lmxopcua/tests/ZB.MOM.WW.OtOpcUa.Driver.Galaxy.ParityTests/ParityHarness.cs

using System.Diagnostics;
using System.Net.Sockets;
using System.Reflection;
using System.Security.Principal;
using Xunit;
using ZB.MOM.WW.OtOpcUa.Core.Abstractions;
using ZB.MOM.WW.OtOpcUa.Driver.Galaxy.Proxy;

namespace ZB.MOM.WW.OtOpcUa.Driver.Galaxy.ParityTests;

/// <summary>
///     Side-by-side fixture that drives both the legacy <see cref="GalaxyProxyDriver"/>
///     (talking to an out-of-process <c>OtOpcUa.Driver.Galaxy.Host.exe</c>) and the new
///     in-process <see cref="GalaxyDriver"/> (talking to a running <c>mxaccessgw</c>
///     gateway) against the same dev Galaxy. Phase 5 scenario tests use this harness
///     to capture comparable snapshots from each backend.
/// </summary>
/// <remarks>
///     Per-backend availability is independent — a developer running just the legacy
///     Galaxy.Host EXE without an mxaccessgw process up will see the legacy driver
///     resolve and the mxgw driver mark itself unavailable. Each test decides how to
///     handle partial availability:
///     <list type="bullet">
///       <item>Strict-parity tests call <see cref="RequireBoth"/> to skip when either side
///         is missing.</item>
///       <item>Single-backend smoke tests call <see cref="GetDriver"/> for the backend they
///         care about and skip with the recorded <c>SkipReason</c>.</item>
///     </list>
///     Endpoint overrides come from environment variables so dev VMs and the central
///     parity host can target the same suite without touching the test source:
///     <list type="bullet">
///       <item><c>OTOPCUA_PARITY_GW_ENDPOINT</c> — defaults to <c>http://localhost:5120</c>
///         (mxaccessgw <c>launchSettings.json</c> http profile).</item>
///       <item><c>OTOPCUA_PARITY_GW_API_KEY</c> — defaults to <c>parity-suite-key</c>.</item>
///       <item><c>OTOPCUA_PARITY_CLIENT_NAME</c> — defaults to <c>OtOpcUa-Parity</c>.</item>
///     </list>
/// </remarks>
public sealed class ParityHarness : IAsyncLifetime
{
    public enum Backend { LegacyHost, MxGateway }

    private const string LegacySecret = "parity-suite-secret";
    private const string DefaultGwEndpoint = "http://localhost:5120";
    private const string DefaultGwApiKey = "parity-suite-key";
    private const string DefaultClientName = "OtOpcUa-Parity";

    public IDriver? LegacyDriver { get; private set; }
    public string? LegacySkipReason { get; private set; }

    public IDriver? MxGatewayDriver { get; private set; }
    public string? MxGatewaySkipReason { get; private set; }

    private Process? _legacyHost;

    public async ValueTask InitializeAsync()
    {
        if (!OperatingSystem.IsWindows())
        {
            LegacySkipReason = "Windows-only";
            MxGatewaySkipReason = "Windows-only";
            return;
        }

        await InitializeLegacyAsync();
        await InitializeMxGatewayAsync();
    }

    public async ValueTask DisposeAsync()
    {
        // Independent teardown — failure on one side must not prevent the other from
        // releasing its resources (esp. the legacy Host EXE subprocess).
        if (LegacyDriver is not null)
        {
            try { await LegacyDriver.ShutdownAsync(CancellationToken.None); } catch { /* shutdown */ }
            (LegacyDriver as IDisposable)?.Dispose();
            LegacyDriver = null;
        }
        if (_legacyHost is not null && !_legacyHost.HasExited)
        {
            try { _legacyHost.Kill(entireProcessTree: true); } catch { /* ignore */ }
            try { _legacyHost.WaitForExit(5_000); } catch { /* ignore */ }
        }
        _legacyHost?.Dispose();
        _legacyHost = null;

        if (MxGatewayDriver is not null)
        {
            try { await MxGatewayDriver.ShutdownAsync(CancellationToken.None); } catch { /* shutdown */ }
            (MxGatewayDriver as IDisposable)?.Dispose();
            MxGatewayDriver = null;
        }
    }

