Zero out client state before processing upgraded client proof

[colin-axner]

Summary

In 07-tendermint, we check the proof against the submitted upgraded client state. If the relayer submits the zeroed out client state wrong, it fails. We should zero out the client state before checking the proof. If the proof works for the zeroed out client state, no need to prevent an upgrade because a relayer implementation was incorrect

---

For Admin Use

• Not duplicate issue
• Appropriate labels applied
• Appropriate contributors tagged/assigned

[crodriguezvega]

So do we need to do something like this?

// Verify client proof
zeroedUpgradedClient := upgradedClient.ZeroCustomFields()
bz, err := cdc.MarshalInterface(zeroedUpgradedClient)
if err != nil {
	return nil, nil, sdkerrors.Wrapf(clienttypes.ErrInvalidClient, "could not marshal client state: %v", err)
}
// construct clientState Merkle path
upgradeClientPath := constructUpgradeClientMerklePath(cs.UpgradePath, lastHeight)
if err := merkleProofClient.VerifyMembership(cs.ProofSpecs, consState.GetRoot(), upgradeClientPath, bz); err != nil {
	return nil, nil, sdkerrors.Wrapf(err, "client state proof failed. Path: %s", upgradeClientPath.Pretty())
}

In this part of the code?

[colin-axner]

Correct, this should be done with the 02-client refactor

[colin-axner]

closed by #1674