Analyzing Regression Test Selection Techniques

1. Analyzing Regression Test Selection Techniques -presented by Xuan Lin

2. Outline • Introduction • Concepts and Assumptions • Analysis Framework • Examples Techniques • Conculsion and Discussion

3. Outline • Introduction • Concepts and Assumptions • Analysis Framework • Examples Techniques • Conculsion and Discussion

4. Introduction • What is Regression Testing? -Everybody knows … • Retest-all strategy VS. Test Selection • Notions: P,P’,S,S’,T,T’,

5. Typical Selective Retest Process 1. Select T'≤T, a set of tests to execute on P' Regression Test Selection Problem 2.Test P’ with T’, to establish the correctness of P’ with respect to T’ Test Suite Execution Problem 3.If necessary, create T’’, a set of new functional or structural tests for P’ Coverage Identification Problem 4.Test P’ with T’’, to establish the correctness of P’ with respect to T’’ Test Suite Execution Problem 5.Create T’’’, a new test suite and test history for P’, from T, T’ and T’’ Test Suite Maintenance

6. Typical Selective Retest Process 1. Select T'≤T, a set of tests to execute on P' Regression Test Selection Problem 2.Test P’ with T’, to establish the correctness of P’ with respect to T’ Test Suite Execution Problem 3.If necessary, create T’’, a set of new functional or structural tests for P’ Coverage Identification Problem 4.Test P’ with T’’, to establish the correctness of P’ with respect to T’’ Test Suite Execution Problem 5.Create T’’’, a new test suite and test history for P’, from T, T’ and T’’ Test Suite Maintenance

7. Test Selection Techniques • Specification-based VS. Code-based • Three distinct goals of code-based test selection techniques -Coverage Techniques Compare and Evaluation !!! -Minimization Techniques -Safe Techniques

8. Outline • Introduction • Concepts and Assumptions • Analysis Framework • Examples Techniques • Conculsion and Discussion

9. Concepts and Assumptions • Fault-realing for P’: cause P’ to fail -No Effective procedure by which to find tests in T that are fault-realing for P’ [1] -Under certain conditions, a technique can select a Superset of the set of fault-revealing for P • Modification-revealing: casue the outputs of P and P’ to differ.

10. Concepts and Assumptions • Modification-revealing = Fault-revealing ??? • P-Correct-for-T Assumption: For each test t in T, when P was tested with t, P halted and produced the correct output • Obsolet-Test-Identification Assumption: There is an effective procedure for determining, for each test in t, whether t is obsolete for P’. Test t is obsolete for P’ if and only if t either specifies an input to P’ that, according to S’, is invalid for P’, or t specifies an invalid input-output relation for P’

11. Concepts and Assumptions • Up to now, we can find the fault-revealing test cases by: 1. Run our procedure for identifying obsolete test in T. 2. Remove them. 3. Find the modification-revealing test cases. - In the set of non-obsolete test cases, modification-revealing=fault-revealing

12. Concepts and Assumptions Obsolete Fault-Revealing Nonbsolete Fault-Revealing Modification-Revealing ???

13. Concepts and Assumptions • Modification-traversing: a test t is modification-traversing for P and P’ if and only if it (a) executes new or modified code in P’, or (b) formerly executed code that has since been deleted

14. Concepts and Assumptions Obsolete Fault-Revealing Nonbsolete Fault-Revealing Modification-Revealing ??? Modification-Traversing

15. Concepts and Assumptions • Controlled Regression Testing Assumption: when P’ is tested with t, we hold all factors that might infuence the output of P’, except for the code in P’, constant with respect to their states when we tested P with t.

