https://github.com/FTBTeam/FTB-Modpack-Issues/issues/6183 Unresolved

[no-name-provided]

Modpack

FTB Evolution

Modpack version

1.8.0

Launcher

FTB App

Has the pack been modified

Yes

Log Files

https://gist.github.com/no-name-provided/cbbb129e00d4045a04f3d4c5f1535ec0

Describe the bug

This bug is still in the pack, and blocking progression in the FTB Pyramid quests. Specifically, I've killed the ender dragon 4 times, and my dragon model was finally saying basic, but teleporting back to my base reverted it to faulty. It refuses to produce predictions.

Steps to reproduce

Place ender dragon model in deep learner, the deep learner in the deep learner curio slot, and kill the ender dragon. Then hop in the portal, then go to the mining dimension.

Expected behaviour

A basic tier ender dragon data model and (shortly) a stack of predictions I can use to make "evolutionary matter".

Screenshots

Again, using alltheleaks. Also implemented the ruined book recipe fix mentioned in related ticket.

Additional information

No response

[eSonOfAnder]

Per the HNN mod dev, this appears to be an issue with the Adorned mod that FTB Evo is using for its Curios implementation. Take the deep learner out of the curio slot and it will work as expected.

Shadows-of-Fire/Hostile-Neural-Networks#75

This is most likely something that the Adorned mod dev is going to have to fix; not much the FTB team will be able to do unless they swap out the Curios mod that they're using. Probably best to report it to the Adorned dev - https://github.com/JDKDigital/Adorned/issues

[no-name-provided]

Those are useful links, but there are a few issues with your interpretation/suggestion:

1. shadows-of-fire said "Not sure what would be causing this. It doesn't appear to happen in isolation", then concluded it was a mod interaction or a dependency mismatch. In short, they don't know. Figuring out if (and which) which mods are fighting is something the FTB team can do - that's presumably why we're told to report problems to pack devs first.
2. This is only an issue with the ender dragon model, and it's a consistent issue with that model. The original reporter on Mob model kill counter going in reverse Shadows-of-Fire/Hostile-Neural-Networks#75 observed more complex behavior.
3. Adorned is no longer receiving updates: Half of the curios ui hides behind EMI's item viewer thingy JDKDigital/Adorned#16 (comment). It's recommended that users switch to the official 1.21 port of the curios mod.

Personally, my guess would be Savage Ender Dragon is doing something server side that causes the server version of Hostile Neural Networks to not register the kill. Still, I'll try putting the deep mob learner on my hotbar and killing the dragon again.

[LeSwede]

2. This is only an issue with the ender dragon model, and it's a consistent issue with that model. The original reporter on Mob model kill counter going in reverse Shadows-of-Fire/Hostile-Neural-Networks#75 observed more complex behavior.

I had this issue in ATM10 as well. The Deep learner doesnt work properly in the curio slot so its not isolated to FTB packs.

[no-name-provided]

I had this issue in ATM10 as well. The Deep learner doesnt work properly in the curio slot so its not isolated to FTB packs.

That's the thing: it does work fine. I didn't notice any problem with the magma cube, warden, wither, etc. models, but the ender dragon model, specifically, reset at least three times (each time roughly coinciding with leaving the end). I didn't even need to take the model out of the learner to make it happen.

It's possible for a mod to have more than one bug. The bug report eSonOfAnder linked was for AllTheMods10, BTW. It's probably the same one you're experiencing. It may or may not be related to this one.

[eSonOfAnder]

I had this issue in ATM10 as well. The Deep learner doesnt work properly in the curio slot so its not isolated to FTB packs.

That's the thing: it does work fine. I didn't notice any problem with the magma cube, warden, wither, etc. models, but the ender dragon model, specifically, reset at least three times (each time roughly coinciding with leaving the end). I didn't even need to take the model out of the learner to make it happen.

It's possible for a mod to have more than one bug. The bug report eSonOfAnder linked was for AllTheMods10, BTW. It's probably the same one you're experiencing. It may or may not be related to this one.

That was just one example that I found of reports of the HNN deep learner not working properly when in the curio slot. I've run into people reporting this type of issue in multiple locations. This is why it was suggested (both by the mod dev and myself) to take the learner out of the curio slot to see if the problem is specific to that, or if the problem lies elsewhere. Particularly in large modpacks like the ATM series, it is important to test things one variable at a time. If you are able to test this with the learner not in a curio slot, and the problem still happens then you can eliminate the curio slot implementation as the source of the conflict.

As for your earlier comment about this, your complete dismissal of the report and statement from the dev, what the mod dev actually said is that things worked when eliminating all other variables and bringing them back to just his implementation of the mod, meaning the problem MUST be an interaction with another mod that is being used at the same time. If the problem goes away when the learner is not in the curio slot, that suggests a very strong likelihood that the problem is the curio implementation since it is different than the one that he explicitly develops against. With development on Adorned being abandoned, I would expect both the FTB and ATM teams will be moving to a new curio implementation moving forward anyway. I included that report here, because both FTB Evo and ATM10 use Adorned as their curio implementation, so if there is some interaction issue going on between them, you would most likely see it in both modpacks.

Have you tested with the learner not in the curio slot? Does it work that way?

It's also possible the problem has to do with The End dimension, specifically. Again, it needs to be a meticulous process of eliminating a single variable at a time.

[no-name-provided]

I looked at various reports. They all described much more general errors (either every model losing all progress on tiering up, or every model experiencing limited progress loss). For some, the loss was instant; for others, it only triggered when removing the data model from the curios slot and/or learner. These are probably related, but aren't quite the same. For example, I have a spider model that's been sitting at something like basic +13 kills for real-life weeks, but my dragon model couldn't survive an hour.

Every report seems to be for a large modpack, so, as you said, testing would need to be done to find the root cause. And that's all the dev said at that time, that they didn't know, but it wasn't happening in their dev environment, so it was possibly a cross mod interaction or a dependency mismatch. Their exact words were:

Not sure what would be causing this. It doesn't appear to happen in isolation with Curios API Continuation (the one I develop against). Possible this is a mod interaction issue with something in ATM10 or an issue with Adorned (the fork of curios that ATM uses).

In short: I didn't dismiss Shadows-of-Fire, I just didn't put words in their mouth. It's only recently, after I reached out to them, that they were able to confirm that the/an issue was a dependency mismatch:

I am almost certain this is a curios issue to some degree. It does not happen with Curios API Continuation, but does happen with Adorned and also happens with mainline Curios.

(presumably after testing this in isolation).

Have you tested with the learner not in the curio slot? Does it work that way?

I changed many things, because I was tired of killing the dragon. Not using the curios slot, removing the model as soon as possible, rushing the dragon kills (rather than collecting them over multiple sessions), making even more sure I killed it with a sword (not a bow/magic), plopping the model in a simulation chamber as soon as possible, etc. This combination of things worked... this time. If the other tickets are to be believed, there's an element of random chance. However, given the latest update from the mod dev, I think it's fair to say repeating this test isn't necessary.