On Wed, 2003-03-19 at 18:14, Thomas Wouters wrote:

> > Do we understand why you were getting these exceptions?  It feels a
> > little like putting a bandaid on an oozing (if not hemorrhaging :)
> > wound.
> 
> No, we don't[*]
> 
> [*]: I think Peer is having race-condition problems. He sent me some broken
> pickles that look a lot like write-races, and Pending.py is writing to
> 'pending.pck.tmp' -- always, and only checks locks once. I strongly, strongly
> suspect some lock-breaking is involved, and this results in both the
> NotLockedErrors and the pending.pck breakage. I'm rewriting Pending.py to be
> more robust.

I like your changes -- thanks!  I suppose we should bite the bullet and
rip out the global pending.pck file.  But not for MM2.1, unless there's
no other way!

Could it really take such a long time to write a pending file?  I wonder
if there isn't some clock skew or instability happening?  Your changes
help make Mailman more defensive so that's +1, but WIBNI we understood
what was going on? :)

-Barry



_______________________________________________
Mailman-checkins mailing list
[EMAIL PROTECTED]
http://mail.python.org/mailman/listinfo/mailman-checkins

Reply via email to