1 /* $OpenBSD: mandocdb.c,v 1.186 2017/01/11 17:39:45 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 <sys/types.h> 20 #include <sys/stat.h> 21 #include <sys/wait.h> 22 23 #include <assert.h> 24 #include <ctype.h> 25 #include <err.h> 26 #include <errno.h> 27 #include <fcntl.h> 28 #include <fts.h> 29 #include <limits.h> 30 #include <stdarg.h> 31 #include <stddef.h> 32 #include <stdio.h> 33 #include <stdint.h> 34 #include <stdlib.h> 35 #include <string.h> 36 #include <unistd.h> 37 38 #include "mandoc_aux.h" 39 #include "mandoc_ohash.h" 40 #include "mandoc.h" 41 #include "roff.h" 42 #include "mdoc.h" 43 #include "man.h" 44 #include "manconf.h" 45 #include "mansearch.h" 46 #include "dba_array.h" 47 #include "dba.h" 48 49 extern const char *const mansearch_keynames[]; 50 51 enum op { 52 OP_DEFAULT = 0, /* new dbs from dir list or default config */ 53 OP_CONFFILE, /* new databases from custom config file */ 54 OP_UPDATE, /* delete/add entries in existing database */ 55 OP_DELETE, /* delete entries from existing database */ 56 OP_TEST /* change no databases, report potential problems */ 57 }; 58 59 struct str { 60 const struct mpage *mpage; /* if set, the owning parse */ 61 uint64_t mask; /* bitmask in sequence */ 62 char key[]; /* rendered text */ 63 }; 64 65 struct inodev { 66 ino_t st_ino; 67 dev_t st_dev; 68 }; 69 70 struct mpage { 71 struct inodev inodev; /* used for hashing routine */ 72 struct dba_array *dba; 73 char *sec; /* section from file content */ 74 char *arch; /* architecture from file content */ 75 char *title; /* title from file content */ 76 char *desc; /* description from file content */ 77 struct mpage *next; /* singly linked list */ 78 struct mlink *mlinks; /* singly linked list */ 79 int name_head_done; 80 enum form form; /* format from file content */ 81 }; 82 83 struct mlink { 84 char file[PATH_MAX]; /* filename rel. to manpath */ 85 char *dsec; /* section from directory */ 86 char *arch; /* architecture from directory */ 87 char *name; /* name from file name (not empty) */ 88 char *fsec; /* section from file name suffix */ 89 struct mlink *next; /* singly linked list */ 90 struct mpage *mpage; /* parent */ 91 int gzip; /* filename has a .gz suffix */ 92 enum form dform; /* format from directory */ 93 enum form fform; /* format from file name suffix */ 94 }; 95 96 typedef int (*mdoc_fp)(struct mpage *, const struct roff_meta *, 97 const struct roff_node *); 98 99 struct mdoc_handler { 100 mdoc_fp fp; /* optional handler */ 101 uint64_t mask; /* set unless handler returns 0 */ 102 int taboo; /* node flags that must not be set */ 103 }; 104 105 106 int mandocdb(int, char *[]); 107 108 static void dbadd(struct dba *, struct mpage *); 109 static void dbadd_mlink(const struct mlink *mlink); 110 static void dbprune(struct dba *); 111 static void dbwrite(struct dba *); 112 static void filescan(const char *); 113 static int fts_compare(const FTSENT **, const FTSENT **); 114 static void mlink_add(struct mlink *, const struct stat *); 115 static void mlink_check(struct mpage *, struct mlink *); 116 static void mlink_free(struct mlink *); 117 static void mlinks_undupe(struct mpage *); 118 static void mpages_free(void); 119 static void mpages_merge(struct dba *, struct mparse *); 120 static void parse_cat(struct mpage *, int); 121 static void parse_man(struct mpage *, const struct roff_meta *, 122 const struct roff_node *); 123 static void parse_mdoc(struct mpage *, const struct roff_meta *, 124 const struct roff_node *); 125 static int parse_mdoc_head(struct mpage *, const struct roff_meta *, 126 const struct roff_node *); 127 static int parse_mdoc_Fd(struct mpage *, const struct roff_meta *, 128 const struct roff_node *); 129 static void parse_mdoc_fname(struct mpage *, const struct roff_node *); 130 static int parse_mdoc_Fn(struct mpage *, const struct roff_meta *, 131 const struct roff_node *); 132 static int parse_mdoc_Fo(struct mpage *, const struct roff_meta *, 133 const struct roff_node *); 134 static int parse_mdoc_Nd(struct mpage *, const struct roff_meta *, 135 const struct roff_node *); 136 static int parse_mdoc_Nm(struct mpage *, const struct roff_meta *, 137 const struct roff_node *); 138 static int parse_mdoc_Sh(struct mpage *, const struct roff_meta *, 139 const struct roff_node *); 140 static int parse_mdoc_Va(struct mpage *, const struct roff_meta *, 141 const struct roff_node *); 142 static int parse_mdoc_Xr(struct mpage *, const struct roff_meta *, 143 const struct roff_node *); 144 static void putkey(const struct mpage *, char *, uint64_t); 145 static void putkeys(const struct mpage *, char *, size_t, uint64_t); 146 static void putmdockey(const struct mpage *, 147 const struct roff_node *, uint64_t, int); 148 static int render_string(char **, size_t *); 149 static void say(const char *, const char *, ...) 150 __attribute__((__format__ (printf, 2, 3))); 151 static int set_basedir(const char *, int); 152 static int treescan(void); 153 static size_t utf8(unsigned int, char [7]); 154 155 static int nodb; /* no database changes */ 156 static int mparse_options; /* abort the parse early */ 157 static int use_all; /* use all found files */ 158 static int debug; /* print what we're doing */ 159 static int warnings; /* warn about crap */ 160 static int write_utf8; /* write UTF-8 output; else ASCII */ 161 static int exitcode; /* to be returned by main */ 162 static enum op op; /* operational mode */ 163 static char basedir[PATH_MAX]; /* current base directory */ 164 static struct mpage *mpage_head; /* list of distinct manual pages */ 165 static struct ohash mpages; /* table of distinct manual pages */ 166 static struct ohash mlinks; /* table of directory entries */ 167 static struct ohash names; /* table of all names */ 168 static struct ohash strings; /* table of all strings */ 169 static uint64_t name_mask; 170 171 static const struct mdoc_handler mdocs[MDOC_MAX] = { 172 { NULL, 0, 0 }, /* Ap */ 173 { NULL, 0, NODE_NOPRT }, /* Dd */ 174 { NULL, 0, NODE_NOPRT }, /* Dt */ 175 { NULL, 0, NODE_NOPRT }, /* Os */ 176 { parse_mdoc_Sh, TYPE_Sh, 0 }, /* Sh */ 177 { parse_mdoc_head, TYPE_Ss, 0 }, /* Ss */ 178 { NULL, 0, 0 }, /* Pp */ 179 { NULL, 0, 0 }, /* D1 */ 180 { NULL, 0, 0 }, /* Dl */ 181 { NULL, 0, 0 }, /* Bd */ 182 { NULL, 0, 0 }, /* Ed */ 183 { NULL, 0, 0 }, /* Bl */ 184 { NULL, 0, 0 }, /* El */ 185 { NULL, 0, 0 }, /* It */ 186 { NULL, 0, 0 }, /* Ad */ 187 { NULL, TYPE_An, 0 }, /* An */ 188 { NULL, TYPE_Ar, 0 }, /* Ar */ 189 { NULL, TYPE_Cd, 0 }, /* Cd */ 190 { NULL, TYPE_Cm, 0 }, /* Cm */ 191 { NULL, TYPE_Dv, 0 }, /* Dv */ 192 { NULL, TYPE_Er, 0 }, /* Er */ 193 { NULL, TYPE_Ev, 0 }, /* Ev */ 194 { NULL, 0, 0 }, /* Ex */ 195 { NULL, TYPE_Fa, 0 }, /* Fa */ 196 { parse_mdoc_Fd, 0, 0 }, /* Fd */ 197 { NULL, TYPE_Fl, 0 }, /* Fl */ 198 { parse_mdoc_Fn, 0, 0 }, /* Fn */ 199 { NULL, TYPE_Ft, 0 }, /* Ft */ 200 { NULL, TYPE_Ic, 0 }, /* Ic */ 201 { NULL, TYPE_In, 0 }, /* In */ 202 { NULL, TYPE_Li, 0 }, /* Li */ 203 { parse_mdoc_Nd, 0, 0 }, /* Nd */ 204 { parse_mdoc_Nm, 0, 0 }, /* Nm */ 205 { NULL, 0, 0 }, /* Op */ 206 { NULL, 0, 0 }, /* Ot */ 207 { NULL, TYPE_Pa, NODE_NOSRC }, /* Pa */ 208 { NULL, 0, 0 }, /* Rv */ 209 { NULL, TYPE_St, 0 }, /* St */ 210 { parse_mdoc_Va, TYPE_Va, 0 }, /* Va */ 211 { parse_mdoc_Va, TYPE_Vt, 0 }, /* Vt */ 212 { parse_mdoc_Xr, 0, 0 }, /* Xr */ 213 { NULL, 0, 0 }, /* %A */ 214 { NULL, 0, 0 }, /* %B */ 215 { NULL, 0, 0 }, /* %D */ 216 { NULL, 0, 0 }, /* %I */ 217 { NULL, 0, 0 }, /* %J */ 218 { NULL, 0, 0 }, /* %N */ 219 { NULL, 0, 0 }, /* %O */ 220 { NULL, 0, 0 }, /* %P */ 221 { NULL, 0, 0 }, /* %R */ 222 { NULL, 0, 0 }, /* %T */ 223 { NULL, 0, 0 }, /* %V */ 224 { NULL, 0, 0 }, /* Ac */ 225 { NULL, 0, 0 }, /* Ao */ 226 { NULL, 0, 0 }, /* Aq */ 227 { NULL, TYPE_At, 0 }, /* At */ 228 { NULL, 0, 0 }, /* Bc */ 229 { NULL, 0, 0 }, /* Bf */ 230 { NULL, 0, 0 }, /* Bo */ 231 { NULL, 0, 0 }, /* Bq */ 232 { NULL, TYPE_Bsx, NODE_NOSRC }, /* Bsx */ 233 { NULL, TYPE_Bx, NODE_NOSRC }, /* Bx */ 234 { NULL, 0, 0 }, /* Db */ 235 { NULL, 0, 0 }, /* Dc */ 236 { NULL, 0, 0 }, /* Do */ 237 { NULL, 0, 0 }, /* Dq */ 238 { NULL, 0, 0 }, /* Ec */ 239 { NULL, 0, 0 }, /* Ef */ 240 { NULL, TYPE_Em, 0 }, /* Em */ 241 { NULL, 0, 0 }, /* Eo */ 242 { NULL, TYPE_Fx, NODE_NOSRC }, /* Fx */ 243 { NULL, TYPE_Ms, 0 }, /* Ms */ 244 { NULL, 0, 0 }, /* No */ 245 { NULL, 0, 0 }, /* Ns */ 246 { NULL, TYPE_Nx, NODE_NOSRC }, /* Nx */ 247 { NULL, TYPE_Ox, NODE_NOSRC }, /* Ox */ 248 { NULL, 0, 0 }, /* Pc */ 249 { NULL, 0, 0 }, /* Pf */ 250 { NULL, 0, 0 }, /* Po */ 251 { NULL, 0, 0 }, /* Pq */ 252 { NULL, 0, 0 }, /* Qc */ 253 { NULL, 0, 0 }, /* Ql */ 254 { NULL, 0, 0 }, /* Qo */ 255 { NULL, 0, 0 }, /* Qq */ 256 { NULL, 0, 0 }, /* Re */ 257 { NULL, 0, 0 }, /* Rs */ 258 { NULL, 0, 0 }, /* Sc */ 259 { NULL, 0, 0 }, /* So */ 260 { NULL, 0, 0 }, /* Sq */ 261 { NULL, 0, 0 }, /* Sm */ 262 { NULL, 0, 0 }, /* Sx */ 263 { NULL, TYPE_Sy, 0 }, /* Sy */ 264 { NULL, TYPE_Tn, 0 }, /* Tn */ 265 { NULL, 0, NODE_NOSRC }, /* Ux */ 266 { NULL, 0, 0 }, /* Xc */ 267 { NULL, 0, 0 }, /* Xo */ 268 { parse_mdoc_Fo, 0, 0 }, /* Fo */ 269 { NULL, 0, 0 }, /* Fc */ 270 { NULL, 0, 0 }, /* Oo */ 271 { NULL, 0, 0 }, /* Oc */ 272 { NULL, 0, 0 }, /* Bk */ 273 { NULL, 0, 0 }, /* Ek */ 274 { NULL, 0, 0 }, /* Bt */ 275 { NULL, 0, 0 }, /* Hf */ 276 { NULL, 0, 0 }, /* Fr */ 277 { NULL, 0, 0 }, /* Ud */ 278 { NULL, TYPE_Lb, NODE_NOSRC }, /* Lb */ 279 { NULL, 0, 0 }, /* Lp */ 280 { NULL, TYPE_Lk, 0 }, /* Lk */ 281 { NULL, TYPE_Mt, NODE_NOSRC }, /* Mt */ 282 { NULL, 0, 0 }, /* Brq */ 283 { NULL, 0, 0 }, /* Bro */ 284 { NULL, 0, 0 }, /* Brc */ 285 { NULL, 0, 0 }, /* %C */ 286 { NULL, 0, 0 }, /* Es */ 287 { NULL, 0, 0 }, /* En */ 288 { NULL, TYPE_Dx, NODE_NOSRC }, /* Dx */ 289 { NULL, 0, 0 }, /* %Q */ 290 { NULL, 0, 0 }, /* br */ 291 { NULL, 0, 0 }, /* sp */ 292 { NULL, 0, 0 }, /* %U */ 293 { NULL, 0, 0 }, /* Ta */ 294 { NULL, 0, 0 }, /* ll */ 295 }; 296 297 298 int 299 mandocdb(int argc, char *argv[]) 300 { 301 struct manconf conf; 302 struct mparse *mp; 303 struct dba *dba; 304 const char *path_arg, *progname; 305 size_t j, sz; 306 int ch, i; 307 308 if (pledge("stdio rpath wpath cpath fattr flock proc exec", NULL) == -1) { 309 warn("pledge"); 310 return (int)MANDOCLEVEL_SYSERR; 311 } 312 313 memset(&conf, 0, sizeof(conf)); 314 315 /* 316 * We accept a few different invocations. 317 * The CHECKOP macro makes sure that invocation styles don't 318 * clobber each other. 319 */ 320 #define CHECKOP(_op, _ch) do \ 321 if (OP_DEFAULT != (_op)) { \ 322 warnx("-%c: Conflicting option", (_ch)); \ 323 goto usage; \ 324 } while (/*CONSTCOND*/0) 325 326 path_arg = NULL; 327 op = OP_DEFAULT; 328 329 while (-1 != (ch = getopt(argc, argv, "aC:Dd:npQT:tu:v"))) 330 switch (ch) { 331 case 'a': 332 use_all = 1; 333 break; 334 case 'C': 335 CHECKOP(op, ch); 336 path_arg = optarg; 337 op = OP_CONFFILE; 338 break; 339 case 'D': 340 debug++; 341 break; 342 case 'd': 343 CHECKOP(op, ch); 344 path_arg = optarg; 345 op = OP_UPDATE; 346 break; 347 case 'n': 348 nodb = 1; 349 break; 350 case 'p': 351 warnings = 1; 352 break; 353 case 'Q': 354 mparse_options |= MPARSE_QUICK; 355 break; 356 case 'T': 357 if (strcmp(optarg, "utf8")) { 358 warnx("-T%s: Unsupported output format", 359 optarg); 360 goto usage; 361 } 362 write_utf8 = 1; 363 break; 364 case 't': 365 CHECKOP(op, ch); 366 dup2(STDOUT_FILENO, STDERR_FILENO); 367 op = OP_TEST; 368 nodb = warnings = 1; 369 break; 370 case 'u': 371 CHECKOP(op, ch); 372 path_arg = optarg; 373 op = OP_DELETE; 374 break; 375 case 'v': 376 /* Compatibility with espie@'s makewhatis. */ 377 break; 378 default: 379 goto usage; 380 } 381 382 argc -= optind; 383 argv += optind; 384 385 if (nodb) { 386 if (pledge("stdio rpath", NULL) == -1) { 387 warn("pledge"); 388 return (int)MANDOCLEVEL_SYSERR; 389 } 390 } 391 392 if (OP_CONFFILE == op && argc > 0) { 393 warnx("-C: Too many arguments"); 394 goto usage; 395 } 396 397 exitcode = (int)MANDOCLEVEL_OK; 398 mchars_alloc(); 399 mp = mparse_alloc(mparse_options, MANDOCLEVEL_BADARG, NULL, NULL); 400 mandoc_ohash_init(&mpages, 6, offsetof(struct mpage, inodev)); 401 mandoc_ohash_init(&mlinks, 6, offsetof(struct mlink, file)); 402 403 if (OP_UPDATE == op || OP_DELETE == op || OP_TEST == op) { 404 405 /* 406 * Most of these deal with a specific directory. 407 * Jump into that directory first. 408 */ 409 if (OP_TEST != op && 0 == set_basedir(path_arg, 1)) 410 goto out; 411 412 dba = nodb ? dba_new(128) : dba_read(MANDOC_DB); 413 if (dba != NULL) { 414 /* 415 * The existing database is usable. Process 416 * all files specified on the command-line. 417 */ 418 if (!nodb) { 419 if (pledge("stdio rpath wpath cpath fattr flock", NULL) == -1) { 420 warn("pledge"); 421 exitcode = (int)MANDOCLEVEL_SYSERR; 422 goto out; 423 } 424 } 425 use_all = 1; 426 for (i = 0; i < argc; i++) 427 filescan(argv[i]); 428 if (nodb == 0) 429 dbprune(dba); 430 } else { 431 /* Database missing or corrupt. */ 432 if (op != OP_UPDATE || errno != ENOENT) 433 say(MANDOC_DB, "%s: Automatically recreating" 434 " from scratch", strerror(errno)); 435 exitcode = (int)MANDOCLEVEL_OK; 436 op = OP_DEFAULT; 437 if (0 == treescan()) 438 goto out; 439 dba = dba_new(128); 440 } 441 if (OP_DELETE != op) 442 mpages_merge(dba, mp); 443 if (nodb == 0) 444 dbwrite(dba); 445 dba_free(dba); 446 } else { 447 /* 448 * If we have arguments, use them as our manpaths. 449 * If we don't, use man.conf(5). 450 */ 451 if (argc > 0) { 452 conf.manpath.paths = mandoc_reallocarray(NULL, 453 argc, sizeof(char *)); 454 conf.manpath.sz = (size_t)argc; 455 for (i = 0; i < argc; i++) 456 conf.manpath.paths[i] = mandoc_strdup(argv[i]); 457 } else 458 manconf_parse(&conf, path_arg, NULL, NULL); 459 460 if (conf.manpath.sz == 0) { 461 exitcode = (int)MANDOCLEVEL_BADARG; 462 say("", "Empty manpath"); 463 } 464 465 /* 466 * First scan the tree rooted at a base directory, then 467 * build a new database and finally move it into place. 468 * Ignore zero-length directories and strip trailing 469 * slashes. 470 */ 471 for (j = 0; j < conf.manpath.sz; j++) { 472 sz = strlen(conf.manpath.paths[j]); 473 if (sz && conf.manpath.paths[j][sz - 1] == '/') 474 conf.manpath.paths[j][--sz] = '\0'; 475 if (0 == sz) 476 continue; 477 478 if (j) { 479 mandoc_ohash_init(&mpages, 6, 480 offsetof(struct mpage, inodev)); 481 mandoc_ohash_init(&mlinks, 6, 482 offsetof(struct mlink, file)); 483 } 484 485 if ( ! set_basedir(conf.manpath.paths[j], argc > 0)) 486 continue; 487 if (0 == treescan()) 488 continue; 489 dba = dba_new(128); 490 mpages_merge(dba, mp); 491 if (nodb == 0) 492 dbwrite(dba); 493 dba_free(dba); 494 495 if (j + 1 < conf.manpath.sz) { 496 mpages_free(); 497 ohash_delete(&mpages); 498 ohash_delete(&mlinks); 499 } 500 } 501 } 502 out: 503 manconf_free(&conf); 504 mparse_free(mp); 505 mchars_free(); 506 mpages_free(); 507 ohash_delete(&mpages); 508 ohash_delete(&mlinks); 509 return exitcode; 510 usage: 511 progname = getprogname(); 512 fprintf(stderr, "usage: %s [-aDnpQ] [-C file] [-Tutf8]\n" 513 " %s [-aDnpQ] [-Tutf8] dir ...