Time  Nick           Message
11:25 sfan5          today in "why you do not reinvent basics": irrlicht's core::string::append(const T*, u32) forgets to null-terminate
11:25 sfan5          I only found out because I just wrote unit tests for it which apparently nobody has ever done before
12:02 MTDiscord      <luatic> it's a feature
14:44 schwarzwald[m] sfan5: Are you planning to merge those? I added a test suite with my glTF loader as well.
15:09 sfan5          sure
15:12 MTDiscord      <exe_virus> So for gtlf, we will add a new file type support for media, will servers have to reject older clients by default?
15:12 MTDiscord      <exe_virus> Should it be treated as a network change, i.e. major version number change?
15:13 sfan5          new features are not an incompatibility
15:14 MTDiscord      <exe_virus> Gotcha, so in those cases the outdated client does what? Reject the media?
15:20 sfan5          probably
15:50 MTDiscord      <exe_virus> I'll take a look, wanted to make a .jxl loader and need to know how we handle the mismatch
17:00 schwarzwald[m] sfan5: I'm using Catch2 and I put my test file in source/Irrlicht/tests. Moving them later should be easy. The part I'm wondering about is whether I should use whatever test framework (homemade maybe) for consistency.
17:05 kilbith        https://www.youtube.com/watch?v=W7JPVcauq9c
17:05 kilbith        pathfinder's sucks but it's just for the demo
21:31 schwarzwald[m] I see the irrString tests. Would it be alright to get rid of Irrlicht's test runner and use something else instead? I would even prefer a runner based on Rubenwardy's to Irrlicht's. doctest is still my ideal test runner, and Catch2 is my top pick because Minetest already requires it.
22:36 sfan5          there's no "irrlicht's test runner", it's just some quick code so we can test things we want to
22:36 sfan5          irrlicht's current tests will either be thrown away or merged with what MT has once it's imported into the repo
22:38 schwarzwald[m] On the topic of glTF models: I'm moving on to unit testing a model exported by Blender, and each of the 5 vertices is listed at least 3 times in the vertex buffer. Why?
22:39 schwarzwald[m] It has 5 vertices, but 16 vertexes in the vertex buffer.
22:40 schwarzwald[m] Naturally I'm thinking someone broke something, but it looks like the exporter is official, so maybe this is some computer graphics thing I don't know about?
22:41 schwarzwald[m] 5 vertices in blender but 16 vertices in the glTF vertex buffer*
23:08 Desour         schwarzwald[m]: vertices don't just have positions, they also have uv coordinates, normals and other attributes. when rendering, you therefore have to use multiple different vertices at the same position
23:09 schwarzwald[m] Ah, interesting!
23:09 Desour         collada has an option to keep it one vertex, idk if gltf supports it
23:10 Desour         which is also why gltf is probably not as well-suited for storing models
23:10 schwarzwald[m] It's ok, I'm just annoyed that I have to unit test the positions of 24 vertices just for a simple cube. xD
23:11 schwarzwald[m] I can copy/paste most of them because they're duplicates.
23:12 Desour         btw. for old client support, you could try doing mesh import serverside, and export it as b3d
23:13 Desour         (if it's not too much work)
23:14 schwarzwald[m] lol
23:41 MTDiscord      <GreenXenith> Gltf ought to be plenty well suited for storing models, Desour. In fact it is designed to be a better format for games in particular. And I would much rather deal with binary and json than xml
23:47 schwarzwald[m] > I would much rather
23:47 schwarzwald[m] You? Maybe you better start helping then, hehe.
23:57 MTDiscord      <GreenXenith> That requires time and energy 😩