Regression in rendering textured meshes (imported via Wavefront OBJ files)

[trevorld]

In {rayvertex} v0.4.11 I'm observing a regression in rendering a textured mesh imported via Wavefront OBJ file:

# generate Wavefront OBJ file with {piecepackr}
library("piecepackr")
envir <- game_systems("sans3d", round=TRUE)
f <- save_piece_obj(piece_side = "tile_back", x = 1.5, y = 1.5, suit = 1, rank = 1,
                    cfg = envir$piecepack)

# import/render Wavefront OBJ file in {rayvertex}
library("rayvertex")
tile <- obj_mesh(f$obj)
table <- sphere_mesh(c(0, 0, -1e3), radius=1e3, material = material_list(diffuse="grey40"))
scene <- add_shape(table, tile)
rasterize_scene(scene, filename = "rayvertex_bug.png",
                lookat = c(4.5, 4, 0), lookfrom=c(4.5, -16, 20),
                light_info = directional_light(c(5, -7, 7), intensity = 2.5))

In contrast things look fine with the latest version of {rayrender} (also look fine with {rgl} via {readobj} or an earlier version of {rayvertex} like v0.3.3)

# contrast with {rayrender}
library("rayrender")
tile <- obj_model(f$obj, load_textures = TRUE, load_material = TRUE)
table <- sphere(z=-1e3, radius=1e3, material=diffuse(color="grey40")) |>
         add_object(sphere(x=5,y=-4, z=30, material=light(intensity=420)))
scene <- add_object(table, tile)
render_scene(scene, filename = "rayrender.png",
             lookat = c(5, 5, 0), lookfrom = c(5, -7, 25),
             width = 500, height = 500, samples=200, clamp_value=8)

[tylermorganwall]

THere's a couple unrelated bugs in the scene construction methods that I've mostly chased down, but I'm still working on why this regression appeared.

[trevorld]

git bisect indicates the regression was introduced in https://github.com/tylermorganwall/rayvertex/commit/cc263f7eb59fdc900dcdf20f816df6431763ab36

I wildly speculate that maybe there was an upstream regression in tinyobjloader? Assuming {rayrender} is using an older version of tinyobjloader that would explain why it still works fine.

[tylermorganwall]

Should be fixed in the latest commit. This issue simply stemming from not including the DEFINE that turned on the new (fixed) tinyobjloader earcuting algorithm in the last update. All works fine on my end now--check it out on yours.

[trevorld]

Looks good on my end!