Make release branches a thing #1259

emilazy · 2025-01-08T23:00:52Z

Finally!

Once this is approved, I will cut a nix-darwin-24.11 branch and PR the appropriate changes.

Left as future work:

Building a seperate manual for each release, moving content from README.md and CHANGELOG to the manual, and linking an index page with the manuals for each version from README.md.
Infrastructure for automated backports.
Writing a checklist for the release process (or better, automating it).
Perhaps having a bot automatically update our flake.lock so we can just tell people to use nix-darwin’s Nixpkgs input rather than bringing their own (and risking mismatches).

Closes: #727

Samasaur1 · 2025-01-08T23:50:15Z

modules/system/version.nix

+    system.darwinRelease = mkOption {
+      type = types.str;
+      default = (lib.importJSON ../../version.json).release;
+      description = "The nix-darwin release (e.g. `24.11`).";
+    };


This should be read-only, right?

Is this equivalent to a NixOS option?

Yes, system.nixos.release, which is read only (so I’ve fixed that). The inconsistent naming here is unfortunate, but I thought it better to stick with the existing convention throughout the file and address all of them at once at some later date. (Though I do admit to feeling a little bad about introducing a new option with a name I don’t love.)

It would be good to just add a comment to document the equivalent option

This is just another entry in the various version options (system.darwinLabel, system.darwinVersion, system.darwinVersionSuffix, etc.) that all mirror various NixOS options. Do you want comments on all of those too? Maybe it should be in a different PR?

Sure a different PR is fine

bew · 2025-01-09T10:39:23Z

modules/examples/flake/flake.nix

@@ -3,7 +3,7 @@

  inputs = {
    nixpkgs.url = "github:NixOS/nixpkgs/nixpkgs-unstable";
-    nix-darwin.url = "github:LnL7/nix-darwin";
+    nix-darwin.url = "github:LnL7/nix-darwin/master";


Hello,
Why this change? Isn't it already the default 🤔

Yeah. The eval-config.nix error tells people to ensure their input looks like github:LnL7/nix-darwin/NIX-DARWIN-BRANCH, and it felt simpler to just align the examples with that. I don’t love how it looks but I couldn’t think of a way to account for it in the eval-config.nix error that wouldn’t confuse people or be even more verbose. But I’m open to alternative suggestions.

My ideal would be to rename our master branch to nix-darwin-unstable to match nixpkgs-unstable, but that requires repo admin privileges so it’s not easy to do immediately.

I do like that it's more explicit, not sure how I feel about changing our default branch to nix-darwin-unstable as I don't think we really need the master and unstable split like nixpkgs

Enzime · 2025-01-10T00:38:01Z

modules/system/version.nix

+    system.darwinRelease = mkOption {
+      type = types.str;
+      default = (lib.importJSON ../../version.json).release;
+      description = "The nix-darwin release (e.g. `24.11`).";
+    };


Is this equivalent to a NixOS option?

Enzime · 2025-01-10T02:54:52Z

modules/examples/flake/flake.nix

@@ -3,7 +3,7 @@

  inputs = {
    nixpkgs.url = "github:NixOS/nixpkgs/nixpkgs-unstable";
-    nix-darwin.url = "github:LnL7/nix-darwin";
+    nix-darwin.url = "github:LnL7/nix-darwin/master";


I do like that it's more explicit, not sure how I feel about changing our default branch to nix-darwin-unstable as I don't think we really need the master and unstable split like nixpkgs

eval-config.nix

Enzime · 2025-01-10T03:12:41Z

eval-config.nix

+  If `darwin` or `nixpkgs` are present in `nix-channel --list` (without
+  `sudo`), you should delete them with `nix-channel --remove NAME`.
+  These can contribute to version mismatch problems.


I think this should live in its own check

As in, add an additional check to make sure the channels are all in order? I think we need to mention it here because having multiple nixpkgs channels that are out of sync with versions is a common source of issues; otherwise people might update one of the channels and still run into the error without ever being able to get to the check that would tell them what’s wrong.

