Tuesday, 2023-03-07

espenflHi there. Did I do the update correctly here: https://review.opendev.org/c/openstack/diskimage-builder/+/876084/1. Sorry for asking, I am just not familiar with gerrit and git-review. I think that the failing tests are mostly fixed now. The remaining one, I know little about.08:01
clarkbespenfl: that created a second followup change rather than updating the existing change. Typically with Gerrit you want to make each commit a logical passing entity. In this case I think both changes should be squashed into one15:52
clarkbespenfl: in this case you can do that by running `git review -d 876084` to download the two changes into a branch, then `git rebase -i HEAD~2` set the second commit line to squash to squash it into its parent. Edit the commit message to preserve the original commit message and its change id. Then run `git review` to push the result abck to gerrit15:53
espenflclarkb: Thanks, I will try that tomorrow.20:45
espenflso maybe the typical approach would be to first do git-review and then on future commits to update that "PR" I just do a regular git commit and push?20:47
JayFSo the pattern I follow, this is not the only pattern20:51
JayFis a single commit, git-review, gets pushed20:51
JayFI need to do a revision. I revise my files, and I `git commit --amend` to amend my previous commit with these new changes. I run git-review again.20:51
JayFFor multi-related-change workflows it gets more complex; but if you're just starting out and not chaining changes, that should work :)20:52
JayFalso, don't sweat pushing the bonus change earlier -- I think it's borderline a rite of passage for people using gerrit for changesets for the first time :D20:52
clarkbjlvillal: I believe the arch is actuall aarch6420:54
JayFyeah, that's what I was about to say :D 20:54
jlvillalTrying to DIB to build a `fedora` image and it fails because it says `arm64` ARCH is not supported20:54
clarkbbut that seems to map it to arm64 intentionally20:54
jlvillalmaybe related to this: https://github.com/openstack/diskimage-builder/blob/174089a6a5f1ff4628a9d103e4328a1b1298e241/diskimage_builder/lib/common-defaults#L29-L3020:54
clarkbjlvillal: we (opendev) build with the container element instead20:54
jlvillaland: https://github.com/openstack/diskimage-builder/blob/174089a6a5f1ff4628a9d103e4328a1b1298e241/diskimage_builder/elements/fedora/root.d/10-fedora-cloud-image#L27-L2820:54
jlvillalIt fails on line 43 of 10-fedora-cloud-image20:54
jlvillalOkay. Finished copying stuff over from previous post on #openstack-infra20:55
clarkboh we don't build fedora for arm though20:55
clarkbso ya you may be blazing a new trail for fedora + arm and finding where assumptions were wrong20:55
jlvillalDarn. I hate blazing! ;)20:55
jlvillalI was afraid of that. Thanks.20:56
clarkbjlvillal: some logs would probably be helpful though.20:56
jlvillalI decided to try Fedora because Ubuntu keeps failing to build today. Was working yesterday. But I think new packages dropped but haven't hit http://ports.ubuntu.com/ :(20:56
clarkbI suspect the normalization there was done without consideration that some things actually use aarch64?20:57
jlvillal2023-03-07 20:57:04.307 | dib-run-parts Running /tmp/dib_build.DSfkK4iq/hooks/root.d/10-fedora-cloud-image20:57
jlvillal2023-03-07 20:57:04.313 | Error: unknown ARCH: arm6420:57
jlvillalThe most important part of the log.20:57
clarkbwe do build https://opendev.org/openstack/project-config/src/branch/master/nodepool/nb04.opendev.org.yaml#L239-L247 though20:57
clarkbjlvillal: oh reading that if you set ARCH explicitly to aarch64 it should skip over things20:58
clarkbbasically ifyou set ARCH then it doesn't try to infer it. I would try that20:58
jlvillalYeah. I should at least try that :)  See if it breaks other things or not.20:58
jlvillalWell it is moving along farther. I'll keep my fingers crossed that it will succeed. 21:00
jlvillalclarkb, JayF Thanks.21:00
clarkbalso you should totally not do this long term, but we mirror ubuntu-ports with reprepro which is supposed to verify that things are valid in the mirror before we publish them21:00
clarkbthis means in theory you can get around broken mirrors using our mirror. But we definitely don't intend for them to be used publicly as we may delete/remove/move/update etc at any time21:00
jlvillalOh cool.21:00
jlvillalI tried to use the DIB_DISTRIBUTION_MIRROR variable. But doesn't seem like it rewrites http://ports.ubuntu.com/ :(21:01
jlvillalNot sure if there should be a DIB_DISTRIBUTION_PORTS_MIRROR variable...21:02
clarkbjlvillal: https://opendev.org/openstack/project-config/src/branch/master/nodepool/nb04.opendev.org.yaml#L20121:04
clarkbthat is what we do21:04
clarkbits possible that isn't working though21:04
clarkbor maybe you have to use ubuntu-minimal to make it work21:04
jlvillalI was assuming that: https://github.com/openstack/diskimage-builder/blob/174089a6a5f1ff4628a9d103e4328a1b1298e241/diskimage_builder/elements/ubuntu/pre-install.d/01-set-ubuntu-mirror#L2021:06
jlvillaldid not handle ports.ubuntu.com. But maybe my regex parsing isn't correct.21:06
clarkbjlvillal: it may not. We don't use the ubuntu element though. We use ubuntu-minimal21:08
jlvillalOkay. I should give that a try. Thanks!21:08
clarkbwith ubuntu minimal you'll need to set up our initial user and add common things like ssh21:09
clarkbits really there to allow you to build up what you want without the assumptions from the ubuntu image builds. 21:10
jlvillalThanks!21:10
jlvillalFYI: Fedora seemed to build. Finished without errors.21:10
jlvillalFYI: Ubuntu for ARM64 is building again. Looks like the packages hit ports.ubuntu.com.21:13
JayFWe'll have to get you a hat fit for a trailblazer, John :D Thanks!21:19
jlvillalJayF, :)21:20

Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!