1 /* $Id: mandocdb.c,v 1.262 2018/12/30 00:49:55 schwarze Exp $ */ 2 /* 3 * Copyright (c) 2011, 2012 Kristaps Dzonsons <kristaps@bsd.lv> 4 * Copyright (c) 2011-2018 Ingo Schwarze <schwarze@openbsd.org> 5 * Copyright (c) 2016 Ed Maste <emaste@freebsd.org> 6 * 7 * Permission to use, copy, modify, and distribute this software for any 8 * purpose with or without fee is hereby granted, provided that the above 9 * copyright notice and this permission notice appear in all copies. 10 * 11 * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHORS DISCLAIM ALL WARRANTIES 12 * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF 13 * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHORS BE LIABLE FOR 14 * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES 15 * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN 16 * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF 17 * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. 18 */ 19 #include "config.h" 20 21 #include <sys/types.h> 22 #include <sys/mman.h> 23 #include <sys/stat.h> 24 25 #include <assert.h> 26 #include <ctype.h> 27 #if HAVE_ERR 28 #include <err.h> 29 #endif 30 #include <errno.h> 31 #include <fcntl.h> 32 #if HAVE_FTS 33 #include <fts.h> 34 #else 35 #include "compat_fts.h" 36 #endif 37 #include <limits.h> 38 #if HAVE_SANDBOX_INIT 39 #include <sandbox.h> 40 #endif 41 #include <stdarg.h> 42 #include <stddef.h> 43 #include <stdio.h> 44 #include <stdint.h> 45 #include <stdlib.h> 46 #include <string.h> 47 #include <unistd.h> 48 49 #include "mandoc_aux.h" 50 #include "mandoc_ohash.h" 51 #include "mandoc.h" 52 #include "roff.h" 53 #include "mdoc.h" 54 #include "man.h" 55 #include "mandoc_parse.h" 56 #include "manconf.h" 57 #include "mansearch.h" 58 #include "dba_array.h" 59 #include "dba.h" 60 61 extern const char *const mansearch_keynames[]; 62 63 enum op { 64 OP_DEFAULT = 0, /* new dbs from dir list or default config */ 65 OP_CONFFILE, /* new databases from custom config file */ 66 OP_UPDATE, /* delete/add entries in existing database */ 67 OP_DELETE, /* delete entries from existing database */ 68 OP_TEST /* change no databases, report potential problems */ 69 }; 70 71 struct str { 72 const struct mpage *mpage; /* if set, the owning parse */ 73 uint64_t mask; /* bitmask in sequence */ 74 char key[]; /* rendered text */ 75 }; 76 77 struct inodev { 78 ino_t st_ino; 79 dev_t st_dev; 80 }; 81 82 struct mpage { 83 struct inodev inodev; /* used for hashing routine */ 84 struct dba_array *dba; 85 char *sec; /* section from file content */ 86 char *arch; /* architecture from file content */ 87 char *title; /* title from file content */ 88 char *desc; /* description from file content */ 89 struct mpage *next; /* singly linked list */ 90 struct mlink *mlinks; /* singly linked list */ 91 int name_head_done; 92 enum form form; /* format from file content */ 93 }; 94 95 struct mlink { 96 char file[PATH_MAX]; /* filename rel. to manpath */ 97 char *dsec; /* section from directory */ 98 char *arch; /* architecture from directory */ 99 char *name; /* name from file name (not empty) */ 100 char *fsec; /* section from file name suffix */ 101 struct mlink *next; /* singly linked list */ 102 struct mpage *mpage; /* parent */ 103 int gzip; /* filename has a .gz suffix */ 104 enum form dform; /* format from directory */ 105 enum form fform; /* format from file name suffix */ 106 }; 107 108 typedef int (*mdoc_fp)(struct mpage *, const struct roff_meta *, 109 const struct roff_node *); 110 111 struct mdoc_handler { 112 mdoc_fp fp; /* optional handler */ 113 uint64_t mask; /* set unless handler returns 0 */ 114 int taboo; /* node flags that must not be set */ 115 }; 116 117 118 int mandocdb(int, char *[]); 119 120 static void dbadd(struct dba *, struct mpage *); 121 static void dbadd_mlink(const struct mlink *mlink); 122 static void dbprune(struct dba *); 123 static void dbwrite(struct dba *); 124 static void filescan(const char *); 125 #if HAVE_FTS_COMPARE_CONST 126 static int fts_compare(const FTSENT *const *, const FTSENT *const *); 127 #else 128 static int fts_compare(const FTSENT **, const FTSENT **); 129 #endif 130 static void mlink_add(struct mlink *, const struct stat *); 131 static void mlink_check(struct mpage *, struct mlink *); 132 static void mlink_free(struct mlink *); 133 static void mlinks_undupe(struct mpage *); 134 static void mpages_free(void); 135 static void mpages_merge(struct dba *, struct mparse *); 136 static void parse_cat(struct mpage *, int); 137 static void parse_man(struct mpage *, const struct roff_meta *, 138 const struct roff_node *); 139 static void parse_mdoc(struct mpage *, const struct roff_meta *, 140 const struct roff_node *); 141 static int parse_mdoc_head(struct mpage *, const struct roff_meta *, 142 const struct roff_node *); 143 static int parse_mdoc_Fa(struct mpage *, const struct roff_meta *, 144 const struct roff_node *); 145 static int parse_mdoc_Fd(struct mpage *, const struct roff_meta *, 146 const struct roff_node *); 147 static void parse_mdoc_fname(struct mpage *, const struct roff_node *); 148 static int parse_mdoc_Fn(struct mpage *, const struct roff_meta *, 149 const struct roff_node *); 150 static int parse_mdoc_Fo(struct mpage *, const struct roff_meta *, 151 const struct roff_node *); 152 static int parse_mdoc_Nd(struct mpage *, const struct roff_meta *, 153 const struct roff_node *); 154 static int parse_mdoc_Nm(struct mpage *, const struct roff_meta *, 155 const struct roff_node *); 156 static int parse_mdoc_Sh(struct mpage *, const struct roff_meta *, 157 const struct roff_node *); 158 static int parse_mdoc_Va(struct mpage *, const struct roff_meta *, 159 const struct roff_node *); 160 static int parse_mdoc_Xr(struct mpage *, const struct roff_meta *, 161 const struct roff_node *); 162 static void putkey(const struct mpage *, char *, uint64_t); 163 static void putkeys(const struct mpage *, char *, size_t, uint64_t); 164 static void putmdockey(const struct mpage *, 165 const struct roff_node *, uint64_t, int); 166 static int render_string(char **, size_t *); 167 static void say(const char *, const char *, ...) 168 __attribute__((__format__ (__printf__, 2, 3))); 169 static int set_basedir(const char *, int); 170 static int treescan(void); 171 static size_t utf8(unsigned int, char [7]); 172 173 static int nodb; /* no database changes */ 174 static int mparse_options; /* abort the parse early */ 175 static int use_all; /* use all found files */ 176 static int debug; /* print what we're doing */ 177 static int warnings; /* warn about crap */ 178 static int write_utf8; /* write UTF-8 output; else ASCII */ 179 static int exitcode; /* to be returned by main */ 180 static enum op op; /* operational mode */ 181 static char basedir[PATH_MAX]; /* current base directory */ 182 static struct mpage *mpage_head; /* list of distinct manual pages */ 183 static struct ohash mpages; /* table of distinct manual pages */ 184 static struct ohash mlinks; /* table of directory entries */ 185 static struct ohash names; /* table of all names */ 186 static struct ohash strings; /* table of all strings */ 187 static uint64_t name_mask; 188 189 static const struct mdoc_handler mdoc_handlers[MDOC_MAX - MDOC_Dd] = { 190 { NULL, 0, NODE_NOPRT }, /* Dd */ 191 { NULL, 0, NODE_NOPRT }, /* Dt */ 192 { NULL, 0, NODE_NOPRT }, /* Os */ 193 { parse_mdoc_Sh, TYPE_Sh, 0 }, /* Sh */ 194 { parse_mdoc_head, TYPE_Ss, 0 }, /* Ss */ 195 { NULL, 0, 0 }, /* Pp */ 196 { NULL, 0, 0 }, /* D1 */ 197 { NULL, 0, 0 }, /* Dl */ 198 { NULL, 0, 0 }, /* Bd */ 199 { NULL, 0, 0 }, /* Ed */ 200 { NULL, 0, 0 }, /* Bl */ 201 { NULL, 0, 0 }, /* El */ 202 { NULL, 0, 0 }, /* It */ 203 { NULL, 0, 0 }, /* Ad */ 204 { NULL, TYPE_An, 0 }, /* An */ 205 { NULL, 0, 0 }, /* Ap */ 206 { NULL, TYPE_Ar, 0 }, /* Ar */ 