --
903 * parses the man page using libmandoc
904 */
905 static void
906 begin_parse(const char *file, struct mparse *mp, mandb_rec *rec,
907 const void *buf, size_t len)
908 {
909 struct roff_man *roff;
910 mparse_reset(mp);
911
912 rec->xr_found = 0;
913
914 if (mparse_readmem(mp, buf, len, file) >= MANDOCLEVEL_BADARG) {
915 /* Printing this warning at verbosity level 2
916 * because some packages from pkgsrc might trigger several
917 * of such warnings.
918 */
919 if (mflags.verbosity == 2)
920 warnx("%s: Parse failure", file);
921 return;
922 }
923
924 mparse_result(mp, &roff, NULL);
925 if (roff == NULL) {
926 if (mflags.verbosity == 2)
927 warnx("Not a roff(7) page");
928 return;
929 }
930
931 if (roff->macroset == MACROSET_MDOC) {
932 mdoc_validate(roff);
933 rec->page_type = MDOC;
934 proff_node(roff->first->child, rec, mdocs);
935 } else if (roff->macroset == MACROSET_MAN) {
936 man_validate(roff);
937 rec->page_type = MAN;
938 proff_node(roff->first->child, rec, mans);
939 } else
940 warnx("Unknown macroset %d", roff->macroset);
941 set_machine(roff, rec);
942 set_section(roff, rec);
943 }
944
945 /*
946 * set_section --
947 * Extracts the section number and normalizes it to only the numeric part
948 * (Which should be the first character of the string).
949 */
950 static void
951 set_section(const struct roff_man *rm, mandb_rec *rec)
952 {
953 if (!rm)
954 return;
955 const struct roff_meta *rm_meta = &rm->meta;
956 const char *s = rm_meta->msec == NULL ? "?" : rm_meta->msec;
957 easprintf(&rec->section, "%s", s);
958 if (rec->section[0] == '?' && mflags.verbosity == 2)
959 warnx("%s: Missing section number", rec->file_path);
960 }
961
962 /*
963 * get_machine --
964 * Extracts the machine architecture information if available.
965 */
966 static void
967 set_machine(const struct roff_man *rm, mandb_rec *rec)
968 {
969 if (rm == NULL)
970 return;
971 const struct roff_meta *rm_meta = &rm->meta;
972 if (rm_meta->arch)
973 rec->machine = estrdup(rm_meta->arch);
974 }
975
976 /*
977 * pmdoc_Nm --
978 * Extracts the Name of the manual page from the .Nm macro
979 */
980 static void
981 pmdoc_Nm(const struct roff_node *n, mandb_rec *rec)
982 {
983 if (n->sec != SEC_NAME)
984 return;
985
986 for (n = n->child; n; n = n->next) {
987 if (n->type == ROFFT_TEXT) {
988 char *escaped_name = parse_escape(n->string);
989 concat(&rec->name, escaped_name);
990 free(escaped_name);
991 }
992 }
993 }
994
995 /*
996 * pmdoc_Nd --
997 * Extracts the one line description of the man page from the .Nd macro
998 */
999 static void
1000 pmdoc_Nd(const struct roff_node *n, mandb_rec *rec)
1001 {
1002 if (n->type == ROFFT_BODY)
1003 deroff(&rec->name_desc, n);
1004
1005 }
1006
1007 /*
1008 * pmdoc_macro_handler--
1009 * This function is a single point of handling all the special macros that we
1010 * want to handle especially. For example the .Xr macro for properly parsing
1011 * the referenced page name along with the section number, or the .Pp macro
1012 * for adding a new line whenever we encounter it.
1013 */
1014 static void
1015 pmdoc_macro_handler(const struct roff_node *n, mandb_rec *rec, int doct)
1016 {
1017 const struct roff_node *sn;
1018 assert(n);
1019
1020 switch (doct) {
1021 /* Parse the man page references.
1022 * Basically the .Xr macros are used like:
1023 * .Xr ls 1
1024 * and formatted like this:
1025 * ls(1)
1026 * Prepare a buffer to format the data like the above example and call
1027 * pmdoc_parse_section to append it.
1028 */
1029 case MDOC_Xr:
1030 n = n->child;
1031 while (n->type != ROFFT_TEXT && n->next)
1032 n = n->next;
1033
1034 if (n && n->type != ROFFT_TEXT)
1035 return;
1036 sn = n;
1037 if (n->next)
1038 n = n->next;
1039
1040 while (n->type != ROFFT_TEXT && n->next)
1041 n = n->next;
1042
1043 if (n && n->type == ROFFT_TEXT) {
1044 char *buf;
1045 easprintf(&buf, "%s(%s)", sn->string, n->string);
1046 mdoc_parse_section(n->sec, buf, rec);
1047 free(buf);
1048 }
1049
1050 break;
1051
1052 /* Parse the .Pp macro to add a new line */
1053 case MDOC_Pp:
1054 if (n->type == ROFFT_TEXT)
1055 mdoc_parse_section(n->sec, "\n", rec);
1056 break;
1057 default:
1058 break;
1059 }
1060
1061 }
1062
1063 /*
1064 * pmdoc_Xr, pmdoc_Pp--
1065 * Empty stubs.
1066 * The parser calls these functions each time it encounters
1067 * a .Xr or .Pp macro. We are parsing all the data from
1068 * the pmdoc_Sh function, so don't do anything here.
1069 * (See if else blocks in pmdoc_Sh.)
