[omniORB] no timeout when connect hangs (2.8.0)

Peter Bauer PBHD@compuserve.com
Wed, 20 Oct 1999 16:20:37 -0400


Hello,

in some real-life-environment it happens sometimes =

(now: once), that the connect in realConnect hangs forever
(at least 10 Minutes), but the timeout for remote calls
is set to a much lower value (30 seconds), so one would
expect an exception to be thrown within a minute or so
after the call was initiated, but this does not happen. =

I think this comes from the fact, that the scavenger =

does not see that socket (it looks for pd_socket, which =

is assigned after return from realConnect in ll_recv or =

ll_send). My question is: would it make sense to =

assign the newly created socket immediately =

after the socket-call inside realConnect to the =

pd_socket-member, so that the scavenger-initiated =

code can see it and can do a shutdown on this socket in
case of timeout?
I have found some discussion related to long connect-
timeouts in the 1998-archives, but I think that in those =

days there was no timeout-facility built into omniORB,
which could nowadays be used for the abort of such
lengthly connects.

(I'll try that change anyway, but this condition took =

more than 3 days to occur in a environment we can't =

reproduce)

Gruss Peter Bauer