| bryceh | *wave* MacSlow | 05:02 |
|---|---|---|
| ahasenack | is there a way to pass DEB_BUILD_OPTIONS to a ppa build? I'm guessing not, but thought I should ask | 14:07 |
| cjwatson | ahasenack: no, you have to actually change the packaging. (We've discussed it in the past, albeit more about profiles than options I think, but it's never got past the discussion phase.) | 14:30 |
| ahasenack | ok, thx. Yeah, profiles would be a way | 14:30 |
| ahasenack | an ubuntu-advantage-tools dep8 test in bionic/arm64 is failing with this: | 14:39 |
| ahasenack | Installing for arm64-efi platform. | 14:39 |
| ahasenack | grub-install: error: efibootmgr: not found. | 14:39 |
| ahasenack | dpkg: error processing package grub-efi-arm64-signed (--configure): | 14:39 |
| ahasenack | installed grub-efi-arm64-signed package post-installation script subprocess returned error exit status 1 | 14:39 |
| ahasenack | Processing triggers for man-db (2.8.3-2ubuntu0.1) ... | 14:39 |
| ahasenack | Errors were encountered while processing: | 14:39 |
| ahasenack | grub-efi-arm64-signed | 14:39 |
| ahasenack | E: Sub-process /usr/bin/dpkg returned an error code (1) | 14:39 |
| ahasenack | ops, sorry, longer past than I thought | 14:39 |
| ahasenack | I know there are grub updates in flight, it this related? Should the test be run with proposed? | 14:40 |
| ahasenack | it's failing like this just in bionic/arm64 | 14:41 |
| jbicha | jawn-smith: since you touched icu last, will we be syncing icu 72 from Debian (starts a big transition)? | 15:56 |
| ahasenack | juliank: is this something known in bionic/arm64 related to the recent grub/signed/shim updates? Excerpt above, but also at the end of https://autopkgtest.ubuntu.com/results/autopkgtest-bionic/bionic/arm64/u/ubuntu-advantage-tools/20230201_124925_331ea@/log.gz | 17:29 |
| juliank | yes | 17:29 |
| ahasenack | should I re-run that with proposed, or just ignore the failure? | 17:30 |
| juliank | temporary mishap as broken 1.52 was released to release | 17:30 |
| juliank | it doesn't work until a new image with 1.54 has been built | 17:30 |
| juliank | or well a new image with the old shim now in updates :) | 17:30 |
| juliank | ah sorry | 17:30 |
| juliank | I read the wrong log | 17:31 |
| juliank | the bionic images are broken somehow because they have UEFI but not efibootmgr installed which is needed to actually install grub, so how did they install grub in the first palce | 17:31 |
| ahasenack | ¯\_(ツ)_/¯ | 17:32 |
| juliank | like there's no fix for that in proposed, that's not a regression | 17:32 |
| juliank | bionic images also can't install kernel upgrades on arm64 | 17:33 |
| ahasenack | it says it's upgrading grub-efi-*, so it must exist in the image already | 17:33 |
| juliank | yes we accidentally dropped the dependency on efibootmgr in 2020 I think | 17:33 |
| ahasenack | but the test did pass in 2023-01-28 01:49:44 UTC | 17:34 |
| ahasenack | https://autopkgtest.ubuntu.com/packages/u/ubuntu-advantage-tools/bionic/arm64 | 17:34 |
| ahasenack | just a few days ago | 17:34 |
| juliank | yes but it never passes when it needs to upgrade a grub | 17:34 |
| ahasenack | let me check some other recent bionic sru | 17:34 |
| juliank | :D | 17:34 |
| cjwatson | I deliberately patched grub-install in 2019 to use libefiboot/libefivar directly rather than efibootmgr | 17:35 |
| cjwatson | But possibly that didn't apply to arm64, I don't remember | 17:35 |
| juliank | we didn't pick that back to bionic | 17:35 |
| cjwatson | That's also post-bionic, so ... | 17:35 |
| cjwatson | Yeah | 17:35 |
| juliank | I think I'll just add a Depends: efibootmgr to grub-common tbh | 17:35 |
| juliank | I need to add support for gzipped kernels to grub-common too | 17:36 |
| juliank | So like that's the lowest impact fix | 17:36 |
| juliank | People without grub-efi stuff installed will get efibootmgr pulled in but meh | 17:36 |
| ahasenack | https://autopkgtest.ubuntu.com/packages/s/shim-signed is also failing, slightly different error | 17:36 |
| ahasenack | /bin/sh: 1: grub-efi-arm64-signed: not found | 17:37 |
| juliank | That's expected | 17:37 |
| juliank | The error is weird | 17:37 |
| juliank | But anyhow it depends on a newer fwupd than we have in bionic yet | 17:37 |
| cjwatson | There's no reason to put the dependency in grub-common rather than in grub-efi-*-bin | 17:38 |
| ahasenack | https://autopkgtest.ubuntu.com/packages/c/containerd/bionic/arm64 last run on the 26th, green, but a few days before the breakage started | 17:38 |
| cjwatson | It certainly used to be in grub-efi-arm64-bin among others | 17:38 |
| juliank | cjwatson: yes, but I don't want to respin all the grub2-unsigned grub2-signed | 17:38 |
| juliank | cjwatson: I only build them in kinetic | 17:38 |
| juliank | cjwatson: They are binary copied to bionic (the unsinged) | 17:38 |
| juliank | So you gotta update all release series when making a change | 17:39 |
| cjwatson | How does adding a dependency to grub-common rather than grub-efi-*-bin help you there? They're both from the same source package, right? | 17:39 |
