fix: align metadataRouteSuffix exemption list with Next.js

[Divkix]

Fixes #963

Summary

• Removed extra robots and manifest exemptions from metadataRouteSuffix to match getMetadataRouteSuffix and Next.js
• Both functions now only exempt sitemap from hash suffixes
• The extra exemptions were latent (robots/manifest are root-only), but the inconsistency could cause bugs if nestable ever changes

Test plan

• Existing tests/metadata-routes.test.ts (69 tests) pass
• Existing tests/app-router.test.ts metadata tests pass

[pkg-pr-new]

Open in StackBlitz

npm i https://pkg.pr.new/vinext@969

commit: c99f4ae

[james-elicx]

/bigbonk review

[ask-bonk]

LGTM. The change is correct and the reasoning is sound.

The old robots and manifest exemptions in metadataRouteSuffix were dead code — both types have nestable: false in METADATA_FILE_MAP, so the scan loop at line 527 (if (!config.nestable && urlPrefix !== "") continue) already ensures they only appear at root with an empty parentSegments. An empty array means hasInvisibleParent is always false, so the function would have returned "" anyway before reaching the hash computation.

Removing the dead exemptions to match getMetadataRouteSuffix (which only exempts sitemap, per the Next.js port) is the right call. One minor comment on the updated wording.

[ask-bonk]

Nit: the comment says "we keep the exemption list matching getMetadataRouteSuffix" — but the whole point of this PR is that we changed the exemption list to match. The comment reads as if we intentionally kept something unchanged. Consider rewording:

Suggested change

	// matching getMetadataRouteSuffix for defensive consistency).
	// matching getMetadataRouteSuffix and Next.js for consistency).

[ask-bonk]

Review submitted. Here's a summary:

Approved with one minor comment suggestion.

The PR correctly removes dead robots and manifest exemptions from metadataRouteSuffix to align it with getMetadataRouteSuffix. The exemptions were unreachable because both types have nestable: false, meaning they only appear at root where parentSegments is always empty — so hasInvisibleParent would be false regardless, and the function would already return "" without hitting the exemption check.

Left one nit on the comment wording at line 467, suggesting a small tweak for clarity.

github run