[Bug]: gasLimit auto increased when sending erc20

[shunjizhan]

Describe the bug

when I send erc20 with MM, it uses gasLimit which 50% greater than eth_estimateGas return. I did some grep around and it seems to be a "feature". https://github.com/MetaMask/metamask-extension/issues/8771#issuecomment-759243828

However this become a bug in our evm network, where gasLimit has some specific encoding, so can't be altered randomly. I hope to make MM use exactly what eth_estimateGas return.

This 50% increase is a very opinionated operation that's only required for networks that has the gas underestimation issue. So I think there should either be a way to turn it off (make it default behaviour, but optional), or force enable it ONLY for certain networks that really need it.

So my question is:

• is there a way to turn this feature off? If not, it's a feature request.
• can someone please clarify when does MM auto increase gasLimit by 50%? Is it only for token transfer, or is it for all contract calls? If it's only for token transfers, I can probably do some workaround in our network, such as "divide gasLimit by 1.5 before eth_estimateGas return when sending erc20"

Thanks!

Steps to reproduce

1. start an evm network with ETH RPC locally
2. connect MM to localhost, try send erc20
3. local RPC logs show that eth_estimateGas returns X
4. on MM UI we can see X * 1.5 as gasLimit

Error messages or log output

No response

Version

10.30.4

Build type

None

Browser

Chrome

Operating system

MacOS

Hardware wallet

No response

Additional context

No response

[anaamolnar]

Hello, @shunjizhan. Thanks for reporting! I will pass this on to the team.

[shunjizhan]

thank you!

[shunjizhan]

From the codebase I do see this feature disabled for optimism.

I think it's better to be "enabled only to certain networks", instead of current "enabled globally, but only disabled for specific networks".

[bschorchit]

Thank you for reporting and for your investigation into this. We'll investigate wether this buffer is still needed (it has been introduced very early on when estimations were less accurate) and if not needed, we'll communicate it and remove it across networks. If it's still needed, we'll research a permissionless way for networks to workaround/opt out of this feature.