Today s Lecture. A Sample Ada Program to Test. Flow Graphs. Lecture 25: Data Flow and Dependence Graphs. White-Box Testing.

Transcription

1 oday s Lecture Lecture 25: Data low and Dependence Graphs Kenneth M. Anderson oundations of Software Engineering CSCI Spring Semester, White-Box esting Data low Graphs Minimum Retesting Program Dependence Graphs Control Dependence Graphs Data Dependence Graphs Lecture 25 Lecture 25 2 Control low low Graphs Graph representation of control flow and data flow relationships he partial order of statement execution, as defined by the semantics of the language Data low he flow of values from definitions of a variable to its uses Lecture 25 3 A Sample Ada Program to est function P return INEGER is begin, : INEGER; READ(); READ(); while ( > ) loop := Ð ; exit when = ; end loop; if ( < 20 and then mod 2 = 0) then := + 20; else := Ð 20; end if; return 2 + ; end P; Lecture 25 4

2 P s Control low Graph (CG) P s CG with a Data low Edge Lecture 25 5 Lecture 25 P s Control/Data low Graph White-box esting Criteria Use Coverage Select a test set such that, by executing P for each d in, all paths leading from each definition of a variable to each use of that variable in P s control/data flow graph are traversed at least once Lecture 25 Lecture 25 8

3 P s Control/Data low Graph Minimizing Retesting est Only What Is Affected by a Change Key: Dependency Analysis Also used for optimization, parallelization, At Coarse Level, Module Relationships Uses, calls, imports, includes, At ine Level, Control and Data low Program dependence graphs How many test cases are needed? Lecture 25 Lecture 25 Program Dependence Graph (PDG) Summary Representation of Dependence Nodes Are Either Statements or Predicates or the Special Node Entry wo Kinds of Edges Control dependence edge Data dependence edge wo Subgraphs Induced by the Edges Control Dependence Graph (CDG) Informal Definition or nodes and in a CG, is control dependent on if, during execution, can directly affect whether is executed Lecture 25 Lecture 25

4 Control Dependence Graph (CDG) P s Control Dependence Graph ormal Definition Let and be nodes in a CG. If appears on every path from to the exit node, where!=, then post-dominates. here is a control dependence from to with label L iff: there is a non-null path p from to, starting with edge L, such that post-dominates every node strictly between and on p; and does not post-dominate. 2,3,4 5 Lecture 25 3 Lecture 25 P s Control Dependence Graph P s Control Dependence Graph Lecture 25 5 Lecture 25

5 Data Dependence Graph (DDG) Informal Definition wo statements are data dependent if they might reference the same memory location and one of the references is an assignment to the memory location Data Dependence Graph (DDG) Informal Definition wo statements are data dependent if they might reference the same memory location and one of the references is an assignment to the memory location Intuition: If the statements cannot be switched without affecting the program, then they are data dependent Lecture 25 Lecture 25 8 Data Dependence Graph (DDG) P s Data Dependence Graph for ormal Definition Let and be nodes in a CG. here is a data dependence from to with respect to a variable v iff there is a non-null path p from to with no intervening definition of v and either: contains a definition of v and a use of v; contains a use of v and a definition of v; or contains a definition of v and a definition of v. Lecture 25 Lecture 25 20

6 P s PDG (DDG for Only) P s PDG (DDG for Only) Lecture 25 2 Lecture P s PDG (DDG for Only) Minimum Regression esting Given program P, its modified version P, and test set used to test P, find a way, making use of, to test P Identify changes to P resulting in P Select, a subset of, related to changes Run on P Lecture Lecture 25 24

7 Safety Goals Every relevant test from must be selected Precision Select only tests that exhibit different behavior Efficiency Cheap to calculate and run Lecture Modifications Adding Statements Deleting Statements Changing Statements heorem Need only tests in that can traverse different regions of statements in P and P, where regions are dependent-equivalent sub-cdgs Lecture 25 2 est Selection Algorithm procedure Selectests Construct CDGs of P and PÕ, with entry nodes E, E2 Õ = Compare (E, E2) procedure Compare (N,N2) mark N and N2 visited if (children of N and N2 differ) then return all tests that traverse N else Õ = NULL for each region or predicate child C of N not yet visited do find C2, the corresponding child of N2 Õ = Õ union Compare (C,C2) Lecture 25 2