Unloading a world does not stop raids.dat from being saved.

[MeltPels]

Expected behavior

When calling:
boolean success = Bukkit.unloadWorld(world, false);
and receiving success = true, we expect the world to be unloaded without being saved.
This means the world will be unloaded as is, without first saving any of it's files.

Observed/Actual behavior

When unloading via
boolean success = Bukkit.unloadWorld(world, false);
the world gets unloaded correctly. But not before the server tries to save raids.dat in the incorrect location. The server thinks the world is in its root.

I've observed this by not having the world folder in the server's root folder and am getting the following error for a world stored in /plugins/pluginname/worlds/MZ:
[09:54:57 ERROR]: Could not save data to raids.dat java.nio.file.NoSuchFileException: ./MZ/data/raids.dat at java.base/sun.nio.fs.UnixException.translateToIOException(UnixException.java:92) ~[?:?] at java.base/sun.nio.fs.UnixException.rethrowAsIOException(UnixException.java:106) ~[?:?] at java.base/sun.nio.fs.UnixException.rethrowAsIOException(UnixException.java:111) ~[?:?] at java.base/sun.nio.fs.UnixFileSystemProvider.newByteChannel(UnixFileSystemProvider.java:261) ~[?:?] at java.base/java.nio.file.spi.FileSystemProvider.newOutputStream(FileSystemProvider.java:482) ~[?:?] at java.base/java.nio.file.Files.newOutputStream(Files.java:227) ~[?:?] at net.minecraft.nbt.NbtIo.writeCompressed(NbtIo.java:76) ~[paper-1.21.4.jar:1.21.4-188-8de7e35] at net.minecraft.world.level.storage.DimensionDataStorage.tryWrite(DimensionDataStorage.java:185) ~[paper-1.21.4.jar:1.21.4-188-8de7e35] at net.minecraft.world.level.storage.DimensionDataStorage.lambda$scheduleSave$3(DimensionDataStorage.java:163) ~[paper-1.21.4.jar:1.21.4-188-8de7e35] at java.base/java.util.concurrent.CompletableFuture$AsyncRun.run(CompletableFuture.java:1804) ~[?:?] at java.base/java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1144) ~[?:?] at java.base/java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:642) ~[?:?] at java.base/java.lang.Thread.run(Thread.java:1583) ~[?:?]

Steps/models to reproduce

Create a world via a plugin, that saves the world in a folder other than the server's root folder.
(So /plugins/pluginfolder/worlds)
IN GAME: hop into the world to load some chunks, teleport out to a different loaded world.
Unload the world with the 'save: false' flag.
View the console.

Plugin and Datapack List

Paper 1.21.4 (latest release)
Plugins: only the one with the behaviour mentioned above.

Paper version

[10:00:56 INFO]: Checking version, please wait...
[10:00:56 INFO]: This server is running Paper version 1.21.4-188-main@8de7e35 (2025-03-03T16:59:40Z) (Implementing API version 1.21.4-R0.1-SNAPSHOT)
You are running the latest version

Other

I'm also deleting the world, but only after three seconds post unload. The issue appears to originate from ChunkHolderManager. Here's a sequence of events triggered by command:

