Fix intermittent AssemblyLoading_MultipleVersions_ExactAndGreaterMatch(LoadStream) failure by Copilot · Pull Request #83044 · dotnet/roslyn

Copilot · 2026-04-02T22:43:46Z

InvokeUtil.Exec snapshots CompilerLoadContext.Assemblies before running the test, then asserts the set is unchanged afterward. AssertEx has static fields that lazily initialize DiffPlex on first use; when that first use fell between the snapshot and the final comparison, DiffPlex (and Microsoft.Win32.Registry as a netstandard2.0 compat shim on .NET 10) appeared only in the post-test list, failing the SetEqual check. Only LoadStream was affected because the other variants ran in process contexts where a prior test had already triggered AssertEx initialization.

Changes

InvokeUtil.cs: Call RuntimeHelpers.RunClassConstructor(typeof(AssertEx).TypeHandle) before taking the compilerContextAssemblies snapshot, ensuring DiffPlex and any compat-shim assemblies it pulls in are already present in both sides of the comparison.

// Before (snapshot taken before AssertEx/DiffPlex are initialized)
var loader = new AnalyzerAssemblyLoader(...);
var compilerContextAssemblies = loader.CompilerLoadContext.Assemblies.SelectAsArray(...);

// After (force initialization first so the snapshot includes infrastructure assemblies)
var loader = new AnalyzerAssemblyLoader(...);
RuntimeHelpers.RunClassConstructor(typeof(AssertEx).TypeHandle);
var compilerContextAssemblies = loader.CompilerLoadContext.Assemblies.SelectAsArray(...);

Microsoft Reviewers: Open in CodeFlow

…okeUtil.Exec Agent-Logs-Url: https://github.com/dotnet/roslyn/sessions/673c28d8-d3e5-490d-b4e2-caff66247118 Co-authored-by: jaredpar <146967+jaredpar@users.noreply.github.com>

jaredpar · 2026-04-02T22:45:39Z

@RikkiGibson, @333fred, @jjonescz PTAL

In another PR I hit a flaky failure in this test method. Decided to see what CCA could do with it and gave it the following prompt:

When running our unit tests on .NET core we hit the following failure in CI:

AnalyzerAssemblyLoaderTests.AssemblyLoading_MultipleVersions_ExactAndGreaterMatch(kind: LoadStream)
The information on the failure is

Code
The failure is an AssertEx.SetEqual assertion. The diff shows 2 unexpected assemblies loaded in the LoadStream variant (marked 
++>):

 1. DiffPlex, Version=1.7.2.0 — loaded early in the list
 2. Microsoft.Win32.Registry, Version=10.0.0.0 — also unexpectedly present

The LoadDirect and ShadowLoad variants of the same test passed — only LoadStream failed.
This failure is in InvokeUtil.cs:line at line 75.

This indicates that some code in our unit tests is loading DiffPlex and Microsoft.Win32.Registry during that test execution. What could be causing that to happen?

Was really curious if it could reason out a race condition around AssemblyLoadContext. Turns out it can.

RikkiGibson

LGTM, just had another conversation going on while I was in the middle of looking up AssertEx.cs.

RikkiGibson · 2026-04-16T00:03:24Z

heads up that I saw this again for a build on commit 93224f2. I branched this off of main today so I think I have the fix from this PR.

Helix log: https://helixr1107v0xdeko0k025g8.blob.core.windows.net/dotnet-roslyn-refs-pull-83185-merge-acddef87dfd04007b0/workitem_4.Attempt.2/1/console.bc7e4310.log?helixlogtype=result&skoid=8eda00af-b5ec-4be9-b69b-0919a2338892&sktid=72f988bf-86f1-41af-91ab-2d7cd011db47&skt=2026-04-15T23%3A58%3A33Z&ske=2026-04-16T00%3A58%3A33Z&sks=b&skv=2026-02-06&sv=2026-02-06&se=2026-04-16T00%3A58%3A33Z&sr=b&sp=rl&sig=vIktUJ6cbaMvBzn3sdbyZGSo7352aixhyisHX%2BotTSc%3D.

