There are 3 changes:
1. Replace `anchorQueue` with `anchorTree` to be able to always walk the
anchors in the order of increasing blockHeights (not possible with the
queue) to prioritise making progress on the lowest block heights
2. Not increment `nextRetryTime` if the request was not sent
3. Reduce the strides in skeleton from `8*192` to `192` to reduce
reliance of the long series of requests to make progress
Co-authored-by: Alex Sharp <alexsharp@Alexs-MacBook-Pro-2.local>
There are two fixes:
1) the blockNumber in the CreateHistoryStateReader should be blockNumber
+ 1 as in the eth_call, eth_createAccessList:
- eth_call: already use blockNumber + 1 see
CreateStateReaderFromBlockNumber where the blockNumber
- passed to the CreateHistoryStateReader is blockNumber + 1
- eth_createAccessList: call CreateHistoryStateReader directly with
blockNumber + 1
2) The eth_callBudle allows to execute the original transaction (here
the sender nonce is assigned to the tx) on a user defined block; the
original transition could be performed in another block . The sender's
nonce when the original transaction was performed had a value that may
be different to the nonce of the sender when the tx is "executed" on new
block. So during the execution of the eth_callBundle must not be
verified that the nonce is progressive (original nonce of the tx + 1:
see preCheck() in state_transition.go that; if the checkNonce fails an
error is arised, nonce too low or nonce to high)
The eth_call does not check the nonce because is using ToMessage(calls
types.NewMessage with checkNonce = false) with checkNonce disabled;
while the eth_callBundle uses AsMessage object (where is built the
Message with checkNonce = true) where the checkNonce is enabled.
Action: Disable the checkNonce as proposed
To eliminate a probably cause of issues related to `tx.Put` and
`tx.Append` getting stuck
Co-authored-by: Alexey Sharp <alexeysharp@Alexeys-iMac.local>
When body gets evicted from the cache, its corresponding entry in
`requestMap` is also removed but was never re-instated when re-request
happens
Co-authored-by: Alexey Sharp <alexeysharp@Alexeys-iMac.local>
This is more reliable criterion because it is harder to spoof
continuously by malicious peers
Co-authored-by: Alexey Sharp <alexeysharp@Alexeys-iMac.local>
This is a fix for
[withdrawal-devnet-3](https://forkmon.withdrawal-devnet-3.ethpandaops.io/).
The problem was that CL calls `NewPayloadV2` for pre-Shanghai blocks
with `nil` withdrawals, which is OK, but that `nil` was erroneously
converted into an empty array, causing block hash mismatch.
