| 8 Jan 2025 |
 ElvishJerricco | antifuchs: The default for both of those is true. | 03:40:38 |
| ElvishJerricco | restartIfChanged = false just means that it won't be stopped / started at all. i.e. It will just be left running from the old configuration | 03:41:04 |
| ElvishJerricco | so it's true by default | 03:41:16 |
| ElvishJerricco | and stopIfChanged means that when restartIfChanged happens, it does so by a stop then a start | 03:41:32 |
 Arian | Also see https://github.com/NixOS/nixpkgs/issues/49528 | 09:32:15 |
 antifuchs | In reply to @elvishjerricco:matrix.org
and stopIfChanged means that when restartIfChanged happens, it does so by a stop then a start Right, so I think we need to set stopIfChanged to false on those foundational units too, udevd, networkd, maybe others too | 12:14:59 |
| Arian | Yep makes sense. In my opinion anything that doesn't have an ExecStop should just be able to use restarts iirc | 12:18:45 |
| ElvishJerricco | In reply to @arianvp:matrix.org
Yep makes sense. In my opinion anything that doesn't have an ExecStop should just be able to use restarts iirc Oh that's a good idea | 15:38:45 |
 @msanft:matrix.org | Does anyone of you want to take a look at this?
https://github.com/NixOS/nixpkgs/pull/364544 | 16:02:49 |
| ElvishJerricco | Moritz Sanft: Would it be plausible to make the documentation not depend on perl instead? | 16:21:35 |
 Sandro 🐧 | posted a link in the thread to where I think it is coming from | 16:40:02 |
| antifuchs | I keep forgetting, should changes touching stuff in nixos/lib apply nixfmt-rfc-style, or no? | 17:31:15 |
| ElvishJerricco | I'm pretty sure the rule of thumb, since the treewide reformat, is "if the file is already formatted, keep it formatted, otherwise, you don't have to worry about it" | 17:32:24 |
| ElvishJerricco | if it's unformatted and you'd like to format, then please format the whole file in a separate commit | 17:33:09 |
| antifuchs | haha, makes sense. The systemd-unit-opitons.nix file definitely doesn't look better post-format, so I will abstain | 17:36:02 |
| ElvishJerricco | antifuchs: the advice I've gotten for files that look awful post-formatting was "it's not the formatter's fault, just structure the code better so the formatting is better" | 17:36:49 |
| ElvishJerricco | which infuriates me | 17:36:51 |
| ElvishJerricco | but it's the best we've got | 17:36:55 |
| antifuchs | lol, yeah | 17:37:00 |
| antifuchs | HELL YES, this functional composition of options thing totally rules. think I have a change that defaults stopIfChanged to false unless ExecStop / ExecStopPost is set | 17:45:18 |
| antifuchs | filed https://github.com/NixOS/nixpkgs/pull/372191, I think this probably needs docs? | 17:49:43 |
| ElvishJerricco | antifuchs: that's iffy | 17:54:06 |
| ElvishJerricco | because of systemd.packages | 17:54:12 |
| ElvishJerricco | units can easily have an ExecStop that can't be known about at eval-tiem | 17:54:30 |
| ElvishJerricco | * units can easily have an ExecStop that can't be known about at eval-time | 17:54:32 |
| ElvishJerricco | this would have to be an stc-level change | 17:54:46 |
| antifuchs | ahhhhhh | 17:54:55 |
| antifuchs | ok, so probably better to explicitly set the stopIfChanged flag for the units I care about now & then see about modding stc | 17:55:34 |
| ElvishJerricco | I think so, yes | 17:56:03 |
| antifuchs | heeeere we go on pt1 of this: https://github.com/NixOS/nixpkgs/pull/372196 | 18:09:34 |
