[libcxx][tests] Continue marking tests XFAIL in clang-22

[ilovepi]

These tests still fail on Windows with clang-22, as reported in
#70225 (comment).
The root cause seems to be the version bump in
https://llvm.googlesource.com/llvm-project/+/01f36b39bd2475a271bbeb95fb9db8ed65e2d065

Bot: https://ci.chromium.org/ui/p/fuchsia/builders/toolchain.ci/clang-windows-x64/b8709267330516788049/overview
Logs: https://logs.chromium.org/logs/fuchsia/buildbucket/cr-buildbucket/8709267330516788049/+/u/clang/test/stdout

[ilovepi]

• [libcxx][tests] Continue marking tests XFAIL in clang-22 #149124 👈 (View in Graphite)
• main

This stack of pull requests is managed by Graphite. Learn more about stacking.

[llvmbot]

@llvm/pr-subscribers-libcxx

Author: Paul Kirth (ilovepi)

Changes

These tests still fail on Windows with clang-22, as reported in
#70225 (comment).
The root cause seems to be the version bump in
https://llvm.googlesource.com/llvm-project/+/01f36b39bd2475a271bbeb95fb9db8ed65e2d065

Bot: https://ci.chromium.org/ui/p/fuchsia/builders/toolchain.ci/clang-windows-x64/b8709267330516788049/overview
Logs: https://logs.chromium.org/logs/fuchsia/buildbucket/cr-buildbucket/8709267330516788049/+/u/clang/test/stdout

---

Full diff: https://github.com/llvm/llvm-project/pull/149124.diff

3 Files Affected:

• (modified) libcxx/test/libcxx/fuzzing/random.pass.cpp (+1-1)
• (modified) libcxx/test/std/depr/depr.c.headers/math_h.pass.cpp (+1-1)
• (modified) libcxx/test/std/numerics/c.math/cmath.pass.cpp (+1-1)

diff --git a/libcxx/test/libcxx/fuzzing/random.pass.cpp b/libcxx/test/libcxx/fuzzing/random.pass.cpp
index cb074bd60fdc8..346ff47d9bb36 100644
--- a/libcxx/test/libcxx/fuzzing/random.pass.cpp
+++ b/libcxx/test/libcxx/fuzzing/random.pass.cpp
@@ -8,7 +8,7 @@
 
 // This test fails because Clang no longer enables -fdelayed-template-parsing
 // by default on Windows with C++20 (#69431).
-// XFAIL: msvc && (clang-18 || clang-19 || clang-20 || clang-21)
+// XFAIL: msvc && (clang-18 || clang-19 || clang-20 || clang-21 || clang-22)
 
 // UNSUPPORTED: c++03, c++11
 
diff --git a/libcxx/test/std/depr/depr.c.headers/math_h.pass.cpp b/libcxx/test/std/depr/depr.c.headers/math_h.pass.cpp
index 1ba0063c1dada..a6fdea524fb83 100644
--- a/libcxx/test/std/depr/depr.c.headers/math_h.pass.cpp
+++ b/libcxx/test/std/depr/depr.c.headers/math_h.pass.cpp
@@ -8,7 +8,7 @@
 
 // This test fails because Clang no longer enables -fdelayed-template-parsing
 // by default on Windows with C++20 (#69431).
-// XFAIL: msvc && (clang-18 || clang-19 || clang-20 || clang-21)
+// XFAIL: msvc && (clang-18 || clang-19 || clang-20 || clang-21 || clang-22)
 
 // <math.h>
 
diff --git a/libcxx/test/std/numerics/c.math/cmath.pass.cpp b/libcxx/test/std/numerics/c.math/cmath.pass.cpp
index 48c2918802fc3..4d1e5c329b388 100644
--- a/libcxx/test/std/numerics/c.math/cmath.pass.cpp
+++ b/libcxx/test/std/numerics/c.math/cmath.pass.cpp
@@ -8,7 +8,7 @@
 
 // This test fails because Clang no longer enables -fdelayed-template-parsing
 // by default on Windows with C++20 (#69431).
-// XFAIL: msvc && (clang-18 || clang-19 || clang-20 || clang-21)
+// XFAIL: msvc && (clang-18 || clang-19 || clang-20 || clang-21 || clang-22)
 
 // <cmath>
 

[ldionne]

I would suggest switching to this:

// This test doesn't work on Windows because the Windows headers <EXPLAIN REASON>
// XFAIL: msvc

IIUC, the test basically doesn't work on Windows with any of our supported compilers, so it's not really useful to keep carrying the clang-XY annotations here. By the way, why does the test fail? I mean, what's special about Windows that makes this test require -fdelayed-template-parsing?

[jyknight]

The problem is explained in issue 70225; please add a link to that issue in the comment here.

[ilovepi]

Done.

[ilovepi]

oops only fixed one file. will address.

[ilovepi]

Actually done this time :)

[frederick-vs-ja]

FYI, I'm working on a "proper" workaround since C++20, see #149234.

[boomanaiden154]

(Sorry, accidentally hit the update branch button on my phone).

[ilovepi]

FYI, I'm working on a "proper" workaround since C++20, see #149234.

That's fine, but I'd like to unblock our CI ASAP. We haven't been able to qualify a toolchain since the version bump, and we're loosing a lot of important signal. If this can land sooner I'd prefer to do that, and then there isn't any urgency to land a proper fix quickly.

[ilovepi]

All tests were passing, but I reworded the comments.

[ilovepi]

@ldionne anything left to address?