[llvm-commits] [llvm] r166157 - /llvm/trunk/lib/VMCore/Verifier.cpp

[Bill Wendling]

On Oct 17, 2012, at 5:01 PM, Eli Friedman <eli.friedman at gmail.com> wrote:

> On Wed, Oct 17, 2012 at 4:54 PM, Bill Wendling <isanbard at gmail.com> wrote:
>> Author: void
>> Date: Wed Oct 17 18:54:19 2012
>> New Revision: 166157
>> 
>> URL: http://llvm.org/viewvc/llvm-project?rev=166157&view=rev
>> Log:
>> Check that the operand of the GEP is not the GEP itself. This occurred during an LTO build of LLVM.
>> 
>> Modified:
>>    llvm/trunk/lib/VMCore/Verifier.cpp
>> 
>> Modified: llvm/trunk/lib/VMCore/Verifier.cpp
>> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/VMCore/Verifier.cpp?rev=166157&r1=166156&r2=166157&view=diff
>> ==============================================================================
>> --- llvm/trunk/lib/VMCore/Verifier.cpp (original)
>> +++ llvm/trunk/lib/VMCore/Verifier.cpp Wed Oct 17 18:54:19 2012
>> @@ -1371,9 +1371,11 @@
>>   Type *TargetTy = GEP.getPointerOperandType()->getScalarType();
>> 
>>   Assert1(isa<PointerType>(TargetTy),
>> -    "GEP base pointer is not a vector or a vector of pointers", &GEP);
>> +          "GEP base pointer is not a vector or a vector of pointers", &GEP);
>>   Assert1(cast<PointerType>(TargetTy)->getElementType()->isSized(),
>>           "GEP into unsized type!", &GEP);
>> +  Assert1(GEP.getPointerOperand() != &GEP,
>> +          "GEP is using the result as the pointer operand!", &GEP);
> 
> This check is bogus.  The following is legal IR:
> 
> define void @f() {
>  ret void
> 
> notreachedbb:
>  %xx = getelementptr i8* %xx, i32 0
>  ret void
> }
> 
I disagree. The code you have there isn't in SSA form, which though there may be holes in the IR format which allows for such non-SSA form, LLVM itself wants to be in SSA form and shouldn't encourage this type of code.

-bw