From 3a9def86c71a9a62c299ca5af9fefd1fb5806425 Mon Sep 17 00:00:00 2001 From: Evgeni Burovski Date: Sat, 31 Oct 2026 14:50:15 +0100 Subject: [PATCH] TST: skip test_matmul on NumPy 1.22 because `@=` fails Since https://github.com/data-apis/array-api-tests/pull/466 `test_linalg::test_matmul` exercises the in-place matmul operator, `@=`. This is expected by the spec, but is not implemented in numpy 1.22. Since this was since implemented between 1.22 and 1.26, just skip the test under `numpy==1.22`. --- numpy-1-22-xfails.txt | 3 +++ 1 file changed, 3 insertions(+) diff --git a/numpy-1-22-xfails.txt b/numpy-1-22-xfails.txt index 9947db89..5f21b99a 100644 --- a/numpy-1-22-xfails.txt +++ b/numpy-1-22-xfails.txt @@ -50,6 +50,9 @@ array_api_tests/test_creation_functions.py::test_linspace # https://github.com/numpy/numpy/issues/21083 array_api_tests/test_linalg.py::test_vector_norm +# in-place matmuls fail on NumPy 1.22 +array_api_tests/test_linalg.py::test_matmul + # NumPy 1.22 doesn't support NPY_PROMOTION_STATE=weak, so many tests fail with # type promotion issues # NOTE: some of these may not fail until one runs array-api-tests with