why subtract one in integer_chip

[JohnWick2ETH]

why subtract one to compute v1 in the integer_chip.add_constraints_for_mul_equation_on_limb0()?

AFAIK, v1*2^2b = u1 + v0.

halo2-snark-aggregator/halo2-ecc-circuit-lib/src/five/integer_chip.rs

Line 194 in 8b01e90

let u1 = v0 - one + limbs[2].value - rem.limbs_le[2].value

[lanbones]

To avoid minus overflow when computing u0, we borrowed one from v1.
Think about
limbs[1].value < rem.limbs_le[1].value || (limbs[1].value == rem.limbs_le[1].value && limbs[0].value < rem.limbs_le[0].value)
see

halo2-snark-aggregator/halo2-ecc-circuit-lib/src/five/integer_chip.rs

Line 190 in 8b01e90

+ self.helper.limb_modulus_exps[2];

,

[JohnWick2ETH]

why would it happen that u0 is negative?