makemandb.c revision 1.54 1 /* $NetBSD: makemandb.c,v 1.54 2017/05/02 13:54:08 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 1.54 2017/05/02 13:54:08 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 *, 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 mdoc_parse_Sh(const struct roff_node *, mandb_rec *);
111 static void pmdoc_Xr(const struct roff_node *, mandb_rec *);
112 static void pmdoc_Pp(const struct roff_node *, mandb_rec *);
113 static void pmdoc_macro_handler(const struct roff_node *, mandb_rec *, int);
114 static void pman_parse_node(const struct roff_node *, secbuff *);
115 static void pman_parse_name(const struct roff_node *, mandb_rec *);
116 static void pman_sh(const struct roff_node *, mandb_rec *);
117 static void pman_block(const struct roff_node *, mandb_rec *);
118 static void traversedir(const char *, const char *, sqlite3 *, struct mparse *);
119 static void mdoc_parse_section(enum roff_sec, const char *, mandb_rec *);
120 static void man_parse_section(enum man_sec, const struct roff_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 void replace_hyph(char *);
128 static makemandb_flags mflags = { .verbosity = 1 };
129
130 static const proff_nf mdocs[MDOC_MAX + 1] = {
131 NULL, /* Ap */
132 NULL, /* Dd */
133 NULL, /* Dt */
134 NULL, /* Os */
135
136 pmdoc_Sh, /* Sh */
137 NULL, /* Ss */
138 pmdoc_Pp, /* Pp */
139 NULL, /* D1 */
140
141 NULL, /* Dl */
142 NULL, /* Bd */
143 NULL, /* Ed */
144 NULL, /* Bl */
145
146 NULL, /* El */
147 NULL, /* It */
148 NULL, /* Ad */
149 NULL, /* An */
150
151 NULL, /* Ar */
152 NULL, /* Cd */
153 NULL, /* Cm */
154 NULL, /* Dv */
155
156 NULL, /* Er */
157 NULL, /* Ev */
158 NULL, /* Ex */
159 NULL, /* Fa */
160
161 NULL, /* Fd */
162 NULL, /* Fl */
163 NULL, /* Fn */
164 NULL, /* Ft */
165
166 NULL, /* Ic */
167 NULL, /* In */
168 NULL, /* Li */
169 pmdoc_Nd, /* Nd */
170
171 pmdoc_Nm, /* Nm */
172 NULL, /* Op */
173 NULL, /* Ot */
174 NULL, /* Pa */
175
176 NULL, /* Rv */
177 NULL, /* St */
178 NULL, /* Va */
179 NULL, /* Vt */
180
181 pmdoc_Xr, /* Xr */
182 NULL, /* %A */
183 NULL, /* %B */
184 NULL, /* %D */
185
186 NULL, /* %I */
187 NULL, /* %J */
188 NULL, /* %N */
189 NULL, /* %O */
190
191 NULL, /* %P */
192 NULL, /* %R */
193 NULL, /* %T */
194 NULL, /* %V */
195
196 NULL, /* Ac */
197 NULL, /* Ao */
198 NULL, /* Aq */
199 NULL, /* At */
200
201 NULL, /* Bc */
202 NULL, /* Bf */
203 NULL, /* Bo */
204 NULL, /* Bq */
205
206 NULL, /* Bsx */
207 NULL, /* Bx */
208 NULL, /* Db */
209 NULL, /* Dc */
210
211 NULL, /* Do */
212 NULL, /* Dq */
213 NULL, /* Ec */
214 NULL, /* Ef */
215
216 NULL, /* Em */
217 NULL, /* Eo */
218 NULL, /* Fx */
219 NULL, /* Ms */
220
221 NULL, /* No */
222 NULL, /* Ns */
223 NULL, /* Nx */
224 NULL, /* Ox */
225
226 NULL, /* Pc */
227 NULL, /* Pf */
228 NULL, /* Po */
229 NULL, /* Pq */
230
231 NULL, /* Qc */
232 NULL, /* Ql */
233 NULL, /* Qo */
234 NULL, /* Qq */
235
236 NULL, /* Re */
237 NULL, /* Rs */
238 NULL, /* Sc */
239 NULL, /* So */
240
241 NULL, /* Sq */
242 NULL, /* Sm */
243 NULL, /* Sx */
244 NULL, /* Sy */
245
246 NULL, /* Tn */
247 NULL, /* Ux */
248 NULL, /* Xc */
249 NULL, /* Xo */
250
251 NULL, /* Fo */
252 NULL, /* Fc */
253 NULL, /* Oo */
254 NULL, /* Oc */
255
256 NULL, /* Bk */
257 NULL, /* Ek */
258 NULL, /* Bt */
259 NULL, /* Hf */
260
261 NULL, /* Fr */
262 NULL, /* Ud */
263 NULL, /* Lb */
264 NULL, /* Lp */
265
266 NULL, /* Lk */
267 NULL, /* Mt */
268 NULL, /* Brq */
269 NULL, /* Bro */
270
271 NULL, /* Brc */
272 NULL, /* %C */
273 NULL, /* Es */
274 NULL, /* En */
275
276 NULL, /* Dx */
277 NULL, /* %Q */
278 NULL, /* br */
279 NULL, /* sp */
280
281 NULL, /* %U */
282 NULL, /* Ta */
283 NULL, /* ll */
284 NULL, /* text */
285 };
286
287 static const proff_nf mans[MAN_MAX] = {
288 NULL, //br
289 NULL, //TH
290 pman_sh, //SH
291 NULL, //SS
292 NULL, //TP
293 NULL, //LP
294 NULL, //PP
295 NULL, //P
296 NULL, //IP
297 NULL, //HP
298 NULL, //SM
299 NULL, //SB
300 NULL, //BI
301 NULL, //IB
302 NULL, //BR
303 NULL, //RB
304 NULL, //R
305 pman_block, //B
306 NULL, //I
307 NULL, //IR
308 NULL, //RI
309 NULL, //sp
310 NULL, //nf
311 NULL, //fi
312 NULL, //RE
313 NULL, //RS
314 NULL, //DT
315 NULL, //UC
316 NULL, //PD
317 NULL, //AT
318 NULL, //in
319 NULL, //ft
320 NULL, //OP
321 NULL, //EX
322 NULL, //EE
323 NULL, //UR
324 NULL, //UE
325 NULL, //ll
326 };
327
328
329 int
330 main(int argc, char *argv[])
331 {
332 FILE *file;
333 const char *sqlstr, *manconf = NULL;
334 char *line, *command;
335 char *errmsg;
336 int ch;
337 struct mparse *mp;
338 sqlite3 *db;
339 ssize_t len;
340 size_t linesize;
341 struct roff_mandb_rec rec;
342
343 while ((ch = getopt(argc, argv, "C:floQqv")) != -1) {
344 switch (ch) {
345 case 'C':
346 manconf = optarg;
347 break;
348 case 'f':
349 mflags.recreate = 1;
350 break;
351 case 'l':
352 mflags.limit = 1;
353 break;
354 case 'o':
355 mflags.optimize = 1;
356 break;
357 case 'Q':
358 mflags.verbosity = 0;
359 break;
360 case 'q':
361 mflags.verbosity = 1;
362 break;
363 case 'v':
364 mflags.verbosity = 2;
365 break;
366 default:
367 usage();
368 }
369 }
370
371 memset(&rec, 0, sizeof(rec));
372
373 init_secbuffs(&rec);
374 mchars_alloc();
375 mp = mparse_alloc(0, MANDOCLEVEL_BADARG, NULL, NULL);
376
377 if (manconf) {
378 char *arg;
379 size_t command_len = shquote(manconf, NULL, 0) + 1;
380 arg = emalloc(command_len);
381 shquote(manconf, arg, command_len);
382 easprintf(&command, "man -p -C %s", arg);
383 free(arg);
384 } else {
385 command = estrdup("man -p");
386 manconf = MANCONF;
387 }
388
389 if (mflags.recreate) {
390 char *dbp = get_dbpath(manconf);
391 /* No error here, it will fail in init_db in the same call */
392 if (dbp != NULL)
393 remove(dbp);
394 }
395
396 if ((db = init_db(MANDB_CREATE, manconf)) == NULL)
397 exit(EXIT_FAILURE);
398
399 sqlite3_exec(db, "PRAGMA synchronous = 0", NULL, NULL, &errmsg);
400 if (errmsg != NULL) {
401 warnx("%s", errmsg);
402 free(errmsg);
403 close_db(db);
404 exit(EXIT_FAILURE);
405 }
406
407 sqlite3_exec(db, "ATTACH DATABASE \':memory:\' AS metadb", NULL, NULL,
