Added check for libiconv_open when iconv_open isn't found.
[rsync/rsync.git] / util.c
CommitLineData
0f78b815
WD
1/*
2 * Utility routines used in rsync.
5cb37436 3 *
0f78b815
WD
4 * Copyright (C) 1996-2000 Andrew Tridgell
5 * Copyright (C) 1996 Paul Mackerras
6 * Copyright (C) 2001, 2002 Martin Pool <mbp@samba.org>
ba2133d6 7 * Copyright (C) 2003-2007 Wayne Davison
5cb37436 8 *
0ecfbf27 9 * This program is free software; you can redistribute it and/or modify
8e41b68e
WD
10 * it under the terms of the GNU General Public License as published by
11 * the Free Software Foundation; either version 3 of the License, or
12 * (at your option) any later version.
5cb37436 13 *
0ecfbf27
MP
14 * This program is distributed in the hope that it will be useful,
15 * but WITHOUT ANY WARRANTY; without even the implied warranty of
16 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
17 * GNU General Public License for more details.
5cb37436 18 *
e7c67065 19 * You should have received a copy of the GNU General Public License along
4fd842f9 20 * with this program; if not, visit the http://fsf.org website.
0ecfbf27 21 */
c627d613 22
c627d613 23#include "rsync.h"
1b42f628 24#include "ifuncs.h"
c627d613 25
c7c11a0d 26extern int verbose;
bc6ebcd2
WD
27extern int dry_run;
28extern int module_id;
29extern int modify_window;
7e43da81 30extern int relative_paths;
e175fb07 31extern int human_readable;
7c73536c 32extern char *module_dir;
a8167c66 33extern unsigned int module_dirlen;
e9357a2d 34extern mode_t orig_umask;
a7260c40 35extern char *partial_dir;
7842418b 36extern struct filter_list_struct server_filter_list;
c7c11a0d 37
0ecfbf27
MP
38int sanitize_paths = 0;
39
a8167c66
WD
40char curr_dir[MAXPATHLEN];
41unsigned int curr_dir_len;
42int curr_dir_depth; /* This is only set for a sanitizing daemon. */
43
0f78b815 44/* Set a fd into nonblocking mode. */
f0359dd0
AT
45void set_nonblocking(int fd)
46{
47 int val;
48
ff530f04 49 if ((val = fcntl(fd, F_GETFL)) == -1)
f0359dd0
AT
50 return;
51 if (!(val & NONBLOCK_FLAG)) {
52 val |= NONBLOCK_FLAG;
53 fcntl(fd, F_SETFL, val);
54 }
55}
56
0f78b815 57/* Set a fd into blocking mode. */
36349ea0
AT
58void set_blocking(int fd)
59{
60 int val;
61
ff530f04 62 if ((val = fcntl(fd, F_GETFL)) == -1)
36349ea0
AT
63 return;
64 if (val & NONBLOCK_FLAG) {
65 val &= ~NONBLOCK_FLAG;
66 fcntl(fd, F_SETFL, val);
67 }
68}
69
ac13ad10 70/**
0ecfbf27
MP
71 * Create a file descriptor pair - like pipe() but use socketpair if
72 * possible (because of blocking issues on pipes).
5cb37436 73 *
0ecfbf27 74 * Always set non-blocking.
f0359dd0 75 */
08f15335
AT
76int fd_pair(int fd[2])
77{
f0359dd0
AT
78 int ret;
79
4f5b0756 80#ifdef HAVE_SOCKETPAIR
f0359dd0 81 ret = socketpair(AF_UNIX, SOCK_STREAM, 0, fd);
08f15335 82#else
f0359dd0 83 ret = pipe(fd);
08f15335 84#endif
f0359dd0
AT
85
86 if (ret == 0) {
87 set_nonblocking(fd[0]);
88 set_nonblocking(fd[1]);
89 }
0ecfbf27 90
f0359dd0 91 return ret;
08f15335
AT
92}
93
0b515981 94void print_child_argv(const char *prefix, char **cmd)
5ad0e46f 95{
0b515981 96 rprintf(FCLIENT, "%s ", prefix);
5ad0e46f
MP
97 for (; *cmd; cmd++) {
98 /* Look for characters that ought to be quoted. This
99 * is not a great quoting algorithm, but it's
100 * sufficient for a log message. */
101 if (strspn(*cmd, "abcdefghijklmnopqrstuvwxyz"
102 "ABCDEFGHIJKLMNOPQRSTUVWXYZ"
103 "0123456789"
104 ",.-_=+@/") != strlen(*cmd)) {
a3f6dbdf 105 rprintf(FCLIENT, "\"%s\" ", *cmd);
5ad0e46f 106 } else {
a3f6dbdf 107 rprintf(FCLIENT, "%s ", *cmd);
5ad0e46f
MP
108 }
109 }
a3f6dbdf 110 rprintf(FCLIENT, "\n");
5ad0e46f
MP
111}
112
4a19c3b2 113NORETURN void out_of_memory(const char *str)
c627d613 114{
2cd421d8 115 rprintf(FERROR, "ERROR: out of memory in %s [%s]\n", str, who_am_i());
c284f34a 116 exit_cleanup(RERR_MALLOC);
575f2fca
AT
117}
118
4a19c3b2 119NORETURN void overflow_exit(const char *str)
575f2fca 120{
2cd421d8 121 rprintf(FERROR, "ERROR: buffer overflow in %s [%s]\n", str, who_am_i());
c284f34a 122 exit_cleanup(RERR_MALLOC);
c627d613
AT
123}
124
4a19c3b2 125int set_modtime(const char *fname, time_t modtime, mode_t mode)
c627d613 126{
25007999
WD
127#if !defined HAVE_LUTIMES || !defined HAVE_UTIMES
128 if (S_ISLNK(mode))
129 return 1;
130#endif
131
404e813c
MP
132 if (verbose > 2) {
133 rprintf(FINFO, "set modtime of %s to (%ld) %s",
785abd48 134 fname, (long)modtime,
404e813c
MP
135 asctime(localtime(&modtime)));
136 }
5cb37436 137
15778afb
WD
138 if (dry_run)
139 return 0;
140
31e12522 141 {
25007999
WD
142#ifdef HAVE_UTIMES
143 struct timeval t[2];
144 t[0].tv_sec = time(NULL);
145 t[0].tv_usec = 0;
146 t[1].tv_sec = modtime;
147 t[1].tv_usec = 0;
148# ifdef HAVE_LUTIMES
149a78e3 149 if (S_ISLNK(mode)) {
77d4c400
WD
150 if (lutimes(fname, t) < 0 && errno != ENOSYS)
151 return -1;
152 return 0;
149a78e3 153 }
25007999
WD
154# endif
155 return utimes(fname, t);
4f9b139a 156#elif defined HAVE_STRUCT_UTIMBUF
5cb37436 157 struct utimbuf tbuf;
31e12522
AT
158 tbuf.actime = time(NULL);
159 tbuf.modtime = modtime;
160 return utime(fname,&tbuf);
4f5b0756 161#elif defined HAVE_UTIME
31e12522
AT
162 time_t t[2];
163 t[0] = time(NULL);
164 t[1] = modtime;
165 return utime(fname,t);
c627d613 166#else
25007999 167#error No file-time-modification routine found!