\n" 514 " %s [-DnpQ] [-Tutf8] -d dir [file ...]\n" 515 " %s [-Dnp] -u dir [file ...]\n" 516 " %s [-Q] -t file ...\n", 517 progname, progname, progname, progname, progname); 518 519 return (int)MANDOCLEVEL_BADARG; 520 } 521 522 /* 523 * To get a singly linked list in alpha order while inserting entries 524 * at the beginning, process directory entries in reverse alpha order. 525 */ 526 static int 527 fts_compare(const FTSENT **a, const FTSENT **b) 528 { 529 return -strcmp((*a)->fts_name, (*b)->fts_name); 530 } 531 532 /* 533 * Scan a directory tree rooted at "basedir" for manpages. 534 * We use fts(), scanning directory parts along the way for clues to our 535 * section and architecture. 536 * 537 * If use_all has been specified, grok all files. 538 * If not, sanitise paths to the following: 539 * 540 * [./]man*[/<arch>]/<name>.<section> 541 * or 542 * [./]cat<section>[/<arch>]/<name>.0 543 * 544 * TODO: accommodate for multi-language directories. 545 */ 546 static int 547 treescan(void) 548 { 549 char buf[PATH_MAX]; 550 FTS *f; 551 FTSENT *ff; 552 struct mlink *mlink; 553 int gzip; 554 enum form dform; 555 char *dsec, *arch, *fsec, *cp; 556 const char *path; 557 const char *argv[2]; 558 559 argv[0] = "."; 560 argv[1] = (char *)NULL; 561 562 f = fts_open((char * const *)argv, FTS_PHYSICAL | FTS_NOCHDIR, 563 fts_compare); 564 if (f == NULL) { 565 exitcode = (int)MANDOCLEVEL_SYSERR; 566 say("", "&fts_open"); 567 return 0; 568 } 569 570 dsec = arch = NULL; 571 dform = FORM_NONE; 572 573 while ((ff = fts_read(f)) != NULL) { 574 path = ff->fts_path + 2; 575 switch (ff->fts_info) { 576 577 /* 578 * Symbolic links require various sanity checks, 579 * then get handled just like regular files. 580 */ 581 case FTS_SL: 582 if (realpath(path, buf) == NULL) { 583 if (warnings) 584 say(path, "&realpath"); 585 continue; 586 } 587 if (strstr(buf, basedir) != buf) { 588 if (warnings) say("", 589 "%s: outside base directory", buf); 590 continue; 591 } 592 /* Use logical inode to avoid mpages dupe. */ 593 if (stat(path, ff->fts_statp) == -1) { 594 if (warnings) 595 say(path, "&stat"); 596 continue; 597 } 598 /* FALLTHROUGH */ 599 600 /* 601 * If we're a regular file, add an mlink by using the 602 * stored directory data and handling the filename. 603 */ 604 case FTS_F: 605 if ( ! strcmp(path, MANDOC_DB)) 606 continue; 607 if ( ! use_all && ff->fts_level < 2) { 608 if (warnings) 609 say(path, "Extraneous file"); 610 continue; 611 } 612 gzip = 0; 613 fsec = NULL; 614 while (fsec == NULL) { 615 fsec = strrchr(ff->fts_name, '.'); 616 if (fsec == NULL || strcmp(fsec+1, "gz")) 617 break; 618 gzip = 1; 619 *fsec = '\0'; 620 fsec = NULL; 621 } 622 if (fsec == NULL) { 623 if ( ! use_all) { 624 if (warnings) 625 say(path, 626 "No filename suffix"); 627 continue; 628 } 629 } else if ( ! strcmp(++fsec, "html")) { 630 if (warnings) 631 say(path, "Skip html"); 632 continue; 633 } else if ( ! strcmp(fsec, "ps")) { 634 if (warnings) 635 say(path, "Skip ps"); 636 continue; 637 } else if ( ! strcmp(fsec, "pdf")) { 638 if (warnings) 639 say(path, "Skip pdf"); 640 continue; 641 } else if ( ! use_all && 642 ((dform == FORM_SRC && 643 strncmp(fsec, dsec, strlen(dsec))) || 644 (dform == FORM_CAT && strcmp(fsec, "0")))) { 645 if (warnings) 646 say(path, "Wrong filename suffix"); 647 continue; 648 } else 649 fsec[-1] = '\0'; 650 651 mlink = mandoc_calloc(1, sizeof(struct mlink)); 652 if (strlcpy(mlink->file, path, 653 sizeof(mlink->file)) >= 654 sizeof(mlink->file)) { 655 say(path, "Filename too long"); 656 free(mlink); 657 continue; 658 } 659 mlink->dform = dform; 660 mlink->dsec = dsec; 661 mlink->arch = arch; 662 mlink->name = ff->fts_name; 663 mlink->fsec = fsec; 664 mlink->gzip = gzip; 665 mlink_add(mlink, ff->fts_statp); 666 continue; 667 668 case FTS_D: 669 case FTS_DP: 670 break; 671 672 default: 673 if (warnings) 674 say(path, "Not a regular file"); 675 continue; 676 } 677 678 switch (ff->fts_level) { 679 case 0: 680 /* Ignore the root directory. */ 681 break; 682 case 1: 683 /* 684 * This might contain manX/ or catX/. 685 * Try to infer this from the name. 686 * If we're not in use_all, enforce it. 687 */ 688 cp = ff->fts_name; 689 if (ff->fts_info == FTS_DP) { 690 dform = FORM_NONE; 691 dsec = NULL; 692 break; 693 } 694 695 if ( ! strncmp(cp, "man", 3)) { 696 dform = FORM_SRC; 697 dsec = cp + 3; 698 } else if ( ! strncmp(cp, "cat", 3)) { 699 dform = FORM_CAT; 700 dsec = cp + 3; 701 } else { 702 dform = FORM_NONE; 703 dsec = NULL; 704 } 705 706 if (dsec != NULL || use_all) 707 break; 708 709 if (warnings) 710 say(path, "Unknown directory part"); 711 fts_set(f, ff, FTS_SKIP); 712 break; 713 case 2: 714 /* 715 * Possibly our architecture. 716 * If we're descending, keep tabs on it. 717 */ 718 if (ff->fts_info != FTS_DP && dsec != NULL) 719 arch = ff->fts_name; 720 else 721 arch = NULL; 722 break; 723 default: 724 if (ff->fts_info == FTS_DP || use_all) 725 break; 726 if (warnings) 727 say(path, "Extraneous directory part"); 728 fts_set(f, ff, FTS_SKIP); 729 break; 730 } 731 } 732 733 fts_close(f); 734 return 1; 735 } 736 737 /* 738 * Add a file to the mlinks table. 739 * Do not verify that it's a "valid" looking manpage (we'll do that 740 * later). 741 * 742 * Try to infer the manual section, architecture, and page name from the 743 * path, assuming it looks like 744 * 745 * [./]man*[/<arch>]/<name>.<section> 746 * or 747 * [./]cat<section>[/<arch>]/<name>.0 748 * 749 * See treescan() for the fts(3) version of this. 750 */ 751 static void 752 filescan(const char *file) 753 { 754 char buf[PATH_MAX]; 755 struct stat st; 756 struct mlink *mlink; 757 char *p, *start; 758 759 assert(use_all); 760 761 if (0 == strncmp(file, "./", 2)) 762 file += 2; 763 764 /* 765 * We have to do lstat(2) before realpath(3) loses 766 * the information whether this is a symbolic link. 767 * We need to know that because for symbolic links, 768 * we want to use the orginal file name, while for 769 * regular files, we want to use the real path. 770 */ 771 if (-1 == lstat(file, &st)) { 772 exitcode = (int)MANDOCLEVEL_BADARG; 773 say(file, "&lstat"); 774 return; 775 } else if (0 == ((S_IFREG | S_IFLNK) & st.st_mode)) { 776 exitcode = (int)MANDOCLEVEL_BADARG; 777 say(file, "Not a regular file"); 778 return; 779 } 780 781 /* 782 * We have to resolve the file name to the real path 783 * in any case for the base directory check. 784 */ 785 if (NULL == realpath(file, buf)) { 786 exitcode = (int)MANDOCLEVEL_BADARG; 787 say(file, "&realpath"); 788 return; 789 } 790 791 if (OP_TEST == op) 792 start = buf; 793 else if (strstr(buf, basedir) == buf) 794 start = buf + strlen(basedir); 795 else { 796 exitcode = (int)MANDOCLEVEL_BADARG; 797 say("", "%s: outside base directory", buf); 798 return; 799 } 800 801 /* 802 * Now we are sure the file is inside our tree. 803 * If it is a symbolic link, ignore the real path 804 * and use the original name. 805 * This implies passing stuff like "cat1/../man1/foo.1" 806 * on the command line won't work. So don't do that. 807 * Note the stat(2) can still fail if the link target 808 * doesn't exist. 