Mojang has flaws in their logic about chunks being concurrently
wrote to. So we constantly see crashes around multiple threads writing.
Additionally, java has optimized synchronization so well that its
in many times faster than trying to manage read write locks for low
contention situations.
And this is extremely a low contention situation.
When the active item is not cleared, the item is still readied
for use and will repeatedly trigger the PlayerItemConsumeEvent
till their item is switched.
This patch clears the active item when the event is cancelled
Removes a layer of indirection for EnumDirection.getAdjacent(X|Y|Z)(), which is in the
critical section for much of the server, including the lighting engine.
RegistryAccess is independant from CraftServer and
doesn't require one to be created allowing the
org.bukkit.Registry class to be loaded earlier.
== AT ==
public net.minecraft.server.RegistryLayer STATIC_ACCESS
== AT ==
public net.minecraft.world.entity.monster.Vindicator DOOR_BREAKING_PREDICATE
public net.minecraft.world.entity.monster.Zombie DOOR_BREAKING_PREDICATE
Co-authored-by: Doc <nachito94@msn.com>
* Call PlayerInteractEvent when left-clicking on a block in adventure
mode.
* Call PlayerInteractEvent when left-clicking an Entity that is out of
range in adventure/survival (entity reach is 3.0).
Co-authored-by: Moulberry <james.jenour@protonmail.com>
The maps did convert successfully, but had noisy logs due to Spigot
implementing this logic incorrectly.
This stops the spam by converting the old format to new before
requesting the world.
Track spigot issue to see when fixed: https://hub.spigotmc.org/jira/browse/SPIGOT-6181
This fixes item position desync (MC-4) by running the item coordinates
through the encode/decode methods of the packet that causes the precision
loss, which forces the server to lose the same precision as the client
keeping them in sync.
get was calling toString() on every NBT object that was ever asked for an optional
key from the object to build a string for the error text.
When done on large NBT objects, this was using a ton of computation time building the
JSON representation of the NBT object.
Now we will just skip the value when 99.9999% of the time the text is never even printed.
This patch implements a noise cache for the end which significantly reduces the computation time of generation. This results in about a 3x improvement.
Original code by SuperCoder7979 and Gegy in Lithium, licensed under LGPL-3.0 (Source: https://github.com/jellysquid3/lithium-fabric)
Co-authored-by: Gegy <gegy1000@gmail.com>
Co-authored-by: Dylan Xaldin <Puremin0rez515@gmail.com>
Co-authored-by: pop4959 <pop4959@gmail.com>
Adds config options for modifying the spawn rates of Wandering Traders.
These values are all easy to understand and configure after a quick read of this
page on the Minecraft wiki: https://minecraft.wiki/wiki/Wandering_Trader#Spawning
Usages of the vanilla WanderingTraderSpawnDelay and WanderingTraderSpawnChance values
in IWorldServerData are removed as they were only used in certain places, with hardcoded
values used in other places.
BellBlockEntity has a list of entities (entitiesAtRing) that was not being cleared at the right time, causing leaks whenever a bell would be rung near a crowd of entities.
Fixes twisted vines not connecting properly when grown via
bonemeal by a player.
In general, look at making this logic more robust (i.e properly handling
cases where a captured entry is overriden) - but for now this will do.
Crystals can end up in a bad state in certain conditions which causes
an exception on the expected number of crystals going negative.
This ensures the crystals/pillars are in expected state when the dragon spawns.
See #3522
Mojang precaches every single potential rewrite rule that could ever
exist on server startup. This includes rules from all the way back to versions from 6+ years ago.
This is the source of why the server hogs every CPU core at 100% every start.
For anyone who hard resets for updates or has force upgraded their entire world, this
results in completely wasted cpu cycles.
This massive CPU usage also delays server startup time.
We improve this by making "min version to precache" that defaults to a future version
so that no rewrite rules are precached.
someone who expects to be converting a lot chunks could theoretically set
-DPaper.minPrecachedDatafixVersion=<dataVersionConvertingFrom> as a startup
parameter and only build from that point on.
