Hi.
This bug was fixed in trixie/sid, but I still would like to see
it fixed it in bookworm via team upload.
Upstream said (at the time) that "parallel code has been
completely re-written". Backporting new code to stable would
be a very bad idea, but we can still make the package buildable
from source for everybody.
The attached patch uses os.cpu_count() and disables the tests
only when we know they will fail. This is an improvement over
the patch I initially proposed, which I admit was not good,
and at the same time better than nothing.
So, my plan (no action required from your side) is to create
a "bookworm" branch, add the last two commits from my fork:
https://salsa.debian.org/sanvila/reproject/-/commits/bookworm
and take care of the related bureaucracy (hopefully for
Debian 12.11, since 12.10 will be out tomorrow).
(So, similar to some extent to what I did with "ndcube" and "sunpy").
Thanks.
Description: Skip some tests on single-CPU systems
Author: Santiago Vila <sanv...@debian.org>
Bug-Debian: https://bugs.debian.org/1031194
Forwarded: https://github.com/astropy/reproject/issues/337
--- a/reproject/spherical_intersect/tests/test_reproject.py
+++ b/reproject/spherical_intersect/tests/test_reproject.py
@@ -2,6 +2,7 @@
import numpy as np
import pytest
+import os
from astropy.io import fits
from astropy.utils.data import get_pkg_data_filename
from astropy.wcs import WCS
@@ -10,6 +11,7 @@
from ..core import _reproject_celestial
+@pytest.mark.skipif(os.cpu_count() == 1, reason="Known not to work on single-CPU systems")
def test_reproject_celestial_slices_2d():
header_in = fits.Header.fromtextfile(get_pkg_data_filename("../../tests/data/gc_ga.hdr"))
@@ -68,6 +70,7 @@
)
+@pytest.mark.skipif(os.cpu_count() == 1, reason="Known not to work on single-CPU systems")
def test_reproject_celestial_consistency():
# Consistency between the different modes
