[cfe-commits] [PATCH] getLValueElement

[Ted Kremenek]

On Oct 23, 2008, at 5:52 PM, Zhongxing Xu wrote:

>
>
> On Fri, Oct 24, 2008 at 8:09 AM, Ted Kremenek <kremenek at apple.com>  
> wrote:
>
> On Oct 22, 2008, at 9:21 PM, Zhongxing Xu wrote:
>
>> New patch attached.
>>
>> On Thu, Oct 23, 2008 at 9:28 AM, Ted Kremenek <kremenek at apple.com>  
>> wrote:
>>
>> On Oct 22, 2008, at 6:26 PM, Zhongxing Xu wrote:
>>
>>
>> This assertion will always be true.  Did you mean assert(false &&  
>> "...")?
>>
>> No, I just don't want the program to crash when things happen, but  
>> also put reminder there.
>>
>> Then it's not really an assertion.  Just use a comment.
>>
>> OK. I'll make it a comment. But I do see others doing this in Clang.
>>
>> Those places are probably miswritten assertions.  When you spot  
>> them feel free to bring it to everyone's attention or submit a patch.
>>
>> Thanks Zhongxing!
>>
>> <element2.patch>
>
> +SVal RegionStoreManager::ArrayToPointer(SVal Array) {
> +  const MemRegion* ArrayR = cast<loc::MemRegionVal>(&Array)- 
> >getRegion();
> +
> +  const VarDecl* D = cast<VarRegion>(ArrayR)->getDecl();
> +
> +  const ConstantArrayType* CAT =
> +    cast<ConstantArrayType>(D->getType().getTypePtr());
>
> Won't this cast<> fail for VLAs?
>
> I fix with this:
>
>  +SVal RegionStoreManager::ArrayToPointer(SVal Array) {
> +  const MemRegion* ArrayR = cast<loc::MemRegionVal>(&Array)- 
> >getRegion();
> +
> +  const VarDecl* D = cast<VarRegion>(ArrayR)->getDecl();
> +
> +  if (const ConstantArrayType* CAT =
> +      dyn_cast<ConstantArrayType>(D->getType().getTypePtr())) {
> +
> +    BasicValueFactory& BasicVals = StateMgr.getBasicVals();
> +
> +    nonloc::ConcreteInt Idx(BasicVals.getValue(0, CAT->getSize 
> ().getBitWidth(),
> +                                               false));
> +
> +    ElementRegion* ER = MRMgr.getElementRegion(Idx, ArrayR);
> +
> +    return loc::MemRegionVal(ER);
> +  }
> +
> +  return Array;
> +}

Looks good, except you need a return value for the VLA case (UnknownVal 
()?).

> I am sure we'll iterate on this when we decide to handle VLAs.

Definitely; we just don't want it to crash in the meantime.  ;-)

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-commits/attachments/20081023/20d1281a/attachment.html>