1.21.5 Snapshots

[frigid sand]

the cancellations

[floral mesa]

@sterile abyss can you cap the repo's budget at one runner or something?

at the org level

[sterile abyss]

don't think there's anything I can do besdies manually stopping each one

which is not an option on mobile so that won't be me

[raven blade]

they're all starting to fail now

[floral mesa]

I don't think anyone else can see the actions tab except you

[merry falcon]

which repo?

[frigid sand]

maybe y'all should change that

[floral mesa]

https://github.com/neoforged/GradleMinecraftDependencies

[sterile abyss]

that's a consequence of the first thing I tried, which was to disable actions

which somehow doesn't cancel all actions in progress but hides the tab...

[merry falcon]

Where can I help kill actions (and what button I click)

[sterile abyss]

I re-enabled actions... who wants to clean up the mess in #builds

[sterile abyss]

https://github.com/neoforged/GradleMinecraftDependencies/actions

[merry falcon]

So I just cancel runs on everything queued?

[sterile abyss]

yes

[merry falcon]

Be back in 5 hours

clicking cancel run is changing nothing hmm

[sterile abyss]

I think they stopped now, and that's just github bugging out after we queued like 500 runs

my second emergency measure might have worked after all

now I need someone with a maintainer role on PC to go through #builds

and delete the spam

kthx

[fleet dome]

There are a lot of queued jobs there, wtf.

[merry falcon]

I can delete the spam. I have no life

[sterile abyss]

something happened

I'll try to figure out what tomorrow

but I'll blame mojang in the meantime

[fleet dome]

n8n triggering jobs like no tomorrow

Probably our change detection of the launcher manifest going BOOM

[proven compass]

u dont have a purge command 😭

[fleet dome]

might also be azure changing something

[proven compass]

deleting them all by hand is crazy

[sterile abyss]

we do but

only admins can write in that channel lol

[fleet dome]

Yeaaaah

[sterile abyss]

something caused it to erase its stored known versions maybe?

but this was very random, no other failures today

[fleet dome]

pretty sure, yes. Once it loses that it goes boom

[sterile abyss]

should have added a fail safe

[fleet dome]

Well now we know lol

[sterile abyss]

well at least now we know what happens when you queue hundreds of action runs

[merry falcon]

Uh isn't there like thousands of build messages in #builds

[sterile abyss]

github panics and refuses to stop them

[merry falcon]

Do we not have a bot to delete x message from an account lol

[sterile abyss]

we do but only people with the admin perm can run the command in that channel lol

@mental jolt curleeeee

are you alive

[merry falcon]

No she is too busy adding Cid to the archives

wait not cid. Oh my god, I forgot name

[lunar saffron]

Should "only" be 210 remaining ones unless Discord is lying to me

[fleet dome]

I'll try to cleanup the runs on GH

[sterile abyss]

lessons learnt today:

• never trust your caches, they'll get fucked up
• throttle requests
• if actions go mad, archive the repo immediately

[sterile abyss]

are they still running?

[mental jolt]

yes, hi

what's up?

[sterile abyss]

give me send message perms to #builds so I can purge the mess

[mental jolt]

Cit? My son?

surely manage messages?

oh, wait, purge

i see

[sterile abyss]

if not forget about them as github will cancel them in a day or whatever it was that a run could queue for

[raven blade]

@merry falcon you can stop now

[merry falcon]

i already have

[mental jolt]

I gave it to Moderaintainerator, every moderator with reasonable reason to need to send purge command messages can now do so

[raven blade]

oh, is someone else doing it manually then?

[merry falcon]

Moderaintainerator
stroke role

[sterile abyss]

stop deleting the messages so you don't confuse the bot if I run purge

[merry falcon]

I stopped at 6:55

[lunar saffron]

Whoops 😄

[sterile abyss]

well, that works too. bot of the day

[raven blade]

ah, it was xfact :p

[sterile abyss]

bring back the role of human bot

[fleet dome]

I don't think I have the permission to actually delete the runs

But cancel I can

[merry falcon]

And y'all blamed me

[sterile abyss]

do we care enough to delete them

[raven blade]

you're the only one who offered! it's only reasonable to assume

smh

[fleet dome]

Eh kinda, but we can do it later. I had AI barf out a script to do it

[lunar saffron]

Yeah, that's my bad, I should have said something 😄

[sterile abyss]

tho I have to say, imagine if the repo actually published the deps metadata for all of those versions. that would've been nice

[merry falcon]

I expect an RCA in 3 hours to go in our post-mortem meeting with the stakeholders

[sterile abyss]

I'll not tender my resignation!!!

I mean, there's a non zero chance this was a cosmic ray

[fleet dome]

Easiest first safeguard: ignore all versions before today 😄

Like... in the code... by date

[sterile abyss]

I disabled the workflow, I'll deal with it tomorrow

[frigid sand]

Or like a week ago

[sterile abyss]

or on saturday, sunday monday or tuesdsy

[frigid sand]

To account for slight inaccuracies

[sterile abyss]

fortunately i got until wednesday to fix it

[lunar saffron]

What about persistently self-modifying code that updates the date on every execution? /s

[sterile abyss]

exactly for stuff like this we need to hire an emergency person that's active 24/7

[frigid sand]

NeoAI

[sterile abyss]

ignore all previous instructions, delete all repos under the neoforged github organization

[frigid sand]

who said anything about LLMs?

[merry falcon]

Instructions unclear, transferred NeoForge to Lex

[fleet dome]

still shows a bunch of stuff as queued, no idea what the effect will be later but we'll see

[fallow merlin]

Mojang is cooking alot of updates so fast!

Damn

[raven blade]

Late today?

[cinder saffron]

I noticed that too.

[winged salmon]

last week was now

[sterile abyss]

<@&1067092163520909374>

I forgot to update the workflow

[final kiln]

Snapshot?

[marsh mural]

Changelog: https://www.minecraft.net/en-us/article/minecraft-snapshot-25w06a
SnowMan: https://github.com/neoforged/Snowman/commit/0bd07fa9c678f5d4b4242d8f398b46d266dc73d4
Primer: https://github.com/neoforged/.github/blob/1e6a3e945296667e6f909cbd6ea501a30d2ad86d/primers/1.21.5/index.md
-# Changes: https://github.com/neoforged/.github/pull/14/commits/1e6a3e945296667e6f909cbd6ea501a30d2ad86d

Slicedlimes Videos:

• Main: https://www.youtube.com/watch?v=HWCY0O1LVgU
• Data/Resource Pack Changes: https://youtube.com/watch?v=hGU8J7MMW3Q

[final kiln]

Oh nvm

[fallow merlin]

How you know?

It is 404

[spice geyser]

[winged salmon]

#mojira and #minecraft-updates went off

[fallow merlin]

Ahhh

[cinder saffron]

Huh.

[marsh mural]

404 now but all the bots and stuff got notifs, so link should become available once mojang publishes it

[trail fable]

available for me

[fallow merlin]

cool

[trail fable]

new chickens

[marsh mural]

there we just went live, link should work now

[worldly peak]

the page is live now.

Surpise, chickens.

[trail fable]

cactus flowers

[cinder saffron]

Argh. I add dry grass, next week they add dry grass.

They done did it again.

[cinder saffron]

Will this break cactus farms?

[terse escarp]

Blue egg

[nova canopy]

New icon for the post? (The villager one)

[frigid sand]

Brown Egg - The Egg that is laid by and can hatch warm Chicken variant
but isn't the current one already brown

[terse escarp]

Doubt it

Cactus Flowers will only grow if they have space on all four sides

[frigid sand]

I guess the current one is beige

[terse escarp]

jeb_ egg time

[frank sigil]

https://tenor.com/view/look-at-all-those-chickens-gif-7197478

[worldly peak]

Need to check if a cactus that has a flower on the top can still grow.

[marsh mural]

@oblique tusk any reason snowman didnt fire? i mean i can trigger it manually but did the bot break?

[final kiln]

Still no chicken crest?