207 { NULL, TYPE_Cd, 0 }, /* Cd */ 208 { NULL, TYPE_Cm, 0 }, /* Cm */ 209 { NULL, TYPE_Dv, 0 }, /* Dv */ 210 { NULL, TYPE_Er, 0 }, /* Er */ 211 { NULL, TYPE_Ev, 0 }, /* Ev */ 212 { NULL, 0, 0 }, /* Ex */ 213 { parse_mdoc_Fa, 0, 0 }, /* Fa */ 214 { parse_mdoc_Fd, 0, 0 }, /* Fd */ 215 { NULL, TYPE_Fl, 0 }, /* Fl */ 216 { parse_mdoc_Fn, 0, 0 }, /* Fn */ 217 { NULL, TYPE_Ft | TYPE_Vt, 0 }, /* Ft */ 218 { NULL, TYPE_Ic, 0 }, /* Ic */ 219 { NULL, TYPE_In, 0 }, /* In */ 220 { NULL, TYPE_Li, 0 }, /* Li */ 221 { parse_mdoc_Nd, 0, 0 }, /* Nd */ 222 { parse_mdoc_Nm, 0, 0 }, /* Nm */ 223 { NULL, 0, 0 }, /* Op */ 224 { NULL, 0, 0 }, /* Ot */ 225 { NULL, TYPE_Pa, NODE_NOSRC }, /* Pa */ 226 { NULL, 0, 0 }, /* Rv */ 227 { NULL, TYPE_St, 0 }, /* St */ 228 { parse_mdoc_Va, TYPE_Va, 0 }, /* Va */ 229 { parse_mdoc_Va, TYPE_Vt, 0 }, /* Vt */ 230 { parse_mdoc_Xr, 0, 0 }, /* Xr */ 231 { NULL, 0, 0 }, /* %A */ 232 { NULL, 0, 0 }, /* %B */ 233 { NULL, 0, 0 }, /* %D */ 234 { NULL, 0, 0 }, /* %I */ 235 { NULL, 0, 0 }, /* %J */ 236 { NULL, 0, 0 }, /* %N */ 237 { NULL, 0, 0 }, /* %O */ 238 { NULL, 0, 0 }, /* %P */ 239 { NULL, 0, 0 }, /* %R */ 240 { NULL, 0, 0 }, /* %T */ 241 { NULL, 0, 0 }, /* %V */ 242 { NULL, 0, 0 }, /* Ac */ 243 { NULL, 0, 0 }, /* Ao */ 244 { NULL, 0, 0 }, /* Aq */ 245 { NULL, TYPE_At, 0 }, /* At */ 246 { NULL, 0, 0 }, /* Bc */ 247 { NULL, 0, 0 }, /* Bf */ 248 { NULL, 0, 0 }, /* Bo */ 249 { NULL, 0, 0 }, /* Bq */ 250 { NULL, TYPE_Bsx, NODE_NOSRC }, /* Bsx */ 251 { NULL, TYPE_Bx, NODE_NOSRC }, /* Bx */ 252 { NULL, 0, 0 }, /* Db */ 253 { NULL, 0, 0 }, /* Dc */ 254 { NULL, 0, 0 }, /* Do */ 255 { NULL, 0, 0 }, /* Dq */ 256 { NULL, 0, 0 }, /* Ec */ 257 { NULL, 0, 0 }, /* Ef */ 258 { NULL, TYPE_Em, 0 }, /* Em */ 259 { NULL, 0, 0 }, /* Eo */ 260 { NULL, TYPE_Fx, NODE_NOSRC }, /* Fx */ 261 { NULL, TYPE_Ms, 0 }, /* Ms */ 262 { NULL, 0, 0 }, /* No */ 263 { NULL, 0, 0 }, /* Ns */ 264 { NULL, TYPE_Nx, NODE_NOSRC }, /* Nx */ 265 { NULL, TYPE_Ox, NODE_NOSRC }, /* Ox */ 266 { NULL, 0, 0 }, /* Pc */ 267 { NULL, 0, 0 }, /* Pf */ 268 { NULL, 0, 0 }, /* Po */ 269 { NULL, 0, 0 }, /* Pq */ 270 { NULL, 0, 0 }, /* Qc */ 271 { NULL, 0, 0 }, /* Ql */ 272 { NULL, 0, 0 }, /* Qo */ 273 { NULL, 0, 0 }, /* Qq */ 274 { NULL, 0, 0 }, /* Re */ 275 { NULL, 0, 0 }, /* Rs */ 276 { NULL, 0, 0 }, /* Sc */ 277 { NULL, 0, 0 }, /* So */ 278 { NULL, 0, 0 }, /* Sq */ 279 { NULL, 0, 0 }, /* Sm */ 280 { NULL, 0, 0 }, /* Sx */ 281 { NULL, TYPE_Sy, 0 }, /* Sy */ 282 { NULL, TYPE_Tn, 0 }, /* Tn */ 283 { NULL, 0, NODE_NOSRC }, /* Ux */ 284 { NULL, 0, 0 }, /* Xc */ 285 { NULL, 0, 0 }, /* Xo */ 286 { parse_mdoc_Fo, 0, 0 }, /* Fo */ 287 { NULL, 0, 0 }, /* Fc */ 288 { NULL, 0, 0 }, /* Oo */ 289 { NULL, 0, 0 }, /* Oc */ 290 { NULL, 0, 0 }, /* Bk */ 291 { NULL, 0, 0 }, /* Ek */ 292 { NULL, 0, 0 }, /* Bt */ 293 { NULL, 0, 0 }, /* Hf */ 294 { NULL, 0, 0 }, /* Fr */ 295 { NULL, 0, 0 }, /* Ud */ 296 { NULL, TYPE_Lb, NODE_NOSRC }, /* Lb */ 297 { NULL, 0, 0 }, /* Lp */ 298 { NULL, TYPE_Lk, 0 }, /* Lk */ 299 { NULL, TYPE_Mt, NODE_NOSRC }, /* Mt */ 300 { NULL, 0, 0 }, /* Brq */ 301 { NULL, 0, 0 }, /* Bro */ 302 { NULL, 0, 0 }, /* Brc */ 303 { NULL, 0, 0 }, /* %C */ 304 { NULL, 0, 0 }, /* Es */ 305 { NULL, 0, 0 }, /* En */ 306 { NULL, TYPE_Dx, NODE_NOSRC }, /* Dx */ 307 { NULL, 0, 0 }, /* %Q */ 308 { NULL, 0, 0 }, /* %U */ 309 { NULL, 0, 0 }, /* Ta */ 310 }; 311 312 313 int 314 mandocdb(int argc, char *argv[]) 315 { 316 struct manconf conf; 317 struct mparse *mp; 318 struct dba *dba; 319 const char *path_arg, *progname; 320 size_t j, sz; 321 int ch, i; 322 323 #if HAVE_PLEDGE 324 if (pledge("stdio rpath wpath cpath", NULL) == -1) { 325 warn("pledge"); 326 return (int)MANDOCLEVEL_SYSERR; 327 } 328 #endif 329 330 #if HAVE_SANDBOX_INIT 331 if (sandbox_init(kSBXProfileNoInternet, SANDBOX_NAMED, NULL) == -1) { 332 warnx("sandbox_init"); 333 return (int)MANDOCLEVEL_SYSERR; 334 } 335 #endif 336 337 memset(&conf, 0, sizeof(conf)); 338 339 /* 340 * We accept a few different invocations. 341 * The CHECKOP macro makes sure that invocation styles don't 342 * clobber each other. 343 */ 344 #define CHECKOP(_op, _ch) do \ 345 if (OP_DEFAULT != (_op)) { \ 346 warnx("-%c: Conflicting option", (_ch)); \ 347 goto usage; \ 348 } while (/*CONSTCOND*/0) 349 350 mparse_options = MPARSE_VALIDATE; 351 path_arg = NULL; 352 op = OP_DEFAULT; 353 354 while (-1 != (ch = getopt(argc, argv, "aC:Dd:npQT:tu:v"))) 355 switch (ch) { 356 case 'a': 357 use_all = 1; 358 break; 359 case 'C': 360 CHECKOP(op, ch); 361 path_arg = optarg; 362 op = OP_CONFFILE; 363 break; 364 case 'D': 365 debug++; 366 break; 367 case 'd': 368 CHECKOP(op, ch); 369 path_arg = optarg; 370 op = OP_UPDATE; 371 break; 372 case 'n': 373 nodb = 1; 374 break; 375 case 'p': 376 warnings = 1; 377 break; 378 case 'Q': 379 mparse_options |= MPARSE_QUICK; 380 break; 381 case 'T': 382 if (strcmp(optarg, "utf8")) { 383 warnx("-T%s: Unsupported output format", 384 optarg); 385 goto usage; 386 } 387 write_utf8 = 1; 388 break; 389 case 't': 390 CHECKOP(op, ch); 391 dup2(STDOUT_FILENO, STDERR_FILENO); 392 op = OP_TEST; 393 nodb = warnings = 1; 394 break; 395 case 'u': 396 CHECKOP(op, ch); 397 path_arg = optarg; 398 op = OP_DELETE; 399 break; 400 case 'v': 401 /* Compatibility with espie@'s makewhatis. */ 402 break; 403 default: 404 goto usage; 405 } 406 407 argc -= optind; 408 argv += optind; 409 410 #if HAVE_PLEDGE 411 if (nodb) { 412 if (pledge("stdio rpath", NULL) == -1) { 413 warn("pledge"); 414 return (int)MANDOCLEVEL_SYSERR; 415 } 416 } 417 #endif 418 419 if (OP_CONFFILE == op && argc > 0) { 420 warnx("-C: Too many arguments"); 421 goto usage; 422 } 423 424 exitcode = (int)MANDOCLEVEL_OK; 425 mchars_alloc(); 426 mp = mparse_alloc(mparse_options, MANDOC_OS_OTHER, NULL); 427 mandoc_ohash_init(&mpages, 6, offsetof(struct mpage, inodev)); 428 mandoc_ohash_init(&mlinks, 6, offsetof(struct mlink, file)); 429 430 if (OP_UPDATE == op || OP_DELETE == op || OP_TEST == op) { 431 432 /* 433 * Most of these deal with a specific directory. 434 * Jump into that directory first. 435 */ 436 if (OP_TEST != op && 0 == set_basedir(path_arg, 1)) 437 goto out; 438 439 dba = nodb ? dba_new(128) : dba_read(MANDOC_DB); 440 if (dba != NULL) { 441 /* 442 * The existing database is usable. Process 443 * all files specified on the command-line. 444 */ 445 use_all = 1; 446 for (i = 0; i < argc; i++) 447 filescan(argv[i]); 448 if (nodb == 0) 449 dbprune(dba); 450 } else { 451 /* Database missing or corrupt. */ 452 if (op != OP_UPDATE || errno != ENOENT) 453 say(MANDOC_DB, "%s: Automatically recreating" 454 " from scratch", strerror(errno)); 455 exitcode = (int)MANDOCLEVEL_OK; 456 op = OP_DEFAULT; 457 if (0 == treescan()) 458 goto out; 459 dba = dba_new(128); 460 } 461 if (OP_DELETE != op) 462 mpages_merge(dba, mp); 463 if (nodb == 0) 464 dbwrite(dba); 465 dba_free(dba); 466 } else { 467 /* 468 * If we have arguments, use them as our manpaths. 469 * If we don't, use man.conf(5). 470 */ 471 if (argc > 0) { 472 conf.manpath.paths = mandoc_reallocarray(NULL, 473 argc, sizeof(char *)); 474 conf.manpath.sz = (size_t)argc; 475 for (i = 0; i < argc; i++) 476 conf.manpath.paths[i] = mandoc_strdup(argv[i]); 477 } else 478 manconf_parse(&conf, path_arg, NULL, NULL); 479 480 if (conf.manpath.sz == 0) { 481 exitcode = (int)MANDOCLEVEL_BADARG; 482 say("", "Empty manpath"); 483 } 484 485 /* 486 * First scan the tree rooted at a base directory, then 487 * build a new database and finally move it into place. 488 * Ignore zero-length directories and strip trailing 489 * slashes. 