Browse Prior Art Database

Creating Evolving and Using an Online Large Software System Test Base

IP.com Disclosure Number: IPCOM000083202D
Original Publication Date: 1975-Apr-01
Included in the Prior Art Database: 2005-Mar-01
Document File: 6 page(s) / 152K

Publishing Venue

IBM

Related People

Shils, AJ: AUTHOR

Abstract

PROBLEMS SOLVED. System development release test cycles consist of running test decks against all the executives and application programs. These test decks are now manually controlled. Because the application is so large and complex, the testing becomes complex; therefore, a semiautomatic test method is proposed. Features in this method are: (1) STCE; Standard Test Characteristic Element (A Data Base). (2) TSAP; Test Selection Analysis Program (Interactive Program). (3) TFEP; Test Finalize and Execute Program (Background Program). (4) LTLM: Latest Test Link Matrix (A Data Set).

This text was extracted from a PDF file.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 40% of the total text.

Page 1 of 6

Creating Evolving and Using an Online Large Software System Test Base

PROBLEMS SOLVED. System development release test cycles consist of running test decks against all the executives and application programs. These test decks are now manually controlled. Because the application is so large and complex, the testing becomes complex; therefore, a semiautomatic test method is proposed. Features in this method are:
(1) STCE; Standard Test Characteristic Element (A Data Base).
(2) TSAP; Test Selection Analysis Program (Interactive Program).
(3) TFEP; Test Finalize and Execute Program (Background

Program).
(4) LTLM: Latest Test Link Matrix (A Data Set).
(5) TLMM; Test Link Matrix Modify (Interactive Program to

update LTLM)

The Specific Problems treated by this method, programs, and data sets are:
(1) Because so many decks are manually controlled per test cycle,

numerous job Control Language (JCL) errors occur. These

require

an average of nearly one run per test to debug. Although

most

runs are only a few minutes, a tester may get only one or two

runs per day. By virtually eliminating JCL type errors,

several

man-days could be saved per test cycle.
(2) There is no central control of test parameters. Deck

duplication and interpretation

for slight modification of a test is manual, slow,

subject to error, and results in 'cannibalization' of decks.

This creates additional JCL errors and test data errors which

add to debug time. Manual parameter control means the tester

cannot know (except intuitively or by hand checking the decks)

the use extent of a parameter or option. Such knowledge would

be valuable to reduce duplicate testing and indicate areas of

inadequate testing.
(3) A tester sets up his cases during the 1st shift, usually to be

executed during the 2nd or 3rd shifts. Test decks use numerous

data sets, some of which change rapidly during a test cycle. A

data set named by the tester during 1st shift may be renamed or

made unnecessary by the developer of the data set before the test

executes. This method includes a process, TFEP, to name at the

last minute the latest data sets.
(4) A small testing group that must use its limited resources to

test manually, precludes development of test tools to generate

data, analyze flow paths, compare programs, and automate

comparison of design output. With human resources freed by

some

initial automatic test facility, this group could further

develop its test practices.
(5) Future releases will be more complex, requiring more complex

1

Page 2 of 6

testing, making automation even more necessary.

DESCRIPTION:

The initial use of this automated test method is for a very large (greater than 1500 modules) system design process. It should apply to any large scientific, design, financial, or operating system. Fig. 1 is an overview of the process.

The Test Data Base - Assume there are now m (m 150) test decks.

The proposed method would require (n+2), (n<<m) data set members. The `weeding' of m down to n is (1...