[Bug]: change in behavior of `personal_sign` on Ledger, in v10.31.0

[firnprotocol]

Describe the bug

as of v10.31.0, but not for versions strictly before it, there was a change in the output of personal_sign, which takes place only on Ledger devices.

prior to the upgrade, the final v byte of the signature was either 0x00 or 0x01. now, it is either 0x1b or 0x1c.

this appears to bring the behavior on Ledger in line with for regular EOAs, which are either 0x1b or 0x1c both before and after v10.31.0.

can someone please explain this? thank you very much.

Steps to reproduce

go to https://etherscan.io/verifiedSignatures#. click "Sign Message". follow the steps to log in using MetaMask. enter an arbitrary string for your message.

observe the last byte (i.e., last two hexits) of the raw signature output. if you're signing using a Ledger, it will be either 0x00 or 0x01 pre-upgrade, and either 0x1b or 0x1c post. for a non-ledger account, it'll be either 0x1b or 0x1c both pre-and post.

Error messages or log output

No response

Version

v10.31.0

Build type

None

Browser

Chrome

Operating system

MacOS

Hardware wallet

Ledger

Additional context

No response

[firnprotocol]

to clarify, it's not clear that the new behavior is the bug—indeed, perhaps the old behavior was the bug. i would just like an explanation, if possible—i am not able to find anything in the release description.

[plasmacorral]

Apologies that this was missed in the release description. This was addressed in a change to the keyring, you can find more information here: MetaMask/eth-ledger-bridge-keyring#152