tree ed48a6a73b7abbdcfba84fdf5adca48cc1f64d0d
parent 43bb3b702e5a85b1b2ba8b7ed3aa2ab0523a9c5e
author Dariusz Luksza <dariusz.luksza@gmail.com> 1702479013 +0000
committer Dariusz Luksza <dariusz.luksza@gmail.com> 1702480597 +0000

Fallback to public organization information

GitHub uses various access scopes for its REST API and users can
choose how much information they allow to access for the integrations.
The narrowest scope is `USER_EMAILl`, which grants only access to the
user's email address. It's important to note that the _Reviewer_ scope
on eclipse.gerrithub.io and review.gerrithub.io is only asking for
access to the email address.

To give users proper permissions in Gerrit we load information about
user teams and organizations. To read those GitHub requires one of
two scopes `org:read` or `user`. None of those are included in the
_Reviewer_ scope. Which led to an exception being thrown.

As we access organizations and teams API's from a cache loader, getting
an exception meant that the value for the given user was never stored.
This then means that each call to user group membership would result in
yet another attempt to load a cache entry and yet another exception. In
the end group membership cache for a user with scope _Reviewer_ will
never be computed. Which will result in poor page load time.

We already have a mitigation mechanism in place, that would fall to
"public organizations" only, when teams cannot be accessed.
Unfortunately, the GitHub REST API endpoint for "public organizations",
was still requiring "org:read" or "user" scopes. This means that for the
_Reviewer_ scope this fallback never worked.

This change adds another fallback, this time to the real "public
organizations" endpoint that doesn't require additional scopes. It also
ensures that we never use `getMyOrganizations()` call directly but
always go through the fallback route.

Bug: Issue 40014763
Change-Id: I9647b9a1b6e30547f7780c785de37aa6ceb8001d