Biome writes at y = world max (320) → ArrayIndexOutOfBoundsException on Paper 26.2 in /plot setbiome and plot clear

[minoneer]

Server Implementation

Paper

Server Version

Paper 26.2 (build 121, 26.2-121-main@a2a42c5; also reproduced on build 112). WorldEdit 7.4.5 (7.4.5+7590-b8dc4c1).

Describe the bug

Since moving the server from Paper 1.21.11 to Paper 26.2, every /plot setbiome <biome> and every plot clear (/plot delete → /plot confirm) prints one ArrayIndexOutOfBoundsException: Index 24 out of bounds for length 24 per chunk of the plot. The world is a standard -64..319 world (24 sections).

The throw site is WorldEdit's adapter-26.2 PaperweightAdapter.setBiome, but the cause is PlotSquared asking it to set a biome at y = 320, one block above the world. Two independent code paths do this:

1. /plot setbiome — Plot.getRegions() uses the exclusive max build height as an inclusive region bound.

PlotSquared/Core/src/main/java/com/plotsquared/core/plot/Plot.java

Lines 2483 to 2486 in b49bd00

	if (!this.isMerged()) {
	Location pos1 = this.getBottomAbs().withY(getArea().getMinBuildHeight());
	Location pos2 = this.getTopAbs().withY(getArea().getMaxBuildHeight());
	CuboidRegion rg = new CuboidRegion(pos1.getBlockVector3(), pos2.getBlockVector3());

if (!this.isMerged()) {
    Location pos1 = this.getBottomAbs().withY(getArea().getMinBuildHeight());
    Location pos2 = this.getTopAbs().withY(getArea().getMaxBuildHeight());   // exclusive (default versionMaxHeight() + 1 == 320)
    CuboidRegion rg = new CuboidRegion(pos1.getBlockVector3(), pos2.getBlockVector3());

PlotArea.getMaxBuildHeight() is documented "Exclusive" and defaults to versionMaxHeight() + 1 (320). The merged-plot branch of the same method already subtracts 1 (int maxHeight = getArea().getMaxBuildHeight() - 1;, line 2563); the unmerged branch does not. RegionManager.setBiome → WorldUtil.setBiomes then does region.forEach(bv -> world.setBiome(bv, biome)), and CuboidRegion iteration is inclusive, so the last layer visited is y = 320.

2. Plot clear — BukkitQueueCoordinator.enqueue rebuilds biome Y from the layer index without the negative-section offset.

PlotSquared/Bukkit/src/main/java/com/plotsquared/bukkit/queue/BukkitQueueCoordinator.java

Line 180 in b49bd00

int y = ChunkUtil.getY(layer + localChunk.getMinSection(), j);

PlotSquared/Bukkit/src/main/java/com/plotsquared/bukkit/queue/BukkitQueueCoordinator.java

Line 201 in b49bd00

int y = ChunkUtil.getY(layer, j);

// blocks (line 180) — correct
int y = ChunkUtil.getY(layer + localChunk.getMinSection(), j);
...
// biomes (line 201) — missing the offset
int y = ChunkUtil.getY(layer, j);

LocalChunk.setBiome stores biomes with getLayerIndex(y) = (y >> 4) - minSection, exactly like blocks. Reading them back without adding minSection back shifts every biome up by 64 blocks in a -64..319 world: a biome queued for y = -64 is written at y = 0, and a biome queued for y ≥ 256 is written at y ≥ 320. HybridPlotManager.clearPlot queues setBiomeCuboid(minGenHeight .. maxGenHeight), so layer 20 (queued y = 256) is the first to land at y = 320 → index 24.

Why it only shows up on 26.2. WorldEdit's adapter-1.21.11 implemented setBiome via vanilla ChunkAccess.setBiome, which clamps Y into the world. The adapter-26.1/adapter-26.2 implementation does chunk.getSection(chunk.getSectionIndex(y)).getBiomes() directly, with no clamp. So both PlotSquared bugs are older than 26.2 and were previously silently swallowed by the clamp (which also means path 2 has been writing biomes 64 blocks too high, and never writing y < 0, in every extended-height world).

Since BukkitChunkCoordinator catches the throwable per chunk and continues, the visible effect is: /plot setbiome applies the biome to every valid layer but skips the refreshChunk for the chunk that threw (players see the old biome until the chunk reloads); a plot clear sets the biome for y ≥ 0 only and skips the tile-entity / entity restoration for that chunk.

To Reproduce

1. Paper 26.2 + WorldEdit 7.4.5 + PlotSquared, plot world with default heights (world.max_height: 320, world.max_gen_height: 319).
2. Stand in an unmerged plot, run /plot setbiome flower_forest.
3. Console prints one ArrayIndexOutOfBoundsException: Index 24 out of bounds for length 24 per chunk.
4. /plot delete + /plot confirm on the same plot prints the same trace per chunk (via BukkitQueueCoordinator.lambda$enqueue$2).

Expected behaviour

No biome writes outside the world's Y range; the region top should be getMaxBuildHeight() - 1 like the merged branch, and the queue readback should be ChunkUtil.getY(layer + localChunk.getMinSection(), j) like the block loop directly above it.

