gitbutler: init at 0.10.11 #289664

hacker1024 · 2024-02-18T05:35:13Z

Description of changes

This PR adds a GitButler package that builds from source.

Unfortunately quite a lot needs to be vendored, including the Yarn lockfile (#231513), which needs to be regenerated and may differ from the upstream pnpm lockfile.

Alternative to #289131.
Closes #288567.

macOS is untested as webkitgtk has been broken for years.

If you're interested in maintaining as well, feel free to let me know or make a PR into my branch adding yourself.

Things done

Add a 👍 reaction to pull requests you find important.

Stunkymonkey

LGTM

pkgs/by-name/gi/gitbutler-ui/package.nix

bobvanderlinden

Works quite well 👍 Better than the deb-based package (#289131).

That said, I'm a bit unsure about the bin/gitbutler-app, where-as the released version is named bin/git-butler. I couldn't figure out where in the release process this new name is applied, but it might be a good idea to align the Nix output with the released naming. Renaming in fixup could be enough if there is no proper way to do it.

pkgs/by-name/gi/gitbutler-ui/package.nix

ambroisie · 2024-02-18T19:51:09Z

pkgs/by-name/gi/gitbutler/package.nix

+
+  meta = gitbutler-ui.meta // {
+    description = "A Git client for simultaneous branches on top of your existing workflow.";
+    platforms = with lib.platforms; linux ++ darwin;


Is that an actual limitation of the program, or does it work on any platform which Rust can target (e.g: BSD)?

If the latter, don't set meta.platforms as buildRustPackage already sets it to the correct value.

Upstream only claims to support Linux and macOS, so I marked it as such. Is there a precedent for doing otherwise?

Suggested change

platforms = with lib.platforms; linux ++ darwin;

buildRustPackage already defaults to this

Stunkymonkey · 2024-02-18T19:55:45Z

Result of nixpkgs-review pr 289664 run on x86_64-linux 1

2 packages built:

gitbutler
gitbutler-ui

geri1701 · 2024-02-18T20:40:14Z

Result of nixpkgs-review pr 289664 run on x86_64-linux 1

2 packages built:

gitbutler
gitbutler-ui

This is permitted according to the license.

This allows backend API requests, such as those needed for logging in, to work.

…xport" This reverts commit e7d5f72d5049bab57d341c0fa8855332fed7a022.

ambroisie · 2024-02-22T09:37:32Z

@getchoo just took a look, using a FOD which calls pnpm directly is not the way to go about packaging pnpm dependencies, and until #231513 is fixed, we have to get around that issue by using yarn or npm packaging.

getchoo · 2024-02-22T21:23:43Z

@getchoo just took a look, using a FOD which calls pnpm directly is not the way to go about packaging pnpm dependencies

why not?

and until #231513 is fixed, we have to get around that issue by using yarn or npm packaging.

why? there is already precedent for this in other packages, with the former even being another tauri package using it for the same purpose. i genuinely don't see a reason as to why this shouldn't be used, or how not having standard tooling should force us into dealing with the pretty big maintenance burden (and error prone process) of keeping a lockfile at parity with upstream.

if there's a more robust solution that doesn't drastically increase maintenance, i'd love to see it; but having our own lockfile from a different package manager is not it.

ambroisie · 2024-02-22T21:37:21Z

why not?

@getchoo for the exact reason explained in the linked issue. If and when pnpm changes their store format, suddenly every FOD derivation based on it is invalid and won't build anymore.

there is already precedent for this in other packages

This package shouldn't have been approved and merged in that state.

how not having standard tooling should force us into dealing with the pretty big maintenance burden (and error prone process) of keeping a lockfile at parity with upstream.

