help-development

[thorn isle]

🤡

[sly topaz]

surely no issues could come from doing that

[thorn isle]

My preferred approach which I recommended in the linked thread is to kill the player so their inventory isn't loaded

Or well it is loaded but presumably not going to be interacted with

[sly topaz]

was about to say, it should still be loaded due to Spigot and derivatives preserving the instance even on death

[thorn isle]

Wonky soulbound plugins aside that fuck with a dead player's inv, it'd probably be fine

[sly topaz]

I don't remember if inventory data is written as soon as you modify it

if it is then that kind of plugin can become a bigger issue, if not then just pray and hope someone isn't doing some hacky shit I guess

[thorn isle]

To disk? I think it's a staggered autosave task defaulting to a few minute interval

[sly topaz]

then it should be fine

[quaint basin]

So, simply do this:

AsyncPlayerPreLoginEvent called on any server:

If the player, in this event, is in connected_players (Redis Map), then set the event result to not_allowed and the player will be kicked. The Redis query is done in the same thread as the async_player_pre_login_event to allow setting the result as not_allowed.

Otherwise, add the player to connected_players along with the server ID and fetch the inventory data, etc. (all done in the async_player_pre_login_event thread).

If the playerquitevent is processed and the server ID from which this event is called is the same as the one in Redis, then remove it from the Redis Map in another thread (in a single-thread executor specifically for Redis operations, for example).

This way, I don't block the main thread in playerquitevent, and I have a guarantee that the inventory won't change between server transitions.

Isn't that all that's necessary? I believe having a connected_players Map with the server_id value isn't even necessary; a Redis Set is sufficient

[sly topaz]

you wouldn't kick on async player pre login, you would kick/kill on player quit event of the previous server if I understood what vcs2 said

otherwise the player would never be able to access the server they're being transferred to

though this assumes you're going with this approach, which well, you'll have to play around to see how safe it is for your server (assuming you don't have anything that relies on this weird behavior by accident)

other than that, the approach sounds fine

[quaint basin]

I don't understand why I can't kick the player in an async login event. There's a method like async_pre_login_event#setResult(not_allowed) or something similar that prevents the player from logging in

[sly topaz]

kicking the player from the server they're being transferred to would imply you don't want them to be transferred

[quaint basin]

So I can have another Set structure called connected_servers that will be atomically updated with connected_players.

Thus, every minute, for example in Velocity, I iterate through all the connected_servers and check if the server in question is offline. If so, then atomically update connected_servers to remove the offline server. To avoid having to iterate through all the connected_players and check if the value matches the offline server, instead I can have a Map<Server_ID, Set<Connected_players> structure in Redis. Obviously, this structure would also be atomically updated with connected_players and connected_servers

[quaint basin]

The problem with this approach is that checking if the server is offline wouldn't be possible atomically by updating connected_servers and connected_players. I don't know if this is really a problem, probably yes. I have to think about it a bit more to see if it's actually a problem

[sly topaz]

that shouldn't be necessary given that you would always remove them from connected_players as soon as PlayerQuitEvent is called, which would happen when you transfer them out of the offline server or for other reason

[quaint basin]

The problem is that if server A crashes, the playerquitevent will not be called, and the connected_players will always contain the player in question

[sly topaz]

if the server crashes then you have bigger issues

you can probably handle that at the proxy level in that scenario

[quaint basin]

What I can do is: when the plugin starts, iterate through all connected_players and if the server_id of the connecting server is in the values ​​of the map structure, then remove them from connected_players.

For this, I need to assume that the server starts automatically if the server crashes

This way I avoid having multiple cache structures that need to be updated atomically and performing fetches every minute on the proxy side

[quaint basin]

I still don't understand the problem xd

Ah, I understand the problem. When a player transfers from server A to server B, the removal of connected_players on server A may not have been processed yet because is done async on player quit event

So yeah, I think that approach isn't valid then

[thorn isle]

i recommend just going with the existing solution

[quaint basin]

The current solution has 1001 problems, and I bet that's why

[sly topaz]

more like, there is no need to work out an already solved problem

HuskSync does it fine

[quaint basin]

Perhaps a valid approach is to use the Velocity EventTask API for this. When ServerPreConnectEvent is called, check if the player is contained in a Velocity network server (done with the Velocity API). If not, then simply check in Redis if connected_players is contained (a Redis search). If so, then schedule a check in 1 second that verifies if the player is contained in connected_players (Redis operation) and if so, fail the event. The check to see if the player is already on the network or not is to avoid queries in Redis if the player joins the network for the first time, and also to make the player's entry into the server faster.

The 1-second timeout is to give the playerquitevent on the player's old server time to process the removal of connected_players. If it takes more than 1 second, then the player will not be able to switch servers and will remain on the same server (I don't know if additional logic needs to be added to remain on the same server).

[quaint basin]

I'm only experiencing rollbacks and duplication issues with this plugin

[sly topaz]

you'd rather find the cause of those and report them rather than developing a new solution that would probably end up having other duplication and rollback issues

[quaint basin]

This plugin is not actively supported anymore

[sly topaz]

last commit was just two months ago

[quaint basin]

Yes, but check out their announcement

[sly topaz]

their website doesn't seem to mention it or I am blind

ah, there it is

[tender shard]

that's been merged, hasn't it?

[sly topaz]

they also have a post on the english fandub of harumi susumiya which sounds random lol

[tender shard]

if you're on latest, you can use the default config

[quaint basin]

where?

[sly topaz]

https://william278.net/posts/important-update-on-william278-plugins-for-minecraft

[quaint basin]

im using the latest version but im experiencing rollback/dupe issues yet

[tender shard]

are you shutting down servers without /save-all sometimes?

[quaint basin]

I just do /stop and for me to stop servers, I have to kick all the players

[tender shard]

that might be the issue

[quaint basin]

But that's the least of the problems. The biggest problem is players suffering rollback when players switch between servers

[tender shard]

I suggest to open another issue on HuskSync's github

[sly topaz]

do you have a good way to reproduce it

[tender shard]

we have not experienced any dupe bugs since the following thing was merged (gimme a sec to find it):

[quaint basin]

reproduce what? server shutdown thing or server transfer?

[sly topaz]

the dupe and/or rollback issues

[quaint basin]

Unfortunately not, that only happens sometimes

[sly topaz]

I would start finding out why it happens then

[quaint basin]

I've already tried, but it's a pattern I still haven't been able to figure out

[chrome beacon]

Add some debug messages and stuff to HuskSync

[quaint basin]

Is it possible to force a server's TPS to 5?

[tender shard]

https://github.com/WiIIiam278/HuskSync/commit/5c4111b6a767f0f04a7ac6e6fc7ac684899f4cf2
This was the PR that fixed dupes for us

[quaint basin]

latest version already includes that

[tender shard]

then I don't know, you'll have to contact the author then

[quaint basin]

Do you have a plugin to perform a /save-all and then /stop immediately afterwards?

[thorn isle]

the cleanest way is to have players go through a hub instead of directly jumping from one server to another

[tender shard]

no, we're doing restarts via simple bash scripts

[quaint basin]

Because if I enter /save-all and /stop, some player will edit their inventory between two commands

[sly topaz]

the good ol' way

[tender shard]

yeah we enable maintenance mode beforehand

[thorn isle]

that way only one backend server has the inventory and data loaded at a time

[tender shard]

we usually enable maintenance mode, which means players have 60 seconds to leave the server or get kicked to fallback server. Only after that we do /save-all and then /restart

[quaint basin]

i see

And why is the save-all necessary? Aren't the players' inventories updated in Redis?

[tender shard]

We have disabled auto world saves, so we do that do save block inventories etc

[quaint basin]

But the server automatically saves by default when it shuts down

[tender shard]

otherwise you might get dupes etc if a player takes out an item out of a chest, then husksync saves player inventories, but world might not have been saved yet -> dupe

[tender shard]

better be safe then sorry

[quaint basin]

This is about trusting science now

If the server shuts down, then the save-all will occur

[tender shard]

yea maybe it's pointless, but as said - better save than sorry

[quaint basin]

If the save-all command fails during shutdown, then the save-all command would also fail

[tender shard]

that's why we do it beforehand

[quaint basin]

And the server freezes until it saves successfully?

I'm doing it manually. /whitelist on, kick @a and then I type /stop

[tender shard]

This our restart script

#!/bin/bash

set -u

SERVERS=(
    # name         host     save
    "proxy-1       delta    0"
    "proxy-2       zeta     0"
    "proxy-3       sigma    0"

    "lobby-1       sigma    0"
    "lobby-2       sigma    0"
    "lobby-3       sigma    0"
    "lobby-4       sigma    0"
    "lobby-5       sigma    0"

    "cb-1          delta  1"
    "cb-1_fw-1     delta  1"
    "cb-1_fw-2     delta  1"
    "cb-1_fw-3     delta  1"
    "cb-1_fw-4     delta  1"
    "cb-1_fw-5     delta  1"
    "cb-1_fw-6     delta  1"

    "cb-2          zeta     1"
    "cb-2_fw-1     zeta     1"
    "cb-2_fw-2     zeta     1"
    "cb-2_fw-3     zeta     1"
    "cb-2_fw-4     zeta     1"
    "cb-2_fw-5     zeta     1"
    "cb-2_fw-6     zeta     1"

    "cb-3          kappa    1"
    "cb-3_fw-1     kappa    1"
    "cb-3_fw-2     kappa    1"
    "cb-3_fw-3     kappa    1"
    "cb-3_fw-4     kappa    1"
    "cb-3_fw-5     kappa    1"
    "cb-3_fw-6     kappa    1"

