Smarten Up Your Pipeline and Reduce Test Execution Time｜
summary
Claims a 30% cut in test runtime by building a coverage-driven filter: instrumented JVMs with jacoco, generated per-test coverage, stored mappings in a service called Timmy, and run only the minimal subset of tests for changed Java classes. Validation uses shadow branches and shows 100% detection of broken tests, supports ~330 feature branches a week (~70 PRs/day), drops typical branch test time from 45 to 30 minutes, and meets a 500 ms SLO for subset computation.
AI-written summary · about
chapters (10)
- 0:00Start
- 1:45Execution Times
- 2:54When Do We Run Our Arrow Tests
- 6:14Long Term Solutions
- 11:41The Testing Policies
- 13:11Generate Code Coverage Report
- 13:52Jacoco as Jvm Agent
- 24:59Return on Investment
- 25:28Pie Chart
- 30:26Takeaways
related talks
28mWhen Can You Skip a Test? Tracking Test Impact
46mSpeed Up PR Tests with Smart Code Coverage-Based Selection
51mAre Your Test Slowing You Down?
55mAre your tests slowing you down?
46mTesting the untestable - patterns and use cases analysis
44mSimplifying and Scaling Software Delivery
35mUsing the DevOps Pipeline to Generate Continuous, Valuable Insights
38mMutants to the rescue: How effective are your unit tests?