| -queuebot:#lubuntu-devel- Unapproved: lubuntu-installer-prompt (noble-proposed/universe) [24.04.5 => 24.04.5.1] (lubuntu) | 05:29 | |
| mkukri | hello, i am investigating LP: #2060624 , and i do wonder, in the calamares based installer, what is/should be responsible for configuring grub-pc/efi's debconf entries during/after installation | 12:22 |
|---|---|---|
| -ubottu:#lubuntu-devel- Launchpad bug 2060624 in grub2 (Ubuntu) "lubuntu/xubuntu reinstall (& install) on dual boot system, grub does appear & offer OS choice" [Undecided, New] https://launchpad.net/bugs/2060624 | 12:22 | |
| mkukri | subiquity has logic to this itself, and there is also a cloud-init module that suppossed to do this (although it's less complete and more buggy) | 12:23 |
| mkukri | s/subiquity/curtin in this case just be clear, it applies to both desktop isntaller and server one | 12:24 |
| * guiverc can't help there sorry.. you need a developer. I did get same response on ubuntu-desktop-installer (xubuntu) BEFORE the format was forced. | 12:34 | |
| mkukri | than this issue might strictly be unrelated but it seems like calamares based ubuntu installs do not fill the grub-pc/efi debconf fields correctly | 12:59 |
| mkukri | interestingly grub2/enable_os_prober is still false in debconf, even in installations were the alternative OS shows up in the menu | 13:12 |
| mkukri | hmm in fact regenerating grub debconf with update-grub after the os booted reproduces the bad behavior | 13:14 |
| mkukri | guiverc, i seem to have reproduced this on the desktop installer too, and i have a hunch what's going on but i am gonna do a few more tests before proclaiming it for sure | 13:41 |
| guiverc | :| at creating it with ubuntu-desktop-installer (I like confirmation of what i experiened, but :( at it still being possible as the forced-format b/c of another issue I thought prevented it appearing with ubuntu-desktop-installer... | 14:03 |
| tsimonq2 | mkukri: There's debconf prompts for that? | 14:04 |
| mkukri | grub-{efi,pc}/install-devices needs to be filled in, otherwise grub is not reinstalled on package update | 14:05 |
| mkukri | this is what curtin does https://github.com/canonical/curtin/blob/204562c277b096229ca20ba546149deef1d85e38/curtin/commands/curthooks.py#L632 | 14:06 |
| mkukri | tsimonq2 ^ | 14:06 |
| mkukri | the os-prober one i think im wrong about | 14:06 |
| mkukri | it exists, but there is an ubuntu patch that should force it enabled if the existing grub.cfg was generated with os-prober enabled | 14:07 |
| mkukri | and btw definitely not release critical, there were similar bugs (grub not reinstalling) in cloud images for years. but it should eventually be fixed | 14:08 |
| tsimonq2 | mkukri: Ah okay, I see, thanks :) | 14:11 |
| tsimonq2 | If I had to guess off the top of my head, this is where the modification should be made: https://git.lubuntu.me/Lubuntu/calamares-settings-ubuntu/src/branch/ubuntu/noble/common/modules/before_bootloader_context.conf | 14:12 |
| tsimonq2 | Either that, or a dedicated contextualprocess to set the debconf prompts | 14:13 |
| * tsimonq2 would be interested in seeing what entries this results in *exactly* for the main Desktop Installer | 14:13 | |
| mkukri | tsimonq2 i can give you an example from a simple install, the more annyoing case is raid where you have to put multiple things into the multiselect | 14:58 |
| mkukri | so for example, a simple *BIOS* qemu VM results in *grub-pc/install_devices: /dev/disk/by-id/ata-QEMU_HARDDISK_QM00001 | 15:02 |
| mkukri | also i am not sure if lubuntu uses cloud-init at all, but the following recent bug might be relevant too https://bugs.launchpad.net/ubuntu/+source/cloud-init/+bug/2060695 | 15:05 |
