[clang] [clang] Skip host default include paths for GPU triples (PR #192821)

[Yaxun Liu via cfe-commits]

================
@@ -0,0 +1,19 @@
+// Check that GPU targets do not get the host system's default C include
+// paths (e.g. /usr/include, /usr/local/include) appended to the search list.
+// GPU toolchains are freestanding and manage their own include paths.
----------------
yxsamliu wrote:

This is already covered by existing driver tests. `cuda-detect.cu` checks that a synthetic sysroot’s C++ include paths are passed to both the host and device compilations, and `hip-include-path.hip` checks that the host system include path is passed to the AMDGPU device compilation. These tests do not depend on the build machine’s installed C++ library.

https://github.com/llvm/llvm-project/pull/192821