| 6 Mar 2025 |
 gdamjan | the host has an address of 2 always | 20:11:24 |
 raitobezarius | but this is exactly the wizardry you need when you run a datacenter at home | 20:11:23 |
| gdamjan | address or CID | 20:11:30 |
| gdamjan | the only stupid issue is, you have to manage guest CIDs manually :/ | 20:12:03 |
| gdamjan | if using low-level tools like qemy/cloud-hypervisor etc | 20:12:19 |
 ElvishJerricco | lol I just mean that wizardry indicates there's a lot of stuff to learn | 20:12:38 |
 Arian | Also known as a fun time | 20:14:16 |
| gdamjan | ЧД | 20:14:22 |
| gdamjan | * :) | 20:14:27 |
| Arian | Speaking of fun times. Networking in my Linux vm running in macos virtualization framework is fast *except for* S3 | 20:15:23 |
| Arian | When I fetch something from S3 in the vm is like kbps slow | 20:15:42 |
| Arian | Everything else is fast. How do I even | 20:15:54 |
| ElvishJerricco | I love incomprehensible edge cases | 20:16:18 |
| Arian | https://github.com/aws/aws-cli/issues/7299 | 20:16:45 |
| gdamjan | someone claims
Running sudo sysctl net.inet.tcp.tso=0 in the CI machine appears to have resolved the issue adequately.
| 20:23:00 |
| raitobezarius | seems like the 90s again when the acceleration HW stacks were completely fucked | 20:30:42 |
 @marie:marie.cologne | i have basically the same thing included in my pr for ssh-generator :3 | 20:40:28 |
 eliasp | In reply to @eliasp:kde.org
unfortunately, systemctl edit doesn't understand --transient, so this has to be handled manually…
but we might want to rework the patch 0016-systemctl-edit-suggest-systemdctl-edit-runtime-on-sy.patch pointing people towards --runtime …and simply dropping things into /run/systemd/transient doesn't seem to work either, even if that's the one that should take precedence over /etc/systemd/system and /run/systemd/system… 😩 | 20:54:04 |
| gdamjan | imho that's for units pid1 stores itself (for units created by systemd-run/dbus), probably doesn't read from there | 21:58:58 |
| gdamjan | perhaps nixos should put units in /usr instead of /etc | 21:59:29 |
 antifuchs | do we even have a /usr 😹 | 22:00:40 |
 Alyssa Ross | Yes | 22:02:00 |
| gdamjan | IIRC /usr/bin/env is required | 22:02:28 |
 Ramses 🇵🇸 | It's not a crazy idea to put them in /usr, imho, but we'll have to do considerable work to make that happen. We will need to generalise all the stuff used to set up /etc currently, and we'll lose some atomicity when switching generations | 22:25:21 |
| Alyssa Ross | We could just install a symlink to a /etc (or /run/current-system) subdirectory in /usr | 22:26:29 |
| eliasp | gdamjan the thing is, it still evaluates the unit in transient, but apparently doesn't apply all directives... Documentation= works, ExecStart= doesn't
nonetheless, I agree that it would be nice to have them all in /usr instead of /etc, but yeah - a looot of work | 23:20:04 |
| eliasp | but while we're at it - anyone feels like picking up my systemd-localectl-archive-path branch (fix for #267101 and successor of #375062)
I don't have time for it right now and I can't figure out, why my patch shows now effect… | 23:22:06 |
| eliasp | but while we're at it - anyone feels like picking up my systemd-localectl-archive-path branch (fix for #267101 and successor of #375062)?
I don't have time for it right now and I can't figure out, why my patch shows now effect… | 23:22:13 |
| eliasp | but while we're at it - anyone feels like picking up my systemd-localectl-archive-path branch (fix for #267101 and successor of #375062)?
I don't have time for it right now and I can't figure out, why my patch shows no effect… | 23:22:24 |
| 7 Mar 2025 |
| Arian | I think the fix here is that we shouldn't put nix generated config in /etc we should put it in /nix/store and extend SYSTEMD_UNIT_PATH | 07:47:07 |
