fix Macro error: expected Expr #19168

[asukaminato0721]

fix #19168

now tags every Fragment::Tokens with a TokensOrigin (raw token vs parsed AST) so we can keep track of which fragments need to stay wrapped as a single tt chunk during transcription.

records that ty/pat/pat_param/stmt/block/meta/item/vis metavars originate from AST parses while ident/tt/lifetime/literal stay raw, so bindings carry the right origin metadata from the moment they’re matched.

respects the new origin flag: ${concat} handling destructures the updated variant, and expand_var now leaves AST-derived fragments enclosed (preventing the stray 'static token that used to cause expected Expr), while raw fragments keep the previous strip-logic.

add test

[Copilot]

Pull request overview

This PR fixes a macro expansion bug (#19168) where AST-parsed metavariable fragments (like $t:ty) were being incorrectly stripped of invisible delimiters during transcription, causing individual tokens (e.g., 'static) to leak out and trigger "expected Expr" errors.

Key Changes:

• Introduced TokensOrigin enum to distinguish between raw token fragments and AST-parsed fragments
• Updated Fragment::Tokens from a tuple variant to a struct variant with an origin field
• Modified matcher logic to tag fragments with appropriate origin during matching
• Updated transcriber to respect origin metadata and skip stripping invisible delimiters for AST-derived fragments

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 2 comments.

File	Description
crates/mbe/src/expander.rs	Adds TokensOrigin enum and converts Fragment::Tokens to struct variant with origin tracking
crates/mbe/src/expander/matcher.rs	Tags matched fragments with appropriate origin (Raw for ident/tt/lifetime/literal, Ast for ty/pat/stmt/block/meta/item/vis)
crates/mbe/src/expander/transcriber.rs	Conditionally applies strip_invisible() based on fragment origin during expansion
crates/hir-def/src/macro_expansion_tests/mbe/regression.rs	Adds regression test for ty fragment transcription preserving structure

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

The TokensOrigin enum should have documentation explaining the distinction between Raw and Ast variants. Consider adding a doc comment like:

/// Tracks whether token fragments come from raw tokens or parsed AST.
/// This distinction is important during transcription: AST-derived fragments
/// should not have their invisible delimiters stripped to preserve their
/// structure as complete syntactic units.

This would help future maintainers understand the purpose of this enum and when to use each variant.

[Copilot]

The documentation comment for the Tokens variant should be updated to explain the new origin field. The current comment "token fragments are just copy-pasted into the output" is incomplete. Consider updating it to:

/// Token fragments are transcribed into the output. The `origin` field tracks whether
/// these tokens come from raw input (ident/tt/lifetime/literal metavars) or parsed AST
/// (ty/pat/stmt/block/etc metavars). AST-derived fragments preserve their structure
/// during transcription by not stripping invisible delimiters.

This provides context for the new field and explains why it matters.