1/* $Id: mandocdb.c,v 1.253 2017/07/28 14:48:25 schwarze Exp $ */ 2/* 3 * Copyright (c) 2011, 2012 Kristaps Dzonsons <kristaps@bsd.lv> 4 * Copyright (c) 2011-2017 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/stat.h> 23#include <sys/wait.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 "manconf.h" 56#include "mansearch.h" 57#include "dba_array.h" 58#include "dba.h" 59 60extern const char *const mansearch_keynames[]; 61 62enum op { 63 OP_DEFAULT = 0, /* new dbs from dir list or default config */ 64 OP_CONFFILE, /* new databases from custom config file */ 65 OP_UPDATE, /* delete/add entries in existing database */ 66 OP_DELETE, /* delete entries from existing database */ 67 OP_TEST /* change no databases, report potential problems */ 68}; 69 70struct str { 71 const struct mpage *mpage; /* if set, the owning parse */ 72 uint64_t mask; /* bitmask in sequence */ 73 char key[]; /* rendered text */ 74}; 75 76struct inodev { 77 ino_t st_ino; 78 dev_t st_dev; 79}; 80 81struct mpage { 82 struct inodev inodev; /* used for hashing routine */ 83 struct dba_array *dba; 84 char *sec; /* section from file content */ 85 char *arch; /* architecture from file content */ 86 char *title; /* title from file content */ 87 char *desc; /* description from file content */ 88 struct mpage *next; /* singly linked list */ 89 struct mlink *mlinks; /* singly linked list */ 90 int name_head_done; 91 enum form form; /* format from file content */ 92}; 93 94struct mlink { 95 char file[PATH_MAX]; /* filename rel. to manpath */ 96 char *dsec; /* section from directory */ 97 char *arch; /* architecture from directory */ 98 char *name; /* name from file name (not empty) */ 99 char *fsec; /* section from file name suffix */ 100 struct mlink *next; /* singly linked list */ 101 struct mpage *mpage; /* parent */ 102 int gzip; /* filename has a .gz suffix */ 103 enum form dform; /* format from directory */ 104 enum form fform; /* format from file name suffix */ 105}; 106 107typedef int (*mdoc_fp)(struct mpage *, const struct roff_meta *, 108 const struct roff_node *); 109 110struct mdoc_handler { 111 mdoc_fp fp; /* optional handler */ 112 uint64_t mask; /* set unless handler returns 0 */ 113 int taboo; /* node flags that must not be set */ 114}; 115 116 117int mandocdb(int, char *[]); 118 119static void dbadd(struct dba *, struct mpage *); 120static void dbadd_mlink(const struct mlink *mlink); 121static void dbprune(struct dba *); 122static void dbwrite(struct dba *); 123static void filescan(const char *); 124#if HAVE_FTS_COMPARE_CONST 125static int fts_compare(const FTSENT *const *, const FTSENT *const *); 126#else 127static int fts_compare(const FTSENT **, const FTSENT **); 128#endif 129static void mlink_add(struct mlink *, const struct stat *); 130static void mlink_check(struct mpage *, struct mlink *); 131static void mlink_free(struct mlink *); 132static void mlinks_undupe(struct mpage *); 133static void mpages_free(void); 134static void mpages_merge(struct dba *, struct mparse *); 135static void parse_cat(struct mpage *, int); 136static void parse_man(struct mpage *, const struct roff_meta *, 137 const struct roff_node *); 138static void parse_mdoc(struct mpage *, const struct roff_meta *, 139 const struct roff_node *); 140static int parse_mdoc_head(struct mpage *, const struct roff_meta *, 141 const struct roff_node *); 142static int parse_mdoc_Fd(struct mpage *, const struct roff_meta *, 143 const struct roff_node *); 144static void parse_mdoc_fname(struct mpage *, const struct roff_node *); 145static int parse_mdoc_Fn(struct mpage *, const struct roff_meta *, 146 const struct roff_node *); 147static int parse_mdoc_Fo(struct mpage *, const struct roff_meta *, 148 const struct roff_node *); 149static int parse_mdoc_Nd(struct mpage *, const struct roff_meta *, 150 const struct roff_node *); 151static int parse_mdoc_Nm(struct mpage *, const struct roff_meta *, 152 const struct roff_node *); 153static int parse_mdoc_Sh(struct mpage *, const struct roff_meta *, 154 const struct roff_node *); 155static int parse_mdoc_Va(struct mpage *, const struct roff_meta *, 156 const struct roff_node *); 157static int parse_mdoc_Xr(struct mpage *, const struct roff_meta *, 158 const struct roff_node *); 159static void putkey(const struct mpage *, char *, uint64_t); 160static void putkeys(const struct mpage *, char *, size_t, uint64_t); 161static void putmdockey(const struct mpage *, 162 const struct roff_node *, uint64_t, int); 163static int render_string(char **, size_t *); 164static void say(const char *, const char *, ...) 165 __attribute__((__format__ (__printf__, 2, 3))); 166static int set_basedir(const char *, int); 167static int treescan(void); 168static size_t utf8(unsigned int, char [7]); 169 170static int nodb; /* no database changes */ 171static int mparse_options; /* abort the parse early */ 172static int use_all; /* use all found files */ 173static int debug; /* print what we're doing */ 174static int warnings; /* warn about crap */ 175static int write_utf8; /* write UTF-8 output; else ASCII */ 176static int exitcode; /* to be returned by main */ 177static enum op op; /* operational mode */ 178static char basedir[PATH_MAX]; /* current base directory */ 179static struct mpage *mpage_head; /* list of distinct manual pages */ 180static struct ohash mpages; /* table of distinct manual pages */ 181static struct ohash mlinks; /* table of directory entries */ 182static struct ohash names; /* table of all names */ 183static struct ohash strings; /* table of all strings */ 184static uint64_t name_mask; 185 186static const struct mdoc_handler __mdocs[MDOC_MAX - MDOC_Dd] = { 187 { NULL, 0, NODE_NOPRT }, /* Dd */ 188 { NULL, 0, NODE_NOPRT }, /* Dt */ 189 { NULL, 0, NODE_NOPRT }, /* Os */ 190 { parse_mdoc_Sh, TYPE_Sh, 0 }, /* Sh */ 191 { parse_mdoc_head, TYPE_Ss, 0 }, /* Ss */ 192 { NULL, 0, 0 }, /* Pp */ 193 { NULL, 0, 0 }, /* D1 */ 194 { NULL, 0, 0 }, /* Dl */ 195 { NULL, 0, 0 }, /* Bd */ 196 { NULL, 0, 0 }, /* Ed */ 197 { NULL, 0, 0 }, /* Bl */ 198 { NULL, 0, 0 }, /* El */ 199 { NULL, 0, 0 }, /* It */ 200 { NULL, 0, 0 }, /* Ad */ 201 { NULL, TYPE_An, 0 }, /* An */ 202 { NULL, 0, 0 }, /* Ap */ 203 { NULL, TYPE_Ar, 0 }, /* Ar */ 204 { NULL, TYPE_Cd, 0 }, /* Cd */ 205 { NULL, TYPE_Cm, 0 }, /* Cm */ 206 { NULL, TYPE_Dv, 0 }, /* Dv */ 207 { NULL, TYPE_Er, 0 }, /* Er */ 208 { NULL, TYPE_Ev, 0 }, /* Ev */ 209 { NULL, 0, 0 }, /* Ex */ 210 { NULL, TYPE_Fa, 0 }, /* Fa */ 211 { parse_mdoc_Fd, 0, 0 }, /* Fd */ 212 { NULL, TYPE_Fl, 0 }, /* Fl */ 213 { parse_mdoc_Fn, 0, 0 }, /* Fn */ 214 { NULL, TYPE_Ft, 0 }, /* Ft */ 215 { NULL, TYPE_Ic, 0 }, /* Ic */ 216 { NULL, TYPE_In, 0 }, /* In */ 217 { NULL, TYPE_Li, 0 }, /* Li */ 218 { parse_mdoc_Nd, 0, 0 }, /* Nd */ 219 { parse_mdoc_Nm, 0, 0 }, /* Nm */ 220 { NULL, 0, 0 }, /* Op */ 221 { NULL, 0, 0 }, /* Ot */ 222 { NULL, TYPE_Pa, NODE_NOSRC }, /* Pa */ 223 { NULL, 0, 0 }, /* Rv */ 224 { NULL, TYPE_St, 0 }, /* St */ 225 { parse_mdoc_Va, TYPE_Va, 0 }, /* Va */ 226 { parse_mdoc_Va, TYPE_Vt, 0 }, /* Vt */ 227 { parse_mdoc_Xr, 0, 0 }, /* Xr */ 228 { NULL, 0, 0 }, /* %A */ 229 { NULL, 0, 0 }, /* %B */ 230 { NULL, 0, 0 }, /* %D */ 231 { NULL, 0, 0 }, /* %I */ 232 { NULL, 0, 0 }, /* %J */ 233 { NULL, 0, 0 }, /* %N */ 234 { NULL, 0, 0 }, /* %O */ 235 { NULL, 0, 0 }, /* %P */ 236 { NULL, 0, 0 }, /* %R */ 237 { NULL, 0, 0 }, /* %T */ 238 { NULL, 0, 0 }, /* %V */ 239 { NULL, 0, 0 }, /* Ac */ 240 { NULL, 0, 0 }, /* Ao */ 241 { NULL, 0, 0 }, /* Aq */ 242 { NULL, TYPE_At, 0 }, /* At */ 243 { NULL, 0, 0 }, /* Bc */ 244 { NULL, 0, 0 }, /* Bf */ 245 { NULL, 0, 0 }, /* Bo */ 246 { NULL, 0, 0 }, /* Bq */ 247 { NULL, TYPE_Bsx, NODE_NOSRC }, /* Bsx */ 248 { NULL, TYPE_Bx, NODE_NOSRC }, /* Bx */ 249 { NULL, 0, 0 }, /* Db */ 250 { NULL, 0, 0 }, /* Dc */ 251 { NULL, 0, 0 }, /* Do */ 252 { NULL, 0, 0 }, /* Dq */ 253 { NULL, 0, 0 }, /* Ec */ 254 { NULL, 0, 0 }, /* Ef */ 255 { NULL, TYPE_Em, 0 }, /* Em */ 256 { NULL, 0, 0 }, /* Eo */ 257 { NULL, TYPE_Fx, NODE_NOSRC }, /* Fx */ 258 { NULL, TYPE_Ms, 0 }, /* Ms */ 259 { NULL, 0, 0 }, /* No */ 260 { NULL, 0, 0 }, /* Ns */ 261 { NULL, TYPE_Nx, NODE_NOSRC }, /* Nx */ 262 { NULL, TYPE_Ox, NODE_NOSRC }, /* Ox */ 263 { NULL, 0, 0 }, /* Pc */ 264 { NULL, 0, 0 }, /* Pf */ 265 { NULL, 0, 0 }, /* Po */ 266 { NULL, 0, 0 }, /* Pq */ 267 { NULL, 0, 0 }, /* Qc */ 268 { NULL, 0, 0 }, /* Ql */ 269 { NULL, 0, 0 }, /* Qo */ 270 { NULL, 0, 0 }, /* Qq */ 271 { NULL, 0, 0 }, /* Re */ 272 { NULL, 0, 0 }, /* Rs */ 273 { NULL, 0, 0 }, /* Sc */ 274 { NULL, 0, 0 }, /* So */ 275 { NULL, 0, 0 }, /* Sq */ 276 { NULL, 0, 0 }, /* Sm */ 277 { NULL, 0, 0 }, /* Sx */ 278 { NULL, TYPE_Sy, 0 }, /* Sy */ 279 { NULL, TYPE_Tn, 0 }, /* Tn */ 280 { NULL, 0, NODE_NOSRC }, /* Ux */ 281 { NULL, 0, 0 }, /* Xc */ 282 { NULL, 0, 0 }, /* Xo */ 283 { parse_mdoc_Fo, 0, 0 }, /* Fo */ 284 { NULL, 0, 0 }, /* Fc */ 285 { NULL, 0, 0 }, /* Oo */ 286 { NULL, 0, 0 }, /* Oc */ 287 { NULL, 0, 0 }, /* Bk */ 288 { NULL, 0, 0 }, /* Ek */ 289 { NULL, 0, 0 }, /* Bt */ 290 { NULL, 0, 0 }, /* Hf */ 291 { NULL, 0, 0 }, /* Fr */ 292 { NULL, 0, 0 }, /* Ud */ 293 { NULL, TYPE_Lb, NODE_NOSRC }, /* Lb */ 294 { NULL, 0, 0 }, /* Lp */ 295 { NULL, TYPE_Lk, 0 }, /* Lk */ 296 { NULL, TYPE_Mt, NODE_NOSRC }, /* Mt */ 297 { NULL, 0, 0 }, /* Brq */ 298 { NULL, 0, 0 }, /* Bro */ 299 { NULL, 0, 0 }, /* Brc */ 300 { NULL, 0, 0 }, /* %C */ 301 { NULL, 0, 0 }, /* Es */ 302 { NULL, 0, 0 }, /* En */ 303 { NULL, TYPE_Dx, NODE_NOSRC }, /* Dx */ 304 { NULL, 0, 0 }, /* %Q */ 305 { NULL, 0, 0 }, /* %U */ 306 { NULL, 0, 0 }, /* Ta */ 307}; 308static const struct mdoc_handler *const mdocs = __mdocs - MDOC_Dd; 309 310 311int 312mandocdb(int argc, char *argv[]) 313{ 314 struct manconf conf; 315 struct mparse *mp; 316 struct dba *dba; 317 const char *path_arg, *progname; 318 size_t j, sz; 319 int ch, i; 320 321#if HAVE_PLEDGE 322 if (pledge("stdio rpath wpath cpath fattr flock proc exec", NULL) == -1) { 323 warn("pledge"); 324 return (int)MANDOCLEVEL_SYSERR; 325 } 326#endif 327 328#if HAVE_SANDBOX_INIT 329 if (sandbox_init(kSBXProfileNoInternet, SANDBOX_NAMED, NULL) == -1) { 330 warnx("sandbox_init"); 331 return (int)MANDOCLEVEL_SYSERR; 332 } 333#endif 334 335 memset(&conf, 0, sizeof(conf)); 336 337 /* 338 * We accept a few different invocations. 339 * The CHECKOP macro makes sure that invocation styles don't 340 * clobber each other. 341 */ 342#define CHECKOP(_op, _ch) do \ 343 if (OP_DEFAULT != (_op)) { \ 344 warnx("-%c: Conflicting option", (_ch)); \ 345 goto usage; \ 346 } while (/*CONSTCOND*/0) 347 348 path_arg = NULL; 349 op = OP_DEFAULT; 350 351 while (-1 != (ch = getopt(argc, argv, "aC:Dd:npQT:tu:v"))) 352 switch (ch) { 353 case 'a': 354 use_all = 1; 355 break; 356 case 'C': 357 CHECKOP(op, ch); 358 path_arg = optarg; 359 op = OP_CONFFILE; 360 break; 361 case 'D': 362 debug++; 363 break; 364 case 'd': 365 CHECKOP(op, ch); 366 path_arg = optarg; 367 op = OP_UPDATE; 368 break; 369 case 'n': 370 nodb = 1; 371 break; 372 case 'p': 373 warnings = 1; 374 break; 375 case 'Q': 376 mparse_options |= MPARSE_QUICK; 377 break; 378 case 'T': 379 if (strcmp(optarg, "utf8")) { 380 warnx("-T%s: Unsupported output format", 381 optarg); 382 goto usage; 383 } 384 write_utf8 = 1; 385 break; 386 case 't': 387 CHECKOP(op, ch); 388 dup2(STDOUT_FILENO, STDERR_FILENO); 389 op = OP_TEST; 390 nodb = warnings = 1; 391 break; 392 case 'u': 393 CHECKOP(op, ch); 394 path_arg = optarg; 395 op = OP_DELETE; 396 break; 397 case 'v': 398 /* Compatibility with espie@'s makewhatis. */ 399 break; 400 default: 401 goto usage; 402 } 403 404 argc -= optind; 405 argv += optind; 406 407#if HAVE_PLEDGE 408 if (nodb) { 409 if (pledge("stdio rpath", NULL) == -1) { 410 warn("pledge"); 411 return (int)MANDOCLEVEL_SYSERR; 412 } 413 } 414#endif 415 416 if (OP_CONFFILE == op && argc > 0) { 417 warnx("-C: Too many arguments"); 418 goto usage; 419 } 420 421 exitcode = (int)MANDOCLEVEL_OK; 422 mchars_alloc(); 423 mp = mparse_alloc(mparse_options, MANDOCERR_MAX, NULL, 424 MANDOC_OS_OTHER, NULL); 425 mandoc_ohash_init(&mpages, 6, offsetof(struct mpage, inodev)); 426 mandoc_ohash_init(&mlinks, 6, offsetof(struct mlink, file)); 427 428 if (OP_UPDATE == op || OP_DELETE == op || OP_TEST == op) { 429 430 /* 431 * Most of these deal with a specific directory. 432 * Jump into that directory first. 433 */ 434 if (OP_TEST != op && 0 == set_basedir(path_arg, 1)) 435 goto out; 436 437 dba = nodb ? dba_new(128) : dba_read(MANDOC_DB); 438 if (dba != NULL) { 439 /* 440 * The existing database is usable. Process 441 * all files specified on the command-line. 442 */ 443#if HAVE_PLEDGE 444 if (!nodb) { 445 if (pledge("stdio rpath wpath cpath fattr flock", NULL) == -1) { 446 warn("pledge"); 447 exitcode = (int)MANDOCLEVEL_SYSERR; 448 goto out; 449 } 450 } 451#endif 452 use_all = 1; 453 for (i = 0; i < argc; i++) 454 filescan(argv[i]); 455 if (nodb == 0) 456 dbprune(dba); 457 } else { 458 /* Database missing or corrupt. */ 459 if (op != OP_UPDATE || errno != ENOENT) 460 say(MANDOC_DB, "%s: Automatically recreating" 461 " from scratch", strerror(errno)); 462 exitcode = (int)MANDOCLEVEL_OK; 463 op = OP_DEFAULT; 464 if (0 == treescan()) 465 goto out; 466 dba = dba_new(128); 467 } 468 if (OP_DELETE != op) 469 mpages_merge(dba, mp); 470 if (nodb == 0) 471 dbwrite(dba); 472 dba_free(dba); 473 } else { 474 /* 475 * If we have arguments, use them as our manpaths. 476 * If we don't, use man.conf(5). 477 */ 478 if (argc > 0) { 479 conf.manpath.paths = mandoc_reallocarray(NULL, 480 argc, sizeof(char *)); 481 conf.manpath.sz = (size_t)argc; 482 for (i = 0; i < argc; i++) 483 conf.manpath.paths[i] = mandoc_strdup(argv[i]); 484 } else 485 manconf_parse(&conf, path_arg, NULL, NULL); 486 487 if (conf.manpath.sz == 0) { 488 exitcode = (int)MANDOCLEVEL_BADARG; 489 say("", "Empty manpath"); 490 } 491 492 /* 493 * First scan the tree rooted at a base directory, then 494 * build a new database and finally move it into place. 495 * Ignore zero-length directories and strip trailing 496 * slashes. 497 */ 498 for (j = 0; j < conf.manpath.sz; j++) { 499 sz = strlen(conf.manpath.paths[j]); 500 if (sz && conf.manpath.paths[j][sz - 1] == '/') 501 conf.manpath.paths[j][--sz] = '\0'; 502 if (0 == sz) 503 continue; 504 505 if (j) { 506 mandoc_ohash_init(&mpages, 6, 507 offsetof(struct mpage, inodev)); 508 mandoc_ohash_init(&mlinks, 6, 509 offsetof(struct mlink, file)); 510 } 511 512 if ( ! set_basedir(conf.manpath.paths[j], argc > 0)) 513 continue; 514 if (0 == treescan()) 515 continue; 516 dba = dba_new(128); 517 mpages_merge(dba, mp); 518 if (nodb == 0) 519 dbwrite(dba); 520 dba_free(dba); 521 522 if (j + 1 < conf.manpath.sz) { 523 mpages_free(); 524 ohash_delete(&mpages); 525 ohash_delete(&mlinks); 526 } 527 } 528 } 529out: 530 manconf_free(&conf); 531 mparse_free(mp); 532 mchars_free(); 533 mpages_free(); 534 ohash_delete(&mpages); 535 ohash_delete(&mlinks); 536 return exitcode; 537usage: 538 progname = getprogname(); 539 fprintf(stderr, "usage: %s [-aDnpQ] [-C file] [-Tutf8]\n" 540 " %s [-aDnpQ] [-Tutf8] dir ...\n" 541 " %s [-DnpQ] [-Tutf8] -d dir [file ...]