Add regression tests for nullable dynamic properties

[JulianThoms]

Pull Request

The issue or feature being addressed

Fixes #4181.

In v10.2.3, marking a nullable object property as nullable gives its untyped schema a null-only type. Microsoft.OpenApi 2.12.0 serializes that as enum: [null] in OpenAPI 3.0, restricting the property to null and breaking downstream client generation.

The generator fix from #3976 is already on master, but is absent from v10.2.3. This PR adds regression coverage for the direct-property scenario reported in #4181.

Details on the issue fix or feature implementation

• Cover nullable dynamic properties and array items with System.Text.Json (object, JsonDocument, and JsonElement?) and Newtonsoft.Json (object, JToken, JObject, and JArray), using the existing fixture with nullable reference types disabled and default generator options.
• Check that serialized object properties and array items have neither type nor enum in Swagger 2.0 and OpenAPI 3.0/3.1, while preserving the array's type and nullability.
• Reuse the existing schema generation and document serialization helpers.

Validation:

• dotnet pack --configuration Release: passed with no warnings or errors.
• dotnet test --configuration Release: 4,033 passed, zero failed or skipped across .NET 8, 9, and 10, including integration and UI tests.
• The six new System.Text.Json cases also pass with Microsoft.OpenApi 2.12.0 substituted into an isolated test output directory at runtime.
• Restoring the v10.2.3 SetNullable logic in an isolated build makes five of those six cases fail with Microsoft.OpenApi 2.12.0, including the OpenAPI 3.0 enum assertion. The source was restored after the check.

---

PR Checklist

• I have read and followed the Contributing guide.
• I have added tests for my changes where appropriate.
• I have the experience and knowledge necessary to understand, review, and validate all content in this PR.
• This PR has content that I did not fully write myself.
  • I used AI and I have read and followed the AI Agent Usage guidelines.

Note

This PR's code and description were generated with OpenAI Codex. The contributor review checkbox is left for the human author to complete.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 95.19%. Comparing base (341a21d) to head (dc0ce50).
✅ All tests successful. No failed tests found.

Additional details and impacted files

@@           Coverage Diff           @@
##           master    #4182   +/-   ##
=======================================
  Coverage   95.19%   95.19%           
=======================================
  Files         111      111           
  Lines        4165     4165           
  Branches      852      852           
=======================================
  Hits         3965     3965           
  Misses        200      200           

Flag	Coverage Δ
Linux	95.19% <ø> (ø)
Windows	95.19% <ø> (ø)
macOS	95.19% <ø> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[Copilot]

Copilot review overview

🟢 Approval recommended

The focused tests accurately cover the reported regression and serializer parity without changing production behavior.

Review effort: Balanced
Findings: None

What changed in this PR

Adds regression coverage ensuring nullable dynamic JSON properties remain untyped across serializers and OpenAPI versions.

Changes:

• Tests System.Text.Json dynamic properties, array items, and serialization.
• Adds equivalent Newtonsoft.Json schema-generation coverage.

File	Description
JsonSerializerSchemaGeneratorTests.cs	Covers nullable dynamic schemas and OpenAPI 2.0/3.0/3.1 output.
NewtonsoftSchemaGeneratorTests.cs	Covers nullable Newtonsoft dynamic types and arrays.

---

💡 Configure MCP servers for context-aware, tailored reviews. Learn more in the docs.