[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Ccrtp-devel] Proof-of-concept RTP softphone now working on OpenBSD
Re: [Ccrtp-devel] Proof-of-concept RTP softphone now working on OpenBSD 3.9
Mon, 03 Jul 2006 10:24:04 -0400
Thunderbird 22.214.171.124 (X11/20060614)
It sounds like we need to work on some of the configure tests in
commoncpp2 to better support OpenBSD. I would be happy to look into
that, and I would be happy to incorporate changes for OpenBSD for the
next release (1.4.3). This does not directly tell us where Twinkle is
misbehaving on OpenBSD, although again I suspect offhand it might be
Semaphores (and if so that the underlying implementation of conditionals
on OpenBSD pthread has issues) since none of the default tests and
sample apps stress those, but Twinkle certainly does.
Michael Grigoni wrote:
> Greetings Federico and developers:
> As per previous posts, I have been working on
> pthreads issues on OpenBSD 3.9. I had previously
> reported that commoncpp2 passed all of its tests. In
> order to compile it on OpenBSD, I needed to patch
> the configure script to recognize many headers and libs
> (I didn't fix it in autoconf...) and later I noticed
> that configure had failed to find pthread_suspend
> (which we have as pthread_suspend_np) and also atomic.h
> and atomicity.h headers. I patched the generated
> include/config.h and include/cc++/config.h, rebuilt
> the package and retested. Once again it passed all
> of its tests and most of the demos and things seemed
> a bit 'snappier'. I rebuilt ccrtp and twinkle;
> twinkle was not fixed; it still has all of the issues
> I previously reported.
> However, I felt that with the improvements in commoncpp2
> I could verify that an RTP softphone is possible on OpenBSD
> based on ccrtp. I have a working full-duplex proof-of-concept
> softphone which works with 'speakfreely', based on the demo
> programs 'audiotx' and 'audiorx' with some minor changes.
> I ran the phone with 8000bps 8bit LE u-law audio for
> several hours connected to a speakfreely client on a Win2k
> machine; latency is essentially zero from speakfreely to
> the phone and about two seconds from the phone to speakfreely
> (due to processing delays in 'sox'), but the audio is clear
> and uninterrupted in both directions. Performance was
> constant under varying system loads. CPU usage was about
> 1% for the talker and listener combined and about 6%
> for sox.
> If this says anything about threads on OpenBSD,
> I invite comment.
> I put the source files and shell script in a tarball at:
> Please keep in mind that this was a very quick hack; I did
> not add any command-line parsing for destination ip, audio
> parameters or anything else.
> I will appreciate everyone's help to get twinkle working.
> Ccrtp-devel mailing list