top-level/all-packages: remove overwriting by-name callPackage by wolfgangwalther · Pull Request #453948 · NixOS/nixpkgs

wolfgangwalther · 2025-10-20T18:40:01Z

In #444420, we changed the approach regarding the override interface for packages in by-name. This allows us to move a lot more packages to by-name - and also to get rid of callPackage in all-packages.nix for by-name packages.

This PR removes all cases where a by-name package was overwritten, aka with the same name. There are still plenty of callPackage into by-name folders in all-packages.nix, but these introduce new attributes.

This first step will allow us to disallow overwriting these by-name attributes and then simplify nixpkgs-vet accordingly.

Things done

Ensure 0 rebuilds.
Fits CONTRIBUTING.md, pkgs/README.md, maintainers/README.md and other READMEs.

Add a 👍 reaction to pull requests you find important.

emilazy

I would like to see the changes in “by-name/various: move default values to override interface”, “by-name/various: use scopes as arguments”, “aegisub: remove overrides from all-package.nix”, and “by-name/various: move overrides to package.nix” consistently move the overrides into a let … in … in package.nix rather than adjusting all the references in the files.

This will not always produce what I think is a superior end result, but in some cases it will, and it minimizes churn – both review and git blame – while making it easier to review for correctness. Since the packages were written this way to begin with, we should by default assume that the identifiers they chose are appropriate – and if we think we should remove that layer of indirection, that should be a separate refactoring change, distinct from this mechanical lifting.

pkgs/applications/editors/vscode/vscode.nix

pkgs/applications/editors/vscode/vscodium.nix

pkgs/top-level/python-packages.nix

emilazy · 2025-10-20T19:01:52Z

pkgs/by-name/el/element-web/package.nix

-  conf ? { },
+  config,
+  conf ? config.element-web.conf or { },


I don’t think we should be using config for this. (Not a blocker, because it was like that beforehand – but I’d like to drop this, as you did with config.aegisub. OTOH, conf is really not a great name for this anyway, so eh…)

The problem here is that conf is currently used from within nixpkgs, IIRC. So I can't drop it as easily. But I agree with you 100%.

Scrubbing most ad-hoc config. from pkgs is on my giant notebook of "changes-to-make".

pkgs/by-name/re/rescript-language-server/package.nix

pkgs/by-name/pl/plfit/package.nix

pkgs/by-name/ba/bazel_7/package.nix

pkgs/by-name/at/atf/package.nix

pkgs/by-name/xc/xcbuild/package.nix

pkgs/by-name/go/gopro-tool/package.nix

wolfgangwalther · 2025-10-20T19:43:28Z

I would like to see the changes in “by-name/various: move default values to override interface”, “by-name/various: use scopes as arguments”, “aegisub: remove overrides from all-package.nix”, and “by-name/various: move overrides to package.nix” consistently move the overrides into a let … in … in package.nix rather than adjusting all the references in the files.

Will look into these things tomorrow, but I addressed all other feedback, so far.

wolfgangwalther · 2025-10-21T09:11:09Z

I would like to see the changes in “by-name/various: move default values to override interface”, “by-name/various: use scopes as arguments”, “aegisub: remove overrides from all-package.nix”, and “by-name/various: move overrides to package.nix” consistently move the overrides into a let … in … in package.nix rather than adjusting all the references in the files.

This will not always produce what I think is a superior end result, but in some cases it will, and it minimizes churn – both review and git blame – while making it easier to review for correctness. Since the packages were written this way to begin with, we should by default assume that the identifiers they chose are appropriate – and if we think we should remove that layer of indirection, that should be a separate refactoring change, distinct from this mechanical lifting.

Unfortunately, the addition of that let block is not easily automated with the ast-grep based scripts, I have.

I now addressed this feedback - in a compromise that is fair, I think. The majority of packages still uses the simple replacement, because they are really simple cases.

Everything that I felt was either more complex or where intent to do things differently in the expression was visible, uses the let pattern now.

There is one review comment open, but I just left that for visibility. I don't intend to do something about it in here.

From my perspective, this is ready.

Overriding the argument with the same default value as in package.nix is pointless.

The contributing guidelines have been changed to encourage breaking the override interface to benefit from explicit conflict resolution on upgrades.

Improves the override interface, because `withPython` does not conflict with the old `python` package.

The config option was introduced in 9361acf, this should just be done via overlays today.

This has the additional benefit, that the tests will run with `finalPackage`, i.e. taking overrides into account.

fmt_11 is now the default.

wolfgangwalther · 2025-10-21T19:26:34Z

just rebased to resolve merge conflicts

MattSturgeon

Diff LGTM, minus merge conflict

wolfgangwalther · 2025-10-23T09:36:42Z

Current status: @emilazy is hacking on the ast-grep script I used for the initial changes to see whether we can make this more automated and less manual - and then scale this to tons of other packages that are not in by-name yet, but need the same treatment eventually.

reckenrode · 2025-10-23T15:54:43Z

pkgs/by-name/at/atf/package.nix

+
+let
+  # atf is a dependency of libiconv. Avoid an infinite recursion with `pkgsStatic` by using a bootstrap stdenv.
+  stdenv = if args.stdenv.hostPlatform.isDarwin then darwin.bootstrapStdenv else args.stdenv;


This means that it’s impossible to override the stdenv on Darwin (without knowing implementation details of the derivation). Is there a way to do this that preserves the ability to override it? That’s why I originally did it in all-packages.nix.

I think @emilazy's comment in #453948 (comment) applies to that?