\n" 542 " %s [-Dnp] -u dir [file ...]\n" 543 " %s [-Q] -t file ...\n", 544 progname, progname, progname, progname, progname); 545 546 return (int)MANDOCLEVEL_BADARG; 547} 548 549/* 550 * To get a singly linked list in alpha order while inserting entries 551 * at the beginning, process directory entries in reverse alpha order. 552 */ 553static int 554#if HAVE_FTS_COMPARE_CONST 555fts_compare(const FTSENT *const *a, const FTSENT *const *b) 556#else 557fts_compare(const FTSENT **a, const FTSENT **b) 558#endif 559{ 560 return -strcmp((*a)->fts_name, (*b)->fts_name); 561} 562 563/* 564 * Scan a directory tree rooted at "basedir" for manpages. 565 * We use fts(), scanning directory parts along the way for clues to our 566 * section and architecture. 567 * 568 * If use_all has been specified, grok all files. 569 * If not, sanitise paths to the following: 570 * 571 * [./]man*[/<arch>]/<name>.<section> 572 * or 573 * [./]cat<section>[/<arch>]/<name>.0 574 * 575 * TODO: accommodate for multi-language directories. 576 */ 577static int 578treescan(void) 579{ 580 char buf[PATH_MAX]; 581 FTS *f; 582 FTSENT *ff; 583 struct mlink *mlink; 584 int gzip; 585 enum form dform; 586 char *dsec, *arch, *fsec, *cp; 587 const char *path; 588 const char *argv[2]; 589 590 argv[0] = "."; 591 argv[1] = NULL; 592 593 f = fts_open((char * const *)argv, FTS_PHYSICAL | FTS_NOCHDIR, 594 fts_compare); 595 if (f == NULL) { 596 exitcode = (int)MANDOCLEVEL_SYSERR; 597 say("", "&fts_open"); 598 return 0; 599 } 600 601 dsec = arch = NULL; 602 dform = FORM_NONE; 603 604 while ((ff = fts_read(f)) != NULL) { 605 path = ff->fts_path + 2; 606 switch (ff->fts_info) { 607 608 /* 609 * Symbolic links require various sanity checks, 610 * then get handled just like regular files. 611 */ 612 case FTS_SL: 613 if (realpath(path, buf) == NULL) { 614 if (warnings) 615 say(path, "&realpath"); 616 continue; 617 } 618 if (strstr(buf, basedir) != buf 619#ifdef HOMEBREWDIR 620 && strstr(buf, HOMEBREWDIR) != buf 621#endif 622 ) { 623 if (warnings) say("", 624 "%s: outside base directory", buf); 625 continue; 626 } 627 /* Use logical inode to avoid mpages dupe. */ 628 if (stat(path, ff->fts_statp) == -1) { 629 if (warnings) 630 say(path, "&stat"); 631 continue; 632 } 633 /* FALLTHROUGH */ 634 635 /* 636 * If we're a regular file, add an mlink by using the 637 * stored directory data and handling the filename. 638 */ 639 case FTS_F: 640 if ( ! strcmp(path, MANDOC_DB)) 641 continue; 642 if ( ! use_all && ff->fts_level < 2) { 643 if (warnings) 644 say(path, "Extraneous file"); 645 continue; 646 } 647 gzip = 0; 648 fsec = NULL; 649 while (fsec == NULL) { 650 fsec = strrchr(ff->fts_name, '.'); 651 if (fsec == NULL || strcmp(fsec+1, "gz")) 652 break; 653 gzip = 1; 654 *fsec = '\0'; 655 fsec = NULL; 656 } 657 if (fsec == NULL) { 658 if ( ! use_all) { 659 if (warnings) 660 say(path, 661 "No filename suffix"); 662 continue; 663 } 664 } else if ( ! strcmp(++fsec, "html")) { 665 if (warnings) 666 say(path, "Skip html"); 667 continue; 668 } else if ( ! strcmp(fsec, "ps")) { 669 if (warnings) 670 say(path, "Skip ps"); 671 continue; 672 } else if ( ! strcmp(fsec, "pdf")) { 673 if (warnings) 674 say(path, "Skip pdf"); 675 continue; 676 } else if ( ! use_all && 677 ((dform == FORM_SRC && 678 strncmp(fsec, dsec, strlen(dsec))) || 679 (dform == FORM_CAT && strcmp(fsec, "0")))) { 680 if (warnings) 681 say(path, "Wrong filename suffix"); 682 continue; 683 } else 684 fsec[-1] = '\0'; 685 686 mlink = mandoc_calloc(1, sizeof(struct mlink)); 687 if (strlcpy(mlink->file, path, 688 sizeof(mlink->file)) >= 689 sizeof(mlink->file)) { 690 say(path, "Filename too long"); 691 free(mlink); 692 continue; 693 } 694 mlink->dform = dform; 695 mlink->dsec = dsec; 696 mlink->arch = arch; 697 mlink->name = ff->fts_name; 698 mlink->fsec = fsec; 699 mlink->gzip = gzip; 700 mlink_add(mlink, ff->fts_statp); 701 continue; 702 703 case FTS_D: 704 case FTS_DP: 705 break; 706 707 default: 708 if (warnings) 709 say(path, "Not a regular file"); 710 continue; 711 } 712 713 switch (ff->fts_level) { 714 case 0: 715 /* Ignore the root directory. */ 716 break; 717 case 1: 718 /* 719 * This might contain manX/ or catX/. 720 * Try to infer this from the name. 721 * If we're not in use_all, enforce it. 722 */ 723 cp = ff->fts_name; 724 if (ff->fts_info == FTS_DP) { 725 dform = FORM_NONE; 726 dsec = NULL; 727 break; 728 } 729 730 if ( ! strncmp(cp, "man", 3)) { 731 dform = FORM_SRC; 732 dsec = cp + 3; 733 } else if ( ! strncmp(cp, "cat", 3)) { 734 dform = FORM_CAT; 735 dsec = cp + 3; 736 } else { 737 dform = FORM_NONE; 738 dsec = NULL; 739 } 740 741 if (dsec != NULL || use_all) 742 break; 743 744 if (warnings) 745 say(path, "Unknown directory part"); 746 fts_set(f, ff, FTS_SKIP); 747 break; 748 case 2: 749 /* 750 * Possibly our architecture. 751 * If we're descending, keep tabs on it. 752 */ 753 if (ff->fts_info != FTS_DP && dsec != NULL) 754 arch = ff->fts_name; 755 else 756 arch = NULL; 757 break; 758 default: 759 if (ff->fts_info == FTS_DP || use_all) 760 break; 761 if (warnings) 762 say(path, "Extraneous directory part"); 763 fts_set(f, ff, FTS_SKIP); 764 break; 765 } 766 } 767 768 fts_close(f); 769 return 1; 770} 771 772/* 773 * Add a file to the mlinks table. 774 * Do not verify that it's a "valid" looking manpage (we'll do that 775 * later). 776 * 777 * Try to infer the manual section, architecture, and page name from the 778 * path, assuming it looks like 779 * 780 * [./]man*[/<arch>]/<name>.<section> 781 * or 782 * [./]cat<section>[/<arch>]/<name>.0 783 * 784 * See treescan() for the fts(3) version of this. 785 */ 786static void 787filescan(const char *file) 788{ 789 char buf[PATH_MAX]; 790 struct stat st; 791 struct mlink *mlink; 792 char *p, *start; 793 794 assert(use_all); 795 796 if (0 == strncmp(file, "./", 2)) 797 file += 2; 798 799 /* 800 * We have to do lstat(2) before realpath(3) loses 801 * the information whether this is a symbolic link. 802 * We need to know that because for symbolic links, 803 * we want to use the orginal file name, while for 804 * regular files, we want to use the real path. 805 */ 806 if (-1 == lstat(file, &st)) { 807 exitcode = (int)MANDOCLEVEL_BADARG; 808 say(file, "&lstat"); 809 return; 810 } else if (0 == ((S_IFREG | S_IFLNK) & st.st_mode)) { 811 exitcode = (int)MANDOCLEVEL_BADARG; 812 say(file, "Not a regular file"); 813 return; 814 } 815 816 /* 817 * We have to resolve the file name to the real path 818 * in any case for the base directory check. 819 */ 820 if (NULL == realpath(file, buf)) { 821 exitcode = (int)MANDOCLEVEL_BADARG; 822 say(file, "&realpath"); 823 return; 824 } 825 826 if (OP_TEST == op) 827 start = buf; 828 else if (strstr(buf, basedir) == buf) 829 start = buf + strlen(basedir); 830#ifdef HOMEBREWDIR 831 else if (strstr(buf, HOMEBREWDIR) == buf) 832 start = buf; 833#endif 834 else { 835 exitcode = (int)MANDOCLEVEL_BADARG; 836 say("", "%s: outside base directory", buf); 837 return; 838 } 839 840 /* 841 * Now we are sure the file is inside our tree. 842 * If it is a symbolic link, ignore the real path 843 * and use the original name. 844 * This implies passing stuff like "cat1/../man1/foo.1" 845 * on the command line won't work. So don't do that. 846 * Note the stat(2) can still fail if the link target 847 * doesn't exist. 