[Bug] Chunk gen NPE when resources are reloaded before safe stronghold data creation #331
Comments
Akiranya
added
Status: Pending
Akiranya
changed the title
Akiranya
changed the title
|
More information: the config of Terra is all default. |
astrsh
added
Note: Help Wanted
|
latest.log |
|
Slightly troubleshooted to determine what the issue was: Was able to determine that issue is not replicatable on the v6.1.2 branch/release. |
|
Troubleshooted further (really just 'disabled half' testing on plugin similarities between affected users). Issue 100% replicatable when both Terra 6.2.0 and CommandAPI (https://www.spigotmc.org/resources/api-commandapi-1-13-1-19-2.62353/) are installed at the same time (no other plugins in combination required.) |
|
@Akiranya Temporary fix is disable CommandAPI. |
This comment was marked as outdated.
This comment was marked as outdated.
|
Not everyone reporting this issue have commandapi installed, fairly sure me and @Akiranya had it due to CommandAPI, but on the discord: 'Barni#6991' reports it with these plugins (AngelChest, AntiCombatLog, AntiNetherRoof*, AugmentedHardcore, BanItem, Breaker, Chunky, ChunkyBorder, Citizens, DiscordSRV, Essentials, EssentialsChat, EssentialsSpawn, LevelledMobs, LuckPerms, MythicMobs, Oraxen, PlaceholderAPI, ProtocolLib, Sentinel, ThirstPlugin, Tweakin, UltimateAutoRestart, WorldEdit, WorldGuard) 'Shiuki#7238' reports it with these plugins (Plugins: { ForcePack v1.2.1 com.convallyria.forcepack.spigot.ForcePackSpigot [SamB440], Skript v2.6.3 ch.njol.skript.Skript [Njol, Mirreski, SkriptLang Team, Contributors], SkBee v2.3.0 com.shanebeestudios.skbee.SkBee [ShaneBee], PL-Hide v1.5.21 me.nononitas.plhide.PLHide [Nononitas], Tether v2.4.1 me.naspo.tether.core.Tether [Naspo], Ships v6.0.0 org.core.implementation.bukkit.platform.plugin.boot.TranslateCoreBoot [], SkQuery v4.1.7 com.w00tmast3r.skquery.SkQuery [LimeGlass, w00tmast3r], VampireRevamp v1.0.BETA-17 com.clanjhoo.vampire.VampireRevamp [Cayorion, Mowstyl], OfflinePlayers v1.0 de.snap20lp.offlineplayers.OfflinePlayers [snap20lp], ProtocolSupport v1.18.2-1-dev protocolsupport.ProtocolSupport [Shevchik, 7kasper, MrPowerGamerBR, gdude2002], ProtocolLib v5.0.0-SNAPSHOT-b600 com.comphenix.protocol.ProtocolLib [dmulloy2, comphenix], Terra v6.2.0-BETA+7d056bd88 com.dfsek.terra.bukkit.TerraBukkitPlugin [dfsek], skRayFall v1.9.28 net.rayfall.eyesniper2.skrayfall.Core [eyesniper2], Sshac v6.19 shiuki.Sshac.Sshac [sh],}) |
|
Hi, I saw this issue via CommandAPI's GitHub and thought I'd check it out. I was surprised that the CommandAPI, which focuses on the command system, could have done anything to disrupt chunk generation like this. Looking into Terra's code by following this error message: it looks like these methods are causing the problem: package com.dfsek.terra.bukkit.nms.v1_19_R1;
public class NMSChunkGeneratorDelegate extends ChunkGenerator {
@Override
public List<ChunkPos> getRingPositionsFor(@NotNull ConcentricRingsStructurePlacement structurePlacement,
@NotNull RandomState noiseConfig) {
ensureStructuresGenerated(noiseConfig);
// error thrown by this line
return ringPositions.get(structurePlacement).value();
}
@Override
public void ensureStructuresGenerated(@NotNull RandomState noiseConfig) {
if(!this.rings) {
super.ensureStructuresGenerated(noiseConfig);
this.populateStrongholdData(noiseConfig);
this.rings = true;
}
}
private void populateStrongholdData(RandomState noiseConfig) {
LOGGER.info("Generating safe stronghold data. This may take up to a minute.");
Set<Holder<Biome>> set = this.biomeSource.possibleBiomes();
possibleStructureSets().map(Holder::value).forEach((holder) -> {
boolean match = false;
for(StructureSelectionEntry structureset_a : holder.structures()) {
Structure structure = structureset_a.structure().value();
Stream<Holder<Biome>> stream = structure.biomes().stream();
if(stream.anyMatch(set::contains)) {
match = true;
}
}
if(match) {
if(holder.placement() instanceof ConcentricRingsStructurePlacement concentricringsstructureplacement) {
this.ringPositions.put(concentricringsstructureplacement, Lazy.lazy(
() -> this.generateRingPositions(holder, noiseConfig, concentricringsstructureplacement)));
}
}
});
}
}Specifically, the error is thrown when I propose that the CommandAPI/other plugins are not causing this issue and it should be replicable with just Terra. I can't test this myself right now, but I plan to try as soon as I can. |