However this will likely never be needed as the server will still run
the same cache logic on demand when it's actually needed. The only
cost would be some delay on the FIRST chunk conversion, but paper already
runs chunk conversions on another thread so this will likely never be
a concern for TPS.
This patch will significantly reduce CPU use on startup, reduce memory usage,
and improve server startup time.
Uses correct setPositionRotation for Entity teleporting instead of setLocation
as this is how Vanilla teleports entities.
Cancel any pending motion when teleported.
This patch buffers the number of logins which will attempt to join
the world per tick, this attempts to reduce the impact that join floods
has on the server
If the server attempts to load a chunk generated by a newer version of
the game, immediately stop the server to prevent data corruption.
You can override this functionality at your own peril.
Datapacks check this on load and are built concurrently. This was breaking them badly due
to race conditions.
Plus, .canUse we want to be safe for async anyways.
This forces the despawn counter to start ticking regardless of
state after the arrow has been alive for 200 ticks (10 seconds)
instead of getting stuck in a never despawn state (bubble columns,
etc).
Fixes per world difficulty with /difficulty command and also
makes it so that the server keeps the last difficulty used instead
of restoring the server.properties every single load.
Caused the server to revert to the player's overworld coordinates
after teleporting into the end.
Sidenote: The underlying issue is that the move call can teleport
entities and do other things like kill the entity. In the future,
to fix all exploits derieved from this usually unexpected
behaviour, we need to move all of this dangerous logic outside
of the move call and into an appropriate place in the tick method.
This patch fixes a bug in the EndIslandDensityFunction class where the distance
from 0,0 squared overflows the maximum size of an integer. The overflow leads
to hard chunk borders around 370,000 blocks from 0,0. After this cutoff there
is a few hundred thousand block gap before end land resuming to generate at
530,000 blocks from spawn. This is due to the integer flipping back and forth.
The fix for the issue is quite simple, casting chunk coordinates to longs
allows the distance calculation to avoid overflow and work as intended.
This issue is being tracked in Mojira ticket MC-159283
Pistons invoke physics when they move blocks. The physics can cause
tnt blocks to ignite. However, pistons (when storing the blocks they "moved")
don't actually go back to the world state sometimes to check if something
like that happened. As a result they end up moving the tnt like it was
never ignited. This resulted in the ability to create machines
that can duplicate tnt, called "world eaters".
This patch makes the piston logic retrieve the block state from the world
prevent this from occuring.
This patch also sets the moved pos to air immediately after creating
the moving piston TE. This prevents the block from being updated from
other physics calls by the piston.
Tested against the following tnt duper design:
https://www.youtube.com/watch?v=mS7xxNGhjxs
This patch also affects every type of machine that utilises
this mechanic. For example, dead coral is removed by a physics
update when being moved while it is attached to slimeblocks.
Standard piston machines that don't destroy or modify the
blocks they move by physics updates should be entirely
unaffected.
This patch fixes https://bugs.mojang.com/browse/MC-188840
This patch also fixes rail duping and carpet duping.
Previously maps would load all chunks in a certain radius depending on
their scale when trying to update their content. This would result in
main thread chunk loads when they weren't really necessary, especially
on low view distances or "slow" async chunk loads after teleports or
other prioritisation.
This changes it to only try to render already loaded chunks based on
the assumption that the chunks around the player will get loaded
eventually anyways and that maps will get checked for update every
five ticks that movement occur in anyways.
Server.reload() had this logic to give time for tasks to shutdown,
however shutdown did not...
Adds a 5 second grace period for any async tasks to finish and warns
if any are still running after that delay just as reload does.
Adds an option for maximum exp value when merging orbs
Adds ExperienceOrbMergeEvent
Fired when the server is about to merge 2 experience orbs
as entities. Plugins can cancel it if they want to ensure experience orbs do not lose important
metadata such as spawn reason, or conditionally move data from source to target.
