@kornel HTML, CSS and JS are prime exhibits of browser-specific behavior, we've been through years of pain. Compatibility issues even contributed to the death of IE. Maybe it's just me but historically, binary formats seem to be less error prone. WGSL as a language is pretty neat though.

@kornel @xoofx Text is not the issue, but vendor-specific parsers are. GLSL is a prime example, it has to be parsed by the driver. In the demoscene, since most people have nVidia hw, hardly any OpenGL demos work on AMD lately. I'm afraid we'll see the same issue with Tint (Chrome) and Naga (Firefox).

Trying #leptos for frontend dev these days. Having static types on FE is great, elminates most of the guesswork. I finally know which fields & methods these objects have without trial & error.

I just published a new blog post about my experience working on developer tools and large-scale codebase migrations at Google. I hope some will find it interesting.

https://laurent.le-brun.eu/blog/evolving-a-codebase-at-google-scale

@llb Great article! Finally had the time to read it. Brings back a few memories, too. :)

First render after rewriting my toy engine from Vulkan to #WebGPU. And I'm awestruck, again.

It's upside down, GUI is missing, shadow map is buggy. But it's not a black screen. In fact, most things work well on the first render.

When I wrote 3D in C++, black screens and crashes were quite common. C++ tells you how you _can_ use an API, but #rust tells you how you _should_. And the validation checks of both #wgpu and #vulkano are awesome.

#windsurf made me a "2024 wrapped" page to make me feel good. Super thoughtful, even though I only started using it yesterday. Why not "top 100%" though?

I clicked on the "update docker" button and it uninstalled itself. I admire the honesty.

My displacement shader just applies normal mapping on the original surface normals as if no displacement was happening.

I was worried about that when I wrote it. Turns out, it's still visually acceptable.

Adjusting the gradient would be a bit more involved since it needs need some info about mesh topology, but the displacement is implemented in the vertex shader where there's no such info by default. I'll need to find a solution.

@ekuber So good. :) Combined with #[non_exhaustive], the ".." syntax is a now very convenient way to maintain API compatibility. Typing "..Default::default()" was driving me nuts. :)

The Microsoft Weather app on Windows. WTH is going on? Why so many renderer processes?

@chainq To be fair, some of the Rust libraries have nice reports when there's a missing library telling you what you need to install. I don't know how that works in FPC.

"Default fields in structs" RFC is in "final comment period", meaning the feature has been accepted. Thanks to t-lang for going over their concerns and discussing them so quickly!

Soon, you'll be able to write in nightly

#[derive(Default)]
struct Foo {
bar: i32 = 42, // `Foo::default()` will give you a `Foo { bar: 42 }`
}

and

struct Foo {
mandatory: i32,
optional: i32 = 42,
}

// you don't need to specify optional
let foo = Foo { mandatory: 0, .. };

#RustLang #Rust

@ekuber That simple `..` syntax, oh my, so cool, so much cleaner than before. Love it.

the biggest problem in computer graphics programming is that the word 'height' is one character longer than the word 'width'

@lumecolca I just called the police on you

We made a demo!
Released at qbparty, using #rust + #vulkan, fully open source.

https://www.youtube.com/watch?v=QJ0BTwXt8sA

@ekuber I'm so happy this is happening. <3 Feel free to go overboard, these conflicts are rare and not trivial.

Remember to load normal maps in linear space! Unlike albedo maps or most regular images, they are not sRGB. PBR materials in PNG format often have no clue how to interpret channel data in an image file. Here's a comparison between linear interpretation (correct) and sRGB (incorrect).

Nice bug: I probably messed up HDRi envmap image loading.