11 Oct 2024 |
ElvishJerricco | oh | 04:23:41 |
ElvishJerricco | was joke | 04:23:42 |
ElvishJerricco | am in critical mode :P | 04:23:47 |
emily | well I mean I did assume that unit config was generally """better"" in some amorphous way | 04:23:57 |
emily | i.e. more flexible, more systemd-native, etc. | 04:24:01 |
emily | being able to specify deps and so on | 04:24:04 |
ElvishJerricco | You can do a lot of it in fstab, but yea, stuff like that is when you would reach for custom mount units | 04:24:23 |
mjm | deps you can get via options in fstab tho can't you? | 04:24:23 |
mjm | at least to some degree | 04:24:34 |
emily | In reply to @mjm:midna.dev deps you can get via options in fstab tho can't you? in a way that's different to every other systemd thing | 04:24:40 |
mjm | sure | 04:24:44 |
ElvishJerricco | but generally it's better to stick to the thing that all the other tooling in the world (like just plain mount ) understands | 04:24:44 |
emily | that's what systemd-mount is for | 04:24:58 |
emily | anyway ElvishJerricco maybe you should open an issue about the problems with the preservation implementation? | 04:25:11 |
emily | impermanence is popular and it's nice to have an implementation of it that works systemd-native and doesn't involve interpreters, even if it might need some refinement | 04:25:30 |
ElvishJerricco | like, we could have migrated fileSystems to pre-generated mount units long ago if we had wanted to but it's just good to have fstab | 04:25:30 |
ElvishJerricco | In reply to @emilazy:matrix.org anyway ElvishJerricco maybe you should open an issue about the problems with the preservation implementation? I'd have to like.... actually review and understand more of it :P | 04:25:44 |
ElvishJerricco | like presumably there's some purpose to the tmpfiles stuff that I'm not understanding | 04:25:58 |
mjm | i assume it's similar to whatever impermanence's bad create dirs needed for boot thing is for | 04:26:29 |
mjm | but idk | 04:26:36 |
ElvishJerricco | oh, that might be a problem, yea | 04:26:45 |
ElvishJerricco | I mean personally I don't think the create-dirs thing is needed; creating the original persistent dirs seems like the sort of thing one should be able to tolerate during formatting | 04:27:13 |
mjm | will a bind mount create every parent directory needed for the target? | 04:27:13 |
ElvishJerricco | systemd always creates parent dirs for the mountpoints of mount units | 04:27:34 |
mjm | ok | 04:27:38 |
ElvishJerricco | (unless a parent is ro ) | 04:27:39 |
mjm | then yeah i don't really get why you would need tmpfiles to come first | 04:27:58 |
ElvishJerricco | I think there's an open issue about creating the parents for bind mount sources as well | 04:28:07 |
mjm | seems like it could happen after all the mounts are done | 04:28:07 |
ElvishJerricco | but that's trickier because you don't necessarily intend for the source to be a dir | 04:28:20 |