Because when pnpm is updated and breaks our FODs, we have more churn to go through. Whereas creating lock files can easily be automated through update scripts if needed.

getchoo · 2024-02-22T22:09:06Z

If and when pnpm changes their store format

this isn't often from my experience with vesktop; seems to be working well in pot as well.
i'm not trying to say this isn't an issue at all, but it's not one big enough to go and maintain our own separate lockfile

This package shouldn't have been approved and merged in that state.

this is very subjective

Because when pnpm is updated and breaks our FODs, we have more churn to go through. Whereas creating lock files can easily be automated through update scripts if needed.

i would argue the churn required to update the hash is less than or equal to - as the comment in says in the current revision - "[the work required] to test the result thoroughly as dependency versions may differ from the release". we shouldn't be sacrificing the use of upstream's tested set of dependencies for marginal at best productivity gains through only some automation (that hasn't even been implemented)

couldn't we also update the FOD hash through an update script?

hacker1024 · 2024-02-25T04:58:21Z

I do not think there's much point in discussing the pros and cons of the pnpm technique here - that's why we have #231513 and #290715.

Until those discussions are resolved, it does not seem appropriate to copy the contested process once again. I'd prefer to stick with the tried and true Yarn method, despite potential problems arising due to mismatched lockfiles.

nixos-discourse · 2024-03-04T09:32:29Z

This pull request has been mentioned on NixOS Discourse. There might be relevant details there:

https://discourse.nixos.org/t/prs-ready-for-review/3032/3574

SuperSandro2000

Please squash all the reeview comment for each package into the init commit.

SuperSandro2000 · 2024-03-04T21:41:13Z

pkgs/by-name/gi/gitbutler-ui/package.nix

+  distPhase = "true";
+
+  meta = rec {
+    description = "The UI for GitButler.";


Suggested change

description = "The UI for GitButler.";

description = "Git client for simultaneous branches on top of your existing workflow";

SuperSandro2000 · 2024-03-04T21:41:24Z

pkgs/by-name/gi/gitbutler-ui/package.nix

+    license = lib.licenses.fsl-10-mit;
+    maintainers = with lib.maintainers; [ hacker1024 ];
+    platforms = with lib.platforms; all;
+    sourceProvenance = with lib.sourceTypes; [ fromSource ];


Suggested change

sourceProvenance = with lib.sourceTypes; [ fromSource ];

That doesn't need to be set

SuperSandro2000 · 2024-03-04T21:41:29Z

pkgs/by-name/gi/gitbutler-ui/package.nix

+    changelog = "https://github.com/gitbutlerapp/gitbutler/releases/tag/release/${version}";
+    license = lib.licenses.fsl-10-mit;
+    maintainers = with lib.maintainers; [ hacker1024 ];
+    platforms = with lib.platforms; all;


Suggested change

platforms = with lib.platforms; all;

platforms = lib.platforms.all;

SuperSandro2000 · 2024-03-04T21:41:42Z

pkgs/by-name/gi/gitbutler-ui/package.nix

+  meta = rec {
+    description = "The UI for GitButler.";
+    homepage = "https://gitbutler.com";
+    downloadPage = homepage;


Suggested change

downloadPage = homepage;

there is no point in repeating this.
Also this is used wrongly, since it is not a direct link to the download page

SuperSandro2000 · 2024-03-04T21:42:50Z

pkgs/by-name/gi/gitbutler/package.nix

+let
+  version = "0.10.11";
+in
+assert lib.assertMsg (version == gitbutler-ui.version) "The GitButler version does not match the GitButler UI version!";


Suggested change

let

version = "0.10.11";

in

assert lib.assertMsg (version == gitbutler-ui.version) "The GitButler version does not match the GitButler UI version!";

if we inherit the version anyway, then there is not really a point doing this

I think it might be useful to prevent improper overrides. But the current version will not do that, as you point out.