1070 */
1071 static void
1072 pmdoc_Xr(const struct roff_node *n, mandb_rec *rec)
1073 {
1074 }
1075
1076 static void
1077 pmdoc_Pp(const struct roff_node *n, mandb_rec *rec)
1078 {
1079 }
1080
1081 /*
1082 * pmdoc_Sh --
1083 * Called when a .Sh macro is encountered and loops through its body, calling
1084 * mdoc_parse_section to append the data to the section specific buffer.
1085 * Two special macros which may occur inside the body of Sh are .Nm and .Xr and
1086 * they need special handling, thus the separate if branches for them.
1087 */
1088 static void
1089 pmdoc_Sh(const struct roff_node *n, mandb_rec *rec)
1090 {
1091 if (n == NULL || (n->type != ROFFT_TEXT && n->tok == MDOC_MAX))
1092 return;
1093 int xr_found = 0;
1094
1095 if (n->type == ROFFT_TEXT) {
1096 mdoc_parse_section(n->sec, n->string, rec);
1097 } else if (mdocs[n->tok] == pmdoc_Nm && rec->name != NULL) {
1098 /*
1099 * When encountering a .Nm macro, substitute it
1100 * with its previously cached value of the argument.
1101 */
1102 mdoc_parse_section(n->sec, rec->name, rec);
1103 } else if (mdocs[n->tok] == pmdoc_Xr) {
1104 /*
1105 * When encountering other inline macros,
1106 * call pmdoc_macro_handler.
1107 */
1108 pmdoc_macro_handler(n, rec, MDOC_Xr);
1109 xr_found = 1;
1110 } else if (mdocs[n->tok] == pmdoc_Pp) {
1111 pmdoc_macro_handler(n, rec, MDOC_Pp);
1112 }
1113
1114 /*
1115 * If an Xr macro was encountered then the child node has
1116 * already been explored by pmdoc_macro_handler.
1117 */
1118 if (xr_found == 0)
1119 pmdoc_Sh(n->child, rec);
1120 pmdoc_Sh(n->next, rec);
1121 }
1122
1123 /*
1124 * mdoc_parse_section--
1125 * Utility function for parsing sections of the mdoc type pages.
1126 * Takes two params:
1127 * 1. sec is an enum which indicates the section in which we are present
1128 * 2. string is the string which we need to append to the secbuff for this
1129 * particular section.
1130 * The function appends string to the global section buffer and returns.
1131 */
1132 static void
1133 mdoc_parse_section(enum roff_sec sec, const char *string, mandb_rec *rec)
1134 {
1135 /*
1136 * If the user specified the 'l' flag, then parse and store only the
1137 * NAME section. Ignore the rest.
1138 */
1139 if (mflags.limit)
1140 return;
1141
1142 switch (sec) {
1143 case SEC_LIBRARY:
1144 append(&rec->lib, string);
1145 break;
1146 case SEC_RETURN_VALUES:
1147 append(&rec->return_vals, string);
1148 break;
1149 case SEC_ENVIRONMENT:
1150 append(&rec->env, string);
1151 break;
1152 case SEC_FILES:
1153 append(&rec->files, string);
1154 break;
1155 case SEC_EXIT_STATUS:
1156 append(&rec->exit_status, string);
1157 break;
1158 case SEC_DIAGNOSTICS:
1159 append(&rec->diagnostics, string);
1160 break;
1161 case SEC_ERRORS:
1162 append(&rec->errors, string);
1163 break;
1164 case SEC_NAME:
1165 case SEC_SYNOPSIS:
1166 case SEC_EXAMPLES:
1167 case SEC_STANDARDS:
1168 case SEC_HISTORY:
1169 case SEC_AUTHORS:
1170 case SEC_BUGS:
1171 break;
1172 default:
1173 append(&rec->desc, string);
1174 break;
1175 }
1176 }
1177
1178 static void
1179 proff_node(const struct roff_node *n, mandb_rec *rec, const proff_nf *func)
1180 {
1181 if (n == NULL)
1182 return;
1183
1184 switch (n->type) {
1185 case (ROFFT_BODY):
1186 /* FALLTHROUGH */
1187 case (ROFFT_BLOCK):
1188 /* FALLTHROUGH */
1189 case (ROFFT_ELEM):
1190 if (func[n->tok] != NULL)
1191 (*func[n->tok])(n, rec);
1192 break;
1193 default:
1194 break;
1195 }
1196
1197 proff_node(n->child, rec, func);
1198 proff_node(n->next, rec, func);
1199 }
1200
1201 /*
1202 * pman_parse_name --
1203 * Parses the NAME section and puts the complete content in the name_desc
1204 * variable.
1205 */
1206 static void
1207 pman_parse_name(const struct roff_node *n, mandb_rec *rec)
1208 {
1209 if (n == NULL)
1210 return;
1211
1212 if (n->type == ROFFT_TEXT) {
1213 char *tmp = parse_escape(n->string);
1214 concat(&rec->name_desc, tmp);
1215 free(tmp);
1216 }
1217
1218 if (n->child)
1219 pman_parse_name(n->child, rec);
1220
1221 if(n->next)
1222 pman_parse_name(n->next, rec);
1223 }
1224
1225 /*
1226 * A stub function to be able to parse the macros like .B embedded inside
1227 * a section.
1228 */
1229 static void
1230 pman_block(const struct roff_node *n, mandb_rec *rec)
1231 {
1232 }
1233
1234 /*
1235 * pman_sh --
1236 * This function does one of the two things:
1237 * 1. If the present section is NAME, then it will:
1238 * (a) Extract the name of the page (in case of multiple comma separated
1239 * names, it will pick up the first one).
1240 * (b) Build a space spearated list of all the symlinks/hardlinks to
1241 * this page and store in the buffer 'links'. These are extracted from
1242 * the comma separated list of names in the NAME section as well.
