Add back noarch condition

[isuruf]

No description provided.

[jakirkham]

Oh is this needed again?

[scopatz]

LGTM!

[isuruf]

See @mbargull's explanation at conda/conda#7713 (comment)

wheel which comes lexicographically after is fine. chardet which comes lexicographically before is not

[jakirkham]

Should we then qualify this with the lexicographic point? Asking as we have made some Conda dependencies noarch: python so this may be confusing to people.

[mbargull]

Oh is this needed again?

conda still doesn't handle noarch: python packages perfectly on Windows.
Ideally it would have some separate conda-python-package-installer (long name to indicate that I didn't think of any actual name) package that would contain the linked console-script .exe* so noarch: python packages won't have that awkward implicit dependency on the base environment's conda.

_{* (and, IMHO, handle the creation of the entrypoints as well as .pyc compilation etc. -- but that's for another discussion)}

[mbargull]

Should we then qualify this with the lexicographic point?

No, I don't think so. The lexicographical thing is only an implementation detail of conda but not something we should rely on.

[jakirkham]

So should we revert any noarch: python conversions made to Conda's dependencies?

[isuruf]

Dependencies like tqdm have been noarch for 3 months, so I don't think we should go back.

[isuruf]

Even defaults ship tqdm as noarch.

[jakirkham]

I guess my point is we have now added guidance here that doesn't align with what we are doing. It would be good to either improve the guidance or align with it. No strong feelings on which, but clarity for others would be good 🙂

[mbargull]

Even defaults ship tqdm as noarch.

Haham, I'm not sure if that was intentional or not..

I can't guarantee that relying on lexicographic order works in every case, though. I'd have to look that up.