bug[next]: Fix codegen in gtfn for unused vertical offset provider #1746
Conversation
|
Let's add a test, it's not obvious to me how to trigger that bug. |
| @@ -229,6 +229,7 @@ def simple_mesh() -> MeshDescriptor: | |||
| C2E.value: gtx.NeighborTableOffsetProvider( | |||
| c2e_arr, Cell, Edge, 4, has_skip_values=False | |||
| ), | |||
| Koff.value: KDim, | |||
There was a problem hiding this comment.
Why is this change here? I don't understand how this is related. Is it required to make the test below fail before the fix?
There was a problem hiding this comment.
This is an easy way to trigger the error.
There was a problem hiding this comment.
But not the best I guess. An isolated test would be one that has an Cartesian offset for a dimension that is not in any field?
There was a problem hiding this comment.
any updates on this? I would need it for C2SM/icon4py#564
There was a problem hiding this comment.
@nfarabullini Let's introduce a new unstructured_case_3d fixture that starting from the unstructured_case fixture adds the default size and offset provider for the KDim.
@jcanton I think we can merge this soon, but this will likely not help you much since icon4py is pinned to an older version of GT4PY until #1738 is merged.
tehrengruber
changed the title
tests/next_tests/integration_tests/feature_tests/ffront_tests/test_execution.py
Outdated
Show resolved
Hide resolved
tests/next_tests/integration_tests/feature_tests/ffront_tests/test_execution.py
Outdated
Show resolved
Hide resolved
…test_execution.py Co-authored-by: Till Ehrengruber <[email protected]>
…test_execution.py Co-authored-by: Till Ehrengruber <[email protected]>
…into koff_small_fix
Co-authored-by: Till Ehrengruber <[email protected]>
Co-authored-by: Till Ehrengruber <[email protected]>
Providing an offest provider for a vertical dimension without using that dimension in a program, e.g. no arguments are fields defined on K, resulted in erroneous C++ code.