makemandb.c revision 1.45 1 /* $NetBSD: makemandb.c,v 1.45 2016/12/17 17:04:38 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 1.45 2016/12/17 17:04:38 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 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, *parent;
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 parent = estrdup(line);
454 char *pdir = estrdup(dirname(parent));
455 free(parent);
456 /* Traverse the man page directories and parse the pages */
457 traversedir(pdir, line, db, mp);
458 free(pdir);
459 }
460 free(line);
461
462 if (pclose(file) == -1) {
463 close_db(db);
464 cleanup(&rec);
465 free_secbuffs(&rec);
466 err(EXIT_FAILURE, "pclose error");
467 }
468
469 if (mflags.verbosity)
470 printf("Performing index update\n");
471 update_db(db, mp, &rec);
472 mparse_free(mp);
473 mchars_free();
474 free_secbuffs(&rec);
475
476 /* Commit the transaction */
477 sqlite3_exec(db, "COMMIT", NULL, NULL, &errmsg);
478 if (errmsg != NULL) {
479 warnx("%s", errmsg);
480 free(errmsg);
481 close_db(db);
482 exit(EXIT_FAILURE);
483 }
484
485 if (mflags.optimize)
486 optimize(db);
487
488 close_db(db);
489 return 0;
490 }
491
492 /*
493 * traversedir --
494 * Traverses the given directory recursively and passes all the man page files
495 * in the way to build_file_cache()
496 */
497 static void
498 traversedir(const char *parent, const char *file, sqlite3 *db,
499 struct mparse *mp)
500 {
501 struct stat sb;
502 struct dirent *dirp;
503 DIR *dp;
504 char *buf;
505
506 if (stat(file, &sb) < 0) {
507 if (mflags.verbosity)
508 warn("stat failed: %s", file);
509 return;
510 }
511
512 /* If it is a directory, traverse it recursively */
513 if (S_ISDIR(sb.st_mode)) {
514 if ((dp = opendir(file)) == NULL) {
515 if (mflags.verbosity)
516 warn("opendir error: %s", file);
517 return;
518 }
519
520 while ((dirp = readdir(dp)) != NULL) {
521 /* Avoid . and .. entries in a directory */
522 if (strncmp(dirp->d_name, ".", 1)) {
523 easprintf(&buf, "%s/%s", file, dirp->d_name);
524 traversedir(parent, buf, db, mp);
525 free(buf);
526 }
527 }
528 closedir(dp);
529 return;
530 }
531
532 if (!S_ISREG(sb.st_mode))
533 return;
534
535 if (sb.st_size == 0) {
536 if (mflags.verbosity)
537 warnx("Empty file: %s", file);
538 return;
539 }
540 build_file_cache(db, parent, file, &sb);
541 }
542
543 /* build_file_cache --
544 * This function generates an md5 hash of the file passed as its 2nd parameter
545 * and stores it in a temporary table file_cache along with the full file path.
546 * This is done to support incremental updation of the database.
547 * The temporary table file_cache is dropped thereafter in the function
548 * update_db(), once the database has been updated.
549 */
550 static void
551 build_file_cache(sqlite3 *db, const char *parent, const char *file,
552 struct stat *sb)
553 {
554 const char *sqlstr;
555 sqlite3_stmt *stmt = NULL;
556 int rc, idx;
557 assert(file != NULL);
558 dev_t device_cache = sb->st_dev;
559 ino_t inode_cache = sb->st_ino;
560 time_t mtime_cache = sb->st_mtime;
561
562 sqlstr = "INSERT INTO metadb.file_cache VALUES (:device, :inode,"
563 " :mtime, :parent, :file)";
564 rc = sqlite3_prepare_v2(db, sqlstr, -1, &stmt, NULL);
565 if (rc != SQLITE_OK) {
566 if (mflags.verbosity)
567 warnx("%s", sqlite3_errmsg(db));
568 return;
569 }
570
571 idx = sqlite3_bind_parameter_index(stmt, ":device");
572 rc = sqlite3_bind_int64(stmt, idx, device_cache);
573 if (rc != SQLITE_OK) {
574 if (mflags.verbosity)
575 warnx("%s", sqlite3_errmsg(db));
576 sqlite3_finalize(stmt);
577 return;
578 }
579
580 idx = sqlite3_bind_parameter_index(stmt, ":inode");
581 rc = sqlite3_bind_int64(stmt, idx, inode_cache);
582 if (rc != SQLITE_OK) {
583 if (mflags.verbosity)
584 warnx("%s", sqlite3_errmsg(db));
585 sqlite3_finalize(stmt);
586 return;
587 }
588
589 idx = sqlite3_bind_parameter_index(stmt, ":mtime");
590 rc = sqlite3_bind_int64(stmt, idx, mtime_cache);
591 if (rc != SQLITE_OK) {
592 if (mflags.verbosity)
593 warnx("%s", sqlite3_errmsg(db));
594 sqlite3_finalize(stmt);
595 return;
596 }
597
598 idx = sqlite3_bind_parameter_index(stmt, ":parent");
599 rc = sqlite3_bind_text(stmt, idx, parent, -1, NULL);
600 if (rc != SQLITE_OK) {
601 if (mflags.verbosity)
602 warnx("%s", sqlite3_errmsg(db));
603 sqlite3_finalize(stmt);
604 return;
605 }
606
607 idx = sqlite3_bind_parameter_index(stmt, ":file");
608 rc = sqlite3_bind_text(stmt, idx, file, -1, NULL);
609 if (rc != SQLITE_OK) {
610 if (mflags.verbosity)
611 warnx("%s", sqlite3_errmsg(db));
612 sqlite3_finalize(stmt);
613 return;
614 }
615
616 sqlite3_step(stmt);
617 sqlite3_finalize(stmt);
618 }
619
620 static void
621 update_existing_entry(sqlite3 *db, const char *file, const char *hash,
622 mandb_rec *rec, int *new_count, int *link_count, int *err_count)
623 {
624 int update_count, rc, idx;
625 const char *inner_sqlstr;
626 sqlite3_stmt *inner_stmt;
627
628 update_count = sqlite3_total_changes(db);
629 inner_sqlstr = "UPDATE mandb_meta SET device = :device,"
630 " inode = :inode, mtime = :mtime WHERE"
631 " md5_hash = :md5 AND file = :file AND"
632 " (device <> :device2 OR inode <> "
633 " :inode2 OR mtime <> :mtime2)";
634 rc = sqlite3_prepare_v2(db, inner_sqlstr, -1, &inner_stmt, NULL);
635 if (rc != SQLITE_OK) {
636 if (mflags.verbosity)
637 warnx("%s", sqlite3_errmsg(db));
638 return;
639 }
640 idx = sqlite3_bind_parameter_index(inner_stmt, ":device");
641 sqlite3_bind_int64(inner_stmt, idx, rec->device);
642 idx = sqlite3_bind_parameter_index(inner_stmt, ":inode");
643 sqlite3_bind_int64(inner_stmt, idx, rec->inode);
644 idx = sqlite3_bind_parameter_index(inner_stmt, ":mtime");
645 sqlite3_bind_int64(inner_stmt, idx, rec->mtime);
646 idx = sqlite3_bind_parameter_index(inner_stmt, ":md5");
647 sqlite3_bind_text(inner_stmt, idx, hash, -1, NULL);
648 idx = sqlite3_bind_parameter_index(inner_stmt, ":file");
649 sqlite3_bind_text(inner_stmt, idx, file, -1, NULL);
650 idx = sqlite3_bind_parameter_index(inner_stmt, ":device2");
651 sqlite3_bind_int64(inner_stmt, idx, rec->device);
652 idx = sqlite3_bind_parameter_index(inner_stmt, ":inode2");
653 sqlite3_bind_int64(inner_stmt, idx, rec->inode);
654 idx = sqlite3_bind_parameter_index(inner_stmt, ":mtime2");
655 sqlite3_bind_int64(inner_stmt, idx, rec->mtime);
656
657 rc = sqlite3_step(inner_stmt);
658 if (rc == SQLITE_DONE) {
659 /* Check if an update has been performed. */
660 if (update_count != sqlite3_total_changes(db)) {
661 if (mflags.verbosity == 2)
662 printf("Updated %s\n", file);
663 (*new_count)++;
664 } else {
665 /* Otherwise it was a hardlink. */
666 (*link_count)++;
667 }
668 } else {
669 if (mflags.verbosity == 2)
670 warnx("Could not update the meta data for %s", file);
671 (*err_count)++;
672 }
673 sqlite3_finalize(inner_stmt);
674 }
675
676 /* read_and_decompress --
677 * Reads the given file into memory. If it is compressed, decompress
678 * it before returning to the caller.
