namecache: numneg > 0 but ncneg is empty

[Peter Holm]

On Thu, Dec 19, 2013 at 03:46:39PM +0200, Andriy Gapon wrote:
> on 19/12/2013 10:12 Peter Holm said the following:
> > On Thu, Dec 19, 2013 at 09:56:28AM +0200, Andriy Gapon wrote:
> >> Peter,
> >>
> >> I am curious about what ideology is behind vfs testing in stress2.  I know that
> >> I can just look at the code myself, but hope that asking you could be faster.
> >> Does stress2 exercise a certain set of scenarios?  Or does it have an element of
> >> randomness?
> >>
> > 
> > The tests found in stress2/testcases does everything in a random
> > fashion.
> 
> Could you please add a few words about what kind of randomness is that?
> E.g. I looked at testcases/rename and it seems to do pretty predictable and
> linear renaming of files within the same directory.  Also, it seems that the
> test would be aborted should a rename operation fail.  But that would be a valid
> outcome in a truly random / chaotic testing.
> 
> > Test found in stress2/misc are for the most part scenarios that has
> > been used for finding specific problems.
> 
> 
> -- 
> Andriy Gapon

For testcases/rename the number of files to rename is controlled by
the random number of invocations of this test. Two new rename
scenarios was added recently by jmg@ to address specific SU+J issues.
More rename scenarios can be found in stress2/misc/rename*

-- 
Peter