| 14 Aug 2025 |
 jade_ | that's correct, yeah, i don't know where that's supposed to be documented | 17:03:09 |
 puck | In reply to @jade_:matrix.org
this seems like a good idea, yes. it will break some of puck 's eval-breaking art projects, but that's good actually. i expected it to break some day, tbh | 17:03:10 |
| puck | and i think i no longer depend on it in any thing | 17:03:36 |
 emily | btw https://github.com/NixOS/nixpkgs/pull/433729 is also funny for CI | 17:03:57 |
| emily | turns out some Nixpkgs lib tests were not in fact testing the built interpreter | 17:04:11 |
| jade_ | unrelated question: why is lix not also codeowners on those files? i mean, i think we would have to convince someone to make us a group but ..... | 17:05:21 |
 Rutile (Commentator2.0) feel free to ping | In reply to @emilazy:matrix.org
btw what's the intended way to update functional2 snapshots with a just-built Lix? The env variable and just test-functional2 --accept-tests | 17:05:28 |
| emily | probably just because nobody did it? | 17:05:42 |
| emily | there's already a lib.teams entry | 17:05:55 |
| emily | any committer can make a GitHub team | 17:06:01 |
| jade_ | wait really?? okay let's fix tha | 17:06:28 |
| emily | right, it was the "just-built Lix" part | 17:06:46 |
| emily | btw | 17:06:48 |
| emily | you can't --accept-tests a change from an eval failure to success or vice versa | 17:06:55 |
| emily | would be nice to have | 17:06:59 |
| emily | https://github.com/NixOS/nix/pull/13013 | 17:15:38 |
| emily | joy | 17:15:40 |
| emily | does anyone else feel like backporting that so I don't spend literally all day dealing with overflow bugs :P | 17:17:40 |
| jade_ | https://github.com/orgs/NixOS/teams/lix-maintainers/members done | 17:18:37 |
| jade_ | woof. yeah i can do it | 17:18:55 |
| emily | the src → lix rename is tremendously annoying for Nix backports because of Git's broken handling of renames. can't wait for Jujutsu's rebase algorithm to handle moves/copies. | 17:19:52 |
| emily | does anyone have any tricks that they use to make it not be useless at that | 17:20:13 |
| emily | I tried merge.directoryRenames | 17:20:31 |
 Charles | i wonder if a similar trick for cherry-picking across formatting changes would help? | 17:21:10 |
| emily | yeah, but it would require jj run rather than jj fix sadly | 17:21:43 |
| emily | maybe a Git mergetool could do it but I don't feel like writing one of those | 17:21:51 |
| emily | I guess I can just duplicate, move, rebase, but ugh | 17:22:01 |
| Charles | 🪦 | 17:22:48 |
| jade_ | yeah. idk. i find so many bugs in code review of stuff that got accepted by cppnix that i think that most code has to be rewritten anyway. i think it's possible to rename the directory in a commit that you can base the cherry-picked one on, or something like that. | 17:25:50 |
| jade_ | for example, i found straight up wrong code in here (in addition to error messages that refer to concepts that only exist in C++ and not in user facing places)
i am still happy to port the fix, but it's going to experience some rewrites | 17:27:16 |
