AssertionError: DataFrame are different with dask 2024.5.1 and python 3.12

[dschwoerer]

Describe the issue:
Some unit tests fails on dask 2024.5.1

FAILED ../dataframe/io/tests/test_parquet.py::test_append[pyarrow-True] - AssertionError: DataFrame are different
FAILED ../dataframe/io/tests/test_parquet.py::test_append_wo_index[pyarrow-True] - AssertionError: DataFrame are different
FAILED ../dataframe/io/tests/test_parquet.py::test_append_known_divisions_to_unknown_divisions_works[pyarrow] - AssertionError: DataFrame are different
FAILED ../dataframe/io/tests/test_parquet.py::test_custom_filename_works_with_pyarrow_when_append_is_true - AssertionError: DataFrame are different
= 4 failed, 10489 passed, 1628 skipped, 265 xfailed, 282 xpassed, 131 warnings in 545.24s (0:09:05) =

Minimal Complete Verifiable Example:

pytest dask

Anything else we need to know?:
Full log is here

Environment:

• Dask version: 2024.5.1
• Python version: 3.12
• Operating System: fedora
• Install method (conda, pip, source): dnf / source

[fjetter]

Can you provide a little more context to what you are doing?

Our CI is passing fine on python 3.12 and your minimal example is referencing xarray.

[dschwoerer]

I am sorry about the MWE, I meant to say dask rather then xarray. I have fixed it now.

I am trying to update the dask package for fedora. As part of the build process, the test suite is run against the packages present in fedora. This is where the error is raised. I am however not sure whether this is an error in dask, and it's usage of pyarrow, or rather in pyarrow. Any help and hints would be appreciated.

[phofl]

You installed dask-expr in version 1.1.0, we fixed the bug that causes these tests to fail in version 1.1.1, so upgrading dask-expr will work