| 8 Apr 2022 |
 tilpner | yeah, that's fair :) | 21:06:32 |
 ma27 | In reply to @tilpner:tx0.co
I have private = true in my config, and it seems to return a 403 unless signed in for every page I implemented https://github.com/NixOS/hydra/pull/919 for private projects a while ago | 21:09:24 |
| ma27 | In reply to @tilpner:tx0.co
Hi ma27, which plugin did you write https://github.com/NixOS/hydra/pull/922 for? You mention e.g. the gitea status plugin, which (as you pointed out) requires a git input, and then to set gitea_status_repo to the name of that input. Is there some hidden flake input, or is that still incompatible with the flakes mode? no it isn't. I add a second git input, so each repo is cloned twice. Dirty, but it gets the job done for now. | 21:10:14 |
| tilpner | ohh! I assumed "Jobset Inputs don't take any effect for flakes" meant that any non-String inputs would just be ignored, I'll try adding a second input for now | 21:11:51 |
| ma27 | see for instance https://hydra.ist.nicht-so.sexy/jobset/nvim-nix/master#tabs-configuration and the inputs configured for that jobset in https://hydra.ist.nicht-so.sexy/eval/699#tabs-inputs. | 21:14:00 |
| tilpner | In reply to @ma27:nicht-so.sexy
see for instance https://hydra.ist.nicht-so.sexy/jobset/nvim-nix/master#tabs-configuration and the inputs configured for that jobset in https://hydra.ist.nicht-so.sexy/eval/699#tabs-inputs. thank you for the reference! I can't seem to make hydra report any status yet, but that could be a misconfigured gitea token/user account, or these evaluation errors... | 21:41:54 |
| 9 Apr 2022 |
| tilpner | Are there any workarounds for automatic updates in flake mode, or does everyone with that requirement still use legacy mode and flake-compat? | 13:04:24 |
 Artturin | In reply to @tilpner:tx0.co
Are there any workarounds for automatic updates in flake mode, or does everyone with that requirement still use legacy mode and flake-compat? automatic updates of what? | 13:10:55 |
 Rick (Mindavi) | NixOS, I assume | 13:24:49 |
| tilpner | In reply to @artturin:matrix.org
automatic updates of what? Automatic updates of flake inputs | 13:37:17 |
 Amanda (she/her) | I use a systemd timer on my homelab that does a regular clone, nix flake update commit and push | 13:38:31 |
| tilpner | * Automatic updates of flake inputs, e.g. the source code of an application, or even nixpkgs inputs for a nixosConfiguration flake | 13:40:38 |
| Amanda (she/her) | If your useing GitHub you can probably make GitHub actions do it | 13:42:00 |
| Amanda (she/her) | But I want to self host my stuff so it's just a systemd timer that updates the flake lock every Wednesday at midnight, then my hydra will build it all so in the morning I can see if it's okay to update | 13:43:16 |
| Amanda (she/her) | If not, I can just revert the lock update | 13:43:44 |
| Amanda (she/her) | With flakes the inputs become part of the source code, so usual source code workflows apply | 13:45:00 |
| tilpner | I have considered that, but it seems odd to have add-on tasks outside of hydra for something that hydra could already do in legacy mode | 13:46:40 |
| tilpner | Also, I'd like much lower git-push-to-deployed latencies than would be possible with even hourly nix flake update timers | 13:47:45 |
| Amanda (she/her) | Flakes are meant to provide more reproducibility. I suppose hydra could have a mode to do the git updates for you, but that seems out of scope | 13:47:50 |
| tilpner | I understand automatic updates don't exactly fit the design of flakes, but this reems to be a recurring problem:
- https://www.reddit.com/r/NixOS/comments/js5fxv/using_hydra_with_flakes_without_lock/
- https://www.reddit.com/r/NixOS/comments/negjsu/people_who_use_nix_flakes_for_their_system_config/
| 13:48:21 |
| Amanda (she/her) | They don't just "not exactly" fit, they're pretty much antithetical | 13:49:15 |
| Amanda (she/her) | The idea with flakes is you can pull a git repo, and get exactly the same output no matter when you do it | 13:49:52 |
| tilpner | and I appreciate that property for manual builds and deployments, it's just with hydra where it becomes a problem for automatic deployments :) | 13:50:53 |
| tilpner | sure, I could whip up a small webhook receiver to automatically forward pushed in one repo to flake updates in another, but that feels very inelegant | 13:52:08 |
| Amanda (she/her) | You probably don't want stuff changing it from under you for automatic deployments, see: node and the recent "protestware" package updates | 13:52:08 |
| Amanda (she/her) | * You probably don't want stuff changing out from under you for automatic deployments, see: node and the recent "protestware" package updates | 13:52:25 |
| tilpner | I have two flakes, one for the system definitions, and one for an application. If I push a new commit to the application, I'd like hydra to build and check it, and then deploy it | 13:53:17 |
| tilpner | That would work fine even in flakes mode, if it were a monorepo. As both repos are under my control, I'm not worried about malicious package updates, though of course these automatic updates should be selective | 13:54:50 |
| tilpner | I suppose I could configure hydra to execute a script after a build of the application repo finishes, which pushes a commit to the systems repo to trigger a deployment of that | 13:57:20 |
| tilpner | (https://github.com/NixOS/hydra/pull/1103 would be neat for that) | 13:59:03 |
