COM3529 Software Testing and Analysis

Introduction to
Software Testing
Professor Phil McMinn
Who here likes software
testing?
Who here thinks testing is
hard?
 Who thinks testing and
debugging are basically the
same thing?
Who thinks that testing is
to show that software
works?
Who thinks that testing is
to show that software
doesn’t work?
Who thinks that the idea
behind testing is to
reduce risks involved in
using software?
 Who thinks that the
testing is to help in the
development of
higher quality software?
Beizer’s Maturity Model
Level 0:
Testing is the same as debugging
The basic, least mature view of testing is that of Level 0 – testing is the
same as debugging.
At Level 0 thinking, programmers get their programs to compile, then
debug the programs with a few arbitrary inputs.
This view does not distinguish between a program's incorrect behaviour
and a mistake within the program. It also does very little to help develop
software that is reliable or safe.
(… and testing is not the same as debugging, as demonstrated later.)
Level 1:
The purpose of software testing
is to show software works
A signi cant step up from the naive Level 0.
But, in any but the most trivial of programs, correctness is virtually
impossible to either achieve or demonstrate!
(… another point we will return to later in the lecture.)
fi
Level 2:
The purpose of software testing
is to show software doesn’t work
Although looking for failures is certainly a valid goal, it is also a
negative goal.
If a company is organised where testers and developers are on di erent
teams, you may have a situation where testers may enjoy nding
problems, but the developers never want to nd problems – they want the
software to work!
fi
fi
ff
“ Excellent testing
can make you
unpopular with
almost everyone! ”
— Bill McKeeman
Level 2 testing puts testers and developers into an adversarial
relationship, which can be bad for team morale.

Beyond that, when our primary goal is to look for failures, we are still
left wondering what to do if no failures are found:
Is our work done?
and if so…
Is our software very good or is our testing poor?
Level 3:
The purpose of software testing
is not to show anything in
particular, but just to reduce the
risk of using software
Level 3 thinking lets us accept the fact that whenever we use software,
we incur some risk.
Level 3 thinking lets us accept the fact that whenever we use software,
we incur some risk.

The risk may be small, and the consequences unimportant, or the risk
may be great and the consequences catastrophic, but risk is always
there.

This allows us to realise that the entire development team wants the same
thing – to reduce the risk of using the software.

In Level 3 testing, testing and developing go hand to hand to reduce

risk.
Level 4:
Testing is a mental discipline
that helps all IT professionals
develop higher quality software
Once the testers and developers are on the same “team”, or, testing is
regarded as equally or even more important to development, an
organisation can progress to Level 4 testing.

Level 4 testing de nes testing as a mental discipline that increases

quality.
fi
that helps all IT professionals
develop higher quality software
Once the testers and developers are on the same “team”, or, testing is
regarded as equally or even more important to development, an
organisation can progress to Level 4 testing.

Level 4 testing de nes testing as a mental discipline that increases

quality.

In the same way, Level 4 testing means that the purpose of testing is
to improve the ability of the developers to produce higher quality
software.
fi
 The best use of a spellchecker is not just to
nd misspelled words, but to improve our
ability to spell.
Every time the spell checker nds an
incorrectly spelled word, we have the
opportunity to learn how to spell the word
correctly.

The spell checker is the “expert” on

spelling quality.
fi
fi
Beizer’s Maturity Model
0 The same activity as debugging

1 Purpose is to show software works

2 Purpose is to show software doesn’t work

3 Purpose is to reduce the risk of using software

4 Purpose is to help all IT professionals engineer better software

 Why?

Why Finding ALL Bugs is Impossible

or
Why Software Testing is Hard
The Problem of Exhaustive Testing

The range of an int in Java is -2,147,483,648 to 2,147,483,647 (or 2 )

32

With six ints that’s 232x6 or 2192 (≈ 6 x 1057) unique inputs to try!
Suppose each input takes ≈ 1 nanosecond to execute.
It would take 1041 years to try them all!
The Halting Problem and
Software Testing
The Halting Problem in Computer Science is basically the problem of not
knowing if a program will terminate given some input.

If we give an arbitrary program an input, it has been proven that no program can
be written that can say whether that original program will terminate.

And this is true in software testing: we don’t know if, given our test inputs,
whether the program being tested will get stuck in an in nite loop!

Meaning that in general, exhaustive testing is not be just intractable, it’s

uncomputable too.

fi
“ Software testing
can only show the
presence, not the
absence of bugs
”
— Edsger Dijkstra
 Uncomputable
Tractable problems Intractable problems
problems

Method for solving Cannot be solved

Description Can be solved e ciently exists, but is hopelessly by any computer
time consuming program

Computable in theory
😎 😎 😭
Computable in practice
😎 🤔 😭
u g s i n a
Find the shortest route
i n g a ll b
Example Decryption F i nd ro g r a m
on a map t e r p
compu
ffi
The Oracle Problem
Even if we could
1) execute all software with all inputs
(i.e., if software testing was a tractable problem)
2) guarantee the software terminated with each input
(i.e., if software testing was a computable problem)

We would still need to solve the oracle problem – how to know, given
some input to a software system, that the output it gives is the
correct one.
The Oracle Problem a human being makes a
manual judgment
e.g. an assertion in JUnit

In software testing an oracle is something or someone who can

determine if a software output is correct

No, not the company!

