FML Clean-up | The NeoForged Project | Page 7

timber ingot Jul 10, 2024, 1:19 PM

#

I don't think thats really conveyed in the PR. What alternatives have been considered?

haughty copper Jul 10, 2024, 1:22 PM

#

???

#

Alternatives to what

#

And detached from what?

timber ingot Jul 10, 2024, 1:23 PM

#

haughty copper Alternatives to what

adding a whole load of NeoForge version specific shit into a minecraft detached project

haughty copper Jul 10, 2024, 1:23 PM

#

Dude, MOD_CLASSES is in FML

#

😄

#

Or what do you mean by version specifics

#

Not to mention all launch targets use minecraft class names

timber ingot Jul 10, 2024, 1:24 PM

#

Im aware that MOD_CLASSES is in FML, its core to the mod loading infrastructure and has absolutely fucking nothing to do with Minecraft

haughty copper Jul 10, 2024, 1:24 PM

#

That statement is far removed from reality

#

In at least two ways:

timber ingot Jul 10, 2024, 1:24 PM

#

is it?

haughty copper Jul 10, 2024, 1:25 PM

#

The launch targets start Minecraft classes BY NAME

#

FML makes assumptions about which libraries are present at launch based on the set of platform libraries provided by Minecraft

#

Oh I found a third one 😄

#

The locators locate minecraft jars by name

timber ingot Jul 10, 2024, 1:26 PM

#

Counter points:

FML contains some defaults that should be moved to NeoForge.
FML requires some libraries to run, which happen to be transitively pulled from minecraft. It also requires libraries that aren't included with minecraft

haughty copper Jul 10, 2024, 1:26 PM

#

In what world is that not minecraft specific

#

FML is currently downgrading libraries provided by MC

glacial crag Jul 10, 2024, 1:26 PM

#

FML is definitely MC specific

chrome mortar Jul 10, 2024, 1:26 PM

#

haughty copper Oh I found a third one 😄

make it 4

#

https://github.com/neoforged/FancyModLoader/blob/af00d44e71b410d6d4a6159265b97dfa26081c2d/loader/src/main/java/net/neoforged/fml/loading/moddiscovery/locators/MinecraftModInfo.java#L20

haughty copper Jul 10, 2024, 1:26 PM

#

Seriously covers how do you arrive at that statement

#

given it was previously in NF itself

#

or rather, Forge

glacial crag Jul 10, 2024, 1:27 PM

#

Did you really say that FML is detached from Minecraft -_-

chrome mortar Jul 10, 2024, 1:27 PM

#

the early window literally says Bootstrapping Minecraft

haughty copper Jul 10, 2024, 1:27 PM

#

I am just wowed

timber ingot Jul 10, 2024, 1:27 PM

#

I was under the impression it used to be, maybe im thinking of ModLauncher

haughty copper Jul 10, 2024, 1:27 PM

#

You are

#

ModLauncher is far more removed from it

timber ingot Jul 10, 2024, 1:27 PM

#

alright, dont be a cunt about it next time

haughty copper Jul 10, 2024, 1:27 PM

#

It's fine sorry, I was just baffled 😄

#

seriously

timber ingot Jul 10, 2024, 1:28 PM

#

I still am not comfortable with NeoForge specific tools being inside FML

haughty copper Jul 10, 2024, 1:28 PM

#

It's a NF project

chrome mortar Jul 10, 2024, 1:28 PM

#

https://github.com/neoforged/FancyModLoader/blob/af00d44e71b410d6d4a6159265b97dfa26081c2d/loader/src/main/java/net/neoforged/fml/ModLoadingContext.java#L24 this is my favourite

haughty copper Jul 10, 2024, 1:28 PM

#

It references the neoforge module by name

#

and crashes

#

if it doesn't find a neoforge module

chrome mortar Jul 10, 2024, 1:28 PM

#

"Where is minecraft???!"

haughty copper Jul 10, 2024, 1:28 PM

#

that is the status quo before the PR

timber ingot Jul 10, 2024, 1:28 PM

#

matty but out please

haughty copper Jul 10, 2024, 1:29 PM

#

check the earlydisplay game layer handover

chrome mortar Jul 10, 2024, 1:29 PM

#

timber ingot matty but out please

uhm no, i'm documenting it for everyone else that is basing decisions on the wrong assumption in the future

#

you can use nicer words btw

timber ingot Jul 10, 2024, 1:29 PM

#

i can and choose to fucking not

chrome mortar Jul 10, 2024, 1:29 PM

#

what was your wording again

#

dont be a cunt about it next time

haughty copper Jul 10, 2024, 1:30 PM

#

https://github.com/neoforged/FancyModLoader/blob/main/earlydisplay/src/main/java/net/neoforged/fml/earlydisplay/DisplayWindow.java#L602

chrome mortar Jul 10, 2024, 1:30 PM

#

cut the crap

glacial crag Jul 10, 2024, 1:30 PM

#

He's Australian...

timber ingot Jul 10, 2024, 1:30 PM

#

fuck off matty

#

seriously

#

im done with your shit

glacial crag Jul 10, 2024, 1:30 PM

#

cunt is the first word newborns learn there

timber ingot Jul 10, 2024, 1:30 PM

#

i was confused about something

haughty copper Jul 10, 2024, 1:30 PM

#

var fm = layer.findModule("neoforge").orElseThrow();

timber ingot Jul 10, 2024, 1:30 PM

#

then get fucking dog piled

haughty copper Jul 10, 2024, 1:30 PM

#

I am just referencing this:

I still am not comfortable with NeoForge specific tools being inside FML
😄

glacial crag Jul 10, 2024, 1:30 PM

#

Okay covers now you're overdoing it I think thinkies

errant yew Jul 10, 2024, 1:31 PM

#

I think we should face the reality of the situation, that FML is by-and-large made for, tailored to, and nearly unequivocally runs with NeoForge
I'd think it's a logical step to at least add easier testing infrastructure to test changes to FML in a NeoForge dev env

timber ingot Jul 10, 2024, 1:31 PM

#

haughty copper I am just referencing this: > I still am not comfortable with NeoForge specific ...

