bump linuxkit version, init version, place etc/issue in good place, fix linuxkit_cmdline parsing in make-roots #4161
Conversation
deitch
force-pushed
the
lkt-version-issue
branch
from
86a0d2b to
69b03fa
Compare
| @@ -1,9 +1,9 @@ | |||
| kernel: | |||
| image: KERNEL_TAG | |||
| cmdline: "rootdelay=3" | |||
| # the unified_cgroup_hierarchy forces cgroupsv1, which is required until pillar is ready to support v2 | |||
| cmdline: "rootdelay=3 linuxkit.unified_cgroup_hierarchy=0" | |||
There was a problem hiding this comment.
@rouming will this impact your RT work? I think you talked about cgroupv2.
There was a problem hiding this comment.
Shouldn't be a problem. Everything now is v1, this makes it explicit. Actually not only does pillar have an issue with v2, runc needs some kernel settings that we don't have enabled. @christoph-zededa knows about them I think.
Either way, when we are ready, just remove that line and it runs v2. You even can do it interactively at grub time.
There was a problem hiding this comment.
We do not support v2, so as @deitch said, this makes it explicit. I'm fine with the change.
|
Just tried rerunning a failing test, to see if it was transient. Unlikely, since all of the eden tests failed, but worth trying. No luck, it still failed at the same point. |
deitch
force-pushed
the
lkt-version-issue
branch
from
69b03fa to
4b4aaea
Compare
|
I had made an error in the PR. That is what CI is for! Needs a new approval to kick off eden tests again, please. |
|
I really like the PR description; it explains the changes in the way I would like to see it in the commit messages. But the commit messages are empty =( Once the PR is merged this valuable info will be lost and not appear in the git history =( |
|
@OhmSpectator those are fair comments. I don't want to make a change to the commits - even just messages - in a way that will cause the whole long CI to run again. If these pass (still have 2 spurious failures, which I restarted), then once those are green, I will change the commit messages, push out, and merge immediately; no need to wait for full hours-long CI if the only change is the commit message content. If I have to make changes because of actual failures, I will do it then. |
|
Finally, all clean. I will update the commit messages and then merge. |
…text until actual rootfs build Signed-off-by: Avi Deitcher <[email protected]>
… cgroups v1; fix bug in kernel cmdline parsing in make-rootfs Signed-off-by: Avi Deitcher <[email protected]>
deitch
force-pushed
the
lkt-version-issue
branch
from
4b4aaea to
337901e
Compare
|
Done @OhmSpectator , much more detailed commit comments. Nothing changed in code, so I am just going to merge this in. CI was 100% green before. |
Several small changes in this PR, split across two commits, as each stands on its own. These also help prepare the groundwork for the installer rearchitecture, which almost is done, and should have a real PR in the coming days.
/etc/issue- which contains EVE logo and name on login - topkg/eve/installer/etc_issue, where it gets placed in thedist/<arch>/<version>/installer/directory, and then placed in the rootfs by afilesentry inrootfs.yml.in. This makes it easier to find and edit, and more importantly, reusable in other builds.make-rootfsscripts that consumed that had a bug, so fixed those as well.I could have held off with these until installer PR is ready, but these are much better on their own. It makes the installer PR smaller and simpler, and if there are issues with the installer PR, these will not be affected.