Fixes an issue where the stacked count was not taking into account
for mending repairs and when merging with spigot's merge-on-spawn
logic
== AT ==
public net.minecraft.world.entity.ExperienceOrb count
Co-authored-by: Aikar <aikar@aikar.co>
Co-authored-by: Jake Potrebic <jake.m.potrebic@gmail.com>
== AT ==
public net.minecraft.world.entity.ai.gossip.GossipContainer$EntityGossips
public net.minecraft.world.entity.ai.gossip.GossipContainer$EntityGossips <init>()V
public net.minecraft.world.entity.ai.gossip.GossipContainer gossips
This notably fixes the newest "Donkey Dupe", but also fixes a lot
of dupe bugs in general around nether portals and entity world transfer
We also fix item duplication generically by anytime we clone an item
to drop it on the ground, destroy the source item.
This avoid an itemstack ever existing twice in the world state pre
clean up stage.
So even if something NEW comes up, it would be impossible to drop the
same item twice because the source was destroyed.
If the playerdata contains an invalid world (missing, unloaded, invalid,
etc.), spawn the player at the spawn point of the main world.
Co-authored-by: Wyatt Childers <wchilders@nearce.com>
Co-authored-by: Jake Potrebic <jake.m.potrebic@gmail.com>
For years, plugin developers have had to delay many things they do
inside of the PlayerJoinEvent by 1 tick to make it actually work.
This all boiled down to 1 reason why: The event fired before the
player was fully ready and joined to the world!
Additionally, if that player logged out on a vehicle, the event
fired before the vehicle was even loaded, so that plugins had no
access to the vehicle during this event either.
This change finally fixes this issue, fully preparing the player
into the world as a fully ready entity, vehicle included.
There should be no plugins that break because of this change, but might
improve consistency with other plugins instead.
For example, if 2 plugins listens to this event, and the first one
teleported the player in the event, then the 2nd plugin actually
would be getting a valid player!
This was very non deterministic. This change will ensure every plugin
receives a deterministic result, and should no longer require 1 tick
delays anymore.
== AT ==
public net.minecraft.server.level.ChunkMap addEntity(Lnet/minecraft/world/entity/Entity;)V
Will not run if:
Max entity cramming is disabled and the max collisions per entity is less than or equal to 0.
Entity#isPushable() returns false, meaning all entities will not be able to collide with this
entity anyways.
The entity's current team collision rule causes them to NEVER collide.
Co-authored-by: Owen1212055 <23108066+Owen1212055@users.noreply.github.com>
Fastutil maps are going to have a lower memory footprint - which
is important because we clone chunk data after reading it for safety.
So, reduce the impact of the clone on GC.
This can cause a nasty server lag the spawn chunks are not kept loaded
or they aren't finished loading yet, or if the world spawn radius is
larger than the keep loaded range.
By skipping this, we avoid potential for a large spike on server start.
== AT ==
public net.minecraft.server.level.ServerPlayer fudgeSpawnLocation(Lnet/minecraft/server/level/ServerLevel;)V
This adds config options for defining the spawn chance, spawn delay and
spawn start day as well as toggles for handling the spawn delay and
start day per player. (Based on the time played statistic)
When not per player it will use the Vanilla mechanic of one delay per
world and the world age for the start day.
Removing the try catch and generally reducing ops should make it
faster on its own, however removing the try catch makes it
easier to inline due to code size
This allows you to solve an issue in vanilla behavior where:
* On easy difficulty your villagers will NEVER get infected, meaning they will always die.
* On normal difficulty they will have a 50% of getting infected or dying.
This patch adds a tool to find calls to getChunkAt which would load
chunks, however it must be enabled by setting the startup flag
-Dpaper.debug-sync-loads=true
- To get a debug log for sync loads, the command is
/paper syncloadinfo
- To clear clear the currently stored sync load info, use
/paper syncloadinfo clear
Adds bool parameter to play world effect on block break
Adds bool parameter to drop xp from blocks
Fixes fluid-logged blocks not leaving fluid behind if
broken
Handles special cases for ice and turtle eggs
== AT ==
public net.minecraft.world.level.block.TurtleEggBlock decreaseEggs(Lnet/minecraft/world/level/Level;Lnet/minecraft/core/BlockPos;Lnet/minecraft/world/level/block/state/BlockState;)V
Co-authored-by: William Blake Galbreath <Blake.Galbreath@GMail.com>
Item entities only have their gravity ticked every 4 ticks when on ground.
