=== cpaelzer__ is now known as cpaelzer | ||
=== didrocks999 is now known as didrocks | ||
LocutusOfBorg | juliank, hello, you there? | 09:38 |
---|---|---|
LocutusOfBorg | I tried to upgrade an Ubuntu 20.04 installation and got this | 09:38 |
LocutusOfBorg | https://paste.ubuntu.com/p/wh7p4b7ng4/ | 09:38 |
juliank | LocutusOfBorg: yes | 09:38 |
LocutusOfBorg | I'm pinging you because looks suspiciusly related to 1872077 | 09:39 |
juliank | that's suboptimal | 09:39 |
LocutusOfBorg | but I have a recent grub | 09:39 |
juliank | LocutusOfBorg: new disk? | 09:39 |
LocutusOfBorg | mmm that usb key is not connected | 09:39 |
LocutusOfBorg | it was probably connected when the system was installed | 09:39 |
juliank | it should have prompted you for the missing disk, though | 09:40 |
LocutusOfBorg | I run apt-get dist-upgrade, no graphics | 09:40 |
LocutusOfBorg | oh well there is a grub in proposed... | 09:41 |
juliank | Well it's a debconf prompt | 09:43 |
xnox | i would try dpkg-reconfigure..... but not sure on which package =) | 09:43 |
LocutusOfBorg | I tried dpkg-reconfigure, and failed, as well as the one in proposed | 09:44 |
LocutusOfBorg | I even tried a grub-install | 09:45 |
LocutusOfBorg | sudo dpkg-reconfigure grub-efi-amd64-signed | 09:45 |
LocutusOfBorg | fails because not fully installed | 09:45 |
LocutusOfBorg | sudo dpkg-reconfigure grub-pc does something | 09:47 |
LocutusOfBorg | and dpkg-reconfigure on every grub-package seems to do mostly nothing | 09:47 |
juliank | gotta dpkg --configure -a I assume | 09:48 |
juliank | well dpkg --configure grub-efi-amd64-signed | 09:48 |
LocutusOfBorg | fails with that error | 09:48 |
LocutusOfBorg | https://paste.ubuntu.com/p/QGmQWkzPTH/ | 09:49 |
juliank | Set set -x to /var/lib/dpkg/grub-efi-amd64-signed.postinst and /usr/lib/grub/grub-multi-install and run it again | 09:49 |
LocutusOfBorg | sorry for LANG | 09:49 |
juliank | and attach a dump of debconf-get-selections | grep grub-efi | 09:49 |
juliank | xnox: I wish we were using UUIDs for grub disks rather than the physical ids :/ | 09:50 |
juliank | Such that I can copy my disk to a new one and it still works | 09:50 |
LocutusOfBorg | https://paste.ubuntu.com/p/f6DDkVckST/ | 09:51 |
juliank | but I just copied the existing code rather than trying to improve it :) | 09:51 |
juliank | LocutusOfBorg: It seems to me like you have no ESPs on that system | 09:52 |
LocutusOfBorg | I don't even know what esp is | 09:52 |
juliank | The EFI system partition | 09:52 |
juliank | INPUT critical grub-efi/install_devices_disks_changed | 09:53 |
juliank | is skipped | 09:53 |
juliank | but then it also had no choices | 09:53 |
juliank | one of the partition types should be 0xef | 09:54 |
juliank | a FAT partition | 09:54 |
juliank | I think what you have here is a weird system with a wrong ESP | 09:54 |
juliank | Like, it's a FAT partition that contains ESP contents | 09:54 |
juliank | but it has the wrong partition type in the partition table | 09:55 |
juliank | some systems can boot from those | 09:55 |
LocutusOfBorg | here my mount https://paste.ubuntu.com/p/DkZT39kVqh/ | 09:55 |
cjwatson | juliank: disks don't have useful UUIDs - filesystems do | 09:55 |
LocutusOfBorg | I have a /boot/efi | 09:56 |
juliank | cjwatson: hmm yes, for ESPs UUIDs would work, but I copied code that installed to disks/partitions | 09:56 |
juliank | LocutusOfBorg: Yes, so you selected that USB stick at one point during upgrading, and that ESP mounted to /boot/efi was migrated to the debconf database too | 09:56 |
juliank | But it's not actually a proper ESP | 09:57 |
cjwatson | That code was specifically designed for BIOS | 09:57 |
juliank | cjwatson: I'm not sure how much it matters, though, because who copies their entire ESP? | 09:57 |
juliank | :D | 09:57 |
LocutusOfBorg | juliank, I did a upgrade and dist-upgrade without touching anything... do you have an hotfix? | 09:57 |
juliank | LocutusOfBorg: Open fdisk, and change the paritition type of /dev/sda1 from 0xb to 0xef | 09:58 |
juliank | 0xb is 32-bit FAT, 0xef is ESP (which is also 32-bit FAT, well technically a subset, but who cares) | 09:59 |
LocutusOfBorg | well, it worked!!! | 10:00 |
LocutusOfBorg | looks like it prompted me for the change and only one selection was available | 10:00 |
LocutusOfBorg | now, let me add some background | 10:00 |
LocutusOfBorg | we use systemback to install systems easily | 10:00 |
juliank | cjwatson: I guess debconf skips multichoice questions with no choices? | 10:00 |
LocutusOfBorg | so that might have broken with 20.04 even if really tested | 10:00 |
