1 /* $NetBSD: makemandb.c,v 1.6 2012/02/27 16:51:06 joerg Exp $ */ 2 /* 3 * Copyright (c) 2011 Abhinav Upadhyay <er.abhinav.upadhyay@gmail.com> 4 * Copyright (c) 2011 Kristaps Dzonsons <kristaps@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 1.6 2012/02/27 16:51:06 joerg Exp $"); 21 22 #include <sys/stat.h> 23 #include <sys/types.h> 24 25 #include <assert.h> 26 #include <ctype.h> 27 #include <dirent.h> 28 #include <err.h> 29 #include <archive.h> 30 #include <libgen.h> 31 #include <md5.h> 32 #include <stdio.h> 33 #include <stdlib.h> 34 #include <string.h> 35 #include <unistd.h> 36 #include <util.h> 37 38 #include "apropos-utils.h" 39 #include "man.h" 40 #include "mandoc.h" 41 #include "mdoc.h" 42 #include "sqlite3.h" 43 44 #define BUFLEN 1024 45 #define MDOC 0 //If the page is of mdoc(7) type 46 #define MAN 1 //If the page is of man(7) type 47 48 /* 49 * A data structure for holding section specific data. 50 */ 51 typedef struct secbuff { 52 char *data; 53 size_t buflen; //Total length of buffer allocated initially 54 size_t offset; // Current offset in the buffer. 55 } secbuff; 56 57 typedef struct makemandb_flags { 58 int optimize; 59 int limit; // limit the indexing to only NAME section 60 int recreate; // Database was created from scratch 61 int verbosity; // 0: quiet, 1: default, 2: verbose 62 } makemandb_flags; 63 64 typedef struct mandb_rec { 65 /* Fields for mandb table */ 66 char *name; // for storing the name of the man page 67 char *name_desc; // for storing the one line description (.Nd) 68 secbuff desc; // for storing the DESCRIPTION section 69 secbuff lib; // for the LIBRARY section 70 secbuff return_vals; // RETURN VALUES 71 secbuff env; // ENVIRONMENT 72 secbuff files; // FILES 73 secbuff exit_status; // EXIT STATUS 74 secbuff diagnostics; // DIAGNOSTICS 75 secbuff errors; // ERRORS 76 char section[2]; 77 78 int xr_found; 79 80 /* Fields for mandb_meta table */ 81 char *md5_hash; 82 dev_t device; 83 ino_t inode; 84 time_t mtime; 85 86 /* Fields for mandb_links table */ 87 char *machine; 88 char *links; //all the links to a page in a space separated form 89 char *file_path; 90 91 /* Non-db fields */ 92 int page_type; //Indicates the type of page: mdoc or man 93 } mandb_rec; 94 95 static void append(secbuff *sbuff, const char *src); 96 static void init_secbuffs(mandb_rec *); 97 static void free_secbuffs(mandb_rec *); 98 static int check_md5(const char *, sqlite3 *, const char *, char **, void *, size_t); 99 static void cleanup(mandb_rec *); 100 static void set_section(const struct mdoc *, const struct man *, mandb_rec *); 101 static void set_machine(const struct mdoc *, mandb_rec *); 102 static int insert_into_db(sqlite3 *, mandb_rec *); 103 static void begin_parse(const char *, struct mparse *, mandb_rec *, 104 const void *, size_t len); 105 static void pmdoc_node(const struct mdoc_node *, mandb_rec *); 106 static void pmdoc_Nm(const struct mdoc_node *, mandb_rec *); 107 static void pmdoc_Nd(const struct mdoc_node *, mandb_rec *); 108 static void pmdoc_Sh(const struct mdoc_node *, mandb_rec *); 109 static void pmdoc_Xr(const struct mdoc_node *, mandb_rec *); 110 static void pmdoc_Pp(const struct mdoc_node *, mandb_rec *); 111 static void pmdoc_macro_handler(const struct mdoc_node *, mandb_rec *, 112 enum mdoct); 113 static void pman_node(const struct man_node *n, mandb_rec *); 114 static void pman_parse_node(const struct man_node *, secbuff *); 115 static void pman_parse_name(const struct man_node *, mandb_rec *); 116 static void pman_sh(const struct man_node *, mandb_rec *); 117 static void pman_block(const struct man_node *, mandb_rec *); 118 static void traversedir(const char *, const char *, sqlite3 *, struct mparse *); 119 static void mdoc_parse_section(enum mdoc_sec, const char *, mandb_rec *); 120 static void man_parse_section(enum man_sec, const struct man_node *, mandb_rec *); 121 static void build_file_cache(sqlite3 *, const char *, const char *, 122 struct stat *); 123 static void update_db(sqlite3 *, struct mparse *, mandb_rec *); 124 __dead static void usage(void); 125 static void optimize(sqlite3 *); 126 static char *parse_escape(const char *); 127 static makemandb_flags mflags = { .verbosity = 1 }; 128 129 typedef void (*pman_nf)(const struct man_node *n, mandb_rec *); 130 typedef void (*pmdoc_nf)(const struct mdoc_node *n, mandb_rec *); 131 static const pmdoc_nf mdocs[MDOC_MAX] = { 132 NULL, /* Ap */ 133 NULL, /* Dd */ 134 NULL, /* Dt */ 135 NULL, /* Os */ 136 pmdoc_Sh, /* Sh */ 137 NULL, /* Ss */ 138 pmdoc_Pp, /* Pp */ 139 NULL, /* D1 */ 140 NULL, /* Dl */ 141 NULL, /* Bd */ 142 NULL, /* Ed */ 143 NULL, /* Bl */ 144 NULL, /* El */ 145 NULL, /* It */ 146 NULL, /* Ad */ 147 NULL, /* An */ 148 NULL, /* Ar */ 149 NULL, /* Cd */ 150 NULL, /* Cm */ 151 NULL, /* Dv */ 152 NULL, /* Er */ 153 NULL, /* Ev */ 154 NULL, /* Ex */ 155 NULL, /* Fa */ 156 NULL, /* Fd */ 157 NULL, /* Fl */ 158 NULL, /* Fn */ 159 NULL, /* Ft */ 160 NULL, /* Ic */ 161 NULL, /* In */ 162 NULL, /* Li */ 163 pmdoc_Nd, /* Nd */ 164 pmdoc_Nm, /* Nm */ 165 NULL, /* Op */ 166 NULL, /* Ot */ 167 NULL, /* Pa */ 168 NULL, /* Rv */ 169 NULL, /* St */ 170 NULL, /* Va */ 171 NULL, /* Vt */ 172 pmdoc_Xr, /* Xr */ 173 NULL, /* %A */ 174 NULL, /* %B */ 175 NULL, /* %D */ 176 NULL, /* %I */ 177 NULL, /* %J */ 178 NULL, /* %N */ 179 NULL, /* %O */ 180 NULL, /* %P */ 181 NULL, /* %R */ 182 NULL, /* %T */ 183 NULL, /* %V */ 184 NULL, /* Ac */ 185 NULL, /* Ao */ 186 NULL, /* Aq */ 187 NULL, /* At */ 188 NULL, /* Bc */ 189 NULL, /* Bf */ 190 NULL, /* Bo */ 191 NULL, /* Bq */ 192 NULL, /* Bsx */ 193 NULL, /* Bx */ 194 NULL, /* Db */ 195 NULL, /* Dc */ 196 NULL, /* Do */ 197 NULL, /* Dq */ 198 NULL, /* Ec */ 199 NULL, /* Ef */ 200 NULL, /* Em */ 201 NULL, /* Eo */ 202 NULL, /* Fx */ 203 NULL, /* Ms */ 204 NULL, /* No */ 205 NULL, /* Ns */ 206 NULL, /* Nx */ 207 NULL, /* Ox */ 208 NULL, /* Pc */ 209 NULL, /* Pf */ 210 NULL, /* Po */ 211 NULL, /* Pq */ 212 NULL, /* Qc */ 213 NULL, /* Ql */ 214 NULL, /* Qo */ 215 NULL, /* Qq */ 216 NULL, /* Re */ 217 NULL, /* Rs */ 218 NULL, /* Sc */ 219 NULL, /* So */ 220 NULL, /* Sq */ 221 NULL, /* Sm */ 222 NULL, /* Sx */ 223 NULL, /* Sy */ 224 NULL, /* Tn */ 225 NULL, /* Ux */ 226 NULL, /* Xc */ 227 NULL, /* Xo */ 228 NULL, /* Fo */ 229 NULL, /* Fc */ 230 NULL, /* Oo */ 231 NULL, /* Oc */ 232 NULL, /* Bk */ 233 NULL, /* Ek */ 234 NULL, /* Bt */ 235 NULL, /* Hf */ 236 NULL, /* Fr */ 237 NULL, /* Ud */ 238 NULL, /* Lb */ 239 NULL, /* Lp */ 240 NULL, /* Lk */ 241 NULL, /* Mt */ 242 NULL, /* Brq */ 243 NULL, /* Bro */ 244 NULL, /* Brc */ 245 NULL, /* %C */ 246 NULL, /* Es */ 247 NULL, /* En */ 248 NULL, /* Dx */ 249 NULL, /* %Q */ 250 NULL, /* br */ 251 NULL, /* sp */ 252 NULL, /* %U */ 253 NULL, /* Ta */ 254 }; 255 256 static const pman_nf mans[MAN_MAX] = { 257 NULL, //br 258 NULL, //TH 259 pman_sh, //SH 260 NULL, //SS 261 NULL, //TP 262 NULL, //LP 263 NULL, //PP 264 NULL, //P 265 NULL, //IP 266 NULL, //HP 267 NULL, //SM 268 NULL, //SB 269 NULL, //BI 270 NULL, //IB 271 NULL, //BR 272 NULL, //RB 