c627d613 168#endif
31e12522 169 }
c627d613 170}
94481d91 171
e9357a2d
WD
172/* This creates a new directory with default permissions. Since there
173 * might be some directory-default permissions affecting this, we can't
174 * force the permissions directly using the original umask and mkdir(). */
175int mkdir_defmode(char *fname)
176{
177 int ret;
178
179 umask(orig_umask);
180 ret = do_mkdir(fname, ACCESSPERMS);
181 umask(0);
182
183 return ret;
184}
185
85c41757
WD
186/* Create any necessary directories in fname. Any missing directories are
187 * created with default permissions. */
e9357a2d 188int create_directory_path(char *fname)
6574b4f7 189{
6574b4f7 190 char *p;
85c41757 191 int ret = 0;
6574b4f7 192
c284f34a
WD
193 while (*fname == '/')
194 fname++;
195 while (strncmp(fname, "./", 2) == 0)
196 fname += 2;
6574b4f7 197
85c41757 198 umask(orig_umask);
6574b4f7 199 p = fname;
c284f34a 200 while ((p = strchr(p,'/')) != NULL) {
85c41757
WD
201 *p = '\0';
202 if (do_mkdir(fname, ACCESSPERMS) < 0 && errno != EEXIST)
203 ret = -1;
204 *p++ = '/';
6574b4f7 205 }
85c41757
WD
206 umask(0);
207
208 return ret;
6574b4f7 209}
950ab32d 210
ac13ad10
MP
211/**
212 * Write @p len bytes at @p ptr to descriptor @p desc, retrying if
213 * interrupted.
214 *
215 * @retval len upon success
216 *
217 * @retval <0 write's (negative) error code
218 *
219 * Derived from GNU C's cccp.c.
220 */
4a19c3b2 221int full_write(int desc, const char *ptr, size_t len)
950ab32d
AT
222{
223 int total_written;
5cb37436 224
950ab32d
AT
225 total_written = 0;
226 while (len > 0) {
5c1b7bfd 227 int written = write(desc, ptr, len);
950ab32d 228 if (written < 0) {
950ab32d
AT
229 if (errno == EINTR)
230 continue;
950ab32d
AT
231 return written;
232 }
233 total_written += written;
234 ptr += written;
235 len -= written;
236 }
237 return total_written;
238}
239
ac13ad10
MP
240/**
241 * Read @p len bytes at @p ptr from descriptor @p desc, retrying if
242 * interrupted.
243 *
244 * @retval >0 the actual number of bytes read
245 *
246 * @retval 0 for EOF
247 *
248 * @retval <0 for an error.
249 *
250 * Derived from GNU C's cccp.c. */
9dd891bb 251static int safe_read(int desc, char *ptr, size_t len)
950ab32d
AT
252{
253 int n_chars;
5cb37436 254
9dd891bb 255 if (len == 0)
950ab32d 256 return len;
5cb37436 257
950ab32d
AT
258 do {
259 n_chars = read(desc, ptr, len);
260 } while (n_chars < 0 && errno == EINTR);
5cb37436 261
950ab32d
AT
262 return n_chars;
263}
264
2cce7545
WD
265/* Copy a file. If ofd < 0, copy_file unlinks and opens the "dest" file.
266 * Otherwise, it just writes to and closes the provided file descriptor.
ac13ad10 267 *
3e13004b
WD
268 * This is used in conjunction with the --temp-dir, --backup, and
269 * --copy-dest options. */
2cce7545
WD
270int copy_file(const char *source, const char *dest, int ofd,
271 mode_t mode, int create_bak_dir)
950ab32d
AT
272{
273 int ifd;
950ab32d
AT
274 char buf[1024 * 8];
275 int len; /* Number of bytes read into `buf'. */
276
f5b532b1 277 if ((ifd = do_open(source, O_RDONLY, 0)) < 0) {
3f0211b6 278 rsyserr(FERROR_XFER, errno, "open %s", full_fname(source));
950ab32d
AT
279 return -1;
280 }
281
2cce7545
WD
282 if (ofd < 0) {
283 if (robust_unlink(dest) && errno != ENOENT) {
3f0211b6 284 rsyserr(FERROR_XFER, errno, "unlink %s", full_fname(dest));
2cce7545
WD
285 return -1;
286 }
950ab32d 287
2cce7545
WD
288 if ((ofd = do_open(dest, O_WRONLY | O_CREAT | O_TRUNC | O_EXCL, mode)) < 0
289 && (!create_bak_dir || errno != ENOENT || make_bak_dir(dest) < 0
290 || (ofd = do_open(dest, O_WRONLY | O_CREAT | O_TRUNC | O_EXCL, mode)) < 0)) {
3f0211b6 291 rsyserr(FERROR_XFER, errno, "open %s", full_fname(dest));
2cce7545
WD
292 close(ifd);
293 return -1;
294 }
950ab32d
AT
295 }
296
5cb37436 297 while ((len = safe_read(ifd, buf, sizeof buf)) > 0) {
950ab32d 298 if (full_write(ofd, buf, len) < 0) {
3f0211b6 299 rsyserr(FERROR_XFER, errno, "write %s", full_fname(dest));
950ab32d
AT
300 close(ifd);
301 close(ofd);
302 return -1;
303 }
304 }
305
8b602edd 306 if (len < 0) {
3f0211b6 307 rsyserr(FERROR_XFER, errno, "read %s", full_fname(source));
8b602edd
WD
308 close(ifd);
309 close(ofd);
310 return -1;
311 }
312
9f27cd8c 313 if (close(ifd) < 0) {
3f0211b6 314 rsyserr(FWARNING, errno, "close failed on %s",
d62bcc17 315 full_fname(source));
9f27cd8c
WD
316 }
317
318 if (close(ofd) < 0) {
3f0211b6 319 rsyserr(FERROR_XFER, errno, "close failed on %s",
d62bcc17 320 full_fname(dest));
9f27cd8c
WD
321 return -1;
322 }
950ab32d 323
950ab32d
AT
324 return 0;
325}
feaa89c4 326
c7c11a0d
DD
327/* MAX_RENAMES should be 10**MAX_RENAMES_DIGITS */
328#define MAX_RENAMES_DIGITS 3
329#define MAX_RENAMES 1000
330
ac13ad10 331/**
b4235b31
MP
332 * Robust unlink: some OS'es (HPUX) refuse to unlink busy files, so
333 * rename to <path>/.rsyncNNN instead.
334 *
335 * Note that successive rsync runs will shuffle the filenames around a
336 * bit as long as the file is still busy; this is because this function
337 * does not know if the unlink call is due to a new file coming in, or
338 * --delete trying to remove old .rsyncNNN files, hence it renames it
339 * each time.
340 **/
63cf5ae7 341int robust_unlink(const char *fname)
c7c11a0d
DD
342{
343#ifndef ETXTBSY
344 return do_unlink(fname);
345#else
346 static int counter = 1;
347 int rc, pos, start;
348 char path[MAXPATHLEN];
349
350 rc = do_unlink(fname);
c284f34a 351 if (rc == 0 || errno != ETXTBSY)
c7c11a0d
DD
352 return rc;
353
c284f34a
WD
354 if ((pos = strlcpy(path, fname, MAXPATHLEN)) >= MAXPATHLEN)
355 pos = MAXPATHLEN - 1;
c7c11a0d 356
c284f34a
WD
357 while (pos > 0 && path[pos-1] != '/')
358 pos--;
5cb37436 359 pos += strlcpy(path+pos, ".rsync", MAXPATHLEN-pos);
c7c11a0d
DD
360
361 if (pos > (MAXPATHLEN-MAX_RENAMES_DIGITS-1)) {
362 errno = ETXTBSY;
363 return -1;
364 }
365
366 /* start where the last one left off to reduce chance of clashes */
367 start = counter;
368 do {
434c4098 369 snprintf(&path[pos], MAX_RENAMES_DIGITS+1, "%03d", counter);
c7c11a0d
DD
370 if (++counter >= MAX_RENAMES)
371 counter = 1;
c284f34a 372 } while ((rc = access(path, 0)) == 0 && counter != start);
c7c11a0d 373
4791825d 374 if (verbose > 0) {
3f0211b6 375 rprintf(FWARNING, "renaming %s to %s because of text busy\n",
785abd48 376 fname, path);
4791825d 377 }
c7c11a0d
DD
378
379 /* maybe we should return rename()'s exit status? Nah. */
380 if (do_rename(fname, path) != 0) {
381 errno = ETXTBSY;
382 return -1;
383 }
384 return 0;
385#endif
386}
387
630f548f 388/* Returns 0 on successful rename, 1 if we successfully copied the file
3ed8eafc
WD
389 * across filesystems, -2 if copy_file() failed, and -1 on other errors.