| -ubottu:#lubuntu-devel- Launchpad bug 2060695 in cloud-init (Ubuntu) "24.04 grub-pc cannot upgrade on mirrored software RAID root disk" [Medium, Triaged] | 15:05 | |
| mkukri | tl;dr cloud-init's cc_grub_dpkg module can mess things up on first boot if enabled | 15:06 |
| tsimonq2 | mkukri: ack looking thanks :) | 15:51 |
| tsimonq2 | mkukri: I don't believe Lubuntu uses cloud-init, no | 15:52 |
| mkukri | seems to be my observation as well because, ive done some lubuntu installs to (try to) track down the above bug and grub-pc/install-devices was always empty | 15:55 |
| mkukri | so yeah i assume it should be set by your installer | 15:55 |
| tsimonq2 | mkukri: Calamares doesn't support RAID yet, interesting | 16:05 |
| arraybolt3 | I just recently did a side-by-side install of Kubuntu (which also uses Calamares) and Windows 10 and os-prober did the right thing... | 16:06 |
| arraybolt3 | does having the debconf fields not filled in result in an intermittent issue in this area? | 16:07 |
| tsimonq2 | arraybolt3: To my understanding, when the grub package is updated it looks for those debconf config values to install GRUB to the right place | 16:07 |
| tsimonq2 | postinst or similar | 16:07 |
| arraybolt3 | right | 16:07 |
| mkukri | arraybolt3, yeah the os-prober issue is seperate from this | 16:08 |
| arraybolt3 | I've seen very strange GRUB prompts during bootloader updates (or possibly even just kernel updates) that looked like something was awry with debconf | 16:08 |
| arraybolt3 | and that is NOT a good UX | 16:08 |
| mkukri | i just noticed that this _whle_ looking at tthat | 16:08 |
| arraybolt3 | so I'd say this is worth calling a release blocker in the event we can get a fix for it | 16:08 |
| mkukri | those should go away if debconf is set | 16:08 |
| mkukri | or there is grub-pc/cloud_style_installation but i recommend not setting that outside cloud images, that uses grub-probe instead | 16:08 |
| arraybolt3 | right | 16:08 |
| tsimonq2 | arraybolt3: What kind of errors did you see? just for the sake of justifying the fix | 16:09 |
| mkukri | i dont think this should give you errors, just debconf questions | 16:09 |
| tsimonq2 | ah okay that follows | 16:09 |
| mkukri | and in non-ineractive mode it should just silently leave the old grub installed on disk | 16:10 |
| arraybolt3 | yep, debconf questions | 16:10 |
| arraybolt3 | so you'd upgrade and then see "Where eddo you want to install GRUB?" | 16:10 |
| arraybolt3 | (or something like that) | 16:10 |
| arraybolt3 | s/eddo/do/ | 16:10 |
| arraybolt3 | (lag is horrible right now because I'm zsync'ing an ISO) | 16:10 |
| mkukri | the reason this didnt come up in 23.10 btw is because grub-pc was *never* updated due to a multi year workaround.... | 16:10 |
| mkukri | that i finally removed in 24.04 | 16:10 |
| arraybolt3 | aha | 16:11 |
| arraybolt3 | that makes a lot of sense | 16:11 |
| mkukri | and that's why cloud_style_installation was added, and why the cloud-init issue with subiquity was discovered too... | 16:12 |
| mkukri | things seem okay in noble, but i still have many discussions to have with various folks over the issues that were papered over for years | 16:12 |
| tsimonq2 | mkukri: Can I ask why you think this may not be a 24.04 release blocker? | 16:12 |
| tsimonq2 | (If I'm reading your earlier messages correctly, that is.) | 16:13 |
| mkukri | i guess i didnt have the debconf questions popping up in mind earlier | 16:13 |
| mkukri | i was only thinking of noninteractive upgrades | 16:13 |
| tsimonq2 | arraybolt3: oh... wait... does Kubuntu's update notifier even support debconf prompts? | 16:14 |
| arraybolt3 | Kubuntu's does not, Lubuntu's does. | 16:14 |
| arraybolt3 | (because I made very sure that it did :P) | 16:14 |
| tsimonq2 | I'd definitely call it a release blocker for Kubuntu then, what do you think? | 16:14 |
