AllBottom: simplify and fix meetWith

[langston-barrett]

Previously, AllBottom would throw an exception when meetWith was called with a user-supplied EdgeFunction. The proper behavior is to return AllBottom unconditionally.

[ericbodden]

I am sorry but the check exists for a reason: when the exception is thrown then you are most likely doing something wrong. IIRC the solver should never get into a situation where it needs to compose AllBottom with a user-supplied function. When does that happen in your case?

[ericbodden]

I had another look to make sure. I am reasonably sure that AllBottom is only used in the case in which IDE is used to instantiate an IFDS problem - and then the exception actually is a sensible precaution. I am not sure AllBottom ever makes sense to be used within another analysis. So can you please elaborate on your use case? Thanks!

[langston-barrett]

Ah, I was under the impression that AllBottom was a general purpose utility class, not an internal class to be used only in the solver. The reason I encountered the exception was that I returned an instance of AllBottom from an implementation of EdgeFunctions, which it sounds like I should not have tried.

To avoid such a misuse in the future, would it make sense to make AllBottom (and perhaps AllTop) into private or protected classes?

[ericbodden]

I can't 100% rule out the fact that AllBottom may be useful

The reason I encountered the exception was that I returned an instance of AllBottom from an implementation of EdgeFunctions, which it sounds like I should not have tried.

@langston-barrett that's interesting. So far I would not have thought this use case to ever arise, but I'd be happy to have you convince me otherwise. Would it be possible for you to paste the relevant part of the edge-function implementation here? If AllBottom seems generally useful then I'd be happy to remove the checks.