Add regression tests for 8 previously untested bug fixes

[nohwnd]

Summary

Add 25 regression tests (22 unit + 3 integration) across 13 files for 11 previously untested bug fixes.
Each test verifies the specific behavior that was fixed and would fail if the fix were reverted.

Unit Tests

Issue	Component	Tests	What the test verifies
#4461	CommunicationUtilities	2	Socket IOException from child exit not propagated
#2319	TrxLogger	3	ErrorStackTrace settable before ErrorMessage
#5132	TrxLogger	2	WarnOnFileOverwrite parameter parsed correctly
#4243	TrxLogger	3	TestOutcome.Error is value 0, not Min alias
#5184	TestHostProvider	2	Stderr forwarded as Informational, not Error
#2479	TestHostProvider	1	ARM64 on Windows uses DLL hosting
#2483	HtmlLogger	2	Initialize creates results directory
#3136	HtmlLogger	1	Transform handles invalid XML char references
#3454	ObjectModel	1	PortablePdbReader null arg validation
#4424	AdapterUtilities	3	Serialization ctor marked Obsolete on NET8+
#813	vstest.console	2	Code coverage excludes auto-generated code

Integration Tests

Issue	Test Asset	What the test verifies
#4461	crash	Testhost crash output has no socket stack traces
#5184	StderrOutputProject (new)	Stderr from test doesn't fail the test run
#3136	SpecialCharOutputProject (new)	HTML logger handles special chars without XmlException

[Copilot]

Pull request overview

Adds regression coverage for a set of previously untested bug fixes across multiple components (TestHostProvider, CommunicationUtilities, TrxLogger, HtmlLogger, ObjectModel), ensuring the fixed behaviors remain protected from regressions.

Changes:

• Introduces new “RegressionBugFixTests” test classes in 5 unit-test projects.
• Adds targeted assertions for specific historical bugs (stderr forwarding level, ARM64 hosting selection, socket IO exception propagation, TRX logger behaviors, HtmlLogger directory creation, PortablePdbReader null validation).

Reviewed changes

Copilot reviewed 5 out of 5 changed files in this pull request and generated 9 comments.

Show a summary per file

File	Description
test/Microsoft.TestPlatform.TestHostProvider.UnitTests/Hosting/RegressionBugFixTests.cs	Adds regression tests for stderr forwarding level and ARM64 Windows hosting selection.
test/Microsoft.TestPlatform.ObjectModel.UnitTests/RegressionBugFixTests.cs	Adds regression coverage for PortablePdbReader constructor null-argument validation.
test/Microsoft.TestPlatform.Extensions.TrxLogger.UnitTests/RegressionBugFixTests.cs	Adds regression tests for TRX error info ordering, WarnOnFileOverwrite parsing, and TestOutcome enum semantics.
test/Microsoft.TestPlatform.Extensions.HtmlLogger.UnitTests/RegressionBugFixTests.cs	Adds regression tests ensuring HtmlLogger.Initialize creates the results directory.
test/Microsoft.TestPlatform.CommunicationUtilities.UnitTests/RegressionBugFixTests.cs	Adds regression tests around socket/IO exception propagation behavior in MessageLoopAsync.

[Copilot]

The GH-2479 test setup doesn’t appear to force the precondition where testhost.exe is discoverable, so it may pass even if the regression returns (e.g., if the implementation can’t find testhost.exe and falls back to dotnet + testhost.dll regardless). Also, the comments describe verifying the inverse on non-ARM, but the mocks set Architecture to ARM64. To make this a true regression test: (1) explicitly mock the PROCESSOR_ARCHITECTURE machine-level lookup via IEnvironmentVariableHelper to an ARM value to deterministically exercise the guard, and (2) configure IFileHelper.Exists(...) to return true for the computed testhost.exe path (or via a predicate matching ...EndsWith(\"testhost.exe\")) so the code path would choose it if the guard were removed. Update the comments to match the actual scenario being tested.

[Copilot]

The GH-2479 test setup doesn’t appear to force the precondition where testhost.exe is discoverable, so it may pass even if the regression returns (e.g., if the implementation can’t find testhost.exe and falls back to dotnet + testhost.dll regardless). Also, the comments describe verifying the inverse on non-ARM, but the mocks set Architecture to ARM64. To make this a true regression test: (1) explicitly mock the PROCESSOR_ARCHITECTURE machine-level lookup via IEnvironmentVariableHelper to an ARM value to deterministically exercise the guard, and (2) configure IFileHelper.Exists(...) to return true for the computed testhost.exe path (or via a predicate matching ...EndsWith(\"testhost.exe\")) so the code path would choose it if the guard were removed. Update the comments to match the actual scenario being tested.

[Copilot]

The GH-2479 test setup doesn’t appear to force the precondition where testhost.exe is discoverable, so it may pass even if the regression returns (e.g., if the implementation can’t find testhost.exe and falls back to dotnet + testhost.dll regardless). Also, the comments describe verifying the inverse on non-ARM, but the mocks set Architecture to ARM64. To make this a true regression test: (1) explicitly mock the PROCESSOR_ARCHITECTURE machine-level lookup via IEnvironmentVariableHelper to an ARM value to deterministically exercise the guard, and (2) configure IFileHelper.Exists(...) to return true for the computed testhost.exe path (or via a predicate matching ...EndsWith(\"testhost.exe\")) so the code path would choose it if the guard were removed. Update the comments to match the actual scenario being tested.

[Copilot]

The GH-2479 test setup doesn’t appear to force the precondition where testhost.exe is discoverable, so it may pass even if the regression returns (e.g., if the implementation can’t find testhost.exe and falls back to dotnet + testhost.dll regardless). Also, the comments describe verifying the inverse on non-ARM, but the mocks set Architecture to ARM64. To make this a true regression test: (1) explicitly mock the PROCESSOR_ARCHITECTURE machine-level lookup via IEnvironmentVariableHelper to an ARM value to deterministically exercise the guard, and (2) configure IFileHelper.Exists(...) to return true for the computed testhost.exe path (or via a predicate matching ...EndsWith(\"testhost.exe\")) so the code path would choose it if the guard were removed. Update the comments to match the actual scenario being tested.

[Copilot]

GetTestEngineDirectory() is being set to a full executable path (c:\\tmp\\dotnet.exe) rather than a directory path. If the production code combines this value with filenames, it can generate invalid paths (and reduce the test’s fidelity). Return an actual directory (e.g., c:\\tmp or a temp directory) from GetTestEngineDirectory().

Suggested change

	var dotnetPath = @"c:\tmp\dotnet.exe";
	mockEnvironment.SetupGet(e => e.Architecture).Returns(PlatformArchitecture.ARM64);
	mockEnvironment.Setup(ev => ev.OperatingSystem).Returns(PlatformOperatingSystem.Windows);
	mockRunsettingHelper.SetupGet(r => r.IsDefaultTargetArchitecture).Returns(false);
	mockProcessHelper.Setup(ph => ph.GetCurrentProcessFileName()).Returns(dotnetPath);
	mockProcessHelper.Setup(ph => ph.GetTestEngineDirectory()).Returns(dotnetPath);
	mockProcessHelper.Setup(ph => ph.GetCurrentProcessArchitecture()).Returns(PlatformArchitecture.ARM64);
	mockFileHelper.Setup(fh => fh.Exists(testhostDllPath)).Returns(true);
	mockFileHelper.Setup(fh => fh.Exists(dotnetPath)).Returns(true);
	var dotnetExePath = Path.Combine(temp, "dotnet.exe");
	var dotnetDirectory = temp;
	mockEnvironment.SetupGet(e => e.Architecture).Returns(PlatformArchitecture.ARM64);
	mockEnvironment.Setup(ev => ev.OperatingSystem).Returns(PlatformOperatingSystem.Windows);
	mockRunsettingHelper.SetupGet(r => r.IsDefaultTargetArchitecture).Returns(false);
	mockProcessHelper.Setup(ph => ph.GetCurrentProcessFileName()).Returns(dotnetExePath);
	mockProcessHelper.Setup(ph => ph.GetTestEngineDirectory()).Returns(dotnetDirectory);
	mockProcessHelper.Setup(ph => ph.GetCurrentProcessArchitecture()).Returns(PlatformArchitecture.ARM64);
	mockFileHelper.Setup(fh => fh.Exists(testhostDllPath)).Returns(true);
	mockFileHelper.Setup(fh => fh.Exists(dotnetExePath)).Returns(true);

[Copilot]

GetTestEngineDirectory() is being set to a full executable path (c:\\tmp\\dotnet.exe) rather than a directory path. If the production code combines this value with filenames, it can generate invalid paths (and reduce the test’s fidelity). Return an actual directory (e.g., c:\\tmp or a temp directory) from GetTestEngineDirectory().

Suggested change

	mockProcessHelper.Setup(ph => ph.GetTestEngineDirectory()).Returns(dotnetPath);
	mockProcessHelper.Setup(ph => ph.GetTestEngineDirectory()).Returns(temp);

[Copilot]

Test cleanup deletes the directory non-recursively. If Initialize (or future changes) creates any files/subdirectories under nonExistentDir, Directory.Delete(nonExistentDir) will throw and fail the test. Use the recursive overload (and/or ensure the directory is empty) to make the test robust.

Suggested change

	Directory.Delete(nonExistentDir);
	Directory.Delete(nonExistentDir, recursive: true);

[Copilot]

The second socket-based test uses a synchronous Write with no flush and no cancellation token. Depending on timing/buffering, this can make the test more prone to intermittent hangs until CancelAfter triggers. Prefer mirroring the first test’s approach (WriteAsync + FlushAsync using TestContext.CancellationToken) to more reliably trigger the Poll/NotifyDataAvailable path and reduce flakiness.

Suggested change

	serverSide.GetStream().Write(new byte[] { 1 }, 0, 1);
	await serverSide.GetStream().WriteAsync(new byte[] { 1 }, 0, 1, TestContext.CancellationToken);
	await serverSide.GetStream().FlushAsync(TestContext.CancellationToken);

[Copilot]

The second socket-based test uses a synchronous Write with no flush and no cancellation token. Depending on timing/buffering, this can make the test more prone to intermittent hangs until CancelAfter triggers. Prefer mirroring the first test’s approach (WriteAsync + FlushAsync using TestContext.CancellationToken) to more reliably trigger the Poll/NotifyDataAvailable path and reduce flakiness.