Removed code duplication (function was introduced in D8010)

Comment and readonly

Extracted the shared code (here and not in D8013, because that's how I have my git stack set up).

Change the function to take a object of params, fix

I think you could test the migration code by mocking the migrations variable with jest, but not sure if that's worth it.

Could we just add it to ModalOverlay so it applies to every modal automatically?

Are you sure about this? Don't we need both web (js code) version and desktop (electron)?

Lgtm

Do we still have places where we don't use hasMinCodeVersion function? Do we have to update these?

Also move tests for id conversion to lib

I think you could merge this one too. Thanks!

Not a fan of this from an accessibility stand point. I found something like this: focus-trap, we could wrap all our modals with it. But I'm not sure if that's worth it. @ashoat any thoughts?

Accepting, but please fix the icon before landing!

Just a few more points about the buttons, otherwise looks good!

As far as I know, diffs should be small, and do only one thing. The division between those diffs is simple: one diff adds the button itself to the tooltip, and the other adds its functionality (entering edit state).
Also, this diff cannot depend on previous diffs, because it is first in the stack.

Rebase, validate uploadMediaMetadataResponder, fix bug with uploadDownloadResponder

Sorry, I should have noticed that during rebase, I will put up a diff with the validators.

These //TODO comments make it hard to review this stack because now the diffs depend on both previous and future diffs. This diff in particular seem like it would be easy to merge with D7824. Is there a reason why you did it this way?

Use FUTURE_CODE_VERSION

Add newline

• Validate also when viewer is null (when error socket message is sent)
• APIResponseServerSocketMessage can have no payload (for example when calling deviceTokenUpdateResponder which returns void). Updated the socket type and the validator to match this

Remove unnecessary changes

Are you sure this is correct? Shouldn't we also convert a response from the server?

If I'm not misunderstanding, the responses are validated in D7749.

Instead of extracting the regexes, instead move the session functions. Only invalidSessionRecovery had to be moved but I moved both of them to keep them together, I can move invalidSessionDowngrade back if that's prefered.

• Updated the test plan
• Also converted role ids. This required the conversion logic to handle tComb refinements, so I added that and tests

Changed the output validation from server error to a warning. This will be temporary, but with this we can safely check for validator <-> flow-types/runtime-objects mismatches without breaking clients. We can change it back once we activate the conversion logic. Included the id in the error message of serverID -> clientID conversion.

Sorry should have mentioned this. There are added because flow types and validators are not in sync. These properties were removed at one point and aren't used. They still exist in the validators so we can support older clients.
Now that tShape<T> is checking if properties count and names match between the flow type and the validator we get an error. Adding empty makes them match, while still blocks the rest of the code from using these values, because they are empty.

I've talked with @ashoat about report ids, and in the future, we probably will send them to some centralized server, so we don't need to convert ids or payloads. Currently the ids aren't used anywhere on the client side (except for being displayed in an alert once).

Fix one test that was being skipped by mistake

Also convert roles

Also convert roles

Also convert media ids