848 */ 849 if (S_IFLNK & st.st_mode) { 850 if (-1 == stat(buf, &st)) { 851 exitcode = (int)MANDOCLEVEL_BADARG; 852 say(file, "&stat"); 853 return; 854 } 855 if (strlcpy(buf, file, sizeof(buf)) >= sizeof(buf)) { 856 say(file, "Filename too long"); 857 return; 858 } 859 start = buf; 860 if (OP_TEST != op && strstr(buf, basedir) == buf) 861 start += strlen(basedir); 862 } 863 864 mlink = mandoc_calloc(1, sizeof(struct mlink)); 865 mlink->dform = FORM_NONE; 866 if (strlcpy(mlink->file, start, sizeof(mlink->file)) >= 867 sizeof(mlink->file)) { 868 say(start, "Filename too long"); 869 free(mlink); 870 return; 871 } 872 873 /* 874 * In test mode or when the original name is absolute 875 * but outside our tree, guess the base directory. 876 */ 877 878 if (op == OP_TEST || (start == buf && *start == '/')) { 879 if (strncmp(buf, "man/", 4) == 0) 880 start = buf + 4; 881 else if ((start = strstr(buf, "/man/")) != NULL) 882 start += 5; 883 else 884 start = buf; 885 } 886 887 /* 888 * First try to guess our directory structure. 889 * If we find a separator, try to look for man* or cat*. 890 * If we find one of these and what's underneath is a directory, 891 * assume it's an architecture. 892 */ 893 if (NULL != (p = strchr(start, '/'))) { 894 *p++ = '\0'; 895 if (0 == strncmp(start, "man", 3)) { 896 mlink->dform = FORM_SRC; 897 mlink->dsec = start + 3; 898 } else if (0 == strncmp(start, "cat", 3)) { 899 mlink->dform = FORM_CAT; 900 mlink->dsec = start + 3; 901 } 902 903 start = p; 904 if (NULL != mlink->dsec && NULL != (p = strchr(start, '/'))) { 905 *p++ = '\0'; 906 mlink->arch = start; 907 start = p; 908 } 909 } 910 911 /* 912 * Now check the file suffix. 913 * Suffix of `.0' indicates a catpage, `.1-9' is a manpage. 914 */ 915 p = strrchr(start, '\0'); 916 while (p-- > start && '/' != *p && '.' != *p) 917 /* Loop. */ ; 918 919 if ('.' == *p) { 920 *p++ = '\0'; 921 mlink->fsec = p; 922 } 923 924 /* 925 * Now try to parse the name. 926 * Use the filename portion of the path. 927 */ 928 mlink->name = start; 929 if (NULL != (p = strrchr(start, '/'))) { 930 mlink->name = p + 1; 931 *p = '\0'; 932 } 933 mlink_add(mlink, &st); 934} 935 936static void 937mlink_add(struct mlink *mlink, const struct stat *st) 938{ 939 struct inodev inodev; 940 struct mpage *mpage; 941 unsigned int slot; 942 943 assert(NULL != mlink->file); 944 945 mlink->dsec = mandoc_strdup(mlink->dsec ? mlink->dsec : ""); 946 mlink->arch = mandoc_strdup(mlink->arch ? mlink->arch : ""); 947 mlink->name = mandoc_strdup(mlink->name ? mlink->name : ""); 948 mlink->fsec = mandoc_strdup(mlink->fsec ? mlink->fsec : ""); 949 950 if ('0' == *mlink->fsec) { 951 free(mlink->fsec); 952 mlink->fsec = mandoc_strdup(mlink->dsec); 953 mlink->fform = FORM_CAT; 954 } else if ('1' <= *mlink->fsec && '9' >= *mlink->fsec) 955 mlink->fform = FORM_SRC; 956 else 957 mlink->fform = FORM_NONE; 958 959 slot = ohash_qlookup(&mlinks, mlink->file); 960 assert(NULL == ohash_find(&mlinks, slot)); 961 ohash_insert(&mlinks, slot, mlink); 962 963 memset(&inodev, 0, sizeof(inodev)); /* Clear padding. */ 964 inodev.st_ino = st->st_ino; 965 inodev.st_dev = st->st_dev; 966 slot = ohash_lookup_memory(&mpages, (char *)&inodev, 967 sizeof(struct inodev), inodev.st_ino); 968 mpage = ohash_find(&mpages, slot); 969 if (NULL == mpage) { 970 mpage = mandoc_calloc(1, sizeof(struct mpage)); 971 mpage->inodev.st_ino = inodev.st_ino; 972 mpage->inodev.st_dev = inodev.st_dev; 973 mpage->form = FORM_NONE; 974 mpage->next = mpage_head; 975 mpage_head = mpage; 976 ohash_insert(&mpages, slot, mpage); 977 } else 978 mlink->next = mpage->mlinks; 979 mpage->mlinks = mlink; 980 mlink->mpage = mpage; 981} 982 983static void 984mlink_free(struct mlink *mlink) 985{ 986 987 free(mlink->dsec); 988 free(mlink->arch); 989 free(mlink->name); 990 free(mlink->fsec); 991 free(mlink); 992} 993 994static void 995mpages_free(void) 996{ 997 struct mpage *mpage; 998 struct mlink *mlink; 999 1000 while ((mpage = mpage_head) != NULL) { 1001 while ((mlink = mpage->mlinks) != NULL) { 1002 mpage->mlinks = mlink->next; 1003 mlink_free(mlink); 1004 } 1005 mpage_head = mpage->next; 1006 free(mpage->sec); 1007 free(mpage->arch); 1008 free(mpage->title); 1009 free(mpage->desc); 1010 free(mpage); 1011 } 1012} 1013 1014/* 1015 * For each mlink to the mpage, check whether the path looks like 1016 * it is formatted, and if it does, check whether a source manual 1017 * exists by the same name, ignoring the suffix. 1018 * If both conditions hold, drop the mlink. 1019 */ 1020static void 1021mlinks_undupe(struct mpage *mpage) 1022{ 1023 char buf[PATH_MAX]; 1024 struct mlink **prev; 1025 struct mlink *mlink; 1026 char *bufp; 1027 1028 mpage->form = FORM_CAT; 1029 prev = &mpage->mlinks; 1030 while (NULL != (mlink = *prev)) { 1031 if (FORM_CAT != mlink->dform) { 1032 mpage->form = FORM_NONE; 1033 goto nextlink; 1034 } 1035 (void)strlcpy(buf, mlink->file, sizeof(buf)); 1036 bufp = strstr(buf, "cat"); 1037 assert(NULL != bufp); 1038 memcpy(bufp, "man", 3); 1039 if (NULL != (bufp = strrchr(buf, '.'))) 1040 *++bufp = '\0'; 1041 (void)strlcat(buf, mlink->dsec, sizeof(buf)); 1042 if (NULL == ohash_find(&mlinks, 1043 ohash_qlookup(&mlinks, buf))) 1044 goto nextlink; 1045 if (warnings) 1046 say(mlink->file, "Man source exists: %s", buf); 1047 if (use_all) 1048 goto nextlink; 1049 *prev = mlink->next; 1050 mlink_free(mlink); 1051 continue; 1052nextlink: 1053 prev = &(*prev)->next; 1054 } 1055} 1056 1057static void 1058mlink_check(struct mpage *mpage, struct mlink *mlink) 1059{ 1060 struct str *str; 1061 unsigned int slot; 1062 1063 /* 1064 * Check whether the manual section given in a file 1065 * agrees with the directory where the file is located. 1066 * Some manuals have suffixes like (3p) on their 1067 * section number either inside the file or in the 1068 * directory name, some are linked into more than one 1069 * section, like encrypt(1) = makekey(8). 1070 */ 1071 1072 if (FORM_SRC == mpage->form && 1073 strcasecmp(mpage->sec, mlink->dsec)) 1074 say(mlink->file, "Section \"%s\" manual in %s directory", 1075 mpage->sec, mlink->dsec); 1076 1077 /* 1078 * Manual page directories exist for each kernel 1079 * architecture as returned by machine(1). 1080 * However, many manuals only depend on the 1081 * application architecture as returned by arch(1). 1082 * For example, some (2/ARM) manuals are shared 1083 * across the "armish" and "zaurus" kernel 1084 * architectures. 1085 * A few manuals are even shared across completely 1086 * different architectures, for example fdformat(1) 1087 * on amd64, i386, and sparc64. 1088 */ 1089 1090 if (strcasecmp(mpage->arch, mlink->arch)) 1091 say(mlink->file, "Architecture \"%s\" manual in " 1092 "\"%s\" directory", mpage->arch, mlink->arch); 1093 1094 /* 1095 * XXX 1096 * parse_cat() doesn't set NAME_TITLE yet. 1097 */ 1098 1099 if (FORM_CAT == mpage->form) 1100 return; 1101 1102 /* 1103 * Check whether this mlink 1104 * appears as a name in the NAME section. 1105 */ 1106 1107 slot = ohash_qlookup(&names, mlink->name); 1108 str = ohash_find(&names, slot); 1109 assert(NULL != str); 1110 if ( ! (NAME_TITLE & str->mask)) 1111 say(mlink->file, "Name missing in NAME section"); 1112} 1113 1114/* 1115 * Run through the files in the global vector "mpages" 1116 * and add them to the database specified in "basedir". 