[pallet-revive] ecrecover

[pgherveou]

Add ECrecover 0x1 precompile and remove the unstable equivalent host function.

• depend on [pallet-revive] tracing should wrap around call stack execution #7676

[pgherveou]

/cmd prdoc --audience runtime_dev --bump minor

[athei]

Please make this less invasive. We know all precompiles we are adding here are pure functions. Just call them directly from the Ext::call. Don't push a frame for them or anything. This will remove all the special casing you are doing right now and reduce the diff to a minimum. Please also rename precompiles.rs to basic_precompiles.rs.

Both of those points will keep the merge conflicts I have to deal with down.

[pgherveou]

Please make this less invasive. We know all precompiles we are adding here are pure functions. Just call them directly from the Ext::call. Don't push a frame for them or anything. This will remove all the special casing you are doing right now and reduce the diff to a minimum. Please also rename precompiles.rs to basic_precompiles.rs.

Both of those points will keep the merge conflicts I have to deal with down.

that's what I did initially, but you also need to duplicate all the tracing logic if you do it, and handle value transfer

[athei]

I understand. But please still just duplicate the logic into precompiles.rs then. I want this PR to have minimal impact on exec.rs. First, because making contract_info optional generates a lot of changes I don't want to backport. Secondly, the merge conflicts I will get from this will break my brain. I checked my branch and it will be bad.

[athei]

I would rather just duplicate the value transfer logic here to keep the diff easy.

[pgherveou]

I understand. But please still just duplicate the logic into precompiles.rs then. I want this PR to have minimal impact on exec.rs. First, because making contract_info optional generates a lot of changes I don't want to backport. Secondly, the merge conflicts I will get from this will break my brain. I checked my branch and it will be bad.

Sounds good then, you have a branch somewhere with another version of precompile implemented?

[athei]

I understand. But please still just duplicate the logic into precompiles.rs then. I want this PR to have minimal impact on exec.rs. First, because making contract_info optional generates a lot of changes I don't want to backport. Secondly, the merge conflicts I will get from this will break my brain. I checked my branch and it will be bad.

Sounds good then, you have a branch somewhere with another version of precompile implemented?

Yes its at/precompiles. But its a bit old and needs rebasing. And I am not 100% sure that the way I am doing it is how we want it. But my point is that it needs more thought and I want to defer the decision for after the release. Hence the requested copy paste solution here.

[pgherveou]

Ok did the refactoring yesterday
Should be fine now might have missed a few things will come back to check ci logs later

[athei]

This is correct. But we also need to not charge the contract call gas in case of a pre-compile. Otherwise it will be super expensive since the contract call gas does include accessing the contract code storage item.

[pgherveou]

do we want to keep the diff small here too and do it naively like this

54f6105

[paritytech-workflow-stopper]

All GitHub workflows were cancelled due to failure one of the required jobs.
Failed workflow url: https://github.com/paritytech/polkadot-sdk/actions/runs/13502890532
Failed job name: fmt