10 Jul 2024 |
charuto | In reply to @raboof:matrix.org it took a few fixes, but the minimal 24.05 ISO is reproducible again: https://discourse.nixos.org/t/nixos-reproducible-builds-minimal-installation-iso-successfully-independently-rebuilt/34756/9 i'm confused, isn't both jfsutils not reproducible and part of the minimal iso? | 22:19:06 |
charuto | In reply to @raboof:matrix.org it took a few fixes, but the minimal 24.05 ISO is reproducible again: https://discourse.nixos.org/t/nixos-reproducible-builds-minimal-installation-iso-successfully-independently-rebuilt/34756/9 * i'm confused, isn't jfsutils both not reproducible and part of the minimal iso? | 22:19:18 |
Atemu | charuto: It regressed | 22:55:37 |
11 Jul 2024 |
charuto | In reply to @atemu12:matrix.org charuto: It regressed ?? #276433 is open since december and was never closed. | 01:08:22 |
Atemu | In that case I may have randomly worked | 01:13:45 |
Atemu | We don't actually know whether the ISO is reproducible or not | 01:14:00 |
Atemu | And we will never find out | 01:14:06 |
Atemu | We can only take snapshots | 01:14:34 |
Atemu | Something working in one snapshot is no guarantee that it works the next time | 01:15:00 |
raboof | In this case 'the ISO is reproducible' meant 'we rebuilt it from source and it produced the same binary' | 06:09:50 |
raboof | we have seen a nondeterminism in jfsutils in the past and haven't fixed it, so we suspect it is still there, but it 'often' does produce the same binary | 06:10:58 |
bbenno | q | 08:43:50 |
emily | JFS is basically unmaintained and has been on the long road to probable removal from the kernel for years, is it even worth keeping? I guess if it helps shake out toolchain bugs | 10:27:57 |
emily | Amazon Linux and probably other distros have already disabled it | 10:28:08 |
Atemu | What's even pulling it in? | 10:29:01 |
emily | I assumed jfsutils was just being included explicitly so you can install on JFS | 10:32:18 |
emily | since it's enabled in the kernel | 10:32:23 |
emily | yeah, nixos/modules/tasks/filesystems/jfs.nix | 10:32:31 |
emily | disabling it to increase the reproducibility stats does feel a little Goodhart's law though :) | 10:35:07 |
Atemu | Given jfs' status, I think it's time to disable it anyways | 10:35:42 |
Atemu | We don't need to carry such baggage forever | 10:35:54 |
emily | FWIW it's not "officially" deprecated upstream, there was just a suggestion to do so years ago, and once in a blue moon it gets a bugfix, but my understanding is that it sees much less attention than any "active" general-purpose filesystems and its reliability might not be in a great place | 10:37:13 |
emily | it might start getting removed if there are major reworks to the FS layer that would be a waste of time to migrate JFS over for but otherwise I expect it'll just continue to rot | 10:37:44 |
Atemu | If even the kernel is considering removing it, we should have probably removed it years ago | 10:39:03 |
emily | looking at nixos/modules/profiles/base.nix , reiserfs probably wants removing; that one is actually officially deprecated and on the path to removal! | 10:39:10 |
emily | it's actually planned for removal next year even | 10:47:15 |
raboof | In reply to @emilazy:matrix.org disabling it to increase the reproducibility stats does feel a little Goodhart's law though :) Somewhat, though keeping the base system small is also a useful goal, and if r-b helps as a mechanism to find candidates for removal that's still a win imo 😃 | 11:20:16 |
emily | ideally there'd be another known package people actually care about that triggers whatever cursed toolchain bug is going on there | 11:21:01 |
raboof | (in the mean time I'm having fun with reproducibility issues in the Scala compiler: https://github.com/scala/scala3/pull/20593, https://github.com/scala/scala3/issues/21154) | 13:27:29 |
| @echobc:matrix.org joined the room. | 15:21:19 |