New test - rpm-ostree-container-bootc + related improvements #1096
Conversation
- pass KSTEST_* variables into the container env - ensure disk images are kept with $KSTESTS_KEEP == 2
The ostree environments have most of the filesystem read-only (including /usr), but this can be overcome via an overlay. In such case the test script intended for run in the system resides in /var/lib/extensions/kickstart-tests/usr/libexec instead of the usual location (/usr/libexec).
![github-advanced-security[bot]](https://avatars.githubusercontent.com/in/57789?s=60&v=4){kind=small}
jikortus
force-pushed
the
ostreecontainer-bootc
branch
from
fc5599e to
cad936d
Compare
|
/test-platforms |
1 similar comment
|
/test-platforms |
It tests installation of a bootable container image (with bootc/bootupd) and boot of the installed system, as well as --remote and --stateroot parameters of ostreecontainer kickstart command. It's based on previous work from Vladimir Slavik.
jikortus
force-pushed
the
ostreecontainer-bootc
branch
from
cad936d to
11bd404
Compare
|
It seems that the test was failing with RHEL-9 due to |
|
Sorry, I started to look at the patchset, did some wrong review and deleted it. |
|
Nice work. I am suggesting updates upon your great work in #1116. Mostly because our common handling of os (rhel / fedora) substitutions at various places is not very clear at first sight: We do all the per-os substitutions in the .ks.in. I guess you did them also in .sh because you were not aware of the fragments mechanism we use for kickstart commands ( We should even do it more concisely as in the follow-up commit of the PR: f965f8b I would also suggest submitting the first commit of this PR as a separate PR as it is not required by the test and reviewing it separately will simplify the process of getting the changes in (I would maybe even split it to 2 PRs). |
| # Check of the remote URL is appended by the test script | ||
|
|
||
| # Checks after boot | ||
| cat >> /opt/kickstart-tests/kickstart-test.sh << 'EOF' |
There was a problem hiding this comment.
This part was failing for me with file not found, updated here: #1116
|
@rvykydal, thanks a lot for your feedback and for the improvements you've made in #1116. At this point, I think we could abandon this PR in favour of yours. I'm mostly aware of how the fragments work and I like introducing the new 'variable' for the container image, I didn't really think of approaching it this way :-). I'll create a new PR for passing the variables (5da51cc) to make it more granular and easier to review. BTW, should I drop 56f198b altogether? It wouldn't likely be needed anymore with the new way to pass the container image URL, OTOH the two variables aren't (IIUIC) passed to the container test environment when executing the tests via |
Okay, let's continue with #1116, feel free to add a review if you want.
Thank you.
Yes, I'd drop it at this point. |
The test is based on previous work from Vladimir Slavik.