408 &errmsg);
409 if (errmsg != NULL) {
410 warnx("%s", errmsg);
411 free(errmsg);
412 close_db(db);
413 exit(EXIT_FAILURE);
414 }
415
416
417 /* Call man -p to get the list of man page dirs */
418 if ((file = popen(command, "r")) == NULL) {
419 close_db(db);
420 err(EXIT_FAILURE, "fopen failed");
421 }
422 free(command);
423
424 /* Begin the transaction for indexing the pages */
425 sqlite3_exec(db, "BEGIN", NULL, NULL, &errmsg);
426 if (errmsg != NULL) {
427 warnx("%s", errmsg);
428 free(errmsg);
429 close_db(db);
430 exit(EXIT_FAILURE);
431 }
432
433 sqlstr = "CREATE TABLE metadb.file_cache(device, inode, mtime, parent,"
434 " file PRIMARY KEY);"
435 "CREATE UNIQUE INDEX metadb.index_file_cache_dev"
436 " ON file_cache (device, inode)";
437
438 sqlite3_exec(db, sqlstr, NULL, NULL, &errmsg);
439 if (errmsg != NULL) {
440 warnx("%s", errmsg);
441 free(errmsg);
442 close_db(db);
443 exit(EXIT_FAILURE);
444 }
445
446 if (mflags.verbosity)
447 printf("Building temporary file cache\n");
448 line = NULL;
449 linesize = 0;
450 while ((len = getline(&line, &linesize, file)) != -1) {
451 /* Replace the new line character at the end of string with '\0' */
452 line[len - 1] = '\0';
453 char *pdir = estrdup(dirname(line));
454 /* Traverse the man page directories and parse the pages */
455 traversedir(pdir, line, db, mp);
456 free(pdir);
457 }
458 free(line);
459
460 if (pclose(file) == -1) {
461 close_db(db);
462 cleanup(&rec);
463 free_secbuffs(&rec);
464 err(EXIT_FAILURE, "pclose error");
465 }
466
467 if (mflags.verbosity)
468 printf("Performing index update\n");
469 update_db(db, mp, &rec);
470 mparse_free(mp);
471 mchars_free();
472 free_secbuffs(&rec);
473
474 /* Commit the transaction */
475 sqlite3_exec(db, "COMMIT", NULL, NULL, &errmsg);
476 if (errmsg != NULL) {
477 warnx("%s", errmsg);
478 free(errmsg);
479 close_db(db);
480 exit(EXIT_FAILURE);
481 }
482
483 if (mflags.optimize)
484 optimize(db);
485
486 close_db(db);
487 return 0;
488 }
489
490 /*
491 * traversedir --
492 * Traverses the given directory recursively and passes all the man page files
493 * in the way to build_file_cache()
494 */
495 static void
496 traversedir(const char *parent, const char *file, sqlite3 *db,
497 struct mparse *mp)
498 {
499 struct stat sb;
500 struct dirent *dirp;
501 DIR *dp;
502 char *buf;
503
504 if (stat(file, &sb) < 0) {
505 if (mflags.verbosity)
506 warn("stat failed: %s", file);
507 return;
508 }
509
510 /* If it is a directory, traverse it recursively */
511 if (S_ISDIR(sb.st_mode)) {
512 if ((dp = opendir(file)) == NULL) {
513 if (mflags.verbosity)
514 warn("opendir error: %s", file);
515 return;
516 }
517
518 while ((dirp = readdir(dp)) != NULL) {
519 /* Avoid . and .. entries in a directory */
520 if (dirp->d_name[0] != '.') {
521 easprintf(&buf, "%s/%s", file, dirp->d_name);
522 traversedir(parent, buf, db, mp);
523 free(buf);
524 }
525 }
526 closedir(dp);
527 return;
528 }
529
530 if (!S_ISREG(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_filter_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_free(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_free(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_free(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, &md5sum, buf, buflen);
809 assert(md5sum != NULL);
810 if (md5_status == -1) {
811 if (mflags.verbosity)
812 warnx("An error occurred in checking md5 value"
813 " for file %s", file);
814 err_count++;
815 continue;
816 }
817
818 if (md5_status == 0) {
819 /*
820 * The MD5 hash is already present in the database,
821 * so simply update the metadata.
822 */
823 update_existing_entry(db, file, md5sum, rec,
824 &new_count, &link_count, &err_count);
825 free(md5sum);
826 continue;
827 }
828
829 if (md5_status == 1) {
830 /*
831 * The MD5 hash was not present in the database.
832 * This means is either a new file or an updated file.
833 * We should go ahead with parsing.
834 */
835 if (chdir(parent) == -1) {
836 if (mflags.verbosity)
837 warn("chdir failed for `%s', could "
838 "not index `%s'", parent, file);
839 err_count++;
840 free(md5sum);
841 continue;
842 }
843
844 if (mflags.verbosity == 2)
845 printf("Parsing: %s\n", file);
846 rec->md5_hash = md5sum;
847 rec->file_path = estrdup(file);
848 // file_path is freed by insert_into_db itself.
849 begin_parse(file, mp, rec, buf, buflen);
850 if (insert_into_db(db, rec) < 0) {
851 if (mflags.verbosity)
852 warnx("Error in indexing `%s'", file);
853 err_count++;
854 } else {
855 new_count++;
856 }
857 }
858 }
859
860 if (mflags.verbosity == 2) {
861 printf("Number of new or updated pages encountered: %d\n"
862 "Number of hard links found: %d\n"
863 "Number of pages that were successfully"
864 " indexed or updated: %d\n"
865 "Number of pages that could not be indexed"
866 " due to errors: %d\n",
867 total_count - link_count, link_count, new_count, err_count);
868 }
869
870 if (mflags.recreate)
871 return;
872
873 if (mflags.verbosity == 2)
874 printf("Deleting stale index entries\n");
875
876 sqlstr = "DELETE FROM mandb_meta WHERE file NOT IN"
877 " (SELECT file FROM metadb.file_cache);"
878 "DELETE FROM mandb_links WHERE md5_hash NOT IN"
879 " (SELECT md5_hash from mandb_meta);"
880 "DROP TABLE metadb.file_cache;"
881 "DELETE FROM mandb WHERE rowid NOT IN"
882 " (SELECT id FROM mandb_meta);";
883
884 sqlite3_exec(db, sqlstr, NULL, NULL, &errmsg);
885 if (errmsg != NULL) {
886 warnx("Removing old entries failed: %s", errmsg);
887 warnx("Please rebuild database from scratch with -f.");
888 free(errmsg);
889 return;
890 }
891 }
892
893 /*
894 * begin_parse --
895 * parses the man page using libmandoc
896 */
897 static void
898 begin_parse(const char *file, struct mparse *mp, mandb_rec *rec,
899 const void *buf, size_t len)
900 {
901 struct roff_man *roff;
902 mparse_reset(mp);
903
904 rec->xr_found = 0;
905
906 if (mparse_readmem(mp, buf, len, file) >= MANDOCLEVEL_BADARG) {
907 /* Printing this warning at verbosity level 2
908 * because some packages from pkgsrc might trigger several
909 * of such warnings.
