1. What is black box/white box testing?

Black-box and white-box are test design methods. Black-box test design treats the system as a “black-box”, so it
doesn’t explicitly use knowledge of the internal structure. Black-box test design is usually described as focusing on
testing functional requirements. Synonyms for black-box include: behavioral, functional, opaque-box, and closed-
box. White-box test design allows one to peek inside the “box”, and it focuses specifically on using internal
knowledge of the software to guide the selection of test data. Synonyms for white-box include: structural, glass-box
and clear-box.

While black-box and white-box are terms that are still in popular use, many people prefer the terms "behavioral" and
"structural". Behavioral test design is slightly different from black-box test design because the use of internal
knowledge isn't strictly forbidden, but it's still discouraged. In practice, it hasn't proven useful to use a single test
design method. One has to use a mixture of different methods so that they aren't hindered by the limitations of a
particular one. Some call this "gray-box" or "translucent-box" test design, but others wish we'd stop talking about
boxes altogether.

It is important to understand that these methods are used during the test design phase, and their influence is hard to
see in the tests once they're implemented. Note that any level of testing (unit testing, system testing, etc.) can use
any test design methods. Unit testing is usually associated with structural test design, but this is because testers
usually don't have well-defined requirements at the unit level to validate.

2. What are unit, component and integration testing?

Note that the definitions of unit, component, integration, and integration testing are recursive:

Unit. The smallest compliable component. A unit typically is the work of one programmer (At least in principle).
As defined, it does not include any called sub-components (for procedural languages) or communicating
components in general.

Unit Testing: in unit testing called components (or communicating components) are replaced with stubs,
simulators, or trusted components. Calling components are replaced with drivers or trusted super-components. The
unit is tested in isolation.

component: a unit is a component. The integration of one or more components is a component.

Note: The reason for "one or more" as contrasted to "Two or more" is to allow for components that call themselves
recursively.

component testing: same as unit testing except that all stubs and simulators are replaced with the real thing.
Two components (actually one or more) are said to be integrated when:

a. They have been compiled, linked, and loaded together.

b. They have successfully passed the integration tests at the interface between them.

Thus, components A and B are integrated to create a new, larger, component (A,B). Note that this does not conflict
with the idea of incremental integration—it just means that A is a big component and B, the component added, is a
small one.

Integration testing: carrying out integration tests.

Integration tests (After Leung and White) for procedural languages. This is easily generalized for OO languages by
using the equivalent constructs for message passing. In the following, the word "call" is to be understood in the
most general sense of a data flow and is not restricted to just formal subroutine calls and returns – for example,
passage of data through global data structures and/or the use of pointers.

Let A and B be two components in which A calls B.

Let Ta be the component level tests of A
Testing 1
 Let Tb be the component level tests of B
Tab The tests in A's suite that cause A to call B.
Tbsa The tests in B's suite for which it is possible to sensitize A -- the inputs
are to A, not B.
Tbsa + Tab == the integration test suite (+ = union).

Note: Sensitize is a technical term. It means inputs that will cause a routine to go down a specified path. The
inputs are to A. Not every input to A will cause A to traverse a path in which B is called. Tbsa is the set of tests
which do cause A to follow a path in which B is called. The outcome of the test of B may or may not be affected.

There have been variations on these definitions, but the key point is that it is pretty darn formal and there's a goodly
hunk of testing theory, especially as concerns integration testing, OO testing, and regression testing, based on
them.

As to the difference between integration testing and system testing. System testing specifically goes after behaviors
and bugs that are properties of the entire system as distinct from properties attributable to components (unless, of
course, the component in question is the entire system). Examples of system testing issues:
Resource loss bugs, throughput bugs, performance, security, recovery,
Transaction synchronization bugs (often misnamed "timing bugs").

3. What's the difference between load and stress testing ?

One of the most common, but unfortunate misuse of terminology is treating “load testing” and “stress testing” as
synonymous. The consequence of this ignorant semantic abuse is usually that the system is neither properly “load
tested” nor subjected to a meaningful stress test.

Stress testing is subjecting a system to an unreasonable load while denying it the resources
(e.g., RAM, disc, mips, interrupts, etc.) needed to process that load. The idea is to stress a
system to the breaking point in order to find bugs that will make that break potentially harmful.
The system is not expected to process the overload without adequate resources, but to behave
(e.g., fail) in a decent manner (e.g., not corrupting or losing data). Bugs and failure modes
discovered under stress testing may or may not be repaired depending on the application, the
failure mode, consequences, etc. The load (incoming transaction stream) in stress testing is
often deliberately distorted so as to force the system into resource depletion.

Load testing is subjecting a system to a statistically representative (usually) load. The two main
reasons for using such loads is in support of software reliability testing and in performance
testing. The term "load testing" by itself is too vague and imprecise to warrant use. For example,
do you mean representative load," "overload," "high load," etc. In performance testing, load is
varied from a minimum (zero) to the maximum level the system can sustain without running out
of resources or having, transactions >suffer (application-specific) excessive delay.

A third use of the term is as a test whose objective is to determine the maximum sustainable
load the system can handle. In this usage, "load testing" is merely testing at the highest
transaction arrival rate in performance testing.

4. What's the difference between QA and testing?

QA is more a preventive thing, ensuring quality in the company and therefore the product rather
than just testing the product for software bugs?

TESTING means "quality control"

QUALITY CONTROL measures the quality of a product
QUALITY ASSURANCE measures the quality of processes used to create a
quality product.

Testing 2
5. What is the best tester to developer ratio?

Reported tester: developer ratios range from 10:1 to 1:10.

There's no simple answer. It depends on so many things, Amount of reused code, number and
type of interfaces, platform, quality goals, etc.

It also can depend on the development model. The more specs, the less testers. The roles can
play a big part also. Does QA own beta? Do you include process auditors or planning
activities?

These figures can all vary very widely depending on how you define "tester" and "developer". In
some organizations, a "tester" is anyone who happens to be testing software at the time -- such
as their own. In other organizations, a "tester" is only a member of an independent test group.

It is better to ask about the test labor content than it is to ask about the tester/developer ratio.
The test labor content, across most applications is generally accepted as 50%, when people do
honest accounting. For life-critical software, this can go up to 80%.

6. What is Software Quality Assurance?

Software QA involves the entire software development PROCESS - monitoring and improving
the process, making sure that any agreed-upon standards and procedures are followed, and
ensuring that problems are found and dealt with. It is oriented to 'prevention'.

7. What is Software Testing?

Testing involves operation of a system or application under controlled conditions and evaluating
the results (eg, 'if the user is in interface A of the application while using hardware B, and does
C, then D should happen'). The controlled conditions should include both normal and abnormal
conditions. Testing should intentionally attempt to make things go wrong to determine if things
happen when they shouldn't or things don't happen when they should. It is oriented to 'detection'.

Organizations vary considerably in how they assign responsibility for QA and testing. Sometimes
they're the combined responsibility of one group or individual. Also common are project teams
that include a mix of testers and developers who work closely together, with overall QA
processes monitored by project managers. It will depend on what best fits an organization's size
and business structure.

8. What are some recent major computer system failures caused by Software bugs?

 In March of 2002 it was reported that software bugs in Britain's national tax system
resulted in more than 100,000 erroneous tax overcharges. The problem was partly
attibuted to the difficulty of testing the integration of multiple systems.
 A newspaper columnist reported in July 2001 that a serious flaw was found in off-the-
shelf software that had long been used in systems for tracking certain U.S. nuclear
materials. The same software had been recently donated to another country to be
used in tracking their own nuclear materials, and it was not until scientists in that
country discovered the problem, and shared the information, that U.S. officials
became aware of the problems.

Testing 3
 According to newspaper stories in mid-2001, a major systems development contractor
was fired and sued over problems with a large retirement plan management system.
According to the reports, the client claimed that system deliveries were late, the
software had excessive defects, and it caused other systems to crash.
 In January of 2001 newspapers reported that a major European railroad was hit by the
aftereffects of the Y2K bug. The company found that many of their newer trains would
not run due to their inability to recognize the date '31/12/2000'; the trains were started
by altering the control system's date settings.
 News reports in September of 2000 told of a software vendor settling a lawsuit with a
large mortgage lender; the vendor had reportedly delivered an online mortgage
processing system that did not meet specifications, was delivered late, and didn't
work.
 In early 2000, major problems were reported with a new computer system in a large
suburban U.S. public school district with 100,000+ students; problems included 10,000
erroneous report cards and students left stranded by failed class registration systems;
the district's CIO was fired. The school district decided to reinstate it's original 25-year
old system for at least a year until the bugs were worked out of the new system by the
software vendors.
 In October of 1999 the $125 million NASA Mars Climate Orbiter spacecraft was
believed to be lost in space due to a simple data conversion error. It was determined
that spacecraft software used certain data in English units that should have been in
metric units. Among other tasks, the orbiter was to serve as a communications relay
for the Mars Polar Lander mission, which failed for unknown reasons in December
1999. Several investigating panels were convened to determine the process failures
that allowed the error to go undetected.
 Bugs in software supporting a large commercial high-speed data network affected
70,000 business customers over a period of 8 days in August of 1999. Among those
affected was the electronic trading system of the largest U.S. futures exchange, which
was shut down for most of a week as a result of the outages.
 In April of 1999 a software bug caused the failure of a $1.2 billion military satellite
launch, the costliest unmanned accident in the history of Cape Canaveral launches.
The failure was the latest in a string of launch failures, triggering a complete military
and industry review of U.S. space launch programs, including software integration and
testing processes. Congressional oversight hearings were requested.
 A small town in Illinois received an unusually large monthly electric bill of $7 million in
March of 1999. This was about 700 times larger than its normal bill. It turned out to be
due to bugs in new software that had been purchased by the local power company to
deal with Y2K software issues.
 In early 1999 a major computer game company recalled all copies of a popular new
product due to software problems. The company made a public apology for releasing
a product before it was ready.
 The computer system of a major online U.S. stock trading service failed during trading
hours several times over a period of days in February of 1999 according to nationwide
news reports. The problem was reportedly due to bugs in a software upgrade intended
to speed online trade confirmations.
 In April of 1998 a major U.S. data communications network failed for 24 hours,
crippling a large part of some U.S. credit card transaction authorization systems as
well as other large U.S. bank, retail, and government data systems. The cause was
eventually traced to a software bug.

Testing 4
  January 1998 news reports told of software problems at a major U.S.
telecommunications company that resulted in no charges for long distance calls for a
month for 400,000 customers. The problem went undetected until customers called up
with questions about their bills.
 In November of 1997 the stock of a major health industry company dropped 60% due
to reports of failures in computer billing systems, problems with a large database
conversion, and inadequate software testing. It was reported that more than
$100,000,000 in receivables had to be written off and that multi-million dollar fines
were levied on the company by government agencies.
 A retail store chain filed suit in August of 1997 against a transaction processing system
vendor (not a credit card company) due to the software's inability to handle credit
cards with year 2000 expiration dates.
 In August of 1997 one of the leading consumer credit reporting companies reportedly
shut down their new public web site after less than two days of operation due to
software problems. The new site allowed web site visitors instant access, for a small
fee, to their personal credit reports. However, a number of initial users ended up
viewing each others' reports instead of their own, resulting in irate customers and
nationwide publicity. The problem was attributed to "...unexpectedly high demand from
consumers and faulty software that routed the files to the wrong computers."
 In November of 1996, newspapers reported that software bugs caused the 411
telephone information system of one of the U.S. RBOC's to fail for most of a day. Most
of the 2000 operators had to search through phone books instead of using their
13,000,000-listing database. The bugs were introduced by new software modifications
and the problem software had been installed on both the production and backup
systems. A spokesman for the software vendor reportedly stated that 'It had nothing to
do with the integrity of the software. It was human error.'
 On June 4 1996 the first flight of the European Space Agency's new Ariane 5 rocket
failed shortly after launching, resulting in an estimated uninsured loss of a half billion
dollars. It was reportedly due to the lack of exception handling of a floating-point error
in a conversion from a 64-bit integer to a 16-bit signed integer.
 Software bugs caused the bank accounts of 823 customers of a major U.S. bank to be
credited with $924,844,208.32 each in May of 1996, according to newspaper reports.
The American Bankers Association claimed it was the largest such error in banking
history. A bank spokesman said the programming errors were corrected and all funds
were recovered.
 Software bugs in a Soviet early-warning monitoring system nearly brought on nuclear
war in 1983, according to news reports in early 1999. The software was supposed to
filter out false missile detections caused by Soviet satellites picking up sunlight
reflections off cloud-tops, but failed to do so. Disaster was averted when a Soviet
commander, based on a what he said was a '...funny feeling in my gut', decided the
apparent missile attack was a false alarm. The filtering software code was rewritten.

9. Why is it often hard for management to get serious about quality assurance?

Solving problems is a high-visibility process; preventing problems is low-visibility. This is

illustrated by an old parable:
In ancient China there was a family of healers, one of whom was known throughout the land and
employed as a physician to a great lord. The physician was asked which of his family was the
most skillful healer. He replied,
"I tend to the sick and dying with drastic and dramatic treatments, and on occasion someone is
Testing 5
cured and my name gets out among the lords."
"My elder brother cures sickness when it just begins to take root, and his skills are known among
the local peasants and neighbors."
"My eldest brother is able to sense the spirit of sickness and eradicate it before it takes form. His
name is unknown outside our home."

10. Why does Software have bugs?

 Miscommunication or no communication - as to specifics of what an application should

or shouldn't do (the application's requirements).
 Software complexity - the complexity of current software applications can be difficult to
comprehend for anyone without experience in modern-day software development.
Windows-type interfaces, client-server and distributed applications, data
communications, enormous relational databases, and sheer size of applications have
all contributed to the exponential growth in software/system complexity. And the use of
object-oriented techniques can complicate instead of simplify a project unless it is
well-engineered.
 Programming errors - programmers, like anyone else, can make mistakes.
 changing requirements - the customer may not understand the effects of changes, or
may understand and request them anyway - redesign, rescheduling of engineers,
effects on other projects, work already completed that may have to be redone or
thrown out, hardware requirements that may be affected, etc. If there are many minor
changes or any major changes, known and unknown dependencies among parts of
the project are likely to interact and cause problems, and the complexity of keeping
track of changes may result in errors. Enthusiasm of engineering staff may be
affected. In some fast-changing business environments, continuously modified
requirements may be a fact of life. In this case, management must understand the
resulting risks, and QA and test engineers must adapt and plan for continuous
extensive testing to keep the inevitable bugs from running out of control.
 time pressures - scheduling of software projects is difficult at best, often requiring a lot
of guesswork. When deadlines loom and the crunch comes, mistakes will be made.

egos - people prefer to say things like:

'no problem'
'piece of cake'
'I can whip that out in a few hours'
'it should be easy to update that old code'

instead of:
'that adds a lot of complexity and we could end up
making a lot of mistakes'
'we have no idea if we can do that; we'll wing it'
'I can't estimate how long it will take, until I
take a close look at it'
'we can't figure out what that old spaghetti code
did in the first place'

If there are too many unrealistic 'no problem's', the result is bugs.

Testing 6
  poorly documented code - it's tough to maintain and modify code that is badly written or
poorly documented; the result is bugs. In many organizations management provides no
incentive for programmers to document their code or write clear, understandable code. In
fact, it's usually the opposite: they get points mostly for quickly turning out code, and
there's job security if nobody else can understand it ('if it was hard to write, it should be
hard to read').
 software development tools - visual tools, class libraries, compilers, scripting tools, etc.
often introduce their own bugs or are poorly documented, resulting in added bugs.

11. How can new Software QA processes be introduced in an existing organization?

 A lot depends on the size of the organization and the risks involved. For large
organizations with high-risk (in terms of lives or property) projects, serious management
buy-in is required and a formalized QA process is necessary.
 Where the risk is lower, management and organizational buy-in and QA implementation
may be a slower, step-at-a-time process. QA processes should be balanced with
productivity so as to keep bureaucracy from getting out of hand.
 For small groups or projects, a more ad-hoc process may be appropriate, depending on
the type of customers and projects. A lot will depend on team leads or managers,
feedback to developers, and ensuring adequate communications among customers,
managers, developers, and testers.
 In all cases the most value for effort will be in requirements management processes, with
a goal of clear, complete, testable requirement specifications or expectations.

