got-unwrapped: init at 0.100; got: wrap got-unwrapped with ssh

[eclairevoyant]

Description of changes

Created as an alternative to #297154 to prevent superfluous rebuilds.
TBH I'm not a big fan of this approach, a rebuild takes a few seconds, and this increases maintenance burden, but this does address the comments on the previous PR.

Fixes #322043.
Closes #297154.

Things done

• Built on platform(s)
  • x86_64-linux
  • aarch64-linux
  • x86_64-darwin
  • aarch64-darwin
• For non-Linux: Is sandboxing enabled in nix.conf? (See Nix manual)
  • sandbox = relaxed
  • sandbox = true
• Tested, as applicable:
  • NixOS test(s) (look inside nixos/tests)
  • and/or package tests
  • or, for functions and "core" functionality, tests in lib/tests or pkgs/test
  • made sure NixOS tests are linked to the relevant packages
• Tested compilation of all packages that depend on this change using nix-shell -p nixpkgs-review --run "nixpkgs-review rev HEAD". Note: all changes have to be committed, also see nixpkgs-review usage
• Tested basic functionality of all binary files (usually in ./result/bin/)
• 24.11 Release Notes (or backporting 23.11 and 24.05 Release notes)
  • (Package updates) Added a release notes entry if the change is major or breaking
  • (Module updates) Added a release notes entry if the change is significant
  • (Module addition) Added a release notes entry if adding a new NixOS module
• Fits CONTRIBUTING.md.

---

Add a 👍 reaction to pull requests you find important.

[afh]

Thanks for this, @eclairevoyant, very much appreciated. Currently I'm a bit confused by the naming—not that there is anything wrong with it per se, could just be my current understanding of nixpkgs best practices.

The help clear this up, @eclairevoyant, could you provide more context or explanation around the motivation to call the "original"/"unmodified" package that offers no "customization" of ssh in got got-unwrapped and the one that does got?

Is it just because the "new" got is a wrapper around the "old" one?

[eclairevoyant]

It's how every other wrapped package I've seen is named. See nvim, mpv, sway, etc.

[afh]

The more I go over this, the more I'm beginning to understand, and starting to really liking the change, @eclairevoyant, very well done!

Prior to review just a minor question more out of interest than anything else:
What is the motivation / pro / con for patching the source over using substituteInPlace in package.nix to replace execv(GOT_DIAL_PATH_SSH with execvp(GOT_DIAL_PATH_SSH in lib/dial.c?

[eclairevoyant]

I don't have a strong opinion either way; I can switch it to substituteInPlace if that's easier to maintain.

[wahjava]

Since you're also overridding GOT_TAG_PATH_SSH_KEYGEN could you patch its use to execvp as well.

And similarly there is occurence for /usr/bin/signify (and /usr/bin/vi). Maybe we can patch that (those) too ?

Thanks!

[afh]

I think that substituteInPlace provides more flexibility if upstream changes the code around the execv calls; a patch is likely to fail just because some surrounding code changes, e.g. added or deleted lines in the source file that do not directly affect the execv call…

[eclairevoyant]

Didn't notice those earlier. I find that it's weird that they hardcode the editor at build-time instead of using an envvar with a default of vi or something.

[eclairevoyant]

Yeah nvm I think this is just the completely wrong approach

[afh]

Thanks for the work on this, @eclairevoyant, too bad this doesn't seem to head in the right direction for you.
What would be a better approach? Is this something that is better raised and addressed upstream?

[eclairevoyant]

If they're just shelling out, this is something best handled at runtime via some envvar, not via buildtime #define. And yes upstream would be best to address this IMO

[afh]

ℹ️ It seems the changes introducing GOT_DEFAULT_EDITOR were motivated by someone thinking of NixOS (see https://marc.gameoftrees.org/mail/1710762601.52948_0.html)

[eclairevoyant]

There's really no reason that they should require users to recompile got, just to use another editor. It's a bad design.

git does the sensible thing and reads VISUAL or EDITOR envvars by default.

[afh]

got does indeed check and use the VISUAL and EDITOR envvars (see cvg/cvg.c:275f), only its default editor is hardcoded, which I believe make sense to change for nixpkgs if people so desire.

How about continuing the conversation over on #322600, @eclairevoyant?