Doing this will produce incorrect results for ed25519. As I have said, your bundler converts bigints to something else and then does Math.pow. Which loses precision if the number is more than 53 bits.

Doing this will produce incorrect results for ed25519. As I have said, your bundler converts bigints to something else and then does Math.pow. Which loses precision if the number is more than 53 bits.

Thanks for the warning!

so far, this change won't break any existing test cases (both in this repo and noble-ed25519). That's why we did it this way.

I will check again and prepare some test cases to cover the scenario that number is more than 53 bits and verifying it.