To be clear, I wasn't worried about a race for updateBaseValue. I was worried that it was passing zero and expecting that to *always* set the ref. This is purely for future proofing changes to checkpointNeeded and reducing "magic" values. Another option would be a checkpointNeeded that takes no timestamp and just unconditionally sets the ref.
To be clear, I wasn't worried about a race for updateBaseValue. I was worried that it was passing zero and expecting that to *always* set the ref. This is purely for future proofing changes to checkpointNeeded and reducing "magic" values. Another option would be a checkpointNeeded that takes no timestamp and just unconditionally sets the ref.