1117 * 1118 * This handles the parsing scheme itself, using the cues of directory 1119 * and filename to determine whether the file is parsable or not. 1120 */ 1121static void 1122mpages_merge(struct dba *dba, struct mparse *mp) 1123{ 1124 struct mpage *mpage, *mpage_dest; 1125 struct mlink *mlink, *mlink_dest; 1126 struct roff_man *man; 1127 char *sodest; 1128 char *cp; 1129 int fd; 1130 1131 for (mpage = mpage_head; mpage != NULL; mpage = mpage->next) { 1132 mlinks_undupe(mpage); 1133 if ((mlink = mpage->mlinks) == NULL) 1134 continue; 1135 1136 name_mask = NAME_MASK; 1137 mandoc_ohash_init(&names, 4, offsetof(struct str, key)); 1138 mandoc_ohash_init(&strings, 6, offsetof(struct str, key)); 1139 mparse_reset(mp); 1140 man = NULL; 1141 sodest = NULL; 1142 1143 if ((fd = mparse_open(mp, mlink->file)) == -1) { 1144 say(mlink->file, "&open"); 1145 goto nextpage; 1146 } 1147 1148 /* 1149 * Interpret the file as mdoc(7) or man(7) source 1150 * code, unless it is known to be formatted. 1151 */ 1152 if (mlink->dform != FORM_CAT || mlink->fform != FORM_CAT) { 1153 mparse_readfd(mp, fd, mlink->file); 1154 close(fd); 1155 fd = -1; 1156 mparse_result(mp, &man, &sodest); 1157 } 1158 1159 if (sodest != NULL) { 1160 mlink_dest = ohash_find(&mlinks, 1161 ohash_qlookup(&mlinks, sodest)); 1162 if (mlink_dest == NULL) { 1163 mandoc_asprintf(&cp, "%s.gz", sodest); 1164 mlink_dest = ohash_find(&mlinks, 1165 ohash_qlookup(&mlinks, cp)); 1166 free(cp); 1167 } 1168 if (mlink_dest != NULL) { 1169 1170 /* The .so target exists. */ 1171 1172 mpage_dest = mlink_dest->mpage; 1173 while (1) { 1174 mlink->mpage = mpage_dest; 1175 1176 /* 1177 * If the target was already 1178 * processed, add the links 1179 * to the database now. 1180 * Otherwise, this will 1181 * happen when we come 1182 * to the target. 1183 */ 1184 1185 if (mpage_dest->dba != NULL) 1186 dbadd_mlink(mlink); 1187 1188 if (mlink->next == NULL) 1189 break; 1190 mlink = mlink->next; 1191 } 1192 1193 /* Move all links to the target. */ 1194 1195 mlink->next = mlink_dest->next; 1196 mlink_dest->next = mpage->mlinks; 1197 mpage->mlinks = NULL; 1198 } 1199 goto nextpage; 1200 } else if (man != NULL && man->macroset == MACROSET_MDOC) { 1201 mdoc_validate(man); 1202 mpage->form = FORM_SRC; 1203 mpage->sec = man->meta.msec; 1204 mpage->sec = mandoc_strdup( 1205 mpage->sec == NULL ? "" : mpage->sec); 1206 mpage->arch = man->meta.arch; 1207 mpage->arch = mandoc_strdup( 1208 mpage->arch == NULL ? "" : mpage->arch); 1209 mpage->title = mandoc_strdup(man->meta.title); 1210 } else if (man != NULL && man->macroset == MACROSET_MAN) { 1211 man_validate(man); 1212 if (*man->meta.msec != '\0' || 1213 *man->meta.title != '\0') { 1214 mpage->form = FORM_SRC; 1215 mpage->sec = mandoc_strdup(man->meta.msec); 1216 mpage->arch = mandoc_strdup(mlink->arch); 1217 mpage->title = mandoc_strdup(man->meta.title); 1218 } else 1219 man = NULL; 1220 } 1221 1222 assert(mpage->desc == NULL); 1223 if (man == NULL) { 1224 mpage->form = FORM_CAT; 1225 mpage->sec = mandoc_strdup(mlink->dsec); 1226 mpage->arch = mandoc_strdup(mlink->arch); 1227 mpage->title = mandoc_strdup(mlink->name); 1228 parse_cat(mpage, fd); 1229 } else if (man->macroset == MACROSET_MDOC) 1230 parse_mdoc(mpage, &man->meta, man->first); 1231 else 1232 parse_man(mpage, &man->meta, man->first); 1233 if (mpage->desc == NULL) { 1234 mpage->desc = mandoc_strdup(mlink->name); 1235 if (warnings) 1236 say(mlink->file, "No one-line description, " 1237 "using filename \"%s\"", mlink->name); 1238 } 1239 1240 for (mlink = mpage->mlinks; 1241 mlink != NULL; 1242 mlink = mlink->next) { 1243 putkey(mpage, mlink->name, NAME_FILE); 1244 if (warnings && !use_all) 1245 mlink_check(mpage, mlink); 1246 } 1247 1248 dbadd(dba, mpage); 1249 1250nextpage: 1251 ohash_delete(&strings); 1252 ohash_delete(&names); 1253 } 1254} 1255 1256static void 1257parse_cat(struct mpage *mpage, int fd) 1258{ 1259 FILE *stream; 1260 struct mlink *mlink; 1261 char *line, *p, *title, *sec; 1262 size_t linesz, plen, titlesz; 1263 ssize_t len; 1264 int offs; 1265 1266 mlink = mpage->mlinks; 1267 stream = fd == -1 ? fopen(mlink->file, "r") : fdopen(fd, "r"); 1268 if (stream == NULL) { 1269 if (fd != -1) 1270 close(fd); 1271 if (warnings) 1272 say(mlink->file, "&fopen"); 1273 return; 1274 } 1275 1276 line = NULL; 1277 linesz = 0; 1278 1279 /* Parse the section number from the header line. */ 1280 1281 while (getline(&line, &linesz, stream) != -1) { 1282 if (*line == '\n') 1283 continue; 1284 if ((sec = strchr(line, '(')) == NULL) 1285 break; 1286 if ((p = strchr(++sec, ')')) == NULL) 1287 break; 1288 free(mpage->sec); 1289 mpage->sec = mandoc_strndup(sec, p - sec); 1290 if (warnings && *mlink->dsec != '\0' && 1291 strcasecmp(mpage->sec, mlink->dsec)) 1292 say(mlink->file, 1293 "Section \"%s\" manual in %s directory", 1294 mpage->sec, mlink->dsec); 1295 break; 1296 } 1297 1298 /* Skip to first blank line. */ 1299 1300 while (line == NULL || *line != '\n') 1301 if (getline(&line, &linesz, stream) == -1) 1302 break; 1303 1304 /* 1305 * Assume the first line that is not indented 1306 * is the first section header. Skip to it. 1307 */ 1308 1309 while (getline(&line, &linesz, stream) != -1) 1310 if (*line != '\n' && *line != ' ') 1311 break; 1312 1313 /* 1314 * Read up until the next section into a buffer. 1315 * Strip the leading and trailing newline from each read line, 1316 * appending a trailing space. 1317 * Ignore empty (whitespace-only) lines. 1318 */ 1319 1320 titlesz = 0; 1321 title = NULL; 1322 1323 while ((len = getline(&line, &linesz, stream)) != -1) { 1324 if (*line != ' ') 1325 break; 1326 offs = 0; 1327 while (isspace((unsigned char)line[offs])) 1328 offs++; 1329 if (line[offs] == '\0') 1330 continue; 1331 title = mandoc_realloc(title, titlesz + len - offs); 1332 memcpy(title + titlesz, line + offs, len - offs); 1333 titlesz += len - offs; 1334 title[titlesz - 1] = ' '; 1335 } 1336 free(line); 1337 1338 /* 1339 * If no page content can be found, or the input line 1340 * is already the next section header, or there is no 1341 * trailing newline, reuse the page title as the page 1342 * description. 1343 */ 1344 1345 if (NULL == title || '\0' == *title) { 1346 if (warnings) 1347 say(mlink->file, "Cannot find NAME section"); 1348 fclose(stream); 1349 free(title); 1350 return; 1351 } 1352 1353 title[titlesz - 1] = '\0'; 1354 1355 /* 1356 * Skip to the first dash. 1357 * Use the remaining line as the description (no more than 70 1358 * bytes). 1359 */ 1360 1361 if (NULL != (p = strstr(title, "- "))) { 1362 for (p += 2; ' ' == *p || '\b' == *p; p++) 1363 /* Skip to next word. */ ; 1364 } else { 1365 if (warnings) 1366 say(mlink->file, "No dash in title line, " 1367 "reusing \"%s\" as one-line description", title); 1368 p = title; 1369 } 1370 1371 plen = strlen(p); 1372 1373 /* Strip backspace-encoding from line. */ 1374 1375 while (NULL != (line = memchr(p, '\b', plen))) { 1376 len = line - p; 1377 if (0 == len) { 1378 memmove(line, line + 1, plen--); 1379 continue; 1380 } 1381 memmove(line - 1, line + 1, plen - len); 1382 plen -= 2; 1383 } 1384 1385 mpage->desc = mandoc_strdup(p); 1386 fclose(stream); 1387 free(title); 1388} 1389 1390/* 1391 * Put a type/word pair into the word database for this particular file. 1392 */ 1393static void 1394putkey(const struct mpage *mpage, char *value, uint64_t type) 1395{ 1396 putkeys(mpage, value, strlen(value), type); 1397} 1398 1399/* 1400 * Grok all nodes at or below a certain mdoc node into putkey(). 