809 */ 810 if (S_IFLNK & st.st_mode) { 811 if (-1 == stat(buf, &st)) { 812 exitcode = (int)MANDOCLEVEL_BADARG; 813 say(file, "&stat"); 814 return; 815 } 816 if (strlcpy(buf, file, sizeof(buf)) >= sizeof(buf)) { 817 say(file, "Filename too long"); 818 return; 819 } 820 start = buf; 821 if (OP_TEST != op && strstr(buf, basedir) == buf) 822 start += strlen(basedir); 823 } 824 825 mlink = mandoc_calloc(1, sizeof(struct mlink)); 826 mlink->dform = FORM_NONE; 827 if (strlcpy(mlink->file, start, sizeof(mlink->file)) >= 828 sizeof(mlink->file)) { 829 say(start, "Filename too long"); 830 free(mlink); 831 return; 832 } 833 834 /* 835 * First try to guess our directory structure. 836 * If we find a separator, try to look for man* or cat*. 837 * If we find one of these and what's underneath is a directory, 838 * assume it's an architecture. 839 */ 840 if (NULL != (p = strchr(start, '/'))) { 841 *p++ = '\0'; 842 if (0 == strncmp(start, "man", 3)) { 843 mlink->dform = FORM_SRC; 844 mlink->dsec = start + 3; 845 } else if (0 == strncmp(start, "cat", 3)) { 846 mlink->dform = FORM_CAT; 847 mlink->dsec = start + 3; 848 } 849 850 start = p; 851 if (NULL != mlink->dsec && NULL != (p = strchr(start, '/'))) { 852 *p++ = '\0'; 853 mlink->arch = start; 854 start = p; 855 } 856 } 857 858 /* 859 * Now check the file suffix. 860 * Suffix of `.0' indicates a catpage, `.1-9' is a manpage. 861 */ 862 p = strrchr(start, '\0'); 863 while (p-- > start && '/' != *p && '.' != *p) 864 /* Loop. */ ; 865 866 if ('.' == *p) { 867 *p++ = '\0'; 868 mlink->fsec = p; 869 } 870 871 /* 872 * Now try to parse the name. 873 * Use the filename portion of the path. 874 */ 875 mlink->name = start; 876 if (NULL != (p = strrchr(start, '/'))) { 877 mlink->name = p + 1; 878 *p = '\0'; 879 } 880 mlink_add(mlink, &st); 881 } 882 883 static void 884 mlink_add(struct mlink *mlink, const struct stat *st) 885 { 886 struct inodev inodev; 887 struct mpage *mpage; 888 unsigned int slot; 889 890 assert(NULL != mlink->file); 891 892 mlink->dsec = mandoc_strdup(mlink->dsec ? mlink->dsec : ""); 893 mlink->arch = mandoc_strdup(mlink->arch ? mlink->arch : ""); 894 mlink->name = mandoc_strdup(mlink->name ? mlink->name : ""); 895 mlink->fsec = mandoc_strdup(mlink->fsec ? mlink->fsec : ""); 896 897 if ('0' == *mlink->fsec) { 898 free(mlink->fsec); 899 mlink->fsec = mandoc_strdup(mlink->dsec); 900 mlink->fform = FORM_CAT; 901 } else if ('1' <= *mlink->fsec && '9' >= *mlink->fsec) 902 mlink->fform = FORM_SRC; 903 else 904 mlink->fform = FORM_NONE; 905 906 slot = ohash_qlookup(&mlinks, mlink->file); 907 assert(NULL == ohash_find(&mlinks, slot)); 908 ohash_insert(&mlinks, slot, mlink); 909 910 memset(&inodev, 0, sizeof(inodev)); /* Clear padding. */ 911 inodev.st_ino = st->st_ino; 912 inodev.st_dev = st->st_dev; 913 slot = ohash_lookup_memory(&mpages, (char *)&inodev, 914 sizeof(struct inodev), inodev.st_ino); 915 mpage = ohash_find(&mpages, slot); 916 if (NULL == mpage) { 917 mpage = mandoc_calloc(1, sizeof(struct mpage)); 918 mpage->inodev.st_ino = inodev.st_ino; 919 mpage->inodev.st_dev = inodev.st_dev; 920 mpage->form = FORM_NONE; 921 mpage->next = mpage_head; 922 mpage_head = mpage; 923 ohash_insert(&mpages, slot, mpage); 924 } else 925 mlink->next = mpage->mlinks; 926 mpage->mlinks = mlink; 927 mlink->mpage = mpage; 928 } 929 930 static void 931 mlink_free(struct mlink *mlink) 932 { 933 934 free(mlink->dsec); 935 free(mlink->arch); 936 free(mlink->name); 937 free(mlink->fsec); 938 free(mlink); 939 } 940 941 static void 942 mpages_free(void) 943 { 944 struct mpage *mpage; 945 struct mlink *mlink; 946 947 while ((mpage = mpage_head) != NULL) { 948 while ((mlink = mpage->mlinks) != NULL) { 949 mpage->mlinks = mlink->next; 950 mlink_free(mlink); 951 } 952 mpage_head = mpage->next; 953 free(mpage->sec); 954 free(mpage->arch); 955 free(mpage->title); 956 free(mpage->desc); 957 free(mpage); 958 } 959 } 960 961 /* 962 * For each mlink to the mpage, check whether the path looks like 963 * it is formatted, and if it does, check whether a source manual 964 * exists by the same name, ignoring the suffix. 965 * If both conditions hold, drop the mlink. 966 */ 967 static void 968 mlinks_undupe(struct mpage *mpage) 969 { 970 char buf[PATH_MAX]; 971 struct mlink **prev; 972 struct mlink *mlink; 973 char *bufp; 974 975 mpage->form = FORM_CAT; 976 prev = &mpage->mlinks; 977 while (NULL != (mlink = *prev)) { 978 if (FORM_CAT != mlink->dform) { 979 mpage->form = FORM_NONE; 980 goto nextlink; 981 } 982 (void)strlcpy(buf, mlink->file, sizeof(buf)); 983 bufp = strstr(buf, "cat"); 984 assert(NULL != bufp); 985 memcpy(bufp, "man", 3); 986 if (NULL != (bufp = strrchr(buf, '.'))) 987 *++bufp = '\0'; 988 (void)strlcat(buf, mlink->dsec, sizeof(buf)); 989 if (NULL == ohash_find(&mlinks, 990 ohash_qlookup(&mlinks, buf))) 991 goto nextlink; 992 if (warnings) 993 say(mlink->file, "Man source exists: %s", buf); 994 if (use_all) 995 goto nextlink; 996 *prev = mlink->next; 997 mlink_free(mlink); 998 continue; 999 nextlink: 1000 prev = &(*prev)->next; 1001 } 1002 } 1003 1004 static void 1005 mlink_check(struct mpage *mpage, struct mlink *mlink) 1006 { 1007 struct str *str; 1008 unsigned int slot; 1009 1010 /* 1011 * Check whether the manual section given in a file 1012 * agrees with the directory where the file is located. 1013 * Some manuals have suffixes like (3p) on their 1014 * section number either inside the file or in the 1015 * directory name, some are linked into more than one 1016 * section, like encrypt(1) = makekey(8). 1017 */ 1018 1019 if (FORM_SRC == mpage->form && 1020 strcasecmp(mpage->sec, mlink->dsec)) 1021 say(mlink->file, "Section \"%s\" manual in %s directory", 1022 mpage->sec, mlink->dsec); 1023 1024 /* 1025 * Manual page directories exist for each kernel 1026 * architecture as returned by machine(1). 1027 * However, many manuals only depend on the 1028 * application architecture as returned by arch(1). 1029 * For example, some (2/ARM) manuals are shared 1030 * across the "armish" and "zaurus" kernel 1031 * architectures. 1032 * A few manuals are even shared across completely 1033 * different architectures, for example fdformat(1) 1034 * on amd64, i386, and sparc64. 1035 */ 1036 1037 if (strcasecmp(mpage->arch, mlink->arch)) 1038 say(mlink->file, "Architecture \"%s\" manual in " 1039 "\"%s\" directory", mpage->arch, mlink->arch); 1040 1041 /* 1042 * XXX 1043 * parse_cat() doesn't set NAME_TITLE yet. 1044 */ 1045 1046 if (FORM_CAT == mpage->form) 1047 return; 1048 1049 /* 1050 * Check whether this mlink 1051 * appears as a name in the NAME section. 1052 */ 1053 1054 slot = ohash_qlookup(&names, mlink->name); 1055 str = ohash_find(&names, slot); 1056 assert(NULL != str); 1057 if ( ! (NAME_TITLE & str->mask)) 1058 say(mlink->file, "Name missing in NAME section"); 1059 } 1060 1061 /* 1062 * Run through the files in the global vector "mpages" 1063 * and add them to the database specified in "basedir". 