490 */ 491 for (j = 0; j < conf.manpath.sz; j++) { 492 sz = strlen(conf.manpath.paths[j]); 493 if (sz && conf.manpath.paths[j][sz - 1] == '/') 494 conf.manpath.paths[j][--sz] = '\0'; 495 if (0 == sz) 496 continue; 497 498 if (j) { 499 mandoc_ohash_init(&mpages, 6, 500 offsetof(struct mpage, inodev)); 501 mandoc_ohash_init(&mlinks, 6, 502 offsetof(struct mlink, file)); 503 } 504 505 if ( ! set_basedir(conf.manpath.paths[j], argc > 0)) 506 continue; 507 if (0 == treescan()) 508 continue; 509 dba = dba_new(128); 510 mpages_merge(dba, mp); 511 if (nodb == 0) 512 dbwrite(dba); 513 dba_free(dba); 514 515 if (j + 1 < conf.manpath.sz) { 516 mpages_free(); 517 ohash_delete(&mpages); 518 ohash_delete(&mlinks); 519 } 520 } 521 } 522 out: 523 manconf_free(&conf); 524 mparse_free(mp); 525 mchars_free(); 526 mpages_free(); 527 ohash_delete(&mpages); 528 ohash_delete(&mlinks); 529 return exitcode; 530 usage: 531 progname = getprogname(); 532 fprintf(stderr, "usage: %s [-aDnpQ] [-C file] [-Tutf8]\n" 533 " %s [-aDnpQ] [-Tutf8] dir ...\n" 534 " %s [-DnpQ] [-Tutf8] -d dir [file ...]\n" 535 " %s [-Dnp] -u dir [file ...]\n" 536 " %s [-Q] -t file ...\n", 537 progname, progname, progname, progname, progname); 538 539 return (int)MANDOCLEVEL_BADARG; 540 } 541 542 /* 543 * To get a singly linked list in alpha order while inserting entries 544 * at the beginning, process directory entries in reverse alpha order. 545 */ 546 static int 547 #if HAVE_FTS_COMPARE_CONST 548 fts_compare(const FTSENT *const *a, const FTSENT *const *b) 549 #else 550 fts_compare(const FTSENT **a, const FTSENT **b) 551 #endif 552 { 553 return -strcmp((*a)->fts_name, (*b)->fts_name); 554 } 555 556 /* 557 * Scan a directory tree rooted at "basedir" for manpages. 558 * We use fts(), scanning directory parts along the way for clues to our 559 * section and architecture. 560 * 561 * If use_all has been specified, grok all files. 562 * If not, sanitise paths to the following: 563 * 564 * [./]man*[/<arch>]/<name>.<section> 565 * or 566 * [./]cat<section>[/<arch>]/<name>.0 567 * 568 * TODO: accommodate for multi-language directories. 569 */ 570 static int 571 treescan(void) 572 { 573 char buf[PATH_MAX]; 574 FTS *f; 575 FTSENT *ff; 576 struct mlink *mlink; 577 int gzip; 578 enum form dform; 579 char *dsec, *arch, *fsec, *cp; 580 const char *path; 581 const char *argv[2]; 582 583 argv[0] = "."; 584 argv[1] = NULL; 585 586 f = fts_open((char * const *)argv, FTS_PHYSICAL | FTS_NOCHDIR, 587 fts_compare); 588 if (f == NULL) { 589 exitcode = (int)MANDOCLEVEL_SYSERR; 590 say("", "&fts_open"); 591 return 0; 592 } 593 594 dsec = arch = NULL; 595 dform = FORM_NONE; 596 597 while ((ff = fts_read(f)) != NULL) { 598 path = ff->fts_path + 2; 599 switch (ff->fts_info) { 600 601 /* 602 * Symbolic links require various sanity checks, 603 * then get handled just like regular files. 604 */ 605 case FTS_SL: 606 if (realpath(path, buf) == NULL) { 607 if (warnings) 608 say(path, "&realpath"); 609 continue; 610 } 611 if (strstr(buf, basedir) != buf 612 #ifdef HOMEBREWDIR 613 && strstr(buf, HOMEBREWDIR) != buf 614 #endif 615 ) { 616 if (warnings) say("", 617 "%s: outside base directory", buf); 618 continue; 619 } 620 /* Use logical inode to avoid mpages dupe. */ 621 if (stat(path, ff->fts_statp) == -1) { 622 if (warnings) 623 say(path, "&stat"); 624 continue; 625 } 626 /* FALLTHROUGH */ 627 628 /* 629 * If we're a regular file, add an mlink by using the 630 * stored directory data and handling the filename. 631 */ 632 case FTS_F: 633 if ( ! strcmp(path, MANDOC_DB)) 634 continue; 635 if ( ! use_all && ff->fts_level < 2) { 636 if (warnings) 637 say(path, "Extraneous file"); 638 continue; 639 } 640 gzip = 0; 641 fsec = NULL; 642 while (fsec == NULL) { 643 fsec = strrchr(ff->fts_name, '.'); 644 if (fsec == NULL || strcmp(fsec+1, "gz")) 645 break; 646 gzip = 1; 647 *fsec = '\0'; 648 fsec = NULL; 649 } 650 if (fsec == NULL) { 651 if ( ! use_all) { 652 if (warnings) 653 say(path, 654 "No filename suffix"); 655 continue; 656 } 657 } else if ( ! strcmp(++fsec, "html")) { 658 if (warnings) 659 say(path, "Skip html"); 660 continue; 661 } else if ( ! strcmp(fsec, "ps")) { 662 if (warnings) 663 say(path, "Skip ps"); 664 continue; 665 } else if ( ! strcmp(fsec, "pdf")) { 666 if (warnings) 667 say(path, "Skip pdf"); 668 continue; 669 } else if ( ! use_all && 670 ((dform == FORM_SRC && 671 strncmp(fsec, dsec, strlen(dsec))) || 672 (dform == FORM_CAT && strcmp(fsec, "0")))) { 673 if (warnings) 674 say(path, "Wrong filename suffix"); 675 continue; 676 } else 677 fsec[-1] = '\0'; 678 679 mlink = mandoc_calloc(1, sizeof(struct mlink)); 680 if (strlcpy(mlink->file, path, 681 sizeof(mlink->file)) >= 682 sizeof(mlink->file)) { 683 say(path, "Filename too long"); 684 free(mlink); 685 continue; 686 } 687 mlink->dform = dform; 688 mlink->dsec = dsec; 689 mlink->arch = arch; 690 mlink->name = ff->fts_name; 691 mlink->fsec = fsec; 692 mlink->gzip = gzip; 693 mlink_add(mlink, ff->fts_statp); 694 continue; 695 696 case FTS_D: 697 case FTS_DP: 698 break; 699 700 default: 701 if (warnings) 702 say(path, "Not a regular file"); 703 continue; 704 } 705 706 switch (ff->fts_level) { 707 case 0: 708 /* Ignore the root directory. */ 709 break; 710 case 1: 711 /* 712 * This might contain manX/ or catX/. 713 * Try to infer this from the name. 714 * If we're not in use_all, enforce it. 715 */ 716 cp = ff->fts_name; 717 if (ff->fts_info == FTS_DP) { 718 dform = FORM_NONE; 719 dsec = NULL; 720 break; 721 } 722 723 if ( ! strncmp(cp, "man", 3)) { 724 dform = FORM_SRC; 725 dsec = cp + 3; 726 } else if ( ! strncmp(cp, "cat", 3)) { 727 dform = FORM_CAT; 728 dsec = cp + 3; 729 } else { 730 dform = FORM_NONE; 731 dsec = NULL; 732 } 733 734 if (dsec != NULL || use_all) 735 break; 736 737 if (warnings) 738 say(path, "Unknown directory part"); 739 fts_set(f, ff, FTS_SKIP); 740 break; 741 case 2: 742 /* 743 * Possibly our architecture. 744 * If we're descending, keep tabs on it. 745 */ 746 if (ff->fts_info != FTS_DP && dsec != NULL) 747 arch = ff->fts_name; 748 else 749 arch = NULL; 750 break; 751 default: 752 if (ff->fts_info == FTS_DP || use_all) 753 break; 754 if (warnings) 755 say(path, "Extraneous directory part"); 756 fts_set(f, ff, FTS_SKIP); 757 break; 758 } 759 } 760 761 fts_close(f); 762 return 1; 763 } 764 765 /* 766 * Add a file to the mlinks table. 767 * Do not verify that it's a "valid" looking manpage (we'll do that 768 * later). 769 * 770 * Try to infer the manual section, architecture, and page name from the 771 * path, assuming it looks like 772 * 773 * [./]man*[/<arch>]/<name>.<section> 774 * or 775 * [./]cat<section>[/<arch>]/<name>.0 776 * 777 * See treescan() for the fts(3) version of this. 778 */ 779 static void 780 filescan(const char *file) 781 { 782 char buf[PATH_MAX]; 783 struct stat st; 784 struct mlink *mlink; 785 char *p, *start; 786 787 assert(use_all); 788 789 if (0 == strncmp(file, "./", 2)) 790 file += 2; 791 792 /* 793 * We have to do lstat(2) before realpath(3) loses 794 * the information whether this is a symbolic link. 795 * We need to know that because for symbolic links, 796 * we want to use the orginal file name, while for 797 * regular files, we want to use the real path. 798 */ 799 if (-1 == lstat(file, &st)) { 800 exitcode = (int)MANDOCLEVEL_BADARG; 801 say(file, "&lstat"); 802 return; 803 } else if (0 == ((S_IFREG | S_IFLNK) & st.st_mode)) { 804 exitcode = (int)MANDOCLEVEL_BADARG; 805 say(file, "Not a regular file"); 806 return; 807 } 808 809 /* 810 * We have to resolve the file name to the real path 811 * in any case for the base directory check. 812 */ 813 if (NULL == realpath(file, buf)) { 814 exitcode = (int)MANDOCLEVEL_BADARG; 815 say(file, "&realpath"); 816 return; 817 } 818 819 if (OP_TEST == op) 820 start = buf; 821 else if (strstr(buf, basedir) == buf) 822 start = buf + strlen(basedir); 823 #ifdef HOMEBREWDIR 824 else if (strstr(buf, HOMEBREWDIR) == buf) 825 start = buf; 826 #endif 827 else { 828 exitcode = (int)MANDOCLEVEL_BADARG; 829 say("", "%s: outside base directory", buf); 830 return; 831 } 832 833 /* 834 * Now we are sure the file is inside our tree. 835 * If it is a symbolic link, ignore the real path 836 * and use the original name. 837 * This implies passing stuff like "cat1/../man1/foo.1" 838 * on the command line won't work. So don't do that. 839 * Note the stat(2) can still fail if the link target 840 * doesn't exist. 