test: deflake load-sensitive tests (lmxopcua#769)
v2-ci / unit-tests (push) Has been cancelled

- NodeManagerLockTimeBenchmark: wall-clock budget enforced only with OTOPCUA_BENCH=1; default run keeps the state-correctness asserts and logs timings.
- OpcUaClientDriverWire host-status: signal (semaphore) wait that only accepts a non-Running state observed after the server is disposed.
- AbCip PollLatencyComparison: retry once on a fixture connection reset; p95 criterion enforced only with AB_LAT_ENFORCE=1.
- ScriptRunGuardTests: orphan scripts held by a ctx.Released signal instead of a 250 ms wall-clock spin that could finish inside the 60 ms budget under load; stuck-orphan wall-clock rule disabled in the tests that pin the strike counting.
- DependencyMuxReplayTests: split the inside-grace positive control (1 h grace) from the release test (control asserted before unregister), removing the 200 ms grace race.

Claude-Session: https://claude.ai/code/session_01GT67AEHd2Acv1YPwGfXEoV
This commit is contained in:
Joseph Doherty
2026-11-10 05:39:22 -05:00
parent 000d7f5591
commit 11a5c1d4e2
6 changed files with 128 additions and 59 deletions
@@ -212,20 +212,26 @@ public sealed class OpcUaClientDriverWireTests
await drv.InitializeAsync("{}", ct);
drv.GetHostStatuses().Single().State.ShouldBe(HostState.Running);
var transitioned = new TaskCompletionSource<HostState>(TaskCreationOptions.RunContinuationsAsynchronously);
drv.OnHostStatusChanged += (_, e) =>
{
if (e.NewState != HostState.Running) transitioned.TrySetResult(e.NewState);
};
// Signal-based (lmxopcua#769): every host-status event releases the semaphore, and the loop re-reads the
// driver's state. A transient non-Running tick BEFORE the server is gone (a starved keep-alive) must not
// satisfy the wait, so the loop only accepts a non-Running state observed AFTER the rig is disposed; the
// 120 s ceiling only matters on a starved box and costs nothing when the signal arrives.
using var signal = new SemaphoreSlim(0);
drv.OnHostStatusChanged += (_, _) => signal.Release();
await rig.DisposeAsync();
rig = null;
signal.Release(); // re-evaluate once now that the server is certainly gone
var deadline = TimeSpan.FromSeconds(120);
var sw = System.Diagnostics.Stopwatch.StartNew();
while (drv.GetHostStatuses().Single().State == HostState.Running)
{
var remaining = deadline - sw.Elapsed;
(await signal.WaitAsync(remaining > TimeSpan.Zero ? remaining : TimeSpan.Zero, ct))
.ShouldBeTrue("the host probe never left Running after the server stopped");
}
// Signal-based (lmxopcua#769): the wait completes the instant the keep-alive loop reports the loss, so
// the ceiling only matters on a starved box, where server teardown plus the 10 s request timeout plus a
// 1 s keep-alive tick exceeded the old 30 s. A generous ceiling costs nothing when the signal arrives.
var state = await transitioned.Task.WaitAsync(TimeSpan.FromSeconds(120), ct);
state.ShouldNotBe(HostState.Running);
drv.GetHostStatuses().Single().State.ShouldNotBe(HostState.Running);
}
finally