This one already had a proper explanation why it fails, which is due to
differences by design in MSVC mode. This isn't a fixme, so degrade the
annotation to a more permanent "XFAIL: msvc" instead.
Details
- Reviewers
ldionne - Group Reviewers
Restricted Project - Commits
- rGbb2a92e66e04: [libcxx] [test] Convert an XFAIL: LIBCXX-WINDOWS-FIXME into XFAIL: msvc
Diff Detail
- Repository
- rG LLVM Github Monorepo
Event Timeline
libcxx/test/libcxx/atomics/diagnose_invalid_memory_order.verify.cpp | ||
---|---|---|
9–13 | Should msvc simply undefine diagnose-if-support? $ git grep msvc std/input.output/iostreams.base/ios.base/ios.base.storage/iword.pass.cpp:// UNSUPPORTED: msvc std/input.output/iostreams.base/ios.base/ios.base.storage/pword.pass.cpp:// UNSUPPORTED: msvc std/language.support/cmp/cmp.categories.pre/zero_type.verify.cpp:// UNSUPPORTED: msvc std/thread/thread.mutex/thread.lock.algorithm/lock.pass.cpp:// UNSUPPORTED: libstdc++, msvc std/utilities/meta/meta.trans/meta.trans.other/underlying_type.fail.cpp:// UNSUPPORTED: msvc $ git grep diagnose-if-support libcxx/atomics/diagnose_invalid_memory_order.verify.cpp:// REQUIRES: diagnose-if-support libcxx/containers/associative/non_const_comparator.verify.cpp:// REQUIRES: diagnose-if-support libcxx/containers/unord/non_const_comparator.verify.cpp:// REQUIRES: diagnose-if-support libcxx/containers/unord/unord.set/missing_hash_specialization.fail.cpp:// REQUIRES: diagnose-if-support |
libcxx/test/libcxx/atomics/diagnose_invalid_memory_order.verify.cpp | ||
---|---|---|
9–13 | I don't see the value in doing that - out of those 4 REQUIRES: diagnose-if-support, this is the only one that has problems running with MSVC configurations. And the fact that we only have a handful of tests that check the msvc flag doesn't mean it that flag should be discontinued - we right now still have 66 unsorted LIBXX-WINDOWS-FIXME - out of them, 10 still seem to be MSVC specific that might warrant changing into UNSUPPORTED/XFAIL: msvc or something like that, after analyzing their root causes. |
Should msvc simply undefine diagnose-if-support?
Right now we don't have many tests that check either of these flags: