[cfe-commits] r61069 - /cfe/trunk/lib/Analysis/RegionStore.cpp

[Zhongxing Xu]

On Wed, Dec 17, 2008 at 2:40 AM, Ted Kremenek <kremenek at apple.com> wrote:

> Hi Zhongxing,
> Thanks for doing this.  A couple comments inline.
>
> On Dec 15, 2008, at 6:36 PM, Zhongxing Xu wrote:
>
> URL: http://llvm.org/viewvc/llvm-project?rev=61069&view=rev
> Log:
> Implement RegionStoreManager::Remove().
>
> Modified:
>    cfe/trunk/lib/Analysis/RegionStore.cpp
>
> Modified: cfe/trunk/lib/Analysis/RegionStore.cpp
> URL:
> http://llvm.org/viewvc/llvm-project/cfe/trunk/lib/Analysis/RegionStore.cpp?rev=61069&r1=61068&r2=61069&view=diff
>
>
> ==============================================================================
> --- cfe/trunk/lib/Analysis/RegionStore.cpp (original)
> +++ cfe/trunk/lib/Analysis/RegionStore.cpp Mon Dec 15 20:36:30 2008
> @@ -111,10 +111,7 @@
>
> +Store RegionStoreManager::Remove(Store store, Loc L) {
> +  RegionBindingsTy B = GetRegionBindings(store);
> +
> +  const MemRegion* R = cast<loc::MemRegionVal>(L).getRegion();
> +  assert(R);
>
>
> This assertion seems bogus.  If Remove should always expect a MemRegion*
> then we should change its interface to accept that as an argument instead of
> a Loc value.  Otherwise, Remove should handle the case when 'L' isn't a
> region.
>
> +
> +  return RBFactory.Remove(B, R).getRoot();
> +}
>
>
> Do we plan on inserting Unknown into the map to marked the value as killed?
>  If so, we should only do that if it was in the map in the first place.
>

I am doing an overhaul on the bind/remove methods of storemanager to
implement the lazy binding idea. These are fixed there.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-commits/attachments/20081217/e1ba84c0/attachment.html>