1/* $Id: mandocdb.c,v 1.263 2019/05/03 18:17:12 schwarze Exp $ */ 2/* 3 * Copyright (c) 2011, 2012 Kristaps Dzonsons <kristaps@bsd.lv> 4 * Copyright (c) 2011-2019 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 61extern const char *const mansearch_keynames[]; 62 63enum 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 71struct 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 77struct inodev { 78 ino_t st_ino; 79 dev_t st_dev; 80}; 81 82struct 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 95struct 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 108typedef int (*mdoc_fp)(struct mpage *, const struct roff_meta *, 109 const struct roff_node *); 110 111struct 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 118int mandocdb(int, char *[]); 119 120static void dbadd(struct dba *, struct mpage *); 121static void dbadd_mlink(const struct mlink *mlink); 122static void dbprune(struct dba *); 123static void dbwrite(struct dba *); 124static void filescan(const char *); 125#if HAVE_FTS_COMPARE_CONST 126static int fts_compare(const FTSENT *const *, const FTSENT *const *); 127#else 128static int fts_compare(const FTSENT **, const FTSENT **); 129#endif 130static void mlink_add(struct mlink *, const struct stat *); 131static void mlink_check(struct mpage *, struct mlink *); 132static void mlink_free(struct mlink *); 133static void mlinks_undupe(struct mpage *); 134static void mpages_free(void); 135static void mpages_merge(struct dba *, struct mparse *); 136static void parse_cat(struct mpage *, int); 137static void parse_man(struct mpage *, const struct roff_meta *, 138 const struct roff_node *); 139static void parse_mdoc(struct mpage *, const struct roff_meta *, 140 const struct roff_node *); 141static int parse_mdoc_head(struct mpage *, const struct roff_meta *, 142 const struct roff_node *); 143static int parse_mdoc_Fa(struct mpage *, const struct roff_meta *, 144 const struct roff_node *); 145static int parse_mdoc_Fd(struct mpage *, const struct roff_meta *, 146 const struct roff_node *); 147static void parse_mdoc_fname(struct mpage *, const struct roff_node *); 148static int parse_mdoc_Fn(struct mpage *, const struct roff_meta *, 149 const struct roff_node *); 150static int parse_mdoc_Fo(struct mpage *, const struct roff_meta *, 151 const struct roff_node *); 152static int parse_mdoc_Nd(struct mpage *, const struct roff_meta *, 153 const struct roff_node *); 154static int parse_mdoc_Nm(struct mpage *, const struct roff_meta *, 155 const struct roff_node *); 156static int parse_mdoc_Sh(struct mpage *, const struct roff_meta *, 157 const struct roff_node *); 158static int parse_mdoc_Va(struct mpage *, const struct roff_meta *, 159 const struct roff_node *); 160static int parse_mdoc_Xr(struct mpage *, const struct roff_meta *, 161 const struct roff_node *); 162static void putkey(const struct mpage *, char *, uint64_t); 163static void putkeys(const struct mpage *, char *, size_t, uint64_t); 164static void putmdockey(const struct mpage *, 165 const struct roff_node *, uint64_t, int); 166static int render_string(char **, size_t *); 167static void say(const char *, const char *, ...) 168 __attribute__((__format__ (__printf__, 2, 3))); 169static int set_basedir(const char *, int); 170static int treescan(void); 171static size_t utf8(unsigned int, char [7]); 172 173static int nodb; /* no database changes */ 174static int mparse_options; /* abort the parse early */ 175static int use_all; /* use all found files */ 176static int debug; /* print what we're doing */ 177static int warnings; /* warn about crap */ 178static int write_utf8; /* write UTF-8 output; else ASCII */ 179static int exitcode; /* to be returned by main */ 180static enum op op; /* operational mode */ 181static char basedir[PATH_MAX]; /* current base directory */ 182static struct mpage *mpage_head; /* list of distinct manual pages */ 183static struct ohash mpages; /* table of distinct manual pages */ 184static struct ohash mlinks; /* table of directory entries */ 185static struct ohash names; /* table of all names */ 186static struct ohash strings; /* table of all strings */ 187static uint64_t name_mask; 188 189static 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 313int 314mandocdb(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 } 522out: 