| 3 Jun 2021 |
 hexa | this is a local privesc and the current staging cycle is like two weeks at best | 20:00:00 |
| 4 Jun 2021 |
 lukegb (he/him) | https://hydra.nixos.org/eval/1675207 for master | 00:54:29 |
 stigo | https://github.com/NixOS/nixpkgs/pull/125646 | 10:27:53 |
| hexa | looks like the polkit change went into master without a hitch | 12:05:31 |
| hexa | stigo: the darwin ofborg builder is somewhat backed up with ~100 jobs in the queue fyi | 12:05:49 |
| lukegb (he/him) | hexa: yeah, just trying to unwedge hydra | 12:06:08 |
| stigo | :-\ | 12:06:18 |
| lukegb (he/him) | It should be fine now, I kicked off another eval | 12:06:20 |
| lukegb (he/him) | https://hydra.nixos.org/eval/1675342 | 12:06:42 |
 philipp | I'm wondering how feasible it would be to leech onto debian or centos for a stable package stack. E.g. take the php/nginx/mariadb/postgresql version from debian stable and port all their patches to nixpkgs and try to support it until the support for that debian version runs out. | 14:18:25 |
 Sandro | Why do we want to downgrade our packages? | 14:25:15 |
| Sandro | Interesting. Our default postgres is also at 11 like debian stable | 14:27:57 |
| Sandro | but I saw a PR that bumps that. | 14:28:06 |
| Sandro | Found a better example: Debian Stable has nginx 1.14.2. We are already at 1.20.1 for nginx stable/mainline which is the default for nginx | 14:29:50 |
| Sandro | or another example is that our glibc is 2.32 while Debian stable is only on 2.28 | 14:31:21 |
| Sandro | so I think this will only cause issues and requires us to patch more especially around software not being compatible with newer/older compilers | 14:32:11 |
| philipp | I don't want to downgrade packages. I would introduce a separate LTS attribute set. | 14:51:48 |
| Sandro | I don't see the advantage of that nor the time to maintain that tbh | 15:13:31 |
| Sandro | maintaining two versions of core packages like glibc6 involves probably a lot of work | 15:14:52 |
 andi- | philipp: what makes you think that sticking to the Debian model is easier? Usually upstreams provide new versions (or patches for the latest versions). I think we need less actual work right now than we would need if we used older versions. Sure, we could pick patches from Debian but that would establish a dependency on them actually updating before us. | 15:46:19 |
| philipp | It's less about making it easier and more allowing for longer support intervals. | 17:40:38 |
 ris_ | quite a few security-related PRs needing review right now | 17:46:00 |
| ris_ | i think it's an interesting idea philipp i'd just wonder how much the result would end up disconnected from our non-LTS branches. andi- patches can certainly flow both ways between the two projects. i know some of my backport patches have made it back into debian | 17:49:44 |
| ris_ | i'd quite like to lure debian developers over to our side because i get the impression that debian's processes and infra for maintaining packages is a nightmare | 17:51:38 |
| ris_ | like, versions of things all over the place, separate source trees, the security team not pushing their patches to sources.debian.org or the package maintainer's source control 😰 | 17:53:25 |
| andi- | I still don't see our gain adding patches to old libraries instead of bumping them - as long as the dependencies don't break. We do not have to retain ABI stability as we are a) rebuilding all depenndencies b) have a proper build system that covers a) :) | 17:54:35 |
| ris_ | well... what is "our" in this case? are "we" just a bunch of people who have self-selected as people who don't care about supporting old software? | 17:56:13 |
| ris_ | there is certainly a need for LTS, otherwise it wouldn't exist | 17:56:30 |
| Sandro | In reply to @philipp:xndr.de
It's less about making it easier and more allowing for longer support intervals. If we find more people which have an high interest in doing that or commercial support we can do that.
Or when we are bored but in my opinion we are not at that level yet. We have enough things to do and supporting more versions is a lot more work. | 17:57:11 |
| ris_ | otherwise my organization wouldn't be paying $x,000 to canonical for continued support of 16.04 | 17:58:03 |