273 NULL, //R 274 pman_block, //B 275 NULL, //I 276 NULL, //IR 277 NULL, //RI 278 NULL, //na 279 NULL, //sp 280 NULL, //nf 281 NULL, //fi 282 NULL, //RE 283 NULL, //RS 284 NULL, //DT 285 NULL, //UC 286 NULL, //PD 287 NULL, //AT 288 NULL, //in 289 NULL, //ft 290 }; 291 292 293 int 294 main(int argc, char *argv[]) 295 { 296 FILE *file; 297 const char *sqlstr, *manconf = NULL; 298 char *line, *command, *parent; 299 char *errmsg; 300 int ch; 301 struct mparse *mp; 302 sqlite3 *db; 303 ssize_t len; 304 size_t linesize; 305 struct mandb_rec rec; 306 307 while ((ch = getopt(argc, argv, "C:floqv")) != -1) { 308 switch (ch) { 309 case 'C': 310 manconf = optarg; 311 break; 312 case 'f': 313 remove(DBPATH); 314 mflags.recreate = 1; 315 break; 316 case 'l': 317 mflags.limit = 1; 318 break; 319 case 'o': 320 mflags.optimize = 1; 321 break; 322 case 'q': 323 mflags.verbosity = 0; 324 break; 325 case 'v': 326 mflags.verbosity = 2; 327 break; 328 default: 329 usage(); 330 } 331 } 332 333 memset(&rec, 0, sizeof(rec)); 334 335 init_secbuffs(&rec); 336 mp = mparse_alloc(MPARSE_AUTO, MANDOCLEVEL_FATAL, NULL, NULL); 337 338 if ((db = init_db(MANDB_CREATE)) == NULL) 339 exit(EXIT_FAILURE); 340 341 sqlite3_exec(db, "PRAGMA synchronous = 0", NULL, NULL, &errmsg); 342 if (errmsg != NULL) { 343 warnx("%s", errmsg); 344 free(errmsg); 345 close_db(db); 346 exit(EXIT_FAILURE); 347 } 348 349 sqlite3_exec(db, "ATTACH DATABASE \':memory:\' AS metadb", NULL, NULL, 350 &errmsg); 351 if (errmsg != NULL) { 352 warnx("%s", errmsg); 353 free(errmsg); 354 close_db(db); 355 exit(EXIT_FAILURE); 356 } 357 358 if (manconf) { 359 char *arg; 360 size_t command_len = shquote(manconf, NULL, 0) + 1; 361 arg = malloc(command_len ); 362 shquote(manconf, arg, command_len); 363 easprintf(&command, "man -p -C %s", arg); 364 free(arg); 365 } else { 366 command = estrdup("man -p"); 367 } 368 369 /* Call man -p to get the list of man page dirs */ 370 if ((file = popen(command, "r")) == NULL) { 371 close_db(db); 372 err(EXIT_FAILURE, "fopen failed"); 373 } 374 free(command); 375 376 /* Begin the transaction for indexing the pages */ 377 sqlite3_exec(db, "BEGIN", NULL, NULL, &errmsg); 378 if (errmsg != NULL) { 379 warnx("%s", errmsg); 380 free(errmsg); 381 exit(EXIT_FAILURE); 382 } 383 384 sqlstr = "CREATE TABLE metadb.file_cache(device, inode, mtime, parent," 385 " file PRIMARY KEY);" 386 "CREATE UNIQUE INDEX metadb.index_file_cache_dev" 387 " ON file_cache (device, inode)"; 388 389 sqlite3_exec(db, sqlstr, NULL, NULL, &errmsg); 390 if (errmsg != NULL) { 391 warnx("%s", errmsg); 392 free(errmsg); 393 close_db(db); 394 exit(EXIT_FAILURE); 395 } 396 397 if (mflags.verbosity) 398 printf("Building temporary file cache\n"); 399 line = NULL; 400 linesize = 0; 401 while ((len = getline(&line, &linesize, file)) != -1) { 402 /* Replace the new line character at the end of string with '\0' */ 403 line[len - 1] = '\0'; 404 parent = estrdup(line); 405 char *pdir = estrdup(dirname(parent)); 406 free(parent); 407 /* Traverse the man page directories and parse the pages */ 408 traversedir(pdir, line, db, mp); 409 free(pdir); 410 } 411 free(line); 412 413 if (pclose(file) == -1) { 414 close_db(db); 415 cleanup(&rec); 416 free_secbuffs(&rec); 417 err(EXIT_FAILURE, "pclose error"); 418 } 419 420 update_db(db, mp, &rec); 421 mparse_free(mp); 422 free_secbuffs(&rec); 423 424 /* Commit the transaction */ 425 sqlite3_exec(db, "COMMIT", NULL, NULL, &errmsg); 426 if (errmsg != NULL) { 427 warnx("%s", errmsg); 428 free(errmsg); 429 exit(EXIT_FAILURE); 430 } 431 432 if (mflags.optimize) 433 optimize(db); 434 435 close_db(db); 436 return 0; 437 } 438 439 /* 440 * traversedir -- 441 * Traverses the given directory recursively and passes all the man page files 442 * in the way to build_file_cache() 443 */ 444 static void 445 traversedir(const char *parent, const char *file, sqlite3 *db, 446 struct mparse *mp) 447 { 448 struct stat sb; 449 struct dirent *dirp; 450 DIR *dp; 451 char *buf; 452 453 if (stat(file, &sb) < 0) { 454 warn("stat failed: %s", file); 455 return; 456 } 457 458 /* If it is a regular file or a symlink, pass it to build_cache() */ 459 if (S_ISREG(sb.st_mode) || S_ISLNK(sb.st_mode)) { 460 build_file_cache(db, parent, file, &sb); 461 return; 462 } 463 464 /* If it is a directory, traverse it recursively */ 465 if (S_ISDIR(sb.st_mode)) { 466 if ((dp = opendir(file)) == NULL) { 467 warn("opendir error: %s", file); 468 return; 469 } 470 471 while ((dirp = readdir(dp)) != NULL) { 472 /* Avoid . and .. entries in a directory */ 473 if (strncmp(dirp->d_name, ".", 1)) { 474 easprintf(&buf, "%s/%s", file, dirp->d_name); 475 traversedir(parent, buf, db, mp); 476 free(buf); 477 } 478 } 479 closedir(dp); 480 } 481 } 482 483 /* build_file_cache -- 484 * This function generates an md5 hash of the file passed as it's 2nd parameter 485 * and stores it in a temporary table file_cache along with the full file path. 486 * This is done to support incremental updation of the database. 487 * The temporary table file_cache is dropped thereafter in the function 488 * update_db(), once the database has been updated. 489 */ 490 static void 491 build_file_cache(sqlite3 *db, const char *parent, const char *file, 492 struct stat *sb) 493 { 494 const char *sqlstr; 495 sqlite3_stmt *stmt = NULL; 496 int rc, idx; 497 assert(file != NULL); 498 dev_t device_cache = sb->st_dev; 499 ino_t inode_cache = sb->st_ino; 500 time_t mtime_cache = sb->st_mtime; 501 502 sqlstr = "INSERT INTO metadb.file_cache VALUES (:device, :inode," 503 " :mtime, :parent, :file)"; 504 rc = sqlite3_prepare_v2(db, sqlstr, -1, &stmt, NULL); 505 if (rc != SQLITE_OK) { 506 warnx("%s", sqlite3_errmsg(db)); 507 return; 508 } 509 510 idx = sqlite3_bind_parameter_index(stmt, ":device"); 511 rc = sqlite3_bind_int64(stmt, idx, device_cache); 512 if (rc != SQLITE_OK) { 513 warnx("%s", sqlite3_errmsg(db)); 514 sqlite3_finalize(stmt); 515 return; 516 } 517 518 idx = sqlite3_bind_parameter_index(stmt, ":inode"); 519 rc = sqlite3_bind_int64(stmt, idx, inode_cache); 520 if (rc != SQLITE_OK) { 521 warnx("%s", sqlite3_errmsg(db)); 522 sqlite3_finalize(stmt); 523 return; 524 } 525 526 idx = sqlite3_bind_parameter_index(stmt, ":mtime"); 527 rc = sqlite3_bind_int64(stmt, idx, mtime_cache); 528 if (rc != SQLITE_OK) { 529 warnx("%s", sqlite3_errmsg(db)); 530 sqlite3_finalize(stmt); 531 return; 532 } 533 534 idx = sqlite3_bind_parameter_index(stmt, ":parent"); 535 rc = sqlite3_bind_text(stmt, idx, parent, -1, NULL); 536 if (rc != SQLITE_OK) { 537 warnx("%s", sqlite3_errmsg(db)); 538 sqlite3_finalize(stmt); 539 return; 540 } 541 542 idx = sqlite3_bind_parameter_index(stmt, ":file"); 543 rc = sqlite3_bind_text(stmt, idx, file, -1, NULL); 544 if (rc != SQLITE_OK) { 545 warnx("%s", sqlite3_errmsg(db)); 546 sqlite3_finalize(stmt); 547 return; 548 } 549 550 sqlite3_step(stmt); 551 sqlite3_finalize(stmt); 552 } 553 554 static void 555 update_existing_entry(sqlite3 *db, const char *file, const char *hash, 556 mandb_rec *rec, int *new_count, int *link_count, int *err_count) 557 { 558 int update_count, rc, idx; 559 const char *inner_sqlstr; 560 sqlite3_stmt *inner_stmt; 561 562 update_count = sqlite3_total_changes(db); 563 inner_sqlstr = "UPDATE mandb_meta SET device = :device," 564 " inode = :inode, mtime = :mtime WHERE" 565 " md5_hash = :md5 AND file = :file AND" 566 " (device <> :device2 OR inode <> " 567 " :inode2 OR mtime <> :mtime2)"; 568 rc = sqlite3_prepare_v2(db, inner_sqlstr, -1, &inner_stmt, NULL); 569 if (rc != SQLITE_OK) { 570 warnx("%s", sqlite3_errmsg(db)); 571 return; 572 } 573 idx = sqlite3_bind_parameter_index(inner_stmt, ":device"); 574 sqlite3_bind_int64(inner_stmt, idx, rec->device); 575 idx = sqlite3_bind_parameter_index(inner_stmt, ":inode"); 576 sqlite3_bind_int64(inner_stmt, idx, rec->inode); 577 idx = sqlite3_bind_parameter_index(inner_stmt, ":mtime"); 578 sqlite3_bind_int64(inner_stmt, idx, rec->mtime); 579 idx = sqlite3_bind_parameter_index(inner_stmt, ":md5"); 580 sqlite3_bind_text(inner_stmt, idx, hash, -1, NULL); 581 idx = sqlite3_bind_parameter_index(inner_stmt, ":file"); 582 sqlite3_bind_text(inner_stmt, idx, file, -1, NULL); 583 idx = sqlite3_bind_parameter_index(inner_stmt, ":device2"); 584 sqlite3_bind_int64(inner_stmt, idx, rec->device); 585 idx = sqlite3_bind_parameter_index(inner_stmt, ":inode2"); 586 sqlite3_bind_int64(inner_stmt, idx, rec->inode); 587 idx = sqlite3_bind_parameter_index(inner_stmt, ":mtime2"); 588 sqlite3_bind_int64(inner_stmt, idx, rec->mtime); 589 590 rc = sqlite3_step(inner_stmt); 591 if (rc == SQLITE_DONE) { 592 /* Check if an update has been performed. */ 593 if (update_count != sqlite3_total_changes(db)) { 594 if (mflags.verbosity) 595 printf("Updated %s\n", file); 596 (*new_count)++; 597 } else { 598 /* Otherwise it was a hardlink. */ 599 (*link_count)++; 600 } 601 } else { 602 warnx("Could not update the meta data for %s", file); 603 (*err_count)++; 604 } 605 sqlite3_finalize(inner_stmt); 606 } 607 608 /* read_and_decompress -- 609 * Reads the given file into memory. If it is compressed, decompres 610 * it before returning to the caller. 611 */ 612 static int 613 read_and_decompress(const char *file, void **buf, size_t *len) 614 { 615 size_t off; 616 ssize_t r; 617 struct archive *a; 618 struct archive_entry *ae; 619 620 if ((a = archive_read_new()) == NULL) 621 errx(EXIT_FAILURE, "memory allocation failed"); 622 623 if (archive_read_support_compression_all(a) != ARCHIVE_OK || 624 archive_read_support_format_raw(a) != ARCHIVE_OK || 625 archive_read_open_filename(a, file, 65536) != ARCHIVE_OK || 626 archive_read_next_header(a, &ae) != ARCHIVE_OK) 627 goto archive_error; 628 *len = 65536; 629 *buf = emalloc(*len); 630 off = 0; 631 for (;;) { 632 r = archive_read_data(a, (char *)*buf + off, *len - off); 633 if (r == ARCHIVE_OK) { 634 archive_read_close(a); 635 *len = off; 636 return 0; 637 } 638 if (r <= 0) { 639 free(*buf); 640 break; 641 } 642 off += r; 643 if (off == *len) { 644 *len *= 2; 645 if (*len < off) { 646 warnx("File too large: %s", file); 647 free(*buf); 648 archive_read_close(a); 649 return -1; 650 } 651 *buf = erealloc(*buf, *len); 652 } 653 } 654 655 archive_error: 656 warnx("Error while reading `%s': %s", file, archive_error_string(a)); 657 archive_read_close(a); 658 return -1; 659 } 660 661 /* update_db -- 662 * Does an incremental updation of the database by checking the file_cache. 663 * It parses and adds the pages which are present in file_cache, 664 * but not in the database. 665 * It also removes the pages which are present in the databse, 666 * but not in the file_cache. 667 */ 668 static void 669 update_db(sqlite3 *db, struct mparse *mp, mandb_rec *rec) 670 { 671 const char *sqlstr; 672 sqlite3_stmt *stmt = NULL; 673 const char *file; 674 const char *parent; 675 char *errmsg = NULL; 676 char *md5sum; 677 void *buf; 678 size_t buflen; 679 int new_count = 0; /* Counter for newly indexed/updated pages */ 680 int total_count = 0; /* Counter for total number of pages */ 681 int err_count = 0; /* Counter for number of failed pages */ 682 int link_count = 0; /* Counter for number of hard/sym links */ 683 int md5_status; 684 int rc; 685 686 sqlstr = "SELECT device, inode, mtime, parent, file" 687 " FROM metadb.file_cache fc" 688 " WHERE NOT EXISTS(SELECT 1 FROM mandb_meta WHERE" 689 " device = fc.device AND inode = fc.inode AND " 690 " mtime = fc.mtime AND file = fc.file)"; 691 692 rc = sqlite3_prepare_v2(db, sqlstr, -1, &stmt, NULL); 693 if (rc != SQLITE_OK) { 694 warnx("%s", sqlite3_errmsg(db)); 695 close_db(db); 696 errx(EXIT_FAILURE, "Could not query file cache"); 697 } 698 699 buf = NULL; 700 while (sqlite3_step(stmt) == SQLITE_ROW) { 701 free(buf); 702 total_count++; 703 rec->device = sqlite3_column_int64(stmt, 0); 704 rec->inode = sqlite3_column_int64(stmt, 1); 705 rec->mtime = sqlite3_column_int64(stmt, 2); 706 parent = (const char *) sqlite3_column_text(stmt, 3); 707 file = (const char *) sqlite3_column_text(stmt, 4); 708 if (read_and_decompress(file, &buf, &buflen)) { 709 err_count++; 710 buf = NULL; 711 continue; 712 } 713 md5_status = check_md5(file, db, "mandb_meta", &md5sum, buf, buflen); 714 assert(md5sum != NULL); 715 if (md5_status == -1) { 716 warnx("An error occurred in checking md5 value" 717 " for file %s", file); 718 err_count++; 719 continue; 720 } 721 722 if (md5_status == 0) { 723 /* 724 * The MD5 hash is already present in the database, 725 * so simply update the metadata, ignoring symlinks. 726 */ 727 struct stat sb; 728 stat(file, &sb); 729 if (S_ISLNK(sb.st_mode)) { 730 free(md5sum); 731 link_count++; 732 continue; 733 } 734 update_existing_entry(db, file, md5sum, rec, 735 &new_count, &link_count, &err_count); 736 free(md5sum); 737 continue; 738 } 739 740 if (md5_status == 1) { 741 /* 742 * The MD5 hash was not present in the database. 743 * This means is either a new file or an updated file. 744 * We should go ahead with parsing. 745 */ 746 if (mflags.verbosity > 1) 747 printf("Parsing: %s\n", file); 748 rec->md5_hash = md5sum; 749 rec->file_path = estrdup(file); 750 // file_path is freed by insert_into_db itself. 