Jim;

There are a number of different approaches I have taken to software testing depending upon policies and procedures for whom I am working for. I must say that I use three sets of data for testing – clean (good) data, dirty (bad) data, and new data entries. I cannot stress the importance of known data sources. You must know what to expect and everyone testing must use the same data source. I am speaking from experience.


Here are the scenarios:

1. Independent consultant. I test my own work. I attempt to test for every possible type of data entry that is possible as well as every condition. Typically, my comments are longer then the code I write.

2. Full time employee

A. No testing allowed – company policy. I mean NO TESTING! The company went belly up but I quit well before that occurred. The owner stated, “Let the customers report bugs”! Six programmers.

B. Test only your own work and nothing else! Two programmers

C. Test your own work and other related functional items. Pass it to SQA (Software Quality Assurance) who has a team of professionals to test the product. Five programmers. Very professional approach. One SQA person (college degreed) for each programmer.

D. Test your work and bring in a qualified SQA tester. Three programmers.

E. Test everything possible after creating or modifying code. One-man full time SQA. This person refused to use the same data we used and continuously found “problems”. Many hours were wasted trying to write code to make the testers data “work”. Example: No child records. Five programmers.

I find it impossible to give a number as to how much of the code I have been associated with has been tested. It really depends upon where I worked.

In the hardware world we use different metrics to find non-conformance. The hardware world is perhaps better-established and vastly different from software. However, in the software world there seems to be a huge variety of approaches to testing. Not all approaches are of a professional nature or realistic.

I have worked for companies that had one of two attitudes: Ship product regardless if it works or not! Or deliver only the finest product! Which is the best approach? It depends. Quality costs money and not every organization can afford to approach what you might term a satisfactory level. If it works for the customer then everything is all right. However if there is a problem, that is an entirely different story. It is called “a business decision”.

Tom




>The assertion has been made that typical applications have between 40%-50% of their code actually run during testing.
>
>I am curious to see what percentage of code you test (actually run during development/unit test/system testing) for:
>
>1) The development of a system from scratch.
>
>
>2) The development of a new/replacement system but using code parts (including classes, functions, etc.) from existing and well-running systems.
>
>
>3) Developing revisions to an existing application.
>
>
>Thanks for your reply.