1401 */ 1402static void 1403putmdockey(const struct mpage *mpage, 1404 const struct roff_node *n, uint64_t m, int taboo) 1405{ 1406 1407 for ( ; NULL != n; n = n->next) { 1408 if (n->flags & taboo) 1409 continue; 1410 if (NULL != n->child) 1411 putmdockey(mpage, n->child, m, taboo); 1412 if (n->type == ROFFT_TEXT) 1413 putkey(mpage, n->string, m); 1414 } 1415} 1416 1417static void 1418parse_man(struct mpage *mpage, const struct roff_meta *meta, 1419 const struct roff_node *n) 1420{ 1421 const struct roff_node *head, *body; 1422 char *start, *title; 1423 char byte; 1424 size_t sz; 1425 1426 if (n == NULL) 1427 return; 1428 1429 /* 1430 * We're only searching for one thing: the first text child in 1431 * the BODY of a NAME section. Since we don't keep track of 1432 * sections in -man, run some hoops to find out whether we're in 1433 * the correct section or not. 1434 */ 1435 1436 if (n->type == ROFFT_BODY && n->tok == MAN_SH) { 1437 body = n; 1438 if ((head = body->parent->head) != NULL && 1439 (head = head->child) != NULL && 1440 head->next == NULL && 1441 head->type == ROFFT_TEXT && 1442 strcmp(head->string, "NAME") == 0 && 1443 body->child != NULL) { 1444 1445 /* 1446 * Suck the entire NAME section into memory. 1447 * Yes, we might run away. 1448 * But too many manuals have big, spread-out 1449 * NAME sections over many lines. 1450 */ 1451 1452 title = NULL; 1453 deroff(&title, body); 1454 if (NULL == title) 1455 return; 1456 1457 /* 1458 * Go through a special heuristic dance here. 1459 * Conventionally, one or more manual names are 1460 * comma-specified prior to a whitespace, then a 1461 * dash, then a description. Try to puzzle out 1462 * the name parts here. 1463 */ 1464 1465 start = title; 1466 for ( ;; ) { 1467 sz = strcspn(start, " ,"); 1468 if ('\0' == start[sz]) 1469 break; 1470 1471 byte = start[sz]; 1472 start[sz] = '\0'; 1473 1474 /* 1475 * Assume a stray trailing comma in the 1476 * name list if a name begins with a dash. 1477 */ 1478 1479 if ('-' == start[0] || 1480 ('\\' == start[0] && '-' == start[1])) 1481 break; 1482 1483 putkey(mpage, start, NAME_TITLE); 1484 if ( ! (mpage->name_head_done || 1485 strcasecmp(start, meta->title))) { 1486 putkey(mpage, start, NAME_HEAD); 1487 mpage->name_head_done = 1; 1488 } 1489 1490 if (' ' == byte) { 1491 start += sz + 1; 1492 break; 1493 } 1494 1495 assert(',' == byte); 1496 start += sz + 1; 1497 while (' ' == *start) 1498 start++; 1499 } 1500 1501 if (start == title) { 1502 putkey(mpage, start, NAME_TITLE); 1503 if ( ! (mpage->name_head_done || 1504 strcasecmp(start, meta->title))) { 1505 putkey(mpage, start, NAME_HEAD); 1506 mpage->name_head_done = 1; 1507 } 1508 free(title); 1509 return; 1510 } 1511 1512 while (isspace((unsigned char)*start)) 1513 start++; 1514 1515 if (0 == strncmp(start, "-", 1)) 1516 start += 1; 1517 else if (0 == strncmp(start, "\\-\\-", 4)) 1518 start += 4; 1519 else if (0 == strncmp(start, "\\-", 2)) 1520 start += 2; 1521 else if (0 == strncmp(start, "\\(en", 4)) 1522 start += 4; 1523 else if (0 == strncmp(start, "\\(em", 4)) 1524 start += 4; 1525 1526 while (' ' == *start) 1527 start++; 1528 1529 mpage->desc = mandoc_strdup(start); 1530 free(title); 1531 return; 1532 } 1533 } 1534 1535 for (n = n->child; n; n = n->next) { 1536 if (NULL != mpage->desc) 1537 break; 1538 parse_man(mpage, meta, n); 1539 } 1540} 1541 1542static void 1543parse_mdoc(struct mpage *mpage, const struct roff_meta *meta, 1544 const struct roff_node *n) 1545{ 1546 1547 for (n = n->child; n != NULL; n = n->next) { 1548 if (n->tok == TOKEN_NONE || 1549 n->tok < ROFF_MAX || 1550 n->flags & mdocs[n->tok].taboo) 1551 continue; 1552 assert(n->tok >= MDOC_Dd && n->tok < MDOC_MAX); 1553 switch (n->type) { 1554 case ROFFT_ELEM: 1555 case ROFFT_BLOCK: 1556 case ROFFT_HEAD: 1557 case ROFFT_BODY: 1558 case ROFFT_TAIL: 1559 if (mdocs[n->tok].fp != NULL && 1560 (*mdocs[n->tok].fp)(mpage, meta, n) == 0) 1561 break; 1562 if (mdocs[n->tok].mask) 1563 putmdockey(mpage, n->child, 1564 mdocs[n->tok].mask, mdocs[n->tok].taboo); 1565 break; 1566 default: 1567 continue; 1568 } 1569 if (NULL != n->child) 1570 parse_mdoc(mpage, meta, n); 1571 } 1572} 1573 1574static int 1575parse_mdoc_Fd(struct mpage *mpage, const struct roff_meta *meta, 1576 const struct roff_node *n) 1577{ 1578 char *start, *end; 1579 size_t sz; 1580 1581 if (SEC_SYNOPSIS != n->sec || 1582 NULL == (n = n->child) || 1583 n->type != ROFFT_TEXT) 1584 return 0; 1585 1586 /* 1587 * Only consider those `Fd' macro fields that begin with an 1588 * "inclusion" token (versus, e.g., #define). 1589 */ 1590 1591 if (strcmp("#include", n->string)) 1592 return 0; 1593 1594 if ((n = n->next) == NULL || n->type != ROFFT_TEXT) 1595 return 0; 1596 1597 /* 1598 * Strip away the enclosing angle brackets and make sure we're 1599 * not zero-length. 1600 */ 1601 1602 start = n->string; 1603 if ('<' == *start || '"' == *start) 1604 start++; 1605 1606 if (0 == (sz = strlen(start))) 1607 return 0; 1608 1609 end = &start[(int)sz - 1]; 1610 if ('>' == *end || '"' == *end) 1611 end--; 1612 1613 if (end > start) 1614 putkeys(mpage, start, end - start + 1, TYPE_In); 1615 return 0; 1616} 1617 1618static void 1619parse_mdoc_fname(struct mpage *mpage, const struct roff_node *n) 1620{ 1621 char *cp; 1622 size_t sz; 1623 1624 if (n->type != ROFFT_TEXT) 1625 return; 1626 1627 /* Skip function pointer punctuation. */ 1628 1629 cp = n->string; 1630 while (*cp == '(' || *cp == '*') 1631 cp++; 1632 sz = strcspn(cp, "()"); 1633 1634 putkeys(mpage, cp, sz, TYPE_Fn); 1635 if (n->sec == SEC_SYNOPSIS) 1636 putkeys(mpage, cp, sz, NAME_SYN); 1637} 1638 1639static int 1640parse_mdoc_Fn(struct mpage *mpage, const struct roff_meta *meta, 1641 const struct roff_node *n) 1642{ 1643 1644 if (n->child == NULL) 1645 return 0; 1646 1647 parse_mdoc_fname(mpage, n->child); 1648 1649 for (n = n->child->next; n != NULL; n = n->next) 1650 if (n->type == ROFFT_TEXT) 1651 putkey(mpage, n->string, TYPE_Fa); 1652 1653 return 0; 1654} 1655 1656static int 1657parse_mdoc_Fo(struct mpage *mpage, const struct roff_meta *meta, 1658 const struct roff_node *n) 1659{ 1660 1661 if (n->type != ROFFT_HEAD) 1662 return 1; 1663 1664 if (n->child != NULL) 1665 parse_mdoc_fname(mpage, n->child); 1666 1667 return 0; 1668} 1669 1670static int 1671parse_mdoc_Va(struct mpage *mpage, const struct roff_meta *meta, 1672 const struct roff_node *n) 1673{ 1674 char *cp; 1675 1676 if (n->type != ROFFT_ELEM && n->type != ROFFT_BODY) 1677 return 0; 1678 1679 if (n->child != NULL && 1680 n->child->next == NULL && 1681 n->child->type == ROFFT_TEXT) 1682 return 1; 1683 1684 cp = NULL; 1685 deroff(&cp, n); 1686 if (cp != NULL) { 1687 putkey(mpage, cp, TYPE_Vt | (n->tok == MDOC_Va || 1688 n->type == ROFFT_BODY ? TYPE_Va : 0)); 1689 free(cp); 1690 } 1691 1692 return 0; 1693} 1694 1695static int 1696parse_mdoc_Xr(struct mpage *mpage, const struct roff_meta *meta, 1697 const struct roff_node *n) 1698{ 1699 char *cp; 1700 1701 if (NULL == (n = n->child)) 1702 return 0; 1703 1704 if (NULL == n->next) { 1705 putkey(mpage, n->string, TYPE_Xr); 1706 return 0; 1707 } 1708 1709 mandoc_asprintf(&cp, "%s(%s)", n->string, n->next->string); 1710 putkey(mpage, cp, TYPE_Xr); 1711 free(cp); 1712 return 0; 1713} 1714 1715static int 1716parse_mdoc_Nd(struct mpage *mpage, const struct roff_meta *meta, 1717 const struct roff_node *n) 1718{ 1719 1720 if (n->type == ROFFT_BODY) 1721 deroff(&mpage->desc, n); 1722 return 0; 1723} 1724 1725static int 1726parse_mdoc_Nm(struct mpage *mpage, const struct roff_meta *meta, 1727 const struct roff_node *n) 1728{ 1729 1730 if (SEC_NAME == n->sec) 1731 putmdockey(mpage, n->child, NAME_TITLE, 0); 1732 else if (n->sec == SEC_SYNOPSIS && n->type == ROFFT_HEAD) { 1733 if (n->child == NULL) 1734 putkey(mpage, meta->name, NAME_SYN); 1735 else 1736 putmdockey(mpage, n->child, NAME_SYN, 0); 1737 } 1738 if ( ! (mpage->name_head_done || 1739 n->child == NULL || n->child->string == NULL || 1740 strcasecmp(n->child->string, meta->title))) { 1741 putkey(mpage, n->child->string, NAME_HEAD); 1742 mpage->name_head_done = 1; 1743 } 1744 return 0; 1745} 1746 1747static int 1748parse_mdoc_Sh(struct mpage *mpage, const struct roff_meta *meta, 1749 const struct roff_node *n) 1750{ 1751 1752 return n->sec == SEC_CUSTOM && n->type == ROFFT_HEAD; 1753} 1754 1755static int 1756parse_mdoc_head(struct mpage *mpage, const struct roff_meta *meta, 1757 const struct roff_node *n) 1758{ 1759 1760 return n->type == ROFFT_HEAD; 1761} 1762 1763/* 1764 * Add a string to the hash table for the current manual. 