| 21 Oct 2023 |
 schmittlauch (he/him) | * The easiest workaround would probably be to use the poetry2nix from the nixpkgs input instead of the additional poetry2nix flake input 😬 | 23:10:45 |
| 22 Oct 2023 |
 cpcloud | Re nixpkgs-unstable, I was able to get the "simple" wheel tests passing by using pypaInstallHook instead of pipInstallHook. The latter causes infinite recursion (of the flavor which --show-trace can help with) | 10:08:42 |
| schmittlauch (he/him) | In reply to @schmittlauch:ohai.is
I could move my configured instanciation of nixpkgs to a separate flake, and export the configured version as an output from there. Then in my main flake, I set poetry2nix to follow that configured flake. Update: Configuring and re-exporting nixpkgs in an external flake did the trick. I'll post details in an issue about this. | 13:19:21 |
| cpcloud | Turns out that the new bootstrapping infrastructure is actually enforcing build requirements now. I am currently sifting through the failing packages' requirements and unpinning them in postPatch. No promises, but I think I am on the way to resolving the bootstrapping issue. | 14:05:36 |
 K900 | Yes, that is also good | 14:09:13 |
| cpcloud | Part of the solution so far is to inherit the build, packaging, and installer packages from nixpkgs | 14:22:09 |
| cpcloud | which is what we are currently doing for wheel | 14:22:27 |
| K900 | Probably also flit-core | 14:22:32 |
| K900 | As it's part of the bootstrap chain too | 14:22:38 |
| cpcloud | oh yeah, sorry that one too | 14:22:40 |
| cpcloud | i was going from memory while waiting for mypy to build :) | 14:22:56 |
| cpcloud | wow, pip has a dependency on sphinx apparently | 14:26:28 |
| K900 | Yep | 14:27:25 |
| K900 | A full-sized pip depends on like half the world | 14:27:34 |
| cpcloud | because of all the vendoring? | 14:27:44 |
| K900 | Mostly because of sphinx honestly | 14:29:15 |
| cpcloud | well, i have it so that pip can be in build-systems.json now, which simplifies these packages that declare it in their build requirements | 14:30:00 |
| K900 | Uhh | 14:30:27 |
| cpcloud | seems to be a bit less magic around bootstrapping python in nixpkgs now | 14:30:27 |
| K900 | Who the fuck | 14:30:30 |
| K900 | Does that | 14:30:32 |
| cpcloud | https://github.com/ansible/molecule/blob/3.5.2/pyproject.toml#L3 | 14:31:34 |
| cpcloud | And a few others, but they all seem to be in the closure of this absolutely cursed ansible-moleule package | 14:32:26 |
| K900 | Should have guessed | 14:32:31 |
| cpcloud | I honestly have no idea if what I am doing is going to be backwards compatible with older nixpkgs | 14:34:06 |
| cpcloud | * And a few others, but they all seem to be in the closure of this absolutely cursed ansible-molecule package | 14:35:39 |
| K900 | No shot | 14:36:42 |
| K900 | Will probably need a separate 23.05 branch | 14:36:54 |
| K900 | And then just kill it in a few months | 14:37:00 |
| cpcloud | Ok, I'm down to 14 tests that won't evaluate, all with the same error: ``` | 14:41:55 |
