Fault Localization via Efficient Probabilistic Modeling of Program Semantics¶
1 Motivation & Innovation¶
1.1 Motivation¶
SBFL: Spectrum-based Fault Localization¶
No Semantic
\[Tarantula = \frac{\frac{e_f}{e_f + n_f}}{\frac{e_p}{e_p + n_p} + \frac{e_f}{e_f + n_f}}\]
Conditions to cause failure¶
MBFL & Angelic debugging¶
Low Effectiveness
MBFL: Mutation-based fault localization
Angelic Debugging
- Symbolic analysis
- The result of an expression can be modified to reverse the results of failing test while maintaining the results of the passing test
high suspicious score
1.2 Innovation - SmartFL¶
Challenges:
- How to model the effect from the control statements?
- Static analysis + dynamic analysis
- Scalability
- Reducing the size of traces
- Selecting test cases