Not sure why nixf-diagnose --auto-fix doesn't pick up a lot of unused meta = with ... patterns. Also eval seems to fail, will take a look later.

Not sure why nixf-diagnose --auto-fix doesn't pick up a lot of unused meta = with ... patterns. Also eval seems to fail, will take a look later.

I think you'll need to specify a long disable list for nixf-diagnose otherwise, it will run all the other checks (and auto-fixes) as well, right? Maybe some other check broke something.

I wonder whether we can pin the right version of nixf-diagnose in the refactor-tractor repo and add it to the lib-prefix command (turning it into remove-meta-with-lib or so), aka doing both steps at once?

Then it would be fully reproducible, and runnable via a single flake reference - which, importantly, is outside the nixpkgs repo, which makes it much easier to use when doing the staging merges later.

Generated using 80c42444a7613782290df3ab5d2e346805620071 commit from: https://github.com/technowledgy/refactor-tractor/pull/15/commits

Changes:

• hydraPlatforms and badPlatforms are now also handled;
• variables bound inside lets are now considered, avoiding spurious overrides when their names match our heuristic patterns like platforms;
• all files with meta = with lib; are now processed, even when no ast-grep modifications happened;
• nixpkgs used for variable resolution is now pulled from the same directory as the target for modification (not from system <nixpkgs>).

$ git grep 'meta = with lib;' | grep '.nix:' | wc -l
90

$ ./diff_all_meta.sh pre post
== Diffing x86_64-linux ==

== Diffing x86_64-darwin ==

== Diffing aarch64-linux ==

== Diffing aarch64-darwin ==

No meta changes detected.

I've checked ~half of remaining pattern occurrences and all of them had a good explanation for why they were not picked up by automation (mostly just actual usage of functions from lib that we'll clean up manually). I got tired so I haven't checked all.

@wolfgangwalther, are we getting close to merge? I think this would be nice to land.

Not quite at the top of my list, yet, but it's getting there. I certainly have not forgotten and would like to finish it off soon.

Thank you, @booxter, for dealing with all my requests. This is great work, I'll follow through with it now.

• Merge this PR into master
• Merge into staging-next
• Merge into staging
• Merge into staging-nixos
• Merge into haskell-updates
• Create new PR for release-25.11 [25.11] treewide: clean up 'meta = with' pattern #469667
• Merge into staging-next-25.11
• Merge into staging-25.11
• Add .git-blame-ignore-revs entry on master .git-blame-ignore-revs: add meta = with cleanup #469665
• Add .git-blame-ignore-revs entry on release-25.11 [25.11] .git-blame-ignore-revs: add meta = with cleanup #469671

Not doing it for 25.05 anymore, because that's close to EOL.

@wolfgangwalther thanks a lot for support with the PR. It was a pleasure working with you on the change and I learned a lot. ❤️

Wow! This is awesome! Thank you!