| 19 Jan 2025 |
 K900 ⚡️ | But it's funny how presumably the systemd update just makes it too fast now | 12:08:08 |
 m1cr0man | Is there a way for me to get notified on every hydra failure of the test? | 12:26:40 |
| K900 ⚡️ | No | 12:27:05 |
| m1cr0man | What annoys me the most is that in practice the module is stable, yet the test suite is forever flaking on things that I would expect to break in the wild also. I think at this point I need to rewrite the whole thing, and deeply analyze all the systemd service relation chains. | 12:42:49 |
| m1cr0man | I might break the test suite up by web server, and factor out all that code which generates a test suite for each one into its own file. That should allow the test to be further paralleled, and allow isolation of failures | 14:25:50 |
| m1cr0man | Infact generally breaking up the test by test case might be a good idea for debugging | 14:26:03 |
| m1cr0man | * I might break the test suite up by web server, and factor out all that code which generates a test suite for each one into its own file. That should allow the test to be further parallelised, and allow isolation of failures | 14:28:56 |
| m1cr0man | Posted a plan here https://github.com/NixOS/nixpkgs/issues/374792#issuecomment-2601075090 | 23:50:16 |
| 21 Jan 2025 |
| m1cr0man | What's the audience's feeling on sentinel values vs new options? https://github.com/NixOS/nixpkgs/issues/370825#issuecomment-2599645510 | 21:39:26 |
| m1cr0man | There's good arguments for both. I'm happy to implement either. | 21:39:52 |
| 28 Jan 2025 |
 Arian | Morning | 13:06:54 |
| Arian | Heard about the hydra flakiness. That sucks. | 13:07:03 |
| Arian | I am currently very low bandwidth on nixpkgs issues that are not directly related to my day job unfortunately. And we don't use acme anymore | 13:07:27 |
| Arian | But happy to review anyhing | 13:07:35 |
| 29 Jan 2025 |
| K900 ⚡️ | Test is flaking a lot today again :( | 07:50:47 |
| K900 ⚡️ | It seems like it's just correlated with machine loa | 07:50:57 |
| K900 ⚡️ | * It seems like it's just correlated with machine load | 07:50:58 |
| Arian | :(( | 13:02:12 |
| Arian | Nuclear option: do we wanna disable the tests on at least unstable for now? | 13:02:26 |
| Arian | I feel bad for this being a channel blocker | 13:02:32 |
| K900 ⚡️ | I don't think we should | 13:19:47 |
| K900 ⚡️ | If it actually breaks, we'll get a bunch of people offline | 13:20:02 |
| K900 ⚡️ | I'm fine kicking it every now and then to make sure that doesn't happen | 13:20:22 |
| m1cr0man | I nearly have the test suite rewritten - working on webserver test isolation now. It will be a lot more reliable, and we can disable tests piecemeal instead of disabling the whole suite if it gets flakey again. | 17:28:03 |
| K900 ⚡️ | ❤️ | 18:03:25 |
| m1cr0man | Are you KIDDING me? There's an option on pebble that sets a percentage failure for cert validation? https://github.com/letsencrypt/pebble?tab=readme-ov-file#invalid-anti-replay-nonce-errors | 19:04:31 |
| m1cr0man | It's been in there for 8 years apparently 🫠 probably not the source of the main problems but still, I've disabled it | 19:07:39 |
| K900 ⚡️ | Uhh | 19:13:13 |
| K900 ⚡️ | I think that's a good thing actually? | 19:13:17 |
| K900 ⚡️ | It seems useful to verify lego behaves correctly in that case | 19:13:30 |