1064 * 1065 * This handles the parsing scheme itself, using the cues of directory 1066 * and filename to determine whether the file is parsable or not. 1067 */ 1068 static void 1069 mpages_merge(struct dba *dba, struct mparse *mp) 1070 { 1071 struct mpage *mpage, *mpage_dest; 1072 struct mlink *mlink, *mlink_dest; 1073 struct roff_man *man; 1074 char *sodest; 1075 char *cp; 1076 int fd; 1077 1078 for (mpage = mpage_head; mpage != NULL; mpage = mpage->next) { 1079 mlinks_undupe(mpage); 1080 if ((mlink = mpage->mlinks) == NULL) 1081 continue; 1082 1083 name_mask = NAME_MASK; 1084 mandoc_ohash_init(&names, 4, offsetof(struct str, key)); 1085 mandoc_ohash_init(&strings, 6, offsetof(struct str, key)); 1086 mparse_reset(mp); 1087 man = NULL; 1088 sodest = NULL; 1089 1090 if ((fd = mparse_open(mp, mlink->file)) == -1) { 1091 say(mlink->file, "&open"); 1092 goto nextpage; 1093 } 1094 1095 /* 1096 * Interpret the file as mdoc(7) or man(7) source 1097 * code, unless it is known to be formatted. 1098 */ 1099 if (mlink->dform != FORM_CAT || mlink->fform != FORM_CAT) { 1100 mparse_readfd(mp, fd, mlink->file); 1101 close(fd); 1102 mparse_result(mp, &man, &sodest); 1103 } 1104 1105 if (sodest != NULL) { 1106 mlink_dest = ohash_find(&mlinks, 1107 ohash_qlookup(&mlinks, sodest)); 1108 if (mlink_dest == NULL) { 1109 mandoc_asprintf(&cp, "%s.gz", sodest); 1110 mlink_dest = ohash_find(&mlinks, 1111 ohash_qlookup(&mlinks, cp)); 1112 free(cp); 1113 } 1114 if (mlink_dest != NULL) { 1115 1116 /* The .so target exists. */ 1117 1118 mpage_dest = mlink_dest->mpage; 1119 while (1) { 1120 mlink->mpage = mpage_dest; 1121 1122 /* 1123 * If the target was already 1124 * processed, add the links 1125 * to the database now. 1126 * Otherwise, this will 1127 * happen when we come 1128 * to the target. 1129 */ 1130 1131 if (mpage_dest->dba != NULL) 1132 dbadd_mlink(mlink); 1133 1134 if (mlink->next == NULL) 1135 break; 1136 mlink = mlink->next; 1137 } 1138 1139 /* Move all links to the target. */ 1140 1141 mlink->next = mlink_dest->next; 1142 mlink_dest->next = mpage->mlinks; 1143 mpage->mlinks = NULL; 1144 } 1145 goto nextpage; 1146 } else if (man != NULL && man->macroset == MACROSET_MDOC) { 1147 mdoc_validate(man); 1148 mpage->form = FORM_SRC; 1149 mpage->sec = man->meta.msec; 1150 mpage->sec = mandoc_strdup( 1151 mpage->sec == NULL ? "" : mpage->sec); 1152 mpage->arch = man->meta.arch; 1153 mpage->arch = mandoc_strdup( 1154 mpage->arch == NULL ? "" : mpage->arch); 1155 mpage->title = mandoc_strdup(man->meta.title); 1156 } else if (man != NULL && man->macroset == MACROSET_MAN) { 1157 man_validate(man); 1158 mpage->form = FORM_SRC; 1159 mpage->sec = mandoc_strdup(man->meta.msec); 1160 mpage->arch = mandoc_strdup(mlink->arch); 1161 mpage->title = mandoc_strdup(man->meta.title); 1162 } else { 1163 mpage->form = FORM_CAT; 1164 mpage->sec = mandoc_strdup(mlink->dsec); 1165 mpage->arch = mandoc_strdup(mlink->arch); 1166 mpage->title = mandoc_strdup(mlink->name); 1167 } 1168 1169 assert(mpage->desc == NULL); 1170 if (man != NULL && man->macroset == MACROSET_MDOC) 1171 parse_mdoc(mpage, &man->meta, man->first); 1172 else if (man != NULL) 1173 parse_man(mpage, &man->meta, man->first); 1174 else 1175 parse_cat(mpage, fd); 1176 if (mpage->desc == NULL) 1177 mpage->desc = mandoc_strdup(mpage->mlinks->name); 1178 1179 if (warnings && !use_all) 1180 for (mlink = mpage->mlinks; mlink; 1181 mlink = mlink->next) 1182 mlink_check(mpage, mlink); 1183 1184 dbadd(dba, mpage); 1185 mlink = mpage->mlinks; 1186 1187 nextpage: 1188 ohash_delete(&strings); 1189 ohash_delete(&names); 1190 } 1191 } 1192 1193 static void 1194 parse_cat(struct mpage *mpage, int fd) 1195 { 1196 FILE *stream; 1197 char *line, *p, *title; 1198 size_t linesz, plen, titlesz; 1199 ssize_t len; 1200 int offs; 1201 1202 stream = (-1 == fd) ? 1203 fopen(mpage->mlinks->file, "r") : 1204 fdopen(fd, "r"); 1205 if (NULL == stream) { 1206 if (-1 != fd) 1207 close(fd); 1208 if (warnings) 1209 say(mpage->mlinks->file, "&fopen"); 1210 return; 1211 } 1212 1213 line = NULL; 1214 linesz = 0; 1215 1216 /* Skip to first blank line. */ 1217 1218 while (getline(&line, &linesz, stream) != -1) 1219 if (*line == '\n') 1220 break; 1221 1222 /* 1223 * Assume the first line that is not indented 1224 * is the first section header. Skip to it. 1225 */ 1226 1227 while (getline(&line, &linesz, stream) != -1) 1228 if (*line != '\n' && *line != ' ') 1229 break; 1230 1231 /* 1232 * Read up until the next section into a buffer. 1233 * Strip the leading and trailing newline from each read line, 1234 * appending a trailing space. 1235 * Ignore empty (whitespace-only) lines. 1236 */ 1237 1238 titlesz = 0; 1239 title = NULL; 1240 1241 while ((len = getline(&line, &linesz, stream)) != -1) { 1242 if (*line != ' ') 1243 break; 1244 offs = 0; 1245 while (isspace((unsigned char)line[offs])) 1246 offs++; 1247 if (line[offs] == '\0') 1248 continue; 1249 title = mandoc_realloc(title, titlesz + len - offs); 1250 memcpy(title + titlesz, line + offs, len - offs); 1251 titlesz += len - offs; 1252 title[titlesz - 1] = ' '; 1253 } 1254 free(line); 1255 1256 /* 1257 * If no page content can be found, or the input line 1258 * is already the next section header, or there is no 1259 * trailing newline, reuse the page title as the page 1260 * description. 1261 */ 1262 1263 if (NULL == title || '\0' == *title) { 1264 if (warnings) 1265 say(mpage->mlinks->file, 1266 "Cannot find NAME section"); 1267 fclose(stream); 1268 free(title); 1269 return; 1270 } 1271 1272 title[titlesz - 1] = '\0'; 1273 1274 /* 1275 * Skip to the first dash. 1276 * Use the remaining line as the description (no more than 70 1277 * bytes). 1278 */ 1279 1280 if (NULL != (p = strstr(title, "- "))) { 1281 for (p += 2; ' ' == *p || '\b' == *p; p++) 1282 /* Skip to next word. */ ; 1283 } else { 1284 if (warnings) 1285 say(mpage->mlinks->file, 1286 "No dash in title line"); 1287 p = title; 1288 } 1289 1290 plen = strlen(p); 1291 1292 /* Strip backspace-encoding from line. */ 1293 1294 while (NULL != (line = memchr(p, '\b', plen))) { 1295 len = line - p; 1296 if (0 == len) { 1297 memmove(line, line + 1, plen--); 1298 continue; 1299 } 1300 memmove(line - 1, line + 1, plen - len); 1301 plen -= 2; 1302 } 1303 1304 mpage->desc = mandoc_strdup(p); 1305 fclose(stream); 1306 free(title); 1307 } 1308 1309 /* 1310 * Put a type/word pair into the word database for this particular file. 1311 */ 1312 static void 1313 putkey(const struct mpage *mpage, char *value, uint64_t type) 1314 { 1315 putkeys(mpage, value, strlen(value), type); 1316 } 1317 1318 /* 1319 * Grok all nodes at or below a certain mdoc node into putkey(). 1320 */ 1321 static void 1322 putmdockey(const struct mpage *mpage, 1323 const struct roff_node *n, uint64_t m, int taboo) 1324 { 1325 1326 for ( ; NULL != n; n = n->next) { 1327 if (n->flags & taboo) 1328 continue; 1329 if (NULL != n->child) 1330 putmdockey(mpage, n->child, m, taboo); 1331 if (n->type == ROFFT_TEXT) 1332 putkey(mpage, n->string, m); 1333 } 1334 } 1335 1336 static void 1337 parse_man(struct mpage *mpage, const struct roff_meta *meta, 1338 const struct roff_node *n) 1339 { 1340 const struct roff_node *head, *body; 1341 char *start, *title; 1342 char byte; 1343 size_t sz; 1344 1345 if (n == NULL) 1346 return; 1347 1348 /* 1349 * We're only searching for one thing: the first text child in 1350 * the BODY of a NAME section. Since we don't keep track of 1351 * sections in -man, run some hoops to find out whether we're in 1352 * the correct section or not. 1353 */ 1354 1355 if (n->type == ROFFT_BODY && n->tok == MAN_SH) { 1356 body = n; 1357 if ((head = body->parent->head) != NULL && 1358 (head = head->child) != NULL && 1359 head->next == NULL && 1360 head->type == ROFFT_TEXT && 1361 strcmp(head->string, "NAME") == 0 && 1362 body->child != NULL) { 1363 1364 /* 1365 * Suck the entire NAME section into memory. 