| 8 Jun 2026 |
 maralorn | 
Download image.png | 10:58:46 |
| maralorn | ^ Everyone have a look at the fucking awesome integration tests, that Primrose (== Léana) wrote at zurihac. | 10:59:14 |
| maralorn | 
Download image.png | 10:59:31 |
| maralorn | Immediately catched a regression caused by a change in the output of newer nix versions. | 11:00:02 |
 Primrose (== Léana) | That’s interesting because it doesn’t catch the regression with lix (we need to change the test to streaming) | 11:17:48 |
 K900 | Have you figured out what's causing the regression? | 11:18:43 |
| K900 | That's very cool | 11:18:48 |
| maralorn | The "regression" only affects the human-readable nix-output parser. Which was a bad idea from the beginning. Nix simply changed how it formats it’s "build failed" message. Lix doesn’t have that "regression" because it didn’t adopt that patch I assume. | 11:28:22 |
| maralorn | I will look into it more closely shortly when I fix it. | 11:28:36 |
| Primrose (== Léana) | I'm thinking maybe it would be interesting to split the oldstyle tests from the newstyle tests?
Also adding a proper logging mechanism to the test suite can be useful, we can copy the test result to $out, which will help debugging CI tests | 11:46:06 |
| Primrose (== Léana) | * I'm thinking maybe it would be interesting to split the oldstyle tests from the newstyle (json) tests?
Also adding a proper logging mechanism to the test suite can be useful, we can copy the test result to $out, which will help debugging CI tests | 11:46:14 |
| maralorn | You mean having different cabal components for that? | 11:57:36 |
| maralorn | Yes, better logging would improve a lot. | 12:00:21 |
| Primrose (== Léana) | Not necessarily, but a way to pass an argument to test only old style and new style, so we separate the execution | 12:01:13 |
| Primrose (== Léana) | In CI you can then glance and see which one failed, but I'm not sure how much this is worth the hassle. | 12:01:32 |
| maralorn | Yeah, in theory it would improve the CI to basically forward the test granularity to a matrix of workflow jobs. So that you can see in CI immediately "old-style parsing failed on nix 2.34". | 12:21:01 |
| maralorn | But otoh, it’s now only a few clicks to figure out. | 12:21:13 |
| Primrose (== Léana) | True, I'll keep things the current way then | 12:21:39 |
| maralorn | If you like to do it, I wouldn’t be exposed. | 12:22:43 |
| maralorn | * If you like to do it, I wouldn’t be opposed. | 12:22:48 |
| maralorn | But I don’t think it’s high priory. | 12:22:56 |
| maralorn | * But I don’t think it’s high priority. | 12:22:58 |
|  @appleptree:private.coffee left the room. | 12:37:50 |
| maralorn | Primrose (== Léana): I kind of remember that the VM tests took about 10s on your machine. On mine they take at least a minute. Do you have that much more power or do I maybe have something misconfigured? | 12:39:30 |
 hexa | have you looked at container tests? | 12:43:00 |
| maralorn | No | 12:46:36 |
| maralorn | Do they have an independent nix-store/ nix-daemon? | 12:46:49 |
| maralorn | * No (at least not me) Primrose (== Léana) is the expert here. 😄 | 12:47:12 |
| Primrose (== Léana) | What containers are we talking about? | 12:47:34 |
| hexa | nixos test containers | 12:48:00 |