910 */
911 if (mflags.verbosity == 2)
912 warnx("%s: Parse failure", file);
913 return;
914 }
915
916 mparse_result(mp, &roff, NULL);
917 if (roff == NULL) {
918 if (mflags.verbosity == 2)
919 warnx("Not a roff(7) page");
920 return;
921 }
922
923 if (roff->macroset == MACROSET_MDOC) {
924 mdoc_validate(roff);
925 rec->page_type = MDOC;
926 proff_node(roff->first->child, rec, mdocs);
927 } else if (roff->macroset == MACROSET_MAN) {
928 man_validate(roff);
929 rec->page_type = MAN;
930 proff_node(roff->first->child, rec, mans);
931 } else
932 warnx("Unknown macroset %d", roff->macroset);
933 set_machine(roff, rec);
934 set_section(roff, rec);
935 }
936
937 /*
938 * set_section --
939 * Extracts the section number and normalizes it to only the numeric part
940 * (Which should be the first character of the string).
941 */
942 static void
943 set_section(const struct roff_man *rm, mandb_rec *rec)
944 {
945 if (!rm)
946 return;
947 const struct roff_meta *rm_meta = &rm->meta;
948 const char *s = rm_meta->msec == NULL ? "?" : rm_meta->msec;
949 easprintf(&rec->section, "%s", s);
950 if (rec->section[0] == '?' && mflags.verbosity == 2)
951 warnx("%s: Missing section number", rec->file_path);
952 }
953
954 /*
955 * get_machine --
956 * Extracts the machine architecture information if available.
957 */
958 static void
959 set_machine(const struct roff_man *rm, mandb_rec *rec)
960 {
961 if (rm == NULL)
962 return;
963 const struct roff_meta *rm_meta = &rm->meta;
964 if (rm_meta->arch)
965 rec->machine = estrdup(rm_meta->arch);
966 }
967
968 /*
969 * pmdoc_Nm --
970 * Extracts the Name of the manual page from the .Nm macro
971 */
972 static void
973 pmdoc_Nm(const struct roff_node *n, mandb_rec *rec)
974 {
975 if (n->sec != SEC_NAME)
976 return;
977
978 for (n = n->child; n; n = n->next) {
979 if (n->type == ROFFT_TEXT) {
980 char *escaped_name = parse_escape(n->string);
981 concat(&rec->name, escaped_name);
982 free(escaped_name);
983 }
984 }
985 }
986
987 /*
988 * pmdoc_Nd --
989 * Extracts the one line description of the man page from the .Nd macro
990 */
991 static void
992 pmdoc_Nd(const struct roff_node *n, mandb_rec *rec)
993 {
994 if (n->type == ROFFT_BODY)
995 deroff(&rec->name_desc, n);
996 if (rec->name_desc)
997 replace_hyph(rec->name_desc);
998
999 }
1000
1001 /*
1002 * pmdoc_macro_handler--
1003 * This function is a single point of handling all the special macros that we
1004 * want to handle especially. For example the .Xr macro for properly parsing
1005 * the referenced page name along with the section number, or the .Pp macro
1006 * for adding a new line whenever we encounter it.
1007 */
1008 static void
1009 pmdoc_macro_handler(const struct roff_node *n, mandb_rec *rec, int doct)
1010 {
1011 const struct roff_node *sn;
1012 assert(n);
1013
1014 switch (doct) {
1015 /* Parse the man page references.
1016 * Basically the .Xr macros are used like:
1017 * .Xr ls 1
1018 * and formatted like this:
1019 * ls(1)
1020 * Prepare a buffer to format the data like the above example and call
1021 * pmdoc_parse_section to append it.
1022 */
1023 case MDOC_Xr:
1024 n = n->child;
1025 while (n->type != ROFFT_TEXT && n->next)
1026 n = n->next;
1027
1028 if (n && n->type != ROFFT_TEXT)
1029 return;
1030 sn = n;
1031 if (n->next)
1032 n = n->next;
1033
1034 while (n->type != ROFFT_TEXT && n->next)
1035 n = n->next;
1036
1037 if (n && n->type == ROFFT_TEXT) {
1038 char *buf;
1039 easprintf(&buf, "%s(%s)", sn->string, n->string);
1040 mdoc_parse_section(n->sec, buf, rec);
1041 free(buf);
1042 }
1043
1044 break;
1045
1046 /* Parse the .Pp macro to add a new line */
1047 case MDOC_Pp:
1048 if (n->type == ROFFT_TEXT)
1049 mdoc_parse_section(n->sec, "\n", rec);
1050 break;
1051 default:
1052 break;
1053 }
1054
1055 }
1056
1057 /*
1058 * pmdoc_Xr, pmdoc_Pp--
1059 * Empty stubs.
1060 * The parser calls these functions each time it encounters
1061 * a .Xr or .Pp macro. We are parsing all the data from
1062 * the pmdoc_Sh function, so don't do anything here.
1063 * (See if else blocks in pmdoc_Sh.)
1064 */
1065 static void
1066 pmdoc_Xr(const struct roff_node *n, mandb_rec *rec)
1067 {
1068 }
1069
1070 static void
1071 pmdoc_Pp(const struct roff_node *n, mandb_rec *rec)
1072 {
1073 }
1074
1075 /*
1076 * pmdoc_Sh --
1077 * Called when a .Sh macro is encountered and tries to parse its body
1078 */
1079 static void
1080 pmdoc_Sh(const struct roff_node *n, mandb_rec *rec)
1081 {
1082 if (n == NULL)
1083 return;
1084
1085 switch (n->sec) {
1086 case SEC_NAME:
1087 case SEC_SYNOPSIS:
1088 case SEC_EXAMPLES:
1089 case SEC_STANDARDS:
1090 case SEC_HISTORY:
1091 case SEC_AUTHORS:
1092 case SEC_BUGS:
1093 /*
1094 * We don't care about text from these sections
1095 */
1096 return;
1097 default:
1098 break;
1099 }
1100
1101 if (n->type == ROFFT_BLOCK)
1102 mdoc_parse_Sh(n->body, rec);
1103 }
1104
1105 /*
1106 * Called from pmdoc_Sh to parse body of a .Sh macro. It calls
1107 * mdoc_parse_section to append the data to the section specific buffer.
1108 * The .Xr macro needs special handling, thus the separate if branch for it.
1109 */
1110 static void
1111 mdoc_parse_Sh(const struct roff_node *n, mandb_rec *rec)
1112 {
1113 if (n == NULL || (n->type != ROFFT_TEXT && n->tok == MDOC_MAX))
1114 return;
1115 int xr_found = 0;
1116
1117 if (n->type == ROFFT_TEXT) {
1118 mdoc_parse_section(n->sec, n->string, rec);
1119 } else if (mdocs[n->tok] == pmdoc_Xr) {
1120 /*
1121 * When encountering other inline macros,
1122 * call pmdoc_macro_handler.
