Looking through my 4 invalids on various machines there appears to be a commonality. All of them are WUs that have been restarted after a BOINC restart. The WUs that process without interruption are always marked valid. A bug in the client or validator?

Looking at some of the top computers shows that the above pattern holds. All the WUs marked invalid are ones that have been restarted at some point. All the valid WUs I've checked (a lot) are ones that have been run to completion without being interrupted in any way. Anyway if this is a known problem let me know and I'll not investigate further...

Hi,

Apologies for late reply.

A BOINC restart probably didn't allow enough time for the wrapper and client to save their work and shutdown cleanly but rather killed them abruptly. This left the current work in the checkpoint file, which is where they are supposed to be loaded from next start.. but for some reason current client is ignoring them so the packets get lost and never get crunched by you and returned. As the returned work unit doesn't have all the packets done, it gets thrown back by the validator. (And eventually you should get partial credit and that task is marked as invalid.)

So this is somewhat a known problem or actually two but haven't yet found fixes for them. First is the rapid shutdown (BOINC Client and wrapper play here) and second is the ignored checkpoint (Distributed.net Client is behaving oddly).

-w