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; 530usage: 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 */ 546static int 547#if HAVE_FTS_COMPARE_CONST 548fts_compare(const FTSENT *const *a, const FTSENT *const *b) 549#else 550fts_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 */ 570static int 571treescan(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 */ 779static void 780filescan(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 929static void 930mlink_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 976static void 977mlink_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 987static void 988mpages_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 */ 1013static void 1014mlinks_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; 1045nextlink: 1046 prev = &(*prev)->next; 1047 } 1048} 1049 1050static void 1051mlink_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 */ 1114static void 1115mpages_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 goto nextpage; 1190 } 1191 meta->macroset = MACROSET_NONE; 1192 } 1193 if (meta != NULL && meta->macroset == MACROSET_MDOC) { 1194 mpage->form = FORM_SRC; 1195 mpage->sec = meta->msec; 1196 mpage->sec = mandoc_strdup( 1197 mpage->sec == NULL ? "" : mpage->sec); 1198 mpage->arch = meta->arch; 1199 mpage->arch = mandoc_strdup( 1200 mpage->arch == NULL ? "" : mpage->arch); 1201 mpage->title = mandoc_strdup(meta->title); 1202 } else if (meta != NULL && meta->macroset == MACROSET_MAN) { 1203 if (*meta->msec != '\0' || *meta->title != '\0') { 1204 mpage->form = FORM_SRC; 1205 mpage->sec = mandoc_strdup(meta->msec); 1206 mpage->arch = mandoc_strdup(mlink->arch); 1207 mpage->title = mandoc_strdup(meta->title); 1208 } else 1209 meta = NULL; 1210 } 1211 1212 assert(mpage->desc == NULL); 1213 if (meta == NULL || meta->sodest != NULL) { 1214 mpage->sec = mandoc_strdup(mlink->dsec); 1215 mpage->arch = mandoc_strdup(mlink->arch); 1216 mpage->title = mandoc_strdup(mlink->name); 1217 if (meta == NULL) { 1218 mpage->form = FORM_CAT; 1219 parse_cat(mpage, fd); 1220 } else 1221 mpage->form = FORM_SRC; 1222 } else if (meta->macroset == MACROSET_MDOC) 1223 parse_mdoc(mpage, meta, meta->first); 1224 else 1225 parse_man(mpage, meta, meta->first); 1226 if (mpage->desc == NULL) { 1227 mpage->desc = mandoc_strdup(mlink->name); 1228 if (warnings) 1229 say(mlink->file, "No one-line description, " 1230 "using filename \"%s\"", mlink->name); 1231 } 1232 1233 for (mlink = mpage->mlinks; 1234 mlink != NULL; 1235 mlink = mlink->next) { 1236 putkey(mpage, mlink->name, NAME_FILE); 1237 if (warnings && !use_all) 1238 mlink_check(mpage, mlink); 1239 } 1240 1241 dbadd(dba, mpage); 1242 1243nextpage: 1244 ohash_delete(&strings); 1245 ohash_delete(&names); 1246 } 1247} 1248 1249static void 1250parse_cat(struct mpage *mpage, int fd) 1251{ 1252 FILE *stream; 1253 struct mlink *mlink; 1254 char *line, *p, *title, *sec; 1255 size_t linesz, plen, titlesz; 1256 ssize_t len; 1257 int offs; 1258 1259 mlink = mpage->mlinks; 1260 stream = fd == -1 ? fopen(mlink->file, "r") : fdopen(fd, "r"); 1261 if (stream == NULL) { 1262 if (fd != -1) 1263 close(fd); 1264 if (warnings) 1265 say(mlink->file, "&fopen"); 1266 return; 1267 } 1268 1269 line = NULL; 1270 linesz = 0; 1271 1272 /* Parse the section number from the header line. */ 1273 1274 while (getline(&line, &linesz, stream) != -1) { 1275 if (*line == '\n') 1276 continue; 1277 if ((sec = strchr(line, '(')) == NULL) 1278 break; 1279 if ((p = strchr(++sec, ')')) == NULL) 1280 break; 1281 free(mpage->sec); 1282 mpage->sec = mandoc_strndup(sec, p - sec); 1283 if (warnings && *mlink->dsec != '\0' && 1284 strcasecmp(mpage->sec, mlink->dsec)) 1285 say(mlink->file, 1286 "Section \"%s\" manual in %s directory", 1287 mpage->sec, mlink->dsec); 1288 break; 1289 } 1290 1291 /* Skip to first blank line. */ 1292 1293 while (line == NULL || *line != '\n') 1294 if (getline(&line, &linesz, stream) == -1) 1295 break; 1296 1297 /* 1298 * Assume the first line that is not indented 1299 * is the first section header. Skip to it. 1300 */ 1301 1302 while (getline(&line, &linesz, stream) != -1) 1303 if (*line != '\n' && *line != ' ') 1304 break; 1305 1306 /* 1307 * Read up until the next section into a buffer. 