WAI reason
We wanted to do changes to the worldgen system to make it simple/clear and possible to datadrive it in the future.
While doing so we noticed at a certain moment that we generated some areas of the terrain differently than before (not meaning the things placed as decoration, but the actual shape of the land). It took a while to figure out why, but it turned out that in the old code, there had been a bug from the beginning. We forgot to reset the seed of a random when we merged rivers and the biomes.
In the new code there was no way to reproduce this wrong behavior and this is why [this issue] has been resolved as Working as Intended.
We're not happy that we ended up changing the world where the intention was not to. We found no way to keep the bug while also keeping the new functionality we've added. So we chose to keep the new bug-free implementation and hopefully be able to offer datadriven biome definitions in the future because of it.
– @unknown in this Reddit comment
The new world generator changed some biome and landscape generation compared to 1.12.2. Examples:
Changed terrain and missing ice spikes and hill of flower forest biomes (Seed 5178019746038571832,
[media][media]/tp @p 340 104 -3085 40 20)
Missing Desert M and Sunflower Plains biomes (Seed 1443894742985852992,
[media][media]/tp @p 100 200 100 0 90)
Changed terrain and new Savanna M biome (Seed 1443894742985852992,
[media][media]/tp @p 195 115 555 60 27)
Missing flower forest (Seed 554762737523904,
[media][media]/tp @p 394 118 -353)
Related issues
is duplicated by
MC-125057
More or less inconsistent terrain generation in comparison between Stable 1.12.2 and Snapshot 18w06a
MC-126345
Flower Forests are Erased in Worlds Generated Pre 1.12 in 1.13 Snapshots
MC-131025
Terrain generation sometimes generates different land biomes between 1.12.2 and 1.13
MC-131986
Terrain generation boundaries
MC-133433
World gen Inconsistency
MC-137007
Terrain bug in 1.13.1
Attachments
Comments
migrated
NeunEinser
Can confirm. Quoting @unknown from MC-124950, this seems to be an unintentional change.
Heightmaps should be identical, but everything produced post initial heightmap generation might show variations.
Irbis
edited to replace my screenshots to make differences more visible. Screenshots made with 32 render distance from /tp @p 340 104 -3085 40 20
Irbis
First time looked into the code with MCP to find how biomes works.
It seems that problem is caused by changed placement of Modified version of biomes. To be more precise - it's caused by modified biomes at edges of normal biomes. In 1.12 there was ability to generate Modified versions of 2 biomes next to each other. Probably caused by the same "Modifier". Now the Modifier doesn't spread to other biomes and doesn't enlarge the biome itself.
This will affect all modified versions of biomes: Sunflower plains, Savanna M, Flower Forest, Ice Spikes etc... Hope that will help finding a problem in the code.
NeunEinser
@unknown The issue was marked as works as intended, you do not need to keep it updated 😉
Irbis
Hope dies the last. 😛
migrated
Nathan Adams