1243 * (c) Move on to the one line description section, which is after the list
1244 * of names in the NAME section.
1245 * 2. Otherwise, it will check the section name and call the man_parse_section
1246 * function, passing the enum corresponding to that section.
1247 */
1248 static void
1249 pman_sh(const struct roff_node *n, mandb_rec *rec)
1250 {
1251 static const struct {
1252 enum man_sec section;
1253 const char *header;
1254 } mapping[] = {
1255 { MANSEC_DESCRIPTION, "DESCRIPTION" },
1256 { MANSEC_SYNOPSIS, "SYNOPSIS" },
1257 { MANSEC_LIBRARY, "LIBRARY" },
1258 { MANSEC_ERRORS, "ERRORS" },
1259 { MANSEC_FILES, "FILES" },
1260 { MANSEC_RETURN_VALUES, "RETURN VALUE" },
1261 { MANSEC_RETURN_VALUES, "RETURN VALUES" },
1262 { MANSEC_EXIT_STATUS, "EXIT STATUS" },
1263 { MANSEC_EXAMPLES, "EXAMPLES" },
1264 { MANSEC_EXAMPLES, "EXAMPLE" },
1265 { MANSEC_STANDARDS, "STANDARDS" },
1266 { MANSEC_HISTORY, "HISTORY" },
1267 { MANSEC_BUGS, "BUGS" },
1268 { MANSEC_AUTHORS, "AUTHORS" },
1269 { MANSEC_COPYRIGHT, "COPYRIGHT" },
1270 };
1271 const struct roff_node *head;
1272 char *name_desc;
1273 size_t sz;
1274 size_t i;
1275
1276 if ((head = n->parent->head) == NULL || (head = head->child) == NULL ||
1277 head->type != ROFFT_TEXT)
1278 return;
1279
1280 /*
1281 * Check if this section should be extracted and
1282 * where it should be stored. Handled the trival cases first.
1283 */
1284 for (i = 0; i < sizeof(mapping) / sizeof(mapping[0]); ++i) {
1285 if (strcmp(head->string, mapping[i].header) == 0) {
1286 man_parse_section(mapping[i].section, n, rec);
1287 return;
1288 }
1289 }
1290
1291 if (strcmp(head->string, "NAME") == 0) {
1292 /*
1293 * We are in the NAME section.
1294 * pman_parse_name will put the complete content in name_desc.
1295 */
1296 pman_parse_name(n, rec);
1297
1298 name_desc = rec->name_desc;
1299 if (name_desc == NULL)
1300 return;
1301
1302 /* Remove any leading spaces. */
1303 while (name_desc[0] == ' ')
1304 name_desc++;
1305
1306 /* If the line begins with a "\&", avoid those */
1307 if (name_desc[0] == '\\' && name_desc[1] == '&')
1308 name_desc += 2;
1309
1310 /* Now name_desc should be left with a comma-space
1311 * separated list of names and the one line description
1312 * of the page:
1313 * "a, b, c \- sample description"
1314 * Take out the first name, before the first comma
1315 * (or space) and store it in rec->name.
1316 * If the page has aliases then they should be
1317 * in the form of a comma separated list.
1318 * Keep looping while there is a comma in name_desc,
1319 * extract the alias name and store in rec->links.
1320 * When there are no more commas left, break out.
1321 */
1322 int has_alias = 0; // Any more aliases left?
1323 while (*name_desc) {
1324 /* Remove any leading spaces or hyphens. */
1325 if (name_desc[0] == ' ' || name_desc[0] == '-') {
1326 name_desc++;
1327 continue;
1328 }
1329 sz = strcspn(name_desc, ", ");
1330
1331 /* Extract the first term and store it in rec->name. */
1332 if (rec->name == NULL) {
1333 if (name_desc[sz] == ',')
1334 has_alias = 1;
1335 rec->name = estrndup(name_desc, sz);
1336 /* XXX This would only happen with a poorly
1337 * written man page, maybe warn? */
1338 if (name_desc[sz] == '\0')
1339 break;
1340 name_desc += sz + 1;
1341 continue;
1342 }
1343
1344 /*
1345 * Once rec->name is set, rest of the names
1346 * are to be treated as links or aliases.
1347 */
1348 if (rec->name && has_alias) {
1349 if (name_desc[sz] != ',') {
1350 /* No more commas left --> no more
1351 * aliases to take out */
1352 has_alias = 0;
1353 }
1354 concat2(&rec->links, name_desc, sz);
1355 /* XXX This would only happen with a poorly
1356 * written man page, maybe warn? */
1357 if (name_desc[sz] == '\0')
1358 break;
1359 name_desc += sz + 1;
1360 continue;
1361 }
1362 break;
1363 }
1364
1365 /* Parse any escape sequences that might be there */
1366 char *temp = parse_escape(name_desc);
1367 free(rec->name_desc);
1368 rec->name_desc = temp;
1369 temp = parse_escape(rec->name);
1370 free(rec->name);
1371 rec->name = temp;
1372 return;
1373 }
1374
1375 /* The RETURN VALUE section might be specified in multiple ways */
1376 if (strcmp(head->string, "RETURN") == 0 &&
1377 head->next != NULL && head->next->type == ROFFT_TEXT &&
1378 (strcmp(head->next->string, "VALUE") == 0 ||
1379 strcmp(head->next->string, "VALUES") == 0)) {
1380 man_parse_section(MANSEC_RETURN_VALUES, n, rec);
1381 return;
1382 }
1383
1384 /*
1385 * EXIT STATUS section can also be specified all on one line or on two
1386 * separate lines.