    "event-1a      gamma    1"
    "event-1b      gamma    0"
    "event-1c      gamma    0"
    "event-2a      gamma    0"
    "event-2b      gamma    0"
    "event-2c      gamma    0"
    "event-3a      kappa    0"
    "event-3b      kappa    0"
    "event-3c      kappa    0"





    "creative      kappa    1"

    "vistas_server sigma    0"
)


errcho() { 
  >&2 echo $@;
}

esleep() {
  echo "Waiting $1 s"
  sleep $1
}

exec_ssh() {
  ssh $1@$2 "source .profile && $3"
}

start_one_server() {
    echo "start: $2"
    sleep 0.05
    exec_ssh mc $1 "cd $2 && ./start.sh || >&2 echo COULDNT START $2"
}

stop_one_server() {
    echo "stop: $2"
    sleep 0.05
    exec_ssh mc $1 "screen -XS $2 quit"
}

save_one_server() {
    echo "save-all: $2"
    sleep 0.05
    exec_ssh mc $1 "screen -XS $2 stuff '\nsave-all\n'"
}

start_all_server() {
  for entry in "${SERVERS[@]}"; do
    read -r server host save <<< "$entry"
    start_one_server $host $server
  done
}

clean_all_hosts() {
  /home/mc/bin/clean_all_hosts.sh
}

flush_all_redis() {
  /home/mc/bin/reset-redis.sh sigma <redacted>
  /home/mc/bin/reset-redis.sh delta <redacted>
  /home/mc/bin/reset-redis.sh zeta i<redacted>
  /home/mc/bin/reset-redis.sh kappa <redacted>
}

stop_all_server() {  
  # Iteration 1: Save
  for (( idx=${#SERVERS[@]}-1 ; idx>=0 ; idx-- )); do
    entry=${SERVERS[$idx]}
    read -r server host save <<< "$entry"
    if [[ $save -eq 1 ]]; then
      save_one_server $host $server
    fi
  done
  esleep 3
  # Iteration 2: Actual stop
  for (( idx=${#SERVERS[@]}-1 ; idx>=0 ; idx-- )); do
    entry=${SERVERS[$idx]}
    read -r server host save <<< "$entry"
    stop_one_server $host $server
  done
  esleep 3
  # Flush Redis
  flush_all_redis
  clean_all_hosts
}

restart_all_server() {
  stop_all_server
  esleep 5
  start_all_server
}

main() {
    case "$action" in
    start)
        start_all_server   
        ;;
    stop)
        stop_all_server   
        ;;
    restart)
        restart_all_server 
        ;;
    *)
        errcho "Unknown action: $action"
        exit 1
    esac
}

##############################

action="${1:-}"
if [[ -z "$action" ]]; then
    errcho "Missing arguments. Usage: $0 <action>"
    exit 2
fi

main $action

[quaint basin]

And I haven't had any more problems with restarts. The problem now is with rollbacks when players switch between servers

[tender shard]

you will have to flush redis

shutdown all servers, then reset redis (FLUSHALL)

then start again

[quaint basin]

when?

[tender shard]

once all servers are shutdown

[quaint basin]

why

[tender shard]

because husksync keeps inventory data in redis

[quaint basin]

the plugin that should handle this

There's nothing to say about that in their documentation

[tender shard]

there is

wait a sec

https://william278.net/docs/husksync/troubleshooting#unable-to-reset-my-server-wipe-all-player-data

[quaint basin]

But when I restart all the servers, the date that will be in Redis is supposedly the most up-to-date. They probably say this so that memory usage doesn't grow indefinitely

[tender shard]

yeah idk, I didnt write husksync. imho it's a pretty shitty plugin

we're only using it because it's "always been there"

I'm just trying to help, not to justify husksync's stupid behaviour

[quaint basin]

yes i see

Access your MySQL Database using MySQL Workbench or similar. DROP your husksync database, or at the very least DROP ALL husksync tables.
Double check you have done this

[quaint basin]

This documentation is for resetting the server, not for restarting all servers

[tender shard]

you can simply FLUSHALL your redis, unless you're using it for anything else than husksync

[quaint basin]

But the HuskSync documentation says it's not necessary. However, it's good practice to do this so the memory doesn't grow indefinitely, I'd say

[tender shard]

yeah their docs are shit

the whole plugin is shit imho

why do you think I made so many PRs in that repo

[quaint basin]

You have more scripts than that, right?

[tender shard]

of course

[quaint basin]

like clean_allhosts.sh etc

[tender shard]

clean_allhosts only removes plugins that we didnt install through our auto plugin updater

nothing interesting

[quaint basin]

but flush_all_redis does not solve the rollback issue when players are changing the server

[tender shard]

I can't help you with that, if the linked github doesn't help, then you have a different issue than what we had

[quaint basin]

What I've noticed is that when a player switches servers, their inventory gets rolled back. This happens rarely. It might be related to low server tps

[tender shard]

husksync is dead, the author doesnt do shit. he sometimes accepts PRs from strangers, but he hasnt worked on that plugin in ages himself

[quaint basin]

That's why I was looking into making a sync plugin

[quaint basin]

If the save-all fails, the script will still shut down the server

I'm going to force the server TPS to 5 and see if the inventory rollback is related to the low server TPS

yes, the rollback/dupe issue has to do with tps

[thorn isle]

lol

[quaint basin]

i'll send u a video

https://imgur.com/y1pmJIY @tender shard see this lol

You were right in saying it was the TPS server causing this

[thorn isle]

that is so sketch

[hybrid spoke]

https://giphy.com/gifs/Coddies-slippers-coddies-fish-flip-flops-pzoQawOBngFGC53N9X

[quaint basin]

The update to the player's database is likely done asynchronously, and what should happen is that the asynchronous save should occur after the database is fetched from the new server

[hybrid spoke]

so as always, a race condition. is the database call synced?

[quaint basin]

https://pastes.dev/wW9qwfxzIt This is called in the playerquitevent. This should be called after the player has already changed the server

Since the lock is not global (but local), this happens

[thorn isle]

so the ideal way of doing this is:

• velocity starts transferring player from A -> B
• B fires async prelogin; data can be optimistically prefetched here
• B fires login; velocity starts disconnecting the player from A
• concurrently, B goes through config/play protocol phases, and A fires quitevent and disconnects the player
• A unlocks database row
• B waits on the lock
• B fires player join event; the player loads into the world

the problem is that there is no sane way of doing "B waits on the lock" at that stage that don't block the main thread, without doing weird protocol hacks, like delaying some config phase packets; one subpar but practical solution is to just block for the lock in the join event at LOWEST priority, as in practice, the quit event almost always fires before the join event; but if the origin server crashes mid-quit or there is a networking issue, this can hang the destination server

one thing that is probably tolerable is to put a sane cap of say 50ms on the blocking in the event, and kicking the player if the lock isn't released in time

[umbral ridge]

isnt this where transactions come into play

[thorn isle]

in a perfect world

[quaint basin]

yea

[thorn isle]

unfortunately just about every paradigm under the sky ends up in little pieces all over the floor when it runs into real world constraints in a system not designed for it, like the minecraft server and protocol

[tender shard]

I am not sure what I'm supposed to see in that video

[sly topaz]

I think they're showing the dupe

they keep the item they just threw when changing back and forth between the servers

so they could abuse of lag spikes to cause that, but sounds complicated to reproduce even for players lol

[thorn isle]

ezpz just never have lag

if you think about it, it's like an automatic SLA

whenever your server starts shitting bricks, players get free stuff as recompensation

[sly topaz]

@young knoll

[young knoll]

?ban @kindred crater Spam bot

[undone axle]

Done. That felt good.

[sly topaz]

those have been increasing lately

[thorn isle]

you'd think the guys at discord could put together a regex matcher or something after all these years

or if that's not technologically feasible, perhaps hire an AI (Actually Indian) to do string matching instead

[lilac dagger]

I used to do h+e+c+k as a kid and it actually worked if i remember correctly

I didn't know what regex was back then

[quaint basin]

rollback/dupe happens when server tps is low

What's happening is that the save is done asynchronously when the player quits; if the player leaves server A with the tps on 5, then the scheduling will be 4 times slower and they will go to server B with the tps at 20. The login event on server B is processed 4 times faster than on server A, and therefore the database fetch is performed before the save on server A

[quaint basin]

My server was experiencing some low TPS spikes due to a memory leak, which I believe I've already resolved

[hot temple]

This is the best plugin for authentication

[chrome beacon]

Don't advertise your stuff here

[thorn isle]

where'd that even come from

and in the dev channel rather than the server admin channel too 🤡

[quaint basin]

Was the check-in petition used to resolve this error? @tropic knoll

@tender shard SORRY

wrong person

[chrome beacon]

AI ahh plugin

Couldn't even be bothered to fix the unhandled color codes in the preview images

[thorn isle]

fucking lol

[buoyant viper]

the preview image of the config having butchered ass text 😭

[thorn isle]

how does that even happen

[sly topaz]

mfw &c[ViperAuth]

[buoyant viper]

⚡ Optimized TPS

[thorn isle]

send me the source code and give me $25 and i'll perform an Independent Assessment & Certification of your codebase (normally priced at $75 but as a promising first time customer i'll give you a discount)

you can put the certificate next to the mangled chat screenshots

[kind hatch]

You're the goat

[quaint basin]

check-in-petition=true solve this yea

[echo basalt]

You can block B at the prelogin

no?

or after prelogin fires but before join fires on the netty thread or something

I don't recall the exact protocol but the client should signal it's done receiving chunks or something

[thorn isle]

no; blocking on prelogin or anything before login, causes the player to never disconnect from the origin server

velocity maintains both connections until the login protocol phase completes

so you must block in either play or config; blocking play will block the main thread, since that's where those packets are processed, and there are no events by which you can block in config

