jetbrains: switch builder to use extendMkDerivation pattern by theCapypara · Pull Request #478575 · NixOS/nixpkgs

theCapypara · 2026-01-09T23:58:57Z

Follow-up to #475183

I recommend disabling whitespaces for reviewing.

⚠️ The Darwin and aarch64 builds need attention. I can't test them and I can't explain why only these specific 5 IDEs are rebuilt on Darwin

Things done

Add a 👍 reaction to pull requests you find important.

pkgs/applications/editors/jetbrains/builder/default.nix

MattSturgeon

Looks like a much cleaner implementation overall!

I've scrolled through the diff a few times and left a mixture of design questions/comments and impl nitpicks. I'll stop scrolling now, to avoid overwhelming you with too much review spam. 🙈

pkgs/applications/editors/jetbrains/builder/darwin.nix

MattSturgeon · 2026-01-10T12:38:11Z

pkgs/applications/editors/jetbrains/builder/default.nix

-  meta = meta // {
-    teams = [ lib.teams.jetbrains ];
-  };
+      meta = (lib.optionalAttrs (args ? meta) args.meta) // {


Same feedback(s) as the other meta

pkgs/applications/editors/jetbrains/builder/linux.nix

MattSturgeon · 2026-01-10T12:56:58Z

pkgs/applications/editors/jetbrains/builder/default.nix

+    # Args to not pass to mkDerivation in the base builders. Since both get the same args
+    # passed in, both have the same list of args to ignore, even if they don't both use
+    # all of them.
+    excludeDrvArgNames = [
+      "product"
+      "productShort"
+      "wmClass"
+      "libdbm"
+      "fsnotifier"
+      "extraLdPath"
+      "extraWrapperArgs"
+    ];


Hm. This intuitively feels a bit ugly, but I get why you've taken this approach... I currently don't have a better suggestion.

If the goal is being DRY, you either do this or have both implementations import ./exclude-drv-arg-names.nix. If the goal is to be explicit about which args each implementation expects, then calling code would need to apply args conditionally.

Maybe related:

This also makes me question: does the current structure (common builder -> calls platform base builder) make more sense than the alternative (platform builder -> calls common base builder)... What do you think?

Regading your quote: I think which order we do this in kinda makes no difference, since we'll end up with the Darwin builder getting most of these attrs, even though it doesn't use them, regardless, unless we would specify separate drvs for Linux and Darwin for all IDEs, which doesn't make any sense really.

I'll think about this a bit, because I also kinda don't like that there isn't one place that documents the arguments. I guess having the excludeDrvArgNames here in this file at least allows seeing all of them in one file and we could even add documentation comments for them here as well.

I couldn't really find any example of how this is solved elsewhere, the pattern of needing to have a completely different base drv for Darwin and Linux seems to be somewhat unique.

We could maybe put all the Linux-specific attributes into a sub-attrset linux but this seems very confusing and may cause issues in the future, if we need some of them for the Darwin builder as well.

MattSturgeon · 2026-01-10T14:49:34Z

pkgs/applications/editors/jetbrains/ides/pycharm-oss.nix

+  mkPyCharmProduct,
  mkJetBrainsSource,
-  pyCharmCommonOverrides,
+  ...


These are callPackage args, right? Why the ...?

It doesn't use mkJetBrainsProduct anymore and I didn't want to introduce an extra function just for PyCharm in the default.nix, so it also gets mkPyCharmProduct passed. Happy for any suggestions to make this cleaner!

I'll probably need to clone this locally to fully understand, but my instinct is that unnecessary ... is a code smell. How come mkJetBrainsProduct and mkPyCharmProduct get supplied unconditionally?

MattSturgeon · 2026-01-10T14:49:59Z

pkgs/applications/editors/jetbrains/ides/pycharm.nix

  fsnotifier,
-  pyCharmCommonOverrides,
  musl,
+  ...


pkgs/applications/editors/jetbrains/builder/darwin.nix

pkgs/applications/editors/jetbrains/builder/linux.nix

theCapypara · 2026-01-11T10:44:47Z

The fact that there are now more rebuilds than before really confuses me. I can't really figure out why? I guess maybe because of the default attrs and how they are merged? Or some interactions with the overrides in the IDEs? But then why is mps rebuild on Darwin?