1308 * Strip the leading and trailing newline from each read line, 1309 * appending a trailing space. 1310 * Ignore empty (whitespace-only) lines. 1311 */ 1312 1313 titlesz = 0; 1314 title = NULL; 1315 1316 while ((len = getline(&line, &linesz, stream)) != -1) { 1317 if (*line != ' ') 1318 break; 1319 offs = 0; 1320 while (isspace((unsigned char)line[offs])) 1321 offs++; 1322 if (line[offs] == '\0') 1323 continue; 1324 title = mandoc_realloc(title, titlesz + len - offs); 1325 memcpy(title + titlesz, line + offs, len - offs); 1326 titlesz += len - offs; 1327 title[titlesz - 1] = ' '; 1328 } 1329 free(line); 1330 1331 /* 1332 * If no page content can be found, or the input line 1333 * is already the next section header, or there is no 1334 * trailing newline, reuse the page title as the page 1335 * description. 1336 */ 1337 1338 if (NULL == title || '\0' == *title) { 1339 if (warnings) 1340 say(mlink->file, "Cannot find NAME section"); 1341 fclose(stream); 1342 free(title); 1343 return; 1344 } 1345 1346 title[titlesz - 1] = '\0'; 1347 1348 /* 1349 * Skip to the first dash. 1350 * Use the remaining line as the description (no more than 70 1351 * bytes). 1352 */ 1353 1354 if (NULL != (p = strstr(title, "- "))) { 1355 for (p += 2; ' ' == *p || '\b' == *p; p++) 1356 /* Skip to next word. */ ; 1357 } else { 1358 if (warnings) 1359 say(mlink->file, "No dash in title line, " 1360 "reusing \"%s\" as one-line description", title); 1361 p = title; 1362 } 1363 1364 plen = strlen(p); 1365 1366 /* Strip backspace-encoding from line. */ 1367 1368 while (NULL != (line = memchr(p, '\b', plen))) { 1369 len = line - p; 1370 if (0 == len) { 1371 memmove(line, line + 1, plen--); 1372 continue; 1373 } 1374 memmove(line - 1, line + 1, plen - len); 1375 plen -= 2; 1376 } 1377 1378 /* 1379 * Cut off excessive one-line descriptions. 1380 * Bad pages are not worth better heuristics. 1381 */ 1382 1383 mpage->desc = mandoc_strndup(p, 150); 1384 fclose(stream); 1385 free(title); 1386} 1387 1388/* 1389 * Put a type/word pair into the word database for this particular file. 1390 */ 1391static void 1392putkey(const struct mpage *mpage, char *value, uint64_t type) 1393{ 1394 putkeys(mpage, value, strlen(value), type); 1395} 1396 1397/* 1398 * Grok all nodes at or below a certain mdoc node into putkey(). 1399 */ 1400static void 1401putmdockey(const struct mpage *mpage, 1402 const struct roff_node *n, uint64_t m, int taboo) 1403{ 1404 1405 for ( ; NULL != n; n = n->next) { 1406 if (n->flags & taboo) 1407 continue; 1408 if (NULL != n->child) 1409 putmdockey(mpage, n->child, m, taboo); 1410 if (n->type == ROFFT_TEXT) 1411 putkey(mpage, n->string, m); 1412 } 1413} 1414 1415static void 1416parse_man(struct mpage *mpage, const struct roff_meta *meta, 1417 const struct roff_node *n) 1418{ 1419 const struct roff_node *head, *body; 1420 char *start, *title; 1421 char byte; 1422 size_t sz; 1423 1424 if (n == NULL) 1425 return; 1426 1427 /* 1428 * We're only searching for one thing: the first text child in 1429 * the BODY of a NAME section. Since we don't keep track of 1430 * sections in -man, run some hoops to find out whether we're in 1431 * the correct section or not. 1432 */ 1433 1434 if (n->type == ROFFT_BODY && n->tok == MAN_SH) { 1435 body = n; 1436 if ((head = body->parent->head) != NULL && 1437 (head = head->child) != NULL && 1438 head->next == NULL && 1439 head->type == ROFFT_TEXT && 1440 strcmp(head->string, "NAME") == 0 && 1441 body->child != NULL) { 1442 1443 /* 1444 * Suck the entire NAME section into memory. 1445 * Yes, we might run away. 1446 * But too many manuals have big, spread-out 1447 * NAME sections over many lines. 1448 */ 1449 1450 title = NULL; 1451 deroff(&title, body); 1452 if (NULL == title) 1453 return; 1454 1455 /* 1456 * Go through a special heuristic dance here. 1457 * Conventionally, one or more manual names are 1458 * comma-specified prior to a whitespace, then a 1459 * dash, then a description. Try to puzzle out 1460 * the name parts here. 1461 */ 1462 1463 start = title; 1464 for ( ;; ) { 1465 sz = strcspn(start, " ,"); 1466 if ('\0' == start[sz]) 1467 break; 1468 1469 byte = start[sz]; 1470 start[sz] = '\0'; 1471 1472 /* 1473 * Assume a stray trailing comma in the 1474 * name list if a name begins with a dash. 1475 */ 1476 1477 if ('-' == start[0] || 1478 ('\\' == start[0] && '-' == start[1])) 