refactor(runtime): delete the dead in-place ApplyDelta seam; config change = respawn only (#665)

Since #516 a DriverConfig change is stop + respawn, so DriverSpawnPlan.ToApplyDelta
was always empty and the in-place path had no producer. #663 and #664 were both
latent bugs in it. This deletes it:
- host: ApplyChildDelta, _pendingDelta, HandleApplyResult and HandleDeltaApplied
- plan: DriverSpawnPlan.ToApplyDelta
- actor: the ApplyDelta, ApplyResult and DeltaApplied messages, the Connected
  reinit handler and the mid-connect AdoptConfigDuringInit
- tests: the actor tests that drove the seam

IDriver.ReinitializeAsync and each driver's implementation and unit tests stay.
Comments and docs that cited AdoptConfigDuringInit as the reason a driver must
release prior state at init now cite the IDriver contract and overlapping
retries, which still apply.

Claude-Session: https://claude.ai/code/session_01Ef2ozxS7QPanGT8uoATbTe
This commit is contained in:
Joseph Doherty
2026-10-09 11:23:16 -04:00
parent f622d0be2d
commit 1ea1da8236
33 changed files with 112 additions and 559 deletions
@@ -5,8 +5,8 @@ using ZB.MOM.WW.OtOpcUa.Driver.AbLegacy.PlcFamilies;
namespace ZB.MOM.WW.OtOpcUa.Driver.AbLegacy.Tests;
/// <summary>
/// #630 — <c>InitializeAsync</c> on a live driver (<c>DriverInstanceActor.AdoptConfigDuringInit</c> runs
/// it with no teardown first) overwrote <c>_devices[host]</c> and orphaned the previous state's
/// #630 — <c>InitializeAsync</c> on a live driver with no teardown first (the <c>IDriver</c> contract
/// allows it) overwrote <c>_devices[host]</c> and orphaned the previous state's
/// probe loop + runtime. It now releases them first.
/// </summary>
[Trait("Category", "Unit")]
@@ -396,7 +396,7 @@ public sealed class CalculationDriverTests
}
/// <summary>The skip ledger must be REBUILT on ReinitializeAsync, not just at construction — an
/// ApplyDelta that drops a tag's scriptId (or adds a new tag with none) must be reflected immediately,
/// in-place config change that drops a tag's scriptId (or adds a new tag with none) must be reflected immediately,
/// since RebuildTagTable is the one routine both the ctor and reinit share.</summary>
[Fact]
public async Task GetSkippedTags_updates_after_reinitialize()
@@ -414,9 +414,9 @@ public sealed class CalculationDriverTests
}
/// <summary>
/// Gitea #664 — <c>InitializeAsync</c> re-runs on a LIVE driver with a new config when an ApplyDelta lands
/// mid-connect (<c>DriverInstanceActor.AdoptConfigDuringInit</c>). It used to ignore the JSON, so the
/// driver kept its constructor tag set while the deploy reported the change applied. A config with a body
/// Gitea #664 — <c>InitializeAsync</c> must serve the config document it is handed (the <c>IDriver</c>
/// contract). It used to ignore the JSON, so a live driver re-initialised with a new config kept its
/// constructor tag set. A config with a body
/// now rebuilds the tag table (and the dependency-ref set the host re-registers from).
/// </summary>
[Fact]
@@ -5,8 +5,8 @@ using ZB.MOM.WW.OtOpcUa.Driver.FOCAS;
namespace ZB.MOM.WW.OtOpcUa.Driver.FOCAS.Tests;
/// <summary>
/// #630 — <c>InitializeAsync</c> on a live driver (<c>DriverInstanceActor.AdoptConfigDuringInit</c> runs
/// it with no teardown first) overwrote <c>_devices[host]</c> and orphaned the previous state's
/// #630 — <c>InitializeAsync</c> on a live driver with no teardown first (the <c>IDriver</c> contract
/// allows it) overwrote <c>_devices[host]</c> and orphaned the previous state's
/// probe loop + client. It now releases them first.
/// </summary>
[Trait("Category", "Unit")]
@@ -209,9 +209,8 @@ public sealed class GalaxyDriverFactoryTests
}
/// <summary>
/// Gitea #664 — <c>InitializeAsync</c> re-runs on a live driver with a NEW config when an ApplyDelta lands
/// mid-connect. Galaxy cannot hot-swap its options, and it used to ignore the JSON — connecting on the OLD
/// endpoint while the deploy reported the change applied. A non-equivalent config is now refused, exactly
/// Gitea #664 — <c>InitializeAsync</c> must not silently ignore the config it is handed. Galaxy cannot
/// hot-swap its options, and it used to ignore the JSON — connecting on the OLD endpoint. A non-equivalent config is now refused, exactly
/// as <c>ReinitializeAsync</c> refuses it; an equivalent one still initialises.
/// </summary>
[Fact]
@@ -265,9 +265,8 @@ public sealed class MTConnectDriverLifecycleTests