1387 */
1388 if (strcmp(head->string, "EXIT") == 0 &&
1389 head->next != NULL && head->next->type == ROFFT_TEXT &&
1390 strcmp(head->next->string, "STATUS") == 0) {
1391 man_parse_section(MANSEC_EXIT_STATUS, n, rec);
1392 return;
1393 }
1394
1395 /* Store the rest of the content in desc. */
1396 man_parse_section(MANSEC_NONE, n, rec);
1397 }
1398
1399 /*
1400 * pman_parse_node --
1401 * Generic function to iterate through a node. Usually called from
1402 * man_parse_section to parse a particular section of the man page.
1403 */
1404 static void
1405 pman_parse_node(const struct roff_node *n, secbuff *s)
1406 {
1407 if (n == NULL)
1408 return;
1409
1410 if (n->type == ROFFT_TEXT)
1411 append(s, n->string);
1412
1413 pman_parse_node(n->child, s);
1414 pman_parse_node(n->next, s);
1415 }
1416
1417 /*
1418 * man_parse_section --
1419 * Takes two parameters:
1420 * sec: Tells which section we are present in
1421 * n: Is the present node of the AST.
1422 * Depending on the section, we call pman_parse_node to parse that section and
1423 * concatenate the content from that section into the buffer for that section.
1424 */
1425 static void
1426 man_parse_section(enum man_sec sec, const struct roff_node *n, mandb_rec *rec)
1427 {
1428 /*
1429 * If the user sepecified the 'l' flag then just parse
1430 * the NAME section, ignore the rest.
1431 */
1432 if (mflags.limit)
1433 return;
1434
1435 switch (sec) {
1436 case MANSEC_LIBRARY:
1437 pman_parse_node(n, &rec->lib);
1438 break;
1439 case MANSEC_RETURN_VALUES:
1440 pman_parse_node(n, &rec->return_vals);
1441 break;
1442 case MANSEC_ENVIRONMENT:
1443 pman_parse_node(n, &rec->env);
1444 break;
1445 case MANSEC_FILES:
1446 pman_parse_node(n, &rec->files);
1447 break;
1448 case MANSEC_EXIT_STATUS:
1449 pman_parse_node(n, &rec->exit_status);
1450 break;
1451 case MANSEC_DIAGNOSTICS:
1452 pman_parse_node(n, &rec->diagnostics);
1453 break;
1454 case MANSEC_ERRORS:
1455 pman_parse_node(n, &rec->errors);
1456 break;
1457 case MANSEC_NAME:
1458 case MANSEC_SYNOPSIS:
1459 case MANSEC_EXAMPLES:
1460 case MANSEC_STANDARDS:
1461 case MANSEC_HISTORY:
1462 case MANSEC_BUGS:
1463 case MANSEC_AUTHORS:
1464 case MANSEC_COPYRIGHT:
1465 break;
1466 default:
1467 pman_parse_node(n, &rec->desc);
1468 break;
1469 }
1470
1471 }
1472
1473 /*
1474 * insert_into_db --
1475 * Inserts the parsed data of the man page in the Sqlite databse.
1476 * If any of the values is NULL, then we cleanup and return -1 indicating
1477 * an error.
1478 * Otherwise, store the data in the database and return 0.
1479 */
1480 static int
1481 insert_into_db(sqlite3 *db, mandb_rec *rec)
1482 {
1483 int rc = 0;
1484 int idx = -1;
1485 const char *sqlstr = NULL;
1486 sqlite3_stmt *stmt = NULL;
1487 char *ln = NULL;
1488 char *errmsg = NULL;
1489 long int mandb_rowid;
1490
1491 /*
1492 * At the very minimum we want to make sure that we store
1493 * the following data:
1494 * Name, one line description, and the MD5 hash
1495 */
1496 if (rec->name == NULL || rec->name_desc == NULL ||
1497 rec->md5_hash == NULL) {
1498 cleanup(rec);
1499 return -1;
1500 }
1501
1502 /* Write null byte at the end of all the sec_buffs */
1503 rec->desc.data[rec->desc.offset] = 0;
1504 rec->lib.data[rec->lib.offset] = 0;
1505 rec->env.data[rec->env.offset] = 0;
1506 rec->return_vals.data[rec->return_vals.offset] = 0;
1507 rec->exit_status.data[rec->exit_status.offset] = 0;
1508 rec->files.data[rec->files.offset] = 0;
1509 rec->diagnostics.data[rec->diagnostics.offset] = 0;
1510 rec->errors.data[rec->errors.offset] = 0;
1511
1512 /*
1513 * In case of a mdoc page: (sorry, no better place to put this code)
1514 * parse the comma separated list of names of man pages,
1515 * the first name will be stored in the mandb table, rest will be
1516 * treated as links and put in the mandb_links table.