[09:54:57 INFO]: [[Maze]] Preparing to unload world: MZ [09:54:57 ERROR]: Could not save data to raids.dat java.nio.file.NoSuchFileException: ./MZ/data/raids.dat at java.base/sun.nio.fs.UnixException.translateToIOException(UnixException.java:92) ~[?:?] at java.base/sun.nio.fs.UnixException.rethrowAsIOException(UnixException.java:106) ~[?:?] at java.base/sun.nio.fs.UnixException.rethrowAsIOException(UnixException.java:111) ~[?:?] at java.base/sun.nio.fs.UnixFileSystemProvider.newByteChannel(UnixFileSystemProvider.java:261) ~[?:?] at java.base/java.nio.file.spi.FileSystemProvider.newOutputStream(FileSystemProvider.java:482) ~[?:?] at java.base/java.nio.file.Files.newOutputStream(Files.java:227) ~[?:?] at net.minecraft.nbt.NbtIo.writeCompressed(NbtIo.java:76) ~[paper-1.21.4.jar:1.21.4-188-8de7e35] at net.minecraft.world.level.storage.DimensionDataStorage.tryWrite(DimensionDataStorage.java:185) ~[paper-1.21.4.jar:1.21.4-188-8de7e35] at net.minecraft.world.level.storage.DimensionDataStorage.lambda$scheduleSave$3(DimensionDataStorage.java:163) ~[paper-1.21.4.jar:1.21.4-188-8de7e35] at java.base/java.util.concurrent.CompletableFuture$AsyncRun.run(CompletableFuture.java:1804) ~[?:?] at java.base/java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1144) ~[?:?] at java.base/java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:642) ~[?:?] at java.base/java.lang.Thread.run(Thread.java:1583) ~[?:?] [09:54:57 INFO]: [ChunkHolderManager] Waiting 60s for chunk system to halt for world 'MZ' [09:54:57 INFO]: [ChunkHolderManager] Halted chunk system for world 'MZ' [09:54:57 INFO]: [ChunkHolderManager] Waiting 60s for chunk I/O to halt for world 'MZ' [09:54:57 INFO]: [ChunkHolderManager] Halted I/O scheduler for world 'MZ' [09:54:57 INFO]: [[Maze]] Successfully unloaded world: MZ [09:55:05 INFO]: [[Maze]] Deleted temporary world folder: MZ

[Warriorrrr]

How did you get the world folder into your plugin's directory? That seems to be the important thing here but you haven't provided any code for it.

[electronicboy]

there is a fun annoying issue where you can technically shove a folder path into create world and it works, we should probably figure out how we want to handle that, but, that's an issue for another time.

The issue here is that Mojang doesn't support multiple 'worlds' in the same way that the API pretends it does, and a lot of the per-dimension data is basically a global state as far as the server is really concerned. There would need to be a huge effort to clean this up, but I feel that one of the first goals before we start fighting this would be to remove spigots, half-maintained world dimension changes and revert back to vanilla logic here, but that's another entire issue to contend with...

[MeltPels]

This is how I'm creating a new world:

`Bukkit.getScheduler().runTaskAsynchronously(PluginName.getInstance(), () -> {

        File tempWorldsFolder = new File(Bukkit.getWorldContainer(), "plugins/PluginName/worlds/");
        File worldFolder = new File(Bukkit.getWorldContainer(), worldName);

        if (!tempWorldsFolder.exists()) {
            tempWorldsFolder.mkdirs();
        }

        Bukkit.getScheduler().runTask(PluginName.getInstance(), () -> {
            WorldCreator creator = new WorldCreator(worldName);
            creator.environment(World.Environment.NORMAL);
            creator.generator(new Generator()); // Generic flat world generator as placeholder
            creator.seed(new Random().nextLong());

            World world = Bukkit.createWorld(creator);

            if (world != null) {
                world.setAutoSave(false); // Attempt to prevent saving these worlds

                // Move the world to the temporary directory immediately
                File targetFolder = new File(tempWorldsFolder, worldName);
                if (worldFolder.exists() && worldFolder.isDirectory()) {
                    worldFolder.renameTo(targetFolder);
                }
            } else {
                sender.sendMessage("Failed to create maze world: " + worldName);
            }
        });
    });`

So basically, I'm creating it and then moving it. I wasn't aware I could shove a folder path into world creation.
Maybe I'm mistaken, I haven't seen papers' code handling this, but moving the world is (i believe) only the reason I noticed it's saving raids.dat. It's the cause for the error, yes, but not the cause for the bug report.

The bug being that unloading a world with 'save:false' flag, will still save raids.dat.

[electronicboy]

raids.dat is not unloaded, that has been the case for years, this is well known issue with how world management was implemented; unloading a world without saving is also in a really weird area in which that just controls if the chunk source is asked to save chunks during the process, it does not control other aspects of saving; Mojang has 0 mechanms of allowing for read only worlds