Fix that and also remove Spigot's arbitrary tick skipping. It's a terribly
cheap way of getting extra performance that doesn't really work at all.
Hoppers call this to I guess "get the primary side" of a double sided chest.
If the double sided chest crosses chunk lines, it causes the chunk to load.
This will end up causing sync chunk loads, which will unload with Chunk GC,
only to be reloaded again the next tick.
This of course is undesirable, so just return the loaded side as "primary"
and treat it as a single chest if the other sides are unloaded
Due to a bug in 2e29af3df0
which was added all the way back in March of 2016, it was unknown (potentially not at the time)
that an entity might actually change the seed of the random object.
At some point, EntitySquid did start setting the seed. Due to this shared random, this caused
every entity to use a Random object with a predictable seed.
This has caused entities to potentially generate with the same UUID....
Over the years, servers have had entities disappear, but no sign of trouble
because CraftBukkit removed the log lines indicating that something was wrong.
We have fixed the root issue causing duplicate UUID's, however we now have chunk
files full of entities that have the same UUID as another entity!
When these chunks load, the 2nd entity will not be added to the world correctly.
If that chunk loads in a different order in the future, then it will reverse and the
missing one is now the one added to the world and not the other. This results in very
inconsistent entity behavior.
This change allows you to recover any duplicate entity by generating a new UUID for it.
This also lets you delete them instead if you don't want to risk having new entities added to
the world that you previously did not see.
But for those who are ok with leaving this inconsistent behavior, you may use WARN or NOTHING options.
It is recommended you regenerate the entities, as these were legit entities, and deserve your love.
This patch adds an option "disable relative projectile velocity", which, when
enabled, will cause projectiles to ignore the shooter's current velocity,
like they did in Minecraft 1.8 and prior.
If a player is falling, for example, their shooting range will be drastically
reduced, as a downwards velocity is applied to the projectile. This prevents
players from saving themselves from falling off floating islands, for example,
as a thrown ender pearl will not make it back to the island, while it would
have in 1.8.
While this could easily be done with plugins, too, there are multiple problems:
P1) If multiple plugins cancel the velocity by subtracting the shooter's velocity
from the projectile's velocity, the projectile's velocity would be different.
As there's no way to detect whether the projectile's velocity has already been
adjusted to ignore the player's velocity, plugins can't not do it if it's not
necessary.
P2) I've noticed some inconsistencies, e.g. weird velocity when shooting while
using an elytra. Checking for those inconsistencies is possible, but not as
efficient as just not applying the velocity in the first place.
P3) Solutions for 1) and especially 2) might not be future-proof, while this
server-internal fix makes this change future-proof.
This resolves the super common complaint about mobs not spawning.
This was ultimately a flaw in the vanilla count algorithim that allows
spawners and other misc mobs to count against the mob limit, which are
not bounded, and can prevent the entire world from spawning new.
I believe Bukkits changes around persistence may of actually made it
worse than vanilla.
This should fully solve all of the issues around it so that only natural
influences natural spawns.
== AT ==
public net.minecraft.world.level.BaseSpawner isNearPlayer(Lnet/minecraft/world/level/Level;Lnet/minecraft/core/BlockPos;)Z
public net.minecraft.world.level.BaseSpawner delay(Lnet/minecraft/world/level/Level;Lnet/minecraft/core/BlockPos;)V
public net.minecraft.world.level.BaseSpawner setNextSpawnData(Lnet/minecraft/world/level/Level;Lnet/minecraft/core/BlockPos;Lnet/minecraft/world/level/SpawnData;)V
upstream was doing a containsKey/get pattern, and always doing it at that.
that scenario is only even valid if were in the middle of a block place.
