Accessibility · Cluster
Testing Accessibility
Automated checks and manual verification.
- 01 What an Automated Accessibility Check Can Decide Which accessibility questions resolve from the serialized DOM and computed style, which require judging meaning, and why an engine cannot tell the difference on its own.
- 02 The Accessibility Tree as a Test Oracle How browsers expose roles, names, states, and relations through the accessibility API, and how tests can assert on that projection instead of on markup.
- 03 How an Accessibility Rule Engine Decides The parse, match, evaluate, and report stages inside a checker like axe-core, how rule results are graded, and why the same page yields different reports from different tools.
- 04 Why Zero Violations Is Not a Pass Automated checks cover a minority of success criteria, report a large incomplete queue, and cannot observe interaction, so a clean run bounds failure rather than proving conformance.
- 05 A Sequential Focus Test Protocol How to record and verify tab order, focus visibility, and focus containment, and what tabindex values, inert subtrees, and dialogs do to the sequence.
- 06 Measuring Contrast From Rendered Pixels Why the contrast ratio is simple arithmetic but the colour inputs are not, how effective backgrounds are composited, and what sampling rendered pixels can and cannot settle.
- 07 Verifying What a Screen Reader Actually Says How speech output is produced from accessibility events, why announcement assertions are brittle across reader versions, and which properties can be tested stably.
- 08 Testing Reflow, Zoom, and Text Spacing How the reflow, resize-text, and text-spacing criteria turn layout questions into arithmetic on viewport width, scroll extents, and computed font metrics.
- 09 Contract Tests for Stateful ARIA Widgets A role-specific contract of states, key bindings, and announcements that can be asserted mechanically, and where the ARIA authoring practices stop being testable.
- 10 Conformance Evidence and Defect Triage How to assemble a defensible accessibility claim: sampling by template, recording scope and exceptions, ranking defects by user impact, and keeping the evidence current.