On Fri, Nov 7, 2008 at 2:01 PM, Zhongxing Xu <[EMAIL PROTECTED]> wrote:
> > > On Fri, Nov 7, 2008 at 1:54 PM, Ted Kremenek <[EMAIL PROTECTED]> wrote: > >> >> On Nov 6, 2008, at 9:45 PM, Zhongxing Xu wrote: >> >> I actually don't think a hybrid approach is not all that difficult to >>> implement. >>> >>> Do you mean " ... don't think ... is all that difficult ...", since two >>> negation imply no negation. >>> >> >> Yes. >> >> >>> >>> A switch statement on the kind of a region should be enough to determine >>> when we can determine the extent of a region simply from the MemRegion >>> object itself or we need to look in a side map. >>> >>> I like (b1), too. Then we have a clear direction to go now. Thanks! >>> >> >> Great. >> > > Also I think we should have a clear client for the extent before/when we > implement the extent mechanism. One direct client is the array element > access checker. It should be in the EvalLocation() method to check if the > array access is out of bound. But not all store manager support this check. > Shall we make a new transfer function to do this check? > So I feel these are what we should do: 1. overhaul the transfer function interface. 2. add specific checker (e.g. array bound checker) that will use the region extent property. 3. implement the region extent interface.
_______________________________________________ cfe-commits mailing list [email protected] http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
