16 May
2009
16 May
'09
12:31 a.m.
On Fri, 2009-05-15 at 11:21 +0200, Roberto Gordo Saez wrote:
On Fri, May 15, 2009 at 02:02:11PM +0900, Dietrich
Bollmann wrote:
If I just would be able to list other details than that ugly "krrk!"
every some seconds :) You can be sure that I would send them (...but to
whom?).
what I am dreaming about is some kind of "micro-level task logger" which
would react to the same pattern Pure Data's "DIO errors" button reacts
to. It would save a list of all processes running in the kernel when
some DIO error occurs. Letting this tool run for some time and looking
at the intersection of all logged processes should reveal the bad one :)
Probably this is a much too naive idea, I know ... but it never hurts to
ask.
Another idea I thought about would be some "dedicated sound / signal
processor" for the audio signal processing. The processor on my laptop
would only be responsible for controlling the dedicated processor and
there would be no way for any other process to cause ugly glitches. I
once thought a sound card would be exactly this kind of thing. But I
seem to have been mistaken. (I have to admit that I still don't really
understand what beyond a DAC a soundcard atually is.) Another problem
is that I probably would have to use USB to connect such a thing to my
IBM/Lenovo X60s and that this might cause another (latency) problem?
Sorry for spamming the list with this kind of uneducated dilettantish
ideas :)
Dietrich
So I now have a running realtime kernel - but
unfortunately the glitches
remain...
I wonder if there is some tool which allows to trace them down to some
application / driver / ... ?
Yeah, one of my systems periodically crashes with the -rt kernel even
with no realtime processes running and I'm still struggling to find
where the problem is, so unfortunately the glitches of the -rt kernel
are common. If the failure is at run time, it is usually harder to find
the problem, probably the best "tool" to fix those bugs is to send a
detailed report to the developers :P