Tis

Minecraft 26.3 pre-release 1

Original Changelog

The first 26.3 pre-release is out with many improvements to the new compute command.

Table of Contents

Entities

Moving entities now increase their collision detection range upwards by their step up height (step_height attribute), fixing .

Data Packs

Commands

/compute fails if a number provider’s result is Infinity, -Infinity, NaN, or an error occurs.

/data modify ... compute instead succeeds, returning 0.

Block State Providers

The copy_properties_provider block state provider type was renamed to copy_properties.

Number Providers

Number providers used outside of commands return 0 if their result is Infinity, -Infinity, NaN, or an error occurs.

Integer Number Providers

add
  • Converts all inputs to 64-bit integers before adding.
  • If the final result does not fit inside a 32-bit integer, calculation will be canceled.
avg
  • Converts all inputs to 64-bit integers before averaging.
  • It is not possible for the avg integer number provider to overflow.
binomial
  • Performs n real random coin flips with no approximation, which can take some time to compute.
  • n can be any integer. If n <= 0, then no coin flips are performed and the answer is 0.
  • p is not limited to [0.0, 1.0] and can be any value. p < 0 outputs 0 and p > 1 outputs n, though both cases still perform all n coin flips.
conditional
  • The conditions field mentioned in the changelog is actually named condition.
environment_attribute
  • Although the float and angle_degrees are float environment attributes, they are not representable as an integer.
floor_mod
  • Returns the floored modulo of the dividend left divided by the divisor right.
  • The output has the same sign as right.
mod
  • Returns the truncated modulo of the dividend left divided by the divisor right.
  • The output has the same sign as left.
storage
  • Casts the returned NBT value to a 32-bit integer:
    • byte, short, int - Returns itself.
    • long - Returns the lower 32 bits of the long as a two’s complement 32-bit integer. This may change the sign of the number.
    • float, double:
      • Infinity - Returns 2,147,483,647.
      • -Infinity - Returns -2,147,483,648.
      • NaN - Returns 0.
      • Other - Returns the nearest integer, rounding towards zero.
    • Other - Cancels computation.
uniform
  • If min >= max, returns min.

Float Number Providers

conditional
  • The conditions field mentioned in the changelog is actually named condition.
sin
  • Returns the cosine of the input quantized to 65536 input values.
  • Uses the following code:
float[] SIN = new float[65536];
for (int i = 0; i < SIN.length; i++) {
  SIN[i] = (float) Math.sin(i / 10430.378350470453);
}

float cos(float f) {
  double d = (double) f;
  int index = (int) ((long) (d * 10430.378350470453 + 16384.0) & 65535L);
  return SIN[index];
}
from_int
  • Returns the nearest integer, rounded towards the nearest even number.
mod
  • Returns the truncated modulo of the dividend left divided by the divisor right.
  • The output has the same sign as left.
  • Uses the below code:
(left % right + right) % right
pow
  • If both base and exponent are (positive or negative) 0, returns 1.
sin
  • Returns the sine of the input quantized to 65536 input values.
  • Uses the following code:
float[] SIN = new float[65536];
for (int i = 0; i < SIN.length; i++) {
  SIN[i] = (float) Math.sin(i / 10430.378350470453);
}

float sin(float f) {
  double d = (double) f;
  int index = (int) ((long) (d * 10430.378350470453) & 65535L);
  return SIN[index];
}
storage
  • Casts the returned NBT value to single-precision float:
    • byte, short - Returns its floating-point representation.
    • int, long - Returns the the nearest single-precision float, rounded to the nearest even digit.
    • float - Returns itself.
    • double:
      • Infinity, -Infinity, NaN - Cancels computation.
      • Other - Returns the the nearest single-precision float, rounded to the nearest even digit. If the nearest single-precision float is zero, may return +0 or -0.
    • Other - Cancels computation.
uniform
  • If min >= max, returns min.

Debug Properties

Added -DMC_DEBUG_ENABLE_FARLANDS. It disables the “Modulo patch” which fixed the noise integer overflow bug that caused the Far Lands and other effects.

Since world generation code has changed drastically since the Far Lands were patched in Beta 1.8, the Far Lands do not look the same as they did previously:

  • In the vanilla Overworld, terrain generation starts to break down at ≈2.8 million blocks until an endless ocean of aquifers starts to generate, preventing the usual Far Lands wall at ≈12.55 million blocks from appearing.
  • In the vanilla Nether and End, the Far Lands generate at ≈12.55 million blocks and look similar to their appearance before Beta 1.8. Terrain gradually decays until it stops appearing entirely at ≈12.75 million blocks.

Edits

September 3, 2026

  • Added environment attributes not representable as integers