1 /* Id: mandocdb.c,v 1.244 2017/02/17 14:45:55 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 "main.h" 55 #include "man.h" 56 #include "manconf.h" 57 #include "mansearch.h" 58 #include "dba_array.h" 59 #include "dba.h" 60 61 extern const char *const mansearch_keynames[]; 62 63 enum op { 64 OP_DEFAULT = 0, /* new dbs from dir list or default config */ 65 OP_CONFFILE, /* new databases from custom config file */ 66 OP_UPDATE, /* delete/add entries in existing database */ 67 OP_DELETE, /* delete entries from existing database */ 68 OP_TEST /* change no databases, report potential problems */ 69 }; 70 71 struct str { 72 const struct mpage *mpage; /* if set, the owning parse */ 73 uint64_t mask; /* bitmask in sequence */ 74 char key[]; /* rendered text */ 75 }; 76 77 struct inodev { 78 ino_t st_ino; 79 dev_t st_dev; 80 }; 81 82 struct mpage { 83 struct inodev inodev; /* used for hashing routine */ 84 struct dba_array *dba; 85 char *sec; /* section from file content */ 86 char *arch; /* architecture from file content */ 87 char *title; /* title from file content */ 88 char *desc; /* description from file content */ 89 struct mpage *next; /* singly linked list */ 90 struct mlink *mlinks; /* singly linked list */ 91 int name_head_done; 92 enum form form; /* format from file content */ 93 }; 94 95 struct mlink { 96 char file[PATH_MAX]; /* filename rel. to manpath */ 97 char *dsec; /* section from directory */ 98 char *arch; /* architecture from directory */ 99 char *name; /* name from file name (not empty) */ 100 char *fsec; /* section from file name suffix */ 101 struct mlink *next; /* singly linked list */ 102 struct mpage *mpage; /* parent */ 103 int gzip; /* filename has a .gz suffix */ 104 enum form dform; /* format from directory */ 105 enum form fform; /* format from file name suffix */ 106 }; 107 108 typedef int (*mdoc_fp)(struct mpage *, const struct roff_meta *, 109 const struct roff_node *); 110 111 struct mdoc_handler { 112 mdoc_fp fp; /* optional handler */ 113 uint64_t mask; /* set unless handler returns 0 */ 114 int taboo; /* node flags that must not be set */ 115 }; 116 117 118 int mandocdb(int, char *[]); 119 120 static void dbadd(struct dba *, struct mpage *); 121 static void dbadd_mlink(const struct mlink *mlink); 122 static void dbprune(struct dba *); 123 static void dbwrite(struct dba *); 124 static void filescan(const char *); 125 #if HAVE_FTS_COMPARE_CONST 126 static int fts_compare(const FTSENT *const *, const FTSENT *const *); 127 #else 128 static int fts_compare(const FTSENT **, const FTSENT **); 129 #endif 130 static void mlink_add(struct mlink *, const struct stat *); 131 static void mlink_check(struct mpage *, struct mlink *); 132 static void mlink_free(struct mlink *); 133 static void mlinks_undupe(struct mpage *); 134 static void mpages_free(void); 135 static void mpages_merge(struct dba *, struct mparse *); 136 static void parse_cat(struct mpage *, int); 137 static void parse_man(struct mpage *, const struct roff_meta *, 138 const struct roff_node *); 139 static void parse_mdoc(struct mpage *, const struct roff_meta *, 140 const struct roff_node *); 141 static int parse_mdoc_head(struct mpage *, const struct roff_meta *, 142 const struct roff_node *); 143 static int parse_mdoc_Fd(struct mpage *, const struct roff_meta *, 144 const struct roff_node *); 145 static void parse_mdoc_fname(struct mpage *, const struct roff_node *); 146 static int parse_mdoc_Fn(struct mpage *, const struct roff_meta *, 147 const struct roff_node *); 148 static int parse_mdoc_Fo(struct mpage *, const struct roff_meta *, 149 const struct roff_node *); 150 static int parse_mdoc_Nd(struct mpage *, const struct roff_meta *, 151 const struct roff_node *); 152 static int parse_mdoc_Nm(struct mpage *, const struct roff_meta *, 153 const struct roff_node *); 154 static int parse_mdoc_Sh(struct mpage *, const struct roff_meta *, 155 const struct roff_node *); 156 static int parse_mdoc_Va(struct mpage *, const struct roff_meta *, 157 const struct roff_node *); 158 static int parse_mdoc_Xr(struct mpage *, const struct roff_meta *, 159 const struct roff_node *); 160 static void putkey(const struct mpage *, char *, uint64_t); 161 static void putkeys(const struct mpage *, char *, size_t, uint64_t); 162 static void putmdockey(const struct mpage *, 163 const struct roff_node *, uint64_t, int); 164 static int render_string(char **, size_t *); 165 static void say(const char *, const char *, ...) 166 __attribute__((__format__ (__printf__, 2, 3))); 167 static int set_basedir(const char *, int); 168 static int treescan(void); 169 static size_t utf8(unsigned int, char [7]); 170 171 static int nodb; /* no database changes */ 172 static int mparse_options; /* abort the parse early */ 173 static int use_all; /* use all found files */ 174 static int debug; /* print what we're doing */ 175 static int warnings; /* warn about crap */ 176 static int write_utf8; /* write UTF-8 output; else ASCII */ 177 static int exitcode; /* to be returned by main */ 178 static enum op op; /* operational mode */ 179 static char basedir[PATH_MAX]; /* current base directory */ 180 static struct mpage *mpage_head; /* list of distinct manual pages */ 181 static struct ohash mpages; /* table of distinct manual pages */ 182 static struct ohash mlinks; /* table of directory entries */ 183 static struct ohash names; /* table of all names */ 184 static struct ohash strings; /* table of all strings */ 185 static uint64_t name_mask; 186 187 static const struct mdoc_handler mdocs[MDOC_MAX] = { 188 { NULL, 0, 0 }, /* Ap */ 189 { NULL, 0, NODE_NOPRT }, /* Dd */ 190 { NULL, 0, NODE_NOPRT }, /* Dt */ 191 { NULL, 0, NODE_NOPRT }, /* Os */ 192 { parse_mdoc_Sh, TYPE_Sh, 0 }, /* Sh */ 193 { parse_mdoc_head, TYPE_Ss, 0 }, /* Ss */ 194 { NULL, 0, 0 }, /* Pp */ 195 { NULL, 0, 0 }, /* D1 */ 196 { NULL, 0, 0 }, /* Dl */ 197 { NULL, 0, 0 }, /* Bd */ 198 { NULL, 0, 0 }, /* Ed */ 199 { NULL, 0, 0 }, /* Bl */ 200 { NULL, 0, 0 }, /* El */ 201 { NULL, 0, 0 }, /* It */ 202 { NULL, 0, 0 }, /* Ad */ 203 { NULL, TYPE_An, 0 }, /* An */ 204 { NULL, TYPE_Ar, 0 }, /* Ar */ 205 { NULL, TYPE_Cd, 0 }, /* Cd */ 206 { NULL, TYPE_Cm, 0 }, /* Cm */ 207 { NULL, TYPE_Dv, 0 }, /* Dv */ 208 { NULL, TYPE_Er, 0 }, /* Er */ 209 { NULL, TYPE_Ev, 0 }, /* Ev */ 210 { NULL, 0, 0 }, /* Ex */ 211 { NULL, TYPE_Fa, 0 }, /* Fa */ 212 { parse_mdoc_Fd, 0, 0 }, /* Fd */ 213 { NULL, TYPE_Fl, 0 }, /* Fl */ 214 { parse_mdoc_Fn, 0, 0 }, /* Fn */ 215 { NULL, TYPE_Ft, 0 }, /* Ft */ 216 { NULL, TYPE_Ic, 0 }, /* Ic */ 217 { NULL, TYPE_In, 0 }, /* In */ 218 { NULL, TYPE_Li, 0 }, /* Li */ 219 { parse_mdoc_Nd, 0, 0 }, /* Nd */ 220 { parse_mdoc_Nm, 0, 0 }, /* Nm */ 221 { NULL, 0, 0 }, /* Op */ 222 { NULL, 0, 0 }, /* Ot */ 223 { NULL, TYPE_Pa, NODE_NOSRC }, /* Pa */ 224 { NULL, 0, 0 }, /* Rv */ 225 { NULL, TYPE_St, 0 }, /* St */ 226 { parse_mdoc_Va, TYPE_Va, 0 }, /* Va */ 227 { parse_mdoc_Va, TYPE_Vt, 0 }, /* Vt */ 228 { parse_mdoc_Xr, 0, 0 }, /* Xr */ 229 { NULL, 0, 0 }, /* %A */ 230 { NULL, 0, 0 }, /* %B */ 231 { NULL, 0, 0 }, /* %D */ 232 { NULL, 0, 0 }, /* %I */ 233 { NULL, 0, 0 }, /* %J */ 234 { NULL, 0, 0 }, /* %N */ 235 { NULL, 0, 0 }, /* %O */ 236 { NULL, 0, 0 }, /* %P */ 