Fix PlcTransportAdapter: unreliable reconnect guard and leaked stream on disconnect

[Copilot]

ConnectAsync had a broken TcpClient reset condition that could throw ObjectDisposedException on _client.Client and never correctly detected a "weird state". DisconnectAsync closed the stream without nulling _stream, leaving DataAvailable, ReadAsync, and WriteAsync operating on a disposed NetworkStream, and used the async/await Task.CompletedTask antipattern.

These fixes are applied to Services.Adapters.PlcTransportAdapter — the implementation registered in DI. The unreferenced Services.Plc.Adapters.PlcTransportAdapter duplicate has been removed, leaving a single source of truth.

Changes

• ConnectAsync: Replaced the unreliable multi-part condition with an early-return on _client?.Connected == true, then unconditionally disposes the previous _stream (setting it to null) and _client before allocating a fresh TcpClient — ensuring clean state on every connection attempt.

• DisconnectAsync: Disposes _stream and immediately sets it to null before tearing down the client. Drops the async modifier; returns Task.CompletedTask directly.

• Removed duplicate: Deleted the unused Services.Plc.Adapters.PlcTransportAdapter class that was never registered in DI and never referenced, eliminating the split source of truth.

// Before — stream never nulled; broken reset condition; pointless async
public async Task DisconnectAsync(CancellationToken cancellationToken = default)
{
    _stream?.Close();          // closed but not nulled
    _client?.Close();
    _client = new TcpClient();
    await Task.CompletedTask;  // antipattern
}

// After
public Task DisconnectAsync(CancellationToken cancellationToken = default)
{
    _stream?.Dispose();
    _stream = null;            // guards ReadAsync/WriteAsync/DataAvailable
    _client?.Dispose();
    _client = new TcpClient();
    return Task.CompletedTask;
}

---

⚡ Quickly spin up Copilot coding agent tasks from anywhere on your macOS or Windows machine with Raycast.

[Copilot]

Pull request overview

This PR aims to harden the PLC TCP transport adapter by ensuring clean reconnect state and preventing use-after-dispose of a NetworkStream during disconnects.

Changes:

• Simplifies ConnectAsync guard logic and forces a fresh TcpClient per connection attempt (disposing prior client/stream).
• Updates DisconnectAsync to dispose and null _stream, dispose the client, and drop the unnecessary async/await Task.CompletedTask pattern.

[Copilot]

This fix is applied to S7Tools.Services.Plc.Adapters.PlcTransportAdapter, but the DI registration currently binds IPlcTransport to S7Tools.Services.Adapters.PlcTransportAdapter (see ServiceCollectionExtensions.cs). As a result, the reconnect/stream-disposal fix may not affect the transport implementation actually used at runtime. Consider either applying the same changes to the registered adapter (recommended) or updating DI/removing the duplicate adapter to ensure there is a single source of truth.

[efargas]

@copilot apply changes based on the comments in this thread

[Copilot]

@copilot apply changes based on the comments in this thread

Fixed in commit dd8ea1f. Applied the same ConnectAsync/DisconnectAsync fixes to Services.Adapters.PlcTransportAdapter — the class registered in DI (services.TryAddTransient<IPlcTransport, Services.Adapters.PlcTransportAdapter>()). Also deleted the unreferenced Services.Plc.Adapters.PlcTransportAdapter duplicate, leaving a single source of truth.