[Openais] first cut at defect 188 for the B apis

Mark Haverkamp markh at osdl.org
Fri Feb 18 08:37:11 PST 2005


On Wed, 2005-02-16 at 15:32 -0700, Steven Dake wrote:
> Ok so you recommend I delete evt.c and evt.h from the patch I sent you,
> then pull latest bk, then apply my patch and your patch to get
> consolidated patch?
> 
> Thanks
> -steve
> 
> On Wed, 2005-02-16 at 15:25, Mark Haverkamp wrote:
> > On Wed, 2005-02-16 at 13:40 -0800, Mark Haverkamp wrote:
> > > On Wed, 2005-02-16 at 13:05 -0800, Mark Haverkamp wrote:
> > > > On Wed, 2005-02-16 at 11:29 -0700, Steven Dake wrote:
> > > > > Mark,
> > > > > I understand...  I can work the evt against latest bk if you have such a
> > > > > patch.
> > > > > 
> > > > 
> > > > I have the patch integrated and sort of working.  It passes my testevt
> > > > program.  But, when I started running my publish/subscribe stress test
> > > > and returned from a meeting an hour later, two of the execs had
> > > > asserted:
> > > 
> > > I have an idea on this.  I think that I must not have the patch applied
> > > quite right and a subscribed channel is leaking memory when receiving
> > > events.  The systems ran out of memory after the applications consumed
> > > it all.  I'll track it down.
> > 
> > I found the leak.  I have been running for some time now so here are the
> > event service related files.  They apply to the latest checked in files.
> > Take a look.  I'll let my tests run overnight and see how they do.

My tests have been running OK for a couple nights now.  I have had a
couple token timeout/config changes but things settled out OK without
any crashes/hangs/asserts/segfaults, etc.

Mark.

-- 
Mark Haverkamp <markh at osdl.org>




More information about the Openais mailing list