Derek isn't subscribed to squeak-dev anymore, forwarding his reply:

Begin forwarded message from Derek O'Connell <doc at doconnel.f9.co.uk>
> 
> Hi Bert, Tim,
> 
> the fact that it works once makes it smells like an "EINTR" problem. If
> you can "strace" to locate where it's occurring then it's simply a
> matter of wrapping it in a retry-loop. There will be other places where
> sys-calls are similarly wrapped (if not there then definitely in the
> ALSA/PA plug-in's and probably in the VM itself).
> 
> In case you're not aware, a change in the Linux kernel several years ago
> dictated that anything could be interrupted at any time. I forget the
> details but it had a big impact on user-land code. It has took a long
> time for coders to adapt their code, not helped by the unpredictable
> nature of such interrupts and that there is no one-shot solution, hence
> some code could run for years without problem. To make matters worse you
> can end up plugging one bit of code only for EINTR to start popping up
> elsewhere. OTOH it seems insane to have to have to wrap every sys-call.
> 
> Re camera code: std v4l code taken straight from examples plus
> non-circular buffering. Anything else can only be to do pixel formatting.
> 
> Best wishes,
> 
> -D

-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/pkcs7-signature
Size: 4115 bytes
Desc: not available
Url : http://lists.squeakfoundation.org/pipermail/squeak-dev/attachments/20150209/93aae188/smime.bin