Add extended test support to multi-arch CI workflows#4649
Conversation
Port run_extended_tests from the non-multi-arch pipeline into the multi-arch CI scripts and workflows. Extended tests (functional + benchmarks) are enabled for schedule triggers and workflow_dispatch with test labels. Signed-off-by: Lenine Ajagappane <Lenine.Ajagappane@amd.com>
benchmark_runs_on and run_extended_tests are not declared in test_artifacts.yml until #3986 merges. Passing undefined inputs causes workflow validation failure. Comment them out with TODO(#3986) markers to uncomment once the inputs are available. Signed-off-by: Lenine Ajagappane <Lenine.Ajagappane@amd.com>
benchmark_runs_on and run_extended_tests. Uncomment the inputs that were disabled in 293abdf. Signed-off-by: Lenine Ajagappane <Lenine.Ajagappane@amd.com>
…hub.com/ROCm/TheRock into users/lajagapp/extended-test-multi-arch
|
Sample run with extended_test label included -- https://github.com/ROCm/TheRock/actions/runs/24903491216/job/72955238298 |
- Merge benchmark-runs-on presence and type checks into a single test with accurate docstring (gfx94x has the key, gfx110x does not). - Remove duplicate test_default_extended_tests_is_false (identical to test_pr_disables_extended_tests). - Collapse write_outputs enabled/disabled tests into one serialisation check; keep skipped-CI edge case separate. Signed-off-by: Lenine Ajagappane <Lenine.Ajagappane@amd.com>
Scheduled run with all extended_tests labels included (functional + benchmarks) -- https://github.com/ROCm/TheRock/actions/runs/24929653222/job/73019262986 |
geomin12
left a comment
geomin12
left a comment
There was a problem hiding this comment.
overall looks good, few comments! merge conflict too
| # TODO: Consolidate test_type, test labels, and run_functional_tests | ||
| # (from the single-arch pipeline) into a per-platform test config object | ||
| # (e.g. linux_test_config JSON) instead of separate top-level outputs. |
There was a problem hiding this comment.
add back comment, this work does not finish this
| matrix: | ||
| family_info: ${{ fromJSON(inputs.build_config).per_family_info }} | ||
| uses: ./.github/workflows/test_artifacts.yml | ||
| secrets: inherit |
| matrix: | ||
| family_info: ${{ fromJSON(inputs.build_config).per_family_info }} | ||
| uses: ./.github/workflows/test_artifacts.yml | ||
| secrets: inherit |
There was a problem hiding this comment.
let's remove this, last i remember, the API no longer worsk
FYI this is a workflow dispatch, scheduled run will work with #4384 |
2 participants
Summary
run_extended_testssupport from the non-multi-arch CI pipeline (configure_ci.py,setup.yml) into the multi-arch equivalents (configure_multi_arch_ci.py,setup_multi_arch.yml), as discussed in Consolidate extended test flags into a single run_extended_tests boolean #3986 (comment).scheduletriggers andworkflow_dispatchwith test labels; disabled for PRs and push events.benchmark-runs-onrunner labels are included in per-family build config and passed through totest_artifacts.yml.Dependency
The downstream consumers (
test_artifacts.ymlandfetch_test_configurations.py) that act onrun_extended_testsandbenchmark_runs_oninputs will be available once #3986 is merged. Until then, the values are passed but not yet consumed bytest_artifacts.yml.Test plan
python -m pytest configure_multi_arch_ci_test.py)