[MC-197883] Jigsaw Structures causing DataFixer spam

The bug

Similar to MC-191388 but a new issue (this is a not a duplicate or a "wasn't fixed correctly" - this is a new issue)

Jigsaw World Gen Features is triggering codec errors like the following:

[19:34:04] [Server thread/ERROR]: Input does not contain a key [type]: MapLike[\{name:"minecraft:no_op",config:{}}]
 [19:34:04] [Server thread/ERROR]: Input does not contain a key [type]: MapLike[{name:"minecraft:block_pile",config:{stat
 e_provider:{state:{Properties:{axis:"y"},Name:"minecraft:hay_block"},type:"minecraft:simple_state_provider"}}}]
 [19:34:04] [Server thread/ERROR]: Input does not contain a key [type]: MapLike[\{name:"minecraft:no_op",config:{}}]
 [19:34:04] [Server thread/ERROR]: Input does not contain a key [type]: MapLike[{name:"minecraft:block_pile",config:{stat
 e_provider:{state:{Properties:{axis:"y"},Name:"minecraft:hay_block"},type:"minecraft:simple_state_provider"}}}]
 [19:34:04] [Server thread/ERROR]: Input does not contain a key [type]: MapLike[\{name:"minecraft:no_op",config:{}}]
 [19:34:04] [Server thread/ERROR]: Input does not contain a key [type]: MapLike[{name:"minecraft:flower",config:{tries:64
 ,yspread:3,xspread:7,need_water:0b,zspread:7,blacklist:[L;],project:1b,block_placer:{type:"minecraft:simple_block_placer"},state_provider:{type:"minecraft:plain_flower_provider"},whitelist:[L;],can_replace:0b}}]
 [19:34:04] [Server thread/ERROR]: Input does not contain a key [type]: MapLike[\{name:"minecraft:no_op",config:{}}]

Using Spigot/Paper mappings, the callstack of where this is triggered is

https://gist.github.com/aikar/56b0c2beff6d8c70e18893a72889eb76

Code triggering it is this:

this.b = new BlockPosition(nbttagcompound.getInt("PosX"), nbttagcompound.getInt("PosY"), nbttagcompound.getInt("PosZ"));
this.e = nbttagcompound.getInt("ground_level_delta");
DataResult dataresult = WorldGenFeatureDefinedStructurePoolStructure.e.parse(DynamicOpsNBT.a, nbttagcompound.getCompound("pool_element"));

Unsure about data corruption, but can be spammy.

Reproduction steps

Create a world in 1.15.2 and teleport to a village; or use the attached world

[media]

Open the world in the latest version
❌ Multiple errors are logged

Root Cause

The Root Cause is that there are items in the FEATURE registry that are specifying "name" as their identifier instead of "type", and these are mixed into the same collection.

This code in Spigot mappings is the source point:

public class WorldGenFeatureConfigured<FC extends WorldGenFeatureConfiguration, F extends WorldGenerator<FC>> implements IDecoratable<WorldGenFeatureConfigured<?, ?>> {
    public static final Codec<WorldGenFeatureConfigured<?, ?>> a = IRegistry.FEATURE.dispatch((worldgenfeatureconfigured) -> {
        return worldgenfeatureconfigured.e;
    }, WorldGenerator::a);
    public static final Codec<Supplier<WorldGenFeatureConfigured<?, ?>>> b = RegistryFileCodec.a(IRegistry.au, WorldGenFeatureConfigured.a);
    public static final Codec<List<Supplier<WorldGenFeatureConfigured<?, ?>>>> c = RegistryFileCodec.b(IRegistry.au, WorldGenFeatureConfigured.a);
    public static final Logger LOGGER = LogManager.getLogger();

Changing this to .dispatch("name", (....

Fixes the errors for the ones listed in the ticket, however then breaks every thing using type.

The data sets need to be updated to consistently use type.

