__is_constructible return true for abstract types PR19178

[Alp Toker]

On 14/04/2014 11:59, Nikola Smiljanic wrote:
> Please review.
>
> pr19178.patch
>
>
> diff --git a/lib/Sema/SemaExprCXX.cpp b/lib/Sema/SemaExprCXX.cpp
> index 8b9c0e2..2ff501f 100644
> --- a/lib/Sema/SemaExprCXX.cpp
> +++ b/lib/Sema/SemaExprCXX.cpp
> @@ -3656,6 +3656,12 @@ static bool evaluateTypeTrait(Sema &S, TypeTrait Kind, SourceLocation KWLoc,
>       if (Args[0]->getType()->isIncompleteType())
>         return false;
>   
> +    // Make sure the first argument is not an abstract type.
> +
> +    CXXRecordDecl *RD = Args[0]->getType()->getAsCXXRecordDecl();
> +    if (RD && RD->isAbstract())
> +      return false;
> +

The fix is correct. Perhaps you could use RequireNonAbstractType() 
instead of checking for CXXRecordDecl directly, something like this 
taken from the counterpart in BTT_IsConvertible:


-    // Make sure the first argument is a complete type.
-    if (Args[0]->getType()->isIncompleteType())
+    // Make sure the first argument is a complete non-abstract type.
+    if (S.RequireCompleteType(KWLoc, Args[0]->getType(), 0) ||
+        S.RequireNonAbstractType(KWLoc, Args[0]->getType(), 0))
        return false;

Either way looks good to me, thanks!

Alp.


>       SmallVector<OpaqueValueExpr, 2> OpaqueArgExprs;
>       SmallVector<Expr *, 2> ArgExprs;
>       ArgExprs.reserve(Args.size() - 1);
> diff --git a/test/SemaCXX/type-traits.cpp b/test/SemaCXX/type-traits.cpp
> index 28fb8dc..b8557c4 100644
> --- a/test/SemaCXX/type-traits.cpp
> +++ b/test/SemaCXX/type-traits.cpp
> @@ -1964,6 +1964,8 @@ void constructible_checks() {
>   
>     { int arr[T(__is_constructible(NonPOD, int))]; }
>     { int arr[F(__is_nothrow_constructible(NonPOD, int))]; }
> +
> +  { int arr[F(__is_constructible(Abstract))]; } // PR19178
>   }
>   
>   // Instantiation of __is_trivially_constructible
>
>
> _______________________________________________
> cfe-commits mailing list
> cfe-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits

-- 
http://www.nuanti.com
the browser experts