-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/16411/#review30746
-----------------------------------------------------------

Ship it!


Ship It!

- Gordon Sim


On Dec. 20, 2013, 2:50 p.m., Alan Conway wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/16411/
> -----------------------------------------------------------
> 
> (Updated Dec. 20, 2013, 2:50 p.m.)
> 
> 
> Review request for qpid and Gordon Sim.
> 
> 
> Bugs: QPID-5431
>     https://issues.apache.org/jira/browse/QPID-5431
> 
> 
> Repository: qpid
> 
> 
> Description
> -------
> 
> Avoid double reset(), alternative to Gordon's suggestion.
> 
> I think this is a simpler fix, check the diff r1 r2. If checkDisconnected 
> returns true then we know that reset() has already been called so no need to 
> call it in check()' Do you see any flaws?
> 
> 
> QPID-5431: Qpid c++ client hangs / crashes during reception failover in HA 
> environment (mutual recursion)
> 
> Bug in AMQP 1.0 retry code caused an infinite recursion when failing over.
> The recursion was in messaging::amqp::ConnectionContext, where the following
> recursive cycle could occur:
>  
> check()->autoconnect()->tryConnect(Url)->tryConnect(Address)->wait()->check()->...
> 
> Re-organized the code to avoid the recursion, specifically avoid calling 
> check()
> in tryConnect(Address). A disconnect detected in tryConnect results in 
> continuing
> the retry rather than calling autoconnect again.
> 
> 
> Diffs
> -----
> 
>   /trunk/qpid/cpp/examples/messaging/drain.cpp 1552476 
>   /trunk/qpid/cpp/src/qpid/ha/BrokerReplicator.cpp 1552476 
>   /trunk/qpid/cpp/src/qpid/messaging/amqp/ConnectionContext.h 1552476 
>   /trunk/qpid/cpp/src/qpid/messaging/amqp/ConnectionContext.cpp 1552476 
> 
> Diff: https://reviews.apache.org/r/16411/diff/
> 
> 
> Testing
> -------
> 
> Bug reproducer, full make test.
> 
> 
> Thanks,
> 
> Alan Conway
> 
>

Reply via email to