| arraybolt3 | Even on Kubuntu though this issue should trigger if the user upgrades via apt, which I did for a while and still do sometimes. | 16:14 |
| mkukri | is kubuntu and lubuntu the only flavours not using the new desktop installer? | 16:15 |
| arraybolt3 | Ubuntu Unity also. | 16:15 |
| tsimonq2 | mkukri: and Ubuntu Unity | 16:15 |
| tsimonq2 | arraybolt3: jinx :)_ | 16:15 |
| arraybolt3 | The main difficulty I'm seeing is, how do we get Calamares to tell us what device the bootloader was installed to? | 16:15 |
| * arraybolt3 digs in scripts | 16:15 | |
| tsimonq2 | arraybolt3: already there | 16:15 |
| tsimonq2 | globalstorage | 16:15 |
| arraybolt3 | right but I mean in a contextual process | 16:16 |
| arraybolt3 | Calamares doesn't expose everything to the shell script thingies | 16:16 |
| arraybolt3 | we could write a module in Python... | 16:16 |
| tsimonq2 | Calamares exposes all globalStorage values to contextualprocesses :) | 16:16 |
| arraybolt3 | but let's exhaust the other options first for sanity's sake | 16:16 |
| arraybolt3 | oh it does? Perfect. | 16:16 |
| mkukri | should i file a bug about this. and what package is appopriate? | 16:17 |
| tsimonq2 | mkukri: calamares-settings-ubuntu and yes please | 16:18 |
| arraybolt3 | tsimonq2: I see a contextualprocess can decide what code to run based on the value of a global storage key, but does it let you use the value of a global storage key as a variable? | 16:20 |
| tsimonq2 | ...oh? | 16:20 |
| arraybolt3 | The thing is we need to be able to say "take the global storage key, find the corresponding by-id device, stuff it in this debconf entry". | 16:21 |
| mkukri | https://bugs.launchpad.net/ubuntu/+source/calamares-settings-ubuntu/+bug/2063354 | 16:22 |
| -ubottu:#lubuntu-devel- Launchpad bug 2063354 in calamares-settings-ubuntu (Ubuntu) "Calamares based flavours do not set grub-{efi,pc}/install_devices debconf" [Undecided, New] | 16:22 | |
| tsimonq2 | arraybolt3: hey guess what, the bootloader module itself is written in Python :P | 16:22 |
| tsimonq2 | mkukri: Thank you! | 16:22 |
| arraybolt3 | tsimonq2: heh, well there you go | 16:22 |
| tsimonq2 | arraybolt3: do we want to move to Matrix? | 16:23 |
| arraybolt3 | tsimonq2: mkukri is here so I think this is OK here for now | 16:23 |
| tsimonq2 | arraybolt3: ok cool :) | 16:24 |
| mkukri | im end of day currently, but if you need help please tag my nick, ill check on IRC periodically | 16:24 |
| arraybolt3 | will do, thank you for your help! | 16:25 |
| arraybolt3 | tsimonq2: want me to tackle the patch or are you already doing it? | 16:26 |
| tsimonq2 | arraybolt3: I'm on it, will check in | 16:26 |
| tsimonq2 | arraybolt3: In the meantime any chance you could update RT and the bug report on what's happening? | 16:26 |
| tsimonq2 | (probs release blocker for all three flavors, eh?) | 16:27 |
| arraybolt3 | sure thing | 16:27 |
| tsimonq2 | thank you :) | 16:27 |
| arraybolt3 | and yes, probably a release blocker | 16:27 |
| mkukri | i assume the desktop installer based ones are fine? | 16:27 |
| mkukri | and maybe talk to the release team please | 16:27 |
| arraybolt3 | I do not know if the desktop installer based ones are fine or not. If they don't set the debconf prompt, they're probably not fine. | 16:29 |
| arraybolt3 | er, the debconf variable - you know what I mean | 16:29 |
| mkukri | curtin sets it | 16:29 |
| arraybolt3 | Either the user will get unsightly debconf prompts, OR GRUB won't end up updating most likely, which would leave a gaping security hole and opportunity for Secure Boot bypasses and potential failure to boot if the SBAT key is rotated out. | 16:30 |
| mkukri | ubuntu desktop itself is definitely fine | 16:30 |