1479 break; 1480 1481 putkey(mpage, start, NAME_TITLE); 1482 if ( ! (mpage->name_head_done || 1483 strcasecmp(start, meta->title))) { 1484 putkey(mpage, start, NAME_HEAD); 1485 mpage->name_head_done = 1; 1486 } 1487 1488 if (' ' == byte) { 1489 start += sz + 1; 1490 break; 1491 } 1492 1493 assert(',' == byte); 1494 start += sz + 1; 1495 while (' ' == *start) 1496 start++; 1497 } 1498 1499 if (start == title) { 1500 putkey(mpage, start, NAME_TITLE); 1501 if ( ! (mpage->name_head_done || 1502 strcasecmp(start, meta->title))) { 1503 putkey(mpage, start, NAME_HEAD); 1504 mpage->name_head_done = 1; 1505 } 1506 free(title); 1507 return; 1508 } 1509 1510 while (isspace((unsigned char)*start)) 1511 start++; 1512 1513 if (0 == strncmp(start, "-", 1)) 1514 start += 1; 1515 else if (0 == strncmp(start, "\\-\\-", 4)) 1516 start += 4; 1517 else if (0 == strncmp(start, "\\-", 2)) 1518 start += 2; 1519 else if (0 == strncmp(start, "\\(en", 4)) 1520 start += 4; 1521 else if (0 == strncmp(start, "\\(em", 4)) 1522 start += 4; 1523 1524 while (' ' == *start) 1525 start++; 1526 1527 /* 1528 * Cut off excessive one-line descriptions. 1529 * Bad pages are not worth better heuristics. 1530 */ 1531 1532 mpage->desc = mandoc_strndup(start, 150); 1533 free(title); 1534 return; 1535 } 1536 } 1537 1538 for (n = n->child; n; n = n->next) { 1539 if (NULL != mpage->desc) 1540 break; 1541 parse_man(mpage, meta, n); 1542 } 1543} 1544 1545static void 1546parse_mdoc(struct mpage *mpage, const struct roff_meta *meta, 1547 const struct roff_node *n) 1548{ 1549 const struct mdoc_handler *handler; 1550 1551 for (n = n->child; n != NULL; n = n->next) { 1552 if (n->tok == TOKEN_NONE || n->tok < ROFF_MAX) 1553 continue; 1554 assert(n->tok >= MDOC_Dd && n->tok < MDOC_MAX); 1555 handler = mdoc_handlers + (n->tok - MDOC_Dd); 1556 if (n->flags & handler->taboo) 1557 continue; 1558 1559 switch (n->type) { 1560 case ROFFT_ELEM: 1561 case ROFFT_BLOCK: 1562 case ROFFT_HEAD: 1563 case ROFFT_BODY: 1564 case ROFFT_TAIL: 1565 if (handler->fp != NULL && 1566 (*handler->fp)(mpage, meta, n) == 0) 1567 break; 1568 if (handler->mask) 1569 putmdockey(mpage, n->child, 1570 handler->mask, handler->taboo); 1571 break; 1572 default: 1573 continue; 1574 } 1575 if (NULL != n->child) 1576 parse_mdoc(mpage, meta, n); 1577 } 1578} 1579 1580static int 1581parse_mdoc_Fa(struct mpage *mpage, const struct roff_meta *meta, 1582 const struct roff_node *n) 1583{ 1584 uint64_t mask; 1585 1586 mask = TYPE_Fa; 1587 if (n->sec == SEC_SYNOPSIS) 1588 mask |= TYPE_Vt; 1589 1590 putmdockey(mpage, n->child, mask, 0); 1591 return 0; 1592} 1593 1594static int 1595parse_mdoc_Fd(struct mpage *mpage, const struct roff_meta *meta, 1596 const struct roff_node *n) 1597{ 1598 char *start, *end; 1599 size_t sz; 1600 1601 if (SEC_SYNOPSIS != n->sec || 1602 NULL == (n = n->child) || 1603 n->type != ROFFT_TEXT) 1604 return 0; 1605 1606 /* 1607 * Only consider those `Fd' macro fields that begin with an 1608 * "inclusion" token (versus, e.g., #define). 1609 */ 1610 1611 if (strcmp("#include", n->string)) 1612 return 0; 1613 1614 if ((n = n->next) == NULL || n->type != ROFFT_TEXT) 1615 return 0; 1616 1617 /* 1618 * Strip away the enclosing angle brackets and make sure we're 1619 * not zero-length. 1620 */ 1621 1622 start = n->string; 1623 if ('<' == *start || '"' == *start) 1624 start++; 1625 1626 if (0 == (sz = strlen(start))) 1627 return 0; 1628 1629 end = &start[(int)sz - 1]; 1630 if ('>' == *end || '"' == *end) 1631 end--; 1632 1633 if (end > start) 1634 putkeys(mpage, start, end - start + 1, TYPE_In); 1635 return 0; 1636} 1637 1638static void 1639parse_mdoc_fname(struct mpage *mpage, const struct roff_node *n) 1640{ 1641 char *cp; 1642 size_t sz; 1643 1644 if (n->type != ROFFT_TEXT) 1645 return; 1646 1647 /* Skip function pointer punctuation. */ 1648 1649 cp = n->string; 1650 while (*cp == '(' || *cp == '*') 1651 cp++; 1652 sz = strcspn(cp, "()"); 1653 1654 putkeys(mpage, cp, sz, TYPE_Fn); 1655 if (n->sec == SEC_SYNOPSIS) 1656 putkeys(mpage, cp, sz, NAME_SYN); 1657} 1658 1659static int 1660parse_mdoc_Fn(struct mpage *mpage, const struct roff_meta *meta, 1661 const struct roff_node *n) 1662{ 1663 uint64_t mask; 1664 1665 if (n->child == NULL) 1666 return 0; 1667 1668 parse_mdoc_fname(mpage, n->child); 