Evidence policy

Testing methodology

Every conclusion begins with an original minimal reproduction and ends with a reviewable test record.

01

Build fixtures

Create failing and fixed versions with the same original content.

02

Define assertions

Record the expected bounds, scroll widths, focus visibility, or layout shifts.

03

Run the matrix

Execute controlled viewports in Chromium, Firefox, and WebKit.

04

Review evidence

Inspect JSON, screenshots, sources, limitations, and revision dates.

What a test can prove

A test can show that a specific fixture produced a measured result in a recorded environment. It does not prove that one CSS declaration fixes every site, browser version, assistive technology, or content shape.

Fixture rules

Fixtures use code and content written for this project. Failing and fixed states differ only where needed to isolate the mechanism. Third-party pages, screenshots, tutorials, and proprietary code are not copied into the evidence set.

Measurements

Overflow cases record viewport, document and key element dimensions. Reflow and overlay cases add focus visibility and occlusion checks. Layout-shift cases capture browser performance entries and the exact action that triggered the change.

Test environment

A formal run records the date, operating system, Playwright version, browser versions, viewport, device pixel ratio, color scheme, locale, fonts, and fixture revision. Results from different environments are not silently merged.

Corrections and updates

When evidence or standards change, the case keeps a revision entry describing the old conclusion, the new result, and the reason for the change. A current date is never added solely to make content appear fresh.