Right, lets get to this statement

#

What are your requirements for this system in the first place?

haughty copper Jul 10, 2024, 1:31 PM

#

It's hard tied to NeoForge and I don't see why we should put up with the ball ache trying to pretend it's not

timber ingot Jul 10, 2024, 1:31 PM

#

put me on the same page, at the start

haughty copper Jul 10, 2024, 1:31 PM

#

The PR you mean?

timber ingot Jul 10, 2024, 1:31 PM

#

yes

haughty copper Jul 10, 2024, 1:31 PM

#

I am trying to get rid of the requirement to manage three different sets of classpaths in dev

#

which tend to result in various errors for our users

#

from the infamous "two module xyz is read by module abc" or whatever it is

#

to the various issues with people trying to shove dependencies into their dev runs and failing

#

due to the stupid "legacyClassPath" crap

#

There's an additional sideeffect we are currently just chosing to ignore

#

Which has nothing to do with the classpaths themselves, but with us not considering to allow --add-opens for base Java modules that target modules we create after boot

#

namely: Netty in Vanilla uses JDK unsafe for some optimizations, but cannot (and never could) in modular forge, since we load it into a module layer after boot, which would require us to dynamically at runtime open java.lang to it

#

that is use case #2 for instrumentation, but can be solved with other hacks too

#

instrumentation is just currently the only real supported way of doing that

timber ingot Jul 10, 2024, 1:34 PM

#

Okay, so what parts of this are implemented or worked towards by the PR?

haughty copper Jul 10, 2024, 1:35 PM

#

The PR already launches ATM10 without the need to pass a module path or LCP

timber ingot Jul 10, 2024, 1:35 PM

#

How oes the PR help solve those problems

#

we are, and have always been, talking about this PR, you are confused: https://github.com/neoforged/FancyModLoader/pull/177

haughty copper Jul 10, 2024, 1:35 PM

#

Fuck me I am

timber ingot Jul 10, 2024, 1:35 PM

#

:D

haughty copper Jul 10, 2024, 1:36 PM

#

Christ that is just a gradle test tool 😅

timber ingot Jul 10, 2024, 1:36 PM

#

Right, so what does it do and how does it help?

haughty copper Jul 10, 2024, 1:36 PM

#

I was writing an essay about the startup-experiments branch 😛

#

Oh that's easy

#

If you change easydisplay

#

Or any of the production locators

#

Your only way to test those

#

is to crush your own balls

timber ingot Jul 10, 2024, 1:37 PM

#

alright, expand> "crush your own balls"

haughty copper Jul 10, 2024, 1:37 PM

#

This PR introduces a way to press "play" and get it tested

#

So, honestly? I think everyone has unique approaches to this

#

For Production, I think I'd publish a FML PR

#

Then I'd go and make a NF PR

#

that makes use of the FML PR via PR publishing

#

Then I'd wait for NF to publish

#

download the installer

#

run the installer

#

start the vanilla launcher

#

and try to run it

timber ingot Jul 10, 2024, 1:38 PM

#

Okay, so can you augment that flow with publishing to maven local?

haughty copper Jul 10, 2024, 1:38 PM

#

Can you though?

timber ingot Jul 10, 2024, 1:38 PM

#

i was asking

haughty copper Jul 10, 2024, 1:38 PM

#

Can the Vanilla launcher via its JSON actually pull from there

#

I don't think it can, but may be mistaken

timber ingot Jul 10, 2024, 1:39 PM

#

it doesn't need to, it will never pull if the file is already in the libraries folder

#

that has always been the case

haughty copper Jul 10, 2024, 1:39 PM

#

The FML PR there allows you to press "installClient" and then "runClient"

#

and get a debugger attached right then and there

#

I hope we're on the same page that FML is intended to be the modloader for NeoForge and will ultimately be used in a Minecraft environment

glacial crag Jul 10, 2024, 1:40 PM

#

I suppose that hot reloading and the likes also "just work"

haughty copper Jul 10, 2024, 1:40 PM

#

So getting a test run set up to test FML in it's production environment

#

seems like a natural thing to do?

#

As stated above, you must pull in NeoForge for that

#

since FML doesn't start at all without it

timber ingot Jul 10, 2024, 1:41 PM

#

So, what about includeBuild in gradle, you can just include the FML project into NeoForge and hit the run button, develop both in the same IDE

haughty copper Jul 10, 2024, 1:41 PM

#

No

timber ingot Jul 10, 2024, 1:41 PM

#

No?

haughty copper Jul 10, 2024, 1:41 PM

#

You do not get a production env like that

#

The production locators will not run

#

and remain untested

#

you do not even have the ability to run a production env from NeoForge dev at the moment

timber ingot Jul 10, 2024, 1:42 PM

#

Whats the difference with the production locators? what do they do differently?

haughty copper Jul 10, 2024, 1:42 PM

#

Buckle up, it's gonna be a fun ride 😄

timber ingot Jul 10, 2024, 1:42 PM

#

fuck

haughty copper Jul 10, 2024, 1:42 PM

#

Take a guess why FML requires you to pass --fml.neoForgeVersion, --fml.neoFormVersion and --fml.mcVersion on the command line, or otherwise refuses to run...

#

When you are in prod, it also requires you to pass -DlibrariesDirectory pointing to the MC Launcher libraries directory

timber ingot Jul 10, 2024, 1:43 PM

#

mother fucker

#

yep

haughty copper Jul 10, 2024, 1:43 PM

#

When it launches a production launch target

#

It will go pick up jars based on those version numbers from that folder and assemble SecureJar instances

#

And it's a bit... nuts

#

It uses a MC jar that does not have patched files from NeoFOrge

#

then layers another partial MC jar on top that has the NeoForge patches

#

then layers the resources on top

#

and that's your MC jar

#

That works completely differently in dev

#

(not that it has to, by the way, it just does)

#

The primary point of that FML PR is that it's currently a pain in the ass to test those parts of the code

timber ingot Jul 10, 2024, 1:46 PM

#

could we perhaps instead move the prod testing stuff NeoForge side? it seems quite useful to test NeoForge in prod

haughty copper Jul 10, 2024, 1:46 PM

#

Don't get me started on the earlydisplay madness

timber ingot Jul 10, 2024, 1:46 PM

#

then you could link via includeBuild FML

haughty copper Jul 10, 2024, 1:47 PM

#

I thought about it, but do you actually want that

timber ingot Jul 10, 2024, 1:47 PM

#

and achieve the same setup for both projects

haughty copper Jul 10, 2024, 1:47 PM

#

There are benefits to at least being able to do it that way

#

but I am not sure how well we can do the prod-install for neo dev

#

Since you need a working installer

#

and don't get one since the files are all local there

#

BTW it's a separate issue I also thought about

#

we'd kinda want the ability to test the installer even

#

in Neo

#

since currently we don't. and had broken ones. 😐

timber ingot Jul 10, 2024, 1:48 PM

#

Why can't the installer extract files from the jar? Its historically always been able to do that, making local installers was quite common

haughty copper Jul 10, 2024, 1:48 PM

#

Don't ask me I didn't write it 😅

#

I think it has the ability to do so, but for FML it expects it externally IIRC. I haven't really delved into the installer too much yet

glad cobalt Jul 10, 2024, 1:49 PM

#

timber ingot Why can't the installer extract files from the jar? Its historically always been...

