TonemappingTest.gltf is rendered with incorrect texture filtering

[rparrett]

Bevy version

main, since #17875

Relevant system information

SystemInfo { os: "macOS 15.3.2 Sequoia", kernel: "24.3.0", cpu: "Apple M4 Max", core_count: "16", memory: "64.0 GiB" }
AdapterInfo { name: "Apple M4 Max", vendor: 0, device: 0, device_type: IntegratedGpu, driver: "", driver_info: "", backend: Metal }

What you did

cargo run --example color_grading

What went wrong

Debug texture on cubes uses linear filtering, but should be nearest.

Poking around in the glTF, it seems to be set to nearest as far as I can tell, so there seems to be some sort of issue with the glTF loader. Though perhaps it's just a problem with this particular glTF.

Additional information

Caught by pixel eagle.

https://pixel-eagle.com/project/b25a040a-a980-4602-b90c-d480ab84076d/run/10231/compare/10224?screenshot=3D%20Rendering/color_grading.png

[rparrett]

cc @axlitEels @thepackett

[greeble-dev]

I think this condition might be wrong? I can make a PR if I get a confirmation.

 /// Extracts the texture sampler data from the glTF [`Texture`].
 pub(crate) fn texture_sampler(
     texture: &Texture<'_>,
     default_sampler: &ImageSamplerDescriptor,
 ) -> ImageSamplerDescriptor {
     let gltf_sampler = texture.sampler();
     let mut sampler = default_sampler.clone();
 
     sampler.address_mode_u = address_mode(&gltf_sampler.wrap_s());
     sampler.address_mode_v = address_mode(&gltf_sampler.wrap_t());
 
     // Shouldn't parse filters when anisotropic filtering is on, because trilinear is then required by wgpu.
     // We also trust user to have provided a valid sampler.
-    if sampler.anisotropy_clamp != 1 {
+    if sampler.anisotropy_clamp == 1 {

[axlitEels]

@greeble-dev I'm afraid you're right. Pixel Eagle didn't catch that on PR validation because color_grading just isn't there.

[axlitEels]

Just tested it. flipping the sign does fix the issue. Now I'm concerned if we made some false assumptions for #18267.

[greeble-dev]

PR: #19118