679 */
680 static int
681 read_and_decompress(const char *file, void **bufp, size_t *len)
682 {
683 size_t off;
684 ssize_t r;
685 struct archive *a;
686 struct archive_entry *ae;
687 char *buf;
688
689 if ((a = archive_read_new()) == NULL)
690 errx(EXIT_FAILURE, "memory allocation failed");
691
692 *bufp = NULL;
693 if (archive_read_support_compression_all(a) != ARCHIVE_OK ||
694 archive_read_support_format_raw(a) != ARCHIVE_OK ||
695 archive_read_open_filename(a, file, 65536) != ARCHIVE_OK ||
696 archive_read_next_header(a, &ae) != ARCHIVE_OK)
697 goto archive_error;
698 *len = 65536;
699 buf = emalloc(*len);
700 off = 0;
701 for (;;) {
702 r = archive_read_data(a, buf + off, *len - off);
703 if (r == ARCHIVE_OK) {
704 archive_read_finish(a);
705 *bufp = buf;
706 *len = off;
707 return 0;
708 }
709 if (r <= 0) {
710 free(buf);
711 break;
712 }
713 off += r;
714 if (off == *len) {
715 *len *= 2;
716 if (*len < off) {
717 if (mflags.verbosity)
718 warnx("File too large: %s", file);
719 free(buf);
720 archive_read_finish(a);
721 return -1;
722 }
723 buf = erealloc(buf, *len);
724 }
725 }
726
727 archive_error:
728 warnx("Error while reading `%s': %s", file, archive_error_string(a));
729 archive_read_finish(a);
730 return -1;
731 }
732
733 /* update_db --
734 * Does an incremental updation of the database by checking the file_cache.
735 * It parses and adds the pages which are present in file_cache,
736 * but not in the database.
737 * It also removes the pages which are present in the databse,
738 * but not in the file_cache.
739 */
740 static void
741 update_db(sqlite3 *db, struct mparse *mp, mandb_rec *rec)
742 {
743 const char *sqlstr;
744 sqlite3_stmt *stmt = NULL;
745 char *file;
746 char *parent;
747 char *errmsg = NULL;
748 char *md5sum;
749 void *buf;
750 size_t buflen;
751 struct sql_row {
752 struct sql_row *next;
753 dev_t device;
754 ino_t inode;
755 time_t mtime;
756 char *parent;
757 char *file;
758 } *rows, *row;
759 int new_count = 0; /* Counter for newly indexed/updated pages */
760 int total_count = 0; /* Counter for total number of pages */
761 int err_count = 0; /* Counter for number of failed pages */
762 int link_count = 0; /* Counter for number of hard/sym links */
763 int md5_status;
764 int rc;
765
766 sqlstr = "SELECT device, inode, mtime, parent, file"
767 " FROM metadb.file_cache fc"
768 " WHERE NOT EXISTS(SELECT 1 FROM mandb_meta WHERE"
769 " device = fc.device AND inode = fc.inode AND "
770 " mtime = fc.mtime AND file = fc.file)";
771
772 rc = sqlite3_prepare_v2(db, sqlstr, -1, &stmt, NULL);
773 if (rc != SQLITE_OK) {
774 if (mflags.verbosity)
775 warnx("%s", sqlite3_errmsg(db));
776 close_db(db);
777 errx(EXIT_FAILURE, "Could not query file cache");
778 }
779
780 buf = NULL;
781 rows = NULL;
782 while (sqlite3_step(stmt) == SQLITE_ROW) {
783 row = emalloc(sizeof(struct sql_row));
784 row->device = sqlite3_column_int64(stmt, 0);
785 row->inode = sqlite3_column_int64(stmt, 1);
786 row->mtime = sqlite3_column_int64(stmt, 2);
787 row->parent = estrdup((const char *) sqlite3_column_text(stmt, 3));
788 row->file = estrdup((const char *) sqlite3_column_text(stmt, 4));
789 row->next = rows;
790 rows = row;
791 total_count++;
792 }
793 sqlite3_finalize(stmt);
794
795 for ( ; rows != NULL; free(parent), free(file), free(buf)) {
796 row = rows;
797 rows = rows->next;
798
799 rec->device = row->device;
800 rec->inode = row->inode;
801 rec->mtime = row->mtime;
802 parent = row->parent;
803 file = row->file;
804 free(row);
805
806 if (read_and_decompress(file, &buf, &buflen)) {
807 err_count++;
808 continue;
809 }
810 md5_status = check_md5(file, db, "mandb_meta", &md5sum, buf,
811 buflen);
812 assert(md5sum != NULL);
813 if (md5_status == -1) {
814 if (mflags.verbosity)
815 warnx("An error occurred in checking md5 value"
816 " for file %s", file);
817 err_count++;
818 continue;
819 }
820
821 if (md5_status == 0) {
822 /*
823 * The MD5 hash is already present in the database,
824 * so simply update the metadata.
825 */
826 update_existing_entry(db, file, md5sum, rec,
827 &new_count, &link_count, &err_count);
828 free(md5sum);
829 continue;
830 }
831
832 if (md5_status == 1) {
833 /*
834 * The MD5 hash was not present in the database.
835 * This means is either a new file or an updated file.
836 * We should go ahead with parsing.
837 */
838 if (chdir(parent) == -1) {
839 if (mflags.verbosity)
840 warn("chdir failed for `%s', could "
841 "not index `%s'", parent, file);
842 err_count++;
843 free(md5sum);
844 continue;
845 }
846
847 if (mflags.verbosity == 2)
848 printf("Parsing: %s\n", file);
849 rec->md5_hash = md5sum;
850 rec->file_path = estrdup(file);
851 // file_path is freed by insert_into_db itself.
