Professional Documents
Culture Documents
Software Testing
Objectives and principles Techniques Process Object-oriented testing Test workbenches and frameworks
Penn State University Last updated: Friday, February 17, 2012
Lecture Objectives
Understand
ENGINEERING DIVISION, PENN STATE UNIVERSITY
testing objectives and principles Testing techniques black-box and white-box Testing process unit and integration Object-oriented testing Test workbenches and frameworks
Software
There are 250 billion unique paths between A and B. If each set of possible data is used, and a single run takes 1 millisecond to execute, it would take 8 years to test all paths.
Such faults are called Bohrbugs (an allusion to Niels Bohrs simple and intelligible atomic model) Under seemingly exact conditions, the actions that a test case specifies can sometimes, but not always, lead to a failure Software engineers sometimes refer to faults with this property as Mandelbugs (an allusion to Benoit Mandelbrot, a leading researcher in fractal geometry)
Example: the software fault in the Patriot missile defense system responsible for the Scud incident in Dhahran
To project a targets trajectory, the weapons control computer required its velocity and the time as real values The system, however, kept time internally as an integer, counting tenths of seconds and storing them in a 24 bit register The necessary conversion into a real value caused imprecision in the calculated range where a detected target was expected next For a given velocity of the target, these inaccuracies were proportional to the length of time the system had been continuously running
Testing Objectives
ENGINEERING DIVISION, PENN STATE UNIVERSITY
Software testing can show the presence of bugs, but it can never show their absence. Therefore,
Testing
is the process of exercising a program with the specific intent of finding errors prior to delivery to the end user. A good test case is one that has a high probability of finding an error. A successful test is one that uncovers an error.
Testing Principles
All tests should be traceable to customer requirements Tests should be planned long before testing begins
ENGINEERING DIVISION, PENN STATE UNIVERSITY
The Pareto principle applies to software testing Testing should begin in the small and progress toward testing in the large Exhaustive testing is not possible To be most effective, testing should be conducted by an independent third party
6
Testing must be planned and performed systematicallynot ad hoc or random. Testing can be performed in two ways:
1.
2.
Knowing the specified function that a product has been designed to perform black-box testing. Knowing the internal workings of the product and testing to ensure all parts are exercised adequately white-box testing.
7
Black-box Testing
ENGINEERING DIVISION, PENN STATE UNIVERSITY
An approach to testing where the program is considered as a black-box The program test cases are based on the system specification Test planning can begin early in the software process
8
Equivalence Partitioning
ENGINEERING DIVISION, PENN STATE UNIVERSITY
Divide the input domain into classes of data from which test cases can be derived. Strives to define a test that uncovers classes of errors reducing total number of test cases required.
Example
ENGINEERING DIVISION, PENN STATE UNIVERSITY
Specifications for DBMS state that product must handle any number of records between 1 and 16,383 (2 14 1) If system can handle 34 records and 14,870 records, then probably will work fine for 8,252 records, say. If system works for any one test case in range (1..16,383), then it will probably work for any other test case in range Range (1..16,383) constitutes an equivalence class
Any one member is as good a test case as any other member of the class
10
Example
ENGINEERING DIVISION, PENN STATE UNIVERSITY
Class 1: Fewer than 1 record Equivalence Class 2: Between 1 and 16,383 records Equivalence Class 3: More than 16,383 records
11
Technique that leads to selection of test cases that exercise bounding values. Selecting test case on or just to one side of boundary of equivalence class increases probability of detecting fault.
"Bugs lurk in corners and congregate at boundaries"
12
DBMS Example
Test case 1: 0 records
ENGINEERING DIVISION, PENN STATE UNIVERSITY
Member of equivalence class 1 (& adjacent to boundary value) Test case 2: 1 record Boundary value Test case 3: 2 records Adjacent to boundary value Test case 4: 723 records Member of equivalence class 2 Test case 5: 16,382 records Adjacent to boundary value Test case 6: 16,383 records Boundary value Test case 7: 16,384 records Member of equivalence class 3 (& adjacent to boundary value)
13
Example In 1994, minimum Social Security (FICA) deduction from any one paycheck was $0.00, and the maximum was $3571.20
cases must include input data which should result in deductions of exactly $0.00 and exactly $3571.20 test data that might result in deductions of less than $0.00 or more than $3571.20
14
test
White-box Testing
ENGINEERING DIVISION, PENN STATE UNIVERSITY
Test case design method that uses the control structure of the procedural design to derive test cases. Can derive tests that:
Guarantee all
at least once Exercise all logical decisions on their true and false sides Execute all loops at their boundaries and within operational bounds Exercise internal data structures to ensure validity
15
Proposed by Tom McCabe. Use cyclomatic complexity measure as guide for defining a basis set of execution paths. Test cases derived to exercise the basis set are guaranteed to execute every statement at least once.