390 * If partialptr is not NULL and we need to do a copy, copy the file into
391 * the active partial-dir instead of over the destination file. */
4a19c3b2 392int robust_rename(const char *from, const char *to, const char *partialptr,
3ed8eafc 393 int mode)
c7c11a0d 394{
62c9e6b3
WD
395 int tries = 4;
396
397 while (tries--) {
398 if (do_rename(from, to) == 0)
399 return 0;
400
401 switch (errno) {
402#ifdef ETXTBSY
403 case ETXTBSY:
404 if (robust_unlink(to) != 0)
405 return -1;
406 break;
c7c11a0d 407#endif
62c9e6b3 408 case EXDEV:
3ed8eafc
WD
409 if (partialptr) {
410 if (!handle_partial_dir(partialptr,PDIR_CREATE))
411 return -1;
412 to = partialptr;
413 }
2cce7545 414 if (copy_file(from, to, -1, mode, 0) != 0)
62c9e6b3
WD
415 return -2;
416 do_unlink(from);
630f548f 417 return 1;
62c9e6b3
WD
418 default:
419 return -1;
420 }
421 }
422 return -1;
feaa89c4 423}
3ba62a83 424
3ba62a83
AT
425static pid_t all_pids[10];
426static int num_pids;
427
4cf64834 428/** Fork and record the pid of the child. **/
3ba62a83
AT
429pid_t do_fork(void)
430{
431 pid_t newpid = fork();
5cb37436 432
4cf64834 433 if (newpid != 0 && newpid != -1) {
3ba62a83
AT
434 all_pids[num_pids++] = newpid;
435 }
436 return newpid;
437}
438
4cf64834
MP
439/**
440 * Kill all children.
441 *
442 * @todo It would be kind of nice to make sure that they are actually
443 * all our children before we kill them, because their pids may have
444 * been recycled by some other process. Perhaps when we wait for a
445 * child, we should remove it from this array. Alternatively we could
446 * perhaps use process groups, but I think that would not work on
447 * ancient Unix versions that don't support them.
448 **/
3ba62a83
AT
449void kill_all(int sig)
450{
451 int i;
4cf64834
MP
452
453 for (i = 0; i < num_pids; i++) {
454 /* Let's just be a little careful where we
455 * point that gun, hey? See kill(2) for the
456 * magic caused by negative values. */
457 pid_t p = all_pids[i];
458
459 if (p == getpid())
460 continue;
461 if (p <= 0)
462 continue;
463
464 kill(p, sig);
3ba62a83
AT
465 }
466}
9486289c 467
ac13ad10 468/** Turn a user name into a uid */
4a19c3b2 469int name_to_uid(const char *name, uid_t *uid)
8ef4ffd6
AT
470{
471 struct passwd *pass;
b5bd5542
WD
472 if (!name || !*name)
473 return 0;
8ef4ffd6
AT
474 pass = getpwnam(name);
475 if (pass) {
476 *uid = pass->pw_uid;
477 return 1;
478 }
479 return 0;
480}
481
ac13ad10 482/** Turn a group name into a gid */
4a19c3b2 483int name_to_gid(const char *name, gid_t *gid)
8ef4ffd6
AT
484{
485 struct group *grp;
b5bd5542
WD
486 if (!name || !*name)
487 return 0;
8ef4ffd6
AT
488 grp = getgrnam(name);
489 if (grp) {
490 *gid = grp->gr_gid;
491 return 1;
492 }
493 return 0;
494}
495
ac13ad10 496/** Lock a byte range in a open file */
31593dd6 497int lock_range(int fd, int offset, int len)
0c515f17 498{
31593dd6 499 struct flock lock;
0c515f17 500
31593dd6
AT
501 lock.l_type = F_WRLCK;
502 lock.l_whence = SEEK_SET;
503 lock.l_start = offset;
504 lock.l_len = len;
505 lock.l_pid = 0;
5cb37436 506
31593dd6 507 return fcntl(fd,F_SETLK,&lock) == 0;
0c515f17 508}
874895d5 509
7842418b 510static int filter_server_path(char *arg)
4791825d
WD
511{
512 char *s;
4791825d 513
7842418b 514 if (server_filter_list.head) {
4791825d
WD
515 for (s = arg; (s = strchr(s, '/')) != NULL; ) {
516 *s = '\0';
7842418b 517 if (check_filter(&server_filter_list, arg, 1) < 0) {
4791825d
WD
518 /* We must leave arg truncated! */
519 return 1;
520 }
521 *s++ = '/';
522 }
523 }
524 return 0;
525}
874895d5 526
044dc293 527void glob_expand(char *s, char ***argv_ptr, int *argc_ptr, int *maxargs_ptr)
874895d5 528{
b7061c82 529 char **argv = *argv_ptr;
b5bd5542 530 int argc = *argc_ptr;
b7061c82 531 int maxargs = *maxargs_ptr;
4f5b0756 532#if !defined HAVE_GLOB || !defined HAVE_GLOB_H
b7061c82
WD
533 if (argc == maxargs) {
534 maxargs += MAX_ARGS;
535 if (!(argv = realloc_array(argv, char *, maxargs)))
044dc293 536 out_of_memory("glob_expand");
b7061c82
WD
537 *argv_ptr = argv;
538 *maxargs_ptr = maxargs;
539 }
4135d091
WD
540 if (!*s)
541 s = ".";
b5bd5542 542 s = argv[argc++] = strdup(s);
7842418b 543 filter_server_path(s);
874895d5
AT
544#else
545 glob_t globbuf;
874895d5 546
b5bd5542
WD
547 if (maxargs <= argc)
548 return;
4135d091
WD
549 if (!*s)
550 s = ".";
e42c9458 551
4135d091 552 if (sanitize_paths)
29bca53f 553 s = sanitize_path(NULL, s, "", 0);
84a63795
WD
554 else
555 s = strdup(s);
044dc293
WD
556 if (!s)
557 out_of_memory("glob_expand");
087bf010 558
5cb37436 559 memset(&globbuf, 0, sizeof globbuf);
7842418b 560 if (!filter_server_path(s))
4791825d 561 glob(s, 0, NULL, &globbuf);
b7061c82
WD
562 if (MAX((int)globbuf.gl_pathc, 1) > maxargs - argc) {
563 maxargs += globbuf.gl_pathc + MAX_ARGS;
564 if (!(argv = realloc_array(argv, char *, maxargs)))
044dc293 565 out_of_memory("glob_expand");
b7061c82
WD
566 *argv_ptr = argv;
567 *maxargs_ptr = maxargs;
568 }
b5bd5542
WD
569 if (globbuf.gl_pathc == 0)
570 argv[argc++] = s;
571 else {
c8933031 572 int i;
b5bd5542 573 free(s);
c8933031 574 for (i = 0; i < (int)globbuf.gl_pathc; i++) {
b5bd5542 575 if (!(argv[argc++] = strdup(globbuf.gl_pathv[i])))
044dc293 576 out_of_memory("glob_expand");
b5bd5542 577 }
874895d5
AT
578 }
579 globfree(&globbuf);
874895d5 580#endif
b5bd5542 581 *argc_ptr = argc;
874895d5 582}
5a96ee05 583
4791825d 584/* This routine is only used in daemon mode. */
044dc293 585void glob_expand_module(char *base1, char *arg, char ***argv_ptr, int *argc_ptr, int *maxargs_ptr)
087bf010 586{
044dc293 587 char *p, *s;
ba5e128d 588 char *base = base1;
