| 16 May 2024 |
 delroth | (also if you'd rather have no ping here for stuff like that because you process bug / CL notifications async, lmk, I have no clue how y'all work so I'm defaulting to over-notifying...) | 22:50:28 |
 Qyriad | Failing at parse time is the solution we envisioned | 22:53:59 |
| Qyriad | We had the E/easy on there since we imagined it would be as simply as changing the assertion to a user-facing error message, but seem to have underestimated the difficulty in that itself, oops | 22:54:40 |
| Qyriad | (And yeah pinging us like this is fine) | 22:55:42 |
 jade_ | Qyriad: ok if we stage the libnixexpr->liblixexpr etc change? | 22:57:23 |
| jade_ | (pinging as assignee) | 22:57:42 |
| delroth | In reply to @qyriad:katesiria.org
We had the E/easy on there since we imagined it would be as simply as changing the assertion to a user-facing error message, but seem to have underestimated the difficulty in that itself, oops nah I think that if you had explained your envisioned fix it would have been fairly trivial but having to independently come to the same conclusion without having the same background was harder :P | 22:57:50 |
| delroth | still, happy it was the same conclusion (roughly - I think rejecting at parsing time is better than failing when making the worker goals, but that's tiny details) | 22:58:22 |
| Qyriad | In reply to@jade_:matrix.org
Qyriad: ok if we stage the libnixexpr->liblixexpr etc change? Yes, absolutely. If you are getting to it before us feel free! | 22:58:53 |
| jade_ | alrighty | 22:58:58 |
| Qyriad | In reply to@delroth:delroth.net
nah I think that if you had explained your envisioned fix it would have been fairly trivial but having to independently come to the same conclusion without having the same background was harder :P Yes we absolutely should have done that. I think in retrospect we may also have been somewhat confused, since the crash doesn't happen on Nix 2.3 because of course Nix 2.3 didn't have the ^ feature | 22:59:37 |
| delroth | also re: docs thanks for the pointer, I checked many manual pages (incl. "nix build"), I checked the glossary, etc. and of course didn't think of checking the generic "nix CLI" page | 22:59:55 |
| delroth | doesn't even really make sense that it's there because I think it's only a "build" feature anyway? | 23:00:06 |
| delroth | ah no, there's examples there with path-info | 23:00:19 |
| Qyriad | Yeah, the nix build and co pages now at least have a link to the generic "nix CLI page" saying that that's where installables are documented, but it needs to be so much more | 23:00:28 |
| Qyriad | and actually things like nix shell and nix develop can also use this syntax | 23:00:55 |
| Qyriad | Basically any nix3 command that might also realise in the process can do that | 23:01:11 |
| delroth | what does it mean to nix develop an output | 23:03:09 |
| Qyriad | fantastic question. | 23:03:51 |
| delroth | yeah. | 23:03:58 |
 puck | it uhhhhhhhhh | 23:04:00 |
| puck | sources $stdenv/setup, then does some churning to output the list of all variables and functions | 23:04:40 |
| puck | why is this part of Nix. this is nixpkgs-specific | 23:04:51 |
| delroth | yeah but all that is derivation specific, nothing that has to do with outputs | 23:05:27 |
| puck | oh! so | 23:06:04 |
| delroth | I'd guess it's just being ignored? | 23:06:17 |
| puck | you can nix develop a path that ends with -env and it'll .. just return an existing env file | 23:06:21 |
| Qyriad | just ran a nix develop on openssl^bin and openssl^lib and envd in both of them and there was zero difference | 23:06:22 |
| Qyriad | wait what | 23:06:45 |
| puck | okay, sec, this is a bit tricky to show | 23:06:57 |