[Fact]
public async Task Initialize_honours_the_config_json_over_the_constructor_options()
{
// The runtime delivers a config change ONLY through this argument (DriverInstanceActor
// ApplyDelta -> ReinitializeAsync(json) on the LIVE instance). Ignoring it would make every
// MTConnect config change a silent no-op that still seals the deployment green.
// The IDriver contract: the driver serves the config document it is handed, not only its
// constructor options — ignoring it would turn an in-place config change into a silent no-op.
MTConnectDriverOptions? seen = null;
var driver = new MTConnectDriver(Opts(), "mt1", o =>
{
@@ -533,7 +532,7 @@ public sealed class MTConnectDriverLifecycleTests
[Fact]
public async Task Reinitialize_with_an_unreadable_config_keeps_the_running_driver_serving()
{
// Blast radius: an unparseable config edit fails the DEPLOYMENT (ApplyResult(false)); it must
// Blast radius: an unparseable in-place config edit fails the reinitialize call; it must
// not down a driver that is happily serving its previous, valid configuration.
var (driver, client) = await InitializedDriverAsync();
@@ -225,9 +225,8 @@ public sealed class ModbusLifecycleHygieneTests
// -------------------- Re-initialize transport hygiene --------------------
/// <summary>
/// A second <c>InitializeAsync</c> on a live driver — the shape
/// <c>DriverInstanceActor.AdoptConfigDuringInit</c> produces when an ApplyDelta lands while the
/// driver is Connecting/Reconnecting — must dispose the transport the previous initialize
/// A second <c>InitializeAsync</c> on a live driver with no teardown between (the <c>IDriver</c>
/// contract allows it) must dispose the transport the previous initialize
/// published before publishing a new one. Pre-fix <c>_transport</c> was simply overwritten,
/// stranding a connected transport no later dispose could reach.
/// </summary>
@@ -251,7 +250,7 @@ public sealed class ModbusLifecycleHygieneTests
});
await drv.InitializeAsync("{}", CancellationToken.None);
// No ShutdownAsync in between — exactly what AdoptConfigDuringInit does.
// No ShutdownAsync in between.
await drv.InitializeAsync("{}", CancellationToken.None);
handed.Count.ShouldBe(2, "each initialize must build its own transport");
@@ -415,8 +414,8 @@ public sealed class ModbusLifecycleHygieneTests
/// <summary>
/// Two initializes racing on the same driver — the shape <c>DriverInstanceActor</c> produces by
/// dispatching initialize on a fire-and-forget <c>Task.Run</c> with <c>AdoptConfigDuringInit</c>
/// firing twice — must serialize on the lifecycle gate. Ungated, the teardown at the top of one
/// dispatching initialize on a fire-and-forget <c>Task.Run</c>, where a reconnect retry tick can
/// overlap a still-running attempt — must serialize on the lifecycle gate. Ungated, the teardown at the top of one
/// initialize disposes the transport the other just published (or nulls <c>_transport</c>
/// mid-connect), leaving the driver holding a disposed transport.
/// </summary>
@@ -28,10 +28,9 @@ public sealed class MqttDriverSessionLifecycleTests
/// </summary>
/// <remarks>
/// <para>
/// Reachable with no teardown of the driver's own: <c>DriverInstanceActor</c> re-runs
/// <c>InitializeAsync</c> with no <c>ShutdownAsync</c> when an <c>ApplyDelta</c> arrives while
/// the driver is Connecting/Reconnecting (<c>AdoptConfigDuringInit</c>). The
/// <c>teardownFirst</c> flag on the #527 retry path covers only the retry path.
/// The <c>IDriver</c> contract allows re-initialising a live driver with no
/// <c>ShutdownAsync</c>, and the #527 retry path's <c>teardownFirst</c> flag does not stop a
/// retry overlapping a still-running attempt, so the driver must not rely on a teardown.
/// </para>
/// <para>
/// The assertion is the broker's <b>live socket count</b>, not the driver's own view: a leak
@@ -483,8 +483,8 @@ public sealed class S7DriverReconnectTests
}
/// <summary>
/// #630 — a second <c>InitializeAsync</c> on a live driver (no <c>ShutdownAsync</c> between, as
/// <c>DriverInstanceActor.AdoptConfigDuringInit</c> does) must stop the first probe loop and
/// #630 — a second <c>InitializeAsync</c> on a live driver (no <c>ShutdownAsync</c> between — the
/// <c>IDriver</c> contract allows it) must stop the first probe loop and
/// dispose the first connection rather than orphan them.
/// </summary>
[Fact]