Also, as points out, if you are willing to install browsers and other large applications ‘natively’ (Homebrew cask or drag & drop), Nix works really great for managing a Unix environment on macOS.ĭarwin/MacOS is important. A while ago when I was traveling I made some changes to PyTorch and every build was 1.5-2 hours.īut I think Darwin support is important, because it is a gateway for a lot of people. I still have a MacBook Air, it’s ok for reviewing (or fixing) derivations that do not have long build times or that do not have many reverse dependencies. Another issue is that beefy Mac hardware is quite expensive. I haven’t regretted it since, but that also means there is not much incentive to fixing macOS specific issues anymore.

However once I got deeper into the topic I was really eager to completely switch to NixOS to get all of the reproducibility benefits. Me personally, I started using nix on OS X and did my first contributions to nixpkgs via that. It is nice to get desktop apps sorted out, and I look forward to the day when I can manage everything I use with Nix, but I also don’t feel particularly put-out by having to install a few desktop apps this way. I get the attraction of One Config to Rule Them All, but forcing these into Nix smells weird to me (an anti-pattern? reproducibility theater? trying to fit a square peg in a round hole?) Some people use a Nix wrapper to download+install ( or even to run Homebrew), but many of these apps auto-update, or stop working when the client is no longer compatible with a centralized back-end. It doesn’t directly apply to Chromium, but enough of my desktop apps are closed-source and/or Mac-only that I don’t see any near-term escape from needing Homebrew or some other ritual for downloading and installing binaries. Perhaps I’m an outlier, but I haven’t experienced missing desktop apps as a showstopper on macOS. Many likely newcomers likely will abandon evaluating Nix on Darwin without some of them.

Chromium macos