False-positive of arrow-parens with generic arrow function

[mnn]

const t = <A>(a: A): A => {
    const b = a + 1;
    return b;
};

I am getting:

error Expected parentheses around arrow function argument having a body with curly braces arrow-parens

which doesn't seem right. Applying fix results in invalid code:

const t = (<)A>(a: A): A => {

Workaround is rewrite arrow function to classic function.

[gajus]

Fix would be welcome.

[gajus]

🎉 This issue has been resolved in version 3.12.0 🎉

The release is available on:

• npm package (@latest dist-tag)
• GitHub release

Your semantic-release bot 📦🚀

[Mr-Wallet]

Still getting this problem on [email protected] and [email protected].

[gajus]

Please raise a PR with a broken test case.

[Mr-Wallet]

@gajus Question about writing tests: is the assumption that we always be targeting the latest version of flow, and changes to the Flow API are supposed to be handled in future PRs? I don't see any way to target specific versions of flow.

After 90 minutes, I'm totally at a loss of how to reproduce the error in the tests. I've ruled out an interaction with other plugins in my own project, and I've copy-pasted a few examples that fail in my project and it doesn't show an error in the tests. The issue only comes up when defining functions that use Generics, and usually only when those functions are defined as arguments to other functions. Using the function keyword instead of arrow functions is a reasonable work-around (we don't use this very much in our code base).

I'm prepared to leave this for now and move on to other things. But if you happen have any advice for why the tests might have different results than an actual project, I would be eager to try some more experiments to try and nail down why this is happening.

[gajus]

@gajus Question about writing tests: is the assumption that we always be targeting the latest version of flow, and changes to the Flow API are supposed to be handled in future PRs? I don't see any way to target specific versions of flow.

That is correct.

I'm prepared to leave this for now and move on to other things. But if you happen have any advice for why the tests might have different results than an actual project, I would be eager to try some more experiments to try and nail down why this is happening.

If you can produce a minimal repository for replicating the issue, that would be enough.