Monitoring Tools

The app itself

Accessibility audit

A named checklist for keyboard use, accessible names, target size and contrast, measured on the built app. It runs on the same tour as the Material 3 audit: every page, window and state, in the dark and the light theme (about 340 scans, 16,000 control sightings and 39,000 text sightings per run). Findings go to audit-ui.json.

Code: UI/UiAudit.cs, UI/SelfTest.Audit.cs, UI/AccessibleLabels.cs.

#Checklist

RowHow it is measuredResult
Every control has an accessible nameThe automation peer's name of every visible button, check box, text box, combo box, slider, tab and menu item must not be empty. A name of only symbols is allowed when a text tooltip explains it (the regex token buttons).Pass after two fixes. 431 buttons and 21 inputs had no name. A button whose content is a panel now takes its name from the text it shows; an icon-only button takes its tooltip; an input takes the label before it or its tooltip. The names follow the text, so the language and personal wording reach them.
Keyboard reach: staticEach enabled control must be focusable, a tab stop and have no ancestor that turns Tab off.Pass. A list is entered through its items. The command palette is excluded, see below.
Keyboard reach: real traversalTab is pressed from the first element through the whole window (326 traversals, about 19,000 steps per run) and every enabled control must receive focus.Pass after one real fix: committing the Display name or logo colour field rebuilt the Appearance page, so tabbing out of the field lost focus. The page now keeps its place (Appearance.KeepPage). The Minecraft players list now lets Tab reach the switches inside each row.
Target sizeEvery enabled target at least 24 by 24 (WCAG 2.2, 2.5.8), or a 24 diameter circle on its centre touches no other target.Pass. About 800 small targets pass by the spacing exception; the rest are 32 or larger. Most buttons are 40 high; the 48 by 48 recommendation is not required.
Text contrastText colour against the real backdrop (the nearest opaque fills, including template borders beneath the content) at least 4.5:1, or 3:1 for large text, in both themes. Disabled text and text in the middle of a fade are skipped.Pass. Fixed: muted text on tonal surfaces (the strongest surface now reaches 4.9:1), warning, accent and error text in the light theme, text on selected rows and chips, the date picker box in the dark theme.
Palette contrastEvery text role on every surface and every filled container pair, in all five seeds, light and dark.Pass.
Focus visibleEvery interactive style has a keyboard focus ring.Pass (see the state layer row of the Material 3 audit).
Reduced motion, live regions, text size 90 to 150 percent, 200 percent display scaleExisting features, covered by their own checks and captures.Unchanged.

#Accepted exceptions

  • The command palette rows hold a switch that is not a tab stop on purpose: the palette is driven by Up, Down, Enter and, new in this audit, Ctrl+Space to operate the selected row's switch or Unlock button.
  • A list or combo box with no items has nothing to focus and is not required to be reached.
  • Text in the middle of a fade (opacity under 0.98) is skipped, because its settled colour is what a person reads.

#What the audit does not prove

  • Real assistive technology (Narrator, NVDA) was not run. The names are the automation peer's, which is what they read.
  • Contrast is measured on solid fills. A gradient or picture behind text would be skipped; none was met.
  • Target size does not account for text size scaling beyond the scale factor of the window.

#Verification

MonitoringTools.exe --selftest <folder> runs the checks named accessibility: .... The quick loop is MT_ONLY=audit.

View this article's Markdown source on GitHub

Suggested articles

All The app itself articles