| juliank | grub-common is in src:grub2, not src:grub2-unsigned | 17:39 |
| juliank | src:grub2 is not shared between releases | 17:39 |
| cjwatson | Ah I see, that got moved. Ugh. I think that's a bad reason for a spurious dep, but not my call | 17:40 |
| juliank | We split that up back in 2020 to have a single grub version for security supported platforms :) | 17:40 |
| juliank | cjwatson: Right we could backport libefivars support too but then it ends up in grub-common too, so the difference isn't large. | 17:40 |
| juliank | vorlon: What's your opinion? Add efibootmgr Depends to grub-common/bionic on arm64 amd64 (i386?), or respin the whole grub2-unsigned batch with a Depends: efibootmgr | grub2-common (>= 2.04-something?), or backport libefivars support to grub-common? | 17:43 |
| juliank | Like if users upgrade from bionic to focal their grub-common depends on libefivars anyhow, having bionic depend on efibootmgr there seems fair | 17:44 |
| juliank | That avoids the annoying dependency in the shared package just for bionic :) | 17:45 |
| juliank | Basically I lean towards grub-common as I try to converge on whatever we do in later series, so there's less surprises when backporting grub2-unsigned | 17:45 |
| juliank | And we can fast track the src:grub2/bionic SRU | 17:46 |
| juliank | vorlon: have a look here, that SRU I think we can fast track, dannf could verify and we could just release it same-day https://code.launchpad.net/~juliank/grub/+git/bionic/+merge/436705 | 17:50 |
| juliank | with the patch, grub-check-signatures is identical to all other series | 17:50 |
| vorlon | juliank: so, I need to step out for food, back in ~1.5h. I haven't had a chance to digest this and think about it seriously. But I think my default position is that we should not work around having to do a separate build of grub2-unsigned for bionic with ugly packaging tricks | 17:53 |
| juliank | We absolutely should not do separate builds, and we'd still need to respin all series to accomodate | 17:54 |
| juliank | so the versions don't go backward :) | 17:55 |
| juliank | We tried that once to make a change in bionic only and it was lost again in the tumult of the next security update so that's just risk | 17:55 |
| juliank | (that became https://bugs.launchpad.net/ubuntu/+source/grub2/+bug/1995751) | 17:57 |
| -ubottu:#ubuntu-devel- Launchpad bug 1995751 in grub2-unsigned (Ubuntu Bionic) "update to 2.04-1ubuntu47.4 drops zz-update-grub" [Undecided, Confirmed] | 17:57 | |
| juliank | xenial ESM too | 17:59 |
| vorlon | we'd have to do respins, yes; but I don't particularly agree that the greater risk is from having deltas across series, we should just have a policy of not clobbering them in updates like we do for everything else! trying to make a single package compatible across all series is a worse risk because of things like this | 17:59 |
| vorlon | anyway, food, argue more later :) | 17:59 |
| * juliank uploaded that branch to the queue just in case :) | 18:02 | |
| ahasenack | oh, you found an existing bug about this | 18:04 |
| ahasenack | https://bugs.launchpad.net/ubuntu/+source/grub2-signed/+bug/1936857 | 18:04 |
| -ubottu:#ubuntu-devel- Launchpad bug 1936857 in grub2-signed (Ubuntu Bionic) "grub-install: error: efibootmgr: not found." [Undecided, In Progress] | 18:04 | |
| ahasenack | also https://bugs.launchpad.net/ubuntu/+source/linux-hwe-5.4/+bug/1991676, more specific to autopkgtests | 18:05 |
| -ubottu:#ubuntu-devel- Launchpad bug 1991676 in grub2-signed (Ubuntu Bionic) "Package grub-efi-arm64-signed 1.173.2~18.04.1+2.04-1ubuntu47.4 from bionic-proposed fails to install/upgrade (grub-install: error: efibootmgr: not found.)" [High, Triaged] | 18:05 | |
| juliank | ahasenack: yeah dann reported it again this morning :) | 18:05 |
| juliank | dannf: fixes for the bugs you complained about yesterday: https://code.launchpad.net/~juliank/grub/+git/bionic/+merge/436705 | 18:06 |
| ahasenack | juliank: lunar is also failing in a related area in arm64, with proposed already it seems: https://autopkgtest.ubuntu.com/results/autopkgtest-lunar/lunar/arm64/u/update-motd/20230201_130652_2066a@/log.gz | 18:27 |
| ahasenack | update-alternatives: error: alternative link /usr/lib/shim/shimaa64.efi.signed is already managed by shimx64.efi.signed | 18:27 |
| juliank | ahasenack: yes we know, see my first reply to you, that was supposed to be about this | 18:34 |
| ahasenack | ok, cool | 18:34 |
| juliank | 🥺 | 19:41 |
| arrayboltCBook | [] | 19:43 |
| arrayboltCBook | Hmm... that emoji renders in Element but not at all in WeeChat on Lubuntu... | 19:44 |
| arrayboltCBook | (the one above, not my [], that was just silliness) | 19:44 |
| ogra | i only see two brackets here | 19:44 |
| ogra | ah 🙂 | 19:44 |
| ogra | lol | 19:44 |
| arrayboltCBook | I saw what looked like a block and was like "ok so we're sending blocks to the screen, ok, []" :P | 19:45 |
| jbicha | 🫠 | 21:25 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!