751 chdir(parent); 752 begin_parse(file, mp, rec, buf, buflen); 753 if (insert_into_db(db, rec) < 0) { 754 warnx("Error in indexing %s", file); 755 err_count++; 756 } else { 757 new_count++; 758 } 759 } 760 } 761 free(buf); 762 763 sqlite3_finalize(stmt); 764 765 if (mflags.verbosity) { 766 printf("Total Number of new or updated pages enountered = %d\n" 767 "Total number of pages that were successfully" 768 " indexed/updated = %d\n" 769 "Total number of (hard or symbolic) links found = %d\n" 770 "Total number of pages that could not be indexed" 771 " due to errors = %d\n", 772 total_count, new_count, link_count, err_count); 773 } 774 775 if (mflags.recreate == 0) 776 return; 777 778 if (mflags.verbosity) 779 printf("Deleting stale index entries\n"); 780 781 sqlstr = "DELETE FROM mandb_meta WHERE file NOT IN" 782 " (SELECT file FROM metadb.file_cache);" 783 "DROP TABLE metadb.file_cache;" 784 "DELETE FROM mandb WHERE rowid NOT IN" 785 " (SELECT id FROM mandb_meta);"; 786 787 sqlite3_exec(db, sqlstr, NULL, NULL, &errmsg); 788 if (errmsg != NULL) { 789 warnx("Removing old entries failed: %s", errmsg); 790 warnx("Please rebuild database from scratch with -f."); 791 free(errmsg); 792 return; 793 } 794 } 795 796 /* 797 * begin_parse -- 798 * parses the man page using libmandoc 799 */ 800 static void 801 begin_parse(const char *file, struct mparse *mp, mandb_rec *rec, 802 const void *buf, size_t len) 803 { 804 struct mdoc *mdoc; 805 struct man *man; 806 mparse_reset(mp); 807 808 rec->xr_found = 0; 809 810 if (mparse_readmem(mp, buf, len, file) >= MANDOCLEVEL_FATAL) { 811 warnx("%s: Parse failure", file); 812 return; 813 } 814 815 mparse_result(mp, &mdoc, &man); 816 if (mdoc == NULL && man == NULL) { 817 warnx("Not a man(7) or mdoc(7) page"); 818 return; 819 } 820 821 set_machine(mdoc, rec); 822 set_section(mdoc, man, rec); 823 if (mdoc) { 824 rec->page_type = MDOC; 825 pmdoc_node(mdoc_node(mdoc), rec); 826 } else { 827 rec->page_type = MAN; 828 pman_node(man_node(man), rec); 829 } 830 } 831 832 /* 833 * set_section -- 834 * Extracts the section number and normalizes it to only the numeric part 835 * (Which should be the first character of the string). 836 */ 837 static void 838 set_section(const struct mdoc *md, const struct man *m, mandb_rec *rec) 839 { 840 if (md) { 841 const struct mdoc_meta *md_meta = mdoc_meta(md); 842 rec->section[0] = md_meta->msec[0]; 843 } else if (m) { 844 const struct man_meta *m_meta = man_meta(m); 845 rec->section[0] = m_meta->msec[0]; 846 } 847 } 848 849 /* 850 * get_machine -- 851 * Extracts the machine architecture information if available. 852 */ 853 static void 854 set_machine(const struct mdoc *md, mandb_rec *rec) 855 { 856 if (md == NULL) 857 return; 858 const struct mdoc_meta *md_meta = mdoc_meta(md); 859 if (md_meta->arch) 860 rec->machine = estrdup(md_meta->arch); 861 } 862 863 static void 864 pmdoc_node(const struct mdoc_node *n, mandb_rec *rec) 865 { 866 867 if (n == NULL) 868 return; 869 870 switch (n->type) { 871 case (MDOC_BODY): 872 /* FALLTHROUGH */ 873 case (MDOC_TAIL): 874 /* FALLTHROUGH */ 875 case (MDOC_ELEM): 876 if (mdocs[n->tok] == NULL) 877 break; 878 (*mdocs[n->tok])(n, rec); 879 break; 880 default: 881 break; 882 } 883 884 pmdoc_node(n->child, rec); 885 pmdoc_node(n->next, rec); 886 } 887 888 /* 889 * pmdoc_Nm -- 890 * Extracts the Name of the manual page from the .Nm macro 891 */ 892 static void 893 pmdoc_Nm(const struct mdoc_node *n, mandb_rec *rec) 894 { 895 if (n->sec != SEC_NAME) 896 return; 897 898 for (n = n->child; n; n = n->next) { 899 if (n->type == MDOC_TEXT) { 900 concat(&rec->name, n->string); 901 } 902 } 903 } 904 905 /* 906 * pmdoc_Nd -- 907 * Extracts the one line description of the man page from the .Nd macro 908 */ 909 static void 910 pmdoc_Nd(const struct mdoc_node *n, mandb_rec *rec) 911 { 912 /* 913 * A static variable for keeping track of whether a Xr macro was seen 914 * previously. 915 */ 916 char *buf = NULL; 917 char *temp; 918 919 if (n == NULL) 920 return; 921 922 if (n->type == MDOC_TEXT) { 923 if (rec->xr_found && n->next) { 924 /* 925 * An Xr macro was seen previously, so parse this 926 * and the next node. 927 */ 928 temp = estrdup(n->string); 929 n = n->next; 930 easprintf(&buf, "%s(%s)", temp, n->string); 931 concat(&rec->name_desc, buf); 932 free(buf); 933 free(temp); 934 } else { 935 concat(&rec->name_desc, n->string); 936 } 937 rec->xr_found = 0; 938 } else if (mdocs[n->tok] == pmdoc_Xr) { 939 /* Remember that we have encountered an Xr macro */ 940 rec->xr_found = 1; 941 } 942 943 if (n->child) 944 pmdoc_Nd(n->child, rec); 945 946 if(n->next) 947 pmdoc_Nd(n->next, rec); 948 } 949 950 /* 951 * pmdoc_macro_handler-- 952 * This function is a single point of handling all the special macros that we 953 * want to handle especially. For example the .Xr macro for properly parsing 954 * the referenced page name along with the section number, or the .Pp macro 955 * for adding a new line whenever we encounter it. 956 */ 957 static void 958 pmdoc_macro_handler(const struct mdoc_node *n, mandb_rec *rec, enum mdoct doct) 959 { 960 const struct mdoc_node *sn; 961 assert(n); 962 963 switch (doct) { 964 /* Parse the man page references. 965 * Basically the .Xr macros are used like: 966 * .Xr ls 1 967 * and formatted like this: 968 * ls(1) 969 * Prepare a buffer to format the data like the above example and call 970 * pmdoc_parse_section to append it. 971 */ 972 case MDOC_Xr: 973 n = n->child; 974 while (n->type != MDOC_TEXT && n->next) 975 n = n->next; 976 977 if (n && n->type != MDOC_TEXT) 978 return; 979 sn = n; 980 if (n->next) 981 n = n->next; 982 983 while (n->type != MDOC_TEXT && n->next) 984 n = n->next; 985 986 if (n && n->type == MDOC_TEXT) { 987 size_t len = strlen(sn->string); 988 char *buf = emalloc(len + 4); 989 memcpy(buf, sn->string, len); 990 buf[len] = '('; 991 buf[len + 1] = n->string[0]; 992 buf[len + 2] = ')'; 993 buf[len + 3] = 0; 994 mdoc_parse_section(n->sec, buf, rec); 995 free(buf); 996 } 997 998 break; 999 1000 /* Parse the .Pp macro to add a new line */ 1001 case MDOC_Pp: 1002 if (n->type == MDOC_TEXT) 1003 mdoc_parse_section(n->sec, "\n", rec); 1004 break; 1005 default: 1006 break; 1007 } 1008 1009 } 1010 1011 /* 1012 * pmdoc_Xr, pmdoc_Pp-- 1013 * Empty stubs. 1014 * The parser calls these functions each time it encounters 1015 * a .Xr or .Pp macro. We are parsing all the data from 1016 * the pmdoc_Sh function, so don't do anything here. 1017 * (See if else blocks in pmdoc_Sh.) 1018 */ 1019 static void 1020 pmdoc_Xr(const struct mdoc_node *n, mandb_rec *rec) 1021 { 1022 } 1023 1024 static void 1025 pmdoc_Pp(const struct mdoc_node *n, mandb_rec *rec) 1026 { 1027 } 1028 1029 /* 1030 * pmdoc_Sh -- 1031 * Called when a .Sh macro is encountered and loops through its body, calling 1032 * mdoc_parse_section to append the data to the section specific buffer. 1033 * Two special macros which may occur inside the body of Sh are .Nm and .Xr and 1034 * they need special handling, thus the separate if branches for them. 1035 */ 1036 static void 1037 pmdoc_Sh(const struct mdoc_node *n, mandb_rec *rec) 1038 { 1039 if (n == NULL) 1040 return; 1041 int xr_found = 0; 1042 1043 if (n->type == MDOC_TEXT) { 1044 mdoc_parse_section(n->sec, n->string, rec); 1045 } else if (mdocs[n->tok] == pmdoc_Nm && rec->name != NULL) { 1046 /* 1047 * When encountering a .Nm macro, substitute it 1048 * with its previously cached value of the argument. 1049 */ 1050 mdoc_parse_section(n->sec, rec->name, rec); 1051 } else if (mdocs[n->tok] == pmdoc_Xr) { 1052 /* 1053 * When encountering other inline macros, 1054 * call pmdoc_macro_handler. 1055 */ 1056 pmdoc_macro_handler(n, rec, MDOC_Xr); 1057 xr_found = 1; 1058 } else if (mdocs[n->tok] == pmdoc_Pp) { 1059 pmdoc_macro_handler(n, rec, MDOC_Pp); 1060 } 1061 1062 /* 1063 * If an Xr macro was encountered then the child node has 1064 * already been explored by pmdoc_macro_handler. 