[cfe-commits] [Review Request]Polymorphic Pointers Checker

[Douglas Gregor]

On May 20, 2011, at 4:34 PM, Jim Goodnow II wrote:

> Well, you could ask that about any static analysis checker, but in this 
> case, it is detecting a particular situation that is prone to errors, 
> but is not itself an actual error. The intent is to have a set of C++ 
> checks that give warnings about potentially problematic code. While 
> these two are fairly simple, others will be more involved. I'm just 
> trying to increase the amount of checking that is C++ specific. Adding 
> these two doubles the number of C++ specific checkers.


These generally aren't the heuristics we apply when we decide whether to put a warning in the compiler vs. in the static analyzer. Typically, we put cheap-to-detect issues in the compiler under the control of an appropriate warning flag. When a warning requires more involved flow- or path-sensitive analysis, we put it in the static analyzer.

This kind of problem seems like a solid candidate for going into the compiler.

	- Doug