1123 */
1124 pmdoc_macro_handler(n, rec, MDOC_Xr);
1125 xr_found = 1;
1126 } else if (mdocs[n->tok] == pmdoc_Pp) {
1127 pmdoc_macro_handler(n, rec, MDOC_Pp);
1128 }
1129
1130 /*
1131 * If an Xr macro was encountered then the child node has
1132 * already been explored by pmdoc_macro_handler.
1133 */
1134 if (xr_found == 0)
1135 mdoc_parse_Sh(n->child, rec);
1136 mdoc_parse_Sh(n->next, rec);
1137 }
1138
1139 /*
1140 * mdoc_parse_section--
1141 * Utility function for parsing sections of the mdoc type pages.
1142 * Takes two params:
1143 * 1. sec is an enum which indicates the section in which we are present
1144 * 2. string is the string which we need to append to the secbuff for this
1145 * particular section.
1146 * The function appends string to the global section buffer and returns.
1147 */
1148 static void
1149 mdoc_parse_section(enum roff_sec sec, const char *string, mandb_rec *rec)
1150 {
1151 /*
1152 * If the user specified the 'l' flag, then parse and store only the
1153 * NAME section. Ignore the rest.
1154 */
1155 if (mflags.limit)
1156 return;
1157
1158 switch (sec) {
1159 case SEC_LIBRARY:
1160 append(&rec->lib, string);
1161 break;
1162 case SEC_RETURN_VALUES:
1163 append(&rec->return_vals, string);
1164 break;
1165 case SEC_ENVIRONMENT:
1166 append(&rec->env, string);
1167 break;
1168 case SEC_FILES:
1169 append(&rec->files, string);
1170 break;
1171 case SEC_EXIT_STATUS:
1172 append(&rec->exit_status, string);
1173 break;
1174 case SEC_DIAGNOSTICS:
1175 append(&rec->diagnostics, string);
1176 break;
1177 case SEC_ERRORS:
1178 append(&rec->errors, string);
1179 break;
1180 default:
1181 append(&rec->desc, string);
1182 break;
1183 }
1184 }
1185
1186 static void
1187 proff_node(const struct roff_node *n, mandb_rec *rec, const proff_nf *func)
1188 {
1189 if (n == NULL)
1190 return;
1191
1192 switch (n->type) {
1193 case (ROFFT_BODY):
1194 /* FALLTHROUGH */
1195 case (ROFFT_BLOCK):
1196 /* FALLTHROUGH */
1197 case (ROFFT_ELEM):
1198 if (func[n->tok] != NULL)
1199 (*func[n->tok])(n, rec);
1200 break;
1201 default:
1202 break;
1203 }
1204
1205 proff_node(n->child, rec, func);
1206 proff_node(n->next, rec, func);
1207 }
1208
1209 /*
1210 * pman_parse_name --
1211 * Parses the NAME section and puts the complete content in the name_desc
1212 * variable.
1213 */
1214 static void
1215 pman_parse_name(const struct roff_node *n, mandb_rec *rec)
1216 {
1217 if (n == NULL)
1218 return;
1219
1220 if (n->type == ROFFT_TEXT) {
1221 char *tmp = parse_escape(n->string);
1222 concat(&rec->name_desc, tmp);
1223 free(tmp);
1224 }
1225
1226 if (n->child)
1227 pman_parse_name(n->child, rec);
1228
1229 if(n->next)
1230 pman_parse_name(n->next, rec);
1231 }
1232
1233 /*
1234 * A stub function to be able to parse the macros like .B embedded inside
1235 * a section.
1236 */
1237 static void
1238 pman_block(const struct roff_node *n, mandb_rec *rec)
1239 {
1240 }
1241
1242 /*
1243 * pman_sh --
1244 * This function does one of the two things:
1245 * 1. If the present section is NAME, then it will:
1246 * (a) Extract the name of the page (in case of multiple comma separated
1247 * names, it will pick up the first one).
1248 * (b) Build a space spearated list of all the symlinks/hardlinks to
1249 * this page and store in the buffer 'links'. These are extracted from
1250 * the comma separated list of names in the NAME section as well.
1251 * (c) Move on to the one line description section, which is after the list
1252 * of names in the NAME section.
1253 * 2. Otherwise, it will check the section name and call the man_parse_section
1254 * function, passing the enum corresponding to that section.
1255 */
1256 static void
1257 pman_sh(const struct roff_node *n, mandb_rec *rec)
1258 {
1259 static const struct {
1260 enum man_sec section;
1261 const char *header;
1262 } mapping[] = {
1263 { MANSEC_DESCRIPTION, "DESCRIPTION" },
1264 { MANSEC_SYNOPSIS, "SYNOPSIS" },
1265 { MANSEC_LIBRARY, "LIBRARY" },
1266 { MANSEC_ERRORS, "ERRORS" },
1267 { MANSEC_FILES, "FILES" },
1268 { MANSEC_RETURN_VALUES, "RETURN VALUE" },
1269 { MANSEC_RETURN_VALUES, "RETURN VALUES" },
1270 { MANSEC_EXIT_STATUS, "EXIT STATUS" },
1271 { MANSEC_EXAMPLES, "EXAMPLES" },
1272 { MANSEC_EXAMPLES, "EXAMPLE" },
1273 { MANSEC_STANDARDS, "STANDARDS" },
1274 { MANSEC_HISTORY, "HISTORY" },
1275 { MANSEC_BUGS, "BUGS" },
1276 { MANSEC_AUTHORS, "AUTHORS" },
1277 { MANSEC_COPYRIGHT, "COPYRIGHT" },
1278 };
1279 const struct roff_node *head;
1280 char *name_desc;
1281 size_t sz;
1282 size_t i;
1283
1284 if ((head = n->parent->head) == NULL || (head = head->child) == NULL ||
1285 head->type != ROFFT_TEXT)
1286 return;
1287
1288 /*
1289 * Check if this section should be extracted and
1290 * where it should be stored. Handled the trival cases first.
1291 */
1292 for (i = 0; i < sizeof(mapping) / sizeof(mapping[0]); ++i) {
1293 if (strcmp(head->string, mapping[i].header) == 0) {
1294 man_parse_section(mapping[i].section, n, rec);
1295 return;
1296 }
1297 }
1298
1299 if (strcmp(head->string, "NAME") == 0) {
1300 /*
1301 * We are in the NAME section.
1302 * pman_parse_name will put the complete content in name_desc.
1303 */
1304 pman_parse_name(n, rec);
1305
1306 name_desc = rec->name_desc;
1307 if (name_desc == NULL)
1308 return;
1309
1310 /* Remove any leading spaces. */
1311 while (name_desc[0] == ' ')
1312 name_desc++;
1313
1314 /* If the line begins with a "\&", avoid those */
1315 if (name_desc[0] == '\\' && name_desc[1] == '&')
1316 name_desc += 2;
1317
1318 /* Now name_desc should be left with a comma-space
1319 * separated list of names and the one line description
1320 * of the page:
1321 * "a, b, c \- sample description"
1322 * Take out the first name, before the first comma
1323 * (or space) and store it in rec->name.
1324 * If the page has aliases then they should be
1325 * in the form of a comma separated list.
1326 * Keep looping while there is a comma in name_desc,
1327 * extract the alias name and store in rec->links.
1328 * When there are no more commas left, break out.
1329 */
1330 int has_alias = 0; // Any more aliases left?