841 */ 842 if (S_IFLNK & st.st_mode) { 843 if (-1 == stat(buf, &st)) { 844 exitcode = (int)MANDOCLEVEL_BADARG; 845 say(file, "&stat"); 846 return; 847 } 848 if (strlcpy(buf, file, sizeof(buf)) >= sizeof(buf)) { 849 say(file, "Filename too long"); 850 return; 851 } 852 start = buf; 853 if (OP_TEST != op && strstr(buf, basedir) == buf) 854 start += strlen(basedir); 855 } 856 857 mlink = mandoc_calloc(1, sizeof(struct mlink)); 858 mlink->dform = FORM_NONE; 859 if (strlcpy(mlink->file, start, sizeof(mlink->file)) >= 860 sizeof(mlink->file)) { 861 say(start, "Filename too long"); 862 free(mlink); 863 return; 864 } 865 866 /* 867 * In test mode or when the original name is absolute 868 * but outside our tree, guess the base directory. 869 */ 870 871 if (op == OP_TEST || (start == buf && *start == '/')) { 872 if (strncmp(buf, "man/", 4) == 0) 873 start = buf + 4; 874 else if ((start = strstr(buf, "/man/")) != NULL) 875 start += 5; 876 else 877 start = buf; 878 } 879 880 /* 881 * First try to guess our directory structure. 882 * If we find a separator, try to look for man* or cat*. 883 * If we find one of these and what's underneath is a directory, 884 * assume it's an architecture. 885 */ 886 if (NULL != (p = strchr(start, '/'))) { 887 *p++ = '\0'; 888 if (0 == strncmp(start, "man", 3)) { 889 mlink->dform = FORM_SRC; 890 mlink->dsec = start + 3; 891 } else if (0 == strncmp(start, "cat", 3)) { 892 mlink->dform = FORM_CAT; 893 mlink->dsec = start + 3; 894 } 895 896 start = p; 897 if (NULL != mlink->dsec && NULL != (p = strchr(start, '/'))) { 898 *p++ = '\0'; 899 mlink->arch = start; 900 start = p; 901 } 902 } 903 904 /* 905 * Now check the file suffix. 906 * Suffix of `.0' indicates a catpage, `.1-9' is a manpage. 907 */ 908 p = strrchr(start, '\0'); 909 while (p-- > start && '/' != *p && '.' != *p) 910 /* Loop. */ ; 911 912 if ('.' == *p) { 913 *p++ = '\0'; 914 mlink->fsec = p; 915 } 916 917 /* 918 * Now try to parse the name. 919 * Use the filename portion of the path. 920 */ 921 mlink->name = start; 922 if (NULL != (p = strrchr(start, '/'))) { 923 mlink->name = p + 1; 924 *p = '\0'; 925 } 926 mlink_add(mlink, &st); 927 } 928 929 static void 930 mlink_add(struct mlink *mlink, const struct stat *st) 931 { 932 struct inodev inodev; 933 struct mpage *mpage; 934 unsigned int slot; 935 936 assert(NULL != mlink->file); 937 938 mlink->dsec = mandoc_strdup(mlink->dsec ? mlink->dsec : ""); 939 mlink->arch = mandoc_strdup(mlink->arch ? mlink->arch : ""); 940 mlink->name = mandoc_strdup(mlink->name ? mlink->name : ""); 941 mlink->fsec = mandoc_strdup(mlink->fsec ? mlink->fsec : ""); 942 943 if ('0' == *mlink->fsec) { 944 free(mlink->fsec); 945 mlink->fsec = mandoc_strdup(mlink->dsec); 946 mlink->fform = FORM_CAT; 947 } else if ('1' <= *mlink->fsec && '9' >= *mlink->fsec) 948 mlink->fform = FORM_SRC; 949 else 950 mlink->fform = FORM_NONE; 951 952 slot = ohash_qlookup(&mlinks, mlink->file); 953 assert(NULL == ohash_find(&mlinks, slot)); 954 ohash_insert(&mlinks, slot, mlink); 955 956 memset(&inodev, 0, sizeof(inodev)); /* Clear padding. */ 957 inodev.st_ino = st->st_ino; 958 inodev.st_dev = st->st_dev; 959 slot = ohash_lookup_memory(&mpages, (char *)&inodev, 960 sizeof(struct inodev), inodev.st_ino); 961 mpage = ohash_find(&mpages, slot); 962 if (NULL == mpage) { 963 mpage = mandoc_calloc(1, sizeof(struct mpage)); 964 mpage->inodev.st_ino = inodev.st_ino; 965 mpage->inodev.st_dev = inodev.st_dev; 966 mpage->form = FORM_NONE; 967 mpage->next = mpage_head; 968 mpage_head = mpage; 969 ohash_insert(&mpages, slot, mpage); 970 } else 971 mlink->next = mpage->mlinks; 972 mpage->mlinks = mlink; 973 mlink->mpage = mpage; 974 } 975 976 static void 977 mlink_free(struct mlink *mlink) 978 { 979 980 free(mlink->dsec); 981 free(mlink->arch); 982 free(mlink->name); 983 free(mlink->fsec); 984 free(mlink); 985 } 986 987 static void 988 mpages_free(void) 989 { 990 struct mpage *mpage; 991 struct mlink *mlink; 992 993 while ((mpage = mpage_head) != NULL) { 994 while ((mlink = mpage->mlinks) != NULL) { 995 mpage->mlinks = mlink->next; 996 mlink_free(mlink); 997 } 998 mpage_head = mpage->next; 999 free(mpage->sec); 1000 free(mpage->arch); 1001 free(mpage->title); 1002 free(mpage->desc); 1003 free(mpage); 1004 } 1005 } 1006 1007 /* 1008 * For each mlink to the mpage, check whether the path looks like 1009 * it is formatted, and if it does, check whether a source manual 1010 * exists by the same name, ignoring the suffix. 1011 * If both conditions hold, drop the mlink. 1012 */ 1013 static void 1014 mlinks_undupe(struct mpage *mpage) 1015 { 1016 char buf[PATH_MAX]; 1017 struct mlink **prev; 1018 struct mlink *mlink; 1019 char *bufp; 1020 1021 mpage->form = FORM_CAT; 1022 prev = &mpage->mlinks; 1023 while (NULL != (mlink = *prev)) { 1024 if (FORM_CAT != mlink->dform) { 1025 mpage->form = FORM_NONE; 1026 goto nextlink; 1027 } 1028 (void)strlcpy(buf, mlink->file, sizeof(buf)); 1029 bufp = strstr(buf, "cat"); 1030 assert(NULL != bufp); 1031 memcpy(bufp, "man", 3); 1032 if (NULL != (bufp = strrchr(buf, '.'))) 1033 *++bufp = '\0'; 1034 (void)strlcat(buf, mlink->dsec, sizeof(buf)); 1035 if (NULL == ohash_find(&mlinks, 1036 ohash_qlookup(&mlinks, buf))) 1037 goto nextlink; 1038 if (warnings) 1039 say(mlink->file, "Man source exists: %s", buf); 1040 if (use_all) 1041 goto nextlink; 1042 *prev = mlink->next; 1043 mlink_free(mlink); 1044 continue; 1045 nextlink: 1046 prev = &(*prev)->next; 1047 } 1048 } 1049 1050 static void 1051 mlink_check(struct mpage *mpage, struct mlink *mlink) 1052 { 1053 struct str *str; 1054 unsigned int slot; 1055 1056 /* 1057 * Check whether the manual section given in a file 1058 * agrees with the directory where the file is located. 1059 * Some manuals have suffixes like (3p) on their 1060 * section number either inside the file or in the 1061 * directory name, some are linked into more than one 1062 * section, like encrypt(1) = makekey(8). 1063 */ 1064 1065 if (FORM_SRC == mpage->form && 1066 strcasecmp(mpage->sec, mlink->dsec)) 1067 say(mlink->file, "Section \"%s\" manual in %s directory", 1068 mpage->sec, mlink->dsec); 1069 1070 /* 1071 * Manual page directories exist for each kernel 1072 * architecture as returned by machine(1). 1073 * However, many manuals only depend on the 1074 * application architecture as returned by arch(1). 1075 * For example, some (2/ARM) manuals are shared 1076 * across the "armish" and "zaurus" kernel 1077 * architectures. 1078 * A few manuals are even shared across completely 1079 * different architectures, for example fdformat(1) 1080 * on amd64, i386, and sparc64. 1081 */ 1082 1083 if (strcasecmp(mpage->arch, mlink->arch)) 1084 say(mlink->file, "Architecture \"%s\" manual in " 1085 "\"%s\" directory", mpage->arch, mlink->arch); 1086 1087 /* 1088 * XXX 1089 * parse_cat() doesn't set NAME_TITLE yet. 1090 */ 1091 1092 if (FORM_CAT == mpage->form) 1093 return; 1094 1095 /* 1096 * Check whether this mlink 1097 * appears as a name in the NAME section. 1098 */ 1099 1100 slot = ohash_qlookup(&names, mlink->name); 1101 str = ohash_find(&names, slot); 1102 assert(NULL != str); 1103 if ( ! (NAME_TITLE & str->mask)) 1104 say(mlink->file, "Name missing in NAME section"); 1105 } 1106 1107 /* 1108 * Run through the files in the global vector "mpages" 1109 * and add them to the database specified in "basedir". 