4791825d 589 int base_len = strlen(base);
087bf010 590
044dc293 591 if (!arg || !*arg)
b5bd5542 592 return;
087bf010 593
044dc293
WD
594 if (strncmp(arg, base, base_len) == 0)
595 arg += base_len;
e42c9458 596
044dc293
WD
597 if (!(arg = strdup(arg)))
598 out_of_memory("glob_expand_module");
087bf010 599
b5bd5542 600 if (asprintf(&base," %s/", base1) <= 0)
044dc293 601 out_of_memory("glob_expand_module");
4791825d 602 base_len++;
ba5e128d 603
044dc293
WD
604 for (s = arg; *s; s = p + base_len) {
605 if ((p = strstr(s, base)) != NULL)
b5bd5542 606 *p = '\0'; /* split it at this point */
044dc293 607 glob_expand(s, argv_ptr, argc_ptr, maxargs_ptr);
b5bd5542
WD
608 if (!p)
609 break;
087bf010
AT
610 }
611
044dc293 612 free(arg);
ba5e128d 613 free(base);
087bf010 614}
5a96ee05 615
ac13ad10
MP
616/**
617 * Convert a string to lower case
618 **/
5a96ee05
AT
619void strlower(char *s)
620{
621 while (*s) {
2dc7b8bd
WD
622 if (isUpper(s))
623 *s = toLower(s);
5a96ee05
AT
624 s++;
625 }
626}
e42c9458 627
368ad70e
WD
628/* Join strings p1 & p2 into "dest" with a guaranteed '/' between them. (If
629 * p1 ends with a '/', no extra '/' is inserted.) Returns the length of both
a8f7e4b8
WD
630 * strings + 1 (if '/' was inserted), regardless of whether the null-terminated
631 * string fits into destsize. */
368ad70e
WD
632size_t pathjoin(char *dest, size_t destsize, const char *p1, const char *p2)
633{
634 size_t len = strlcpy(dest, p1, destsize);
635 if (len < destsize - 1) {
636 if (!len || dest[len-1] != '/')
637 dest[len++] = '/';
638 if (len < destsize - 1)
639 len += strlcpy(dest + len, p2, destsize - len);
640 else {
641 dest[len] = '\0';
642 len += strlen(p2);
643 }
644 }
645 else
646 len += strlen(p2) + 1; /* Assume we'd insert a '/'. */
647 return len;
648}
649
650/* Join any number of strings together, putting them in "dest". The return
a8f7e4b8
WD
651 * value is the length of all the strings, regardless of whether the null-
652 * terminated whole fits in destsize. Your list of string pointers must end
653 * with a NULL to indicate the end of the list. */
368ad70e
WD
654size_t stringjoin(char *dest, size_t destsize, ...)
655{
5cb37436 656 va_list ap;
368ad70e
WD
657 size_t len, ret = 0;
658 const char *src;
659
660 va_start(ap, destsize);
661 while (1) {
662 if (!(src = va_arg(ap, const char *)))
663 break;
664 len = strlen(src);
665 ret += len;
666 if (destsize > 1) {
667 if (len >= destsize)
668 len = destsize - 1;
669 memcpy(dest, src, len);
670 destsize -= len;
671 dest += len;
672 }
673 }
674 *dest = '\0';
675 va_end(ap);
676
677 return ret;
678}
679
1d6b8f9a
WD
680int count_dir_elements(const char *p)
681{
682 int cnt = 0, new_component = 1;
683 while (*p) {
684 if (*p++ == '/')
a8167c66 685 new_component = (*p != '.' || (p[1] != '/' && p[1] != '\0'));
1d6b8f9a
WD
686 else if (new_component) {
687 new_component = 0;
688 cnt++;
689 }
690 }
691 return cnt;
692}
693
6bb82fe0
WD
694/* Turns multiple adjacent slashes into a single slash, drops interior "."
695 * elements, drops an intial "./" unless CFN_KEEP_LEADING_DOT_DIR is flagged,
696 * will even drop a trailing '.' after a '/' if CFN_DROP_TRAILING_DOT_DIR is
697 * flagged, removes a trailing slash (perhaps after removing the aforementioned
698 * dot) unless CFN_KEEP_TRAILING_SLASH is flagged, will even collapse ".."
699 * elements (except at the start of the string) if CFN_COLLAPSE_DOT_DOT_DIRS
700 * is flagged. If the resulting name would be empty, we return ".". */
701unsigned int clean_fname(char *name, int flags)
5243c216 702{
e012b94f 703 char *limit = name - 1, *t = name, *f = name;
ebdd24d6 704 int anchored;
5243c216 705
b5bd5542 706 if (!name)
3104620c 707 return 0;
5243c216 708
ebdd24d6
WD
709 if ((anchored = *f == '/') != 0)
710 *t++ = *f++;
6bb82fe0
WD
711 else if (flags & CFN_KEEP_LEADING_DOT_DIR && *f == '.' && f[1] == '/') {
712 *t++ = *f++;
713 *t++ = *f++;
714 }
ebdd24d6
WD
715 while (*f) {
716 /* discard extra slashes */
717 if (*f == '/') {
718 f++;
719 continue;
5243c216 720 }
ebdd24d6 721 if (*f == '.') {
6bb82fe0 722 /* discard interior "." dirs */
ebdd24d6 723 if (f[1] == '/') {
e012b94f 724 f += 2;
ebdd24d6
WD
725 continue;
726 }
6bb82fe0
WD
727 if (f[1] == '\0' && flags & CFN_DROP_TRAILING_DOT_DIR)
728 break;
ebdd24d6 729 /* collapse ".." dirs */
6bb82fe0
WD
730 if (flags & CFN_COLLAPSE_DOT_DOT_DIRS
731 && f[1] == '.' && (f[2] == '/' || !f[2])) {
ebdd24d6
WD
732 char *s = t - 1;
733 if (s == name && anchored) {
734 f += 2;
735 continue;
736 }
737 while (s > limit && *--s != '/') {}
e012b94f 738 if (s != t - 1 && (s < name || *s == '/')) {
ebdd24d6
WD
739 t = s + 1;
740 f += 2;
741 continue;
742 }
f55c2dfc 743 limit = t + 2;
5243c216
AT
744 }
745 }
ebdd24d6 746 while (*f && (*t++ = *f++) != '/') {}
5243c216 747 }
ebdd24d6 748
6bb82fe0 749 if (t > name+anchored && t[-1] == '/' && !(flags & CFN_KEEP_TRAILING_SLASH))
ebdd24d6
WD
750 t--;
751 if (t == name)
752 *t++ = '.';
753 *t = '\0';
3104620c
WD
754
755 return t - name;
5243c216
AT
756}
757
84a63795
WD
758/* Make path appear as if a chroot had occurred. This handles a leading
759 * "/" (either removing it or expanding it) and any leading or embedded
760 * ".." components that attempt to escape past the module's top dir.
b4235b31 761 *
1d6b8f9a
WD
762 * If dest is NULL, a buffer is allocated to hold the result. It is legal
763 * to call with the dest and the path (p) pointing to the same buffer, but
764 * rootdir will be ignored to avoid expansion of the string.
b4235b31 765 *
1d6b8f9a 766 * The rootdir string contains a value to use in place of a leading slash.