1669 1670 n = n->child->next; 1671 if (n != NULL && n->type == ROFFT_TEXT) { 1672 mask = TYPE_Fa; 1673 if (n->sec == SEC_SYNOPSIS) 1674 mask |= TYPE_Vt; 1675 putmdockey(mpage, n, mask, 0); 1676 } 1677 1678 return 0; 1679} 1680 1681static int 1682parse_mdoc_Fo(struct mpage *mpage, const struct roff_meta *meta, 1683 const struct roff_node *n) 1684{ 1685 1686 if (n->type != ROFFT_HEAD) 1687 return 1; 1688 1689 if (n->child != NULL) 1690 parse_mdoc_fname(mpage, n->child); 1691 1692 return 0; 1693} 1694 1695static int 1696parse_mdoc_Va(struct mpage *mpage, const struct roff_meta *meta, 1697 const struct roff_node *n) 1698{ 1699 char *cp; 1700 1701 if (n->type != ROFFT_ELEM && n->type != ROFFT_BODY) 1702 return 0; 1703 1704 if (n->child != NULL && 1705 n->child->next == NULL && 1706 n->child->type == ROFFT_TEXT) 1707 return 1; 1708 1709 cp = NULL; 1710 deroff(&cp, n); 1711 if (cp != NULL) { 1712 putkey(mpage, cp, TYPE_Vt | (n->tok == MDOC_Va || 1713 n->type == ROFFT_BODY ? TYPE_Va : 0)); 1714 free(cp); 1715 } 1716 1717 return 0; 1718} 1719 1720static int 1721parse_mdoc_Xr(struct mpage *mpage, const struct roff_meta *meta, 1722 const struct roff_node *n) 1723{ 1724 char *cp; 1725 1726 if (NULL == (n = n->child)) 1727 return 0; 1728 1729 if (NULL == n->next) { 1730 putkey(mpage, n->string, TYPE_Xr); 1731 return 0; 1732 } 1733 1734 mandoc_asprintf(&cp, "%s(%s)", n->string, n->next->string); 1735 putkey(mpage, cp, TYPE_Xr); 1736 free(cp); 1737 return 0; 1738} 1739 1740static int 1741parse_mdoc_Nd(struct mpage *mpage, const struct roff_meta *meta, 1742 const struct roff_node *n) 1743{ 1744 1745 if (n->type == ROFFT_BODY) 1746 deroff(&mpage->desc, n); 1747 return 0; 1748} 1749 1750static int 1751parse_mdoc_Nm(struct mpage *mpage, const struct roff_meta *meta, 1752 const struct roff_node *n) 1753{ 1754 1755 if (SEC_NAME == n->sec) 1756 putmdockey(mpage, n->child, NAME_TITLE, 0); 1757 else if (n->sec == SEC_SYNOPSIS && n->type == ROFFT_HEAD) { 1758 if (n->child == NULL) 1759 putkey(mpage, meta->name, NAME_SYN); 1760 else 1761 putmdockey(mpage, n->child, NAME_SYN, 0); 1762 } 1763 if ( ! (mpage->name_head_done || 1764 n->child == NULL || n->child->string == NULL || 1765 strcasecmp(n->child->string, meta->title))) { 1766 putkey(mpage, n->child->string, NAME_HEAD); 1767 mpage->name_head_done = 1; 1768 } 1769 return 0; 1770} 1771 1772static int 1773parse_mdoc_Sh(struct mpage *mpage, const struct roff_meta *meta, 1774 const struct roff_node *n) 1775{ 1776 1777 return n->sec == SEC_CUSTOM && n->type == ROFFT_HEAD; 1778} 1779 1780static int 1781parse_mdoc_head(struct mpage *mpage, const struct roff_meta *meta, 1782 const struct roff_node *n) 1783{ 1784 1785 return n->type == ROFFT_HEAD; 1786} 1787 1788/* 1789 * Add a string to the hash table for the current manual. 1790 * Each string has a bitmask telling which macros it belongs to. 1791 * When we finish the manual, we'll dump the table. 1792 */ 1793static void 1794putkeys(const struct mpage *mpage, char *cp, size_t sz, uint64_t v) 1795{ 1796 struct ohash *htab; 1797 struct str *s; 1798 const char *end; 1799 unsigned int slot; 1800 int i, mustfree; 1801 1802 if (0 == sz) 1803 return; 1804 1805 mustfree = render_string(&cp, &sz); 1806 1807 if (TYPE_Nm & v) { 1808 htab = &names; 1809 v &= name_mask; 1810 if (v & NAME_FIRST) 1811 name_mask &= ~NAME_FIRST; 1812 if (debug > 1) 1813 say(mpage->mlinks->file, 1814 "Adding name %*s, bits=0x%llx", (int)sz, cp, 1815 (unsigned long long)v); 1816 } else { 1817 htab = &strings; 1818 if (debug > 1) 1819 for (i = 0; i < KEY_MAX; i++) 1820 if ((uint64_t)1 << i & v) 1821 say(mpage->mlinks->file, 1822 "Adding key %s=%*s", 1823 mansearch_keynames[i], (int)sz, cp); 1824 } 1825 1826 end = cp + sz; 1827 slot = ohash_qlookupi(htab, cp, &end); 1828 s = ohash_find(htab, slot); 1829 1830 if (NULL != s && mpage == s->mpage) { 1831 s->mask |= v; 1832 return; 1833 } else if (NULL == s) { 1834 s = mandoc_calloc(1, sizeof(struct str) + sz + 1); 1835 memcpy(s->key, cp, sz); 1836 ohash_insert(htab, slot, s); 1837 } 1838 s->mpage = mpage; 1839 s->mask = v; 1840 1841 if (mustfree) 1842 free(cp); 1843} 1844 1845/* 1846 * Take a Unicode codepoint and produce its UTF-8 encoding. 1847 * This isn't the best way to do this, but it works. 1848 * The magic numbers are from the UTF-8 packaging. 