| 7 Jun 2022 |
cole-h | Could you say more...? | 15:08:00 |
Sandro | The ofborg check which checks for aliases is green but we are clearly using aliases in the tree. I know that this happens for unfree packages which already got tried to fix multiple times but there are also free packages in there and modules. | 15:12:46 |
Sandro | Maybe this is also dead code but I would expect ofborg to catch such cases. Maybe we need to evaluate the entire documentation? | 15:13:27 |
Alyssa Ross | sandro: I don't think that's an OfBorg problem, really — the code in NixOS modules is largely conditional on module options, so to be guaranteed to catch aliasing violations you'd have to evaluate every possible combination of NixOS options | 15:29:22 |
Alyssa Ross | in fact, as soon as stuff like with is involved, I'd assume it would be undecidable | 15:30:03 |
Sandro | hmmm... you're probably right. Would be cool if we could find a way around this | 15:31:01 |
Alyssa Ross | I think what we're looking for is coverage checking | 15:31:39 |
Alyssa Ross | but I think that would have to be integrated into Nix | 15:31:53 |
Alyssa Ross | then you could see that these lines aren't evaluated by OfBorg, and try to add something that would evaluate them | 15:32:13 |
Sandro | that would be really useful | 15:32:29 |
7c6f434c | Rest assured it is undecidable even without with whether a line of code is dead or not… Coverage tracker sounds useful, yes. I wonder what fraction of the dead code would come from optional feature flags never set in Nixpkgs, of course. | 19:48:49 |
dotlambda | ofborg also didn't flag the usage of an alias in https://github.com/NixOS/nixpkgs/pull/176734 | 21:29:02 |
| 9 Jun 2022 |
dotlambda | ofborg didn't flag the usage of an alias in https://github.com/NixOS/nixpkgs/pull/175637 but nixpkgs-review does detect it | 00:13:06 |
Sandro | *I think nixpkgs-review only uses it for one eval and it is a bit buggy | 07:37:21 |
Sandro | Best to turn aliases of in config. Nix | 07:37:21 |
| 21 Jun 2022 |
ncfavier | running @ofborg test networking wireguard gives instant success with a weird 404 log, while @ofborg build nixosTests.networking nixosTests.wireguard is fine | 21:13:21 |
ncfavier | @ofborg test networking.networkd.dhcpSimple is fine so it's probably a problem with attrsets | 21:15:13 |
ncfavier | looks like https://github.com/NixOS/ofborg/issues/504 | 21:18:30 |
| 22 Jun 2022 |
hexa | indeed, ofborg seems to be behaving weirdly for tests 😕 | 14:34:21 |
hexa | many queued jobs, all unfinished | 14:35:50 |
| 25 Jun 2022 |
Sandro | ofborg logs are currently not loading for me | 22:25:59 |
Sandro | the websocket is only sending empty 1 bytes messages after the subscribe | 22:26:46 |
| 27 Jun 2022 |
| tinybronca joined the room. | 08:21:13 |
| 28 Jun 2022 |
| ma27 changed their profile picture. | 11:38:05 |
| 29 Jun 2022 |
Sandro | How did it happen that we lose aarch64-darwin? | 12:28:04 |
Alyssa Ross | hmm, OfBorg doesn't seem to be noticing my @OfBorg build comments here https://github.com/NixOS/nixpkgs/pull/179556#issuecomment-1170407607 | 20:17:29 |
| 7 Jul 2022 |
Winter (she/her) | Is there a reason that failures during test eval are reported as neutral and not failing? | 04:17:39 |
7c6f434c | Yes. | 04:18:02 |
7c6f434c | There are flaky tests, failed-before-anyway, disk full issues, timeouts… | 04:18:49 |
7c6f434c | (also complicated issues with some subset of the builders…) | 04:19:26 |