[cfe-dev] Symbolic Extents

[Zhongxing Xu]

On Fri, Jun 25, 2010 at 5:18 AM, Jordy Rose <jediknil at belkadan.com> wrote:
> The start of work on symbolic extents. I wanted to ask what the best way
> to handle this was. It seems we should have a general EvaluateSymExpr
> method somewhere that simplifies a SymExpr as much as possible, which is
> what the getKnownExtent() method (see patch) would eventually be doing. Or
> perhaps EvaluateSVal, to handle ConcreteInts as well.

I think this patch is a good small improvement over current implementation.

>
> It's also going to soon become helpful to ask for the minimum or maximum
> value of a SymExpr -- knowing the upper bound on an array size would
> certainly be better than no information at all. What should the API for
> that be?

Symbolic extents could go really wild. It's hard to design the
interface without concrete test cases that we want to solve.

>
> (One goal of mine is to be able to model that the length of argv is argc.)
>
> Jordy
> _______________________________________________
> cfe-dev mailing list
> cfe-dev at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/cfe-dev
>
>