Skip to content

flake: add albert-sourceos package (fork w/ SourceOS plugin) - #50

Closed
mdheller wants to merge 4 commits into
mainfrom
feat/nix-albert-sourceos
Closed

flake: add albert-sourceos package (fork w/ SourceOS plugin)#50
mdheller wants to merge 4 commits into
mainfrom
feat/nix-albert-sourceos

Conversation

@mdheller

@mdhellermdheller commented Apr 19, 2026

Copy link
Copy Markdown
Contributor

Closed: Albert is not OSI-open and is classified as unfree by nixpkgs.

See discussion in comments: SociOS-Linux/albert uses "Albert license v1.1" which permits redistribution only without modification and includes platform-targeting restrictions, making it incompatible with SourceOS constraints and with shipping a modified build including plugins/sourceos.

Replacement PR will pivot to an open-source launcher palette (Wayland-first; fuzzel/wofi/rofi) bound to Super+Space via sourceos palette.

@mdhellerChatGPT Codex Connector

Copy link
Copy Markdown
ContributorAuthor

CI shows albert-sourceos cannot build under nixpkgs defaults because Albert is classified as unfree (evaluation refuses without allowUnfree).

More importantly, upstream Albert is not OSI-open: SociOS-Linux/albert ships LICENSE.md as "Albert license v1.1" which:

  • permits redistribution without modification
  • does not permit redistribution of modifications
  • includes explicit platform targeting restrictions

That’s incompatible with SourceOS constraints (open-source only) and with our need to ship a modified build including plugins/sourceos.

Recommendation:

  • Close this PR and pivot to an open-source launcher lane (Wayland-first). Fedora already packages fuzzel (MIT) and wofi (GPL-3.0-only). We can bind Super+Space to sourceos palette implemented with fuzzel/wofi/rofi dmenu mode.

I’ll open a replacement PR that removes Albert dependency and implements the open-source launcher palette.

@mdheller
mdheller deleted the feat/nix-albert-sourceos branch July 19, 2026 02:07
mdheller added a commit that referenced this pull request Jul 29, 2026
)
The health check ran `sourceos-boot rollback execute --execute || true`. That `|| true`
turned the last line of defence into a reporting lie: the subcommand was dropped from
sourceos-boot by PR #48, so on any host tracking HEAD the call is argparse exit 2, the
unit logs "triggering rollback", and then reports success having done nothing.
It survived only because flake.lock pins sourceos-boot-src to bc6dd8c, the commit before
the removal — one `nix flake update` from disarming auto-rollback on stable-x86_64,
canary-x86_64, exit-x86_64 and builder-aarch64, all of which set rollbackOnFailure = true.
This branch runs precisely when the host is ALREADY unhealthy, so a rollback that cannot
run is the worst possible thing to hide. Failure now logs what happened and exits 3 —
distinct from the plain-unhealthy exit 2, so "unhealthy" and "unhealthy AND did not roll
back" are different signals to whatever supervises the unit. Success is logged too, so the
happy path is observable rather than inferred from silence.
The subcommand itself is restored in sourceos-boot #50; this change is what stops the next
such removal from being invisible.
Sign up for freeto join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant

@mdheller