juliank | xnox: So ... I guess we also need to show any VFAT partition on the system, not just proper ESPs | 10:01 |
juliank | xnox: Because clearly LocutusOfBorg's system booted from that ESP despite it having the wrong partition type in the partition table | 10:02 |
juliank | which raises more questions | 10:03 |
juliank | There are hidden 32-bit, hidden 16-bit, and of course 16-bit FAT partitions | 10:03 |
juliank | and some more strange partition types | 10:03 |
juliank | i found this list https://thestarman.pcministry.com/asm/mbr/PartTypes.htm | 10:03 |
juliank | Maybe we need to look at the file system | 10:04 |
juliank | rather than the partition | 10:04 |
juliank | At least, if that's been a choice previously, it should be shown | 10:04 |
juliank | LocutusOfBorg: Could you put the data into a bug report? | 10:06 |
juliank | this clearly needs fixing | 10:06 |
cjwatson | juliank: Uh, I don't remember. Do you need me to dig? | 10:06 |
juliank | cjwatson: Nah, i was just curious a bit | 10:07 |
cjwatson | "Unlike select lists, multiselect questions are visible if there is just one choice." | 10:07 |
cjwatson | say the docs | 10:07 |
juliank | right, we had no choices at all here which is not a useful choice to have :D | 10:07 |
juliank | But that means I miss a prompt for systems that have no suitable partition | 10:08 |
cjwatson | They're actually visible if there are zero choices, which seems to me to be a possible bug | 10:08 |
juliank | Like "Your system does not have a suitable EFI system partition. Cannot continue." | 10:08 |
juliank | Though | 10:08 |
juliank | Hmm | 10:08 |
juliank | It should ask the _empty question | 10:08 |
cjwatson | Oh of course, multiselect obviously has to be visible with one choice because multiselect is more like a set of checkboxes than a radio button or whatever | 10:09 |
juliank | yup | 10:09 |
juliank | So I need to see why the code never prompted grub-efi/install_devices_empty despite there being no disks found | 10:09 |
juliank | Ah | 10:10 |
juliank | hmm | 10:10 |
LocutusOfBorg | juliank, against which package? | 10:18 |
LocutusOfBorg | grub2 right? | 10:18 |
juliank | LocutusOfBorg: grub2 | 10:18 |
LocutusOfBorg | ack, affecting groovy and focal | 10:18 |
juliank | yeah | 10:18 |
LocutusOfBorg | I'll point this link to the bug report | 10:19 |
LocutusOfBorg | https://irclogs.ubuntu.com/2020/10/12/%23ubuntu-devel.html | 10:19 |
juliank | LocutusOfBorg: Helpfully it's the first discussion today, even :D | 10:21 |
LocutusOfBorg | it is :D | 10:21 |
LocutusOfBorg | #1899462 | 10:22 |
LocutusOfBorg | LP: #1899462 | 10:25 |
ubottu | Launchpad bug 1899462 in grub2 (Ubuntu Groovy) "grub2, cannot upgrade due to EFI partition type mismatch" [High,New] https://launchpad.net/bugs/1899462 | 10:25 |
juliank | LocutusOfBorg: thanks | 10:31 |
LocutusOfBorg | thanks to you for the great help! | 10:31 |
LocutusOfBorg | I admit this seems a serious bug to me, it prevents upgrades... | 10:31 |
juliank | ack | 10:32 |
juliank | It affects a few people | 10:33 |
juliank | But it's obviously not a problem for most | 10:33 |
LocutusOfBorg | I don't know if the system was still bootable after this "incomplete upgrade" | 10:33 |
LocutusOfBorg | this is why I just set up high severity | 10:33 |
juliank | It probably was, yes | 10:33 |
LocutusOfBorg | oh this is less worrysome then :P | 10:33 |
juliank | The EFI binaries are single binaries, they don't need external modules, and will continue to work if not updated | 10:34 |
juliank | Well, except the new shim does not trust old grubs | 10:34 |
juliank | um | 10:34 |
juliank | old grubs were revoked | 10:34 |
seb128 | bdmurray, hey, did you have any chance to poke at the retracers issue on 20.10? | 15:01 |
=== _hc[m] is now known as _hc | ||
bdmurray | seb128: A little bit and then I discovered an issue with the Launchpad retracers which I sorted out. I'll look more at the Error Tracker ones today. | 15:15 |
seb128 | bdmurray, thanks | 15:25 |
seb128 | bdmurray, do buckets with a failed retrace get stucked on the status until someone reset them or something or do they keep trying to retrace new reports until they get a successful backtrace? | 15:27 |
bdmurray | seb128: They keep retrying for 20.10 as of last week or so - I'd missed that. | 15:29 |
bdmurray | seb128: I do see a lot more "Create" links on the errors.ubuntu.com page for 20.10 | 15:31 |
seb128 | bdmurray, right, still not for the top of the list though, still somewhat weird | 15:32 |
bdmurray | yeah, gnome-shell has quite a few failures | 15:34 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!