16
Independent Paths
CC = 5 So 5 independent paths
ENGINEERING DIVISION, PENN STATE UNIVERSITY
1. 2. 3. 4. 5.
a, c, f a, d, c, f a, b, e, f a, b, e, a, a, b, e, b, e,
a b e f
17
Graph Matrices
ENGINEERING DIVISION, PENN STATE UNIVERSITY
By forming a graph matrix the cyclomatic complexity and the basis path set can be found automatically.
a b c d e f a 1 1 1 b 1 c 1 1 d e 1 1 1 f
3-1 1-1 1-1 1-1 3-1 = = = = = 2 0 0 0 2
18
CC= 2+2+1=5
The Flowgraph
ENGINEERING DIVISION, PENN STATE UNIVERSITY
Before the cyclomatic complexity can be calculated, and the paths determined, the flowgraph must be created. Done by translating the source code into flowgraph notation:
sequence
if
while
until
case
19
Example
PROCEDURE average INTERFACE RETURNS average, total.input, total.valid; INTERFACE ACCEPTS value, minimum, maximum; TYPE value[1:100] IS SCALAR ARRAY; TYPE average, total.input, total.valid; minimum, maximum, sum IS SCALAR; TYPE i IS INTEGER;
i = 1; 2 total.input = total.valid = 0; sum = 0; DO WHILE value[i] <> -999 AND total.input < 100 increment total.input by 1;
3 6
5 7 8
ENDIF increment i by 1;
9 ENDDO
IF total.valid > 0
10
12
13 ENDIF
20
Example
Flowgraph for average
ENGINEERING DIVISION, PENN STATE UNIVERSITY
2 3 10
12 13
11
4 5
6 7
2. Independent paths
9
21
Condition Testing
ENGINEERING DIVISION, PENN STATE UNIVERSITY
Exercises logical conditions contained within a program module. Types of errors found include;
Boolean
operator error (OR, AND, NOT) Boolean variable error Boolean parenthesis error Relational operator error (>,<,=,!=,) Arithmetic expression error
22
Loop Testing
ENGINEERING DIVISION, PENN STATE UNIVERSITY
Focus exclusively on the validity of loop constructs. 4 types of loop can be defined:
Simple Nested Concatenated Unstructured
23
Loop Types
ENGINEERING DIVISION, PENN STATE UNIVERSITY
Unstructured
24
Simple Loops
ENGINEERING DIVISION, PENN STATE UNIVERSITY
Where n is the max number of passes, the following test can be applied:
loop entirely Only one pass 2 passes m passes (where m<n) n-1, n, n+1 passes
25
Skip
Nested Loops
ENGINEERING DIVISION, PENN STATE UNIVERSITY
If the approach for simple loops is extended, number of possible tests would grow geometrically impractical. Instead:
Start at innermost loop. Set all other loops to minimum values. Conduct simple loop test for innermost loop while holding outer loops at minimum loop counter values. Add other test for out-of-range or excluded values. Work outward, conducting tests for next loop, but keeping all other outer lops at minimum values and other nested loops to typical values. Continue until all loops tested.
26
Concatenated Loops
ENGINEERING DIVISION, PENN STATE UNIVERSITY
Test as simple loops provided each loop is independent. If two loops are concatenated and loop counter for loop 1 is used as initial value for loop 2, then test as nested loops.
27
Unstructured Loops
ENGINEERING DIVISION, PENN STATE UNIVERSITY
Cant test unstructured loops effectively. Reflects very bad practice and should be redesigned.
28
The Tester
Who does the testing?
