Thursday, 23 January 2014

WHITE BOX TESTING (WBT)


Entire WBT is done by developers. It is the testing of each and every line of code in the program. Developers do WBT, sends the s/w to testing team. The testing team does black box testing and checks the s/w against requirements and finds any defects and sends it to the developer. The developers fixes the defect and does WBT and sends it to the testing team. Fixing defect means the defect is removed and the feature is working fine.

Test engineers should not be involved in fixing the bug because,
1) if they spend time in fixing the bug, they lose time to catch some more other defects in the s/w
2) fixing a defect might break a lot of other features. Thus, testers should always identify defects and developers should always be involved in fixing defects.


WBT consists of the following tests :

a) Path testing
Write flow graphs and test all the independent paths.

Writing flow graphs means – flow graphs means representing the flow of the  program, how each program is interlinked with one another.



Test all independent paths – Consider a path from main( ) to function 7. Set the parameters and test if the program is correctly in that path. Similarly test all other paths and fix defects.

b) Condition testing
Test all the logical conditions for both true and false values i.e, we check for both “if” and “else” condition.
If( condition) - true
{
                …….
                …….
}
Else - false
{
                …..
                …..
}
The program should work correctly for both conditions i.e, if condition is true, then else should be false and vice-versa

c) Loop testing
Test the loops(for, while, do-while, etc) for all the cycles and also check for terminating condition if working properly and if the size of the condition is sufficient enough.
For ex, let us consider a program where in the developer has given about 1lakh loops.
{
                While ( 1,00,000 )
                …….
                …….
}
We cannot test this manually for all 1lakh cycles. So we write a small program,
Test A
{
                ……
                ……             }

Which checks for all  1lakh loops. This Test A is known as unit test. The test program is written in the same language as the source code program. Developers only write the test program.


Let us consider the above case as shown in the figure. Suppose the project consists of Requirement A,B,C,D,E. Then the developer writes the Codes A,B,C,D,E for the corresponding requirements. The program consists of 100s of lines of code.
As  developers do WBT, they test the 5 programs line by line of code for defects. If in any the program, there is a defect, the developers identify the defect and rectifies it, and they again have to test the program and all the programs again. This involves a lot of time and effort and slows down the project completion time.
Let us consider another case. Suppose the client asks for requirement changes, then the developers have to do the necessary changes and test the entire 5 programs again. This again involves a lot of time and effort.

This drawback can be corrected in the following manner.
We write a test program for the corresponding program. The developers write these test programs in the same language as the source code. The developers then run these test programs also called as unit test programs. These test programs connect to the main programs and run as the programs. Thus if there is any requirement change or defects in the program, then the developers simply make the changes either in the test program / main program and run the test program as a whole. 



The collection of test programs is known as test suite.
As early as we catch the bug, cost of fixing the bug will be less. If we delay in catching the bug, cost of fixing the bug increases exponentially.

Let us consider a program for the calculator. Let us consider the addition function.


The other test programs are,
Y = callAdd.myAdd (10.5, 5.2)
If ( Y = 15.7)
{
            Print (“test is pass”)
}
Else
{
            Print(“test is fail”)
}

Similarly we do for,
Z = 9,00,000 and 1,00,000
P = -10, -5
In all the above test programs, the variables of X, Y, Z, P are calling the function add.java and getting the add. We do the addition manually and check for the remaining results. We can see that the test programs are more tedious and lengthy and time consuming.
Thus, we have readymade programs.

Like, for example, we have a ready-made program named Junit which has a feature named Assert which runs the above and compares the 2 values and automatically prints pass or fail.
Thus we can replace the test programs with these simple programs,
Import Junit
X = callAdd.myAdd (10, 5)
            Call Assert (X, 15)
It automatically prints pass or fail. Similarly we can do for the other tests for Y, Z and P.

d) From memory point of view of testing
The size of code of a program increases because,

i) The logic used by the programmer may vary. If one developer writes a code of 300kb file size, then another developer may write the same program using different logic and write of code of 200kb file size.

ii ) Reuse of code is not there : for example, if for 5 programs of the same s/w, the 1st ten lines of the program is the same. Then we can write these ten lines as a separate function and write a single function call using which the 5programs can access this function. Also, if any defect is there we can simply change the line of code in the function rather than all the programs.

iii ) Developers declare so many variables, functions which may never be used in any part of the program. Thus, the size of the program increases.

For ex,
Int i = 10 ;
Int j = 15 ;
String S = “hello” ;
. . . . .
. . . . .
. . . . .
.  . . .
. . . . .
Int a = j ;
Create user
{
            . . . .
            . . . .                 100 lines of code
            . . . .
} 

In the above program, we can see that the integer i has never been called anywhere in the program and also the function create user has  never been called anywhere in the program. Thus this leads to memory wastage.

We cannot recognize this error manually by checking the program because of the large program. Thus we have ready-made tools to check for unnecessary variables and functions.


Program A
Int i = 10
Int j = 15
String S = “hello”
. . . .
. . . .
Int a = j
Create user ( )
{
            . . . .
            . . . .
            . . . .
}
We have a tool by name Rational Purify


Programs A, B, and C is given as input to D. D goes into the programs and checks for unused variables. It then gives the result. The developer can then click on the various results and call or delete the unused variables and functions.
This tool is specific only for C, C++ languages. For other languages, we have other similar tools.
iv ) The developer doesn’t use already existing inbuilt functions and sits and writes the entire function using his logic. Thus leads to waste of time and also delays.
Let us consider there is an already inbuilt function sort ( ). Instead the developer sits and writes his own program mysort ( ). This leads to waste of time and effort. Instead he could have used a single function call sort ( ).

e ) Test for response time/ speed/ performance of the program
The reasons for slow performance could be,
i ) Logic used
ii ) Switch case – Do not use nested if, instead use switch case
iii ) Also there must be appropriate use of “or” and “and” for conditional cases.




As the developer is doing WBT, he sees that the program is running slow or the performance of the program is slow. The developer cannot go manually through the code and check which line of the code is slowing the program.

We have a tool by name Rational Quantify to do this job automatically. As soon as all the programs are ready. This tool will go into the program and runs all the programs. The result is shown in the result sheet in the form of thick and thin lines. Thick lines indicate that the piece of code is taking longer time to run. When we double-click on the thick line, then the tool automatically takes us to the line/piece of code which is also colored differently. We can modify that code and again use rational quantify. When the sequence of lines are all thin, we know that the performance of the program has improved.

Developers do a lot of WBT automatically rather than manually because it saves time.

Difference between White Box Testing and Black Box testing
1) White Box Testing                       2) Black Box Testing

a) 1) Done by developers
    2) Done by test engineers
b) 1) Look into the source code and test the logic of the code
     2) Verifying the functionality of the application against requirement specifications
c)  1) Should have knowledge of internal design of the code
     2) No need to have knowledge of internal design of the code
d) 1) Should have knowledge of programming
     2) No need to have knowledge of programming



0 comments:

Post a Comment