1517 */
1518 if (rec->page_type == MDOC) {
1519 char *tmp;
1520 rec->links = estrdup(rec->name);
1521 free(rec->name);
1522 size_t sz = strcspn(rec->links, " \0");
1523 rec->name = emalloc(sz + 1);
1524 memcpy(rec->name, rec->links, sz);
1525 if(rec->name[sz - 1] == ',')
1526 rec->name[sz - 1] = 0;
1527 else
1528 rec->name[sz] = 0;
1529 while (rec->links[sz] == ' ')
1530 ++sz;
1531 tmp = estrdup(rec->links + sz);
1532 free(rec->links);
1533 rec->links = tmp;
1534 }
1535
1536 /*------------------------ Populate the mandb table---------------------------*/
1537 sqlstr = "INSERT INTO mandb VALUES (:section, :name, :name_desc, :desc,"
1538 " :lib, :return_vals, :env, :files, :exit_status,"
1539 " :diagnostics, :errors, :md5_hash, :machine)";
1540
1541 rc = sqlite3_prepare_v2(db, sqlstr, -1, &stmt, NULL);
1542 if (rc != SQLITE_OK)
1543 goto Out;
1544
1545 idx = sqlite3_bind_parameter_index(stmt, ":name");
1546 rc = sqlite3_bind_text(stmt, idx, rec->name, -1, NULL);
1547 if (rc != SQLITE_OK) {
1548 sqlite3_finalize(stmt);
1549 goto Out;
1550 }
1551
1552 idx = sqlite3_bind_parameter_index(stmt, ":section");
1553 rc = sqlite3_bind_text(stmt, idx, rec->section, -1, NULL);
1554 if (rc != SQLITE_OK) {
1555 sqlite3_finalize(stmt);
1556 goto Out;
1557 }
1558
1559 idx = sqlite3_bind_parameter_index(stmt, ":name_desc");
1560 rc = sqlite3_bind_text(stmt, idx, rec->name_desc, -1, NULL);
1561 if (rc != SQLITE_OK) {
1562 sqlite3_finalize(stmt);
1563 goto Out;
1564 }
1565
1566 idx = sqlite3_bind_parameter_index(stmt, ":desc");
1567 rc = sqlite3_bind_text(stmt, idx, rec->desc.data,
1568 rec->desc.offset + 1, NULL);
1569 if (rc != SQLITE_OK) {
1570 sqlite3_finalize(stmt);
1571 goto Out;
1572 }
1573
1574 idx = sqlite3_bind_parameter_index(stmt, ":lib");
1575 rc = sqlite3_bind_text(stmt, idx, rec->lib.data,
1576 rec->lib.offset + 1, NULL);
1577 if (rc != SQLITE_OK) {
1578 sqlite3_finalize(stmt);
1579 goto Out;
1580 }
1581
1582 idx = sqlite3_bind_parameter_index(stmt, ":return_vals");
1583 rc = sqlite3_bind_text(stmt, idx, rec->return_vals.data,
1584 rec->return_vals.offset + 1, NULL);
1585 if (rc != SQLITE_OK) {
1586 sqlite3_finalize(stmt);
1587 goto Out;
1588 }
1589
1590 idx = sqlite3_bind_parameter_index(stmt, ":env");
1591 rc = sqlite3_bind_text(stmt, idx, rec->env.data,
1592 rec->env.offset + 1, NULL);
1593 if (rc != SQLITE_OK) {
1594 sqlite3_finalize(stmt);
1595 goto Out;
1596 }
1597
1598 idx = sqlite3_bind_parameter_index(stmt, ":files");
1599 rc = sqlite3_bind_text(stmt, idx, rec->files.data,
1600 rec->files.offset + 1, NULL);
1601 if (rc != SQLITE_OK) {
1602 sqlite3_finalize(stmt);
1603 goto Out;
1604 }
1605
1606 idx = sqlite3_bind_parameter_index(stmt, ":exit_status");
1607 rc = sqlite3_bind_text(stmt, idx, rec->exit_status.data,
1608 rec->exit_status.offset + 1, NULL);
1609 if (rc != SQLITE_OK) {
1610 sqlite3_finalize(stmt);
1611 goto Out;
1612 }
1613
1614 idx = sqlite3_bind_parameter_index(stmt, ":diagnostics");
1615 rc = sqlite3_bind_text(stmt, idx, rec->diagnostics.data,
1616 rec->diagnostics.offset + 1, NULL);
1617 if (rc != SQLITE_OK) {
1618 sqlite3_finalize(stmt);
1619 goto Out;
1620 }
1621
1622 idx = sqlite3_bind_parameter_index(stmt, ":errors");
1623 rc = sqlite3_bind_text(stmt, idx, rec->errors.data,
1624 rec->errors.offset + 1, NULL);
1625 if (rc != SQLITE_OK) {
1626 sqlite3_finalize(stmt);
1627 goto Out;
1628 }
1629
1630 idx = sqlite3_bind_parameter_index(stmt, ":md5_hash");
1631 rc = sqlite3_bind_text(stmt, idx, rec->md5_hash, -1, NULL);
1632 if (rc != SQLITE_OK) {
1633 sqlite3_finalize(stmt);
1634 goto Out;
1635 }
1636
1637 idx = sqlite3_bind_parameter_index(stmt, ":machine");
1638 if (rec->machine)
1639 rc = sqlite3_bind_text(stmt, idx, rec->machine, -1, NULL);
1640 else
1641 rc = sqlite3_bind_null(stmt, idx);
1642 if (rc != SQLITE_OK) {
1643 sqlite3_finalize(stmt);
1644 goto Out;
1645 }
1646
1647 rc = sqlite3_step(stmt);
1648 if (rc != SQLITE_DONE) {
1649 sqlite3_finalize(stmt);
1650 goto Out;
1651 }
1652
1653 sqlite3_finalize(stmt);
1654
1655 /* Get the row id of the last inserted row */
1656 mandb_rowid = sqlite3_last_insert_rowid(db);
1657
1658 /*------------------------Populate the mandb_meta table-----------------------*/
1659 sqlstr = "INSERT INTO mandb_meta VALUES (:device, :inode, :mtime,"
1660 " :file, :md5_hash, :id)";
1661 rc = sqlite3_prepare_v2(db, sqlstr, -1, &stmt, NULL);
1662 if (rc != SQLITE_OK)
1663 goto Out;
1664
1665 idx = sqlite3_bind_parameter_index(stmt, ":device");
1666 rc = sqlite3_bind_int64(stmt, idx, rec->device);
1667 if (rc != SQLITE_OK) {
1668 sqlite3_finalize(stmt);
1669 goto Out;
1670 }