7c73536c 767 * Specify NULL to get the default of "module_dir".
ac13ad10 768 *
a8167c66 769 * The depth var is a count of how many '..'s to allow at the start of the
29bca53f 770 * path.
ac13ad10 771 *
b92693da 772 * We also clean the path in a manner similar to clean_fname() but with a
0f78b815 773 * few differences:
b92693da
WD
774 *
775 * Turns multiple adjacent slashes into a single slash, gets rid of "." dir
776 * elements (INCLUDING a trailing dot dir), PRESERVES a trailing slash, and
777 * ALWAYS collapses ".." elements (except for those at the start of the
778 * string up to "depth" deep). If the resulting name would be empty,
779 * change it into a ".". */
29bca53f 780char *sanitize_path(char *dest, const char *p, const char *rootdir, int depth)
1b8e662a 781{
29bca53f 782 char *start, *sanp;
7e43da81 783 int rlen = 0, leave_one_dotdir = relative_paths;
84a63795
WD
784
785 if (dest != p) {
786 int plen = strlen(p);
1d6b8f9a
WD
787 if (*p == '/') {
788 if (!rootdir)
7c73536c 789 rootdir = module_dir;
1d6b8f9a
WD
790 rlen = strlen(rootdir);
791 depth = 0;
84a63795
WD
792 p++;
793 }
794 if (dest) {
795 if (rlen + plen + 1 >= MAXPATHLEN)
796 return NULL;
797 } else if (!(dest = new_array(char, rlen + plen + 1)))
798 out_of_memory("sanitize_path");
799 if (rlen) {
1d6b8f9a 800 memcpy(dest, rootdir, rlen);
84a63795
WD
801 if (rlen > 1)
802 dest[rlen++] = '/';
803 }
804 }
cb13abfe 805
84a63795 806 start = sanp = dest + rlen;
29bca53f 807 while (*p) {
2d41264e
WD
808 /* discard leading or extra slashes */
809 if (*p == '/') {
810 p++;
811 continue;
812 }
b5f9e67d 813 /* this loop iterates once per filename component in p.
44e2e578 814 * both p (and sanp if the original had a slash) should
b5f9e67d
DD
815 * always be left pointing after a slash
816 */
c284f34a 817 if (*p == '.' && (p[1] == '/' || p[1] == '\0')) {
35812ea1 818 if (leave_one_dotdir && p[1])
7e43da81
WD
819 leave_one_dotdir = 0;
820 else {
821 /* skip "." component */
822 p++;
823 continue;
824 }
cb13abfe 825 }
c284f34a 826 if (*p == '.' && p[1] == '.' && (p[2] == '/' || p[2] == '\0')) {
cb13abfe 827 /* ".." component followed by slash or end */
8e5f029e
WD
828 if (depth <= 0 || sanp != start) {
829 p += 2;
830 if (sanp != start) {
831 /* back up sanp one level */
832 --sanp; /* now pointing at slash */
833 while (sanp > start && sanp[-1] != '/') {
834 /* skip back up to slash */
835 sanp--;
836 }
b5f9e67d 837 }
8e5f029e 838 continue;
1b8e662a 839 }
8e5f029e
WD
840 /* allow depth levels of .. at the beginning */
841 depth--;
842 /* move the virtual beginning to leave the .. alone */
843 start = sanp + 3;
1b8e662a 844 }
2d41264e
WD
845 /* copy one component through next slash */
846 while (*p && (*sanp++ = *p++) != '/') {}
1b8e662a 847 }
84a63795 848 if (sanp == dest) {
b5f9e67d 849 /* ended up with nothing, so put in "." component */
44e2e578 850 *sanp++ = '.';
b5f9e67d 851 }
44e2e578 852 *sanp = '\0';
1b8e662a 853
84a63795 854 return dest;
14b61c63 855}
5243c216 856
a8167c66 857/* Like chdir(), but it keeps track of the current directory (in the
a16d8f2b 858 * global "curr_dir"), and ensures that the path size doesn't overflow.
a8167c66 859 * Also cleans the path using the clean_fname() function. */
2c708476 860int push_dir(const char *dir, int set_path_only)
5243c216 861{
5243c216 862 static int initialised;
4af8fe4e 863 unsigned int len;
5243c216
AT
864
865 if (!initialised) {
866 initialised = 1;
5cb37436 867 getcwd(curr_dir, sizeof curr_dir - 1);
4af8fe4e 868 curr_dir_len = strlen(curr_dir);
5243c216
AT
869 }
870
4af8fe4e
WD
871 if (!dir) /* this call was probably just to initialize */
872 return 0;
c226b7c2 873
4af8fe4e
WD
874 len = strlen(dir);
875 if (len == 1 && *dir == '.')
876 return 1;
5243c216 877
7c73536c
WD
878 if ((*dir == '/' ? len : curr_dir_len + 1 + len) >= sizeof curr_dir) {
879 errno = ENAMETOOLONG;
4af8fe4e 880 return 0;
7c73536c 881 }
4af8fe4e 882
ba081be3 883 if (!set_path_only && chdir(dir))
4af8fe4e 884 return 0;
5243c216
AT
885
886 if (*dir == '/') {
4af8fe4e
WD
887 memcpy(curr_dir, dir, len + 1);
888 curr_dir_len = len;
889 } else {
890 curr_dir[curr_dir_len++] = '/';
891 memcpy(curr_dir + curr_dir_len, dir, len + 1);
892 curr_dir_len += len;
5243c216
AT
893 }
894
6bb82fe0 895 curr_dir_len = clean_fname(curr_dir, CFN_COLLAPSE_DOT_DOT_DIRS);
a8167c66
WD
896 if (sanitize_paths) {
897 if (module_dirlen > curr_dir_len)
898 module_dirlen = curr_dir_len;
899 curr_dir_depth = count_dir_elements(curr_dir + module_dirlen);
900 }
5243c216 901
af5ed0f2
WD
902 if (verbose >= 5)
903 rprintf(FINFO, "[%s] dir is now %s\n", who_am_i(), curr_dir);
904
4af8fe4e 905 return 1;
5243c216
AT
906}
907
a16d8f2b
WD
908/**
909 * Reverse a push_dir() call. You must pass in an absolute path
910 * that was copied from a prior value of "curr_dir".
911 **/
2c708476 912int pop_dir(const char *dir)
5243c216 913{
4af8fe4e
WD
914 if (chdir(dir))
915 return 0;
5243c216 916
4af8fe4e
WD
917 curr_dir_len = strlcpy(curr_dir, dir, sizeof curr_dir);
918 if (curr_dir_len >= sizeof curr_dir)
919 curr_dir_len = sizeof curr_dir - 1;
a8167c66
WD
920 if (sanitize_paths)
921 curr_dir_depth = count_dir_elements(curr_dir + module_dirlen);
5243c216 922
4af8fe4e 923 return 1;
5243c216 924}
aa9b77a5 925
eb61be19
WD
926/**
927 * Return a quoted string with the full pathname of the indicated filename.
928 * The string " (in MODNAME)" may also be appended. The returned pointer
929 * remains valid until the next time full_fname() is called.
