|
|
| |
What
is conditional coverage? Conditional coverage
is a white-box testing method that measures the percentage
of exercised branches in the code. Conditional coverage is
a special case of test diversity.
|
|
| |
Why
conditional coverage? An un-exercised branch
could contain a defect that will not be
detected. An exercised branch increases the confidence that
the branch contains no defects.
|
|
Conditional
Coverage
Conditional coverage is only a special
case of test diversity. Conditional coverage, also known as branch
testing or decision coverage, measures the percentage of exercised
conditional expressions in conditional statements, multi-branching
statements, and/or loops. A conditional expression is covered when
testing results in both true and false evaluation
of the expression. For example, in the C statement:
if(x == 1 || y == 1)
the conditional expression x
== 1 || y == 1 is completely conditionally covered with values
1 for x and 1 for
y, and 0 for x
and 0 for y.
In a multi-branching statement, such as a switch statement in C++,
conditional coverage measures the percentage of exercised branches
in the multi-branching statement. A multi-branching statement is conditionally
covered when all the branches get hit by testing. For example,
in the C switch statement:
| |
switch(choice) {
case 1: x=x+y;
break;
case 2: x=x*y;
break;
default: x=0;
} |
choice=1 covers
the case 1 branch, choice=2 covers
the case 2 branch, and any other value for choice
covers the default branch.
Conditional
Coverage vs. Test Diversity
Test criteria, such as coverage criteria, neither
make a distinction between the potential infinity of test suites
that satisfy the given criteria nor they make a distinction between
test suites that do not satisfy the given criteria. Typically, some
of the large number of suits that satisfy a criterion detect defects
while other suites that satisfy the criterion do not detect defects.
And vice versa, some of the large number of suits that do not satisfy
a criterion detect defects while other suites that do not satisfy
the criterion do not detect defects. That is, even if you have obtained
100% branch coverage, or 100% of any other type of coverage, you
could have done that in a restricted and concentrated manner with
respect to internal control and data. Such concentrated coverage
has no chance of detecting defects outside of this restricted area.
Smaller the restricted area is smaller the chance are that all the
defects in the code will be detected.
Regardless of the test criterion
used or the reliability model used, test cases that yield similar
control-flows and data-flows exercise the program in a restricted,
confined, dependent, and concentrated manner, whereas test cases with
different control flows and data flows increase the chance
of defect detection by exercising the program in fundamentally different,
dispersed, and independent ways. For
more information on test diversity and the benefits
of diversity over code coverage go to diversity.
|