1331 while (*name_desc) {
1332 /* Remove any leading spaces or hyphens. */
1333 if (name_desc[0] == ' ' || name_desc[0] == '-') {
1334 name_desc++;
1335 continue;
1336 }
1337 sz = strcspn(name_desc, ", ");
1338
1339 /* Extract the first term and store it in rec->name. */
1340 if (rec->name == NULL) {
1341 if (name_desc[sz] == ',')
1342 has_alias = 1;
1343 rec->name = estrndup(name_desc, sz);
1344 /* XXX This would only happen with a poorly
1345 * written man page, maybe warn? */
1346 if (name_desc[sz] == '\0')
1347 break;
1348 name_desc += sz + 1;
1349 continue;
1350 }
1351
1352 /*
1353 * Once rec->name is set, rest of the names
1354 * are to be treated as links or aliases.
1355 */
1356 if (rec->name && has_alias) {
1357 if (name_desc[sz] != ',') {
1358 /* No more commas left --> no more
1359 * aliases to take out */
1360 has_alias = 0;
1361 }
1362 concat2(&rec->links, name_desc, sz);
1363 /* XXX This would only happen with a poorly
1364 * written man page, maybe warn? */
1365 if (name_desc[sz] == '\0')
1366 break;
1367 name_desc += sz + 1;
1368 continue;
1369 }
1370 break;
1371 }
1372
1373 /* Parse any escape sequences that might be there */
1374 char *temp = parse_escape(name_desc);
1375 free(rec->name_desc);
1376 rec->name_desc = temp;
1377 temp = parse_escape(rec->name);
1378 free(rec->name);
1379 rec->name = temp;
1380 return;
1381 }
1382
1383 /* The RETURN VALUE section might be specified in multiple ways */
1384 if (strcmp(head->string, "RETURN") == 0 &&
1385 head->next != NULL && head->next->type == ROFFT_TEXT &&
1386 (strcmp(head->next->string, "VALUE") == 0 ||
1387 strcmp(head->next->string, "VALUES") == 0)) {
1388 man_parse_section(MANSEC_RETURN_VALUES, n, rec);
1389 return;
1390 }
1391
1392 /*
1393 * EXIT STATUS section can also be specified all on one line or on two
1394 * separate lines.
1395 */
1396 if (strcmp(head->string, "EXIT") == 0 &&
1397 head->next != NULL && head->next->type == ROFFT_TEXT &&
1398 strcmp(head->next->string, "STATUS") == 0) {
1399 man_parse_section(MANSEC_EXIT_STATUS, n, rec);
1400 return;
1401 }
1402
1403 /* Store the rest of the content in desc. */
1404 man_parse_section(MANSEC_NONE, n, rec);
1405 }
1406
1407 /*
1408 * pman_parse_node --
1409 * Generic function to iterate through a node. Usually called from
1410 * man_parse_section to parse a particular section of the man page.
1411 */
1412 static void
1413 pman_parse_node(const struct roff_node *n, secbuff *s)
1414 {
1415 if (n == NULL)
1416 return;
1417
1418 if (n->type == ROFFT_TEXT)
1419 append(s, n->string);
1420
1421 pman_parse_node(n->child, s);
1422 pman_parse_node(n->next, s);
1423 }
1424
1425 /*
1426 * man_parse_section --
1427 * Takes two parameters:
1428 * sec: Tells which section we are present in
1429 * n: Is the present node of the AST.
1430 * Depending on the section, we call pman_parse_node to parse that section and
1431 * concatenate the content from that section into the buffer for that section.
1432 */
1433 static void
1434 man_parse_section(enum man_sec sec, const struct roff_node *n, mandb_rec *rec)
1435 {
1436 /*
1437 * If the user sepecified the 'l' flag then just parse
1438 * the NAME section, ignore the rest.
1439 */
1440 if (mflags.limit)
1441 return;
1442
1443 switch (sec) {
1444 case MANSEC_LIBRARY:
1445 pman_parse_node(n, &rec->lib);
1446 break;
1447 case MANSEC_RETURN_VALUES:
1448 pman_parse_node(n, &rec->return_vals);
1449 break;
1450 case MANSEC_ENVIRONMENT:
1451 pman_parse_node(n, &rec->env);
1452 break;
1453 case MANSEC_FILES:
1454 pman_parse_node(n, &rec->files);
1455 break;
1456 case MANSEC_EXIT_STATUS:
1457 pman_parse_node(n, &rec->exit_status);
1458 break;
1459 case MANSEC_DIAGNOSTICS:
1460 pman_parse_node(n, &rec->diagnostics);
1461 break;
1462 case MANSEC_ERRORS:
1463 pman_parse_node(n, &rec->errors);
1464 break;
1465 case MANSEC_NAME:
1466 case MANSEC_SYNOPSIS:
1467 case MANSEC_EXAMPLES:
1468 case MANSEC_STANDARDS:
1469 case MANSEC_HISTORY:
1470 case MANSEC_BUGS:
1471 case MANSEC_AUTHORS:
1472 case MANSEC_COPYRIGHT:
1473 break;
1474 default:
1475 pman_parse_node(n, &rec->desc);
1476 break;
1477 }
1478
1479 }
1480
1481 /*
1482 * insert_into_db --
1483 * Inserts the parsed data of the man page in the Sqlite databse.
1484 * If any of the values is NULL, then we cleanup and return -1 indicating
1485 * an error.
1486 * Otherwise, store the data in the database and return 0.
1487 */
1488 static int
1489 insert_into_db(sqlite3 *db, mandb_rec *rec)
1490 {
1491 int rc = 0;
1492 int idx = -1;
1493 const char *sqlstr = NULL;
1494 sqlite3_stmt *stmt = NULL;
1495 char *ln = NULL;
1496 char *errmsg = NULL;
1497 long int mandb_rowid;
1498
1499 /*
1500 * At the very minimum we want to make sure that we store
1501 * the following data:
1502 * Name, one line description, and the MD5 hash
1503 */
1504 if (rec->name == NULL || rec->name_desc == NULL ||
1505 rec->md5_hash == NULL) {
1506 cleanup(rec);
1507 return -1;
1508 }
1509
1510 /* Write null byte at the end of all the sec_buffs */
1511 rec->desc.data[rec->desc.offset] = 0;
1512 rec->lib.data[rec->lib.offset] = 0;
1513 rec->env.data[rec->env.offset] = 0;
1514 rec->return_vals.data[rec->return_vals.offset] = 0;
1515 rec->exit_status.data[rec->exit_status.offset] = 0;
1516 rec->files.data[rec->files.offset] = 0;
1517 rec->diagnostics.data[rec->diagnostics.offset] = 0;
1518 rec->errors.data[rec->errors.offset] = 0;
1519
1520 /*
1521 * In case of a mdoc page: (sorry, no better place to put this code)
1522 * parse the comma separated list of names of man pages,
1523 * the first name will be stored in the mandb table, rest will be
1524 * treated as links and put in the mandb_links table.
