> The last gump run (which has come on-line about 30 mins ago)
>
>      Start Date/Time (UTC) Thu, 08 Jul 2004 07:00:49 (UTC)
>      End Date/Time (UTC) Thu, 08 Jul 2004 13:26:41 (UTC)
>
> What is happening between 13:26 and 17:15 (about 4 hours). There appears
> to be a really big delay between the end of a gump run and the
> publication of results.

There are some serious memory leaks w/ the CVS HEAD Gump code. Python is
non-trivial in keeping healthy (is my experience, or maybe I just trod on
land mines). Anyways, the simple documentation of results (generating pages)
has some awful memory leaks. I think the process bogs down to a crawl.

The 'CleanUp' branch started as adding lots of __del__ to try to break
circular dependencies and reduce these leaks. I think the situation is
better. A lot more went into this branch (moving to DOM from Pythonic
pseudo-(D)OM). Also, the update get performed right before the build, and
the e-mails get sent right after the build completes (given the largest
window to fix before next build). A bunch of stuff like that.

I asked yesterday about merging CleanUp into CVS HEAD in the main 'cos you
are working so diligently on this, and I am nervous I'll break you. After
tonight I'll be done w/ my EMT refresher & will be able to commit to fixing
what breaks. I am tempted to perform the merge, but turn off notification
(which might send duplicates for some odd reason) and see what transpires.

regards,

Adam


---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]

Reply via email to