I’m not opposed to having a separate check for this that triggers even if the versions are all in order but it should maybe be a separate PR? This will become a bigger deal as I push out stage 1 of The Plan, since people will be running darwin-rebuild as root and (I think? not sure) not have access to their regular user’s channels any more, especially since our installation instructions previously recommended adding darwin to the user’s channels (while nixpkgs gets automatically added to root’s channels). But in that case maybe the check should be in darwin-rebuild itself.

(The version check itself is deliberately outside of the module system and as early as possible in evaluation, because there can be changes to Nixpkgs that would prevent an entire configuration from evaluating and building in the first place.)

Yeah, I was referring to checking that the channels are owned by root not by the current user

I think we should try to avoid additional nix-darwin specific login in darwin-rebuild, instead we could put it in the apply script as a check just to ensure that root has nixpkgs channel and that no other users have a nixpkgs channel

I don’t think it could live in an apply script, because the problem I anticipate is sudo darwin-rebuild not being able to find the darwin channel to build the configuration at all, thanks to our historical installation instructions. But I’ll have to do some testing. Anyway, seems more in scope for the The Plan PR than this one?

.github/workflows/test.yml

Enzime

LGTM

emilazy · 2025-01-14T01:47:02Z

Here goes nothing.

flake: pin Nixpkgs explicitly

0ef91bc

Samasaur1 reviewed Jan 8, 2025

View reviewed changes

bew reviewed Jan 9, 2025

View reviewed changes

Enzime reviewed Jan 10, 2025

View reviewed changes

version: implement nix-darwin release versions

8a3ea96

emilazy force-pushed the push-xztopxsynztu branch from 50bc034 to 41fb294 Compare January 11, 2025 15:38

Samasaur1 approved these changes Jan 11, 2025

View reviewed changes

emilazy mentioned this pull request Jan 13, 2025

checks: check for macOS ≥ 11.3 #1265

Open

emilazy force-pushed the push-xztopxsynztu branch from 41fb294 to 85f128e Compare January 13, 2025 00:05

ci: only test one version

c7b33c1

emilazy force-pushed the push-xztopxsynztu branch from 85f128e to e93e67a Compare January 13, 2025 00:09

eval-config: implement release branch checks

efba351

emilazy force-pushed the push-xztopxsynztu branch from e93e67a to 5a68c1c Compare January 13, 2025 02:46

emilazy mentioned this pull request Jan 13, 2025

checks: check /etc/nix/nix.custom.conf hash #1266

Draft

Enzime previously approved these changes Jan 14, 2025

View reviewed changes

emilazy added 2 commits January 14, 2025 01:19

checks: recommend sudo nix-channel

d5aeb4e

{readme,examples/flake}: update for release branches

e33d37c

emilazy dismissed Enzime’s stale review via e33d37c January 14, 2025 01:19

emilazy force-pushed the push-xztopxsynztu branch from 5a68c1c to e33d37c Compare January 14, 2025 01:19

Enzime approved these changes Jan 14, 2025

View reviewed changes

emilazy merged commit bd92122 into LnL7:master Jan 14, 2025
3 checks passed

emilazy deleted the push-xztopxsynztu branch January 14, 2025 01:49

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Make release branches a thing #1259

Make release branches a thing #1259

emilazy commented Jan 8, 2025

Samasaur1 Jan 8, 2025

Enzime Jan 10, 2025

emilazy Jan 11, 2025

Enzime Jan 13, 2025

emilazy Jan 13, 2025

Enzime Jan 14, 2025

bew Jan 9, 2025

emilazy Jan 10, 2025

Enzime Jan 10, 2025

Enzime Jan 10, 2025

Enzime Jan 10, 2025

Enzime Jan 10, 2025

emilazy Jan 11, 2025

Enzime Jan 13, 2025

emilazy Jan 13, 2025

Enzime left a comment

emilazy commented Jan 14, 2025

Make release branches a thing #1259

Make release branches a thing #1259

Conversation

emilazy commented Jan 8, 2025

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Enzime left a comment

Choose a reason for hiding this comment

emilazy commented Jan 14, 2025