incorrect rounded of negative coordinates

In negative coordinates on edge of chunk layers of glass rendered incorrectly (missing some layers, Which should be visible)
This happens because in Sodium Used wrong rounding method in CameraTransform.java (Should use Math.floor(Value) instead of (Int)value)
Because of this -9.5 Interpreted as -9, when should as -10
Pls fix it

Reproduction Steps

Stay on edge of chunk in negative coordinates, and look on layered glass

Log File

None

Crash Report

None

I don't believe there is any bug with the code you talk about, and cannot reproduce the issues you describe. Because you say this problem doesn't occur for you in newer versions, it is very likely you are experiencing an older bug that has since been fixed, and there is in fact no problem with the code now.

I don't understand the assertion being made. The point of that code is to create two vectors (one which is the integer part, and the other which is the fractional part) so that when they're added together, the result in the original value but with a fixed amount of precision across the entire range. If this wasn't behaving correctly, then all camera movement, not just translucency effects, would be extremely glitchy in the negative quadrant.

I also verified: fractional(-5.43) returns -0.43 and integral(-5.43) returns -5.00. And of course, adding those values back together produces -5.43 again.

Please send us your game log and a screenshot of the problem with the F3 screen open.

Have you made sure you're using the latest version of Sodium? You might be seeing a bug that was present in older versions which we recently fixed, which involved translucency sorting using the wrong vectors.

I dont see this bug on newer version of sodium, but in code i still see this rounding method on latest version
And i don't know why this thing is displayed only on glass, but i just change private static int integral (double value){ return (Int) value;} to private static int integral (double value){ return Math.floor(value);} in CameraTransform.java and that fixed this bug
most likely this bug was fixed in the new version, but via another change (in float fractional method for example), and if that thing didnt use anywhere but here, everything is fine, but it would be better if correct rounding would be used in future versions

I don't understand the assertion being made. The point of that code is to create two vectors (one which is the integer part, and the other which is the fractional part) so that when they're added together, the result in the original value but with a fixed amount of precision across the entire range. If this wasn't behaving correctly, then all camera movement, not just translucency effects, would be extremely glitchy in the negative quadrant.

I also verified: fractional(-5.43) returns -0.43 and integral(-5.43) returns -5.00. And of course, adding those values back together produces -5.43 again.

Mabe it matters cause 15+0.5 it absolute 15.5 and 15.5 in first chunk, while 16-0.5 it 15.5 absolute and -0.5 in second chunk