237 { NULL, 0, 0 }, /* %R */ 238 { NULL, 0, 0 }, /* %T */ 239 { NULL, 0, 0 }, /* %V */ 240 { NULL, 0, 0 }, /* Ac */ 241 { NULL, 0, 0 }, /* Ao */ 242 { NULL, 0, 0 }, /* Aq */ 243 { NULL, TYPE_At, 0 }, /* At */ 244 { NULL, 0, 0 }, /* Bc */ 245 { NULL, 0, 0 }, /* Bf */ 246 { NULL, 0, 0 }, /* Bo */ 247 { NULL, 0, 0 }, /* Bq */ 248 { NULL, TYPE_Bsx, NODE_NOSRC }, /* Bsx */ 249 { NULL, TYPE_Bx, NODE_NOSRC }, /* Bx */ 250 { NULL, 0, 0 }, /* Db */ 251 { NULL, 0, 0 }, /* Dc */ 252 { NULL, 0, 0 }, /* Do */ 253 { NULL, 0, 0 }, /* Dq */ 254 { NULL, 0, 0 }, /* Ec */ 255 { NULL, 0, 0 }, /* Ef */ 256 { NULL, TYPE_Em, 0 }, /* Em */ 257 { NULL, 0, 0 }, /* Eo */ 258 { NULL, TYPE_Fx, NODE_NOSRC }, /* Fx */ 259 { NULL, TYPE_Ms, 0 }, /* Ms */ 260 { NULL, 0, 0 }, /* No */ 261 { NULL, 0, 0 }, /* Ns */ 262 { NULL, TYPE_Nx, NODE_NOSRC }, /* Nx */ 263 { NULL, TYPE_Ox, NODE_NOSRC }, /* Ox */ 264 { NULL, 0, 0 }, /* Pc */ 265 { NULL, 0, 0 }, /* Pf */ 266 { NULL, 0, 0 }, /* Po */ 267 { NULL, 0, 0 }, /* Pq */ 268 { NULL, 0, 0 }, /* Qc */ 269 { NULL, 0, 0 }, /* Ql */ 270 { NULL, 0, 0 }, /* Qo */ 271 { NULL, 0, 0 }, /* Qq */ 272 { NULL, 0, 0 }, /* Re */ 273 { NULL, 0, 0 }, /* Rs */ 274 { NULL, 0, 0 }, /* Sc */ 275 { NULL, 0, 0 }, /* So */ 276 { NULL, 0, 0 }, /* Sq */ 277 { NULL, 0, 0 }, /* Sm */ 278 { NULL, 0, 0 }, /* Sx */ 279 { NULL, TYPE_Sy, 0 }, /* Sy */ 280 { NULL, TYPE_Tn, 0 }, /* Tn */ 281 { NULL, 0, NODE_NOSRC }, /* Ux */ 282 { NULL, 0, 0 }, /* Xc */ 283 { NULL, 0, 0 }, /* Xo */ 284 { parse_mdoc_Fo, 0, 0 }, /* Fo */ 285 { NULL, 0, 0 }, /* Fc */ 286 { NULL, 0, 0 }, /* Oo */ 287 { NULL, 0, 0 }, /* Oc */ 288 { NULL, 0, 0 }, /* Bk */ 289 { NULL, 0, 0 }, /* Ek */ 290 { NULL, 0, 0 }, /* Bt */ 291 { NULL, 0, 0 }, /* Hf */ 292 { NULL, 0, 0 }, /* Fr */ 293 { NULL, 0, 0 }, /* Ud */ 294 { NULL, TYPE_Lb, NODE_NOSRC }, /* Lb */ 295 { NULL, 0, 0 }, /* Lp */ 296 { NULL, TYPE_Lk, 0 }, /* Lk */ 297 { NULL, TYPE_Mt, NODE_NOSRC }, /* Mt */ 298 { NULL, 0, 0 }, /* Brq */ 299 { NULL, 0, 0 }, /* Bro */ 300 { NULL, 0, 0 }, /* Brc */ 301 { NULL, 0, 0 }, /* %C */ 302 { NULL, 0, 0 }, /* Es */ 303 { NULL, 0, 0 }, /* En */ 304 { NULL, TYPE_Dx, NODE_NOSRC }, /* Dx */ 305 { NULL, 0, 0 }, /* %Q */ 306 { NULL, 0, 0 }, /* br */ 307 { NULL, 0, 0 }, /* sp */ 308 { NULL, 0, 0 }, /* %U */ 309 { NULL, 0, 0 }, /* Ta */ 310 { NULL, 0, 0 }, /* ll */ 311 }; 312 313 314 int 315 mandocdb(int argc, char *argv[]) 316 { 317 struct manconf conf; 318 struct mparse *mp; 319 struct dba *dba; 320 const char *path_arg, *progname; 321 size_t j, sz; 322 int ch, i; 323 324 #if HAVE_PLEDGE 325 if (pledge("stdio rpath wpath cpath fattr flock proc exec", NULL) == -1) { 326 warn("pledge"); 327 return (int)MANDOCLEVEL_SYSERR; 328 } 329 #endif 330 331 #if HAVE_SANDBOX_INIT 332 if (sandbox_init(kSBXProfileNoInternet, SANDBOX_NAMED, NULL) == -1) { 333 warnx("sandbox_init"); 334 return (int)MANDOCLEVEL_SYSERR; 335 } 336 #endif 337 338 memset(&conf, 0, sizeof(conf)); 339 340 /* 341 * We accept a few different invocations. 342 * The CHECKOP macro makes sure that invocation styles don't 343 * clobber each other. 344 */ 345 #define CHECKOP(_op, _ch) do \ 346 if (OP_DEFAULT != (_op)) { \ 347 warnx("-%c: Conflicting option", (_ch)); \ 348 goto usage; \ 349 } while (/*CONSTCOND*/0) 350 351 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, MANDOCLEVEL_BADARG, NULL, 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 #if HAVE_PLEDGE 446 if (!nodb) { 447 if (pledge("stdio rpath wpath cpath fattr flock", NULL) == -1) { 448 warn("pledge"); 449 exitcode = (int)MANDOCLEVEL_SYSERR; 450 goto out; 451 } 452 } 453 #endif 454 use_all = 1; 455 for (i = 0; i < argc; i++) 456 filescan(argv[i]); 457 if (nodb == 0) 458 dbprune(dba); 459 } else { 460 /* Database missing or corrupt. */ 461 if (op != OP_UPDATE || errno != ENOENT) 462 say(MANDOC_DB, "%s: Automatically recreating" 463 " from scratch", strerror(errno)); 464 exitcode = (int)MANDOCLEVEL_OK; 465 op = OP_DEFAULT; 466 if (0 == treescan()) 467 goto out; 468 dba = dba_new(128); 469 } 470 if (OP_DELETE != op) 471 mpages_merge(dba, mp); 472 if (nodb == 0) 473 dbwrite(dba); 474 dba_free(dba); 475 } else { 476 /* 477 * If we have arguments, use them as our manpaths. 478 * If we don't, use man.conf(5). 479 */ 480 if (argc > 0) { 481 conf.manpath.paths = mandoc_reallocarray(NULL, 482 argc, sizeof(char *)); 483 conf.manpath.sz = (size_t)argc; 484 for (i = 0; i < argc; i++) 485 conf.manpath.paths[i] = mandoc_strdup(argv[i]); 486 } else 487 manconf_parse(&conf, path_arg, NULL, NULL); 488 489 if (conf.manpath.sz == 0) { 490 exitcode = (int)MANDOCLEVEL_BADARG; 491 say("", "Empty manpath"); 492 } 493 494 /* 495 * First scan the tree rooted at a base directory, then 496 * build a new database and finally move it into place. 497 * Ignore zero-length directories and strip trailing 498 * slashes. 499 */ 500 for (j = 0; j < conf.manpath.sz; j++) { 501 sz = strlen(conf.manpath.paths[j]); 502 if (sz && conf.manpath.paths[j][sz - 1] == '/') 503 conf.manpath.paths[j][--sz] = '\0'; 504 if (0 == sz) 505 continue; 506 507 if (j) { 508 mandoc_ohash_init(&mpages, 6, 509 offsetof(struct mpage, inodev)); 510 mandoc_ohash_init(&mlinks, 6, 511 offsetof(struct mlink, file)); 512 } 513 514 if ( ! set_basedir(conf.manpath.paths[j], argc > 0)) 515 continue; 516 if (0 == treescan()) 517 continue; 518 dba = dba_new(128); 519 mpages_merge(dba, mp); 520 if (nodb == 0) 521 dbwrite(dba); 522 dba_free(dba); 523 524 if (j + 1 < conf.manpath.sz) { 525 mpages_free(); 526 ohash_delete(&mpages); 527 ohash_delete(&mlinks); 528 } 529 } 530 } 531 out: 532 manconf_free(&conf); 533 mparse_free(mp); 534 mchars_free(); 535 mpages_free(); 536 ohash_delete(&mpages); 537 ohash_delete(&mlinks); 538 return exitcode; 539 usage: 540 progname = getprogname(); 541 fprintf(stderr, "usage: %s [-aDnpQ] [-C file] [-Tutf8]\n" 542 " %s [-aDnpQ] [-Tutf8] dir ...\n" 543 " %s [-DnpQ] [-Tutf8] -d dir [file ...]\n" 544 " %s [-Dnp] -u dir [file ...]\n" 545 " %s [-Q] -t file ...\n", 546 progname, progname, progname, progname, progname); 547 548 return (int)MANDOCLEVEL_BADARG; 549 } 550 551 /* 552 * To get a singly linked list in alpha order while inserting entries 553 * at the beginning, process directory entries in reverse alpha order. 554 */ 555 static int 556 #if HAVE_FTS_COMPARE_CONST 557 fts_compare(const FTSENT *const *a, const FTSENT *const *b) 558 #else 559 fts_compare(const FTSENT **a, const FTSENT **b) 560 #endif 561 { 562 return -strcmp((*a)->fts_name, (*b)->fts_name); 563 } 564 565 /* 566 * Scan a directory tree rooted at "basedir" for manpages. 567 * We use fts(), scanning directory parts along the way for clues to our 568 * section and architecture. 569 * 570 * If use_all has been specified, grok all files. 571 * If not, sanitise paths to the following: 572 * 573 * [./]man*[/<arch>]/<name>.<section> 574 * or 575 * [./]cat<section>[/<arch>]/<name>.0 576 * 577 * TODO: accommodate for multi-language directories. 