Author Rhamphoryncus
Recipients Rhamphoryncus, amaury.forgeotdarc, gregory.p.smith, gvanrossum, kbk, larsimmisch
Date 2008-05-07.18:11:43
SpamBayes Score 0.00941091
Marked as misclassified No
Message-id <1210183915.75.0.0624230391664.issue1856@psf.upfronthosting.co.za>
In-reply-to
Content
Cleaned up version of Amaury's patch.  I stop releasing the GIL after
sys.exitfunc is called, which protects threads from the ensuing teardown.

I also grab the import lock (and never release it).  This should prevent
the nasty issue with daemon threads doing imports mentioned in issue
1720705, so that part of the documentation change can be removed.  I
didn't do any significant testing though.

Importing raises a potential issue with this approach.  The standard
meme of "release GIL; grab lock; acquire GIL" becomes invalid.  A child
may grab the lock, then block on the GIL, while the main thread (which
never releases the GIL) will block on acquiring the lock.  I worked
around it with the import lock due to it's recursive behaviour, but if
it exists anywhere else in CPython it may cause a problem if used during
shutdown.
History
Date User Action Args
2008-05-07 18:11:56Rhamphoryncussetspambayes_score: 0.00941091 -> 0.00941091
recipients: + Rhamphoryncus, gvanrossum, gregory.p.smith, kbk, amaury.forgeotdarc, larsimmisch
2008-05-07 18:11:55Rhamphoryncussetspambayes_score: 0.00941091 -> 0.00941091
messageid: <1210183915.75.0.0624230391664.issue1856@psf.upfronthosting.co.za>
2008-05-07 18:11:54Rhamphoryncuslinkissue1856 messages
2008-05-07 18:11:52Rhamphoryncuscreate