SuperSandro2000 · 2024-03-04T21:43:46Z

pkgs/by-name/gi/gitbutler/package.nix

+  '';
+
+  postInstall = ''
+    mv "$out"/bin/{gitbutler-app,'${meta.mainProgram}'}


Suggested change

mv "$out"/bin/{gitbutler-app,'${meta.mainProgram}'}

mv "$out"/bin/{gitbutler-app,'git-butler'}

a change to meta should not trigger a rebuild

SuperSandro2000 · 2024-03-04T21:43:55Z

pkgs/by-name/gi/gitbutler/package.nix

+      genericName = "Git client";
+      categories = [ "Development" ];
+      comment = meta.description;
+      exec = meta.mainProgram;


Suggested change

exec = meta.mainProgram;

exec = "git-butler";

SuperSandro2000 · 2024-03-04T21:45:00Z

pkgs/by-name/gi/gitbutler/package.nix

+      desktopName = "GitButler";
+      genericName = "Git client";
+      categories = [ "Development" ];
+      comment = meta.description;


Suggested change

comment = meta.description;

comment = "Git client for simultaneous branches";

most DEs don't render so long comments that great

SuperSandro2000 · 2024-03-04T21:45:21Z

pkgs/by-name/gi/gitbutler/package.nix

+
+  meta = gitbutler-ui.meta // {
+    description = "A Git client for simultaneous branches on top of your existing workflow.";
+    platforms = with lib.platforms; linux ++ darwin;


Suggested change

platforms = with lib.platforms; linux ++ darwin;

buildRustPackage already defaults to this

SuperSandro2000 · 2024-03-04T21:46:01Z

pkgs/by-name/gi/gitbutler/package.nix

+  ];
+
+  meta = gitbutler-ui.meta // {
+    description = "A Git client for simultaneous branches on top of your existing workflow.";


With my other review comment the description gets already set to this plus it will be compliant with the contributing guide which states that no A in the front or dot at the end should be.

Where is "no A in the front" stated?

https://github.com/NixOS/nixpkgs/tree/master/pkgs#meta-attributes doesn't.

https://nixos.org/manual/nixpkgs/unstable/#var-meta-description even puts a right example with this.

I don't think the A at the beginning caused any problems. In fact, in most cases it even makes the description read more fluently.

hallettj · 2024-03-19T18:57:12Z

How did you generate yarn.lock? I'm attempting to update the package for 0.10.24, but it seems I used a different method, and the yarn.lock diff is very large. Also I'm not sure my yarn.lock works correctly. What I tried was:

nix run nixpkgs#pnpm-lock-export -- --schema yarn.lock@v1

Edit: Oh I'm sorry - I finally saw the comment:

To generate the Yarn lockfile, run yarn install.

Also while I'm here, this package does not run on Wayland without XWayland. (Gitbutler's appimage has the same issue.) There is an error, "(git-butler:2957473): Gtk-WARNING **: 11:53:42.101: cannot open display: ". But from what I've read it seems like Tauri apps should run in Wayland? Anyone happen to know if there is a cargo feature that needs to be enabled, or an env var that needs to be set or something?

Edit: Updating gitbutler to 0.10.24 fixed the Wayland issue for me.

Thanks for working on the package!

ambroisie · 2024-03-23T11:22:43Z

nix run nixpkgs#pnpm-lock-export -- --schema yarn.lock@v1
Edit: Oh I'm sorry - I finally saw the comment:

To generate the Yarn lockfile, run yarn install.

If pnpm-lock-export works on this package, I'd recommend using it as it would allow us to use the same version of the packages as upstream specified.

hallettj · 2024-03-23T22:32:26Z

If pnpm-lock-export works on this package, I'd recommend using it as it would allow us to use the same version of the packages as upstream specified.

You're right, and it's probably important to match versions especially with the dependencies fetched via git. Unfortunately pnpm-lock-export is not handling git dependencies correctly when I try it. For example when I use yarn install I get this in yarn.lock:

"tauri-plugin-log-api@github:tauri-apps/tauri-plugin-log#v1":
  version "0.0.0"
  resolved "https://codeload.github.com/tauri-apps/tauri-plugin-log/tar.gz/db7255ca2e07fc4d3e6cc5d93f9ccfceacb28901"
  dependencies:
    "@tauri-apps/api" "1.5.3"

When I use pnpm-lock-export I get this instead:

"tauri-plugin-log-api@0.0.0":
  version "0.0.0"
  resolved "https://codeload.github.com/tauri-apps/tauri-plugin-log/tar.gz/19f5dcc0425e9127d2c591780e5047b83e77a7c2"
  integrity "undefined"
  dependencies:
    "@tauri-apps/api" "1.5.3"

Note that the git hash in the second matches the hash in the original lock file.