It does

timber ingot Jul 10, 2024, 1:49 PM

#

the installer used to just have code to when evaluating for a dependency, either pull from remote, or pull from the /maven/ folder inside

glad cobalt Jul 10, 2024, 1:49 PM

#

It does and it will

haughty copper Jul 10, 2024, 1:49 PM

#

Yes but it will not bundle FML by default since it tries to reuse it

timber ingot Jul 10, 2024, 1:49 PM

#

Alright

#

So, can we add a makeDevInstaller task which does this?

haughty copper Jul 10, 2024, 1:50 PM

#

You'd likely need to build some special casing around it. Might be doable, but the value in NF is more about testing the installer than testing Neo itself

#

Since Neo largely behaves the same now

#

It's mostly FML that has major differences

#

the same between dev and prod I mean

#

For testing the installer I'd rather have had a github workflow based testing step to test the real installer that we hand out, ultimately

timber ingot Jul 10, 2024, 1:51 PM

#

I really don't like the wishy washy runtime back link from NeoForge to FML, it just irks me and i think we should improve the testability of both projects instead of just FML

haughty copper Jul 10, 2024, 1:51 PM

#

I stand by my opinion it'd be easier to pull FML back into NF 😛

#

if there's some startup component that is ultimately less coupled

#

that can be pulled out

#

but currently it's pretty tightly coupled and having it in two repos is just a pain in the ass

timber ingot Jul 10, 2024, 1:52 PM

#

the point of being detached is that FML is meant to be stable and not really have 'much' iteration happening between mc versions, it can be shared, etc

#

recently thats not really the case, there is a lot of iteration, thats fine, but the end goal is FML is stable and is used for more than one minecraft cycle

#

which i think is feasable

haughty copper Jul 10, 2024, 1:53 PM

#

I mean we had the discussion about Neo-as-a-mod. I think in that scenario you'd have 80% of what FML is today in Neo itself

glacial crag Jul 10, 2024, 1:53 PM

#

It's not a reasonable goal

haughty copper Jul 10, 2024, 1:53 PM

#

and maybe 20% that's actually independent

glacial crag Jul 10, 2024, 1:54 PM

#

A lot of the FML code could be moved into NF just fine

haughty copper Jul 10, 2024, 1:54 PM

#

Which has less to do with the coupling and more with: ideally you'd never have to patch the base install that loads neo-as-a-mod over the full lifecycle of a MC version

glacial crag Jul 10, 2024, 1:54 PM

#

Every FML feature development is currently a PITA, precisely because you have to go back and forth between NF and FML

haughty copper Jul 10, 2024, 1:55 PM

#

making it possible for the end user to not care about any kind of "version" for that base install

glacial crag Jul 10, 2024, 1:55 PM

#

haughty copper Which has less to do with the coupling and more with: ideally you'd never have t...

That too. FML should be stripped down to what we can actually ensure will remain untouched. Rather than what "just happens to be there at the moment".

haughty copper Jul 10, 2024, 1:56 PM

#

No idea that is really some future stuff

#

Currently I am just trying to improve the existing process as much as possible. Being able to test in prod quickly helps a lot

#

Even testing non prod helps alot

#

Without having to make a full NF checkout + setup + includeBuild dance

#

It's currently not possible to start or test earlydisplay in isolation at all, sadly

glacial crag Jul 10, 2024, 1:57 PM

#

Yes we need to work on that testing setup

#

Most outsider FML PRs are not even properly tested because of how annoying it is to setup atm

haughty copper Jul 10, 2024, 1:58 PM

#

I mean at least the FML Loader aspects I did try to write as many unit tests for as possible where I do simulate a production install. But you know how it goes, that only gets you so far 😅

timber ingot Jul 10, 2024, 1:58 PM

#

I really dont like the split in tools we have now, you guys have gone off and made ModDevGradle and NFRT stuff, and We still have NeoGradle. I'm wondering when those are going to converge back together. I can see why this PR is using NFRT, it has some nice tools

haughty copper Jul 10, 2024, 1:59 PM

#

Have you actually looked at NeoGradle code?

#

Like, for real

#

I do not have the mental capacity for it 😐

chrome mortar Jul 10, 2024, 1:59 PM

#

haughty copper I mean at least the FML Loader aspects I did try to write as many unit tests for...

even then, the FML tests have a lot of glue because FML has a non-zero amount of arguments/processes that are tied to mc/neo directly

timber ingot Jul 10, 2024, 2:00 PM

#

Not with a fine tooth comb, but I personally would have spent my time developing NG v++ instead. But thats just me, having a split is more confusing

haughty copper Jul 10, 2024, 2:00 PM

#

NFRT / MDG are largely an attempt at "divide and conquer" applied to the complexity of the overall process

timber ingot Jul 10, 2024, 2:00 PM

#

Do you plan to propose turning MDG into NG v++?

haughty copper Jul 10, 2024, 2:01 PM

#

By having a hard process boundary between the MCP/NeoForm process and the insanity of Gradles "my way or the highway" approach to API design

haughty copper Jul 10, 2024, 2:01 PM

#

timber ingot Do you plan to propose turning MDG into NG v++?

