| 28 Dec 2025 |
 leona | then I agree | 08:33:18 |
 hexa | uvloop is stubbornly flaky | 08:33:34 |
| hexa | but it just built on hydra on aarch64-darwin | 08:33:41 |
| hexa | now it fails on numpy | 08:33:53 |
| hexa | numpy is cached | 08:34:19 |
| hexa | nvm | 08:34:21 |
| hexa | ok, I'm disabling a test on uvloop | 09:30:54 |
| hexa | * ok, I'm disabling a test on uvloop on darwin | 09:30:57 |
| hexa | not sure why it appeared cached to me, maybe I got lucky | 09:31:11 |
| hexa | https://github.com/NixOS/nixpkgs/pull/474758 | 09:41:09 |
 Grimmauld (any/all) | with c23, some things (e.g. nasm) fail in a creative way: They check whether stdbool.h is available, and include it if so. Which obviously breaks on C23, where bool can't be typedefed. No idea why this didn't break on glibc, i didn't look at that 🤷 | 19:22:55 |
| 29 Dec 2025 |
 @ghpzin:envs.net | typedefs in stdbool.h are conditional on __STDC_VERSION__ > 201710L
So you can safely include it there.
In musl version they are not. | 03:04:01 |
| @ghpzin:envs.net | typedefs in stdbool.h are conditional on __STDC_VERSION__ > 201710L
So you can safely include it there.
In musl version they are not:
https://git.musl-libc.org/cgit/musl/tree/include/stdbool.h | 03:06:02 |
| @ghpzin:envs.net | true/false in stdbool.h are conditional on __STDC_VERSION__ > 201710L
So you can safely include it.
In musl they are not:
https://git.musl-libc.org/cgit/musl/tree/include/stdbool.h | 07:58:27 |
| @ghpzin:envs.net | true/false in stdbool.h are conditional on __STDC_VERSION__ > 201710L:
https://github.com/gcc-mirror/gcc/blob/1b995214830669b96a19a6b6463aa6c1647cea9b/gcc/ginclude/stdbool.h#L33
So you can safely include it.
In musl they are not:
https://git.musl-libc.org/cgit/musl/tree/include/stdbool.h | 07:59:00 |
| @ghpzin:envs.net | true/false in stdbool.h are conditional on __STDC_VERSION__ < 201710L:
https://github.com/gcc-mirror/gcc/blob/1b995214830669b96a19a6b6463aa6c1647cea9b/gcc/ginclude/stdbool.h#L33
So you can safely include it.
In musl they are not:
https://git.musl-libc.org/cgit/musl/tree/include/stdbool.h | 07:59:31 |
 juliusfreudenberger | Go 1.24.11 would be there, which is required for teleport. But it is not security relevant, so I might just skip this last update for oldstable as well. | 14:26:48 |
 Vladimír Čunát | The current release-25.05..staging-25.05 will probably never get merged anymore. | 14:29:38 |
| Vladimír Čunát | (i.e. no staging-next-25.05 cycle anymore) | 14:31:07 |
| juliusfreudenberger | That's what I already figured. Thanks for clarifying nonetheless! | 14:36:35 |
| Vladimír Čunát | staging-next-25.11 was running builds in backgroud for a few days already. Now I opened a PR:
https://github.com/NixOS/nixpkgs/pull/475079 | 15:53:15 |
| Vladimír Čunát | * staging-next-25.11 has been running builds in backgroud for a few days already. Now I opened a PR:
https://github.com/NixOS/nixpkgs/pull/475079 | 15:54:50 |
| 30 Dec 2025 |
| Vladimír Čunát | I'm not sure about staging-next. While I'm not aware of any particular blockers, the amount of regressions does remain pretty high.
One way to look is https://hydra.nixos.org/eval/1821537?compare=unstable
i.e. 4.3k jobs newly failing - 1.1k newly succeeding. If we wanted to take into account the rotation of non-default python version (which appear as new and removed packages in there), we could look at total successes and failures.
Successes decrease by 15k (!) and failures increase by 7k.
| 07:43:57 |
| Vladimír Čunát | (PRs currently open against staging-next seem to fix at most a couple hundred jobs) | 07:46:45 |
| Vladimír Čunát | * I'm not sure about staging-next. While I'm not aware of any particular blockers, the amount of regressions does remain pretty high.
One way to look is https://hydra.nixos.org/eval/1821537?compare=unstable
i.e. 4.3k jobs newly failing - 1.1k newly succeeding. If we wanted to take into account the rotation of non-default python version (which appear as new and removed packages in there), we could look at total successes and failures.
Successes decrease by 17k (!) and failures increase by 7k.
| 07:55:51 |
 K900 | I see a lot of potentially flaky stuff in the reports | 07:59:53 |
| K900 | twisted uvloop etc | 07:59:56 |
| Vladimír Čunát | They've been retried a few times already. | 08:00:37 |
| Grimmauld (any/all) | didn't hexa disable tests on uvloop darwin? | 08:00:54 |
| Vladimír Čunát | If they're that much flaky, I consider them as good as always failing. | 08:00:56 |