The pnpm-lock-export version leads to a build error. I can fix it by doing some fix-up by hand: deleting the integrity "undefined" line, and changing the specifier (the part after the @ on the first line). But then I'm getting an error saying that an esbuild package can't be found. I make experiment some more.

gitbutler-ui-modules> yarn config v1.22.19
gitbutler-ui-modules> success Set "yarn-offline-mirror" to "/nix/store/w8rw695xiggyqj1j46lc7v1iz77fh1a9-offline".
gitbutler-ui-modules> Done in 0.01s.
gitbutler-ui-modules> yarn install v1.22.19
gitbutler-ui-modules> [1/4] Resolving packages...
gitbutler-ui-modules> error Can't make a request in offline mode ("https://github.com/tauri-apps/tauri-plugin-log")
gitbutler-ui-modules> info Visit https://yarnpkg.com/en/docs/cli/install for documentation about this command.
gitbutler-ui-modules> Error: Can't make a request in offline mode ("https://github.com/tauri-apps/tauri-plugin-store")
gitbutler-ui-modules>     at MessageError.ExtendableBuiltin (/nix/store/f8nf1q0lffdni21lfw9frd5f9z231nzp-yarn-1.22.19/libexec/yarn/lib/cli.js:721:66)
gitbutler-ui-modules>     at new MessageError (/nix/store/f8nf1q0lffdni21lfw9frd5f9z231nzp-yarn-1.22.19/libexec/yarn/lib/cli.js:750:123)
gitbutler-ui-modules>     at RequestManager.request (/nix/store/f8nf1q0lffdni21lfw9frd5f9z231nzp-yarn-1.22.19/libexec/yarn/lib/cli.js:65887:59)
gitbutler-ui-modules>     at GitHubResolver.<anonymous> (/nix/store/f8nf1q0lffdni21lfw9frd5f9z231nzp-yarn-1.22.19/libexec/yarn/lib/cli.js:33912:48)
gitbutler-ui-modules>     at Generator.next (<anonymous>)
gitbutler-ui-modules>     at step (/nix/store/f8nf1q0lffdni21lfw9frd5f9z231nzp-yarn-1.22.19/libexec/yarn/lib/cli.js:310:30)
gitbutler-ui-modules>     at /nix/store/f8nf1q0lffdni21lfw9frd5f9z231nzp-yarn-1.22.19/libexec/yarn/lib/cli.js:328:14
gitbutler-ui-modules>     at new Promise (<anonymous>)
gitbutler-ui-modules>     at new F (/nix/store/f8nf1q0lffdni21lfw9frd5f9z231nzp-yarn-1.22.19/libexec/yarn/lib/cli.js:5305:28)
gitbutler-ui-modules>     at GitHubResolver.<anonymous> (/nix/store/f8nf1q0lffdni21lfw9frd5f9z231nzp-yarn-1.22.19/libexec/yarn/lib/cli.js:307:12)

Edit: I found out the esbuild problem comes up because pnpm-lock-export does not copy optionalDependencies to the generated yarn.lock

hallettj · 2024-03-25T01:11:48Z

It turns out that pnpm-lock-export hasn't been updated in quite a while. I dusted it off, updated it for the latest pnpm-lock.yaml format, and got it into good shape for converting gitbutler's lock file to a yarn.lock. (I didn't do much to update conversions to package-lock.json - that might need some more work.)

I updated my PR off of this PR, hacker1024#2, to use the converted lock file.

My pnpm-lock-export branch is here: https://github.com/hallettj/pnpm-lock-export

There is a nixpkgs package in progress at NixOS/nixpkgs#289664

ambroisie · 2024-03-25T11:11:18Z

@hallettj the current packaged version is a fork by @adamcstephens to support the v6 lock format, feel free to open a PR to update the package if it improves on it in any way.

hacker1024 · 2024-03-26T05:31:33Z

Sorry for the long silence, I'll update this PR shortly. Thanks everyone!

luftmensch-luftmensch · 2024-04-20T13:13:57Z

Any updates on this PR?

xyven1 · 2024-05-12T02:33:42Z

Bumping this.