578 */ 579 static int 580 treescan(void) 581 { 582 char buf[PATH_MAX]; 583 FTS *f; 584 FTSENT *ff; 585 struct mlink *mlink; 586 int gzip; 587 enum form dform; 588 char *dsec, *arch, *fsec, *cp; 589 const char *path; 590 const char *argv[2]; 591 592 argv[0] = "."; 593 argv[1] = NULL; 594 595 f = fts_open((char * const *)argv, FTS_PHYSICAL | FTS_NOCHDIR, 596 fts_compare); 597 if (f == NULL) { 598 exitcode = (int)MANDOCLEVEL_SYSERR; 599 say("", "&fts_open"); 600 return 0; 601 } 602 603 dsec = arch = NULL; 604 dform = FORM_NONE; 605 606 while ((ff = fts_read(f)) != NULL) { 607 path = ff->fts_path + 2; 608 switch (ff->fts_info) { 609 610 /* 611 * Symbolic links require various sanity checks, 612 * then get handled just like regular files. 613 */ 614 case FTS_SL: 615 if (realpath(path, buf) == NULL) { 616 if (warnings) 617 say(path, "&realpath"); 618 continue; 619 } 620 if (strstr(buf, basedir) != buf 621 #ifdef HOMEBREWDIR 622 && strstr(buf, HOMEBREWDIR) != buf 623 #endif 624 ) { 625 if (warnings) say("", 626 "%s: outside base directory", buf); 627 continue; 628 } 629 /* Use logical inode to avoid mpages dupe. */ 630 if (stat(path, ff->fts_statp) == -1) { 631 if (warnings) 632 say(path, "&stat"); 633 continue; 634 } 635 /* FALLTHROUGH */ 636 637 /* 638 * If we're a regular file, add an mlink by using the 639 * stored directory data and handling the filename. 640 */ 641 case FTS_F: 642 if ( ! strcmp(path, MANDOC_DB)) 643 continue; 644 if ( ! use_all && ff->fts_level < 2) { 645 if (warnings) 646 say(path, "Extraneous file"); 647 continue; 648 } 649 gzip = 0; 650 fsec = NULL; 651 while (fsec == NULL) { 652 fsec = strrchr(ff->fts_name, '.'); 653 if (fsec == NULL || strcmp(fsec+1, "gz")) 654 break; 655 gzip = 1; 656 *fsec = '\0'; 657 fsec = NULL; 658 } 659 if (fsec == NULL) { 660 if ( ! use_all) { 661 if (warnings) 662 say(path, 663 "No filename suffix"); 664 continue; 665 } 666 } else if ( ! strcmp(++fsec, "html")) { 667 if (warnings) 668 say(path, "Skip html"); 669 continue; 670 } else if ( ! strcmp(fsec, "ps")) { 671 if (warnings) 672 say(path, "Skip ps"); 673 continue; 674 } else if ( ! strcmp(fsec, "pdf")) { 675 if (warnings) 676 say(path, "Skip pdf"); 677 continue; 678 } else if ( ! use_all && 679 ((dform == FORM_SRC && 680 strncmp(fsec, dsec, strlen(dsec))) || 681 (dform == FORM_CAT && strcmp(fsec, "0")))) { 682 if (warnings) 683 say(path, "Wrong filename suffix"); 684 continue; 685 } else 686 fsec[-1] = '\0'; 687 688 mlink = mandoc_calloc(1, sizeof(struct mlink)); 689 if (strlcpy(mlink->file, path, 690 sizeof(mlink->file)) >= 691 sizeof(mlink->file)) { 692 say(path, "Filename too long"); 693 free(mlink); 694 continue; 695 } 696 mlink->dform = dform; 697 mlink->dsec = dsec; 698 mlink->arch = arch; 699 mlink->name = ff->fts_name; 700 mlink->fsec = fsec; 701 mlink->gzip = gzip; 702 mlink_add(mlink, ff->fts_statp); 703 continue; 704 705 case FTS_D: 706 case FTS_DP: 707 break; 708 709 default: 710 if (warnings) 711 say(path, "Not a regular file"); 712 continue; 713 } 714 715 switch (ff->fts_level) { 716 case 0: 717 /* Ignore the root directory. */ 718 break; 719 case 1: 720 /* 721 * This might contain manX/ or catX/. 722 * Try to infer this from the name. 723 * If we're not in use_all, enforce it. 724 */ 725 cp = ff->fts_name; 726 if (ff->fts_info == FTS_DP) { 727 dform = FORM_NONE; 728 dsec = NULL; 729 break; 730 } 731 732 if ( ! strncmp(cp, "man", 3)) { 733 dform = FORM_SRC; 734 dsec = cp + 3; 735 } else if ( ! strncmp(cp, "cat", 3)) { 736 dform = FORM_CAT; 737 dsec = cp + 3; 738 } else { 739 dform = FORM_NONE; 740 dsec = NULL; 741 } 742 743 if (dsec != NULL || use_all) 744 break; 745 746 if (warnings) 747 say(path, "Unknown directory part"); 748 fts_set(f, ff, FTS_SKIP); 749 break; 750 case 2: 751 /* 752 * Possibly our architecture. 753 * If we're descending, keep tabs on it. 754 */ 755 if (ff->fts_info != FTS_DP && dsec != NULL) 756 arch = ff->fts_name; 757 else 758 arch = NULL; 759 break; 760 default: 761 if (ff->fts_info == FTS_DP || use_all) 762 break; 763 if (warnings) 764 say(path, "Extraneous directory part"); 765 fts_set(f, ff, FTS_SKIP); 766 break; 767 } 768 } 769 770 fts_close(f); 771 return 1; 772 } 773 774 /* 775 * Add a file to the mlinks table. 776 * Do not verify that it's a "valid" looking manpage (we'll do that 777 * later). 778 * 779 * Try to infer the manual section, architecture, and page name from the 780 * path, assuming it looks like 781 * 782 * [./]man*[/<arch>]/<name>.<section> 783 * or 784 * [./]cat<section>[/<arch>]/<name>.0 785 * 786 * See treescan() for the fts(3) version of this. 787 */ 788 static void 789 filescan(const char *file) 790 { 791 char buf[PATH_MAX]; 792 struct stat st; 793 struct mlink *mlink; 794 char *p, *start; 795 796 assert(use_all); 797 798 if (0 == strncmp(file, "./", 2)) 799 file += 2; 800 801 /* 802 * We have to do lstat(2) before realpath(3) loses 803 * the information whether this is a symbolic link. 804 * We need to know that because for symbolic links, 805 * we want to use the orginal file name, while for 806 * regular files, we want to use the real path. 807 */ 808 if (-1 == lstat(file, &st)) { 809 exitcode = (int)MANDOCLEVEL_BADARG; 810 say(file, "&lstat"); 811 return; 812 } else if (0 == ((S_IFREG | S_IFLNK) & st.st_mode)) { 813 exitcode = (int)MANDOCLEVEL_BADARG; 814 say(file, "Not a regular file"); 815 return; 816 } 817 818 /* 819 * We have to resolve the file name to the real path 820 * in any case for the base directory check. 821 */ 822 if (NULL == realpath(file, buf)) { 823 exitcode = (int)MANDOCLEVEL_BADARG; 824 say(file, "&realpath"); 825 return; 826 } 827 828 if (OP_TEST == op) 829 start = buf; 830 else if (strstr(buf, basedir) == buf) 831 start = buf + strlen(basedir); 832 #ifdef HOMEBREWDIR 833 else if (strstr(buf, HOMEBREWDIR) == buf) 834 start = buf; 835 #endif 836 else { 837 exitcode = (int)MANDOCLEVEL_BADARG; 838 say("", "%s: outside base directory", buf); 839 return; 840 } 841 842 /* 843 * Now we are sure the file is inside our tree. 844 * If it is a symbolic link, ignore the real path 845 * and use the original name. 846 * This implies passing stuff like "cat1/../man1/foo.1" 847 * on the command line won't work. So don't do that. 848 * Note the stat(2) can still fail if the link target 849 * doesn't exist. 850 */ 851 if (S_IFLNK & st.st_mode) { 852 if (-1 == stat(buf, &st)) { 853 exitcode = (int)MANDOCLEVEL_BADARG; 854 say(file, "&stat"); 855 return; 856 } 857 if (strlcpy(buf, file, sizeof(buf)) >= sizeof(buf)) { 858 say(file, "Filename too long"); 859 return; 860 } 861 start = buf; 862 if (OP_TEST != op && strstr(buf, basedir) == buf) 863 start += strlen(basedir); 864 } 865 866 mlink = mandoc_calloc(1, sizeof(struct mlink)); 867 mlink->dform = FORM_NONE; 868 if (strlcpy(mlink->file, start, sizeof(mlink->file)) >= 869 sizeof(mlink->file)) { 870 say(start, "Filename too long"); 871 free(mlink); 872 return; 873 } 874 875 /* 876 * In test mode or when the original name is absolute 877 * but outside our tree, guess the base directory. 878 */ 879 880 if (op == OP_TEST || (start == buf && *start == '/')) { 881 if (strncmp(buf, "man/", 4) == 0) 882 start = buf + 4; 883 else if ((start = strstr(buf, "/man/")) != NULL) 884 start += 5; 885 else 886 start = buf; 887 } 888 889 /* 890 * First try to guess our directory structure. 891 * If we find a separator, try to look for man* or cat*. 892 * If we find one of these and what's underneath is a directory, 893 * assume it's an architecture. 