[libc-commits] [clang] [libc] [openmp] [Clang] Remove the old LLVM offloading driver entirely (PR #211924)

[Yaxun Liu via libc-commits]

================
@@ -1,35 +0,0 @@
-// RUN: %clang -### --target=x86_64-linux-gnu \
-// RUN:   -x hip --cuda-gpu-arch=gfx803 -flto-partitions=42 \
-// RUN:   --no-offload-new-driver --emit-static-lib -nogpulib \
-// RUN:   -fuse-ld=lld -B%S/Inputs/lld -fgpu-rdc -nogpuinc \
-// RUN:   %S/Inputs/hip_multiple_inputs/a.cu \
-// RUN:   %S/Inputs/hip_multiple_inputs/b.hip \
-// RUN: 2>&1 | FileCheck %s --check-prefix=FIXED-PARTS
-
-// FIXED-PARTS-NOT: "*.llvm-link"
-// FIXED-PARTS-NOT: ".*opt"
-// FIXED-PARTS-NOT: ".*llc"
-// FIXED-PARTS: [[LLD: ".*lld.*"]] {{.*}} "-plugin-opt=-amdgpu-internalize-symbols"
-// FIXED-PARTS-SAME: "--lto-partitions=42"
-// FIXED-PARTS-SAME: "-plugin-opt=mcpu=gfx803"
-// FIXED-PARTS-SAME: "-o" "{{.*out}}" "{{.*o}}"
-
-// RUN: not %clang -### --target=x86_64-linux-gnu \
-// RUN:   -x hip --cuda-gpu-arch=gfx803 -flto-partitions=a \
----------------
yxsamliu wrote:

Maybe keep one negative test for `-flto-partitions` on the new-driver path?

The option is still valid there, and the PR keeps valid-value coverage. It would be good to also keep coverage for bad values like `-flto-partitions=a` or `-flto-partitions=0`.

https://github.com/llvm/llvm-project/pull/211924