| 7 Oct 2025 |
 Sergei Zimmerman (xokdvium) | Merged for now :) Sorry for being a pain. We can do incremental improvements | 23:39:11 |
 lovesegfault | not a pain at all, i share your goals and concerns here :) | 23:39:31 |
| lovesegfault | btw, does this help with perf concerns? https://github.com/NixOS/nix/pull/14171/files#diff-d6730623c42446aa1f33e3177f7146d342ca8d5103e0814bb0b7ac212af5576dR90-R99 | 23:41:43 |
| lovesegfault | idk if the str scan is fast enough to help | 23:41:54 |
| Sergei Zimmerman (xokdvium) | I was thinking that we could just special-case '.' and '..': '.' does nothing and '..' does a pop. Also I don't think the rhs of this operator is ever supposed to have a '/' there | 23:43:07 |
| Sergei Zimmerman (xokdvium) | I see you've added the tests for such cases, but isn't that supposed to be handled by CanonPath::extend? | 23:45:46 |
| Sergei Zimmerman (xokdvium) | There's the comment:
/**
* Add a path component to this one. It must not contain any slashes.
*/
| 23:46:27 |
| Sergei Zimmerman (xokdvium) | Maybe the right thing is to instead fix the user code? | 23:46:38 |
| lovesegfault | i wasn't sure what could end up there realistically, just wanted to avoid having that weird assert thrown at users | 23:46:50 |
| lovesegfault | yeah, okay, maybe we only handle dots | 23:46:57 |
|  @adam:robins.wtf left the room. | 23:48:18 |
| Sergei Zimmerman (xokdvium) | Also maybe it's the downstream code that needs fixing. I think the idea for CanonPath (and this function) was to not accept . .. at all | 23:48:19 |
| lovesegfault | that i'm not sure about. I'll do your idea of special-casing ./.. and we can bring it up in the next meeting? | 23:48:55 |
| lovesegfault | done | 23:54:15 |
| lovesegfault | now onto the ci work | 23:54:41 |
| 8 Oct 2025 |
| lovesegfault | oh, that was easy | 00:03:18 |
| lovesegfault | i think | 00:03:20 |
| lovesegfault | https://github.com/NixOS/nix/pull/14184 | 00:05:40 |
| lovesegfault | Sergei Zimmerman (xokdvium): i think i achieved what you wanted, unless you wanted there to only be one allTests attrset that has componentTests and vmTests | 00:11:17 |
| lovesegfault | nice, addressing the review | 00:13:22 |
 Mic92 | @xokdvium:matrix.org: you will notice something about this merge: https://github.com/NixOS/nix/pull/14180 | 04:02:51 |
| Mic92 | https://github.com/NixOS/nix/actions/runs/18333686106/job/52213704796?pr=14187#step:8:28 yet another unwrapped std::filesystem exception. | 06:33:49 |
| lovesegfault | alright, with the CI changes done i've rebased the next little bit of the s3 work and it's ready for review: https://github.com/NixOS/nix/pull/14170 | 15:12:13 |
| lovesegfault | we're almost there now 🥳 | 15:12:59 |
|  siraben joined the room. | 15:25:02 |
| siraben | Looking to gauge interest in the ability to break a nix store lock manually, what do people think?
https://github.com/NixOS/nix/pull/13813 revisiting this PR right now | 15:25:17 |
|  frontear joined the room. | 15:36:50 |
| lovesegfault | John Ericson: i saw your review comments on #14170 just now, but i don't get the motivation? why introduce this usernameAuth stuff? | 15:45:23 |
| lovesegfault | are you thinking that something else will leverage that later? | 15:45:42 |
 John Ericson | lovesegfault: yes, something else might leverage it later, and also just trying to reduce the mount of stuff we have under the S3 ifdef | 15:46:14 |