1110 * 1111 * This handles the parsing scheme itself, using the cues of directory 1112 * and filename to determine whether the file is parsable or not. 1113 */ 1114 static void 1115 mpages_merge(struct dba *dba, struct mparse *mp) 1116 { 1117 struct mpage *mpage, *mpage_dest; 1118 struct mlink *mlink, *mlink_dest; 1119 struct roff_meta *meta; 1120 char *cp; 1121 int fd; 1122 1123 for (mpage = mpage_head; mpage != NULL; mpage = mpage->next) { 1124 mlinks_undupe(mpage); 1125 if ((mlink = mpage->mlinks) == NULL) 1126 continue; 1127 1128 name_mask = NAME_MASK; 1129 mandoc_ohash_init(&names, 4, offsetof(struct str, key)); 1130 mandoc_ohash_init(&strings, 6, offsetof(struct str, key)); 1131 mparse_reset(mp); 1132 meta = NULL; 1133 1134 if ((fd = mparse_open(mp, mlink->file)) == -1) { 1135 say(mlink->file, "&open"); 1136 goto nextpage; 1137 } 1138 1139 /* 1140 * Interpret the file as mdoc(7) or man(7) source 1141 * code, unless it is known to be formatted. 1142 */ 1143 if (mlink->dform != FORM_CAT || mlink->fform != FORM_CAT) { 1144 mparse_readfd(mp, fd, mlink->file); 1145 close(fd); 1146 fd = -1; 1147 meta = mparse_result(mp); 1148 } 1149 1150 if (meta != NULL && meta->sodest != NULL) { 1151 mlink_dest = ohash_find(&mlinks, 1152 ohash_qlookup(&mlinks, meta->sodest)); 1153 if (mlink_dest == NULL) { 1154 mandoc_asprintf(&cp, "%s.gz", meta->sodest); 1155 mlink_dest = ohash_find(&mlinks, 1156 ohash_qlookup(&mlinks, cp)); 1157 free(cp); 1158 } 1159 if (mlink_dest != NULL) { 1160 1161 /* The .so target exists. */ 1162 1163 mpage_dest = mlink_dest->mpage; 1164 while (1) { 1165 mlink->mpage = mpage_dest; 1166 1167 /* 1168 * If the target was already 1169 * processed, add the links 1170 * to the database now. 1171 * Otherwise, this will 1172 * happen when we come 1173 * to the target. 1174 */ 1175 1176 if (mpage_dest->dba != NULL) 1177 dbadd_mlink(mlink); 1178 1179 if (mlink->next == NULL) 1180 break; 1181 mlink = mlink->next; 1182 } 1183 1184 /* Move all links to the target. */ 1185 1186 mlink->next = mlink_dest->next; 1187 mlink_dest->next = mpage->mlinks; 1188 mpage->mlinks = NULL; 1189 } 1190 goto nextpage; 1191 } else if (meta != NULL && meta->macroset == MACROSET_MDOC) { 1192 mpage->form = FORM_SRC; 1193 mpage->sec = meta->msec; 1194 mpage->sec = mandoc_strdup( 1195 mpage->sec == NULL ? "" : mpage->sec); 1196 mpage->arch = meta->arch; 1197 mpage->arch = mandoc_strdup( 1198 mpage->arch == NULL ? "" : mpage->arch); 1199 mpage->title = mandoc_strdup(meta->title); 1200 } else if (meta != NULL && meta->macroset == MACROSET_MAN) { 1201 if (*meta->msec != '\0' || *meta->title != '\0') { 1202 mpage->form = FORM_SRC; 1203 mpage->sec = mandoc_strdup(meta->msec); 1204 mpage->arch = mandoc_strdup(mlink->arch); 1205 mpage->title = mandoc_strdup(meta->title); 1206 } else 1207 meta = NULL; 1208 } 1209 1210 assert(mpage->desc == NULL); 1211 if (meta == NULL) { 1212 mpage->form = FORM_CAT; 1213 mpage->sec = mandoc_strdup(mlink->dsec); 1214 mpage->arch = mandoc_strdup(mlink->arch); 1215 mpage->title = mandoc_strdup(mlink->name); 1216 parse_cat(mpage, fd); 1217 } else if (meta->macroset == MACROSET_MDOC) 1218 parse_mdoc(mpage, meta, meta->first); 1219 else 1220 parse_man(mpage, meta, meta->first); 1221 if (mpage->desc == NULL) { 1222 mpage->desc = mandoc_strdup(mlink->name); 1223 if (warnings) 1224 say(mlink->file, "No one-line description, " 1225 "using filename \"%s\"", mlink->name); 1226 } 1227 1228 for (mlink = mpage->mlinks; 1229 mlink != NULL; 1230 mlink = mlink->next) { 1231 putkey(mpage, mlink->name, NAME_FILE); 1232 if (warnings && !use_all) 1233 mlink_check(mpage, mlink); 1234 } 1235 1236 dbadd(dba, mpage); 1237 1238 nextpage: 1239 ohash_delete(&strings); 1240 ohash_delete(&names); 1241 } 1242 } 1243 1244 static void 1245 parse_cat(struct mpage *mpage, int fd) 1246 { 1247 FILE *stream; 1248 struct mlink *mlink; 1249 char *line, *p, *title, *sec; 1250 size_t linesz, plen, titlesz; 1251 ssize_t len; 1252 int offs; 1253 1254 mlink = mpage->mlinks; 1255 stream = fd == -1 ? fopen(mlink->file, "r") : fdopen(fd, "r"); 1256 if (stream == NULL) { 1257 if (fd != -1) 1258 close(fd); 1259 if (warnings) 1260 say(mlink->file, "&fopen"); 1261 return; 1262 } 1263 1264 line = NULL; 1265 linesz = 0; 1266 1267 /* Parse the section number from the header line. */ 1268 1269 while (getline(&line, &linesz, stream) != -1) { 1270 if (*line == '\n') 1271 continue; 1272 if ((sec = strchr(line, '(')) == NULL) 1273 break; 1274 if ((p = strchr(++sec, ')')) == NULL) 1275 break; 1276 free(mpage->sec); 1277 mpage->sec = mandoc_strndup(sec, p - sec); 1278 if (warnings && *mlink->dsec != '\0' && 1279 strcasecmp(mpage->sec, mlink->dsec)) 1280 say(mlink->file, 1281 "Section \"%s\" manual in %s directory", 1282 mpage->sec, mlink->dsec); 1283 break; 1284 } 1285 1286 /* Skip to first blank line. */ 1287 1288 while (line == NULL || *line != '\n') 1289 if (getline(&line, &linesz, stream) == -1) 1290 break; 1291 1292 /* 1293 * Assume the first line that is not indented 1294 * is the first section header. Skip to it. 1295 */ 1296 1297 while (getline(&line, &linesz, stream) != -1) 1298 if (*line != '\n' && *line != ' ') 1299 break; 1300 1301 /* 1302 * Read up until the next section into a buffer. 1303 * Strip the leading and trailing newline from each read line, 1304 * appending a trailing space. 1305 * Ignore empty (whitespace-only) lines. 1306 */ 1307 1308 titlesz = 0; 1309 title = NULL; 1310 1311 while ((len = getline(&line, &linesz, stream)) != -1) { 1312 if (*line != ' ') 1313 break; 1314 offs = 0; 1315 while (isspace((unsigned char)line[offs])) 1316 offs++; 1317 if (line[offs] == '\0') 1318 continue; 1319 title = mandoc_realloc(title, titlesz + len - offs); 1320 memcpy(title + titlesz, line + offs, len - offs); 1321 titlesz += len - offs; 1322 title[titlesz - 1] = ' '; 1323 } 1324 free(line); 1325 1326 /* 1327 * If no page content can be found, or the input line 1328 * is already the next section header, or there is no 1329 * trailing newline, reuse the page title as the page 1330 * description. 1331 */ 1332 1333 if (NULL == title || '\0' == *title) { 1334 if (warnings) 1335 say(mlink->file, "Cannot find NAME section"); 1336 fclose(stream); 1337 free(title); 1338 return; 1339 } 1340 1341 title[titlesz - 1] = '\0'; 1342 1343 /* 1344 * Skip to the first dash. 1345 * Use the remaining line as the description (no more than 70 1346 * bytes). 1347 */ 1348 1349 if (NULL != (p = strstr(title, "- "))) { 1350 for (p += 2; ' ' == *p || '\b' == *p; p++) 1351 /* Skip to next word. */ ; 1352 } else { 1353 if (warnings) 1354 say(mlink->file, "No dash in title line, " 1355 "reusing \"%s\" as one-line description", title); 1356 p = title; 1357 } 1358 1359 plen = strlen(p); 1360 1361 /* Strip backspace-encoding from line. */ 1362 1363 while (NULL != (line = memchr(p, '\b', plen))) { 1364 len = line - p; 1365 if (0 == len) { 1366 memmove(line, line + 1, plen--); 1367 continue; 1368 } 1369 memmove(line - 1, line + 1, plen - len); 1370 plen -= 2; 1371 } 1372 1373 /* 1374 * Cut off excessive one-line descriptions. 1375 * Bad pages are not worth better heuristics. 1376 */ 1377 1378 mpage->desc = mandoc_strndup(p, 150); 1379 fclose(stream); 1380 free(title); 1381 } 1382 1383 /* 1384 * Put a type/word pair into the word database for this particular file. 1385 */ 1386 static void 1387 putkey(const struct mpage *mpage, char *value, uint64_t type) 1388 { 1389 putkeys(mpage, value, strlen(value), type); 1390 } 1391 1392 /* 1393 * Grok all nodes at or below a certain mdoc node into putkey(). 