852 begin_parse(file, mp, rec, buf, buflen);
853 if (insert_into_db(db, rec) < 0) {
854 if (mflags.verbosity)
855 warnx("Error in indexing `%s'", file);
856 err_count++;
857 } else {
858 new_count++;
859 }
860 }
861 }
862
863 if (mflags.verbosity == 2) {
864 printf("Number of new or updated pages encountered: %d\n"
865 "Number of hard links found: %d\n"
866 "Number of pages that were successfully"
867 " indexed or updated: %d\n"
868 "Number of pages that could not be indexed"
869 " due to errors: %d\n",
870 total_count - link_count, link_count, new_count, err_count);
871 }
872
873 if (mflags.recreate)
874 return;
875
876 if (mflags.verbosity == 2)
877 printf("Deleting stale index entries\n");
878
879 sqlstr = "DELETE FROM mandb_meta WHERE file NOT IN"
880 " (SELECT file FROM metadb.file_cache);"
881 "DELETE FROM mandb_links WHERE md5_hash NOT IN"
882 " (SELECT md5_hash from mandb_meta);"
883 "DROP TABLE metadb.file_cache;"
884 "DELETE FROM mandb WHERE rowid NOT IN"
885 " (SELECT id FROM mandb_meta);";
886
887 sqlite3_exec(db, sqlstr, NULL, NULL, &errmsg);
888 if (errmsg != NULL) {
889 warnx("Removing old entries failed: %s", errmsg);
890 warnx("Please rebuild database from scratch with -f.");
891 free(errmsg);
892 return;
893 }
894 }
895
896 /*
897 * begin_parse --
898 * parses the man page using libmandoc
899 */
900 static void
901 begin_parse(const char *file, struct mparse *mp, mandb_rec *rec,
902 const void *buf, size_t len)
903 {
904 struct roff_man *roff;
905 mparse_reset(mp);
906
907 rec->xr_found = 0;
908
909 if (mparse_readmem(mp, buf, len, file) >= MANDOCLEVEL_BADARG) {
910 /* Printing this warning at verbosity level 2
911 * because some packages from pkgsrc might trigger several
912 * of such warnings.
913 */
914 if (mflags.verbosity == 2)
915 warnx("%s: Parse failure", file);
916 return;
917 }
918
919 mparse_result(mp, &roff, NULL);
920 if (roff == NULL) {
921 if (mflags.verbosity == 2)
922 warnx("Not a roff(7) page");
923 return;
924 }
925
926 if (roff->macroset == MACROSET_MDOC) {
927 mdoc_validate(roff);
928 rec->page_type = MDOC;
929 proff_node(roff->first->child, rec, mdocs);
930 } else if (roff->macroset == MACROSET_MAN) {
931 man_validate(roff);
932 rec->page_type = MAN;
933 proff_node(roff->first->child, rec, mans);
934 } else
935 warnx("Unknown macroset %d", roff->macroset);
936 set_machine(roff, rec);
937 set_section(roff, rec);
938 }
939
940 /*
941 * set_section --
942 * Extracts the section number and normalizes it to only the numeric part
943 * (Which should be the first character of the string).
944 */
945 static void
946 set_section(const struct roff_man *rm, mandb_rec *rec)
947 {
948 if (!rm)
949 return;
950 const struct roff_meta *rm_meta = &rm->meta;
951 const char *s = rm_meta->msec == NULL ? "?" : rm_meta->msec;
952 easprintf(&rec->section, "%s", s);
953 if (rec->section[0] == '?' && mflags.verbosity == 2)
954 warnx("%s: Missing section number", rec->file_path);
955 }
956
957 /*
958 * get_machine --
959 * Extracts the machine architecture information if available.
960 */
961 static void
962 set_machine(const struct roff_man *rm, mandb_rec *rec)
963 {
964 if (rm == NULL)
965 return;
966 const struct roff_meta *rm_meta = &rm->meta;
967 if (rm_meta->arch)
968 rec->machine = estrdup(rm_meta->arch);
969 }
970
971 /*
972 * pmdoc_Nm --
973 * Extracts the Name of the manual page from the .Nm macro
974 */
975 static void
976 pmdoc_Nm(const struct roff_node *n, mandb_rec *rec)
977 {
978 if (n->sec != SEC_NAME)
979 return;
980
981 for (n = n->child; n; n = n->next) {
982 if (n->type == ROFFT_TEXT) {
983 char *escaped_name = parse_escape(n->string);
984 concat(&rec->name, escaped_name);
985 free(escaped_name);
986 }
987 }
988 }
989
990 /*
991 * pmdoc_Nd --
992 * Extracts the one line description of the man page from the .Nd macro
993 */
994 static void
995 pmdoc_Nd(const struct roff_node *n, mandb_rec *rec)
996 {
997 if (n->type == ROFFT_BODY)
998 deroff(&rec->name_desc, n);
999
1000 }
1001
1002 /*
1003 * pmdoc_macro_handler--
1004 * This function is a single point of handling all the special macros that we
1005 * want to handle especially. For example the .Xr macro for properly parsing
1006 * the referenced page name along with the section number, or the .Pp macro
1007 * for adding a new line whenever we encounter it.
1008 */
1009 static void
1010 pmdoc_macro_handler(const struct roff_node *n, mandb_rec *rec, int doct)
1011 {
1012 const struct roff_node *sn;
1013 assert(n);
1014
1015 switch (doct) {
1016 /* Parse the man page references.
1017 * Basically the .Xr macros are used like:
1018 * .Xr ls 1
1019 * and formatted like this:
1020 * ls(1)
1021 * Prepare a buffer to format the data like the above example and call
1022 * pmdoc_parse_section to append it.
1023 */
1024 case MDOC_Xr:
1025 n = n->child;
1026 while (n->type != ROFFT_TEXT && n->next)
1027 n = n->next;
1028
1029 if (n && n->type != ROFFT_TEXT)
1030 return;
1031 sn = n;
1032 if (n->next)
1033 n = n->next;
1034
1035 while (n->type != ROFFT_TEXT && n->next)
1036 n = n->next;
1037
1038 if (n && n->type == ROFFT_TEXT) {
1039 char *buf;
1040 easprintf(&buf, "%s(%s)", sn->string, n->string);
1041 mdoc_parse_section(n->sec, buf, rec);
1042 free(buf);
1043 }
1044
1045 break;
1046
1047 /* Parse the .Pp macro to add a new line */
1048 case MDOC_Pp:
1049 if (n->type == ROFFT_TEXT)
1050 mdoc_parse_section(n->sec, "\n", rec);
1051 break;
1052 default:
1053 break;
1054 }
1055
1056 }
1057
1058 /*
1059 * pmdoc_Xr, pmdoc_Pp--
1060 * Empty stubs.
1061 * The parser calls these functions each time it encounters
1062 * a .Xr or .Pp macro. We are parsing all the data from
1063 * the pmdoc_Sh function, so don't do anything here.
1064 * (See if else blocks in pmdoc_Sh.)
1065 */
1066 static void
1067 pmdoc_Xr(const struct roff_node *n, mandb_rec *rec)
1068 {
1069 }
1070
1071 static void
1072 pmdoc_Pp(const struct roff_node *n, mandb_rec *rec)
1073 {
1074 }
1075
1076 /*
1077 * pmdoc_Sh --
1078 * Called when a .Sh macro is encountered and tries to parse its body
1079 */
1080 static void
1081 pmdoc_Sh(const struct roff_node *n, mandb_rec *rec)
1082 {
1083 if (n == NULL)
1084 return;
1085
1086 switch (n->sec) {
1087 case SEC_NAME:
1088 case SEC_SYNOPSIS:
1089 case SEC_EXAMPLES:
1090 case SEC_STANDARDS:
1091 case SEC_HISTORY:
1092 case SEC_AUTHORS:
1093 case SEC_BUGS:
1094 /*
1095 * We don't care about text from these sections
1096 */
1097 return;
1098 default:
1099 break;
1100 }
1101
1102 if (n->type == ROFFT_BLOCK)
1103 mdoc_parse_Sh(n->body, rec);
1104 }
1105
1106 /*
1107 * Called from pmdoc_Sh to parse body of a .Sh macro. It calls
1108 * mdoc_parse_section to append the data to the section specific buffer.
1109 * Two special macros which may occur inside the body of Sh are .Nm and .Xr and
1110 * they need special handling, thus the separate if branches for them.
