Prev | Current Page 91 | Next

Peter Farrell-Vinay

"Manage Software Testing"


Story
on??™t be ridiculous,??? said my boss, when I told him that we were spending at least 33%
of the project??™s money on testing. ???Go and look at the time sheets.???5
So I did. I was wrong. We weren??™t spending 33% of our resources on testing: it was 35%.
5 The company sensibly had all staff fill out time sheets showing how much time was spent on various project
phases. I included unit and system testing.
???D
Introduction 11
Remember however, that testing does not, and cannot prove that some software conforms to
specification. By the nature of proof itself, all that a test can show is that:
??? (rather negatively) An attempt to make the software fail with respect to some specification has
itself failed
??? (rather positively) That the software worked
This can be taken only as an implication that the software has passed that test, and therefore
conforms to that specification on which the test was based. By passing some test, all that can
reasonably be deduced is that the test failed to find any bugs in that software.
Example: a very expensive motor-car is being road-tested. The baselines of the test are the
specifications for that car plus all the last-minute tweaks made to its software, suspension, fuel
injection system etc. It is a bright summer??™s day. The test consists of the following procedure:
1. Start up the car.
2. Drive for 100 miles at speeds of up to 70 mph.
3. Stop the car.
Assuming that the car passes such a test, what can we conclude from it? We can assume that
under the circumstances of a bright summer??™s day the car can be persuaded to start, be driven,
and stop.


Pages:
79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103