1525 */
1526 if (rec->page_type == MDOC) {
1527 char *tmp;
1528 rec->links = estrdup(rec->name);
1529 free(rec->name);
1530 size_t sz = strcspn(rec->links, " \0");
1531 rec->name = emalloc(sz + 1);
1532 memcpy(rec->name, rec->links, sz);
1533 if(rec->name[sz - 1] == ',')
1534 rec->name[sz - 1] = 0;
1535 else
1536 rec->name[sz] = 0;
1537 while (rec->links[sz] == ' ')
1538 ++sz;
1539 tmp = estrdup(rec->links + sz);
1540 free(rec->links);
1541 rec->links = tmp;
1542 }
1543
1544 /*------------------------ Populate the mandb table---------------------------*/
1545 sqlstr = "INSERT INTO mandb VALUES (:section, :name, :name_desc, :desc,"
1546 " :lib, :return_vals, :env, :files, :exit_status,"
1547 " :diagnostics, :errors, :md5_hash, :machine)";
1548
1549 rc = sqlite3_prepare_v2(db, sqlstr, -1, &stmt, NULL);
1550 if (rc != SQLITE_OK)
1551 goto Out;
1552
1553 idx = sqlite3_bind_parameter_index(stmt, ":name");
1554 rc = sqlite3_bind_text(stmt, idx, rec->name, -1, NULL);
1555 if (rc != SQLITE_OK) {
1556 sqlite3_finalize(stmt);
1557 goto Out;
1558 }
1559
1560 idx = sqlite3_bind_parameter_index(stmt, ":section");
1561 rc = sqlite3_bind_text(stmt, idx, rec->section, -1, NULL);
1562 if (rc != SQLITE_OK) {
1563 sqlite3_finalize(stmt);
1564 goto Out;
1565 }
1566
1567 idx = sqlite3_bind_parameter_index(stmt, ":name_desc");
1568 rc = sqlite3_bind_text(stmt, idx, rec->name_desc, -1, NULL);
1569 if (rc != SQLITE_OK) {
1570 sqlite3_finalize(stmt);
1571 goto Out;
1572 }
1573
1574 idx = sqlite3_bind_parameter_index(stmt, ":desc");
1575 rc = sqlite3_bind_text(stmt, idx, rec->desc.data,
1576 rec->desc.offset + 1, NULL);
1577 if (rc != SQLITE_OK) {
1578 sqlite3_finalize(stmt);
1579 goto Out;
1580 }
1581
1582 idx = sqlite3_bind_parameter_index(stmt, ":lib");
1583 rc = sqlite3_bind_text(stmt, idx, rec->lib.data,
1584 rec->lib.offset + 1, NULL);
1585 if (rc != SQLITE_OK) {
1586 sqlite3_finalize(stmt);
1587 goto Out;
1588 }
1589
1590 idx = sqlite3_bind_parameter_index(stmt, ":return_vals");
1591 rc = sqlite3_bind_text(stmt, idx, rec->return_vals.data,
1592 rec->return_vals.offset + 1, NULL);
1593 if (rc != SQLITE_OK) {
1594 sqlite3_finalize(stmt);
1595 goto Out;
1596 }
1597
1598 idx = sqlite3_bind_parameter_index(stmt, ":env");
1599 rc = sqlite3_bind_text(stmt, idx, rec->env.data,
1600 rec->env.offset + 1, NULL);
1601 if (rc != SQLITE_OK) {
1602 sqlite3_finalize(stmt);
1603 goto Out;
1604 }
1605
1606 idx = sqlite3_bind_parameter_index(stmt, ":files");
1607 rc = sqlite3_bind_text(stmt, idx, rec->files.data,
1608 rec->files.offset + 1, NULL);
1609 if (rc != SQLITE_OK) {
1610 sqlite3_finalize(stmt);
1611 goto Out;
1612 }
1613
1614 idx = sqlite3_bind_parameter_index(stmt, ":exit_status");
1615 rc = sqlite3_bind_text(stmt, idx, rec->exit_status.data,
1616 rec->exit_status.offset + 1, NULL);
1617 if (rc != SQLITE_OK) {
1618 sqlite3_finalize(stmt);
1619 goto Out;
1620 }
1621
1622 idx = sqlite3_bind_parameter_index(stmt, ":diagnostics");
1623 rc = sqlite3_bind_text(stmt, idx, rec->diagnostics.data,
1624 rec->diagnostics.offset + 1, NULL);
1625 if (rc != SQLITE_OK) {
1626 sqlite3_finalize(stmt);
1627 goto Out;
1628 }
1629
1630 idx = sqlite3_bind_parameter_index(stmt, ":errors");
1631 rc = sqlite3_bind_text(stmt, idx, rec->errors.data,
1632 rec->errors.offset + 1, NULL);
1633 if (rc != SQLITE_OK) {
1634 sqlite3_finalize(stmt);
1635 goto Out;
1636 }
1637
1638 idx = sqlite3_bind_parameter_index(stmt, ":md5_hash");
1639 rc = sqlite3_bind_text(stmt, idx, rec->md5_hash, -1, NULL);
1640 if (rc != SQLITE_OK) {
1641 sqlite3_finalize(stmt);
1642 goto Out;
1643 }
1644
1645 idx = sqlite3_bind_parameter_index(stmt, ":machine");
1646 if (rec->machine)
1647 rc = sqlite3_bind_text(stmt, idx, rec->machine, -1, NULL);
1648 else
1649 rc = sqlite3_bind_null(stmt, idx);
1650 if (rc != SQLITE_OK) {
1651 sqlite3_finalize(stmt);
1652 goto Out;
1653 }
1654
1655 rc = sqlite3_step(stmt);
1656 if (rc != SQLITE_DONE) {
1657 sqlite3_finalize(stmt);
1658 goto Out;
1659 }
1660
1661 sqlite3_finalize(stmt);
1662
1663 /* Get the row id of the last inserted row */
1664 mandb_rowid = sqlite3_last_insert_rowid(db);
1665
1666 /*------------------------Populate the mandb_meta table-----------------------*/
1667 sqlstr = "INSERT INTO mandb_meta VALUES (:device, :inode, :mtime,"
1668 " :file, :md5_hash, :id)";
1669 rc = sqlite3_prepare_v2(db, sqlstr, -1, &stmt, NULL);
1670 if (rc != SQLITE_OK)
1671 goto Out;
1672
1673 idx = sqlite3_bind_parameter_index(stmt, ":device");
1674 rc = sqlite3_bind_int64(stmt, idx, rec->device);
1675 if (rc != SQLITE_OK) {
1676 sqlite3_finalize(stmt);
1677 goto Out;
1678 }
1679
1680 idx = sqlite3_bind_parameter_index(stmt, ":inode");
1681 rc = sqlite3_bind_int64(stmt, idx, rec->inode);
1682 if (rc != SQLITE_OK) {
1683 sqlite3_finalize(stmt);
1684 goto Out;
1685 }
1686
1687 idx = sqlite3_bind_parameter_index(stmt, ":mtime");
1688 rc = sqlite3_bind_int64(stmt, idx, rec->mtime);
1689 if (rc != SQLITE_OK) {
1690 sqlite3_finalize(stmt);
1691 goto Out;
1692 }
1693
1694 idx = sqlite3_bind_parameter_index(stmt, ":file");
1695 rc = sqlite3_bind_text(stmt, idx, rec->file_path, -1, NULL);
1696 if (rc != SQLITE_OK) {
1697 sqlite3_finalize(stmt);
1698 goto Out;
1699 }
1700
1701 idx = sqlite3_bind_parameter_index(stmt, ":md5_hash");
1702 rc = sqlite3_bind_text(stmt, idx, rec->md5_hash, -1, NULL);
1703 if (rc != SQLITE_OK) {
1704 sqlite3_finalize(stmt);
1705 goto Out;
1706 }
1707
1708 idx = sqlite3_bind_parameter_index(stmt, ":id");
1709 rc = sqlite3_bind_int64(stmt, idx, mandb_rowid);
1710 if (rc != SQLITE_OK) {
1711 sqlite3_finalize(stmt);
1712 goto Out;
1713 }
1714
1715 rc = sqlite3_step(stmt);
1716 sqlite3_finalize(stmt);
1717 if (rc == SQLITE_CONSTRAINT_UNIQUE) {
1718 /* The *most* probable reason for reaching here is that
1719 * the UNIQUE contraint on the file column of the mandb_meta
1720 * table was violated.