894 */ 895 if (NULL != (p = strchr(start, '/'))) { 896 *p++ = '\0'; 897 if (0 == strncmp(start, "man", 3)) { 898 mlink->dform = FORM_SRC; 899 mlink->dsec = start + 3; 900 } else if (0 == strncmp(start, "cat", 3)) { 901 mlink->dform = FORM_CAT; 902 mlink->dsec = start + 3; 903 } 904 905 start = p; 906 if (NULL != mlink->dsec && NULL != (p = strchr(start, '/'))) { 907 *p++ = '\0'; 908 mlink->arch = start; 909 start = p; 910 } 911 } 912 913 /* 914 * Now check the file suffix. 915 * Suffix of `.0' indicates a catpage, `.1-9' is a manpage. 916 */ 917 p = strrchr(start, '\0'); 918 while (p-- > start && '/' != *p && '.' != *p) 919 /* Loop. */ ; 920 921 if ('.' == *p) { 922 *p++ = '\0'; 923 mlink->fsec = p; 924 } 925 926 /* 927 * Now try to parse the name. 928 * Use the filename portion of the path. 929 */ 930 mlink->name = start; 931 if (NULL != (p = strrchr(start, '/'))) { 932 mlink->name = p + 1; 933 *p = '\0'; 934 } 935 mlink_add(mlink, &st); 936 } 937 938 static void 939 mlink_add(struct mlink *mlink, const struct stat *st) 940 { 941 struct inodev inodev; 942 struct mpage *mpage; 943 unsigned int slot; 944 945 assert(NULL != mlink->file); 946 947 mlink->dsec = mandoc_strdup(mlink->dsec ? mlink->dsec : ""); 948 mlink->arch = mandoc_strdup(mlink->arch ? mlink->arch : ""); 949 mlink->name = mandoc_strdup(mlink->name ? mlink->name : ""); 950 mlink->fsec = mandoc_strdup(mlink->fsec ? mlink->fsec : ""); 951 952 if ('0' == *mlink->fsec) { 953 free(mlink->fsec); 954 mlink->fsec = mandoc_strdup(mlink->dsec); 955 mlink->fform = FORM_CAT; 956 } else if ('1' <= *mlink->fsec && '9' >= *mlink->fsec) 957 mlink->fform = FORM_SRC; 958 else 959 mlink->fform = FORM_NONE; 960 961 slot = ohash_qlookup(&mlinks, mlink->file); 962 assert(NULL == ohash_find(&mlinks, slot)); 963 ohash_insert(&mlinks, slot, mlink); 964 965 memset(&inodev, 0, sizeof(inodev)); /* Clear padding. */ 966 inodev.st_ino = st->st_ino; 967 inodev.st_dev = st->st_dev; 968 slot = ohash_lookup_memory(&mpages, (char *)&inodev, 969 sizeof(struct inodev), inodev.st_ino); 970 mpage = ohash_find(&mpages, slot); 971 if (NULL == mpage) { 972 mpage = mandoc_calloc(1, sizeof(struct mpage)); 973 mpage->inodev.st_ino = inodev.st_ino; 974 mpage->inodev.st_dev = inodev.st_dev; 975 mpage->form = FORM_NONE; 976 mpage->next = mpage_head; 977 mpage_head = mpage; 978 ohash_insert(&mpages, slot, mpage); 979 } else 980 mlink->next = mpage->mlinks; 981 mpage->mlinks = mlink; 982 mlink->mpage = mpage; 983 } 984 985 static void 986 mlink_free(struct mlink *mlink) 987 { 988 989 free(mlink->dsec); 990 free(mlink->arch); 991 free(mlink->name); 992 free(mlink->fsec); 993 free(mlink); 994 } 995 996 static void 997 mpages_free(void) 998 { 999 struct mpage *mpage; 1000 struct mlink *mlink; 1001 1002 while ((mpage = mpage_head) != NULL) { 1003 while ((mlink = mpage->mlinks) != NULL) { 1004 mpage->mlinks = mlink->next; 1005 mlink_free(mlink); 1006 } 1007 mpage_head = mpage->next; 1008 free(mpage->sec); 1009 free(mpage->arch); 1010 free(mpage->title); 1011 free(mpage->desc); 1012 free(mpage); 1013 } 1014 } 1015 1016 /* 1017 * For each mlink to the mpage, check whether the path looks like 1018 * it is formatted, and if it does, check whether a source manual 1019 * exists by the same name, ignoring the suffix. 1020 * If both conditions hold, drop the mlink. 1021 */ 1022 static void 1023 mlinks_undupe(struct mpage *mpage) 1024 { 1025 char buf[PATH_MAX]; 1026 struct mlink **prev; 1027 struct mlink *mlink; 1028 char *bufp; 1029 1030 mpage->form = FORM_CAT; 1031 prev = &mpage->mlinks; 1032 while (NULL != (mlink = *prev)) { 1033 if (FORM_CAT != mlink->dform) { 1034 mpage->form = FORM_NONE; 1035 goto nextlink; 1036 } 1037 (void)strlcpy(buf, mlink->file, sizeof(buf)); 1038 bufp = strstr(buf, "cat"); 1039 assert(NULL != bufp); 1040 memcpy(bufp, "man", 3); 1041 if (NULL != (bufp = strrchr(buf, '.'))) 1042 *++bufp = '\0'; 1043 (void)strlcat(buf, mlink->dsec, sizeof(buf)); 1044 if (NULL == ohash_find(&mlinks, 1045 ohash_qlookup(&mlinks, buf))) 1046 goto nextlink; 1047 if (warnings) 1048 say(mlink->file, "Man source exists: %s", buf); 1049 if (use_all) 1050 goto nextlink; 1051 *prev = mlink->next; 1052 mlink_free(mlink); 1053 continue; 1054 nextlink: 1055 prev = &(*prev)->next; 1056 } 1057 } 1058 1059 static void 1060 mlink_check(struct mpage *mpage, struct mlink *mlink) 1061 { 1062 struct str *str; 1063 unsigned int slot; 1064 1065 /* 1066 * Check whether the manual section given in a file 1067 * agrees with the directory where the file is located. 1068 * Some manuals have suffixes like (3p) on their 1069 * section number either inside the file or in the 1070 * directory name, some are linked into more than one 1071 * section, like encrypt(1) = makekey(8). 1072 */ 1073 1074 if (FORM_SRC == mpage->form && 1075 strcasecmp(mpage->sec, mlink->dsec)) 1076 say(mlink->file, "Section \"%s\" manual in %s directory", 1077 mpage->sec, mlink->dsec); 1078 1079 /* 1080 * Manual page directories exist for each kernel 1081 * architecture as returned by machine(1). 1082 * However, many manuals only depend on the 1083 * application architecture as returned by arch(1). 1084 * For example, some (2/ARM) manuals are shared 1085 * across the "armish" and "zaurus" kernel 1086 * architectures. 1087 * A few manuals are even shared across completely 1088 * different architectures, for example fdformat(1) 1089 * on amd64, i386, and sparc64. 1090 */ 1091 1092 if (strcasecmp(mpage->arch, mlink->arch)) 1093 say(mlink->file, "Architecture \"%s\" manual in " 1094 "\"%s\" directory", mpage->arch, mlink->arch); 1095 1096 /* 1097 * XXX 1098 * parse_cat() doesn't set NAME_TITLE yet. 1099 */ 1100 1101 if (FORM_CAT == mpage->form) 1102 return; 1103 1104 /* 1105 * Check whether this mlink 1106 * appears as a name in the NAME section. 1107 */ 1108 1109 slot = ohash_qlookup(&names, mlink->name); 1110 str = ohash_find(&names, slot); 1111 assert(NULL != str); 1112 if ( ! (NAME_TITLE & str->mask)) 1113 say(mlink->file, "Name missing in NAME section"); 1114 } 1115 1116 /* 1117 * Run through the files in the global vector "mpages" 1118 * and add them to the database specified in "basedir". 1119 * 1120 * This handles the parsing scheme itself, using the cues of directory 1121 * and filename to determine whether the file is parsable or not. 1122 */ 1123 static void 1124 mpages_merge(struct dba *dba, struct mparse *mp) 1125 { 1126 struct mpage *mpage, *mpage_dest; 1127 struct mlink *mlink, *mlink_dest; 1128 struct roff_man *man; 1129 char *sodest; 1130 char *cp; 1131 int fd; 1132 1133 for (mpage = mpage_head; mpage != NULL; mpage = mpage->next) { 1134 mlinks_undupe(mpage); 1135 if ((mlink = mpage->mlinks) == NULL) 1136 continue; 1137 1138 name_mask = NAME_MASK; 1139 mandoc_ohash_init(&names, 4, offsetof(struct str, key)); 1140 mandoc_ohash_init(&strings, 6, offsetof(struct str, key)); 1141 mparse_reset(mp); 1142 man = NULL; 1143 sodest = NULL; 1144 1145 if ((fd = mparse_open(mp, mlink->file)) == -1) { 1146 say(mlink->file, "&open"); 1147 goto nextpage; 1148 } 1149 1150 /* 1151 * Interpret the file as mdoc(7) or man(7) source 1152 * code, unless it is known to be formatted. 