[Vortex] Listener notification of channel creation?

[Francis Brosnan Blazquez]

Hi Skye,

> > See test_01c at test/vortex-regression-client.c (and how it is
> > implemented the profile used at the listener side:
> > test/vortex-regression-listener.c). 
> > 
> 
> That's great, just what I needed!  I'm just having one problem now,
> the
> initiator is hanging for around 30 seconds when I try to
> vortex_connection_close()... here's the debug output:
> 
> (proc 2851): (debug) vortex-channel: called vortex_cond_timedwait from
> vortex_channel.c:4284
> (proc 2851): (debug) vortex-thread: waiting from: 1224544082.745357000
> (proc 2851): (debug) vortex-thread: to (microseconds=500000):
> 1224544083.245357000
> (proc 2851): (warning) vortex-thread: timeout reached for conditional
> variable (system call pthread_cond_wait finished)
> (proc 2851): (debug) vortex-channel: finished vortex_cond_timedwait
> (result:
> 1) from vortex_channel.c:4284
> (proc 2851): (warning) vortex-channel: we still didn't sent replies
> for
> connection=1 channel=3, block until all replies are sent (RPY -1 !=
> MSG 0)
> 
> As specified in RFC 3195 for the RAW profile:
> 
> http://www.faqs.org/rfcs/rfc3195.html
> 
> The listener sends a "welcome" message, and the initiator sends a
> bunch of
> ANS, then a NUL, then closes.  Does that not count as a proper
> reply? ;)

There is a bug on the vortex_sequencer.c which causes to not properly
notify "reply written" when handling NUL replies. It is fixed but still
not committed to svn.

In the meantime the next release is published you can use
vortex_connection_shutdown before calling to vortex_connection_close to
"force" connection close without taking into consideration pending
replies to be sent.

> I'll probably finish this up tomorrow and release it under the GPL if
> you're
> interested.

Sure. I'm sure other users might be interested too. Let me know the
project link to add a reference. 

Cheers!

> Skye
-- 
Francis Brosnan Blazquez <francis at aspl.es>
Advanced Software Production Line, S.L.