Advanced Test Coverage Criteria: Specify and Measure, Cover and Unmask
summary
Labels specify many white‑box coverage criteria as concrete (location,predicate) objectives; an optimized dynamic symbolic execution variant plus iterative label deletion cuts instrumentation overhead from about 272× to ~7× in experiments; a combined static analysis (value analysis + WP) flags many infeasible objectives (~49.5% detection rate reported); an implemented toolset built on framesi and exposed as LTest supports measurement and automatic generation; an extended language HTOL encodes MCDC, dataflow and other advanced criteria and is partially implemented.
AI-written summary · about
chapters (17)
- 0:00Introduction
- 1:13White box testing overview
- 2:40Coverage criteria challenges
- 4:09Labels as specification
- 7:04Simulating coverage with labels
- 9:55Weak mutations and criteria
- 11:19Framesi and PathCrawler
- 12:25Dynamic symbolic execution
- 14:11Optimizing DSE for labels
- 16:54Iterative label deletion
- 20:11Detecting infeasible objectives
- 23:09Static analysis for labels
- 26:26Infeasible label example
- 29:16Handling duplicate objectives
- 30:19Limitations and hyper labels
- 33:14Summary
- 34:26Q&A
related talks
03mTest coverage
01h 00mRequirement Coverage: Myth or an Ultimate Goal
45mUsing AI, statistics, and set theory to know what to test
26mImproving Coverage Analysis
58mCoverage, quality, and Goodhart's Law
49mTest Coverage Visualization
19mTesting Your Tests With Code Coverage
01h 40mTesting, Preconditions, Coverage and Templates in Safety-Critical C++ Code