Add auto-retry for flaky PyPerf test to improve CI stability

[mlim19]

Summary

This PR adds automatic retry functionality for the test_dso_name_in_pyperf_profile test, which occasionally fails due to PyPerf timeouts in CI environments.

Problem

The test test_dso_name_in_pyperf_profile[2.7-glibc-python-True-pyperf-True] sporadically fails with:

WARNING  gprofiler.profilers.python_ebpf:python_ebpf.py:301 PyPerf dead/not responding, killing it
TimeoutError

This occurs when PyPerf cannot generate profile files within the 5-second timeout, particularly:

• In resource-constrained CI environments
• With Python 2.7 (slower than newer versions)
• When there's system load or contention

Example failure: https://github.com/intel/gprofiler/actions/runs/19146096101/job/54734480910

Solution

Implement auto-retry using the pytest-rerunfailures plugin:

1. Add dependency: pytest-rerunfailures==15.0 to dev-requirements.txt
2. Mark test as flaky: Add @pytest.mark.flaky(reruns=3, reruns_delay=2) decorator

The test will now:

• Automatically retry up to 3 times if it fails
• Wait 2 seconds between retry attempts (gives system time to stabilize)
• Only fail if ALL attempts fail (catches genuine issues)
• Apply to all 4 parameter combinations of this test

Impact

• Scope: Only affects test_dso_name_in_pyperf_profile (4 parameter combinations)
• Other tests: Run normally without retries
• CI time: Minimal increase (only when test fails)
• Reliability: Significantly improved CI stability for this flaky test

Changes

+ pytest-rerunfailures==15.0  # Added to dev-requirements.txt

+ @pytest.mark.flaky(reruns=3, reruns_delay=2)  # Added to test function
  @pytest.mark.parametrize("in_container", [True])
  @pytest.mark.parametrize("profiler_type", ["pyperf"])
  def test_dso_name_in_pyperf_profile(...):

Test Plan

• Test will retry automatically on failure
• Other tests remain unaffected
• CI should show "RERUN" status when retry occurs
• Verify in GitHub Actions that the test passes after retry

🤖 Generated with Claude Code