1671
1672 idx = sqlite3_bind_parameter_index(stmt, ":inode");
1673 rc = sqlite3_bind_int64(stmt, idx, rec->inode);
1674 if (rc != SQLITE_OK) {
1675 sqlite3_finalize(stmt);
1676 goto Out;
1677 }
1678
1679 idx = sqlite3_bind_parameter_index(stmt, ":mtime");
1680 rc = sqlite3_bind_int64(stmt, idx, rec->mtime);
1681 if (rc != SQLITE_OK) {
1682 sqlite3_finalize(stmt);
1683 goto Out;
1684 }
1685
1686 idx = sqlite3_bind_parameter_index(stmt, ":file");
1687 rc = sqlite3_bind_text(stmt, idx, rec->file_path, -1, NULL);
1688 if (rc != SQLITE_OK) {
1689 sqlite3_finalize(stmt);
1690 goto Out;
1691 }
1692
1693 idx = sqlite3_bind_parameter_index(stmt, ":md5_hash");
1694 rc = sqlite3_bind_text(stmt, idx, rec->md5_hash, -1, NULL);
1695 if (rc != SQLITE_OK) {
1696 sqlite3_finalize(stmt);
1697 goto Out;
1698 }
1699
1700 idx = sqlite3_bind_parameter_index(stmt, ":id");
1701 rc = sqlite3_bind_int64(stmt, idx, mandb_rowid);
1702 if (rc != SQLITE_OK) {
1703 sqlite3_finalize(stmt);
1704 goto Out;
1705 }
1706
1707 rc = sqlite3_step(stmt);
1708 sqlite3_finalize(stmt);
1709 if (rc == SQLITE_CONSTRAINT_UNIQUE) {
1710 /* The *most* probable reason for reaching here is that
1711 * the UNIQUE contraint on the file column of the mandb_meta
1712 * table was violated.
1713 * This can happen when a file was updated/modified.
1714 * To fix this we need to do two things:
1715 * 1. Delete the row for the older version of this file
1716 * from mandb table.
1717 * 2. Run an UPDATE query to update the row for this file
1718 * in the mandb_meta table.
1719 */
1720 warnx("Trying to update index for %s", rec->file_path);
1721 char *sql = sqlite3_mprintf("DELETE FROM mandb "
1722 "WHERE rowid = (SELECT id"
1723 " FROM mandb_meta"
1724 " WHERE file = %Q)",
1725 rec->file_path);
1726 sqlite3_exec(db, sql, NULL, NULL, &errmsg);
1727 sqlite3_free(sql);
1728 if (errmsg != NULL) {
1729 if (mflags.verbosity)
1730 warnx("%s", errmsg);
1731 free(errmsg);
1732 }
1733 sqlstr = "UPDATE mandb_meta SET device = :device,"
1734 " inode = :inode, mtime = :mtime, id = :id,"
1735 " md5_hash = :md5 WHERE file = :file";
1736 rc = sqlite3_prepare_v2(db, sqlstr, -1, &stmt, NULL);
1737 if (rc != SQLITE_OK) {
1738 if (mflags.verbosity)
1739 warnx("Update failed with error: %s",
1740 sqlite3_errmsg(db));
1741 close_db(db);
1742 cleanup(rec);
1743 errx(EXIT_FAILURE,
1744 "Consider running makemandb with -f option");
1745 }
1746
1747 idx = sqlite3_bind_parameter_index(stmt, ":device");
1748 sqlite3_bind_int64(stmt, idx, rec->device);
1749 idx = sqlite3_bind_parameter_index(stmt, ":inode");
1750 sqlite3_bind_int64(stmt, idx, rec->inode);
1751 idx = sqlite3_bind_parameter_index(stmt, ":mtime");
1752 sqlite3_bind_int64(stmt, idx, rec->mtime);
1753 idx = sqlite3_bind_parameter_index(stmt, ":id");
1754 sqlite3_bind_int64(stmt, idx, mandb_rowid);
1755 idx = sqlite3_bind_parameter_index(stmt, ":md5");
1756 sqlite3_bind_text(stmt, idx, rec->md5_hash, -1, NULL);
1757 idx = sqlite3_bind_parameter_index(stmt, ":file");
1758 sqlite3_bind_text(stmt, idx, rec->file_path, -1, NULL);
1759 rc = sqlite3_step(stmt);
1760 sqlite3_finalize(stmt);
1761
1762 if (rc != SQLITE_DONE) {
1763 if (mflags.verbosity)
1764 warnx("%s", sqlite3_errmsg(db));
1765 close_db(db);
1766 cleanup(rec);
1767 errx(EXIT_FAILURE,
1768 "Consider running makemandb with -f option");
1769 }
1770 } else if (rc != SQLITE_DONE) {
1771 /* Otherwise make this error fatal */
1772 warnx("Failed at %s\n%s", rec->file_path, sqlite3_errmsg(db));
1773 cleanup(rec);
1774 close_db(db);
1775 exit(EXIT_FAILURE);
1776 }
1777
1778 /*------------------------ Populate the mandb_links table---------------------*/
1779 char *str = NULL;
1780 char *links;
1781 if (rec->links && strlen(rec->links)) {
1782 links = rec->links;
1783 for(ln = strtok(links, " "); ln; ln = strtok(NULL, " ")) {
1784 if (ln[0] == ',')
1785 ln++;
1786 if(ln[strlen(ln) - 1] == ',')
1787 ln[strlen(ln) - 1] = 0;
1788
1789 str = sqlite3_mprintf("INSERT INTO mandb_links"
1790 " VALUES (%Q, %Q, %Q, %Q, %Q)",
1791 ln, rec->name, rec->section,
1792 rec->machine, rec->md5_hash);
1793 sqlite3_exec(db, str, NULL, NULL, &errmsg);
1794 sqlite3_free(str);
1795 if (errmsg != NULL) {
1796 warnx("%s", errmsg);
1797 cleanup(rec);
1798 free(errmsg);
1799 return -1;
1800 }
1801 }
1802 }
1803
1804 cleanup(rec);
1805 return 0;
1806
1807 Out:
1808 if (mflags.verbosity)
1809 warnx("%s", sqlite3_errmsg(db));
1810 cleanup(rec);
1811 return -1;
1812 }
1813
1814 /*
1815 * check_md5--
1816 * Generates the md5 hash of the file and checks if it already doesn't exist
1817 * in the table (passed as the 3rd parameter).