12. What is verification? validation?

Verification typically involves reviews and meetings to evaluate documents, plans, code,
requirements, and specifications. This can be done with checklists, issues lists, walkthroughs,
and inspection meetings. Validation typically involves actual testing and takes place after
verifications are completed. The term 'IV & V' refers to Independent Verification and Validation.

13. What is a 'walkthrough'?

A 'walkthrough' is an informal meeting for evaluation or informational purposes. Little or no

preparation is usually required.

14. What's an 'inspection'?

An inspection is more formalized than a 'walkthrough', typically with 3-8 people including a
moderator, reader, and a recorder to take notes. The subject of the inspection is typically a
document such as a requirements spec or a test plan, and the purpose is to find problems and
see what's missing, not to fix anything. Attendees should prepare for this type of meeting by
reading thru the document; most problems will be found during this preparation. The result of the
inspection meeting should be a written report. Thorough preparation for inspections is difficult,
painstaking work, but is one of the most cost effective methods of ensuring quality. Employees
who are most skilled at inspections are like the 'eldest brother' in the parable in 'Why is it often
hard for management to get serious about quality assurance?'. Their skill may have low visibility
but they are extremely valuable to any software development organization, since bug prevention
is far more cost-effective than bug detection.

Testing 7
15. What kinds of testing should be considered?

 Black box testing - not based on any knowledge of internal design or code. Tests are
based on requirements and functionality.
 White box testing - based on knowledge of the internal logic of an application's code.
Tests are based on coverage of code statements, branches, paths, conditions.
 unit testing - the most 'micro' scale of testing; to test particular functions or code modules.
Typically done by the programmer and not by testers, as it requires detailed knowledge of
the internal program design and code. Not always easily done unless the application has
a well-designed architecture with tight code; may require developing test driver modules
or test harnesses.
 incremental integration testing - continuous testing of an application as new functionality
is added; requires that various aspects of an application's functionality be independent
enough to work separately before all parts of the program are completed, or that test
drivers be developed as needed; done by programmers or by testers.
 integration testing - testing of combined parts of an application to determine if they
function together correctly. The 'parts' can be code modules, individual applications, client
and server applications on a network, etc. This type of testing is especially relevant to
client/server and distributed systems.
 functional testing - black-box type testing geared to functional requirements of an
application; this type of testing should be done by testers. This doesn't mean that the
programmers shouldn't check that their code works before releasing it (which of course
applies to any stage of testing.)
 system testing - black-box type testing that is based on overall requirements
specifications; covers all combined parts of a system.
 end-to-end testing - similar to system testing; the 'macro' end of the test scale; involves
testing of a complete application environment in a situation that mimics real-world use,
such as interacting with a database, using network communications, or interacting with
other hardware, applications, or systems if appropriate.
 sanity testing - typically an initial testing effort to determine if a new software version is
performing well enough to accept it for a major testing effort. For example, if the new
software is crashing systems every 5 minutes, bogging down systems to a crawl, or
destroying databases, the software may not be in a 'sane' enough condition to warrant
further testing in its current state.
 regression testing - re-testing after fixes or modifications of the software or its
environment. It can be difficult to determine how much re-testing is needed, especially
near the end of the development cycle. Automated testing tools can be especially useful
for this type of testing.
 acceptance testing - final testing based on specifications of the end-user or customer, or
based on use by end-users/customers over some limited period of time.
 load testing - testing an application under heavy loads, such as testing of a web site
under a range of loads to determine at what point the system's response time degrades
or fails.
 stress testing - term often used interchangeably with 'load' and 'performance' testing. Also
used to describe such tests as system functional testing while under unusually heavy
loads, heavy repetition of certain actions or inputs, input of large numerical values, large
complex queries to a database system, etc.

Testing 8
  performance testing - term often used interchangeably with 'stress' and 'load' testing.
Ideally 'performance' testing (and any other 'type' of testing) is defined in requirements
documentation or QA or Test Plans.
 usability testing - testing for 'user-friendliness'. Clearly this is subjective, and will depend
on the targeted end-user or customer. User interviews, surveys, video recording of user
sessions, and other techniques can be used. Programmers and testers are usually not
appropriate as usability testers.
 install/uninstall testing - testing of full, partial, or upgrade install/uninstall processes.
 recovery testing - testing how well a system recovers from crashes, hardware failures, or
other catastrophic problems.
 security testing - testing how well the system protects against unauthorized internal or
external access, willful damage, etc; may require sophisticated testing techniques.
 compatability testing - testing how well software performs in a particular
hardware/software/operating system/network/etc. environment.
 exploratory testing - often taken to mean a creative, informal software test that is not
based on formal test plans or test cases; testers may be learning the software as they test
it.
 ad-hoc testing - similar to exploratory testing, but often taken to mean that the testers
have significant understanding of the software before testing it.
 user acceptance testing - determining if software is satisfactory to an end-user or
customer.
 comparison testing - comparing software weaknesses and strengths to competing
products.
 alpha testing - testing of an application when development is nearing completion; minor
design changes may still be made as a result of such testing. Typically done by end-users
or others, not by programmers or testers.
 beta testing - testing when development and testing are essentially completed and final
bugs and problems need to be found before final release. Typically done by end-users or
others, not by programmers or testers.
 mutation testing - a method for determining if a set of test data or test cases is useful, by
deliberately introducing various code changes ('bugs') and retesting with the original test
data/cases to determine if the 'bugs' are detected. Proper implementation requires large
computational resources.

16. What are 5 common problems in the software development process?

 poor requirements - if requirements are unclear, incomplete, too general, or not

testable, there will be problems.
 unrealistic schedule - if too much work is crammed in too little time, problems are
inevitable.
 inadequate testing - no one will know whether or not the program is any good until the
customer complains or systems crash.
 featuritis - requests to pile on new features after development is underway; extremely
common.
 miscommunication - if developers don't know what's needed or customer's have
erroneous expectations, problems are guaranteed.

17. What are 5 common solutions to software development problems?

Testing 9
  solid requirements - clear, complete, detailed, cohesive, attainable, testable
requirements that are agreed to by all players. Use prototypes to help nail down
requirements.
 realistic schedules - allow adequate time for planning, design, testing, bug fixing, re-
testing, changes, and documentation; personnel should be able to complete the
project without burning out.
 adequate testing - start testing early on, re-test after fixes or changes, plan for
adequate time for testing and bug-fixing.
 stick to initial requirements as much as possible - be prepared to defend against
changes and additions once development has begun, and be prepared to explain
consequences. If changes are necessary, they should be adequately reflected in
related schedule changes. If possible, use rapid prototyping during the design phase
so that customers can see what to expect. This will provide them a higher comfort
level with their requirements decisions and minimize changes later on.
 communication - require walkthroughs and inspections when appropriate; make
extensive use of group communication tools - e-mail, groupware, networked bug-
tracking tools and change management tools, intranet capabilities, etc.; insure that
documentation is available and up-to-date - preferably electronic, not paper; promote
teamwork and cooperation; use prototypes early on so that customers' expectations
are clarified.

18. What is software 'quality'?

Quality software is reasonably bug-free, delivered on time and within budget, meets
requirements and/or expectations, and is maintainable. However, quality is obviously a
subjective term. It will depend on who the 'customer' is and their overall influence in the scheme
of things. A wide-angle view of the 'customers' of a software development project might include
end-users, customer acceptance testers, customer contract officers, customer management, the
development organization's management/accountants/testers/salespeople, future software
maintenance engineers, stockholders, magazine columnists, etc. Each type of 'customer' will
have their own slant on 'quality' - the accounting department might define quality in terms of
profits while an end-user might define quality as user-friendly and bug-free.

19. What is 'good code'?

'Good code' is code that works, is bug free, and is readable and maintainable. Some
organizations have coding 'standards' that all developers are supposed to adhere to, but
everyone has different ideas about what's best, or what is too many or too few rules. There are
also various theories and metrics, such as McCabe Complexity metrics. It should be kept in mind
that excessive use of standards and rules can stifle productivity and creativity. 'Peer reviews',
'buddy checks' code analysis tools, etc. can be used to check for problems and enforce
standards.
For C and C++ coding, here are some typical ideas to consider in setting rules/standards; these
may or may not apply to a particular situation:

 minimize or eliminate use of global variables.

 use descriptive function and method names - use both upper and lower case, avoid
abbreviations, use as many characters as necessary to be adequately descriptive (use of
more than 20 characters is not out of line); be consistent in naming conventions.

Testing 10
  use descriptive variable names - use both upper and lower case, avoid abbreviations, use
as many characters as necessary to be adequately descriptive (use of more than 20
characters is not out of line); be consistent in naming conventions.
 function and method sizes should be minimized; less than 100 lines of code is good, less
than 50 lines is preferable.
 function descriptions should be clearly spelled out in comments preceding a function's
code.
 organize code for readability.
 use whitespace generously - vertically and horizontally
 each line of code should contain 70 characters max.
 one code statement per line.
 coding style should be consistent throught a program (eg, use of brackets, indentations,
naming conventions, etc.)
 in adding comments, err on the side of too many rather than too few comments; a
common rule of thumb is that there should be at least as many lines of comments
(including header blocks) as lines of code.
 no matter how small, an application should include documentaion of the overall program
function and flow (even a few paragraphs is better than nothing); or if possible a separate
flow chart and detailed program documentation.
 make extensive use of error handling procedures and status and error logging.
 for C++, to minimize complexity and increase maintainability, avoid too many levels of
inheritance in class heirarchies (relative to the size and complexity of the application).
Minimize use of multiple inheritance, and minimize use of operator overloading (note that
the Java programming language eliminates multiple inheritance and operator
overloading.)
 for C++, keep class methods small, less than 50 lines of code per method is preferable.
 for C++, make liberal use of exception handlers

20. What is 'good design'?

'Design' could refer to many things, but often refers to 'functional design' or 'internal design'.
Good internal design is indicated by software code whose overall structure is clear,
understandable, easily modifiable, and maintainable; is robust with sufficient error-handling and
status logging capability; and works correctly when implemented. Good functional design is
indicated by an application whose functionality can be traced back to customer and end-user
requirements. For programs that have a user interface, it's often a good idea to assume that the
end user will have little computer knowledge and may not read a user manual or even the on-
line help; some common rules-of-thumb include:

 the program should act in a way that least surprises the user
 it should always be evident to the user what can be done next and how to exit
 the program shouldn't let the users do something stupid without warning them.

21. What is SEI? CMM? ISO? IEEE? ANSI? Will it help?

 SEI = 'Software Engineering Institute' at Carnegie-Mellon University; initiated by the

U.S. Defense Department to help improve software development processes.
 CMM = 'Capability Maturity Model', developed by the SEI. It's a model of 5 levels of
organizational 'maturity' that determine effectiveness in delivering quality software. It is
Testing 11
 geared to large organizations such as large U.S. Defense Department contractors.
However, many of the QA processes involved are appropriate to any organization, and
if reasonably applied can be helpful. Organizations can receive CMM ratings by
undergoing assessments by qualified auditors.

Level 1 - characterized by chaos, periodic panics, and heroic

efforts required by individuals to successfully
complete projects. Few if any processes in place;
successes may not be repeatable.

Level 2 - software project tracking, requirements management,

realistic planning, and configuration management
processes are in place; successful practices can
be repeated.

Level 3 - standard software development and maintenance processes

are integrated throughout an organization; a Software
Engineering Process Group is is in place to oversee
software processes, and training programs are used to
ensure understanding and compliance.

Level 4 - metrics are used to track productivity, processes,

and products. Project performance is predictable,
and quality is consistently high.

Level 5 - the focus is on continouous process improvement. The

impact of new processes and technologies can be
predicted and effectively implemented when required.

Perspective on CMM ratings: During 1997-2001, 1018 organizations

were assessed. Of those, 27% were rated at Level 1, 39% at 2,
23% at 3, 6% at 4, and 5% at 5. (For ratings during the period
1992-96, 62% were at Level 1, 23% at 2, 13% at 3, 2% at 4, and
0.4% at 5.) The median size of organizations was 100 software
engineering/maintenance personnel; 32% of organizations were
U.S. federal contractors or agencies. For those rated at
Level 1, the most problematical key process area was in
Software Quality Assurance.

 ISO = 'International Organisation for Standardization' - The ISO 9001:2000 standard

(which replaces the previous standard of 1994) concerns quality systems that are
assessed by outside auditors, and it applies to many kinds of production and
manufacturing organizations, not just software. It covers documentation, design,
development, production, testing, installation, servicing, and other processes. The full set
of standards consists of: (a)Q9001-2000 - Quality Management Systems: Requirements;
(b)Q9000-2000 - Quality Management Systems: Fundamentals and Vocabulary;
(c)Q9004-2000 - Quality Management Systems: Guidelines for Performance
Improvements. To be ISO 9001 certified, a third-party auditor assesses an organization,
and certification is typically good for about 3 years, after which a complete reassessment
is required. Note that ISO certification does not necessarily indicate quality products - it
indicates only that documented processes are followed.
Testing 12
  IEEE = 'Institute of Electrical and Electronics Engineers' - among other things, creates
standards such as 'IEEE Standard for Software Test Documentation' (IEEE/ANSI
Standard 829), 'IEEE Standard of Software Unit Testing (IEEE/ANSI Standard 1008),
'IEEE Standard for Software Quality Assurance Plans' (IEEE/ANSI Standard 730), and
others.
 ANSI = 'American National Standards Institute', the primary industrial standards body in
the U.S.; publishes some software-related standards in conjunction with the IEEE and
ASQ (American Society for Quality).
 Other software development process assessment methods besides CMM and ISO 9000
include SPICE, Trillium, TickIT. and Bootstrap.

22. What is the 'software life cycle'?

The life cycle begins when an application is first conceived and ends when it is no longer in use.
It includes aspects such as initial concept, requirements analysis, functional design, internal
design, documentation planning, test planning, coding, document preparation, integration,
testing, maintenance, updates, retesting, phase-out, and other aspects.

23. Will automated testing tools make testing easier?

 Possibly. For small projects, the time needed to learn and implement them may not be
worth it. For larger projects, or on-going long-term projects they can be valuable.
 A common type of automated tool is the 'record/playback' type. For example, a tester
could click through all combinations of menu choices, dialog box choices, buttons, etc.
in an application GUI and have them 'recorded' and the results logged by a tool. The
'recording' is typically in the form of text based on a scripting language that is
interpretable by the testing tool. If new buttons are added, or some underlying code in
the application is changed, etc. the application can then be retested by just 'playing
back' the 'recorded' actions, and comparing the logging results to check effects of the
changes. The problem with such tools is that if there are continual changes to the
system being tested, the 'recordings' may have to be changed so much that it
becomes very time-consuming to continuously update the scripts. Additionally,
interpretation of results (screens, data, logs, etc.) can be a difficult task. Note that
there are record/playback tools for text-based interfaces also, and for all types of
platforms.
 Other automated tools can include:

code analyzers - monitor code complexity, adherence to

standards, etc.

coverage analyzers - these tools check which parts of the

code have been exercised by a test, and may
be oriented to code statement coverage,
condition coverage, path coverage, etc.

memory analyzers - such as bounds-checkers and leak detectors.

load/performance test tools - for testing client/server

and web applications under various load
levels.
Testing 13
 web test tools - to check that links are valid, HTML code
usage is correct, client-side and
server-side programs work, a web site's
interactions are secure.

other tools - for test case management, documentation

management, bug reporting, and configuration
management.

24. What makes a good test engineer?

A good test engineer has a 'test to break' attitude, an ability to take the point of view of the
customer, a strong desire for quality, and an attention to detail. Tact and diplomacy are useful in
maintaining a cooperative relationship with developers, and an ability to communicate with both
technical (developers) and non-technical (customers, management) people is useful. Previous
software development experience can be helpful as it provides a deeper understanding of the
software development process, gives the tester an appreciation for the developers' point of view,
and reduce the learning curve in automated test tool programming. Judgment skills are needed
to assess high-risk areas of an application on which to focus testing efforts when time is limited.

25. What makes a good Software QA engineer?

The same qualities a good tester has are useful for a QA engineer. Additionally, they must be
able to understand the entire software development process and how it can fit into the business
approach and goals of the organization. Communication skills and the ability to understand
various sides of issues are important. In organizations in the early stages of implementing QA
processes, patience and diplomacy are especially needed. An ability to find problems as well as
to see 'what's missing' is important for inspections and reviews.

