Re: Adding native audio support to ports

[adr <adr%sdf.org@localhost>]

On Sun, 28 Jan 2024, Brian Buhrow wrote:
> 	Hello.  When I worked on this issue, I discovered the problem wasn't with the size of the
> audio buffer in the audio  subsystem of the kernel, but that the audio driver, in Flite in my
> case, confused drain and flush.  Drain tells the audio subsystem to finish what it's doing and
> return.  Flush tells the audio system to stop what it's doing, *NOW*, and return.

Yes, it is well documented in audio(4), and I'm pretty sure espeak-ng
is using it, because I wrote code in pcaudiolib to use the netbsd
audio api, and I can assure you, it is using flush correctly.

> On netBSD, at least until NetBSD-9.9977, the flush ioctl works as it should and audio can be
> stopped very quickly.  I see no reason why this should not be true under NetBSD-10 or -current

Yes, in fact flite is working ok. I've been using your patched
sources, I'll will take a look to the latest release when I'm
finishing with emacspeak - espeak-ng - pcaudiolib.

> sources for this error, but if you have not, I'm guessing the issue you're seeing is in one of those, not NetBSD.

Oh, I don' think the fault is in NetBSD. I'm going to use the water
marks to constraint the buffer size, it should affect the latency
as it's done in the alsa code (in pcaudiolib).

I'm not sure, but my gut tells me this could be the issue.

adr.