[omniORB] Server hangs when client crashes

[Jan Lessner]

Hi
> When running two or more clients against one server, and one of the client
> crashes, the server blocks.
> 
 When one of the client stops, it seems like the server blocks in the
> return-statement while trying to send to the client that recently crashed.

We sometimes found the same behaviour on Solaris which seems to be
related to the operating system's socket handling. If an application
crashes without being able to cleanup its IP port allocations, it turned
out that sending data into its open socket connections still works for
quite a while without complaining. However this is true for sending only
in our case and therefore happend only with oneway calls until the
operating system cleaned up by itself (seems to be done about once every
minute or so). Attempting to read from the same socket immediately
produced an error.
So it's a littel different from your problems but I could imaging that
it's similar.

Regards,

	Jan Lessner, C-LAB