| 15 Jun 2023 |
 @uep:matrix.org | Hm. Helping someone with an install, they did:
fileSystems."/boot" =
{
device = "/dev/disk/by-uuid/7CBF-1A69";
fsType = "zfs";
};
(wrong fstype)
| 16:12:37 |
| @uep:matrix.org | but the result is that it inferred a pool name of "" | 16:12:54 |
| @uep:matrix.org | and added a job to import that pool, which is of course failing. | 16:13:13 |
| @uep:matrix.org | seems like there's room for an assertion that the pool name isn't blank | 16:13:28 |
 ElvishJerricco | Oh that's... interesting | 16:13:34 |
| ElvishJerricco | Yea for sure | 16:13:37 |
| @uep:matrix.org | somewhere in here:
datasetToPool = x: elemAt (splitString "/" x) 0;
fsToPool = fs: datasetToPool fs.device;
zfsFilesystems = filter (x: x.fsType == "zfs") config.system.build.fileSystems;
allPools = unique ((map fsToPool zfsFilesystems) ++ cfgZfs.extraPools);
rootPools = unique (map fsToPool (filter fsNeededForBoot zfsFilesystems));
| 16:22:40 |
| @uep:matrix.org | or, misplaced in the thing that creates the import service, I guess | 16:23:00 |
| @uep:matrix.org | * or, misplaced in the thing that creates the import service, just below, I guess | 16:23:08 |
| ElvishJerricco | I'd just add a NixOS style assertion like assertions = [{ assertion = !(elem "" allPools); message = "pls no"; }] | 16:24:53 |
| @uep:matrix.org | well, that will fail but not helpfully? | 16:26:15 |
| ElvishJerricco | I mean failure is the only useful outcome | 16:26:31 |
| @uep:matrix.org | but yes, it does fit there | 16:26:33 |
| ElvishJerricco | So the message just needs to be helpful | 16:26:37 |
| ElvishJerricco | I just didn't bother to write one because I'm on my phone :P | 16:26:50 |
| @uep:matrix.org | yeah but it cant point to the dataset | 16:27:05 |
| ElvishJerricco | Sure it can | 16:27:12 |
| ElvishJerricco | Oh | 16:27:16 |
| ElvishJerricco | Wait I see what you mean | 16:27:20 |
| ElvishJerricco | It can name the mountpoint though | 16:27:28 |
| @uep:matrix.org | yeah, all it can point to is the blank at that point | 16:27:32 |
| ElvishJerricco | It's not uncommon to have more expensive, redundant computations in the message field, since it's lazily evaluated. Sometimes people throw in calculations that would error if the assertion isn't false | 16:28:23 |
| @uep:matrix.org | fair | 16:28:55 |
| ElvishJerricco | So the message can calculate which mountpoints are busted | 16:28:57 |
| @uep:matrix.org | i'll look into this later, when it's not 02.30 | 16:29:33 |
| @uep:matrix.org | it's probably enough to just have "must not start with /" in the message | 16:31:24 |
 @lily:lily.flowers | In reply to @elvishjerricco:matrix.org
lol yea it'd probably be helpful for testing it https://gist.github.com/lilyinstarlight/0597b656045ed7042fd200d0f0e9879f | 16:58:59 |
 K900 (deprecated) | Sadness | 17:18:35 |
| K900 (deprecated) | I still can't get amdgpu to not modeset on my desktop | 17:18:45 |
| K900 (deprecated) | It's triple monitors though | 17:18:49 |
