If the risk of a system failure is in some way catastrophic
then YES. Additionally a spreadsheet like the one shown above will show you clearly which:
TABLE 15.8 Example chart plotting the system features against the various system tests
System test name/
feature name 023-56 023-57 023-58 023-59 023-60 023-61 023-62 023-66
file: print.d
collect x
yrstrncmp x x x
print BA table x x
print literals x x x x
print msg x x
print prompt x
print range status x x
print test pages
user range x x x
valid response x x x x
System and Acceptance Testing 257
??? Objects are under- or over-tested
??? Tests cover which objects
??? Scenarios have no, too few or too many tests against them
??? Scenarios have a higher priority (and which tests mapped against them should be in the regression
test set)
??? Scenarios cover the same objects too often and could be split
??? States cannot (or should not be able to) be reached
??? Scenarios provoke the most bugs (if you map the bugs found, against first the tests and then the
scenarios)
Other questions remain:
1. How can I show that I have exercised all the code using this level of coverage? Get the source code
instrumented to show which statements have been executed, hope that the system??™s timing is
sufficiently robust to cope with the delays this instrumentation will impose, run all your tests,
and then look at the results. You will be lucky to hit more than 90% since some code can only be
exercised by errors.
Pages:
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493