26. What makes a good QA or Test manager?

A good QA, test, or QA/Test(combined) manager should:

 be familiar with the software development process

 be able to maintain enthusiasm of their team and promote a positive atmosphere,
despite what is a somewhat 'negative' process (e.g., looking for or preventing
problems)
 be able to promote teamwork to increase productivity
 be able to promote cooperation between software, test, and QA engineers
 have the diplomatic skills needed to promote improvements in QA processes
 have the ability to withstand pressures and say 'no' to other managers when quality is
insufficient or QA processes are not being adhered to
 have people judgement skills for hiring and keeping skilled personnel
 be able to communicate with technical and non-technical people, engineers,
managers, and customers.
 be able to run meetings and keep them focused

27. What's the role of documentation in QA?

Testing 14
Critical. (Note that documentation can be electronic, not necessarily paper.) QA practices should
be documented such that they are repeatable. Specifications, designs, business rules,
inspection reports, configurations, code changes, test plans, test cases, bug reports, user
manuals, etc. should all be documented. There should ideally be a system for easily finding and
obtaining documents and determining what documentation will have a particular piece of
information. Change management for documentation should be used if possible.

28. What's the big deal about 'requirements'?

One of the most reliable methods of insuring problems, or failure, in a complex software project
is to have poorly documented requirements specifications. Requirements are the details
describing an application's externally-perceived functionality and properties. Requirements
should be clear, complete, reasonably detailed, cohesive, attainable, and testable. A non-
testable requirement would be, for example, 'user-friendly' (too subjective). A testable
requirement would be something like 'the user must enter their previously-assigned password to
access the application'. Determining and organizing requirements details in a useful and efficient
way can be a difficult effort; different methods are available depending on the particular project.
Many books are available that describe various approaches to this task.

Care should be taken to involve ALL of a project's significant 'customers' in the requirements
process. 'Customers' could be in-house personnel or out, and could include end-users, customer
acceptance testers, customer contract officers, customer management, future software
maintenance engineers, salespeople, etc. Anyone who could later derail the project if their
expectations aren't met should be included if possible.

Organizations vary considerably in their handling of requirements specifications. Ideally, the

requirements are spelled out in a document with statements such as 'The product shall.....'.
'Design' specifications should not be confused with 'requirements'; design specifications should
be traceable back to the requirements.

In some organizations requirements may end up in high level project plans, functional
specification documents, in design documents, or in other documents at various levels of detail.
No matter what they are called, some type of documentation with detailed requirements will be
needed by testers in order to properly plan and execute tests. Without such documentation,
there will be no clear-cut way to determine if a software application is performing correctly.

29. What steps are needed to develop and run software tests?

The following are some of the steps to consider:

 Obtain requirements, functional design, and internal design specifications and other
necessary documents
 Obtain budget and schedule requirements
 Determine project-related personnel and their responsibilities, reporting requirements,
required standards and processes (such as release processes, change processes, etc.)
 Identify application's higher-risk aspects, set priorities, and determine scope and
limitations of tests
 Determine test approaches and methods - unit, integration, functional, system, load,
usability tests, etc.
 Determine test environment requirements (hardware, software, communications, etc.)
Testing 15
  Determine testware requirements (record/playback tools, coverage analyzers, test
tracking, problem/bug tracking, etc.)
 Determine test input data requirements
 Identify tasks, those responsible for tasks, and labor requirements
 Set schedule estimates, timelines, milestones
 Determine input equivalence classes, boundary value analyses, error classes
 Prepare test plan document and have needed reviews/approvals
 Write test cases
 Have needed reviews/inspections/approvals of test cases
 Prepare test environment and testware, obtain needed user manuals/reference
documents/configuration guides/installation guides, set up test tracking processes, set up
logging and archiving processes, set up or obtain test input data
 Obtain and install software releases
 Perform tests
 Evaluate and report results
 Track problems/bugs and fixes
 Retest as needed
 Maintain and update test plans, test cases, test environment, and testware through life
cycle

30. What's a 'test plan'?

A software project test plan is a document that describes the objectives, scope, approach, and
focus of a software testing effort. The process of preparing a test plan is a useful way to think
through the efforts needed to validate the acceptability of a software product. The completed
document will help people outside the test group understand the 'why' and 'how' of product
validation. It should be thorough enough to be useful but not so thorough that no one outside the
test group will read it. The following are some of the items that might be included in a test plan,
depending on the particular project:

 Title
 Identification of software including version/release numbers
 Revision history of document including authors, dates, approvals
 Table of Contents
 Purpose of document, intended audience
 Objective of testing effort
 Software product overview
 Relevant related document list, such as requirements, design documents, other test
plans, etc.
 Relevant standards or legal requirements
 Traceability requirements
 Relevant naming conventions and identifier conventions
 Overall software project organization and personnel/contact-info/responsibilties
 Test organization and personnel/contact-info/responsibilities
 Assumptions and dependencies
 Project risk analysis
 Testing priorities and focus
 Scope and limitations of testing
Testing 16
  Test outline - a decomposition of the test approach by test type, feature, functionality,
process, system, module, etc. as applicable
 Outline of data input equivalence classes, boundary value analysis, error classes
 Test environment - hardware, operating systems, other required software, data
configurations, interfaces to other systems
 Test environment validity analysis - differences between the test and production systems
and their impact on test validity.
 Test environment setup and configuration issues
 Software migration processes
 Software CM processes
 Test data setup requirements
 Database setup requirements
 Outline of system-logging/error-logging/other capabilities, and tools such as screen
capture software, that will be used to help describe and report bugs
 Discussion of any specialized software or hardware tools that will be used by testers to
help track the cause or source of bugs
 Test automation - justification and overview
 Test tools to be used, including versions, patches, etc.
 Test script/test code maintenance processes and version control
 Problem tracking and resolution - tools and processes
 Project test metrics to be used
 Reporting requirements and testing deliverables
 Software entrance and exit criteria
 Initial sanity testing period and criteria
 Test suspension and restart criteria
 Personnel allocation
 Personnel pre-training needs
 Test site/location
 Outside test organizations to be utilized and their purpose, responsibilities, deliverables,
contact persons, and coordination issues
 Relevant proprietary, classified, security, and licensing issues.
 Open issues
 Appendix - glossary, acronyms, etc.

31. What's a 'test case'?

 A test case is a document that describes an input, action, or event and an expected
response, to determine if a feature of an application is working correctly. A test case
should contain particulars such as test case identifier, test case name, objective, test
conditions/setup, input data requirements, steps, and expected results.
 Note that the process of developing test cases can help find problems in the requirements
or design of an application, since it requires completely thinking through the operation of
the application. For this reason, it's useful to prepare test cases early in the development
cycle if possible.

32. What should be done after a bug is found?

Testing 17
The bug needs to be communicated and assigned to developers that can fix it. After the problem
is resolved, fixes should be re-tested, and determinations made regarding requirements for
regression testing to check that fixes didn't create problems elsewhere. If a problem-tracking
system is in place, it should encapsulate these processes. A variety of commercial problem-
tracking/management software tools are available. The following are items to consider in the
tracking process:

 Complete information such that developers can understand the bug, get an idea of it's
severity, and reproduce it if necessary.
 Bug identifier (number, ID, etc.)
 Current bug status (e.g., 'Released for Retest', 'New', etc.)
 The application name or identifier and version
 The function, module, feature, object, screen, etc. where the bug occurred
 Environment specifics, system, platform, relevant hardware specifics
 Test case name/number/identifier
 One-line bug description
 Full bug description
 Description of steps needed to reproduce the bug if not covered by a test case or if the
developer doesn't have easy access to the test case/test script/test tool
 Names and/or descriptions of file/data/messages/etc. used in test
 File excerpts/error messages/log file excerpts/screen shots/test tool logs that would be
helpful in finding the cause of the problem
 Severity estimate (a 5-level range such as 1-5 or 'critical'-to-'low' is common)
 Was the bug reproducible?
 Tester name
 Test date
 Bug reporting date
 Name of developer/group/organization the problem is assigned to
 Description of problem cause
 Description of fix
 Code section/file/module/class/method that was fixed
 Date of fix
 Application version that contains the fix
 Tester responsible for retest
 Retest date
 Retest results
 Regression testing requirements
 Tester responsible for regression tests
 Regression testing results

A reporting or tracking process should enable notification of appropriate personnel at various

stages. For instance, testers need to know when retesting is needed, developers need to know
when bugs are found and how to get the needed information, and reporting/summary
capabilities are needed for managers.

33. What is 'configuration management'?

Testing 18
Configuration management covers the processes used to control, coordinate, and track: code,
requirements, documentation, problems, change requests, designs,
tools/compilers/libraries/patches, changes made to them, and who makes the changes.

34. What if the software is so buggy it can't really be tested at all?

The best bet in this situation is for the testers to go through the process of reporting whatever
bugs or blocking-type problems initially show up, with the focus being on critical bugs. Since this
type of problem can severely affect schedules, and indicates deeper problems in the software
development process (such as insufficient unit testing or insufficient integration testing, poor
design, improper build or release procedures, etc.) managers should be notified, and provided
with some documentation as evidence of the problem.

35. How can it be known when to stop testing?

This can be difficult to determine. Many modern software applications are so complex, and run in
such an interdependent environment, that complete testing can never be done. Common factors
in deciding when to stop are:

 Deadlines (release deadlines, testing deadlines, etc.)

 Test cases completed with certain percentage passed
 Test budget depleted
 Coverage of code/functionality/requirements reaches a specified point
 Bug rate falls below a certain level
 Beta or alpha testing period ends

36. What if there isn't enough time for thorough testing?

Use risk analysis to determine where testing should be focused.

Since it's rarely possible to test every possible aspect of an application, every possible
combination of events, every dependency, or everything that could go wrong, risk analysis is
appropriate to most software development projects. This requires judgement skills, common
sense, and experience. (If warranted, formal methods are also available.) Considerations can
include:

 Which functionality is most important to the project's intended purpose?

 Which functionality is most visible to the user?
 Which functionality has the largest safety impact?
 Which functionality has the largest financial impact on users?
 Which aspects of the application are most important to the customer?
 Which aspects of the application can be tested early in the development cycle?
 Which parts of the code are most complex, and thus most subject to errors?
 Which parts of the application were developed in rush or panic mode?
 Which aspects of similar/related previous projects caused problems?
 Which aspects of similar/related previous projects had large maintenance expenses?
 Which parts of the requirements and design are unclear or poorly thought out?
 What do the developers think are the highest-risk aspects of the application?
 What kinds of problems would cause the worst publicity?
 What kinds of problems would cause the most customer service complaints?
Testing 19
  What kinds of tests could easily cover multiple functionalities?
 Which tests will have the best high-risk-coverage to time-required ratio?

37. What can be done if requirements are changing continuously?

A common problem and a major headache.

 Work with the project's stakeholders early on to understand how requirements might
change so that alternate test plans and strategies can be worked out in advance, if
possible.
 It's helpful if the application's initial design allows for some adaptability so that later
changes do not require redoing the application from scratch.
 If the code is well-commented and well-documented this makes changes easier for the
developers.
 Use rapid prototyping whenever possible to help customers feel sure of their
requirements and minimize changes.
 The project's initial schedule should allow for some extra time commensurate with the
possibility of changes.
 Try to move new requirements to a 'Phase 2' version of an application, while using the
original requirements for the 'Phase 1' version.
 Negotiate to allow only easily-implemented new requirements into the project, while
moving more difficult new requirements into future versions of the application.
 Be sure that customers and management understand the scheduling impacts, inherent
risks, and costs of significant requirements changes. Then let management or the
customers (not the developers or testers) decide if the changes are warranted - after all,
that's their job.
 Balance the effort put into setting up automated testing with the expected effort required
to re-do them to deal with changes.
 Try to design some flexibility into automated test scripts.
 Focus initial automated testing on application aspects that are most likely to remain
unchanged.
 Devote appropriate effort to risk analysis of changes to minimize regression testing
needs.
 Design some flexibility into test cases (this is not easily done; the best bet might be to
minimize the detail in the test cases, or set up only higher-level generic-type test plans)
 Focus less on detailed test plans and test cases and more on ad hoc testing (with an
understanding of the added risk that this entails).

38. What if the project isn't big enough to justify extensive testing?

Consider the impact of project errors, not the size of the project. However, if extensive testing is
still not justified, risk analysis is again needed and the same considerations as described
previously in 'What if there isn't enough time for thorough testing?' apply. The tester might then
do ad hoc testing, or write up a limited test plan based on the risk analysis.

39. What if the application has functionality that wasn't in the requirements?

It may take serious effort to determine if an application has significant unexpected or hidden
functionality, and it would indicate deeper problems in the software development process. If the

Testing 20
functionality isn't necessary to the purpose of the application, it should be removed, as it may
have unknown impacts or dependencies that were not taken into account by the designer or the
customer. If not removed, design information will be needed to determine added testing needs or
regression testing needs. Management should be made aware of any significant added risks as
a result of the unexpected functionality. If the functionality only effects areas such as minor
improvements in the user interface, for example, it may not be a significant risk.

40. How can Software QA processes be implemented without stifling productivity?

By implementing QA processes slowly over time, using consensus to reach agreement on
processes, and adjusting and experimenting as an organization grows and matures, productivity
will be improved instead of stifled. Problem prevention will lessen the need for problem
detection, panics and burn-out will decrease, and there will be improved focus and less wasted
effort. At the same time, attempts should be made to keep processes simple and efficient,
minimize paperwork, promote computer-based processes and automated tracking and reporting,
minimize time required in meetings, and promote training as part of the QA process. However,
no one - especially talented technical types - likes rules or bureacracy, and in the short run
things may slow down a bit. A typical scenario would be that more days of planning and
development will be needed, but less time will be required for late-night bug-fixing and calming
of irate customers.

41. What if an organization is growing so fast that fixed QA processes are impossible?

This is a common problem in the software industry, especially in new technology areas. There is
no easy solution in this situation, other than:

 Hire good people

 Management should 'ruthlessly prioritize' quality issues and maintain focus on the
customer
 Everyone in the organization should be clear on what 'quality' means to the customer

42. How does a client/server environment affect testing?

Client/server applications can be quite complex due to the multiple dependencies among clients,
data communications, hardware, and servers. Thus testing requirements can be extensive.
When time is limited (as it usually is) the focus should be on integration and system testing.
Additionally, load/stress/performance testing may be useful in determining client/server
application limitations and capabilities. There are commercial tools to assist with such testing.

43. How can World Wide Web sites be tested?

Web sites are essentially client/server applications - with web servers and 'browser' clients.
Consideration should be given to the interactions between html pages, TCP/IP communications,
Internet connections, firewalls, applications that run in web pages (such as applets, javascript,
plug-in applications), and applications that run on the server side (such as cgi scripts, database
interfaces, logging applications, dynamic page generators, asp, etc.). Additionally, there are a
wide variety of servers and browsers, various versions of each, small but sometimes significant
differences between them, variations in connection speeds, rapidly changing technologies, and
multiple standards and protocols. The end result is that testing for web sites can become a major
ongoing effort. Other considerations might include:

Testing 21
  What are the expected loads on the server (e.g., number of hits per unit time?), and what
kind of performance is required under such loads (such as web server response time,
database query response times). What kinds of tools will be needed for performance
testing (such as web load testing tools, other tools already in house that can be adapted,
web robot downloading tools, etc.)?
 Who is the target audience? What kind of browsers will they be using? What kind of
connection speeds will they by using? Are they intra- organization (thus with likely high
connection speeds and similar browsers) or Internet-wide (thus with a wide variety of
connection speeds and browser types)?
 What kind of performance is expected on the client side (e.g., how fast should pages
appear, how fast should animations, applets, etc. load and run)?
 Will down time for server and content maintenance/upgrades be allowed? how much?
 What kinds of security (firewalls, encryptions, passwords, etc.) will be required and what
is it expected to do? How can it be tested?
 How reliable are the site's Internet connections required to be? And how does that affect
backup system or redundant connection requirements and testing?
 What processes will be required to manage updates to the web site's content, and what
are the requirements for maintaining, tracking, and controlling page content, graphics,
links, etc.?
 Which HTML specification will be adhered to? How strictly? What variations will be
allowed for targeted browsers?
 Will there be any standards or requirements for page appearance and/or graphics
throughout a site or parts of a site??
 How will internal and external links be validated and updated? how often?
 Can testing be done on the production system, or will a separate test system be
required? How are browser caching, variations in browser option settings, dial-up
connection variabilities, and real-world internet 'traffic congestion' problems to be
accounted for in testing?
 How extensive or customized are the server logging and reporting requirements; are they
considered an integral part of the system and do they require testing?
 How are cgi programs, applets, javascripts, ActiveX components, etc. to be maintained,
tracked, controlled, and tested?
 Pages should be 3-5 screens max unless content is tightly focused on a single topic. If
larger, provide internal links within the page.
 The page layouts and design elements should be consistent throughout a site, so that it's
clear to the user that they're still within a site.
 Pages should be as browser-independent as possible, or pages should be provided or
generated based on the browser-type.
 All pages should have links external to the page; there should be no dead-end pages.
 The page owner, revision date, and a link to a contact person or organization should be
included on each page.

