[omniORB] Differences on RH and DEC

[Matthew Berry]

I am using omni3.0.3 with -WbBOA on RH 6.1 and DEC alpha OSF 1.4

I have managed to replicate the problem I am having using the example, eg2.
I modify the implementation by inserting abort() in the method echoString.

On Red Hat, a COMM FAILURE is caught when the eg2_server "crashes", but on
DEC the client retries indefinitely to call the server after it crashes. The
client is never notified on the server "crashing".

Here is a trace from the DEC eg2_client (-ORBtraceLevel
25 -ORBtraceInvocations)

omniORB: throw omniConnectionBroken (minor 36) from
tcpSocketMTfactory.cc:1076
omniORB: tcpSocketStrand::~Strand() close socket no. 3
omniORB: defaultTransientExceptionHandler: retry 0th times.
omniORB: Invoke 'echoString' on remote: root<0>
omniORB: throw omniConnectionBroken (minor 32) from
tcpSocketMTfactory.cc:1132
omniORB: tcpSocketStrand::~Strand() close socket no. 3
omniORB: throw COMM_FAILURE from remoteIdentity.cc:178
omniORB: Reverting object reference to original profile
omniORB: strand Rope::incrRefCount: old value = 1
omniORB: strand Rope::decrRefCount: old value = 2
omniORB: defaultTransientExceptionHandler: retry 1th times
omniORB: Invoke 'echoString' on remote: root<0>
omniORB: throw omniConnectionBroken (minor 32) from
tcpSocketMTfactory.cc:1132
omniORB: tcpSocketStrand::~Strand() close socket no. 3
omniORB: throw COMM_FAILURE from remoteIdentity.cc:178
omniORB: Reverting object reference to original profile
omniORB: strand Rope::incrRefCount: old value = 1
omniORB: strand Rope::decrRefCount: old value = 2
omniORB: defaultTransientExceptionHandler: retry 2th times