[light-client] remove pending parentchain tx-inclusion check #1455
Conversation
clangenb
changed the title
| block.extrinsics().iter().for_each(|xt| { | ||
| if let Some(index) = relay.verify_tx_inclusion.iter().position(|xt_opaque| { | ||
| <HashingFor<Block>>::hash_of(xt) == <HashingFor<Block>>::hash_of(xt_opaque) | ||
| }) { | ||
| found_xts.push(index); | ||
| } | ||
| }); |
There was a problem hiding this comment.
This was the piece of code that didn't work and that caused a delay in block import.
| self.last_finalized_block_header.number(), | ||
| self.current_validator_set, | ||
| self.current_validator_set_id, | ||
| self.verify_tx_inclusion.len() | ||
| self.unjustified_headers.len() |
There was a problem hiding this comment.
It should always be 0, but I still want to log this, just in case.
clangenb
added
A0-core
clangenb
changed the title
There was a problem hiding this comment.
this PR actually removes functionality. Please describe this loss of functionality in the PR description, along with the follow up issues to get the functionality again.
My understanding: We can no longer verify that an extrinsic which was sent by the enclave has actually been successfully included in the parentchain. Is that so?
|
I have updated the PR description. Yes, you are right that this feature is removed, but it was never complete. Because we have checked in the light-client that the extrinsic was included, but we did not check there whether the extrinsic was a success or not, nor did we have some kind of reaction implemented if the extrinsic was found or if it hasn't been found for a long time. So we should probably integrate these checks somewhere else in the business logic, like in the indirect executor, where we are currently starting to implement checks if parentchain transfers were successful for the privacy sidechain case. |
This PR removes the parentchain tx-inclusion check, which didn't work. This led to an infinitely growing list of extrinsics with pending inclusion check in the light-client, which slowed down the block import tremendously. As the light-client was never the correct place to do that, I just removed the feature, and added a tracking issue for proper re-introduction: #1457.