C++: Suggestions to #15343

[MathiasVP]

@bdrodes these are my suggestions to github#15343.

Commit-by-commit review recommended - especially if you're not yet comfortable with parameterized modules!

• The first commit does a straightforward refactoring of the interface of FlowFromFree: instead of passing two predicates as parameters to the module I wrap the "input predicates" into a module signature and change the FlowFromFree module to take such an "input module" as a parameter.
• The next commit is really the meat of my suggestion: it adds a new predicate to the "input module" that represents additional requirements on the relationship between the source and the sink. In particular, for the github/codeql queries this can be the dominance/post-domination relationship, and for microsoft/codeql this defaults to any() meaning that no additional relationship is required.
• The next commit makes cpp/use-after-free and cpp/double-free conform to this new interface.
• The final commit accepts the test changes. This basically undoes the changes you saw in Generalization of FlowAfterFree github/codeql#15343.

The change I'm proposing here should mean that there are no performance changes compared to what's running currently in Code Scanning 🤞. But obviously we need to run DCA on this to confirm

[bdrodes]

Interesting solution. I'm going to run with it. My only concern is that an 'any' predicate would be somehow inefficient, but I assume there are optimizations to avoid explosion with these?

[MathiasVP]

Interesting solution. I'm going to run with it. My only concern is that an 'any' predicate would be somehow inefficient, but I assume there are optimizations to avoid explosion with these?

Note that the predicate is marked with a bindingset annotation (which means that the arguments are already bound). So there's no explosion. If we removed that annotation it would produce a cartesian product.