Fix cryptography WithSerialization aliases

[hasier]

Align with changes in pyca/cryptography@6a8c0b5, still present in latest commit in main.

[github-actions]

According to mypy_primer, this change has no effect on the checked open source code. 🤖🎉

[hasier]

@AlexWaygood I see you have been one of the latest contributors to cryptography, could you help me understand why the 3rd party stubtest is failing? I'm not too sure I understand why it's not finding private_bytes and private_numbers, they are indeed part of the interfaces in the latest commit. Does it have to do with the stubtest_allowlist.txt (I'm not sure how this interacts in the tests either tbh 😅 )?

[AlexWaygood]

Hi, @hasier -- thanks for the PR!

I think stubtest is probably failing because the version field in the METADATA.toml file for these stubs is pinned to 3.3:

typeshed/stubs/cryptography/METADATA.toml

Line 1 in 592707b

version = "3.3.*"

. This is because cryptography has shipped with inline type annotations for over a year now, so the stubs typeshed ships are now obsolete -- you should use the type hints provided by the upstream library instead.

Usually in a situation like this, we would have deleted the stubs from typeshed by now. However, in this case things are a little more complicated due to the fact that other stubs packages in typeshed depend on these stubs (#5618).

If you'd like to help out with typeshed, a really great contribution to make would be to put some work into tweaking our testing infrastructure so that we can include non-stub packages as dependencies for typeshed stubs. Once we're able to do that, we'll be able to remove these out-of-date cryptography stubs (#5768).

For now, I think it's probably not worth putting the effort into improving these long-obsolete stubs, sadly :(

[hasier]

Ah I see. I had not realised they already had inline annotations because we are not in the most recent version of cryptography, so some of them were missing 🤦 I'll close this PR then, thanks for explaining @AlexWaygood!

[AlexWaygood]

Ah I see. I had not realised they already had inline annotations because we are not in the most recent version of cryptography, so some of them were missing 🤦 I'll close this PR then, thanks for explaining @AlexWaygood!

No problem — and thanks again for the PR!