add patches for glibc-detection, numpy segfault on win, get_python_lib

[h-vetinari]

Fixing https://foss.heptapod.net/pypy/pypy/-/issues/3502 as suggested by @mattip:

PyPy issue is fixed. Is it worth making a patch and releasing a new pypy7.3.5 build? The next PyPy release will probably be a few months coming, and I don't know how common it is to use PyPy + centos7

[conda-forge-linter]

Hi! This is the friendly automated conda-forge-linting service.

I just wanted to let you know that I linted all conda-recipes in your PR (recipe) and found it was in an excellent condition.

[mattip]

While the patch fixes the os.confstr incompatibility, I don't know if the cost to users and/or conda-forge to update the build number for this is worth it. Does it dsrupt anything to accept it?

[h-vetinari]

I don't know if the cost to users and/or conda-forge to update the build number for this is worth it. Does it dsrupt anything to accept it?

It's a normal part of the process. It means people using pypy download the new package when they update their environment (creating some more CDN traffic), but otherwise, the infrastructural costs are negligible (still takes reviewer time, of course, but you've done >90% of the work there by implementing the fix and approving it 🙃).

And of course, it helps with not having intermittent numpy build failures (like in conda-forge/numpy-feedstock#238), or someone wanting to run the numpy test suite on an AVX512 machine, etc. 😉

[mattip]

run the numpy test suite on an AVX512 machine

... but still stuck with CentOS6 which went EOL 7 months ago

[mattip]

I am working on more fixes for PyPy + Windows + NumPy, so maybe let's wait a day or two to see if I can add another patch here

[conda-forge-linter]

Hi! This is the friendly automated conda-forge-linting service.

I wanted to let you know that I linted all conda-recipes in your PR (recipe) and found some lint.

Here's what I've got...

For recipe:

• Selectors are suggested to take a <two spaces>#<one space>[<expression>] form. See lines [25]

[mattip]

I pushed some more fixes:

• renamed the patch @h-vetinari added since it will be part of pypyv7.3.6
• added another patch that is also part of v7.3.6. It seems to solve the segfault on win64 when testing numpy
• added a patch to fix the schema used in get_python_lib from distutils.sysconfig to make it match CPython's ersion. I limited it to windows, but I think the schema changes in both the distutil* patches should also apply to linux. @isuruf thoughts?

[conda-forge-linter]

Hi! This is the friendly automated conda-forge-linting service.

I just wanted to let you know that I linted all conda-recipes in your PR (recipe) and found it was in an excellent condition.

[mattip]

The get_python_libs function is not commonly uses. NumPy uses it in runtests.py

[h-vetinari]

Thanks Matti, LGTM! :)

[h-vetinari]

@isuruf, could you take a look here please? I'd love to unblock the win-pypy numpy build, which will likely still take another iteration after this.

[h-vetinari]

Thanks! :)