[thorn isle]

this is so that velocity can print login phase disconnect messages, e.g. "you're not whitelisted on this server", in chat while the player is still on the origin server, without disconnecting them from the whole network

in other words until velocity is sure that the player can actually join the destination server, it won't disconnect from the origin server

[echo basalt]

isn't there that one velocity plugin that keeps connections alive and "reconnects" people

what if you disconnect in the backend, use velocity to keep the connection alive

block the other connection

Or you can make your transfer command / event listener trigger a manual save instead of tying it all to the quit event

use a distributed lock and yeet the player

[thorn isle]

velocity kinda sorta does this or something similar to this built in; if you get kicked from a backend, it will keep the connection alive and try connecting you to a fallback server

but it's kind of ass because to use this, you always have to kick the player when they want to change servers

doing it with a velocity plugin like you mentioned might work better, definitely worth a try

though what i ended up with was using protocollib to delay config phase packets to "block" on waiting for the database row to be unlocked

[quaint basin]

anyway husksync is working - just need enable the check-in-petition option

https://pastes.dev/v3E5SXwg1N

this is husksync solution:

 // Another server is saving the player's data after they've logged out - we'll try fetching the data in a little while
            if (server.isPresent() && !server.get().equals(plugin.getServerName())) {
                if (plugin.getSettings().getSynchronization().isCheckinPetitions()) {
                    getRedis().petitionServerCheckin(server.get(), user);
                }
                return false;
            }

[quaint basin]

If the method returns false, the lambda code will occur in next tick (with the default configuration) with a maximum of 18 attempts. If Redis is offline during server transitions, then rollbacks/dupes will occur

and checkinpetitions need be true

[thorn isle]

the problem with this is that while you might save all the data before the quit event manually, other plugins are very liable to modify the data on quit event, and the data goes out of sync

suppose e.g. a transient item that gets deleted from your inventory when you quit; the plugin will remove it on quit, but you've already saved the inventory before quit

when the player returns, the deleted item reappears

the converse is sort of true with what husksync does, except that window occurs on the destination server; premature access to the inventory or other data might cause those modifications to be lost or for you to read stale data, if you access it before the server receives the data saved/sent by the origin server

husksync tries to get around this by listening to and blocking typical player interactions that'd access the data, e.g. inventory interactions

but other plugins must explicitly check for and respect whether the data is ready to be accessed, so you end up with your entire tech stack depending on husksync

[tender shard]

yeah, that PR about the checking petitions did fix all our dupe issues

we only got 2 problems left with husksync:

1. map sync is random (doesn't really work)
2. admin commands such as /inventory may cause inventory desyncs if that person is online on another server

we had like 3 other issues but I've fixed them myself through PRs to HuskSync's repo

the authors seems to be doing nothing, we're glad he's still accepting PRs so we at least don't have to maintain a fork

[quaint basin]

but u enabled the option right?

They're telling me to disable the option, I don't know why

https://imgur.com/a/cCNElmA see: check-in-petition= true solve the problem

[tender shard]

ughm lemme check

[quaint basin]

yea then make sense now

[quaint basin]

For some reason they tell you to leave the option disabled

[tender shard]

who says that?

[quaint basin]

https://github.com/WiIIiam278/HuskSync/issues/585

https://imgur.com/a/ZELrnBz

[tender shard]

but it was fixed

the main issue was when switching servers with different TPS, e.g. going from a 5 tps server to a 20 tps server

which happens to us once a month when we got 500 players

but since that "checkin petition" thing got merged, this has never happened again

so I suggest that you use the default config, having checkin petitions enabled

[quaint basin]

default config have this option disabled idk why

and they are talking to disable the option that's why i dont understand

[tender shard]

enable it

[quaint basin]

With check-in petitions, this happens:

Player leaves server A -> disassociates the player from server A in Redis -> as soon as the disassociation is done, it is saved in the database -> Player enters server B -> checks if there is any associated server (if server A has TPS=5, then it will probably say yes) --> if there is an associated server, then it checks if the associated server is different from server B (in this case, yes, because the associated server is A), then it sends a petition message to the associated server (in this case, A), and server A, as soon as it receives this message, disassociates the player if the player is offline and unlocked (I don't understand why there's this additional message logic) -> as soon as server A disassociates the player, server B, since it is constantly checking if the player is disassociated, as soon as they are disassociated, associates the player with server B and fetches the database as soon as the association is made. (Server B constantly checks for up to 18 attempts and by default keeps trying every tick)

The disassociation is done as soon as the save to the database is performed in the player quit. I don't understand the reason for this message additional logic

[tender shard]

well you do you. All I know is that, since we merged that PR in our own fork, no dupe/item loss has ever been reported again

[quaint basin]

Yes, there are indeed no dupes/rollbacks in the flow, but there is redundancy

[sly topaz]

does it matter

as long as it works

[quaint basin]

https://imgur.com/a/9sM8G7S The code inside the red square is redundant and can be removed

[thorn isle]

checks for up to 18 attempts and by default keeps trying every tick
this is exactly what i was talking about earlier

that's an 18 tick time window where all reads and writes to the local data will be wrong/overwritten

[quaint basin]

For this to happen, Redis needs to be offline during the transition or very slow

But Redis being slow is much less likely than the server TPS be low

[hot temple]

Bruh its called self lose by self mistake

[quaint basin]

In other words, the check-in-petition option is useless. The duplicate was happening because the server-id was the same on both servers 🤡

[thorn isle]

doesn't matter, what matters is that it can still happen

[quaint basin]

Yes, of course, but for now I'm fine with it

[quaint basin]

If I have a class with a variable that has a Block, will the world associated with that block be deleted in the gc?

[thorn isle]

public class CraftBlock implements Block {
    private final net.minecraft.world.level.LevelAccessor world;
    private final BlockPos position;

    private CraftBlock(LevelAccessor world, BlockPos position) {
        this.world = world;
        this.position = position.immutable();
    }

it's a strong reference

are you trying to diagnose another memory leak? the heap dump allows you to find the ref path from an object to gc roots; you don't need to guesstimate or xy ask here

[jagged thicket]

Shi bro, after the Intro to OOP 3 credit course in college

the way i look at java code is completely different

life changing,

college courses are actually useful

[lilac dagger]

what did you find?

oop is great

[light kettle]

does anyone know any plugins that can regenerate big arenas automatically and remove player placed blocks for free that works on 1.21.11?

[light kettle]

i need help so bad about this

[kind hatch]

I think I see what you mean about file permissions.
I used my other computer to run buildtools, copied the the output file over to my main, ran diffoscope on it and got this output.

I'm not 100% certain which line represents which file, but my best guess is that the red text is the first jar and the green text is the second jar.
In which case, this would be the jar I build on my main pc (red), and the jar I build on my second one (green)

From what I can tell, this is actually really good/close to working as intended. Because there is nothing else from the output suggesting that anything else within these files are different.

Granted, this is on two linux systems. (Linux Mint 22.2 & Zorin OS 17)
Not sure how windows would be differ (if at all)

[quaint basin]

yea

I saw that the heapdump and offline craftplayers were in GC Roots because of Block - I just confirmed it here

[sullen marlin]

Looks like <dependencySet> in assembly.xml also supports fileMode which should fix that

https://hub.spigotmc.org/stash/projects/SPIGOT/repos/craftbukkit/browse/src/assembly/bootstrap.xml

I think we just gotta add fileMode to everything

[mortal hare]

we really need ssh equivalent of minecraft's rcon

[slender elbow]

what

[winter viper]

what can you recommend to contain data? I need to log info about ores that players breaking. I think JSON and XML are too bad for this. SQLite with local .db files, I think, the best variant

[kind hatch]

Just tested that theory and I have matching hashes.

Now, this is fantastic, but I still wonder if there is any variance in Windows enviornments.

Then we might have another issue with Java variants. (OpenJDK, Temurin, Oracle, etc)

But, I'm still pretty happy with the progress right now.

[hybrid spoke]

XML?

sure, SQLite would fit the usecase. in the end it's up to you, what you want to use. Best you can do is modularize your architecture, so you have the input layer, a cache layer and an persistence layer. With that you can later swap the persistence impl if you found something better without breaking your whole architecture

[native gate]

You mean https://www.spigotmc.org/resources/minecraft-sshd-ssh-for-your-minecraft-server-universal-support.71887/?

[winter viper]

oh, thats my first plugin so i dont know about all these layers
thanks

[quaint basin]

Is it possible to view the Spark Heap summary to see if there are more Craftplayer instances than online ones?

[pseudo hazel]

idk but you shouldnt be storing players as an object anyways

[slender elbow]

by taking a heapdump and analysing it, sure

just throw it at the promptitute and ask it to read it for you

[thorn isle]

the summary should probably show it, but you'll have to scroll down quite far since there will be so few of those objects

i don't remember off the top of my head if there's a search bar

knowing spark there probably isn't or it doesn't work

[quaint basin]

The problem is that there's no Ctrl+F

i'll need use heap dump yea

[quaint basin]

It's not my plugin with the memo leak

it is huskhomes

@tender shard you are using huskhomes?

[pseudo hazel]

oof

[thorn isle]

lol

[umbral ridge]

no its donald trump white homes

[winter jungle]

Is there a way to stop the totem pop sound?

[summer scroll]

Yes, probably by intercepting the sound packet

[winter jungle]

Yeah did it by stopping the sound at the PLAYERS sound category

[summer scroll]

I think you can check for specific sound so it doesn't mess up other sounds

[silent cove]

getAge() of an adult ageable shouldn't be zero, right?