1153 */ 1154 if (mlink->dform != FORM_CAT || mlink->fform != FORM_CAT) { 1155 mparse_readfd(mp, fd, mlink->file); 1156 close(fd); 1157 fd = -1; 1158 mparse_result(mp, &man, &sodest); 1159 } 1160 1161 if (sodest != NULL) { 1162 mlink_dest = ohash_find(&mlinks, 1163 ohash_qlookup(&mlinks, sodest)); 1164 if (mlink_dest == NULL) { 1165 mandoc_asprintf(&cp, "%s.gz", sodest); 1166 mlink_dest = ohash_find(&mlinks, 1167 ohash_qlookup(&mlinks, cp)); 1168 free(cp); 1169 } 1170 if (mlink_dest != NULL) { 1171 1172 /* The .so target exists. */ 1173 1174 mpage_dest = mlink_dest->mpage; 1175 while (1) { 1176 mlink->mpage = mpage_dest; 1177 1178 /* 1179 * If the target was already 1180 * processed, add the links 1181 * to the database now. 1182 * Otherwise, this will 1183 * happen when we come 1184 * to the target. 1185 */ 1186 1187 if (mpage_dest->dba != NULL) 1188 dbadd_mlink(mlink); 1189 1190 if (mlink->next == NULL) 1191 break; 1192 mlink = mlink->next; 1193 } 1194 1195 /* Move all links to the target. */ 1196 1197 mlink->next = mlink_dest->next; 1198 mlink_dest->next = mpage->mlinks; 1199 mpage->mlinks = NULL; 1200 } 1201 goto nextpage; 1202 } else if (man != NULL && man->macroset == MACROSET_MDOC) { 1203 mdoc_validate(man); 1204 mpage->form = FORM_SRC; 1205 mpage->sec = man->meta.msec; 1206 mpage->sec = mandoc_strdup( 1207 mpage->sec == NULL ? "" : mpage->sec); 1208 mpage->arch = man->meta.arch; 1209 mpage->arch = mandoc_strdup( 1210 mpage->arch == NULL ? "" : mpage->arch); 1211 mpage->title = mandoc_strdup(man->meta.title); 1212 } else if (man != NULL && man->macroset == MACROSET_MAN) { 1213 man_validate(man); 1214 if (*man->meta.msec != '\0' || 1215 *man->meta.msec != '\0') { 1216 mpage->form = FORM_SRC; 1217 mpage->sec = mandoc_strdup(man->meta.msec); 1218 mpage->arch = mandoc_strdup(mlink->arch); 1219 mpage->title = mandoc_strdup(man->meta.title); 1220 } else 1221 man = NULL; 1222 } 1223 1224 assert(mpage->desc == NULL); 1225 if (man == NULL) { 1226 mpage->form = FORM_CAT; 1227 mpage->sec = mandoc_strdup(mlink->dsec); 1228 mpage->arch = mandoc_strdup(mlink->arch); 1229 mpage->title = mandoc_strdup(mlink->name); 1230 parse_cat(mpage, fd); 1231 } else if (man->macroset == MACROSET_MDOC) 1232 parse_mdoc(mpage, &man->meta, man->first); 1233 else 1234 parse_man(mpage, &man->meta, man->first); 1235 if (mpage->desc == NULL) { 1236 mpage->desc = mandoc_strdup(mlink->name); 1237 if (warnings) 1238 say(mlink->file, "No one-line description, " 1239 "using filename \"%s\"", mlink->name); 1240 } 1241 1242 for (mlink = mpage->mlinks; 1243 mlink != NULL; 1244 mlink = mlink->next) { 1245 putkey(mpage, mlink->name, NAME_FILE); 1246 if (warnings && !use_all) 1247 mlink_check(mpage, mlink); 1248 } 1249 1250 dbadd(dba, mpage); 1251 1252 nextpage: 1253 ohash_delete(&strings); 1254 ohash_delete(&names); 1255 } 1256 } 1257 1258 static void 1259 parse_cat(struct mpage *mpage, int fd) 1260 { 1261 FILE *stream; 1262 struct mlink *mlink; 1263 char *line, *p, *title, *sec; 1264 size_t linesz, plen, titlesz; 1265 ssize_t len; 1266 int offs; 1267 1268 mlink = mpage->mlinks; 1269 stream = fd == -1 ? fopen(mlink->file, "r") : fdopen(fd, "r"); 1270 if (stream == NULL) { 1271 if (fd != -1) 1272 close(fd); 1273 if (warnings) 1274 say(mlink->file, "&fopen"); 1275 return; 1276 } 1277 1278 line = NULL; 1279 linesz = 0; 1280 1281 /* Parse the section number from the header line. */ 1282 1283 while (getline(&line, &linesz, stream) != -1) { 1284 if (*line == '\n') 1285 continue; 1286 if ((sec = strchr(line, '(')) == NULL) 1287 break; 1288 if ((p = strchr(++sec, ')')) == NULL) 1289 break; 1290 free(mpage->sec); 1291 mpage->sec = mandoc_strndup(sec, p - sec); 1292 if (warnings && *mlink->dsec != '\0' && 1293 strcasecmp(mpage->sec, mlink->dsec)) 1294 say(mlink->file, 1295 "Section \"%s\" manual in %s directory", 1296 mpage->sec, mlink->dsec); 1297 break; 1298 } 1299 1300 /* Skip to first blank line. */ 1301 1302 while (line == NULL || *line != '\n') 1303 if (getline(&line, &linesz, stream) == -1) 1304 break; 1305 1306 /* 1307 * Assume the first line that is not indented 1308 * is the first section header. Skip to it. 1309 */ 1310 1311 while (getline(&line, &linesz, stream) != -1) 1312 if (*line != '\n' && *line != ' ') 1313 break; 1314 1315 /* 1316 * Read up until the next section into a buffer. 1317 * Strip the leading and trailing newline from each read line, 1318 * appending a trailing space. 1319 * Ignore empty (whitespace-only) lines. 1320 */ 1321 1322 titlesz = 0; 1323 title = NULL; 1324 1325 while ((len = getline(&line, &linesz, stream)) != -1) { 1326 if (*line != ' ') 1327 break; 1328 offs = 0; 1329 while (isspace((unsigned char)line[offs])) 1330 offs++; 1331 if (line[offs] == '\0') 1332 continue; 1333 title = mandoc_realloc(title, titlesz + len - offs); 1334 memcpy(title + titlesz, line + offs, len - offs); 1335 titlesz += len - offs; 1336 title[titlesz - 1] = ' '; 1337 } 1338 free(line); 1339 1340 /* 1341 * If no page content can be found, or the input line 1342 * is already the next section header, or there is no 1343 * trailing newline, reuse the page title as the page 1344 * description. 1345 */ 1346 1347 if (NULL == title || '\0' == *title) { 1348 if (warnings) 1349 say(mlink->file, "Cannot find NAME section"); 1350 fclose(stream); 1351 free(title); 1352 return; 1353 } 1354 1355 title[titlesz - 1] = '\0'; 1356 1357 /* 1358 * Skip to the first dash. 1359 * Use the remaining line as the description (no more than 70 1360 * bytes). 1361 */ 1362 1363 if (NULL != (p = strstr(title, "- "))) { 1364 for (p += 2; ' ' == *p || '\b' == *p; p++) 1365 /* Skip to next word. */ ; 1366 } else { 1367 if (warnings) 1368 say(mlink->file, "No dash in title line, " 1369 "reusing \"%s\" as one-line description", title); 1370 p = title; 1371 } 1372 1373 plen = strlen(p); 1374 1375 /* Strip backspace-encoding from line. */ 1376 1377 while (NULL != (line = memchr(p, '\b', plen))) { 1378 len = line - p; 1379 if (0 == len) { 1380 memmove(line, line + 1, plen--); 1381 continue; 1382 } 1383 memmove(line - 1, line + 1, plen - len); 1384 plen -= 2; 1385 } 1386 1387 mpage->desc = mandoc_strdup(p); 1388 fclose(stream); 1389 free(title); 1390 } 1391 1392 /* 1393 * Put a type/word pair into the word database for this particular file. 1394 */ 1395 static void 1396 putkey(const struct mpage *mpage, char *value, uint64_t type) 1397 { 1398 putkeys(mpage, value, strlen(value), type); 1399 } 1400 1401 /* 1402 * Grok all nodes at or below a certain mdoc node into putkey(). 1403 */ 1404 static void 1405 putmdockey(const struct mpage *mpage, 1406 const struct roff_node *n, uint64_t m, int taboo) 1407 { 1408 1409 for ( ; NULL != n; n = n->next) { 1410 if (n->flags & taboo) 1411 continue; 1412 if (NULL != n->child) 1413 putmdockey(mpage, n->child, m, taboo); 1414 if (n->type == ROFFT_TEXT) 1415 putkey(mpage, n->string, m); 1416 } 1417 } 1418 1419 static void 1420 parse_man(struct mpage *mpage, const struct roff_meta *meta, 1421 const struct roff_node *n) 1422 { 1423 const struct roff_node *head, *body; 1424 char *start, *title; 1425 char byte; 1426 size_t sz; 1427 1428 if (n == NULL) 1429 return; 1430 1431 /* 1432 * We're only searching for one thing: the first text child in 1433 * the BODY of a NAME section. Since we don't keep track of 1434 * sections in -man, run some hoops to find out whether we're in 1435 * the correct section or not. 1436 */ 1437 1438 if (n->type == ROFFT_BODY && n->tok == MAN_SH) { 1439 body = n; 1440 if ((head = body->parent->head) != NULL && 1441 (head = head->child) != NULL && 1442 head->next == NULL && 1443 head->type == ROFFT_TEXT && 1444 strcmp(head->string, "NAME") == 0 && 1445 body->child != NULL) { 1446 1447 /* 1448 * Suck the entire NAME section into memory. 