Conversation
Edited 15 days ago
Tfw I test my small Pleroma patches more than pipewire devs test their entire fucking releases.

Audio buffer issues have been a thing for years, they always appear out of nowhere after half a year and then magically fix themselves few releases later only to come back again.

RT: https://fluffytail.org/objects/64a6bf66-2256-429f-bd82-dd62d9caa9ea
3
0
1
@phnt isn't it normal that using 1 sample buffer would fuck your audio?
1
0
0
And guess what barely had buffer issues, pulseaudio. Because the buffer was so big that it barely had a chance to occur. But no, we must have ultra low latency by default where setting arbitrary unit values barely does anything.
1
0
2
@snacks It shouldn't be 1 sample because that's is dumb. The audio buffer size is supposed to be dynamic in pipewire and was in pulseaudio.

Pulseaudio even conveniently allowed you to see the latency in ms and set the buffer size accordingly. Pipewire has some arbitrary unit that has no explanation what it does, a magic number basically.
1
1
1

@phnt@fluffytail.org the average FDO developer literally doesn't even use L'Eunuchs (or was that GNOME? I don't care, same thing)

1
0
1

@phnt “it works on my machine”

look inside
IT’S A FUCKING MAC

0
0
1
@phnt isn't clock.force-quantum exactly to disable any dynamic buffer size? And quantum seems to just be samples
1
0
0
@phnt > than ... devs test their entire fucking releases.

i think that's more accurate.
0
0
0
@snacks quantum is supposed to be the "default middle ground" where min-quantum is for when the buffer is too large and max-quantum is for when the buffer is too small. Is it supposed to be between max/min, averaged around default. At least that's what it was years ago. If they changed it, they are idiots.

To force a specific buffer size, you use force-quantum.
2
0
0
@snacks which is where I guess you got the 1 sample. I forgot that ain't a boolean that enforces clock.quantum.
0
0
1
@phnt as far as i got it today applications just pick one of the three quantum settings
1
0
1
@phnt hmm, seems you're right that it can just pick something between all the values looking at pw-top
0
0
0