Hi,
I've thinking about a change in the PulseAudio driver, changing
bufattr.maxlength from period_bytes to -1. What this means is basically:
if PulseAudio finds out that the specified buffer length (i e latency)
can't be satisfied (usually it tries and receives a drop-out), it will
use a bigger latency. The current behavior is to make drop-outs instead.
This is somewhat a matter of taste with pros and cons in both camps, but
I'm thinking that since PulseAudio isn't low-latency as in good live
real-time playing anyway, the PulseAudio driver will probably be used
more when FluidSynth is just a MIDI file player, and in that case,
bigger latency is probably preferable to drop-outs.
What do you think?
// David