44. How is testing affected by object-oriented designs?

Well-engineered object-oriented design can make it easier to trace from code to internal design
to functional design to requirements. While there will be little affect on black box testing (where
an understanding of the internal design of the application is unnecessary), white-box testing can
be oriented to the application's objects. If the application was well-designed this can simplify test
design.
Testing 22
45. What is Extreme Programming and what's it got to do with testing?

Extreme Programming (XP) is a software development approach for small teams on risk-prone
projects with unstable requirements. It was created by Kent Beck who described the approach in
his book 'Extreme Programming Explained'. Testing ('extreme testing') is a core aspect of
Extreme Programming. Programmers are expected to write unit and functional test code first -
before the application is developed. Test code is under source control along with the rest of the
code. Customers are expected to be an integral part of the project team and to help develope
scenarios for acceptance/black box testing. Acceptance tests are preferably automated, and are
modified and rerun for each of the frequent development iterations. QA and test personnel are
also required to be an integral part of the project team. Detailed requirements documentation is
not used, and frequent re-scheduling, re-estimating, and re-prioritizing is expected.

46. Common Software Errors

Introduction

This document takes you through whirl-wind tour of common software errors. This is an excellent
aid for software testing. It helps you to identify errors systematically and increases the efficiency
of software testing and improves testing productivity. For more information, please refer Testing
Computer Software, Wiley Edition.

Type of Errors

• User Interface Errors

• Error Handling

• Boundary related errors

• Calculation errors

• Initial and Later states

• Control flow errors

• Errors in Handling or Interpreting Data

• Race Conditions

• Load Conditions

• Hardware

• Source, Version and ID Control

• Testing Errors

Let us go through details of each kind of error.

User Interface Errors

Testing 23
 Functionality
Sl No Possible Error Conditions
1 Excessive Functionality
2 Inflated impression of functionality
3 Inadequacy for the task at hand
4 Missing function
5 Wrong function
6 Functionality must be created by user
7 Doesn't do what the user expects

Communication
Missing Information
Sl No Possible Error Conditions
1 No on Screen instructions
2 Assuming printed documentation is already available.
3 Undocumented features
4 States that appear impossible to exit
5 No cursor
6 Failure to acknowledge input
7 Failure to show activity during long delays
8 Failure to advise when a change will take effect
9 Failure to check for the same document being opened twice
Wrong, misleading, confusing information
10 Simple factual errors
11 Spelling errors
12 Inaccurate simplifications
13 Invalid metaphors
14 Confusing feature names
15 More than one name for the same feature
16 Information overland
17 When are data saved
18 Wrong function
19 Functionality must be created by user
20 Poor external modularity
Help text and error messages
21 Inappropriate reading levels
22 Verbosity
23 Inappropriate emotional tone
24 Factual errors
25 Context errors
26 Failure to identify the source of error
27 Forbidding a resource without saying why
28 Reporting non-errors
29 Failure to highlight the part of the screen
30 Failure to clear highlighting
31 Wrong/partial string displayed
32 Message displayed for too long or not long enough
Display Layout
Testing 24
33 Poor aesthetics in screen layout
34 Menu Layout errors
35 Dialog box layout errors
36 Obscured Instructions
37 Misuse of flash
38 Misuse of color
39 Heavy reliance on color
40 Inconsistent with the style of the environment
41 Cannot get rid of on screen information
Output
42 Can't output certain data
43 Can't redirect output
44 Format incompatible with a follow-up process
45 Must output too little or too much
46 Can't control output layout
47 Absurd printout level of precision
48 Can't control labeling of tables or figures
49 Can't control scaling of graphs
Performance
50 Program Speed
51 User Throughput
52 Can't redirect output
53 Perceived performance
54 Slow program
55 slow echoing
56 how to reduce user throughput
57 Poor responsiveness
58 No type ahead
59 No warning that the operation takes long time
60 No progress reports
61 Problems with time-outs
62 Program pesters you

Program Rigidity
User tailorability
Sl No Possible Error Conditions
1 Can't turn off case sensitivity
2 Can't tailor to hardware at hand
3 Can't change device initialization
4 Can't turn off automatic changes
5 Can't slow down/speed up scrolling
6 Can't do what you did last time
7 Failure to execute a customization commands
8 Failure to save customization commands
9 Side effects of feature changes
10 Can't turn off the noise
11 Infinite tailorability
Who is in control?
12 Unnecessary imposition of a conceptual style
Testing 25
13 Novice friendly, experienced hostile
14 Surplus or redundant information required
15 Unnecessary repetition of steps
16 Unnecessary limits

Command Structure and Rigidity

Inconsistencies
Sl No Possible Error Conditions
1 Optimizations
2 Inconsistent syntax
3 Inconsistent command entry style
4 Inconsistent abbreviations
5 Inconsistent termination rule
6 Inconsistent command options
7 Similarly named commands
8 Inconsistent Capitalization
9 Inconsistent menu position
10 Inconsistent function key usage
11 Inconsistent error handling rules
12 Inconsistent editing rules
13 Inconsistent data saving rules
Time Wasters
14 Garden paths
15 choice can't be taken
16 Are you really, really sure
17 Obscurely or idiosyncratically named commands
Menus
18 Excessively complex menu hierarchy
19 Inadequate menu navigation options
20 Too many paths to the same place
21 You can't get there from here
22 Related commands relegated to unrelated menus
23 Unrelated commands tossed under the same menu
Command Lines
24 Forced distinction between uppercase and lowercase
25 Reversed parameters
26 Full command names are not allowed
27 Abbreviations are not allowed
28 Demands complex input on one line
29 no batch input
30 can't edit commands
Inappropriate use of key board
31 Failure to use cursor, edit, or function keys
32 Non std use of cursor and edit keys
33 non-standard use of function keys
34 Failure to filter invalid keys
35 Failure to indicate key board state changes

Missing Commands
Testing 26
State transitions
Sl No Possible Error Conditions
1 Can't do nothing and leave
2 Can't quit mid-program
3 Can't stop mid-command
4 Can't pause
Disaster prevention
5 No backup facility
6 No undo
7 No are you sure
8 No incremental saves
Disaster prevention
9 Inconsistent menu position
10 Inconsistent function key usage
11 Inconsistent error handling rules
12 Inconsistent editing rules
13 Inconsistent data saving rules
Error handling by the user
14 No user specifiable filters
15 Awkward error correction
16 Can't include comments
17 Can't display relationships between variables
Miscellaneous
18 Inadequate privacy or security
19 Obsession with security
20 Can't hide menus
21 Doesn't support standard OS features
22 Doesn't allow long names

Error Handling

Error prevention
Sl No Possible Error Conditions
1 Inadequate initial state validation
2 Inadequate tests of user input
3 Inadequate protection against corrupted data
4 Inadequate tests of passed parameters
5 Inadequate protection against operating system bugs
6 Inadequate protection against malicious use
7 Inadequate version control

Error Detection
Sl No Possible Error Conditions
1 ignores overflow
2 ignores impossible values
3 ignores implausible values
4 ignores error flag
5 ignores hardware fault or error conditions
6 data comparison
Testing 27
 Error Recovery
Sl No Possible Error Conditions
1 automatic error detection
2 failure to report error
3 failure to set an error flag
4 where does the program go back to
5 aborting errors
6 recovery from hardware problems
7 no escape from missing disks

Boundary related errors

Sl No Possible Error Conditions

1 Numeric boundaries
2 Equality as boundary
3 Boundaries on numerosity
4 Boundaries in space
5 Boundaries in time
6 Boundaries in loop
7 Boundaries in memory
8 Boundaries with data structure
9 Hardware related boundaries
10 Invisible boundaries
11 Mishandling of boundary case
12 Wrong boundary
13 Mishandling of cases outside boundary

Calculation Errors

Sl No Possible Error Conditions

1 Bad Logic
2 Bad Arithmetic
3 Imprecise Calculations
4 Outdated constants
5 Calculation errors
6 Impossible parenthesis
7 Wrong order of calculations
8 Bad underlying functions
9 Overflow and Underflow
10 Truncation and Round-off error
11 Confusion about the representation of data
12 Incorrect conversion from one data representation to another
13 Wrong Formula
14 Incorrect Approximation

Race Conditions
Testing 28
Sl No Possible Error Conditions
1 Races in updating data
2 Assumption that one event or task finished before another begins
3 Assumptions that one event or task has finished before another
begins
4 Assumptions that input won't occur during a brief processing interval
5 Assumptions that interrupts won't occur during brief interval
6 Resource races
7 Assumptions that a person, device or process will respond quickly
8 Options out of sync during display changes
9 Tasks starts before its prerequisites are met
10 Messages cross or don't arrive in the order sent

Initial and Later States

Sl No Possible Error Conditions

1 Failure to set data item to zero
2 Failure to initialize a loop-control variable
3 Failure to initialize a or re-initialize a pointer
4 Failure to clear a string
5 Failure to initialize a register
6 Failure to clear a flag
7 Data were supposed to be initialized elsewhere
8 Failure to re-initialize
9 Assumption that data were not re-initialized
10 Confusion between static and dynamic storage
11 Data modifications by side effect
12 Incorrect initialization

Control Flow Errors

Program runs amok

Sl No Possible Error Conditions
1 Jumping to a routine that isn't resident
2 Re-entrance
3 Variables contains embedded command names
4 Wrong returning state assumed
5 Exception handling based exits

Return to wrong place

Sl No Possible Error Conditions
1 Corrupted Stack
2 Stack underflow/overflow
3 GOTO rather than RETURN from sub-routine
Interrupts
Sl No Possible Error Conditions
1 Wrong interrupt vector
Testing 29
2 Failure to restore or update interrupt vector
3 Invalid restart after an interrupt
4 Failure to block or un-block interrupts

Program Stops
Sl No Possible Error Conditions
1 Dead crash
2 Syntax error reported at run time
3 Waiting for impossible condition or combinations of conditions
4 Wrong user or process priority

Error Detection
Sl No Possible Error Conditions
1 infinite loop
2 Wrong starting value for the loop control variables
3 Accidental change of loop control variables
4 Command that do or don't belong inside the loop
5 Command that do or don't belong inside the loop
6 Improper loop nesting

If Then Else , Or may not

Sl No Possible Error Conditions
1 Wrong inequalities
2 Comparison sometimes yields wrong result
3 Not equal verses equal when there are three cases
4 Testing floating point values for equality
5 confusion between inclusive and exclusive OR
6 Incorrectly negating a logical expression
7 Assignment equal instead of test equal
8 Commands being inside the THEN or ELSE clause
9 Commands that don't belong either case
10 Failure to test a flag
11 Failure to clear a flag

Multiple Cases
Sl No Possible Error Conditions
1 Missing default
2 Wrong default
3 Missing cases
4 Overlapping cases
5 Invalid or impossible cases
6 Commands being inside the THEN or ELSE clause
7 Case should be sub-divided

Testing 30
Errors Handling or Interpreting Data

Problems in passing data between routines

Sl No Possible Error Conditions
1 Parameter list variables out of order or missing
2 Data Type errors
3 Aliases and shifting interpretations of the same area of memory
4 Misunderstood data values
5 inadequate error information
6 Failure to clean up data on exception handling
7 Outdated copies of data
8 Related variable get out of synch
9 Local setting of global data
10 Global use of local variables
11 Wrong mask in bit fields
12 Wrong value from table

Data boundaries
Sl No Possible Error Conditions
1 Un-terminated null strings
2 Early end of string
3 Read/Write past end of data structure or an element in it

Read outside the limits of message buffer

Sl No Possible Error Conditions
1 Complier padding to word boundaries
2 value stack underflow/overflow
3 Trampling another process's code or data

Messaging Problems
Sl No Possible Error Conditions
1 Messages sent to wrong process or port
2 Failure to validate an incoming message
3 Lost or out of synch messages
4 Message sent to only N of N+1 processes

Data Storage corruption

Sl No Possible Error Conditions
1 Overwritten changes
2 Data entry not saved
3 Too much data for receiving process to handle
4 Overwriting a file after an error exit or user abort

Load Conditions

Sl No Possible Error Conditions

Testing 31
1 Required resources are not available
2 No available large memory area
3 Input buffer or queue not deep enough
4 Doesn't clear item from queue, buffer or stock
5 Lost Messages
6 Performance costs
7 Race condition windows expand
8 Doesn't abbreviate under load
9 Doesn't recognize that another process abbreviates output under load
10 Low priority tasks not put off
11 Low priority tasks never done

Doesn't return a resource

Sl No Possible Error Conditions
1 Doesn't indicate that it's done with a device
2 Doesn't erase old files from mass storage
3 Doesn't return unused memory
4 Wastes computer time
Hardware
Sl No Possible Error Conditions
1 Wrong Device
2 Wrong Device Address
3 Device unavailable
4 Device returned to wrong type of pool
5 Device use forbidden to caller
6 Specifies wrong privilege level for the device
7 Noisy Channel
8 Channel goes down
9 Time-out problems
10 Wrong storage device
11 Doesn't check the directory of current disk
12 Doesn't close the file
13 Unexpected end of file
14 Disk sector bug and other length dependent errors
15 Wrong operation or instruction codes
16 Misunderstood status or return code
17 Underutilizing device intelligence
18 Paging mechanism ignored or misunderstood
19 Ignores channel throughput limits
20 Assuming device is or isn't or should be or shouldn't be initialized
21 Assumes programmable function keys are programmed correctly
Source, Version, ID Control
Sl No Possible Error Conditions
1 Old bugs mysteriously re appear
2 Failure to update multiple copies of data or program files
3 No title
4 No version ID

Testing 32
5 Wrong version number of title screen
6 No copy right message or bad one
7 Archived source doesn't compile into a match for shipping code
8 Manufactured disks don't work or contain wrong code or data
Testing Errors
Missing bugs in the program
Sl No Possible Error Conditions
1 Failure to notice a problem
2 You don't know what the correct test results are
3 You are bored or inattentive
4 Misreading the Screen
5 Failure to report problem
6 Failure to execute a planned test
7 Failure to use the most promising test case
8 Ignoring programmer's suggestions

Finding bugs that aren't in the program

Sl No Possible Error Conditions
1 Errors in testing programs
2 Corrupted data files
3 Misinterpreted specifications or documentation
Poor reporting
Sl No Possible Error Conditions
1 Illegible reports
2 Failure to make it clear how to reproduce the problem
3 Failure to say you can't reproduce the problem
4 Failure to check your report
5 Failure to report timing dependencies
6 Failure to simplify conditions
7 Concentration on trivia
8 Abusive language
Poor Tracking and follow-up
Sl No Possible Error Conditions
1 Failure to provide summary report
2 Failure to re-report serious bug
3 Failure to check for unresolved problems just before release
4 Failure to verify fixes
47. Designing Unit Test Cases

Executive Summary

Producing a test specification, including the design of test cases, is the level of test design which
has the highest degree of creative input. Furthermore, unit test specifications will usually be
produced by a large number of staff with a wide range of experience, not just a few experts.

This paper provides a general process for developing unit test specifications and then describes
some specific design techniques for designing unit test cases. It serves as a tutorial for

Testing 33
developers who are new to formal testing of software, and as a reminder of some finer points for
experienced software testers.

A. Introduction

The design of tests is subject to the same basic engineering principles as the design of software.
Good design consists of a number of stages which progressively elaborate the design. Good
test design consists of a number of stages which progressively elaborate the design of tests:

 Test strategy;
 Test planning;
 Test specification;
 Test procedure.

These four stages of test design apply to all levels of testing, from unit testing through to system
testing. This paper concentrates on the specification of unit tests; i.e. the design of individual unit
test cases within unit test specifications. A more detailed description of the four stages of test
design can be found in the IPL paper "An Introduction to Software Testing".

