Ongen sapling trees get cut off at chunk borders

[livphi]

Hi, I was playing around in my custom mapgen and noticed some of the moretrees trees were cut off. I thought it was a problem with my mapgen, so I went into v7 and the trees were still getting cut off. I see the trees grow completely intact, then they get cut off at the chunk edge when the chunk finishes loading. Sometimes the trees grow, then they get overwritten and become ongen saplings again that just awkwardly sit there. I assume this has something to do with the biome refactoring of 5.12.0. I haven't noticed this happening with any other trees, just moretrees.

I'm not sure exactly why this is happening? V7 mapgen checks if the content id of the node == IGNORE, if it doesn't, then skip this node and dont place anything. So logically it should ignore the already placed leaves because their content id is not equal to IGNORE. My mapgen does the same thing. But it appears to overwrite the tree nodes with air.

Luanti 5.12.0
Only using moretrees biome mod
Happens on v7 and probably all the other ones (i havent tested)

[wsor4035]

i would be mildly interested to see if #51 solves your issue for trees handled in that pr (DONT RUN IT IN PRODUCTION) - but also not at all surprised if it didnt help at all

[SwissalpS]

Thanks for opening an issue on this.

This has nothing to do with 5.12. it's a known fact for years.

[livphi]

Here's another example.

[livphi]

i would be mildly interested to see if #51 solves your issue for trees handled in that pr (DONT RUN IT IN PRODUCTION) - but also not at all surprised if it didnt help at all

I checked, unless I'm misunderstanding something, sequoia trees already use an L-system definition. And they still get cut off. Is this a problem with the engine? Does this happen with every mod's l-system large trees?

[livphi]

Right now I have a fix that fixes it most of the time, but they still get cut off sometimes. It checks if the 3x5x3 mapblock column surrounding the sapling is active, if it is, then it can grow. I have no idea why its still getting cut off after this "fix", but it appears to happen a lot less.

[livphi]

i would be mildly interested to see if #51 solves your issue for trees handled in that pr (DONT RUN IT IN PRODUCTION) - but also not at all surprised if it didnt help at all

A snippet from your code actually does fix the problem. But we still have to figure out how to deal with birch trees and the non-lsystem trees.

	if minetest.features.lsystem_decoration_type then
		deco_def.deco_type = "lsystem"
		deco_def.treedef = moretrees[treename .. "_model"]
	else
		deco_def.deco_type = "simple"
		deco_def.decoration = "moretrees:"..treename.."_sapling_ongen"
	end

Placing the lsystem tree directly instead of a sapling that grows into an lsystem tree fixes this bug.

I'll see if I can write a general, non-invasive fix that can be used in production.

[SwissalpS]

I thought this had to do with chunks not being generated in time as I usually see cut of trees in y direction. I mostly dig either straight up or down. (mars caves have trees in the game I play)