1849 * They're not as scary as they seem: read the UTF-8 spec for details. 1850 */ 1851static size_t 1852utf8(unsigned int cp, char out[7]) 1853{ 1854 size_t rc; 1855 1856 rc = 0; 1857 if (cp <= 0x0000007F) { 1858 rc = 1; 1859 out[0] = (char)cp; 1860 } else if (cp <= 0x000007FF) { 1861 rc = 2; 1862 out[0] = (cp >> 6 & 31) | 192; 1863 out[1] = (cp & 63) | 128; 1864 } else if (cp <= 0x0000FFFF) { 1865 rc = 3; 1866 out[0] = (cp >> 12 & 15) | 224; 1867 out[1] = (cp >> 6 & 63) | 128; 1868 out[2] = (cp & 63) | 128; 1869 } else if (cp <= 0x001FFFFF) { 1870 rc = 4; 1871 out[0] = (cp >> 18 & 7) | 240; 1872 out[1] = (cp >> 12 & 63) | 128; 1873 out[2] = (cp >> 6 & 63) | 128; 1874 out[3] = (cp & 63) | 128; 1875 } else if (cp <= 0x03FFFFFF) { 1876 rc = 5; 1877 out[0] = (cp >> 24 & 3) | 248; 1878 out[1] = (cp >> 18 & 63) | 128; 1879 out[2] = (cp >> 12 & 63) | 128; 1880 out[3] = (cp >> 6 & 63) | 128; 1881 out[4] = (cp & 63) | 128; 1882 } else if (cp <= 0x7FFFFFFF) { 1883 rc = 6; 1884 out[0] = (cp >> 30 & 1) | 252; 1885 out[1] = (cp >> 24 & 63) | 128; 1886 out[2] = (cp >> 18 & 63) | 128; 1887 out[3] = (cp >> 12 & 63) | 128; 1888 out[4] = (cp >> 6 & 63) | 128; 1889 out[5] = (cp & 63) | 128; 1890 } else 1891 return 0; 1892 1893 out[rc] = '\0'; 1894 return rc; 1895} 1896 1897/* 1898 * If the string contains escape sequences, 1899 * replace it with an allocated rendering and return 1, 1900 * such that the caller can free it after use. 1901 * Otherwise, do nothing and return 0. 1902 */ 1903static int 1904render_string(char **public, size_t *psz) 1905{ 1906 const char *src, *scp, *addcp, *seq; 1907 char *dst; 1908 size_t ssz, dsz, addsz; 1909 char utfbuf[7], res[6]; 1910 int seqlen, unicode; 1911 1912 res[0] = '\\'; 1913 res[1] = '\t'; 1914 res[2] = ASCII_NBRSP; 1915 res[3] = ASCII_HYPH; 1916 res[4] = ASCII_BREAK; 1917 res[5] = '\0'; 1918 1919 src = scp = *public; 1920 ssz = *psz; 1921 dst = NULL; 1922 dsz = 0; 1923 1924 while (scp < src + *psz) { 1925 1926 /* Leave normal characters unchanged. */ 1927 1928 if (strchr(res, *scp) == NULL) { 1929 if (dst != NULL) 1930 dst[dsz++] = *scp; 1931 scp++; 1932 continue; 1933 } 1934 1935 /* 1936 * Found something that requires replacing, 1937 * make sure we have a destination buffer. 1938 */ 1939 1940 if (dst == NULL) { 1941 dst = mandoc_malloc(ssz + 1); 1942 dsz = scp - src; 1943 memcpy(dst, src, dsz); 1944 } 1945 1946 /* Handle single-char special characters. */ 1947 1948 switch (*scp) { 1949 case '\\': 1950 break; 1951 case '\t': 1952 case ASCII_NBRSP: 1953 dst[dsz++] = ' '; 1954 scp++; 1955 continue; 1956 case ASCII_HYPH: 1957 dst[dsz++] = '-'; 1958 /* FALLTHROUGH */ 1959 case ASCII_BREAK: 1960 scp++; 1961 continue; 1962 default: 1963 abort(); 1964 } 1965 1966 /* 1967 * Found an escape sequence. 1968 * Read past the slash, then parse it. 1969 * Ignore everything except characters. 1970 */ 1971 1972 scp++; 1973 if (mandoc_escape(&scp, &seq, &seqlen) != ESCAPE_SPECIAL) 1974 continue; 1975 1976 /* 1977 * Render the special character 1978 * as either UTF-8 or ASCII. 1979 */ 1980 1981 if (write_utf8) { 1982 unicode = mchars_spec2cp(seq, seqlen); 1983 if (unicode <= 0) 1984 continue; 1985 addsz = utf8(unicode, utfbuf); 1986 if (addsz == 0) 1987 continue; 1988 addcp = utfbuf; 1989 } else { 1990 addcp = mchars_spec2str(seq, seqlen, &addsz); 1991 if (addcp == NULL) 1992 continue; 1993 if (*addcp == ASCII_NBRSP) { 1994 addcp = " "; 1995 addsz = 1; 1996 } 1997 } 1998 1999 /* Copy the rendered glyph into the stream. */ 2000 2001 ssz += addsz; 2002 dst = mandoc_realloc(dst, ssz + 1); 2003 memcpy(dst + dsz, addcp, addsz); 2004 dsz += addsz; 2005 } 2006 if (dst != NULL) { 2007 *public = dst; 2008 *psz = dsz; 2009 } 2010 2011 /* Trim trailing whitespace and NUL-terminate. */ 2012 2013 while (*psz > 0 && (*public)[*psz - 1] == ' ') 2014 --*psz; 2015 if (dst != NULL) { 2016 (*public)[*psz] = '\0'; 2017 return 1; 2018 } else 2019 return 0; 2020} 2021 2022static void 2023dbadd_mlink(const struct mlink *mlink) 2024{ 2025 dba_page_alias(mlink->mpage->dba, mlink->name, NAME_FILE); 2026 dba_page_add(mlink->mpage->dba, DBP_SECT, mlink->dsec); 2027 dba_page_add(mlink->mpage->dba, DBP_SECT, mlink->fsec); 2028 dba_page_add(mlink->mpage->dba, DBP_ARCH, mlink->arch); 2029 dba_page_add(mlink->mpage->dba, DBP_FILE, mlink->file); 2030} 2031 2032/* 2033 * Flush the current page's terms (and their bits) into the database. 