[clang] Fix Lambda Mangling in Namespace-Scope Variable Initializers. (PR #159115)

[Tom Honermann via cfe-commits]

================
@@ -0,0 +1,40 @@
+// RUN: %clang_cc1 -triple x86_64-unknown-linux-gnu \
+// RUN: -aux-triple nvptx64-nvidia-cuda -fcuda-is-device \
+// RUN: -x cuda -emit-llvm %s -o - | FileCheck %s --check-prefix=DEVICE
+
+// RUN: %clang_cc1 -triple x86_64-unknown-linux-gnu \
+// RUN: -aux-triple nvptx64-nvidia-cuda  \
+// RUN: -x cuda -emit-llvm %s -o - | FileCheck %s --check-prefix=HOST
+
+int cudaConfigureCall(int, int, decltype(sizeof(int)) = 0, void* = nullptr);
----------------
tahonermann wrote:

There are local CUDA support header files for tests that can be used here instead of manually declaring `cudaConfigureCall`.
```suggestion
#include "Inputs/cuda.h"
```

https://github.com/llvm/llvm-project/pull/159115