ENGINEERING DIVISION, PENN STATE UNIVERSITY
a) b) c) d)
29
Strictly speaking, testing should be performed by an independent group (SQA or 3rd party) Members of the development team are inclined to be more interested in meeting the rapidlyapproaching due-date. The developer of the code is prone to test gently. Must remember that the objective is to find errors, not to complete test without finding them (because theyre always there!)
30
Successful Testing
The success of testing can be measured by applying a simple metric:
ENGINEERING DIVISION, PENN STATE UNIVERSITY
3
31
of individual program components Often performed by the component developer Tests often derived from the developers experience! Increased productivity possible with xUnit framework
Testing
Integration testing
of groups of components integrated to create a system or sub-system The responsibility of an independent testing team Tests are based on a system specification
32
Testing
Testing Phases
ENGINEERING DIVISION, PENN STATE UNIVERSITY
Unit testing
Software developer
Integration testing
Development team/ SQA/ Independent Test Group
33
Integration Testing
ENGINEERING DIVISION, PENN STATE UNIVERSITY
Tests complete systems or subsystems composed of integrated components Integration testing should be black-box testing with tests derived from the specification Main difficulty is localizing errors Incremental integration testing reduces this problem
34
T1
T1 T2 T2 B T3 T3 C T4
T5
35
with high-level system and integrate from the top-down replacing individual components by stubs where appropriate components in levels until the complete system is created
Bottom-up testing
Integrate individual
Top-down Testing
Level 1
ENGINEERING DIVISION, PENN STATE UNIVERSITY
Testing sequence
Level 1
...
Level 2
Level 2
Level 2
Level 3 stubs
37
Bottom-up Testing
Test drivers
ENGINEERING DIVISION, PENN STATE UNIVERSITY
Level N
Level N
Level N
Level N
Level N
Testing sequence
Test drivers
Level N1
Level N1
Level N1
38
Which is Best?
In bottom-up testing:
Test
ENGINEERING DIVISION, PENN STATE UNIVERSITY
time. Integration errors are found later rather than earlier. Systems-level design flaws that could require major reconstruction are found last. There is no visible, working system until the last stage so is harder to demonstrate progress to clients.
39
Interface Testing
ENGINEERING DIVISION, PENN STATE UNIVERSITY
Takes place when modules or sub-systems are integrated to create larger systems Objectives are to detect faults due to interface errors or invalid assumptions about interfaces Particularly important for object-oriented development as objects are defined by their interfaces
40
Interface Testing
Test cases
ENGINEERING DIVISION, PENN STATE UNIVERSITY
C
41
Interfaces Types
Parameter interfaces
Data
ENGINEERING DIVISION, PENN STATE UNIVERSITY
Procedural interfaces
Sub-system
encapsulates a set of procedures to be called by other sub-systems request services from other sub42
systems
Interface Errors
Interface misuse
ENGINEERING DIVISION, PENN STATE UNIVERSITY
A calling component calls another component and makes an error in its use of its interface e.g. parameters in the wrong order A calling component embeds assumptions about the behaviour of the called component which are incorrect The called and the calling component operate at different speeds and out-of-date information is accessed
Interface misunderstanding
Timing errors
43
Design tests so that parameters to a called procedure are at the extreme ends of their ranges Always test pointer parameters with null pointers Use stress testing in message passing systems In shared memory systems, vary the order in which components are activated Design tests which cause the component to fail
44
Stress Testing
Exercises the system beyond its maximum design load.
ENGINEERING DIVISION, PENN STATE UNIVERSITY
Stressing
light
should not fail catastrophically. Stress testing checks for unacceptable loss of service or data
Particularly relevant to distributed systems which can exhibit severe degradation as a network becomes overloaded
45
Object-Oriented Testing
ENGINEERING DIVISION, PENN STATE UNIVERSITY
The components to be tested are object classes that are instantiated as objects Larger grain than individual functions so approaches to white-box testing have to be extended No obvious top to the system for top-down integration and testing
46
Testing Levels
ENGINEERING DIVISION, PENN STATE UNIVERSITY
Test object classes Test clusters of cooperating objects Test the complete OO system
47
all operations associated with an object Setting and interrogating all object attributes Exercising the object in all possible states
Inheritance makes it more difficult to design object class tests as the information to be tested is not localized
48
Object Integration
ENGINEERING DIVISION, PENN STATE UNIVERSITY
Levels of integration are less distinct in objectoriented systems Cluster testing is concerned with integrating and testing clusters of cooperating objects Identify clusters using knowledge of the operation of objects and the system features that are implemented by these clusters
49
Testing is based on a user interactions with the system Has the advantage that it tests system features as experienced by users
Thread testing
A thread consists of all the classes needed to respond to a single external input. Each class is unit tested, and then the thread set is exercised. Tests sequences of object interactions that stop when an object operation does not call on services from another object
Uses-based testing
Begins by testing classes that use few or no server classes. Next, classes that use the first group of classes are tested, followed by classes that use the second group, and so on.