[ivory sleet]

dont babies have negative age, and then zero means its an adult

[young knoll]

I forgot this was a minecraft discord and was very confused at first

[umbral ridge]

yeah i was o
confused at first too xD

[slate mortar]

is it "intended" that the InventoryClickEvent#action seems to completely malfunction for any events in the creative inventory (no matter the view)?

It seems to somehow categorize anything (really) as PLACE_ALL

*getAction, kotlin destroyed my brain

[rotund ravine]

Creative does what Creative wants

There’s a specific event for Creative Inventory stuff though

[slate mortar]

that even does the exact same behavior

in case you mean InventoryCreativeEvent

[rotund ravine]

Creative does what Creative wants 🤷🏽‍♂️ and yeah.

[slate mortar]

meh

[winter viper]

anyone know how World.getBlockAt() can load server if use it on unloaded chunks?

or there are no difference between loaded and unloaded chunks?

[rotund ravine]

Are you asking why it loads the chunk?

[mortal vortex]

getBlockAt() on an unloaded chunk will load it synchronously i think?? blocking main thread

idk what u mean "how", how it works?

just guard with isChunkLoaded

that will take a chunk tho, u can do location.getBlockX() >> 4 to get the chunk from an X

[winter viper]

no, will it load the chunk
I want to use this method but I don't know how much load it will put on the server if it loads chunks when trying to get a block from an unloaded one

[mortal vortex]

if (world.isChunkLoaded(blockX >> 4, blockZ >> 4)) {
    Block block = world.getBlockAt(x, y, z);
}

i dont think us spigoters get async loaders like paper

also @winter viper do you plan to modify it, or do you just want to know what it IS? You can use ChunkSnapshots which are thread-safe (if you ARENT planning on modifying it) https://hub.spigotmc.org/javadocs/spigot/org/bukkit/ChunkSnapshot.html

[winter viper]

oh thank you

[winter viper]

ok ill check it

[rotund ravine]

Iirc it might also not until he calls any functions on the block, cause it’s a glorified wrapper. Can’t recall though

[mortal vortex]

ur right, i think CraftBlock is just a BlockPosition right????

but again, i cant imagine calling the getBlockAt, and then doing nothing with it.

The minute u call getType or anything its gonna access the chunk, its just a matter of when, so the guard is still needed, and the chunk will need to be loaded synchronously, again, just a matter of when rather if.

[winter viper]

i need just get block type at coordinates and put it in db

[winter viper]

so there are will be better to use chunksnapshot?

[wraith magnet]

Hey guys, I'm trying to create a "care package" dropping from the sky (invisible armor stand with a chest as helmet), and it has a parachute of chickens, connected to the armorstand with leads, but I can't seem to get the connection right. The leads should be connected to the chest, instead of currently being connected to like the side of the armor stand.. I cant send an image to clarify, but if anyone has any ideas how I can make it look clean lmk

[mortal vortex]

yessir

guard it still

[winter viper]

thanks

[mortal vortex]

Send an image, with either imgur or verify your account

[wraith magnet]

https://imgur.com/a/lTdKCrX

[mortal vortex]

code?

[quaint basin]

my velocity have this memory usage:

Memory(process)
1.2 GB
/
4 GB
29.19%

why is the velocity using 2.7 gbs swap? I'm on linux

[thorn isle]

creating a chunk snapshot to access a single block is extreme overkill and still has to be done sync; not only does it not help in any way, it is much slower

chunk snapshots are useful when you have to look at hundreds or thousands of blocks in bulk or have complex, expensive logic happening in between your block checks, which you want to do off the main thread

e.g. matching large structures to the world, or checking if a spot is a valid spot for a structure to be generated at

to clarify, the snapshot itself is thread safe to access, but creating it still requires syncing to main; so it is not at all an async substitute for World::getBlockAt

[grave solar]

OK, I have a data pack that includes a scoreboard on the side bar. But whenever I have a certain plug-in on that I need for my server the sidebar disappears. I need help figuring out how to keep the plug-in, but actually be able to see the scoreboard for the data pack.

[lilac dagger]

this should work no?

arraycopy doesn't like it when i give the second component as Jpanel

even if i know for a fact what elements there's in the panel

actually wait

might not be safe

there's a getparent

[slender elbow]

that'd only work if components contains only jpanels

[lilac dagger]

yeah

it would've worked if getparent didn't exists

[slender elbow]

Throws:
ArrayStoreException - if an element in the src array could not be stored into the dest array because of a type mismatch.

[lilac dagger]

technically it should be fine

as i know what elements this component holds

[wraith magnet]

yo guys, any way to not show a wither's shield particle when its health is >= 50% (spigot 1.8.8)? its for a countdown bossbar

[lilac dagger]

i just put the wither really far

location.getDirection().multiply(60)

or wait

can you just set a very small size to the wither?

what's the behaviour to the shield if the size approaches 0?

[wraith magnet]

its still noticable even from a distance tho, also when a player looks down the bossbar dissapears

[wraith magnet]

hmm that might be a good idea no idea if its possible tho

[lilac dagger]

it should be

worth a test

let me know how it goes

[wraith magnet]

hell yeah

alright will do

[lilac dagger]

i can also give you the method

because 1.8.8 mappings are sadly hard to work with

[wraith magnet]

yeah I already got it, was just wondering how to remove the shield etc

since on hypixel it doesnt show so I guess there's a way to completly hide it but idk

[lilac dagger]

yes

i think it's this

r(500); if small wither

either 1000 or 10 is small

make sure to send the datawatcher

[placid rivet]

i'm making a custom gui and going into a slot a highlight appears, is there any way to disable it? i'm on 1.21.10

[lilac dagger]

i assume it can be done via resource pack

[hybrid spoke]

a highlight? like a tooltip of the filler item, or clientside hover effect?

[placid rivet]

Yea i got it, but i meant the hover effect on a slot

[quaint basin]

If I have an Entity instance where the chunk it's in has been unloaded and loaded, does Entity#isValid still return true?

[chrome beacon]

I doubt it

[lilac dagger]

which client version do you guys think i should use for these options?

i was thinking http/1.1

it should support more servers this way no?

[quaint basin]

should I use bukkit#getEntity right

and check if is not null

[kind hatch]

@sullen marlin I'm still looking into the reproducibility stuff and I've found that there is consistency between java vendors within the operating systems, but not across them. (Both Linux & Windows with the <fileMode> changes)
This is good since it's one less variable, but I actually think that maven is what's controlling most of the factors here.

The diffoscope output between linux and windows had much more this time, but the differences between the two are, from what I can tell, small inconsistencies.
First being a small 53 byte difference between the jar files themselves. There's also bunch of stuff at the end of the comparisons that just shouldn't be there.

I don't know how far you got initially, or if you even came to this conclusion already, but this seems to point towards OS specific line endings being another contributing factor. (LF vs CRLF)
This makes some sense because the compilation is done on the end user's machine which might pull in/auto format to the OS specific line ending. (At least on windows that is)
I'm trying to look into how we can make maven enforce one specific type, at least to test, but I'm coming up short at the moment.

[sullen marlin]

I never tried different OSs

but looks like line endings is an issue

https://hakanozbay.github.io/2017/02/15/maven-assembly-line-ending.html

looks like a line ending option for the .list files?

[kind hatch]

@sullen marlin I have matching hashes across operating systems now.

There is some stuff to consider with these specific changes.

• The <project.build.outputTimestamp> flag needs to be set for this to work correctly.
  • Currently, this variable is used to set the Vendor name so that CraftBukkit can read it for the outdated check it does. (Still don't fully understand why it is the way it is btw)
    • A different flag should be found and used if at all possible.
  • Setting this to the maven wiki example (2023-01-01T00:00:00Z) will cause spigot to not load. (Unless you have the IReallyKnow flag set)
    • This is due to CraftBukkit reading this and expecting it to be an integer. Setting the value to 1 let's spigot load as expected.
  • This flag should be set dynamically somehow. I would assume this needs to be done through the jenkins pipeline.
• Automated testing for each build would need to be figured out.
  • Also assuming that this would be done in jenkins, as I don't think we would want to include diffoscope as a dependency for BuildTools.
• I have no idea if maven profiles are affected by this.

My testing still shows that OpenJDK, Azul/Zulu, & Oracle JDKs will produce the same hashes. (This likely extends to other vendors)
I also have no idea how to push these would be changes to stash. I am extremely confused on how that works for spigot itself.

[wet breach]

<assembly
xmlns="http://maven.apache.org/plugins/maven-assembly-plugin/assembly/1.1.2"
xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
xsi:schemaLocation="http://maven.apache.org/plugins/maven-assembly-plugin/assembly/1.1.2 http://maven.apache.org/xsd/assembly-1.1.2.xsd">

<id>x</id>
<formats>
    <format>jar</format>
</formats>
<includeBaseDirectory>false</includeBaseDirectory>
<fileSets>
    <fileSet>
        <directory>../../deploy/script/linux_x86</directory>
        <outputDirectory>${project.version}/bin</outputDirectory>
        <fileMode>0755</fileMode>
        <!-- Options are "keep" - Preserve all line endings

"unix" - Use Unix-style line endings (i.e. "\n")

"lf" - Use a single line-feed line endings (i.e. "\n")

"dos" - Use DOS-/Windows-style line endings (i.e. "\r\n")

"windows" - Use DOS-/Windows-style line endings (i.e. "\r\n")

"crlf" - Use carriage-return, line-feed line endings (i.e. "\r\n") -->
        <lineEnding>unix</lineEnding>
    </fileSet>
</fileSets>

to fix line endings of files

[stoic surge]

dm for custom plugins

[mortal vortex]

?services

[undone axle]

If you wish to request or offer development/art/building/administration services, please do so at https://www.spigotmc.org/forums/services-recruitment-v2.54/

[sullen marlin]

Not sure your point about timestamps is right, scriptus already sets timestamps based on git hash and they're definitely static in all the diffs both you and I have done

[kind hatch]

Just removed the flag and the compilation shows no differences. My bad, I guess I put too much emphasis on that part. I had just gotten matching hashes working when I sent that message and had those timestamps set.
I wasn't aware that scriptus was actually setting the timestamp. :p

[echo basalt]

hi frosty

nvm I'm late asf

[tender shard]

no, of course I don't

[quaint basin]

When LivingEntity#getEquipment can return null?