theCapypara · 2026-01-11T10:44:50Z

@ofborg build jetbrains.idea-oss.tests.plugins.stored-correctly

MattSturgeon · 2026-01-12T17:03:40Z

Testing locally, I was able to avoid rebuilds for jetbrains.idea with the following patch:

diff --git a/pkgs/applications/editors/jetbrains/builder/default.nix b/pkgs/applications/editors/jetbrains/builder/default.nix
index c0bf69d462c2..b710bb9cc0d9 100644
--- a/pkgs/applications/editors/jetbrains/builder/default.nix
+++ b/pkgs/applications/editors/jetbrains/builder/default.nix
@@ -23,6 +23,7 @@ lib.extendMkDerivation {
     excludeDrvArgNames = [
       "product"
       "productShort"
+      "buildNumber"
       "wmClass"
       "libdbm"
       "fsnotifier"
diff --git a/pkgs/applications/editors/jetbrains/builder/linux.nix b/pkgs/applications/editors/jetbrains/builder/linux.nix
index 76a08a4220a9..5194df00de23 100644
--- a/pkgs/applications/editors/jetbrains/builder/linux.nix
+++ b/pkgs/applications/editors/jetbrains/builder/linux.nix
@@ -81,21 +81,19 @@ lib.extendMkDerivation {
     {
       inherit desktopItem vmoptsIDE vmoptsFile;
 
-      buildInputs = [
+      buildInputs = buildInputs ++ [
         stdenv.cc.cc
         fontconfig
         libGL
         libX11
-      ]
-      ++ buildInputs;
+      ];
 
-      nativeBuildInputs = [
+      nativeBuildInputs = nativeBuildInputs ++ [
         makeWrapper
         patchelf
         unzip
         autoPatchelfHook
-      ]
-      ++ nativeBuildInputs;
+      ];
 
       postPatch = ''
         rm -rf jbr

buildInputs order was accidentally changed (not important, but does technically change the drv)
buildNumber was accidentally added as a derivation attr

Maybe this is enough for all IDEs?

I'm not suggesting that avoiding all rebuilds should be the end goal, but it does make the initial switch easier to review 🙂

To fully take advantage of fixpoint-args args, making them more accessible from finalAttrs and overridable via overrideAttrs, we'll want to allow more args to propagate into derivationAttrs (rather than being excluded). But it probably makes sense to do this after verifying that the initial change is rebuild-free.

theCapypara · 2026-01-12T17:11:51Z

Thanks! I will test later, this may get rid of all rebuilds.

If it does, I'll add additional commits to reduce the list of excluded attrs again, we'll have to see what makes sense there. This will cause rebuilds again, but we'll know why.

Everything else would then go in follow-up PRs.

theCapypara · 2026-01-17T21:59:43Z

Alright, got rid of all rebuilds now except for idea-oss and pycharm-oss. I diffed their drvAttrs and the only difference is that preferLocalBuild is now false instead of being undefined. I think that's fine?

So we can now talk about changing excludeDrvArgNames. I'm not entirely sure which of these attributes it would actually make sense to be override-able...

Edit: On Darwin there are more rebuilds: plugins is no longer part of drvAttrs, it was always empty before and was unused.

theCapypara · 2026-01-17T22:46:19Z

I now made it so that the first commit does not cause any rebuilds* (see the fact that the labels changed to "rebuild-*: 0". I'll now push a second commit that removes the unused plugins again on Darwin and refactors preferLocalBuild on Linux.

I also had to disable the buildInputs for dataspell on Darwin. I'm not sure anybody uses DataSpell on Darwin, I can't imagine those missing is actually correct, but the old builder for Darwin just ignored them, so I'll keep it like this for now and added a note.

*This first commit allegedly removes 6 packages, see the diff generated by CI:

{"added":[],"changed":["release-checks"],"rebuilds":["release-checks"],"removed":["jetbrains.idea.aarch64-darwin","jetbrains.phpstorm.aarch64-darwin","jetbrains.pycharm.aarch64-darwin","jetbrains.rider.aarch64-darwin","jetbrains.ruby-mine.aarch64-darwin","jetbrains.webstorm.aarch64-darwin"]}

It doesn't actually remove these.

MattSturgeon · 2026-01-20T12:55:53Z

*This first commit technically removes 6 packages, see the diff generated by CI:

Odd. Testing locally, all of these are still present. Not sure why the eval comparison would find them missing.

This commit does not cause any rebuilds.

MattSturgeon

Let's get this merged to unblock further refactoring. LGTM!

MattSturgeon · 2026-01-20T13:04:46Z

pkgs/applications/editors/jetbrains/builder/default.nix

+        inherit
+          buildNumber
+          product
+          libdbm
+          fsnotifier
+          ;


At this layer, these arguments are only used to populate passthru. I understand the motivation to avoid duplicating the passthru wiring in both linux.nix and darwin.nix, but it does mean this wrapper needs to be aware of arguments it otherwise doesn't consume. That's a bit of an abstraction leak: ideally each layer would fully own the arguments it accepts.

That said, this isn't something introduced by this change. I'm happy to defer that to a follow-up.

Stepping back, builder/default.nix itself is fairly thin, and it's worth reconsidering whether this level of indirection is buying us much at all.

MattSturgeon · 2026-02-04T11:20:19Z

For the record, I'm still not sure why the eval comparison thinks these packages have been removed; they are still there:

Added packages (0)

Removed packages (6)

Changed packages (9)

cc @NixOS/nixpkgs-ci

theCapypara · 2026-02-08T07:48:29Z

@MattSturgeon Seems that the packages got removed on Darwin because (somehow) musl got added as a dependency. So CI was right.