1111 */
1112 static void
1113 mdoc_parse_Sh(const struct roff_node *n, mandb_rec *rec)
1114 {
1115 if (n == NULL || (n->type != ROFFT_TEXT && n->tok == MDOC_MAX))
1116 return;
1117 int xr_found = 0;
1118
1119 if (n->type == ROFFT_TEXT) {
1120 mdoc_parse_section(n->sec, n->string, rec);
1121 } else if (mdocs[n->tok] == pmdoc_Nm && rec->name != NULL) {
1122 /*
1123 * When encountering a .Nm macro, substitute it
1124 * with its previously cached value of the argument.
1125 */
1126 mdoc_parse_section(n->sec, rec->name, rec);
1127 } else if (mdocs[n->tok] == pmdoc_Xr) {
1128 /*
1129 * When encountering other inline macros,
1130 * call pmdoc_macro_handler.
1131 */
1132 pmdoc_macro_handler(n, rec, MDOC_Xr);
1133 xr_found = 1;
1134 } else if (mdocs[n->tok] == pmdoc_Pp) {
1135 pmdoc_macro_handler(n, rec, MDOC_Pp);
1136 }
1137
1138 /*
1139 * If an Xr macro was encountered then the child node has
1140 * already been explored by pmdoc_macro_handler.
1141 */
1142 if (xr_found == 0)
1143 mdoc_parse_Sh(n->child, rec);
1144 mdoc_parse_Sh(n->next, rec);
1145 }
1146
1147 /*
1148 * mdoc_parse_section--
1149 * Utility function for parsing sections of the mdoc type pages.
1150 * Takes two params:
1151 * 1. sec is an enum which indicates the section in which we are present
1152 * 2. string is the string which we need to append to the secbuff for this
1153 * particular section.
1154 * The function appends string to the global section buffer and returns.
1155 */
1156 static void
1157 mdoc_parse_section(enum roff_sec sec, const char *string, mandb_rec *rec)
1158 {
1159 /*
1160 * If the user specified the 'l' flag, then parse and store only the
1161 * NAME section. Ignore the rest.
1162 */
1163 if (mflags.limit)
1164 return;
1165
1166 switch (sec) {
1167 case SEC_LIBRARY:
1168 append(&rec->lib, string);
1169 break;
1170 case SEC_RETURN_VALUES:
1171 append(&rec->return_vals, string);
1172 break;
1173 case SEC_ENVIRONMENT:
1174 append(&rec->env, string);
1175 break;
1176 case SEC_FILES:
1177 append(&rec->files, string);
1178 break;
1179 case SEC_EXIT_STATUS:
1180 append(&rec->exit_status, string);
1181 break;
1182 case SEC_DIAGNOSTICS:
1183 append(&rec->diagnostics, string);
1184 break;
1185 case SEC_ERRORS:
1186 append(&rec->errors, string);
1187 break;
1188 default:
1189 append(&rec->desc, string);
1190 break;
1191 }
1192 }
1193
1194 static void
1195 proff_node(const struct roff_node *n, mandb_rec *rec, const proff_nf *func)
1196 {
1197 if (n == NULL)
1198 return;
1199
1200 switch (n->type) {
1201 case (ROFFT_BODY):
1202 /* FALLTHROUGH */
1203 case (ROFFT_BLOCK):
1204 /* FALLTHROUGH */
1205 case (ROFFT_ELEM):
1206 if (func[n->tok] != NULL)
1207 (*func[n->tok])(n, rec);
1208 break;
1209 default:
1210 break;
1211 }
1212
1213 proff_node(n->child, rec, func);
1214 proff_node(n->next, rec, func);
1215 }
1216
1217 /*
1218 * pman_parse_name --
1219 * Parses the NAME section and puts the complete content in the name_desc
1220 * variable.
1221 */
1222 static void
1223 pman_parse_name(const struct roff_node *n, mandb_rec *rec)
1224 {
1225 if (n == NULL)
1226 return;
1227
1228 if (n->type == ROFFT_TEXT) {
1229 char *tmp = parse_escape(n->string);
1230 concat(&rec->name_desc, tmp);
1231 free(tmp);
1232 }
1233
1234 if (n->child)
1235 pman_parse_name(n->child, rec);
1236
1237 if(n->next)
1238 pman_parse_name(n->next, rec);
1239 }
1240
1241 /*
1242 * A stub function to be able to parse the macros like .B embedded inside
1243 * a section.
1244 */
1245 static void
1246 pman_block(const struct roff_node *n, mandb_rec *rec)
1247 {
1248 }
1249
1250 /*
1251 * pman_sh --
1252 * This function does one of the two things:
1253 * 1. If the present section is NAME, then it will:
1254 * (a) Extract the name of the page (in case of multiple comma separated
1255 * names, it will pick up the first one).
1256 * (b) Build a space spearated list of all the symlinks/hardlinks to
1257 * this page and store in the buffer 'links'. These are extracted from
1258 * the comma separated list of names in the NAME section as well.
1259 * (c) Move on to the one line description section, which is after the list
1260 * of names in the NAME section.
1261 * 2. Otherwise, it will check the section name and call the man_parse_section
1262 * function, passing the enum corresponding to that section.
1263 */
1264 static void
1265 pman_sh(const struct roff_node *n, mandb_rec *rec)
1266 {
1267 static const struct {
1268 enum man_sec section;
1269 const char *header;
1270 } mapping[] = {
1271 { MANSEC_DESCRIPTION, "DESCRIPTION" },
1272 { MANSEC_SYNOPSIS, "SYNOPSIS" },
1273 { MANSEC_LIBRARY, "LIBRARY" },
1274 { MANSEC_ERRORS, "ERRORS" },
1275 { MANSEC_FILES, "FILES" },
1276 { MANSEC_RETURN_VALUES, "RETURN VALUE" },
1277 { MANSEC_RETURN_VALUES, "RETURN VALUES" },
1278 { MANSEC_EXIT_STATUS, "EXIT STATUS" },
1279 { MANSEC_EXAMPLES, "EXAMPLES" },
1280 { MANSEC_EXAMPLES, "EXAMPLE" },
1281 { MANSEC_STANDARDS, "STANDARDS" },
1282 { MANSEC_HISTORY, "HISTORY" },
1283 { MANSEC_BUGS, "BUGS" },
1284 { MANSEC_AUTHORS, "AUTHORS" },
1285 { MANSEC_COPYRIGHT, "COPYRIGHT" },
1286 };
1287 const struct roff_node *head;
1288 char *name_desc;
1289 size_t sz;
1290 size_t i;
1291
1292 if ((head = n->parent->head) == NULL || (head = head->child) == NULL ||
1293 head->type != ROFFT_TEXT)
1294 return;
1295
1296 /*
1297 * Check if this section should be extracted and
1298 * where it should be stored. Handled the trival cases first.
1299 */
1300 for (i = 0; i < sizeof(mapping) / sizeof(mapping[0]); ++i) {
1301 if (strcmp(head->string, mapping[i].header) == 0) {
1302 man_parse_section(mapping[i].section, n, rec);
1303 return;
1304 }
1305 }
1306
1307 if (strcmp(head->string, "NAME") == 0) {
1308 /*
1309 * We are in the NAME section.
1310 * pman_parse_name will put the complete content in name_desc.
1311 */
1312 pman_parse_name(n, rec);
1313
1314 name_desc = rec->name_desc;
1315 if (name_desc == NULL)
1316 return;
1317
1318 /* Remove any leading spaces. */
1319 while (name_desc[0] == ' ')
1320 name_desc++;
1321
1322 /* If the line begins with a "\&", avoid those */
1323 if (name_desc[0] == '\\' && name_desc[1] == '&')
1324 name_desc += 2;
1325
1326 /* Now name_desc should be left with a comma-space
1327 * separated list of names and the one line description
1328 * of the page:
1329 * "a, b, c \- sample description"
1330 * Take out the first name, before the first comma
1331 * (or space) and store it in rec->name.