The design of tests has to be driven by the specification of the software. For unit testing, tests
are designed to verify that an individual unit implements all design decisions made in the unit's
design specification. A thorough unit test specification should include positive testing, that the
unit does what it is supposed to do, and also negative testing, that the unit does not do anything
that it is not supposed to do.

Producing a test specification, including the design of test cases, is the level of test design which
has the highest degree of creative input. Furthermore, unit test specifications will usually be
produced by a large number of staff with a wide range of experience, not just a few experts.

This paper provides a general process for developing unit test specifications, and then describes
some specific design techniques for designing unit test cases. It serves as a tutorial for
developers who are new to formal testing of software, and as a reminder of some finer points for
experienced software testers.

B. Developing Unit Test Specifications

Once a unit has been designed, the next development step is to design the unit tests. An
important point here is that it is more rigorous to design the tests before the code is written. If the
code was written first, it would be too tempting to test the software against what it is observed to
do (which is not really testing at all), rather than against what it is specified to do.

A unit test specification comprises a sequence of unit test cases. Each unit test case should
include four essential elements:

 A statement of the initial state of the unit, the starting point of the test case (this is only
applicable where a unit maintains state between calls);
 The inputs to the unit, including the value of any external data read by the unit;
 What the test case actually tests, in terms of the functionality of the unit and the analysis
used in the design of the test case (for example, which decisions within the unit are
tested);
Testing 34
  The expected outcome of the test case (the expected outcome of a test case should
always be defined in the test specification, prior to test execution).

The following subsections of this paper provide a six step general process for developing a unit
test specification as a set of individual unit test cases. For each step of the process, suitable test
case design techniques are suggested. (Note that these are only suggestions. Individual
circumstances may be better served by other test case design techniques). Section 3 of this
paper then describes in detail a selection of techniques which can be used within this process to
help design test cases.

B.1 Step 1 - Make it Run

The purpose of the first test case in any unit test specification should be to execute the unit
under test in the simplest way possible. When the tests are actually executed, knowing that at
least the first unit test will execute is a good confidence boost. If it will not execute, then it is
preferable to have something as simple as possible as a starting point for debugging.

Suitable techniques:

- Specification derived tests

- Equivalence partitioning

B.2 Step 2 - Positive Testing

Test cases should be designed to show that the unit under test does what it is supposed to do.
The test designer should walk through the relevant specifications; each test case should test
one or more statements of specification. Where more than one specification is involved, it is best
to make the sequence of test cases correspond to the sequence of statements in the primary
specification for the unit.

Suitable techniques:

- Specification derived tests

- Equivalence partitioning
- State-transition testing

B.3. Step 3 - Negative Testing

Existing test cases should be enhanced and further test cases should be designed to show that
the software does not do anything that it is not specified to do. This step depends primarily upon
error guessing, relying upon the experience of the test designer to anticipate problem areas.

Suitable techniques:

- Error guessing
- Boundary value analysis
- Internal boundary value testing
- State-transition testing

B.4. Step 4 - Special Considerations

Testing 35
Where appropriate, test cases should be designed to address issues such as performance,
safety requirements and security requirements. Particularly in the cases of safety and security, it
can be convenient to give test cases special emphasis to facilitate security analysis or safety
analysis and certification. Test cases already designed which address security issues or safety
hazards should be identified in the unit test specification. Further test cases should then be
added to the unit test specification to ensure that all security issues and safety hazards
applicable to the unit will be fully addressed.

Suitable techniques:

- Specification derived tests

B.5. Step 5 - Coverage Tests

The test coverage likely to be achieved by the designed test cases should be visualised. Further
test cases can then be added to the unit test specification to achieve specific test coverage
objectives. Once coverage tests have been designed, the test procedure can be developed and
the tests executed.

Suitable techniques:

- Branch testing
- Condition testing
- Data definition-use testing
- State-transition testing

B.6. Test Execution

A test specification designed using the above five steps should in most cases provide a thorough
test for a unit. At this point the test specification can be used to develop an actual test
procedure, and the test procedure used to execute the tests. For users of AdaTEST or Cantata,
the test procedure will be an AdaTEST or Cantata test script.

Execution of the test procedure will identify errors in the unit which can be corrected and the unit
re-tested. Dynamic analysis during execution of the test procedure will yield a measure of test
coverage, indicating whether coverage objectives have been achieved. There is therefore a
further coverage completion step in the process of designing test specifications.

B.7. Step 6 - Coverage Completion

Depending upon an organization’s standards for the specification of a unit, there may be no
structural specification of processing within a unit other than the code itself. There are also likely
to have been human errors made in the development of a test specification. Consequently, there
may be complex decision conditions, loops and branches within the code for which coverage
targets may not have been met when tests were executed. Where coverage objectives are not
achieved, analysis must be conducted to determine why. Failure to achieve a coverage objective
may be due to:

Testing 36
  Infeasible paths or conditions - the corrective action should be to annotate the test
specification to provide a detailed justification of why the path or condition is not tested.
AdaTEST provides some facilities to help exclude infeasible conditions from Boolean
coverage metrics.
 Unreachable or redundant code - the corrective action will probably be to delete the
offending code. It is easy to make mistakes in this analysis, particularly where defensive
programming techniques have been used. If there is any doubt, defensive programming
should not be deleted.
 Insufficient test cases - test cases should be refined and further test cases added to a test
specification to fill the gaps in test coverage.

Ideally, the coverage completion step should be conducted without looking at the actual
code. However, in practice some sight of the code may be necessary in order to achieve
coverage targets. It is vital that all test designers should recognize that use of the coverage
completion step should be minimized. The most effective testing will come from analysis and
specification, not from experimentation and over dependence upon the coverage completion
step to cover for sloppy test design.

Suitable techniques:

- Branch testing
- Condition testing
- Data definition-use testing
- State-transition testing

B.8. General Guidance

Note that the first five steps in producing a test specification can be achieved:

 Solely from design documentation;

 Without looking at the actual code;
 Prior to developing the actual test procedure.

It is usually a good idea to avoid long sequences of test cases which depend upon the outcome
of preceding test cases. An error identified by a test case early in the sequence could cause
secondary errors and reduce the amount of real testing achieved when the tests are executed.

The process of designing test cases, including executing them as "thought experiments", often
identifies bugs before the software has even been built. It is not uncommon to find more bugs
when designing tests than when executing tests.

Throughout unit test design, the primary input should be the specification documents for the unit
under test. While use of actual code as an input to the test design process may be necessary in
some circumstances, test designers must take care that they are not testing the code against
itself. A test specification developed from the code will only prove that the code does what the
code does, not that it does what it is supposed to do.

C. Test Case Design Techniques

Testing 37
The preceding section of this paper has provided a "recipe" for developing a unit test
specification as a set of individual test cases. In this section a range of techniques which can be
to help define test cases are described.

Test case design techniques can be broadly split into two main categories. Black box techniques
use the interface to a unit and a description of functionality, but do not need to know how the
inside of a unit is built. White box techniques make use of information about how the inside of a
unit works. There are also some other techniques which do not fit into either of the above
categories. Error guessing falls into this category.

The most important ingredients of any test design are experience and common sense. Test
designers should not let any of the given techniques obstruct the application of experience and
common sense.

The selection of test case design techniques described in the following subsections is by no
means exhaustive. Further information on techniques for test case design can be found in
"Software Testing Techniques" 2nd Edition, B Beizer,Van Nostrand Reinhold, New York 1990.

C.1. Specification Derived Tests

As the name suggests, test cases are designed by walking through the relevant specifications.
Each test case should test one or more statements of specification. It is often practical to make
the sequence of test cases correspond to the sequence of statements in the specification for the
unit under test. For example, consider the specification for a function to calculate the square root
of a real number, shown in figure 3.1.

Testing 38
There are three statements in this specification, which can be addressed by two test cases. Note
that the use of Print_Line conveys structural information in the specification.

Test Case 1: Input 4, Return 2

- Exercises the first statement in the specification

("When given an input of 0 or greater, the positive square
root of the input shall be returned.").

Test Case 2: Input -10, Return 0, Output "Square root error - illegal negative input" using
Print_Line.

- Exercises the second and third statements in the specification

("When given an input of less than 0, the error message

"Square root error - illegal negative input" shall be displayed
and a value of 0 returned. The library routine Print_Line shall
be used to display the error message.").

Specification derived test cases can provide an excellent correspondence to the sequence of
statements in the specification for the unit under test, enhancing the readability and
maintainability of the test specification. However, specification derived testing is a positive test
case design technique. Consequently, specification derived test cases have to be supplemented
by negative test cases in order to provide a thorough unit test specification.

A variation of specification derived testing is to apply a similar technique to a security analysis,

safety analysis, software hazard analysis, or other document which provides supplementary
information to the unit's specification.

C.2. Equivalence Partitioning

Testing 39
Equivalence partitioning is a much more formalised method of test case design. It is based upon
splitting the inputs and outputs of the software under test into a number of partitions, where the
behaviour of the software is equivalent for any value within a particular partition. Data which
forms partitions is not just routine parameters. Partitions can also be present in data accessed
by the software, in time, in input and output sequence, and in state.

Equivalence partitioning assumes that all values within any individual partition are equivalent for
test purposes. Test cases should therefore be designed to test one value in each partition.
Consider again the square root function used in the previous example. The square root function
has two input partitions and two output partitions, as shown in table 3.2.

These four partitions can be tested with two test cases:

Test Case 1: Input 4, Return 2

- Exercises the >=0 input partition (ii)
- Exercises the >=0 output partition (a)

Test Case 2: Input -10, Return 0, Output "Square root error - illegal negative input" using
Print_Line.

- Exercises the <0 input partition (i)

- Exercises the "error" output partition (b)

For a function like square root, we can see that equivalence partitioning is quite simple. One test
case for a positive number and a real result; and a second test case for a negative number and
an error result. However, as software becomes more complex, the identification of partitions and
the inter-dependencies between partitions becomes much more difficult, making it less
convenient to use this technique to design test cases. Equivalence partitioning is still basically a
positive test case design technique and needs to be supplemented by negative tests.

C.3. Boundary Value Analysis

Boundary value analysis uses the same analysis of partitions as equivalence partitioning.
However, boundary value analysis assumes that errors are most likely to exist at the boundaries
between partitions. Boundary value analysis consequently incorporates a degree of negative
testing into the test design, by anticipating that errors will occur at or near the partition
boundaries. Test cases are designed to exercise the software on and at either side of boundary
values. Consider the two input partitions in the square root example, as illustrated by figure 3.2.

Testing 40
The zero or greater partition has a boundary at 0 and a boundary at the most positive real
number. The less than zero partition shares the boundary at 0 and has another boundary at the
most negative real number. The output has a boundary at 0, below which it cannot go.

Test Case 1: Input {the most negative real number}, Return 0, Output "Square root error - illegal
negative input" using Print_Line

-Exercises the lower boundary of partition (i).

Test Case 2: Input {just less than 0}, Return 0, Output "Square root error - illegal
negative input" using Print_Line

- Exercises the upper boundary of partition (i).

Test Case 3: Input 0, Return 0

- Exercises just outside the upper boundary of partition (i),

the lower boundary of partition (ii) and the lower boundary
of partition (a).

Test Case 4: Input {just greater than 0}, Return {the positive square root of the input}

- Exercises just inside the lower boundary of partition (ii).

Test Case 5: Input {the most positive real number}, Return {the positive square root of the input}

- Exercises the upper boundary of partition (ii) and the upper boundary of
partition (a).

As for equivalence partitioning, it can become impractical to use boundary value analysis
thoroughly for more complex software. Boundary value analysis can also be meaningless for
non scalar data, such as enumeration values. In the example, partition (b) does not really have
boundaries. For purists, boundary value analysis requires knowledge of the underlying
representation of the numbers. A more pragmatic approach is to use any small values above
and below each boundary and suitably big positive and negative numbers

C.4. State-Transition Testing

Testing 41
State transition testing is particularly useful where either the software has been designed as a
state machine or the software implements a requirement that has been modelled as a state
machine. Test cases are designed to test the transitions between states by creating the events
which lead to transitions.

When used with illegal combinations of states and events, test cases for negative testing can be
designed using this approach. Testing state machines is addressed in detail by the IPL paper
"Testing State Machines with AdaTEST and Cantata".
C.5. Branch Testing
In branch testing, test cases are designed to exercise control flow branches or decision points in
a unit. This is usually aimed at achieving a target level of Decision Coverage. Given a functional
specification for a unit, a "black box" form of branch testing is to "guess" where branches may be
coded and to design test cases to follow the branches. However, branch testing is really a "white
box" or structural test case design technique. Given a structural specification for a unit,
specifying the control flow within the unit, test cases can be designed to exercise branches.
Such a structural unit specification will typically include a flowchart or PDL.

Returning to the square root example, a test designer could assume that there would be a
branch between the processing of valid and invalid inputs, leading to the following test cases:

Test Case 1: Input 4, Return 2

- Exercises the valid input processing branch

Test Case 2: Input -10, Return 0, Output "Square root error - illegal negative input" using
Print_Line.

- Exercises the invalid input processing branch

However, there could be many different structural implementations of the square root function.
The following structural specifications are all valid implementations of the square root function,
but the above test cases would only achieve decision coverage of the first and third versions of
the specification.

Testing 42
Testing 43
It can be seen that branch testing works best with a structural specification for the unit. A
structural unit specification will enable branch test cases to be designed to achieve decision
coverage, but a purely functional unit specification could lead to coverage gaps.

One thing to beware of is that by concentrating upon branches, a test designer could loose sight
of the overall functionality of a unit. It is important to always remember that it is the overall
functionality of a unit that is important, and that branch testing is a means to an end, not an end
in itself. Another consideration is that branch testing is based solely on the outcome of decisions.
It makes no allowances for the complexity of the logic which leads to a decision.

C.6. Condition Testing

There are a range of test case design techniques which fall under the general title of condition
testing, all of which try to allay the weaknesses of branch testing when complex logical
conditions are encountered. The object of condition testing is to design test cases to show that
the individual components of logical conditions and combinations of the individual components
are correct.

Test cases are designed to test the individual elements of logical expressions, both within
branch conditions and within other expressions in a unit. As for branch testing, condition testing
could be used as a "black box" technique, where the test designer makes intelligent guesses
about the implementation of a functional specification for a unit. However, condition testing is
more suited to "white box" test design from a structural specification for a unit.

Testing 44
The test cases should be targeted at achieving a condition coverage metric, such as Modified
Condition Decision Coverage (available as Boolean Operand Effectiveness in AdaTEST). The
IPL paper entitled "Structural Coverage Metrics" provides more detail of condition coverage
metrics.

To illustrate condition testing, consider the example specification for the square root function
which uses successive approximation (figure 3.3(d) - Specification 4). Suppose that the designer
for the unit made a decision to limit the algorithm to a maximum of 10 iterations, on the grounds
that after 10 iterations the answer would be as close as it would ever get. The PDL specification
for the unit could specify an exit condition like that given in figure 3.4.

If the coverage objective is Modified Condition Decision Coverage, test cases have to prove that
both error<desired accuracy and iterations=10 can independently affect the outcome of the
decision.

Test Case 1: 10 iterations, error>desired accuracy for all iterations.

- Both parts of the condition are false for the first 9

iterations. On the tenth iteration, the first part of the
condition is false and the second part becomes true,
showing that the iterations=10 part of the condition can
independently affect its outcome.

Test Case 2: 2 iterations, error>=desired accuracy for the first iteration, and
error<desired accuracy for the second iteration.

- Both parts of the condition are false for the first iteration.
On the second iteration, the first part of the condition
becomes true and the second part remains false, showing
that the error<desired accuracy part of the condition can
independently affect its outcome.

Condition testing works best when a structural specification for the unit is available. It provides a
thorough test of complex conditions, an area of frequent programming and design error and an
area which is not addressed by branch testing. As for branch testing, it is important for test

Testing 45
designers to beware that concentrating on conditions could distract a test designer from the
overall functionality of a unit.

C.7. Data Definition-Use Testing

Data definition-use testing designs test cases to test pairs of data definitions and uses. A data
definition is anywhere that the value of a data item is set, and a data use is anywhere that a data
item is read or used. The objective is to create test cases which will drive execution through
paths between specific definitions and uses.

Like decision testing and condition testing, data definition-use testing can be used in
combination with a functional specification for a unit, but is better suited to use with a structural
specification for a unit.

