In <[EMAIL PROTECTED]> Jeroen van Wolffelaar  wrote:
> So... I think the best way to do is:
> 1) timeouts severly reduced, 30 seconds?
>  (if one goes to look after a failed test, one will reasearch further
>  anyway, and see the timeout and see if that might be the issue)
> 2) kill -KILL in the timeout.
> 
> Especially since we know now a lot of tests fail on some architectures,
> I don't see any harm in running the risk of actidently cutting off a
> test that would otherwise succeed in 30-300 seconds.

hmm. if it passes, it should pass ! If thread-related tests fail, 
probably there is more a dead-lock problem or something of the kind when 
the machine is reasonably fast. It would be better to investigate WHY of 
some slowness than hacking the tests and giving us false positive. We 
have already enough failures on some platforms.
Once things work, the test suites passes in a reasonable amount of 
minutes even on a 200-Mhz range system.

_R


_______________________________________________
kaffe mailing list
[email protected]
http://kaffe.org/cgi-bin/mailman/listinfo/kaffe

Reply via email to