[CodeGenCXX] XFAIL test for ASAN on Darwin.

[vsapsai]

Summary

The test hits stack overflow trying to instantiate recursive templates.
It is observed with ASAN and not with a regular build because ASAN
increases stack frame size.

rdar://problem/45009892

[george.karpenkov]

clang/test/CodeGenCXX/castexpr-basepathsize-threshold.cpp

8

Do we actually want UNSUPPORTED here? We don't want to fail if ASAN stack usage decreases?

[vsapsai]

clang/test/CodeGenCXX/castexpr-basepathsize-threshold.cpp

8

If ASAN stack usage decreases or template instantiation stack usage decreases, I'd like to know that and to remove XFAIL. My reason to prefer XFAIL over UNSUPPORTED is that currently the test fails due to specific implementation of Clang and ASAN, not due to conceptual incompatibility. But I don't have any evidence to show that my suggestion is actually better, so if my argument doesn't look convincing, most likely UNSUPPORTED would be better.

[george.karpenkov]

clang/test/CodeGenCXX/castexpr-basepathsize-threshold.cpp

8

Either one works for me.

clang/test/CodeGenCXX/castexpr-basepathsize-threshold.cpp

 // RUN: %clang_cc1 %s -emit-llvm-only -o -
 
 // https://bugs.llvm.org/show_bug.cgi?id=38356
 // We only check that we do not crash.
 
+// ASAN increases stack usage, so we are hitting stack overflow before reaching
+// recursive template instantiation limit.
+// XFAIL: darwin && asan
+
 template <typename a, a b(unsigned), int c, unsigned...>
 struct d : d<a, b, c - 1> {};
 template <typename a, a b(unsigned), unsigned... e>
 struct d<a, b, 0, e...> {
   a f[0];
 };
 struct g {
   static g h(unsigned);