1366 * Yes, we might run away. 1367 * But too many manuals have big, spread-out 1368 * NAME sections over many lines. 1369 */ 1370 1371 title = NULL; 1372 deroff(&title, body); 1373 if (NULL == title) 1374 return; 1375 1376 /* 1377 * Go through a special heuristic dance here. 1378 * Conventionally, one or more manual names are 1379 * comma-specified prior to a whitespace, then a 1380 * dash, then a description. Try to puzzle out 1381 * the name parts here. 1382 */ 1383 1384 start = title; 1385 for ( ;; ) { 1386 sz = strcspn(start, " ,"); 1387 if ('\0' == start[sz]) 1388 break; 1389 1390 byte = start[sz]; 1391 start[sz] = '\0'; 1392 1393 /* 1394 * Assume a stray trailing comma in the 1395 * name list if a name begins with a dash. 1396 */ 1397 1398 if ('-' == start[0] || 1399 ('\\' == start[0] && '-' == start[1])) 1400 break; 1401 1402 putkey(mpage, start, NAME_TITLE); 1403 if ( ! (mpage->name_head_done || 1404 strcasecmp(start, meta->title))) { 1405 putkey(mpage, start, NAME_HEAD); 1406 mpage->name_head_done = 1; 1407 } 1408 1409 if (' ' == byte) { 1410 start += sz + 1; 1411 break; 1412 } 1413 1414 assert(',' == byte); 1415 start += sz + 1; 1416 while (' ' == *start) 1417 start++; 1418 } 1419 1420 if (start == title) { 1421 putkey(mpage, start, NAME_TITLE); 1422 if ( ! (mpage->name_head_done || 1423 strcasecmp(start, meta->title))) { 1424 putkey(mpage, start, NAME_HEAD); 1425 mpage->name_head_done = 1; 1426 } 1427 free(title); 1428 return; 1429 } 1430 1431 while (isspace((unsigned char)*start)) 1432 start++; 1433 1434 if (0 == strncmp(start, "-", 1)) 1435 start += 1; 1436 else if (0 == strncmp(start, "\\-\\-", 4)) 1437 start += 4; 1438 else if (0 == strncmp(start, "\\-", 2)) 1439 start += 2; 1440 else if (0 == strncmp(start, "\\(en", 4)) 1441 start += 4; 1442 else if (0 == strncmp(start, "\\(em", 4)) 1443 start += 4; 1444 1445 while (' ' == *start) 1446 start++; 1447 1448 mpage->desc = mandoc_strdup(start); 1449 free(title); 1450 return; 1451 } 1452 } 1453 1454 for (n = n->child; n; n = n->next) { 1455 if (NULL != mpage->desc) 1456 break; 1457 parse_man(mpage, meta, n); 1458 } 1459 } 1460 1461 static void 1462 parse_mdoc(struct mpage *mpage, const struct roff_meta *meta, 1463 const struct roff_node *n) 1464 { 1465 1466 assert(NULL != n); 1467 for (n = n->child; NULL != n; n = n->next) { 1468 if (n->flags & mdocs[n->tok].taboo) 1469 continue; 1470 switch (n->type) { 1471 case ROFFT_ELEM: 1472 case ROFFT_BLOCK: 1473 case ROFFT_HEAD: 1474 case ROFFT_BODY: 1475 case ROFFT_TAIL: 1476 if (NULL != mdocs[n->tok].fp) 1477 if (0 == (*mdocs[n->tok].fp)(mpage, meta, n)) 1478 break; 1479 if (mdocs[n->tok].mask) 1480 putmdockey(mpage, n->child, 1481 mdocs[n->tok].mask, mdocs[n->tok].taboo); 1482 break; 1483 default: 1484 assert(n->type != ROFFT_ROOT); 1485 continue; 1486 } 1487 if (NULL != n->child) 1488 parse_mdoc(mpage, meta, n); 1489 } 1490 } 1491 1492 static int 1493 parse_mdoc_Fd(struct mpage *mpage, const struct roff_meta *meta, 1494 const struct roff_node *n) 1495 { 1496 char *start, *end; 1497 size_t sz; 1498 1499 if (SEC_SYNOPSIS != n->sec || 1500 NULL == (n = n->child) || 1501 n->type != ROFFT_TEXT) 1502 return 0; 1503 1504 /* 1505 * Only consider those `Fd' macro fields that begin with an 1506 * "inclusion" token (versus, e.g., #define). 1507 */ 1508 1509 if (strcmp("#include", n->string)) 1510 return 0; 1511 1512 if ((n = n->next) == NULL || n->type != ROFFT_TEXT) 1513 return 0; 1514 1515 /* 1516 * Strip away the enclosing angle brackets and make sure we're 1517 * not zero-length. 1518 */ 1519 1520 start = n->string; 1521 if ('<' == *start || '"' == *start) 1522 start++; 1523 1524 if (0 == (sz = strlen(start))) 1525 return 0; 1526 1527 end = &start[(int)sz - 1]; 1528 if ('>' == *end || '"' == *end) 1529 end--; 1530 1531 if (end > start) 1532 putkeys(mpage, start, end - start + 1, TYPE_In); 1533 return 0; 1534 } 1535 1536 static void 1537 parse_mdoc_fname(struct mpage *mpage, const struct roff_node *n) 1538 { 1539 char *cp; 1540 size_t sz; 1541 1542 if (n->type != ROFFT_TEXT) 1543 return; 1544 1545 /* Skip function pointer punctuation. */ 1546 1547 cp = n->string; 1548 while (*cp == '(' || *cp == '*') 1549 cp++; 1550 sz = strcspn(cp, "()"); 1551 1552 putkeys(mpage, cp, sz, TYPE_Fn); 1553 if (n->sec == SEC_SYNOPSIS) 1554 putkeys(mpage, cp, sz, NAME_SYN); 1555 } 1556 1557 static int 1558 parse_mdoc_Fn(struct mpage *mpage, const struct roff_meta *meta, 1559 const struct roff_node *n) 1560 { 1561 1562 if (n->child == NULL) 1563 return 0; 1564 1565 parse_mdoc_fname(mpage, n->child); 1566 1567 for (n = n->child->next; n != NULL; n = n->next) 1568 if (n->type == ROFFT_TEXT) 1569 putkey(mpage, n->string, TYPE_Fa); 1570 1571 return 0; 1572 } 1573 1574 static int 1575 parse_mdoc_Fo(struct mpage *mpage, const struct roff_meta *meta, 1576 const struct roff_node *n) 1577 { 1578 1579 if (n->type != ROFFT_HEAD) 1580 return 1; 1581 1582 if (n->child != NULL) 1583 parse_mdoc_fname(mpage, n->child); 1584 1585 return 0; 1586 } 1587 1588 static int 1589 parse_mdoc_Va(struct mpage *mpage, const struct roff_meta *meta, 1590 const struct roff_node *n) 1591 { 1592 char *cp; 1593 1594 if (n->type != ROFFT_ELEM && n->type != ROFFT_BODY) 1595 return 0; 1596 1597 if (n->child != NULL && 1598 n->child->next == NULL && 1599 n->child->type == ROFFT_TEXT) 1600 return 1; 1601 1602 cp = NULL; 1603 deroff(&cp, n); 1604 if (cp != NULL) { 1605 putkey(mpage, cp, TYPE_Vt | (n->tok == MDOC_Va || 1606 n->type == ROFFT_BODY ? TYPE_Va : 0)); 1607 free(cp); 1608 } 1609 1610 return 0; 1611 } 1612 1613 static int 1614 parse_mdoc_Xr(struct mpage *mpage, const struct roff_meta *meta, 1615 const struct roff_node *n) 1616 { 1617 char *cp; 1618 1619 if (NULL == (n = n->child)) 1620 return 0; 1621 1622 if (NULL == n->next) { 1623 putkey(mpage, n->string, TYPE_Xr); 1624 return 0; 1625 } 1626 1627 mandoc_asprintf(&cp, "%s(%s)", n->string, n->next->string); 1628 putkey(mpage, cp, TYPE_Xr); 1629 free(cp); 1630 return 0; 1631 } 1632 1633 static int 1634 parse_mdoc_Nd(struct mpage *mpage, const struct roff_meta *meta, 1635 const struct roff_node *n) 1636 { 1637 1638 if (n->type == ROFFT_BODY) 1639 deroff(&mpage->desc, n); 1640 return 0; 1641 } 1642 1643 static int 1644 parse_mdoc_Nm(struct mpage *mpage, const struct roff_meta *meta, 1645 const struct roff_node *n) 1646 { 1647 1648 if (SEC_NAME == n->sec) 1649 putmdockey(mpage, n->child, NAME_TITLE, 0); 1650 else if (n->sec == SEC_SYNOPSIS && n->type == ROFFT_HEAD) { 1651 if (n->child == NULL) 1652 putkey(mpage, meta->name, NAME_SYN); 1653 else 1654 putmdockey(mpage, n->child, NAME_SYN, 0); 1655 } 1656 if ( ! (mpage->name_head_done || 1657 n->child == NULL || n->child->string == NULL || 1658 strcasecmp(n->child->string, meta->title))) { 1659 putkey(mpage, n->child->string, NAME_HEAD); 1660 mpage->name_head_done = 1; 1661 } 1662 return 0; 1663 } 1664 1665 static int 1666 parse_mdoc_Sh(struct mpage *mpage, const struct roff_meta *meta, 1667 const struct roff_node *n) 1668 { 1669 1670 return n->sec == SEC_CUSTOM && n->type == ROFFT_HEAD; 1671 } 1672 1673 static int 1674 parse_mdoc_head(struct mpage *mpage, const struct roff_meta *meta, 1675 const struct roff_node *n) 1676 { 1677 1678 return n->type == ROFFT_HEAD; 1679 } 1680 1681 /* 1682 * Add a string to the hash table for the current manual. 