dino: docs: leave myself a TODO about niceness/priority
This commit is contained in:
parent
4909127ec7
commit
4a498ef1a9
|
@ -16,13 +16,16 @@
|
||||||
# outstanding bugs:
|
# outstanding bugs:
|
||||||
# - mic is sometimes disabled at call start despite presenting as enabled
|
# - mic is sometimes disabled at call start despite presenting as enabled
|
||||||
# - fix is to toggle it off -> on in the Dino UI
|
# - fix is to toggle it off -> on in the Dino UI
|
||||||
|
# - default mic gain is WAY TOO MUCH (heavily distorted)
|
||||||
|
# - TODO: dino should have more optimal niceness/priority to ensure it can process its buffers
|
||||||
|
#
|
||||||
|
# probably fixed:
|
||||||
# - once per 1-2 minutes dino will temporarily drop mic input:
|
# - once per 1-2 minutes dino will temporarily drop mic input:
|
||||||
# - `rtp-WRNING: plugin.vala:148: Warning in pipeline: Can't record audio fast enough
|
# - `rtp-WRNING: plugin.vala:148: Warning in pipeline: Can't record audio fast enough
|
||||||
# - this was *partially* fixed by bumping the pipewire mic buffer to 2048 samples (from ~512)
|
# - this was *partially* fixed by bumping the pipewire mic buffer to 2048 samples (from ~512)
|
||||||
# - that fix can't extend to Dino itself except by patching its code (perhaps)
|
# - this was further fixed by setting PULSE_LATENCY_MSEC=20.
|
||||||
# - patching every `info.rate / 100` -> `info.rate / 50` in Dino to double the buffer size *seems* to have no effect
|
# - possibly Dino should be updated internally: `info.rate / 100` -> `info.rate / 50`.
|
||||||
# - possible the 10ms buffer constant is inside `webrtc` itself
|
# - i think that affects the batching for echo cancellation, adaptive gain control, etc.
|
||||||
# - it's using gstreamer, so maybe other ways to introspect that/fix it
|
|
||||||
#
|
#
|
||||||
{ config, lib, ... }:
|
{ config, lib, ... }:
|
||||||
let
|
let
|
||||||
|
|
Loading…
Reference in New Issue
Block a user