1818 * This function is being used to avoid hardlinks.
1819 * On successful completion it will also set the value of the fourth parameter
1820 * to the md5 hash of the file (computed previously). It is the responsibility
1821 * of the caller to free this buffer.
1822 * Return values:
1823 * -1: If an error occurs somewhere and sets the md5 return buffer to NULL.
1824 * 0: If the md5 hash does not exist in the table.
1825 * 1: If the hash exists in the database.
1826 */
1827 static int
1828 check_md5(const char *file, sqlite3 *db, const char *table, char **md5sum,
1829 void *buf, size_t buflen)
1830 {
1831 int rc = 0;
1832 int idx = -1;
1833 char *sqlstr = NULL;
1834 sqlite3_stmt *stmt = NULL;
1835
1836 assert(file != NULL);
1837 *md5sum = MD5Data(buf, buflen, NULL);
1838 if (*md5sum == NULL) {
1839 if (mflags.verbosity)
1840 warn("md5 failed: %s", file);
1841 return -1;
1842 }
1843
1844 easprintf(&sqlstr, "SELECT * FROM %s WHERE md5_hash = :md5_hash",
1845 table);
1846 rc = sqlite3_prepare_v2(db, sqlstr, -1, &stmt, NULL);
1847 if (rc != SQLITE_OK) {
1848 free(sqlstr);
1849 free(*md5sum);
1850 *md5sum = NULL;
1851 return -1;
1852 }
1853
1854 idx = sqlite3_bind_parameter_index(stmt, ":md5_hash");
1855 rc = sqlite3_bind_text(stmt, idx, *md5sum, -1, NULL);
1856 if (rc != SQLITE_OK) {
1857 if (mflags.verbosity)
1858 warnx("%s", sqlite3_errmsg(db));
1859 sqlite3_finalize(stmt);
1860 free(sqlstr);
1861 free(*md5sum);
1862 *md5sum = NULL;
1863 return -1;
1864 }
1865
1866 if (sqlite3_step(stmt) == SQLITE_ROW) {
1867 sqlite3_finalize(stmt);
1868 free(sqlstr);
1869 return 0;
1870 }
1871
1872 sqlite3_finalize(stmt);
1873 free(sqlstr);
1874 return 1;
1875 }
1876
1877 /* Optimize the index for faster search */
1878 static void
1879 optimize(sqlite3 *db)
1880 {
1881 const char *sqlstr;
1882 char *errmsg = NULL;
1883
1884 if (mflags.verbosity == 2)
1885 printf("Optimizing the database index\n");
1886 sqlstr = "INSERT INTO mandb(mandb) VALUES (\'optimize\');"
1887 "VACUUM";
1888 sqlite3_exec(db, sqlstr, NULL, NULL, &errmsg);
1889 if (errmsg != NULL) {
1890 if (mflags.verbosity)
1891 warnx("%s", errmsg);
1892 free(errmsg);
1893 return;
1894 }
1895 }
1896
1897 /*
1898 * cleanup --
1899 * cleans up the global buffers
1900 */
1901 static void
1902 cleanup(mandb_rec *rec)
1903 {
1904 rec->desc.offset = 0;
1905 rec->lib.offset = 0;
1906 rec->return_vals.offset = 0;
1907 rec->env.offset = 0;
1908 rec->exit_status.offset = 0;
1909 rec->diagnostics.offset = 0;
1910 rec->errors.offset = 0;
1911 rec->files.offset = 0;
1912
1913 free(rec->machine);
1914 rec->machine = NULL;
1915
1916 free(rec->links);
1917 rec->links = NULL;
1918
1919 free(rec->file_path);
1920 rec->file_path = NULL;
1921
1922 free(rec->name);
1923 rec->name = NULL;
1924
1925 free(rec->name_desc);
1926 rec->name_desc = NULL;
1927
1928 free(rec->md5_hash);
1929 rec->md5_hash = NULL;
1930
1931 free(rec->section);
1932 rec->section = NULL;
1933 }
1934
1935 /*
1936 * init_secbuffs--
1937 * Sets the value of buflen for all the sec_buff field of rec. And then
1938 * allocate memory to each sec_buff member of rec.
1939 */
1940 static void
1941 init_secbuffs(mandb_rec *rec)
1942 {
1943 /*
1944 * Some sec_buff might need more memory, for example desc,
1945 * which stores the data of the DESCRIPTION section,
1946 * while some might need very small amount of memory.
1947 * Therefore explicitly setting the value of buflen field for
1948 * each sec_buff.
