@isidentical: Status check is done, and it's a success ✅ .

Thanks @isidentical for the PR 🌮🎉.. I'm working now to backport this PR to: 3.8, 3.9.
🐍🍒⛏🤖 I'm not a witch! I'm not a witch!

GH-20675 is a backport of this pull request to the 3.9 branch.

GH-20676 is a backport of this pull request to the 3.8 branch.

@pablogsal the backports passed tests and are ready to merge, but need coredev approval. I don't know if the automerge label is supposed to be carried back or not, but it was not.

Thanks @terryjreedy for the ping. I have merged the backports :)

@isidentical , @pablogsal: Is it clear that this should be backported to 3.8? This will break some programs out in the wild unexpectedly. While I don't know the discussion leading to this PR (I suppose that ast.Name("None", ast.Load()) could be considered invalid usage in the first place), I wouldn't have expected such a change without a major version bump.

In particular, I encountered this for https://github.com/beetbox/beets/ which uses ast to implement a simple templating engine.

Is it clear that this should be backported to 3.8?

Yes, is a bug. ast.Name("None") is invalid in Load/Store contexts because those cannot be variable names you can load or assign to. In Python3.8 the singletons None, True and False are ast.Constant so trying to use them as names is invalid.

I wouldn't have expected such a change without a major version bump.

Although we do our best to reduce breakage, the ast module reflects the internal CPython AST and that can change between major version bumps (from 3.7 to 3.8 for example) as if this was not possible, it would prevent us to optimize the CPython VM.