1065 */ 1066 if (xr_found == 0) 1067 pmdoc_Sh(n->child, rec); 1068 pmdoc_Sh(n->next, rec); 1069 } 1070 1071 /* 1072 * mdoc_parse_section-- 1073 * Utility function for parsing sections of the mdoc type pages. 1074 * Takes two params: 1075 * 1. sec is an enum which indicates the section in which we are present 1076 * 2. string is the string which we need to append to the secbuff for this 1077 * particular section. 1078 * The function appends string to the global section buffer and returns. 1079 */ 1080 static void 1081 mdoc_parse_section(enum mdoc_sec sec, const char *string, mandb_rec *rec) 1082 { 1083 /* 1084 * If the user specified the 'l' flag, then parse and store only the 1085 * NAME section. Ignore the rest. 1086 */ 1087 if (mflags.limit) 1088 return; 1089 1090 switch (sec) { 1091 case SEC_LIBRARY: 1092 append(&rec->lib, string); 1093 break; 1094 case SEC_RETURN_VALUES: 1095 append(&rec->return_vals, string); 1096 break; 1097 case SEC_ENVIRONMENT: 1098 append(&rec->env, string); 1099 break; 1100 case SEC_FILES: 1101 append(&rec->files, string); 1102 break; 1103 case SEC_EXIT_STATUS: 1104 append(&rec->exit_status, string); 1105 break; 1106 case SEC_DIAGNOSTICS: 1107 append(&rec->diagnostics, string); 1108 break; 1109 case SEC_ERRORS: 1110 append(&rec->errors, string); 1111 break; 1112 case SEC_NAME: 1113 case SEC_SYNOPSIS: 1114 case SEC_EXAMPLES: 1115 case SEC_STANDARDS: 1116 case SEC_HISTORY: 1117 case SEC_AUTHORS: 1118 case SEC_BUGS: 1119 break; 1120 default: 1121 append(&rec->desc, string); 1122 break; 1123 } 1124 } 1125 1126 static void 1127 pman_node(const struct man_node *n, mandb_rec *rec) 1128 { 1129 if (n == NULL) 1130 return; 1131 1132 switch (n->type) { 1133 case (MAN_BODY): 1134 /* FALLTHROUGH */ 1135 case (MAN_TAIL): 1136 /* FALLTHROUGH */ 1137 case (MAN_BLOCK): 1138 /* FALLTHROUGH */ 1139 case (MAN_ELEM): 1140 if (mans[n->tok] != NULL) 1141 (*mans[n->tok])(n, rec); 1142 break; 1143 default: 1144 break; 1145 } 1146 1147 pman_node(n->child, rec); 1148 pman_node(n->next, rec); 1149 } 1150 1151 /* 1152 * pman_parse_name -- 1153 * Parses the NAME section and puts the complete content in the name_desc 1154 * variable. 1155 */ 1156 static void 1157 pman_parse_name(const struct man_node *n, mandb_rec *rec) 1158 { 1159 if (n == NULL) 1160 return; 1161 1162 if (n->type == MAN_TEXT) { 1163 char *tmp = parse_escape(n->string); 1164 concat(&rec->name_desc, tmp); 1165 free(tmp); 1166 } 1167 1168 if (n->child) 1169 pman_parse_name(n->child, rec); 1170 1171 if(n->next) 1172 pman_parse_name(n->next, rec); 1173 } 1174 1175 /* 1176 * A stub function to be able to parse the macros like .B embedded inside 1177 * a section. 1178 */ 1179 static void 1180 pman_block(const struct man_node *n, mandb_rec *rec) 1181 { 1182 } 1183 1184 /* 1185 * pman_sh -- 1186 * This function does one of the two things: 1187 * 1. If the present section is NAME, then it will: 1188 * (a) Extract the name of the page (in case of multiple comma separated 1189 * names, it will pick up the first one). 1190 * (b) Build a space spearated list of all the symlinks/hardlinks to 1191 * this page and store in the buffer 'links'. These are extracted from 1192 * the comma separated list of names in the NAME section as well. 1193 * (c) Move on to the one line description section, which is after the list 1194 * of names in the NAME section. 1195 * 2. Otherwise, it will check the section name and call the man_parse_section 1196 * function, passing the enum corresponding that section. 1197 */ 1198 static void 1199 pman_sh(const struct man_node *n, mandb_rec *rec) 1200 { 1201 static const struct { 1202 enum man_sec section; 1203 const char *header; 1204 } mapping[] = { 1205 { MANSEC_DESCRIPTION, "DESCRIPTION" }, 1206 { MANSEC_SYNOPSIS, "SYNOPSIS" }, 1207 { MANSEC_LIBRARY, "LIBRARY" }, 1208 { MANSEC_ERRORS, "ERRORS" }, 1209 { MANSEC_FILES, "FILES" }, 1210 { MANSEC_RETURN_VALUES, "RETURN VALUE" }, 1211 { MANSEC_RETURN_VALUES, "RETURN VALUES" }, 1212 { MANSEC_EXIT_STATUS, "EXIT STATUS" }, 1213 { MANSEC_EXAMPLES, "EXAMPLES" }, 1214 { MANSEC_EXAMPLES, "EXAMPLE" }, 1215 { MANSEC_STANDARDS, "STANDARDS" }, 1216 { MANSEC_HISTORY, "HISTORY" }, 1217 { MANSEC_BUGS, "BUGS" }, 1218 { MANSEC_AUTHORS, "AUTHORS" }, 1219 { MANSEC_COPYRIGHT, "COPYRIGHT" }, 1220 }; 1221 const struct man_node *head; 1222 char *name_desc; 1223 int sz; 1224 size_t i; 1225 1226 if ((head = n->parent->head) == NULL || (head = head->child) == NULL || 1227 head->type != MAN_TEXT) 1228 return; 1229 1230 /* 1231 * Check if this section should be extracted and 1232 * where it should be stored. Handled the trival cases first. 1233 */ 1234 for (i = 0; i < sizeof(mapping) / sizeof(mapping[0]); ++i) { 1235 if (strcmp(head->string, mapping[i].header) == 0) { 1236 man_parse_section(mapping[i].section, n, rec); 1237 return; 1238 } 1239 } 1240 1241 if (strcmp(head->string, "NAME") == 0) { 1242 /* 1243 * We are in the NAME section. 1244 * pman_parse_name will put the complete content in name_desc. 1245 */ 1246 pman_parse_name(n, rec); 1247 1248 name_desc = rec->name_desc; 1249 1250 /* Remove any leading spaces. */ 1251 while (name_desc[0] == ' ') 1252 name_desc++; 1253 1254 /* If the line begins with a "\&", avoid those */ 1255 if (name_desc[0] == '\\' && name_desc[1] == '&') 1256 name_desc += 2; 1257 1258 /* Now name_desc should be left with a comma-space 1259 * separated list of names and the one line description 1260 * of the page: 1261 * "a, b, c \- sample description" 1262 * Take out the first name, before the first comma 1263 * (or space) and store it in rec->name. 1264 * If the page has aliases then they should be 1265 * in the form of a comma separated list. 1266 * Keep looping while there is a comma in name_desc, 1267 * extract the alias name and store in rec->links. 1268 * When there are no more commas left, break out. 1269 */ 1270 int has_alias = 0; // Any more aliases left? 1271 while (*name_desc) { 1272 /* Remove any leading spaces. */ 1273 if (name_desc[0] == ' ') { 1274 name_desc++; 1275 continue; 1276 } 1277 sz = strcspn(name_desc, ", "); 1278 1279 /* Extract the first term and store it in rec->name. */ 1280 if (rec->name == NULL) { 1281 if (name_desc[sz] == ',') 1282 has_alias = 1; 1283 name_desc[sz] = 0; 1284 rec->name = emalloc(sz + 1); 1285 memcpy(rec->name, name_desc, sz + 1); 1286 name_desc += sz + 1; 1287 continue; 1288 } 1289 1290 /* 1291 * Once rec->name is set, rest of the names 1292 * are to be treated as links or aliases. 1293 */ 1294 if (rec->name && has_alias) { 1295 if (name_desc[sz] != ',') { 1296 /* No more commas left --> 1297 * no more aliases to take out 1298 */ 1299 has_alias = 0; 1300 } 1301 name_desc[sz] = 0; 1302 concat2(&rec->links, name_desc, sz); 1303 name_desc += sz + 1; 1304 continue; 1305 } 1306 break; 1307 } 1308 1309 /* Parse any escape sequences that might be there */ 1310 char *temp = parse_escape(name_desc); 1311 free(rec->name_desc); 1312 rec->name_desc = temp; 1313 temp = parse_escape(rec->name); 1314 free(rec->name); 1315 rec->name = temp; 1316 return; 1317 } 1318 1319 /* The RETURN VALUE section might be specified in multiple ways */ 1320 if (strcmp(head->string, "RETURN") == 0 && 1321 head->next != NULL && head->next->type == MAN_TEXT && 1322 (strcmp(head->next->string, "VALUE") == 0 || 1323 strcmp(head->next->string, "VALUES") == 0)) { 1324 man_parse_section(MANSEC_RETURN_VALUES, n, rec); 1325 return; 1326 } 1327 1328 /* 1329 * EXIT STATUS section can also be specified all on one line or on two 1330 * separate lines. 