mattermost/e2e-tests/playwright/docs/accessibility/automated_scan_testing.md
sabril 18d2e8da9f
MM-63700 E2E/Playwright: Add accessibility testing guidelines (#33997)
* add accessibility testing guidelines

* use accessibility locators

* address comments
2025-10-15 12:00:01 +08:00

4 KiB

Automated Accessibility Scan Testing Guidelines

Overview

This guide focuses on automated accessibility testing using a simplified, comprehensive approach. Use axe-core's default rule set and scope testing to specific elements, disabling rules only when necessary. This maximizes coverage while minimizing maintenance overhead.

Testing Philosophy

Core Principle: Scope and Analyze

Use axe-core's comprehensive default rule set and scope testing to specific elements.

Benefits:

  • Maximum Coverage - All WCAG 2.1 AA rules by default
  • Low Maintenance - No rule lists to maintain
  • Future-Proof - New axe-core rules automatically included
  • Clear Intent - Disabled rules are explicit and documented

Basic Pattern

test('page accessibility', async ({axe, page}) => {
    const results = await axe.builder(page).analyze(); // All applicable rules

    expect(results.violations).toHaveLength(0);
});
test('component accessibility', async ({axe, page}) => {
    const results = await axe
        .builder(page)
        .include('#element') // Scope to element
        .analyze(); // All applicable rules

    expect(results.violations).toHaveLength(0);
});

What Can Be Automated

Fully Automatable

  • HTML Structure & Semantics - Markup validity, semantic elements, heading hierarchy
  • ARIA Implementation - Attributes, states, relationships, roles
  • Form Accessibility - Labels, validation, error messages
  • Interactive Elements - Focus capability, accessible names
  • Color and Contrast - WCAG AA/AAA contrast ratios
  • Alternative Text - Image alt text presence
  • Table Structure - Headers, captions, relationships
  • Navigation - Skip links, landmarks, focus order

Cannot Be Fully Automated

  • Cognitive Accessibility - Content readability, comprehension
  • Contextual Appropriateness - Alt text quality, meaningful link text
  • Real User Experience - Actual keyboard workflows, screen reader UX
  • Content Quality - Clear instructions, effective error messaging

Testing Patterns

1. Component Testing

test('modal accessibility', async ({axe, page}) => {
    await page.getByRole('button', {name: 'Open modal'}).click();

    const results = await axe.builder(page).include('[role="dialog"]').analyze();

    expect(results.violations).toHaveLength(0);
});

2. State Testing

test('form error state accessibility', async ({axe, page}) => {
    // Test normal state
    let results = await axe.builder(page).include('form').analyze();
    expect(results.violations).toHaveLength(0);

    // Test error state
    await page.getByRole('button', {name: 'Submit'}).click();
    results = await axe.builder(page).include('form').analyze();
    expect(results.violations).toHaveLength(0);
});

Rule Management

When to Disable Rules

Disable Rules Only When:

  • Known UI Framework Limitations - Theme-related contrast issues
  • Context-Inappropriate Rules - Page-level rules in modal dialogs
  • Temporary Workarounds - With clear documentation and tracking

Examples

// Good - Documented limitation
test('dark theme modal', async ({axe, page}) => {
    const results = await axe
        .builder(page)
        .include('[role="dialog"]')
        .disableRules([
            'color-contrast', // TODO: MM-nnn - Color contrast improvement
        ])
        .analyze();
});

// Good - Context on comment
test('modal dialog', async ({axe, page}) => {
    const results = await axe
        .builder(page)
        .include('[role="dialog"]')
        .disableRules([
            'page-has-heading-one', // Not applicable to modals
            'landmark-one-main', // Not applicable to modals
        ])
        .analyze();
});

Don't Disable Rules For:

  • Convenience - "This rule is annoying"
  • Lack of Understanding - "I don't know what this rule does"
  • Time Pressure - "We'll fix it later" (without tracking)