got a report on "python-nanobind fails test in 1-core VM"

[hsk17]

dear wenzel jacob,

i have packaged nanobind for opensuse and it is now contained in tumbleweed (the rolling release), and hopefully will be in the upcoming leap 16 (the stable release).

now i got a bug report that "test fails in 1-core vm", please see https://bugzilla.suse.com/show_bug.cgi?id=1239153:
++++++
Steps to reproduce:
osc checkout openSUSE:Factory/python-nanobind && cd $_
osc build --vm-type=kvm --noservice --clean -j1 standard

actual results:
[ 369s] 'please double-check that the change makes sense.',
[ 369s] file=sys.stderr
[ 369s] )
[ 369s] > assert False
[ 369s] E assert False
[ 369s]
[ 369s] tests/test_stubs.py:60: AssertionError
[ 369s] ----------------------------- Captured stderr call -----------------------------
[ 369s] --- /home/abuild/rpmbuild/BUILD/build_python3.12/build/tests/test_typing_ext.pyi.ref
[ 369s] +++ /home/abuild/rpmbuild/BUILD/build_python3.12/build/tests/test_typing_ext.pyi
[ 369s] @@ -1,6 +1,7 @@
[ 369s] from collections.abc import Iterable
[ 369s] -import py_stub_test
[ 369s] from typing import Generic, Optional, Self, TypeAlias, TypeVar
[ 369s] +
[ 369s] +import py_stub_test.AClass
[ 369s]
[ 369s] from . import submodule as submodule
[ 369s] from .submodule import F as F, f as f2
[ 369s]
[ 369s] Warning: generated stubs do not match their references. If you
[ 369s] intentionally changed a test suite extension, it may be necessary
[ 369s] to replace the .pyi.ref file with the generated .pyi file. But
[ 369s] please double-check that the change makes sense.
[ 369s] =========================== short test summary info ============================
[ 369s] FAILED tests/test_stubs.py::test01_check_stub_refs[p_ref6] - assert False
[ 369s] ================== 1 failed, 434 passed, 22 skipped in 5.14s ===================
[ 369s] error: Bad exit status from /var/tmp/rpm-tmp.pIPzuZ (%check)

Please work with upstream on a fix for this (it is also an option to skip the test, if it is not meant to work in 1-core-VMs)
++++++

i never observed such failure, neither on suse's openbuild service nor on my own test machines. but indeed: reconfigured one of my virtual machines to have only 1 cpu, ran build-and-test of nanobind on that machine, and test_stubs.py failed as described in the bug report. looking at the last sentence of the bug report, i feel this just a matter of skipping the test in the .spec file for this special case - do you have an opinion on this?

kind regards,
fritz

[hsk17]

could do some further investigation and opened bug issue github.com//issues/970