I propose we offer it as the default moddev plugin for NeoForge, ofcourse. That's what that thing is made for

timber ingot Jul 10, 2024, 2:01 PM

#

So is NFRT designed to encapsulate using NeoForm?

haughty copper Jul 10, 2024, 2:01 PM

#

Yes

#

Hence it's name 😄

#

NeoForm Runtime

timber ingot Jul 10, 2024, 2:02 PM

#

huh, always read it as NeoForgeRunTime, dunno why

haughty copper Jul 10, 2024, 2:02 PM

#

NF and NF having the same acronym is a bit of a problem sometimes

timber ingot Jul 10, 2024, 2:02 PM

#

Then realistically its just a replacement for InstallerTools?

haughty copper Jul 10, 2024, 2:02 PM

#

InstallerTools doesn't do NeoForm

#

just the binpatches

timber ingot Jul 10, 2024, 2:02 PM

#

Or do the encapsulate different sets of tools?

haughty copper Jul 10, 2024, 2:02 PM

#

while NFRT only does NeoForm, but no binpatches

#

NeoForm is ultimately only used at dev-time directly

chrome mortar Jul 10, 2024, 2:02 PM

#

haughty copper just the binpatches

installertools is a.. weird thing

#

haughty copper Jul 10, 2024, 2:03 PM

#

installer downstream only uses the binary result of diffing a compiled NeoForm vs. the original

chrome mortar Jul 10, 2024, 2:03 PM

#

it's a collection of tasks the installer runs in steps

timber ingot Jul 10, 2024, 2:03 PM

#

It has a bit more than bin patches last i checked, has stuff for manually downloading things, and all sorts

haughty copper Jul 10, 2024, 2:03 PM

#

well yes, that is true

#

But it doesn't (or shouldn't) process the NeoForm data files

chrome mortar Jul 10, 2024, 2:03 PM

#

recently it became a collection of tools the installer runs, so it's installertools+binpatch+inject+jarsplit

timber ingot Jul 10, 2024, 2:03 PM

#

its fucking disgusting, i had to hack around the download tasks in InstallerTools in the FTBApp, not fun

haughty copper Jul 10, 2024, 2:04 PM

#

There is some overlap for sure though, that much is true

#

But OTOH NFRT ships stuff we'd never want to ship in installertools

#

such as a javac and eclipse compiler driver 😄

chrome mortar Jul 10, 2024, 2:04 PM

#

or JST

haughty copper Jul 10, 2024, 2:04 PM

#

You could make it load these optionally

timber ingot Jul 10, 2024, 2:04 PM

#

personally i fucking hate the eclipse compiler, and would rather see it rot in hell and nowhere near our tools

haughty copper Jul 10, 2024, 2:04 PM

#

That's also true

chrome mortar Jul 10, 2024, 2:04 PM

#

~~the intellij psi is too many megabytes~~

haughty copper Jul 10, 2024, 2:04 PM

#

/sigh I mean I agree

#

But it's multi-threaded

#

and saves me 10-20 secs of project reload time

timber ingot Jul 10, 2024, 2:05 PM

#

at the cost of less language acuracy

haughty copper Jul 10, 2024, 2:05 PM

#

So I did include it as an opt-in alternative to javac

timber ingot Jul 10, 2024, 2:05 PM

#

but anyway

haughty copper Jul 10, 2024, 2:05 PM

#

Yeah let me make one point though 😅

#

The aim was to pull out all that into an external tool

chrome mortar Jul 10, 2024, 2:06 PM

#

timber ingot at the cost of less language acuracy

yeah it did break some mixins, but for the average user it's a better choice usually

haughty copper Jul 10, 2024, 2:06 PM

#

with the goal of having the Gradle plugin be as simple as we can make it

chrome mortar Jul 10, 2024, 2:06 PM

#

at least in terms of speed

haughty copper Jul 10, 2024, 2:06 PM

#

Since maintaining a Gradle plugin sucks

#

especially with how they evolve their API and put new requirements on plugin authors constantly

timber ingot Jul 10, 2024, 2:06 PM

#

Matty sorry for getting angry before, i was triggered that people dog piled on me and took it out on the next person who pissed me off

#

I still really dont feel comfortable with this FML PR, and I'm not sure what the solution is

haughty copper Jul 10, 2024, 2:08 PM

#

Deleting the Github repo is the only option 😄

chrome mortar Jul 10, 2024, 2:08 PM

#

timber ingot Matty sorry for getting angry before, i was triggered that people dog piled on m...

it's fine, maybe I was a bit passive aggressive

#

~~start from scratch, delete meta inf~~

timber ingot Jul 10, 2024, 2:08 PM

#

chrome mortar it's fine, maybe I was a bit passive aggressive

I think so. If you work on that, ill work on not getting angry in the first place. deal? :D

haughty copper Jul 10, 2024, 2:08 PM

#

So what's absolutly clear: that's a lot of gradle code

#

I don't like a lot of Gradle code

chrome mortar Jul 10, 2024, 2:09 PM

#

tbh fml being split has one benefit I noticed.. it reduced our build times by a significant amount

timber ingot Jul 10, 2024, 2:09 PM

#

haughty copper So what's absolutly clear: that's a lot of gradle code

I think thats part of whats irking me, we have basically all of this in NG

#

which is why it feels better to have these tools around the other way

haughty copper Jul 10, 2024, 2:09 PM

#

It's ultimately half-half roughly I'd say between:

Provide a "generic" way to define Gradle run and IDE run in one go which is completely detached from the minecraft aspect and could be pulled out into a separate plugin
Running the NF installer, which could theoretically also be pulled out

#

Yeah but there's a trade-off since those things are internal in NG and making bespoke tooling in NG will cause you large headaches with tooling dependencies too

#

We had that a while ago with NG changes that were intended for moddev ending up breaking neodev if I am not mistaken

#

but i am 50:50 on that. there was just something up with it

chrome mortar Jul 10, 2024, 2:10 PM

#

NG also has to deal with gradle

#

which is and has been.. a problem

haughty copper Jul 10, 2024, 2:11 PM

#

In general I don't think the experience of making "large" kitchen sink plugins has been very good

chrome mortar Jul 10, 2024, 2:11 PM

#

if we want a maintainable toolchain we need to decrease our gradle footprint

#

not increase and abuse it

haughty copper Jul 10, 2024, 2:11 PM

#

So, ONE THING that was extremely shitty

#

This shit: https://github.com/neoforged/FancyModLoader/pull/177/files#diff-397214a37146f55b5a423f3a0962ee54078bac42bb32f1293db0a882b6ea91f4R220

#

I think that was the third or fourth time I wrote a damn parser / evaluator for a Minecraft version manifest JSON

#

Since we have no lib for that

#

Since that has absolutly no relationship with gradle whatsoever, we might really benefit by pulling out a lib for:

reading NeoForge version JSONs
reading NeoForm version JSONs
reading Minecraft launcher manifests
reading Mincraft version manifests (and evaluating rules)

