Any test that merely shows that, under some conditions
software functions as expected, is only useful as a regression test, that is, to demonstrate that the
system is minimally capable of functioning.
??? A successful test is one which finds a bug. There are 5 kinds of bugs:
1
a. The bug which exists in terms of some specification. (The specification says that
having pressed
F1 a Help menu will always be displayed
. On pressing F1 a help menu is not displayed. Therefore
there is a bug.)
b. The bug which can be imagined. (The early DOS message that said
keyboard not present press
F1
.)
c. The bug which can??™t be imagined. (
The specification says that
having pressed F1 a Help menu
will always be displayed.
I have just pressed F1 and while a menu was displayed, it is now obscured
by the smoke billowing out of the machine.
) These bugs depend on some interaction between
the system and the ???real??? world. The latter cannot be completely specified.
2
d. The bug which has been foreseen but not sufficiently allowed for. (
The possibility of an aircraft
destroying a building because a turbine compressor disk broke off and caused the aircraft to crash,
because a turbine blade melted, because the turbine blade heat sensor didn??™t sound the alarm in
time, because there was a delay in the software.
)
e. The non-functional bug. By non-functional I mean some part of the system where the failure
criterion can only be specified by a (set of) range(s) of values, such as (say) performance or
usability.
Pages:
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156