Functional Testing - Software Testing: Testing Across the Entire Software Development Life Cycle - page 118

both data and processing from the error and allow the end user to continue error-free
operation.
The richness of the error recovery code is limited in part to the programmer's
exposure to end users of similar kinds of applications or support technologies. The
richness of error recovery testing is also limited in part to the tester's exposure to
er ror recover y techniques that do not work well regardless of the business application
or support technologies.
In an ideal development project, the programmers maintain a single list of errors
that end users can encounter. This list would contain items like

Application X—message list contents

unique error code/ID
error message text
error description
code module that detects the error condition
severity of the error encountered
I Informational

W Wa r n ing

Abort current activity and attempt to recover to preerror operation
user action required
application action required
likelihood of data loss

A

Such a list becomes the basis for planning 100% error handling validation. The
testing challenge is to repeatably (remember SPRAE) cause the errors on demand.
The sad truth is that many development projects let their programmers imbed
most of this error message information in the source code. The practice of scattering
error handling information throughout the source code is so prevalent in Internet
applications that software vendors have found a thriving niche market in providing
reverse engineering tools that fi nd and list out all error messages in an application.
The testing challenge is to determine what circumstances can cause each reverse
engineered error message to appear.

7.6.4.2 Data Type Conversions

Any time user inputs are converted from one data type to another, for example,
alphabetic to numeric, there is the possibility of an incorrect or failed conver-
sion. Depending on the programming language(s) used to develop the new ap-
plication, this failed conversion can cause a range of execution responses from
benign return codes that the application can intercept and interpret to malignant
execution halts. The corollary to failed conversion is the application's ability to
identify incorrect conversions (alphas in a strictly numeric fi eld) before they are
attempted.