Merge pull request #110107 from aaronfranke/fix-range-overflow

Fix Range scale overflow
This commit is contained in:
Thaddeus Crews 2025-09-01 12:25:47 -05:00
commit 89f32c6ead
No known key found for this signature in database
GPG key ID: 8C6E5FEB5FC03CCC

View file

@ -44,7 +44,7 @@ double Range::_snapped_r128(double p_value, double p_step) {
// R128 is fixed-precision with 64 bits after the decimal point, but double already uses 53 of those,
// so a step size finer than 2^-11 will lose precision, and in practice even 1e-3 can be problematic.
// By rescaling the value and step, we can shift precision into the higher bits (effectively turning R128 into a makeshift float).
const int decimals = 14 - Math::floor(std::log10(MAX(Math::abs(p_value), p_step)));
const int decimals = MIN(18, 14 - Math::floor(std::log10(MAX(Math::abs(p_value), p_step))));
const double scale = Math::pow(10.0, decimals);
p_value *= scale;
p_step *= scale;