[quaint basin]

Is your plugin?

or is a public plugin or smth

[thorn isle]

i think it returns null for entities that don't support equipment, like cows and such

[quaint basin]

oh i see

[hybrid spoke]

cow armor when?

[young knoll]

All LivingEntities support equipment

Most just don’t render it

if (entity instanceof Mob || entity instanceof ArmorStand || entity instanceof Mannequin) {
            equipment = new CraftEntityEquipment(this);
        }```

I don’t think any vanilla entities will fail that if statement

Okay players fail it but they have a different override for getEquipment anyway

[tender shard]

no

[umbral ridge]

having an ups is so cool

made a http json api thats kind of a wrapper of snmp data.. a plugin that checks the api and can read the whole ups data.. one of the coolest things is it can read the ups state and can say like "Lost electricity. Running on battery" in the game.. can send a warning and stuff XD if theres no electricity

[young knoll]

Wait is an correct there

[umbral ridge]

i dont know

[young knoll]

ups does start with a vowel sound but it just sounds wrong

Am I crazy

[umbral ridge]

a is correct

a sounds weird to me

[quaint basin]

then?

[remote swallow]

i imagine its made by the same person who made husksync and both are open source so you can just check the github or any of its plugin pages

[quaint basin]

https://github.com/MrPippi/HuskSync-26.1/releases/tag/3.8.8.2 @tender shard

[jagged thicket]

An is indeed wrong

if we say UPS it must be "a UPS"

but if we say Uninterrupted Power Supply then we must use "an Uninterrupted Power Supply"

cuz UPS starts with yoo sound like university

we say a university not an university

[young knoll]

So the yoo sound isn’t a vowel sound

[jagged thicket]

Yea

[young knoll]

It has to be a uhh sound

[jagged thicket]

a e i o u

uh ae e ou uh

[young knoll]

English moment

[jagged thicket]

smn like that

i studied ts when i was in 5th std

they made us remember how words sound

based on phoenetics

[buoyant viper]

does the scheduler run tasks near the beginning or the end of the tick?

?stash

[undone axle]

https://hub.spigotmc.org/stash/

[buoyant viper]

looks? like near the end of the tick

eh doesnt actually matter for what im using this info for

[drowsy helm]

Uh i dont think its something you can reliable determine

[buoyant viper]

what do u mean, like across versions?

my question ended up being pretty irrelevant for my use, since one way or another, ill get the data i need

(im just crudely calculating mspt using a repeating 0 delay synchronous task)

(API for exposing TPS data when??)

[buoyant viper]

yeah good enough for me

laggy ahh server

[fathom fable]

Interesting

[young knoll]

IP leaked!!!!1

[buoyant viper]

:(

[pseudo hazel]

im gonna ddos that ip so hard you wont even see it coming

[thorn isle]

you can't really calculate mspt from the tps since the tps is clamped to 20, while mspt isn't

and the reason why we prefer mspt over tps is specifically that it's not clamped

on paper, you could compute it from the tick start vs tick end event, but on paper you also just have a getter for it

on spigot you might be able to hack something together by finding out some event that runs at tick start, and then use the scheduler for the tick end

although now that i think about it, i don't remember if scheduler tasks run at the start or the end of the tick

i want to say end because that'd make more sense for "0 delay" tasks that get scheduled and executed in the same tick, but god only knows

[buoyant viper]

yeah im doing it backwards, calculating TPS from MSPT, where TPS = MIN(1000.0 / MSPT, ServerTickManager#getTickRate())

me calling it MSPT is misleading though, as it is the time between ticks (+ the time a tick spends), rather than just the time of a tick itself

[buoyant viper]

MIN only to avoid the visual "error" where TPS may exceed the target tick rate due to a time delta of <1000.0 / ServerTickManager#getTickRate()ms (or else i may see something like 20.04 or some other odd value)

[slender elbow]

scheduled tasks run at the beginning of the tick

0-delay tasks are 1-delay tasks

[buoyant viper]

good enough for me

[thorn isle]

sadge

[buoyant viper]

its just visual flair anyway so the accuracy of it doesnt have to be super high, im not using this calculation for anything critical

it just has to look "correct enough"

[quaint basin]

EntityRemoveFromWorldEvent is not called when a world unloads? (for the loaded entities)

[slender elbow]

sure should

[quaint basin]

And with the Entities Load Event, can I be sure that they will all be in the same chunk?
Or the list of entities in that event may be in different chunks

[slender elbow]

i mean, not if i go and modify the region files manually before loading the world again

like, generally yeah but there are no guarantees about that stuff

[quaint basin]

One more thing: For some reason, when I teleport to a chunk in the world, entitiesloadevent and entityremovefromworld are called multiple times

[thorn isle]

once per chunk, once per entity, respectively

i feel like we're having this conversation every week now

and have had for months

[slender elbow]

just let natural selection do its thing

[rotund ravine]

Doesn’t work for spigot devs

[slender elbow]

oh yeah because he's developing for spigot

[thorn isle]

lol

a dying breed

[chrome beacon]

I mean they are developing for Paper

They're just already banned in the paper discord

[chrome beacon]

also this 💀

[young knoll]

Was gonna say

There is no EntityRemoveFromWorldEvent on spigot

[kind hatch]

What if there was tho?

[buoyant viper]

what are u waiting for? get scribing 📜

[buoyant viper]

atp spigot should just Steal events from paper lowkirkenuinely

@mortal vortex write these PRs for me

[thorn isle]

no need to write them or steal them

choco already has pr's for all of them and all of them are in ice and limbo forever

[umbral ridge]

test

weird

i can message some people and some people i cant.. 50 50.. russian roulette

[thorn isle]

yeah putler is currently in the discord datacenters pulling random ethernet cables so that the ookranians don't organize to nuke his victory parade tomorrow

[umbral ridge]

yea

[mortal vortex]

Yes mommy :c

[buoyant viper]

they're in 🧊 custody?!

[sullen marlin]

https://hub.spigotmc.org/stash/projects/SPIGOT/repos/craftbukkit/commits/b5ccd9cda721a1469c17a233677d22f8529c1561

Now my local matches Jenkins

probably just need to fix line endings now for cross platform support

[sly topaz]

that's fun

file permissions will never not bite me in the ass when it comes to reproducible builds

[kind hatch]

Now for the hard part.

[bold lotus]

does anyone know a koth plugin that has inbuilt option to add boss fights in the koth snd stuff like that i mean i want warzone events that i can add to different different warzones that im adding in my server, but im unable to find a suitable plugin for it, anyone knows any plugin pls tell

[lilac dagger]

guys, i cooked up, i made an amazing login backend

[buoyant viper]

waow

[quaint basin]

So, is the entities loadevent called for the same entity multiple times?

[tender shard]

no

[quaint basin]

This is what's happening

I'm using player#teleportAsync for a chunk containing entity X, and EntitiesLoadEvent and EntityRemoveFromWorldEvent are called multiple times for entity X

[tender shard]

teleportAsync is paper API, you'll have to ask on paper's discord

[quaint basin]

im banned there xd

[tender shard]

yeah me too but nonetheless

?whereami

[undone axle]

https://cdn.discordapp.com/attachments/864180468680360027/944415049462714469/whereami.mp4

[tender shard]

maybe show your code

[quaint basin]

Okay, but if you use paper and if u know why this happens maybe someone here could help

everyone uses paper so yea

[buoyant viper]

i use spigot :-(

[quaint basin]

The code won't help at all ig

[quaint basin]

I literally only do that

[tender shard]

print out the stacktrace in the respective events

then you'll know why that event is called / where it's called from

e.g. just do new Exception().printStackTrace()

[quaint basin]

alr

I think the EntitiesLoadEvent is called multiple times because the chunk state constantly changes during chunk loading

[tender shard]

idk, never happened to me

[buoyant viper]

oh thats a cursed way of doing that, i love it

[tender shard]

how would you do it lol

[buoyant viper]

well i probably wouldnt bc im stupid

[tender shard]

I use that very often when I don't know why the heck something is happening

[quaint basin]

yewa this is smart lol

[buoyant viper]

i once tried to use Reflection.getCallerClass or whatever the method is for something

[quaint basin]

This doesn't cause me any problems either, but I'm doing unnecessary database fetches this way

[buoyant viper]

but java didnt like that i was trying to use a sun class

i just wanted to be cool and do the same stuff jre internals do :-(

[tender shard]

internals are cursed

[buoyant viper]

yeah

[tender shard]

oh god I just found my "basics" project again https://github.com/SpigotBasics/basics/tree/main

some day I'll finish that

already 2 years ago, felt like yesterday

[buoyant viper]

oh hey i contributed to that

[tender shard]

everyone contributed except @rotund ravine

[lilac dagger]

Does anyone wanna contribute to bot creator? Noone? Alright 🥹

Oh, this is sort of like essentials

[tender shard]

yeah but modular and with only 4% of Essentials's features

it was my project to start learning kotlin and gradle

[lilac dagger]

I see

Pretty cool

[tender shard]

thx

[grave hill]

What's that?

[lilac dagger]

You can use bots to test your server

[rotund ravine]

Ahhh i was wondering when i’d get pinged for that again

[quaint basin]

Is it possible to spawn an entity with a predefined UUID?