1331 */ 1332 if (strcmp(head->string, "EXIT") == 0 && 1333 head->next != NULL && head->next->type == MAN_TEXT && 1334 strcmp(head->next->string, "STATUS") == 0) { 1335 man_parse_section(MANSEC_EXIT_STATUS, n, rec); 1336 return; 1337 } 1338 1339 /* Store the rest of the content in desc. */ 1340 man_parse_section(MANSEC_NONE, n, rec); 1341 } 1342 1343 /* 1344 * pman_parse_node -- 1345 * Generic function to iterate through a node. Usually called from 1346 * man_parse_section to parse a particular section of the man page. 1347 */ 1348 static void 1349 pman_parse_node(const struct man_node *n, secbuff *s) 1350 { 1351 if (n == NULL) 1352 return; 1353 1354 if (n->type == MAN_TEXT) 1355 append(s, n->string); 1356 1357 pman_parse_node(n->child, s); 1358 pman_parse_node(n->next, s); 1359 } 1360 1361 /* 1362 * man_parse_section -- 1363 * Takes two parameters: 1364 * sec: Tells which section we are present in 1365 * n: Is the present node of the AST. 1366 * Depending on the section, we call pman_parse_node to parse that section and 1367 * concatenate the content from that section into the buffer for that section. 1368 */ 1369 static void 1370 man_parse_section(enum man_sec sec, const struct man_node *n, mandb_rec *rec) 1371 { 1372 /* 1373 * If the user sepecified the 'l' flag then just parse 1374 * the NAME section, ignore the rest. 1375 */ 1376 if (mflags.limit) 1377 return; 1378 1379 switch (sec) { 1380 case MANSEC_LIBRARY: 1381 pman_parse_node(n, &rec->lib); 1382 break; 1383 case MANSEC_RETURN_VALUES: 1384 pman_parse_node(n, &rec->return_vals); 1385 break; 1386 case MANSEC_ENVIRONMENT: 1387 pman_parse_node(n, &rec->env); 1388 break; 1389 case MANSEC_FILES: 1390 pman_parse_node(n, &rec->files); 1391 break; 1392 case MANSEC_EXIT_STATUS: 1393 pman_parse_node(n, &rec->exit_status); 1394 break; 1395 case MANSEC_DIAGNOSTICS: 1396 pman_parse_node(n, &rec->diagnostics); 1397 break; 1398 case MANSEC_ERRORS: 1399 pman_parse_node(n, &rec->errors); 1400 break; 1401 case MANSEC_NAME: 1402 case MANSEC_SYNOPSIS: 1403 case MANSEC_EXAMPLES: 1404 case MANSEC_STANDARDS: 1405 case MANSEC_HISTORY: 1406 case MANSEC_BUGS: 1407 case MANSEC_AUTHORS: 1408 case MANSEC_COPYRIGHT: 1409 break; 1410 default: 1411 pman_parse_node(n, &rec->desc); 1412 break; 1413 } 1414 1415 } 1416 1417 /* 1418 * insert_into_db -- 1419 * Inserts the parsed data of the man page in the Sqlite databse. 1420 * If any of the values is NULL, then we cleanup and return -1 indicating 1421 * an error. 1422 * Otherwise, store the data in the database and return 0. 1423 */ 1424 static int 1425 insert_into_db(sqlite3 *db, mandb_rec *rec) 1426 { 1427 int rc = 0; 1428 int idx = -1; 1429 const char *sqlstr = NULL; 1430 sqlite3_stmt *stmt = NULL; 1431 char *ln = NULL; 1432 char *errmsg = NULL; 1433 long int mandb_rowid; 1434 1435 /* 1436 * At the very minimum we want to make sure that we store 1437 * the following data: 1438 * Name, one line description, and the MD5 hash 1439 */ 1440 if (rec->name == NULL || rec->name_desc == NULL || 1441 rec->md5_hash == NULL) { 1442 cleanup(rec); 1443 return -1; 1444 } 1445 1446 /* Write null byte at the end of all the sec_buffs */ 1447 rec->desc.data[rec->desc.offset] = 0; 1448 rec->lib.data[rec->lib.offset] = 0; 1449 rec->env.data[rec->env.offset] = 0; 1450 rec->return_vals.data[rec->return_vals.offset] = 0; 1451 rec->exit_status.data[rec->exit_status.offset] = 0; 1452 rec->files.data[rec->files.offset] = 0; 1453 rec->diagnostics.data[rec->diagnostics.offset] = 0; 1454 rec->errors.data[rec->errors.offset] = 0; 1455 1456 /* 1457 * In case of a mdoc page: (sorry, no better place to put this code) 1458 * parse the comma separated list of names of man pages, 1459 * the first name will be stored in the mandb table, rest will be 1460 * treated as links and put in the mandb_links table. 1461 */ 1462 if (rec->page_type == MDOC) { 1463 char *tmp; 1464 rec->links = estrdup(rec->name); 1465 free(rec->name); 1466 int sz = strcspn(rec->links, " \0"); 1467 rec->name = emalloc(sz + 1); 1468 memcpy(rec->name, rec->links, sz); 1469 if(rec->name[sz - 1] == ',') 1470 rec->name[sz - 1] = 0; 1471 else 1472 rec->name[sz] = 0; 1473 while (rec->links[sz] == ' ') 1474 ++sz; 1475 tmp = estrdup(rec->links + sz); 1476 free(rec->links); 1477 rec->links = tmp; 1478 } 1479 1480 /*------------------------ Populate the mandb table---------------------------*/ 1481 sqlstr = "INSERT INTO mandb VALUES (:section, :name, :name_desc, :desc," 1482 " :lib, :return_vals, :env, :files, :exit_status," 1483 " :diagnostics, :errors, :md5_hash, :machine)"; 1484 1485 rc = sqlite3_prepare_v2(db, sqlstr, -1, &stmt, NULL); 1486 if (rc != SQLITE_OK) 1487 goto Out; 1488 1489 idx = sqlite3_bind_parameter_index(stmt, ":name"); 1490 rc = sqlite3_bind_text(stmt, idx, rec->name, -1, NULL); 1491 if (rc != SQLITE_OK) { 1492 sqlite3_finalize(stmt); 1493 goto Out; 1494 } 1495 1496 idx = sqlite3_bind_parameter_index(stmt, ":section"); 1497 rc = sqlite3_bind_text(stmt, idx, rec->section, -1, NULL); 1498 if (rc != SQLITE_OK) { 1499 sqlite3_finalize(stmt); 1500 goto Out; 1501 } 1502 1503 idx = sqlite3_bind_parameter_index(stmt, ":name_desc"); 1504 rc = sqlite3_bind_text(stmt, idx, rec->name_desc, -1, NULL); 1505 if (rc != SQLITE_OK) { 1506 sqlite3_finalize(stmt); 1507 goto Out; 1508 } 1509 1510 idx = sqlite3_bind_parameter_index(stmt, ":desc"); 1511 rc = sqlite3_bind_text(stmt, idx, rec->desc.data, 1512 rec->desc.offset + 1, NULL); 1513 if (rc != SQLITE_OK) { 1514 sqlite3_finalize(stmt); 1515 goto Out; 1516 } 1517 1518 idx = sqlite3_bind_parameter_index(stmt, ":lib"); 1519 rc = sqlite3_bind_text(stmt, idx, rec->lib.data, rec->lib.offset + 1, NULL); 1520 if (rc != SQLITE_OK) { 1521 sqlite3_finalize(stmt); 1522 goto Out; 1523 } 1524 1525 idx = sqlite3_bind_parameter_index(stmt, ":return_vals"); 1526 rc = sqlite3_bind_text(stmt, idx, rec->return_vals.data, 1527 rec->return_vals.offset + 1, NULL); 1528 if (rc != SQLITE_OK) { 1529 sqlite3_finalize(stmt); 1530 goto Out; 1531 } 1532 1533 idx = sqlite3_bind_parameter_index(stmt, ":env"); 1534 rc = sqlite3_bind_text(stmt, idx, rec->env.data, rec->env.offset + 1, NULL); 1535 if (rc != SQLITE_OK) { 1536 sqlite3_finalize(stmt); 1537 goto Out; 1538 } 1539 1540 idx = sqlite3_bind_parameter_index(stmt, ":files"); 1541 rc = sqlite3_bind_text(stmt, idx, rec->files.data, 1542 rec->files.offset + 1, NULL); 1543 if (rc != SQLITE_OK) { 1544 sqlite3_finalize(stmt); 1545 goto Out; 1546 } 1547 1548 idx = sqlite3_bind_parameter_index(stmt, ":exit_status"); 1549 rc = sqlite3_bind_text(stmt, idx, rec->exit_status.data, 1550 rec->exit_status.offset + 1, NULL); 1551 