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:

• Failed to even lint the recipe, probably because of a conda-smithy bug 😢. This likely indicates a problem in your meta.yaml, though. To get a traceback to help figure out what's going on, install conda-smithy and run conda smithy recipe-lint . from the recipe directory.

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.

I do have some suggestions for making it better though...

For recipe:

• License is not an SPDX identifier (or a custom LicenseRef) nor an SPDX license expression.

-1 on this. I know that there are some downstream software which use sys.version to detect a conda installed python.

Patching sys.version to add brand the banner is hacky at best (and the fact that platform.py much be patched as well shows that it has side effects at least). Issue #330 shows that has negative consequences.

I really prefer this approach over making the regex more tolerant.

Playing devil's advocate here but given

our PyPI manylinux wheel is statically linked with libpython.a. This causes the banner to not be the one from conda-forge an to cause a mismatch with conda-forge's patched validation regex.

Isn't this the thing we should fix here first. Then branding or no branding is untangled from your issue.

Agree with Ray 🙂

Playing devil's advocate here but given

our PyPI manylinux wheel is statically linked with libpython.a. This causes the banner to not be the one from conda-forge an to cause a mismatch with conda-forge's patched validation regex.

Isn't this the thing we should fix here first. Then branding or no branding is untangled from your issue.

@mingwandroid @jakirkham Quite the contrary. The conda package for xeus-python is dynamically linked with libpython.so (and we prefer it so), but in the case of a wheel, some distributions of Python don't come with a shared object - and we must link statically. This is how several other packages operate to distribute PyPI wheels.

My point is that conda-forge's Python brands the Python interpreter's banner by changing the sys.version string, which has several side effects as the string is validated at runtime. I don't think this is a good thing to do at all.

I must not have made myself clear here.

That branding works with the dylib but not the static archive seems to me a serious bug and entirely orthogonal to whether we should brand or not.

Now if everyone was in agreement to ditch branding then OK, let's do that, delete all the changes and move on.

Bur that's not the situation and if we did do that we'd need a deprecation cycle for it.

I'm happy to look into using the anaconda branding code for conda forge, which I believe works in both cases. In my opinion this is the best path forward.

That branding works with the dylib but not the static archive seems to me a serious bug and entirely orthogonal to whether we should brand or not.

Actually, there is no "bug", since the manylinux wheel is linked against the libpython.a from manylinux and not conda, which is why it does not get the branding.

we'd need a deprecation cycle for it.

I think we can start with #331 which makes the regex compatible with both the branding and the absence thereof, and discuss the complete removal at greater length.

What manylinux libpython.a do you mean? AFAIK this doesn't exist. You are linking against some random libpython the distribution you ran your build upon happened to have installed I guess?

I don't want to spend time taking about random software.

And is it true that you are using this random python but with a python syslib from conda-forge?

You are linking against some random libpython the distribution you ran your build upon happened to have installed I guess?

No, we are linking against the right libpython.a, but preventing the manylinux docker file from deleting it. This is something several other packages are doing to embed python, such as Panda3D.