Consider one of the earlier PDL specifications for the square root function which sent every input
to the maths co-processor and used the co-processor status to determine the validity of the
result. (Figure 3.3(c) - Specification 3). The first step is to list the pairs of definitions and uses. In
this specification there are a number of definition-use pairs, as shown in table 3.3.

These pairs of definitions and uses can then be used to design test cases. Two test cases are
required to test all six of these definition-use pairs:

Test Case 1: Input 4, Return 2

- Tests definition-use pairs 1, 2, 5, 6
-
Test Case 2: Input -10, Return 0, Output "Square root error - illegal negative input" using
Print_Line.

- Tests definition-use pairs 1, 2, 3, 4

The analysis needed to develop test cases using this design technique can also be useful for
identifying problems before the tests are even executed; for example, identification of situations
where data is used without having been defined. This is the sort of data flow analysis that some
static analysis tool can help with. The analysis of data definition-use pairs can become very

Testing 46
complex, even for relatively simple units. Consider what the definition-use pairs would be for the
successive approximation version of square root!

It is possible to split data definition-use tests into two categories: uses which affect control flow
(predicate uses) and uses which are purely computational. Refer to "Software Testing
Techniques" 2nd Edition, B Beizer,Van Nostrand Reinhold, New York 1990, for a more detailed
description of predicate and computational uses.

C.8. Internal Boundary Value Testing

In many cases, partitions and their boundaries can be identified from a functional
specification for a unit, as described under equivalence partitioning and boundary value analysis
above. However, a unit may also have internal boundary values which can only be identified
from a structural specification. Consider a fragment of the successive approximation version of
the square root unit specification, as shown in figure 3.5 ( derived from figure 3.3(d) -
Specification 4).

The calculated error can be in one of two partitions about the desired accuracy, a feature of the
structural design for the unit which is not apparent from a purely functional specification. An
analysis of internal boundary values yields three conditions for which test cases need to be
designed.

Test Case 1: Error just greater than the desired accuracy

Test Case 2: Error equal to the desired accuracy
Test Case 3: Error just less than the desired accuracy

Internal boundary value testing can help to bring out some elusive bugs. For example, suppose
"<=" had been coded instead of the specified "<". Nevertheless, internal boundary value testing
is a luxury to be applied only as a final supplement to other test case design techniques.

C.9. Error Guessing

Testing 47
Error guessing is based mostly upon experience, with some assistance from other techniques
such as boundary value analysis. Based on experience, the test designer guesses the types of
errors that could occur in a particular type of software and designs test cases to uncover them.
For example, if any type of resource is allocated dynamically, a good place to look for errors is in
the deallocation of resources. Are all resources correctly deallocated, or are some lost as the
software executes?

Error guessing by an experienced engineer is probably the single most effective method of
designing tests which uncover bugs. A well placed error guess can show a bug which could
easily be missed by many of the other test case design techniques presented in this paper.
Conversely, in the wrong hands error guessing can be a waste of time.

To make the maximum use of available experience and to add some structure to this test case
design technique, it is a good idea to build a check list of types of errors. This check list can then
be used to help "guess" where errors may occur within a unit. The check list should be
maintained with the benefit of experience gained in earlier unit tests, helping to improve the
overall effectiveness of error guessing.

D. Conclusion

Experience has shown that a conscientious approach to unit testing will detect many bugs at a
stage of the software development where they can be corrected economically. A rigorous
approach to unit testing requires:

 That the design of units is documented in a specification before coding

begins;
 That unit tests are designed from the specification for the unit, also
preferably before coding begins;
 That the expected outcomes of unit test cases are specified in the unit test
specification.

The process for developing unit test specifications presented in this paper is generic, in that it
can be applied to any level of testing. Nevertheless, there will be circumstances where it has to
be tailored to specific situations. Tailoring of the process and the use of test case design
techniques should be documented in the overall test strategy.

48. LITERATURE REVIEW

2.1 Introduction

The purpose of this dissertation is to increase understanding of how experienced practitioners

as individuals evaluate diagrammatic models in Formal Technical Review (FTR). In this
research, those aspects of FTR relating to evaluation of an artifact by practitioners as individuals
are referred to as Practitioner Evaluation (PE). The relevant FTR literature is reviewed for theory
and research applicable to PE. However, FTR developed pragmatically without relation to
underlying cognitive theory, and the literature consists primarily of case studies with a very
limited number of controlled experiments.

Other work on the evaluation of diagrams and graphs is also reviewed for possible theoretical
models that could be used in the current research. Human-Computer Interaction (HCI) is an

Testing 48
Information Systems area that has drawn extensively on cognitive science to develop and
evaluate Graphical User Interfaces (GUIs). A brief overview of cognitive-based approaches
utilized in HCI is presented. One of these approaches, the Human Information Processing
System model, in which the human mind is treated as an information-processing system,
provides the cognitive theoretical model for this research and is discussed separately because
of its importance. Work on attention and the comprehension of graphics is also briefly reviewed.

Two further areas are identified as necessary for the development of the research task and
tools: (1) types of diagrammatic models and (2) types of software defects. Relevant work in each
of these areas is briefly reviewed and, since typologies appropriate to this research were not
located, appropriate typologies are developed.

2.2 Formal Technical Review

Software review as a technique to detect software defects is not new -- it has been used since
the earliest days of programming. For example, Babbage and von Neumann regularly asked
colleagues to examine their programs [Freedman and Weinberg 1990], and in the 1950s and
1960s, large software projects often included some type of software review [Knight and Myers
1993]. However, the first significant formalization of software review practice is generally
considered to be the development by Michael Fagan [1976] of a species of FTR that he called
"inspection."

Following Tjahjono [1996, 2], Formal Technical Review may be defined as any "evaluation
technique that involves the bringing together of a group of technical [and sometimes non-
technical] personnel to analyze a software artifact, typically with the goal of discovering errors or
other anomalies." As such, FTR has the following distinguishing characteristics:

1. Formal process.
2. Use of groups or teams. Most FTR techniques involve real groups, but nominal groups are
used as well.
3. Review by knowledgeable individuals or practitioners.
4. Focus on detection of defects.

2.2.1 Types of Formal Technical Review

While the focus of this research is on the individual evaluation aspects of reviews, for context
several other FTR techniques are discussed as well. Among the most common forms of FTR are
the following:

1.Desk Checking, or reading over a program by hand while sitting at one's desk, is the oldest
software review technique [Adrion et al. 1982]. Strictly speaking, desk checking is not a form of
FTR since it does not involve a formal process or a group. Moreover, desk checking is generally
perceived as ineffective and unproductive due to (a) its lack of discipline and (b) the general
ineffectiveness of people in detecting their own errors. To correct for the second problem,
programmers often swap programs and check each other's work. Since desk checking is an
individual process not involving group dynamics, research in this area would be relevant but
none applicable to the current research was found.

It should be noted that Humphrey [1995] has developed a review method, called Personal
Review (PR), which is similar to desk checking. In PR, each programmer examines his own

Testing 49
products to find as many defects as possible utilizing a disciplined process in conjunction with
Humphrey's Personal Software Process (PSP) to improve his own work. The review strategy
includes the use of checklists to guide the review process, review metrics to improve the
process, and defect causal analysis to prevent the same defects from recurring in the future. The
approach taken in developing the Personal Review process is an engineering one; no reference
is made in Humphrey [1995] to cognitive theory.

2. Peer Rating is a technique in which anonymous programs are evaluated in terms of their
overall quality, maintainability, extensibility, usability and clarity by selected programmers who
have similar backgrounds [Myers 1979]. Shneiderman [1980] suggests that peer ratings of
programs are productive, enjoyable, and non-threatening experiences. The technique is often
referred to as Peer Reviews [Shneiderman 1980], but some authors use the term peer
reviews for generic review methods involving peers [Paulk et al 1993; Humphrey 1989].

3. Walkthroughs are presentation reviews in which a review participant, usually the software
author, narrates a description of the software and the other members of the review group
provide feedback throughout the presentation [Freedman and Weinberg 1990; Gilb and
Graham 1993]. It should be noted that the term "walkthrough" has been used in the literature
variously. Some authors unite it with "structured" and treat it as a disciplined, formal review
process [Myers 1979; Yourdon 1989; Adrion et al. 1982]. However, the literature generally
describes walkthrough as an undisciplined process without advance preparation on the part
of reviewers and with the meeting focus on education of participants [Fagan 1976].

4. Round-robin Review is a evaluation process in which a copy of the review materials is

made available and routed to each participant; the reviewers then write their
comments/questions concerning the materials and pass the materials with comments to
another reviewer and to the moderator or author eventually [Hart 1982].

5. Inspection was developed by Fagan [1976, 1986] as a well-planned and well-defined group
review process to detect software defects – defect repair occurs outside the scope of the
process. The original Fagan Inspection (FI) is the most cited review method in the literature
and is the source for a variety of similar inspection techniques [Tjahjono 1996]. Among the
FI-derived techniques are Active Design Review [Parnas and Weiss 1987], Phased
Inspection [Knight and Myers 1993], N-Fold Inspection [Schneider et al. 1992], and FTArm
[Tjahjono 1996]. Unlike the review techniques previously discussed, inspection is often used
to control the quality and productivity of the development process.

A Fagan Inspection consists of six well-defined phases:

i. Planning. Participants are selected and the materials to be reviewed are prepared and
checked for review suitability.
ii. Overview. The author educates the participants about the review materials through a
presentation.
iii. Preparation. The participants learn the materials individually.
iv. Meeting. The reader (a participant other than the author) narrates or paraphrases the
review materials statement by statement, and the other participants raise issues and
questions. Questions continue on a point only until an error is recognized or the item is
deemed correct.
v. Rework. The author fixes the defects identified in the meeting.
vi. Follow-up. The "corrected" products are reinspected.

Testing 50
Practitioner Evaluation is primarily associated with the Preparation phase.

In addition to classification by technique-type, FTR may also be classified on other

dimensions, including the following:

A. Small vs. Large Team Reviews. Siy [1996] classifies reviews into those conducted by
small (1-4 reviewers) [Bisant and Lyle 1996] and large (more than 4 reviewers) [Fagan
1976, 1986] teams. If each reviewer depends on different expertise and experiences, a
large team should allow a wider variety of defects to be detected and thus better
coverage. However, a large team requires more effort due to more individuals inspecting
the artifact, generally involves greater scheduling problems [Ballman and Votta 1994],
and may make it more difficult for all participants to participate fully.

B. No vs. Single vs. Multiple Session Reviews. The traditional Fagan Inspection provided
for one session to inspect the software artifact, with the possibility of a follow-up session
to inspect corrections. However, variants have been suggested.

Humphrey [1989] comments that three-quarters of the errors found in well-run inspections
are found during preparation. Based on an economic analysis of a series of inspections at
AT&T, Votta [1993] argues that inspection meetings are generally not economic and
should be replaced with depositions, where the author and (optionally) the moderator
meet separately with inspectors to collect their results.

On the other hand, some authors [Knight and Myers 1993; Schneider et al. 1992] have
argued for multiple sessions, conducted either in series or parallel. Gilb and Graham
[1993] do not use multiple inspection sessions but add a root cause analysis session
immediately after the inspection meeting.
C. Nonsystematic vs. Systematic Defect-Detection Technique Reviews. The most
frequently used detection methods (ad hoc and checklist) rely on nonsystematic techniques,
and reviewer responsibilities are general and not differentiated for single session reviews [Siy
1996]. However, some methods employ more prescriptive techniques, such as
questionnaires [Parnas and Weiss 1987] and correctness proofs [Britcher 1988].

D.Single Site vs. Multiple Site Reviews. The traditional FTR techniques have assumed that
the group-meeting component would occur face-to-face at a single site. However, with
improved telecommunications, and especially with computer support (see item F below), it
has become increasingly feasible to conduct even the group meeting from multiple sites.

E. Synchronous vs. Asynchronous Reviews. The traditional FTR techniques have also
assumed that the group meeting component would occur in real-time; i.e., synchronously.
However, some newer techniques that eliminate the group meeting or are based on
computer support utilize asynchronous reviews.
F. Manual vs. Computer-supported Reviews. In recent years, several computer supported
review systems have been developed [Brothers et al. 1990; Johnson and Tjahjono 1993;
Gintell et al. 1993; Mashayekhi et al 1994]. The type of support varies from simple
augmentation of the manual practices [Brothers et al. 1990; Gintell et al. 1993] to totally
new review methods [Johnson and Tjahjono 1993].

Testing 51
2.2.2 Economic Analyses of Formal Technical Review

Wheeler et al. [1996], after reviewing a number of studies that support the economic benefit of
FTR, conclude that inspections reduce the number of defects throughout development, cause
defects to be found earlier in the development process where they are less expensive to correct,
and uncover defects that would be difficult or impossible to discover by testing. They also note
"these benefits are not without their costs, however. Inspections require an investment of
approximately 15 percent of the total development cost early in the process [p. 11]."
In discussing overall economic effects, Wheeler et al. cite Fagan [1986] to the effect that
investment in inspections has been reported to yield a 25-to-35 percent overall increase in
productivity. They also reproduce a graphical analysis from Boehm [1987] that indicates
inspections reduce total development cost by approximately 30%.

The Wheeler et al. [1996] analysis does not specify the relative value of Practitioner Evaluation
to FTR, but two recent economic analyses provide indications.

•Votta [1993]. After analyzing data collected from 13 traditional inspections conducted at AT&T,
Votta reports that the approximately 4% increase in faults found at collection meetings (synergy)
does not economically justify the development delays caused by the need to schedule meetings
and the additional developer time associated with the actual meetings. He also argues that it is
not cost-effective to use the collection meeting to reduce the number of items incorrectly
identified as defective prior to the meeting ("false positives"). Based on these findings, he
concludes that almost all inspection meetings requiring all reviewers to be present should be
replaced with Depositions, which are three person meetings with only the author, moderator, and
one reviewer present.

• Siy [1996]. In his analysis of the factors driving inspection costs and benefits, Siy reports that
changes in FTR structural elements, such as group size, number of sessions, and
coordination of multiple sessions, were largely ineffective in improving the effectiveness of
inspections. Instead, inputs into the process (reviewers and code units) accounted for more
outcome variation than structural factors. He concludes by stating "better techniques by
which reviewers detect defects, not better process structures, are the key to improving
inspection effectiveness [Abstract, p. 2]." (emphasis added)

Votta's analysis effectively attributes most of the economic benefit of FTR to PE, and Siy's
explicitly states that better PE techniques "are the key to improving inspection effectiveness."
These findings, if supported by additional research, would further support the contention that a
better understanding of Practitioner Evaluation is necessary.
2.2.3 Psychological Aspects of FTR
Work on the psychological aspects of FTR can be categorized into four groups.
1.Egoless Programming. Gerald Weinberg [1971] began the examination of psychological
issues associated with software review in his work on egoless programming. According to
Weinberg, programmers are often reluctant to allow their programs to be read by other
programmers because the programs are often considered to be an extension of the self and
errors discovered in the programs to be a challenge to one's self-image. Two implications of
this theory are as follows:
i. The ability of a programmer to find errors in his own work tends to be impaired since he
tends to justify his own actions, and it is therefore more effective to have other people
check his work.
Testing 52
 ii. Each programmer should detach himself from his own work. The work should be
considered a public property where other people can freely criticize, and thus, improve its
quality; otherwise, one tends to become defensive, and reluctant to expose one's own
failures.

These two concepts have led to the justification of FTR groups, as well as the establishment
of independent quality assurance groups that specialize in finding software defects in many
software organizations [Humphrey 1989].

2. Role of Management. Another psychological aspect of FTR that has been examined is the
recording of data and its dissemination to management. According to Dobbins [1987], this
must be done in such a way that individual programmers will not feel intimidated or
threatened.

3. Positive Psychological Impacts. Hart [1982] observes that reviews can make one more
careful in writing programs (e.g., double checking code) in anticipation of having to present or
share the programs with other participants. Thus, errors are often eliminated even before the
actual review sessions.

4.Group Process. Most FTR methods are implemented using small groups. Therefore, several
key issues from small group theory apply to FTR, such as group think (tendency to suppress
dissent in the interests of group harmony), group deviants (influence by minority), and
domination of the group by a single member. Other key issues include social facilitation
(presence of others boosts one's performance) and social loafing (one member free rides on the
group's effort) [Myers 1990]. The issue of moderator domination in inspections is also
documented in the literature [Tjahjono 1996].
Perhaps the most interesting research from the perspective of the current study is that of
Sauer et al. [2000]. This research is unusual in that it has an explicit theoretical basis and
outlines a behaviorally motivated program of research into the effectiveness of software
development technical reviews. The finding that most of the variation in effectiveness of
software development technical reviews is the result of variations in expertise among the
participants provides additional motivation for developing a solid understanding of Formal
Technical Review at the individual level.

