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

https://discourse.nixos.org/t/successful-pleroma-configuration-example/9871/8

Thanks for the feedback.

Addressed and forced push.

Just forced push a fix for some issues I noticed:

1. Add aarch64 support. It's a bindist, we need to fetch a different distribution dependending of the target arch (for x compilation). Thanks to Flokli for the help.
2. Added a read-only option pointing to the state directory.
3. Added a guard making sure we're running on a aarch64 or x86-64 system for the tests.

Hey there, I'm running a source-built version of Pleroma since a few months. A binary version is already a big improvement, but also feel free to use my source-built version if you can make it fit:
https://git.petabyte.dev/petabyteboy/nixfiles/src/branch/master/pkgs/pleroma

Waaat, that's great news!

Can I cherry pick your source build and default to it instead of this pleroma bindist half measure?

I see v2.2.0 has just been released I'll bump that this weekend.

I bumped my package to 2.2.0, but it's still running on Elixir 1.10, because it complained about import_config statements with Elixir 1.11 despite my config file not containing any (probably caused by elixir-lang/elixir@b227069).
It might be problematic to upstream my source-built package, because it relies on some unmaintained Elixir tooling that hasn't made its way to nixpkgs for a reason: For example it currently uses builtins.fetchGit, which will not work on Hydra. Someone would need to change mix2nix to prefetch the Git repositories, insert the hash into mix.nix and use fetchgit instead of builtins.fetchGit.
I'd be happy to make changes and take care of updating the Pleroma package in the future, but I'm not able to maintain the new Elixir tooling that would be required to add source-built Pleroma.

Bumped the PR to pleroma latest 2.2.0 release.

Ack @petabyteboy . To be franc, I don't really have any personal use of the elixir/otp ecosystem outside of Pleroma. I'm not really willing to invest a lot of the little hobby time I have in this ecosystem for now.

Let's move on with the bindist for now and see if somebody is willing to improve the mix.nix situation.

I wish I discovered your repo sooner, it'd have been very helpful. I just posted a comment on #62820 pointing to your source build to help with discoverability (in the secret hope to trigger more interest in upstreaming that :) ).

Dear reader, feel free to review/test further this PR ;)

Looks good to me, would love to see it merged :)

I'm not sure what happened on ofborg here. The x86 test & aarch64 tests ran successfully but the x86 pleroma-otp build seemed to have failed. It makes no sense to me: I'm running this PR on my Pleroma instance running on a x86 machine.

I've been running this PR for the last 3 months on my personal Pleroma instance without any issue. This PR also received an approval.

I'm going to merge merge this in a week from now. If you have any objections, do holler now.

Hello, I just tried to build with the files and got an interesting error

building '/nix/store/mq7iw7yjpqqbs6pxihcqmx5jdzdc7gm7-download.drv'...

trying https://git.pleroma.social/pleroma/pleroma/-/jobs/175284/artifacts/download
  % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current
                                 Dload  Upload   Total   Spent    Left  Speed
100 55.9M  100 55.9M    0     0   9.7M      0  0:00:05  0:00:05 --:--:-- 12.4M
hash mismatch in fixed-output derivation '/nix/store/wivbhgbr5l9izc9nm13dch4b2qg7wk75-download':
  wanted: sha256:0z93qvyda5kh794svk00pgbzhm5pv663rjkzkwwb6sfk7f9287ir
  got:    sha256:1c6l04gga9iigm249ywwcrjg6wzy8iiid652mws3j9dnl71w2sim

nix-prefetch-url gives the same result:

# nix-prefetch-url https://git.pleroma.social/pleroma/pleroma/-/jobs/175284/artifacts/download
[55.9 MiB DL]
path is '/nix/store/wivbhgbr5l9izc9nm13dch4b2qg7wk75-download'
1c6l04gga9iigm249ywwcrjg6wzy8iiid652mws3j9dnl71w2sim

Can you confirm?

Yup, I can confirm, thanks for the heads up!

I forgot or failed the copy paste while updating to 2.2.2 on x86_64. The FOD did not fail on my computer since I had 2.2.1 with the correct hash in my nix-store. Maybe Ofborg was failing because of hash mismatch in the end.

Thank you! Looking forward to the merge :)

You might be interested to know that an initial effort has been started to package mix project.
#112477
It's not quite a mix nix, as it is a little more simplistic, but it should be available faster.
I wish I had known about this PR earlier, there are lots of interesting things, great work!