Reinstate EIP-712 Cosmos AnteHandler path

[dogemos]

Context

In Ethermint EVM, there used to be a path for handling ExtensionOptionsWeb3Tx as a normal Cosmos SDK tx, except that signature verification uses the Legacy EIP-712 representation (via an AnteHandler decorator). This was removed (prior to Evmos → Cosmos EVM transition) from the antehandler, and has caused significant issues with users not being able to use hardware wallet to sign, and therefore fragmentation across Cosmos SDK + EVM chains attempting to fix the issue in their own ways vs a standardized way.

Requested Change

We’d like to propose bringing the equivalent behavior back into cosmos/evm at:
• ante/ante.go around this switch (currently missing the Web3Tx legacy EIP-712 handler):

case "/cosmos.evm.types.v1.ExtensionOptionsWeb3Tx":
  // Deprecated: Handle as normal Cosmos SDK tx, except signature is checked for Legacy EIP712 representation
  anteHandler = NewLegacyCosmosAnteHandlerEip712(options)

Rationale

1. “All Cosmos messages must be handled on the EVM contract side” is not scalable

For chains and apps, forcing every Cosmos-native message flow (staking, gov, bank sends, authz, feegrant, IBC, etc.) to be re-implemented as EVM contract calls doesn’t scale operationally or socially (audits, maintenance, divergent behaviors, long-tail bugs). It also makes hardware-wallet-friendly UX much harder in practice.

2. It pushes implementation burden to wallets (sometimes impossible)

If the chain doesn’t support a high-compat signing path, each wallet must implement chain-specific/custom signing logic for every chain/every message/etc. In cases where code is closed, it can be effectively impossible for wallets to support safely and consistently.

3. Legacy EIP-712 can coexist as a “high compatibility” fallback

EIP-712 doesn’t have to replace existing Cosmos sign modes. This can be an additional compatibility option that lets wallets say: “Use the cosmos/evm standard and your Ledger signing will work with Keplr (and others).” while allowing more complex/well integrated integrations to continue to exist

[linear]

EVM-411

[vladjdk]

Hey @dogemos, I don't personally see an issue with doing this. Do you know why it was removed in the first place? Maybe @yihuang can give some light to this as well.

[yihuang]

I think we don't need the ante handler to do eip-712 signing, it's already builtin the ethsecp256k1.
The original PR: evmos/ethermint#1397