930 **/
9a5ade18 931char *full_fname(const char *fn)
eb61be19 932{
eb61be19
WD
933 static char *result = NULL;
934 char *m1, *m2, *m3;
935 char *p1, *p2;
936
937 if (result)
938 free(result);
939
940 if (*fn == '/')
941 p1 = p2 = "";
942 else {
4cff5fa4 943 p1 = curr_dir + module_dirlen;
bc83274a
WD
944 for (p2 = p1; *p2 == '/'; p2++) {}
945 if (*p2)
946 p2 = "/";
eb61be19
WD
947 }
948 if (module_id >= 0) {
949 m1 = " (in ";
950 m2 = lp_name(module_id);
951 m3 = ")";
eb61be19
WD
952 } else
953 m1 = m2 = m3 = "";
954
c98ad3df
WD
955 if (asprintf(&result, "\"%s%s%s\"%s%s%s", p1, p2, fn, m1, m2, m3) <= 0)
956 out_of_memory("full_fname");
eb61be19
WD
957
958 return result;
959}
960
a7260c40
WD
961static char partial_fname[MAXPATHLEN];
962
963char *partial_dir_fname(const char *fname)
964{
965 char *t = partial_fname;
966 int sz = sizeof partial_fname;
967 const char *fn;
968
969 if ((fn = strrchr(fname, '/')) != NULL) {
970 fn++;
971 if (*partial_dir != '/') {
972 int len = fn - fname;
973 strncpy(t, fname, len); /* safe */
974 t += len;
975 sz -= len;
976 }
977 } else
978 fn = fname;
979 if ((int)pathjoin(t, sz, partial_dir, fn) >= sz)
980 return NULL;
5aa7b20a 981 if (server_filter_list.head) {
89363676
WD
982 t = strrchr(partial_fname, '/');
983 *t = '\0';
5aa7b20a
WD
984 if (check_filter(&server_filter_list, partial_fname, 1) < 0)
985 return NULL;
89363676 986 *t = '/';
5aa7b20a
WD
987 if (check_filter(&server_filter_list, partial_fname, 0) < 0)
988 return NULL;
989 }
a7260c40
WD
990
991 return partial_fname;
992}
993
994/* If no --partial-dir option was specified, we don't need to do anything
995 * (the partial-dir is essentially '.'), so just return success. */
996int handle_partial_dir(const char *fname, int create)
997{
998 char *fn, *dir;
999
1000 if (fname != partial_fname)
1001 return 1;
1002 if (!create && *partial_dir == '/')
1003 return 1;
1004 if (!(fn = strrchr(partial_fname, '/')))
1005 return 1;
1006
1007 *fn = '\0';
1008 dir = partial_fname;
1009 if (create) {
1010 STRUCT_STAT st;
a7260c40 1011 int statret = do_lstat(dir, &st);
a7260c40
WD
1012 if (statret == 0 && !S_ISDIR(st.st_mode)) {
1013 if (do_unlink(dir) < 0)
1014 return 0;
1015 statret = -1;
1016 }
1017 if (statret < 0 && do_mkdir(dir, 0700) < 0)
1018 return 0;
1019 } else
1020 do_rmdir(dir);
1021 *fn = '/';
1022
1023 return 1;
1024}
1025
ac13ad10
MP
1026/**
1027 * Determine if a symlink points outside the current directory tree.
036e70b0
MP
1028 * This is considered "unsafe" because e.g. when mirroring somebody
1029 * else's machine it might allow them to establish a symlink to
1030 * /etc/passwd, and then read it through a web server.
1031 *
4e5db0ad
MP
1032 * Null symlinks and absolute symlinks are always unsafe.
1033 *
1034 * Basically here we are concerned with symlinks whose target contains
1035 * "..", because this might cause us to walk back up out of the
1036 * transferred directory. We are not allowed to go back up and
1037 * reenter.
1038 *
036e70b0
MP
1039 * @param dest Target of the symlink in question.
1040 *
25d34a5c 1041 * @param src Top source directory currently applicable. Basically this
036e70b0 1042 * is the first parameter to rsync in a simple invocation, but it's
25d34a5c 1043 * modified by flist.c in slightly complex ways.
036e70b0
MP
1044 *
1045 * @retval True if unsafe
1046 * @retval False is unsafe
4e5db0ad
MP
1047 *
1048 * @sa t_unsafe.c
ac13ad10 1049 **/
7afa3a4a 1050int unsafe_symlink(const char *dest, const char *src)
4b957c22 1051{
7afa3a4a 1052 const char *name, *slash;
4b957c22
AT
1053 int depth = 0;
1054
1055 /* all absolute and null symlinks are unsafe */
b5bd5542
WD
1056 if (!dest || !*dest || *dest == '/')
1057 return 1;
4b957c22
AT
1058
1059 /* find out what our safety margin is */
7afa3a4a
WD
1060 for (name = src; (slash = strchr(name, '/')) != 0; name = slash+1) {
1061 if (strncmp(name, "../", 3) == 0) {
c284f34a 1062 depth = 0;
7afa3a4a 1063 } else if (strncmp(name, "./", 2) == 0) {
4b957c22
AT
1064 /* nothing */
1065 } else {
1066 depth++;
1067 }
1068 }
7afa3a4a
WD
1069 if (strcmp(name, "..") == 0)
1070 depth = 0;
4b957c22 1071
7afa3a4a
WD
1072 for (name = dest; (slash = strchr(name, '/')) != 0; name = slash+1) {
1073 if (strncmp(name, "../", 3) == 0) {
1074 /* if at any point we go outside the current directory
1075 then stop - it is unsafe */
1076 if (--depth < 0)
1077 return 1;
1078 } else if (strncmp(name, "./", 2) == 0) {
4b957c22
AT
1079 /* nothing */
1080 } else {
1081 depth++;
1082 }
4b957c22 1083 }
7afa3a4a
WD
1084 if (strcmp(name, "..") == 0)
1085 depth--;
4b957c22 1086
4b957c22
AT
1087 return (depth < 0);
1088}
375a4556 1089
e175fb07
WD
1090/* Return the int64 number as a string. If the --human-readable option was
1091 * specified, we may output the number in K, M, or G units. We can return
1092 * up to 4 buffers at a time. */
1093char *human_num(int64 num)
1094{
1095 static char bufs[4][128]; /* more than enough room */
1096 static unsigned int n;
1097 char *s;
1098
1099 n = (n + 1) % (sizeof bufs / sizeof bufs[0]);
1100
1101 if (human_readable) {
1102 char units = '\0';
7794db7c 1103 int mult = human_readable == 1 ? 1000 : 1024;
e175fb07
WD
1104 double dnum = 0;
1105 if (num > mult*mult*mult) {
1106 dnum = (double)num / (mult*mult*mult);
1107 units = 'G';
1108 } else if (num > mult*mult) {
1109 dnum = (double)num / (mult*mult);
1110 units = 'M';
1111 } else if (num > mult) {
1112 dnum = (double)num / mult;
1113 units = 'K';
1114 }
1115 if (units) {
83078af5 1116 snprintf(bufs[n], sizeof bufs[0], "%.2f%c", dnum, units);
e175fb07
WD
1117 return bufs[n];
1118 }
1119 }
1120
1121 s = bufs[n] + sizeof bufs[0] - 1;
1122 *s = '\0';
1123
1124 if (!num)
1125 *--s = '0';
1126 while (num) {
26404276 1127 *--s = (char)(num % 10) + '0';
e175fb07
WD
1128 num /= 10;
1129 }
1130 return s;
1131}
1132
1133/* Return the double number as a string. If the --human-readable option was
1134 * specified, we may output the number in K, M, or G units. We use a buffer
1135 * from human_num() to return our result. */
1136char *human_dnum(double dnum, int decimal_digits)
1137{
1138 char *buf = human_num(dnum);
1139 int len = strlen(buf);
2dc7b8bd 1140 if (isDigit(buf + len - 1)) {
e175fb07
WD
1141 /* There's extra room in buf prior to the start of the num. */
1142 buf -= decimal_digits + 1;
1143 snprintf(buf, len + decimal_digits + 2, "%.*f", decimal_digits, dnum);
1144 }
1145 return buf;
1146}
f7632fc6 1147
4a19c3b2 1148/* Return the date and time as a string. Some callers tweak returned buf. */
f7632fc6
AT
1149char *timestring(time_t t)
1150{
1151 static char TimeBuf[200];
1152 struct tm *tm = localtime(&t);
afa73c75 1153 char *p;
f7632fc6 1154
4f5b0756 1155#ifdef HAVE_STRFTIME
5cb37436 1156 strftime(TimeBuf, sizeof TimeBuf - 1, "%Y/%m/%d %H:%M:%S", tm);
f7632fc6 1157#else
5cb37436 1158 strlcpy(TimeBuf, asctime(tm), sizeof TimeBuf);
f7632fc6
AT
1159#endif
1160
afa73c75
WD
1161 if ((p = strchr(TimeBuf, '\n')) != NULL)
1162 *p = '\0';
f7632fc6 1163
afa73c75 1164 return TimeBuf;
f7632fc6
AT
1165}
1166
e1bd49d6
MP
1167/**
1168 * Sleep for a specified number of milliseconds.