It should be noted that all of this work, while based on psychological theory, does not address
the issue of how practitioners actually evaluate software artifacts.
2.3 Approaches to the Evaluation of Diagrammatic Models
The focus of this dissertation is the exploration of how practitioners as individuals evaluate
diagrammatic models for semantic errors that would cause the resulting system not to meet the
functionality, performance, security, usability, maintainability, testability or other requirements
necessary to the purposes of the system [Bass et al. 1998; Boehm et al. 1978].

2.3.1 General Approaches

Information Systems is an applied discipline that traditionally adapts concepts and techniques
from reference disciplines such as management, psychology, and engineering to solve
information systems problems. In searching for a theoretical model that could be used in the
current research, three separate approaches were explored.

Testing 53
1. Computer Aided Design (CAD). Since CAD uses diagrams to specify the design and
construction of physical entities [Yoshikawa and Warman 1987], it seemed reasonable to
assume that techniques developed to evaluate CAD diagrams might be adapted for the
evaluation of diagrams used to specify software systems. However, a review of the literature
found relatively little literature on the evaluation of CAD diagrams, and that which was found
pertained to the formal (i.e., "mathematical") evaluation of circuit designs. Discussion with
William Miller of the University of South Florida Engineering faculty supported this conclusion
[Miller 2000], and this approach was abandoned.

2.Radiological Images. While x-rays are not technically diagrams and do not specify a system,
they are visual artifacts and do convey information. Therefore, it was reasoned that rules for
reading radiological images might provide insights into the evaluation of software diagrammatic
models. Review of the literature found nothing appropriate. More importantly, as further
conceptual work was done regarding the purposes of evaluating software diagrammatic models,
it became apparent that the reading of x-rays was not an appropriate analog. This approach was
therefore also abandoned.

3.Human-Computer Interaction (HCI). In reviewing the HCI literature, the following facts were
noted:

•The language, concepts, and purposes of HCI are very similar to those of information
systems, and it is arguable that HCI is a part of information systems. (See, for example, the
Huber [1983] and Robey [1983] debate on cognitive style and DSS design.)
•HCI is solidly rooted in psychology, a traditional information systems reference discipline.
•Computer user-interfaces almost always have a visual component and are increasingly
diagrammatic in design.
•User-interfaces can be and are evaluated in terms of the semantic error criteria described
above; i.e., defects in functionality, performance, efficiency, etc.

Based on these facts, a decision was made to attempt to identify an HCI evaluation
technique that could be adapted for evaluation of software diagrammatic models.

2.3.2 Human-Computer Interaction

Human-computer interaction (HCI) has been defined as "the processes, dialogues . . . and
actions that a user employs to interact with a computer environment [Baecker and Buxton 1987,
40]."

2.3.2.1 HCI Evaluation Techniques

Mack and Nielsen [1994] identify eight usability inspection techniques:

1. Heuristic Evaluation. Heuristic evaluation is an informal method that involves having

usability specialists judge whether each dialogue element conforms to established usability
principles or heuristics. Nielsen, the author of the technique, recommends that evaluators go
through the interface twice and notes that "[t]his two-pass approach is similar in nature to the
phased inspection method for code inspection (Knight and Myers 1993) [Nielsen 1994, 29]."

2. Guideline Reviews. Guideline reviews are inspections where an interface is checked for
conformance with a comprehensive list of guidelines. Nielsen and Mack note that "since
Testing 54
 guideline documents contain on the order of 1,000 guidelines, guideline reviews require a
high degree of expertise and are fairly rare in practice [Nielsen and Mack 1994, 5]."

3. Pluralistic Walkthroughs. A pluralistic walkthrough is a meeting in which users, developers,

and human factors experts step through a scenario, discussing usability issues associated
with dialogue elements involved in the scenario steps.

4. Consistency Inspections. Consistency inspections have designers representing multiple

projects inspect an interface to see whether it consistent with other interfaces in the "family"
of products.

5. Standards Inspections. In a standards inspection, an expert on some interface standard

checks the interface for compliance with that standard.

6. Cognitive Walkthroughs. Cognitive walkthroughs use an explicitly detailed procedure to

simulate a user's problem-solving process at each step in the human-computer dialog,
checking to see if the simulated user's goals and memory for actions can be assumed to lead
to the next correct action.

7. Formal Usability Inspections. Formal usability inspections are designed to be very similar
to the Fagan Inspection used in code reviews.

8. Feature Inspections. In feature inspections the focus is on the functionality provided by the
software system being inspected; i.e., whether the function as designed meets the needs of
the intended end users.

These HCI evaluation techniques are clearly similar to FTR in that they involve the use of
knowledgeable individuals to detect defects in a software artifact; most also involve a formal
process and a group.

2.3.2.2 Cognitive Psychology and HCI

To assist in the design of better dialogues, HCI researchers have attempted to apply the findings
of cognitive psychology since, all other factors being equal, an interface that requires less short-
term memory resources or can be manipulated more quickly because fewer cognitive steps are
required should be superior. The following is a brief overview of cognitive-based approaches
utilized in HCI.

•Human Information Processing System (HIPS). During the 1960s and 1970s, the main
paradigm in cognitive psychology was to characterize humans as information processors that
processed information much like a computer. While some of the assumptions of the original
model proved to be overly restrictive and other approaches have become popular, updated HIPS
models continue to be useful for HCI research. Given the importance of this model for this
research, a more complete treatment is provided in Section 2.4.1 below.

•Computational approaches also adopt the computer metaphor as a theoretical framework but
conceptualize the cognitive system in terms of the goals, planning, and action involved in task
performance. Tasks are analyzed not in terms of the amount of information processed in the
various stages but in terms of how the system deals with new information [Preece et al. 1994].

Testing 55
• Connectionist approaches simulate behavior through neural network or Parallel Distributed
Processing (PDP) models in which cognition is represented as a web of interconnected
nodes. Connectionist models have become increasingly accepted in cognitive psychology
[Ashcraft 1994], and this fact has been reflected in HCI research [Preece et al. 1994].

• Human Factors/Actors. Bannon [1991, 28] argues that the term human factors should be
replaced with the term human actors to indicate "emphasis is placed on the person as an
autonomous agent that has the capacity to regulate and coordinate his or her behavior,
rather than being a simple passive element in a human-machine system." The change is
supposed to facilitate focusing on the way people act in real work settings instead of viewing
them as information processors.
• Distributed Cognition. An emerging theoretical framework is distributed cognition. The goal
of distributed cognition is to conceptualize cognitive activities as embodied and situated
within the work context in which they occur [Hutchins 1990; Hutchins and Klausen 1992].

The human factors/actors and distributed cognition models are not appropriate to the current
study. The connectionist models show great promise but are not yet sufficiently developed to be
useful for this research. The information processor models are however appropriate and
sufficiently mature; they provide the primary cognitive theoretical base for the dissertation.
Computational approaches are also utilized in that the study analyzes the cognitive system in
terms of the task planning involved in task performance.

2.4 Human Information Processing System (HIPS) Models and Related Topics

2.4.1 General Model

One of the major paradigms in cognitive science is the Human Information Processing System
model. In this model, humans are characterized as information processors, in which information
enters the mind, is processed in a series of ordered stages, and then exits [Preece et al. 1994].
Figure 2.1 summarizes one version of the basic model [Barber 1988].

Figure 2.1 Human Information Processing Stages (adapted from Barber [1988])
An early attempt to apply the model was Card et al.'s The Psychology of Human-Computer
Interaction [1983]. In that work, the authors stated that the human mind is also an information-
processing system and developed a simplified model of it that they called the Model Human
Processor. Based on this model, they made predictions about the usability of various user
interfaces, performed experiments, and reported their findings. The results were equivocal, and
subsequent cognitive psychology research has shown that the serial stage approach to
cognition of the original model is overly simplistic.
The original model also did not include memory and attention. Later versions do include these
processes, and Cowan [1995], in his exhaustive examination of the intersection of memory and
attention, discusses a number of these. Figure 2.2 summarizes a model that does include
memory and attention [Barber 1988].
Testing 56
Figure 2.2 Extended Stages of the Information Processing Model (adapted from Barber
[1988])
HIPS models, such as Anderson's ACT-R [1993], continue to be developed and are useful.
Further, the information processing approach has recently been described as the primary
metatheory of cognitive psychology [Ashcraft 1994].
2.4.2 Coping with Attention as a Limited Resource
One of the earliest psychological definitions of attention is that of William James [1890, vol. 1,
403-404]:
Everyone knows what attention is. It is the taking possession of the mind, in clear and
vivid form, of one out of what seem several simultaneously possible objects or trains of
thought. Focalization, concentration of consciousness are of its essence. It implies
withdrawal from some things in order to deal more effectively with others . . .
(emphasis added)

This appeal to intuition explicitly states that attention is a limited resource.

In reaction to the introspection methodology of James, the Behaviorist movement asserted that
the study of internal representations and processes was unscientific. Since behaviorists
dominated American psychological thought during the first half of the Twentieth Century, little or
no work was done on attention in America during this period. In Europe, Gestalt psychology
became dominant at this time and that school, while not actively hostile to attention studies, did
not encourage work in the area. World War II however led to a rethinking of psychological
approaches and acceptance of using the experimental techniques developed by the behaviorists
to study internal states and processes [Cowan 1995].

An example of this rethinking is the work of Broadbent [1952] and Cherry [1953]. They used a
technique to study attention in which different spoken messages are presented to a subject's two
ears at the same time. Their research shows that subjects are able to attend to one message if

Testing 57
the messages are distinguished by physical (rather than merely semantic) cues, but recall
almost nothing of the nonattended channel. In 1956, Miller reviewed a series of experiments that
utilized a different methodology and noted that, across many domains, subjects could keep in
mind no more than about seven "chunks" simultaneously. These findings were among the first
experimental evidence that attentional capacity is a limited resource.
More recent experimental work continues to indicate that attention is a limited resource
[Cowan 1995]. Even those cognitive psychologists who have recently challenged the very
concept of attention assume their "attention" analog is limited. One example of this would
be Allport [1980] and Wickens [1984], who argue that the concept of attention should be
replaced with the concept of multiple limited processing resources.
Based on an examination of the exhaustive review by Cowan [1995] of the intersection of
memory and attention, the Shiffrin [1988, 739] definition appears to be representative of
contemporary thought:

Attention has been used to refer to all those aspects of human cognition that the
subject can control . . . and to all aspects of cognition having to do with limited
resources or capacity, and methods of dealing with such constraints. (emphasis added)

Since human cognitive resources are limited, cognitively complex tasks may overload these
resources and decrease the quality and/or quantity of outputs. Various approaches to measuring
the cognitive complexity of tasks have been developed. In HCI, an informal view of complexity is
often utilized. For example, Grant [1990, sec. 1.3] defines a complex task as “one for which
there are a large number of potential practical strategies.” This definitions is not inconsistent with
the measure assumed by Simon [1962] in his paper on the use of hierarchical decomposition to
decrease the complexity of problemsolving.

Simon [1990] argues that humans develop mechanisms to enable them to deal with complex,
real-life situations despite their limited cognitive resources. One such mechanism is task
planning. According to Fredericksen and Breuleaux [1990], task planning is a cognitive bargain
in which the time and effort spent working with an abstract, and therefore, smaller problem
space during planning minimizes actual work on the task in the original, detailed problem space.

Earley and Perry [1987, 279] define a task plan as "a cognitively based routine for attaining a
particular objective and consists of multiple steps." Newell and Simon [1972] identify planning
from verbal protocols as those passages in which:

1. a person is considering abstract specifications of the action/information transformations

required to achieve goals;
2. a person considers sequences of two or more such actions or transformations; and
3. after developing the sequences, some or all of them are actually performed.

Two further items should be noted regarding planning:

1. Not all planning is original. Successful plans learned from others or by experience may be
stored in memory or externally [Newell and Simon 1972; Wood and Locke 1990]. Without the
recall, modification, and use of previous plans, the development of expertise would be
impossible.

2. Planning is not complete before action. Both theory and analysis of verbal protocols indicate
that periods of planning are interleaved with action [McDermott 1978; Newell and Simon
Testing 58
 1972]. In other words, practitioners will often plan a response to part of a task, complete
some or all of the actions specified in that plan, plan a new response incorporating
information acquired during prior action period(s), complete the new actions, etc.

2.4.3 Application of the HIPS Model to This Research

In the HIPS model, the nature and amount of stimuli impact both information processing and
output. This research uses a key concept of the HIPS model,
attention, in two ways:

1. Attention is a critical and limited resource, and when attention is overloaded, outputs
decrease in quality and quantity; therefore, a meta-cognitive strategy such as task planning
that minimizes attentional load should improve outputs.

2. Patterns are another meta-cognitive strategy for minimizing attentional load; therefore,
understanding which patterns better support the cognitive processing associated with
evaluation of diagrammatic models may allow individuals to be trained to use these better
patterns, thus lessening their attentional load and improving their outputs.

2.5 Research On the Comprehension of Graphics

Larkin and Simon [1987] consider why diagrams can be superior to a verbal description for
solving problems, and suggest the following reasons:

•Diagrams can group together all information that is used together, thus avoiding large amounts
of search for the elements needed to make a problem-solving inference.
•Diagrams typically use location to group information about a single element, avoiding the need
to match symbolic labels.
•Diagrams automatically support a large number of perceptual inferences, which are extremely
easy for humans.
As noted in Chapter 1, two of these depend on spatial patterns.
Winn [1994] presents an overview of how the symbol system of graphics interacts with the
viewers' perceptual and cognitive processes, which is summarized in figure 2.3. In his
description, the graphical symbol system consists of two elements: (1) Symbols that bear an
unambiguous one-to-one relationship to objects in the domain of reference, and (2) The spatial
relations of the symbols to each other. Thus, how symbols are configured spatially will affect the
way viewers understand how the associated objects are related and interact. For the purposes
of this dissertation, a particularly interesting finding is that biases based on reading direction
(left-to-right for English) affect the interpretation of graphics.

Testing 59
Figure 2.3. Winn [1994] Processes Involved in the Perception and Comprehension of
Graphics

Zhang [1997] proposes a theoretical framework for external representation based problem
solving. In an experiment she conducted using a Tic-Tac-Toe board and its logical isomorphs,
the results show that Tic-Tac-Toe behavior is determined by the configuration of the board.
External representations are thus shown to be more than just memory aids and a
representational determinism is suggested. This last point is particularly relevant to this
dissertation since it states that the form of representation determines what information can be
perceived in a diagram.

2.6 Types of Diagrammatic Models

Selection of diagrammatic models to be included in the research task requires an appropriate

typology. Two diagrammatic model typologies were examined, Wieringa [1998] and Visible
Systems [1999].

2.6.1 Wieringa 1998

Wieringa, in his discussion of graphical structures or models that may be used in software
specification techniques, lists four general classes:

1. Decomposition Specification Techniques. These represent the conceptual structure of

data in a database system. Examples include Entity-Relationship Diagrams (ERDs) and such
ERD extensions as OO class diagrams.

Testing 60
2. Communication Specification Techniques. These show how the conceptual components
interact to realize external system interactions. Examples include Dataflow Diagrams (DFDs),
Context Diagrams, SADT Activity Diagrams, Object Communication Diagrams, SDL Block
Diagrams, Sequence Diagrams, and Collaboration Diagrams.

3. Function Specification Techniques. These specify the external functions of a system or the
functions of system components. Examples Function Refinement Trees, Event-Response
Specifications, and Use Case Diagrams.

4. Behavior Specification Techniques. These show how functions of a system or its

components are ordered in time. Examples include Process Graphs, JSD Process Structure
Diagrams, Finite (and Extended Finite) State Diagrams, Mealy Machines, Moore Machines,
Statecharts, and Process Dependency Diagrams.

2.6.2 Visible Systems

The methods listing in Visible Systems [1999] was examined as a representative of practitioner-
oriented, CASE-tools-based typologies. Seven models are listed; of these, six are diagrammatic
in nature.

1. Functional Decomposition Model. Shows the business functions and the processes they
support drawn in a hierarchical structure; also known as the Business Model. This type of
model is of a high-level functional nature and specifically applies to functions and not to the
data that those functions use. It is generally appropriate for defining the overall functioning of
an enterprise, not for individual projects.