50
Scenario-Based Testing
ENGINEERING DIVISION, PENN STATE UNIVERSITY
Identify scenarios from use-cases and supplement these with interaction diagrams that show the objects involved in the scenario Consider the scenario in the weather station system where a report is generated
51
:WeatherStation
:WeatherData
52
of report request with associated acknowledge and a final output of a report be tested by creating raw data and ensuring that it is summarized properly the same raw data to test the WeatherData object
53
Inheritance means never having to say your sorry Reuse means never having to say your sorry Black box testing is sufficient
54
Implications of Inheritance
Myth:
ENGINEERING DIVISION, PENN STATE UNIVERSITY
specializing
from tested superclasses means subclasses will be correct create new ways to misuse inherited features
Different test
Reality:
Subclasses
cases needed for each context Need to retest inherited methods, even if unchanged.
55
Implications of Reuse
Myth:
Reusing
ENGINEERING DIVISION, PENN STATE UNIVERSITY
a tested class means that the behavior of the server object is trustworthy new usage provides ways to misuse a server.
if many server object of a given class function correctly, nothing is to prevent a new client class from using it incorrectly we can't automatically trust a server because it performs correctly for one client
Even
Reality:
Every
56
Implication of Encapsulation
Myth:
White-box
ENGINEERING DIVISION, PENN STATE UNIVERSITY
testing violates encapsulation, surely black-box testing (of class interfaces) is sufficient.
Reality:
Studies
indicate that thorough BBT sometimes exercises only 1/3 of code. BBT exercises all specified behaviors, what about unspecified behaviors?!
Need
to examine implementation.
57
58
Testing Workbenches
ENGINEERING DIVISION, PENN STATE UNIVERSITY
Testing is an expensive process phase. Testing workbenches provide a range of tools to reduce the time required and total testing costs Most testing workbenches are open systems because testing needs are organization-specific Difficult to integrate with closed design and analysis workbenches
59
Source code
ENGINEERING DIVISION, PENN STATE UNIVERSITY
Test manager
Test data
Oracle
Dynamic analyser
Test results
Test predictions
Execution report
Simulator
File comparator
Report generator
Workbench Components
ENGINEERING DIVISION, PENN STATE UNIVERSITY
Test manager: manages the running of program tests Test data generator: selects test data from database or uses patterns to generate random data of correct form Oracle: Predicts expected results (may be previous version/prototype) Comparator: compare results of oracle and program, or program and previous version (regression test) Dynamic analyzer: counts number of times each statement is executed during test. Simulator: simulates environment (target platform, user interaction, etc)
61
xUnit Framework
ENGINEERING DIVISION, PENN STATE UNIVERSITY
Developed by Kent Beck Makes object-oriented unit testing more accessible. Freeware versions available for most objectoriented languages
www.xprogramming.com/software.htm
62
jUnit successful
63
jUnit unsuccessful
64
Creates objects necessary for the test (fixture) Exercises objects in the fixture Verifies the result.
public String currency() { return fCurrency; } public Money add(Money m) { return new Money(amount()+m.amount(), currency()); } }
66
Key Points
Exhaustive testing is not possible Testing must be done systematically using black-box and white-box testing techniques Testing must be done at both unit and integration levels Object-oriented programming offers its own challenges for testing Testing workbenches and frameworks can help with the testing process
67
References
M. Grottke and K.S. Trivedi. Fighting Bugs: Remove, Retry, Replicate and Rejuvinate. IEEE Computer, February 2007, pp. 107 109. R. Pressman. Software Engineering: A Practitioners Approach, New York, NY: McGraw-Hill, 6th Ed, 2004. I. Sommerville. Software Engineering, 6th Ed. New York, NY: Addison-Wesley, 2000.
68