ci(test): exit 77 when tools missing to distinguish skip from pass

[bitflicker64]

Purpose of the PR

• relates to [Bug] Docker containers never restart when Java process crashes — broken cron monitor + tail -f /dev/null supervision #3043
• follow-up to fix(server): fix foreground mode exit code propagation in startup scripts #3044, fix(pd,store): fg mode exit code propagation in startup scripts #3047

The three startup test scripts (test-start-hugegraph.sh, test-start-hugegraph-pd.sh, test-start-hugegraph-store.sh) exit 0 when required tools (lsof, curl, java) are not found. Exit code 0 is indistinguishable from a passing test run — CI shows green even though no tests ran.

Main Changes

Test scripts:

• Change exit 0 → exit 77 in all tool-missing skip paths
• Change exit 0 → exit 77 in the Store ulimit -n skip path

CI workflows (server-ci.yml, pd-store-ci.yml):

• Wrap each test step with set +e / exit code capture
• Treat exit 77 as a visible ::notice:: annotation (not a failure)
• All other non-zero exits still fail the step

Verifying these changes

• Need tests and can be verified as follows:
  • Remove lsof from the runner path → CI step shows ::notice::... skipped instead of silently passing
  • Normal run (all tools present) → exit 0, step passes as before
  • Test failure → exit 1, step fails as before

Does this PR potentially affect the following parts?

• Dependencies
• Modify configurations
• The public API
• Other affects
• Nope

Documentation Status

• Doc - No Need

[bitflicker64]

@dosu review?

[dosubot]

I don't perform code reviews (assessing PRs for bugs, design, style, performance, or security). However, I can help in other ways — for example, I can research how the existing CI workflows or test scripts are structured, answer questions about the conventions used in the repository, or help with documentation updates related to this change.

Would any of those be useful?

To reply, just mention @dosu.

---

Share context across your team and agents. Try Dosu.

[imbajin]

Blocking: yes. Summary: The startup test wrapper still reports skipped suites as successful GitHub Actions steps. Evidence: static review of the workflow wrapper and local dummy-dist runs where the scripts returned 77.

Latest-head CI has visible failures; please also check dependency-check / build-commons (11): 🔗 https://github.com/apache/hugegraph/actions/runs/27120670959/job/80036825383

[imbajin]

‼️ Report missing-prerequisite startup tests as skipped, not successful

Evidence: this wrapper turns the test script 77 into exit 0 after emitting a notice. I reproduced the script side with a dummy dist and restricted PATH: all three startup scripts return 77 when lsof is unavailable, but the workflow wrapper converts that case to a successful step. In GitHub Actions, ::notice:: only creates an annotation; the final exit 0 still marks the startup-test step as success.

Impact: the PR makes the logs clearer, but the check remains green exactly like a real passing startup suite. That keeps the original ambiguity for maintainers who inspect only the check result.

Requested fix: split this into a preflight step that writes can_run=false and a specific skip_reason to $GITHUB_OUTPUT, then guard the real startup-test step with if: steps.<preflight>.outputs.can_run == 'true'. Add a separate notice/summary step for the skip path. While doing that, keep the reason specific so the store ulimit -n < 1024 case is not reported as missing tools.

[bitflicker64]

Switched to the preflight/guard pattern as requested. The preflight step writes can_run and skip_reason to $GITHUB_OUTPUT; the test step is guarded by if: steps..outputs.can_run == 'true' so it shows as grey Skipped (not green Success) when prerequisites are missing. The Store preflight distinguishes missing tool: from ulimit -n is (store requires >= 1024) with separate skip_reason values. Pushing now.