| mkukri | (ubuntu server too) i do test test thoseregularly | 16:30 |
| arraybolt3 | kk | 16:30 |
| mkukri | and i think i tested xubuntu too yesteday, but dont hold me to that | 16:31 |
| * mkukri is going afk for a bit but will be back later | 16:31 | |
| tsimonq2 | arraybolt3: At this point, just wondering if setting it to /dev/disk/by-id/SOMETHING is the right thing compared to just /dev/sda1? | 16:48 |
| tsimonq2 | thoughts? | 16:48 |
| arraybolt3 | tsimonq2: /dev/sdXY are unstable and *will* change if the user changes drives (by, e.g., booting with a USB drive inserted). | 16:48 |
| arraybolt3 | Learned this the hard way some time ago, and it's pretty common advice on the Interwebs. | 16:48 |
| arraybolt3 | Figuring out the link between /dev/whatever and /dev/disk/by-id/SOMETHING should be easy enough in theory since the /dev/disk/by-id thingamajigs are just symlinks. | 16:49 |
| tsimonq2 | That follows, now just wondering how it works with the main desktop installer, if there are multiple /dev/disk/by-id entries that match to one disk | 16:49 |
| tsimonq2 | (can happen, from what I'm reading) | 16:49 |
| arraybolt3 | https://www.tutorialspoint.com/python/os_readlink.htm | 16:49 |
| arraybolt3 | oh really? That's not good... | 16:50 |
| arraybolt3 | mmm, I see that I have that situation on *this* system I'm typing on. | 16:50 |
| tsimonq2 | Bright side, the debconf config value is of type "multiselect" | 16:50 |
| arraybolt3 | for me nvme-eui.00253857214071d7 and nvme-Samsung_SSD_970_EVO_Plus_1TB_S6S1NJ0T713093F point to the same partition | 16:50 |
| arraybolt3 | er, the same NVME namespace I mean | 16:50 |
| arraybolt3 | gosh I wish that there weren't NVME namespaces *and* partitions :P | 16:50 |
| arraybolt3 | tsimonq2: maybe by-uuid would work too/instead? Just a thought. | 16:51 |
| tsimonq2 | arraybolt3: the other thing is that in this virtual machine I can only see one by-id mapping to /dev/sda1 | 16:52 |
| tsimonq2 | Do you have test hardware? Would be great to see what those entries are actually set to in such a case with the Ubuntu Desktop installer | 16:53 |
| arraybolt3 | tsimonq2: https://termbin.com/8n2f | 16:53 |
| arraybolt3 | hrm... I can make the hardware I have test hardware. | 16:53 |
| tsimonq2 | If you're okay with that, go for it | 16:54 |
| arraybolt3 | I lost my two best testing laptops, one to video card failure, and one to whatever power mess it suffers from, so I'm stuck using one desktop for everything test-wise :P | 16:55 |
| arraybolt3 | but anyway, I was going to test Xubuntu Minimal anyway so I'll give it a shot there | 16:55 |
| tsimonq2 | awesome, thanks | 16:55 |
| arraybolt3 | and maybe also try it in BIOS mode so we get a good picture | 16:55 |
| tsimonq2 | BIOS and UEFI would be great for full testing coverage, probably looking for `sudo debconf-show grub2` on first boot | 16:57 |
| arraybolt3 | nifty | 16:57 |
| tsimonq2 | Testing some code now which just sets an example value for that debconf setting - to test syntax before we have actual values to work with :) | 17:07 |
| tsimonq2 | arraybolt3: `sudo debconf-show grub-pc` not `sudo debconf-show grub2` my bad :) | 17:11 |
| tsimonq2 | I just did a `sudo dpkg-reconfigure grub-pc` on an installed system which gave me some ideas, would still like to see the output of that for both GRUB and EFI to double check though | 17:17 |
| arraybolt3 | tsimonq2: Might take me a bit, I'm just now starting to install Xubuntu for testing. | 17:21 |
| arraybolt3 | Also I'm multitasking between work duties and also Matrix Council activity | 17:21 |
| arraybolt3 | Got the installation going now. | 17:23 |
| tsimonq2 | arraybolt3: sweet thanks | 17:26 |
| arraybolt3 | tsimonq2: alright, so, from an EFI installation: | 17:28 |
