Today’s obnoxious pain in the ass: BOINC and Time Machine.

The PS1 Optical Galaxy Survey recently moved from their temporary EC2 hostname to their current hostname, and the recommended course of action was, of course, to remove and re-add the project. So I decided to cheat and kludge my way around it.

That didn’t work, so naturally I wound up putting back the old files from Time Machine. And naturally, I forgot that Time Machine adds both flags and extended attributes to files and directories which have the net result of making the files and/or directories unreadable.

No, really: the com.apple.metadata:_kTimeMachineNewestSnapshot and com.apple.metadata:_kTimeMachineOldestSnapshot flags get in the way somehow. I’m not even sure how. I suspect that there’s some magic hooked from Time Machine into the filesystem code itself (!); if this were FreeBSD, I’d say it was intercepting vfsops.

Anyhow, removing those flags fixed up the POGS project, and got rid of quite possibly the weirdest BOINC error I’ve ever seen:

Tue  6 Aug 06:57:20 2013 |  | [error] Can't create HTTP response output file projects/pogs.theskynet.org_pogs/wrapper_x86_64-apple-darwin_340.gzt
Tue  6 Aug 06:57:20 2013 | pogs | Backing off 5 min 17 sec on download of wrapper_x86_64-apple-darwin_340

But other projects were still broken: because I’d changed my email address from ancient ones (that I’m not even sure exist any more), the BOINC weak authenticators had changed, so things broke, and I was nearly contemplating mailing the SETI@Home team to ask if I could manually submit workunits somehow.

So I manually wedged the new authenticator key into the accounts config for the SETI@Home project, and lo! everything worked.