Aleksanaa · 2024-05-14T09:23:11Z

Hi, @DataHearth has opened a new pr #289664. Do you still want to work on this or prefer their PR instead?

DataHearth · 2024-05-14T09:28:44Z

Hi, @DataHearth has opened a new pr #289664. Do you still want to work on this or prefer their PR instead?

Hi ! The only prerequisite I need for my PR is someone with a MacOS system to validate the unpack phase. If the current way is not valide, I already have the backup to make it works.

For linux, I'm using the appimage. I didn't see any runtime dependencies required for now that wasn't bundled with the appimage (Like jdk for nosql-workbench I've already bundled).

I've already tested and used it on x64 linux.

Cheers!

luftmensch-luftmensch · 2024-05-14T09:48:11Z

Hi, @DataHearth has opened a new pr #289664. Do you still want to work on this or prefer their PR instead?

It hasn't been decided to use the source based version instead of the AppImage?

alyssais · 2024-05-14T09:56:10Z

Generally in Nixpkgs we prefer to build from source where possible.

luftmensch-luftmensch · 2024-05-14T11:35:57Z

Generally in Nixpkgs we prefer to build from source where possible.

So #289664 should be preferred, and hopefully closed shortly.

SuperSandro2000 · 2024-06-06T11:59:40Z

pkgs/by-name/gi/gitbutler-ui/package.nix

+  # The package.json must use spaces instead of upstream's tabs to pass Nixpkgs
+  # CI.
+  # To generate the Yarn lockfile, run `yarn install`.
+  # There is no way to import the tagged pnpm lockfile, so make sure to test the
+  # result thoughly as dependency versions may differ from the release.
+  packageJSON = ./package.json;
+  yarnLock = ./yarn.lock;
+  offlineCache = fetchYarnDeps {
+    inherit yarnLock;
+    hash = "sha256-rggtkfE6An8It0Rvgfk0J8JHpg0NbLiweRsz0nM/tzM=";
+  };


#290715 👀

hacker1024 · 2024-06-09T15:43:57Z

Closing in favour of #318101.

hacker1024 marked this pull request as draft February 18, 2024 05:35

hacker1024 changed the title ~~gitbutler: iniit at 0.10.10~~ gitbutler: init at 0.10.10 Feb 18, 2024

ofborg bot added 8.has: package (new) 11.by: package-maintainer 10.rebuild-darwin: 1-10 10.rebuild-linux: 1-10 labels Feb 18, 2024

hacker1024 marked this pull request as ready for review February 18, 2024 06:49

hacker1024 mentioned this pull request Feb 18, 2024

Package request: GitButler #288567

Closed

Stunkymonkey reviewed Feb 18, 2024

View reviewed changes

pkgs/by-name/gi/gitbutler-ui/package.nix Outdated Show resolved Hide resolved

bobvanderlinden reviewed Feb 18, 2024

View reviewed changes

bobvanderlinden mentioned this pull request Feb 18, 2024

gitbutler: init at 0.10.6 #289131

Closed

13 tasks

ambroisie reviewed Feb 18, 2024

View reviewed changes

pkgs/by-name/gi/gitbutler-ui/package.nix Show resolved Hide resolved

ambroisie reviewed Feb 18, 2024

View reviewed changes

hacker1024 added 15 commits February 20, 2024 19:12

gitbutler-ui: init at 0.10.10

30cae9b

gitbutler: init at 0.10.10

1fc5716

gitbutler-ui: Do not use URI in desktop file

15a627b

gitbutler: Use wrapGAppsHook

bb8955f

gitbutler-ui, gitbutler: Define meta.sourceProvenance in a list

470c08e

gitbutler, gitbutler-ui: Pass EditorConfig test

2a07ea7

gitbutler-ui, gitbutler: Allow redistribution

b66babc

This is permitted according to the license.

