Ansible request for EL10 distributions

[sxa]

Please put the name of the software product (and affected platforms if relevant) in the title of this issue

Delete as appropriate from this list:

• Bug in ansible playbook

Details:

• Current playbooks do not execute out of the box on CentOS Stream 10 and will therefore likely fail on RHEL10 and other derivative distributions. It may be due to issues caused with a newer version of ansible as CS9 does not have the same problem:

0.930 + ansible-playbook -i hosts ansible/playbooks/AdoptOpenJDK_Unix_Playbook/main.yml -e git_sha=123456 --skip-tags=debug,hosts_file,hostname,adoptopenjdk,jenkins,nagios,superuser,docker,swap_file,crontab,nvidia_cuda_toolkit                                      
1.462 ERROR! couldn't resolve module/action 'homebrew'. This often indicates a misspelling, missing collection, or incorrect module path.
1.462 
1.462 The error appears to be in '/ansible/ansible/playbooks/AdoptOpenJDK_Unix_Playbook/roles/curl/tasks/main.yml': line 121, column 3, but may
1.462 be elsewhere in the file depending on the exact syntax problem.
1.462 
1.462 The offending line appears to be:
1.462 
1.462 
1.462 - name: Install latest version of curl for MacOS
1.462   ^ here
------

[sxa]

Seems to be even more than this if you skip that particular part. Seen similar on RHEL9 (real installation) and there were at least three similarly problematic roles there out of the box.

[sxa]

Noting that this also requires the changes described in #3851
At present, running test jobs requires you to specify &&sw.os.rhel.10 onto any hostname you specify in a test job's LABEL otherwise this code from this PR will try to start Xvfb.

[sxa]

Most of the required modules sych as homebrew, authorized_key zypper mount sysctl and alternatives seem to be provided by adding the ansible-collection-community-general package

[sxa]

CentOS Stream 10 test results

I'm going to summarise CS10 here, but I'll also use the opportunity to summarise the CS9 runs in a follow-on comment.
AQA_Test_Pipeline: https://ci.adoptium.net/job/AQA_Test_Pipeline/419/
Re-grind links are on alternate machines for comparison.

Unstable jobs:

• https://ci.adoptium.net/job/Test_openjdk24_hs_sanity.openjdk_aarch64_linux/50/testReport (8 failures - some TZ related - test material mismatch likely: run-time=2025a, tests=2025b) Re-grind
• https://ci.adoptium.net/job/Test_openjdk24_hs_extended.openjdk_aarch64_linux/50/testReport (202 failures) Most, but not all, display related. Weston issue Re-grind
• https://ci.adoptium.net/job/Test_openjdk24_hs_dev.functional_aarch64_linux/32/testReport (6 failures) Re-grind passed SSL test failures
• https://ci.adoptium.net/job/Test_openjdk24_hs_special.openjdk_aarch64_linux/32/testReport (Unclear what failed on first look) Re-grind Same issue

[sxa]

CentOS Stream 9 test results:

Unstable jobs (Note did not run the full set from CS10 above):

• https://ci.adoptium.net/job/Test_openjdk24_hs_extended.openjdk_aarch64_linux/47/testReport (4 failures) Re-grind in U18.04 4 failures
• https://ci.adoptium.net/job/Test_openjdk24_hs_sanity.openjdk_aarch64_linux/48/testReport (7 failures - some TZ related as with CS10) Re-grind on another machine TIMED OUT Try again

[sxa]

centosstream10 X11 errors are due to:

00:46:43  [19:46:42.898] fatal: environment variable XDG_RUNTIME_DIR is not set.

I have created a /home/jenkins/xdg, mode 0700 (required) and then set XDG_RUNTIME_DIR on the jenkins agent definition to that directory, and am trying again: https://ci.adoptium.net/job/Grinder/13088/console
weston segfaulted - I also needed /tmp/.X11-unix to be created (as the jenkins user is ok) otherwise weston segfaults.

Running at https://ci.adoptium.net/job/Grinder/13089/

EDIT: The above test had 7 failures, so the weston/XDG fixes were successful.