[Alsaplayer-devel] midi plugin

Andy Lo A Foe andy@alsaplayer.org
Sun, 14 Apr 2002 03:21:57 +0200

On Sat, Apr 13, 2002 at 03:11:35PM -1000, Greg Lee wrote:
> > Can you do a backtrace (bt) after the crash?
> (gdb) run
> Starting program: /b/SOUND/APLAYER/alsaplayer/app/alsaplayer 
> [New Thread 1024 (LWP 6585)]
> [New Thread 2049 (LWP 6586)]
> [New Thread 1026 (LWP 6587)]
> [New Thread 2051 (LWP 6588)]
> ptrace: No such process.
> (gdb) bt
> fetch_registers: td_ta_map_id2thr: no thread to satisfy query
> (gdb) 

Did you "make install" after the build? I just tested current CVS on 3
different boxes (local Debian sid box, other sid box, remote Redhat 7.1
box). I'm going to try it on Redhat 7.3 in Vmware, but so far I couldn't
crash or segfault it...

> As far as I know, there will be only one instance of midi-in, if it's
> to work.  (Is the cross-fade actually implemented, so there will actually
> be more than one instance?  If so, I could experiment with trying
> to get multiple instances to work.  But as you know, midi-in is
> not reentrant.)
> Why should data be deallocated in shutdown()?  I'm willing to work
> on this if there is some real point to it.  Does something bad happen
> if I leak?

Not really, shutdown() is called just before the plugin gets unloaded.
And in most cases AlsaPlayer will exit right after that so leaked memory is
not really a problem. Note that I will comment out the shutdown() call
for now, since init() is not called from the host either. So I wouldn't
worry about it for now...