    /// <summary>Skip the test if either backend isn't available — strict-parity scenarios.</summary>
    public void RequireBoth()
    {
        if (LegacySkipReason is not null) Assert.Skip($"legacy backend unavailable: {LegacySkipReason}");
        if (MxGatewaySkipReason is not null) Assert.Skip($"mxgateway backend unavailable: {MxGatewaySkipReason}");
    }

    /// <summary>Get a backend driver or skip if it's unavailable.</summary>
    public IDriver GetDriver(Backend backend)
    {
        return backend switch
        {
            Backend.LegacyHost when LegacyDriver is not null => LegacyDriver,
            Backend.LegacyHost => SkipAndThrow($"legacy backend unavailable: {LegacySkipReason}"),
            Backend.MxGateway when MxGatewayDriver is not null => MxGatewayDriver,
            Backend.MxGateway => SkipAndThrow($"mxgateway backend unavailable: {MxGatewaySkipReason}"),
            _ => throw new ArgumentOutOfRangeException(nameof(backend), backend, null),
        };
    }

    /// <summary>
    ///     Drive the same closure against every available backend. Tests use the
    ///     returned dictionary to diff snapshots — keys are the backends that
    ///     successfully resolved during <see cref="InitializeAsync"/>. If neither
    ///     resolved, the result is empty and the test should skip.
    /// </summary>
    public async Task<IReadOnlyDictionary<Backend, T>> RunOnAvailableAsync<T>(
        Func<IDriver, CancellationToken, Task<T>> scenario, CancellationToken cancellationToken)
    {
        ArgumentNullException.ThrowIfNull(scenario);
        var results = new Dictionary<Backend, T>();
        if (LegacyDriver is not null)
        {
            results[Backend.LegacyHost] = await scenario(LegacyDriver, cancellationToken).ConfigureAwait(false);
        }
        if (MxGatewayDriver is not null)
        {
            results[Backend.MxGateway] = await scenario(MxGatewayDriver, cancellationToken).ConfigureAwait(false);
        }
        return results;
    }

    [System.Runtime.Versioning.SupportedOSPlatform("windows")]
    private async Task InitializeLegacyAsync()
    {
        if (!await ZbReachableAsync())
        {
            LegacySkipReason = "Galaxy ZB SQL not reachable on localhost:1433";
            return;
        }
        var hostExe = FindLegacyHostExe();
        if (hostExe is null)
        {
            LegacySkipReason = "Galaxy.Host EXE not built — run `dotnet build src/ZB.MOM.WW.OtOpcUa.Driver.Galaxy.Host`";
            return;
        }

        var pipe = $"OtOpcUaGalaxyParity-{Guid.NewGuid():N}";
        using var identity = WindowsIdentity.GetCurrent();
        var sid = identity.User!.Value;

        var psi = new ProcessStartInfo(hostExe)
        {
            UseShellExecute = false,
            CreateNoWindow = true,
            RedirectStandardOutput = true,
            RedirectStandardError = true,
            EnvironmentVariables =
            {
                ["OTOPCUA_GALAXY_PIPE"] = pipe,
                ["OTOPCUA_ALLOWED_SID"] = sid,
                ["OTOPCUA_GALAXY_SECRET"] = LegacySecret,
                // PR 5.W triage 2026-04-30: db-backend is Discover-only. The parity
                // matrix needs Read / Write / Subscribe over a real MxAccess session,
                // so use the mxaccess backend. ZB conn string is still consulted for
                // the discovery path (the mxaccess backend layers MxAccess on top of
                // the same DB).
                ["OTOPCUA_GALAXY_BACKEND"] = "mxaccess",
                ["OTOPCUA_GALAXY_ZB_CONN"] = "Server=localhost;Database=ZB;Integrated Security=True;TrustServerCertificate=True;Encrypt=False;",
            },
        };

        try
        {
            _legacyHost = Process.Start(psi)
                ?? throw new InvalidOperationException("Failed to spawn Galaxy.Host EXE");
            await Task.Delay(2_000); // PipeServer warm-up — ParityFixture's settled value