1332 * If the page has aliases then they should be
1333 * in the form of a comma separated list.
1334 * Keep looping while there is a comma in name_desc,
1335 * extract the alias name and store in rec->links.
1336 * When there are no more commas left, break out.
1337 */
1338 int has_alias = 0; // Any more aliases left?
1339 while (*name_desc) {
1340 /* Remove any leading spaces or hyphens. */
1341 if (name_desc[0] == ' ' || name_desc[0] == '-') {
1342 name_desc++;
1343 continue;
1344 }
1345 sz = strcspn(name_desc, ", ");
1346
1347 /* Extract the first term and store it in rec->name. */
1348 if (rec->name == NULL) {
1349 if (name_desc[sz] == ',')
1350 has_alias = 1;
1351 rec->name = estrndup(name_desc, sz);
1352 /* XXX This would only happen with a poorly
1353 * written man page, maybe warn? */
1354 if (name_desc[sz] == '\0')
1355 break;
1356 name_desc += sz + 1;
1357 continue;
1358 }
1359
1360 /*
1361 * Once rec->name is set, rest of the names
1362 * are to be treated as links or aliases.
1363 */
1364 if (rec->name && has_alias) {
1365 if (name_desc[sz] != ',') {
1366 /* No more commas left --> no more
1367 * aliases to take out */
1368 has_alias = 0;
1369 }
1370 concat2(&rec->links, name_desc, sz);
1371 /* XXX This would only happen with a poorly
1372 * written man page, maybe warn? */
1373 if (name_desc[sz] == '\0')
1374 break;
1375 name_desc += sz + 1;
1376 continue;
1377 }
1378 break;
1379 }
1380
1381 /* Parse any escape sequences that might be there */
1382 char *temp = parse_escape(name_desc);
1383 free(rec->name_desc);
1384 rec->name_desc = temp;
1385 temp = parse_escape(rec->name);
1386 free(rec->name);
1387 rec->name = temp;
1388 return;
1389 }
1390
1391 /* The RETURN VALUE section might be specified in multiple ways */
1392 if (strcmp(head->string, "RETURN") == 0 &&
1393 head->next != NULL && head->next->type == ROFFT_TEXT &&
1394 (strcmp(head->next->string, "VALUE") == 0 ||
1395 strcmp(head->next->string, "VALUES") == 0)) {
1396 man_parse_section(MANSEC_RETURN_VALUES, n, rec);
1397 return;
1398 }
1399
1400 /*
1401 * EXIT STATUS section can also be specified all on one line or on two
1402 * separate lines.
1403 */
1404 if (strcmp(head->string, "EXIT") == 0 &&
1405 head->next != NULL && head->next->type == ROFFT_TEXT &&
1406 strcmp(head->next->string, "STATUS") == 0) {
1407 man_parse_section(MANSEC_EXIT_STATUS, n, rec);
1408 return;
1409 }
1410
1411 /* Store the rest of the content in desc. */
1412 man_parse_section(MANSEC_NONE, n, rec);
1413 }
1414
1415 /*
1416 * pman_parse_node --
1417 * Generic function to iterate through a node. Usually called from
1418 * man_parse_section to parse a particular section of the man page.
1419 */
1420 static void
1421 pman_parse_node(const struct roff_node *n, secbuff *s)
1422 {
1423 if (n == NULL)
1424 return;
1425
1426 if (n->type == ROFFT_TEXT)
1427 append(s, n->string);
1428
1429 pman_parse_node(n->child, s);
1430 pman_parse_node(n->next, s);
1431 }
1432
1433 /*
1434 * man_parse_section --
1435 * Takes two parameters:
1436 * sec: Tells which section we are present in
1437 * n: Is the present node of the AST.
1438 * Depending on the section, we call pman_parse_node to parse that section and
1439 * concatenate the content from that section into the buffer for that section.
1440 */
1441 static void
1442 man_parse_section(enum man_sec sec, const struct roff_node *n, mandb_rec *rec)
1443 {
1444 /*
1445 * If the user sepecified the 'l' flag then just parse
1446 * the NAME section, ignore the rest.
1447 */
1448 if (mflags.limit)
1449 return;
1450
1451 switch (sec) {
1452 case MANSEC_LIBRARY:
1453 pman_parse_node(n, &rec->lib);
1454 break;
1455 case MANSEC_RETURN_VALUES:
1456 pman_parse_node(n, &rec->return_vals);
1457 break;
1458 case MANSEC_ENVIRONMENT:
1459 pman_parse_node(n, &rec->env);
1460 break;
1461 case MANSEC_FILES:
1462 pman_parse_node(n, &rec->files);
1463 break;
1464 case MANSEC_EXIT_STATUS:
1465 pman_parse_node(n, &rec->exit_status);
1466 break;
1467 case MANSEC_DIAGNOSTICS:
1468 pman_parse_node(n, &rec->diagnostics);
1469 break;
1470 case MANSEC_ERRORS:
1471 pman_parse_node(n, &rec->errors);
1472 break;
1473 case MANSEC_NAME:
1474 case MANSEC_SYNOPSIS:
1475 case MANSEC_EXAMPLES:
1476 case MANSEC_STANDARDS:
1477 case MANSEC_HISTORY:
1478 case MANSEC_BUGS:
1479 case MANSEC_AUTHORS:
1480 case MANSEC_COPYRIGHT:
1481 break;
1482 default:
1483 pman_parse_node(n, &rec->desc);
1484 break;
1485 }
1486
1487 }
1488
1489 /*
1490 * insert_into_db --
1491 * Inserts the parsed data of the man page in the Sqlite databse.
1492 * If any of the values is NULL, then we cleanup and return -1 indicating
1493 * an error.
1494 * Otherwise, store the data in the database and return 0.
1495 */
1496 static int
1497 insert_into_db(sqlite3 *db, mandb_rec *rec)
1498 {
1499 int rc = 0;
1500 int idx = -1;
1501 const char *sqlstr = NULL;
1502 sqlite3_stmt *stmt = NULL;
1503 char *ln = NULL;
1504 char *errmsg = NULL;
1505 long int mandb_rowid;
1506
1507 /*
1508 * At the very minimum we want to make sure that we store
1509 * the following data:
1510 * Name, one line description, and the MD5 hash
1511 */
1512 if (rec->name == NULL || rec->name_desc == NULL ||
1513 rec->md5_hash == NULL) {
1514 cleanup(rec);
1515 return -1;
1516 }
1517
1518 /* Write null byte at the end of all the sec_buffs */
1519 rec->desc.data[rec->desc.offset] = 0;
1520 rec->lib.data[rec->lib.offset] = 0;
1521 rec->env.data[rec->env.offset] = 0;
1522 rec->return_vals.data[rec->return_vals.offset] = 0;
1523 rec->exit_status.data[rec->exit_status.offset] = 0;
1524 rec->files.data[rec->files.offset] = 0;
1525 rec->diagnostics.data[rec->diagnostics.offset] = 0;
1526 rec->errors.data[rec->errors.offset] = 0;
1527
1528 /*
1529 * In case of a mdoc page: (sorry, no better place to put this code)
1530 * parse the comma separated list of names of man pages,
1531 * the first name will be stored in the mandb table, rest will be
1532 * treated as links and put in the mandb_links table.
