[07:35] Saviq: oh, by the way, I worked out (half of) why you need the aufs `sed`dery in your local spread thing. [07:37] If the lxd storage is on zfs, then it lacks a feature needed for schroot to use an overlay filesystem. [07:38] Ah, so that's the same you encountered [07:39] LXD seems to no longer set up a zfs loop mount by default, so CI works and my more recent setup works [07:40] There's ways to get schroot to not try overlay filesystems, and it's not useful in ephemeral LXD containers. [07:40] I'm actually on ZFS root, do that's what's used, but same problem [08:03] Could be simply --type=directory for mk-sbuild? [08:06] Oh here's a new one from the lab: [08:06] > Mir fatal error: mir::graphics::wayland::Display not initialized [08:14] "Could be simply --type=..." <- Nope. [10:27] > <@saviq:matrix.org> Oh here's a new one from the lab: [10:27] > > Mir fatal error: mir::graphics::wayland::Display not initialized [10:27] Hmm this is happening consistently now on particular HW in the lab, but I can't see anything that's changed software-wise. [10:28] If you have logs I'll take a look once I'm out of current rabbithole [10:34] Failing: [10:34] http://10.102.156.15:8080/job/cert-focal-mir-smoke-mir-kiosk-edge-intel-gfx/774/consoleText [10:34] Last success on this hardware: [10:34] http://10.102.156.15:8080/job/cert-focal-mir-smoke-mir-kiosk-edge-intel-gfx/763/consoleText [10:35] But downgrading mir* to the versions from the successful one doesn't solve. [10:35] Also another (non-nested) test run on the same hardware is fine: [10:35] http://10.102.156.15:8080/job/cert-focal-mir-smoke-mir-test-tools-edge-intel-gfx/708/console [10:36] So yeah. So far no clue. [10:37] I was able to reproduce, and non-nested failed as well. Only difference at that point was mir-kiosk being installed (but not running). Next I'll start from the working non-nested setup and see if/how I can get to a broken state [10:37] But that's in ~2h [12:29] Oh I see, it's because mir-kiosk dies [12:30] That "Display not initialized" is b/c the tests don't get access to the device [16:16] OK. This is GAH. Based on my findings, the offending mir-kiosk changes are… drumroll… [16:16] https://github.com/MirServer/mir/compare/0593df6a8f...987869cbef [16:17] 🤦 [16:26] Everything else is the same between the two snaps. The dependencies Mir built with. The things packaged into the snaps. Yet, one reliably (4 tries now) fails, the other, I've not seen fail yet. [16:34] -GitHub[m]:#mir-server- **[MirServer/mir]** Saviq assigned AlanGriffiths to [issue #2632](https://github.com/MirServer/mir/issues/2632): Miriway frequently crashes to GDM after exiting powersave [16:34] -GitHub[m]:#mir-server- **[MirServer/mir]** Saviq unassigned RAOF from [issue #2632](https://github.com/MirServer/mir/issues/2632): Miriway frequently crashes to GDM after exiting powersave [17:10] Why am I getting std::exception::what: Failed to find any platforms for current system when trying to run a simple Mir-based server? I have mir-platform-graphics-x installed. [17:11] I feel like this should be super obvious [17:12] The rest of the log should show which platforms are tried and (hopefully) why they aren't usable [17:15] * sophie-w sent a code block: https://libera.ems.host/_matrix/media/v3/download/libera.chat/d58cc978c66c7d3de02a0c9ef59cb25280bc674a [17:16] $ ls /usr/lib/x86_64-linux-gnu/mir/server-platform/ [17:16] graphics-gbm-kms.so.19 graphics-wayland.so.19 input-evdev.so.8 server-x11.so.19 [17:16] " mirserver: Found graphics driver: mir:x11 (version 2.7.0) Support priority: 0" - no DISPLAY variable? [17:17] wtf [17:17] ok fixed it [17:18] DISPLAY was set to wayland-1 [17:18] Ah! Mate likes to do that for weird reasons [17:18] "weird" [17:19] It seems to happen only when right-clicking in thunar and clicking "open terminal here" [17:19] So XFCE does it too? [17:19] And thunar is only opened when I open the downloads directory from FF [17:20] Apparently