| 22 Oct 2025 |
|  @abaris99f:matrix.org removed their display name JannikH (Abaris). | 10:35:49 |
| @abaris99f:matrix.org left the room. | 10:36:09 |
| 28 Oct 2025 |
|  @dawnofmidnight:catgirl.cloud joined the room. | 18:01:33 |
| 31 Oct 2025 |
 raboof | reproducible builds report for the build closure of minimal iso: https://reproducibility.nixos.social/reports/nixos-minimal-25.11pre878042.544961dfcce8-x86_64-linux.iso.drv-build-closure | 17:08:15 |
| raboof | it's mostly the haskell stuff, and the upstream -fobject-determinism work is looking very promising, but will still take quite a while to make it into nixpkgs proper | 17:09:14 |
| raboof | * it's looking really good: the main remaining issues are mostly the haskell stuff. The upstream -fobject-determinism work is looking very promising, but will still take quite a while to make it into nixpkgs proper | 17:12:28 |
 emily | it may be possible to backport to older GHCs | 17:16:30 |
| emily | since major GHC upgrades happen fairly infrequently in Nixpkgs | 17:16:41 |
| emily | llvm.dev reproducibility is something I have an in-flight patch for a similar issue (though only really relevant on Darwin since Linux tends to use a fixed build directory), I can probably take at fixing that one too. | 17:17:55 |
| emily | (but no promises, I am overallocated time-wise to begin with) | 17:18:06 |
| raboof | as long as it's "only" the build closure I'm kinda fine with fixes taking a long time to land, as long as they're on their way. | 17:21:17 |
| raboof | I can relate ;) | 17:21:46 |
| raboof | Luckily I can justify setting aside some time for this latest push in reporting improvements through https://discourse.nixos.org/t/transforming-global-software-distribution-with-nixpkgs/64989 | 17:24:32 |
| emily | if you want to feel true despair, try building stuff with a different sandbox-build-dir sometime – there is a whole lot of stuff that is unreproducible if not for sandboxed Linux fixing the build directory by default | 17:24:51 |
| emily | which is much of what makes Darwin reproducibility a mess right now (and which I have WIP patches for) | 17:25:12 |
| emily | might be possible to use a randomized build directory on Linux in the future to surface more of that stuff if I can get the fixes that have helped for Darwin landed | 17:25:52 |
| raboof | yeah you shared that before... I'm a bit torn on it as I essentially don't care about mac 😆, but from a holistic point of view it's probably the wise thing to do. | 17:30:17 |
| emily | it's also relevant for e.g. Linux builds inside environments that can't use the sandbox, although those situations are pretty cursed | 17:45:45 |
| emily | (some rootless container environments, for instance) | 17:45:54 |
| emily | since those already cannot enforce a coherent build directory | 17:46:07 |
 dish [Fox/It/She] | I'm wondering if it would be reasonable to disable the checkPhase in writeShellApplication for pre-switch-checks, which is the only user of haskell in the default system closure. While having shellcheck is a good idea, system.preSwitchChecks is only used in two repos outside of nixpkgs(according to a quick sourcegraph search), and I feel like if this is something you're using, you're generally better at shell scripting and don't need to worry as much about shellcheck. Just a thought though, and would allow us to cut off the entire ghc repro issue for minmal isos | 19:00:34 |
| dish [Fox/It/She] | I'm wondering if it would be reasonable to disable the checkPhase in writeShellApplication for pre-switch-checks, which is the only user of haskell in the default system closure. While having shellcheck is a good idea, system.preSwitchChecks is only used in two repos outside of nixpkgs(according to a quick sourcegraph search), and I feel like if this is something you're using, you're generally better at shell scripting and don't need to worry as much about shellcheck. Just a thought though, and would allow us to cut off the entire ghc/haskell repro issue for minmal isos | 19:00:40 |
| emily | how much would that shrink the minimal ISO? | 22:26:25 |
 Grimmauld (any/all) | it would save me annoying ghc compiles on my minimalism adventures, i approve | 22:26:55 |
| emily | I am frankly not sure we need preSwitchChecks at all | 22:27:23 |
| Grimmauld (any/all) | Do we use buildShellApplication outside systemd.enableStrictShellChecks = true in the default closure? | 22:27:57 |
| Grimmauld (any/all) | actually wait no that is false by default huh | 22:28:25 |
| emily | --set PRE_SWITCH_CHECK ${lib.escapeShellArg config.system.preSwitchChecksScript} \
| 22:29:56 |
| emily | I think we actually just need to conditionalize this on config.system.preSwitchChecks… | 22:30:19 |
| emily | the same way it is in nixos/modules/system/activation/top-level.nix | 22:30:24 |