I tried moving this into darwin's stdenv bootstrap, but failed. I'm sure you'd be capable of doing that, though :)

I can try to do it.

It can’t be in the bootstrap because then pkgsStatic will have an infinite recursion. To make things behave the same between native and cross, the “overrides” have to be in the package definitions not in the bootstrap overlay.

I see, that would explain me failing to do it.

We could introduce a withStdenv ? if ... argument to allow overriding.

Putting as draft for the moment. I am currently looking into a way to make this case better to solve here - it's really unsatisfactory, that we can't override stdenv for these packages afterwards anymore.

wolfgangwalther · 2025-12-10T20:16:04Z

Not going to pursue this anymore myself (#469692).

wolfgangwalther requested a review from emilazy October 20, 2025 18:40

nix-owners bot requested review from Profpatsch and natsukium October 20, 2025 18:45

emilazy reviewed Oct 20, 2025

View reviewed changes

wolfgangwalther force-pushed the by-name-explicit-defaults branch from cbdfc0c to 136c599 Compare October 20, 2025 19:43

wolfgangwalther mentioned this pull request Oct 21, 2025

code-cursor: use generic vscode package builder #413785

Merged

13 tasks

wolfgangwalther force-pushed the by-name-explicit-defaults branch from 136c599 to 17d2489 Compare October 21, 2025 09:08

wolfgangwalther mentioned this pull request Oct 21, 2025

top-level/stage: assert no by-name overwrites #454147

Closed

3 tasks

nixpkgs-ci bot added the 2.status: merge conflict This PR has merge conflicts with the target branch label Oct 21, 2025

wolfgangwalther added 11 commits October 21, 2025 21:23

by-name/various: remove unused argument overrides

485848c

Overriding the argument with the same default value as in package.nix is pointless.

by-name/various: move default values to override interface

1aafd74

The contributing guidelines have been changed to encourage breaking the override interface to benefit from explicit conflict resolution on upgrades.

by-name/various: use explicit default values

a059b9f

by-name/various: use scopes as arguments

43bc90d

plfit: python -> withPython

8017563

Improves the override interface, because `withPython` does not conflict with the old `python` package.

bazel_7: improve override interface

c26212f

gfal2-util: remove outdated override

7cbeba8

atf: move darwin stdenv conditional into package.nix

f78015b

xcbuild: move darwin stdenv conditional into package nix

622bbe6

aegisub: remove overrides from all-package.nix

821c2f6

The config option was introduced in 9361acf, this should just be done via overlays today.

by-name/various: move overrides to package.nix

168125e

wolfgangwalther and others added 4 commits October 21, 2025 21:24

glib: move glib-untested overrides into package.nix

27706e3

This has the additional benefit, that the tests will run with `finalPackage`, i.e. taking overrides into account.

buildVscode: init

b10a6ad

meshlabl{,-unstable}: move darwin conditionals to package.nix

a22b922

saunafs: remove outdated override

866eee1

fmt_11 is now the default.

wolfgangwalther force-pushed the by-name-explicit-defaults branch from 17d2489 to 866eee1 Compare October 21, 2025 19:25

nixpkgs-ci bot removed the 2.status: merge conflict This PR has merge conflicts with the target branch label Oct 21, 2025

wolfgangwalther mentioned this pull request Oct 22, 2025

top-level/stage: assert no by-name overwrites #454525

Closed

3 tasks

MattSturgeon approved these changes Oct 22, 2025

View reviewed changes

nixpkgs-ci bot added 2.status: merge conflict This PR has merge conflicts with the target branch 12.approvals: 1 This PR was reviewed and approved by one person. labels Oct 22, 2025

reckenrode reviewed Oct 23, 2025

View reviewed changes

wolfgangwalther marked this pull request as draft October 26, 2025 17:43

wolfgangwalther mentioned this pull request Nov 3, 2025

botan3: fix static build #458084

Merged

13 tasks

wolfgangwalther closed this Dec 10, 2025

wolfgangwalther deleted the by-name-explicit-defaults branch December 10, 2025 20:16

This was referenced Dec 11, 2025

treewide: remove all-packages.nix overrides of packages defined in by-name (A-C) #469771

Merged

treewide: remove all-packages.nix overrides of packages defined in by-name (H) #469776

Merged

This was referenced Dec 26, 2025

various: remove by-name overrides #474456

Closed

buildVscode: init #474458

Merged

gepbird mentioned this pull request Jan 12, 2026

[RFC 0192] version pins for the pkgs/by-name structure NixOS/rfcs#192

Open

gepbird mentioned this pull request Jan 26, 2026

various: move to pkgs/by-name #483842

Merged

13 tasks

gepbird mentioned this pull request Feb 3, 2026

xcbuild: move top level override into package.nix #485704

Merged

2 tasks

Uh oh!

Conversation

wolfgangwalther commented Oct 20, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Things done

Uh oh!

emilazy left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

wolfgangwalther commented Oct 20, 2025

Uh oh!

wolfgangwalther commented Oct 21, 2025

Uh oh!

wolfgangwalther commented Oct 21, 2025

Uh oh!

MattSturgeon left a comment

Choose a reason for hiding this comment

Uh oh!

wolfgangwalther commented Oct 23, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

reckenrode Oct 23, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

wolfgangwalther commented Dec 10, 2025

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

wolfgangwalther commented Oct 20, 2025 •

edited

Loading

wolfgangwalther commented Oct 23, 2025 •

edited

Loading

reckenrode Oct 23, 2025 •

edited

Loading