You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
CI: failure for "Oldest GCC ..." Linux CI job (related to pre-release NumPy) #21957
I noticed the CI failure while reviewing Ralf's PR at gh-21943. It is reproducible locally on Linux with gcc 9.5 and the new NumPy pre-release, 2.2.0rc1. This CI job builds with the latest NumPy pre-release and then switches to 1.23.5 at test time after building.
Reproduced locally via a build command similar to the CI job in question: python -m build --wheel --no-isolation -Ccompile-args=-j32
Failure can be reproduced on main, so not related to Ralf's branch:
<snip>
Program pythran found: YES 0.17.0 0.17.0 (/home/treddy/python_venvs/py_310_scipy/bin/pythran)
Did not find CMake 'cmake'
Found CMake: NO
Run-time dependency xsimd found: NO (tried pkgconfig and cmake)
Run-time dependency threads found: YES
../scipy/meson.build:50:17: ERROR: Command `/home/treddy/python_venvs/py_310_scipy/bin/python -c 'import os
os.chdir(os.path.join("..", "tools"))
import numpy as np
try:
incdir = os.path.relpath(np.get_include())
except Exception:
incdir = np.get_include()
print(incdir)
'` failed with status 1.
A full log can be found at /home/treddy/github_projects/scipy/.mesonpy-dl97fuyc/meson-logs/meson-log.txt
ERROR Backend subprocess exited when trying to invoke build_wheel
changed the title [-]CI: failure for "Oldest GCC ..." Linux CI job[/-][+]CI: failure for "Oldest GCC ..." Linux CI job (related to pre-release NumPy)[/+]on Nov 26, 2024
I bisected NumPy against SciPy main for the failure to build on x86_64 Ubuntu Linux via python -m build --wheel --no-isolation -Ccompile-args=-j32 and found:
b47bd25e3caba52321a6ba420650fc0eff97f473 is the first bad commit
commit b47bd25e3caba52321a6ba420650fc0eff97f473
Author: Clément Robert <[email protected]>
Date: Fri Nov 22 11:56:31 2024 +0100
API: register NEP 35 functions as array_functions (#27490)
Register `like=` functions as array-functions. This is currently decided based on finding the magic `${ARRAY_FUNCTION_LIKE}` string in the documentation.
Close gh-27451
numpy/_core/_add_newdocs.py | 62 ++++++++-----------------------------
numpy/_core/_asarray.py | 4 +--
numpy/_core/function_base.py | 2 ++
numpy/_core/numeric.py | 10 +++---
numpy/_core/overrides.py | 13 ++++----
numpy/_core/tests/test_overrides.py | 24 +++++++++++++-
numpy/lib/_npyio_impl.py | 6 ++--
numpy/lib/_twodim_base_impl.py | 6 ++--
8 files changed, 58 insertions(+), 69 deletions(-)
Looks like the explanation is incredibly simple after all, import numpy as np fails when PYTHONOPTIMIZE=2 for the latest NumPy pre-release... I thought I was just getting tired or something...
Activity
[-]CI: failure for "Oldest GCC ..." Linux CI job[/-][+]CI: failure for "Oldest GCC ..." Linux CI job (related to pre-release NumPy)[/+]tylerjereddy commentedon Nov 26, 2024
The version of the GNU compiler/toolchain is irrelevant to reproduction of the issue.
tylerjereddy commentedon Nov 26, 2024
I bisected NumPy against SciPy
mainfor the failure to build on x86_64 Ubuntu Linux viapython -m build --wheel --no-isolation -Ccompile-args=-j32and found:tylerjereddy commentedon Nov 26, 2024
Reported upstream for clarification.
tylerjereddy commentedon Nov 27, 2024
Looks like the explanation is incredibly simple after all,
import numpy as npfails whenPYTHONOPTIMIZE=2for the latest NumPy pre-release... I thought I was just getting tired or something...lucascolley commentedon Nov 29, 2024
what is the fix needed here @rgommers ?
CI: use stable NumPy for "old build"
CI: use stable NumPy for "old build" (#21975)