#

since we have probably 4-5 copies of that lying around

timber ingot Jul 10, 2024, 2:13 PM

#

Yes, the rule evaluation is terrible, i had to do it in the FTBApp

haughty copper Jul 10, 2024, 2:14 PM

#

I tried disassembling the launcherlib to figure out what the other values for arch are

#

no dice 😄

timber ingot Jul 10, 2024, 2:14 PM

#

I think the IDE+Gradle runs is a good idea for a plugin, go make that

haughty copper Jul 10, 2024, 2:14 PM

#

I think they only use 'x86' currently, but didn't check all versions

timber ingot Jul 10, 2024, 2:16 PM

#

So when your testing FML in prod, you will probably also want NF in prod to test changes to FML, since the whole point is to test the prod stuff

#

i really want to find a way that benifits both projects

haughty copper Jul 10, 2024, 2:17 PM

#

Yes you'll want both approaches

#

There are changes to i.e. earlydisplay

#

where you do not need to have NF changes

#

but conversely

#

the config changes need NF adaption and you'd want includeBuild to work

#

but for those, you do not need production

#

neodev is likely enough

#

includeBuild fun is where legacyClasspath comes into play and bite you. It should be better now, but I haven't tested it in a bit

timber ingot Jul 10, 2024, 2:19 PM

#

To me, logically, it seems benificial at the moment to pivot and do this from the NF side

#

we should be able to build fat installer jars

#

we should have a task which allows you to test NF in prod from NF. If we end up moving 80% of FML to NF for the neo-as-a-mod stuff, we will need it there anyway

#

we should make that work with includeBuild

haughty copper Jul 10, 2024, 2:20 PM

#

I'll be honest. Doing this in NF is likely going to involve 3-4x the amount of code

#

and will not be done in 1-2 months

timber ingot Jul 10, 2024, 2:21 PM

#

Alright, why's that?

haughty copper Jul 10, 2024, 2:21 PM

#

And it'll still be an inferior process

#

Since you do need an up-to-date neoforge checkout with setup for any change you want to test on FML

#

But going back to your question

#

https://github.com/neoforged/NeoGradle/blob/NG_7.0/platform/src/main/java/net/neoforged/gradle/platform/extensions/DynamicProjectExtension.java

#

I don't even know where to start 😅

glacial crag Jul 10, 2024, 2:22 PM

#

That's comparatively tame... at least the amount of indirection is quite low in that class

timber ingot Jul 10, 2024, 2:23 PM

#

Alright, this doesn't really answer my question, what specifically about this makes it a pain?

#

what modifications are needed here to achive the setup i have described?

glacial crag Jul 10, 2024, 2:26 PM

#

covers I feel like you're missing the point. If the Gradle setup for prod were easy we'd already have it in MDG

timber ingot Jul 10, 2024, 2:26 PM

#

Im trying to understand the problem

glacial crag Jul 10, 2024, 2:26 PM

#

Serious question: what does it buy you? And what does it buy us

timber ingot Jul 10, 2024, 2:27 PM

#

You have linked me to a large blob of code, fantastic. Its Gradle crap. What I don't understand is why I can't get an answer as to why this code is a problem

timber ingot Jul 10, 2024, 2:28 PM

#

glacial crag Serious question: what does it buy you? And what does it buy us

I think i have made it pretty clear what I don't want at the moment, and that PR irks me and I would like to see the alternatives, and thats what we are discussing, the alternatives

glacial crag Jul 10, 2024, 2:29 PM

#

Is there anything to your argument besides that you'd prefer this to be in NG?

#

And that this should ideally be in NF and not FML

haughty copper Jul 10, 2024, 2:29 PM

#

timber ingot You have linked me to a large blob of code, fantastic. Its Gradle crap. What I d...

You did answer it yourself

#

That gradle blob is the build logic for NeoForge

#

Why is that in an external plugin, making it extremely hard to actually evolve it

#

Hiding our skeletons in a Git-repository-sized closet doesn't make them go away

timber ingot Jul 10, 2024, 2:30 PM

#

haughty copper Why is that in an external plugin, making it extremely hard to actually evolve i...

Fantastic question, so why don't we do something about that? Move chunks of it back to NF?

haughty copper Jul 10, 2024, 2:30 PM

#

Because Orion doesn't want me to

glacial crag Jul 10, 2024, 2:31 PM

#

Why do you think we made a new plugin lol

timber ingot Jul 10, 2024, 2:31 PM

#

So it sounds like you have had a disagreement with another maintainer and spawned off another project out of their control?

timber ingot Jul 10, 2024, 2:32 PM

#

glacial crag Is there anything to your argument besides that you'd prefer this to be in NG?

I havent said i'd prefer this to be in NG, I just don't like that its in FML.

#

(shit, ping)

haughty copper Jul 10, 2024, 2:32 PM

#

The consequence of it being in NF is it being in NG

#

But as I also said, that is not really the main concern

#

The main concern is that you do have to have an up-to-date workspace that has been setup

#

*NeoForge workspace

#

And I am concerned it'll diminish the benefits

#

Since I'll be honest, I do not keep my NF workspace up to date that much since I mostly work on other repos at the moment

timber ingot Jul 10, 2024, 2:33 PM

#

I fail to see how that is a serious problem, you clone NF, its a single git pull

#

your working with remote repos, this is not something forigen

#

It really sounds like your putting it where it is, because you don't have to go through Orion to make it happen

#

and imo, thats not okay

#

we are a team

glacial crag Jul 10, 2024, 2:36 PM

#

We are putting it in FML because that's where it makes the most sense now

#

Since its primary purpose is to test FML and changes to it

haughty copper Jul 10, 2024, 2:36 PM

#

timber ingot I fail to see how that is a serious problem, you clone NF, its a single git pull

and then you run gradlew setup and wait 😛

glacial crag Jul 10, 2024, 2:36 PM

#

If we wanted to bypass Orion we'd put it in MDG not FML

haughty copper Jul 10, 2024, 2:36 PM

#

timber ingot It really sounds like your putting it where it is, because you don't have to go ...