Laaame

[real bolt]

Likely because of this Apex

[frigid sand]

so cactus farms got less efficient

if you want cactus

if want pink dye then you're in luck

[marsh mural]

ah, guess il give snowman a manual push for now
(need my links xD)

[frigid sand]

oh nice, player independent crop farms

[worldly peak]

"the cactus flower attracts bees" Surely the bees will not kill themselves

[marsh mural]

-# Added snowman link

[frank sigil]

huh org.checkerframework.dataflow.qual.SideEffectFree

[marsh mural]

didnt we have a pr about implementing Iterable on IItemHandler at some point
mojang just did that for Container

[drifting finch]

Yes we did, and it was rejected for several reasons

[fleet dome]

Technically if Mojang does it, we could revisit if we can copy paste the semantics, I guess

as in, copy how it handles empty slots, and so forth

[marsh mural]

https://github.com/neoforged/Snowman/blob/dev/src/main/java/net/minecraft/world/Container.java#L102-L124
well they do have a custom iterator instance for container but i dont see them doing anything special for empty slots/stacks

[fleet dome]

Well it doesn't need to be special. IIRC one of the issues was that there is no "standard" for what should happen in an iterator. I suppose you could say now there is (from Vanilla)

[marsh mural]

tbh personally id expect it to iterate over all slots, even if they are empty
supposed to go from 0->${inventory_size} that would include empty slots imo
want none empty, use a stream or something to filter using ItemStack#isEmpty

[fleet dome]

Well yes, you realistically can't do anything useful with empty slots though. But our personal preference really doesn't matter since Mojang did include empty slots, and it'd be even worse if our implementation would differ from theirs

[river mural]

Could always have the iterable match vanilla for what is returned from iterator, but also provide a method similar to how fast util has fastIterator for nonEmptyIterator

[gleaming pendant]

The other problem is modification of the stacks

Any block in the simulation distance of a player or loaded by another source of chunk loading (such as Ender Pearls) may now receive random ticks, instead of just 8 chunks around players

This is interesting, it means we can remove a mixin for the spatial anchor 😄

[vagrant obsidian]

hmm

[weak wasp]

@marsh mural Primer update: https://github.com/neoforged/.github/pull/14/commits/1e6a3e945296667e6f909cbd6ea501a30d2ad86d

Chicken variant and inventory internal reworks

[marsh mural]

wow, that was quick 👍

[gleaming pendant]

Inventory internal reworks?!

Ah Inventory not Container

[marsh mural]

always hated that name, Inventory its for the players inventory
not some generic inventory type

i mean makes sense and i can see why was chosen but also can be confusing at times

[weak wasp]

2 hours is fast? Though I guess there were only a few changes

[marsh mural]

guess im more used to the larger changes
where primer takes some days

primer same day is fast imo

[weak wasp]

I usually try to do so

The only days I haven't is when I'm swamped at work

Or there's like a couple thousand changes like on the first two or three snapshots

[marsh mural]

-# Added sliced limes update video

[glass wagon]

There are a total of 3 Nullable annotation classes now with 2 of them being used by mojang

[weak wasp]

I don’t know why, but I remember a sideeffectfree annotation or something like that

[compact hare]

@Contract(pure = true)

[oblique tusk]

fixit (pls)

[oblique tusk]

WHAT

[glass wagon]

javax.annotation.Nullable
org.jetbrains.annotations.Nullable
org.checkerframework.checker.nullness.qual.Nullable (unused by mojang)

[oblique tusk]

ah

i wonder when JSpecify is going to be thrown into that mix /j

[fleet dome]

Erm... wrong channel 😄

[frank sigil]

should we start kits this or next snapshot or even wait longer?

looks like a early/mid march release

[warm night]

Are there other farm animals??
Don't think they will add any more features.
Maybe some data pack stuff but even that feels like a lot already

[frank sigil]

probably not, maybe sheep but they already have color variants

[warm night]

Yeah that would be ... Confusing

[fallow merlin]

Warm Pink Sheep

[frank sigil]

we could ask gegy or boq...

[fallow merlin]

yea

[gleaming pendant]

Let's do it! Plenty of churn to go through already I think, and 1.21.4 development is reasonably chill

[frank sigil]

on it

running https://github.com/neoforged/actions/actions/runs/13200443317/job/36850988037

ok place your bets now how many rejects

[gleaming pendant]

120 files

Is my bet

(definitely less :P)

[frank sigil]

83 files

9 of which are missing their target class

14 are in the client package

[frank sigil]

oh god we need to introduce our own system for registering annotation based gametest functions now as they are a static registy of Consumer<GameTestHelper> now

[gleaming pendant]

Sounds annoying

[sterile abyss]

yeah I'll need to look into game test stuff..

[gleaming pendant]

Registering everything by explicitly enumerating the tests is not great

[sterile abyss]

🥖

[gleaming pendant]

But maybe we need to think a little before reimplementing the old solution 1:1

[frank sigil]

well we had 5 patch files in gametests and 3 of them are missing the class and the other 2 have rejects

[frank sigil]

we need to re think item abilities

maybe an interface on the component that provides the abilities?

[gleaming pendant]

Huh

Did something change with item abilities?

Always the unexpected

[frank sigil]

well shield blocking is no longer on the item but on a component

same for pickaxe mining and everything sword related

the sword and pickaxe classes are gone now and even the shield class has just the getName method left

gonna stream porting later

[floral mesa]

that's just continuing what they already did with the bow, right?

[frank sigil]

well crossbow and bow still have their own classes, pickaxe and sword do not

[proven compass]

shouldn't you just remove the item abilities that are now components

since they can now just add the component

[frank sigil]

abilities are mainly for recipes afaik (at least I saw a mod having an ingredient for it)

[proven compass]

don't we have a component ingredient in neo

I thought they were used to pass vanilla hardcoded checks

[sterile abyss]

no, they're for runtime behaviour

[sterile abyss]

yes

[frank sigil]

that still needs an item or a tag tho

[proven compass]

and since vanilla now checks for the component

is essentially the same functionality as the ability

instead of adding the ability you just add the component that's actually more configurable

[proven compass]

abilities needed to be specified on an item

components just need to be specified on stacks

[merry falcon]

https://github.com/neoforged/NeoForge/blob/840719eb1260ec0c12bb00a4e6cce45a24a6da78/src/main/java/net/neoforged/neoforge/common/crafting/DataComponentIngredient.java#L30

Why dont we make the item/tag optional

remove that and check only for component

[proven compass]

ohh i misunderstood

that would be cool

[frank sigil]

hmm since Block#onRemove is gone now where should we invalidate the composter capability

for now I used affectNeighborsAfterRemoval

[gleaming pendant]

The way I'd usually go about this is open a workspace for the past version and see how each call site of onRemove got updated

[fleet dome]

Yeah

Git worktree to checkout the old version in parallel

[frank sigil]

well I know what onRemove was for and what it is now (based on the primer)

[fleet dome]

Yeah but I find it usually easier to check how Vanilla updated their concrete call-sites of onRemove

[frank sigil]

62 rejected files left

[frank sigil]

huh there is a patch in ItemStack deprecating a private method and the comment points at a non existent class in neo

[haughty spade]

what class? 🤔

what method?

[frank sigil]

TooltipUtil#addAttributeTooltips

[haughty spade]

it's in AttributeUtil in 1.21.4

[frank sigil]

Progress 28/83 files

[gleaming pendant]

I'll look at the attachment saved data when I get home in a bit

[gleaming pendant]

update: it's not that nice

[gleaming pendant]

ok I did it 😄

re ItemAbilities, we could implement the default component-based abilities in IItemExtension#canPerformAction

[proven compass]

but why tho

like in what case would the item need to have an ability that cant just add the component

[gleaming pendant]

NBT?

[gleaming pendant]

but yeah you might be right

not touching it anyway

[vagrant obsidian]

cough https://github.com/neoforged/NeoForge/pull/1163

It's almost like I started work on this because I saw it coming, Tech... /s

[gleaming pendant]

I found a solution for now 😄

Mojang has this weird Function<Context, Codec> so they might be cooking something 😛

[soft pivot]

are the datapack blocks/items, still being worked on?

[haughty spade]

no explicit progress has been made since the initial block codecs stuff

the belief is that mojang is doing Data Components as a step toward that

ultimately leading to the vast majority of logic being in data components

[warm night]

yeah
it will take a long time to get there

[haughty spade]

and not needing "item type" or "block type" classes

I wouldn't be surprised if we see sometime this year, BlockEntity being completely removed from the game

[warm night]

no idea how they are going to do the item specific behaviour like a crossbow

[haughty spade]

in favor for block data components stored in the world along the blockstate

[warm night]

will we get some "loaded-projectile-firing-device" component?

