[Fluxus] fluxa fixes

[David Griffiths]

Hi Kassen,

On 24/11/12 18:03, Kassen wrote:
>> Thanks for this, I wasn't getting the warning so I think this is a
>> 32/64bit issue, thanks for tracking it down (and fixing the other
>> related bug)...
> 
> From what I gathered online this is a general problem these days,
> affecting people across compilers and OS's.
> 
>> libsndfile still seems to be randomly crashing on some of my samples - I
>> haven't looked into it properly yet, it could be that the async loading
>> thread isn't working properly and libsndfile does more allocation that
>> causes jack to fall over for me.
> 
> This is really weird. LibSndFile has been good for me and it's good
> for our friends over at SC, CK, PD, etc, etc, etc too. Could it be
> that your samples have some sort of issue? I stubbornly put it back
> here, if you mail me one of your samples that cause issues (gmail
> takes up to 10MB or so) I could test that on my end.

Not sure if its a particular sample or some sort of non-realtime
allocation or processing timing issue, which means we need to fix the
loader thread (more likely this is the case than a libsndfile bug)...
it's intermittent which is why the simple wav loader tends to
accidentally get committed around gig times :)

>> This could be the problem, I haven't been testing seq properly since
>> moving to temporal recursion. Let me do some tests and get back to you...
> 
> In your other mail you ask for a example. Here is a simple one that
> still gives issues;
> 
> (seq (λ (time clock)
>         (when (zmod clock 4) (play time (sample "TR909/kk.wav" 440)))
>         (when (even? clock) (play time (* .5 (sample
> "TR909/clht.wav" 440))))
>         .1))
> 
> ...that's intentionally "my first Fluxa" material. Because this one
> does not allocate so many nodes it takes its time to start being
> problematic. I don't think that one involves a lot of redacted stuff.

This one seems fine here (on master and redacted branch).

> This is my regular test case, it needs a LOT of redacted stuff. Starts
> giving trouble after half a minute or so;

Just tried this, running for 20 mins now...

So the problem must be either another 32/26bit issue, or..?

Are you getting any messages to stderr from fluxa when it goes wrong? If
the ringbuffer isn't being emptied fast enough it will tell you when it
runs out...

cheers,

dave