[pulseaudio-discuss] GStreamer + "Frozen" streams + memory eating

[Colin Guthrie]

Lennart Poettering wrote:
> On Tue, 04.03.08 09:45, Colin Guthrie (gmane at colin.guthr.ie) wrote:
> 
>> I initially put this down to g-p-m doing something silly, but looking at
>> the sound code there is very little in the way of loops and it just
>> off-loads to gstreamer.
>>
>> Last night, I noticed that a similar thing had happened but this time I
>> had lots of pidgin streams "saved up". I started killing them and the
>> same thing happened, pidgin started going nuts and eating ram.
>>
>> So the common factor is gstreamer. Does anyone have any info on this?
>>
>> I'm using gstreamer0.10-pulse-0.9.7 + pulseaudio 0.9.8
> 
> To my knowledge this is indeed an ALSA issue: after resume the PCM
> device is not properly restarted by the kernel drivers. PA doesn't
> notice that, ALSA never asks for new audio data, all current and new
> audio streams appear stuck.
> 
> In theory ALSA defines a proper state machine for dealing with system
> suspends. The issue is just that it doesn't work on some drivers on
> some hardware.
> 
> Please report this issue to ALSA upstream if it persists with the
> newest ALSA driver.

Thanks for the info Lennart. I've not really had issues of late so I
think a kernel update has indeed fixed the issue for me.

The problem of GStreamer causing a deadlock (#188, #251) I think still
exists, but as it is not triggered very often for me it's less of an issue.

Cheers

Col