Repository navigation
GH-39848: [Python][Packaging] Build pyarrow wheels with numpy RC instead of nightly - #41097
Conversation
|
|
github-actions
Bot
added
Component: Python
awaiting committer review
|
@github-actions crossbow submit -g wheel |
This comment was marked as outdated.
This comment was marked as outdated.
|
Ah, we are also running into the liblzma vcpkg issue now (microsoft/vcpkg#37839 / microsoft/vcpkg#37957), so this is blocked by #41100 |
e257d49 to
16fd1ab
Compare
|
@github-actions crossbow submit -g wheel |
|
Rebased for #41091. |
|
Revision: 16fd1abcadc2be034141b72807b47bb7f77463ae Submitted crossbow builds: ursacomputing/crossbow @ actions-506eae40cb
|
16fd1ab to
e257d49
Compare
|
@github-actions crossbow submit -g wheel |
|
Revision: 74e0a93 Submitted crossbow builds: ursacomputing/crossbow @ actions-faa2d90c72
|
|
The macOS wheels failure is the same issue that happens on test-build-vcpkg-win. I am not sure why is it failing on installing grpc. See comment here: #41124 (comment) |
|
Wheels have been fixed on main, can you rebase to have green CI before merge? |
|
@github-actions crossbow submit -g wheel |
|
Revision: f579e27 Submitted crossbow builds: ursacomputing/crossbow @ actions-4606dc8ba2
|
raulcd
left a comment
raulcd
left a comment
There was a problem hiding this comment.
Java JNI / AMD64 manylinux2014 Java JNI is timing out after 90 minutes because is rebuilding the docker manylinux image. I am going to merge because it has nothing to do with this PR. The timeout was modified from 500 minutes to 90 minutes on this PR: https://github.com/apache/arrow/pull/40609/files#diff-d4336cc8304d9967e1e645bb86368d266ff4b96503aaea4aab2ded3bb398cd8e
We might have to increase it again.
github-actions
Bot
added
awaiting merge
|
@jorisvandenbossche Should we create an issue to update the Numpy version from RC1 to the final released once numpy releases? |
…ead of nightly (#41097) ### Rationale for this change Now NumPy has released a first RC for 2.0, we should update our PyArrow wheels to build with this released version instead of with the nightly numpy packages, to ensure we don't build our release wheels with an unstable numpy version. ### What changes are included in this PR? Increased the version requirement for numpy for the installed packages at build time to `numpy>=2.0.0rc1`, to force installing this RC instead of numpy 1.26 ### Are these changes tested? The wheel tests ensure that those wheels still work with older versions of numpy * GitHub Issue: #39848 Authored-by: Joris Van den Bossche <[email protected]> Signed-off-by: Raúl Cumplido <[email protected]>
|
Checked a subset of the builds in the logs to verify they indeed build with numpy 2.0, on CI we test the wheels runs with older numpy, and locally I tested one wheel that it imports fine with numpy 2.0.
Yep, good idea, will do -> #41203 |
|
After merging your PR, Conbench analyzed the 7 benchmarking runs that have been run so far on merge-commit 0b33c65. There were no benchmark performance regressions. 🎉 The full Conbench report has more details. It also includes information about 4 possible false positives for unstable benchmarks that are known to sometimes produce them. |
…C instead of nightly (apache#41097) ### Rationale for this change Now NumPy has released a first RC for 2.0, we should update our PyArrow wheels to build with this released version instead of with the nightly numpy packages, to ensure we don't build our release wheels with an unstable numpy version. ### What changes are included in this PR? Increased the version requirement for numpy for the installed packages at build time to `numpy>=2.0.0rc1`, to force installing this RC instead of numpy 1.26 ### Are these changes tested? The wheel tests ensure that those wheels still work with older versions of numpy * GitHub Issue: apache#39848 Authored-by: Joris Van den Bossche <[email protected]> Signed-off-by: Raúl Cumplido <[email protected]>
3 participants
Rationale for this change
Now NumPy has released a first RC for 2.0, we should update our PyArrow wheels to build with this released version instead of with the nightly numpy packages, to ensure we don't build our release wheels with an unstable numpy version.
What changes are included in this PR?
Increased the version requirement for numpy for the installed packages at build time to
numpy>=2.0.0rc1, to force installing this RC instead of numpy 1.26Are these changes tested?
The wheel tests ensure that those wheels still work with older versions of numpy