| 1 Jul 2025 |
 Paul Meyer (katexochen) | Anyone interested in reviewing https://github.com/NixOS/nixpkgs/pull/395494? | 16:31:06 |
| 6 Jul 2025 |
|  @jammie:matrix.org left the room. | 02:27:18 |
|  Cathal changed their display name from CJ to Cathal. | 17:17:27 |
| 7 Jul 2025 |
|  @jackleightcap:matrix.org left the room. | 18:09:42 |
| 13 Jul 2025 |
|  timschumi joined the room. | 18:30:07 |
|  Marie changed their profile picture. | 20:11:49 |
| 25 Jul 2025 |
|  Vuks joined the room. | 18:04:29 |
| 26 Jul 2025 |
|  Jarkad joined the room. | 17:51:35 |
| 2 Aug 2025 |
|  @life00:tchncs.de joined the room. | 11:14:27 |
|  @Dan:matrix.org left the room. | 21:17:43 |
| 3 Aug 2025 |
|  prescientmoon changed their display name from Adriel to prescientmoon. | 14:31:36 |
| 5 Aug 2025 |
|  isabel changed their profile picture. | 09:49:47 |
| 6 Aug 2025 |
 emily | fyi I am actively working on detecting and fixing leaks of the build directory into derivation outputs | 13:06:31 |
| emily | which should make a big difference for package reproducibility on Darwin | 13:06:48 |
| emily | (modulo --rebuild still being somewhat busted due to the way it changes store paths and the cryptographic hashes in aarch64-darwin binaries) | 13:07:39 |
| emily | Linux's current use of a fixed namespaced build directory masks reproducibility issues that surface on Darwin due to the lack of working chroot there. the leaks are also often just upstream or packaging bugs when they're not benign ones we can remap away like __FILE__ | 13:09:48 |
| emily | I'm working on a fix for an unrelated Darwin build directory issue that involves injecting more entropy into the paths than we already do, so I'm preemptively building a bunch of stuff and fixing leaks that were happening on Linux too | 13:11:04 |
| emily | hoping to get the build directory changed to something more distinguishable than /build on Linux too so that we can expand the existing tmpdir auditing hook to catch more of this | 13:12:11 |
| emily | (and ideally get some entropy in there in future too, so that --rebuild can reliably detect reproducibility issues they cause on Linux and we don't have to deal with the use-mention problems of the scanner detecting a false leak from documentation that mentions a fixed path) | 13:13:46 |
| 7 Aug 2025 |
|  @sbc64:matrix.org left the room. | 09:41:11 |
| 8 Aug 2025 |
|  0x4A6F joined the room. | 06:59:31 |
|  Picnoir changed their display name from Picnoir to Picnoir DECT 7426. | 12:34:01 |
| 13 Aug 2025 |
| Picnoir changed their display name from Picnoir DECT 7426 to Picnoir. | 13:25:09 |
| 19 Aug 2025 |
 raboof | https://github.com/NixOS/nixpkgs/issues/434930 seems like nice low-hanging fruit regression | 08:05:50 |
| 20 Aug 2025 |
|  jopejoe1 changed their display name from jopejoe1 (4094@GPN23) to jopejoe1. | 18:51:42 |
| 22 Aug 2025 |
|  elikoga changed their profile picture. | 17:28:32 |
| 25 Aug 2025 |
| emily | Alyssa Ross: I remember you having reproducibility objections to PGO, so flagging up https://discourse.nixos.org/t/transparently-supporting-pgo-fdo-plo-optimized-builds/68420 | 03:17:10 |
| emily | though unclear if there is going to be serious work on this or if it is just an ad for the crowdfunding platform… | 03:17:32 |
| raboof | It sounds like they're proposing to do the profiling on the users' machine, in which case reproducibility is less essential I suppose (and it could even be reproducible if they store the profiling data and use it as an input) | 06:11:32 |
| emily | yeah, I don't know. "I believe the correct high-level model for NixOS is binary substitution. The derivation the users specifies does not change. The binary used to satisfy that specification changes." is pretty vague. not sure they understand the model | 06:12:27 |