1683 * Each string has a bitmask telling which macros it belongs to. 1684 * When we finish the manual, we'll dump the table. 1685 */ 1686 static void 1687 putkeys(const struct mpage *mpage, char *cp, size_t sz, uint64_t v) 1688 { 1689 struct ohash *htab; 1690 struct str *s; 1691 const char *end; 1692 unsigned int slot; 1693 int i, mustfree; 1694 1695 if (0 == sz) 1696 return; 1697 1698 mustfree = render_string(&cp, &sz); 1699 1700 if (TYPE_Nm & v) { 1701 htab = &names; 1702 v &= name_mask; 1703 if (v & NAME_FIRST) 1704 name_mask &= ~NAME_FIRST; 1705 if (debug > 1) 1706 say(mpage->mlinks->file, 1707 "Adding name %*s, bits=0x%llx", (int)sz, cp, 1708 (unsigned long long)v); 1709 } else { 1710 htab = &strings; 1711 if (debug > 1) 1712 for (i = 0; i < KEY_MAX; i++) 1713 if ((uint64_t)1 << i & v) 1714 say(mpage->mlinks->file, 1715 "Adding key %s=%*s", 1716 mansearch_keynames[i], (int)sz, cp); 1717 } 1718 1719 end = cp + sz; 1720 slot = ohash_qlookupi(htab, cp, &end); 1721 s = ohash_find(htab, slot); 1722 1723 if (NULL != s && mpage == s->mpage) { 1724 s->mask |= v; 1725 return; 1726 } else if (NULL == s) { 1727 s = mandoc_calloc(1, sizeof(struct str) + sz + 1); 1728 memcpy(s->key, cp, sz); 1729 ohash_insert(htab, slot, s); 1730 } 1731 s->mpage = mpage; 1732 s->mask = v; 1733 1734 if (mustfree) 1735 free(cp); 1736 } 1737 1738 /* 1739 * Take a Unicode codepoint and produce its UTF-8 encoding. 1740 * This isn't the best way to do this, but it works. 1741 * The magic numbers are from the UTF-8 packaging. 1742 * They're not as scary as they seem: read the UTF-8 spec for details. 1743 */ 1744 static size_t 1745 utf8(unsigned int cp, char out[7]) 1746 { 1747 size_t rc; 1748 1749 rc = 0; 1750 if (cp <= 0x0000007F) { 1751 rc = 1; 1752 out[0] = (char)cp; 1753 } else if (cp <= 0x000007FF) { 1754 rc = 2; 1755 out[0] = (cp >> 6 & 31) | 192; 1756 out[1] = (cp & 63) | 128; 1757 } else if (cp <= 0x0000FFFF) { 1758 rc = 3; 1759 out[0] = (cp >> 12 & 15) | 224; 1760 out[1] = (cp >> 6 & 63) | 128; 1761 out[2] = (cp & 63) | 128; 1762 } else if (cp <= 0x001FFFFF) { 1763 rc = 4; 1764 out[0] = (cp >> 18 & 7) | 240; 1765 out[1] = (cp >> 12 & 63) | 128; 1766 out[2] = (cp >> 6 & 63) | 128; 1767 out[3] = (cp & 63) | 128; 1768 } else if (cp <= 0x03FFFFFF) { 1769 rc = 5; 1770 out[0] = (cp >> 24 & 3) | 248; 1771 out[1] = (cp >> 18 & 63) | 128; 1772 out[2] = (cp >> 12 & 63) | 128; 1773 out[3] = (cp >> 6 & 63) | 128; 1774 out[4] = (cp & 63) | 128; 1775 } else if (cp <= 0x7FFFFFFF) { 1776 rc = 6; 1777 out[0] = (cp >> 30 & 1) | 252; 1778 out[1] = (cp >> 24 & 63) | 128; 1779 out[2] = (cp >> 18 & 63) | 128; 1780 out[3] = (cp >> 12 & 63) | 128; 1781 out[4] = (cp >> 6 & 63) | 128; 1782 out[5] = (cp & 63) | 128; 1783 } else 1784 return 0; 1785 1786 out[rc] = '\0'; 1787 return rc; 1788 } 1789 1790 /* 1791 * If the string contains escape sequences, 1792 * replace it with an allocated rendering and return 1, 1793 * such that the caller can free it after use. 1794 * Otherwise, do nothing and return 0. 1795 */ 1796 static int 1797 render_string(char **public, size_t *psz) 1798 { 1799 const char *src, *scp, *addcp, *seq; 1800 char *dst; 1801 size_t ssz, dsz, addsz; 1802 char utfbuf[7], res[6]; 1803 int seqlen, unicode; 1804 1805 res[0] = '\\'; 1806 res[1] = '\t'; 1807 res[2] = ASCII_NBRSP; 1808 res[3] = ASCII_HYPH; 1809 res[4] = ASCII_BREAK; 1810 res[5] = '\0'; 1811 1812 src = scp = *public; 1813 ssz = *psz; 1814 dst = NULL; 1815 dsz = 0; 1816 1817 while (scp < src + *psz) { 1818 1819 /* Leave normal characters unchanged. */ 1820 1821 if (strchr(res, *scp) == NULL) { 1822 if (dst != NULL) 1823 dst[dsz++] = *scp; 1824 scp++; 1825 continue; 1826 } 1827 1828 /* 1829 * Found something that requires replacing, 1830 * make sure we have a destination buffer. 1831 */ 1832 1833 if (dst == NULL) { 1834 dst = mandoc_malloc(ssz + 1); 1835 dsz = scp - src; 1836 memcpy(dst, src, dsz); 1837 } 1838 1839 /* Handle single-char special characters. */ 1840 1841 switch (*scp) { 1842 case '\\': 1843 break; 1844 case '\t': 1845 case ASCII_NBRSP: 1846 dst[dsz++] = ' '; 1847 scp++; 1848 continue; 1849 case ASCII_HYPH: 1850 dst[dsz++] = '-'; 1851 /* FALLTHROUGH */ 1852 case ASCII_BREAK: 1853 scp++; 1854 continue; 1855 default: 1856 abort(); 1857 } 1858 1859 /* 1860 * Found an escape sequence. 1861 * Read past the slash, then parse it. 1862 * Ignore everything except characters. 1863 */ 1864 1865 scp++; 1866 if (mandoc_escape(&scp, &seq, &seqlen) != ESCAPE_SPECIAL) 1867 continue; 1868 1869 /* 1870 * Render the special character 1871 * as either UTF-8 or ASCII. 1872 */ 1873 1874 if (write_utf8) { 1875 unicode = mchars_spec2cp(seq, seqlen); 1876 if (unicode <= 0) 1877 continue; 1878 addsz = utf8(unicode, utfbuf); 1879 if (addsz == 0) 1880 continue; 1881 addcp = utfbuf; 1882 } else { 1883 addcp = mchars_spec2str(seq, seqlen, &addsz); 1884 if (addcp == NULL) 1885 continue; 1886 if (*addcp == ASCII_NBRSP) { 1887 addcp = " "; 1888 addsz = 1; 1889 } 1890 } 1891 1892 /* Copy the rendered glyph into the stream. */ 1893 1894 ssz += addsz; 1895 dst = mandoc_realloc(dst, ssz + 1); 1896 memcpy(dst + dsz, addcp, addsz); 1897 dsz += addsz; 1898 } 1899 if (dst != NULL) { 1900 *public = dst; 1901 *psz = dsz; 1902 } 1903 1904 /* Trim trailing whitespace and NUL-terminate. */ 1905 1906 while (*psz > 0 && (*public)[*psz - 1] == ' ') 1907 --*psz; 1908 if (dst != NULL) { 1909 (*public)[*psz] = '\0'; 1910 return 1; 1911 } else 1912 return 0; 1913 } 1914 1915 static void 1916 dbadd_mlink(const struct mlink *mlink) 1917 { 1918 dba_page_alias(mlink->mpage->dba, mlink->name, NAME_FILE); 1919 dba_page_add(mlink->mpage->dba, DBP_SECT, mlink->dsec); 1920 dba_page_add(mlink->mpage->dba, DBP_SECT, mlink->fsec); 1921 dba_page_add(mlink->mpage->dba, DBP_ARCH, mlink->arch); 1922 dba_page_add(mlink->mpage->dba, DBP_FILE, mlink->file); 1923 } 1924 1925 /* 1926 * Flush the current page's terms (and their bits) into the database. 1927 * Also, handle escape sequences at the last possible moment. 