[Gllug] memtest86 running time

Nix nix at esperi.org.uk
Mon May 10 22:39:04 UTC 2004


On Sun, 09 May 2004, John Hearns said:
> We also burn our nodes in doing an endless loop of kernel compiles,
> so it might be worth doing that too.

GCC bootstrap-and-test runs are much more effective, I think, because of
the three-stage process and the subsequent self-tests: if it fails,
you're likely to *know*. With kernel compiles, a lot of errors will be
spotted (if they make GCC segfault), but if they merely make it produce
invalid results by bad luck you might miss it.

(I'd define `failure' of the self-tests here as `the test results
change', not `there are any failures': there are always a few.)


(Mind you the only broken boxes I've had a chance to try this on were
sufficiently badly off that they froze halfway through stage1 of the
first run. You can't miss that, even with a mere kernel compile. ;) )

-- 
`If you believe in strong AI, then death is no longer a mystery,
 but merely a lack of adequate backups.' --- Steven McDougall
-- 
Gllug mailing list  -  Gllug at gllug.org.uk
http://lists.gllug.org.uk/mailman/listinfo/gllug




More information about the GLLUG mailing list