On Sun, Apr 24, 2005 at 07:05:45AM -0700, Ben Collver wrote:
> What was in patch-an was a workaround for GNU PTH.  The whole Midi.C
> needs to be re-designed.  As it was, when it failed to open the device
> path, it looped tightly in the do read() while loop.  This made the
> application unresponsive under PTH.
> 
> A better fix would be for me to alter MidiReaderCallback so that checks
> whether m_MidiFd is negative, and in that case avoid polling.  But that
> would not help with the responsiveness.

Actually, a better fix would be for me to alter Open() to properly test
whether open() fails.  Then the m_MidiReader thread would not be created
if there was nothing to read.

I have updated audio/spiralsynth/patches/patch-an to do this.

Ben