Usage of the nixFlakes attribute obscures the fact that it's an
experimental feature. Providing a nixExperimental attribute (which
people will inevitably start using on their production machines) makes
it too easy to enable all experimental features, when you should
explicitly opt in to the features that you want to try out.
Also, upstream Nix doesn't provide an "enable all experimental
features" patch so neither should Nixpkgs.
This includes an overhaul of the install script, using patchelf to set
the interpreter and the libraries. This reflects a lot of what I've
learned about electron + nixos over the last few months.
This change is motivated by the fact that 1Password's version of
electron changed, probably leading to crashes that I've been seeing when
I try to update the Nix derivation to the latest versions.
Co-authored-by: Daniël de Kok <me@github.danieldk.eu>
Remove elements of the PR template that have a low signal/noise ratio,
and add one that I think would have a good signal/noise ratio.
-----
Remove:
Determined the impact on package closure size (by running `nix path-info
-S` before and after)
-----
Rationale:
This is rarely done in practice, and apart from for specific packages
this is usually not a good indicator of anything useful
It might make sense to re-introduce it with two holes to fill, but then
we would have to make a serious decision to never land without these two
numbers filled in or with too big a regression, because in practice this
box has been a no-op in many cases.
Maybe just integrating this check in nixpkgs-review would bring the most
benefit here?
-----
-----
Remove:
Ensured that relevant documentation is up to date
-----
Rationale:
This is fuzzy, “relevant documentation” is way too often hard to find
-----
-----
Add:
Added a release notes entry if the change is major or breaking
-----
Rationale:
This is way too often forgotten, and is also a self-contained easy task
-----