1533 */
1534 if (rec->page_type == MDOC) {
1535 char *tmp;
1536 rec->links = estrdup(rec->name);
1537 free(rec->name);
1538 size_t sz = strcspn(rec->links, " \0");
1539 rec->name = emalloc(sz + 1);
1540 memcpy(rec->name, rec->links, sz);
1541 if(rec->name[sz - 1] == ',')
1542 rec->name[sz - 1] = 0;
1543 else
1544 rec->name[sz] = 0;
1545 while (rec->links[sz] == ' ')
1546 ++sz;
1547 tmp = estrdup(rec->links + sz);
1548 free(rec->links);
1549 rec->links = tmp;
1550 }
1551
1552 /*------------------------ Populate the mandb table---------------------------*/
1553 sqlstr = "INSERT INTO mandb VALUES (:section, :name, :name_desc, :desc,"
1554 " :lib, :return_vals, :env, :files, :exit_status,"
1555 " :diagnostics, :errors, :md5_hash, :machine)";
1556
1557 rc = sqlite3_prepare_v2(db, sqlstr, -1, &stmt, NULL);
1558 if (rc != SQLITE_OK)
1559 goto Out;
1560
1561 idx = sqlite3_bind_parameter_index(stmt, ":name");
1562 rc = sqlite3_bind_text(stmt, idx, rec->name, -1, NULL);
1563 if (rc != SQLITE_OK) {
1564 sqlite3_finalize(stmt);
1565 goto Out;
1566 }
1567
1568 idx = sqlite3_bind_parameter_index(stmt, ":section");
1569 rc = sqlite3_bind_text(stmt, idx, rec->section, -1, NULL);
1570 if (rc != SQLITE_OK) {
1571 sqlite3_finalize(stmt);
1572 goto Out;
1573 }
1574
1575 idx = sqlite3_bind_parameter_index(stmt, ":name_desc");
1576 rc = sqlite3_bind_text(stmt, idx, rec->name_desc, -1, NULL);
1577 if (rc != SQLITE_OK) {
1578 sqlite3_finalize(stmt);
1579 goto Out;
1580 }
1581
1582 idx = sqlite3_bind_parameter_index(stmt, ":desc");
1583 rc = sqlite3_bind_text(stmt, idx, rec->desc.data,
1584 rec->desc.offset + 1, NULL);
1585 if (rc != SQLITE_OK) {
1586 sqlite3_finalize(stmt);
1587 goto Out;
1588 }
1589
1590 idx = sqlite3_bind_parameter_index(stmt, ":lib");
1591 rc = sqlite3_bind_text(stmt, idx, rec->lib.data,
1592 rec->lib.offset + 1, NULL);
1593 if (rc != SQLITE_OK) {
1594 sqlite3_finalize(stmt);
1595 goto Out;
1596 }
1597
1598 idx = sqlite3_bind_parameter_index(stmt, ":return_vals");
1599 rc = sqlite3_bind_text(stmt, idx, rec->return_vals.data,
1600 rec->return_vals.offset + 1, NULL);
1601 if (rc != SQLITE_OK) {
1602 sqlite3_finalize(stmt);
1603 goto Out;
1604 }
1605
1606 idx = sqlite3_bind_parameter_index(stmt, ":env");
1607 rc = sqlite3_bind_text(stmt, idx, rec->env.data,
1608 rec->env.offset + 1, NULL);
1609 if (rc != SQLITE_OK) {
1610 sqlite3_finalize(stmt);
1611 goto Out;
1612 }
1613
1614 idx = sqlite3_bind_parameter_index(stmt, ":files");
1615 rc = sqlite3_bind_text(stmt, idx, rec->files.data,
1616 rec->files.offset + 1, NULL);
1617 if (rc != SQLITE_OK) {
1618 sqlite3_finalize(stmt);
1619 goto Out;
1620 }
1621
1622 idx = sqlite3_bind_parameter_index(stmt, ":exit_status");
1623 rc = sqlite3_bind_text(stmt, idx, rec->exit_status.data,
1624 rec->exit_status.offset + 1, NULL);
1625 if (rc != SQLITE_OK) {
1626 sqlite3_finalize(stmt);
1627 goto Out;
1628 }
1629
1630 idx = sqlite3_bind_parameter_index(stmt, ":diagnostics");
1631 rc = sqlite3_bind_text(stmt, idx, rec->diagnostics.data,
1632 rec->diagnostics.offset + 1, NULL);
1633 if (rc != SQLITE_OK) {
1634 sqlite3_finalize(stmt);
1635 goto Out;
1636 }
1637
1638 idx = sqlite3_bind_parameter_index(stmt, ":errors");
1639 rc = sqlite3_bind_text(stmt, idx, rec->errors.data,
1640 rec->errors.offset + 1, NULL);
1641 if (rc != SQLITE_OK) {
1642 sqlite3_finalize(stmt);
1643 goto Out;
1644 }
1645
1646 idx = sqlite3_bind_parameter_index(stmt, ":md5_hash");
1647 rc = sqlite3_bind_text(stmt, idx, rec->md5_hash, -1, NULL);
1648 if (rc != SQLITE_OK) {
1649 sqlite3_finalize(stmt);
1650 goto Out;
1651 }
1652
1653 idx = sqlite3_bind_parameter_index(stmt, ":machine");
1654 if (rec->machine)
1655 rc = sqlite3_bind_text(stmt, idx, rec->machine, -1, NULL);
1656 else
1657 rc = sqlite3_bind_null(stmt, idx);
1658 if (rc != SQLITE_OK) {
1659 sqlite3_finalize(stmt);
1660 goto Out;
1661 }
1662
1663 rc = sqlite3_step(stmt);
1664 if (rc != SQLITE_DONE) {
1665 sqlite3_finalize(stmt);
1666 goto Out;
1667 }
1668
1669 sqlite3_finalize(stmt);
1670
1671 /* Get the row id of the last inserted row */
1672 mandb_rowid = sqlite3_last_insert_rowid(db);
1673
1674 /*------------------------Populate the mandb_meta table-----------------------*/
1675 sqlstr = "INSERT INTO mandb_meta VALUES (:device, :inode, :mtime,"
1676 " :file, :md5_hash, :id)";
1677 rc = sqlite3_prepare_v2(db, sqlstr, -1, &stmt, NULL);
1678 if (rc != SQLITE_OK)
1679 goto Out;
1680
1681 idx = sqlite3_bind_parameter_index(stmt, ":device");
1682 rc = sqlite3_bind_int64(stmt, idx, rec->device);
1683 if (rc != SQLITE_OK) {
1684 sqlite3_finalize(stmt);
1685 goto Out;
1686 }
1687
1688 idx = sqlite3_bind_parameter_index(stmt, ":inode");
1689 rc = sqlite3_bind_int64(stmt, idx, rec->inode);
1690 if (rc != SQLITE_OK) {
1691 sqlite3_finalize(stmt);
1692 goto Out;
1693 }
1694
1695 idx = sqlite3_bind_parameter_index(stmt, ":mtime");
1696 rc = sqlite3_bind_int64(stmt, idx, rec->mtime);
1697 if (rc != SQLITE_OK) {
1698 sqlite3_finalize(stmt);
1699 goto Out;
1700 }
1701
1702 idx = sqlite3_bind_parameter_index(stmt, ":file");
1703 rc = sqlite3_bind_text(stmt, idx, rec->file_path, -1, NULL);
1704 if (rc != SQLITE_OK) {
1705 sqlite3_finalize(stmt);
1706 goto Out;
1707 }
1708
1709 idx = sqlite3_bind_parameter_index(stmt, ":md5_hash");
1710 rc = sqlite3_bind_text(stmt, idx, rec->md5_hash, -1, NULL);
1711 if (rc != SQLITE_OK) {
1712 sqlite3_finalize(stmt);
1713 goto Out;
1714 }
1715
1716 idx = sqlite3_bind_parameter_index(stmt, ":id");
1717 rc = sqlite3_bind_int64(stmt, idx, mandb_rowid);
1718 if (rc != SQLITE_OK) {
1719 sqlite3_finalize(stmt);
1720 goto Out;
1721 }
1722
1723 rc = sqlite3_step(stmt);
1724 sqlite3_finalize(stmt);
1725 if (rc == SQLITE_CONSTRAINT_UNIQUE) {
1726 /* The *most* probable reason for reaching here is that
1727 * the UNIQUE contraint on the file column of the mandb_meta
1728 * table was violated.
