Sync / Login loop

[fschoell]

The problem is when the user can be authenticated on the remote database but somehow not locally.

I had this problem currently on the entity_mapper branch. The local login fails (username / password not correct because of the prefix change). However, due to the failed local login the LoginComponent get's stuck in a loop. First onLoginFailure() is called, which then calls retryLoginAfterSync(), which subscribes to the database sync status and then (because the sync actually works) will be notified about the completed sync, which then leads to another call to login(), which triggers another remote authentication and because this works (and the local authentication not), this triggers a new sync and the whole procedure from onLoginFailure()starts again.

[sleidig]

Is this the same issue as #31 ? Probably it's at least somehow related.

If there is a radically better/simpler solution for this mixing of sync and login I would be more than happy.

[fschoell]

Ok, found the bug. This line is actually the problem. It resets isRetriedLogin when onLoginFailure() is called. This is a problem, because this way this statement will always return true, no matter how many retries we actually had.

[Lorsbyjm]

@liwde is this still possible?

[liwde]

Recent changes (#130) reintroduced this bug in the master.

fb7123d does work (all logging through AlertService)
d28c720 doesn't work (Merge PR130)

I'm investigating, but currently I don't see where the issue is coming from.

I tried to implement @fschoell's solution, but then the login didn't work at all.
I also tried to always self.retryLoginSubscription.unsubscribe(); regardless of sync state (sync works fine, as described above), but then Login doesn't work at all.

I will continue investigating next week.

[sleidig]

So this is only occurring when the remote login succeeds and the local login fails? That would be a fringe case that only occurs on an inconsistent database, isn't it?
I'd suggest to first finish the redesign of the whole Session/Login system (#78) and only then look back at this to see whether it still occurs.

[sleidig]

closing this as the whole session/login architecture has changed with PR #246 (issue #78). If similar problems come up again we can open a new issue.