if (rc != SQLITE_OK) { 1552 sqlite3_finalize(stmt); 1553 goto Out; 1554 } 1555 1556 idx = sqlite3_bind_parameter_index(stmt, ":diagnostics"); 1557 rc = sqlite3_bind_text(stmt, idx, rec->diagnostics.data, 1558 rec->diagnostics.offset + 1, NULL); 1559 if (rc != SQLITE_OK) { 1560 sqlite3_finalize(stmt); 1561 goto Out; 1562 } 1563 1564 idx = sqlite3_bind_parameter_index(stmt, ":errors"); 1565 rc = sqlite3_bind_text(stmt, idx, rec->errors.data, 1566 rec->errors.offset + 1, NULL); 1567 if (rc != SQLITE_OK) { 1568 sqlite3_finalize(stmt); 1569 goto Out; 1570 } 1571 1572 idx = sqlite3_bind_parameter_index(stmt, ":md5_hash"); 1573 rc = sqlite3_bind_text(stmt, idx, rec->md5_hash, -1, NULL); 1574 if (rc != SQLITE_OK) { 1575 sqlite3_finalize(stmt); 1576 goto Out; 1577 } 1578 1579 idx = sqlite3_bind_parameter_index(stmt, ":machine"); 1580 if (rec->machine) 1581 rc = sqlite3_bind_text(stmt, idx, rec->machine, -1, NULL); 1582 else 1583 rc = sqlite3_bind_null(stmt, idx); 1584 if (rc != SQLITE_OK) { 1585 sqlite3_finalize(stmt); 1586 goto Out; 1587 } 1588 1589 rc = sqlite3_step(stmt); 1590 if (rc != SQLITE_DONE) { 1591 sqlite3_finalize(stmt); 1592 goto Out; 1593 } 1594 1595 sqlite3_finalize(stmt); 1596 1597 /* Get the row id of the last inserted row */ 1598 mandb_rowid = sqlite3_last_insert_rowid(db); 1599 1600 /*------------------------Populate the mandb_meta table-----------------------*/ 1601 sqlstr = "INSERT INTO mandb_meta VALUES (:device, :inode, :mtime," 1602 " :file, :md5_hash, :id)"; 1603 rc = sqlite3_prepare_v2(db, sqlstr, -1, &stmt, NULL); 1604 if (rc != SQLITE_OK) 1605 goto Out; 1606 1607 idx = sqlite3_bind_parameter_index(stmt, ":device"); 1608 rc = sqlite3_bind_int64(stmt, idx, rec->device); 1609 if (rc != SQLITE_OK) { 1610 sqlite3_finalize(stmt); 1611 goto Out; 1612 } 1613 1614 idx = sqlite3_bind_parameter_index(stmt, ":inode"); 1615 rc = sqlite3_bind_int64(stmt, idx, rec->inode); 1616 if (rc != SQLITE_OK) { 1617 sqlite3_finalize(stmt); 1618 goto Out; 1619 } 1620 1621 idx = sqlite3_bind_parameter_index(stmt, ":mtime"); 1622 rc = sqlite3_bind_int64(stmt, idx, rec->mtime); 1623 if (rc != SQLITE_OK) { 1624 sqlite3_finalize(stmt); 1625 goto Out; 1626 } 1627 1628 idx = sqlite3_bind_parameter_index(stmt, ":file"); 1629 rc = sqlite3_bind_text(stmt, idx, rec->file_path, -1, NULL); 1630 if (rc != SQLITE_OK) { 1631 sqlite3_finalize(stmt); 1632 goto Out; 1633 } 1634 1635 idx = sqlite3_bind_parameter_index(stmt, ":md5_hash"); 1636 rc = sqlite3_bind_text(stmt, idx, rec->md5_hash, -1, NULL); 1637 if (rc != SQLITE_OK) { 1638 sqlite3_finalize(stmt); 1639 goto Out; 1640 } 1641 1642 idx = sqlite3_bind_parameter_index(stmt, ":id"); 1643 rc = sqlite3_bind_int64(stmt, idx, mandb_rowid); 1644 if (rc != SQLITE_OK) { 1645 sqlite3_finalize(stmt); 1646 goto Out; 1647 } 1648 1649 rc = sqlite3_step(stmt); 1650 sqlite3_finalize(stmt); 1651 if (rc == SQLITE_CONSTRAINT) { 1652 /* The *most* probable reason for reaching here is that 1653 * the UNIQUE contraint on the file column of the mandb_meta 1654 * table was violated. 1655 * This can happen when a file was updated/modified. 1656 * To fix this we need to do two things: 1657 * 1. Delete the row for the older version of this file 1658 * from mandb table. 1659 * 2. Run an UPDATE query to update the row for this file 1660 * in the mandb_meta table. 1661 */ 1662 warnx("Trying to update index for %s", rec->file_path); 1663 char *sql = sqlite3_mprintf("DELETE FROM mandb " 1664 "WHERE rowid = (SELECT id" 1665 " FROM mandb_meta" 1666 " WHERE file = %Q)", 1667 rec->file_path); 1668 sqlite3_exec(db, sql, NULL, NULL, &errmsg); 1669 sqlite3_free(sql); 1670 if (errmsg != NULL) { 1671 warnx("%s", errmsg); 1672 free(errmsg); 1673 } 1674 sqlstr = "UPDATE mandb_meta SET device = :device," 1675 " inode = :inode, mtime = :mtime, id = :id," 1676 " md5_hash = :md5 WHERE file = :file"; 1677 rc = sqlite3_prepare_v2(db, sqlstr, -1, &stmt, NULL); 1678 if (rc != SQLITE_OK) { 1679 warnx("Update failed with error: %s", 1680 sqlite3_errmsg(db)); 1681 close_db(db); 1682 cleanup(rec); 1683 errx(EXIT_FAILURE, 1684 "Consider running makemandb with -f option"); 1685 } 1686 1687 idx = sqlite3_bind_parameter_index(stmt, ":device"); 1688 sqlite3_bind_int64(stmt, idx, rec->device); 1689 idx = sqlite3_bind_parameter_index(stmt, ":inode"); 1690 sqlite3_bind_int64(stmt, idx, rec->inode); 1691 idx = sqlite3_bind_parameter_index(stmt, ":mtime"); 1692 sqlite3_bind_int64(stmt, idx, rec->mtime); 1693 idx = sqlite3_bind_parameter_index(stmt, ":id"); 1694 sqlite3_bind_int64(stmt, idx, mandb_rowid); 1695 idx = sqlite3_bind_parameter_index(stmt, ":md5"); 1696 sqlite3_bind_text(stmt, idx, rec->md5_hash, -1, NULL); 1697 idx = sqlite3_bind_parameter_index(stmt, ":file"); 1698 sqlite3_bind_text(stmt, idx, rec->file_path, -1, NULL); 1699 rc = sqlite3_step(stmt); 1700 sqlite3_finalize(stmt); 1701 1702 if (rc != SQLITE_DONE) { 1703 warnx("%s", sqlite3_errmsg(db)); 1704 close_db(db); 1705 cleanup(rec); 1706 errx(EXIT_FAILURE, 1707 "Consider running makemandb with -f option"); 1708 } 1709 } else if (rc != SQLITE_DONE) { 1710 /* Otherwise make this error fatal */ 1711 warnx("Failed at %s\n%s", rec->file_path, sqlite3_errmsg(db)); 1712 cleanup(rec); 1713 close_db(db); 1714 exit(EXIT_FAILURE); 1715 } 1716 1717 /*------------------------ Populate the mandb_links table---------------------*/ 1718 char *str = NULL; 1719 char *links; 1720 if (rec->links && strlen(rec->links)) { 1721 links = rec->links; 1722 for(ln = strtok(links, " "); ln; ln = strtok(NULL, " ")) { 1723 if (ln[0] == ',') 1724 ln++; 1725 if(ln[strlen(ln) - 1] == ',') 1726 ln[strlen(ln) - 1] = 0; 1727 1728 str = sqlite3_mprintf("INSERT INTO mandb_links" 1729 " VALUES (%Q, %Q, %Q, %Q)", 1730 ln, rec->name, rec->section, 1731 rec->machine); 1732 sqlite3_exec(db, str, NULL, NULL, &errmsg); 1733 sqlite3_free(str); 1734 if (errmsg != NULL) { 1735 warnx("%s", errmsg); 1736 cleanup(rec); 1737 free(errmsg); 1738 return -1; 1739 } 1740 } 1741 } 1742 1743 cleanup(rec); 1744 return 0; 1745 1746 Out: 1747 warnx("%s", sqlite3_errmsg(db)); 1748 cleanup(rec); 1749 return -1; 1750 } 1751 1752 /* 1753 * check_md5-- 1754 * Generates the md5 hash of the file and checks if it already doesn't exist 1755 * in the table (passed as the 3rd parameter). 1756 * This function is being used to avoid hardlinks. 1757 * On successful completion it will also set the value of the fourth parameter 1758 * to the md5 hash of the file (computed previously). It is the responsibility 1759 * of the caller to free this buffer. 1760 * Return values: 1761 * -1: If an error occurs somewhere and sets the md5 return buffer to NULL. 1762 * 0: If the md5 hash does not exist in the table. 1763 * 1: If the hash exists in the database. 