| arraybolt3 | uh, `sudo debconf-show grub2` on Xubuntu Minimal outputs nothing. | 17:28 |
| arraybolt3 | Not even a newline. | 17:28 |
| tsimonq2 | arraybolt3: 17:11 < tsimonq2> arraybolt3: `sudo debconf-show grub-pc` not `sudo debconf-show grub2` my bad :) | 17:28 |
| tsimonq2 | :) | 17:29 |
| arraybolt3 | oh lol, alrighty | 17:29 |
| arraybolt3 | ok, grub-pc shows this: | 17:29 |
| arraybolt3 | uh, lemme pastebin | 17:29 |
| arraybolt3 | https://termbin.com/bn4o | 17:30 |
| tsimonq2 | can I get the whole thing? | 17:30 |
| arraybolt3 | I have the ls -l from /dev/disk/by-id coming | 17:30 |
| arraybolt3 | and that is the whole thing | 17:30 |
| tsimonq2 | wait, the whole output? | 17:30 |
| arraybolt3 | That's literally the result of `sudo debconf-show grub-pc | nc termbin.com 9999` | 17:30 |
| arraybolt3 | so yes, that's the whole output | 17:30 |
| tsimonq2 | oh wow this is interesting, ok, yes please what is the /dev/disk/by-id output? | 17:31 |
| arraybolt3 | frustratingly it looks like this particular physical machine with a SATA drive does NOT have multiple /dev/disk/by-id entries that point to the same device! | 17:31 |
| arraybolt3 | https://termbin.com/y01a | 17:31 |
| arraybolt3 | I didn't expect that | 17:31 |
| arraybolt3 | And I cannot safely do this test on my primary laptop. | 17:31 |
| tsimonq2 | arraybolt3: On your primary laptop, any chance you could just run `sudo dpkg-reconfigure grub-pc`, make sure all the entries are correct, then get the debconf-show output from that? | 17:33 |
| tsimonq2 | also looking on my end | 17:33 |
| arraybolt3 | grr, well... https://termbin.com/xiuu now you don't get any device at all... because this is a 22.04 installation on my main laptop. | 17:35 |
| tsimonq2 | aaaaaaaHA | 17:35 |
| tsimonq2 | ok it's only one device | 17:35 |
| tsimonq2 | but it picks the first from the list | 17:35 |
| arraybolt3 | ok good whatever works :P | 17:36 |
| tsimonq2 | arraybolt3: I appreciate ya :) | 17:36 |
| arraybolt3 | sorry I wasn't much help | 17:36 |
| arraybolt3 | I tried *really hard* though | 17:36 |
| tsimonq2 | you helped more than you think - only one debconf setting for that package is a good one | 17:36 |
| arraybolt3 | ah ok | 17:40 |
| mkukri | make sure to test cryptsetup, LVM and those types of things if you support it too. ive seen weird bugs elsewhere with that before | 17:50 |
| arraybolt3 | mkukri: we have a pretty immense checklist for the Cala-based flavors that covers almost all those weird edge cases (except for LVM which Doesn't Work(TM) with Calamres yet). | 17:54 |
| arraybolt3 | tsimonq2: you still need the BIOS install output? I'm about to reinstall in BIOS mode for that. | 17:54 |
| arraybolt3 | oh fun, Dell's BIOS has arbitrarily decided to gray out my legacy boot option for who-knows-why | 17:55 |
| arraybolt3 | oh there we go, had to enable legacy option ROMs first | 17:55 |
| tsimonq2 | arraybolt3: That would help :) | 17:58 |
| arraybolt3 | kk, in progress... | 17:58 |
| arraybolt3 | (man this installer takes like time and eternity to load) | 17:59 |
| tsimonq2 | Prototype fix: | 17:59 |
| tsimonq2 | +def get_disk_id(device_name): | 17:59 |
| tsimonq2 | + by_id_path = "/dev/disk/by-id" | 17:59 |
| tsimonq2 | + device_path = os.path.realpath(device_name) | 17:59 |
| tsimonq2 | + | 17:59 |
| tsimonq2 | + for entry in os.listdir(by_id_path): | 17:59 |
| tsimonq2 | + full_entry_path = os.path.join(by_id_path, entry) | 17:59 |
| tsimonq2 | + if os.path.realpath(full_entry_path) == device_path: | 17:59 |
| tsimonq2 | + return full_entry_path | 17:59 |
| tsimonq2 | [...] | 17:59 |
| tsimonq2 | + dev_id_path = get_disk_id(boot_loader["installPath"]) | 17:59 |
| tsimonq2 | + check_target_env_call(["echo 'grub-pc grub-pc/install_devices multiselect " + dev_id_path + "' | debconf-set-selections"]) | 17:59 |