1449 * Yes, we might run away. 1450 * But too many manuals have big, spread-out 1451 * NAME sections over many lines. 1452 */ 1453 1454 title = NULL; 1455 deroff(&title, body); 1456 if (NULL == title) 1457 return; 1458 1459 /* 1460 * Go through a special heuristic dance here. 1461 * Conventionally, one or more manual names are 1462 * comma-specified prior to a whitespace, then a 1463 * dash, then a description. Try to puzzle out 1464 * the name parts here. 1465 */ 1466 1467 start = title; 1468 for ( ;; ) { 1469 sz = strcspn(start, " ,"); 1470 if ('\0' == start[sz]) 1471 break; 1472 1473 byte = start[sz]; 1474 start[sz] = '\0'; 1475 1476 /* 1477 * Assume a stray trailing comma in the 1478 * name list if a name begins with a dash. 1479 */ 1480 1481 if ('-' == start[0] || 1482 ('\\' == start[0] && '-' == start[1])) 1483 break; 1484 1485 putkey(mpage, start, NAME_TITLE); 1486 if ( ! (mpage->name_head_done || 1487 strcasecmp(start, meta->title))) { 1488 putkey(mpage, start, NAME_HEAD); 1489 mpage->name_head_done = 1; 1490 } 1491 1492 if (' ' == byte) { 1493 start += sz + 1; 1494 break; 1495 } 1496 1497 assert(',' == byte); 1498 start += sz + 1; 1499 while (' ' == *start) 1500 start++; 1501 } 1502 1503 if (start == title) { 1504 putkey(mpage, start, NAME_TITLE); 1505 if ( ! (mpage->name_head_done || 1506 strcasecmp(start, meta->title))) { 1507 putkey(mpage, start, NAME_HEAD); 1508 mpage->name_head_done = 1; 1509 } 1510 free(title); 1511 return; 1512 } 1513 1514 while (isspace((unsigned char)*start)) 1515 start++; 1516 1517 if (0 == strncmp(start, "-", 1)) 1518 start += 1; 1519 else if (0 == strncmp(start, "\\-\\-", 4)) 1520 start += 4; 1521 else if (0 == strncmp(start, "\\-", 2)) 1522 start += 2; 1523 else if (0 == strncmp(start, "\\(en", 4)) 1524 start += 4; 1525 else if (0 == strncmp(start, "\\(em", 4)) 1526 start += 4; 1527 1528 while (' ' == *start) 1529 start++; 1530 1531 mpage->desc = mandoc_strdup(start); 1532 free(title); 1533 return; 1534 } 1535 } 1536 1537 for (n = n->child; n; n = n->next) { 1538 if (NULL != mpage->desc) 1539 break; 1540 parse_man(mpage, meta, n); 1541 } 1542 } 1543 1544 static void 1545 parse_mdoc(struct mpage *mpage, const struct roff_meta *meta, 1546 const struct roff_node *n) 1547 { 1548 1549 assert(NULL != n); 1550 for (n = n->child; NULL != n; n = n->next) { 1551 if (n->flags & mdocs[n->tok].taboo) 1552 continue; 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 (NULL != mdocs[n->tok].fp) 1560 if (0 == (*mdocs[n->tok].fp)(mpage, meta, n)) 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 assert(n->type != ROFFT_ROOT); 1568 continue; 1569 } 1570 if (NULL != n->child) 1571 parse_mdoc(mpage, meta, n); 1572 } 1573 } 1574 1575 static int 1576 parse_mdoc_Fd(struct mpage *mpage, const struct roff_meta *meta, 1577 const struct roff_node *n) 1578 { 1579 char *start, *end; 1580 size_t sz; 1581 1582 if (SEC_SYNOPSIS != n->sec || 1583 NULL == (n = n->child) || 1584 n->type != ROFFT_TEXT) 1585 return 0; 1586 1587 /* 1588 * Only consider those `Fd' macro fields that begin with an 1589 * "inclusion" token (versus, e.g., #define). 1590 */ 1591 1592 if (strcmp("#include", n->string)) 1593 return 0; 1594 1595 if ((n = n->next) == NULL || n->type != ROFFT_TEXT) 1596 return 0; 1597 1598 /* 1599 * Strip away the enclosing angle brackets and make sure we're 1600 * not zero-length. 1601 */ 1602 1603 start = n->string; 1604 if ('<' == *start || '"' == *start) 1605 start++; 1606 1607 if (0 == (sz = strlen(start))) 1608 return 0; 1609 1610 end = &start[(int)sz - 1]; 1611 if ('>' == *end || '"' == *end) 1612 end--; 1613 1614 if (end > start) 1615 putkeys(mpage, start, end - start + 1, TYPE_In); 1616 return 0; 1617 } 1618 1619 static void 1620 parse_mdoc_fname(struct mpage *mpage, const struct roff_node *n) 1621 { 1622 char *cp; 1623 size_t sz; 1624 1625 if (n->type != ROFFT_TEXT) 1626 return; 1627 1628 /* Skip function pointer punctuation. */ 1629 1630 cp = n->string; 1631 while (*cp == '(' || *cp == '*') 1632 cp++; 1633 sz = strcspn(cp, "()"); 1634 1635 putkeys(mpage, cp, sz, TYPE_Fn); 1636 if (n->sec == SEC_SYNOPSIS) 1637 putkeys(mpage, cp, sz, NAME_SYN); 1638 } 1639 1640 static int 1641 parse_mdoc_Fn(struct mpage *mpage, const struct roff_meta *meta, 1642 const struct roff_node *n) 1643 { 1644 1645 if (n->child == NULL) 1646 return 0; 1647 1648 parse_mdoc_fname(mpage, n->child); 1649 1650 for (n = n->child->next; n != NULL; n = n->next) 1651 if (n->type == ROFFT_TEXT) 1652 putkey(mpage, n->string, TYPE_Fa); 1653 1654 return 0; 1655 } 1656 1657 static int 1658 parse_mdoc_Fo(struct mpage *mpage, const struct roff_meta *meta, 1659 const struct roff_node *n) 1660 { 1661 1662 if (n->type != ROFFT_HEAD) 1663 return 1; 1664 1665 if (n->child != NULL) 1666 parse_mdoc_fname(mpage, n->child); 1667 1668 return 0; 1669 } 1670 1671 static int 1672 parse_mdoc_Va(struct mpage *mpage, const struct roff_meta *meta, 1673 const struct roff_node *n) 1674 { 1675 char *cp; 1676 1677 if (n->type != ROFFT_ELEM && n->type != ROFFT_BODY) 1678 return 0; 1679 1680 if (n->child != NULL && 1681 n->child->next == NULL && 1682 n->child->type == ROFFT_TEXT) 1683 return 1; 1684 1685 cp = NULL; 1686 deroff(&cp, n); 1687 if (cp != NULL) { 1688 putkey(mpage, cp, TYPE_Vt | (n->tok == MDOC_Va || 1689 n->type == ROFFT_BODY ? TYPE_Va : 0)); 1690 free(cp); 1691 } 1692 1693 return 0; 1694 } 1695 1696 static int 1697 parse_mdoc_Xr(struct mpage *mpage, const struct roff_meta *meta, 1698 const struct roff_node *n) 1699 { 1700 char *cp; 1701 1702 if (NULL == (n = n->child)) 1703 return 0; 1704 1705 if (NULL == n->next) { 1706 putkey(mpage, n->string, TYPE_Xr); 1707 return 0; 1708 } 1709 1710 mandoc_asprintf(&cp, "%s(%s)", n->string, n->next->string); 1711 putkey(mpage, cp, TYPE_Xr); 1712 free(cp); 1713 return 0; 1714 } 1715 1716 static int 1717 parse_mdoc_Nd(struct mpage *mpage, const struct roff_meta *meta, 1718 const struct roff_node *n) 1719 { 1720 1721 if (n->type == ROFFT_BODY) 1722 deroff(&mpage->desc, n); 1723 return 0; 1724 } 1725 1726 static int 1727 parse_mdoc_Nm(struct mpage *mpage, const struct roff_meta *meta, 1728 const struct roff_node *n) 1729 { 1730 1731 if (SEC_NAME == n->sec) 1732 putmdockey(mpage, n->child, NAME_TITLE, 0); 1733 else if (n->sec == SEC_SYNOPSIS && n->type == ROFFT_HEAD) { 1734 if (n->child == NULL) 1735 putkey(mpage, meta->name, NAME_SYN); 1736 else 1737 putmdockey(mpage, n->child, NAME_SYN, 0); 1738 } 1739 if ( ! (mpage->name_head_done || 1740 n->child == NULL || n->child->string == NULL || 1741 strcasecmp(n->child->string, meta->title))) { 1742 putkey(mpage, n->child->string, NAME_HEAD); 1743 mpage->name_head_done = 1; 1744 } 1745 return 0; 1746 } 1747 1748 static int 1749 parse_mdoc_Sh(struct mpage *mpage, const struct roff_meta *meta, 1750 const struct roff_node *n) 1751 { 1752 1753 return n->sec == SEC_CUSTOM && n->type == ROFFT_HEAD; 1754 } 1755 1756 static int 1757 parse_mdoc_head(struct mpage *mpage, const struct roff_meta *meta, 1758 const struct roff_node *n) 1759 { 1760 1761 return n->type == ROFFT_HEAD; 1762 } 1763 1764 /* 1765 * Add a string to the hash table for the current manual. 