Fix Framework-Dependent, RuntimeIdentifier-specific tool execution

[baronfel]

Builds on top of #49501, merge that one first.

Fixes #49494

The new multi-RID tool support relies on using the 'runner' set in the tool manifest to properly discover how to launch the app in the tool package. For this to work, the 'runner' type ('dotnet'/'executable') needs to align with the name of the binary inside the tool package we're running ('toolname.dll'/'toolname[.exe]'). These two were mismatched because we were pivoting the detection of which execution mode to use for tools off of the wrong property - if an apphost exists for the tool then we need to use the executable strategy.

A bit of test refactoring made this much easier to verify.

[Copilot]

Pull Request Overview

This PR corrects how the SDK chooses executable vs. framework‐dependent runners based on whether an apphost is used, and refactors tests to make RID‐specific and trimmed tool packaging easier to assert.

• TestToolBuilder:
  • Propagate a RidSpecific flag from various properties and include it in the tool identifier.
  • Add a collectBinlogs option to pack commands and handle RuntimeIdentifiers centrally.
• End-to-end tests:
  • Introduce new tests covering multiple RID‐specific and trimmed tool packaging.
  • Refactor XML helpers and add assertions for trimmed dependencies.
• PackTool.targets:
  • Preserve publish-related properties (SelfContained, Publish*) during restore.
  • Switch ToolCommandRunner to use UseAppHost instead of just SelfContained.

Reviewed Changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 1 comment.

File	Description
test/.../TestToolBuilder.cs	Updated TestToolSettings auto-properties and CreateTestTool
test/.../EndToEndToolTests.cs	Added RID-specific and trimmed packaging tests and helpers
src/.../Microsoft.NET.PackTool.targets	Adjusted publish property conditions and ToolCommandRunner

Comments suppressed due to low confidence (7)

test/Microsoft.DotNet.PackageInstall.Tests/TestToolBuilder.cs:35

• Auto-properties cannot reference field. You need an explicit backing field (e.g. _nativeAot) and use that in the getter/setter.

            public bool NativeAOT { get; set { field = value; this.RidSpecific = value; } } = false;

test/Microsoft.DotNet.PackageInstall.Tests/TestToolBuilder.cs:36

• Same issue here—auto-properties don't expose field. Introduce a private backing field for SelfContained.

            public bool SelfContained { get; set { field = value; this.RidSpecific = value; } } = false;

test/Microsoft.DotNet.PackageInstall.Tests/TestToolBuilder.cs:37

• This setter also uses field. Define an explicit backing field (e.g. _trimmed) before referencing it.

            public bool Trimmed { get; set { field = value; this.RidSpecific = value; } } = false;

test/Microsoft.DotNet.PackageInstall.Tests/TestToolBuilder.cs:38

• The IncludeAnyRid setter still refers to field. You must back these auto-properties with a private field.

            public bool IncludeAnyRid { get; set { field = value; this.RidSpecific = value; } } = false;

test/Microsoft.DotNet.PackageInstall.Tests/EndToEndToolTests.cs:168

• In this test you reference expectedRids later but never define it. Add var expectedRids = ToolsetInfo.LatestRuntimeIdentifiers.Split(';'); before using it.

        public void PackagesMultipleTrimmedToolsWithASingleInvocation()

src/Tasks/Microsoft.NET.Build.Tasks/targets/Microsoft.NET.PackTool.targets:192

• MSBuild conditions must compare strings. Change the condition to Condition="'$(UseAppHost)' != 'true'" so it parses correctly.

    <ToolCommandRunner Condition="!$(UseAppHost)">dotnet</ToolCommandRunner>

src/Tasks/Microsoft.NET.Build.Tasks/targets/Microsoft.NET.PackTool.targets:193

• This should read Condition="'$(UseAppHost)' == 'true'" to reliably detect when an apphost is used.

    <ToolCommandRunner Condition="$(UseAppHost)">executable</ToolCommandRunner>