feat: add nix-eval workflow for cross-platform flake evaluation

[jerome-benoit]

Replace nix-desktop.yml build workflow with lightweight cross-platform evaluation.

• Evaluate all packages (opencode, desktop) and devShells for all 4 systems
• Single Linux runner (~2-5 min) instead of 4 native runners (~60 min)
• Catches evaluation-time errors (missing outputHashes, syntax errors, missing deps)

Ref #11755

[github-actions]

Thanks for your contribution!

This PR doesn't have a linked issue. All PRs must reference an existing issue.

Please:

1. Open an issue describing the bug/feature (if one doesn't exist)
2. Add Fixes #<number> or Closes #<number> to this PR description

See CONTRIBUTING.md for details.

[github-actions]

The following comment was made by an LLM, it may be inaccurate:

No duplicate PRs found

[Copilot]

Pull request overview

This PR replaces the old (disabled) Nix desktop build workflow with a lightweight, cross-platform flake evaluation workflow to catch evaluation-time issues (like missing outputHashes) earlier and more cheaply.

Changes:

• Add .github/workflows/nix-eval.yml to evaluate opencode and desktop packages and devShells across all four supported systems from a single Linux runner.
• Remove the obsolete .github/workflows/nix-desktop.yml.disabled workflow that previously built the desktop package on multiple native runners.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 1 comment.

File	Description
.github/workflows/nix-eval.yml	New workflow that evaluates flake packages and devShells for all supported systems to detect evaluation-time Nix errors.
.github/workflows/nix-desktop.yml.disabled	Removes the old, disabled Nix desktop build workflow in favor of the new evaluation-based approach.

Comments suppressed due to low confidence (1)

.github/workflows/nix-eval.yml:77

• Similar to the packages loop above, the same nix eval is executed twice here on failure (once in the if condition with output suppressed and again in the else branch). To avoid redundant and potentially expensive Nix evaluations, consider restructuring this logic so that each devShell is only evaluated once while still emitting the error details when it fails.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[ndrwstn]

@jerome-benoit @rekram1-node

So this shouldn't be merged, IMO. Nix does not guarantee correct hashes for other platforms (e.g. nix-darwin) on a linux-runner. This is a repeating issue as we go through a cycle of someone thinking we should just use a linux runner to calculate all the hashes--which may work for a release or two, but then breaks for a major upgrade and we have to reimplement the multiple build platform.

I didn't mean to be rude, if it's interpreted that way, just a little frustrated by this cycle.

[jerome-benoit]

I didn't mean to be rude, if it's interpreted that way, just a little frustrated by this cycle.

You haven't actually read and understood the purpose of that PR:

• no relation at all with the hashes generation, at all, by any means
• standard lightweigth ci for nix packaging to ensure a flake is actually syntactically correct and the build requirements are met.

A check of the flake.lock freshness, etc. is still missing.

[ndrwstn]

You're right that I misunderstood what you were doing.

[rekram1-node]

so is this one good to go?

[jerome-benoit]

so is this one good to go?

as long as you expect to have non blocking failures in the CI for opencode desktop because of #11755. The opencode nix package is now evaluated in the workflow.