[slender elbow]

that sounds like flawed design

[quaint basin]

I have a database with minions where the entities have UUIDs. Now the entities are persistent (before they weren't) and I need to add all the entities to my worlds that are in my database. To do this, I need to generate an entity with the UUID that is in my database

If that's not possible, then I'll have to update the UUID entity that's in my database

[chrome beacon]

If the entities are persistent why are you trying to spawn them again

That will just cause there to be multiple of them

besides even if they weren't persistent it's still a bad design

[quaint basin]

(before they weren't)

[chrome beacon]

still bad design

[quaint basin]

why?

The possible options are: persistent or non-persistent. If persistent is bad design and non-persistent is bad design, what do you do?

[chrome beacon]

That's not what we were saying

Your way of storing data and trying to force a uuid on an entity is a flawed design

Persistent or non-persistent both options are fine as long as they're handled well

[quaint basin]

This will only be done once in the onenable to create the world entities from the data in the database

[chrome beacon]

That would require loading all the chunks that your entities are stored in

[quaint basin]

I know that. Is that bad design?

[chrome beacon]

Ideally you'd migrate things as needed

because too many chunks being loaded and such will take too long

[quaint basin]

I'm just trying to update the world state with the information in my database - I just wanted to know if it was possible to create an entity with a UUID to avoid writing additional code - that's all

[chrome beacon]

It is possible just a bad idea

[quaint basin]

i don't have many minions

The problem here will be loading multiple worlds because I use one world per player, not the number of chunks themselves

But I iterate through all the worlds, and when I finish iterating through one world, I unload it before iterate on to the next

So I don't think it will run out of memory

[chrome beacon]

If you really want to do it that way just update the uuid in the database

[quaint basin]

I can have rows in the database and not have the entity in the world

Ah, yes, I update the database with the new entity uuid

I'm listening to EntitiesLoadEvent, and for each entity that is a minion, I query the database to load the minion cache. Since EntitiesLoadEvent is called for the same entity multiple times during chunk loading, I need to create logic that, for example, schedules a task to be executed in 1 second for each entity in EntitiesLoadEvent, and then uses map#put(entityId, taskId) (to track the last taskid associated with the entityid). If the associated taskid is in the map, then performs the scheduler task, and then loads the database to update the Bukkit minion cache if the entity is still valid?

[quaint basin]

This is simply to avoid creating multiple database fetches multiple times, since this event is called multiple times for the same entity during chunk loading

should I do this?

[slender elbow]

sure, that is generally known as debouncing

[quaint basin]

in this case i'll do that:

check if there's a completablefuture in my future cache, return the result of that if it is
if none of thos exist, create a future and add it to the future map

[umbral ridge]

im making an antenna and i wanna check if im in reach of it

how do you check how far away you are from the block? and if theres anything in between?

[young knoll]

Ray trace

[umbral ridge]

nice

thats so cool

XD

[lilac dagger]

crazy

i'm working on my front end

i can't wait to get it working

[umbral ridge]

man

cant wait to get my antenna plugin fully running. im making repeaters now. so that in tunnels there will be signal repeaters

[lilac dagger]

😮

[umbral ridge]

and system will keep track of minecarts in the tunnels where they are and display them on a map

have to make an image api on my web server so i can use image to map plugin api

that will be sick

I love having tech stuff in minecraft

[quaint basin]

How can I access the task ID within the lambda of a scheduled task?

[chrome beacon]

?scheduling

[undone axle]

https://www.spigotmc.org/wiki/scheduler-programming/

[lilac dagger]

there's runnable

and a task giver

consumer i think?

@quaint basin

[quaint basin]

The problem with this is that if I replace the () with task, the method then doesn't return the taskid (I need to access the task id both inside and outside the task)

https://pastes.dev/czpY6LSIrq

I would like to do that

but line 81 have an error

[lilac dagger]

what about this?

by the time your task runs the field is populated

[quaint basin]

this way I need create a class for that

[lilac dagger]

you can put it in an existing class

you could make the first scheduler an annymous class that holds the task of the second

[quaint basin]

In the current class I can't because it's a Listener and I can't share the same instance for different entity IDs

[lilac dagger]

if you want that either

the listener does accept a custom constructor

you could inject a class that takes a task id

and use that

[quaint basin]

can u give me an example?

[lilac dagger]

it's too much to write

but imagine this

in your listener

you make a constructor

and pass an object

in that object you then do something like object.setTask(Task);

and you can also reuse that elsewhere

to cancel it on server stop for example

[quaint basin]

In this way, the object instance will be shared by all entities

[lilac dagger]

yes

[quaint basin]

That's not what I want

I think this is the only way

final int[] taskIdHolder = new int[1];

taskIdHolder[0] = Bukkit.getScheduler().runTaskLater(plugin, () -> {

    int taskId = taskIdHolder[0]; // acessível aqui

    plugin.getLogger().info("taskId = " + taskId);

}, 20L).getTaskId();

[lilac dagger]

if you want a local variable this works too

[thorn isle]

use bukkitrunnable

any time you see weird array wrappers for mutable inline variables like this in a bukkit task, you probably want to use bukkitrunnable

[young knoll]

Clearly you gotta use an atomic int

https://cdn.discordapp.com/emojis/1017515898870956154.gif?size=48&name=pepe_laugh&quality=lossless

[quaint basin]

chatgpt peak

[serene egret]

is there a way to set a textComponent as a header / footer of a tablist? Im trying to use custom fonts but the #setPlayerListHeader method for example doesn't let me parsing them.

[thorn isle]

i've see that here and called it out more than once

[quaint basin]

how can i do the code above with this?

[thorn isle]

i have no idea what you're doing or trying to do

[quaint basin]

int taskId = new BukkitRunnable() {
@Override
public void run() {

            }
        }.runTaskLater(plugin, 20).getTaskId();

how can I get the taskId inside @run

i need an array yet

[chrome beacon]

Just read the wiki page I linked...

[tender shard]

You gotta use the consumer way of scheduling the task

[thorn isle]

personally i'd make the taskid a field on the class and assign it from this.runTaskLater in the constructor

that said i'm not sure what you need the task id for since you already have the runnable itself

[quaint basin]

how

[thorn isle]

still no clue what you're doing btw

really sounds like a xy question

[tender shard]

ugh I havent used java in ages, but sh like that I guess? Bukkit.getScheduler().runTask(plugin, (myTask) -> { ...

and then you can do myTask.cancel() or whatever

[quaint basin]

this method is Void type

[tender shard]

correct

[quaint basin]

and I need access the task Id outside too

[tender shard]

why?

[quaint basin]

https://pastes.dev/kT7bg1prtT

[quaint basin]

because i need lol

[tender shard]

why?

[quaint basin]

https://pastes.dev/kT7bg1prtT here is the code

I'm listening to EntitiesLoadEvent, and for each entity that is a minion, I query the database to load the minion cache. Since EntitiesLoadEvent is called for the same entity multiple times during chunk loading, I need to create logic that, for example, schedules a task to be executed in 1 second for each entity in EntitiesLoadEvent, and then uses map#put(entityId, taskId) (to track the last taskid associated with the entityid). If the associated taskid is in the map, then performs the scheduler task, and then loads the database to update the Bukkit minion cache if the entity is still valid

[tender shard]

you can keep a reference to the callback inside the callback itself, if you really need to

[quaint basin]

can you give me a example code?

[tender shard]

ScheduledTask myRememberedTask = null;
Bukkit.getScheduler().runTask(plugin, (myTask) -> { ...
myRembembered Task = myTask;
...

sth like this

[thorn isle]

i'd use a uuid -> completablefuture map

[tender shard]

makes little sense though normally

[quaint basin]

this will not works

myRememberedTask will be null until the task is executed

[thorn isle]

uh

i'm leaving now

[quaint basin]

for what

[tender shard]

correct but it will never be executed immedietaly

otherwise it wouldn't be a task

yk

[quaint basin]

I need to access myRememberedTask immediately after the task is scheduled, and in 100% of cases this happens before the task is actually executed

[tender shard]

why do you need that?

[quaint basin]

read this code: https://pastes.dev/kT7bg1prtT

[tender shard]

No, tell me

[quaint basin]

see the line 106

[quaint basin]

^^

[thorn isle]

use a completable future instead of messing with this task nonsense

[tender shard]

your whole logic is flawed if you rely upon both the actually scheduled task, and the ID assigned by the scheduler

[quaint basin]

I don't understand how ur solution works

[thorn isle]

trying to jerryrig the bukkit scheduler to do some sort of async data dependency stuff and callbacks from a database is an exercise in futility'

[quaint basin]

wdy,m

[tender shard]

erm hard to explain, lemme think

[quaint basin]

If it's what I'm thinking, then it doesn't work

[tender shard]

you are trying to code some kind of hypixel-like minion thing? right?

[quaint basin]

yes

entityloadevent -> fetch db and load the minion cache
entityremoveevent -> remove from minion cache

the problem is:

entityloadevent is called multiple times on the chunk loading

[quaint basin]

and i want avoid many useless database fetches

[tender shard]

okay so you should either have

1. one runnable per minion, OR
2. one global runnable that iterates over all existing minions

Which option are you currently doing?

[tender shard]

then you only have to schedule ONE runnable ONCE

and you never have to worry about its task id

[quaint basin]

wdym

[tender shard]

just make some task that does sth like this:

for (Minion minion : allMyMinions) {
 // do stuff
}

and run that e.g. idk once per second

[quaint basin]

i have this

[tender shard]

ok so what's the issue?