gitbutler: Use glib-networking

54c0efb

This allows backend API requests, such as those needed for logging in, to work.

gitbutler: Use correct executable name

7306644

gitbutler: Remove unused scheme handler

7c42262

gitbutler-ui: Update lockfile instructions to use pnpm-lock-export

9b2bd4f

gitbutler-ui: Revert "Update lockfile instructions to use pnpm-lock-e…

2279a50

…xport" This reverts commit e7d5f72d5049bab57d341c0fa8855332fed7a022.

gitbutler, gitbutler-ui: 0.10.10 -> 0.10.11

8b3da89

gitbutler: Use upstream binary name

bbbaf59

licenses: Add fsl-10-mit

ecef428

getchoo mentioned this pull request Feb 22, 2024

pnpm.fetchDeps: init #290715

Merged

16 tasks

hacker1024 mentioned this pull request Mar 4, 2024

gitbutler: use pnpm directly to build hacker1024/nixpkgs#1

Closed

SuperSandro2000 requested changes Mar 4, 2024

View reviewed changes

iynaix mentioned this pull request Mar 23, 2024

Package request: gitbulter #298179

Closed

hallettj mentioned this pull request Mar 23, 2024

update gitbutler to 0.10.27 and update with PR review feedback hacker1024/nixpkgs#2

Open

hallettj added a commit to hallettj/home.nix that referenced this pull request Mar 25, 2024

remove my gitbutler package

28481a8

There is a nixpkgs package in progress at NixOS/nixpkgs#289664

msanft mentioned this pull request May 14, 2024

git-butler: init at 0.11.7 #311361

Closed

13 tasks

SuperSandro2000 reviewed Jun 6, 2024

View reviewed changes

getchoo mentioned this pull request Jun 7, 2024

gitbutler: init at 0.12.2 #318101

Merged

13 tasks

hacker1024 closed this Jun 9, 2024

	description = "The UI for GitButler.";
	description = "Git client for simultaneous branches on top of your existing workflow";

	platforms = with lib.platforms; all;
	platforms = lib.platforms.all;

	mv "$out"/bin/{gitbutler-app,'${meta.mainProgram}'}
	mv "$out"/bin/{gitbutler-app,'git-butler'}

	comment = meta.description;
	comment = "Git client for simultaneous branches";

gitbutler: init at 0.10.11 #289664

gitbutler: init at 0.10.11 #289664

Conversation

hacker1024 commented Feb 18, 2024 • edited Loading

Description of changes

Things done

Stunkymonkey left a comment

Choose a reason for hiding this comment

bobvanderlinden left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

hacker1024 Feb 20, 2024 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Stunkymonkey commented Feb 18, 2024

geri1701 commented Feb 18, 2024

ambroisie commented Feb 22, 2024

getchoo commented Feb 22, 2024

ambroisie commented Feb 22, 2024

getchoo commented Feb 22, 2024

hacker1024 commented Feb 25, 2024

nixos-discourse commented Mar 4, 2024

SuperSandro2000 left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

hallettj commented Mar 19, 2024 • edited Loading

ambroisie commented Mar 23, 2024

hallettj commented Mar 23, 2024 • edited Loading

hallettj commented Mar 25, 2024

ambroisie commented Mar 25, 2024

hacker1024 commented Mar 26, 2024

luftmensch-luftmensch commented Apr 20, 2024

xyven1 commented May 12, 2024

Aleksanaa commented May 14, 2024

DataHearth commented May 14, 2024

luftmensch-luftmensch commented May 14, 2024

alyssais commented May 14, 2024

luftmensch-luftmensch commented May 14, 2024

Choose a reason for hiding this comment

hacker1024 commented Jun 9, 2024

hacker1024 commented Feb 18, 2024 •

edited

Loading

hacker1024 Feb 20, 2024 •

edited

Loading

hallettj commented Mar 19, 2024 •

edited

Loading

hallettj commented Mar 23, 2024 •

edited

Loading