I think that is a gross mischaracterization

#

Because it is going through Orion 😄

#

I just like rapid iteration, and forcing includeBuild into NF as the only way to test FML changes in a real MC client is not very rapid

timber ingot Jul 10, 2024, 2:38 PM

#

Im gonna step out of this entire convosation. I really couldn't care at this point. Its obvious that you guys don't like my suggestions and concerns and belive your solution here is best. It may be.

haughty copper Jul 10, 2024, 2:38 PM

#

Hey that is not true, I do agree it should also be possible to do includeBuild of FML in NF

#

and I also agree that we should try to minimize the loose gradle code that can be reused

#

I just don't think that it should preclude having the ability to run FML directly without a second git workspace

timber ingot Jul 10, 2024, 2:40 PM

#

But I will put on record, that I am very disappointed that beef between developers caused an entire split in our tooling, that is not something that would have happend if we had proper leadership in this entire project, It is my opinion that the democratic system we have here is not working as it should, and i have absolutely no idea how to fix it.

haughty copper Jul 10, 2024, 2:40 PM

#

Covers

#

The best Gradle plugin from a user perspective is gonna get traction

#

Whether it has a NF label on it

#

or not

timber ingot Jul 10, 2024, 2:40 PM

#

If thats MDG then so be it, if its not, whatever

haughty copper Jul 10, 2024, 2:41 PM

#

Well yes

timber ingot Jul 10, 2024, 2:41 PM

#

I just think it was spawned out of bad faith

haughty copper Jul 10, 2024, 2:41 PM

#

But how do you even get the chance to try a new approach

chrome mortar Jul 10, 2024, 2:41 PM

#

timber ingot But I will put on record, that I am very disappointed that beef between develope...

for the record

timber ingot Jul 10, 2024, 2:41 PM

#

and I'm not a fan

haughty copper Jul 10, 2024, 2:41 PM

#

if the answer is always "no"

chrome mortar Jul 10, 2024, 2:41 PM

#

neogradle was not a democratic system when mdg was born

#

it was in orion's full control as a subproject lead

glad cobalt Jul 10, 2024, 2:41 PM

#

chrome mortar it was in orion's full control as a subproject lead

That is plain wrong! For the record!

haughty copper Jul 10, 2024, 2:42 PM

#

If you said no on a PR

#

it aint gonna get merged

glad cobalt Jul 10, 2024, 2:42 PM

#

haughty copper If you said no on a PR

Then get a vote

#

I won't get a Pr merged in mdg if you or tech says no

#

Or when touching capabilities

timber ingot Jul 10, 2024, 2:42 PM

#

chrome mortar neogradle was *not* a democratic system when mdg was born

Then the solution then should have been to see about removing the sub-project lead system(as it is now) as you believed it to not be working, instead of spawning another project just do have your own ways.

haughty copper Jul 10, 2024, 2:43 PM

#

You'd be surprised Orion 😅

#

I already got overruled

chrome mortar Jul 10, 2024, 2:43 PM

#

glad cobalt That is plain wrong! For the record!

They may:

Make final decisions on the implementation and structure of their subproject.

#

we don't need to get into this again

timber ingot Jul 10, 2024, 2:43 PM

#

Anyway, my disapproval is there on record

#

i dont wish to entertain this convosation anymore

#

Thanks Shartte for the great insight into everything, i appreciate your time

haughty copper Jul 10, 2024, 2:44 PM

#

Yeah sorry for the circumstances 😅 I want more people to participate in the technical fringes

#

so thank you for the interest in it, we really need more of that

timber ingot Jul 10, 2024, 2:45 PM

#

I will add another note, We should work on how we describe PR's, as everything we just went through (a condensed form) should be in that PR description

#

so a less informed maintainer can pick up the key points and enquire about those specifically here or in the comments

haughty copper Jul 10, 2024, 2:46 PM

#

Ah yes, I can see that. Ultimately it has no effect on the project artifacts so I didn't quite put in the usual effort into the PR description

timber ingot Jul 10, 2024, 2:47 PM

#

I don't think we should have any legislation for it, but some general guidelines or something would be nice, dont want to add too much bureaucracy to it, but we should try and be better about it overall

haughty copper Jul 10, 2024, 2:47 PM

#

I advocated for that myself, you don't need to remind me 😅

#

Well no, you actually do 😄

#

Since I did get lazy on this description

timber ingot Jul 10, 2024, 2:47 PM

#

:)

glad cobalt Jul 10, 2024, 3:22 PM

#

Technical question: @haughty copper do you use dynamic attach in your new fml system?

haughty copper Jul 10, 2024, 3:24 PM

#

Only as a fallback if the intended normal attach didn't work for whatever reason. I am aware that this will go away next LTS

#

My thinking is: The only situation in which a normal attach will be gone is where ALL jvm args are dropped

#

that would only apply to potentially running JUnit tests from Eclipse, or from IJ

#

I belive we are in the same boat as many enterprise customers there, which should give us an out until they disable it in the next LTS

chrome mortar Jul 10, 2024, 3:28 PM

#

haughty copper My thinking is: The only situation in which a normal attach will be gone is wher...

*and there is no executable jar

haughty copper Jul 10, 2024, 3:29 PM

#

Well even if you don't. agent is just another JVM arg like the many we already need to make it run

#

If you have literally no args at all preserved

#

Then it can be a last-resort

#

So.... I'll bugger off to some more pulumi /sigh

timber ingot Jul 10, 2024, 3:34 PM

#

Yeah I think they made it clear that agents aren't going away, just dynamic attach, if you want an agent, add it at startup

#

not sure what their plan for profilers is, but we'll see

chrome mortar Jul 10, 2024, 3:35 PM

#

even with dynamic attach, there's a non-zero chance they'll end up with a compromise

#

there's a ton of enterprises using bytebuddy and co

#

so it's not really something i'd worry about in general

timber ingot Jul 10, 2024, 3:35 PM

#

nah

#

its fine

haughty copper Jul 10, 2024, 3:35 PM

#

I think it's a tooling problem

#

Gradle has an issue for it already, IJ will have to follow suit

#

Imagine having a built-in config similar to the annotationProcessor one

#

Where you shove in an agent dependency, and it actually goes and just adds that to the JVM args.
IntelliJ would also read that.

