[Python-Dev] buildbot VM exhausted in test_crashers

[Nick Coghlan]

On Sun, Mar 20, 2011 at 9:44 AM, David Bolen <db3l.net at gmail.com> wrote:
> Nick Coghlan <ncoghlan at gmail.com> writes:
>
>> I don't want to give up completely on the idea just yet, but I'll
>> experiment in the sandbox before I turn it back on.
>
> If you get to that point again, I'd also be willing to pick a time to
> manually check out the right branch or whatever and try it manually
> on one of two of my slaves while I'm watching.
>
> Typically I don't pay too much attention to various failures created
> by testing, but this particular case breaks in a way that the slaves don't
> recover, so is probably better to put a toe in the water slowly while
> watching :-)

That would be great. The first thing I'll be trying (once I get back
to it) is simply skipping the crasher I think is the main cause of the
instability (the compiler recursion one). I already deliberately dodge
the two crashers that are known to cause infinite loops, skipping one
more wouldn't be a terrible outcome.

I need to bug Barry about his PPC buildbot as well. It seemed to be
surviving even after I had cranked the expression size in the compiler
recursion crasher to absolutely ridiculous levels, so I'm wondering if
there may be something different about the way PPC handles the stack.

Cheers,
Nick.

-- 
Nick Coghlan   |   ncoghlan at gmail.com   |   Brisbane, Australia