prep 3.8.0 release

[aidandj]

The next round of features required a bigger change, and this fixes the mypy memory issues with generics, so I think it makes sense to cut this release.

[aidandj]

Don't merge until this is added in: #717

[nipunn1313]

yep - on board with it being a minor release!

[nipunn1313]

is there anything non-backwards compatible going on that warrants a 4.0? I think no, but you might know better.

[aidandj]

is there anything non-backwards compatible going on that warrants a 4.0? I think no, but you might know better.

Breaking changes with this generated code is interesting. On one hand, if it still passes mypy, and the resolved types stay the same, it's not breaking. On the other hand, I know that people are doing file manipulation after the fact for things like imports, so I made the following decisions.

• Support editions - not breaking
• concrete stubs, sync/async only - options, not breaking
• ArgType aliases, technically changing the generated code, but I don't foresee it as a common code manipulation path, maybe someone was extracting the type literals or something, but seems unlikely
• __new__ overloads - fixing async stubs that were released in 3.7.0 as non breaking
• Drop support for py_generic_services - On the fence about this one. On one hand, the newer compilers already dropped it, but I guess it technically may be considered breaking.

If we were to do a major version, I think I'd also want to include import mangling as seen here and mangle all imports so they don't have a chance to conflict.

Although, I guess nothing is stopping us from releasing 4.0.0 and then 5.0.0. I think I'm leaning towards that now. Will update the PR to 4.0.0

[aidandj]

If someone really needs it I can release a 3.8.0 with just the __new__ overload fix. But I may wait until that is requested.

[nipunn1313]

yeah fine with me!