Class Specification Implementation Graph By: Njume Njinimbam Chi-Chang Sun

White Box testing A software testing technique whereby explicit knowledge of the internal workings of the item being tested are used to select the test data. White box testing uses specific knowledge of programming code to examine outputs

White Box testing Control flow testing Data flow testing Branch testing Path testing Statement coverage Decision coverage

Black Box testing Black-box testing is a method of software testing that tests the functionality of an application as opposed to its internal structures or workings (see white-box testing). This method of test can be applied to all levels of software testing: unit, integration, system and acceptance. It typically comprises most if not all testing at higher levels, but can also dominate unit testing as well.

Black Box testing Decision table testing All-pairs testing State transition tables Equivalence partitioning Boundary value analysis Control flow testing

Adjective Sufficient testing requires both black- and white- box techniques. Although both tasks are similar in that they have the same objective, namely detecting faults within a program, often black- and white-box techniques are applied separately using different tools. The reason for this is the lack of techniques and tools integrating both tasks.

Object-oriented prototypes VS CSIG Both approach relies on a graphical representation combining both specification and implementation of the class under test. Especially in control flow graph of the implementation by special data flow edges. However, the major different is that CSIG can construction process can be automated.

Benefits 1. The tester has only to be familiar with the concepts underlying one technique and needs training only for one tool. 2. Less maintenance effort is required, since only one tool has to be maintained. 3. Testing can be carried out more efficiently, since an integrated technique can generate test cases covering both the specification and the source code at the same time.

CSIG demonstration Edges Intra- metho control and data flow edges Inter-method control and data flow edges CCFG frame edges

Step 1: Generating method implementation graphs The first step consists of generating a control flow graph for each method based on its implementation. A control flow graph consists of nodes, representing statements, and edges, representing control flow among the statements. Moreover, the generated control flow graph can be augmented with data flow information for data flow testing.

Step 2: Generating a prototype for each specified method

Prototype of a transition if (source) if (guard) action; else throw new ErrorStateException();

Combine

CSIG construction cont'd step 3- Generating method specification graphs Step 4- Generating a CCFG frame Step 5- Inserting the method graphs into a CCFG frame Step 6- Adding data flow edges for black-box testing ▫The identification of def-use pairs for black-box testing can be carried out in 9 steps

Tool support - Sample test process cont’d

Tool support - Using CSIGs for regression testing Objective - provide confidence that modifications have the intended effect and do not affect other unchanged parts of the program. selective regression testing - select those test cases from a given test suite covering changed parts of the program

Tool support - Using CSIGs for regression testing cont’d selective regression testing technique proposed by Rothermel et al. [21] ▫main idea - compare two versions of a class and analyze the changes between these versions In the original approach, class control flow graphs(CCFG) are used to compare the two version. ▫CCFG does not consider specification changes, we have adjusted their selection algorithm to be applicable to CSIGs

Tool support - Sample test process 1. Enter the specification of the class to be tested 2. Generation of executable test oracles 3. CSIG of the class is generated 4. User enters test cases needed for testing first version of the class. 5.Entered test cases are executed and results are compared to those obtained by the test oracle. The tool also determines adequacy of the entered test cases

Tool support - Sample test process cont’d 6. Two modifications are made to the class. first modification concerns its implementation other modification concerns the specification 7. An executable test oracle and the CSIG are generated 8. The two CSIGs are compared using the modified selection algorithm 9. The test cases selected are executed.

Conclusions Contrary to existing representation of classes, a CSIG is not restricted to the specification or implementation of a class.