On Mon, Jul 6, 2009 at 11:23 PM, Fernando Lopez-Lezcano wrote:
On Mon, 2009-07-06 at 23:11 -0400, Orcan Ogetbil wrote:
>
> It seems to me that the new fluidsynth is not that stable. I start
> qjackctl which runs jackd (this is Fedora's jack 0.116.1-5) with
> /usr/bin/jackd -R -dalsa -dhw:0 -r44100 -p1024 -n2 -Xseq
> Then I start qsynth. As soon as I do this jack starts getting very
> many xruns and in about 10 seconds qjackctl exits with a SIGKILL, also
> killing jackd with itself along the way.
>
> Now I start qjackctl once more and click on "Restart" on already
> running qsynth. This time there are no xruns and jackd runs without
> problems.
>
> It's not desirable to use jack always with this workaround and I am
> not pushing the package to stable until the issue is resolved.
>
> Fernando, do you still develop jack?
No, sorry, I don't...
> Do you know where the problem might be?
It would seem to be some sort of caching problem. That is, the first
time fluidsynth is starting up and reading the soundfont, the second
time the soundfont is already in memory. This should not happen, of
course, as the process callback in fluidsynth should not be doing any
file i/o directly (or be influenced by it).
Is this happening with the standard Fedora kernel or the rt kernel?
This is Fedora kernel
2.6.29.4-167.fc11.x86_64
I'll see what happens with the rt kernel.
What happens if you change the soundfont while fluid is running
happily
the second time?
When I select a different soundfont in the second run of fluidsynth,
nothing bad happens. Things continue to work properly.
Another test I have made is: I ran jackd without the -R (realtime)
flag. This time when I start qsynth, I get lots of xruns again but
this time qjackctl does not shut down and continues to report the
xruns.
Can you (or anyone) reproduce any of these results?
Orcan