20 Sep 2024 |
Mic92 | So you would need to change that | 05:23:56 |
aleksana (force me to bed after 18:00 UTC) | In reply to @joerg:thalheim.io I don't think your local nix version will change anything Understood, now I'm asking whether we should bump default version again now or wait a bit? | 05:24:30 |
Mic92 | aleksana (force me to bed after 18:00 UTC): if you could figure out if the error is reproducible by re-running hydra-eval-jobs, that would be useful. | 06:07:50 |
aleksana (force me to bed after 18:00 UTC) | In reply to @joerg:thalheim.io aleksana (force me to bed after 18:00 UTC): if you could figure out if the error is reproducible by re-running hydra-eval-jobs, that would be useful. That hydra issue is always reproducible on hydra, but not on Nix | 06:08:38 |
aleksana (force me to bed after 18:00 UTC) | Also just found another compilation error on git: > src/libfetchers/git-utils.cc:267:13: error: 'git_mempack_write_thin_pack' was not declared in this scope
> 267 | git_mempack_write_thin_pack(mempack_backend, packBuilder.get())
> | ^~~~~~~~~~~~~~~~~~~~~~~~~~~
| 06:09:12 |
aleksana (force me to bed after 18:00 UTC) | In reply to @aleksana:mozilla.org That hydra issue is always reproducible on hydra, but not on Nix We haven't reproduced it on Nix using the same configuration | 06:09:51 |
aleksana (force me to bed after 18:00 UTC) | Also puck was running a NixOS configuration when catching that bug, so we can hopefully guess there's no more similar bugs that could happen in Nix now | 06:11:04 |
aleksana (force me to bed after 18:00 UTC) | In reply to @aleksana:mozilla.org
Also just found another compilation error on git: > src/libfetchers/git-utils.cc:267:13: error: 'git_mempack_write_thin_pack' was not declared in this scope
> 267 | git_mempack_write_thin_pack(mempack_backend, packBuilder.get())
> | ^~~~~~~~~~~~~~~~~~~~~~~~~~~
Just forgot to patch libgit2 | 06:13:07 |
Mic92 | In reply to @aleksana:mozilla.org We haven't reproduced it on Nix using the same configuration You can reproduce it with hydra-eval-jobs at least, right? | 06:20:34 |
Mic92 | That should be sufficient | 06:20:38 |
aleksana (force me to bed after 18:00 UTC) | In reply to @joerg:thalheim.io You can reproduce it with hydra-eval-jobs at least, right? Yes | 06:31:35 |
aleksana (force me to bed after 18:00 UTC) | Bumping 2.24, 2.18 and git. The triggered build is a bit too much for me, so I haven’t run nixpkgs-review yet. https://github.com/NixOS/nixpkgs/pull/343187 | 06:44:10 |
Mic92 | aleksana (force me to bed after 18:00 UTC): merged | 10:27:00 |
Mic92 | https://github.com/NixOS/nixpkgs/pull/343233 | 10:37:18 |
Mic92 | Backport of the same | 10:37:24 |
K900 | https://github.com/NixOS/nix/issues/10202 | 16:45:24 |
K900 | This is probably also blocker-adjacent for reverting to 2.24 by default | 16:45:32 |
K900 | Maybe some sort of migration aid in nixos-rebuild could work | 16:45:40 |
Alyssa Ross | If people want /etc/nixos to be a git repo not owned by root, maybe they should just mark it as safe? | 16:48:39 |
Alyssa Ross | I'm not sure that's even a Nix bug | 16:48:44 |
K900 | It's not just /etc/nixos | 16:49:21 |
K900 | The more common failure case is sudo nixos-rebuild --flake ~/my-cool-nixos | 16:49:37 |
K900 | Which explodes in exactly the opposite way | 16:49:51 |
Alyssa Ross | right, but in general, is this just the thing where git wants you to list repositories in a config file that you want to access with a non-owner? | 16:49:52 |
K900 | Well yes, but maybe some better issue reporting could be good here | 16:50:14 |
K900 | Or some privilege wiggling in nixos-rebuild | 16:50:22 |
Alyssa Ross | Former sounds better | 16:50:29 |
emily | error reporting would be nice but doesn't seem like a blocker to me | 17:14:08 |
aloisw | Given that this is Git behaviour, why did it even regress with the Nix update? | 17:19:19 |
K900 | Because nix switched to libgit2 | 17:20:59 |