For the last year or so, there have been linux-aarch64 binaries at the ziglang download page of the master build. I depend heavily on this, and e.g., recently reported a serious bug I found by testing building Python with this (github issue 13303). Suddenly there are no more linux-aarch64 binaries. Why is this? Is the problem, lack of build support? Is this just a temporarily blip and aarch64 binaries will be back soon? I can offer hardware access if that could help.
I’ve been trying hard to not have to build zig from source all the time, since it takes a while, and despite several attempts I’ve never been able to successfully build the latest version of zig from source (which is weird, since I’m really good at building things from source). Obviously, with sufficient effort I can figure this out.
In any case, I’m just curious what the situation is with binary support.
(I apologize that it’s hard to follow this email since I had to rewrite all the links with “hints” due to your anti-link filter.)
I think the problem is that the CI service that produced those builds is having issues. Ideally we should be able to solve it soon. I’ll personally look into it as soon as I’m done with the event stuff I’m working on right now, as I was interested in trying to move everything to GH Actions anyway.
Can’t promise immediate results though, so I would recommend you take this as an opportunity to try again to build Zig yourself. Note that now you don’t even need to go through a C compiler first. Just use the Zig compiler that you already have and use it to build a new version of the self-hosted compiler.
You will still need a build of LLVM though.
Regarding the spam filtering, I added ziglang.org and github.com to the whitelist so new users should be able to post all the links they need to those hosts without issue.
Thanks for the feedback and updates. I hope switching to GH Actions works for this. There’s some discussion here: weak memcmp symbol from compiler_rt is not overridden when building Python 3.11.0 on Linux and MacOS · Issue #13303 · ziglang/zig · GitHub
Yes, it’s the LLVM part of the build from source that is the most challenging. I’ll figure this out eventually, with another goal being a wasm build of zig with cc/c++ support. I’m already extremely impressed that a build of zig for wasm without cc/c++ support is so easy to create out of the box now. and I’ve had no trouble doing that.