[quaint basin]

I don't understand where you're going with this

[quaint basin]

^^

[quaint basin]

^^

[tender shard]

I have no clue how those two problems would even remotely have sth in common

[quaint basin]

read this

[tender shard]

Keep a Map<Location,MyMinionStuff> in memory

[quaint basin]

bruh

i alr have this

[thorn isle]

the only way i can make sense of it is that he wants to collect all the entities in the loaded chunk and then do a bulk db query for them all, rather than one query per

not sure though

[quaint basin]

no no

[thorn isle]

especially since the entities load event fires on a per chunk basis and already has all the entities in the chunk

[quaint basin]

Have you read this????

[tender shard]

EntitiesLoadEvent definitely does not fire 1000 times on latest paper, nor does it do on latest Spigot

[quaint basin]

they do

[tender shard]

no

[quaint basin]

u alr debugged?

[young knoll]

It should be once per chunk load

[thorn isle]

i don't see what that has to do with task id's

[tender shard]

exactly

[young knoll]

EntityAddedToWorldEvent or whatever it is might fire a lot more

[quaint basin]

but is not

[thorn isle]

and yes, it should be once per chunk load, not multiple times

are you sure you're instrumenting it properly

[tender shard]

show your code

[thorn isle]

^

if it really does fire multiple times, someone should open an issue about it

[quaint basin]

@EventHandler
public void onEntitiesLoad(EntitiesLoadEvent event) {
List<Entity> entities = event.getEntities();
minionController.onMinionLoad(entities);
}

https://pastes.dev/kT7bg1prtT

[thorn isle]

and how exactly are you instrumenting it?

i don't see a sysout in the listener

[tender shard]

Add my suggested way of debugging at the top of the event

E.g. simply printout "listener called" or better, create an exception and print out the stacktrace

you'll see that it will NOT be called multiple times

[young knoll]

“Oops I accidentally have a task running that loads and unloads 50 chunks every tick”

[quaint basin]

bc i alr removed

[tender shard]

yeah probably sth like that

[quaint basin]

@EventHandler
    public void onEntitiesLoad(EntitiesLoadEvent event) {
        
        List<Entity> entities = event.getEntities();
        minionController.onMinionLoad(entities);

        try {
            throw new RuntimeException("test");
        } catch (Exception e) {
            BukkitMinionPlugin.INSTANCE.getLogger().log(Level.INFO, "exception occurred on EntitiesLoadEvent", e);
        }

    }

if I have this, can I see what's causing the event call?

[tender shard]

yes

[quaint basin]

But won't it always be paper?

[tender shard]

but you can also just do new Exception().printStackTrace()

[thorn isle]

it is true that having tasks like this, accessing chunks and triggered by chunk loads or unloads, can result in infinite loops where the chunk loads, schedules the task, unloads, and then gets loaded again by the task, rinse and repeat

[tender shard]

well yes ofc

[young knoll]

Can’t you also do like

Thread.dumpStack or something

[thorn isle]

myes

[quaint basin]

then how this will be help me?

[thorn isle]

we don't believe what you're saying, basically

[quaint basin]

My Minion Tick Loop won't load chunks if the chunk is unloaded

[tender shard]

.....

[thorn isle]

this will help you by proving what you're saying is happening is happening

[tender shard]

by checking out the other lines that are not paper related

[thorn isle]

or alternatively by finding out the issue that is causing it to appear as if it's called multiple times

[tender shard]

I still claim that the event is not called multiple times by the server software; they are causing this themselves. probably by loading a chunk themselves in the chunk load event

[quaint basin]

well my minions plugin does not use chunkloadevent. my other plugins i dont think so too

[tender shard]

ok

go debug

[quaint basin]

yes im starting the server

@Override
    public void onMinionLoad(List<Entity> entities) {

        for (Entity entity : entities) {

            String customName = entity.getCustomName();

            if (customName == null || !customName.contains(KeyConstants.MINION_ENTITY)) {
                continue;
            }

            if (entity.getType() != EntityType.ARMOR_STAND) {
                continue;
            }

            UUID entityUuid = entity.getUniqueId();
            new Exception("Entity Uuid: " + entityUuid).printStackTrace();

i added the exception here

[young knoll]

Oh boy using entity name for identification

[tender shard]

no....

please print the stack trace at the top of the event handler

not somewhere convuluted inside your existing flawed logic

[quaint basin]

what should I use?

[young knoll]

PDC

[quaint basin]

this plugin is 1.8-1.21

[young knoll]

Ew

NBTApi I guess

[quaint basin]

this way we ll dont know if the event is being called multiple times to the same entity

[tender shard]

1.8?

[quaint basin]

I'm running 1.21

[tender shard]

nobody uses 1.8 nowadays

I am out

[quaint basin]

I'm running 1.21 brother

the problem is happening on 1.21

[tender shard]

Please debug as I suggested

Then show us the stacktrace

[quaint basin]

like this?

@EventHandler
public void onEntitiesLoad(EntitiesLoadEvent event) {
new Exception().printStackTrace();
List<Entity> entities = event.getEntities();
minionController.onMinionLoad(entities);
}

[tender shard]

Yep

[young knoll]

Oh wait without the PDC patches the game doesn’t even save custom nbt data on entities so that wouldn’t work

[tender shard]

wdym? It still doesn't do that for everything

e.g. tile entities

(unless I'm wrong)

[young knoll]

I mean, it saves stuff in pdc

Which can hold whatever you want

[tender shard]

IIRC you can write to a chest's PDC but it'll be gone upon chunk unload / reload

[young knoll]

No? That’s what PDC is for

It’s persistent

[tender shard]

yeah but the holder isn't

[young knoll]

Well no, but why would you be using the holder

[quaint basin]

https://pastes.dev/uWgGk0i9fU

I just joined on the server

[tender shard]

you're calling the event yourself by accessing the chunk in the EntitiesLoedEvent, it seems

you must not access the chunk in EntitiesLoadEvent because the chunk might not be loaded at that point

if however you decide to access it at that point, you force load it, causing the EntitiesLoadEvent to be called again

(I think - might be wrong)

Please show your EntityChunkModernListener class

[slender elbow]

I don't see where in those stack traces it is the plugin causing a sync chunk load?

[tender shard]

idk but some stacktraces are longer than others

[slender elbow]

lol

okay

[quaint basin]

https://pastes.dev/yuLR3MazDv

[quaint basin]

Even if a plugin causes chunk loading, won't only paper appear in the stack trace?

[slender elbow]

i mean, no

if it's a sync chunk load, the code that caused that load will show up

[quaint basin]

so we can conclude that the server software is causing the EntitiesLoadEvent to be called multiple times for the same entity during chunk loading

[young knoll]

You have SlimeWorldManager stuff going on

Who knows what that is doing

[quaint basin]

they don't change this

we can test this on paper too if u want

[slender elbow]

you're jud[ asking to get ?whereami'd again

lol

[quaint basin]

I'm banned on paper

But we can test this on spigot too

🤡

[slender elbow]

sure but coming to the spigot discord saying "this happens in paper too" isn't exactly what here is for

[quaint basin]

In my mind, entitiesLoadEvent wasn't even from Spigot, but then I saw that it is

[young knoll]

How many chunks are actually being loaded

And how many times is the event being called

[slender elbow]

anyway, nothing in that paste shows it's happening for one singular chunk, but that is a known "issue", however, and as you mentioned before, the unload event is called as well
https://github.com/PaperMC/Paper/issues/9581

you really just need some minimal debouncing logic, nothing crazy

[quaint basin]

yes unload event is called too

[quaint basin]

yes! i did

[slender elbow]

great

problem solved

[quaint basin]

I just wanted to do this debug because I was almost killed here for being accused of messing things up

[slender elbow]

I'm gonna go out on a limb and say that's a slight exaggeration lol

[young knoll]

That issue doesn’t mention the EntitiesLoadEvent

[quaint basin]

https://pastes.dev/Ohwx4Fv5yt

Well i tested myself

[slender elbow]

the idea is the same

[quaint basin]

see: the event is being called 3 times for the same entity on the chunk loading

[quaint basin]

definitely not

[young knoll]

That implies the game is loading, unloading, loading, unloading, and loading the chunk

[slender elbow]

I doubt you were "almost killed" for it

also, lol

[young knoll]

Either that or the event is being called from places it shouldn’t be

[quaint basin]

should I use EntitiesLoadEvent or EntityAddToWorldEvent?

[slender elbow]

they do different things

[quaint basin]

what is the difference?

[slender elbow]

look up the previous 30 conversations about that exact question

that you yourself had

[quaint basin]

I think i just asked the difference between EntitiesLoadEvent and ChunkLoadEvent + event#getChunk#getEntities

[slender elbow]

one for all entities in a chunk load, one for any individual entity being added for any reason, chunk load or natural mob spawning or dropping an item etc etc

you should write that down on a notebook

[quaint basin]

I will lol

[quaint basin]

entitiesloadevent does not include natural mob spawning etc? entitiesloadevent just include chunk load?

but well. I think my logic is well done then https://pastes.dev/kT7bg1prtT

[slender elbow]

correct

[tender shard]

wow one of my mice escaped and now my bf is upset at me

as if that'd be my fault

[slender elbow]

i blame Demeter

no

I love Demeter

I blame Hermes

[placid gull]

Hello guys

I'm facing a issue with geyser mc can anyone help me ?

I'm getting end of the stream issue

[mortal vortex]

With the API?

[mortal vortex]

show your code

[placid gull]

I’m using a setup with Geyser, Floodgate, JPremium, and a Velocity proxy for cracked + premium support and Bedrock crossplay.
The issue is that Java players can join normally, but Bedrock players joining through Geyser/Floodgate receive an “End of Stream” error during login. Online mode is currently set to false on both the Velocity proxy and all backend servers. Geyser is only installed on the proxy side, not on backend servers.
I want help identifying what configuration conflict or authentication issue could be causing Bedrock players to disconnect with the “End of Stream” error.

