| 9 Apr 2026 |
|  Lisanna changed their profile picture. | 22:12:35 |
 viraptor | Sigh, looks like I got lost in the signing maze. That's what I get for working way past bedtime 🤦♂️ | 22:23:36 |
| Lisanna changed their profile picture. | 23:14:56 |
| 10 Apr 2026 |
 Randy Eckenrode | Apple might have fixed the libiconv issue that was causing problems for gnulib. That’s unexpected if so. | 02:08:43 |
 mall0c | In reply to @viraptor:tchncs.de
Ok, I've got a draft PR which I can't fully test due to some security software. Does anyone feel like rebuilding the bootstrap? https://github.com/NixOS/nixpkgs/pull/508307 To test if it worked properly, you can try building https://github.com/NixOS/nixpkgs/pull/500850 rebased on top of it.
rebuild-linux: 5001+
👀 bro about to make hydra cry | 02:30:28 |
| Randy Eckenrode | https://github.com/NixOS/nixpkgs/pull/508474 splits the source releases out from the SDK. I tried to get it down to zero rebuilds, but the overlaying during the bootstrap messes things up. | 03:38:50 |
| mall0c | I suppose this is part of your effort to get swift sdk updated? | 03:54:50 |
| Randy Eckenrode | It’s a detour to clean up some Darwin stuff. | 04:16:23 |
| Randy Eckenrode | I think I got rid of the rebuilds. | 04:32:12 |
| Randy Eckenrode | I did just confirm though that SwiftPM 5.10.1 supports putting packages in editable mode. That’s how I’m vendoring dependencies. I want to land the breaking stuff for the release, then I can work on the Swift 6.x update. | 04:59:11 |
| Randy Eckenrode | https://github.com/NixOS/nixpkgs/pull/508474 definitely has no rebuilds on either Darwin or Linux. | 05:17:45 |
|  jyap joined the room. | 19:46:58 |
| jyap | so the latest macos tahoe 26.4 seems really hostile now to built CLI apps. they don't even run for testing.
is there a solution for this where the command is just killed by the OS?
e.g.
> _________________ ERROR collecting ffmpeg/tests/test_ffmpeg.py _________________
> ffmpeg/tests/test_ffmpeg.py:29: in <module>
> subprocess.check_call(['/nix/store/6a5nr567sb4a36lisa6gydpp3bfij1vv-ffmpeg-8.0-bin/bin/ffmpeg', '-version'])
> /nix/store/h3l9h1vmbzvgs4x1ffl0pqbmzhlk89yl-python3-3.13.12/lib/python3.13/subprocess.py:419: in check_call
> raise CalledProcessError(retcode, cmd)
> E subprocess.CalledProcessError: Command '['/nix/store/6a5nr567sb4a36lisa6gydpp3bfij1vv-ffmpeg-8.0-bin/bin/ffmpeg', '-version']' died with <Signals.SIGKILL: 9>.
| 19:48:05 |
| Randy Eckenrode | Probably https://github.com/NixOS/nixpkgs/issues/208951. It’s concerning the problem seems to be getting worse. | 19:50:07 |
| jyap | thanks, i'll take a look | 19:52:32 |
| jyap | yeah it's similar to this issue where the fix is to disable tests https://github.com/NixOS/nixpkgs/issues/507531 | 19:58:05 |
| 11 Apr 2026 |
| Randy Eckenrode | Why can’t anything link when using the 26.4 SDK? | 00:43:03 |
| Randy Eckenrode | Oh, crap. | 00:44:20 |
| Randy Eckenrode | Apple dropped arm64 support. | 00:44:25 |
| Randy Eckenrode | The stubs only support arm64e. | 00:44:34 |
| Randy Eckenrode | Well, some of them only support it. libc++ still supports arm64, which may be a mistake. | 00:47:03 |
| Randy Eckenrode | Ugh, this sucks. | 00:47:06 |
| Randy Eckenrode | Is our SDK unpacking mangling the stubs? That somehow seems worse. | 00:53:10 |
| Randy Eckenrode | No, doesn’t look like. | 01:00:44 |
| Randy Eckenrode | * | 01:00:50 |
 emily | wait, what? wasn't the arm64e ABI not stable before? | 01:02:19 |
| emily | isn't it more likely that they just merged the symbol lists into one or something? | 01:02:28 |
| Randy Eckenrode | It was stabilized in macOS 26. | 01:05:11 |
| Randy Eckenrode | Many of the 26.4 headers do not include arm64-macos in the list of targets. | 01:05:28 |
| emily | but you can target way back with the macOS 26 SDK, right? 🤔 | 01:05:37 |
