Estimation for contract call gas amount is too low

[cmichi]

When executing the method of some contracts the default Max Gas Allowed value (the estimated one) is estimated significantly too low.

How to reproduce

We've found that at least the ink! examples delegator and contract-terminate are affected. contract-terminate is the easiest to reproduce this. I've attached the contract which can be used for deployment here: contract_terminate.contract.zip.

The contract has only one method terminateMe, the UI suggests a default Max Gas Allowed of 1417, but the contract then fails with system.ExtrinsicFailed: contracts.OutOfGas.

It works if the max gas is set manually to 2500.

Debugging

We've used canvas-node with the latest Substrate HEAD. canvas-ui is also affected by this.

It might be that the actual bug is in Substrate, but I don't know how the logic of polkadot-js works here. Might be that we need to move the issue over to Substrate.

[jacogr]

On the estimation, it uses the contracts.call RPC and then uses the gasConsumed value as the estimated gas passed to the actual call execute tx. So if this is very low (and in this case it seems off by a huge margin), something is really weird in the contracts call RPC in these specific cases.

The UI itself does round the retrieved value upwards, so the passed value is slightly more than the gasConsumed already, but there may be a case to add an additional safety-factor in there as well, say 5% above the gasConsumed estimation, but certainly not 50%

[jacogr]

Logged the actual gasConsumed value on this specific contract, it is 1,416,325,397 (The UI adjust this upwards and passes 1,417,000,000 with the mega-gas rounding)

Added a 1% factor above that, it still fails. Even at 10%/15%/25%/35%/50% above the received gasConsumed value it still fails.

In this specific case, quite probably due to the termination itself?

[cmichi]

Thanks @jacogr! I'm cc'ing @athei.

[jacogr]

I tried to see if I'm doing something stoopid, but def. do call the read RPC with the selected account, i.e. same one that will execute the actual tx as well.

[athei]

Thanks for your investigation. Looks like a a bug in the pallet_contracts then. I will look into it.

[jacogr]

Happy to hear if I'm doing something wrong here, just cannot find out what it is atm. So eyes elsewhere would help - there is indeed an issue "somewhere".

[polkadot-js-bot]

This issue has been open for 21 days with no activity and is not labelled as an enhancement. It will be closed in 7 days.

[cmichi]

There is paritytech/substrate#8976 now, which should fix this.

[jacogr]

Closing, tracked in Substrate.

[polkadot-js-bot]

This thread has been automatically locked since there has not been any recent activity after it was closed. Please open a new issue if you think you have a related problem or query.