2. Data Model. Shows the data entities of an application and the relationships between the
entities. Entities and relationships can be selected in subsets to produce views of the data
model. The diagramming technique normally used to depict graphically the data model is the
Entity Relationship Diagram (ERD) and the model is sometimes referred to as the Entity-
Relationship Model.

3. Process Model. Shows how things occur in the organization via a sequence of processes,
actions, stores, inputs and outputs. Processes are decomposed into more detail, producing a
layered hierarchical structure. The diagramming technique used for process modeling in
structured analysis is the Data Flow Diagram (DFD). Several notations are available for
representing process modeling, with the most widely used being Yourdon/DeMarco and
Gane & Sarson.

4. Product Model. Shows a hierarchical, top-down design map of how the application is to be
programmed, built, integrated, and tested. The modeling technique used in structured design
is the structure chart. It is a tree or hierarchical diagram that defines the overall architecture
of a program or system by showing the program modules and their interrelationships.

5. State Transition Model (Real Time Model). Shows how objects transition to and from
various states or conditions and the events or triggers that cause them to change between
the different states.

6. Object Class Model. Shows classes of objects, subclasses, aggregations and inheritance
and defines structures and packaging of data for an application.
Testing 61
2.6.3 Evaluation of Typologies in Prior Work

In evaluating these two typologies for this research, two problems were noted:

1.Neither classification scheme includes diagrammatic representations of Graphical User

Interfaces (GUIs). While such representations are not technically graphs (and thus not discussed
by Wieringa) and are not listed in Visible Systems, they may be used to specify parts of a
system and are therefore appropriate to this research.
2. Wieringa's work is based on the theoretical characteristics of graphs while Visible Analyst is
representative of practitioner-oriented, CASE-tool-based typologies. Neither is appropriate to
the research of this dissertation since neither captures factors likely to affect the cognitive
processing of practitioners in evaluating software diagrammatic models.

While it would be relatively easy to add diagrammatic representations of GUIs to Wieringa or

Visible Analyst, it was concluded that the second problem disqualified them for the purposes of
this research. Further review of several leading systems analysis and design texts [Fertuck
1995; Hoffer et al. 1998; Kendall and Kendall 1995] did not yield an appropriate typology of
diagrammatic models, and it was therefore deemed necessary to develop one specifically for
this dissertation.

2.6.4 Diagrammatic Model Typology Development

The first step in the development process was to consult several systems analysis and design
and structured techniques texts for classification insights and to derive lists of commonly used
diagrammatic models. These included Fertuck [1995], Hoffer et al. [1998], Kendall and Kendall
[1995], and Martin and McClure [1985].

Martin and McClure make a major distinction between hierarchical diagrams (i.e., those having
one overall node or root and which do not remerge) and mesh or network diagrams (i.e., those
not having a single overall node or root or which do remerge). For the purposes of this research,
this distinction is operationalized as the categorical variable hierarchical/not hierarchical.

Martin and McClure also make a major distinction between diagrams showing sequence and
those that do not. Sequence usually implies temporal directionality; for this dissertation, the
distinction is broadened to include the possibility of logical and other forms of directionality and
is operationalized as the categorical variable directional/not directional.

A distinction found in all texts referenced is between data-oriented and process-oriented

diagrams. Inspection of diagram types shows that the distinction is actually a data/process
orientation continuum. For the purposes of this dissertation, this continuum is collapsed into the
categorical variable data/hybrid/process oriented.

As a test of the feasibility of the classification scheme, twenty diagram types from Martin and
McClure, UML diagrams from Harmon and Watson [1998], and a model of a "typical" GUI were
then categorized. The results of this categorization are shown in table 2.1.

Table 2.1 Diagrammatic Model Types

Testing 62
Inspection of table 2.1 shows that only seven of the twelve (2 x 2 x 3) possible categories are
actually populated. Table 2.2 shows the categorization of the diagram types after collapsing
unpopulated categories.
HIERARCHICAL NOT HIERARCHICAL
DIRECTIONAL NOT DIRECTIONAL DIRECTIONAL NOT DIRECTIONAL
DATA HYBRID PROCESS DATA HYBRID PROCESS DATA HYBRID PROCESS DATA HYBRID PROCESS
I II III IV V VI VII VIII IX X XI XII
Functional Functional Data Flow Data Analysis “Typical”
Decomposi-tion Decomposi-tion GUI
II I

Structure Charts Flow Charts Entity- UML Use

Relationship Case

HIPO HIPO Data Navigation Inverted-L UML Class

(Overview) (VTC)

HIPO UML Sequence

(Detail)

Warnier- Warnier-Orr UML

Orr (Process) Collaboration
(Data)

Michael Michael Jackson Michael UML State

Jackson System Network Jackson
Data- Program-
Structure Structure

Nassi- UML Activity

Shneiderman
Charts

Action II Action I

Table 2.2 Diagrammatic Model Types (Collapsed)

Testing 63
 HIERARCHICAL NOT HIERARCHICAL
DIRECTIONAL DIRECTIONAL NOT DIRECTIONAL
DATA HYBRID PROCESS HYBRID PROCESS DATA HYBRID
I II III VIII IX X XI
Functional Functional Data Flow Data Analysis “Typical” GUI
Decomposition II Decomposi-tion I

Structure Charts Flow Charts Entity- UML Use Case

Relationship

HIPO HIPO Data Navigation Inverted-L UML Class

(Overview) (VTC)

HIPO UML Sequence

(Detail)

Warnier-Orr Warnier-Orr UML Collaboration

(Data) (Process)

Michael Jackson Michael Jackson Michael Jackson UML State

Data-Structure System Network Program-Structure

Nassi-ShneidermanUML Activity
Charts

Action II Action I

2.7 Types of Software Defects

A semantic software defect (the focus of this research) is defined as a non-syntactic defect that
causes a software artifact or resulting system not to have the functionality, performance,
security, usability, maintainability, testability or other qualities necessary for the purposes of the
system. In other words, software defects are defined in terms of missing qualities. Other
research reviewed is not inconsistent with this approach. For example, Boehm et al. [1978] and
Bass et al. [1998] develop typologies of software qualities, and the definition in Grady [1992,
122] of a defect as "any flaw in the specification, design, or implementation of a product"
inherently includes software qualities. Therefore, the primary focus of the first section below is
on typologies of software qualities. The second section reviews other software defect typologies,
and the third section discusses the development of the typology used in this research.

2.7.1 Software Quality Typologies

An interesting early software qualities typology is the Software Quality Characteristics Tree
(SQCT) of Boehm et al. [1978]. The SQCT is a hierarchical scheme in which the highest-
level construct, General Utility, is determined by two second-level constructs, As-Is Utility
and Maintainability, and one third-level construct, Portability. The second-level constructs
are each in turn determined by three other third-level constructs, Reliability, Efficiency, and
Testing 64
Human Engineering and Testability, Understandability, and Modifiability respectively. The
third-level constructs are determined by various combinations of twelve primitive
characteristics (Device Independence, Completeness, Accuracy, Consistency, Device
Efficiency, Accessibility, Communicativeness, Structuredness, Self-Descriptiveness,
Conciseness, Legibility, and Augmentability), which are strongly differentiated with respect
to each other.

The Software Quality Characteristics Tree is shown in figure 2.4.

Testing 65
Figure 2.4 Boehm et al. [1978] Software Quality Characteristics Tree (adapted)

The Grady [1992] software defect model is shown below in figure 2.5. It is also a hierarchical
model (with the root at the bottom) that classifies defects according to origin, type, and mode.
Grady describes six types of software defects that correspond to the five modes plus a residual
"Other" category:

Testing 66
1. Specifications/Requirements Defect. A mistake in the definition of the customer/target
needs for a system or system component. Such mistakes can be in functional requirements,
performance requirements, test requirements, development standards, and so on.

2. Design Defect. A mistake in the design of a system or system component. Such mistakes
can be in algorithms, control logic, data structures, database access, input/output formats,
interface descriptions, and so on.

3. Code Defect. A mistake in the implementation of a computer program. Such mistakes can be
in product or test code, JCL, build files, and so on.

4. Documentation Defect. A mistake in any non-code product material delivered to a customer.

Such mistakes can be in user manuals, installation instructions, data sheets, product demos,
and so on. Mistakes in requirements specification documents, design documents, or code
listings are assumed to be specification defects, design defects, and coding defects,
respectively.

5. Environmental Support Defect. Defects that arise as a result of the system development
and/or testing environment. Such mistakes can be in the build/configuration process, the
development/integration tools, the testing environment, and so on.

6. Other.

Testing 67
Figure 2.5 Grady [1992] Software Defect Model

Bass et al. [1998] discuss ten technical qualities of software, dividing them into those that are
discernible at runtime (DR) and those not discernible at runtime (NDR). The following is a brief
discussion of the software qualities in their typology:

1. Functionality (DR) is the ability of the system to do the work for which it was intended; it is the
basic statement of the system's capabilities, services, and behavior.

2. Performance (DR) refers to the responsiveness of the system - the time required to respond
to stimuli (events) or the number of events processed in some interval of time.

Bass et al. [1998, 79] note that "For most of the history of software engineering, performance
has been the driving factor in software architecture, and this has frequently compromised the
achievement of other qualities."

Testing 68
 It should be noted that performance is relative to system requirements and that what would
otherwise be a "defect" may be the result of increasing some other quality.

3. Security (DR) is a measure of the system's ability to resist unauthorized attempts at usage
and denial of service while still providing its services to legitimate users.

4. Availability (DR) measures the proportion of time the system is up and running and is
typically defined as

α = (MTF) / (MTF + MTR) ,

where MTF = mean time to failure and

MTR = mean time to repair.

5. Usability (DR) is largely a function of the user interface.

6. Maintainability (NDR). Bass et al. [1998] use the terms modifiability and maintainability
interchangeably and define modifiability as the ability of a system to make changes quickly
and cost effectively. According to them, modifications to a system can be broadly categorized
as follows:

•Extending or changing capabilities. This category includes corrective maintenance and

extensibility.
•Deleting unwanted capabilities.
•Adapting to new operating environments.
•Restructuring.

7. Portability (NDR) is the ability of a system to run under different computing environments.

8. Reusability (NDR) relates to the design of a system so that the system's structure or some of
its components can be reused again in future applications. Bass et al. [1998, 84] note that
"Reusability is actually a special case of modifiability..."

9. Integrability (NDR) is the ability to make the separately developed components of the system
work correctly together.

10. Software testability (NDR) refers to the ease with which software can be made to
demonstrate its faults through (typically execution-based) testing.

This research uses Bass et al. [1998] as the basis for the qualities dimension of the software
defects typology.

2.7.2 Other Defect Dimensions

Review of the literature yields three other dimensions for the classification of software defects.

2.7.2.1 Class
Class refers to whether the defect is the result of logic or other required structure's being
missing (M), incorrect (I), or extra (E) [Ebenau and Strauss 1994].

Testing 69
While extra functionality may increase storage requirements or otherwise decrease efficiency,
the impact on functionality is generally less severe than that caused by the other two types.

2.7.2.2 Severity

The defect severity categories generally listed are major (J), minor (N), and (sometimes) trivial
(T) [Ebenau and Strauss 1994; Gilb and Graham 1993; Kelly et al. 1992].

A major defect is defined as one "that is expected to cause product failure, departure from
specifications, or prevent further correct development of the product[Ebenau and Strauss 1994,
92]." A minor defect is defined as one "that reduces the effectiveness, or confuses a product's
representation, format, or development process characteristics, but is not expected to impact the
operation or further development of the product [p. 92 ]."

2.7.2.3 Cause

Humphrey [1995], following Gale [1990], lists five categories of basic defect causes:

1. Education. You did not understand how to do something.

2. Communication. You were not properly informed about something.
3. Oversight. You omitted doing something.
4. Transcription. You knew what to do but made a mistake in doing it.
5. Process. Your process somehow misdirected your actions.
2.7.3 Development of the Defect Typology

The four dimensions discussed above produce a four-dimensional defect space. However,
examination shows that dimensional simplification is appropriate.

1. Defect cause cannot be determined directly from examination of software diagrammatic

models.

2. Defect severity is defined in terms of impact on system functionality. Given that functionality
is a type of technical quality, a separate dimension would be redundant.

Further simplification is achieved by ignoring extra functionality defects of the class dimension.
The rationale for this reduction is that, while defects associated with extra functionality may
increase storage requirements or otherwise decrease efficiency, the impact on functionality is
generally less severe than that caused by missing and incorrect defects.

Change is also necessary on the qualities dimension. Six of the Bass et al. [1998] qualities are
not readily discernable from diagrammatic models and are consequently not appropriate to the
typology. However, according to Boehm et al. [1978], the primitive quality Structuredness
partially determines three of the six. Similarly, Fenton and Neil [2001] lists Structuredness as an
internal attribute associated with the external attributes reliability (or availability), maintainability,
and reusability. The six non-discernable qualities are listed below. A B indicates a Boehm quality;
an F indicates a Fenton attribute.

•Availability F
•Maintainability B,F
•Portability

Testing 70
 •Reusability B,F
•Integrability
•Testability B

Since Structuredness is associated with four of the six non-discernable qualities and is readily
discernable from a diagrammatic model, it is substituted as a partial proxy.

During the early development of the research task, several subjects noted that the scope of the
diagrammatic models was not consistent. From a theoretical perspective, lack of Scope
Consistency is an instance of a general consistency problem. In the structured approach to IS
development, data and process models are supposed to model the same system but are
fundamentally separate. This separateness leads to multiple problems including lack of
consistency [Repa 2001]. Consideration was given to adding the broader quality consistency to
the topology, but this was rejected because (1) some subjects perceived lack of Scope
Consistency to be a separate issue and (2) lack of Scope Consistency is different in that it can
generally be readily discerned by comparing data and process models, while other consistency
problems are apparent only after significant functional analysis. Lack of Scope Consistency
would be expected to impact negatively on the integrability and maintainability of the specified
system

The resulting matrix is a two-dimensional defect space based on quality affected and class. It
should be noted that Scope Consistency and Structuredness are treated as logical variables; the
quality is either present or missing. Table 2.3 shows the resulting matrix.

Table 2.3 Software Defect Matrix: Qualities vs. Class

QUALITY
Scope Consistency

Structuredness

Functionality

Performance

Usability
Security

CLASS

Missing

Incorrect — —

Testing 71
2.7.4 Diagrammatic Model Type vs. Software Defect Type Matrix

Table 2.4 shows the matrix resulting from combining the Diagrammatic Model Type and Software
Defect Type typologies.

Table 2.4 Diagrammatic Model Type vs. Software Defect Type

Scope Consistency

QUALITY
Structuredness

Functionality

Performance

Usability
Security
MODEL M M M I M I M I M I
Hierarchical-
W-O D1

Directional-
Data (I)
Hierarchical-
StrC2

Directional-
Hybrid (II)
Hierarchical-
W-O P3

Directional-
Process (III)
Not Hierarchical-
DFD4

Directional-
Hybrid (VIII)
Not Hierarchical-
FlowC5

Directional-
Process (IX)
Not Hierarchical-
ERD6

Not Directional-
Data (X)
Not Hierarchical-
GUI7

Not Directional-
Hybrid (XI)

Testing 72
NOTES
M = missing
I = incorrect

Typical Diagram for Each Category

1 = Warnier-Orr (Data) Diagram

2 = Structure Chart
3 = Warnier-Orr (Process) Diagram
4 = Data Flow Diagram
5 = Flow Chart
6 = Entity-Relationship Diagram
7 = “Typical” GUI

2.8 Summary and Conclusions

Prior theory and research that might inform the dissertation are reviewed. A large body of
research exists concerning Formal Technical Review, but review of this work shows that it is not
based on theory and therefore cannot inform this research effort. The first part of the literature
review therefore provides context rather than explicating applicable theory.

Three techniques from non-information systems disciplines for evaluating visual artifacts
conveying meaning are evaluated. While work on the evaluation of human-computer interaction
(HCI) approaches proves not to be directly applicable to this research, one of the HCI
paradigms, the Human Information Processing System (HIPS) model, is found to be relevant.
The HIPS model is reviewed, as is cognitive science work on attention and the comprehension
of graphics.

Two other areas are identified as necessary for the development of the research task and tools:
(1) types of diagrammatic models and (2) types of software defects. The literature is reviewed
and new typologies are developed.

Testing 73