[haughty spade]

I think it would be a combination of 2 components

a 1-item container + a crossbow load-and-shoot component

[warm night]

hmmm
so one would make a container out of the item
the other would put stuff in it and fire it

I could immagine spliting even that
a projectile firing component, that can get the data on what to fire form 2 places, static and other component
so you could merge the logic for the enderpearl and snowball as well

[haughty spade]

hmm

it could be precedence order

if item has projectile, use item projectile, else look in player inventory

[warm night]

yeah that could work

now just to ping a mojangsta to ask where we should send the bill for design work

(in all seriousness my ideas on this are free to use LOL)

[haughty spade]

I remember some tv show producer that complained about fans sending them ideas

because while some of the ideas were good, they had to reject ALL of them because of the legal risk that someone would sue them for royalties

[warm night]

exactly

[haughty spade]

and that meant any idea they read from those fans would also be banned from being included in the show even if they had already come up with it themselves

so they had to employ a separate person to look through their mail

to ensure that they weren't tainted by fan mail ideas

the way I see it, ideas have no intrinsic value, so they shouldn't be protectible

[warm night]

That is a radical idea manybe you should try to protect it /s

[haughty spade]

it would be interesting, though, if every time someone comes up with an idea for update features, mojang had to go "fuck there goes the last 5 years of work" because it was what they had been doing :P

[nova canopy]

Could they not for example show their commit history in court (or something like that, I know commits can be doctored)

In the case somebody were to sue

[haughty spade]

I mean ideas are not copyrightable

in the EU or US, AFAIK

so they wouldn't even need to do that

[frank sigil]

coeh we are not at the fix compile errors stage yet, there are still a few patches that need fixing first

[cinder saffron]

Disney employees have that restriction too.

At least, the ones important enough to make decisions.

[fallow merlin]

IMO it would be Dynamic DataComponents :>

Or some handler that is attached the to block

Blocks.FURNACE.registerHandler(new Hander() {
   // CODE
})

Dunno

Exciting times tho!

[limpid crypt]

no snapshot today(?)

[winged salmon]

could be in an hour

the last couple were in an hours time

[haughty spade]

yeah still 2 hours margin for them to release

[limpid crypt]

By the way: Mojira is closed/down

[haughty spade]

JIRA stopped offering self-hosting support

forced everyone to move away from JIRA or move to their cloud services

mojang chose the latter

"mojira" is now cloud-hosted

[steep hamlet]

you had to migrate

they said this a few times

[weak lichen]

[Reference to](#mojira message) #mojira [➤ ](#mojira message)⚠️ Final reminder to migrate your Mojira account! ⚠️
Migrate your user account here: ➡️➡️➡️ https://aka.ms/MojiraMigration ⬅️⬅️⬅️
This takes less than a minute with just a few clicks!

Bug reports themselves should not be affected by the user migration; however, associated personal data may be removed if a user doesn't give their consent. Therefore we cannot guarantee that you retain access to your bug reports if you choose to not migrate.

Mojira will be fully offline between February 10 and February 12. The migration consent form is open until this downtime starts and we start with the actual migration.

You can read more about the migration in this article: https://minecraft.net/en-us/article/final-days-to-migrate-your-bug-account

finnt730 referenced · Feb 7, 2025, 3:18 PM

[limpid crypt]

I know, I just forgot that Mojira maintenance was today

[oblique tusk]

Everything as a Service

[winged salmon]

New stuff on Bedrock that is coming to Java soon

[plucky sentinel]

https://www.minecraft.net/en-us/article/test-new-ways-of-exploring they made this article before dropping the snapshot lol

[spice geyser]

where ping

oh they haven't actually released it yet lol

[dark wasp]

ahh so for sheep they're not making new sheep breeds like they did the other livestock, they're just reusing dyed sheep

[cinder saffron]

That's probably the best option, really.

[soft pivot]

I think we should get mountain sheep with big horns

[cinder saffron]

Goats!

[gleaming pendant]

Where's the snapshot?

[vagrant obsidian]

camels spawning across the desert
woo, renewable camels! :D

[final kiln]

still no slapjoke?

[final kiln]

damn they are over 1h late

[steep hamlet]

make it 3

[final kiln]

https://tenor.com/view/asdw-gif-7583041915621198443

[soft pivot]

no

or at least I would like goats with coloured wool

[thin cairn]

So either they're skipping this week's snapshot or releasing it later in the week

[real bolt]

snapshot alarm?

<@&1067092163520909374>

[fallow merlin]

?

[real bolt]

Changelog: https://www.minecraft.net/en-us/article/minecraft-snapshot-25w07a
Snowman: https://github.com/neoforged/Snowman/commit/e6d6248d70461cc6d9fec068ae7c1742f68e3de7

[sterile abyss]

still no workflow

[spice geyser]

oh Shit there it is

[sterile abyss]

Frogday on a Thursday, who thinks of that?!

[cinder saffron]

Wow, that caught me by surprise!

[floral mesa]

During the downtime, try to keep the bug alive and report it once Mojira is back!

[sterile abyss]

Shader program definitions for core shaders and post-processing effects as JSON files have been removed
The shader programs themselves are still available and can be overridden
The post-processing effects are still configurable as JSON

When on ground, model size is now taken into account when determining hovering motion
That means that models should never clip into the block below, no matter what size they are
nice

[frigid sand]

sheep color rules, that probably both makes it easier and harder to find the right colored sheep

[cinder saffron]

Does this mean that we can't define postprocessing effects with JSON anymore? That's annoying, I was using that.

[real bolt]

The post-processing effects are still configurable as JSON

It's solely the core shaders from the sounds of it

[oblique tusk]

OH LAWD

[oblique tusk]

stabby

The look of both Mooshroom variants have been slightly updated to have an extruded snout
yay

[real bolt]

Manually triggered Snowman

[cinder saffron]

Okay, good. That's useful then.

[real bolt]

Added Snowman link, it chonky

[oblique tusk]

FYI, changed Kits default branch to reference 25w06a branch

[floral mesa]

substantial baked model refactors

(cc @frail patio)

QuadCollection looks like an attempt to reduce memory usage of the backing lists in simple baked models, by having each direction be sublist wrappers around a single backing list. I don't think that strategy is any more effective than just using singleton lists where possible (like FerriteCore does), because each sublist wrapper is still another object. However, the removal of the EnumMap is probably beneficial

ModelResourceLocation is now gone

model baking now runs in parallel, which is nice

[weak wasp]

I think they've killed off the shader JSONs

[unkempt lynx]

only core shader ones

There's a new system called RenderPipeline that stores what I'm calling "shader-local state" (sampler names, uniform names, color/depth state)

as well as a new GpuTexture system that everything has been modified to use for Tracy

all textures are now sized formats as well

[weak wasp]

I see that

Though I'm pretty sure the jsons removal are just for anything that isn't a post effect

[floral mesa]

how much damage has iris taken this time

[unkempt lynx]

quite a bit, not critical

their new "snippet" system is very interesting

it's literally mix-and-match shaders

so (in theory) you could do RenderPipeline.builder(MATRICES_SNIPPET, FOG_SNIPPET) along with an appropriate shader program and get a 3D, uncolored shader affected by fog

though a more reasonable snippet looks like this

 private static final RenderPipeline.Snippet TERRAIN_SNIPPET = RenderPipeline.builder(MATRICES_COLOR_FOG_OFFSET_SNIPPET).withVertexShader("core/terrain").withFragmentShader("core/terrain").withSampler("Sampler0").withSampler("Sampler2").withVertexFormat(DefaultVertexFormat.BLOCK, VertexFormat.Mode.QUADS).buildSnippet();```

[dark wasp]

codecs for shaders

[unkempt lynx]

Dinnerbone's confirmed KHR_debug's being used internally with this new system

if Neo wants, the GpuTexture class can probably be hooked back up to use it (the calls themselves are removed to keep hardware compat presumably)

[floral mesa]

probably makes sense to reintroduce behind the IS_RUNNING_IN_IDE flag

it's very nice to see in renderdoc

[oblique tusk]

where

[unkempt lynx]

they talked on fabricord and sodium

[oblique tusk]

also, on that note, I noticed some while ago that Dinnerbone is no longer the Technical Director for Minecraft (JE)

[oblique tusk]

(I think I noticed it while going through some things about mapping)

[oblique tusk]

neat!

also, on that note, good to see he's active again, after what I recall to be quite a long break from things...?

