On Sun 11 Mar 2007 at 10:01PM, Timo Sirainen wrote:
On Sun, 2007-03-11 at 21:45 +0200, Timo Sirainen wrote:
So the only problem is if the last uidlist rename() fails. But that I hope won't fail because it's done inside the same directory, and the maildir root directory is pretty static.
/mnt/Maildir/tmp% mv 82 83 mv: cannot move
82' to83': No space left on device/mnt/Maildir/tmp% mv 82 81 /mnt/Maildir/tmp%
I think in general you can't depend upon this-- take for example any filesystem supporting snapshots (like say ZFS). ZFS counts the size of your metadata against the quota (this is a good thing). The size occupied by your snapshot is also counted against the quota-- which means that if you have /home/dp@yesterday and you are near the quota, a rename of something which is in the snapshot will of course need to allocate new data and/or metadata blocks, since the old state-of-the-world must be saved in the snapshot.
-dp-- Daniel Price - Solaris Kernel Engineering - dp@eng.sun.com - blogs.sun.com/dp