Screenshots / Videos

No response

Error log (if applicable)

[WARN]: java.lang.ArrayIndexOutOfBoundsException: Index 24 out of bounds for length 24
[WARN]: 	at net.minecraft.world.level.chunk.ChunkAccess.getSection(ChunkAccess.java:231)
[WARN]: 	at WorldEdit.jar//com.sk89q.worldedit.bukkit.adapter.impl.v26_2.PaperweightAdapter.setBiome(PaperweightAdapter.java:445)
[WARN]: 	at WorldEdit.jar//com.sk89q.worldedit.bukkit.BukkitWorld.setBiome(BukkitWorld.java:589)
[WARN]: 	at PlotSquared.jar//com.plotsquared.core.util.WorldUtil.lambda$setBiomes$0(WorldUtil.java:230)
[WARN]: 	at java.base/java.lang.Iterable.forEach(Iterable.java:75)
[WARN]: 	at PlotSquared.jar//com.plotsquared.core.util.WorldUtil.setBiomes(WorldUtil.java:230)
[WARN]: 	at PlotSquared.jar//com.plotsquared.core.util.WorldUtil.setBiome(WorldUtil.java:83)
[WARN]: 	at PlotSquared.jar//com.plotsquared.core.util.RegionManager.lambda$setBiome$2(RegionManager.java:409)
[WARN]: 	at PlotSquared.jar//com.plotsquared.bukkit.queue.BukkitChunkCoordinator.run(BukkitChunkCoordinator.java:187)
[WARN]: 	at PlotSquared.jar//com.plotsquared.bukkit.util.task.BukkitPlotSquaredTask.runTask(BukkitPlotSquaredTask.java:39)
[WARN]: 	at PlotSquared.jar//com.plotsquared.core.util.task.PlotSquaredTask.run(PlotSquaredTask.java:44)
[WARN]: 	at org.bukkit.craftbukkit.scheduler.CraftTask.run(CraftTask.java:78)
[WARN]: 	at org.bukkit.craftbukkit.scheduler.CraftScheduler.mainThreadHeartbeat(CraftScheduler.java:474)
[WARN]: 	at net.minecraft.server.MinecraftServer.tickChildren(MinecraftServer.java:1767)

Plot clear variant:

[WARN]: java.lang.ArrayIndexOutOfBoundsException: Index 24 out of bounds for length 24
[WARN]: 	at net.minecraft.world.level.chunk.ChunkAccess.getSection(ChunkAccess.java:231)
[WARN]: 	at WorldEdit.jar//com.sk89q.worldedit.bukkit.adapter.impl.v26_2.PaperweightAdapter.setBiome(PaperweightAdapter.java:445)
[WARN]: 	at WorldEdit.jar//com.sk89q.worldedit.bukkit.BukkitWorld.setBiome(BukkitWorld.java:589)
[WARN]: 	at PlotSquared.jar//com.plotsquared.bukkit.queue.BukkitQueueCoordinator.lambda$enqueue$2(BukkitQueueCoordinator.java:203)
[WARN]: 	at PlotSquared.jar//com.plotsquared.bukkit.queue.BukkitChunkCoordinator.run(BukkitChunkCoordinator.java:187)

Plot Debugpaste

Not attached — this is a code-level report; both lines are unchanged on main (b49bd002) and nothing in the paste would add to it. Happy to provide one if you still want it.

PlotSquared Version

Observed on a 7.5.11-SNAPSHOT build; the offending lines are identical in 7.5.11 (where the line numbers in the trace match exactly), 7.6.0 and current main @ b49bd00.

Checklist

• I have included a Plot debugpaste. (see above)
• I am using the newest build from https://www.spigotmc.org/resources/77506/ and the issue still persists. (verified against the source of main, see above)

Anything else?

Proposed fix (two one-line changes):

--- a/Core/src/main/java/com/plotsquared/core/plot/Plot.java
+++ b/Core/src/main/java/com/plotsquared/core/plot/Plot.java
@@ public @NonNull Set<CuboidRegion> getRegions() {
         if (!this.isMerged()) {
             Location pos1 = this.getBottomAbs().withY(getArea().getMinBuildHeight());
-            Location pos2 = this.getTopAbs().withY(getArea().getMaxBuildHeight());
+            Location pos2 = this.getTopAbs().withY(getArea().getMaxBuildHeight() - 1);
--- a/Bukkit/src/main/java/com/plotsquared/bukkit/queue/BukkitQueueCoordinator.java
+++ b/Bukkit/src/main/java/com/plotsquared/bukkit/queue/BukkitQueueCoordinator.java
@@ for (int layer = 0; layer < localChunk.getBiomes().length; layer++) {
                             int x = sx + ChunkUtil.getX(j);
-                            int y = ChunkUtil.getY(layer, j);
+                            int y = ChunkUtil.getY(layer + localChunk.getMinSection(), j);
                             int z = sz + ChunkUtil.getZ(j);

WorldEdit could additionally guard PaperweightAdapter.setBiome against out-of-range Y (as the 1.21.11 adapter effectively did), but the out-of-range coordinates originate here.