"MikkelFJ" <mikkelj-anti-spam / post1.dknet.dk> writes:

> The automated testing sounds nice - and certainly would be visible
> outside the Ruby enviroment.  But with threading issues in
> particular it may be hairy. You can very correct results, but
> verifying fair scheduling is something different where you need some
> kind of statistical measure.

The existing Rubicon tests have some elementary checks for this
(things like higher priority threads get more instructions executed
than lower priority ones). In fact it was one of these tests that
found the Windows problem. I'd welcome a better set of threading tests
if anyone is feeling inspired.


Dave