Skip face culling in shadows for double-sided materials (e.g. plantlike)

[x2048]

At some sun angles, plantlike nodes produce incomplete or no shadow. This is because they render double-sided (no face culling) but shadow renderer does not respect that and culls front faces.

This PR fixes this problem by only overriding culling for nodes where at least one side is culled in the original material.

To do

This PR is Ready for Review.

How to test

1. Use API to set negative tilt for the sun/moon

minetest.register_on_joinplayer(function(player)
	player:set_sky({body_orbit_tilt = -30})
end)

2. Enter the world and look at the shadows of plantlike nodes. Shadows must be correct at all times.
3. Check shadows of other nodes, they should also be correct.

[lhofhansl]

Seems to work. :)

[lhofhansl]

Actually... This seems to also remove shadows from water.

[Zughy]

@x2048 any updates on this?

[SmallJoker]

Did not notice any difference with plantlike and rooted plantlike.

[grorp]

Two screenshots of the problem described by lhofhansl:

Before this PR:

After this PR:

F5 debug info:

Screenshots taken in Minetest Game, but with the dynamic shadow intensity variation stuff removed. Instead:

player:set_lighting({ shadows = { intensity = 0.85 } })

minetest.conf excerpt:

antialiasing = ssaa
shadow_map_color = true
shadow_map_texture_size = 8192
shadow_map_max_distance = 300
enable_dynamic_shadows = true
shadow_levels = 5
shadow_map_texture_32bit = true
enable_waving_plants = false
shadow_filters = 2
enable_waving_water = false
enable_waving_plants = false
enable_waving_leaves = false
tone_mapping = false
enable_shaders = true
mip_map = true
time_speed = 0

/time 18:00

[SmallJoker]

Cannot reproduce the water shadow issue. I am using the following dynamic shadows settings:

shadow_filters = 1
shadow_soft_radius = 5
shadow_map_texture_32bit = true
shadow_map_texture_size = 1024
enable_dynamic_shadows = true
shadow_map_color = false
shadow_map_max_distance = 100

EDIT: also using tone_mapping = true but turning it off does not make a difference either.

[grorp]

Here's a simpler example:

Before:

The player has two shadows: One on the water surface and a smaller one on the bottom of the pond.

After:

The player has only one shadow, the one on the water surface. The player still has two shadows, you just don't see both because of the dark texture. The bottom of the pond is completely covered by a shadow, as if the water also had a shadow.

EDIT: Actually, maybe I'm wrong... The new behavior (that the water has a shadow as well) seems like it should be the expected one with colored shadows.

/time 12:30

[lhofhansl]

With PR:

Without PR:

Looks different to me.
-1

[sfan5]

@grorp or @lhofhansl if you have a moment please re-test

[lhofhansl]

The problem I was seeing is gone now. Shadows on water are correct now.

[grorp]

The problem I was seeing is fixed as well (it was the same problem).

[lhofhansl]

@x2048 Wanna merge?

[srifqi]

I tested this PR and it fixes the issue. Plant-like nodes cast correct shadows. Shadows on water (liquid) surface and inside water look brighter than master branch.

Test	master branch	This PR
Plant-like shadows
Shadows on water (liquid) surface and inside water

Contents of minetest.conf:

enable_shaders = true
antialiasing = ssaa

shadow_filters = 1
shadow_levels = 5
shadow_map_color = false
shadow_map_max_distance = 140
shadow_map_texture_32bit = true
shadow_map_texture_size = 2048

enable_dynamic_shadows = true

[grorp]

Shadows on water (liquid) surface and inside water look brighter than master branch.

On your second screenshot, all shadows look brighter, whether in the water or outside. Could the brightness difference possibly be caused by MTG's shadow intensity randomization (minetest/minetest_game#3024)?

Just tried again, this PR doesn't change anything related to water shadows for me:

	shadow_map_color = false	shadow_map_color = true
current master branch
this PR, rebased to current master branch

All the screenshots look nearly identical to me.

[grorp]

@srifqi Would you be fine with merging this?

[srifqi]

Would you be fine with merging this?

Yes.

---

Could the brightness difference possibly be caused by MTG's shadow intensity randomization (minetest/minetest_game#3024)?

I guess. The problem disappears in other games.

[lhofhansl]

@x2048 Still around?

[x2048]

@lhofhansl still around, not active.