[xUnit.net 00:00:17.64]     Microsoft.CodeAnalysis.UnitTests.AnalyzerAssemblyLoaderTests.PathResolver_MultipleResolvers_CanIntercept_ReturningNull(kind: ShadowLoad) [FAIL]
<trimmed for brevity>
[xUnit.net 00:00:17.65]       Differences:
[xUnit.net 00:00:17.65]       ++> Microsoft.Win32.Registry, Version=10.0.0.0, Culture=neutral, 
[xUnit.net 00:00:17.65]       Stack Trace:
[xUnit.net 00:00:17.65]         /_/src/Compilers/Test/Core/Assert/AssertEx.cs(483,0): at Roslyn.Test.Utilities.AssertEx.SetEqual[T](IEnumerable`1 expected, IEnumerable`1 actual, IEqualityComparer`1 comparer, String message, String itemSeparator, Func`2 itemInspector)
[xUnit.net 00:00:17.65]         /_/src/Compilers/Test/Core/Assert/AssertEx.cs(394,0): at Roslyn.Test.Utilities.AssertEx.SetEqual(IEnumerable`1 expected, IEnumerable`1 actual, IEqualityComparer`1 comparer, String message, String itemSeparator, Func`2 itemInspector)
[xUnit.net 00:00:17.65]         /_/src/Compilers/Core/CodeAnalysisTest/InvokeUtil.cs(84,0): at Microsoft.CodeAnalysis.UnitTests.InvokeUtil.Exec(ITestOutputHelper testOutputHelper, ImmutableArray`1 pathResolvers, ImmutableArray`1 assemblyResolvers, AssemblyLoadTestFixture fixture, AnalyzerTestKind kind, String typeName, String methodName, Object state)
[xUnit.net 00:00:17.65]         /_/src/Compilers/Core/CodeAnalysisTest/AnalyzerAssemblyLoaderTests.cs(122,0): at Microsoft.CodeAnalysis.UnitTests.AnalyzerAssemblyLoaderTests.Run(AnalyzerTestKind kind, Object state, MethodInfo method, ImmutableArray`1 pathResolvers, ImmutableArray`1 assemblyResolvers)
[xUnit.net 00:00:17.65]         /_/src/Compilers/Core/CodeAnalysisTest/AnalyzerAssemblyLoaderTests.cs(94,0): at Microsoft.CodeAnalysis.UnitTests.AnalyzerAssemblyLoaderTests.Run(AnalyzerTestKind kind, Action`2 testAction, ImmutableArray`1 pathResolvers, ImmutableArray`1 assemblyResolvers)
[xUnit.net 00:00:17.65]         /_/src/Compilers/Core/CodeAnalysisTest/AnalyzerAssemblyLoaderTests.cs(1751,0): at Microsoft.CodeAnalysis.UnitTests.AnalyzerAssemblyLoaderTests.PathResolver_MultipleResolvers_CanIntercept_ReturningNull(AnalyzerTestKind kind)
[xUnit.net 00:00:17.65]            at InvokeStub_AnalyzerAssemblyLoaderTests.PathResolver_MultipleResolvers_CanIntercept_ReturningNull(Object, Span`1)
[xUnit.net 00:00:17.65]            at System.Reflection.MethodBaseInvoker.InvokeWithOneArg(Object obj, BindingFlags invokeAttr, Binder binder, Object[] parameters, CultureInfo culture)

I have a branch on the side where I'm trying to solve. Have to focus on other things for now though.

I am wondering if we should, instead of asserting that entire state of the world remains exactly the same before/after, just ensure that assemblies that are "part of the fixture" are not leaking in. As it seems like any work occurring ambiently in the process could perhaps cause new assemblies to be loaded in this context or the default context over time.

AlekseyTs · 2026-04-16T10:41:47Z

heads up that I saw this again for a build on commit

I think this test should do its work in isolation, i.e. in an environment where other tests cannot interfere. And it should not be snapshotting assemblies loaded in the current test process. Perhaps it should execute the scenario in a separate process.

Force AssertEx static init before CompilerLoadContext snapshot in Inv…

abb985b

…okeUtil.Exec Agent-Logs-Url: https://github.com/dotnet/roslyn/sessions/673c28d8-d3e5-490d-b4e2-caff66247118 Co-authored-by: jaredpar <146967+jaredpar@users.noreply.github.com>

Copilot AI assigned Copilot and jaredpar Apr 2, 2026

Copilot created this pull request from a session on behalf of jaredpar April 2, 2026 22:43 View session

jaredpar marked this pull request as ready for review April 2, 2026 22:43

jaredpar requested a review from a team as a code owner April 2, 2026 22:43

github-actions Bot added the Area-Compilers label Apr 2, 2026

333fred approved these changes Apr 3, 2026

View reviewed changes

jaredpar merged commit f6b7abf into main Apr 3, 2026
26 checks passed

jaredpar deleted the copilot/fix-analyzer-assembly-loader-tests branch April 3, 2026 00:16

dotnet-policy-service Bot added this to the Next milestone Apr 3, 2026

RikkiGibson reviewed Apr 3, 2026

View reviewed changes

dotnet-maestro Bot mentioned this pull request Apr 3, 2026

[release/11.0.1xx-preview3] Source code updates from dotnet/roslyn dotnet/dotnet#5858

Merged

dotnet-bot mentioned this pull request Apr 8, 2026

[Automated] PRs inserted in VS build main-11707.340 #83096

Closed

dotnet-maestro Bot mentioned this pull request Apr 10, 2026

[main] Source code updates from dotnet/roslyn dotnet/dotnet#5962

Merged

RikkiGibson mentioned this pull request Apr 20, 2026

Pre-load lazily-loaded assemblies before assembly snapshots #83247

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Fix intermittent AssemblyLoading_MultipleVersions_ExactAndGreaterMatch(LoadStream) failure#83044

Fix intermittent AssemblyLoading_MultipleVersions_ExactAndGreaterMatch(LoadStream) failure#83044
jaredpar merged 1 commit intomainfrom
copilot/fix-analyzer-assembly-loader-tests

Copilot AI commented Apr 2, 2026 •

edited by dotnet-policy-service Bot

Loading

Uh oh!

jaredpar commented Apr 2, 2026

Uh oh!

Uh oh!

RikkiGibson left a comment

Uh oh!

RikkiGibson commented Apr 16, 2026

Uh oh!

AlekseyTs commented Apr 16, 2026 •

edited

Loading

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

Conversation

Copilot AI commented Apr 2, 2026 • edited by dotnet-policy-service Bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Changes

Microsoft Reviewers: Open in CodeFlow

Uh oh!

jaredpar commented Apr 2, 2026

Uh oh!

Uh oh!

RikkiGibson left a comment

Choose a reason for hiding this comment

Uh oh!

RikkiGibson commented Apr 16, 2026

Uh oh!

AlekseyTs commented Apr 16, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

Copilot AI commented Apr 2, 2026 •

edited by dotnet-policy-service Bot

Loading

AlekseyTs commented Apr 16, 2026 •

edited

Loading