# Annex D Numerical examples as a regression test

Annex D is informative and gives worked numerical examples. They are the cheapest regression test in the part, because they turn a formula into a value that either matches or does not. The fourth edition will reissue them for the revised text.

## What it asks

Run the examples against the implementation and keep the output with the release record, and add any local parameter set to the same test. Where an example does not match, record the difference as a defect rather than adjusting the expected value.

## What the revision does

The fourth edition will update the examples. The revised values are not public.

## Who it binds

The test owner and the release manager.

## Evidence an auditor samples

- The example runner and its expected values (test suite)
- The last run output, dated, with the edition it used (release record)
- The defect record for any mismatch (issue tracker)

## The test you can run

Run the examples on a clean build. A pass on one family and a skip on another is a finding, not a partial pass.

## Common failures

- The examples are run once by hand and never in the pipeline.
- A mismatch is resolved by changing the expected value.

## How to close it

Put the examples in the pipeline and make the release record quote the run. It is the highest value test per line of code in a mechanism change, and it catches the arithmetic drift that reviews miss.

## Source

ISO/IEC 9798-5:2009(E), Annex D, publisher preview; replaced by ISO/IEC DIS 9798-5, stage 40.20. Read on 21 October 2026. Our wording, not the instrument's.
https://www.iso.org/standard/89516.html

Not legal advice; does not certify.