transmissions11
This implementation violates the invariant that the VRGDA should never reach 0
Noticed this during invariant testing of an Art Gobblers adjacent project. As we approach infinity, this implementation rounds down. The conjecture is that it should round up to 1 wei.
Credit to @mevbandit @MarcusWentz @horsefacts for the help in discovering this.
i dont think not reaching 0 is necessarily an invariant, whats the use case you have in mind where it needs to round up?
Looking at the equation, the function should never reach zero. I noticed it because we hit divide by zeros during invariant test and that function rounding up was one of the invariants.
we have finite precision, beyond a certain point both rounding up and rounding down are "wrong" behavior mathematically
Agree, but my conjecture is it's always a non-zero value, and so rounding up is more correct. Agree there is no right choice here though :).
i am a little biased cuz its a bit scary changing things to differ from the gobbler implementation, but i do kinda agree with you
specifically if you, say, wanted to sell something only to people who can get access to some token (say the token is normally soulbound or something) if you round down you allow people without access to that token to purchase
@0xAlcibiades if you want to make a PR for this i think i'd be willing to merge it thru 👍