Disposing of a CopilotSession after stopping the client results in exception

[stephentoub]

Opus wrote code that explicitly issued the StopAsync, which is how I noticed. Presumably the StopAsync isn't actually necessary, and that the client ideally shouldn't be stopped while there are active sessions, but disposal also should ideally be tolerant to such things.

Repro:

using GitHub.Copilot.SDK;

await using var client = new CopilotClient();
await using var session = await client.CreateSessionAsync();

await client.StopAsync();

results in this exception from the DisposeAsync from the await using cleanup:

Unhandled exception. System.ObjectDisposedException: Cannot access a disposed object.
Object name: 'StreamJsonRpc.JsonRpc'.
   at Microsoft.Verify.NotDisposed(IDisposableObservable disposedValue, String message)
   at StreamJsonRpc.JsonRpc.InvokeCoreAsync[TResult](RequestId id, String targetName, IReadOnlyList`1 arguments, IReadOnlyList`1 positionalArgumentDeclaredTypes, IReadOnlyDictionary`2 namedArgumentDeclaredTypes, CancellationToken cancellationToken, Boolean isParameterObject)
   at GitHub.Copilot.SDK.CopilotClient.InvokeRpcAsync[T](JsonRpc rpc, String method, Object[] args, CancellationToken cancellationToken)
   at GitHub.Copilot.SDK.CopilotSession.DisposeAsync()
   at Program.<Main>$(String[] args)

[github-actions]

Thank you for reporting this issue! I've labeled this as a bug and sdk/dotnet.

This is a clear disposal ordering issue where CopilotSession.DisposeAsync() attempts to invoke RPC methods on a JsonRpc instance that has already been disposed by StopAsync(). The session disposal should be tolerant to the client being stopped first.

This has been triaged and the team will investigate the proper fix for handling disposal after the client is stopped.

AI generated by Issue Triage Agent