| 14 Aug 2025 |
 Sergei Zimmerman (xokdvium) | https://github.com/NixOS/nix/blob/58eabe7479177dc63cf3c704230a45b9260b3fec/ci/gha/tests/default.nix#L172 | 18:16:22 |
 jade_ | so umm. they seem to have defined this as "throws .. sometimes .. if certain int to float conversions fail .. sometimes" | 18:17:01 |
| jade_ | surely these should just passthrough integers entirely unmodified | 18:17:12 |
| jade_ | right? | 18:17:18 |
 emily | I assume the conservative throw is for the same reason you did one in fromJSON | 18:17:40 |
| emily | keeps options open | 18:17:47 |
| jade_ | I also rewrote the float to int conversion to use the correct function | 18:17:51 |
| emily | but yeah it's hard to imagine a different sensible behaviour to use here | 18:17:56 |
| emily | but I think if you pass through integers the practical value you get is different to what it was historically | 18:18:18 |
| jade_ | yeah idk, i think converting it to a float first is just .. weird and hard to ensure it behaves consistently, comparatively | 18:18:19 |
| emily | because of how the UB would be compiled in practice | 18:18:21 |
| jade_ | oh | 18:18:24 |
| jade_ | ugh | 18:18:26 |
| jade_ | yeah | 18:18:27 |
| jade_ | okay | 18:18:28 |
| emily | so having a deprecation period where it errors seems sensible | 18:18:34 |
| emily | put up a whole bunch of changes for the TOML stuff (both the port of the Nix change for stable branches and the dropping of the experimental feature for main), NUL bytes in JSON/TOML, and overflowing JSON literals | 19:11:24 |
 Rutile (Commentator2.0) feel free to ping | https://gerrit.lix.systems/c/lix/+/3333/47
can someone punch besadii? | 19:39:52 |
| jade_ | hm, waiting for agent looks like something is busted ya | 19:50:52 |
| emily | jade_: do you want to take a look at the current state of https://github.com/NixOS/nixpkgs/pull/433729 as Lix is an out-of-tree consumer? | 20:26:19 |
| 15 Aug 2025 |
| emily | is there anything that can be done to kick flaky CI builds for my own CLs? | 01:02:31 |
| emily | I guess I can just do a nop rebase to bump the commit timestamp | 01:06:25 |
| emily | or not, because it tries too hard to be clever: "Verified-1 has been copied to patch set 4 (copy condition: "changekind:NO_CODE_CHANGE")." | 01:13:50 |
| emily | (…and doing and then undoing a trivial code change breaks the +2. oh well) | 01:47:00 |
 aloisw | You should be able to click on the buildkite link and then there is a button to retry failed builds. | 06:00:09 |
| Rutile (Commentator2.0) feel free to ping | Can someone give us an overview what parts of store relocation work and dont work on darwin? | 07:25:04 |
| emily | chroot does not practically work (though actually I wonder whether we might be able to make it work with Nix specifically, but it probably wouldn't be too useful) | 14:46:04 |
| emily | I assume you mean a diverted store right? I'm bad at all this terminology | 14:46:24 |
| emily | you cannot have a build whose logical build directory differs from its physical one, basically | 14:46:40 |
| emily | In reply to @aloisw:julia0815.de
You should be able to click on the buildkite link and then there is a button to retry failed builds. it has a forbidden cursor and I can't click it. I presume I'd need to log in and have privileges that I expect I don't have? | 14:47:07 |