| arraybolt3 | Install in progress | 18:00 |
| arraybolt3 | I don't get the details of what that function does but it looks vaguely right to me :) | 18:01 |
| tsimonq2 | the one thing that's verylikely to cause issues is check_target_env_call | 18:02 |
| tsimonq2 | and yep just ran into it not DTRT - coooooooooool | 18:02 |
| arraybolt3 | needs sudo rights perhaps? | 18:03 |
| arraybolt3 | or are you not chroot'd in? | 18:03 |
| arraybolt3 | mmm, you probably need to run that command from within a chroot environment | 18:04 |
| arraybolt3 | tsimonq2: alright a BIOS system has some significantly more interesting keys: https://termbin.com/qixa | 18:05 |
| arraybolt3 | not sure which ones you care about (probably things like grub2/kfreebsd_cmdline_default are ignorable :P), but there you go | 18:06 |
| arraybolt3 | and also ls -l from /dev/disk/by-id: https://termbin.com/15h8 | 18:06 |
| tsimonq2 | check_target_env_call means "call this in the target environment and check the result" :) | 18:07 |
| tsimonq2 | arraybolt3: looking thanks! | 18:07 |
| arraybolt3 | ah shoot then you *are* chroot'd in it sounds like | 18:07 |
| arraybolt3 | well, I'm sure you'll figure it out, good luck and I'll be around to try and help | 18:08 |
| tsimonq2 | thanks - also that BIOS output helps thank you :) | 18:08 |
| tsimonq2 | arraybolt3: I have working code for BIOS, testing for EFI | 18:44 |
| arraybolt3 | awesome | 18:52 |
| tsimonq2 | arraybolt3: can I rubber ducky off you real quick? | 18:59 |
| tsimonq2 | BIOS is good, EFI is the last hurdle here | 19:00 |
| tsimonq2 | It seems like in the bootloader module it's being passed as /boot/efi, not the actual partition name | 19:00 |
| arraybolt3 | oooh, ok | 19:00 |
| tsimonq2 | So when grabbing the device ID for EFI, it should grab the ID corresponding to the partition mounted at /boot/efi yes? | 19:01 |
| arraybolt3 | In that instance figure out what partition is mounted at /boot/efi and then use that? | 19:01 |
| arraybolt3 | jinx | 19:01 |
| arraybolt3 | so yeah, that sounds right | 19:01 |
| tsimonq2 | sounds good :) finding some coffee | 19:01 |
| * genii twitches | 19:01 | |
| mkukri | into install_device you need to put the actual disks | 19:36 |
| mkukri | tsimonq2, arraybolt3. grub-multi-install internally finds all the ESPs from that and mounts them | 19:36 |
| mkukri | /boot/efi is just a fallback and that mountpoint is ignored when you set install devices | 19:37 |
| mkukri | and grub-pc also expects the bare disks, but that part i think is obvious | 19:37 |
| tsimonq2 | mkukri: Oh, just the actual disks? | 19:39 |
| tsimonq2 | mkukri: Nothing under like /dev/disk/by-id or /dev/disk/by-uuid, just e.g. /dev/sda1 ? | 19:40 |
| mkukri | seems like i am wrong sorry, i checked a desktop installer install and it seems to have the ESP device node https://imgur.com/a/nVSnPGj | 19:40 |
| mkukri | the by-id and by-uuid part is still there yeah. i was just trying to say that it wasnt expecting the partition but it looks like it is | 19:40 |
| tsimonq2 | No problem! Quick question though... | 19:41 |
| tsimonq2 | Does by-uuid work just as well as by-id in this case? | 19:41 |
| mkukri | well, i do want to switch to by-uuid *eventually* but it's not tested yet | 19:41 |
| mkukri | and just to be clear this is the debconf options i get https://imgur.com/a/MvCEVlQ and the above image is the result in debconf db | 19:42 |
| tsimonq2 | mkukri: The current roadblock I have is, for some reason I can only query by-uuid on EFI and by-id on BIOS | 19:42 |
| tsimonq2 | Otherwise I have working code now | 19:43 |
| mkukri | by-id should work on EFI | 19:44 |
| mkukri | how are you trying to query it? | 19:44 |