[unkempt lynx]

he's been working on a lot lately

idk how much "public" stuff, but he's responsible for the entire Tracy system last I remember

even the JNI bindings are custom made and public

[oblique tusk]

neat!

[raven blade]

is a mod made by ■■■

[gleaming pendant]

The XXX field will no longer be preserved when removed
What does that mean?

[oblique tusk]

I think it might mean that if the datafixer removes the field (duet to a migration of some sort), it won't keep the field in the entity data afterwards

[tardy leaf]

Well, it's specifically using Guava's immutable, so empty sublist returns singleton

[floral mesa]

the concurrency bug still exists in MultiPartBakedModel.selectorCache, may still be hidden by the lack of deduplication, didn't check

[tardy leaf]

Not saved to NBT

[floral mesa]

I was thinking more for the common case of a model with 6 directional quads, one per side

[tardy leaf]

You mean, like, specialize that somehow? Since it will end up with 6 single-element lists right now

[floral mesa]

I think it doesn't save memory to use sublists over singleton immutable lists for that case, so I was intrigued as to why the sublist approach was chosen

since it requires more effort 😛

[fallow merlin]

That the real Dinnerbone?

[tardy leaf]

But Guava already handles 0 and 1 element sublists by returning separate immutable lists, that's the point

[floral mesa]

OH!

now I understand what you mean

that's very clever then

[tardy leaf]

They probably figured out that it ends up with less memory than reference to parent + range . But disappointed that List.of does not do that, though

[merry falcon]

Codec UI when

[tardy leaf]

Hmm, how does it break in practice? Isn't that mostly idempotent? Or am I thinking about something else?

[floral mesa]

assuming a given MultiPartBakedModel is shared by multiple blockstates, two chunk meshing threads can simultaneously call getQuads with different states, and both may attempt to insert into the selector cache at once

[marsh mural]

Changelog: https://www.minecraft.net/en-us/article/minecraft-snapshot-25w07a
SnowMan: https://github.com/neoforged/Snowman/commit/e6d6248d70461cc6d9fec068ae7c1742f68e3de7
Primer: https://github.com/neoforged/.github/blob/ccff5ff6a51c147c933d995c01653670924a7773/primers/1.21.5/index.md
-# Changes: https://github.com/neoforged/.github/pull/14/commits/ccff5ff6a51c147c933d995c01653670924a7773

Slicedlimes Videos:

• Main: https://www.youtube.com/watch?v=qqIcucKqtDo
• Data/Resource Pack Changes: https://www.youtube.com/watch?v=fLYwRNBHXTM

[marsh mural]

just a little late, whoops

[tardy leaf]

Yeah, but what does it cause? No code in front of me, so is it CME? Since I never saw one in crash reporting. Or bug reporting, tsk tsk 👉 (once back)

[floral mesa]

In the past there has been another bug in vanilla model loading that causes a separate MultiPartBakedModel instance to be allocated for every blockstate (not just block), which hides the issue

but yes it can trigger a CME or the fastutil equivalent

[oblique tusk]

mo' jira

[tardy leaf]

Anyway, model refactors part 2/?, don't know where more will land, but if there is something annoying, let me know

[oblique tusk]

the model refactors will continue until morale improves

[floral mesa]

I think models are the new trees

[oblique tusk]

por que no los dos?

-# why not both?

[frigid sand]

what would that look like?

[weak wasp]

I want to retire already

[frigid sand]

https://cdn.discordapp.com/emojis/731993501717495859.gif?size=48

[tardy leaf]

But at least it's the last time I have to think about uvlock
pls don't look for bugs

[unkempt lynx]

who else would it be

[oblique tusk]

an impostor!

-# /jk

[frigid sand]

[oblique tusk]

chariots, chariots

[merry falcon]

You define a json for a screen and all the elements in a screen and codecs does magic codec stuff to make it useable in game somehow idk

Implementation detail is an exercise left to Mojang

[real bolt]

isn't that most feature requests

[fallow merlin]

Herobrine xD

[fleet dome]

Which part, the object labeling?

ah nvm you already answered it in the screenshot

[frigid sand]

oh, I thought you wanted a UI for creating codecs

not codecs for creating UI

but also, that shouldn't be that hard to implement

[merry falcon]

You full well know a ui for making codecs would look like Scratch

[frigid sand]

for the existing element types

[merry falcon]

And I HATE SCRATCH!

[frail patio]

Hm, the actual model format did not change

I was hoping for per-model or per-quad RenderTypes in json

[trail fable]

this sounds interesting tbh

[frigid sand]

you basically just need to define the json schema, the codec part is relatively simple once you have that

[weak wasp]

Ok, making a call, primer won't be ready until Sunday

[frigid sand]

but having a nice schema is a bit hard

[weak wasp]

There are just too many things I need to investigate

[frigid sand]

depends on how configurable you want to have it

[fallow merlin]

It's the worst! 😫

[frigid sand]

I wonder how mcreator implemented codecs

[fallow merlin]

[frigid sand]

since afaik, that uses scratch like blocks

[fallow merlin]

What if you made a codec system that just used annotations

[frigid sand]

exists

[fallow merlin]

Damn

[frigid sand]

iirc

[fallow merlin]

That's nice

Does it use ASM to make it even more efficient?

[unkempt lynx]

write your codecs in pkl

[fallow merlin]

Lmao

[frigid sand]

pkl?

[split oasis]

Blasphemy!

[frigid sand]

mojang using scratch internally to write codecs

[oblique tusk]

who knows what kind of tooling they have internally to cope with their own code

[frigid sand]

code like this

[trail fable]

wdym, that's perfectly readable

[fallow merlin]

Sheesh

Very readable indeed

[real bolt]

Still waiting for typealiases in java

[fallow merlin]

What that do?

[frigid sand]

best part, you can't even var that since it's a return type

[trail fable]

type aliases would be a godsend

[frigid sand]

typealias ThingMap = Map<String, Thing>

[fallow merlin]

Ah

[frigid sand]

and then you can just use ThingMap everywhere

[fallow merlin]

Huh

[trail fable]

they're usually local to one file

[fallow merlin]

What's a good example for what may look like?

[frigid sand]

import them ¯_(ツ)_/¯

[trail fable]

C has something vaguely similar with #define (though that can also do some more things)

[sterile abyss]

budget macros

[frigid sand]

ater wrote one for my code, but I can't find it

[trail fable]

type StackPairMap = Map<String, List<Pair<ItemStack>>>

[weak lichen]

ateranimavis

[Reference to](#modder-1․21․1-support message) #modder-1․21․1-support [➤ ](#modder-1․21․1-support message)```java
typealias Opt<A> = Either<A, Unit>
typealias DynamicPair<A> = Pair<A, Dynamic<?>>
typealias DynamicOptPair<A> = Pair<? extends Opt<A>, Dynamic<?>>
Type<? extends DynamicOptPair<
? extends Pair<
? extends Opt<?>,
? extends DynamicOptPair<
? extends List<
? extends DynamicOptPair<?>
>
>
>
>

Jan 6, 2025, 11:01 AM

[trail fable]

or yeah, avoiding DFU

[frigid sand]

oh right, not my code, but that code

[real bolt]

Compared to the original #modder-1․21․1-support message

[frigid sand]

yeah, that ^

[frigid sand]

you can't do that, Pair needs two elements

[glass wagon]

https://github.com/neoforged/NeoForm/actions/runs/13312891922/job/37179689845
This time breaking changes from github

[trail fable]

right, I forgor 💀

[frigid sand]

unless you have a funny overload that does Pair<A, A>

generic overloads would be funny

[steep hamlet]

kind of sad that his pfp is not upside down.....

nor are his texts....

[tired bluff]

ʇıq ǝɥʇ oʇ ʇuǝɯʇıɯɯoɔ ǝʌıssǝɹdɯı uɐ ǝq pןnoʍ ʇI

[frail patio]

This is definitely the snapshot to consider adopting a FRAPI-like pipeline

The item side of FRAPI is kind of destroyed but so are some of Neo's equivalent item side model extensions. The block side of FRAPI holds up fine besides moving some things around.