Optimize to check if the captured list even has values in it, and also to
just do a get call since the value can never be null.
Expose an entities spawn reason on the entity.
Pre existing entities will return NATURAL if it was a non
persistenting Living Entity, SPAWNER for spawners,
or DEFAULT since data was not stored.
Additionally, add missing spawn reasons.
Co-authored-by: Jake Potrebic <jake.m.potrebic@gmail.com>
Co-authored-by: Doc <nachito94@msn.com>
modified clients can send more data from the client
to the server and it would get stored on the sign as sent.
Mojang has a limit of 384 which is much higher than reasonable.
the client can barely render around 16 characters as-is, but formatting
codes can get it to be more than 16 actual length.
Set a limit of 80 which should give an average of 16 characters 2
sets of legacy formatting codes which should be plenty for all uses.
This does not strip any existing data from the NBT as plugins
may use this for storing data out of the rendered area.
it only impacts data sent from the client.
Set -DPaper.maxSignLength=XX to change limit or -1 to disable
Adds AsyncPlayerSendCommandsEvent
- Allows modifying on a per command basis what command data they see.
Adds CommandRegisteredEvent
- Allows manipulating the CommandNode to add more children/metadata for the client
This adds a custom pool inorder to make sure that they are closed
without much though, as it doesn't matter if the client is not sent
commands if the server is restarting. Using the default async pool caused issues to arise
due to the shutdown logic generally being much later.
Adds an event for when the server is going to destroy a current block,
potentially causing it to drop. This event can be cancelled to avoid
the block destruction, such as preventing signs from popping when
floating in the air.
This can replace many uses of BlockPhysicsEvent
Currently OfflinePlayer#getLastPlayed could more accurately be described
as "OfflinePlayer#getLastTimeTheirDataWasSaved".
The API doc says it should return the last time the server "witnessed"
the player, whilst also saying it should return the last time they
logged in. The current implementation does neither.
Given this interesting contradiction in the API documentation and the
current defacto implementation, I've elected to deprecate (with no
intent to remove) and replace it with two new methods, clearly named and
documented as to their purpose.
This event is invoked when a player has disconnected. It is guaranteed that,
if the server is in online-mode, that the provided uuid and username have been
validated.
The event is invoked for players who have not yet logged into the world, whereas
PlayerQuitEvent is only invoked on players who have logged into the world.
The event is invoked for players who have already logged into the world,
although whether or not the player exists in the world at the time of
firing is undefined. (That is, whether the plugin can retrieve a Player object
using the event parameters is undefined). However, it is guaranteed that this
event is invoked AFTER PlayerQuitEvent, if the player has already logged into
the world.
This event is guaranteed to never fire unless AsyncPlayerPreLoginEvent has
been called beforehand, and this event may not be called in parallel with
AsyncPlayerPreLoginEvent for the same connection.
Cancelling the AsyncPlayerPreLoginEvent guarantees the corresponding
PlayerConnectionCloseEvent is never called.
The event may be invoked asynchronously or synchronously. As it stands,
it is never invoked asynchronously. However, plugins should check
Event#isAsynchronous to be future-proof.
On purpose, the deprecated PlayerPreLoginEvent event is left out of the
API spec for this event. Plugins should not be using that event, and
how PlayerPreLoginEvent interacts with PlayerConnectionCloseEvent
is undefined.
== AT ==
public net.minecraft.server.network.ServerLoginPacketListenerImpl$State
public net.minecraft.server.network.ServerLoginPacketListenerImpl state
Entities must be dismounted before teleportation in order to avoid
multiple issues in the server with regards to teleportation, shamefully,
too many plugins rely on the events firing, which means that not firing
these events caues more issues than it solves;
In order to counteract this, Entity dismount/exit vehicle events have
been modified to supress cancellation (and has a method to allow plugins
to check if this has been set), noting that cancellation will be silently
surpressed given that plugins are not expecting this event to not be cancellable.
This is a far from ideal scenario, however: given the current state of this
event and other alternatives causing issues elsewhere, I believe that
this is going to be the best soultion all around.