| tsimonq2 | $ ls -l /dev/disk/by-id/ | 19:44 |
| tsimonq2 | total 0 | 19:44 |
| tsimonq2 | lrwxrwxrwx 1 root root 9 Apr 24 13:45 ata-QEMU_DVD-ROM_QM00001 -> ../../sr0 | 19:44 |
| mkukri | also if you are using a if=virtio disk in qemu you have to give it a serial number on the command line or it wont show up | 19:45 |
| tsimonq2 | mkukri: Simple ls of /dev/disk/by-id | 19:45 |
| tsimonq2 | oooh | 19:45 |
| tsimonq2 | mkukri: I am using QEMU, how does that part work? | 19:45 |
| mkukri | what command line flag are you using to specify your hard disk? | 19:45 |
| mkukri | or is it virt manager? or LXD? | 19:45 |
| tsimonq2 | virt-managere | 19:45 |
| tsimonq2 | s/managere/manager/ | 19:45 |
| mkukri | ah i am not sure what that does, but i am guessing it doesnt give the disk a serial number | 19:46 |
| mkukri | this is a problem with regular ubuntu as well in that case the debconf wont be filled either | 19:46 |
| mkukri | but the fallback install via /boot/efi still works | 19:46 |
| mkukri | if you want to test it similar to real hardware, try to specify a disk serial number for your VM in the options somewhere | 19:46 |
| mkukri | g | 19:46 |
| mkukri | apparently you can put <serial>some-arbitrary-serial</serial> in the vm's xml inside the <disk> tag | 19:47 |
| tsimonq2 | ok looking then thanks | 19:47 |
| tsimonq2 | mkukri: So to be clear, as for right now you'd recommend against going with by-uuid? | 19:48 |
| mkukri | if you want to have the same bugs as us then yes :) | 19:49 |
| mkukri | by uuid is the future plan but maybe it's better if we all switch at once in the 24.10 or 25.04 cycle | 19:49 |
| tsimonq2 | cool, when we do get to that point it'll be a one-liner :) | 19:50 |
| mkukri | there are problems like UUID's not actually always being unique :/ | 19:50 |
| * mkukri looks at our cloud images in disappointment..... | 19:51 | |
| tsimonq2 | oh wowww, okay, didn't know that | 19:51 |
| mkukri | the issue is, the images are obviously pre generated, so you'd need to replace the uuid with a random one on first boot, but no one has bothered yet | 19:52 |
| tsimonq2 | ahh okay | 19:53 |
| tsimonq2 | mkukri: yes that <serial>ABCDGEF</serial> fix worked \o/ | 19:53 |
| tsimonq2 | ok doing a final round of testing before uploading, should be in the clear in the next hour here | 19:54 |
| mkukri | im glad to hear, that took me many many hours of headbanging to figure out once | 19:54 |
| mkukri | so you said you dont have LVM or RAID. but do you have cryptsetup? | 19:54 |
| mkukri | and maybe ZFS? | 19:54 |
| mkukri | did you test those? | 19:55 |
| tsimonq2 | I can do a round of testing with those, yeah. For now I have just been concerned with making sure regular non-encrypted BIOS and EFI work | 19:57 |
| tsimonq2 | We don't do ZFS, RAID, or LVM at this current point in time, but encrypted installs definitely | 19:57 |
| mkukri | okay cool im just trying to be exhaustive | 19:58 |
| tsimonq2 | no you're okay :) thanks for that | 19:58 |
| mkukri | and amd64 bios and uefi are your only GRUB based platforms? | 19:58 |
| tsimonq2 | yes | 19:58 |
| mkukri | okay i think, i guess that's all the cases than | 19:59 |
| tsimonq2 | cool cool :) | 19:59 |
| -queuebot:#lubuntu-devel- Unapproved: calamares (noble-proposed/universe) [3.3.5-0ubuntu3 => 3.3.5-0ubuntu4] (lubuntu, ubuntustudio) | 20:30 | |
| -queuebot:#lubuntu-devel- Unapproved: rejected lubuntu-installer-prompt [source] (noble-proposed) [24.04.5.1] | 20:42 | |
| -queuebot:#lubuntu-devel- Unapproved: lubuntu-installer-prompt (noble-proposed/universe) [24.04.5 => 24.04.6] (lubuntu) | 20:45 | |
| -queuebot:#lubuntu-devel- Unapproved: accepted lubuntu-installer-prompt [source] (noble-proposed) [24.04.6] | 20:50 | |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!