|
I tried replicating this bug but it never crashed for me... I tried on Paper (build 141) and Purpur (build 1777), with and without CommandAPI, teleporting to strongholds and just moving around, but nothing made it crash. I'm using Terra 6.2.0-BETA, CommandAPI 8.5.1, and Minecraft 1.19.2, and I haven't done anything to the Terra config files. Has anyone else been able to trigger the crash with these latest versions? If not I would conclude that this bug was somehow fixed in the latest versions. Any ideas on how to reproduce this for myself? |
|
????? |
|
Huh, maybe you could upload your server directory (or at least a minimal setup) and I could download and run it on my computer exactly as you have it? It'd be incredibly weird if this was OS related, but for the record I'm running Ubuntu Linux. Additionally I don't think there is a minimum-plugin log file posted here yet so that might be helpful. |
|
Replicatable steps as follows: Fresh install - only Terra active, generate world. |
|
Was able to reproduce by following these steps Log for initial world creation https://mclo.gs/ySYR2Zt |
|
Based on this it seems like the NPE is thrown during chunk gen on an existing world after certain plugins have been added |
astrsh
changed the title
|
Hmm... following those steps I still can't get the server to crash. Here's my logs for comparison: CommandAPI test. I noticed that Astrashh had Chunky as well in their logs, so I tried that too, but still no crash: Chunky test. One thing I noticed in both Astrashh's logs that I didn't have was these lines: It seems that I'm missing the step where Terra takes over the overworld generation. If Terra's code isn't taking over for me, it would make sense that nothing special would be happening for me. I was starting from a completely fresh server, with everything from the previous tests deleted, so maybe you all have some leftover data that's affecting it? |
|
@willkroboth have you told Terra/Bukkit to override the regular worldgen in the 'Bukkit.yml' file? You need to add for it to work as the generator. |
|
This may be completely irrelevant, but I thought I'd mention it anyway: the CommandAPI performs a post-server load resources reload. That is, after the server has finished loading, the CommandAPI triggers a modified internal More info can be found here: https://github.com/JorelAli/CommandAPI/blob/master/commandapi-nms/commandapi-1.19-common/src/main/java/dev/jorel/commandapi/nms/NMS_1_19_Common.java#L634 This doesn't explain the cases where this issue can arise without the CommandAPI, but it might help point things in the right direction? (I'm currently on holiday, so I haven't investigated this issue personally, but this may help!) |
|
Oh, I think that may be the culprit |
Of course, as mentioned above, this doesn't explain other plugins (might be worth looking into ProtocolLib? Both logs above that didn't have the CommandAPI did have ProtocolLib). This also leads to whether Terra naturally supports regular |
|
Upon further testing: Initial Setup
Test A
Test B
Test C
Conclusion If a datapack(?) reload occurs before stronghold data is created, chunk generation will throw NPEs (for Terra worlds)
If my conclusion is correct, then this issue will occur with plugins that trigger resources to be reloaded before Terra is able to generate safe stronghold data. This doesn't happen for initial world creation because that seems to trigger the data gen before said plugins reload resources, hence why it only occurs with existing Terra worlds after server start. |
astrsh
changed the title
|
I'm not sure of the possible repercussions of this, but perhaps we could just return Also, I believe what is possibly happening (just from intuition; I haven't looked at it in a debugger), is that for some reason the reloading is constructing new instances of This may also be a race condition that is, for some reason, causing |
|
Yeah, it might be a race condition that's happening because the pack reload has been triggered, but it hasn't finished so the structures aren't populated properly (?), causing |
|
Actually, it could be these lines which is causing changes to the structure manager. |
|
I tried to solve this issue by returning null when the lazy is null, like this: @Override
