| 26 Jun 2025 |
 aleksana 🏳️⚧️ (force me to bed after 18:00 UTC) | I think we should consider dropping pkg-config (at least as default) soon as it will fail to build and upstream is inactive | 04:03:12 |
 Alyssa Ross | In reply to @aleksana:mozilla.org
So it is trying to standarise this behavior?
pkg-config is patched to ignore Requires.private for --cflags without --static, whereas libpkgconf is not. I believe such a change also has to be made for pkgconf to be viable.
I thought we had a plan to avoid that issue by using absolute paths in .pc files. I think there was even a PR. The change we need made to pkgconf is to have it support overriding variables from .pc files with environment variables like pkg-config does. Until that's implemented, we can't switch. It would likely not be very hard to implement in pkgconf upstream. | 06:10:37 |
 emily | In reply to @aleksana:mozilla.org
https://github.com/NixOS/nixpkgs/pull/420107
Why we didn't switch to pkgconf like Arch Linux did @rosscomputerguy:matrix.org we don't do conditional patches if we can avoid it, the original unconditional + targeting stagimg was better | 11:07:28 |
| Alyssa Ross | I thought we had this documented even but couldn't find it | 11:07:52 |
 Randy Eckenrode | https://github.com/catchorg/Catch2/issues/2991 is broken by Clang 20. Any C++ people here have an idea what the fix is? | 13:18:47 |
| Randy Eckenrode | https://github.com/llvm/llvm-project/issues/140519 Maybe Clang bug? Fun. | 13:24:55 |
| emily | probably just disable the test? | 13:32:19 |
 Tristan Ross | Stdenv team is on the website https://nixos.org/community/teams/stdenv/ | 15:18:52 |
| aleksana 🏳️⚧️ (force me to bed after 18:00 UTC) | In reply to @qyliss:fairydust.space
I thought we had a plan to avoid that issue by using absolute paths in .pc files. I think there was even a PR. The change we need made to pkgconf is to have it support overriding variables from .pc files with environment variables like pkg-config does. Until that's implemented, we can't switch. It would likely not be very hard to implement in pkgconf upstream. So that PR alone isn't sufficient for switching to pkgconf yet? | 15:29:03 |
 K900 | It will break things | 15:29:20 |
| K900 | If we switch with just that PR | 15:29:29 |
|  nhnn joined the room. | 16:04:40 |
| Tristan Ross | Done | 20:42:34 |
 Grimmauld (any/all) | How would we even support the environment thingy? Do we actually need that in the first place or can we migrate packages using it away? | 20:45:43 |
| emily | adding support to pkgconf; the alternatives seemed worse last time we discussed | 20:48:49 |
| Alyssa Ross | it would be such a small patch | 20:49:08 |
| Alyssa Ross | I don't understand why we're spending so much time discussing it and not doing it | 20:49:16 |
| emily | if pkgconf upstream objects then hopefully we can work together to find another way to meet our needs | 20:49:28 |
| Alyssa Ross | yes, but as previously discussed that seems extremely unlikely | 20:49:43 |
| Grimmauld (any/all) | i mean, this expects we have people with the necessary domain knowledge that also have the time and motivation... Seems that never happened in the past year | 20:50:35 |
| emily | it requires little domain knowledge | 20:51:04 |
| Alyssa Ross | well we have plenty people with the time and motivation to keep asking us why we're not using pkgconf and who want to go full steam ahead on changing to it | 20:51:29 |
| Alyssa Ross | but that motivation never seems to extend far enough to submitting a very small PR to bring it to feature parity | 20:51:54 |
| emily | (for my part the reason is just "context switching to even a trivial task is easy for any given individual task but if I don't deliberately apply inertia to ones that don't seem like a fantastic idea to do immediately then bad things result") | 20:52:37 |
| emily | (this does not apply to giving pointers on what needs doing but that does require someone willing to put follow-up at the top of their own stack :p ) | 20:53:07 |
| emily | it is annoying having to keep pkg-config going | 20:53:44 |
| emily | I'm guessing the hook will probably give us unexpected fun when we do switch just on an unknown unknowns basis | 20:54:11 |
| Grimmauld (any/all) | we could already run one staging with the hook and then do the actual switch after | 20:55:21 |
| Grimmauld (any/all) | but that might be a dumb idea | 20:55:32 |
| emily | that won't help unless we revert our hack patch | 20:55:41 |