1765 * Each string has a bitmask telling which macros it belongs to. 1766 * When we finish the manual, we'll dump the table. 1767 */ 1768static void 1769putkeys(const struct mpage *mpage, char *cp, size_t sz, uint64_t v) 1770{ 1771 struct ohash *htab; 1772 struct str *s; 1773 const char *end; 1774 unsigned int slot; 1775 int i, mustfree; 1776 1777 if (0 == sz) 1778 return; 1779 1780 mustfree = render_string(&cp, &sz); 1781 1782 if (TYPE_Nm & v) { 1783 htab = &names; 1784 v &= name_mask; 1785 if (v & NAME_FIRST) 1786 name_mask &= ~NAME_FIRST; 1787 if (debug > 1) 1788 say(mpage->mlinks->file, 1789 "Adding name %*s, bits=0x%llx", (int)sz, cp, 1790 (unsigned long long)v); 1791 } else { 1792 htab = &strings; 1793 if (debug > 1) 1794 for (i = 0; i < KEY_MAX; i++) 1795 if ((uint64_t)1 << i & v) 1796 say(mpage->mlinks->file, 1797 "Adding key %s=%*s", 1798 mansearch_keynames[i], (int)sz, cp); 1799 } 1800 1801 end = cp + sz; 1802 slot = ohash_qlookupi(htab, cp, &end); 1803 s = ohash_find(htab, slot); 1804 1805 if (NULL != s && mpage == s->mpage) { 1806 s->mask |= v; 1807 return; 1808 } else if (NULL == s) { 1809 s = mandoc_calloc(1, sizeof(struct str) + sz + 1); 1810 memcpy(s->key, cp, sz); 1811 ohash_insert(htab, slot, s); 1812 } 1813 s->mpage = mpage; 1814 s->mask = v; 1815 1816 if (mustfree) 1817 free(cp); 1818} 1819 1820/* 1821 * Take a Unicode codepoint and produce its UTF-8 encoding. 1822 * This isn't the best way to do this, but it works. 1823 * The magic numbers are from the UTF-8 packaging. 1824 * They're not as scary as they seem: read the UTF-8 spec for details. 1825 */ 1826static size_t 1827utf8(unsigned int cp, char out[7]) 1828{ 1829 size_t rc; 1830 1831 rc = 0; 1832 if (cp <= 0x0000007F) { 1833 rc = 1; 1834 out[0] = (char)cp; 1835 } else if (cp <= 0x000007FF) { 1836 rc = 2; 1837 out[0] = (cp >> 6 & 31) | 192; 1838 out[1] = (cp & 63) | 128; 1839 } else if (cp <= 0x0000FFFF) { 1840 rc = 3; 1841 out[0] = (cp >> 12 & 15) | 224; 1842 out[1] = (cp >> 6 & 63) | 128; 1843 out[2] = (cp & 63) | 128; 1844 } else if (cp <= 0x001FFFFF) { 1845 rc = 4; 1846 out[0] = (cp >> 18 & 7) | 240; 1847 out[1] = (cp >> 12 & 63) | 128; 1848 out[2] = (cp >> 6 & 63) | 128; 1849 out[3] = (cp & 63) | 128; 1850 } else if (cp <= 0x03FFFFFF) { 1851 rc = 5; 1852 out[0] = (cp >> 24 & 3) | 248; 1853 out[1] = (cp >> 18 & 63) | 128; 1854 out[2] = (cp >> 12 & 63) | 128; 1855 out[3] = (cp >> 6 & 63) | 128; 1856 out[4] = (cp & 63) | 128; 1857 } else if (cp <= 0x7FFFFFFF) { 1858 rc = 6; 1859 out[0] = (cp >> 30 & 1) | 252; 1860 out[1] = (cp >> 24 & 63) | 128; 1861 out[2] = (cp >> 18 & 63) | 128; 1862 out[3] = (cp >> 12 & 63) | 128; 1863 out[4] = (cp >> 6 & 63) | 128; 1864 out[5] = (cp & 63) | 128; 1865 } else 1866 return 0; 1867 1868 out[rc] = '\0'; 1869 return rc; 1870} 1871 1872/* 1873 * If the string contains escape sequences, 1874 * replace it with an allocated rendering and return 1, 1875 * such that the caller can free it after use. 1876 * Otherwise, do nothing and return 0. 1877 */ 1878static int 1879render_string(char **public, size_t *psz) 1880{ 1881 const char *src, *scp, *addcp, *seq; 1882 char *dst; 1883 size_t ssz, dsz, addsz; 1884 char utfbuf[7], res[6]; 1885 int seqlen, unicode; 1886 1887 res[0] = '\\'; 1888 res[1] = '\t'; 1889 res[2] = ASCII_NBRSP; 1890 res[3] = ASCII_HYPH; 1891 res[4] = ASCII_BREAK; 1892 res[5] = '\0'; 1893 1894 src = scp = *public; 1895 ssz = *psz; 1896 dst = NULL; 1897 dsz = 0; 1898 1899 while (scp < src + *psz) { 1900 1901 /* Leave normal characters unchanged. */ 1902 1903 if (strchr(res, *scp) == NULL) { 1904 if (dst != NULL) 1905 dst[dsz++] = *scp; 1906 scp++; 1907 continue; 1908 } 1909 1910 /* 1911 * Found something that requires replacing, 1912 * make sure we have a destination buffer. 1913 */ 1914 1915 if (dst == NULL) { 1916 dst = mandoc_malloc(ssz + 1); 1917 dsz = scp - src; 1918 memcpy(dst, src, dsz); 1919 } 1920 1921 /* Handle single-char special characters. */ 1922 1923 switch (*scp) { 1924 case '\\': 1925 break; 1926 case '\t': 1927 case ASCII_NBRSP: 1928 dst[dsz++] = ' '; 1929 scp++; 1930 continue; 1931 case ASCII_HYPH: 1932 dst[dsz++] = '-'; 1933 /* FALLTHROUGH */ 1934 case ASCII_BREAK: 1935 scp++; 1936 continue; 1937 default: 1938 abort(); 1939 } 1940 1941 /* 1942 * Found an escape sequence. 1943 * Read past the slash, then parse it. 1944 * Ignore everything except characters. 1945 */ 1946 1947 scp++; 1948 if (mandoc_escape(&scp, &seq, &seqlen) != ESCAPE_SPECIAL) 1949 continue; 1950 1951 /* 1952 * Render the special character 1953 * as either UTF-8 or ASCII. 1954 */ 1955 1956 if (write_utf8) { 1957 unicode = mchars_spec2cp(seq, seqlen); 1958 if (unicode <= 0) 1959 continue; 1960 addsz = utf8(unicode, utfbuf); 1961 if (addsz == 0) 1962 continue; 1963 addcp = utfbuf; 1964 } else { 1965 addcp = mchars_spec2str(seq, seqlen, &addsz); 1966 if (addcp == NULL) 1967 continue; 1968 if (*addcp == ASCII_NBRSP) { 1969 addcp = " "; 1970 addsz = 1; 1971 } 1972 } 1973 1974 /* Copy the rendered glyph into the stream. */ 1975 1976 ssz += addsz; 1977 dst = mandoc_realloc(dst, ssz + 1); 1978 memcpy(dst + dsz, addcp, addsz); 1979 dsz += addsz; 1980 } 1981 if (dst != NULL) { 1982 *public = dst; 1983 *psz = dsz; 1984 } 1985 1986 /* Trim trailing whitespace and NUL-terminate. */ 1987 1988 while (*psz > 0 && (*public)[*psz - 1] == ' ') 1989 --*psz; 1990 if (dst != NULL) { 1991 (*public)[*psz] = '\0'; 1992 return 1; 1993 } else 1994 return 0; 1995} 1996 1997static void 1998dbadd_mlink(const struct mlink *mlink) 1999{ 2000 dba_page_alias(mlink->mpage->dba, mlink->name, NAME_FILE); 2001 dba_page_add(mlink->mpage->dba, DBP_SECT, mlink->dsec); 2002 dba_page_add(mlink->mpage->dba, DBP_SECT, mlink->fsec); 2003 dba_page_add(mlink->mpage->dba, DBP_ARCH, mlink->arch); 2004 dba_page_add(mlink->mpage->dba, DBP_FILE, mlink->file); 2005} 2006 2007/* 2008 * Flush the current page's terms (and their bits) into the database. 2009 * Also, handle escape sequences at the last possible moment. 