libm VS rust

Yes I took a look at build.rs. There's nothing here that stands out to me. Building the project on my Ubuntu machine works btw.

The more interesting code is the "kernel" tangent function https://github.com/rust-lang/libm/blob/master/src/math/k_tan.rs . I don't see any numerical analysis of accuracy in either the Rust lib or the MUSL source lib, but from one of the comments it looks like it's intended to be within 1ULP, which is what I'd expect.

It's nowhere near complete, but: https://github.com/rust-lang/libm

Inaccurate fast implementations use polynomial approximation. Accurate implementations use high-powered argument reduction and then polynomial approximation. Everything is polynomial approximation.

It looks like https://github.com/rust-lang/libm/pull/249 may be a fix for this.

In order to be standards-compliant, computer math libraries need to have a lot more digits of pi than should be strictly necessary. Accurately calculating sine in the upper reaches of double precision floats requires computer to have a table of about 100 digits of pi (or rather pi/2 for technically reasons) and 500 digits of 2/pi - here is an example.

So, fixed-point numbers weren't the solution to the problem either. The next thing I came across was the Rapier physics engine. It promises cross-platform determinism all IEEE 754-2008 compliant platforms. Although it's written in Rust, creating C# bindings was fairly easy, and I could use it in Unity after a few hours. But Rapier is still in early stages of development, which means that (at the time of writing this) it lacked many features, like raycasting, which would have been necessary for my project. Also, it's only deterministic on IEEE 754-2008 compliant platforms. But I couldn't really find any information on which processors support it. Even if I did, I still wanted to have deterministic physics on all platforms, not just on "most of them". But I still wondered does how Rapier achieve cross-platform floating point determinism. Supporting only IEEE 754-2008 compliant platforms is one thing, but the mathematical functions I mentioned above (sqrt, trigonometry, etc.) must also be deterministic. I found out that it uses libm, which implements all of these functions in software. It was primarily designed for embedded systems without FPUs, but it turns out it's also pretty useful for making this part of floating point math deterministic. But after all, the basic mathematical operations of floats are still not deterministic. Unless... How about emulating the behavior of floats in software? It should be possible to write a program that works on the byte representation of a float, doing all operations using integer arithmentic... right?

If you haven't dipped your touch-typing fingers into Rust yet, you really owe it to yourself. Rust is a modern programming language with features that make it suitable not only for systems programming -- its original purpose, but just about any other environment, too; there are frameworks that let your build web services, web applications including user interfaces, software for embedded devices, machine learning solutions, and of course, command-line tools. Since a custom GitHub Action is essentially a command-line tool that interacts with the system through files and environment variables, Rust is perfectly suited for that as well.

Here's an example of someone citing a disagreement between CRT and shell32:

https://github.com/rust-lang/rust/issues/44650

This in addition to the Rust CVE mentioned elsewhere in the thread which was rooted in this issue:

https://blog.rust-lang.org/2024/04/09/cve-2024-24576.html

Here are some quick programs to test contrasting approaches. I don't have examples of inputs where they parse differently on hand right now, but I know they exist. This was also a problem that was frequently discussed internally when I worked at MSFT.

    #include 

> instead of choosing a certain numbered version of the random library (if I remember correctly) I let cargo download the latest version which had a completely different API.

Yeah, they didn't follow the instructions and got burned. I still think that multiple things went wrong simultaneously for that experience. I wonder if more prevalent uses of `#[doc(alias = "name")]` being leveraged by https://github.com/rust-lang/rust/pull/120730 (which now that I check only accounts for methods and not functions, I should get on that!) so that when changing APIs around people at least get a slightly better experience.

Almost fixed the compiler: https://github.com/rust-lang/rust/pull/123325

Rust: A secure, efficient, and modern programming language (omitting ten thousand words). You can simply follow the installation instructions provided on the official website.

Recently did something similar in Rust but for generating SVGs. We've adopted it for snapshot testing of cargo and rustc's output. Don't have a good PR handy for showing Github's rendering of changes in the SVG (text, side-by-side, swiping) but https://github.com/rust-lang/rust/pull/121877/files has newly added SVGs.

To see what is supported, see the screenshot in the docs: https://docs.rs/anstyle-svg/latest/anstyle_svg/

We strongly believe in Rust as a powerful language for building production-grade software, especially for systems like ours that run alongside Kubernetes.

The above Assert<{N % 2 == 1}> requires #![feature(generic_const_exprs)] and the nightly toolchain. See https://github.com/rust-lang/rust/issues/76560 for more info.