1394 */ 1395 static void 1396 putmdockey(const struct mpage *mpage, 1397 const struct roff_node *n, uint64_t m, int taboo) 1398 { 1399 1400 for ( ; NULL != n; n = n->next) { 1401 if (n->flags & taboo) 1402 continue; 1403 if (NULL != n->child) 1404 putmdockey(mpage, n->child, m, taboo); 1405 if (n->type == ROFFT_TEXT) 1406 putkey(mpage, n->string, m); 1407 } 1408 } 1409 1410 static void 1411 parse_man(struct mpage *mpage, const struct roff_meta *meta, 1412 const struct roff_node *n) 1413 { 1414 const struct roff_node *head, *body; 1415 char *start, *title; 1416 char byte; 1417 size_t sz; 1418 1419 if (n == NULL) 1420 return; 1421 1422 /* 1423 * We're only searching for one thing: the first text child in 1424 * the BODY of a NAME section. Since we don't keep track of 1425 * sections in -man, run some hoops to find out whether we're in 1426 * the correct section or not. 1427 */ 1428 1429 if (n->type == ROFFT_BODY && n->tok == MAN_SH) { 1430 body = n; 1431 if ((head = body->parent->head) != NULL && 1432 (head = head->child) != NULL && 1433 head->next == NULL && 1434 head->type == ROFFT_TEXT && 1435 strcmp(head->string, "NAME") == 0 && 1436 body->child != NULL) { 1437 1438 /* 1439 * Suck the entire NAME section into memory. 1440 * Yes, we might run away. 1441 * But too many manuals have big, spread-out 1442 * NAME sections over many lines. 1443 */ 1444 1445 title = NULL; 1446 deroff(&title, body); 1447 if (NULL == title) 1448 return; 1449 1450 /* 1451 * Go through a special heuristic dance here. 1452 * Conventionally, one or more manual names are 1453 * comma-specified prior to a whitespace, then a 1454 * dash, then a description. Try to puzzle out 1455 * the name parts here. 1456 */ 1457 1458 start = title; 1459 for ( ;; ) { 1460 sz = strcspn(start, " ,"); 1461 if ('\0' == start[sz]) 1462 break; 1463 1464 byte = start[sz]; 1465 start[sz] = '\0'; 1466 1467 /* 1468 * Assume a stray trailing comma in the 1469 * name list if a name begins with a dash. 1470 */ 1471 1472 if ('-' == start[0] || 1473 ('\\' == start[0] && '-' == start[1])) 1474 break; 1475 1476 putkey(mpage, start, NAME_TITLE); 1477 if ( ! (mpage->name_head_done || 1478 strcasecmp(start, meta->title))) { 1479 putkey(mpage, start, NAME_HEAD); 1480 mpage->name_head_done = 1; 1481 } 1482 1483 if (' ' == byte) { 1484 start += sz + 1; 1485 break; 1486 } 1487 1488 assert(',' == byte); 1489 start += sz + 1; 1490 while (' ' == *start) 1491 start++; 1492 } 1493 1494 if (start == title) { 1495 putkey(mpage, start, NAME_TITLE); 1496 if ( ! (mpage->name_head_done || 1497 strcasecmp(start, meta->title))) { 1498 putkey(mpage, start, NAME_HEAD); 1499 mpage->name_head_done = 1; 1500 } 1501 free(title); 1502 return; 1503 } 1504 1505 while (isspace((unsigned char)*start)) 1506 start++; 1507 1508 if (0 == strncmp(start, "-", 1)) 1509 start += 1; 1510 else if (0 == strncmp(start, "\\-\\-", 4)) 1511 start += 4; 1512 else if (0 == strncmp(start, "\\-", 2)) 1513 start += 2; 1514 else if (0 == strncmp(start, "\\(en", 4)) 1515 start += 4; 1516 else if (0 == strncmp(start, "\\(em", 4)) 1517 start += 4; 1518 1519 while (' ' == *start) 1520 start++; 1521 1522 /* 1523 * Cut off excessive one-line descriptions. 1524 * Bad pages are not worth better heuristics. 1525 */ 1526 1527 mpage->desc = mandoc_strndup(start, 150); 1528 free(title); 1529 return; 1530 } 1531 } 1532 1533 for (n = n->child; n; n = n->next) { 1534 if (NULL != mpage->desc) 1535 break; 1536 parse_man(mpage, meta, n); 1537 } 1538 } 1539 1540 static void 1541 parse_mdoc(struct mpage *mpage, const struct roff_meta *meta, 1542 const struct roff_node *n) 1543 { 1544 const struct mdoc_handler *handler; 1545 1546 for (n = n->child; n != NULL; n = n->next) { 1547 if (n->tok == TOKEN_NONE || n->tok < ROFF_MAX) 1548 continue; 1549 assert(n->tok >= MDOC_Dd && n->tok < MDOC_MAX); 1550 handler = mdoc_handlers + (n->tok - MDOC_Dd); 1551 if (n->flags & handler->taboo) 1552 continue; 1553 1554 switch (n->type) { 1555 case ROFFT_ELEM: 1556 case ROFFT_BLOCK: 1557 case ROFFT_HEAD: 1558 case ROFFT_BODY: 1559 case ROFFT_TAIL: 1560 if (handler->fp != NULL && 1561 (*handler->fp)(mpage, meta, n) == 0) 1562 break; 1563 if (handler->mask) 1564 putmdockey(mpage, n->child, 1565 handler->mask, handler->taboo); 1566 break; 1567 default: 1568 continue; 1569 } 1570 if (NULL != n->child) 1571 parse_mdoc(mpage, meta, n); 1572 } 1573 } 1574 1575 static int 1576 parse_mdoc_Fa(struct mpage *mpage, const struct roff_meta *meta, 1577 const struct roff_node *n) 1578 { 1579 uint64_t mask; 1580 1581 mask = TYPE_Fa; 1582 if (n->sec == SEC_SYNOPSIS) 1583 mask |= TYPE_Vt; 1584 1585 putmdockey(mpage, n->child, mask, 0); 1586 return 0; 1587 } 1588 1589 static int 1590 parse_mdoc_Fd(struct mpage *mpage, const struct roff_meta *meta, 1591 const struct roff_node *n) 1592 { 1593 char *start, *end; 1594 size_t sz; 1595 1596 if (SEC_SYNOPSIS != n->sec || 1597 NULL == (n = n->child) || 1598 n->type != ROFFT_TEXT) 1599 return 0; 1600 1601 /* 1602 * Only consider those `Fd' macro fields that begin with an 1603 * "inclusion" token (versus, e.g., #define). 1604 */ 1605 1606 if (strcmp("#include", n->string)) 1607 return 0; 1608 1609 if ((n = n->next) == NULL || n->type != ROFFT_TEXT) 1610 return 0; 1611 1612 /* 1613 * Strip away the enclosing angle brackets and make sure we're 1614 * not zero-length. 1615 */ 1616 1617 start = n->string; 1618 if ('<' == *start || '"' == *start) 1619 start++; 1620 1621 if (0 == (sz = strlen(start))) 1622 return 0; 1623 1624 end = &start[(int)sz - 1]; 1625 if ('>' == *end || '"' == *end) 1626 end--; 1627 1628 if (end > start) 1629 putkeys(mpage, start, end - start + 1, TYPE_In); 1630 return 0; 1631 } 1632 1633 static void 1634 parse_mdoc_fname(struct mpage *mpage, const struct roff_node *n) 1635 { 1636 char *cp; 1637 size_t sz; 1638 1639 if (n->type != ROFFT_TEXT) 1640 return; 1641 1642 /* Skip function pointer punctuation. */ 1643 1644 cp = n->string; 1645 while (*cp == '(' || *cp == '*') 1646 cp++; 1647 sz = strcspn(cp, "()"); 1648 1649 putkeys(mpage, cp, sz, TYPE_Fn); 1650 if (n->sec == SEC_SYNOPSIS) 1651 putkeys(mpage, cp, sz, NAME_SYN); 1652 } 1653 1654 static int 1655 parse_mdoc_Fn(struct mpage *mpage, const struct roff_meta *meta, 1656 const struct roff_node *n) 1657 { 1658 uint64_t mask; 1659 1660 if (n->child == NULL) 1661 return 0; 1662 1663 parse_mdoc_fname(mpage, n->child); 1664 1665 n = n->child->next; 1666 if (n != NULL && n->type == ROFFT_TEXT) { 1667 mask = TYPE_Fa; 1668 if (n->sec == SEC_SYNOPSIS) 1669 mask |= TYPE_Vt; 1670 putmdockey(mpage, n, mask, 0); 1671 } 1672 1673 return 0; 1674 } 1675 1676 static int 1677 parse_mdoc_Fo(struct mpage *mpage, const struct roff_meta *meta, 1678 const struct roff_node *n) 1679 { 1680 1681 if (n->type != ROFFT_HEAD) 1682 return 1; 1683 1684 if (n->child != NULL) 1685 parse_mdoc_fname(mpage, n->child); 1686 1687 return 0; 1688 } 1689 1690 static int 1691 parse_mdoc_Va(struct mpage *mpage, const struct roff_meta *meta, 1692 const struct roff_node *n) 1693 { 1694 char *cp; 1695 1696 if (n->type != ROFFT_ELEM && n->type != ROFFT_BODY) 1697 return 0; 1698 1699 if (n->child != NULL && 1700 n->child->next == NULL && 1701 n->child->type == ROFFT_TEXT) 1702 return 1; 1703 1704 cp = NULL; 1705 deroff(&cp, n); 1706 if (cp != NULL) { 1707 putkey(mpage, cp, TYPE_Vt | (n->tok == MDOC_Va || 1708 n->type == ROFFT_BODY ? TYPE_Va : 0)); 1709 free(cp); 1710 } 1711 1712 return 0; 1713 } 1714 1715 static int 1716 parse_mdoc_Xr(struct mpage *mpage, const struct roff_meta *meta, 1717 const struct roff_node *n) 1718 { 1719 char *cp; 1720 1721 if (NULL == (n = n->child)) 1722 return 0; 1723 1724 if (NULL == n->next) { 1725 putkey(mpage, n->string, TYPE_Xr); 1726 return 0; 1727 } 1728 1729 mandoc_asprintf(&cp, "%s(%s)", n->string, n->next->string); 1730 putkey(mpage, cp, TYPE_Xr); 1731 free(cp); 1732 return 0; 1733 } 1734 1735 static int 1736 parse_mdoc_Nd(struct mpage *mpage, const struct roff_meta *meta, 1737 const struct roff_node *n) 1738 { 1739 1740 if (n->type == ROFFT_BODY) 1741 deroff(&mpage->desc, n); 1742 return 0; 1743 } 1744 1745 static int 1746 parse_mdoc_Nm(struct mpage *mpage, const struct roff_meta *meta, 1747 const struct roff_node *n) 1748 { 1749 1750 if (SEC_NAME == n->sec) 1751 putmdockey(mpage, n->child, NAME_TITLE, 0); 1752 else if (n->sec == SEC_SYNOPSIS && n->type == ROFFT_HEAD) { 1753 if (n->child == NULL) 1754 putkey(mpage, meta->name, NAME_SYN); 1755 else 1756 putmdockey(mpage, n->child, NAME_SYN, 0); 1757 } 1758 if ( ! (mpage->name_head_done || 1759 n->child == NULL || n->child->string == NULL || 1760 strcasecmp(n->child->string, meta->title))) { 1761 putkey(mpage, n->child->string, NAME_HEAD); 1762 mpage->name_head_done = 1; 1763 } 1764 return 0; 1765 } 1766 1767 static int 1768 parse_mdoc_Sh(struct mpage *mpage, const struct roff_meta *meta, 1769 const struct roff_node *n) 1770 { 1771 1772 return n->sec == SEC_CUSTOM && n->type == ROFFT_HEAD; 1773 } 1774 1775 static int 1776 parse_mdoc_head(struct mpage *mpage, const struct roff_meta *meta, 1777 const struct roff_node *n) 1778 { 1779 1780 return n->type == ROFFT_HEAD; 1781 } 1782 1783 /* 1784 * Add a string to the hash table for the current manual. 