If they solve that, they're kinda set to remove the dynamic attach, I think

chrome mortar Jul 10, 2024, 3:36 PM

#

haughty copper I think it's a tooling problem

ah definitely, the tooling always ends up being the middleman problem

haughty copper Jul 10, 2024, 3:37 PM

#

I mean it kinda is in this case. ByteBuddy (the backend for Mockito) already supports just attaching the agent directly, it's just that it's too hard to do via Gradle / IJ

#

But yeah if we go down this route it should be with attach via JVM arg and dynamic attach only as a fallback for bad tooling in tests

#

Here's the Gradle issue: https://github.com/gradle/gradle/issues/27714

chrome mortar Jul 10, 2024, 3:39 PM

#

i hope it doesn't end up being a "charset take 2" issue

haughty copper Jul 10, 2024, 3:42 PM

#

Hm, yes, but assuming gradle and IJ don't get their shit together, I think the outrage of us corporate drones combined may make OpenJDK wait with the removal heh

#

We'll see

glad cobalt Jul 10, 2024, 3:43 PM

#

Okey

#

Perfect seems that you thought this through

#

So one more checkbox

haughty copper Jul 10, 2024, 3:45 PM

#

There's a workaround btw. which is ultimately --add-opens=java.base/java.lang=ALL-UNNAMED

#

Since we only use the module redefinition API of Instrumentation in startup-experiments at the moment

#

And you kinda can get the same through reflection

#

But that seems even less supported

#

Or rather, not less, since instrumentation is supported.

#

Just not what we are doing with it 😅

chrome mortar Jul 10, 2024, 3:46 PM

#

you can also use Modules directly if you export at compile-time, but in both cases it's less supported than instrumentation

#

not like instrumenting java modules to be open is a supported use-case in any case

#

more like the tools used are supported

haughty copper Jul 10, 2024, 3:50 PM

#

@glad cobalt Another thing I found surprising but very useful: When developing FML itself with an agent, we do not actually need to use the full jar as the -javagent

#

It is sufficient to pass a jar to that argument that only has the MANIFEST.MF entry for the agent class you want to use

#

it'll happily load that class from your normal CP

#

That means we can check in a trivial jar with a fixed name that just contains the correct MANIFEST.MF to launch our agent, and use that jar for IDE runs in FML

plain solar Jul 11, 2024, 12:08 PM

#

@haughty copper I get error: LaunchServiceHandler is not public in cpw.mods.modlauncher trying to compile startup experiments; what did I forget?

glacial crag Jul 11, 2024, 12:08 PM

#

Are you on the right version of ML?

#

You need a custom build of some PR, no?

haughty copper Jul 11, 2024, 12:37 PM

#

Oh sorry yes please clone mod launcher branch of the same name and check the includeBuild in settings.gradle

glacial crag Aug 10, 2024, 12:32 PM

#

what's the status of https://github.com/neoforged/FancyModLoader/pull/177?

GitHub

Add the ability to install and run a production client and server a...

This adds the ability to:

Run :test:installNeoForgeServer to install a local copy of the NeoForge server into build/installs/server using the installers silent-install feature. This also produces ...

haughty copper Aug 10, 2024, 1:27 PM

#

Well, merge it!

glad cobalt Aug 10, 2024, 1:29 PM

#

No

glacial crag Aug 10, 2024, 1:32 PM

#

then what's the status? I already went through all the Gradle stuff extensively

glad cobalt Aug 10, 2024, 1:33 PM

#

NFRT

#

It should simply not be there

#

Period

#

I would like to use its asset downloading features

#

But I don't want to use NFRT

#

Simple as that

glacial crag Aug 10, 2024, 1:33 PM

#

We aren't splitting them out of NFRT. Period.

#

If you want them, use NFRT. If you don't want them, don't

glad cobalt Aug 10, 2024, 1:33 PM

#

glacial crag If you want them, use NFRT. If you don't want them, don't

Hwy not?

#

It is a dedicated task

#

That has nothing to do with neoform

#

It can then be used in the installer

#

In NeoGradle

#

And other locations

glacial crag Aug 10, 2024, 1:34 PM

#

there's no need for them in the installer afaik?

#

and NG should be able to just use NFRT

glad cobalt Aug 10, 2024, 1:35 PM

#

I think the installer could make use of it to create a cleaner and faster setup no?

#

Right now it requires the client to be run first

#

To get all the natives etc in place

#

Instead of forcing it to just look at the vanilla location

#

It would be reasonable that it looks at other locations as well, and potentially copies them over no?

glacial crag Aug 10, 2024, 1:36 PM

#

I mean, the other locations are already "where launchers store their stuff"

#

so I don't think we could invert that

glad cobalt Aug 10, 2024, 1:37 PM

#

No i mean, right now we require the vanilla launcher to be ran

#

Regardless of whether other launchers ran or not

#

It would make sense to allow the installer to grab data from them

glacial crag Aug 10, 2024, 1:37 PM

#

for the assets?

glad cobalt Aug 10, 2024, 1:37 PM

#

Instead of having that requirement build in

#

Wait it just does the assets?

#

Not the natives?

haughty copper Aug 10, 2024, 1:38 PM

#

NFRT is the right tool for the job here

glad cobalt Aug 10, 2024, 1:39 PM

#

It is not

#

It is a runtime implementation of neoform specification

#

It was not supposed to do this in the first place

#

NeoForm has no requirement in its specification that it downloads the assets

haughty copper Aug 10, 2024, 1:39 PM

#

It's a dev support tool and this is exactly what it's supposed to do

glad cobalt Aug 10, 2024, 1:39 PM

#

Or does anything

haughty copper Aug 10, 2024, 1:39 PM

#

This is FML, not NeoForm please stay on topic

glad cobalt Aug 10, 2024, 1:39 PM

#

haughty copper It's a dev support tool and this is exactly what it's supposed to do

Then move it to the original dev support tool: InstallerTools

haughty copper Aug 10, 2024, 1:39 PM

#

The installer has no requirement to download assets either

#

you are being inconsistent

glacial crag Aug 10, 2024, 1:40 PM

#

I don't think there is a strong reason NOT to use NFRT given that it already exists and does what we want

#

in the future it should be possible to reorganize parts of it if it would help with the installer

glad cobalt Aug 10, 2024, 1:40 PM

#

haughty copper This is FML, not NeoForm please stay on topic

I am staying on-topic. This PR adds a tool desgined for use in Mod Development Environments, solely for a functionality that should not be in it in the first place, to a tool that has no bearing on NeoForm

glacial crag Aug 10, 2024, 1:40 PM

#

is there a technical reason not to use NFRT?

