| 9 Apr 2022 |
 tilpner | I don't need changes to the application to trigger rebuilds of the system in this instance, as I can poll-deploy the latest evaluation of the application to a separate nix profile and restart the corresponding systemd service. But that's not a general solution, and wouldn't support e.g. keeping a nixos module for the application in the application flake, because the system flake jobset couldn't automatically update itself to use that new module | 14:06:19 |
| tilpner | (Yes, keeping and automatically applying a nixos module from an application module is a security issue if different sets of people have access to the application and system repositories) | 14:07:39 |
| tilpner | * (Yes, automatically applying a nixos module from an application module is a security issue if different sets of people have access to the application and system repositories) | 14:15:18 |
 @grahamc:nixos.org | we should merge that PR | 18:58:40 |
| 10 Apr 2022 |
 @ulrikstrid:matrix.org | If I want to start playing with hydra, is it best to setup everything on the same machine and then add more builders later? | 12:47:49 |
| tilpner | I don't know about best, but running hydra without remote builders is perfectly fine | 12:48:33 |
| @ulrikstrid:matrix.org | Best was the wrong word, easiest is what I was looking for 😅 | 12:49:16 |
| @ulrikstrid:matrix.org | Maybe another stupid question, can I have the project config separate from the repo I'm building? | 12:49:56 |
| tilpner | that said, I do have localhost registered as a remote builder, just... a local remote builder :P | 12:50:31 |
| tilpner | what do you mean by project config? | 12:51:32 |
| @ulrikstrid:matrix.org | What and how to build a project, I'm not sure about the wording of hydra | 12:52:06 |
| tilpner | The "what projects to build" can be configured declaratively in a jobsets repository: https://hydra.nixos.org/build/172143314/download/1/hydra/plugins/declarative-projects.html | 12:53:59 |
| tilpner | But I don't think that's what you meant, as the non-declarative jobset configuration wouldn't be part of the repo anyway | 12:54:20 |
| @ulrikstrid:matrix.org | So to simplify: can I build any repo with a nix configuration? | 12:55:08 |
| @ulrikstrid:matrix.org | And, does the repo have to know about hydra? | 12:55:30 |
| tilpner | No, the repositories don't have to know about hydra | 12:56:11 |
| tilpner | (but it might be more direct if they do, as otherwise you may have to specify hydraJobs in a separate repo which pulls in the first repo) | 12:57:47 |
| @ulrikstrid:matrix.org | I basically want to play with hydra for some repos before committing to it | 12:58:26 |
| tilpner | before committing to hydra, or before committing to the repos? | 12:58:55 |
| @ulrikstrid:matrix.org | Especially if I want to introduce it at work | 12:58:59 |
| @ulrikstrid:matrix.org | Before committing to hydra | 12:59:03 |
| tilpner | you can always fork the repo and register the fork with hydra? | 12:59:40 |
 tomberek | Ulrik Strid: sure... you can create an "external expression" where the Nix lives apart from the source code. As opposed to an "internal expression" which would live in the same repo. | 13:00:08 |
| @ulrikstrid:matrix.org | Actually, can I do it at any branch? | 13:00:16 |
| @ulrikstrid:matrix.org | I have commit rights/own the repo(s) I want to use this on | 13:01:24 |
| tilpner | yes, you can specify the branch while fetching an input | 13:01:35 |
| @ulrikstrid:matrix.org | Okay, nice 👌 I'll see if I can find some time this evening to play with it | 13:02:23 |
| tilpner | I have an application flake with one derivation for building in staging mode and one for building in production mode. This flake exists on two different branches (you guessed it, staging and production), which I'd both like to deploy automatically. It's only a slight annoyance, but hydra build both branches in both modes, even though the staging build of the production branch and the production build of the staging branch won't be needed. If I could pass arguments to a flake jobset, I could conditionally present a different hydraJobs attrset, but AFAIK there's no way to pass arguments to a flake jobset. Is there an easy workaround here? | 13:14:00 |
| tilpner | * I have an application flake with one derivation for building in staging mode and one for building in production mode. This flake exists on two different branches (you guessed it, staging and production), which I'd both like to deploy automatically. It's only a slight annoyance, but hydra builds both branches in both modes, even though the staging build of the production branch and the production build of the staging branch won't be needed. If I could pass arguments to a flake jobset, I could conditionally present a different hydraJobs attrset, but AFAIK there's no way to pass arguments to a flake jobset. Is there an easy workaround here? | 13:19:31 |
| tomberek | no, no inputs from the outside. But a flake can do something like check the existence or contents of another file: builtins.pathExists. Perhaps something like hydraJobs = if builtins.pathExists x then import x self else null; or something like that | 13:22:09 |