But we don’t need every single
input to ensure the program is
working… right …?
But how do we choose that subset of inputs?
This is the essence of the software testing problem.
We need to choose a set of inputs that will reveal as much
information about the quality of the software as possible.
But we don’t know that we’ve selected all the inputs that will reveal all
of the bugs.
Why Finding ALL Bugs is Impossible, or:
Why Software Testing is Hard
1 Executing all inputs for any non-trivial program is intractable

2 Ensuring the software will terminate with every input is undecidable

3 Recognising correct/incorrect outputs given their corresponding

inputs is at least as hard as building the software in the rst place –
the oracle problem

fi
How do Software Failures
Happen?
A Method and its Tests
A Method and its Tests

PASSED

FAILED
 This method contains a bug. Or more precisely, a defect.
Where is the defect? How does it cause the method to fail?
Defects
Software failures always start with the execution of one or more
defects in the code.
A defect is simply a piece of faulty, incorrect code.
A defect may be a part of or a complete program statement, or may
correspond to statements that don’t exist that should exist.
Although programmers are responsible for making defects in the code,
they may not be technically always be at fault – the problem may have
arisen, for example, from a poorly speci ed set of requirements.
fi
The defect in our
example is with
the second loop
initialiser.

It should start
iterating at i + 1
to check for
duplicates of the
character at i, not
at i itself (which is
guaranteed to be
identical!)
Infections
An infection is what happens when the defect is executed, and the
program’s state is a ected.
When a program’s state is infected it starts to work incorrectly:
• Variables start to take on the wrong values
• Decisions made in the program are evaluated incorrectly, and the
execution path deviates from the correct one.
But at this point, it has not a ected the output of the program (and the
fault, so far, has had no observable e ect).
ff
ff
ff
The infection in
our example
causes the loop
starts iterating an
index in the string
too early.

This further causes

each character in
the string to be
added to the
duplicates set.
But at this point,
there is nothing
observably wrong
with the program.
Failures
A failure occurs when the infection propagates to the output of the
program.
That is, the program is observably behaving incorrectly.
 Failures therefore
depend on when
programs deliver
observable outputs. In
our example, we
interrogate the return
value of the method in
our tests, causing a test
failure.

But during execution of

the whole software, the
method may be used
internally by another
method and a failure
may happen much
later, in a di erent
place.
ff
Failures vs Test Failures
We therefore need to distinguish between failures and test failures.
Failures are when software behaves incorrectly when run as a whole in
production.
Test failures are when tests themselves fail because either:
(a) the test revealed a software failure

(b) the test itself was incorrect, for example it made an incorrect
assertion about the behaviour of the software.
Testing vs Debugging
We can also now debunk the idea that testing and debugging are
the same.
Testing is the process of evaluating software by observing its
execution.
Debugging is the process of tracking failures/test failures back to the
defects that were ultimately responsible for them.
How do Software Failures Happen?
1. The program location
containing a defect is
reached during execution.
😫 Defect
2. The defect infects the
state of the program 🤢 Infection
3. The infection propagates to
the program’s output causing
a failure.
🤮 Failure
Defects are not
always reached
(executed)
Consider what happens
if the inputs string s is
empty.

A good test suite needs

to exercise as much of
the software as possible.

We will come back to this

in later in the module.
 Defects may not
always cause
infections
Consider a defect where
the loop test is reversed (i.e.
“>” is used instead of “<”)

For empty strings, the

defect would be executed,
but no variables take on the
wrong values and the loop
body is not executed – as
normal.

So for this speci c input,

there is no infection.
fi
Infections may
not always
propagate to the
output
Consider what happens
with the original defect if
the inputs string
s=“stst”.

The defect is executed

and the characters “s”
and “t” are entered into
the duplicates set too
early, but the overall
output is correct.
Test cases need The method fails with both tests, but
only one of the tests “revealed” the
to reveal failures failure (by causing a test failure).

PASSED

FAILED
How Software Failures
are Detected by Test Cases.
The RIPR model
Defect Reached 😫
State Infected 🤢
Infection Propagated 🤮
Failure Revealed 🤓
Roadmap of this Module
What we will cover – Weeks 1-5
Theoretical foundations (this lecture) Code Coverage
Structural
Types of testing
Logic
Automated and Manual
Data Flow
Unit, Integration, and System
Input Domain

Unit Testing
Good practices
Test doubles
What we won’t cover…
On the whole, we won’t be covering speci c technologies and
environments.
The practices and techniques we will cover apply to any software
system you are developing.
The course is designed to serve as the foundation of any testing
that you need to do.
Each domain has its own testing practices and tools, however, so you
will need to look for additional resources that focus on those when the
time comes.
fi
Assessment and Feedback
This module is solely assessed by an exam.
Preparation will be via the lab classes, where you will study problems
similar to those that will appear on the exam.
We will provide feedback individually and via model answers.
Computer Laboratory Classes
We will set problem sheets and practical computer-based exercises
that are designed to help you:

• Understand the concepts explained in lectures

• Understand how the lecture material can be applied practically
• Prepare for the exam
You can ask us any questions about the materials.
 Teaching Team

Phil McMinn José Rojas Islam Elgendy Megan Maton Ruizhen (Rui) Gu
Module Lecturer Module Lecturer Teaching Assistant Demonstrator Demonstrator
Weeks 1-5 Weeks 6-10 Practical Sessions Practical Sessions Practical Sessions
Additional Notes / Follow-Up
The Post O ce software scandal mentioned at the start of the
lecture was the subject of an episode of the BBC’s “Panorama”,
which may still be available here:

https://www.bbc.co.uk/iplayer/episode/m0016t20/panorama-the-post-office-scandal
ffi