haughty copper Aug 10, 2024, 1:41 PM

#

Apparently not

glad cobalt Aug 10, 2024, 1:41 PM

#

Yes the architecture is shit

glacial crag Aug 10, 2024, 1:41 PM

#

lol

haughty copper Aug 10, 2024, 1:41 PM

#

Bahahaha sure Orion

glad cobalt Aug 10, 2024, 1:41 PM

#

Well maybe not shit

#

The idea to use a tool is good

#

The idea that that tool should be as re-useable is also good

haughty copper Aug 10, 2024, 1:41 PM

#

So a tool that is used in dev which requires asset downloading is shit

#

but moving it to the installer which does not need this functionality is good

glad cobalt Aug 10, 2024, 1:41 PM

#

haughty copper So a tool that is used in dev which requires asset downloading is shit

No....

haughty copper Aug 10, 2024, 1:41 PM

#

you got your architecture backwards there

glad cobalt Aug 10, 2024, 1:42 PM

#

The fact that this functionality is tagged on to a tool that has no bearing being in FMLs dependency chain in any location is

haughty copper Aug 10, 2024, 1:42 PM

#

I might also add: you are litigating your thoughts on NFRT

#

on a PR that adds testing to FML

#

I find this highly inappropriate

glad cobalt Aug 10, 2024, 1:42 PM

#

If this exact code was not in NFRT it would be merged right here right now

haughty copper Aug 10, 2024, 1:42 PM

#

And I wonder Orion, what will you do when I start reviewing your NG PRs and veto them on architectural grounds

#

for completely unrelated reasons

glad cobalt Aug 10, 2024, 1:42 PM

#

haughty copper on a PR that adds testing to FML

It also adds NFRT to FML

haughty copper Aug 10, 2024, 1:42 PM

#

To it's build script. For testing.

#

It has no bearing on FMLs own architecture

glacial crag Aug 10, 2024, 1:42 PM

#

I think the real answer is that you dislike NFRT because it supports MDG which will probably replace NG long-term

#

which is unfortunately not a valid reason to block this PR

haughty copper Aug 10, 2024, 1:43 PM

#

That is also my take after holding a PR that has no bearing on FMLs actual architecture for so long.

glad cobalt Aug 10, 2024, 1:43 PM

#

It does not on FMLS own, but on how FML is being developed

#

NFRT was designed to be used in MDG

haughty copper Aug 10, 2024, 1:43 PM

#

No?

glad cobalt Aug 10, 2024, 1:43 PM

#

Yes

haughty copper Aug 10, 2024, 1:43 PM

#

That is a lie

#

No it's a lie

#

It was developed before MDG even existed

#

Check the commit history

glad cobalt Aug 10, 2024, 1:44 PM

#

But it was designed to be used in MDG?

haughty copper Aug 10, 2024, 1:44 PM

#

No

glad cobalt Aug 10, 2024, 1:44 PM

#

Was it not?

glacial crag Aug 10, 2024, 1:44 PM

#

MDG came after NFRT

haughty copper Aug 10, 2024, 1:44 PM

#

The idea for MDG didnt even exist when NFRT was started

glacial crag Aug 10, 2024, 1:44 PM

#

the starting idea behind MDG was "hey with this NFRT thing we should be able to build a gradle plugin quite easily"

haughty copper Aug 10, 2024, 1:44 PM

#

MDG was started as a way to test NFRT

glad cobalt Aug 10, 2024, 1:45 PM

#

I understand its idea.

#

Merge it

#

But that also means that we can trivially rip it back out

#

When we have a dedicated library for it no?

haughty copper Aug 10, 2024, 1:45 PM

#

That's the entire argument that you can

#

Look at the PR it is concentrated completely in a separate subproject

glacial crag Aug 10, 2024, 1:46 PM

#

glad cobalt When we have a dedicated library for it no?

of course yes; that library would be pulled out of NFRT anyway (for example if shown that it would help with the installer's performance or something like that)

glad cobalt Aug 10, 2024, 1:46 PM

#

I removed the blocker

glacial crag Aug 10, 2024, 1:47 PM

#

thanks

#

shartte I see that you removed the (broken) IJ runs

#

do we want to add Gradle task IJ runs?

#

or actually, if they were too broken that's not even an option

glacial crag Aug 10, 2024, 7:54 PM

#

important slip up: https://github.com/neoforged/FancyModLoader/pull/193

GitHub

Undo bump of AccessTransformers to 11.0.1 (back to 10.0.1 for now) ...

#177 bumped the AT dependency to major version 11, which removes the ANTLR dependency. This is a good thing, however it is also a breaking change that should be kept for the next BC cycle.

#

the other dep bumps in #177 are fine

haughty copper Aug 10, 2024, 9:04 PM

#

glacial crag do we want to add Gradle task IJ runs?

Nah, doesn'T matter

#

Wait, we never actually bumped AT to the ANTLR-free version?

#

Really? 😄

#

That feels like it was merged ages ago

glacial crag Aug 10, 2024, 9:49 PM

#

Yes

#

But we didn't bump it

glacial crag Aug 11, 2024, 8:47 PM

#

shartte I think you forgot to approve https://github.com/neoforged/FancyModLoader/pull/193

#

given that you enabled auto-merge, I suppose that was your intention

haughty copper Aug 11, 2024, 8:48 PM

#

yes it was

#

thx for the hint ^^

split wasp Aug 12, 2024, 3:00 AM

#

glacial crag important slip up: https://github.com/neoforged/FancyModLoader/pull/193

Ugh, that never actually got bumped beforehand? Lovely. Gotta remember to do that for 1.21.2 I guess

errant yew Aug 18, 2024, 1:34 PM

#

Since we now have a channel dedicated for FML discussions, and I think this thread has outlived its usefulness now that most cleanup has been performed on FML, I'll be closing this.
Redirect all future discussion to #loader-dev, kthx

glacial crag Nov 20, 2024, 4:54 PM

#

How viable would it be to have a single launch target (instead of client/data/server), and pass the main class as a launch argument? Just a thought 😄

haughty copper Nov 20, 2024, 5:11 PM

#

God why are you writing in this necro thread 😄

#

Go to #loader-dev

errant yew Nov 20, 2024, 5:11 PM

#

what the H

haughty copper Nov 20, 2024, 5:17 PM

#

glacial crag How viable would it be to have a single launch target (instead of client/data/se...

Also please do post it in there since the topic is interesting 😛

glacial crag Nov 20, 2024, 5:28 PM

#

Oh oops

#FML Clean-up