[cfe-dev] Allowing checkers to mark symbols as live

[Ted Kremenek]

On Aug 9, 2010, at 12:31 PM, Jordy Rose wrote:

> There is no reference to the symbol besides what's in the GDM. A
> SymbolRegionValue is for region contents, not metadata, and a SymbolDerived
> requires a symbolic parent. SymbolExtent, besides currently having a
> dedicated use, has the wrong semantics in that it isn't dependent on being
> at a certain place in the code.

The extent is a property of the region.  It's only symbolic if the region is symbolic; otherwise it is a value we should be able to compute up front.

For symbolic constraints, the dependence on where we are in the code is modeled by the constraints on the symbol.  The symbol itself doesn't need to record any more context other than being based on the symbolic region (which should capture all the context we need).  Why do we need to capture any more context in the SymbolExtent?