ci: Fail CI if xfails pass unexpectedly (#1888)

Be strict about tests marked with xfail. If they pass unexpectedly, fail the testsuite instead of logging an XPASS.

pyproject.toml

@@ -319,6 +319,7 @@ markers = [
   'uses_sparse_fields_as_output: tests that require backend support for writing sparse fields',
   'uses_strided_neighbor_offset: tests that require backend support for strided neighbor offset',
   'uses_tuple_args: tests that require backend support for tuple arguments',
+  'uses_tuple_args_with_different_but_promotable_dims: test that requires backend support for tuple args with different but promotable dims',
   'uses_tuple_iterator: tests that require backend support to deref tuple iterators',
   'uses_tuple_returns: tests that require backend support for tuple results',
   'uses_zero_dimensional_fields: tests that require backend support for zero-dimensional fields',
@@ -330,6 +331,7 @@ markers = [
 ]
 norecursedirs = ['dist', 'build', 'cpp_backend_tests/build*', '_local/*', '.*']
 testpaths = 'tests'
+xfail_strict = true
 
 # -- ruff --
 [tool.ruff]

tests/next_tests/definitions.py

@@ -116,6 +116,9 @@ class ProgramFormatterId(_PythonObjectIdMixin, str, enum.Enum):
 USES_REDUCTION_WITH_ONLY_SPARSE_FIELDS = "uses_reduction_with_only_sparse_fields"
 USES_STRIDED_NEIGHBOR_OFFSET = "uses_strided_neighbor_offset"
 USES_TUPLE_ARGS = "uses_tuple_args"
+USES_TUPLES_ARGS_WITH_DIFFERENT_BUT_PROMOTABLE_DIMS = (
+    "uses_tuple_args_with_different_but_promotable_dims"
+)
 USES_TUPLE_ITERATOR = "uses_tuple_iterator"
 USES_TUPLE_RETURNS = "uses_tuple_returns"
 USES_ZERO_DIMENSIONAL_FIELDS = "uses_zero_dimensional_fields"
@@ -139,6 +142,7 @@ class ProgramFormatterId(_PythonObjectIdMixin, str, enum.Enum):
     (USES_NEGATIVE_MODULO, XFAIL, UNSUPPORTED_MESSAGE),
     (USES_REDUCTION_WITH_ONLY_SPARSE_FIELDS, XFAIL, REDUCTION_WITH_ONLY_SPARSE_FIELDS_MESSAGE),
     (USES_SPARSE_FIELDS_AS_OUTPUT, XFAIL, UNSUPPORTED_MESSAGE),
+    (USES_TUPLES_ARGS_WITH_DIFFERENT_BUT_PROMOTABLE_DIMS, XFAIL, UNSUPPORTED_MESSAGE),
 ]
 # Markers to skip because of missing features in the domain inference
 DOMAIN_INFERENCE_SKIP_LIST = [
@@ -168,6 +172,7 @@ class ProgramFormatterId(_PythonObjectIdMixin, str, enum.Enum):
 ]
 ROUNDTRIP_SKIP_LIST = DOMAIN_INFERENCE_SKIP_LIST + [
     (USES_SPARSE_FIELDS_AS_OUTPUT, XFAIL, UNSUPPORTED_MESSAGE),
+    (USES_TUPLES_ARGS_WITH_DIFFERENT_BUT_PROMOTABLE_DIMS, XFAIL, UNSUPPORTED_MESSAGE),
 ]
 GTFN_SKIP_TEST_LIST = (
     COMMON_SKIP_TEST_LIST

tests/next_tests/integration_tests/feature_tests/ffront_tests/test_execution.py

@@ -262,9 +262,7 @@ def testee(a: tuple[int32, tuple[int32, cases.IField, int32]]) -> cases.IField:
 
 
 @pytest.mark.uses_tuple_args
-@pytest.mark.xfail(
-    reason="Not implemented in frontend (implicit size arg handling needs to be adopted) and GTIR embedded backend."
-)
+@pytest.mark.uses_tuple_args_with_different_but_promotable_dims
 def test_tuple_arg_with_different_but_promotable_dims(cartesian_case):
     @gtx.field_operator
     def testee(a: tuple[cases.IField, cases.IJField]) -> cases.IJField: