-
Notifications
You must be signed in to change notification settings - Fork 6
Transactions causes split with Parity #55
Comments
might be config issue if it also happens before block 100 |
eip98 is no longer relevant here it does not happen before block 100, so it must be an atlantis issue
|
But reproducible after block 100
|
that is strange that the unverified transaction would be from a Call, which wasn't changed during Atlantis and is tested through the Ethereum tests. I wonder if there is an implicit transition that we haven't implemented, like EIP 684, that isn't specifically handled in Parity? Now looking a bit into it, it seems there may have to be a transition defined in these backwards compatible clients to disable EIP 684 and other unsupported functionality in Atlantis. |
it can't be EIP 649, can it? |
no |
The only reason I think it could be a config discrepancy is because the Ethereum tests submodule all pass (Except for EIP 684 and noted edge case). I will continue to look into it and run my own tests. |
Yes, there is still a chance this is a config issue. I really can't think of anything though Tried setting up a new testnet but somehow failed to make the clients even talked to each other. Need a break 💥 |
I have meticulously gone through config and compared it to existing parity config and I can't see anything that I think is wrong, am very confident our client's config is set up correctly, and went through each commit's changes to anything around Call and did not find any functional differences or implementations. Just an update, I'll start trying to get creative to find out what exactly is going wrong. |
Have you seen this? we can run multi geth on kensington now, this makes it easier to determine which client is wrong |
sounds good! I will look into it in about an hour |
Is this a typo or copy paste issue that |
nvm, it omits on long lines |
So, Multi-Geth using the Parity configuration also fails to process these blocks
|
so multi-geth and parity are behaving in agreement about the "bad block" (for them it is not a bad block?)? have these experiments always been when sending the tx from getc? or does getc also report a bad block when receiving a block with an equivalent tx? (in some of these snippets above, it is not clear to me which client is reporting or doing what, sorry) |
and this
makes me wonder if the transaction is being properly signed? is the nonce being incremented properly? is eip658 engaged, and properly including the |
|
this might be why
where |
Yes, a block with a normal transaction signed by Getc is handled the same way in Multi Geth and Parity Ethereum as "bad block" (see above).
I also tested it the other way around. A block with tx signed by Multi Geth causes the same invalid receipt root hash in Getc. |
@meowsbits Sorry I had not checked on this previously, this, along with the actual problem causing the consensus issue had already been fixed, see PR referenced. Changes have been tested internally and on Kensington and this version has stayed in sync with Kensington since the change |
@austinabell I'm not sure I can concur on the "already been fixed" clause, when the current PR addressing this is still open, but glad to see that it is being resolved. |
okay, sorry on my poor choice of words :) I really just meant to be succinct to create the connection to the change that solves the specific issue of the thread as to not cause any confusion for any reader. |
This:
Caused this:
The text was updated successfully, but these errors were encountered: