This has been happening with the CI tests for the clang package e.g. https://jenkins-continuous-infra.apps.ci.centos.org/view/Fedora%20All%20Packages%20Pipeline/job/fedora-rawhide-pr-pipeline/2855/
The test run completes with SUCCESS even though the clang-rhbz_1657544 test fails.
I do see the FAIL in both the console output as well as the artifacts: https://jenkins-continuous-infra.apps.ci.centos.org/view/Fedora%20All%20Packages%20Pipeline/job/fedora-rawhide-pr-pipeline/2855/artifact/package-tests/logs/FAIL-str_clang-rhbz_1657544.log
06:18:21 TASK [standard-test-basic : Check the results] ********************************* 06:18:21 changed: [/workDir/workspace/fedora-rawhide-pr-pipeline/images/test_subject.qcow2] => changed=true 06:18:21 cmd: |- 06:18:21 log="/tmp/artifacts//test.log" 06:18:21 if [ ! -f "$log" ]; then 06:18:21 echo ERROR 06:18:21 echo "Test results not found." 1>&2 06:18:21 elif grep ^ERROR "$log" 1>&2; then 06:18:21 echo ERROR 06:18:21 elif grep ^FAIL "$log" 1>&2; then 06:18:21 echo FAIL 06:18:21 elif grep -q ^PASS "$log"; then 06:18:21 echo PASS 06:18:21 else 06:18:21 echo ERROR 06:18:21 echo "No test results found." 1>&2 06:18:21 fi 06:18:21 delta: '0:00:00.011938' 06:18:21 end: '2020-03-15 11:18:21.560883' 06:18:21 rc: 0 06:18:21 start: '2020-03-15 11:18:21.548945' 06:18:21 stderr: FAIL clang/rhbz_1657544 06:18:21 stderr_lines: <omitted> 06:18:21 stdout: FAIL 06:18:21 stdout_lines: <omitted> 06:18:21 06:18:21 TASK [standard-test-basic : Set role result] *********************************** 06:18:21 ok: [/workDir/workspace/fedora-rawhide-pr-pipeline/images/test_subject.qcow2] => changed=false 06:18:21 ansible_facts: 06:18:21 role_message: FAIL clang/rhbz_1657544 06:18:21 role_result: FAIL 06:18:21 06:18:21 TASK [include_role : str-common-final] ***************************************** 06:18:21 06:18:21 TASK [str-common-final : Pull out the logs from test environment to test runner] *** 06:18:22 changed: [/workDir/workspace/fedora-rawhide-pr-pipeline/images/test_subject.qcow2] => changed=true 06:18:22 cmd: sshpass -d3 /usr/bin/rsync --delay-updates -F --compress --archive --rsh=/usr/bin/ssh -S none -i /tmp/inventory-cloud2no8g0b5/identity -o Port=4115 -o StrictHostKeyChecking=no -o UserKnownHostsFile=/dev/null -o UserKnownHostsFile=/dev/null --out-format=<<CHANGED>>%i %n%L root@127.0.0.3:/tmp/artifacts// /workDir/workspace/fedora-rawhide-pr-pipeline/logs/ 06:18:22 msg: |- 06:18:22 .d..t...... ./ 06:18:22 >f+++++++++ FAIL-str_clang-rhbz_1657544.log 06:18:22 >f+++++++++ PASS-str_clang-llvm-toolchain.log 06:18:22 >f+++++++++ PASS-str_clang-rhbz_1647130.log 06:18:22 >f+++++++++ PASS-str_llvm-test-suite-abi-test-suite.log 06:18:22 >f+++++++++ PASS-str_llvm-test-suite-test-suite.log 06:18:22 >f+++++++++ PASS-str_rhbz#482491.log 06:18:22 >f+++++++++ results.yml 06:18:22 >f+++++++++ test.log 06:18:22 rc: 0 06:18:22 stdout_lines: 06:18:22 - .d..t...... ./ 06:18:22 - '>f+++++++++ FAIL-str_clang-rhbz_1657544.log' 06:18:22 - '>f+++++++++ PASS-str_clang-llvm-toolchain.log' 06:18:22 - '>f+++++++++ PASS-str_clang-rhbz_1647130.log' 06:18:22 - '>f+++++++++ PASS-str_llvm-test-suite-abi-test-suite.log' 06:18:22 - '>f+++++++++ PASS-str_llvm-test-suite-test-suite.log' 06:18:22 - '>f+++++++++ PASS-str_rhbz#482491.log' 06:18:22 - '>f+++++++++ results.yml' 06:18:22 - '>f+++++++++ test.log' 06:18:22 06:18:22 TASK [str-common-final : Report role result] *********************************** 06:18:22 ok: [/workDir/workspace/fedora-rawhide-pr-pipeline/images/test_subject.qcow2] => 06:18:22 msg: 06:18:22 - 'Result: FAIL' 06:18:22 - FAIL clang/rhbz_1657544 06:18:22 - '' 06:18:22
I know that we did some work to change pipelines to not fail in a red state when a test failed, but it should be yellow I think? And the linked job is green, which seems odd. I need to do some looking, but I just wanted to confirm what you're seeing and share what I found with my initial inspection.
Here is a better example of what I'm seeing:
If you look at this pull request, the CI results reports as Pass: https://src.fedoraproject.org/rpms/clang/pull-request/55
But when I click through to the jenkins job and look at the artifacts, it lists 2 tests as failing:
https://jenkins-continuous-infra.apps.ci.centos.org/blue/organizations/jenkins/fedora-f32-pr-pipeline/detail/fedora-f32-pr-pipeline/35/artifacts
Yeah, I need to see what logic sets the report on the PR as pass/fail vs the jenkins job. The only change we made recently was to add 1 package into the image to support fonts. I'm not familiar with the logic, so I'll need to figure out what's making the decision and why it's not seeing the failures (unless another fedora ci member jumps in before I figure it out). =)
Metadata Update from @jimbair: - Issue assigned to jimbair
Monday follow-up; we are still looking into it, but haven't pin-pointed the issue yet. I'm taking the ticket to ensure we keep up on it, and as we have more updates, I'll post them here!
I know one idea was there are two testing yamls in the repo, though with different names, possibly with logic having one overwrite the others results? That shouldn't happen, but one thought that came up in this morning's IRC discussions.
Proposed fix: https://github.com/CentOS-PaaS-SIG/upstream-fedora-pipeline/pull/216
@jimbair can this issue be closed?
I think we are good. My test pull-request failed (as expected) with the fix in place. So we should be good here.
Metadata Update from @msrb: - Issue status updated to: Closed (was: Open)
Login to comment on this ticket.