c++ broker changes that might affect windows

[Alan Conway <ac...@redhat.com>]

I added a function AsynchIO::stopReading() in revision 743416 to fix a race 
condition with read-credit on Linux and a no-op implementation for windows.

This shouldn't cause any change to existing windows behaviour. I'm raising it in 
case anyone wants to investigate whether windows has the same race. At the 
moment only the cluster uses read credit so it's somewhat hypothetical on 
windows. If someone wants to investigate I can give details of the race and 
pointers on writing a cluster-free test case for it.

Apologies in advance if I broke the windows build, I'm almost certain I didn't.



---------------------------------------------------------------------
Apache Qpid - AMQP Messaging Implementation
Project:      http://qpid.apache.org
Use/Interact: mailto:dev-subscribe@qpid.apache.org

RE: c++ broker changes that might affect windows

[Steve Huston <sh...@riverace.com>]

Hi Alan,

> I added a function AsynchIO::stopReading() in revision 743416 
> to fix a race 
> condition with read-credit on Linux and a no-op 
> implementation for windows.

Ok, thanks for the heads up.

> This shouldn't cause any change to existing windows 
> behaviour. I'm raising it in 
> case anyone wants to investigate whether windows has the same 
> race. At the 
> moment only the cluster uses read credit so it's somewhat 
> hypothetical on 
> windows. If someone wants to investigate I can give details 
> of the race and 
> pointers on writing a cluster-free test case for it.

Ok - it would be nice to investigate this soon.

> Apologies in advance if I broke the windows build, I'm almost 
> certain I didn't.

You didn't.

-Steve


---------------------------------------------------------------------
Apache Qpid - AMQP Messaging Implementation
Project:      http://qpid.apache.org
Use/Interact: mailto:dev-subscribe@qpid.apache.org