| 22 Dec 2025 |
 Randy Eckenrode | Actually, if I had to blame anything, it’s probably UB-related. ld64 can’t be built with libc++ hardening enabled. It’s pretty sloppy with what it does with pointers. Running it with UBSan prints a ton of issues.
Another possible issue is a couple of parts are parallelized. I wouldn’t be if there are races in there.
| 14:04:00 |
| Randy Eckenrode | Oh, and it uses recursion. That’s why it requires a massive stack and why GHC would crash it sometimes when we used the default stack size. | 14:04:38 |
| Randy Eckenrode | I mean, it randomly crashes under Rosetta 2. Switching to LLD can’t happen soon enough. | 14:05:51 |
 ghpzin | Just sent liblqr1 fix for darwin (that I broke fixing it for gcc15):
https://github.com/NixOS/nixpkgs/pull/473298
Would be nice to test whether it indeed builds fine on darwin (I only tested with clangStdenv). | 14:26:59 |
 Yureka (she/her) | pkgsMusl.stdenv fix (it's a world rebuild) | 20:06:02 |
| Yureka (she/her) | https://github.com/NixOS/nixpkgs/pull/473403 | 20:06:05 |
| ghpzin | Doesn't need second bootsrap update (aarch64 one) ? | 20:13:32 |
| ghpzin | Doesn't need second bootsrap update (aarch64 one) ?
(asked in PR) | 21:03:46 |
| 25 Dec 2025 |
 Fabián Heredia | Feliz navidad @ everyone :3 | 03:54:13 |
| Fabián Heredia | * Feliz navidad / nochebuena @ everyone :3 | 03:54:25 |
 K900 | Hit restart all failed on staging-next | 07:38:07 |
 Marcus | Wonder how hard it would be to fix hydra to do proper paging on still failing instead 'show me the rest' | 09:45:06 |
 Vladimír Čunát | Yeah, we have impractically many jobs nowadays. nix-review-tools-reports pages are also quite heavy. | 10:00:04 |
| Marcus | Even if I implement paging in the tt code, it would do a full backend request on every page render due to the diffing against previous eval. But it'd be easier on the browser at least. | 10:13:51 |
| K900 | The entire thing is a mess and needs a full rework tbh | 10:20:51 |
| Marcus | Don't disagree with that. It's using some of our worst ideas from Catalyst development as well 🙂 | 10:22:21 |
| 26 Dec 2025 |
|  isabel changed their profile picture. | 11:38:05 |
|  zitrone 🍋 changed their display name from zitrone to zitrone (39c3 DECT ZITR/9487). | 12:19:22 |
| Vladimír Čunát | Uh, still lots of failures on staging-next, around 18k instead of 5k from unstable. | 13:30:38 |
| Vladimír Čunát | Part of that is the exchange of python 3.14 instead of 3.12, and lots of 3.14 packages being broken. But even without that there are thousands of linux regressions:
https://hydra.nixos.org/eval/1821437?filter=-linux&compare=unstable#tabs-now-fail :-/ | 13:35:15 |
