Rebase

Rebase

Rebase

Rebase

Rebase

Rebase

Rebase

In D7519#224404, @tomek wrote:

Fix the index and delete unnecessary field. Tested this change by droping a table and running the migration. I have to test creating a fresh db - going to do that before
landing.

Do not await report promise

Add some code comments

In D7524#226197, @tomek wrote:

In D7524#223171, @inka wrote:

Would it make sense to add an invariant to check that this modal is not called with `status: 'already_joined'? Currently there is nothing stopping someone from doing that, because there is no type checking. And the modal would be displayed without errors, with a message that would be incorrect and misleading.

Yeah, adding some validation is a good idea!

Do not show a modal in already_joined state

Don't return response directly

In D7524#223171, @inka wrote:

Would it make sense to add an invariant to check that this modal is not called with `status: 'already_joined'? Currently there is nothing stopping someone from doing that, because there is no type checking. And the modal would be displayed without errors, with a message that would be incorrect and misleading.

It is a good idea to check if there are other places where this check is missing. Describing how it was done should extend the test plan.

I believe the z-index is necessary since we're rendering the message in the ModalOverlay, and that has a z-index of 4, so by default I think the tooltip will render below the modal.

Haven't compared validators with the types as it is more time consuming here than in other diffs, where the Flow types were next to tComb ones. If you could make comments with the Flow types, I would take another look.

Could we add a test where avatar is not defined (RawThreadInfo doesn't have this field at all, not just it is set to null or undefined).

Is there a way to avoid specifying z-index? (usually when we use z-index in one place, it quickly becomes necessary in a lot of places) If there's no other way, can we limit the number of places where it has to be specified?

Fix the index and delete unnecessary field. Tested this change by droping a table and running the migration. I have to test creating a fresh db - going to do that before
landing.

Cast an id to a string

Looks great! (requesting changes because addressing the comments might change things significantly)

After reading the explanation and spending a lot more time, I was able to review this. This code is complicated, but I don't see any serious issues in it.

This diff is really hard to review. Wondering if splitting it would make it better - if every smaller diff had changes that are easy to review. I guess it might be challenging, so if you think that it's not possible, request a review again and I will give the review another try.

Licenses and versions look ok

If we move actions to lib, we can simplify the reducer and avoid splitting the login in two places.