Redirecting here from mailman-checkins...

On Fri, 2003-03-21 at 03:30, Thomas Wouters wrote:

On Thu, Mar 20, 2003 at 11:02:29PM -0500, Barry Warsaw wrote:

...

On Wed, 2003-03-19 at 19:31, twouters@users.sourceforge.net wrote:

...

...

Fix another of Peer's annoying bugs: somehow his next_request_id ended up lower than the highest entry in the list config.db, so the assertion was being triggered. We can just deal with the problem instead.

...

Very strange! Does this (and Peer's other problems) make us question the integrity of the LockFile algorithm?

Yes, I'm going to setup a LockFile stresstest when I get a moment (which will probably not be in the next three weeks, unfortunately.)

Note that at the bottom of LockFile.py there's an attempt at a stress test. It would be good to code review this to make sure it's testing what we think it should test and what it ought to test <wink>. It's also never been really run in an NFS environment. I don't know if Peer's using NFS, but if so that would be the first place I'd suspect.

-Barry