| 2 Sep 2021 |
 Florian | W3F - OoO Mon/Tue | In reply to @emmanuelrosa:matrix.org
OK. The problem I see is that it can differ from the Nixpkgs that's pinned at /etc/nixos. I set the NIX_PATH using nix.nixPath, which I was setting to a local copy of Nixpkg. However, nixpkgs is an input to the Nixos configuration (in the flake.nix), so I tried this nix.nixPath = [ "nixpkgs=${nixpkgs}" ]; and now NIX_PATH is being set to a copy of Nixpkgs in the Nix store :) I'm not sure if that path would get garbage collected, but I'll play with this idea some more. If combined with deleting all channels and setting nix.registry.nixpkgs.flake = nixpkgs; you should be all set to use your systems pinned nixpkgs for everything. | 07:33:05 |
| Florian | W3F - OoO Mon/Tue | In reply to @emmanuelrosa:matrix.org
I use nix-shell to create bash scripts using the -i intepreter flag. But I don't see a way to do the same thing with nix shell. Right now, I have NIX_PATH so that nix-shell continues to work, now that I'm using Nix flakes to build NixOS. But of course that means it can point to a different Nixpkgs. What's the Flakes equivalent of nix-shell -i bash? I have not tried this, but maybe nix run nixpkgs#bash might work. | 07:33:48 |
| Florian | W3F - OoO Mon/Tue | In reply to @emmanuelrosa:matrix.org
I use nix-shell to create bash scripts using the -i intepreter flag. But I don't see a way to do the same thing with nix shell. Right now, I have NIX_PATH so that nix-shell continues to work, now that I'm using Nix flakes to build NixOS. But of course that means it can point to a different Nixpkgs. What's the Flakes equivalent of nix-shell -i bash? * I have not tried this, but maybe nix run nixpkgs#bash -- -i might work. | 07:34:11 |
| Florian | W3F - OoO Mon/Tue | * I have not tried this, but maybe nix run nixpkgs#bash -- -i might work. (EDIT: I played around with this, could not make it work with some simple examples. Looks like nix-shell does something special when called as interpreter by the kernel). | 07:42:21 |
| Florian | W3F - OoO Mon/Tue | There is some work on shebang support though in #5189 | 07:46:41 |
|  papojari joined the room. | 15:02:21 |
| 3 Sep 2021 |
|  humancalico joined the room. | 06:34:29 |
 emmanuelrosa | Is there a way to update only a specific input in a flake lock file? | 12:25:47 |
 nan-matrix | In reply to @emmanuelrosa:matrix.org
Is there a way to update only a specific input in a flake lock file? nix flake lock --update-input ... | 12:26:42 |
 edrex | something seems hosed with my local nix store, I think related to switching from 2.3 to 2.4pre (i should have switched immediately).
https://clbin.com/zyElX | 17:58:10 |
| edrex | ok i think my mistake is I installed nixUnstable to my normal user profile and it didn't update the daemon? but now I'm getting a weird error trying to update the daemon: https://clbin.com/CDYla | 18:09:55 |
| edrex | i'm guessing my issues may have stemmed from using the 3.0pre installer linked at https://nixos.wiki/wiki/Flakes#Non-NixOS
| 18:45:14 |
| edrex | Wiped everything and did a simple single user install. | 18:45:32 |
| edrex | Unrelated question: without lots of overhead managing binary caches on both flake producer and consumer side, moving towards a lot more 3rd-party flakes is going to involve a lot more compiling. Is there any work in the community (or core) on making cache management somehow easier for third party flakes? | 18:48:53 |
| edrex | I guess a central build cache for 3rd party flakes is bad since it would allow unverified code in | 18:52:07 |
| edrex | is there something in the flake spec that allows the flake author to specify a cache? | 18:52:08 |
 tomberek | Yes. It’s a lot to go over. But yes, there are mechanisms to address this. | 21:27:46 |
| 4 Sep 2021 |
 bew | In reply to @edrex:matrix.org
is there something in the flake spec that allows the flake author to specify a cache? https://github.com/NixOS/nix/pull/4189 ? | 05:13:09 |
| bew | There's also https://github.com/NixOS/nix/issues/3824 | 05:38:56 |
| bew | Which adds the loading of a local nix.conf file | 05:39:22 |
| edrex | In reply to @bew:matrix.org
https://github.com/NixOS/nix/pull/4189 ? that was merged to master in Nov, awesome! | 19:16:11 |
| edrex | is there a doc describing the Nix branching strategy somewhere? (wondering how commits to master make it to a release) | 19:16:47 |
| edrex | i imagine lots of cherry picking | 19:17:06 |
 Roos | IIRC, you open an MR against master and if you want it backported you open an MR against the release branch. | 19:44:09 |
| Roos | I imagine that release branches never get merged back to master and master is forked prior to ZHF upon the next release. | 19:44:57 |
| edrex | are release branches tagged directly off current master? | 19:50:26 |
| edrex | err, branched rather | 19:50:34 |
| edrex | it looks like releases are tagged off the latest-release branch, which seems to keep a persistent history parallel to master | 19:56:46 |
| edrex | ah, i think it starts from master with each new minor release, ie
git show $(git merge-base master origin/latest-release) -> Sep 4 2019 | 20:01:10 |
| 5 Sep 2021 |
 David Arnold (blaggacao) | Releases are actually "release tracks" which can suffer continous bugfixing and the likes, hence they are modled as branches. | 22:57:42 |