1928 */ 1929 static void 1930 dbadd(struct dba *dba, struct mpage *mpage) 1931 { 1932 struct mlink *mlink; 1933 struct str *key; 1934 char *cp; 1935 uint64_t mask; 1936 size_t i; 1937 unsigned int slot; 1938 int mustfree; 1939 1940 mlink = mpage->mlinks; 1941 1942 if (nodb) { 1943 for (key = ohash_first(&names, &slot); NULL != key; 1944 key = ohash_next(&names, &slot)) 1945 free(key); 1946 for (key = ohash_first(&strings, &slot); NULL != key; 1947 key = ohash_next(&strings, &slot)) 1948 free(key); 1949 if (0 == debug) 1950 return; 1951 while (NULL != mlink) { 1952 fputs(mlink->name, stdout); 1953 if (NULL == mlink->next || 1954 strcmp(mlink->dsec, mlink->next->dsec) || 1955 strcmp(mlink->fsec, mlink->next->fsec) || 1956 strcmp(mlink->arch, mlink->next->arch)) { 1957 putchar('('); 1958 if ('\0' == *mlink->dsec) 1959 fputs(mlink->fsec, stdout); 1960 else 1961 fputs(mlink->dsec, stdout); 1962 if ('\0' != *mlink->arch) 1963 printf("/%s", mlink->arch); 1964 putchar(')'); 1965 } 1966 mlink = mlink->next; 1967 if (NULL != mlink) 1968 fputs(", ", stdout); 1969 } 1970 printf(" - %s\n", mpage->desc); 1971 return; 1972 } 1973 1974 if (debug) 1975 say(mlink->file, "Adding to database"); 1976 1977 cp = mpage->desc; 1978 i = strlen(cp); 1979 mustfree = render_string(&cp, &i); 1980 mpage->dba = dba_page_new(dba->pages, 1981 *mpage->arch == '\0' ? mlink->arch : mpage->arch, 1982 cp, mlink->file, mpage->form); 1983 if (mustfree) 1984 free(cp); 1985 dba_page_add(mpage->dba, DBP_SECT, mpage->sec); 1986 1987 while (mlink != NULL) { 1988 dbadd_mlink(mlink); 1989 mlink = mlink->next; 1990 } 1991 1992 for (key = ohash_first(&names, &slot); NULL != key; 1993 key = ohash_next(&names, &slot)) { 1994 assert(key->mpage == mpage); 1995 dba_page_alias(mpage->dba, key->key, key->mask); 1996 free(key); 1997 } 1998 for (key = ohash_first(&strings, &slot); NULL != key; 1999 key = ohash_next(&strings, &slot)) { 2000 assert(key->mpage == mpage); 2001 i = 0; 2002 for (mask = TYPE_Xr; mask <= TYPE_Lb; mask *= 2) { 2003 if (key->mask & mask) 2004 dba_macro_add(dba->macros, i, 2005 key->key, mpage->dba); 2006 i++; 2007 } 2008 free(key); 2009 } 2010 } 2011 2012 static void 2013 dbprune(struct dba *dba) 2014 { 2015 struct dba_array *page, *files; 2016 char *file; 2017 2018 dba_array_FOREACH(dba->pages, page) { 2019 files = dba_array_get(page, DBP_FILE); 2020 dba_array_FOREACH(files, file) { 2021 if (*file < ' ') 2022 file++; 2023 if (ohash_find(&mlinks, ohash_qlookup(&mlinks, 2024 file)) != NULL) { 2025 if (debug) 2026 say(file, "Deleting from database"); 2027 dba_array_del(dba->pages); 2028 break; 2029 } 2030 } 2031 } 2032 } 2033 2034 /* 2035 * Write the database from memory to disk. 2036 */ 2037 static void 2038 dbwrite(struct dba *dba) 2039 { 2040 char tfn[32]; 2041 int status; 2042 pid_t child; 2043 2044 if (dba_write(MANDOC_DB "~", dba) != -1) { 2045 if (rename(MANDOC_DB "~", MANDOC_DB) == -1) { 2046 exitcode = (int)MANDOCLEVEL_SYSERR; 2047 say(MANDOC_DB, "&rename"); 2048 unlink(MANDOC_DB "~"); 2049 } 2050 return; 2051 } 2052 2053 (void)strlcpy(tfn, "/tmp/mandocdb.XXXXXXXX", sizeof(tfn)); 2054 if (mkdtemp(tfn) == NULL) { 2055 exitcode = (int)MANDOCLEVEL_SYSERR; 2056 say("", "&%s", tfn); 2057 return; 2058 } 2059 2060 (void)strlcat(tfn, "/" MANDOC_DB, sizeof(tfn)); 2061 if (dba_write(tfn, dba) == -1) { 2062 exitcode = (int)MANDOCLEVEL_SYSERR; 2063 say(tfn, "&dba_write"); 2064 goto out; 2065 } 2066 2067 switch (child = fork()) { 2068 case -1: 2069 exitcode = (int)MANDOCLEVEL_SYSERR; 2070 say("", "&fork cmp"); 2071 return; 2072 case 0: 2073 execlp("cmp", "cmp", "-s", tfn, MANDOC_DB, (char *)NULL); 2074 say("", "&exec cmp"); 2075 exit(0); 2076 default: 2077 break; 2078 } 2079 if (waitpid(child, &status, 0) == -1) { 2080 exitcode = (int)MANDOCLEVEL_SYSERR; 2081 say("", "&wait cmp"); 2082 } else if (WIFSIGNALED(status)) { 2083 exitcode = (int)MANDOCLEVEL_SYSERR; 2084 say("", "cmp died from signal %d", WTERMSIG(status)); 2085 } else if (WEXITSTATUS(status)) { 2086 exitcode = (int)MANDOCLEVEL_SYSERR; 2087 say(MANDOC_DB, 2088 "Data changed, but cannot replace database"); 2089 } 2090 2091 out: 2092 *strrchr(tfn, '/') = '\0'; 2093 switch (child = fork()) { 2094 case -1: 2095 exitcode = (int)MANDOCLEVEL_SYSERR; 2096 say("", "&fork rm"); 2097 return; 2098 case 0: 2099 execlp("rm", "rm", "-rf", tfn, (char *)NULL); 2100 say("", "&exec rm"); 2101 exit((int)MANDOCLEVEL_SYSERR); 2102 default: 2103 break; 2104 } 2105 if (waitpid(child, &status, 0) == -1) { 2106 exitcode = (int)MANDOCLEVEL_SYSERR; 2107 say("", "&wait rm"); 2108 } else if (WIFSIGNALED(status) || WEXITSTATUS(status)) { 2109 exitcode = (int)MANDOCLEVEL_SYSERR; 2110 say("", "%s: Cannot remove temporary directory", tfn); 2111 } 2112 } 2113 2114 static int 2115 set_basedir(const char *targetdir, int report_baddir) 2116 { 2117 static char startdir[PATH_MAX]; 2118 static int getcwd_status; /* 1 = ok, 2 = failure */ 2119 static int chdir_status; /* 1 = changed directory */ 2120 char *cp; 2121 2122 /* 2123 * Remember the original working directory, if possible. 2124 * This will be needed if the second or a later directory 2125 * on the command line is given as a relative path. 2126 * Do not error out if the current directory is not 2127 * searchable: Maybe it won't be needed after all. 2128 */ 2129 if (0 == getcwd_status) { 2130 if (NULL == getcwd(startdir, sizeof(startdir))) { 2131 getcwd_status = 2; 2132 (void)strlcpy(startdir, strerror(errno), 2133 sizeof(startdir)); 2134 } else 2135 getcwd_status = 1; 2136 } 2137 2138 /* 2139 * We are leaving the old base directory. 2140 * Do not use it any longer, not even for messages. 2141 */ 2142 *basedir = '\0'; 2143 2144 /* 2145 * If and only if the directory was changed earlier and 2146 * the next directory to process is given as a relative path, 2147 * first go back, or bail out if that is impossible. 2148 */ 2149 if (chdir_status && '/' != *targetdir) { 2150 if (2 == getcwd_status) { 2151 exitcode = (int)MANDOCLEVEL_SYSERR; 2152 say("", "getcwd: %s", startdir); 2153 return 0; 2154 } 2155 if (-1 == chdir(startdir)) { 2156 exitcode = (int)MANDOCLEVEL_SYSERR; 2157 say("", "&chdir %s", startdir); 2158 return 0; 2159 } 2160 } 2161 2162 /* 2163 * Always resolve basedir to the canonicalized absolute 2164 * pathname and append a trailing slash, such that 2165 * we can reliably check whether files are inside. 2166 */ 2167 if (NULL == realpath(targetdir, basedir)) { 2168 if (report_baddir || errno != ENOENT) { 2169 exitcode = (int)MANDOCLEVEL_BADARG; 2170 say("", "&%s: realpath", targetdir); 2171 } 2172 return 0; 2173 } else if (-1 == chdir(basedir)) { 2174 if (report_baddir || errno != ENOENT) { 2175 exitcode = (int)MANDOCLEVEL_BADARG; 2176 say("", "&chdir"); 2177 } 2178 return 0; 2179 } 2180 chdir_status = 1; 2181 cp = strchr(basedir, '\0'); 2182 if ('/' != cp[-1]) { 2183 if (cp - basedir >= PATH_MAX - 1) { 2184 exitcode = (int)MANDOCLEVEL_SYSERR; 2185 say("", "Filename too long"); 2186 return 0; 2187 } 2188 *cp++ = '/'; 2189 *cp = '\0'; 2190 } 2191 return 1; 2192 } 2193 2194 static void 2195 say(const char *file, const char *format, ...) 2196 { 2197 va_list ap; 2198 int use_errno; 2199 2200 if ('\0' != *basedir) 2201 fprintf(stderr, "%s", basedir); 2202 if ('\0' != *basedir && '\0' != *file) 2203 fputc('/', stderr); 2204 if ('\0' != *file) 2205 fprintf(stderr, "%s", file); 2206 2207 use_errno = 1; 2208 if (NULL != format) { 2209 switch (*format) { 2210 case '&': 2211 format++; 2212 break; 2213 case '\0': 2214 format = NULL; 2215 break; 2216 default: 2217 use_errno = 0; 2218 break; 2219 } 2220 } 2221 if (NULL != format) { 2222 if ('\0' != *basedir || '\0' != *file) 2223 fputs(": ", stderr); 2224 va_start(ap, format); 2225 vfprintf(stderr, format, ap); 2226 va_end(ap); 2227 } 2228 if (use_errno) { 2229 if ('\0' != *basedir || '\0' != *file || NULL != format) 2230 fputs(": ", stderr); 2231 perror(NULL); 2232 } else 2233 fputc('\n', stderr); 2234 } 2235