Drop support for < net8.0 in MSTest and MTP

[Youssef1313]

As VSTest latest version supports only net8.0+ and net462, I think we should take the same step in MSTest and MTP.

[Evangelink]

You would want this to be part of v4 as it's a technical breaking change.

[Youssef1313]

@Evangelink The reason I'm putting it in 3.10 is that a similar breaking change was taken already in VSTest in a minor release. https://www.nuget.org/packages/Microsoft.NET.Test.Sdk/17.14.0 is net8.0+ while https://www.nuget.org/packages/Microsoft.NET.Test.Sdk/17.13.0 is netcoreapp3.1+.

Also, at least in theory, older TFMs should still work for the case of MSTest and MTP via netstandard2.0. There could be some problems we are not anticipating, but they will probably be fixable?

[Evangelink]

Yes and I am (was) against the change but VSTest is following VS versioning scheme so they don't guarantee any semantic meaning while we do. I think it's important to respect that, I do expect still a lot of teams to use MSTest with non-supported TFMs.

If we have some tests to prove that netstandard2.0 fallback is good enough then it could be fine. I still think it's not respecting semantic versioning and so not ideal.

[nohwnd]

I don't see how netstandard2.0 would produce a runnable exe for MTP, while technically you can run it, it won't work out of the box, and so it will be a breaking change.

On the other hand it seems like self-imposed limitation to have to make big breaking changes in mstest v4 and mtp v2, maybe those can move to "vNext" and we can follow normal semantic versioning with smaller breaking changes in v4, v5 etc.

[Youssef1313]

Moving to v4 then. We will then have to be careful not to update VSTest dependency to 17.14. 😕