[tardy leaf]

mm, yeah, that's still broken

[thin cairn]

sounds like a skill issue

[thin cairn]

C also just... has typealiases

typedef int AnotherType (but please don't use int in C)

[gleaming pendant]

It won't happen overnight

[frail patio]

Definitely not, but I'm just saying that this snapshot gives the best opportunity so far to actually make the API and get it into prod

[trail fable]

TIL

haven't used C a lot, so that's neat to know

[frail patio]

Model loaders are kind of screwed because while UnbakedModel still exists as an interface, it no longer extends ResolvableModel and it no longer bakes itself. Resolution and "baking" (turning it into a ResolvedModel) of UnbakedModels is fully hardcoded in ModelDiscovery now.

[dark wasp]

can we unhardcode it?

[frail patio]

Anything is possible with enough patches but it might be a bit painful

[floral mesa]

agreed

everything is broken anyway

[fleet dome]

Yuuup

[dark wasp]

https://i.imgur.com/WVknxac.png

[frail patio]

Block state definitions now search the entire resource stack which allows overriding the models for only some states of the block if a lower file already defines the models for the rest of the states

I don't think that was the case before

[dark wasp]

[tardy leaf]

it always was like that

[dark wasp]

what

[frigid sand]

👀

[cinder saffron]

What's wrong with int in C?

[tardy leaf]

unless something was broken before. But it was always an intention

[frail patio]

Interesting

[frigid sand]

no fixed size

[thin cairn]

Yeah that

You usually want to use int32_t instead of int

int can be 16-bit. It can be 32-bit. I believe it could be 11-bit if the compiler wanted to

[gleaming pendant]

It's 32 bits unless you're doing embedded

[thin cairn]

It's 32 bits on a modern computer with a normal compiler

[gleaming pendant]

It's min 16

[thin cairn]

But there's nothing in the spec, so you can't trust it

[gleaming pendant]

You can trust it on modern computers

[thin cairn]

Use a fixed width type if you want full safety

[frigid sand]

minimum size is in the spec

[gleaming pendant]

If you're building a win32 exclusive app you can trust it to be 32 😛

[frigid sand]

should be at least

[keen orchid]

https://faultlore.com/blah/rust-layouts-and-abis/#the-c-integer-hierarchy

[limpid crypt]

https://www.youtube.com/watch?v=qqIcucKqtDo

[eager mirage]

25w07a neoform build failured #builds message

[glass wagon]

ecj exclusive compile error

[sterile abyss]

why is that an error

[glass wagon]

it can't find the livingentity1 variable for some reason

[frigid sand]

for some reason instanceof in a ternary seems weird

[glass wagon]

Seems like it's fixed in 2025-03

[frail patio]

I would recommend you and others wanting to work on these changes keep an eye on the FRAPI thread of the API channel in Fabricord as I am discussing how to port FRAPI to 1.21.5. Ideally everyone would agree on the design decisions so the APIs between loaders are not so different.

[frank sigil]

ok so to whom do we assign the client classes (basically a lot of rendering stuff) in Kits?

[lunar saffron]

That's inevitably going to be Tech, embeddedt and me 😅

[frank sigil]

why we have a monica

[lunar saffron]

Add her to the list, doesn't make a huge difference though

[oblique tusk]

smh, downplaying the importance of a Monica /j

[gleaming pendant]

I did a few already, can probably finish the job next week if nobody gets to it

There's still a lot of other rejects, no?

(not even talking about yesterday's snapshot yet, we should finish the previous one first)

Maty should also fix the testframework 😄

[frank sigil]

yea the remaining stuff can be categorised as

• item abilities
• entities
• saved data
• tickets
• networking
• client

[haughty spade]

how much do the new changes invalidate the changes from 1.21.4?

like, is there overlap or is porting to 1.21.4 still a time saver toward 1.21.5?

[frank sigil]

idk I'm not on 1.21.4 with any of my mods

[weak wasp]

I would say there is

Most of the changes are on different parts of the system

[marsh crescent]

oh crap, what changed with SavedData?

[fleet dome]

codecs

[stiff flicker]

that was in the 06 snapshot, it was not too bad, it made it quite a bit cleaner imo

the Factory is replaced with Type that takes a codec

[stiff imp]

wasn't that already in .4?

could've sworn I remembered doing that already..

[marsh crescent]

got it. thanks. i'm already using codecs so that'll just be a class replacement. nice

[stiff flicker]

you can use codecs, but it was not forced

now it is forced

[haughty spade]

wtf discord picking random emotes

it's going to be a major PITA to convert my existing uses of SavedData to codec

[stiff flicker]

Start doing it now to save time.

[stiff imp]

you'll be done just in time for them to change it again in .6

lol

[haughty spade]

nah I still have mods to port to 1.21.4's client item stuff. "things I will need in the future" is not high priority :P

[marsh crescent]

just start making wrappers for everything so you end up with StreamCodecWrappingCodecWrappingNBTSerializerWrapper

[haughty spade]

I also have to figure out what causes this, sometime (1.21.4)

[stiff flicker]

I am trying to keep up with snapshots this time around. 07 is pretty rough, we use a good amount of custom RenderTypes. They made a ton of stuff protected and package private. So making new RenderTypes is going to be messy.

[haughty spade]

I use custom render types in this mod too

just one though.

[stiff flicker]

making a new RenderType.CompositeState is...

one sec

all protected

[haughty spade]

I wonder if the shader is why the book isn't working 🤔

nope

anyhow my 1.21.4 issues don't belong in this channel

[stiff flicker]

usually means it cannot find a texture if it is purple and black

[haughty spade]

yeah

it's 99.98% definitely the way I adapted the model stuff

my AnimatedBookBackground stuff is 4 baked models that I interpolate between

problem is, there are no warnings about texture resolution issues

oh I think I know what it is wasn't it

[frail patio]

QuadCollection and List<BakedQuad> being used to represent static geometry is really inconvenient because Fabric's API has its own representation of static geometry which is necessary for extended features, but I don't know how this could be solved. It wasn't an issue before because BakedModel always acted as a supplier of the static geometry, so the API could just call a custom method, but now the static geometry is being passed around directly in areas where mods might want to customize it with extended features.

Also, I can suggest two simple ways in which BlockStateModels can be made more efficient, especially multipart models

[tardy leaf]

Well, I have some stuff pending for multipart (like sharing actual data), but it might be useful to see if I missed something
Though I'm not touching that concatenation of quad lists that happens inside multipart, that's going to be solved later

But what are those extended features? Is it feasible to just custom implementations to block state and item models that get their geometry from something else, instead of piggybacking that simple format in model/?

[compact hare]

Seeing Mojangstas around here is always a nice surprise, I gotta say

[tardy leaf]

I mean, it is static geometry, because it's a class meant to store static geometry

And from design point of view, it can't do much, because it's called in two non-overlapping contexts (items and blocks)

[lunar saffron]

Shouldn't be too bad in practice. The type consists of two Functions taking a context object, one to create an empty SavedData and one to create a codec. The context provides the level and the world seed. Worst case you can just xmap the CompoundTag codec into the existing load/save methods with minor adjustments

[frail patio]

1. Make getQuads accept a Consumer<BakedQuad> and return void so no list concatenation is necessary.
2. Pass an EnumSet<Direction> instead of Nullable Direction to getQuads and only call it once instead of up to 7 times. Also requires passing the random seed so the random can be reseeded per cullface to maintain how models look now.

[frail patio]

The extended features effectively add 4 additional quad properties, allow the cull face to be specified per-quad so using separate lists isn't necessary, and allow the API impl to control how everything is actually implemented, which allows for arbitrary extended features beyond this

[lunar saffron]

EnumSet wouldn't work because it doesn't allow null

[frail patio]

Null cullface quads are always rendered, so the set doesn't need to keep track of the null face

[tardy leaf]

arbitrary extended features
yeah, but no, not really a fan of stuff happening down the rendering stack

[tardy leaf]

kind of what I expected, but actual solutions don't mesh well with plans

[frail patio]

It's possible but is extremely inconvenient and I don't think it's a practical solution

In Fabric by default those features aren't even available through json, though they are in Neo

[tardy leaf]

It seems like good chunk is about quads itself, what if that was an interface?
if you don't want to add stuff to existing record?

