Re: [Gcl-devel] Re: Interpretation of random tester output

[Paul F. Dietz]

Camm Maguire wrote:

> Paul, I want to thank you for this very helpful tool, and your
> responsive explanations.  Of course I'm being a bit lazy as I could
> study the code itself and answer many of my questions, but I'm short
> on time right now.  If I might put to you one more -- I take it the
> algorithm goes through 2000 runs looking for random failures in a
> fashion which does not depend on what might have been found earlier in
> the run.  The next 2000 -- are these limited to searching a subspace
> based on errors found in the first set, or is one still covering new
> ground in these tests as well?  When is one done boiling down errors
> found in the first set?

All the individual random tests are independent of each other.

The user interface for this has evolved, and isn't terribly
good.  It was mainly 'good enough' for me to use.  I should
have it prune error cases as soon as they are found, instead
of waiting to do them in batches.

        Paul