Thanks for filling in those details, Corion. Now I can just skip to the crazy ideas:
my $rand= md5_hash_hex( rand() . $$ . $x . "super secret" );
mkdir( "$x/$rand" );
if( -d "$y/$rand" ) {
warn "$x == $y\n";
}
rmdir( "$x/$rand" );
I only created a subdirectory instead of a file because it makes for more concise Perl code. Or you could lock a file you find in $x and see if you hold the lock on that file in $y? (But that assumes that all of your file system exporting technologies convey lock information, which is not always the case, of course.) You could create a pipe in one... etc.
| [reply]
[d/l] |
Or you could lock a file you find in $x and see if you hold the lock on that file in $y
Or, then, I could create a tempfile (File::Temp) in one, and see whether the tempfile exists in the other!!!
--
Ronald Fischer <ynnor@mm.st>
| [reply] |
rovf++ - this may well be the most portable idea of all. There is a very small chance of failure in weird setups like AFS, in which synchronization between the exported and the local dir may happen days apart, but that's a real stretch.
--
Human history becomes more and more a race between education and catastrophe. -- HG Wells
| [reply] |
You should not forget to check, if the file creation was successful. If you don't have permission to create the file and you don't check that, your check for existence might lead to a wrong conclusion!
| [reply] |
You can check the device and inode field of the results of stat for both directories, but that will only work for file systems that follow the idea the inventors originally had.
I guess this means "the usual file systems on Unix and Linux". Do you happen to know to what extent this is fulfilled for networked file systems on Windows?
You can readdir the directories and consider them to be the same if they contain the same files.
In my case, this would mean I would have to actually compare the content of the files, because even the sizes could be the same :-(
Thanks a lot for clarification.
Update: I just made a test on Windows, and the inode number indeed comes up as zero.
--
Ronald Fischer <ynnor@mm.st>
| [reply] |
| [reply] |
Testing for identical content (whether using rsync or something else) will not give you a 100% sure answer. Even if the content is the same, the directories do not have to, for instance:
$ cp -a dir1 dir2
will create a second directory with identical content, but dir1 and dir2 are different. You'd have to look at inode change time to see a difference, but even then, it's not a garantee.
OTOH, even if you think the content is different, the directories may be the same. If the directories are the same, you'll be looking at the same directory twice, but since you cannot garantee you'll be looking twice at the same time, the directory, or any of its files inside it, may have changed between your two inspections.
Having said that, I don't think there's solution that will always work anywhere. But usually you will know a bit of the environment. If you know your filesystems are local, and no device is mounted twice, checking inode and device number will do. If you know the directories to be tested aren't being modified while you make your check, checking for content may work. | [reply]
[d/l] |
| [reply] |