Add cats-effect to the community build

[vasilmkd]

Cats Effect 2 added under "cats-effect-2".
Cats Effect 3 added under "cats-effect-3".

cc @djspiewak

[dottybot]

Hello, and thank you for opening this PR! 🎉

All contributors have signed the CLA, thank you! ❤️

Have an awesome day! ☀️

[vasilmkd]

The cats effect build failed with the following exception:

sbt.librarymanagement.ResolveException: Error downloading ch.epfl.lamp:dotty-library_sjs1_0.28:0.28.0-bin-SNAPSHOT

[smarter]

I think (publishLocal in `dotty-library-bootstrappedJS`).value needs to be added to prepareCommunityBuild in Build.scala: https://github.com/lampepfl/dotty/blob/c104e9b1b7c028f114f8f3c0331198a9587391b5/project/Build.scala#L1253

[vasilmkd]

That did the trick. Thank you @smarter.

[anatoliykmetyuk]

Thanks for contributing this @vasilmkd! All of the Dotty community projects live in forks under dotty-staging organization – so that we have the capability to fix them fast if they break. I've created such a fork for cats-effects: https://github.com/dotty-staging/cats-effect – could you point the submodule to this fork instead?

[vasilmkd]

@anatoliykmetyuk On it right now.

[anatoliykmetyuk]

Looks like this project fails the CI once in a while:

https://github.com/lampepfl/dotty/runs/1131886366
https://github.com/lampepfl/dotty/runs/1128837504

@vasilmkd @djspiewak do you have an idea why this is happening?

[vasilmkd]

The tests that fail are really hard to properly write and we know them to be slightly non-deterministic and therefore flaky. May I ask what environment these tests run in?

[anatoliykmetyuk]

The CI definition of the community build step is here: https://github.com/lampepfl/dotty/blob/04659bec64d98cc483c9288f91b04a5b274989bb/.github/workflows/ci.yaml#L99
And the Docker container used is defined as follows: https://github.com/lampepfl/dotty-ci/blob/master/dotty-docker/Dockerfile

[anatoliykmetyuk]

Is it possible to disable these tests in the dotty-staging fork so that the entire CI doesn't fail non-deterministically?

[vasilmkd]

I will push a commit to do that in a bit, is that ok?

[vasilmkd]

I opened a PR here.
dotty-staging/cats-effect#1

[vasilmkd]

@anatoliykmetyuk @smarter What does the process for keeping the dotty-staging/cats-effect fork up to date with upstream? For example, we have a fix for the flaky tests pushed to typelevel/cats-effect and we'd like to incorporate them into the community build. How do we do that? Is it maybe possible to restore the git histories to be identical, since they have diverged with the commit that disabled the flaky tests. Thank you.

cc @djspiewak

[smarter]

Is it maybe possible to restore the git histories to be identical, since they have diverged with the commit that disabled the flaky tests.

Yes, but we have to keep the old history in a branch in the fork, so that each dotty commit still points to reachable commits in the submodules.

[vasilmkd]

This might be a stupid question, I just want to be sure I understood correctly. If I want to restore the series/2.x (the main branch) to its current state as it is in typelevel/cats-effect, I should first branch off the current series/2.x branch in dotty-staging/cats-effect in order to save the history and force push the main branch?

[smarter]

Yes, and push the saved branch too :) I suggest putitng the date in the name of the branch.

[vasilmkd]

@smarter Does series/2.x-20.09.2020 work?

[smarter]

There's only one true date format: https://xkcd.com/1179/ :)

[vasilmkd]

Agreed. 😆

[vasilmkd]

https://github.com/dotty-staging/cats-effect/tree/series/2.x-2020-09-20

I will go ahead and force push series/2.x.

[vasilmkd]

Thank you @smarter for the quick replies and the help.