Antoine Pitrou
Antoine Pitrou
@jorisvandenbossche Can you give this a final review?
Why only in some tests?
@aucahuasi Did you validate that it fixes the Valgrind issues?
PR looks fine a priori, but can we try to reduce those build times?
We can defer this to ARROW-18039, but I'd like to make sure this isn't forgotten :-)
@raulcd Perhaps try applying this patch? ```diff diff --git a/python/pyproject.toml b/python/pyproject.toml index edbc4ade6..a799dc761 100644 --- a/python/pyproject.toml +++ b/python/pyproject.toml @@ -18,7 +18,7 @@ [build-system] requires = [ "cython >= 0.29.22", -...
This patch should help fix the 3.11 enum issue: ```diff diff --git a/python/pyarrow/_fs.pyx b/python/pyarrow/_fs.pyx index e7b028a07..557c08149 100644 --- a/python/pyarrow/_fs.pyx +++ b/python/pyarrow/_fs.pyx @@ -78,6 +78,12 @@ cdef CFileType _unwrap_file_type(FileType ty) except...
The tests are trying to compile grpcio, can we avoid that? https://github.com/ursacomputing/crossbow/actions/runs/3330588690/jobs/5509267855#step:11:116 Either install the GCS testbench on a different Python (with binary wheels), or don't test GCS at all...
For the record, I posted https://discuss.python.org/t/dependency-resolution-conflict-on-universal2-with-pip-22-3-and-python-3-11/20419
> So it is here that the deployment target should be changed in case of 3.11: But why would that be necessary only for `universal2`, not `arm64`? Note that Numpy...