2010 */ 2011static void 2012dbadd(struct dba *dba, struct mpage *mpage) 2013{ 2014 struct mlink *mlink; 2015 struct str *key; 2016 char *cp; 2017 uint64_t mask; 2018 size_t i; 2019 unsigned int slot; 2020 int mustfree; 2021 2022 mlink = mpage->mlinks; 2023 2024 if (nodb) { 2025 for (key = ohash_first(&names, &slot); NULL != key; 2026 key = ohash_next(&names, &slot)) 2027 free(key); 2028 for (key = ohash_first(&strings, &slot); NULL != key; 2029 key = ohash_next(&strings, &slot)) 2030 free(key); 2031 if (0 == debug) 2032 return; 2033 while (NULL != mlink) { 2034 fputs(mlink->name, stdout); 2035 if (NULL == mlink->next || 2036 strcmp(mlink->dsec, mlink->next->dsec) || 2037 strcmp(mlink->fsec, mlink->next->fsec) || 2038 strcmp(mlink->arch, mlink->next->arch)) { 2039 putchar('('); 2040 if ('\0' == *mlink->dsec) 2041 fputs(mlink->fsec, stdout); 2042 else 2043 fputs(mlink->dsec, stdout); 2044 if ('\0' != *mlink->arch) 2045 printf("/%s", mlink->arch); 2046 putchar(')'); 2047 } 2048 mlink = mlink->next; 2049 if (NULL != mlink) 2050 fputs(", ", stdout); 2051 } 2052 printf(" - %s\n", mpage->desc); 2053 return; 2054 } 2055 2056 if (debug) 2057 say(mlink->file, "Adding to database"); 2058 2059 cp = mpage->desc; 2060 i = strlen(cp); 2061 mustfree = render_string(&cp, &i); 2062 mpage->dba = dba_page_new(dba->pages, 2063 *mpage->arch == '\0' ? mlink->arch : mpage->arch, 2064 cp, mlink->file, mpage->form); 2065 if (mustfree) 2066 free(cp); 2067 dba_page_add(mpage->dba, DBP_SECT, mpage->sec); 2068 2069 while (mlink != NULL) { 2070 dbadd_mlink(mlink); 2071 mlink = mlink->next; 2072 } 2073 2074 for (key = ohash_first(&names, &slot); NULL != key; 2075 key = ohash_next(&names, &slot)) { 2076 assert(key->mpage == mpage); 2077 dba_page_alias(mpage->dba, key->key, key->mask); 2078 free(key); 2079 } 2080 for (key = ohash_first(&strings, &slot); NULL != key; 2081 key = ohash_next(&strings, &slot)) { 2082 assert(key->mpage == mpage); 2083 i = 0; 2084 for (mask = TYPE_Xr; mask <= TYPE_Lb; mask *= 2) { 2085 if (key->mask & mask) 2086 dba_macro_add(dba->macros, i, 2087 key->key, mpage->dba); 2088 i++; 2089 } 2090 free(key); 2091 } 2092} 2093 2094static void 2095dbprune(struct dba *dba) 2096{ 2097 struct dba_array *page, *files; 2098 char *file; 2099 2100 dba_array_FOREACH(dba->pages, page) { 2101 files = dba_array_get(page, DBP_FILE); 2102 dba_array_FOREACH(files, file) { 2103 if (*file < ' ') 2104 file++; 2105 if (ohash_find(&mlinks, ohash_qlookup(&mlinks, 2106 file)) != NULL) { 2107 if (debug) 2108 say(file, "Deleting from database"); 2109 dba_array_del(dba->pages); 2110 break; 2111 } 2112 } 2113 } 2114} 2115 2116/* 2117 * Write the database from memory to disk. 2118 */ 2119static void 2120dbwrite(struct dba *dba) 2121{ 2122 char tfn[32]; 2123 int status; 2124 pid_t child; 2125 2126 /* 2127 * Do not write empty databases, and delete existing ones 2128 * when makewhatis -u causes them to become empty. 2129 */ 2130 2131 dba_array_start(dba->pages); 2132 if (dba_array_next(dba->pages) == NULL) { 2133 if (unlink(MANDOC_DB) == -1 && errno != ENOENT) 2134 say(MANDOC_DB, "&unlink"); 2135 return; 2136 } 2137 2138 /* 2139 * Build the database in a temporary file, 2140 * then atomically move it into place. 2141 */ 2142 2143 if (dba_write(MANDOC_DB "~", dba) != -1) { 2144 if (rename(MANDOC_DB "~", MANDOC_DB) == -1) { 2145 exitcode = (int)MANDOCLEVEL_SYSERR; 2146 say(MANDOC_DB, "&rename"); 2147 unlink(MANDOC_DB "~"); 2148 } 2149 return; 2150 } 2151 2152 /* 2153 * We lack write permission and cannot replace the database 2154 * file, but let's at least check whether the data changed. 2155 */ 2156 2157 (void)strlcpy(tfn, "/tmp/mandocdb.XXXXXXXX", sizeof(tfn)); 2158 if (mkdtemp(tfn) == NULL) { 2159 exitcode = (int)MANDOCLEVEL_SYSERR; 2160 say("", "&%s", tfn); 2161 return; 2162 } 2163 2164 (void)strlcat(tfn, "/" MANDOC_DB, sizeof(tfn)); 2165 if (dba_write(tfn, dba) == -1) { 2166 exitcode = (int)MANDOCLEVEL_SYSERR; 2167 say(tfn, "&dba_write"); 2168 goto out; 2169 } 2170 2171 switch (child = fork()) { 2172 case -1: 2173 exitcode = (int)MANDOCLEVEL_SYSERR; 2174 say("", "&fork cmp"); 2175 return; 2176 case 0: 2177 execlp("cmp", "cmp", "-s", tfn, MANDOC_DB, (char *)NULL); 2178 say("", "&exec cmp"); 2179 exit(0); 2180 default: 2181 break; 2182 } 2183 if (waitpid(child, &status, 0) == -1) { 2184 exitcode = (int)MANDOCLEVEL_SYSERR; 2185 say("", "&wait cmp"); 2186 } else if (WIFSIGNALED(status)) { 2187 exitcode = (int)MANDOCLEVEL_SYSERR; 2188 say("", "cmp died from signal %d", WTERMSIG(status)); 2189 } else if (WEXITSTATUS(status)) { 2190 exitcode = (int)MANDOCLEVEL_SYSERR; 2191 say(MANDOC_DB, 2192 "Data changed, but cannot replace database"); 2193 } 2194 2195out: 2196 *strrchr(tfn, '/') = '\0'; 2197 switch (child = fork()) { 2198 case -1: 2199 exitcode = (int)MANDOCLEVEL_SYSERR; 2200 say("", "&fork rm"); 2201 return; 2202 case 0: 2203 execlp("rm", "rm", "-rf", tfn, (char *)NULL); 2204 say("", "&exec rm"); 2205 exit((int)MANDOCLEVEL_SYSERR); 2206 default: 2207 break; 2208 } 2209 if (waitpid(child, &status, 0) == -1) { 2210 exitcode = (int)MANDOCLEVEL_SYSERR; 2211 say("", "&wait rm"); 2212 } else if (WIFSIGNALED(status) || WEXITSTATUS(status)) { 2213 exitcode = (int)MANDOCLEVEL_SYSERR; 2214 say("", "%s: Cannot remove temporary directory", tfn); 2215 } 2216} 2217 2218static int 2219set_basedir(const char *targetdir, int report_baddir) 2220{ 2221 static char startdir[PATH_MAX]; 2222 static int getcwd_status; /* 1 = ok, 2 = failure */ 2223 static int chdir_status; /* 1 = changed directory */ 2224 char *cp; 2225 2226 /* 2227 * Remember the original working directory, if possible. 2228 * This will be needed if the second or a later directory 2229 * on the command line is given as a relative path. 2230 * Do not error out if the current directory is not 2231 * searchable: Maybe it won't be needed after all. 2232 */ 2233 if (0 == getcwd_status) { 2234 if (NULL == getcwd(startdir, sizeof(startdir))) { 2235 getcwd_status = 2; 2236 (void)strlcpy(startdir, strerror(errno), 2237 sizeof(startdir)); 2238 } else 2239 getcwd_status = 1; 2240 } 2241 2242 /* 2243 * We are leaving the old base directory. 2244 * Do not use it any longer, not even for messages. 2245 */ 2246 *basedir = '\0'; 2247 2248 /* 2249 * If and only if the directory was changed earlier and 2250 * the next directory to process is given as a relative path, 2251 * first go back, or bail out if that is impossible. 2252 */ 2253 if (chdir_status && '/' != *targetdir) { 2254 if (2 == getcwd_status) { 2255 exitcode = (int)MANDOCLEVEL_SYSERR; 2256 say("", "getcwd: %s", startdir); 2257 return 0; 2258 } 2259 if (-1 == chdir(startdir)) { 2260 exitcode = (int)MANDOCLEVEL_SYSERR; 2261 say("", "&chdir %s", startdir); 2262 return 0; 2263 } 2264 } 2265 2266 /* 2267 * Always resolve basedir to the canonicalized absolute 2268 * pathname and append a trailing slash, such that 2269 * we can reliably check whether files are inside. 2270 */ 2271 if (NULL == realpath(targetdir, basedir)) { 2272 if (report_baddir || errno != ENOENT) { 2273 exitcode = (int)MANDOCLEVEL_BADARG; 2274 say("", "&%s: realpath", targetdir); 2275 } 2276 return 0; 2277 } else if (-1 == chdir(basedir)) { 2278 if (report_baddir || errno != ENOENT) { 2279 exitcode = (int)MANDOCLEVEL_BADARG; 2280 say("", "&chdir"); 2281 } 2282 return 0; 2283 } 2284 chdir_status = 1; 2285 cp = strchr(basedir, '\0'); 2286 if ('/' != cp[-1]) { 2287 if (cp - basedir >= PATH_MAX - 1) { 2288 exitcode = (int)MANDOCLEVEL_SYSERR; 2289 say("", "Filename too long"); 2290 return 0; 2291 } 2292 *cp++ = '/'; 2293 *cp = '\0'; 2294 } 2295 return 1; 2296} 2297 2298static void 2299say(const char *file, const char *format, ...) 2300{ 2301 va_list ap; 2302 int use_errno; 2303 2304 if ('\0' != *basedir) 2305 fprintf(stderr, "%s", basedir); 2306 if ('\0' != *basedir && '\0' != *file) 2307 fputc('/', stderr); 2308 if ('\0' != *file) 2309 fprintf(stderr, "%s", file); 2310 2311 use_errno = 1; 2312 if (NULL != format) { 2313 switch (*format) { 2314 case '&': 2315 format++; 2316 break; 2317 case '\0': 2318 format = NULL; 2319 break; 2320 default: 2321 use_errno = 0; 2322 break; 2323 } 2324 } 2325 if (NULL != format) { 2326 if ('\0' != *basedir || '\0' != *file) 2327 fputs(": ", stderr); 2328 va_start(ap, format); 2329 vfprintf(stderr, format, ap); 2330 va_end(ap); 2331 } 2332 if (use_errno) { 2333 if ('\0' != *basedir || '\0' != *file || NULL != format) 2334 fputs(": ", stderr); 2335 perror(NULL); 2336 } else 2337 fputc('\n', stderr); 2338} 2339