16. Why We Need Define These Concepts and Assumptions? • Evaluate test selection techniques in terms of their ablities to select and avoid discarding fault-revealing tests. • Three classes can be used to distinguish techniques even CRTA is not satisfied. • Coverage techniques may omit tests from T’ that may reveal faults in P’

17. Outline • Introduction • Concepts and Assumptions • Analysis Framework • Examples Techniques • Conculsion and Discussion

18. Analysis Framework • Incusiveness • Precision • Efficiency • Generality

19. Analysis Framework-Inclusiveness

20. Analysis Framework-Inclusiveness • There is no algorithm to determine the inclusiveness! • However… • We can prove M is safe. • We can prove M is not safe. • We can compare techinques in terms of inclusiveness • We can experiment to approximate

21. Analysis Framework-Precision

22. Analysis Framework-Precision • There is no algorithm to determine the precision! • However… • We can compare techinques in terms of precision. • We can prove M is not precise • We can show M is precise. • We can experiment to compare.

23. Analysis Framework-Efficiency • Time & Space • Cost of selecting T’ < the cost of running T-T’ • Three Factors 1.preliminary phase vs. critical phase 2.automatability 3. calculation informatin on program modifications 4.ability to handle multiple modifications

24. Analysis Framework-Generality • Should function for some identifiable and practical class of program • Should handle realistic program modifications • Should be independent of assumptions about testing or maintenance enviroments. • Should be independent of particular program analysis tools • Should support intraprocedural or interprocedural test selection

25. Analysis Framework-Tradeoffs • Precision vs. Efficiency - both safe and unsafe • Inclusiveness vs.Efficiency -not safe • Generality vs. Inclusiveness, Efficiency or Precision • Multiple modication vs. Efficiency

26. Outline • Introduction • Concepts and Assumptions • Analysis Framework • Examples Techniques • Conculsion and Discussion

27. Refresh… Obsolete Fault-Revealing Nonbsolete Fault-Revealing Modification-Revealing Modification-Traversing

28. Depiction of inclusiveness and precision

29. Retest-all

30. Optimum

31. Examples: Dataflow • Caculate d-u pairs for both P and P’ • Identify and select d-u pairs that are new in, or modified for P’ • Some techniques also select deleted d-u pairs • Incremental / Nonincremental

32. Examples: Dataflow-Inclusion • Not safe

33. Examples: Dataflow-Precision • Not precise

34. Examples: Dataflow

35. Examples: Dataflow-Effiency • Incremental- O(|T|*|P’|*|P’|) • Nonincremental-

36. Examples: Dataflow-Generity • Applied to procedural programs generally. • Function for all program changes except those that do not alter d-u association • Some techiques applied to intraprocedural programs while others applied to interprocedural programs • Incremental approach requires incremental dataflow analysis tools.

37. Examples: Graph Walk Techniques • Build CFG for P and P’ • Collects traces for tests with CFG edges. • Performs synchronous depth-first traversals of the two graphs, selects those are not lexically identical.

38. Examples: Graph Walk Techniques-Inclusiveness • [1] shows that for controlled regression testing, the techniques will select all modification-traversing test. • So, it is safe.

39. Examples: Graph Walk Techniques-Precision • Not precise • Multiply-visisted-node

40. Examples: Graph Walk Techniques

41. In practice

42. Improved version

43. Examples: Graph Walk Techniques-Efficiency • Generally: • Property not hold[1]:

44. Examples: Graph Walk Techniques-Generality • Apply to procedural languages generally • All type of modifications • Both interprocedure and intraprocedure • No assumption on test suite or coverage • Require tools for constructing dataflow and tools for dataflow analysis

45. Examples: Path Analysis • Takes set of program paths in P’ expressed as an algebraic expression • Manipulates the expression to get a set of cycle-free exemplar paths. • Compare such paths in P with P’ • Tests that traverse modified exemplar paths will be selected

46. Examples: Path Analysis-Inclusiveness • Selects only modified paths and omits the cancel and new paths. • Not safe.

47. Examples: Path Analysis-Precision • It will select all the test cases that are modification-traversing and execute modified exemplar paths.

48. Examples: Path Analysis

49. Examples: Path Analysis-Efficiency • Exponential in |P| and |P’|

50. Examples: Path Analysis-Generality • Assumption: low-level program designs are depicted by language-independent algebraic representations. • Does not handle test cases for additions or deletings of code. • Does not require any coverage criterian or test generation technique. • Require tool for collecting traces at the statement level.