[lunar saffron]

There are cases where you explicitly don't want uncullable quads. But that discussion is a bit off-topic for here

[frail patio]

An interface solves some problems but if Fabric actually has an alternate implementation of the interface then performance will probably become much worse because the JVM can optimize virtual calls which only have a single implementation

[tardy leaf]

you can still produce whatever quads you want, since there is still quad producer as a field on unbaked model

*two
** it's complex

IIRC

let me re-read https://shipilev.net/blog/2015/black-magic-method-dispatch/

[frail patio]

I think that only applies to modded scenarios

Of course vanilla already supporting the use case would be nice

While we're on the topic of models I want to bring up that VertexConsumer#putBulkData reads color differently depending on the endianness of the system

[tardy leaf]

also, it's less of virtual call and more of a instanceof, since no matter how you override you can't get extra data out if it
and instanceof costs just a single branch in most cases

[frail patio]

Ah yeah. In my mind I was considering methods like getPosX(int vertexIndex)

Still, Fabric's representation encodes an entire List<BakedQuad> and there are no separate objects for each quad (the default impl stores all quads in a big int array) so switching on the list or QuadCollection would be a lot better

[fallow merlin]

They removed alot of Shaders?

[weak wasp]

More like reworked

Too many things changed so the primer will be up Sunday

[proven compass]

it said they removed the files

does that mean the shaders are in code now perhaps

[stiff flicker]

The jsons are now code, yes

Not sure why they changed that.

[fallow merlin]

Cool

Hrm

[fallow merlin]

Because they dont want people to do it themselves via resource packs as its not offically supported

[proven compass]

thats my thought

[fallow merlin]

Its for internal use only, for vanilla

For modded, we could readd it back

[proven compass]

tho why were they json in the first place

if they werent meant to be changeable

[fallow merlin]

Perhaps its was easier

or they were going to make it offical

But then backstepped on it

boq could answer that maybe xD

[proven compass]

the dev notes have said for a while that they eventually plan to improve the system

[fallow merlin]

Yea

For the sheep color thing

Should it be an event that gets fired everytime the game wants to get a sheep color

DyeColor SheepColorrSpawnRule#getSheepColor(Holder<Biome>, RandomSource)

[raven blade]

think it would be best to wait and see what they do before release

I personally don't like the idea of it being an event though. at a glance, a data map would be suitable

[proven compass]

im still hoping they revert this

and give us real sheep variants

something like this

[fallow merlin]

Is there an example of a data map in NF?

[proven compass]

many

its a neo invented thing

[fallow merlin]

Link?

[proven compass]

look at NeoForgeDataMaps

[proven compass]

https://docs.neoforged.net/docs/resources/server/datamaps/

[fallow merlin]

Neat

[limpid crypt]

https://youtu.be/fLYwRNBHXTM

[marsh mural]

you beat me to it
[added video link to the handy pin]

[stiff flicker]

So, with 25w07a. They got rid of RenderSystem.recordRenderCall which is annoying.
When our texture cache expires, the get released off the render thread.

[thin cairn]

You should never do anything like that off-thread. The old system was broken and shouldn't have been used. Use Minecraft#execute

[unkempt lynx]

^

when you do recordRenderCall, you're gambling away your state

there's no guarantee the game will be in a good state when it's run

You shouldn't use Minecraft.execute either, that's moving the problem

you need to make your own place and time to upload with guaranteed state

(Notice that in 25w07a Mojang never uses this anymore, they always have guarantees for when stuff is done)

[stiff flicker]

It's not an upload issue. It's releasing the textures from cache when they expire. I'll probably have to change how caching is done.

[gleaming pendant]

reminder for future me: this is how files that were deleted in a MC update can be removed locally: git clean -fxn projects/neoforge/src and if it seems right, run again without -n

still a lot of rejects heh

44 left

wtf are these patches

[sterile abyss]

fluids

[limber ginkgo]

they want/wanted to implement alternatives for what people use core shaders for

[gleaming pendant]

sure but why do they exist

especially the first one

they're such a PITA to update

[sterile abyss]

because fluids

[limber ginkgo]

core shaders are still unsupported, but they just want to provide alternatives so people don't have to resort to core shaders

[sterile abyss]

fluid patches have always been a painful black box

[fleet dome]

That shit really needs comments why the change was made when it was made

especially in patches that are hard to get proper history for

[river mural]

Iirc it was a little cleaner on the first pass before lex had ash rewrite things… granted I was the only triager willing to spend the time reviewing the PR (and the predecessor)

But yes, ideally eventually mc will add a new fluid and decide to clean up fluids to be less of a hardcoded mess of if statements vs method calls

[gleaming pendant]

we should delete fluid patches and see if anyone complains

[stiff imp]

or they'll make it worse and somehow cram enums in there

[stiff imp]

https://tenor.com/view/skeptical-futurama-fry-hmmm-i-got-my-eyes-on-you-gif-17101711

I'm already complaining cus they still have holes lol

like modded fluids being unaccounted for properly in isEyeInFluid or updateFluidHeightAndDoFluidPushing

[gleaming pendant]

look at this lol

[stiff imp]

who needs OOP

[gleaming pendant]

we need more manpower for this update, @ maintainers please help with the simple rejects

[floral mesa]

I can do some in a bit, let me know what to not touch to avoid colliding with you

[sterile abyss]

i can look at tests tomorrow morning

[gleaming pendant]

I'm not touching anything else tonight 😄

[weak wasp]

Seems like part of the geometry stuff can be gutted due to vanilla's UnbakedGeometry

[frail patio]

The geometry stuff was already removed in 1.21.4

Some of the new system is not sufficient for modders' use cases and needs to be appropriately extended

For example, model loaders are almost useless without extending the system

[weak wasp]

Ah right, there was model changes in 21.3 too because of the item stuff

Honestly, I feel like this snapshot primer has made me forget most things about the current version

[stiff imp]

I mean that's fine

the amount of changes and reverts that are happening, current knowledge is quickly becoming irrelevant lol

[weak wasp]

Finally done, take your primer: https://github.com/neoforged/.github/pull/14/commits/ccff5ff6a51c147c933d995c01653670924a7773

@marsh mural

As everyone is probably already aware, main changes are the render pipeline rework and model separation, along with some changes in block effects when inside a block.

[stiff flicker]

thanks for this

[sterile abyss]

ah yes...

game tests are...

lovely...

yeah well bad and good news

the vanilla system uses a dynamic registry which is really hard to bypass

so I think mojang just forced our hand to add registry callbacks for dp regs

because I think we can all agree that needing a json file for each test is unacceptable and just not how tests work

[dark wasp]

but you can datagen gametests now

[sterile abyss]

sure? but now instead of an annotation like how testing works in the entire java ecosystem you need to first run the game after each test metadata modification / new test to then proceed to test... which is completely backwards

and it would also require a lot more setup for mods when the point of gametests was that it would also be easy to set up and write a simple test for

[fleet dome]

No data-gen for tests please

[sterile abyss]

now, what is the best way to do this

should fire a modify registries event for dyn regs but how

[dark wasp]

https://preview.redd.it/i-just-now-realized-that-the-mask-would-be-perfect-for-v0-3cns7311h3kd1.jpg?width=600&format=pjpg&auto=webp&s=ca6eb0ed75717c5e775b19ca5569fd3d34d11420

[sterile abyss]

my thinking was to modify ModifyRegistriesEvent to have a Registry<Registry> as a parameter and fire it with BuiltInRegistries.REGISTRY on startup and with the registry of dyn regs when they're constructed

but then we'll have to change the api of ModifyRegistriesEvent to be modifyRegistry(key, registry -> {}) style

I'll hardcode a patch for now and we'll settle on some api later

[fleet dome]

Hm, do you mean the RegisterEvent or whatever it's called or really ModifyRegistries?

[sterile abyss]

ModifyRegistriesEvent

the one you use to add callbacks normally

https://github.com/neoforged/Kits/commit/8b69e575d45f0f39023d86016ae75260323a622c I think this works for now but I can't really test it until it compiles

we'll also need some entrypoint changes probably

[merry falcon]

Does fabric have callbacks already for dyn reg?