            var driver = new GalaxyProxyDriver(new GalaxyProxyOptions
            {
                DriverInstanceId = "parity-legacy",
                PipeName = pipe,
                SharedSecret = LegacySecret,
                ConnectTimeout = TimeSpan.FromSeconds(5),
            });
            await driver.InitializeAsync(driverConfigJson: "{}", CancellationToken.None);
            LegacyDriver = driver;
        }
        catch (Exception ex)
        {
            LegacySkipReason = $"legacy backend boot failed: {ex.Message}";
            if (_legacyHost is not null && !_legacyHost.HasExited)
            {
                try { _legacyHost.Kill(entireProcessTree: true); } catch { /* ignore */ }
            }
        }
    }

    private async Task InitializeMxGatewayAsync()
    {
        var endpoint = Environment.GetEnvironmentVariable("OTOPCUA_PARITY_GW_ENDPOINT") ?? DefaultGwEndpoint;
        var apiKey = Environment.GetEnvironmentVariable("OTOPCUA_PARITY_GW_API_KEY") ?? DefaultGwApiKey;
        var clientName = Environment.GetEnvironmentVariable("OTOPCUA_PARITY_CLIENT_NAME") ?? DefaultClientName;

        if (!await GwReachableAsync(endpoint))
        {
            MxGatewaySkipReason = $"mxaccessgw not reachable at {endpoint}";
            return;
        }

        var configJson = $$"""
            {
              "Gateway": {
                "Endpoint": "{{endpoint}}",
                "ApiKeySecretRef": "{{apiKey}}",
                "UseTls": {{(endpoint.StartsWith("https") ? "true" : "false")}}
              },
              "MxAccess": { "ClientName": "{{clientName}}" }
            }
            """;

        try
        {
            var driver = GalaxyDriverFactoryExtensions.CreateInstance("parity-mxgw", configJson);
            await driver.InitializeAsync(configJson, CancellationToken.None);
            MxGatewayDriver = driver;
        }
        catch (Exception ex)
        {
            MxGatewaySkipReason = $"mxgateway backend boot failed: {ex.GetType().Name}: {ex.Message}";
        }
    }

    private static IDriver SkipAndThrow(string reason)
    {
        Assert.Skip(reason);
        throw new UnreachableException(); // Assert.Skip throws SkipException; this satisfies the compiler
    }

    private static async Task<bool> ZbReachableAsync()
    {
        try
        {
            using var client = new TcpClient();
            var task = client.ConnectAsync("localhost", 1433);
            return await Task.WhenAny(task, Task.Delay(1_500)) == task && client.Connected;
        }
        catch { return false; }
    }

    private static async Task<bool> GwReachableAsync(string endpoint)
    {
        // Lightweight TCP probe — avoids spending the full gRPC connect timeout when the
        // gateway just isn't running. We can't validate the API-key handshake here without
        // doing a real RPC, so a successful TCP connect is the "available" signal and any
        // auth/protocol failure surfaces during InitializeAsync below.
        try
        {
            var uri = new Uri(endpoint, UriKind.Absolute);
            using var client = new TcpClient();
            var port = uri.Port > 0 ? uri.Port : (uri.Scheme == "https" ? 443 : 80);
            var task = client.ConnectAsync(uri.Host, port);
            return await Task.WhenAny(task, Task.Delay(1_500)) == task && client.Connected;
        }
        catch { return false; }
    }

    private static string? FindLegacyHostExe()
    {
        var asmDir = Path.GetDirectoryName(Assembly.GetExecutingAssembly().Location)!;
        var solutionRoot = asmDir;
        for (var i = 0; i < 8 && solutionRoot is not null; i++)
        {
            if (File.Exists(Path.Combine(solutionRoot, "ZB.MOM.WW.OtOpcUa.slnx"))) break;
            solutionRoot = Path.GetDirectoryName(solutionRoot);
        }
        if (solutionRoot is null) return null;

        var path = Path.Combine(solutionRoot,
            "src", "ZB.MOM.WW.OtOpcUa.Driver.Galaxy.Host", "bin", "Debug", "net48",
            "OtOpcUa.Driver.Galaxy.Host.exe");
        return File.Exists(path) ? path : null;
    }
}

[CollectionDefinition(nameof(ParityCollection))]
public sealed class ParityCollection : ICollectionFixture<ParityHarness> { }