fix(engine): emit ECMA-335 metadata-format type names in TestMethodIdentifierProperty

[thomhurst]

Problem

Microsoft.Testing.Platform specifies that TestMethodIdentifierProperty values are ECMA-335 compliant (managed name format, vstest RFC 0017). TUnit was passing raw reflection Type.FullName for ParameterTypeFullNames and ReturnTypeFullName, which produces:

• Assembly-qualified generic arguments: System.Collections.Generic.List1[[System.String, System.Private.CoreLib, Version=...]]`
• null (via FullName!) for open generic method parameters (T)

Fix

New MetadataTypeNameFormatter (cached per Type) emits the spec format, matching what MSTest's ManagedNameParser produces:

• Constructed generics: System.Collections.Generic.List1<System.String>`
• Generic method parameters: !!0; generic type parameters: !0
• Arrays ([], [,], jagged), byref &, pointer *
• Nested types via +

Deliberately unchanged

TypeName keeps TUnit's class-argument decoration (MyTests(arg1, arg2)) for parameterized test classes — this keeps class-data-source instances distinguishable in Test Explorer. Nested-type + joining, namespace split, and generic backtick arity were already spec-conformant.

No internal consumers read ParameterTypeFullNames/ReturnTypeFullName (reporters only read TypeName/MethodName/Namespace), so no downstream output changes.

Tests

15 new tests in MetadataTypeNameFormatterTests covering simple types, constructed/open generics, nested generics, arrays, nullable, nested types, !!n/!n placeholders, byref.

[claude]

Code review

Reviewed the diff, cross-checked the new MetadataTypeNameFormatter logic against the ECMA-335/vstest RFC-0017 spec described in the PR, and had two independent passes hunt for bugs/security issues in just the introduced code. No issues found.

Notes from the review:

• The recursive formatter correctly handles constructed generics, generic method (!!N) vs. type (!N) parameters, arrays (including jagged/multi-rank), byref (&), pointer (*), and nested types (+) — verified by tracing the logic against all 15 new test cases in MetadataTypeNameFormatterTests.cs.
• The returnTypeFullName change correctly preserves the old System.Void fallback behavior for void returns, while fixing the previous bug where an open generic return type (T) would silently fall back to "System.Void" via the ?? operator instead of producing a spec-correct !!0.
• ParameterTypeFullNames/ReturnTypeFullName are computed once downstream of TestDetails, which is populated uniformly by both source-gen and reflection discovery — so this satisfies the dual-mode requirement without needing separate source-gen logic.
• Caching via ConcurrentDictionary<Type, string> + StringBuilderPool is thread-safe and consistent with the existing TypeNameFormatter pattern in TUnit.Core.Helpers.

Checked for bugs and CLAUDE.md compliance (dual-mode metadata collection, snapshot testing, AOT compatibility, performance).

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 772139f923

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Use TypeInfo for source-gen generic parameters

In source-generated metadata, parameters that depend on generic arguments are stored with Type == typeof(object) and the real signature in TypeInfo (for example the generated snapshots create IEnumerable<TSource> as typeof(object) plus ConstructedGeneric). Because this new formatting path only reads parameters[i].Type, source-gen generic tests still publish System.Object instead of metadata names like !!0 or System.Collections.Generic.IEnumerable1<!!0>`, while reflection mode now publishes the correct placeholders; that leaves MTP method identity/source navigation wrong for those generic tests and violates the dual-mode requirement.

Useful? React with 👍 / 👎.

[codacy-production]

Not up to standards ⛔

🔴 Issues 3 medium · 1 minor

Alerts:
⚠ 4 issues (≤ 0 issues of at least minor severity)

Results:
4 new issues

Category	Results
UnusedCode	1 minor
BestPractice	3 medium

View in Codacy

🟢 Metrics 19 complexity

Metric	Results
Complexity	19

View in Codacy

_{NEW Get contextual insights on your PRs based on Codacy's metrics, along with PR and Jira context, without leaving GitHub. Enable AI reviewer
TIP This summary will be updated as you push new changes.}