[Clang] Fix failing CI with different test case attribute & host triple

[yuxuanchen1997]

As in title, a8f4397 broke CI due to the calling convention not available on certain targets. This patch uses a simpler calling convention and enables the test only when the attribute exists. It's verified that this test crashes the compiler before a8f4397 so it's the same effect as the previous test. Disabling the test on platforms that don't have the calling convention is fine because it's guarding against a frontend bug.

[llvmbot]

@llvm/pr-subscribers-clang

Author: Yuxuan Chen (yuxuanchen1997)

Changes

As in title, a8f4397 broke CI due to the calling convention not available on certain targets. This patch uses a simpler calling convention and enables the test only when the attribute exists. It's verified that this test crashes the compiler before a8f4397 so it's the same effect as the previous test. Disabling the test on platforms that don't have the calling convention is fine because it's guarding against a frontend bug.

---

Full diff: https://github.com/llvm/llvm-project/pull/76863.diff

1 Files Affected:

• (modified) clang/test/SemaCXX/template-instantiation.cpp (+3-1)

diff --git a/clang/test/SemaCXX/template-instantiation.cpp b/clang/test/SemaCXX/template-instantiation.cpp
index e714e070a206f5..a8ecafc401d9d0 100644
--- a/clang/test/SemaCXX/template-instantiation.cpp
+++ b/clang/test/SemaCXX/template-instantiation.cpp
@@ -3,13 +3,15 @@
 
 namespace GH76521 {
 
+#if __has_attribute(preserve_most)
 template <typename T>
 void foo() {
-  auto l = []() __attribute__((pcs("aapcs-vfp"))) {};
+  auto l = []() __attribute__((preserve_most)) {};
 }
 
 void bar() {
   foo<int>();
 }
+#endif
 
 }

[erichkeane]

Is preserve_most enabled for all platforms? It looks like it'll eventually go to the TargetInfo::checkCallingConvention for each target, and I'm pretty sure we don't support it for all targets.

You're likely better off just adding a triple/handful of triples here.

[yuxuanchen1997]

Like this?

[erichkeane]

Looks fine to me, triple of x86_64 + gnu should be enough to make sure this is a supported attribute.