Nix standalone on MacOS's system string is `arm64-apple-darwin` and not `aarch64-darwin`

[doggobit0]

Describe the bug

This is sort-of related to nixpkgs itself, but there's no issue template for non-package-related isssues there.

For MacOS, there seem to be twothree different system strings at the moment: aarch64-darwin, aarch64-apple-darwin and arm64-apple-darwin. I can't find any documentation describing how they differ. As a result, most packages are configured with options for aarch64-darwin and not arm64-apple-darwin. For example: nix-community/home-manager#6878

My system is showing as arm64-apple-darwin when configured as a standalone Nix installation over MacOS (not a nix-darwin setup). I don't have access to an unmanaged mac, thus I can't test a nix-darwin setup sadly.

$ nix eval nixpkgs#legacyPackages.aarch64-darwin.stdenv.hostPlatform.config
"arm64-apple-darwin"

However, when packages specify supported platforms, the system string seems to be aarch64-apple-darwin, so could this be an issue with the package derivation logic itself? (logs from linked issue):

error: Package ‘firefox-137.0.1’ in /nix/store/hzaj4d6ari2wq2cbg1j60n9zw42gnshy-source/pkgs/applications/networking/browsers/firefox/wrapper.nix:504 is not available on the requested hostPlatform:
         hostPlatform.config = "aarch64-apple-darwin"
         package.meta.platforms = [
           "i686-cygwin"
           "x86_64-cygwin"
         ...
         ]

This would be super weird, because it would mean that the hostPlatform.config has different values in make-derivation-nix vs check-meta-nix, as the invocation simply inherits the hostPlatform? Alternatively, there could be some logic that overrides this value that only shows up when doing a build versus evaluating the output?

The main two questions are:

1. Why are they different?
2. Should they be different? If not, how do we fix it?

EDIT: This is the whole extent to which I know how to debug this on my own, I would highly appreciate some guidance forward, as it's highly likely this impacts other users as well (a quick search shows up multiple aarch64-apple-darwin results). I am happy to pour some time into this, but unpicking the different places this value is set (and where it's derived from in the first place) is very elusive to me atm.

Steps To Reproduce

1. Install Nix multi-user as a standalone installation on MacOS

Expected behavior

If the system strings should be different, then have documentation explaining how and how to handle them (and the differences). If not, realign the system strings.

Metadata

nix-env (Nix) 2.26.3

Additional context

Checklist

• checked latest Nix manual (source)
• checked open bug issues and pull requests for possible duplicates

---

Add 👍 to issues you find important.

[RossComputerGuy]

This isn't an error within nix itself, we switched things in nixpkgs on purpose due to Apple using arm64 instead of aarch64.

[grahamc]

Huh! Can you reference issues/prs/rfcs/etc for that change?

…

On Tue, Apr 22, 2025, at 6:55 PM, Tristan Ross wrote: This isn't an error within nix itself, we switched things in nixpkgs on purpose due to Apple using `arm64` instead of `aarch64`. — Reply to this email directly, view it on GitHub <#401364>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAASXLHLZSBLOHHFM44WBBD223CHPAVCNFSM6AAAAAB3T7GUZ6VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDQMRSGY2DKNRVG4>. You are receiving this because you are subscribed to this thread.Message ID: ***@***.***> *RossComputerGuy* left a comment (NixOS/nixpkgs#401364) <#401364> This isn't an error within nix itself, we switched things in nixpkgs on purpose due to Apple using `arm64` instead of `aarch64`. — Reply to this email directly, view it on GitHub <#401364>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAASXLHLZSBLOHHFM44WBBD223CHPAVCNFSM6AAAAAB3T7GUZ6VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDQMRSGY2DKNRVG4>. You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[RossComputerGuy]

#393213

[grahamc]

Ah, cool ok! I misread as thought aarch64-Darwin was renamed, which had me a bit Stressed. Vs the internal triple used. Cool.

…

On Tue, Apr 22, 2025, at 7:24 PM, Tristan Ross wrote: NixOS/nixpkgs#393213 <#393213> — Reply to this email directly, view it on GitHub <#401364>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAASXLCONFFFQGPBBDAKAO3223FTZAVCNFSM6AAAAAB3T7GUZ6VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDQMRSGY4TINBSHA>. You are receiving this because you commented.Message ID: ***@***.***> *RossComputerGuy* left a comment (NixOS/nixpkgs#401364) <#401364> NixOS/nixpkgs#393213 <#393213> — Reply to this email directly, view it on GitHub <#401364>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAASXLCONFFFQGPBBDAKAO3223FTZAVCNFSM6AAAAAB3T7GUZ6VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDQMRSGY4TINBSHA>. You are receiving this because you commented.Message ID: ***@***.***>

[Mic92]

Transferred because nixpkgs does that.

[booxter]

@NixOS/darwin-core does it mean every package now has to list a bunch of platform flavors whenever they'd like to target darwin? (If so, should this be handled in a centralized way, somehow - e.g. by populating meta.platforms with all flavors?)

[Mic92]

meta.platforms still takes "aarch64-darwin" as returned by stdenv.hostPlatform.system

[reckenrode]

To provide a bit more context, Darwin will be updating to LLVM 20 once staging opens up for post-25.05 changes. Having this change now lets us shake out any problems ahead of that change.