Improvements/suggestions welcome!
Splits time updates into incremental updates as well as does
the updates per world, so that we can re-use the same packet
object for every player unless they have per-player time enabled.
Use a simple executor since Fork join is a much more complex pool
type and we are not using its capabilities.
Set thread priorities so main thread has above normal priority over
server threads
Allow usage of a single thread executor by not using ForkJoin so single core CPU's
and reduce worldgen thread worker count for low core count CPUs.
== AT ==
public net.minecraft.Util onThreadException(Ljava/lang/Thread;Ljava/lang/Throwable;)V
Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com>
Use a proper teleport for teleporting to entities in different
worlds.
Implementation improvements authored by Spottedleaf <Spottedleaf@users.noreply.github.com>
Validate that the target entity is valid and deny spectate
requests from frozen players.
Also, make sure the entity is spawned to the client before
sending the camera packet. If the entity isn't spawned clientside
when it receives the camera packet, then the client will not
spectate the target entity.
Co-authored-by: Spottedleaf <Spottedleaf@users.noreply.github.com>
== AT ==
public net.minecraft.world.entity.animal.Turtle getHomePos()Lnet/minecraft/core/BlockPos;
public net.minecraft.world.entity.animal.Turtle setHasEgg(Z)V
public net.minecraft.world.entity.animal.Turtle isGoingHome()Z
public net.minecraft.world.entity.animal.Turtle setGoingHome(Z)V
public net.minecraft.world.entity.animal.Turtle isTravelling()Z
public net.minecraft.world.entity.animal.Turtle setTravelling(Z)V
This adds a separate event before an entity is spawned by a spawner
which contains the location of the spawner too similarly to how the
SpawnerSpawnEvent gets called instead of the CreatureSpawnEvent for
spawners.
Fixed an issue where a furnace's cook-speed multiplier rounds down
to the nearest Integer when updating its current cook time.
== AT ==
public net.minecraft.world.level.block.entity.AbstractFurnaceBlockEntity getTotalCookTime(Lnet/minecraft/server/level/ServerLevel;Lnet/minecraft/world/level/block/entity/AbstractFurnaceBlockEntity;)I
Co-authored-by: Eric Su <ericsu@alumni.usc.edu>
Implements Pathfinding API for mobs
== AT ==
public net.minecraft.world.entity.ai.navigation.PathNavigation pathFinder
public net.minecraft.world.level.pathfinder.PathFinder nodeEvaluator
public net.minecraft.world.level.pathfinder.Path nodes
This adds the ability to cancel the death events and to modify the sound
an entity makes when dying. (In cases were no sound should it will be
called with shouldPlaySound set to false allowing unsilencing of silent
entities)
It makes handling of entity deaths a lot nicer as you no longer need
to listen on the damage event and calculate if the entity dies yourself
to cancel the death which has the benefit of also receiving the dropped
items and experience which is otherwise only properly possible by using
internal code.
== AT ==
public net.minecraft.world.entity.LivingEntity getDeathSound()Lnet/minecraft/sounds/SoundEvent;
public net.minecraft.world.entity.LivingEntity getSoundVolume()F
This is specifically aimed at fixing #471
Using a ConcurrentHashMap because thread safety
The performance benefit of Map over ConcurrentMap is negligabe at best in this scenaio, as most operations will be get and not add or remove
Even without considering the use-case the benefits are still negligable
Original ideas for the system included an expiration policy and/or handler
The simpler solution was to use a computeIfPresent in the get method
This will simultaneously have an O(1) lookup time and automatically expire any values
Since the get method (nor other similar methods) don't seem to have a critical need to flush the map to disk at any of these points further processing is simply wasteful
Meaning the original function expired values unrelated to the current value without actually having any explicit need to
The h method was heavily modified to be much more efficient in its processing
Also instead of being called on every get, it's now called just before a save
This will eliminate stale values being flushed to disk
Modified isEmpty to use the isEmpty() method instead of the slightly confusing size() < 1
The point of this is readability, but does have a side-benefit of a small microptimization
