selftest --strict can never work?

[Martin Pool]

On 8/23/07, Robert Collins <robertc at robertcollins.net> wrote:
> On Thu, 2007-08-23 at 08:48 +1000, Martin Pool wrote:
> >
> > Rather, the design is not fully implemented.  Per my document, strict
> > is meant to fail on missing dependencies that could be fixed by eg
> > installing more software, but not fail because of platform
> > dependencies.  At the moment we do not distinguish those cases with
> > different exception classes.
>
> Ok. I think this is my fault, I used Feature for things that can be
> fixed by installing strace. Certainly TestSkipped is in my head for
> things that can be fixed.

I wouldn't say that's a fault, it's a reasonable way to use what we
have at the moment.

I was thinking about changing Feature to allow setting what kind of
error is raised if the feature is missing.  For different features it
might be PlatformLimitation, or MissingDependency - or conceivably
something more exotic like KnownFailure.

-- 
Martin