[Openais] Re: latest openais tree problems

[Steven Dake]

On Thu, 2006-01-05 at 12:54 -0800, Mark Haverkamp wrote:
> On Thu, 2006-01-05 at 13:48 -0700, Steven Dake wrote:
> > sounds good go ahead and apply it but please make a bug number if you
> > can - i'd still like to use bug numbers for picacho so we can track what
> > changed between releases.
> > 
> 
> OK, I'll get a bug number.
> 
> > I am curious about your ideas as to whether this would effect other
> > services as well - Ie should we rework the dispatch code in other
> > services?
> > 
> 
> The event dispatch is a little different than most.  Since I do my own
> throttling, message delivery is a two step process.  The exec pings the
> library with a small message saying that there is data available, then
> the library requests that data.  When the library received the "data
> available" message and its request for the event gets the "try again"
> error code, that is where the problem happened.  I don't think that the
> other services operate that way.
> 

i see thanks we should be good to go then
regards

> 
> > thanks
> > -steve
> > 
> > On Thu, 2006-01-05 at 11:34 -0800, Mark Haverkamp wrote:
> > > On Thu, 2006-01-05 at 12:22 -0700, Steven Dake wrote:
> > > > Mark
> > > > Should this go into picacho too?  I'm wondering if it is possible that a
> > > > try-again could occur during this dispatch routine locking up the evt
> > > > service like we see with ykd.
> > > 
> > > Good question.  I'm not sure what circumstances would cause the try
> > > again.  It probably wouldn't hurt to put it in though. If you like I can
> > > put it in both places.
> > > 
> > > Mark.
> > > 
> > > > 
> > > > regards
> > > > -steve
> > > > 
> > > > 
> > > > On Thu, 2006-01-05 at 09:43 -0800, Mark Haverkamp wrote:
> > > > > On Wed, 2006-01-04 at 16:06 -0700, Steven Dake wrote:
> > > > > [...]
> > > > > > 
> > > > > > this would show the above scenario that I outline.  Let me work out a
> > > > > > patch tonight for you to try for the first issue.  Could you investigate
> > > > > > the subscription tests for the second issue?  
> > > > > 
> > > > > OK, Here is a patch for the dispatch "try again" problem.
> > > > > 
> > > > > Thanks,
> > > > > Mark.
> > > > >