localhost udp: address in use

[mdberger]

I have some complex multi-thread code in which I do
some internal signaling with udp sockets:
snd: 127.0.0.1:6501
rcv: 127.0.0.1:6502

On my new F15_64 box, I usually get:
errno=98=Address already in use
in response to the bind command when I bring up the snd
socket (which is first up) even when I just come up from
power off.

Rarely, it works. I think that it is after I have been
doing nothing, but with the box running, for a while.
I'm not sure.

Now on my old CentOS 5_32 box, the same code
(but, of course, compiled for 32) works with no
problem. I can start and stop it rapidly with no
problem.

(BTW, this does not account for the problem I
mentioned in my thread on starting SysV daemons
on boot.)

Thanks for your suggestions.
Mike.

[sonoran]

Have you tried asking on the Users mailing list?

https://admin.fedoraproject.org/mailman/listinfo

I think you're more likely to get the attention of a technical expert there.

[mdberger]

Thanks, I just posted it there.

BTW, my opinion that when it works, it does so because
I left it alone for a while, is not correct. I can see no pattern.

Mike.

[mdberger]

This turned out to be a coding problem uncovered
by migrating to a much faster box.
Briefly, the initial call to the getInstance()
of a singleton was not complete before a
second, higher priority call was issued.

Sorry for the distraction.
Mike.