Possible Memory Leak in pd-linux-0.25TEST4.tar

est at hyperreal.org est at hyperreal.org
Wed Mar 31 00:03:24 CEST 1999


Guenter Geiger discourseth:
> est at hyperreal.org writes:
>  > Hmm, I don't even get that far.  I've tried the last 3 versions of pd
>  > on RedHat 5.2 (using all their latest updates) and every one of them
>  > pretty quickly freezes the whole system in a Tk file-open dialog.  I
>  > can't even transfer to my character-based virtual terminals or reboot
>  > with control-at-delete!  I also tried the latest Tcl/Tk (8.05) with
>  > the same results.
> 
> Mhmm, do you mean version 0.23, 0.24, and 0.25 ?

I built and tried the following, the last one is from MP's site:

/usr/archive/pd_0.24.tar.gz
/usr/archive/pd-0.25TEST3linux.tar.gz
/usr/archive/pd-linux-0.25TEST4.tar.gz

> First of all never call pd under user root. (We might add a
> switch for enabling high priority, as this happens quite often)
> The freeze of your machine should go away.

Unfortunately I haven't been calling it as root. :(

> Your problem might be related (puh, .. again ) to the audio drivers.
> (Or possibly the way I access them through very short audio buffers)
> This can cause a hang in the pd main process, which is most commonly
> the problem in these cases (Nothing to do with your Tcl/Tk version). 

I doubt that's the problem in my case.  Coincidentally, I've done a
lot of work with buffer configuration for my audio driver. :)

> You might try the -dac or -nosound switch on version 0.24 or 
> it's corresponding -noadc in version 0.25.

OK..I'll give those a try.

> What soundcard, kernel version, sounddriver do you use ?

It's a PCI128 soundcard.  My kernel is the latest update from redhat:
2.0.36-3.  I *did* modify the sounddriver to fix a bug in the mixer
map.

> As you probably realized, between version 0.25TEST3 for linux and 
> 0.25TEST4, we made a switch of responsibilities for the linux 
> sounddriver code. We therefore have two rather different versions
> (TEST3, which is on ftp://wonk.epy.co.at/pub/pd/linux/ and version
> 0.25TEST4, which is the original version from Miller).
> Both of them are considered to really be TEST versions, as in general
> it is hard to predict how the Linux sounddrivers behave under
> "realtime" conditions, and we need feedback there as we can't possibly 
> test on all soundcards, with all drivers.

If I can get PD working here, I can definitely help.  I'm working on
real-time sound software myself. :)

Eric



More information about the Pd-list mailing list