1785 * Each string has a bitmask telling which macros it belongs to. 1786 * When we finish the manual, we'll dump the table. 1787 */ 1788 static void 1789 putkeys(const struct mpage *mpage, char *cp, size_t sz, uint64_t v) 1790 { 1791 struct ohash *htab; 1792 struct str *s; 1793 const char *end; 1794 unsigned int slot; 1795 int i, mustfree; 1796 1797 if (0 == sz) 1798 return; 1799 1800 mustfree = render_string(&cp, &sz); 1801 1802 if (TYPE_Nm & v) { 1803 htab = &names; 1804 v &= name_mask; 1805 if (v & NAME_FIRST) 1806 name_mask &= ~NAME_FIRST; 1807 if (debug > 1) 1808 say(mpage->mlinks->file, 1809 "Adding name %*s, bits=0x%llx", (int)sz, cp, 1810 (unsigned long long)v); 1811 } else { 1812 htab = &strings; 1813 if (debug > 1) 1814 for (i = 0; i < KEY_MAX; i++) 1815 if ((uint64_t)1 << i & v) 1816 say(mpage->mlinks->file, 1817 "Adding key %s=%*s", 1818 mansearch_keynames[i], (int)sz, cp); 1819 } 1820 1821 end = cp + sz; 1822 slot = ohash_qlookupi(htab, cp, &end); 1823 s = ohash_find(htab, slot); 1824 1825 if (NULL != s && mpage == s->mpage) { 1826 s->mask |= v; 1827 return; 1828 } else if (NULL == s) { 1829 s = mandoc_calloc(1, sizeof(struct str) + sz + 1); 1830 memcpy(s->key, cp, sz); 1831 ohash_insert(htab, slot, s); 1832 } 1833 s->mpage = mpage; 1834 s->mask = v; 1835 1836 if (mustfree) 1837 free(cp); 1838 } 1839 1840 /* 1841 * Take a Unicode codepoint and produce its UTF-8 encoding. 1842 * This isn't the best way to do this, but it works. 1843 * The magic numbers are from the UTF-8 packaging. 1844 * They're not as scary as they seem: read the UTF-8 spec for details. 1845 */ 1846 static size_t 1847 utf8(unsigned int cp, char out[7]) 1848 { 1849 size_t rc; 1850 1851 rc = 0; 1852 if (cp <= 0x0000007F) { 1853 rc = 1; 1854 out[0] = (char)cp; 1855 } else if (cp <= 0x000007FF) { 1856 rc = 2; 1857 out[0] = (cp >> 6 & 31) | 192; 1858 out[1] = (cp & 63) | 128; 1859 } else if (cp <= 0x0000FFFF) { 1860 rc = 3; 1861 out[0] = (cp >> 12 & 15) | 224; 1862 out[1] = (cp >> 6 & 63) | 128; 1863 out[2] = (cp & 63) | 128; 1864 } else if (cp <= 0x001FFFFF) { 1865 rc = 4; 1866 out[0] = (cp >> 18 & 7) | 240; 1867 out[1] = (cp >> 12 & 63) | 128; 1868 out[2] = (cp >> 6 & 63) | 128; 1869 out[3] = (cp & 63) | 128; 1870 } else if (cp <= 0x03FFFFFF) { 1871 rc = 5; 1872 out[0] = (cp >> 24 & 3) | 248; 1873 out[1] = (cp >> 18 & 63) | 128; 1874 out[2] = (cp >> 12 & 63) | 128; 1875 out[3] = (cp >> 6 & 63) | 128; 1876 out[4] = (cp & 63) | 128; 1877 } else if (cp <= 0x7FFFFFFF) { 1878 rc = 6; 1879 out[0] = (cp >> 30 & 1) | 252; 1880 out[1] = (cp >> 24 & 63) | 128; 1881 out[2] = (cp >> 18 & 63) | 128; 1882 out[3] = (cp >> 12 & 63) | 128; 1883 out[4] = (cp >> 6 & 63) | 128; 1884 out[5] = (cp & 63) | 128; 1885 } else 1886 return 0; 1887 1888 out[rc] = '\0'; 1889 return rc; 1890 } 1891 1892 /* 1893 * If the string contains escape sequences, 1894 * replace it with an allocated rendering and return 1, 1895 * such that the caller can free it after use. 1896 * Otherwise, do nothing and return 0. 1897 */ 1898 static int 1899 render_string(char **public, size_t *psz) 1900 { 1901 const char *src, *scp, *addcp, *seq; 1902 char *dst; 1903 size_t ssz, dsz, addsz; 1904 char utfbuf[7], res[6]; 1905 int seqlen, unicode; 1906 1907 res[0] = '\\'; 1908 res[1] = '\t'; 1909 res[2] = ASCII_NBRSP; 1910 res[3] = ASCII_HYPH; 1911 res[4] = ASCII_BREAK; 1912 res[5] = '\0'; 1913 1914 src = scp = *public; 1915 ssz = *psz; 1916 dst = NULL; 1917 dsz = 0; 1918 1919 while (scp < src + *psz) { 1920 1921 /* Leave normal characters unchanged. */ 1922 1923 if (strchr(res, *scp) == NULL) { 1924 if (dst != NULL) 1925 dst[dsz++] = *scp; 1926 scp++; 1927 continue; 1928 } 1929 1930 /* 1931 * Found something that requires replacing, 1932 * make sure we have a destination buffer. 1933 */ 1934 1935 if (dst == NULL) { 1936 dst = mandoc_malloc(ssz + 1); 1937 dsz = scp - src; 1938 memcpy(dst, src, dsz); 1939 } 1940 1941 /* Handle single-char special characters. */ 1942 1943 switch (*scp) { 1944 case '\\': 1945 break; 1946 case '\t': 1947 case ASCII_NBRSP: 1948 dst[dsz++] = ' '; 1949 scp++; 1950 continue; 1951 case ASCII_HYPH: 1952 dst[dsz++] = '-'; 1953 /* FALLTHROUGH */ 1954 case ASCII_BREAK: 1955 scp++; 1956 continue; 1957 default: 1958 abort(); 1959 } 1960 1961 /* 1962 * Found an escape sequence. 1963 * Read past the slash, then parse it. 1964 * Ignore everything except characters. 1965 */ 1966 1967 scp++; 1968 if (mandoc_escape(&scp, &seq, &seqlen) != ESCAPE_SPECIAL) 1969 continue; 1970 1971 /* 1972 * Render the special character 1973 * as either UTF-8 or ASCII. 1974 */ 1975 1976 if (write_utf8) { 1977 unicode = mchars_spec2cp(seq, seqlen); 1978 if (unicode <= 0) 1979 continue; 1980 addsz = utf8(unicode, utfbuf); 1981 if (addsz == 0) 1982 continue; 1983 addcp = utfbuf; 1984 } else { 1985 addcp = mchars_spec2str(seq, seqlen, &addsz); 1986 if (addcp == NULL) 1987 continue; 1988 if (*addcp == ASCII_NBRSP) { 1989 addcp = " "; 1990 addsz = 1; 1991 } 1992 } 1993 1994 /* Copy the rendered glyph into the stream. */ 1995 1996 ssz += addsz; 1997 dst = mandoc_realloc(dst, ssz + 1); 1998 memcpy(dst + dsz, addcp, addsz); 1999 dsz += addsz; 2000 } 2001 if (dst != NULL) { 2002 *public = dst; 2003 *psz = dsz; 2004 } 2005 2006 /* Trim trailing whitespace and NUL-terminate. */ 2007 2008 while (*psz > 0 && (*public)[*psz - 1] == ' ') 2009 --*psz; 2010 if (dst != NULL) { 2011 (*public)[*psz] = '\0'; 2012 return 1; 2013 } else 2014 return 0; 2015 } 2016 2017 static void 2018 dbadd_mlink(const struct mlink *mlink) 2019 { 2020 dba_page_alias(mlink->mpage->dba, mlink->name, NAME_FILE); 2021 dba_page_add(mlink->mpage->dba, DBP_SECT, mlink->dsec); 2022 dba_page_add(mlink->mpage->dba, DBP_SECT, mlink->fsec); 2023 dba_page_add(mlink->mpage->dba, DBP_ARCH, mlink->arch); 2024 dba_page_add(mlink->mpage->dba, DBP_FILE, mlink->file); 2025 } 2026 2027 /* 2028 * Flush the current page's terms (and their bits) into the database. 2029 * Also, handle escape sequences at the last possible moment. 