1949 */
1950 rec->desc.buflen = 10 * BUFLEN;
1951 rec->desc.data = emalloc(rec->desc.buflen);
1952 rec->desc.offset = 0;
1953
1954 rec->lib.buflen = BUFLEN / 2;
1955 rec->lib.data = emalloc(rec->lib.buflen);
1956 rec->lib.offset = 0;
1957
1958 rec->return_vals.buflen = BUFLEN;
1959 rec->return_vals.data = emalloc(rec->return_vals.buflen);
1960 rec->return_vals.offset = 0;
1961
1962 rec->exit_status.buflen = BUFLEN;
1963 rec->exit_status.data = emalloc(rec->exit_status.buflen);
1964 rec->exit_status.offset = 0;
1965
1966 rec->env.buflen = BUFLEN;
1967 rec->env.data = emalloc(rec->env.buflen);
1968 rec->env.offset = 0;
1969
1970 rec->files.buflen = BUFLEN;
1971 rec->files.data = emalloc(rec->files.buflen);
1972 rec->files.offset = 0;
1973
1974 rec->diagnostics.buflen = BUFLEN;
1975 rec->diagnostics.data = emalloc(rec->diagnostics.buflen);
1976 rec->diagnostics.offset = 0;
1977
1978 rec->errors.buflen = BUFLEN;
1979 rec->errors.data = emalloc(rec->errors.buflen);
1980 rec->errors.offset = 0;
1981 }
1982
1983 /*
1984 * free_secbuffs--
1985 * This function should be called at the end, when all the pages have been
1986 * parsed.
1987 * It frees the memory allocated to sec_buffs by init_secbuffs in the starting.
1988 */
1989 static void
1990 free_secbuffs(mandb_rec *rec)
1991 {
1992 free(rec->desc.data);
1993 free(rec->lib.data);
1994 free(rec->return_vals.data);
1995 free(rec->exit_status.data);
1996 free(rec->env.data);
1997 free(rec->files.data);
1998 free(rec->diagnostics.data);
1999 free(rec->errors.data);
2000 }
2001
2002 static void
2003 replace_hyph(char *str)
2004 {
2005 char *iter = str;
2006 while ((iter = strchr(iter, ASCII_HYPH)) != NULL)
2007 *iter = '-';
2008
2009 iter = str;
2010 while ((iter = strchr(iter, ASCII_NBRSP)) != NULL)
2011 *iter = '-';
2012 }
2013
2014 static char *
2015 parse_escape(const char *str)
2016 {
2017 const char *backslash, *last_backslash;
2018 char *result, *iter;
2019 size_t len;
2020
2021 assert(str);
2022
2023 last_backslash = str;
2024 backslash = strchr(str, '\\');
2025 if (backslash == NULL) {
2026 result = estrdup(str);
2027 replace_hyph(result);
2028 return result;
2029 }
2030
2031 result = emalloc(strlen(str) + 1);
2032 iter = result;
2033
2034 do {
2035 len = backslash - last_backslash;
2036 memcpy(iter, last_backslash, len);
2037 iter += len;
2038 if (backslash[1] == '-' || backslash[1] == ' ') {
2039 *iter++ = backslash[1];
2040 last_backslash = backslash + 2;
2041 backslash = strchr(last_backslash, '\\');
2042 } else {
2043 ++backslash;
2044 mandoc_escape(&backslash, NULL, NULL);
2045 last_backslash = backslash;
2046 if (backslash == NULL)
2047 break;
2048 backslash = strchr(last_backslash, '\\');
2049 }
2050 } while (backslash != NULL);
2051 if (last_backslash != NULL)
2052 strcpy(iter, last_backslash);
2053
2054 replace_hyph(result);
2055 return result;
2056 }
2057
2058 /*
2059 * append--
2060 * Concatenates a space and src at the end of sbuff->data (much like concat in
2061 * apropos-utils.c).
2062 * Rather than reallocating space for writing data, it uses the value of the
2063 * offset field of sec_buff to write new data at the free space left in the
2064 * buffer.
2065 * In case the size of the data to be appended exceeds the number of bytes left
2066 * in the buffer, it reallocates buflen number of bytes and then continues.
2067 * Value of offset field should be adjusted as new data is written.
2068 *
2069 * NOTE: This function does not write the null byte at the end of the buffers,
2070 * write a null byte at the position pointed to by offset before inserting data
2071 * in the db.
2072 */
2073 static void
2074 append(secbuff *sbuff, const char *src)
2075 {
2076 short flag = 0;
2077 size_t srclen, newlen;
2078 char *temp;
2079
2080 assert(src != NULL);
2081 temp = parse_escape(src);
2082 srclen = strlen(temp);
2083
2084 if (sbuff->data == NULL) {
2085 sbuff->data = emalloc(sbuff->buflen);
2086 sbuff->offset = 0;
2087 }
2088
2089 newlen = sbuff->offset + srclen + 2;
2090 if (newlen >= sbuff->buflen) {
2091 while (sbuff->buflen < newlen)
2092 sbuff->buflen += sbuff->buflen;
2093 sbuff->data = erealloc(sbuff->data, sbuff->buflen);
2094 flag = 1;
2095 }
2096
2097 /* Append a space at the end of the buffer. */
2098 if (sbuff->offset || flag)
2099 sbuff->data[sbuff->offset++] = ' ';
2100 /* Now, copy src at the end of the buffer. */
2101 memcpy(sbuff->data + sbuff->offset, temp, srclen);
2102 sbuff->offset += srclen;
2103 free(temp);
2104 }
2105
2106 static void
2107 usage(void)
2108 {
2109 fprintf(stderr, "Usage: %s [-floQqv] [-C path]\n", getprogname());
2110 exit(1);
2111 }
2112