| 28 Jan 2026 |
 samasaur | allegedly still doable as of 2021, so I would give that a shot first | 01:02:48 |
 m.a. | that was my plan if it wasn't already implemented in nix-darwin. will probably do sometime tomorrow | 01:03:34 |
| m.a. | thanks for the help :) | 01:04:48 |
 Randy Eckenrode | There appear to have been some changes to CUPS in macOS 26. | 01:05:26 |
| Randy Eckenrode | Minor though. They added httpDeleteCredentials, which doesn’t appear in upstream CUPS. | 01:09:00 |
| Randy Eckenrode | Neat. The failure is non-deterministic. | 01:10:08 |
 Tristan Ross | Oh yeah, I ran into a non deterministic failure with aarch64-darwin today with x265 lol. It said a .o file had a bad size, the linker said 0 bytes. | 01:11:51 |
 Ihar Hrachyshka | so what's the latest on swift 5 being broken in unstable? I have half of my system tied to it, apparently - who new? (nixvim plugins, pre-commit, mpv are just first things that popped up in the chain of dependencies). looking at https://github.com/NixOS/nixpkgs/issues/483584 comments, seems like no specific plans or ideas to fix it? | 02:23:25 |
| Ihar Hrachyshka | * so what's the latest on swift 5 being broken in unstable? I have half of my system tied to it, apparently - who knew? (nixvim plugins, pre-commit, mpv are just first things that popped up in the chain of dependencies). looking at https://github.com/NixOS/nixpkgs/issues/483584 comments, seems like no specific plans or ideas to fix it? | 02:23:37 |
| Randy Eckenrode | I’m not looking at it. Time spent on that is time not spent on Swift 6.2. Someone else will need to debug what is causing the compiler to crash during build. | 02:31:14 |
| Randy Eckenrode | You’ll need to bisect across the last staging-next, which is where it broke. | 02:31:44 |
| m.a. left the room. | 02:35:11 |
| m.a. joined the room. | 02:35:30 |
| Tristan Ross | Oh great, I just ran into that swift bug.... | 04:18:18 |
| samasaur | I'm kicking off a bisect for that Swift bug with jj bisect run --range b579d443b37c9c5373044201ea77604e37e748c8..a1bab9e494f5f4939442a57a58d0449a109593fe -- nom build -A swift (commits grabbed from Hydra). I think it'll be a couple days before I can investigate but maybe that will give enough time for this bisect to finish, given that it will be rebuilding the world a lot | 06:57:29 |
| samasaur | (I haven't used JJ's bisecting before, so if anyone has and sees anything wrong with that command let me know. I think I got it right but I'm not entirely sure) | 06:58:01 |
| samasaur | * I'm kicking off a bisect for that Swift bug with jj bisect run --range b579d443b37c9c5373044201ea77604e37e748c8..a1bab9e494f5f4939442a57a58d0449a109593fe -- nom-build -A swift (commits grabbed from Hydra). I think it'll be a couple days before I can investigate but maybe that will give enough time for this bisect to finish, given that it will be rebuilding the world a lot | 06:58:58 |
| samasaur | I did not realize how much stuff in nixpkgs depends on Swift on Darwin | 07:04:50 |
| samasaur | oh also Randy Eckenrode do you know if there's anything weird you have to do to get llvm-symbolizer to work? I tried setting LLVM_SYMBOLIZER_PATH to lib.getExe' libllvm "llvm-symbolizer" but that didn't seem to get picked up | 07:08:58 |
| Randy Eckenrode | In the swift-frontend stack traces? I don’t think I’ve ever gotten it working. | 10:49:14 |
| Randy Eckenrode | Admittedly, I don’t think I tried more than setting the path in the terminal. | 10:49:58 |
| Ihar Hrachyshka | samasaur: in my experience bare automated bisects that don't consider WHY a failure happens are not very reliable in nixpkgs. it fail in "bison" or "sqlite" or whatever and assume that it's a sign of swift being broken. | 14:50:58 |
| Ihar Hrachyshka | * samasaur: in my experience bare automated bisects that don't consider WHY a failure happens are not very reliable in nixpkgs. it will fail in "bison" or "sqlite" or whatever and assume that it's a sign of swift being broken. | 14:51:18 |
| Ihar Hrachyshka | (writing a script that captures the cause may or may not be easy; if too hard or just lazy, I suspect a AI agent could be instructed to run the bisect loop for you... never tried it myself before though) | 14:55:38 |
 Austin Horstman | everytime swift breaks it feels catastrophic to my setup lol | 14:59:55 |
| Austin Horstman | mixed with firefox not building / caching my macbook is in a weird state lol | 15:00:50 |
| Ihar Hrachyshka | sorry for the latter. fwiw one can take firefox and thunderbird from release branches, they are the same version most of the time. | 15:01:38 |
| Austin Horstman | is it same version just without lto? | 15:01:57 |
| Ihar Hrachyshka | yeah, lto was not backported. | 15:02:07 |
| Austin Horstman | ah, good to know | 15:02:12 |
