CurtHagenlocher opened a new issue, #54489:
URL: https://github.com/apache/arrow/issues/54489
### Describe the bug, including details regarding any error messages,
version, and platform.
Building the `conda-integration` image from scratch now gives Python 3.15.0.
The integration run then fails in `ci/scripts/integration_arrow.sh` at `pip
install jpype1`. jpype1 1.7.1 has no cp315 wheel on PyPI (its newest wheels are
cp314), so pip builds it from source, and the build needs Apache Ant, which the
image doesn't include:
```
+ pip install jpype1
Collecting jpype1
Downloading jpype1-1.7.1.tar.gz (783 kB)
...
CMake Error at CMakeLists.txt:43 (find_program):
Could not find ANT_EXECUTABLE using the following names: ant
...
ERROR: Failed building wheel for jpype1
```
Python is no longer pinned in this image (#51099), so it follows
conda-forge's default, which moved from 3.14 to 3.15 within the last week. This
breaks the Integration job in apache/arrow-dotnet, which builds this image from
apache/arrow `main`. That job has failed on every run since Oct 18; the last
successful run, on Oct 12, used Python 3.14 and a prebuilt `jpype1-1.7.1-cp314`
wheel. Example failure:
https://github.com/apache/arrow-dotnet/actions/runs/39147945839/job/117735520032
The integration runs in this repository still pass, presumably because they
reuse cached images built before the switch, so they'll likely hit the same
failure the next time the image is rebuilt.
conda-forge already publishes jpype1 1.7.1 for Python 3.15 on linux-64 and
linux-aarch64. Installing jpype1 with mamba in `conda-integration.dockerfile`,
as `conda-python-jpype.dockerfile` already does, would avoid the source build.
### Component(s)
Continuous Integration, Integration
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
