Stale/Fresh Leaf mechanism in overseer might not actually get used

[eskimor]

It is not clear whether a chain reversion actually triggers a BlockImport event, as technically no block got imported. We should verify through reading code/docs and trying whether that code is actually needed and if not, get rid of it.

Semantics of BlockImport event should in any case get clarified in code and docs on how it behaves on chain reversions.

[tifecool]

How can I execute a "chain reversion"?

[eskimor]

You need to trigger a dispute that resolves against a candidate. This will trigger a reversion right before the including block. A dispute can be triggered in Zombienet via malus nodes. See here for example.

[bkchr]

A re-org is not triggering any block import event. We also don't directly re-org from the point of view of the database. Babe will just start building on a different fork as being told by the SelectChain implementation in Polkadot. The Substrate db re-orgs when there is a fork that is better than the current best chain (this is happening this way because of some legacy stuff AFAIK).

[eskimor]

Got it, so this answers the question: The code is not actually in use and can be removed.

[tifecool]

I'm assuming the status for the ActivatedLeaves will be Stale as a chain revertion is still occurring, but the leaves are being encountered again?

[ordian]

I'm assuming the status for the ActivatedLeaves will be Stale as a chain revertion is still occurring, but the leaves are being encountered again?

No, LeafStatus::Stale is an unreachable state, so we can remove LeafStatus completely. As mentioned above, on chain reversion, there won't be an block import notification issued for the new (old) best block.

[tifecool]

I'm not sure I'll be able to assist with this issue 😓