Haven't run these for a while on this box but just completed one the other day and I notice they are still not cleaning up after themselves, at least on Linux.

Don't know if they are better behaved on other platforms.

My experience on Win7 64-bit is that they clean up if they are successful, but don't clean up if they fail.

Thanks Jim, Not sure if it was only on Linux that they weren't cleaning up after completing successfully. I also wondered briefly if it was just older ones that weren't cleaning up but on checking the one that didn't is one of the new ones. I know that weren't doing it before but thought it might be resolved by now. I think I have just had other types of tasks for a while so had got out of the habit of checking the projects folder regularly.

There is a new alpha version of BOINC (7.6.1) that cleans up slots better. Whether that fixes anything on CPDN I don't know, since I am not doing the HadCM3 shorts now. And it is Windows-only at the moment.
http://boinc.berkeley.edu/dev/forum_thread.php?id=10204&postid=62325#62325

But it is working fine for me on CPDN otherwise.

I don't think that new version of BOINC will make any difference here. The 'cleaning up slots' problem relates to a failure to delete individual files larger than 4 GB: so far as we know, only one BOINC project - the CMS-dev VirtualBox machine images created by CERN - has ever created files that large in the slot directories.

It's Windows only so far, because it was only a 32-bit Windows API call which was causing the deletions to fail. Other OSs use different file deletion techniques, which weren't subject to the 4 GB limit.

OK, thanks for the clarification.