| 20 Oct 2024 |
 ersei | In reply to @aloisw:kde.org
Well nix develop does set IN_NIX_SHELL. but nix shell doesn't, unfortunately | 18:32:36 |
| ersei | In reply to @9999years:matrix.org
also you should look up the --impure option instead of setting impure unconditionally probably, but maybe wait to see if people are like We Cannot Do THis first I think it'll always be impure, no? since the system's path will be leaking into the environment | 18:33:04 |
| ersei | even in --ignore-environment, PATH will contain /run/current-system stuff | 18:33:35 |
| ersei | but yeah it's up to the maintainers and what they think | 18:35:34 |
| ersei | personally I think that it'll make behaviour more consistent | 18:35:49 |
| ersei | and I like having my custom prompt in fish know it's in a nix3 shell | 18:36:25 |
| ersei | though if you look at the github issue linked to the PR (https://github.com/NixOS/nix/issues/3862), there is an argument to be made that nix shell does not warrant having IN_NIX_SHELL set | 18:37:39 |
| ersei | https://github.com/NixOS/nix/issues/3862#issuecomment-918926654 specifically
Yeah, I think IN_NIX_SHELL should only be set in nix develop. nix shell only adds packages to $PATH so I don't think it really matches the expectations for IN_NIX_SHELL (which is that you're in a stdenv build environment).
| 18:37:55 |
| ersei | it is worth bringing up the purity concerns with building, since if there is an inconsistency, then there could be a source of variance | 18:40:15 |
| ersei | but I'm not familiar enough with the deterministic nature of nix builds to know if this will be an issue or not | 18:41:47 |
 aloisw | None of the shells is really designed for purity (if you want that, use sandboxing), IN_NIX_SHELL also seems to be mostly for user convenience. | 19:39:52 |
 EsperLily [she/her] | hey folks, i finally dealt with work restrictions and can now do maintenance work on Lix in my spare time. I'm getting ready now to submit my first changeset for a trivial issue, never used gerrit before, is there any established etiquette to this? | 19:53:06 |
 just1602 | EsperLily [she/her]: there's this page on the lix wiki that can probably answer a couple of your questions / interrogations: https://wiki.lix.systems/books/lix-contributors/page/gerrit | 20:02:14 |
| EsperLily [she/her] | i read through that already and it talks about the mechanics of it, i just didn't know if there's any etiquette around e.g. grabbing an issue, or asking for review. that page also mentions topics but doesn't say if there's any expectation around how topics are used | 20:09:02 |
| EsperLily [she/her] | * i read through that already and it talks about the mechanics of it, i just didn't know if there's any etiquette around e.g. grabbing an issue, or asking for review. that page also mentions topics but doesn't say if there's any expectation around how topics are used (edit: just read gerrit documentation on topics, sounds like something i should probably ignore) | 20:12:33 |
| just1602 | In reply to @esperlily:matrix.org
i read through that already and it talks about the mechanics of it, i just didn't know if there's any etiquette around e.g. grabbing an issue, or asking for review. that page also mentions topics but doesn't say if there's any expectation around how topics are used I only submit 2 CL and they were link to an issue, but I'm not a core dev at all, so I'll let other people answer since it's gonna be more useful to you. | 20:12:49 |
| EsperLily [she/her] | ok i went ahead and submitted the changelist | 20:16:02 |
 benjamin | In reply to@ersei:ersei.net
what do y'all think about setting IN_NIX_SHELL in nix3? it's the same behaviour as nix2 I would like this a lot as a user. Currently I'm working around the absence of an IN_NIX_SHELL-equivalent by looking for nix store paths in PATH to set my shell prompt | 21:09:24 |
| benjamin | https://github.com/NixOS/nix/issues/6677 <- issue I opened about this back in 2022 | 21:10:18 |
| EsperLily [she/her] | I have SHLVL in my prompt and that's my normal cue for being in a nix shell. Having some env var might be nice, though i'm a little concerned about reusing IN_NIX_SHELL as that currently implies a much larger change to the shell environment than just some PATH additions | 22:00:54 |
| EsperLily [she/her] | i'm looking at the auto-optimize-store darwin bug and reading through the optimize-store code i'm slightly concerned about the possibility of TOCTTOU here. Can multiple calls to LocalStore::optimisePath_ ever occur concurrently (whether it's different processes or different threads in the same process)? There's a minor TOCTTOU bug here though the conditions for triggering it are likely hard (though possibly also triggerable if a GC is occurring at the same time, i'm not sure how GC interacts with optimize store yet) | 22:45:00 |
| 21 Oct 2024 |
| benjamin | In reply to@esperlily:matrix.org
I have SHLVL in my prompt and that's my normal cue for being in a nix shell. Having some env var might be nice, though i'm a little concerned about reusing IN_NIX_SHELL as that currently implies a much larger change to the shell environment than just some PATH additions yeah, IN_NIX3_SHELL or something might be better? | 02:31:30 |
 Linux Hackerman | jade_: I applied your suggestions in https://gerrit.lix.systems/c/lix/+/2057 -- anything else to do? And is it appropriate for me to ping you here if I'm not sure about the status? :) | 11:46:08 |
 KFears (tragedy arc) | I can't figure out the "functional" tests (the ones that are a Bash script) at all. Locally, I get a failure for nix profile something. Remotely, functional-nix-shell fails (and it's the one that's actually supposed to fail). How do I actually run those tests correctly in the local environment? I really want to look at the detailed logs that CI won't show me :( | 13:28:46 |
 piegames | Direnv allow and just test functional-lang should suffice | 13:39:36 |
| KFears (tragedy arc) | I think just test functional-lang literally runs "functional-lang" test | 13:48:19 |
| KFears (tragedy arc) | And just test fails the wrong test but succeeds the one that actually fails | 13:48:49 |
| piegames | In reply to @kfears:matrix.org
I think just test functional-lang literally runs "functional-lang" test Yes, insert you test name here | 13:52:15 |
| piegames | In reply to @kfears:matrix.org
And just test fails the wrong test but succeeds the one that actually fails Try doing a clean build. Sometimes stuff is a bit impure, frustratingly | 13:52:39 |
| KFears (tragedy arc) | That seems to have helped | 13:57:27 |
