Use personal_sign when injected web3provider is TrustWallet

[alfetopito]

Summary

Using personal_sign instead of eth_sign for TrustWallet.

This sort of a follow up to my comment on a similar PR #1285 (comment)

TrustWallet's eth_sign method does not properly encode the data.
With this workaround I'm able to perform the signing as expected.

The flag isTrust is set by TrustWallet's web3 provider https://github.com/trustwallet/trust-web3-provider/blob/93395938bb6838e35e3e358bcf952124131186fa/README.md#L15 similar to isMetaMask and isStatus flags.

According to MM docs, personal_sign is preferred instead of eth_sign.

And now that I went through TrustWallet's web3 provider code I found out the reasoning for different choice of method: trustwallet/trust-web3-provider#102 (comment)

Although, to be honest, I'm not 100% clear on that.

Testing

Well, I would have published a RC npm package with these changes under my NPM account, or at a minimum publish it using yalc locally, but setting up ethers.js is a bit tricky.
So I went with the good old "copy generated files to destination node_modules folder". And it worked as expected :)

If you insist, I can work on a minimum code sample.

[alfetopito]

FYI, we kind of need this change sooner rather than later, so I went ahead and published the providers package under my account https://www.npmjs.com/package/ethers_providers_alfetopito

I made a mistake and screwed the package name, but it's working and makes it explicit for us that this is just a temporary fork.
When this change is merged and released with ethersjs I'll clean that mess up :)

[ricmoo]

I would live to make this change, but am waiting for someone to provide the research outlined here:

#1285 (comment)

I’ll open a new issue to track this.

[ricmoo]

Please see #1544.

Add any environments you have access to. :)

[rhlsthrm]

Any update on this? We need this change as well.

[ricmoo]

I’ll be making the switch on the next minor version bump.

[rhlsthrm]

Any timeline for this?

[ricmoo]

Not yet, I’ve just finished off a few major tasks, and have 2 presentations to prepare for. Maybe within 2 weeks?

[rhlsthrm]

Thanks @ricmoo. Will be watching this and implement a workaround in the meantime.

[rhlsthrm]

Ping @ricmoo any update here?

[ricmoo]

Useful info: WalletConnect/walletconnect-monorepo#1395

Starting in v5, the JsonRpcSigner will use personal_sign, but a new method provider._legacySignMessage has been added for compatibility for those who need it. This legacy method will likely go away entirely in v6 though.

[ricmoo]

This has been updated in 5.5.0.

For those who require the old behaviour, the JsonRpcSigner has a _legacySignMessage method which mimics the old behaviour; but not that this is functionality most clients no longer support, or if they support the method, it performs prefixing breaking legacy anyways. In v6, this method will go away.

Thanks! :)