Subtle failure in UNDO

[Terkwood]

she went back 7 moves and then tried to go forward and it blocked up

Generally undo works, but sometimes it doesn't.

Is this about backend? If so, which service? Does sync impact this?

Or is it the frontend?

Can we describe the failure scenario more precisely? Sometimes after 60 undos you still cannot trigger this bug

• make sure 🐊tinybrain: katago deser & illegal move error #456 doesn't impact this behavior

[Terkwood]

We should create an integration testing app which can communicate with Redis and push undo commands, expect move undone events, push make move commands, expect acceptance. It should push many thousands of requests and detect failure. This would either rule out or blame the undo /changelog flow

[Terkwood]

Advancing #486, then using it in the test harness, should help this

[Terkwood]

Test harness should use #502 instead of #486

[Terkwood]

This occurs infrequently when I test using personal devices. Can it be triggered more often if BUGOUT web client is running on an azure vm ?

[Terkwood]

This failure state does not prevent unrelated games from proceeding. It does not log anything useful in undo service.

one potential plan

Write sync procedures into sync service and browser impl,so that silently failing UNDO requests can be reconciled