public List<ChunkPos> getRingPositionsFor(@NotNull ConcentricRingsStructurePlacement structurePlacement,
@NotNull RandomState noiseConfig) {
ensureStructuresGenerated(noiseConfig);
Lazy<List<ChunkPos>> lazy = ringPositions.get(structurePlacement);
return lazy == null ? null : lazy.value();
}Looking at the method this overrides, I figured this would probably be fine: @Nullable
public List<ChunkPos> getRingPositionsFor(ConcentricRingsStructurePlacement structurePlacement, RandomState noiseConfig) {
this.ensureStructuresGenerated(noiseConfig);
CompletableFuture<List<ChunkPos>> completablefuture = (CompletableFuture) this.ringPositions.get(structurePlacement);
return completablefuture != null ? (List) completablefuture.join() : null;
}While this stops the NPE from occurring, the game instead can't figure out where strongholds are supposed to be when CommandAPI is on the server. Even with this fix, there still seems to be an underlying problem where the |
|
One thing I noticed in my testing is if ring structure generation works correctly (case when CommandAPI isn't on the server), with the exception of the first time the server generates the world, the safe stronghold data generation always happens when the server is ticking and serving players. This causes a large lag spike, potentially timing out players. If this issue ends up changing the stronghold data generation process, it might be good to always generate the data while the server is starting up to avoid this lag spike. Additionally, I'm having some weird exceptions when adding CommandAPI, attempting to generate new chunks, stopping the server, removing CommandAPI, then joining the server again when using my lazy-null-null patch from my previous comment. [07:10:49 ERROR]: Error whilst processing packet net.minecraft.network.protocol.game.PacketPlayInBlockPlace@50d9a14d for willkroboth[/127.0.0.1:58238]
java.lang.IllegalStateException: Somehow tried to find structures for a placement that doesn't exist
at net.minecraft.world.level.chunk.ChunkGenerator.getNearestGeneratedStructure(ChunkGenerator.java:410) ~[paper-1.19.2.jar:git-Paper-141]
at net.minecraft.world.level.chunk.ChunkGenerator.findNearestMapStructure(ChunkGenerator.java:355) ~[paper-1.19.2.jar:git-Paper-141]
at net.minecraft.server.level.ServerLevel.findNearestMapStructure(ServerLevel.java:1904) ~[?:?]
at net.minecraft.world.item.EnderEyeItem.use(EnderEyeItem.java:103) ~[?:?]
at net.minecraft.world.item.ItemStack.use(ItemStack.java:497) ~[paper-1.19.2.jar:git-Paper-141]
at net.minecraft.server.level.ServerPlayerGameMode.useItem(ServerPlayerGameMode.java:472) ~[?:?]
at net.minecraft.server.network.ServerGamePacketListenerImpl.handleUseItem(ServerGamePacketListenerImpl.java:2046) ~[?:?]
at net.minecraft.network.protocol.game.ServerboundUseItemPacket.handle(ServerboundUseItemPacket.java:32) ~[?:?]
at net.minecraft.network.protocol.game.ServerboundUseItemPacket.a(ServerboundUseItemPacket.java:8) ~[?:?]
at net.minecraft.network.protocol.PacketUtils.lambda$ensureRunningOnSameThread$1(PacketUtils.java:51) ~[?:?]
at net.minecraft.server.TickTask.run(TickTask.java:18) ~[paper-1.19.2.jar:git-Paper-141]
at net.minecraft.util.thread.BlockableEventLoop.doRunTask(BlockableEventLoop.java:153) ~[?:?]
at net.minecraft.util.thread.ReentrantBlockableEventLoop.doRunTask(ReentrantBlockableEventLoop.java:24) ~[?:?]
at net.minecraft.server.MinecraftServer.doRunTask(MinecraftServer.java:1361) ~[paper-1.19.2.jar:git-Paper-141]
at net.minecraft.server.MinecraftServer.d(MinecraftServer.java:185) ~[paper-1.19.2.jar:git-Paper-141]
at net.minecraft.util.thread.BlockableEventLoop.pollTask(BlockableEventLoop.java:126) ~[?:?]
at net.minecraft.server.MinecraftServer.pollTaskInternal(MinecraftServer.java:1338) ~[paper-1.19.2.jar:git-Paper-141]
at net.minecraft.server.MinecraftServer.pollTask(MinecraftServer.java:1331) ~[paper-1.19.2.jar:git-Paper-141]
at net.minecraft.util.thread.BlockableEventLoop.runAllTasks(BlockableEventLoop.java:114) ~[?:?]
at net.minecraft.server.MinecraftServer.tickServer(MinecraftServer.java:1465) ~[paper-1.19.2.jar:git-Paper-141]
at net.minecraft.server.MinecraftServer.runServer(MinecraftServer.java:1193) ~[paper-1.19.2.jar:git-Paper-141]
at net.minecraft.server.MinecraftServer.lambda$spin$0(MinecraftServer.java:305) ~[paper-1.19.2.jar:git-Paper-141]
at java.lang.Thread.run(Thread.java:833) ~[?:?]