1169 *
1170 * Always returns TRUE. (In the future it might return FALSE if
1171 * interrupted.)
1172 **/
1173int msleep(int t)
9ec16c83 1174{
c284f34a
WD
1175 int tdiff = 0;
1176 struct timeval tval, t1, t2;
9ec16c83
AT
1177
1178 gettimeofday(&t1, NULL);
5cb37436 1179
9ec16c83
AT
1180 while (tdiff < t) {
1181 tval.tv_sec = (t-tdiff)/1000;
1182 tval.tv_usec = 1000*((t-tdiff)%1000);
5cb37436 1183
9ec16c83
AT
1184 errno = 0;
1185 select(0,NULL,NULL, NULL, &tval);
1186
1187 gettimeofday(&t2, NULL);
5cb37436 1188 tdiff = (t2.tv_sec - t1.tv_sec)*1000 +
9ec16c83
AT
1189 (t2.tv_usec - t1.tv_usec)/1000;
1190 }
e1bd49d6
MP
1191
1192 return True;
9ec16c83
AT
1193}
1194
c2b54076
WD
1195/* Determine if two time_t values are equivalent (either exact, or in
1196 * the modification timestamp window established by --modify-window).
ac13ad10
MP
1197 *
1198 * @retval 0 if the times should be treated as the same
1199 *
1200 * @retval +1 if the first is later
1201 *
1202 * @retval -1 if the 2nd is later
1203 **/
c2b54076 1204int cmp_time(time_t file1, time_t file2)
5b56cc19 1205{
5b56cc19 1206 if (file2 > file1) {
bc6ebcd2
WD
1207 if (file2 - file1 <= modify_window)
1208 return 0;
5b56cc19
AT
1209 return -1;
1210 }
bc6ebcd2
WD
1211 if (file1 - file2 <= modify_window)
1212 return 0;
5b56cc19
AT
1213 return 1;
1214}
1215
1216
1217#ifdef __INSURE__XX
0f8f98c8
AT
1218#include <dlfcn.h>
1219
ac13ad10
MP
1220/**
1221 This routine is a trick to immediately catch errors when debugging
1222 with insure. A xterm with a gdb is popped up when insure catches
1223 a error. It is Linux specific.
1224**/
0f8f98c8
AT
1225int _Insure_trap_error(int a1, int a2, int a3, int a4, int a5, int a6)
1226{
1227 static int (*fn)();
1228 int ret;
8950ac03 1229 char *cmd;
0f8f98c8 1230
5cb37436 1231 asprintf(&cmd, "/usr/X11R6/bin/xterm -display :0 -T Panic -n Panic -e /bin/sh -c 'cat /tmp/ierrs.*.%d ; gdb /proc/%d/exe %d'",
0f8f98c8
AT
1232 getpid(), getpid(), getpid());
1233
1234 if (!fn) {
1235 static void *h;
1236 h = dlopen("/usr/local/parasoft/insure++lite/lib.linux2/libinsure.so", RTLD_LAZY);
1237 fn = dlsym(h, "_Insure_trap_error");
1238 }
1239
1240 ret = fn(a1, a2, a3, a4, a5, a6);
1241
1242 system(cmd);
1243
8950ac03
AT
1244 free(cmd);
1245
0f8f98c8
AT
1246 return ret;
1247}
1248#endif
58cadc86 1249
58cadc86
WD
1250#define MALLOC_MAX 0x40000000
1251
a2dc4d68 1252void *_new_array(unsigned long num, unsigned int size, int use_calloc)
58cadc86
WD
1253{
1254 if (num >= MALLOC_MAX/size)
1255 return NULL;
a2dc4d68 1256 return use_calloc ? calloc(num, size) : malloc(num * size);
58cadc86
WD
1257}
1258
1259void *_realloc_array(void *ptr, unsigned int size, unsigned long num)
1260{
1261 if (num >= MALLOC_MAX/size)
1262 return NULL;
58cadc86
WD
1263 if (!ptr)
1264 return malloc(size * num);
1265 return realloc(ptr, size * num);
1266}
e64ae6d7
WD
1267
1268/* Take a filename and filename length and return the most significant
1269 * filename suffix we can find. This ignores suffixes such as "~",
1270 * ".bak", ".orig", ".~1~", etc. */
1271const char *find_filename_suffix(const char *fn, int fn_len, int *len_ptr)
1272{
1273 const char *suf, *s;
1274 BOOL had_tilde;
1275 int s_len;
1276
1277 /* One or more dots at the start aren't a suffix. */
1278 while (fn_len && *fn == '.') fn++, fn_len--;
1279
1280 /* Ignore the ~ in a "foo~" filename. */
1281 if (fn_len > 1 && fn[fn_len-1] == '~')
1282 fn_len--, had_tilde = True;
1283 else
1284 had_tilde = False;
1285
1286 /* Assume we don't find an suffix. */
1287 suf = "";
1288 *len_ptr = 0;
1289
1290 /* Find the last significant suffix. */
1291 for (s = fn + fn_len; fn_len > 1; ) {
1292 while (*--s != '.' && s != fn) {}
1293 if (s == fn)
1294 break;
1295 s_len = fn_len - (s - fn);
1296 fn_len = s - fn;
6012eaa1 1297 if (s_len == 4) {
e64ae6d7
WD
1298 if (strcmp(s+1, "bak") == 0
1299 || strcmp(s+1, "old") == 0)
1300 continue;
6012eaa1 1301 } else if (s_len == 5) {
e64ae6d7
WD
1302 if (strcmp(s+1, "orig") == 0)
1303 continue;
1304 } else if (s_len > 2 && had_tilde
2dc7b8bd 1305 && s[1] == '~' && isDigit(s + 2))
e64ae6d7
WD
1306 continue;
1307 *len_ptr = s_len;
1308 suf = s;
1309 if (s_len == 1)
1310 break;
1311 /* Determine if the suffix is all digits. */
1312 for (s++, s_len--; s_len > 0; s++, s_len--) {
2dc7b8bd 1313 if (!isDigit(s))
e64ae6d7
WD
1314 return suf;
1315 }
1316 /* An all-digit suffix may not be that signficant. */
1317 s = suf;
1318 }
1319
1320 return suf;
1321}
1322
1323/* This is an implementation of the Levenshtein distance algorithm. It
1324 * was implemented to avoid needing a two-dimensional matrix (to save
1325 * memory). It was also tweaked to try to factor in the ASCII distance
1326 * between changed characters as a minor distance quantity. The normal
1327 * Levenshtein units of distance (each signifying a single change between