1729 * This can happen when a file was updated/modified.
1730 * To fix this we need to do two things:
1731 * 1. Delete the row for the older version of this file
1732 * from mandb table.
1733 * 2. Run an UPDATE query to update the row for this file
1734 * in the mandb_meta table.
1735 */
1736 warnx("Trying to update index for %s", rec->file_path);
1737 char *sql = sqlite3_mprintf("DELETE FROM mandb "
1738 "WHERE rowid = (SELECT id"
1739 " FROM mandb_meta"
1740 " WHERE file = %Q)",
1741 rec->file_path);
1742 sqlite3_exec(db, sql, NULL, NULL, &errmsg);
1743 sqlite3_free(sql);
1744 if (errmsg != NULL) {
1745 if (mflags.verbosity)
1746 warnx("%s", errmsg);
1747 free(errmsg);
1748 }
1749 sqlstr = "UPDATE mandb_meta SET device = :device,"
1750 " inode = :inode, mtime = :mtime, id = :id,"
1751 " md5_hash = :md5 WHERE file = :file";
1752 rc = sqlite3_prepare_v2(db, sqlstr, -1, &stmt, NULL);
1753 if (rc != SQLITE_OK) {
1754 if (mflags.verbosity)
1755 warnx("Update failed with error: %s",
1756 sqlite3_errmsg(db));
1757 close_db(db);
1758 cleanup(rec);
1759 errx(EXIT_FAILURE,
1760 "Consider running makemandb with -f option");
1761 }
1762
1763 idx = sqlite3_bind_parameter_index(stmt, ":device");
1764 sqlite3_bind_int64(stmt, idx, rec->device);
1765 idx = sqlite3_bind_parameter_index(stmt, ":inode");
1766 sqlite3_bind_int64(stmt, idx, rec->inode);
1767 idx = sqlite3_bind_parameter_index(stmt, ":mtime");
1768 sqlite3_bind_int64(stmt, idx, rec->mtime);
1769 idx = sqlite3_bind_parameter_index(stmt, ":id");
1770 sqlite3_bind_int64(stmt, idx, mandb_rowid);
1771 idx = sqlite3_bind_parameter_index(stmt, ":md5");
1772 sqlite3_bind_text(stmt, idx, rec->md5_hash, -1, NULL);
1773 idx = sqlite3_bind_parameter_index(stmt, ":file");
1774 sqlite3_bind_text(stmt, idx, rec->file_path, -1, NULL);
1775 rc = sqlite3_step(stmt);
1776 sqlite3_finalize(stmt);
1777
1778 if (rc != SQLITE_DONE) {
1779 if (mflags.verbosity)
1780 warnx("%s", sqlite3_errmsg(db));
1781 close_db(db);
1782 cleanup(rec);
1783 errx(EXIT_FAILURE,
1784 "Consider running makemandb with -f option");
1785 }
1786 } else if (rc != SQLITE_DONE) {
1787 /* Otherwise make this error fatal */
1788 warnx("Failed at %s\n%s", rec->file_path, sqlite3_errmsg(db));
1789 cleanup(rec);
1790 close_db(db);
1791 exit(EXIT_FAILURE);
1792 }
1793
1794 /*------------------------ Populate the mandb_links table---------------------*/
1795 char *str = NULL;
1796 char *links;
1797 if (rec->links && strlen(rec->links)) {
1798 links = rec->links;
1799 for(ln = strtok(links, " "); ln; ln = strtok(NULL, " ")) {
1800 if (ln[0] == ',')
1801 ln++;
1802 if(ln[strlen(ln) - 1] == ',')
1803 ln[strlen(ln) - 1] = 0;
1804
1805 str = sqlite3_mprintf("INSERT INTO mandb_links"
1806 " VALUES (%Q, %Q, %Q, %Q, %Q)",
1807 ln, rec->name, rec->section,
1808 rec->machine, rec->md5_hash);
1809 sqlite3_exec(db, str, NULL, NULL, &errmsg);
1810 sqlite3_free(str);
1811 if (errmsg != NULL) {
1812 warnx("%s", errmsg);
1813 cleanup(rec);
1814 free(errmsg);
1815 return -1;
1816 }
1817 }
1818 }
1819
1820 cleanup(rec);
1821 return 0;
1822
1823 Out:
1824 if (mflags.verbosity)
1825 warnx("%s", sqlite3_errmsg(db));
1826 cleanup(rec);
1827 return -1;
1828 }
1829
1830 /*
1831 * check_md5--
1832 * Generates the md5 hash of the file and checks if it already doesn't exist
1833 * in the table (passed as the 3rd parameter).
1834 * This function is being used to avoid hardlinks.
1835 * On successful completion it will also set the value of the fourth parameter
1836 * to the md5 hash of the file (computed previously). It is the responsibility
1837 * of the caller to free this buffer.
1838 * Return values:
1839 * -1: If an error occurs somewhere and sets the md5 return buffer to NULL.
1840 * 0: If the md5 hash does not exist in the table.
1841 * 1: If the hash exists in the database.
1842 */
1843 static int
1844 check_md5(const char *file, sqlite3 *db, const char *table, char **md5sum,
1845 void *buf, size_t buflen)
1846 {
1847 int rc = 0;
1848 int idx = -1;
1849 char *sqlstr = NULL;
1850 sqlite3_stmt *stmt = NULL;
1851
1852 assert(file != NULL);
1853 *md5sum = MD5Data(buf, buflen, NULL);
1854 if (*md5sum == NULL) {
1855 if (mflags.verbosity)
1856 warn("md5 failed: %s", file);
1857 return -1;
1858 }
1859
1860 easprintf(&sqlstr, "SELECT * FROM %s WHERE md5_hash = :md5_hash",
1861 table);
1862 rc = sqlite3_prepare_v2(db, sqlstr, -1, &stmt, NULL);
1863 if (rc != SQLITE_OK) {
1864 free(sqlstr);
1865 free(*md5sum);
1866 *md5sum = NULL;
1867 return -1;
1868 }
1869
1870 idx = sqlite3_bind_parameter_index(stmt, ":md5_hash");
1871 rc = sqlite3_bind_text(stmt, idx, *md5sum, -1, NULL);
1872 if (rc != SQLITE_OK) {
1873 if (mflags.verbosity)
1874 warnx("%s", sqlite3_errmsg(db));
1875 sqlite3_finalize(stmt);
1876 free(sqlstr);
1877 free(*md5sum);
1878 *md5sum = NULL;
1879 return -1;
1880 }
1881
1882 if (sqlite3_step(stmt) == SQLITE_ROW) {
1883 sqlite3_finalize(stmt);
1884 free(sqlstr);
1885 return 0;
1886 }
1887
1888 sqlite3_finalize(stmt);
1889 free(sqlstr);
1890 return 1;
1891 }
1892
1893 /* Optimize the index for faster search */
1894 static void
1895 optimize(sqlite3 *db)
1896 {
1897 const char *sqlstr;
1898 char *errmsg = NULL;
1899
1900 if (mflags.verbosity == 2)
1901 printf("Optimizing the database index\n");
1902 sqlstr = "INSERT INTO mandb(mandb) VALUES (\'optimize\');"
1903 "VACUUM";
1904 sqlite3_exec(db, sqlstr, NULL, NULL, &errmsg);
1905 if (errmsg != NULL) {
1906 if (mflags.verbosity)
1907 warnx("%s", errmsg);
1908 free(errmsg);
1909 return;
1910 }
1911 }
1912
1913 /*
1914 * cleanup --
1915 * cleans up the global buffers
1916 */
1917 static void
1918 cleanup(mandb_rec *rec)
1919 {
1920 rec->desc.offset = 0;
1921 rec->lib.offset = 0;
1922 rec->return_vals.offset = 0;
1923 rec->env.offset = 0;
1924 rec->exit_status.offset = 0;
1925 rec->diagnostics.offset = 0;
1926 rec->errors.offset = 0;
1927 rec->files.offset = 0;
1928
1929 free(rec->machine);
1930 rec->machine = NULL;
1931
1932 free(rec->links);
1933 rec->links = NULL;
1934
1935 free(rec->file_path);
1936 rec->file_path = NULL;
1937
1938 free(rec->name);
1939 rec->name = NULL;
1940
1941 free(rec->name_desc);
1942 rec->name_desc = NULL;
1943
1944 free(rec->md5_hash);
1945 rec->md5_hash = NULL;
1946
1947 free(rec->section);
1948 rec->section = NULL;
1949 }
1950
1951 /*
1952 * init_secbuffs--
1953 * Sets the value of buflen for all the sec_buff field of rec. And then
1954 * allocate memory to each sec_buff member of rec.