1766 * Each string has a bitmask telling which macros it belongs to. 1767 * When we finish the manual, we'll dump the table. 1768 */ 1769 static void 1770 putkeys(const struct mpage *mpage, char *cp, size_t sz, uint64_t v) 1771 { 1772 struct ohash *htab; 1773 struct str *s; 1774 const char *end; 1775 unsigned int slot; 1776 int i, mustfree; 1777 1778 if (0 == sz) 1779 return; 1780 1781 mustfree = render_string(&cp, &sz); 1782 1783 if (TYPE_Nm & v) { 1784 htab = &names; 1785 v &= name_mask; 1786 if (v & NAME_FIRST) 1787 name_mask &= ~NAME_FIRST; 1788 if (debug > 1) 1789 say(mpage->mlinks->file, 1790 "Adding name %*s, bits=0x%llx", (int)sz, cp, 1791 (unsigned long long)v); 1792 } else { 1793 htab = &strings; 1794 if (debug > 1) 1795 for (i = 0; i < KEY_MAX; i++) 1796 if ((uint64_t)1 << i & v) 1797 say(mpage->mlinks->file, 1798 "Adding key %s=%*s", 1799 mansearch_keynames[i], (int)sz, cp); 1800 } 1801 1802 end = cp + sz; 1803 slot = ohash_qlookupi(htab, cp, &end); 1804 s = ohash_find(htab, slot); 1805 1806 if (NULL != s && mpage == s->mpage) { 1807 s->mask |= v; 1808 return; 1809 } else if (NULL == s) { 1810 s = mandoc_calloc(1, sizeof(struct str) + sz + 1); 1811 memcpy(s->key, cp, sz); 1812 ohash_insert(htab, slot, s); 1813 } 1814 s->mpage = mpage; 1815 s->mask = v; 1816 1817 if (mustfree) 1818 free(cp); 1819 } 1820 1821 /* 1822 * Take a Unicode codepoint and produce its UTF-8 encoding. 1823 * This isn't the best way to do this, but it works. 1824 * The magic numbers are from the UTF-8 packaging. 1825 * They're not as scary as they seem: read the UTF-8 spec for details. 1826 */ 1827 static size_t 1828 utf8(unsigned int cp, char out[7]) 1829 { 1830 size_t rc; 1831 1832 rc = 0; 1833 if (cp <= 0x0000007F) { 1834 rc = 1; 1835 out[0] = (char)cp; 1836 } else if (cp <= 0x000007FF) { 1837 rc = 2; 1838 out[0] = (cp >> 6 & 31) | 192; 1839 out[1] = (cp & 63) | 128; 1840 } else if (cp <= 0x0000FFFF) { 1841 rc = 3; 1842 out[0] = (cp >> 12 & 15) | 224; 1843 out[1] = (cp >> 6 & 63) | 128; 1844 out[2] = (cp & 63) | 128; 1845 } else if (cp <= 0x001FFFFF) { 1846 rc = 4; 1847 out[0] = (cp >> 18 & 7) | 240; 1848 out[1] = (cp >> 12 & 63) | 128; 1849 out[2] = (cp >> 6 & 63) | 128; 1850 out[3] = (cp & 63) | 128; 1851 } else if (cp <= 0x03FFFFFF) { 1852 rc = 5; 1853 out[0] = (cp >> 24 & 3) | 248; 1854 out[1] = (cp >> 18 & 63) | 128; 1855 out[2] = (cp >> 12 & 63) | 128; 1856 out[3] = (cp >> 6 & 63) | 128; 1857 out[4] = (cp & 63) | 128; 1858 } else if (cp <= 0x7FFFFFFF) { 1859 rc = 6; 1860 out[0] = (cp >> 30 & 1) | 252; 1861 out[1] = (cp >> 24 & 63) | 128; 1862 out[2] = (cp >> 18 & 63) | 128; 1863 out[3] = (cp >> 12 & 63) | 128; 1864 out[4] = (cp >> 6 & 63) | 128; 1865 out[5] = (cp & 63) | 128; 1866 } else 1867 return 0; 1868 1869 out[rc] = '\0'; 1870 return rc; 1871 } 1872 1873 /* 1874 * If the string contains escape sequences, 1875 * replace it with an allocated rendering and return 1, 1876 * such that the caller can free it after use. 1877 * Otherwise, do nothing and return 0. 1878 */ 1879 static int 1880 render_string(char **public, size_t *psz) 1881 { 1882 const char *src, *scp, *addcp, *seq; 1883 char *dst; 1884 size_t ssz, dsz, addsz; 1885 char utfbuf[7], res[6]; 1886 int seqlen, unicode; 1887 1888 res[0] = '\\'; 1889 res[1] = '\t'; 1890 res[2] = ASCII_NBRSP; 1891 res[3] = ASCII_HYPH; 1892 res[4] = ASCII_BREAK; 1893 res[5] = '\0'; 1894 1895 src = scp = *public; 1896 ssz = *psz; 1897 dst = NULL; 1898 dsz = 0; 1899 1900 while (scp < src + *psz) { 1901 1902 /* Leave normal characters unchanged. */ 1903 1904 if (strchr(res, *scp) == NULL) { 1905 if (dst != NULL) 1906 dst[dsz++] = *scp; 1907 scp++; 1908 continue; 1909 } 1910 1911 /* 1912 * Found something that requires replacing, 1913 * make sure we have a destination buffer. 1914 */ 1915 1916 if (dst == NULL) { 1917 dst = mandoc_malloc(ssz + 1); 1918 dsz = scp - src; 1919 memcpy(dst, src, dsz); 1920 } 1921 1922 /* Handle single-char special characters. */ 1923 1924 switch (*scp) { 1925 case '\\': 1926 break; 1927 case '\t': 1928 case ASCII_NBRSP: 1929 dst[dsz++] = ' '; 1930 scp++; 1931 continue; 1932 case ASCII_HYPH: 1933 dst[dsz++] = '-'; 1934 /* FALLTHROUGH */ 1935 case ASCII_BREAK: 1936 scp++; 1937 continue; 1938 default: 1939 abort(); 1940 } 1941 1942 /* 1943 * Found an escape sequence. 1944 * Read past the slash, then parse it. 1945 * Ignore everything except characters. 1946 */ 1947 1948 scp++; 1949 if (mandoc_escape(&scp, &seq, &seqlen) != ESCAPE_SPECIAL) 1950 continue; 1951 1952 /* 1953 * Render the special character 1954 * as either UTF-8 or ASCII. 1955 */ 1956 1957 if (write_utf8) { 1958 unicode = mchars_spec2cp(seq, seqlen); 1959 if (unicode <= 0) 1960 continue; 1961 addsz = utf8(unicode, utfbuf); 1962 if (addsz == 0) 1963 continue; 1964 addcp = utfbuf; 1965 } else { 1966 addcp = mchars_spec2str(seq, seqlen, &addsz); 1967 if (addcp == NULL) 1968 continue; 1969 if (*addcp == ASCII_NBRSP) { 1970 addcp = " "; 1971 addsz = 1; 1972 } 1973 } 1974 1975 /* Copy the rendered glyph into the stream. */ 1976 1977 ssz += addsz; 1978 dst = mandoc_realloc(dst, ssz + 1); 1979 memcpy(dst + dsz, addcp, addsz); 1980 dsz += addsz; 1981 } 1982 if (dst != NULL) { 1983 *public = dst; 1984 *psz = dsz; 1985 } 1986 1987 /* Trim trailing whitespace and NUL-terminate. */ 1988 1989 while (*psz > 0 && (*public)[*psz - 1] == ' ') 1990 --*psz; 1991 if (dst != NULL) { 1992 (*public)[*psz] = '\0'; 1993 return 1; 1994 } else 1995 return 0; 1996 } 1997 1998 static void 1999 dbadd_mlink(const struct mlink *mlink) 2000 { 2001 dba_page_alias(mlink->mpage->dba, mlink->name, NAME_FILE); 2002 dba_page_add(mlink->mpage->dba, DBP_SECT, mlink->dsec); 2003 dba_page_add(mlink->mpage->dba, DBP_SECT, mlink->fsec); 2004 dba_page_add(mlink->mpage->dba, DBP_ARCH, mlink->arch); 2005 dba_page_add(mlink->mpage->dba, DBP_FILE, mlink->file); 2006 } 2007 2008 /* 2009 * Flush the current page's terms (and their bits) into the database. 2010 * Also, handle escape sequences at the last possible moment. 