Code Review Request, JDK-8214418 HttpClient falls in running with 100% cpu usage after an error signalled on channel

[Daniel Fuchs]

Hi Xuelei,

This is not my area of expertise - so I'm going to rephrase
what I understand (which may be wrong):

SSLEngineImpl.java:
   This change makes sure that the SSLEngineResult::getStatus()
   returns Status.CLOSED when closure has been initiated, even
   if the engine is only half-closed at that time.
   This allows the application to complete the closure by looking
   at the state of the inbound/outbound.

TransportContext.java:
    This change makes sure that the appropriate
    HandshakeStatus is returned to allow the completion of the
    closure: reading/sending the closure acknowledgement.

If my understanding is correct - then I think this change is good.
As far as I can see this is exactly what we are expecting in
the java.net.http HttpClient SSLFlowDelegate, so that looks good
to me!

As Chris mentioned, it would be good to have a deterministic test
to verify the behavior.

best regards,

-- daniel

On 22/12/2018 17:20, Xue-Lei Fan wrote:
> Hi,
> 
> Could I get the update reviewed?
>     http://cr.openjdk.java.net/~xuelei/8214418/webrev.00/
> 
> The reproducing testing case passed with the update.
> 
> The issue is caused by the handshake status "NEED_WRAP" while the 
> connection is half-closed. An application may just call wrap() when the 
> handshake status is "NEED_WRAP". For compatibility, I changed the 
> handshake status from NEED_WRAP back to NOT_HANDSHAKING for inbound 
> half-closed connection.  An application can use 
> SSLEngine.isOutboundDone() for the determination if SSLEngine.wrap() 
> should be called.
> 
> Thanks,
> Xuelei
> 
>