1955 */
1956 static void
1957 init_secbuffs(mandb_rec *rec)
1958 {
1959 /*
1960 * Some sec_buff might need more memory, for example desc,
1961 * which stores the data of the DESCRIPTION section,
1962 * while some might need very small amount of memory.
1963 * Therefore explicitly setting the value of buflen field for
1964 * each sec_buff.
1965 */
1966 rec->desc.buflen = 10 * BUFLEN;
1967 rec->desc.data = emalloc(rec->desc.buflen);
1968 rec->desc.offset = 0;
1969
1970 rec->lib.buflen = BUFLEN / 2;
1971 rec->lib.data = emalloc(rec->lib.buflen);
1972 rec->lib.offset = 0;
1973
1974 rec->return_vals.buflen = BUFLEN;
1975 rec->return_vals.data = emalloc(rec->return_vals.buflen);
1976 rec->return_vals.offset = 0;
1977
1978 rec->exit_status.buflen = BUFLEN;
1979 rec->exit_status.data = emalloc(rec->exit_status.buflen);
1980 rec->exit_status.offset = 0;
1981
1982 rec->env.buflen = BUFLEN;
1983 rec->env.data = emalloc(rec->env.buflen);
1984 rec->env.offset = 0;
1985
1986 rec->files.buflen = BUFLEN;
1987 rec->files.data = emalloc(rec->files.buflen);
1988 rec->files.offset = 0;
1989
1990 rec->diagnostics.buflen = BUFLEN;
1991 rec->diagnostics.data = emalloc(rec->diagnostics.buflen);
1992 rec->diagnostics.offset = 0;
1993
1994 rec->errors.buflen = BUFLEN;
1995 rec->errors.data = emalloc(rec->errors.buflen);
1996 rec->errors.offset = 0;
1997 }
1998
1999 /*
2000 * free_secbuffs--
2001 * This function should be called at the end, when all the pages have been
2002 * parsed.
2003 * It frees the memory allocated to sec_buffs by init_secbuffs in the starting.
2004 */
2005 static void
2006 free_secbuffs(mandb_rec *rec)
2007 {
2008 free(rec->desc.data);
2009 free(rec->lib.data);
2010 free(rec->return_vals.data);
2011 free(rec->exit_status.data);
2012 free(rec->env.data);
2013 free(rec->files.data);
2014 free(rec->diagnostics.data);
2015 free(rec->errors.data);
2016 }
2017
2018 static void
2019 replace_hyph(char *str)
2020 {
2021 char *iter = str;
2022 while ((iter = strchr(iter, ASCII_HYPH)) != NULL)
2023 *iter = '-';
2024
2025 iter = str;
2026 while ((iter = strchr(iter, ASCII_NBRSP)) != NULL)
2027 *iter = '-';
2028 }
2029
2030 static char *
2031 parse_escape(const char *str)
2032 {
2033 const char *backslash, *last_backslash;
2034 char *result, *iter;
2035 size_t len;
2036
2037 assert(str);
2038
2039 last_backslash = str;
2040 backslash = strchr(str, '\\');
2041 if (backslash == NULL) {
2042 result = estrdup(str);
2043 replace_hyph(result);
2044 return result;
2045 }
2046
2047 result = emalloc(strlen(str) + 1);
2048 iter = result;
2049
2050 do {
2051 len = backslash - last_backslash;
2052 memcpy(iter, last_backslash, len);
2053 iter += len;
2054 if (backslash[1] == '-' || backslash[1] == ' ') {
2055 *iter++ = backslash[1];
2056 last_backslash = backslash + 2;
2057 backslash = strchr(last_backslash, '\\');
2058 } else {
2059 ++backslash;
2060 mandoc_escape(&backslash, NULL, NULL);
2061 last_backslash = backslash;
2062 if (backslash == NULL)
2063 break;
2064 backslash = strchr(last_backslash, '\\');
2065 }
2066 } while (backslash != NULL);
2067 if (last_backslash != NULL)
2068 strcpy(iter, last_backslash);
2069
2070 replace_hyph(result);
2071 return result;
2072 }
2073
2074 /*
2075 * append--
2076 * Concatenates a space and src at the end of sbuff->data (much like concat in
2077 * apropos-utils.c).
2078 * Rather than reallocating space for writing data, it uses the value of the
2079 * offset field of sec_buff to write new data at the free space left in the
2080 * buffer.
2081 * In case the size of the data to be appended exceeds the number of bytes left
2082 * in the buffer, it reallocates buflen number of bytes and then continues.
2083 * Value of offset field should be adjusted as new data is written.
2084 *
2085 * NOTE: This function does not write the null byte at the end of the buffers,
2086 * write a null byte at the position pointed to by offset before inserting data
2087 * in the db.
2088 */
2089 static void
2090 append(secbuff *sbuff, const char *src)
2091 {
2092 short flag = 0;
2093 size_t srclen, newlen;
2094 char *temp;
2095
2096 assert(src != NULL);
2097 temp = parse_escape(src);
2098 srclen = strlen(temp);
2099
2100 if (sbuff->data == NULL) {
2101 sbuff->data = emalloc(sbuff->buflen);
2102 sbuff->offset = 0;
2103 }
2104
2105 newlen = sbuff->offset + srclen + 2;
2106 if (newlen >= sbuff->buflen) {
2107 while (sbuff->buflen < newlen)
2108 sbuff->buflen += sbuff->buflen;
2109 sbuff->data = erealloc(sbuff->data, sbuff->buflen);
2110 flag = 1;
2111 }
2112
2113 /* Append a space at the end of the buffer. */
2114 if (sbuff->offset || flag)
2115 sbuff->data[sbuff->offset++] = ' ';
2116 /* Now, copy src at the end of the buffer. */
2117 memcpy(sbuff->data + sbuff->offset, temp, srclen);
2118 sbuff->offset += srclen;
2119 free(temp);
2120 }
2121
2122 static void
2123 usage(void)
2124 {
2125 fprintf(stderr, "Usage: %s [-floQqv] [-C path]\n", getprogname());
2126 exit(1);
2127 }
2128