1764 */ 1765 static int 1766 check_md5(const char *file, sqlite3 *db, const char *table, char **md5sum, 1767 void *buf, size_t buflen) 1768 { 1769 int rc = 0; 1770 int idx = -1; 1771 char *sqlstr = NULL; 1772 sqlite3_stmt *stmt = NULL; 1773 1774 assert(file != NULL); 1775 *md5sum = MD5Data(buf, buflen, NULL); 1776 if (*md5sum == NULL) { 1777 warn("md5 failed: %s", file); 1778 return -1; 1779 } 1780 1781 easprintf(&sqlstr, "SELECT * FROM %s WHERE md5_hash = :md5_hash", 1782 table); 1783 rc = sqlite3_prepare_v2(db, sqlstr, -1, &stmt, NULL); 1784 if (rc != SQLITE_OK) { 1785 free(sqlstr); 1786 free(*md5sum); 1787 *md5sum = NULL; 1788 return -1; 1789 } 1790 1791 idx = sqlite3_bind_parameter_index(stmt, ":md5_hash"); 1792 rc = sqlite3_bind_text(stmt, idx, *md5sum, -1, NULL); 1793 if (rc != SQLITE_OK) { 1794 warnx("%s", sqlite3_errmsg(db)); 1795 sqlite3_finalize(stmt); 1796 free(sqlstr); 1797 free(*md5sum); 1798 *md5sum = NULL; 1799 return -1; 1800 } 1801 1802 if (sqlite3_step(stmt) == SQLITE_ROW) { 1803 sqlite3_finalize(stmt); 1804 free(sqlstr); 1805 return 0; 1806 } 1807 1808 sqlite3_finalize(stmt); 1809 free(sqlstr); 1810 return 1; 1811 } 1812 1813 /* Optimize the index for faster search */ 1814 static void 1815 optimize(sqlite3 *db) 1816 { 1817 const char *sqlstr; 1818 char *errmsg = NULL; 1819 1820 if (mflags.verbosity) 1821 printf("Optimizing the database index\n"); 1822 sqlstr = "INSERT INTO mandb(mandb) VALUES (\'optimize\');" 1823 "VACUUM"; 1824 sqlite3_exec(db, sqlstr, NULL, NULL, &errmsg); 1825 if (errmsg != NULL) { 1826 warnx("%s", errmsg); 1827 free(errmsg); 1828 return; 1829 } 1830 } 1831 1832 /* 1833 * cleanup -- 1834 * cleans up the global buffers 1835 */ 1836 static void 1837 cleanup(mandb_rec *rec) 1838 { 1839 rec->desc.offset = 0; 1840 rec->lib.offset = 0; 1841 rec->return_vals.offset = 0; 1842 rec->env.offset = 0; 1843 rec->exit_status.offset = 0; 1844 rec->diagnostics.offset = 0; 1845 rec->errors.offset = 0; 1846 rec->files.offset = 0; 1847 1848 free(rec->machine); 1849 rec->machine = NULL; 1850 1851 free(rec->links); 1852 rec->links = NULL; 1853 1854 free(rec->file_path); 1855 rec->file_path = NULL; 1856 1857 free(rec->name); 1858 rec->name = NULL; 1859 1860 free(rec->name_desc); 1861 rec->name_desc = NULL; 1862 1863 free(rec->md5_hash); 1864 rec->md5_hash = NULL; 1865 } 1866 1867 /* 1868 * init_secbuffs-- 1869 * Sets the value of buflen for all the sec_buff field of rec. And then 1870 * allocate memory to each sec_buff member of rec. 1871 */ 1872 static void 1873 init_secbuffs(mandb_rec *rec) 1874 { 1875 /* 1876 * Some sec_buff might need more memory, for example desc, 1877 * which stores the data of the DESCRIPTION section, 1878 * while some might need very small amount of memory. 1879 * Therefore explicitly setting the value of buflen field for 1880 * each sec_buff. 1881 */ 1882 rec->desc.buflen = 10 * BUFLEN; 1883 rec->desc.data = emalloc(rec->desc.buflen); 1884 rec->desc.offset = 0; 1885 1886 rec->lib.buflen = BUFLEN / 2; 1887 rec->lib.data = emalloc(rec->lib.buflen); 1888 rec->lib.offset = 0; 1889 1890 rec->return_vals.buflen = BUFLEN; 1891 rec->return_vals.data = emalloc(rec->return_vals.buflen); 1892 rec->return_vals.offset = 0; 1893 1894 rec->exit_status.buflen = BUFLEN; 1895 rec->exit_status.data = emalloc(rec->exit_status.buflen); 1896 rec->exit_status.offset = 0; 1897 1898 rec->env.buflen = BUFLEN; 1899 rec->env.data = emalloc(rec->env.buflen); 1900 rec->env.offset = 0; 1901 1902 rec->files.buflen = BUFLEN; 1903 rec->files.data = emalloc(rec->files.buflen); 1904 rec->files.offset = 0; 1905 1906 rec->diagnostics.buflen = BUFLEN; 1907 rec->diagnostics.data = emalloc(rec->diagnostics.buflen); 1908 rec->diagnostics.offset = 0; 1909 1910 rec->errors.buflen = BUFLEN; 1911 rec->errors.data = emalloc(rec->errors.buflen); 1912 rec->errors.offset = 0; 1913 } 1914 1915 /* 1916 * free_secbuffs-- 1917 * This function should be called at the end, when all the pages have been 1918 * parsed. 1919 * It frees the memory allocated to sec_buffs by init_secbuffs in the starting. 1920 */ 1921 static void 1922 free_secbuffs(mandb_rec *rec) 1923 { 1924 free(rec->desc.data); 1925 free(rec->lib.data); 1926 free(rec->return_vals.data); 1927 free(rec->exit_status.data); 1928 free(rec->env.data); 1929 free(rec->files.data); 1930 free(rec->diagnostics.data); 1931 free(rec->errors.data); 1932 } 1933 1934 static void 1935 replace_hyph(char *str) 1936 { 1937 char *iter = str; 1938 while ((iter = strchr(iter, ASCII_HYPH)) != NULL) 1939 *iter = '-'; 1940 } 1941 1942 static char * 1943 parse_escape(const char *str) 1944 { 1945 const char *backslash, *last_backslash; 1946 char *result, *iter; 1947 size_t len; 1948 1949 assert(str); 1950 1951 last_backslash = str; 1952 backslash = strchr(str, '\\'); 1953 if (backslash == NULL) { 1954 result = estrdup(str); 1955 replace_hyph(result); 1956 return result; 1957 } 1958 1959 result = emalloc(strlen(str) + 1); 1960 iter = result; 1961 1962 do { 1963 len = backslash - last_backslash; 1964 memcpy(iter, last_backslash, len); 1965 iter += len; 1966 if (backslash[1] == '-' || backslash[1] == ' ') { 1967 *iter++ = backslash[1]; 1968 last_backslash = backslash + 2; 1969 backslash = strchr(backslash + 2, '\\'); 1970 } else { 1971 ++backslash; 1972 mandoc_escape(&backslash, NULL, NULL); 1973 last_backslash = backslash; 1974 if (backslash == NULL) 1975 break; 1976 backslash = strchr(last_backslash, '\\'); 1977 } 1978 } while (backslash != NULL); 1979 if (last_backslash != NULL) 1980 strcpy(iter, last_backslash); 1981 1982 replace_hyph(result); 1983 return result; 1984 } 1985 1986 /* 1987 * append-- 1988 * Concatenates a space and src at the end of sbuff->data (much like concat in 1989 * apropos-utils.c). 1990 * Rather than reallocating space for writing data, it uses the value of the 1991 * offset field of sec_buff to write new data at the free space left in the 1992 * buffer. 1993 * In case the size of the data to be appended exceeds the number of bytes left 1994 * in the buffer, it reallocates buflen number of bytes and then continues. 1995 * Value of offset field should be adjusted as new data is written. 1996 * 1997 * NOTE: This function does not write the null byte at the end of the buffers, 1998 * write a null byte at the position pointed to by offset before inserting data 1999 * in the db. 2000 */ 2001 static void 2002 append(secbuff *sbuff, const char *src) 2003 { 2004 short flag = 0; 2005 size_t srclen, newlen; 2006 char *temp; 2007 2008 assert(src != NULL); 2009 temp = parse_escape(src); 2010 srclen = strlen(temp); 2011 2012 if (sbuff->data == NULL) { 2013 sbuff->data = emalloc(sbuff->buflen); 2014 sbuff->offset = 0; 2015 } 2016 2017 newlen = sbuff->offset + srclen + 2; 2018 if (newlen >= sbuff->buflen) { 2019 while (sbuff->buflen < newlen) 2020 sbuff->buflen += sbuff->buflen; 2021 sbuff->data = erealloc(sbuff->data, sbuff->buflen); 2022 flag = 1; 2023 } 2024 2025 /* Append a space at the end of the buffer. */ 2026 if (sbuff->offset || flag) 2027 sbuff->data[sbuff->offset++] = ' '; 2028 /* Now, copy src at the end of the buffer. */ 2029 memcpy(sbuff->data + sbuff->offset, temp, srclen); 2030 sbuff->offset += srclen; 2031 free(temp); 2032 } 2033 2034 static void 2035 usage(void) 2036 { 2037 fprintf(stderr, "Usage: %s [-flo]\n", getprogname()); 2038 exit(1); 2039 } 2040