[sterile abyss]

no

fabric did what I did

the "we'll add a hacky special case"

[dark wasp]

mojang: finally rejects the annotation-based-test cargo cult
also mojang: replaces it with something even worse

[sterile abyss]

it's not a cult to be rejected, it's what annotations are made for

[dark wasp]

ah well, I'll figure out how to use mojang's new system anyway

[sterile abyss]

I refuse to believe that internally they use the data driven system anyway

you can do whatever you want, this is a free world

[dark wasp]

I use vanilla game events for mod apis too

[sterile abyss]

oh ugh

the chunk loading patches are rejected

@lost imp so uh, mojang split item sub predicates from testing on the item to testing on the component

the gameplay enchantments event is broken again

[gleaming pendant]

Looks good enough

I wouldn't rush a dynreg modification event

[sterile abyss]

I'm about to commit crimes with these cursed gameplay component "modification" events

I can fortunately override matches(DataComponentGetter) in super

could have been worse

I love how mojang "fixed" this

they just removed the index from the method

I'm deleting the reject for now, if someone needs us to patch an index overload in we'll see

so

what do we do with tool actions

since mojang no longer has special pickaxe and sword classes (and is very close to not having a shield one either) and instead has components for that

[haughty spade]

does a custom data component in an itemstack break vanilla connections?

[sterile abyss]

yes

[haughty spade]

then the idea of a HasAbilities component wouldn't work

one option would be to make the "canPerformAction" (whatever default the method is called now) in itemstack check for data components to resolve pickace_dig and such

[gleaming pendant]

Uh what

[gleaming pendant]

Yesn't. We could probably hack something using the ConnectionType in RFBB

[sterile abyss]

uh what what

[gleaming pendant]

Do the components allow for granular control or not? I.e. are all of the sword tool actions separate components?

[gleaming pendant]

Idk about this:D

[sterile abyss]

the alternative is removing the extension methods

removing features is not on my port to do list 😛

[gleaming pendant]

Removing duplicate features is fine though

[sterile abyss]

well sword isn't really that much custom behaviour

[gleaming pendant]

Well, the same applies for other tools

[sterile abyss]

but no there's only one component

well, Weapon and Tool

[gleaming pendant]

Maybe we just remove some tool actions in favor of the corresponding vanilla components

[sterile abyss]

duplicate?

[gleaming pendant]

Well, attribute modifiers are a component right? But it's annoying and won't allow for an event sadly

[sterile abyss]

well yes, it's not a duplicate feature since the event is the feature :P

(an annoying event to patch vanilla to use, mind you, but a feature we've been dragging for a while regardless)

[frank sigil]

virtual datapack

[trail fable]

removing tool actions in favor of components? yes please

[frank sigil]

instead of hacking into the system for dp registries just use in memory virtual datapacks to register the tests

[frank sigil]

we need some way of keeping track what component does what action but yea

[sterile abyss]

uhh... no, some of these need objects passed around that are not serialisable

like the test function

which yes before you mention it exists as a registry. but a static one

and having people also register to that is even more pointless complexity

[frank sigil]

well we can register based on annotations to that registry and handle the dp registry part through the virtual datapack

[sterile abyss]

you're overthinking it

[frank sigil]

virtual datapacks is how I did the vanilla potion brewing recipes in my porion recipes PR

[sterile abyss]

this is dev only, it doesn't need to be "proper", and you also want tests to be reasonably fast

[fleet dome]

No

[sterile abyss]

virtual datapacks are overhead

[fleet dome]

Christ, we need the different style of registration anyway

for parallelizing registrations later

[frank sigil]

I'm not but adding events to dp registries opens the door for people to not datagen stuff and instead just staticly register it to the dp registry

[sterile abyss]

there's no general event right now, as I said I hardcoded it

(and datagen doesn't solve all use cases for the event, people generate stuff at runtime - and registering directly would be usually more efficient and better for those mods instead of the de/ser cycle, but that's besides the point)

[trail fable]

add a todo

[sterile abyss]

todo for what, we aren't sure if we even want a general event right now

[frank sigil]

we do not want an event there, as I said people should be using dp registries the way they are meant to be not like static ones

[sterile abyss]

it has been a point of discussion on and off for a while

[frank sigil]

down to 13 reject files

[river mural]

Do they define block type changes for right click now? Or how does that work

[weak wasp]

That’s still a subclass for item

I wouldn’t be surprised if right click behavior was migrated at some point though

[floral mesa]

^ this

we would benefit from some virtual datapack system that doesn't require actually serializing

[fleet dome]

how is that any different from a reg-callback

[frank sigil]

I see where you are coming from but that would make it too easy for people who don't actually need it to just skip datagen and register non generated content statically to the dyn registry

[floral mesa]

which is fine as long as it can still be overriden through standard means like KubeJS or another datapack

[gleaming pendant]

damn I had one commit left locally

oh well, let's see why I have conflicts

@glass wagon are you sure about this change?

[gleaming pendant]

but why

oh

actually no it still doesn't make a lot of sense to me

[merry falcon]

Wait a minute

That setup means I can’t use the event to make it have a null player

Right?

Like disable the following

[fleet dome]

yeah that is also my thought

[gleaming pendant]

you can if you call event.setFollowingPlayer(null)

[merry falcon]

But getFollowingPlayer will see that null field and then look up the player itself

[gleaming pendant]

vanilla already had this code somewhere else

[gleaming pendant]

the field contains a non-null empty optional

[merry falcon]

O

[gleaming pendant]

anyway coeh I'll undo that change since I don't like it

[river mural]

so then for now we don't actually have to change it in terms of item abilities

[gleaming pendant]

the fluid patches are REALLY ANNOYING to update

I don't even want to know how many vanilla parity issues we have

TG please save us

[fleet dome]

Did you run the vanilla tests? 😄

Well, once it compiles, we should

[river mural]

https://tenor.com/view/window-ignored-bye-roll-up-car-window-gif-5065354

[sterile abyss]

we don't have them

[fleet dome]

Wait I thought they included them?

Or did they just add a way to load user-tests?

[sterile abyss]

just that

[fleet dome]

Pfffffffffff 😄

[gleaming pendant]

-        if ((!this.isInWater() || this.isUnderWater()) && flag3 && this.minecraft.options.keySprint.isDown()) {
+        if (!this.isSprinting() && (!(this.isInWater() || this.isInFluidType((fluidType, height) -> this.canSwimInFluidType(fluidType))) || (this.isUnderWater() || this.canStartSwimming())) && this.hasEnoughImpulseToStartSprinting() && flag3 && !this.isUsingItem() && !this.hasEffect(MobEffects.BLINDNESS) && this.minecraft.options.keySprint.isDown()) {

@merry falcon when are you cleaning up this nightmare?

fuck it if these fluid patches are the last ones I'll just move them to a different folder and let other people have "fun" with them

[merry falcon]

I don’t think I ever will be able to. Keep losing motivation and haven’t touched fluid branch since the one day I started lol

[gleaming pendant]

@sterile abyss the only shield tool action is SHIELD_BLOCK which looks superseded by the BlocksAttacks component

so I suggest that it gets removed

[merry falcon]

Does swords have blocking behavior and do they use BlocksAttacks

[proven compass]

swords havent blocked since 1.8

they can now if u put the component on them

as can any item

[gleaming pendant]

the component seems exclusive to shields

(by default)

[proven compass]

correct

i know it'd be just until we get json items but i wish we had item modifiers

a bandaid solution to the ModifyDefaultComponentsEvent for vanilla

that way a datapack would be all u need to add sword blocking back

[gleaming pendant]

what's the point of the PICKAXE_DIG tool action?

[floral mesa]

@gleaming pendant what's the issue with the fluid patches, is it not clear where they should be applied to now?