2011 */ 2012 static void 2013 dbadd(struct dba *dba, struct mpage *mpage) 2014 { 2015 struct mlink *mlink; 2016 struct str *key; 2017 char *cp; 2018 uint64_t mask; 2019 size_t i; 2020 unsigned int slot; 2021 int mustfree; 2022 2023 mlink = mpage->mlinks; 2024 2025 if (nodb) { 2026 for (key = ohash_first(&names, &slot); NULL != key; 2027 key = ohash_next(&names, &slot)) 2028 free(key); 2029 for (key = ohash_first(&strings, &slot); NULL != key; 2030 key = ohash_next(&strings, &slot)) 2031 free(key); 2032 if (0 == debug) 2033 return; 2034 while (NULL != mlink) { 2035 fputs(mlink->name, stdout); 2036 if (NULL == mlink->next || 2037 strcmp(mlink->dsec, mlink->next->dsec) || 2038 strcmp(mlink->fsec, mlink->next->fsec) || 2039 strcmp(mlink->arch, mlink->next->arch)) { 2040 putchar('('); 2041 if ('\0' == *mlink->dsec) 2042 fputs(mlink->fsec, stdout); 2043 else 2044 fputs(mlink->dsec, stdout); 2045 if ('\0' != *mlink->arch) 2046 printf("/%s", mlink->arch); 2047 putchar(')'); 2048 } 2049 mlink = mlink->next; 2050 if (NULL != mlink) 2051 fputs(", ", stdout); 2052 } 2053 printf(" - %s\n", mpage->desc); 2054 return; 2055 } 2056 2057 if (debug) 2058 say(mlink->file, "Adding to database"); 2059 2060 cp = mpage->desc; 2061 i = strlen(cp); 2062 mustfree = render_string(&cp, &i); 2063 mpage->dba = dba_page_new(dba->pages, 2064 *mpage->arch == '\0' ? mlink->arch : mpage->arch, 2065 cp, mlink->file, mpage->form); 2066 if (mustfree) 2067 free(cp); 2068 dba_page_add(mpage->dba, DBP_SECT, mpage->sec); 2069 2070 while (mlink != NULL) { 2071 dbadd_mlink(mlink); 2072 mlink = mlink->next; 2073 } 2074 2075 for (key = ohash_first(&names, &slot); NULL != key; 2076 key = ohash_next(&names, &slot)) { 2077 assert(key->mpage == mpage); 2078 dba_page_alias(mpage->dba, key->key, key->mask); 2079 free(key); 2080 } 2081 for (key = ohash_first(&strings, &slot); NULL != key; 2082 key = ohash_next(&strings, &slot)) { 2083 assert(key->mpage == mpage); 2084 i = 0; 2085 for (mask = TYPE_Xr; mask <= TYPE_Lb; mask *= 2) { 2086 if (key->mask & mask) 2087 dba_macro_add(dba->macros, i, 2088 key->key, mpage->dba); 2089 i++; 2090 } 2091 free(key); 2092 } 2093 } 2094 2095 static void 2096 dbprune(struct dba *dba) 2097 { 2098 struct dba_array *page, *files; 2099 char *file; 2100 2101 dba_array_FOREACH(dba->pages, page) { 2102 files = dba_array_get(page, DBP_FILE); 2103 dba_array_FOREACH(files, file) { 2104 if (*file < ' ') 2105 file++; 2106 if (ohash_find(&mlinks, ohash_qlookup(&mlinks, 2107 file)) != NULL) { 2108 if (debug) 2109 say(file, "Deleting from database"); 2110 dba_array_del(dba->pages); 2111 break; 2112 } 2113 } 2114 } 2115 } 2116 2117 /* 2118 * Write the database from memory to disk. 2119 */ 2120 static void 2121 dbwrite(struct dba *dba) 2122 { 2123 char tfn[32]; 2124 int status; 2125 pid_t child; 2126 2127 if (dba_write(MANDOC_DB "~", dba) != -1) { 2128 if (rename(MANDOC_DB "~", MANDOC_DB) == -1) { 2129 exitcode = (int)MANDOCLEVEL_SYSERR; 2130 say(MANDOC_DB, "&rename"); 2131 unlink(MANDOC_DB "~"); 2132 } 2133 return; 2134 } 2135 2136 (void)strlcpy(tfn, "/tmp/mandocdb.XXXXXXXX", sizeof(tfn)); 2137 if (mkdtemp(tfn) == NULL) { 2138 exitcode = (int)MANDOCLEVEL_SYSERR; 2139 say("", "&%s", tfn); 2140 return; 2141 } 2142 2143 (void)strlcat(tfn, "/" MANDOC_DB, sizeof(tfn)); 2144 if (dba_write(tfn, dba) == -1) { 2145 exitcode = (int)MANDOCLEVEL_SYSERR; 2146 say(tfn, "&dba_write"); 2147 goto out; 2148 } 2149 2150 switch (child = fork()) { 2151 case -1: 2152 exitcode = (int)MANDOCLEVEL_SYSERR; 2153 say("", "&fork cmp"); 2154 return; 2155 case 0: 2156 execlp("cmp", "cmp", "-s", tfn, MANDOC_DB, (char *)NULL); 2157 say("", "&exec cmp"); 2158 exit(0); 2159 default: 2160 break; 2161 } 2162 if (waitpid(child, &status, 0) == -1) { 2163 exitcode = (int)MANDOCLEVEL_SYSERR; 2164 say("", "&wait cmp"); 2165 } else if (WIFSIGNALED(status)) { 2166 exitcode = (int)MANDOCLEVEL_SYSERR; 2167 say("", "cmp died from signal %d", WTERMSIG(status)); 2168 } else if (WEXITSTATUS(status)) { 2169 exitcode = (int)MANDOCLEVEL_SYSERR; 2170 say(MANDOC_DB, 2171 "Data changed, but cannot replace database"); 2172 } 2173 2174 out: 2175 *strrchr(tfn, '/') = '\0'; 2176 switch (child = fork()) { 2177 case -1: 2178 exitcode = (int)MANDOCLEVEL_SYSERR; 2179 say("", "&fork rm"); 2180 return; 2181 case 0: 2182 execlp("rm", "rm", "-rf", tfn, (char *)NULL); 2183 say("", "&exec rm"); 2184 exit((int)MANDOCLEVEL_SYSERR); 2185 default: 2186 break; 2187 } 2188 if (waitpid(child, &status, 0) == -1) { 2189 exitcode = (int)MANDOCLEVEL_SYSERR; 2190 say("", "&wait rm"); 2191 } else if (WIFSIGNALED(status) || WEXITSTATUS(status)) { 2192 exitcode = (int)MANDOCLEVEL_SYSERR; 2193 say("", "%s: Cannot remove temporary directory", tfn); 2194 } 2195 } 2196 2197 static int 2198 set_basedir(const char *targetdir, int report_baddir) 2199 { 2200 static char startdir[PATH_MAX]; 2201 static int getcwd_status; /* 1 = ok, 2 = failure */ 2202 static int chdir_status; /* 1 = changed directory */ 2203 char *cp; 2204 2205 /* 2206 * Remember the original working directory, if possible. 2207 * This will be needed if the second or a later directory 2208 * on the command line is given as a relative path. 2209 * Do not error out if the current directory is not 2210 * searchable: Maybe it won't be needed after all. 2211 */ 2212 if (0 == getcwd_status) { 2213 if (NULL == getcwd(startdir, sizeof(startdir))) { 2214 getcwd_status = 2; 2215 (void)strlcpy(startdir, strerror(errno), 2216 sizeof(startdir)); 2217 } else 2218 getcwd_status = 1; 2219 } 2220 2221 /* 2222 * We are leaving the old base directory. 2223 * Do not use it any longer, not even for messages. 2224 */ 2225 *basedir = '\0'; 2226 2227 /* 2228 * If and only if the directory was changed earlier and 2229 * the next directory to process is given as a relative path, 2230 * first go back, or bail out if that is impossible. 2231 */ 2232 if (chdir_status && '/' != *targetdir) { 2233 if (2 == getcwd_status) { 2234 exitcode = (int)MANDOCLEVEL_SYSERR; 2235 say("", "getcwd: %s", startdir); 2236 return 0; 2237 } 2238 if (-1 == chdir(startdir)) { 2239 exitcode = (int)MANDOCLEVEL_SYSERR; 2240 say("", "&chdir %s", startdir); 2241 return 0; 2242 } 2243 } 2244 2245 /* 2246 * Always resolve basedir to the canonicalized absolute 2247 * pathname and append a trailing slash, such that 2248 * we can reliably check whether files are inside. 2249 */ 2250 if (NULL == realpath(targetdir, basedir)) { 2251 if (report_baddir || errno != ENOENT) { 2252 exitcode = (int)MANDOCLEVEL_BADARG; 2253 say("", "&%s: realpath", targetdir); 2254 } 2255 return 0; 2256 } else if (-1 == chdir(basedir)) { 2257 if (report_baddir || errno != ENOENT) { 2258 exitcode = (int)MANDOCLEVEL_BADARG; 2259 say("", "&chdir"); 2260 } 2261 return 0; 2262 } 2263 chdir_status = 1; 2264 cp = strchr(basedir, '\0'); 2265 if ('/' != cp[-1]) { 2266 if (cp - basedir >= PATH_MAX - 1) { 2267 exitcode = (int)MANDOCLEVEL_SYSERR; 2268 say("", "Filename too long"); 2269 return 0; 2270 } 2271 *cp++ = '/'; 2272 *cp = '\0'; 2273 } 2274 return 1; 2275 } 2276 2277 static void 2278 say(const char *file, const char *format, ...) 2279 { 2280 va_list ap; 2281 int use_errno; 2282 2283 if ('\0' != *basedir) 2284 fprintf(stderr, "%s", basedir); 2285 if ('\0' != *basedir && '\0' != *file) 2286 fputc('/', stderr); 2287 if ('\0' != *file) 2288 fprintf(stderr, "%s", file); 2289 2290 use_errno = 1; 2291 if (NULL != format) { 2292 switch (*format) { 2293 case '&': 2294 format++; 2295 break; 2296 case '\0': 2297 format = NULL; 2298 break; 2299 default: 2300 use_errno = 0; 2301 break; 2302 } 2303 } 2304 if (NULL != format) { 2305 if ('\0' != *basedir || '\0' != *file) 2306 fputs(": ", stderr); 2307 va_start(ap, format); 2308 vfprintf(stderr, format, ap); 2309 va_end(ap); 2310 } 2311 if (use_errno) { 2312 if ('\0' != *basedir || '\0' != *file || NULL != format) 2313 fputs(": ", stderr); 2314 perror(NULL); 2315 } else 2316 fputc('\n', stderr); 2317 } 2318