1328 * the two strings) are defined as a "UNIT". */
1329
1330#define UNIT (1 << 16)
1331
1332uint32 fuzzy_distance(const char *s1, int len1, const char *s2, int len2)
1333{
1334 uint32 a[MAXPATHLEN], diag, above, left, diag_inc, above_inc, left_inc;
1335 int32 cost;
1336 int i1, i2;
1337
1338 if (!len1 || !len2) {
1339 if (!len1) {
1340 s1 = s2;
1341 len1 = len2;
1342 }
1343 for (i1 = 0, cost = 0; i1 < len1; i1++)
1344 cost += s1[i1];
1345 return (int32)len1 * UNIT + cost;
1346 }
1347
1348 for (i2 = 0; i2 < len2; i2++)
1349 a[i2] = (i2+1) * UNIT;
1350
1351 for (i1 = 0; i1 < len1; i1++) {
1352 diag = i1 * UNIT;
1353 above = (i1+1) * UNIT;
1354 for (i2 = 0; i2 < len2; i2++) {
1355 left = a[i2];
1356 if ((cost = *((uchar*)s1+i1) - *((uchar*)s2+i2)) != 0) {
1357 if (cost < 0)
1358 cost = UNIT - cost;
1359 else
1360 cost = UNIT + cost;
1361 }
1362 diag_inc = diag + cost;
1363 left_inc = left + UNIT + *((uchar*)s1+i1);
1364 above_inc = above + UNIT + *((uchar*)s2+i2);
1365 a[i2] = above = left < above
1366 ? (left_inc < diag_inc ? left_inc : diag_inc)
1367 : (above_inc < diag_inc ? above_inc : diag_inc);
1368 diag = left;
1369 }
1370 }
1371
1372 return a[len2-1];
1373}
c2b54076
WD
1374
1375#define BB_SLOT_SIZE (16*1024) /* Desired size in bytes */
1376#define BB_PER_SLOT_BITS (BB_SLOT_SIZE * 8) /* Number of bits per slot */
1377#define BB_PER_SLOT_INTS (BB_SLOT_SIZE / 4) /* Number of int32s per slot */
1378
1379struct bitbag {
1380 uint32 **bits;
1381 int slot_cnt;
1382};
1383
1384struct bitbag *bitbag_create(int max_ndx)
1385{
1386 struct bitbag *bb = new(struct bitbag);
1387 bb->slot_cnt = (max_ndx + BB_PER_SLOT_BITS - 1) / BB_PER_SLOT_BITS;
1388
1389 if (!(bb->bits = (uint32**)calloc(bb->slot_cnt, sizeof (uint32*))))
1390 out_of_memory("bitbag_create");
1391
1392 return bb;
1393}
1394
1395void bitbag_set_bit(struct bitbag *bb, int ndx)
1396{
1397 int slot = ndx / BB_PER_SLOT_BITS;
1398 ndx %= BB_PER_SLOT_BITS;
1399
1400 if (!bb->bits[slot]) {
1401 if (!(bb->bits[slot] = (uint32*)calloc(BB_PER_SLOT_INTS, 4)))
1402 out_of_memory("bitbag_set_bit");
1403 }
1404
1405 bb->bits[slot][ndx/32] |= 1u << (ndx % 32);
1406}
1407
1408#if 0 /* not needed yet */
1409void bitbag_clear_bit(struct bitbag *bb, int ndx)
1410{
1411 int slot = ndx / BB_PER_SLOT_BITS;
1412 ndx %= BB_PER_SLOT_BITS;
1413
1414 if (!bb->bits[slot])
1415 return;
1416
1417 bb->bits[slot][ndx/32] &= ~(1u << (ndx % 32));
1418}
1419
1420int bitbag_check_bit(struct bitbag *bb, int ndx)
1421{
1422 int slot = ndx / BB_PER_SLOT_BITS;
1423 ndx %= BB_PER_SLOT_BITS;
1424
1425 if (!bb->bits[slot])
1426 return 0;
1427
1428 return bb->bits[slot][ndx/32] & (1u << (ndx % 32)) ? 1 : 0;
1429}
1430#endif
1431
1432/* Call this with -1 to start checking from 0. Returns -1 at the end. */
1433int bitbag_next_bit(struct bitbag *bb, int after)
1434{
1435 uint32 bits, mask;
1436 int i, ndx = after + 1;
1437 int slot = ndx / BB_PER_SLOT_BITS;
1438 ndx %= BB_PER_SLOT_BITS;
1439
1440 mask = (1u << (ndx % 32)) - 1;
1441 for (i = ndx / 32; slot < bb->slot_cnt; slot++, i = mask = 0) {
1442 if (!bb->bits[slot])
1443 continue;
1444 for ( ; i < BB_PER_SLOT_INTS; i++, mask = 0) {
1445 if (!(bits = bb->bits[slot][i] & ~mask))
1446 continue;
1447 /* The xor magic figures out the lowest enabled bit in
1448 * bits, and the switch quickly computes log2(bit). */
1449 switch (bits ^ (bits & (bits-1))) {
1450#define LOG2(n) case 1u << n: return slot*BB_PER_SLOT_BITS + i*32 + n
1451 LOG2(0); LOG2(1); LOG2(2); LOG2(3);
1452 LOG2(4); LOG2(5); LOG2(6); LOG2(7);
1453 LOG2(8); LOG2(9); LOG2(10); LOG2(11);
1454 LOG2(12); LOG2(13); LOG2(14); LOG2(15);
1455 LOG2(16); LOG2(17); LOG2(18); LOG2(19);
1456 LOG2(20); LOG2(21); LOG2(22); LOG2(23);
1457 LOG2(24); LOG2(25); LOG2(26); LOG2(27);
1458 LOG2(28); LOG2(29); LOG2(30); LOG2(31);
1459 }
1460 return -1; /* impossible... */
1461 }
1462 }
1463
1464 return -1;
1465}
1c3344a1
WD
1466
1467void *expand_item_list(item_list *lp, size_t item_size,
1468 const char *desc, int incr)
1469{
1470 /* First time through, 0 <= 0, so list is expanded. */
1471 if (lp->malloced <= lp->count) {
1472 void *new_ptr;
1473 size_t new_size = lp->malloced;
1474 if (incr < 0)
0b515981 1475 new_size += -incr; /* increase slowly */
1c3344a1
WD
1476 else if (new_size < (size_t)incr)
1477 new_size += incr;
1478 else
1479 new_size *= 2;
1480 new_ptr = realloc_array(lp->items, char, new_size * item_size);
1481 if (verbose >= 4) {
1482 rprintf(FINFO, "[%s] expand %s to %.0f bytes, did%s move\n",
1483 who_am_i(), desc, (double)new_size * item_size,
1484 new_ptr == lp->items ? " not" : "");
1485 }
1486 if (!new_ptr)
1487 out_of_memory("expand_item_list");
1488
1489 lp->items = new_ptr;
1490 lp->malloced = new_size;
1491 }
1492 return (char*)lp->items + (lp->count++ * item_size);
1493}