2034 * Also, handle escape sequences at the last possible moment. 2035 */ 2036static void 2037dbadd(struct dba *dba, struct mpage *mpage) 2038{ 2039 struct mlink *mlink; 2040 struct str *key; 2041 char *cp; 2042 uint64_t mask; 2043 size_t i; 2044 unsigned int slot; 2045 int mustfree; 2046 2047 mlink = mpage->mlinks; 2048 2049 if (nodb) { 2050 for (key = ohash_first(&names, &slot); NULL != key; 2051 key = ohash_next(&names, &slot)) 2052 free(key); 2053 for (key = ohash_first(&strings, &slot); NULL != key; 2054 key = ohash_next(&strings, &slot)) 2055 free(key); 2056 if (0 == debug) 2057 return; 2058 while (NULL != mlink) { 2059 fputs(mlink->name, stdout); 2060 if (NULL == mlink->next || 2061 strcmp(mlink->dsec, mlink->next->dsec) || 2062 strcmp(mlink->fsec, mlink->next->fsec) || 2063 strcmp(mlink->arch, mlink->next->arch)) { 2064 putchar('('); 2065 if ('\0' == *mlink->dsec) 2066 fputs(mlink->fsec, stdout); 2067 else 2068 fputs(mlink->dsec, stdout); 2069 if ('\0' != *mlink->arch) 2070 printf("/%s", mlink->arch); 2071 putchar(')'); 2072 } 2073 mlink = mlink->next; 2074 if (NULL != mlink) 2075 fputs(", ", stdout); 2076 } 2077 printf(" - %s\n", mpage->desc); 2078 return; 2079 } 2080 2081 if (debug) 2082 say(mlink->file, "Adding to database"); 2083 2084 cp = mpage->desc; 2085 i = strlen(cp); 2086 mustfree = render_string(&cp, &i); 2087 mpage->dba = dba_page_new(dba->pages, 2088 *mpage->arch == '\0' ? mlink->arch : mpage->arch, 2089 cp, mlink->file, mpage->form); 2090 if (mustfree) 2091 free(cp); 2092 dba_page_add(mpage->dba, DBP_SECT, mpage->sec); 2093 2094 while (mlink != NULL) { 2095 dbadd_mlink(mlink); 2096 mlink = mlink->next; 2097 } 2098 2099 for (key = ohash_first(&names, &slot); NULL != key; 2100 key = ohash_next(&names, &slot)) { 2101 assert(key->mpage == mpage); 2102 dba_page_alias(mpage->dba, key->key, key->mask); 2103 free(key); 2104 } 2105 for (key = ohash_first(&strings, &slot); NULL != key; 2106 key = ohash_next(&strings, &slot)) { 2107 assert(key->mpage == mpage); 2108 i = 0; 2109 for (mask = TYPE_Xr; mask <= TYPE_Lb; mask *= 2) { 2110 if (key->mask & mask) 2111 dba_macro_add(dba->macros, i, 2112 key->key, mpage->dba); 2113 i++; 2114 } 2115 free(key); 2116 } 2117} 2118 2119static void 2120dbprune(struct dba *dba) 2121{ 2122 struct dba_array *page, *files; 2123 char *file; 2124 2125 dba_array_FOREACH(dba->pages, page) { 2126 files = dba_array_get(page, DBP_FILE); 2127 dba_array_FOREACH(files, file) { 2128 if (*file < ' ') 2129 file++; 2130 if (ohash_find(&mlinks, ohash_qlookup(&mlinks, 2131 file)) != NULL) { 2132 if (debug) 2133 say(file, "Deleting from database"); 2134 dba_array_del(dba->pages); 2135 break; 2136 } 2137 } 2138 } 2139} 2140 2141/* 2142 * Write the database from memory to disk. 2143 */ 2144static void 2145dbwrite(struct dba *dba) 2146{ 2147 struct stat sb1, sb2; 2148 char tfn[33], *cp1, *cp2; 2149 off_t i; 2150 int fd1, fd2; 2151 2152 /* 2153 * Do not write empty databases, and delete existing ones 2154 * when makewhatis -u causes them to become empty. 2155 */ 2156 2157 dba_array_start(dba->pages); 2158 if (dba_array_next(dba->pages) == NULL) { 2159 if (unlink(MANDOC_DB) == -1 && errno != ENOENT) 2160 say(MANDOC_DB, "&unlink"); 2161 return; 2162 } 2163 2164 /* 2165 * Build the database in a temporary file, 2166 * then atomically move it into place. 2167 */ 2168 2169 if (dba_write(MANDOC_DB "~", dba) != -1) { 2170 if (rename(MANDOC_DB "~", MANDOC_DB) == -1) { 2171 exitcode = (int)MANDOCLEVEL_SYSERR; 2172 say(MANDOC_DB, "&rename"); 2173 unlink(MANDOC_DB "~"); 2174 } 2175 return; 2176 } 2177 2178 /* 2179 * We lack write permission and cannot replace the database 2180 * file, but let's at least check whether the data changed. 