#488169

Edit: Yeah I somehow managed to miss the buildInputs of these packages, they need to be marked as Linux-only, since before they also effectively were (since the Darwin builder ignored them. Will fix later today.)

This comment was marked as outdated.

Sign in to view

theCapypara commented Jan 10, 2026

View reviewed changes

pkgs/applications/editors/jetbrains/builder/default.nix Show resolved Hide resolved

theCapypara marked this pull request as ready for review January 10, 2026 11:15

theCapypara requested a review from MattSturgeon January 10, 2026 11:15

nixpkgs-ci bot requested review from GenericNerdyUsername, Mic92, baldo, dritter, gytis-ivaskevicius, jamesward, leona-ya, raphaelr, thiagokokada and tymscar January 10, 2026 11:16

MattSturgeon reviewed Jan 10, 2026

View reviewed changes

theCapypara force-pushed the jetbrains/extendMkDerivation branch 2 times, most recently from dfb133f to 8347f2b Compare January 11, 2026 10:19

This comment was marked as outdated.

Sign in to view

theCapypara force-pushed the jetbrains/extendMkDerivation branch from 8347f2b to 25cff2b Compare January 17, 2026 21:42

nixpkgs-ci bot added 10.rebuild-linux: 1-10 This PR causes between 1 and 10 packages to rebuild on Linux. and removed 10.rebuild-linux: 11-100 This PR causes between 11 and 100 packages to rebuild on Linux. labels Jan 17, 2026

theCapypara force-pushed the jetbrains/extendMkDerivation branch from 25cff2b to b4f6a47 Compare January 17, 2026 22:07

theCapypara force-pushed the jetbrains/extendMkDerivation branch from b4f6a47 to 02c853a Compare January 17, 2026 22:25

nixpkgs-ci bot added 10.rebuild-linux: 0 This PR does not cause any packages to rebuild on Linux. 10.rebuild-darwin: 1 This PR causes 1 package to rebuild on Darwin. and removed 10.rebuild-linux: 1-10 This PR causes between 1 and 10 packages to rebuild on Linux. labels Jan 17, 2026

theCapypara force-pushed the jetbrains/extendMkDerivation branch 2 times, most recently from 69f3bca to 4264290 Compare January 17, 2026 22:39

nixpkgs-ci bot added 10.rebuild-darwin: 0 This PR does not cause any packages to rebuild on Darwin. and removed 10.rebuild-darwin: 1-10 This PR causes between 1 and 10 packages to rebuild on Darwin. 10.rebuild-darwin: 1 This PR causes 1 package to rebuild on Darwin. labels Jan 17, 2026

nixpkgs-ci bot added the 2.status: merge conflict This PR has merge conflicts with the target branch label Jan 30, 2026

theCapypara added 2 commits February 3, 2026 19:10

jetbrains: switch builder to use extendMkDerivation pattern

f55d624

This commit does not cause any rebuilds.

jetbrains: refactor some builder attributes

a6276de

theCapypara force-pushed the jetbrains/extendMkDerivation branch from 69f3bca to a6276de Compare February 3, 2026 18:11

nixpkgs-ci bot removed the 2.status: merge conflict This PR has merge conflicts with the target branch label Feb 3, 2026

MattSturgeon approved these changes Feb 4, 2026

View reviewed changes

MattSturgeon added this pull request to the merge queue Feb 4, 2026

Merged via the queue into NixOS:master with commit d96826b Feb 4, 2026
29 of 31 checks passed

This was referenced Feb 8, 2026

Build failure: jetbrains.idea on darwin #488169

Closed

jetbrains.idea,phpstorm,pycharm,rider,ruby-mine,webstorm: fix buildInputs on darwin #488283

Merged

Uh oh!

Conversation

theCapypara commented Jan 9, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Things done

Uh oh!

This comment was marked as outdated.

Uh oh!

MattSturgeon left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

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!

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!

This comment was marked as outdated.

theCapypara commented Jan 11, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

theCapypara commented Jan 11, 2026

Uh oh!

MattSturgeon commented Jan 12, 2026

Uh oh!

theCapypara commented Jan 12, 2026

Uh oh!

theCapypara commented Jan 17, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

theCapypara commented Jan 17, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

MattSturgeon commented Jan 20, 2026

Uh oh!

MattSturgeon left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

MattSturgeon commented Feb 4, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Uh oh!

theCapypara commented Feb 8, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

theCapypara commented Jan 9, 2026 •

edited

Loading

theCapypara commented Jan 11, 2026 •

edited

Loading

theCapypara commented Jan 17, 2026 •

edited

Loading

theCapypara commented Jan 17, 2026 •

edited

Loading

MattSturgeon commented Feb 4, 2026 •

edited

Loading

theCapypara commented Feb 8, 2026 •

edited

Loading