Bug with rule refinement: Static analysis

[anthonyanjorin]

Refinement does not seem to work correctly for overriding correspondence types.

Subtypes should win, but we should also complain if there is no clear "lowest" subtype. Currently we just pick one of the types without any warning or error.

[anthonyanjorin]

Any progress? Have you been able to reproduce?

[filipgia]

Hello,
@saman and I had two meetings with @NilsWeidmann . We are investigating on this issue. Right now, we have three breakpoints. We are trying to find the way to make a sup-type win and log it with the Logger.

[anthonyanjorin]

ok - so first concentrate on simply reproducing right? That means I believe a clear subtype must win at the moment. The problem is more when there is no clear lowest subtype. This is a specification error and must be flagged as such. Instead at the moment something just happens.

E.g.:

CorrA ...
CorrB extends CorrA
CorrC extends CorrA

If you then have a rule X with CorrB, a rule Y with CorrC, and a rule Z that refines X and Y you should get an error as it is unclear if Z should use CorrB and CorrC (they are not comparable).

Clear?

[filipgia]

Yes it is clear. At the moment we are using
FeatureModelConciseToSafe. But we will also try your example.

[anthonyanjorin]

That’s fine. Any example is ok if you can reproduce the problem.