eth/catalyst: fix canon behind CurrentBlock

[MariusVanDerWijden]

This fixes all hive tests Invalid Ancestor Chain Re-Org, Invalid StateRoot, Invalid P9', Reveal using newPayload (go-ethereum)
where we reveal using newPayload.
The issue is that the head block is marked as canonical but the current block is behind the head block.
I am not really confident in this as I suspect it might not fix the root cause (a block being marked canonical that is not really in our canonical chain).

Maybe @rjl493456442 could take a look at this

[rjl493456442]

The scenario is: the local canonical hash is matched with the block, but local chain head is lower than the block.

I am wondering why it can happen in the first place. Whenever reorg happens, all the canonical hash above the new head will be deleted.

[rjl493456442]

Can you share the concrete test scenario description?

[MariusVanDerWijden]

Invalid Ancestor Re-Org Tests Attempt to re-org to an unknown side chain which at some point contains an invalid payload. The side chain is constructed in parallel while the CL Mock builds the canonical chain, but changing the extraData to simply produce a different hash. At a given point, the side chain invalidates one of the payloads by modifying one of the payload fields. Once the side chain reaches a certain deviation height (N) from the commonAncestor, the CL switches to it by either of the following methods: a) newPayload each of the payloads in the side chain, and the invalid payload shall return INVALID

commonAncestor◄─▲── P1 ◄─ P2 ◄─ P3 ◄─ ... ◄─ Pn
		    │
             └── P1' ◄─ P2' ◄─ ... ◄─ INV_P ◄─ ... ◄─ Pn'
		          ```

[MariusVanDerWijden]

https://github.com/ethereum/hive/blob/f2838a629dcd1a027a260cd10fc84899f89e3ba6/simulators/ethereum/engine/enginetests.go#L220

[rjl493456442]

I think the root cause is: after reorg, the newchain length is 1. The new head won't be processed in reorg function.

So here https://github.com/ethereum/go-ethereum/blob/master/core/blockchain.go#L2075 HeadBlock is not updated and no canonical hash marker get deleted

I will submit a PR to fix it

[holiman]

I will submit a PR to fix it

So, I think @rjl493456442 's PR #24996 was it?
Question then: is this PR still needed?

[MariusVanDerWijden]

I don't thinks so, I'll wait for hive to run and close this PR then