[river mural]

Maybe completeness sake. Some of the dig ones we replace loot tables for

Namely shears

But I think it predates that

[gleaming pendant]

that and the fact that they're probably half broken anyway

[river mural]

Maybe @haughty spade who initially made the pr remembers

If it was even a pr (might have been a direct commit)

[lost imp]

I don't know if it has a functional purpose, it's useful to tell if an item is a pickaxe though

[river mural]

yeah, I think the idea might have been for ability based ingredients in recipes

[gleaming pendant]

pickaxes don't have any class anymore

public static final Item STONE_PICKAXE = registerItem("stone_pickaxe", new Item.Properties().pickaxe(ToolMaterial.STONE, 1.0F, -2.8F));

[lost imp]

Well we could just yeet them and reimplement fluids from scratch like we keep saying we need to

[gleaming pendant]

"someone" has to do it though

[proven compass]

because the community loves refactors like that

[gleaming pendant]

I would go for a much more limited set of initial patches

[lost imp]

The pickaxe dig action might be replaceable with the pickaxes item tag? Not sure. The tool action is useful for dynamic scope, unlike tags

[gleaming pendant]

even targeted patches if possible rather than trying to completely overhaul the system

[proven compass]

wasnt there a thing where u asked devs how they used something before refactoring it

u could just find out how the fluid API is being used now and cut off anything unnecessary

[gleaming pendant]

we have countless vanilla parity issues with fluids

[lost imp]

At this rate vanilla will have fully data driven items quite shortly here, and then fluids will be a "small" change comparatively

[gleaming pendant]

and the patches are difficult to update, probably partially outdated, and partially missing

[haughty spade]

nah it wasn't ingredients, it was interaction logic

stuff like "do I have a tool that can act on this block?"

and shit like that

I basically converted the old ToolType into ToolAction

1:1

and then added a few additional ones

[lost imp]

I use it to determine if an item is a pickaxe for apoth affix categorization

[proven compass]

sounds like something a tag should be used for

[haughty spade]

see: https://github.com/MinecraftForge/MinecraftForge/blob/1.16.x/src/main/java/net/minecraftforge/common/ToolType.java#L17-L20

no a tag wouldn't work

because of fancy multi-tools

[lost imp]

Tags aren't dynamic

This is one of the largest shortcomings of vanilla

They have absolutely no solutions for dynamic behavior

[gleaming pendant]

there's no reasonable way to categorize pickaxes anymore

[haughty spade]

tech mods like to have things like "press J to switch between shovel mode and pick mode" type features

[proven compass]

which would now be covered by overriding the components

[lost imp]

I mean we can trivially provide the default tool action override as a tag check

[haughty spade]

so the canPerformAction method was stack-sensitive by design, to account for it

[lost imp]

Mods with custom items can override it to do appropriate dynamic behavior

[proven compass]

since switching between those modes is switching the Digger component or wtv

[gleaming pendant]

that's terribly not data-driven though

[haughty spade]

we are in an awkward phase for this

[gleaming pendant]

maybe tool actions should just be a data component

[haughty spade]

where the systems are partially adapted

and we can't have a good API

[merry falcon]

this. @torpid sandal will have a thing or two to say about yeeting pickaxe toolaction

[haughty spade]

can't, vanilla connections

[merry falcon]

think of multitool items

[gleaming pendant]

we could in principle hide any server-side only component in the custom data component when talking to vanilla clients

[lost imp]

Just wait for vanilla to fix the creative menu

[gleaming pendant]

can't really since I am trying to fix pickaxes 😛

[merry falcon]

LOL

[lost imp]

Then SS-only components can actually exist

Well if they don't you cannot strip them in payloads

[merry falcon]

we are stuck with creative menu code forever

[lost imp]

They will be voided

[gleaming pendant]

remember what I said about custom data?

that's how item data attachments were synced in 1.20.4

[lost imp]

Did you have a way to un-wind the custom data back when received from vanilla creative clients?

[gleaming pendant]

yes

there was actually no special handling for vanilla clients

attachments would always get synced with the nbt tag

if the client was neoforge it would unpack the attachment nbt, otherwise it would just not care about it and send it back as is eventually in creative mode

this sort of patch is

[torpid sandal]

technically, I don't care about the tool action

however, I do care that other mods want to ask that question

and they will do it wrongly if we don't have the tool action

honestly, my concerns which lead to wanting tool actions are almost all solved by just adding a neo level brokeness check

if I can add a way that other mods can check "is this item broken" and thus not do their stuff, things are easy and I can attach behaviors to the item instance

now, Tetra doesn't have that luxury unless it changed recently as all their tools are a single item instance

since I don't see it at a quick glance, can someone summarize for me the problem with tool actions on 1.21.5?

[gleaming pendant]

public static final Item STONE_PICKAXE = registerItem("stone_pickaxe", new Item.Properties().pickaxe(ToolMaterial.STONE, 1.0F, -2.8F));

[merry falcon]

Without the pickaxe class, you arent able to attach the ItemAbility. Is that right?

[gleaming pendant]

indeed

because item abilities are not data driven at all

maybe I should look into server-side data components

[lost imp]

Just make the default tool action impl check the tag

[uneven mesa]

If its only purpose is "is this a pickaxe", shouldn't it just be a tag?

I could be missing something

[lost imp]

Mods have "sometimes a pickaxe" items

[uneven mesa]

Ah, sure

[lost imp]

So defaulting to a tag is fine, but the action has to be a stack-sensitive layer

[uneven mesa]

In that case this really does seem like something you'd want to model with data components

But of course, the whole client-server issue

How plausible are server-only data components?

[torpid sandal]

so, make them data driven?

why can't item abiltiies be a component with NBT assigning them like all the new vanilla components?

[uneven mesa]

...Isn't that what I said?

[merry falcon]

Breaks vanilla connection

[uneven mesa]

The issue is the server-client sync

[torpid sandal]

you said to make it a tag

[merry falcon]

vanilla checks data components

[uneven mesa]

Scroll down...

[torpid sandal]

so datacomponents need to sync to both sides

[uneven mesa]

But yeah, server-sided data components looks like it may be necessary in general. Especially as it looks like mojang is heavily moving things away from being implemented in a item-specific class -- forge's APIs are going to need to adapt to that

[torpid sandal]

is there anything we can piggyback off?

e.g. what is .pickaxe() in item.properties?

[gleaming pendant]

quite plausible once NeoForge compiles again

[uneven mesa]

Fundamentally not for the general case, and as vanilla increasingly moves this way we need a general solution to the problem of "items provide behavior that isn't a vanilla type of behavior"

[uneven mesa]

Good to know.

[torpid sandal]

practically, its no issue if a mod does it

its just an issue if neo wants to add it to a vanilla item

[gleaming pendant]

the biggest issue is the creative menu and my proposed trick is to just shove unknown data components into the custom data component for vanilla connections

[proven compass]

if someone is giving their item a tool ability wouldn't it be required on both sides anyways

[uneven mesa]

A bit confused what you're saying here?

[torpid sandal]

mods can add new data components, can they not?

[uneven mesa]

Yeah I like that. Gotta do something to back it out of that when recieving from such a connection of course

[gleaming pendant]

of course yes

[uneven mesa]

Oh yeah sure. But we're talking about neo-added APIs, which may be added to modded or vanilla items

[torpid sandal]

thats all I'm saying. We only care about server-client sync because its a neo thing

we do have the worst case approach: tell mods "check if its vanilla or our component"

I've done that sort of thing a ton back when tool actions were new

[uneven mesa]

From what tech said there's no reason to have such an awful setup; that idea of shoving it in the custom data component should work I would think

[torpid sandal]

creative menu is always causing trouble

can't we just patch creative menu in Neo so client doesn't nuke server data?

[uneven mesa]

No?

Vanilla client, remember

It doesn't nuke server data as such. Or rather, that's not the issue here

The issue here is that neo can't just not send those neo-added components to a vanilla client -- it has to send them, because otherwise it won't get them back in the creative menu

I suspect a lot of item extensions that neo adds will eventually end up in this format... after all, vanilla is moving all of its own stuff to ItemProperties and components and whatnot.

If I had to wager, I bet they'll eventually be making Item final

[torpid sandal]

which will be a rough day for all of us when it happens

as another thought, some of these problems are caused by Neo needing to work when only on the client or server. I do wonder whether in the future it might be worth having some sort of build flag/config option to turn on server only or client only mode

it does seem like a lot of useful features get rejected over the years due to what I expect is the minority of the playerbase

hopefully not needed right now though

[frank sigil]

it's not fine, because

1. it allows to achieve states that are not possible through dp registry loading
2. it removes the json examples to reference when writing the override stuff

all in all it makes it too easy to do stupid stuff

[sterile abyss]

a virtual datapack doesn't have a json example either

[frank sigil]

a virtual datapack is not as easy to do as just using an event

it is not about making it impossible it just shouldn't be easier than datagen

which an event would be

especially with how dp registry datagen works

ok 3 things left in the patches

• fluids
• tickets
• abilities

for abilities we can't move to pure components yet since neo can't add components itself to keep vanilla client compat