1721 * This can happen when a file was updated/modified.
1722 * To fix this we need to do two things:
1723 * 1. Delete the row for the older version of this file
1724 * from mandb table.
1725 * 2. Run an UPDATE query to update the row for this file
1726 * in the mandb_meta table.
1727 */
1728 warnx("Trying to update index for %s", rec->file_path);
1729 char *sql = sqlite3_mprintf("DELETE FROM mandb "
1730 "WHERE rowid = (SELECT id"
1731 " FROM mandb_meta"
1732 " WHERE file = %Q)",
1733 rec->file_path);
1734 sqlite3_exec(db, sql, NULL, NULL, &errmsg);
1735 sqlite3_free(sql);
1736 if (errmsg != NULL) {
1737 if (mflags.verbosity)
1738 warnx("%s", errmsg);
1739 free(errmsg);
1740 }
1741 sqlstr = "UPDATE mandb_meta SET device = :device,"
1742 " inode = :inode, mtime = :mtime, id = :id,"
1743 " md5_hash = :md5 WHERE file = :file";
1744 rc = sqlite3_prepare_v2(db, sqlstr, -1, &stmt, NULL);
1745 if (rc != SQLITE_OK) {
1746 if (mflags.verbosity)
1747 warnx("Update failed with error: %s",
1748 sqlite3_errmsg(db));
1749 close_db(db);
1750 cleanup(rec);
1751 errx(EXIT_FAILURE,
1752 "Consider running makemandb with -f option");
1753 }
1754
1755 idx = sqlite3_bind_parameter_index(stmt, ":device");
1756 sqlite3_bind_int64(stmt, idx, rec->device);
1757 idx = sqlite3_bind_parameter_index(stmt, ":inode");
1758 sqlite3_bind_int64(stmt, idx, rec->inode);
1759 idx = sqlite3_bind_parameter_index(stmt, ":mtime");
1760 sqlite3_bind_int64(stmt, idx, rec->mtime);
1761 idx = sqlite3_bind_parameter_index(stmt, ":id");
1762 sqlite3_bind_int64(stmt, idx, mandb_rowid);
1763 idx = sqlite3_bind_parameter_index(stmt, ":md5");
1764 sqlite3_bind_text(stmt, idx, rec->md5_hash, -1, NULL);
1765 idx = sqlite3_bind_parameter_index(stmt, ":file");
1766 sqlite3_bind_text(stmt, idx, rec->file_path, -1, NULL);
1767 rc = sqlite3_step(stmt);
1768 sqlite3_finalize(stmt);
1769
1770 if (rc != SQLITE_DONE) {
1771 if (mflags.verbosity)
1772 warnx("%s", sqlite3_errmsg(db));
1773 close_db(db);
1774 cleanup(rec);
1775 errx(EXIT_FAILURE,
1776 "Consider running makemandb with -f option");
1777 }
1778 } else if (rc != SQLITE_DONE) {
1779 /* Otherwise make this error fatal */
1780 warnx("Failed at %s\n%s", rec->file_path, sqlite3_errmsg(db));
1781 cleanup(rec);
1782 close_db(db);
1783 exit(EXIT_FAILURE);
1784 }
1785
1786 /*------------------------ Populate the mandb_links table---------------------*/
1787 char *str = NULL;
1788 char *links;
1789 if (rec->links && strlen(rec->links)) {
1790 links = rec->links;
1791 for(ln = strtok(links, " "); ln; ln = strtok(NULL, " ")) {
1792 if (ln[0] == ',')
1793 ln++;
1794 if(ln[strlen(ln) - 1] == ',')
1795 ln[strlen(ln) - 1] = 0;
1796
1797 str = sqlite3_mprintf("INSERT INTO mandb_links"
1798 " VALUES (%Q, %Q, %Q, %Q, %Q)",
1799 ln, rec->name, rec->section,
1800 rec->machine, rec->md5_hash);
1801 sqlite3_exec(db, str, NULL, NULL, &errmsg);
1802 sqlite3_free(str);
1803 if (errmsg != NULL) {
1804 warnx("%s", errmsg);
1805 cleanup(rec);
1806 free(errmsg);
1807 return -1;
1808 }
1809 }
1810 }
1811
1812 cleanup(rec);
1813 return 0;
1814
1815 Out:
1816 if (mflags.verbosity)
1817 warnx("%s", sqlite3_errmsg(db));
1818 cleanup(rec);
1819 return -1;
1820 }
1821
1822 /*
1823 * check_md5--
1824 * Generates the md5 hash of the file and checks if it already doesn't exist
1825 * in the table.
1826 * This function is being used to avoid hardlinks.
1827 * On successful completion it will also set the value of the fourth parameter
1828 * to the md5 hash of the file (computed previously). It is the responsibility
1829 * of the caller to free this buffer.
1830 * Return values:
1831 * -1: If an error occurs somewhere and sets the md5 return buffer to NULL.
1832 * 0: If the md5 hash does not exist in the table.
1833 * 1: If the hash exists in the database.
1834 */
1835 static int
1836 check_md5(const char *file, sqlite3 *db, char **md5, void *buf, size_t buflen)
1837 {
1838 int rc = 0;
1839 int idx = -1;
1840 char *sqlstr = NULL;
1841 char *mymd5;
1842 sqlite3_stmt *stmt = NULL;
1843 *md5 = NULL;
1844
1845 assert(file != NULL);
1846 if ((mymd5 = MD5Data(buf, buflen, NULL)) == NULL) {
1847 if (mflags.verbosity)
1848 warn("md5 failed: %s", file);
1849 return -1;
1850 }
1851
1852 easprintf(&sqlstr, "SELECT * FROM mandb_meta WHERE md5_hash = :md5_hash");
1853 rc = sqlite3_prepare_v2(db, sqlstr, -1, &stmt, NULL);
1854 if (rc != SQLITE_OK) {
1855 free(sqlstr);
1856 free(mymd5);
1857 return -1;
1858 }
1859
1860 idx = sqlite3_bind_parameter_index(stmt, ":md5_hash");
1861 rc = sqlite3_bind_text(stmt, idx, mymd5, -1, NULL);
1862 if (rc != SQLITE_OK) {
1863 if (mflags.verbosity)
1864 warnx("%s", sqlite3_errmsg(db));
1865 sqlite3_finalize(stmt);
1866 free(sqlstr);
1867 free(mymd5);
1868 return -1;
1869 }
1870
1871 *md5 = mymd5;
1872 if (sqlite3_step(stmt) == SQLITE_ROW) {
1873 sqlite3_finalize(stmt);
1874 free(sqlstr);
1875 return 0;
1876 }
1877
1878 sqlite3_finalize(stmt);
1879 free(sqlstr);
1880 return 1;
1881 }
1882
1883 /* Optimize the index for faster search */
1884 static void
1885 optimize(sqlite3 *db)
1886 {
1887 const char *sqlstr;
1888 char *errmsg = NULL;
1889
1890 if (mflags.verbosity == 2)
1891 printf("Optimizing the database index\n");
1892 sqlstr = "INSERT INTO mandb(mandb) VALUES (\'optimize\');"
1893 "VACUUM";
1894 sqlite3_exec(db, sqlstr, NULL, NULL, &errmsg);
1895 if (errmsg != NULL) {
1896 if (mflags.verbosity)
1897 warnx("%s", errmsg);
1898 free(errmsg);
1899 return;
1900 }
1901 }
1902
1903 /*
1904 * cleanup --
1905 * cleans up the global buffers
1906 */
1907 static void
1908 cleanup(mandb_rec *rec)
1909 {
1910 rec->desc.offset = 0;
1911 rec->lib.offset = 0;
1912 rec->return_vals.offset = 0;
1913 rec->env.offset = 0;
1914 rec->exit_status.offset = 0;
1915 rec->diagnostics.offset = 0;
1916 rec->errors.offset = 0;
1917 rec->files.offset = 0;
1918
1919 free(rec->machine);
1920 rec->machine = NULL;
1921
1922 free(rec->links);
1923 rec->links = NULL;
1924
1925 free(rec->file_path);
1926 rec->file_path = NULL;
1927
1928 free(rec->name);
1929 rec->name = NULL;
1930
1931 free(rec->name_desc);
1932 rec->name_desc = NULL;
1933
1934 free(rec->md5_hash);
1935 rec->md5_hash = NULL;
1936
1937 free(rec->section);
1938 rec->section = NULL;
1939 }
1940
1941 /*
1942 * init_secbuffs--
1943 * Sets the value of buflen for all the sec_buff field of rec. And then
1944 * allocate memory to each sec_buff member of rec.
