pthreads issues?

[Chris Morgan]

I've been working on a wine driver for the jack audio server, 
jackit.sourceforge.net.  The concept of the server is to open audio ports and 
setup a callback function that will be called whenever the server needs more 
audio data.  Everything works fine until the point of starting up the client 
and starting the calls to the callback function.  With the latest cvs version 
of the jack server and some changes to the driver I haven't been able to get 
the driver to start the client up successfully, it just hangs after the below 
first two pthreads errors although I suspect this may be due to the same 
issues as before.  Previously only about 50% of the time would the client be 
started up correctly, the other half of the time it would report failure to 
connect to the server and I would have to kill wine and restart a few times 
before it finally would work.  I got errors like below:

~/c/Program Files/winamp$ wine Winamp.exe
FIXME:pthread_cond_init
FIXME:pthread_cond_signal
FIXME:pthread_cond_wait
fixme:ntdll:RtlNtStatusToDosError no mapping for 00000102

It appears that something went wrong between wine and wineserver 
communication and that strange 102 got stuck in there.  I've also had strange 
X errors and another person reported:

"I'm having problems getting jack to work with a GTK+ app. 3 times out of 4
when I start I get:
engine has shut down socket; thread exiting
Xlib: unexpected async reply (sequence 0x102)!"

Any ideas?

Thanks,
Chris

[Ove Kaaven]

On Sun, 24 Mar 2002, Chris Morgan wrote:
> ~/c/Program Files/winamp$ wine Winamp.exe
> FIXME:pthread_cond_init
> FIXME:pthread_cond_signal
> FIXME:pthread_cond_wait

Maybe these need to actually be implemented then. Conditions does not map
directly to a Win32 synchronization primitive, but it might be possible to
implement them in terms of counters and semaphores. Condition waits do
some interaction with mutexes as well, so perhaps it has hung waiting for
the mutex that the condition wait is supposed to unlock.

Or, you could do some actual traces to find out if implementing these are
really needed or there's some other problem...