Enable staging repos when DNF_CONTENTDIR is specified #169
Labels
No Label
priority: critical
priority: high
priority: low
priority: medium
status: abandoned
status: duplicate
status: invalid
status: peer review needed
status: wontfix
tag: content
tag: good first issue
tag: help wanted
test suite
type: bug
type: documentation
type: enhancement
type: maintenance
type: question
No Milestone
No project
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies
No dependencies set.
Reference: testing/os-autoinst-distri-rocky-migrated#169
Loading…
Reference in New Issue
Block a user
No description provided.
Delete Branch "enable_contentdir"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
This PR adds no new functionality but re-enables some logic to enable staging repositories when DNF_CONTENTDIR is supplied.
I have tested this using:
Result: Both tests fail in _staging_repos_enable.pm line 11
Test incomplete:
Reason: backend died: No map for '' at /usr/lib/os-autoinst/consoles/VNC.pm line 660.
This is not looking good. I'm getting exact same incomplete tests with 8.7 stg.
Tested using:
Investigation is continuing...
@AlanMarshall Don't spend more time debugging this error issue other than to simply solve a puzzle. I forgot how this contentdir change is implemented and this PR isn't required after all. @akatch please leave this PR open until we discuss as it serves as a good example of how to test this in the production instance.
For reference (unless they get deleted) here are two tests of interest...
...and the code that is currently in place to use supplied
contentdir
without the need to runtests/_staging_repos_enable.pm
which bombs terribly.Likely a better PR will remove the conditional from
main.pm
and also removetests/_staging_repos_enable.pm
andtests/_staging_repos_disable.pm
after verifying they are dead code paths.Yes, thanks @tcooper from tests I am well on the way to a very similar conclusion and was about to report as such.
@tcooper I confirm that your results are exactly what my tests showed. The failure in
script_run sed -i...
was due to having no map for key#0001
which is probably ascii SOH. The failure occurs only after reboot when sed (or rsync) try to replace/etc/yum.repos.d/Rocky-*
and does not occur when the same commands are run at the end of the install phase. However,/etc/yum.repos.d/Rocky-*
can be read after reboot without any failure occurring.@akatch I found that having a modified
tests/_staging_repos_enable.pm
was very useful in working out what was going on here and I therefore suggest retaining and extending PR #169 to provide an ongoing debug facility that only operates when the stg repos are being used viaDNF_CONTENTDIR=stg
. After all we have this existing code that can be modified and used to advantage for development rather than just deleting it as redundant.Does this seem reasonable to everyone?
As a suggestion, this is what I'm thinking of for now:
The results appear as a row in the openqa run details page.
The
dnf repoinfo
result has identified the postinstall enabled repos as including:http://dl.rockylinux.org/stg/rocky/8/BaseOS/x86_64/os/
and similar for
AppStream
andextras
whereas the enabled repos should be:http://dl.rockylinux.org/stg/rocky/8.8-Beta/BaseOS/x86_64/os/
I tried various combinations of VERSION=8.8-Beta and CURREL=8 with same result. Maybe there is another parameter that I haven't discovered yet... If not then this is a bug that needs fixing.
One of the commands that I tried was:
Result always:
http://dl.rockylinux.org/stg/rocky/8/BaseOS/x86_64/os/
I would also rename
_staging_repos_enable.pm
to something more appropriate.@AlanMarshall I like this in principle. My suggestions...
assert_script_run("command | grep <expected_content}")
the test can retain it's pass / fail nature.stg/rocky
and not mirrorlist type we should also double-check that to help us catch early any change in the mirrorlist format thatsed
pattern doesn't work for (unexpected but possible)._staging_repos_verify.pm
? Note a change inmain.pm
is also required if the test name is changed.Good work 👍🏻
@tcooper Thanks Trevor. I've just found a fix for the bug in the postinstall repos that I referred to earlier. I'll give you the details soon. It's a serious bug so needs addressed sooner rather than later. I'll PR when I get it all together.
The bug is that at the moment, postinstall, the repos are configured to stg/rocky/8 whereas they should be to stg/rocky/8.8-BETA as Louis defined a few days ago. I now have a fix for it.
Pull request closed