GeezerVison

test_tools and test_venv fail because of pip: pypa/pip#11103

By default, make install fails on ensurepip, also because of pip.

Temporary workaround: ./configure --with-ensurepip=no: make install doesn't run ensurepip and so doesn't fail.

For now, this PR is a draft, and so I didn't write any documentation (no NEWS entry, no What's New in Python 3.12, etc.).

Docs / Docs (pull_request) Failing

make -C Doc/ PYTHON=../python venv fails because of pip.

Related issue: #92584

@erlend-aasland and I will have the last modules ported to Modules/Setup.stdlib very soon. We won't need distutils any more then. I recommend that we do not rename distutils and instead remove it once and for all.

@tiran:

@erlend-aasland and I will have the last modules ported to Modules/Setup.stdlib very soon. We won't need distutils any more then. I recommend that we do not rename distutils and instead remove it once and for all.

See the issue #92584: peg_generator, c-analyzer tool and test_cppext still use distutils. I would prefer to not break them.

See the issue #92584: peg_generator, c-analyzer tool and test_cppext still use distutils. I would prefer to not break them.

I modified test_cppext to build the C++ extension with setuptools. So far, so good: it works on all buildbots.

We can try a similar approach for test_peg_generator. Right now, test_peg_generator builds a library to make the test faster. I don't know how to port distutils code to setuptools.

@tiran:

@erlend-aasland and I will have the last modules ported to Modules/Setup.stdlib very soon.

How is it going? Is this work tracked somewhere?

For example, configure.ac mentions the array module, but setup.py still has code to build this extension. Is there a plan to remove the setup.py code, if the configure/Makefile code is enough? Or is there still a reason to keep the setup.py code?

I have created GH-93939

tl;dr every module that uses addext() can be built without distutils (except for a quirk in multiprocessing). Modules that use add() are not yet ported. I have an open PR for _ctypes.

It seems like most core devs don't want to keep distutils hidden in the stdlib, but want to directly remove it. So I close my PR. Remaining steps are tracked by other issues.