2181 */ 2182 2183 (void)strlcpy(tfn, "/tmp/mandocdb.XXXXXXXX", sizeof(tfn)); 2184 if (mkdtemp(tfn) == NULL) { 2185 exitcode = (int)MANDOCLEVEL_SYSERR; 2186 say("", "&%s", tfn); 2187 return; 2188 } 2189 cp1 = cp2 = MAP_FAILED; 2190 fd1 = fd2 = -1; 2191 (void)strlcat(tfn, "/" MANDOC_DB, sizeof(tfn)); 2192 if (dba_write(tfn, dba) == -1) { 2193 say(tfn, "&dba_write"); 2194 goto err; 2195 } 2196 if ((fd1 = open(MANDOC_DB, O_RDONLY, 0)) == -1) { 2197 say(MANDOC_DB, "&open"); 2198 goto err; 2199 } 2200 if ((fd2 = open(tfn, O_RDONLY, 0)) == -1) { 2201 say(tfn, "&open"); 2202 goto err; 2203 } 2204 if (fstat(fd1, &sb1) == -1) { 2205 say(MANDOC_DB, "&fstat"); 2206 goto err; 2207 } 2208 if (fstat(fd2, &sb2) == -1) { 2209 say(tfn, "&fstat"); 2210 goto err; 2211 } 2212 if (sb1.st_size != sb2.st_size) 2213 goto err; 2214 if ((cp1 = mmap(NULL, sb1.st_size, PROT_READ, MAP_PRIVATE, 2215 fd1, 0)) == MAP_FAILED) { 2216 say(MANDOC_DB, "&mmap"); 2217 goto err; 2218 } 2219 if ((cp2 = mmap(NULL, sb2.st_size, PROT_READ, MAP_PRIVATE, 2220 fd2, 0)) == MAP_FAILED) { 2221 say(tfn, "&mmap"); 2222 goto err; 2223 } 2224 for (i = 0; i < sb1.st_size; i++) 2225 if (cp1[i] != cp2[i]) 2226 goto err; 2227 goto out; 2228 2229err: 2230 exitcode = (int)MANDOCLEVEL_SYSERR; 2231 say(MANDOC_DB, "Data changed, but cannot replace database"); 2232 2233out: 2234 if (cp1 != MAP_FAILED) 2235 munmap(cp1, sb1.st_size); 2236 if (cp2 != MAP_FAILED) 2237 munmap(cp2, sb2.st_size); 2238 if (fd1 != -1) 2239 close(fd1); 2240 if (fd2 != -1) 2241 close(fd2); 2242 unlink(tfn); 2243 *strrchr(tfn, '/') = '\0'; 2244 rmdir(tfn); 2245} 2246 2247static int 2248set_basedir(const char *targetdir, int report_baddir) 2249{ 2250 static char startdir[PATH_MAX]; 2251 static int getcwd_status; /* 1 = ok, 2 = failure */ 2252 static int chdir_status; /* 1 = changed directory */ 2253 char *cp; 2254 2255 /* 2256 * Remember the original working directory, if possible. 2257 * This will be needed if the second or a later directory 2258 * on the command line is given as a relative path. 2259 * Do not error out if the current directory is not 2260 * searchable: Maybe it won't be needed after all. 2261 */ 2262 if (0 == getcwd_status) { 2263 if (NULL == getcwd(startdir, sizeof(startdir))) { 2264 getcwd_status = 2; 2265 (void)strlcpy(startdir, strerror(errno), 2266 sizeof(startdir)); 2267 } else 2268 getcwd_status = 1; 2269 } 2270 2271 /* 2272 * We are leaving the old base directory. 2273 * Do not use it any longer, not even for messages. 2274 */ 2275 *basedir = '\0'; 2276 2277 /* 2278 * If and only if the directory was changed earlier and 2279 * the next directory to process is given as a relative path, 2280 * first go back, or bail out if that is impossible. 2281 */ 2282 if (chdir_status && '/' != *targetdir) { 2283 if (2 == getcwd_status) { 2284 exitcode = (int)MANDOCLEVEL_SYSERR; 2285 say("", "getcwd: %s", startdir); 2286 return 0; 2287 } 2288 if (-1 == chdir(startdir)) { 2289 exitcode = (int)MANDOCLEVEL_SYSERR; 2290 say("", "&chdir %s", startdir); 2291 return 0; 2292 } 2293 } 2294 2295 /* 2296 * Always resolve basedir to the canonicalized absolute 2297 * pathname and append a trailing slash, such that 2298 * we can reliably check whether files are inside. 2299 */ 2300 if (NULL == realpath(targetdir, basedir)) { 2301 if (report_baddir || errno != ENOENT) { 2302 exitcode = (int)MANDOCLEVEL_BADARG; 2303 say("", "&%s: realpath", targetdir); 2304 } 2305 return 0; 2306 } else if (-1 == chdir(basedir)) { 2307 if (report_baddir || errno != ENOENT) { 2308 exitcode = (int)MANDOCLEVEL_BADARG; 2309 say("", "&chdir"); 2310 } 2311 return 0; 2312 } 2313 chdir_status = 1; 2314 cp = strchr(basedir, '\0'); 2315 if ('/' != cp[-1]) { 2316 if (cp - basedir >= PATH_MAX - 1) { 2317 exitcode = (int)MANDOCLEVEL_SYSERR; 2318 say("", "Filename too long"); 2319 return 0; 2320 } 2321 *cp++ = '/'; 2322 *cp = '\0'; 2323 } 2324 return 1; 2325} 2326 2327static void 2328say(const char *file, const char *format, ...) 2329{ 2330 va_list ap; 2331 int use_errno; 2332 2333 if ('\0' != *basedir) 2334 fprintf(stderr, "%s", basedir); 2335 if ('\0' != *basedir && '\0' != *file) 2336 fputc('/', stderr); 2337 if ('\0' != *file) 2338 fprintf(stderr, "%s", file); 2339 2340 use_errno = 1; 2341 if (NULL != format) { 2342 switch (*format) { 2343 case '&': 2344 format++; 2345 break; 2346 case '\0': 2347 format = NULL; 2348 break; 2349 default: 2350 use_errno = 0; 2351 break; 2352 } 2353 } 2354 if (NULL != format) { 2355 if ('\0' != *basedir || '\0' != *file) 2356 fputs(": ", stderr); 2357 va_start(ap, format); 2358 vfprintf(stderr, format, ap); 2359 va_end(ap); 2360 } 2361 if (use_errno) { 2362 if ('\0' != *basedir || '\0' != *file || NULL != format) 2363 fputs(": ", stderr); 2364 perror(NULL); 2365 } else 2366 fputc('\n', stderr); 2367} 2368