[Driver] Fix Arm/AArch64 Link Argument tests

[lenary]

The openmp-offload-amdgpu-runtime-2 bot specifies default rtlib of compiler-rt, but default unwindlib of libgcc. Change the tests to accept that there may be "--as-needed" "-lgcc_s" "--no-as-needed" between libclang_rt.builtins.a and -lc.

Relates to #121830

[llvmbot]

@llvm/pr-subscribers-clang-driver

@llvm/pr-subscribers-clang

Author: Sam Elliott (lenary)

Changes

The openmp-offload-amdgpu-runtime-2 bot specifies default rtlib of compiler-rt, but default unwindlib of libgcc. Change the tests to accept that there may be "--as-needed" "-lgcc_s" "--no-as-needed" between libclang_rt.builtins.a and -lc.

Relates to #121830

---

Full diff: https://github.com/llvm/llvm-project/pull/144582.diff

2 Files Affected:

• (modified) clang/test/Driver/aarch64-toolchain.c (+2-1)
• (modified) clang/test/Driver/arm-toolchain.c (+2-1)

diff --git a/clang/test/Driver/aarch64-toolchain.c b/clang/test/Driver/aarch64-toolchain.c
index e12107fa2c506..327161b81d9f6 100644
--- a/clang/test/Driver/aarch64-toolchain.c
+++ b/clang/test/Driver/aarch64-toolchain.c
@@ -135,7 +135,8 @@
 
 // AARCH64-BAREMETAL-COMPILER-RT: "{{.*}}crt0.o"
 // AARCH64-BAREMETAL-COMPILER-RT: "{{.*}}clang_rt.crtbegin.o"
-// AARCH64-BAREMETAL-COMPILER-RT: "--start-group" "{{.*}}libclang_rt.builtins.a" "-lc" "-lgloss" "--end-group"
+// AARCH64-BAREMETAL-COMPILER-RT: "--start-group" "{{.*}}libclang_rt.builtins.a"
+// AARCH64-BAREMETAL-COMPILER-RT: "-lc" "-lgloss" "--end-group"
 // AARCH64-BAREMETAL-COMPILER-RT: "{{.*}}clang_rt.crtend.o"
 
 // RUN: %clang -### %s -fuse-ld= \
diff --git a/clang/test/Driver/arm-toolchain.c b/clang/test/Driver/arm-toolchain.c
index d4f9bf2aaf3d5..5368158cdeeda 100644
--- a/clang/test/Driver/arm-toolchain.c
+++ b/clang/test/Driver/arm-toolchain.c
@@ -136,7 +136,8 @@
 
 // ARM-BAREMETAL-COMPILER-RT: "{{.*}}crt0.o"
 // ARM-BAREMETAL-COMPILER-RT: "{{.*}}clang_rt.crtbegin.o"
-// ARM-BAREMETAL-COMPILER-RT: "--start-group" "{{.*}}libclang_rt.builtins.a" "-lc" "-lgloss" "--end-group"
+// ARM-BAREMETAL-COMPILER-RT: "--start-group" "{{.*}}libclang_rt.builtins.a"
+// ARM-BAREMETAL-COMPILER-RT: "-lc" "-lgloss" "--end-group"
 // ARM-BAREMETAL-COMPILER-RT: "{{.*}}clang_rt.crtend.o"
 
 // RUN: %clang -### %s -fuse-ld= \

[quic-garvgupt]

`ToolChain::UnwindLibType ToolChain::GetUnwindLibType(
const ArgList &Args) const {
if (unwindLibType)
return *unwindLibType;

const Arg *A = Args.getLastArg(options::OPT_unwindlib_EQ);
StringRef LibName = A ? A->getValue() : CLANG_DEFAULT_UNWINDLIB;`

If --unwindlib= is not set to empty explicitly, then CLANG_DEFAULT_UNWINDLIB overrides. I think, adding --unwindlib=, might fix the tests.

[quic-garvgupt]

`ToolChain::UnwindLibType ToolChain::GetUnwindLibType( const ArgList &Args) const { if (unwindLibType) return *unwindLibType;

const Arg *A = Args.getLastArg(options::OPT_unwindlib_EQ); StringRef LibName = A ? A->getValue() : CLANG_DEFAULT_UNWINDLIB;`

If --unwindlib= is not set to empty explicitly, then CLANG_DEFAULT_UNWINDLIB overrides. I think, adding --unwindlib=, might fix the tests.

`ToolChain::UnwindLibType ToolChain::GetUnwindLibType( const ArgList &Args) const { if (unwindLibType) return *unwindLibType;

const Arg *A = Args.getLastArg(options::OPT_unwindlib_EQ); StringRef LibName = A ? A->getValue() : CLANG_DEFAULT_UNWINDLIB;`

If --unwindlib= is not set to empty explicitly, then CLANG_DEFAULT_UNWINDLIB overrides. I think, adding --unwindlib=, might fix the tests.

I revisited the tests and realized that for the ARM and RISC-V targets, the flow goes through GetDefaultUnwindLibType, which defaults to UNW_None. Explicitly adding --unwindlib= to the failing instance would make the test redundant, as it would no longer validate the intended behavior. Please disregard my earlier comment—I'm aligned with this fix. Thanks for pushing the change!

[Kewen12]

Thanks for this quick fix!

Hi @quic-garvgupt, could you please help approve this PR if it looks good to you. It would be helpful to unblock our bots and downstream. Thanks!

[lenary]

I'm going to land this, as I'm taking "I'm aligned with this fix" as approval from garvit (I know he's in a different time zone, which may mean he's logged off for the night)

[Kewen12]

Thanks again for your help @lenary