[cfe-commits] r158153 - in /cfe/trunk: lib/CodeGen/CGExpr.cpp test/CodeGen/vla.c

[John McCall]

On Jun 7, 2012, at 10:07 AM, Fariborz Jahanian wrote:
> Author: fjahanian
> Date: Thu Jun  7 12:07:15 2012
> New Revision: 158153
> 
> URL: http://llvm.org/viewvc/llvm-project?rev=158153&view=rev
> Log:
> When emitting compund literal of vla pointer elements, make 
> sure to emit vla size to prevent an irgen crash.
> // rdar://11485774
> 
> Modified:
>    cfe/trunk/lib/CodeGen/CGExpr.cpp
>    cfe/trunk/test/CodeGen/vla.c
> 
> Modified: cfe/trunk/lib/CodeGen/CGExpr.cpp
> URL: http://llvm.org/viewvc/llvm-project/cfe/trunk/lib/CodeGen/CGExpr.cpp?rev=158153&r1=158152&r2=158153&view=diff
> ==============================================================================
> --- cfe/trunk/lib/CodeGen/CGExpr.cpp (original)
> +++ cfe/trunk/lib/CodeGen/CGExpr.cpp Thu Jun  7 12:07:15 2012
> @@ -2114,7 +2114,11 @@
>     llvm::Value *GlobalPtr = CGM.GetAddrOfConstantCompoundLiteral(E);
>     return MakeAddrLValue(GlobalPtr, E->getType());
>   }
> -
> +  if (const PointerType *pointerType = E->getType()->getAs<PointerType>())
> +    if (pointerType->getPointeeType()->isVariableArrayType())
> +      // make sure to emit the VLA size.
> +    EmitVariablyModifiedType(pointerType->getPointeeType());
> +  

Please just test E->getType()->isVariablyModifiedType() instead of looking
specifically for a pointer to VLA.  The test case here would be even more
contrived, but something like:
  int n = 20, **a;
  (int (**)[i]){&a}[0][1][5] = 0;

John.