Linked issues

is duplicated by 1

MC-199174 Input does not contain a key [type] Resolved

relates to 1

MC-191388 "No key position_predicate in MapLike" console spam Resolved

Attachments

MC-197883.zip

Comments 24

James Elliott 2020-08-13T08:20:18Z

Edit: Never mind, this doesn't really add anything, now that I understand the jigsaw block it makes sense that it would be having this issue.

Looks like this may also affect villages in general in upgrades from 1.15.2 -> 1.16.2. Unknown the specific circumstances as to when though.

I've found villages which seem to generate the issue, deleting the chunks associated which don't seem to have pillager outposts seems to resolve it. The specific village was in the savanna I think.

marcono1234 2020-08-13T13:20:57Z

Could you please attach a world which is causing this issue (and mention steps for causing it)?
Or can you describe reproduction steps for creating a world which, when opened in the latest version, causes this issue?

Aikar 2020-08-13T14:11:12Z

Attached a screenshot from a user, appears this is still the same damage as MC-191388

To reproduce, creating a world in 1.15 and finding a village, and I'm not sure if partial generation applies or not, but if it does, stop moving soon as part of a village is seen, and then open that world in 1.16.2

I will try to attach some files I have reproducing this every time after work too, just need to try to filter it to just needed files first and get coords.

I have one report that it was related to partial generation, but in my own test world I get it at my "Home" location which is a village and i know it was fully generated.

Another suspect is if a village doesn't have a pillar outpost, and my village does not... (likely created in 1.12).

Hope this helps.

James Elliott 2020-08-13T14:38:04Z

The issue occurs with fully generated villages too in my testing. The screenshot is due to partial generation, if you regenerate a village that was generated pre 1.16 in a 1.16 world it will look different.

James Elliott 2020-08-13T14:59:57Z

World generated on 1.15.2, upgraded to 1.16.2/when upgraded to 1.16.2 that has the issue: https://drive.google.com/file/d/130h8UJWViF-YL_UKjszo1RJlsJdabzMc/view?usp=sharing

Seed: 1234

Coords: -415 70 114

14 more comments

Michael 2020-08-28T23:21:02Z

Mod turn down my issue, and say it same as this.

MC-199174

Either way @@unknown , yes see my other issue I start. I get Input does not contain a key [type] on vanillia.

Aikar 2020-08-29T02:36:32Z

@Broken-arrow because the logs you pasted in that ticket is in fact a duplicate of this one.

The logs you pasted on Paper's issue tracker looks to be a different issue. the key is "position_predicate" vs "type"

I linked the logs in my last reply

@Marcono, I can say with confidence my fix can not cause a regression, as it only deals with type/name. I question if Michaels issue is a similar issue but different data set with same problem though.

I cant create a ticket for that yet as michael has not provided me enough information to reproduce it yet.

Michael 2020-09-06T09:34:18Z

I have put in lots of afford in this and all thing I can see it is this errors fault. Aikars fix alter the error and I guess it halfway fix the problem.

this error get altered in paper:
[Server thread/ERROR]: Input does not contain a key [type]: MapLike[\
All problem start with 1.15 with hay_pile errors and if electronicboy correct it is because you remove world gen features.

So i guess it better run paper on old world, an use vanillia jar server.

Joshua Nixon 2020-10-15T15:17:26Z

Issue affecting 1.16.3 as well. Every single new village I come across in unexplored map areas triggers a plethora of these errors and ends up screwing up village generation. Instead of playing, I spend half an hour or more manually filling in missing grass/sand/gravel/water blocks under and around houses and other structures just to make the area look normal again.

Aikar 2020-10-28T01:03:28Z

@Pcobald can you remove that comment? It's spam and does not add anything of value to the conversation.

Unless Mojang asks for more information, no more needs to be supplied for this issue.

So please avoid replying with more stuff until Mojang asks, though last I talked with Mojang on this issue, they seem to understand what the issue is.

This is unlikely to be fixed before 1.17.