| 6 Jan 2026 |
 aleksana 🏳️⚧️ | In reply to @ma27:nicht-so.sexy
vcunat any objections to sending https://github.com/NixOS/nixpkgs/pull/379542 (glibc 2.42) now? We'll also need https://github.com/NixOS/nixpkgs/pull/471270. Not sure of other breakages | 13:21:17 |
| aleksana 🏳️⚧️ | * We'll also need https://github.com/NixOS/nixpkgs/pull/471270 to fix qemu. Not sure of other breakages | 13:21:29 |
 ma27 | not aware of any major breakage. In fact, my laptop is currently running on that branch for testing for a few days.
re. the one you're linking here, we agreed to accept this for now: https://github.com/NixOS/nixpkgs/pull/379542#issuecomment-3699768829
Let's not forget that the interface in question (<termio.h>) has been obsolete since the late 90s and from what I've seent his only affects non-native workloads (and apparently even target architectures only that we don't provide full support for). | 13:30:18 |
 dramforever | no, this affects everything except ppc | 18:58:34 |
| dramforever | which is cursed | 18:59:01 |
| dramforever | but qemu only needs staging-nixos, so i think it should be okay to not hold up glibc 2.42 update | 18:59:37 |
 emily | at least some QEMU configuration is pretty load-bearing for NixOS tests | 19:16:05 |
| emily | so we'd presumably want to throw a fix into -next if it blocks basic native VMs | 19:16:20 |
| dramforever | this is linux-user only | 19:39:49 |
| dramforever | so qemu-system-* are all fine | 19:39:56 |
| emily | ah, ok | 19:39:59 |
| emily | binfmt users will be unhappy I guess | 19:40:09 |
| dramforever | very confused | 19:40:33 |
| dramforever | * they'll be very confused | 19:40:44 |
| dramforever | shell just sits there doing nothing | 19:40:46 |
| dramforever | but i was thinking in that case we could also slip this into staging-nixos and land it on master, while the -next -> master is going on | 19:42:00 |
| dramforever | the main problem is that qemu upstream seems to be just sleeping on this | 19:43:27 |
| emily | has no other distro updated glibc? | 20:04:23 |
| ma27 | gentoo has (there are bugreports on that matter on their end), fedora has (already as part of f43) | 20:08:09 |
 vcunat | Ubuntu 25.10 also has 2.42, at a glance. | 20:10:10 |
| ma27 | anyways, left a comment in https://github.com/NixOS/nixpkgs/pull/471270 because I'm not sure if just shipping the patches would be OK here. | 20:31:10 |
 Fabián Heredia | FYI seems like those have already been reviewed by at least one person and sent via de indicated channel.
Left a comment pointing to the upstream processes for inclusion/merging. | 20:37:38 |
| emily | so if it's just termios… are non-interactive shells not affected? | 21:35:35 |
| emily | i.e. actually Nix binfmt builders should be fine and it's just QEMU user emulation of interactive terminal programs? | 21:35:43 |
| emily | that seems niche enough that it's probably not a blocker IMO | 21:35:49 |
| emily | OTOH Gentoo is patching I guess? | 21:37:59 |
| 7 Jan 2026 |
| dramforever | In reply to @emilazy:matrix.org
so if it's just termios… are non-interactive shells not affected? non-interactive programs are not affected. the top level nix build run in a pty and i hope those are unaffected? | 04:25:05 |
| dramforever | and running cross-arch containers feels like it's, quite common among qemu-user users | 04:25:55 |
| emily | sounds like the projected impact shouldn't be world-ending even if we wait for upstream and they're a bit slow at least? | 05:11:56 |
| emily | I've used qemu-user a nontrivial number of times in my life but I think almost never to run interactive terminal programs | 05:12:31 |