[mortal vortex]

Okay so you aren't developing a plugin?

Maybe ask the developer(s) of the plugin.

[placid gull]

Ahh ok

[umbral ridge]

how heavy can it be for the server to make per player world system? but limited to eg 4000x4000

and for eg. 50 players?

same system like skyblock but the actual world, with worldborder set at 4000x4000, fully loaded on request via a command?

[drowsy helm]

Like a virtual world?

Or like an actual world per player

[buoyant viper]

chat how does this perf metric look so far
TPS is updated once per second, MSPT is updated every tick (to explain visual discrepancies)

[weak isle]

So many smart people

I need a dev for my smp

[buoyant viper]

me 2

[umbral ridge]

what do you mean by virtual world?

an actual world

[orchid rapids]

What happened to org.bukkit.Server.getMinecraftVersion()? I see several people have discussed its absence years back as well, but it always been present in the Javadocs as far as I can tell

But now it's fully gone in 26.1

I know getVersion() does the same thing, but I want to know if it was intentionally removed or not

[drowsy helm]

i mean each world is negligble overhead so willl just be similar to having that many chunks loaded normally

something like SWM would definitely help though

[umbral ridge]

slime world manager?

ill check it out

[orchid rapids]

From md_5's post about 26.1:

There are currently no significant API changes since 1.21.11 and no breakages expected for plugins using the API in this release.
So I guess it was not removed intentionally

[buoyant viper]

?jd-s unrelated

[undone axle]

https://hub.spigotmc.org/javadocs/spigot

[buoyant viper]

?stash

[undone axle]

https://hub.spigotmc.org/stash/

[mortal vortex]

#bot-commands

[buoyant viper]

fake news

[chrome beacon]

POV: you're asking about a Paper method on Spigot

[young knoll]

2b2t spawn

[orchid rapids]

Except it's in Spigot's Javadoc for 1.21.11 and older

[chrome beacon]

Do show me that

because it is not

[orchid rapids]

No wait, that's not what I checked, because Spigot doesn't keep hosting old docs

Hrm

[chrome beacon]

helpchat has them

You're free to take a look

[orchid rapids]

Yup, I believe it. I had docs for several versions of Paper open in addition to Spigot's when trying to figure out where the method came from and where it went

Must have confused myself

Too much blood in my caffeine stream today

[solid wave]

anyone good at mythicmob?

pls help me

[chrome beacon]

?ask

[undone axle]

If you have a question, please just ask it. Don't look for staff or topic experts. Don't ask to ask or ask if people are awake or available. Just ask the question to the channel straight out, and wait patiently for a reply. Make sure you use the right channel regarding the topic of your question. Create a thread in case the channel is already in use!

[solid wave]

Does anyone know how to make a mob when it spawn and get kill ,it will send a messenger in chat so everyone could see

[thorn isle]

mythic is kill

[ebon quarry]

How to make auto stasis trigger work in paper servers?

[chrome beacon]

?whereami

[undone axle]

https://cdn.discordapp.com/attachments/864180468680360027/944415049462714469/whereami.mp4

[ebon quarry]

?

[thorn isle]

MrBozo

[ebon quarry]

Pls

[robust helm]

broski

[smoky anchor]

Sire, this is Spigot

[robust helm]

paper is a separate project with a separate discord

[ebon quarry]

Where is paper dc server

[smoky anchor]

Try googling for once

[robust helm]

https://letmegooglethat.com/?q=paper+discord

[hybrid spoke]

@young knoll

[young knoll]

?ban @wheat lance spam bot

[undone axle]

Done. That felt good.

[hybrid spoke]

thx!

[placid birch]

Whats recommended way of making custom currency? Vault, persistent data storage, something else?

[chrome beacon]

Vault is just an API between plugins, and if you want persistent data storage there are multiple ways of going about it depending on your needs

[placid birch]

Anything you would recommend? Kinda new, trying to convert my skripts to java plugins so the currency needs to work across multiple plugins

[thorn isle]

does vault even support multi-currency?

[chrome beacon]

It doesn't

[thorn isle]

i remember it supports per-world currency, but not so sure about multi-currency

to be clear, are you intending to have a "custom" currency in the sense that the server will have "dollars" as a regular currency, and like "gems" or whatever the shit as a different "custom" currency?

or are you just trying to hook your plugin into a currency in general, be that dollars or whatever else the server admin calls it by?

[placid birch]

Yes, multiple custom currencies like gems, tokens etc, not like name matters but yes

[thorn isle]

i think reserve supports multi-currencies, but from what i know its adoption is mostly limited to whoever the towny guys can get their hands on

[sly topaz]

you can have multiple economy providers however the consumers of the economy services always choose the last registered one

because who would go out of their way to for loop each economy service on each transaction

[thorn isle]

currency needs to work across multiple plugins
if these plugins are all yours, then you can just implement the currency stuff yourself

vault is for when all plugins, including those you didn't write, have to play one tune and work with the same currency

you can still expose your own api for other plugins to hook into to do transactions and checks on your custom currencies

[sly topaz]

as far as economy goes, I think Vault is pretty much useless anyway

or was it permissions

[thorn isle]

permissions probably

bukkit has its own permissions system anyway, i think permissions is only ever used for like getting "metadata" like the primary group and nick and other things popularized by role plugins

[sly topaz]

there's also been multiple spinoffs of Vault like VaultUnlocked centered on improving the Economy service, also Vault 2.0 which is like, just an economy provider

[thorn isle]

don't forget reserve

the 20,000 line god class vault replacement

horrible horrible

[sly topaz]

long archived, luckily enough

[thorn isle]

good

did they also dump TheNewEconomy in the same pile?

[sly topaz]

they stopped maintaining it, yeah

but it was maintained for fairly longer than reserve

[thorn isle]

nature is healing

[buoyant viper]

this is me as hell

im out here usin Fairy Dust as my currency

fully custom eco solution with integration in my own plugin suite 🤞

[quaint basin]

block#breakNaturally calls blockbreakevent on some circunstance?

[young knoll]

Don’t think so

Only Player#breakBlock does

[mortal vortex]

Anyone using XSeries? https://github.com/CryptoMorin/XSeries

(i might have to "dont ask to ask" myself)

I mean, does it truly bridge the gap between Spigot and Paper, or is it not as useful as it seems?

i misunderstood the point of the lib, nvm.

[ivory sleet]

icl forgot that existed, wasnt it more for version mapping materials and other registry types

[mortal vortex]

it is yes

i thought it would help with spigot -> paper mappings

[ivory sleet]

:,)

[mortal vortex]

was hoping something like PaperLib still existed

[fathom fable]

i used it just for version compatibilities from 1.8 to 1.19 but then disabled it honestly

[young knoll]

Does it not?

[mortal vortex]

does what not what

[young knoll]

PaperLib still exist

[mortal vortex]

it dont lad

[young knoll]

I mean, it’s archived but all the code is still there

You can just yoink it

[mortal vortex]

wow thanks genius

get this guy a nobel prize

[young knoll]

Skill issue

[mortal vortex]

https://tenor.com/view/an-iq-too-high-raven-crow-tiktok-gif-3869671553367641450

[quaint basin]

how can i add item flags with datacomponenttypes?

[chrome beacon]

https://docs.papermc.io/paper/dev/data-component-api/

[quaint basin]

yes but i can't find itemflags thing

to hide the enchants

i can add enchants but i can't hide them

[chrome beacon]

You can just use the addItemFlags method

[thorn isle]

item flags aren't a thing in nms

see https://minecraft.wiki/w/Data_component_format#tooltip_display

[chrome beacon]

that too

[kind hatch]

This might be as simple as adding a .gitattributes file and doing a little bit of resource filtering for some metadata. (Has to be committted for it to apply)
I added this to the repo and it does a lot of heavy lifting. This file overrides any autocrlf settings as long as the file type is specified. My diffoscope outputs were much smaller because of it.

The issue I had with metadata was it not having the correct line endings. Specifically on these files.
META-INF/libraries/spigot-api-26.1.2-R0.1-SNAPSHOT.jar/META-INF/maven/org.spigotmc/spigot-api/pom.properties
META-INF/versions/spigot-26.1.2-R0.1-SNAPSHOT.jar/META-INF/maven/org.spigotmc/spigot/pom.properties

Apparently, the maven-jar-plugin dynamically generates the pom.properties during execution using system specific line endings.
This is unfortunate since there is no tag that can be used to change this behavior other than <addMavenDescriptor>, but that would end up removing both the pom.properties and the pom.xml.

However, I got around this by adding the file to the project and filtered it in with the resources plugin.

Spigot-API/
└── src/main/resources/META-INF/maven/org.spigotmc/spigot-api/pom.properties

Spigot-Server/
└── src/main/resources/META-INF/maven/org.spigotmc/spigot/pom.properties

I added this to both poms. (spigot & spigot-api)

<resource>
  <directory>src/main/resources</directory>
  <filtering>true</filtering>
</resource>

That just left me with the .list files in the outermost jar which could thankfully be fixed in the assembly file by adding the <lineEnding>unix</lineEnding> tag to the files and filesets.

Those four changes are what I did to get a reproducible build across operating systems.

There is another approach which involved unpacking the entire jar, forcing the correct line endings, and then repacking it all. However, it comes at the cost of more computation and a longer compile time.
I opted for the resource filtering because of that.

[tender shard]

.setData(DataComponentTypes.HIDDEN_COMPONENTS, ...) ? or what do you mean?

[quaint basin]

this use metadata

[quaint basin]

thx

[quaint basin]

is not this? https://minecraft.wiki/w/Data_component_format#tooltip_display

[kind hatch]

Isn't that paper api?