That was probably the problem. I've seen instances on a couple of projects where a file has uploaded completely and properly, but BOINC has only partly noticed: it notices enough to delete the file in question, but doesn't update the housekeeping records to say that the transfer has finished.

Then, it keeps trying to upload the file, but fails because the file isn't there to upload. In some of those test versions of BOINC, the resulting error message suggests that it's the upload handler which is missing, rather than the local file. The newer test versions resolve the vicious circle by giving up on the transfer if there's no file still there waiting to be uploaded, as your new messages show.

I would urge anybody running test versions of BOINC to keep an eye on BOINC 7 Change Log and news, and/or subscribe to the boinc_alpha mailing list, so that they keep up to date with which bugs have been reported, and which have been fixed.

Thanks Richard - seems like I have spent a lot of time trying to track down the cause of a failed upload that hadn't failed. Still I have learned a bit in the process. I have bookmarked the link for the changelog for BOINC7 and will look at it from time to time.