1945 */
1946 static void
1947 init_secbuffs(mandb_rec *rec)
1948 {
1949 /*
1950 * Some sec_buff might need more memory, for example desc,
1951 * which stores the data of the DESCRIPTION section,
1952 * while some might need very small amount of memory.
1953 * Therefore explicitly setting the value of buflen field for
1954 * each sec_buff.
1955 */
1956 rec->desc.buflen = 10 * BUFLEN;
1957 rec->desc.data = emalloc(rec->desc.buflen);
1958 rec->desc.offset = 0;
1959
1960 rec->lib.buflen = BUFLEN / 2;
1961 rec->lib.data = emalloc(rec->lib.buflen);
1962 rec->lib.offset = 0;
1963
1964 rec->return_vals.buflen = BUFLEN;
1965 rec->return_vals.data = emalloc(rec->return_vals.buflen);
1966 rec->return_vals.offset = 0;
1967
1968 rec->exit_status.buflen = BUFLEN;
1969 rec->exit_status.data = emalloc(rec->exit_status.buflen);
1970 rec->exit_status.offset = 0;
1971
1972 rec->env.buflen = BUFLEN;
1973 rec->env.data = emalloc(rec->env.buflen);
1974 rec->env.offset = 0;
1975
1976 rec->files.buflen = BUFLEN;
1977 rec->files.data = emalloc(rec->files.buflen);
1978 rec->files.offset = 0;
1979
1980 rec->diagnostics.buflen = BUFLEN;
1981 rec->diagnostics.data = emalloc(rec->diagnostics.buflen);
1982 rec->diagnostics.offset = 0;
1983
1984 rec->errors.buflen = BUFLEN;
1985 rec->errors.data = emalloc(rec->errors.buflen);
1986 rec->errors.offset = 0;
1987 }
1988
1989 /*
1990 * free_secbuffs--
1991 * This function should be called at the end, when all the pages have been
1992 * parsed.
1993 * It frees the memory allocated to sec_buffs by init_secbuffs in the starting.
1994 */
1995 static void
1996 free_secbuffs(mandb_rec *rec)
1997 {
1998 free(rec->desc.data);
1999 free(rec->lib.data);
2000 free(rec->return_vals.data);
2001 free(rec->exit_status.data);
2002 free(rec->env.data);
2003 free(rec->files.data);
2004 free(rec->diagnostics.data);
2005 free(rec->errors.data);
2006 }
2007
2008 static void
2009 replace_hyph(char *str)
2010 {
2011 char *iter = str;
2012 while ((iter = strchr(iter, ASCII_HYPH)) != NULL)
2013 *iter = '-';
2014
2015 iter = str;
2016 while ((iter = strchr(iter, ASCII_NBRSP)) != NULL)
2017 *iter = '-';
2018 }
2019
2020 static char *
2021 parse_escape(const char *str)
2022 {
2023 const char *backslash, *last_backslash;
2024 char *result, *iter;
2025 size_t len;
2026
2027 assert(str);
2028
2029 last_backslash = str;
2030 backslash = strchr(str, '\\');
2031 if (backslash == NULL) {
2032 result = estrdup(str);
2033 replace_hyph(result);
2034 return result;
2035 }
2036
2037 result = emalloc(strlen(str) + 1);
2038 iter = result;
2039
2040 do {
2041 len = backslash - last_backslash;
2042 memcpy(iter, last_backslash, len);
2043 iter += len;
2044 if (backslash[1] == '-' || backslash[1] == ' ') {
2045 *iter++ = backslash[1];
2046 last_backslash = backslash + 2;
2047 backslash = strchr(last_backslash, '\\');
2048 } else {
2049 ++backslash;
2050 mandoc_escape(&backslash, NULL, NULL);
2051 last_backslash = backslash;
2052 if (backslash == NULL)
2053 break;
2054 backslash = strchr(last_backslash, '\\');
2055 }
2056 } while (backslash != NULL);
2057 if (last_backslash != NULL)
2058 strcpy(iter, last_backslash);
2059
2060 replace_hyph(result);
2061 return result;
2062 }
2063
2064 /*
2065 * append--
2066 * Concatenates a space and src at the end of sbuff->data (much like concat in
2067 * apropos-utils.c).
2068 * Rather than reallocating space for writing data, it uses the value of the
2069 * offset field of sec_buff to write new data at the free space left in the
2070 * buffer.
2071 * In case the size of the data to be appended exceeds the number of bytes left
2072 * in the buffer, it reallocates buflen number of bytes and then continues.
2073 * Value of offset field should be adjusted as new data is written.
2074 *
2075 * NOTE: This function does not write the null byte at the end of the buffers,
2076 * write a null byte at the position pointed to by offset before inserting data
2077 * in the db.
2078 */
2079 static void
2080 append(secbuff *sbuff, const char *src)
2081 {
2082 short flag = 0;
2083 size_t srclen, newlen;
2084 char *temp;
2085
2086 assert(src != NULL);
2087 temp = parse_escape(src);
2088 srclen = strlen(temp);
2089
2090 if (sbuff->data == NULL) {
2091 sbuff->data = emalloc(sbuff->buflen);
2092 sbuff->offset = 0;
2093 }
2094
2095 newlen = sbuff->offset + srclen + 2;
2096 if (newlen >= sbuff->buflen) {
2097 while (sbuff->buflen < newlen)
2098 sbuff->buflen += sbuff->buflen;
2099 sbuff->data = erealloc(sbuff->data, sbuff->buflen);
2100 flag = 1;
2101 }
2102
2103 /* Append a space at the end of the buffer. */
2104 if (sbuff->offset || flag)
2105 sbuff->data[sbuff->offset++] = ' ';
2106 /* Now, copy src at the end of the buffer. */
2107 memcpy(sbuff->data + sbuff->offset, temp, srclen);
2108 sbuff->offset += srclen;
2109 free(temp);
2110 }
2111
2112 static void
2113 usage(void)
2114 {
2115 fprintf(stderr, "Usage: %s [-floQqv] [-C path]\n", getprogname());
2116 exit(1);
2117 }
2118