Updated to v7.
There are 2 changes that have happened here. sgn0 now uses a new method for calculating the sign and the domain length is appended to the DST.
I cannot find what is wrong here. It passes 1 of the 4 hash to curve test vectors. The problem is in sgn0, but I cannot find it. For example in the a512_aaaaaaaaaaaaaaaaaaaaaaaa... test, Q0.x is correct, but Q0.y is incorrect. I cannot trace where this comes from and it has been days... The function is so simple and I've implemented it line for line. I give up.
What was wrong?
A new variant of H2C was released.
How was it fixed?
Updated to v7. There are 2 changes that have happened here.
sgn0
now uses a new method for calculating the sign and the domain length is appended to theDST
.I cannot find what is wrong here. It passes 1 of the 4 hash to curve test vectors. The problem is in sgn0, but I cannot find it. For example in the
a512_aaaaaaaaaaaaaaaaaaaaaaaa...
test,Q0.x
is correct, butQ0.y
is incorrect. I cannot trace where this comes from and it has been days... The function is so simple and I've implemented it line for line. I give up.