[PATCH] D45517: [analyzer] False positive refutation with Z3

[Mikhail Ramalho via Phabricator via cfe-commits]

mikhail.ramalho added inline comments.


================
Comment at: include/clang/StaticAnalyzer/Core/PathSensitive/ConstraintManager.h:183
+  virtual void reset() {}
+  virtual bool isModelFeasible() { return true; }
+  virtual void addRangeConstraints(ConstraintRangeTy) {}
----------------
george.karpenkov wrote:
> Making those virtual does not make much sense to me.
> Returning `true` by default is not correct.
> When we are using the visitor, we should already know we have a `Z3ConstraintsManager`, why can't we just use methods of that class?
Z3ConstraintManager is fully contained inside a .cpp file, so we need isModelFeasible and addRangeConstraints to be exposed via its base class.

Another solution is to split Z3ConstraintManager into a .h and a .cpp file and include the header. We would then be able to use it directly, instead of through a ConstraintManager object.

I honestly prefer the latter. What do you think?


https://reviews.llvm.org/D45517