Royal Pain

James Knott james.knott-bJEeYj9oJeDQT0dZR+AlfA at public.gmane.org
Fri Jun 18 14:28:02 UTC 2004


Lennart Sorensen wrote:
> On Fri, Jun 18, 2004 at 07:27:26AM -0400, James Knott wrote:
> 
>>And of course, it's impossible to prove there are no bugs.  You can only 
>>fail to find some.
> 
> 
> That is actually not true.  If you have a well defined specification of
> what the behaviour of each piece of the program must be for specific
> inputs, you can actually prove the behaviour of each part of the program
> correct.  This is in fact done at some software companies.  It is
> certainly a lot more work and costs more.  It requires proper bounds
> checks, and full coverage testing at the very least.


Of course, you have to make sure you can test for everything that might 
cause a problem.  There may be some circumstance that you hadn't 
considered that may cause problems.  This happened to me once at IBM.  I 
  was doing intengration testing of a new standard desktop system. 
Everything checked out OK.  But, after release, one user had a font 
problem, with one of the applications that I supported, when used with a 
printer that I hadn't tested with.  I hadn't tested with that printer, 
simply because we didn't have one of that model in the test lab.  The 
cause turned out to be a conflict between the installation of two 
applications and reversing the install order solved that problem.  When 
you're doing integration testing of a complete system, it's physically 
impossible to test for all possible conditions, to verify there's 
absolutely no bugs.



--
The Toronto Linux Users Group.      Meetings: http://tlug.ss.org
TLUG requests: Linux topics, No HTML, wrap text below 80 columns
How to UNSUBSCRIBE: http://tlug.ss.org/subscribe.shtml





More information about the Legacy mailing list