Gsf5 fixes #4312
Gsf5 fixes #4312
Conversation
|
|
||
| // edge-case detected on GSF5 - during the transition we want to try process all transition blocks from CL client | ||
| // The last condition: !parentBlockInfo.IsBeaconBody will be true for terminal blocks. Checking _posSwitcher.IsTerminal might not be the best, because we're loading parentHeader with DoNotCalculateTotalDifficulty option | ||
| bool forceProcessing = !_poSSwitcher.TransitionFinished && (_blockTree.Head?.Number ?? 0) + 8 >= block.Number && !parentBlockInfo.IsBeaconBody; |
There was a problem hiding this comment.
We're using 8 here too: https://github.com/NethermindEth/nethermind/blob/master/src/Nethermind/Nethermind.Facade/Eth/EthSyncingInfo.cs#L37
I should change this condition bool weAreCloseToHead = (_blockTree.Head?.Number ?? 0) + 8 >= block.Number;
| public bool ValidateSeal(BlockHeader header, bool force) => | ||
| _poSSwitcher.IsPostMerge(header) || _preMergeSealValidator.ValidateSeal(header, force); | ||
| public bool ValidateSeal(BlockHeader header, bool force) => | ||
| _poSSwitcher.GetBlockConsensusInfo(header, true).IsPostMerge || _preMergeSealValidator.ValidateSeal(header, force); |
There was a problem hiding this comment.
qq: why we don't check the seal if it is post merge?
There was a problem hiding this comment.
It is PoW specific thing, so we don't have it post merge
There was a problem hiding this comment.
| @@ -180,7 +180,8 @@ public void AddNewBlock(Block block, ISyncPeer nodeWhoSentTheBlock) | |||
|
|
|||
| bool isBlockBeforeTheSyncPivot = block.Number < _pivotNumber; | |||
| bool isBlockOlderThanMaxReorgAllows = block.Number < (_blockTree.Head?.Number ?? 0) - Sync.MaxReorgLength; | |||
| bool isBlockTotalDifficultyLow = block.TotalDifficulty < _blockTree.BestSuggestedHeader.TotalDifficulty; | |||
| bool isBlockTotalDifficultyLow = block.TotalDifficulty < _blockTree.BestSuggestedHeader.TotalDifficulty | |||
| && (_specProvider.TerminalTotalDifficulty == null || block.TotalDifficulty < _specProvider.TerminalTotalDifficulty); // terminal blocks with lower TTD might be useful for smooth merge transition | |||
There was a problem hiding this comment.
What if there are two terminal block and the lower one appear later? I assume it must go through NP?
There was a problem hiding this comment.
If terminal block appears later that shouldn't be block choosen by CL clients to build chain on top of it. However, in this case we will get NP(childOfTerminalBlock) we will return SYNCING and we should be able to catch up with MergeBlockDownloader
|
The workflow is not starting the tests, but I run them locally. |
Fix issues that we noticed during transition on GSF5