2030 */ 2031 static void 2032 dbadd(struct dba *dba, struct mpage *mpage) 2033 { 2034 struct mlink *mlink; 2035 struct str *key; 2036 char *cp; 2037 uint64_t mask; 2038 size_t i; 2039 unsigned int slot; 2040 int mustfree; 2041 2042 mlink = mpage->mlinks; 2043 2044 if (nodb) { 2045 for (key = ohash_first(&names, &slot); NULL != key; 2046 key = ohash_next(&names, &slot)) 2047 free(key); 2048 for (key = ohash_first(&strings, &slot); NULL != key; 2049 key = ohash_next(&strings, &slot)) 2050 free(key); 2051 if (0 == debug) 2052 return; 2053 while (NULL != mlink) { 2054 fputs(mlink->name, stdout); 2055 if (NULL == mlink->next || 2056 strcmp(mlink->dsec, mlink->next->dsec) || 2057 strcmp(mlink->fsec, mlink->next->fsec) || 2058 strcmp(mlink->arch, mlink->next->arch)) { 2059 putchar('('); 2060 if ('\0' == *mlink->dsec) 2061 fputs(mlink->fsec, stdout); 2062 else 2063 fputs(mlink->dsec, stdout); 2064 if ('\0' != *mlink->arch) 2065 printf("/%s", mlink->arch); 2066 putchar(')'); 2067 } 2068 mlink = mlink->next; 2069 if (NULL != mlink) 2070 fputs(", ", stdout); 2071 } 2072 printf(" - %s\n", mpage->desc); 2073 return; 2074 } 2075 2076 if (debug) 2077 say(mlink->file, "Adding to database"); 2078 2079 cp = mpage->desc; 2080 i = strlen(cp); 2081 mustfree = render_string(&cp, &i); 2082 mpage->dba = dba_page_new(dba->pages, 2083 *mpage->arch == '\0' ? mlink->arch : mpage->arch, 2084 cp, mlink->file, mpage->form); 2085 if (mustfree) 2086 free(cp); 2087 dba_page_add(mpage->dba, DBP_SECT, mpage->sec); 2088 2089 while (mlink != NULL) { 2090 dbadd_mlink(mlink); 2091 mlink = mlink->next; 2092 } 2093 2094 for (key = ohash_first(&names, &slot); NULL != key; 2095 key = ohash_next(&names, &slot)) { 2096 assert(key->mpage == mpage); 2097 dba_page_alias(mpage->dba, key->key, key->mask); 2098 free(key); 2099 } 2100 for (key = ohash_first(&strings, &slot); NULL != key; 2101 key = ohash_next(&strings, &slot)) { 2102 assert(key->mpage == mpage); 2103 i = 0; 2104 for (mask = TYPE_Xr; mask <= TYPE_Lb; mask *= 2) { 2105 if (key->mask & mask) 2106 dba_macro_add(dba->macros, i, 2107 key->key, mpage->dba); 2108 i++; 2109 } 2110 free(key); 2111 } 2112 } 2113 2114 static void 2115 dbprune(struct dba *dba) 2116 { 2117 struct dba_array *page, *files; 2118 char *file; 2119 2120 dba_array_FOREACH(dba->pages, page) { 2121 files = dba_array_get(page, DBP_FILE); 2122 dba_array_FOREACH(files, file) { 2123 if (*file < ' ') 2124 file++; 2125 if (ohash_find(&mlinks, ohash_qlookup(&mlinks, 2126 file)) != NULL) { 2127 if (debug) 2128 say(file, "Deleting from database"); 2129 dba_array_del(dba->pages); 2130 break; 2131 } 2132 } 2133 } 2134 } 2135 2136 /* 2137 * Write the database from memory to disk. 2138 */ 2139 static void 2140 dbwrite(struct dba *dba) 2141 { 2142 struct stat sb1, sb2; 2143 char tfn[33], *cp1, *cp2; 2144 off_t i; 2145 int fd1, fd2; 2146 2147 /* 2148 * Do not write empty databases, and delete existing ones 2149 * when makewhatis -u causes them to become empty. 2150 */ 2151 2152 dba_array_start(dba->pages); 2153 if (dba_array_next(dba->pages) == NULL) { 2154 if (unlink(MANDOC_DB) == -1 && errno != ENOENT) 2155 say(MANDOC_DB, "&unlink"); 2156 return; 2157 } 2158 2159 /* 2160 * Build the database in a temporary file, 2161 * then atomically move it into place. 2162 */ 2163 2164 if (dba_write(MANDOC_DB "~", dba) != -1) { 2165 if (rename(MANDOC_DB "~", MANDOC_DB) == -1) { 2166 exitcode = (int)MANDOCLEVEL_SYSERR; 2167 say(MANDOC_DB, "&rename"); 2168 unlink(MANDOC_DB "~"); 2169 } 2170 return; 2171 } 2172 2173 /* 2174 * We lack write permission and cannot replace the database 2175 * file, but let's at least check whether the data changed. 2176 */ 2177 2178 (void)strlcpy(tfn, "/tmp/mandocdb.XXXXXXXX", sizeof(tfn)); 2179 if (mkdtemp(tfn) == NULL) { 2180 exitcode = (int)MANDOCLEVEL_SYSERR; 2181 say("", "&%s", tfn); 2182 return; 2183 } 2184 cp1 = cp2 = MAP_FAILED; 2185 fd1 = fd2 = -1; 2186 (void)strlcat(tfn, "/" MANDOC_DB, sizeof(tfn)); 2187 if (dba_write(tfn, dba) == -1) { 2188 say(tfn, "&dba_write"); 2189 goto err; 2190 } 2191 if ((fd1 = open(MANDOC_DB, O_RDONLY, 0)) == -1) { 2192 say(MANDOC_DB, "&open"); 2193 goto err; 2194 } 2195 if ((fd2 = open(tfn, O_RDONLY, 0)) == -1) { 2196 say(tfn, "&open"); 2197 goto err; 2198 } 2199 if (fstat(fd1, &sb1) == -1) { 2200 say(MANDOC_DB, "&fstat"); 2201 goto err; 2202 } 2203 if (fstat(fd2, &sb2) == -1) { 2204 say(tfn, "&fstat"); 2205 goto err; 2206 } 2207 if (sb1.st_size != sb2.st_size) 2208 goto err; 2209 if ((cp1 = mmap(NULL, sb1.st_size, PROT_READ, MAP_PRIVATE, 2210 fd1, 0)) == MAP_FAILED) { 2211 say(MANDOC_DB, "&mmap"); 2212 goto err; 2213 } 2214 if ((cp2 = mmap(NULL, sb2.st_size, PROT_READ, MAP_PRIVATE, 2215 fd2, 0)) == MAP_FAILED) { 2216 say(tfn, "&mmap"); 2217 goto err; 2218 } 2219 for (i = 0; i < sb1.st_size; i++) 2220 if (cp1[i] != cp2[i]) 2221 goto err; 2222 goto out; 2223 2224 err: 2225 exitcode = (int)MANDOCLEVEL_SYSERR; 2226 say(MANDOC_DB, "Data changed, but cannot replace database"); 2227 2228 out: 2229 if (cp1 != MAP_FAILED) 2230 munmap(cp1, sb1.st_size); 2231 if (cp2 != MAP_FAILED) 2232 munmap(cp2, sb2.st_size); 2233 if (fd1 != -1) 2234 close(fd1); 2235 if (fd2 != -1) 2236 close(fd2); 2237 unlink(tfn); 2238 *strrchr(tfn, '/') = '\0'; 2239 rmdir(tfn); 2240 } 2241 2242 static int 2243 set_basedir(const char *targetdir, int report_baddir) 2244 { 2245 static char startdir[PATH_MAX]; 2246 static int getcwd_status; /* 1 = ok, 2 = failure */ 2247 static int chdir_status; /* 1 = changed directory */ 2248 char *cp; 2249 2250 /* 2251 * Remember the original working directory, if possible. 2252 * This will be needed if the second or a later directory 2253 * on the command line is given as a relative path. 2254 * Do not error out if the current directory is not 2255 * searchable: Maybe it won't be needed after all. 2256 */ 2257 if (0 == getcwd_status) { 2258 if (NULL == getcwd(startdir, sizeof(startdir))) { 2259 getcwd_status = 2; 2260 (void)strlcpy(startdir, strerror(errno), 2261 sizeof(startdir)); 2262 } else 2263 getcwd_status = 1; 2264 } 2265 2266 /* 2267 * We are leaving the old base directory. 2268 * Do not use it any longer, not even for messages. 2269 */ 2270 *basedir = '\0'; 2271 2272 /* 2273 * If and only if the directory was changed earlier and 2274 * the next directory to process is given as a relative path, 2275 * first go back, or bail out if that is impossible. 2276 */ 2277 if (chdir_status && '/' != *targetdir) { 2278 if (2 == getcwd_status) { 2279 exitcode = (int)MANDOCLEVEL_SYSERR; 2280 say("", "getcwd: %s", startdir); 2281 return 0; 2282 } 2283 if (-1 == chdir(startdir)) { 2284 exitcode = (int)MANDOCLEVEL_SYSERR; 2285 say("", "&chdir %s", startdir); 2286 return 0; 2287 } 2288 } 2289 2290 /* 2291 * Always resolve basedir to the canonicalized absolute 2292 * pathname and append a trailing slash, such that 2293 * we can reliably check whether files are inside. 2294 */ 2295 if (NULL == realpath(targetdir, basedir)) { 2296 if (report_baddir || errno != ENOENT) { 2297 exitcode = (int)MANDOCLEVEL_BADARG; 2298 say("", "&%s: realpath", targetdir); 2299 } 2300 return 0; 2301 } else if (-1 == chdir(basedir)) { 2302 if (report_baddir || errno != ENOENT) { 2303 exitcode = (int)MANDOCLEVEL_BADARG; 2304 say("", "&chdir"); 2305 } 2306 return 0; 2307 } 2308 chdir_status = 1; 2309 cp = strchr(basedir, '\0'); 2310 if ('/' != cp[-1]) { 2311 if (cp - basedir >= PATH_MAX - 1) { 2312 exitcode = (int)MANDOCLEVEL_SYSERR; 2313 say("", "Filename too long"); 2314 return 0; 2315 } 2316 *cp++ = '/'; 2317 *cp = '\0'; 2318 } 2319 return 1; 2320 } 2321 2322 static void 2323 say(const char *file, const char *format, ...) 2324 { 2325 va_list ap; 2326 int use_errno; 2327 2328 if ('\0' != *basedir) 2329 fprintf(stderr, "%s", basedir); 2330 if ('\0' != *basedir && '\0' != *file) 2331 fputc('/', stderr); 2332 if ('\0' != *file) 2333 fprintf(stderr, "%s", file); 2334 2335 use_errno = 1; 2336 if (NULL != format) { 2337 switch (*format) { 2338 case '&': 2339 format++; 2340 break; 2341 case '\0': 2342 format = NULL; 2343 break; 2344 default: 2345 use_errno = 0; 2346 break; 2347 } 2348 } 2349 if (NULL != format) { 2350 if ('\0' != *basedir || '\0' != *file) 2351 fputs(": ", stderr); 2352 va_start(ap, format); 2353 vfprintf(stderr, format, ap); 2354 va_end(ap); 2355 } 2356 if (use_errno) { 2357 if ('\0' != *basedir || '\0' != *file || NULL != format) 2358 fputs(": ", stderr); 2359 perror(NULL); 2360 } else 2361 fputc('\n', stderr); 2362 } 2363