[07:10:49 INFO]: willkroboth lost connection: Packet processing errorI get a bunch of these when stopping the server, which seems to prevent the server process from finishing properly: [07:11:46 ERROR]: Failed to schedule load callback for chunk [3, -18]
java.util.concurrent.CompletionException: java.lang.IllegalStateException: zip file closed
at java.util.concurrent.CompletableFuture.encodeThrowable(CompletableFuture.java:315) ~[?:?]
at java.util.concurrent.CompletableFuture.completeThrowable(CompletableFuture.java:320) ~[?:?]
at java.util.concurrent.CompletableFuture$AsyncSupply.run(CompletableFuture.java:1770) ~[?:?]
at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1136) ~[?:?]
at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:635) ~[?:?]
at java.lang.Thread.run(Thread.java:833) ~[?:?]
Caused by: java.lang.IllegalStateException: zip file closed
at java.util.zip.ZipFile.ensureOpen(ZipFile.java:831) ~[?:?]
at java.util.zip.ZipFile.getEntry(ZipFile.java:330) ~[?:?]
at java.util.jar.JarFile.getEntry(JarFile.java:518) ~[?:?]
at java.util.jar.JarFile.getJarEntry(JarFile.java:473) ~[?:?]
at org.bukkit.plugin.java.PluginClassLoader.findClass(PluginClassLoader.java:163) ~[paper-api-1.19.2-R0.1-SNAPSHOT.jar:?]
at java.lang.ClassLoader.loadClass(ClassLoader.java:587) ~[?:?]
at org.bukkit.plugin.java.PluginClassLoader.loadClass0(PluginClassLoader.java:108) ~[paper-api-1.19.2-R0.1-SNAPSHOT.jar:?]
at org.bukkit.plugin.java.PluginClassLoader.loadClass(PluginClassLoader.java:103) ~[paper-api-1.19.2-R0.1-SNAPSHOT.jar:?]
at java.lang.ClassLoader.loadClass(ClassLoader.java:520) ~[?:?]
at com.dfsek.terra.bukkit.generator.BukkitChunkGeneratorWrapper.generateNoise(BukkitChunkGeneratorWrapper.java:67) ~[Terra-bukkit-6.2.0-BETA+e8fa5514c-shaded.jar:?]
at org.bukkit.craftbukkit.v1_19_R1.generator.CustomChunkGenerator.lambda$fillFromNoise$0(CustomChunkGenerator.java:254) ~[paper-1.19.2.jar:git-Paper-141]
at org.bukkit.craftbukkit.v1_19_R1.generator.CustomChunkGenerator.lambda$fillFromNoise$1(CustomChunkGenerator.java:259) ~[paper-1.19.2.jar:git-Paper-141]
at java.util.concurrent.CompletableFuture$AsyncSupply.run(CompletableFuture.java:1768) ~[?:?]
... 3 moreHere's my logs for this: https://paste.gg/p/anonymous/5db87fd36cc64c5cb2cea644dd1013f8 |
|
Huh, I actually got those two errors to happen with an unmodified version of Terra as the only plugin. Here's the log: https://paste.gg/p/anonymous/b088f9627abd427b9719a45d6d995135. It seems the pack processing kick happens when an eye of ender is thrown before the safe stronghold generation is complete. The chunk load callback error seems to happen when the server is stopped while chunks are being generated. I seem to have found two new bugs, but it'd be good if someone could replicate this separately. |
|
Paper has recently rewritten it's chuck generation system. As far as I can tell, everything is working the same (same bugs), but noticeably faster. |
|
For now, let's focus on fixing the issue on the old chunk gen system, before looking at the new one, as hopefully a fix for the old one should fix the new one as well. |
|
I'm gonna try and look at the mc sources later to see exactly how mc handles |
|
Was this ever fixed? I have been having this issue on version 6.2.0 |
|
Generating chunks with Chunky always results in server crashing |
Pre-Issue Checklist
closed ones.
is Terra that is causing the issue.
pack I am using.
latest.logfileEnvironment
Issue Description
Hi, I'm new to Terra and I encountered some problem with Terra (really love this plugin). When generating terrain using the plugin Chunky or just moving around, the errors randomly appear. The config of Terra is all default. I searched in the discord and I found another user has the same issue as mine (https://discord.com/channels/715448651786485780/765260067812540416/990301360274669618)
Steps to reproduce
Create a new world using Terra and move around / use Chunky to generate the world. To better reproduce the error, restart the server and continue the chunk generation might help. Sometimes it does not appear but once appear the error is just so long...
Expected behavior
There should be no errors...
Actual behavior
When generating terrain using the plugin Chunky or just moving around, the errors randomly appear.
Updated 7/7/2022: Server crash log -> https://pastebin.com/6VHj33xq (sorry I forgot to put it here in the first place)
Full stacktrace
latest.log: https://pastebin.com/RRybGKv1
Exception Stacktrace
Additional details
The text was updated successfully, but these errors were encountered: