Affected tests: follow NuGet package references; narrow Directory.Packages.props changes

CHANGELOG.md

@@ -19,6 +19,28 @@ All notable changes to DotNetDevMCP are documented here. The format follows
   that forwards the original `Origin`, or a non-browser client that happens to set one. `--http` still has no
   authentication or TLS and still shouldn't be exposed beyond localhost.
 
+### Added
+- `dotnet_test_affected`'s project fallback now follows NuGet package references, not just `ProjectReference`s: when a
+  test project's restored `obj/project.assets.json` references another solution project's package id (a literal
+  `<PackageId>`, one from the nearest `Directory.Build.props`, or the assembly name), that test project is selected
+  even with no `ProjectReference` between them. A change to `Directory.Packages.props` is narrowed, via the same
+  assets data, to the test projects that use the package ids whose version actually moved (diffed against git) -
+  walking from EVERY solution project whose restored assets reference a changed id, not just test projects', so a
+  `PrivateAssets="all"` package (an analyzer or source generator) a source project consumes directly is still
+  tracked to the test projects that reference that source project - but only when `Directory.Packages.props` is the
+  *only* changed file, only `Version` attributes actually changed (any other edit to the file, even alongside a
+  version bump, is treated as "beyond package versions" and not narrowed), and every solution project has been
+  restored; helper libraries with no test method are never selected, and a narrowed set that turns out to cover
+  every runnable test project runs the whole solution in one invocation instead. Any other case runs the whole
+  solution, with a note explaining why (git unavailable, the file doesn't parse, something beyond package versions
+  changed, an unrestored project, or no restored project uses those ids). Test projects with no
+  `obj/project.assets.json` (not restored) are called out in the note when the run falls back to whole test projects.
+
+### Fixed
+- `dotnet_test_affected`: a changed `.txt` or image inside a project folder (test data such as `TestData/expected.txt` or
+  Verify's `*.verified.txt`) was ignored, so nothing ran. It now selects that project's tests. `.md` files, and
+  documentation outside every project, are still ignored; a project at the solution root doesn't make docs count.
+
 ## [0.3.3] - 2026-09-24
 
 Prompted by an external evaluation; each claim was checked against the code first.

README.md

@@ -126,7 +126,7 @@ Measured with BenchmarkDotNet on an i7-10750H, .NET 10.0.9. The orchestration be
 
 After an edit, the agent usually reruns the whole suite. `dotnet_test_affected` asks Roslyn instead: take the symbols declared in the changed files, follow references (up to `maxDepth` hops, default 8) until you land in a method with `[Fact]`, `[Theory]`, `[Test]`, `[TestCase]` or `[TestMethod]`, then run exactly those. Changed files default to the git working tree, or `gitBase: "main"` for a branch. `dryRun: true` lists the tests without running them; `framework: "net10.0"` runs one target framework of multi-targeted test projects.
 
-The walk has a time budget (`maxSelectionSeconds`, default 10). A change to code that everything depends on reaches too much to trace cheaply; then the test projects that reference the changed projects run instead (the whole solution if that's all of them), and the response says so (`selectionComplete: false`, `ranScope`). The same happens when the selection is more than 20% of all tests (`maxSelectedFraction`), where a filtered run is no faster. Changed files the walk can't trace (a `.csproj`, `.razor`, `appsettings.json`, a deleted file) switch to the same project fallback and are listed in `untracedFiles`; a changed `Directory.Build.props`, `global.json` or `.editorconfig` runs the whole solution. You never get a silently partial selection. Runs are killed after `timeoutSeconds` (default 600) so a hanging test cannot hang the agent; the response names the test modules that never finished. `maxDepth: 3` narrows more changes but misses more tests. Works with VSTest and with Microsoft.Testing.Platform (`"test": { "runner": "Microsoft.Testing.Platform" }` in global.json).
+The walk has a time budget (`maxSelectionSeconds`, default 10). A change to code that everything depends on reaches too much to trace cheaply; then the test projects that reference the changed projects run instead (the whole solution if that's all of them), and the response says so (`selectionComplete: false`, `ranScope`). The same happens when the selection is more than 20% of all tests (`maxSelectedFraction`), where a filtered run is no faster. Changed files the walk can't trace (a `.csproj`, `.razor`, `appsettings.json`, a deleted file) switch to the same project fallback and are listed in `untracedFiles`; a changed `Directory.Build.props`, `global.json` or `.editorconfig` runs the whole solution. You never get a silently partial selection. Runs are killed after `timeoutSeconds` (default 600) so a hanging test cannot hang the agent; the response names the test modules that never finished. `maxDepth: 3` narrows more changes but misses more tests. Works with VSTest and with Microsoft.Testing.Platform (`"test": { "runner": "Microsoft.Testing.Platform" }` in global.json). The project fallback also follows restored NuGet package references (a test project whose `obj/project.assets.json` references another solution project's package id, with no `ProjectReference` between them, is still selected). A change to `Directory.Packages.props` is narrowed to the test projects that actually use the package ids whose version moved only when it is the *only* changed file (documentation outside every project aside), only the `Version` attributes actually changed (not a `Condition`, a `GlobalPackageReference`, or any other edit riding along), and every solution project has been restored; any other change to that file, a change alongside another file, or an unrestored project anywhere in the solution runs the whole solution instead, with a note explaining which of those applied.
 
 On this repository, editing `ConcurrentExecutor.cs` selects 22 of 44 tests (the `ConcurrentExecutorTests` plus the `OrchestrationServiceTests` that reach it through `OrchestrationService`). Measured through the MCP tool, build included, i7-10750H:
 
@@ -195,7 +195,7 @@ Built on the official [MCP C# SDK](https://github.com/modelcontextprotocol/cshar
 
 0.3.3. The Roslyn tools are mature (they come from SharpTools). Testing, build, git and orchestration are newer and have been exercised on this repository and a few others; expect rough edges on unusual project layouts. Issues and PRs welcome, see [CONTRIBUTING](https://github.com/csa7mdm/DotNetDevMCP/blob/main/CONTRIBUTING.md).
 
-Known gaps: `dotnet_test_affected` follows C# references only (no reflection, no DI-by-convention, no string-keyed lookups), so a change reached only through those paths will not select the test; use `dryRun` to check what it picks. Calls through an interface or base class are followed. The project fallback follows `ProjectReference`s only: a test project that uses the changed code through a NuGet package is not found. Builds and tests are not sandboxed (see Security). Tests that hang instead of failing are only caught by a run that finishes. Test attribute detection covers xUnit, NUnit and MSTest by attribute name. Past the command-line length limit the filter widens from methods to classes, then to the whole project (more tests, never fewer).
+Known gaps: `dotnet_test_affected` follows C# references only (no reflection, no DI-by-convention, no string-keyed lookups), so a change reached only through those paths will not select the test; use `dryRun` to check what it picks. Calls through an interface or base class are followed. The project fallback follows `ProjectReference`s and NuGet package references: package references are now followed when the test project has been restored; cross-repo consumers are not. Builds and tests are not sandboxed (see Security). Tests that hang instead of failing are only caught by a run that finishes. Test attribute detection covers xUnit, NUnit and MSTest by attribute name. Past the command-line length limit the filter widens from methods to classes, then to the whole project (more tests, never fewer).
 
 ## Credits and license