| 17 Jan 2025 |
 ElvishJerricco | phaer: huh, vfkit tells me rosetta is not installed, but it totally is... | 21:12:10 |
 phaer | Hm.. that worked for me without a hitch. But the VM itself doesn't need rosetta, so should also work if you disable it in base.nix | 21:13:27 |
| ElvishJerricco | Yea I dunno why your journal seemed very incomplete; everything I was expecting is there when I run this VM myself. | 22:00:43 |
| ElvishJerricco | As far as I can tell reading the systemd source, the only reason the log would like the way it does in stage 2 is because CREDENTIALS_DIRECTORY or ENCRYPTED_CREDENTIALS_DIRECTORY is set on PID 1 as it starts | 22:01:21 |
| ElvishJerricco | which doesn't make any sense; I don't know why those would be set | 22:01:31 |
| ElvishJerricco | I feel like it's gotta be something to do with the fact that initrd-nixos-activation.service moves some mounts around like /run, which I think systemd kind of expects to be in charge of. But I just cannot figure out why that would matter... | 22:03:30 |
| 18 Jan 2025 |
 Arian | Phaer note that some things can get quite wonky with mounting MacOS nix store on linix | 00:19:05 |
| Arian | Due to use-casehack | 00:19:11 |
| Arian | I saw some really weird shit break in really obscure ways | 00:19:25 |
| phaer | I didn't solve it yet, but there's indeed something fishy in initrd-switch-root. switch_root.conf is empty while it shouldn't be and looks the closure is okay. But I need to dig deeper and it's 1:22 am here. So not today | 00:22:52 |
| ElvishJerricco | I thought the installer created a case sensitive APFS volume these days. | 00:45:12 |
| ElvishJerricco | note that switch-root.conf only exists during stage 1 and isn't populated till fairly late in the stage 1 process | 00:46:54 |
| phaer | Yes, I am cat'ing it straight in the (patched) unit | 00:48:43 |
| ElvishJerricco | oh, then that is indeed very weird, assuming you put the cat command at the right time | 00:51:46 |
| Arian | So system creds are borked currently basically? | 10:16:26 |
| Arian | That's what I read between the lines. | 10:18:55 |
| phaer | At least when passed (insecurely) via kernelParam. We don't seem to test that scenario yet | 12:46:53 |
 emily | it does not | 14:02:53 |
| emily | it should, but it requires some more work around /tmp | 14:03:08 |
|  syd installs gentoo (they/them) changed their display name from syd 9291 (they/them) to syd installs gentoo (they/them). | 16:26:38 |
| ElvishJerricco | In reply to @arianvp:matrix.org
So system creds are borked currently basically? Maybe? I need to do more testing outside of this one repo | 19:28:03 |
| ElvishJerricco | phaer: Ok, confirmed there's a systemd-initrd-specific bug you're dealing with here: https://gist.github.com/ElvishJerricco/dca95eb4ea9fc410bd525c3b15b68fdd | 20:21:58 |
| ElvishJerricco | that test fails if systemd initrd is enabled, but not if it's disabled | 20:22:12 |
| 19 Jan 2025 |
 jade_ | this is done iirc, both cppnix and lix have temp-dir by now | 03:23:23 |
| jade_ | but maybe there's additional breakage around that | 03:23:37 |
| emily | I think only Lix does | 03:24:03 |
| emily | but I believe there were some further subtleties | 03:24:09 |
| jade_ | probably. i unfortunately do not have a memory | 03:24:20 |
| emily | I run with /nix/tmp and it works okay but I remember there being some missing pieces to put it all together | 03:24:38 |
| emily | it's also kind of an awkward flag day because migrating existing installations isn't practical | 03:24:46 |
