[libc++] Use -fmodules-validate-system-headers in the test suite

[ldionne]

Summary

As discussed in https://reviews.llvm.org/D91311#2431363, this is the
correct fix for making sure that Clang invalidates the cache when system
headers change.

[ldionne]

This doesn't seem to have fixed the issue, since one of the builds is failing with a pattern similar to what I described: https://buildkite.com/llvm-project/libcxx-ci/builds/641#3b5d04ef-40fa-4cb0-b22a-0b1e11c41b0f

While building module 'std' imported from <...>/llvm-project/libcxx-ci/libcxx/test/libcxx/modules/stds_include.sh.cpp:34:
In file included from <module-includes>:31:
In file included from <...>/llvm-project/libcxx-ci/libcxx/include/complex.h:28:
In file included from <...>/llvm-project/libcxx-ci/libcxx/include/ccomplex:20:
In file included from <...>/llvm-project/libcxx-ci/libcxx/include/complex:242:
In file included from <...>/llvm-project/libcxx-ci/libcxx/include/sstream:173:
In file included from <...>/llvm-project/libcxx-ci/libcxx/include/ostream:137:
In file included from <...>/llvm-project/libcxx-ci/libcxx/include/ios:215:
In file included from <...>/llvm-project/libcxx-ci/libcxx/include/__locale:15:
<...>/llvm-project/libcxx-ci/libcxx/include/string:961:5: error: unknown type name '_LIBCPP_DEPRECATED_IN_CXX20'
    _LIBCPP_DEPRECATED_IN_CXX20 _LIBCPP_INLINE_VISIBILITY
    ^

libcxx/utils/libcxx/test/features.py

DEFAULT_FEATURES = [
   Feature(name='libcpp-no-if-constexpr',        when=lambda cfg: '__cpp_if_constexpr' not in featureTestMacros(cfg)),
   Feature(name='libcpp-no-structured-bindings', when=lambda cfg: '__cpp_structured_bindings' not in featureTestMacros(cfg)),
   Feature(name='libcpp-no-deduction-guides',    when=lambda cfg: featureTestMacros(cfg).get('__cpp_deduction_guides', 0) < 201611),
   Feature(name='libcpp-no-concepts',            when=lambda cfg: featureTestMacros(cfg).get('__cpp_concepts', 0) < 201811),
   Feature(name='has-fobjc-arc',                 when=lambda cfg: hasCompileFlag(cfg, '-xobjective-c++ -fobjc-arc') and
                                                                  sys.platform.lower().strip() == 'darwin'), # TODO: this doesn't handle cross-compiling to Apple platforms.
   Feature(name='objective-c++',                 when=lambda cfg: hasCompileFlag(cfg, '-xobjective-c++ -fobjc-arc')),
 
-  # Note: We use a custom modules cache path to make sure that we don't reuse
-  #       the default one, which can be shared across builds. This is important
-  #       because we define macros in headers files, and a change in these macros
-  #       doesn't seem to invalidate modules cache entries, which means we could
-  #       build against now-invalid cached headers from a previous build.
+  # NOTE: We make sure that Clang considers system headers for invalidating the
+  # modules cache, otherwise we could be using stale versions of the headers.
   Feature(name='modules-support',
           when=lambda cfg: hasCompileFlag(cfg, '-fmodules'),
-          actions=lambda cfg: [AddCompileFlag('-fmodules-cache-path=%t/ModuleCache')]),
+          actions=lambda cfg: [AddCompileFlag('-fmodules-validate-system-headers')]),
 
   Feature(name='non-lockfree-atomics',
           when=lambda cfg: sourceBuilds(cfg, """
             #include <atomic>
             struct Large { int storage[100]; };
             std::atomic<Large> x;
             int main(int, char**) { return x.load(), x.is_lock_free(); }
           """)),