build_dot_zig - Use the Zig build system from Elixir to build your NIFs

I would love to give more hearts to this than one. I had been hoping this would become a thing for some time. I might even have a usecase for it at work. Does this work with nerves and their targets?

Once again, the discoverability on libraries is horrible .

We created a similar library for this year’s spawnfest: GitHub - spawnfest/zig_builder · GitHub

Alas, in my opinion the zig build system is still too unreliable to be used in any serious projects, we had a lot of trouble dealing with internal errors from the build system and the lack of documentation didn’t help.

Also they announced they are moving away from integrating well with C compiler(s) and will be doing their own thing, so to speak, which I think alienated a lot of people because they relied on Zig’s toolchain to help them gradually migrate their C projects to Zig, while allowing both to coexist peacefully and seamlessly in the same project. Wrong, see below.

I am not hating on the language – I watched a few demos and I think it’s really simple and to the point and gets things done – but in light of several such recent news I question its fit and place, especially in a world where Rust exists.

Though Rust’s stdlib is quite big so I heard that many people try to code for micro-controllers with Zig (successfully).

So it’s likely not going anywhere but I wonder how widespread it could get.

As far as I remember that was a big misunderstanding.

Oh? Would be glad if I am wrong on this. It was one of Zig’s big draws.

As someone not experienced in systems level languages I found zig quite a bit more approachable than rust. I also really like that they actually want to make things like cross compilation simpler. I once tried to cross compile a rust project we have at work and just gave up after a bit, because I couldn’t figure out what was going wrong. Doing it with a toolchain in the handful of MB makes zig even more appealing in that regard.

Same here, and that would be my reason to go use Zig and not Rust – if I had the time or was paid to do so, that is.

Problem is, Rust offers much more guarantees out of the box, and that’s crucially important in systems programming. But really, I am no expert in Zig and I make no claims about if it’s more useful than Rust in these contexts.

Not really, I never heard about that. Their zig build system however doesn’t plan to be compatible with make, because that eventually breaks cross-compilation, as there are usually flags for targeting a platform.

Indeed, however this is done by constraining the ways you can write code. There are situations, for example MCUs firmware, where you end up circumventing those safety guarantees as they might not make sense. Then you will end up with rust code using unsafe blocks all over the place, hence using zig is more sane in such a situation.

There is also the benefit of easy refactor of c code, just replace c compiler with zig and you end up with a codebase that can be gradually refactored to zig at virtually zero cost.

Though not sure what happened to that. I heard on HN some months ago that this will be… I don’t know what. Not a first class citizen though. But I should stop talking because I don’t keep current there. So you would likely know better.

This was about a proposal for zig to gain more indepenance from third parties it currently depends on. I’ll link to the comment, which should ultimately give some perspective to the initial proposal: make the main zig executable no longer depend on LLVM, LLD, and Clang libraries · Issue #16270 · ziglang/zig · GitHub