Extract rule: template-require-media-caption

[NullVoxPopuli]

Split from #2371.

[NullVoxPopuli-ai-agent]

Review: template-require-media-caption (PR #2610)

Comparison with ember-template-lint source

General correctness:

1. Error message difference: The original ETL rule uses a single error message: "Media elements such as <audio> and <video> must have a <track> for captions." The ESLint version uses a parameterized missingTrack message with {{tag}}. This is fine and arguably better, but worth noting the divergence.

2. Muted attribute handling: The original ETL checks ['MustacheStatement', 'BlockStatement'].includes(mutedAttribute.value.type) to allow dynamic muted values, and separately checks ![false, 'false'].includes(mutedAttribute.value.chars). The ESLint version handles this differently with explicit type checks. The logic appears equivalent, but there's a subtle difference: the ETL code treats muted="false" (string) as NOT muted and falls through, which is correctly replicated. However, the ETL code does NOT check for BooleanLiteral(false) specifically — it relies on MustacheStatement type check to pass through any mustache expression. The ESLint version adds special handling for muted={{false}} (GlimmerBooleanLiteral false), which is a good enhancement.

3. Missing hasChildren check: The original ETL checks !AstNodeInfo.hasChildren(node) before checking for tracks, meaning an element with no children at all is also flagged. The ESLint version's node.children?.some(...) returns false/undefined for empty children, which correctly produces the same result — elements with no children are flagged. Good.

4. Track kind check with dynamic values: The ESLint version only checks kindAttr.value?.type === 'GlimmerTextNode' for the kind attribute on <track>. This means <track kind={{captionType}} /> would NOT be recognized as valid. The original ETL similarly only checks kindAttribute.value.chars, so this matches the original behavior.

Scope analysis (gjs/gts):

Not applicable. This rule only checks HTML element tags (audio, video, track) and attribute values — no helper/component name matching by string is involved. No scope analysis needed.

Tests: Good coverage of both gjs and hbs modes, including muted attribute edge cases.

Overall, this looks like a solid migration. No blocking issues found.

🤖 Automated review comparing with ember-template-lint source