I may shed light onto some of your troubles:

Code:

## BTW.. Following command will work even without -print option

$ find /tmp/test3 -print

Yes, because "-print" is the default clause if no other one is given. When you write it out you just manually recreate the default. I suppose it is in the man page, it is definitely on the POSIX description:

Code:

$ find /tmp/test3 -exec rm -fr {} \; 
find: `/tmp/test3': No such file or directory

Yes, look at the output of your command when you used the "-print" instead of "-exec": the first directory to be listed was - "/tmp/test3", yes? Now, what string do you suppose has gone first into "rm -rf", hmmm? What will a rm -rf /tmp/test3 do to a file /tmp/test3/file or to a directory /tmp/test3/directory or to a file /tmp/test3/directory/file in it?

What you see is "find" being a bit dumbfounded by the fact that the directory structure it was just starting to read is already gone. Your prompt shows "/tmp/test3" being the current directory, but this is just because the shell hasn't noticed now that the directory is already gone. Do a "cd *" and you will probably find that "/tmp/test doesn't exist any more.

This is absolutely true and a valuable advice you should take to heart! It might not be necessary because of the reasons above - before find could even try to follow a link it will already be long gone - but beware once you got your command halfways correct and it starts actually doing what you mean (which might not be what you want).

Fortunately there is a remedy:

Code:

find /tmp/test3/* -prune ! -type l -exec rm -rf {} \;

Because you use rm -r you do not need to descend the directory structure at all, the first layer will suffice. All links are filtered out before find gets to the dangerous stuff.

I hope this helps.

bakunin