[omniORB] COMM_FAILURE for "long" sequences

[Clarke Brunt]

Duncan Grisby wrote:

> If you're definitely not releasing the memory prematurely,
> try editing the send function in 
> src/lib/omniORB/orbcore/tcp/tcpConnection.cc. Right at the 
> start of the function, there's an #ifdef for VMS that limits 
> the send size to 64K. Try enabling that.

I haven't got anything definite to contribute, but a while ago, I posted to
the list that I'd been getting unpredictable failures with large transfers
(ours were 10-20MB, so I'd definitely set giopMaxMsgSize). We were running
on WinNT 4.0. I experimented with limiting the send size as Duncan says, but
in the end, the problem was only observed to happen on my own desktop
machine, and didn't afflict the servers on which we _really_ wanted to run
it, so I reverted to an un-patched omniORB. I think that there's equivalent
code for limiting the 'receive' size as well, if that end of the connection
ever gets the problem.

If relevant, the thread on this mailing list was titled "Problem with large
requests (WinNT)".

-- 
Clarke Brunt, Principal Software Engineer, Trafficmaster