On Tue, Oct 11, 2005 at 11:05:04PM -0400, Charles P. Wright wrote:
> On Sat, 2005-10-08 at 22:41 -0400, Shaya Potter wrote:
> > I don't understand this behavior.
> > 
> > it 
> > 
> > 1) renames the file to the whiteout
> This was done so we would have better atomicity.
> > 2) truncates it.  
> This is so we didn't fill up the disk unnecessarily.

Makes sense to me.

> > This truncation is obviously wrong (i.e. kills any open files).

Hmm, but that's right, too.

> > I would think the better thing to do is 
> > 
> > 1) create whiteout file
> > 2) unlink old file.
> Yes.  Or we do magic with open files to check if we can't truncate. 

True.

> The third option is actually that we remove delete=whiteout.  I didn't
> know anyone that uses it, except you

Uhm, well, actually, I was using this, too, and was not aware of any
problems yet.

I just did not want a file to get deleted on a non-toplevel branch when
doing snapshots with multiple rw branches. WEll, maybe my interpretation
of what the delete=all default does is wrong, but I thought that
delete=whiteout would be the safest opion, which only operates on the
toplevel rw branch. Right?

Regards
-Klaus Knopper
_______________________________________________
unionfs mailing list
[email protected]
http://www.fsl.cs.sunysb.edu/mailman/listinfo/unionfs

Reply via email to