Cowswap trade on Gnosis via Rabby does not work

[koeppelmann]

Overview

Connected with Rabby/ Chrome/ iOS.
Switched to Gnosis in the Shapeshift UI -> Rabby also switches to Gnosis.
Want to do trade via Cowswap.
When initiating the trade as expected message to sign pops up.

Issue 1) It is not formatted
Issue 2) It seems that it is signing a message for mainnet and not for Gnosis

2. is likely the reason why after signing I get an error message.

--

References and additional details

..

Acceptance Criteria

..

Need By Date

No response

Screenshots/Mockups

No response

Estimated effort

No response

[gomesalexandre]

Hi @koeppelmann thank you for reporting this! Please note that we don't support Rabby, I suppose you've used MetaMask and rabby injected at MM.
I'm not familiar with the internals of Rabby, are you currently connected to Gnosis chain on it when attempting the trade? If not, the way things work is there is a first JSON-RPC call to wallet_switchEthereumChain before the actual signature (in the case of CoW)/Tx is prompted for signing, so I would assume Rabby doesn't support this method.

[twblack88]

@koeppelmann thanks so much for bringing this in! We've got some rough edges to smooth on COWswap and we absolutely welcome your feedback. If you're trying to use it as walletconnect we are going to prioritize that at the next roadmap call (hopefully) this might be a great use case.

@gomesalexandre I think Rabby runs their own RPC IIRC, they're stack is 🐐 -ed but that flow makes sense.

[gomesalexandre]

@twblack88 this wouldn't be related to the JSON-RPC endpoint used, but to:

1. wallet_switchEthereumChain method not being triggered. Disregard my previous message about Rabby supposedly not supporting this method, this is irrelevant of the wallet used. The reason why switch chain isn't triggered is because this is a signature, not a transaction. Hence, we don't need to switch chains, see this CoW signature while connected to BSC:

And the matching CoW Tx on Gnosis chain: https://explorer.cow.fi/gc/orders/0xf46f02d45e3c2d3d528105f1448361662a8f9ba61e0a6f61dd456fe834a5e4f45daf465a9ccf64deb146eeae9e7bd40d6761c98664ec666b?tab=overview

@twblack88 I think this raises a valid point for signatures, while the wallet doesn't need to be connected to a specific chain, it can definitely confusing for users. What about we switch chains before signing messages, similar to before signing and broadcasting a Tx?

2. Confirmed with MM our messages are not currently structured, at least not from a user-facing perspective. This is CoW's UI for comparison:

Looking at hdwallet, we indeed use personal_sign vs. eth_signTypedData_v4 if applicable for EIP-712. We should implement hdwallet signing using eth_signTypedData_v4 to make this human-readable.

[twblack88]

We should implement hdwallet signing using eth_signTypedData_v4 to make this human-readable.

@gomesalexandre agreed, slot in when we have the time to reduce UX confusion.

[gomesalexandre]

@twblack88 I wasn't able to repro this - Rabby signing is happy and I was able to complete a CoW swap end-to-end

I suggest we close this and open an issue for eth_signTypedData_v4 instead. This is a relatively high effort with huge risks for regressions (in particular signing with Ledger and Trezor through MM, see https://docs.metamask.io/wallet/how-to/sign-data/). I do believe we need this so users aren't blind signing and are able to see the structured EIP-712 data, but as far as this is concerned, this is a non-issue.

[gomesalexandre]

Closing in favor of #5916 as discussed with @twblack88