fix(mtconnect): Unsupported latch inside the checked section; live-index reads do not clear a write degradation (review r1 N-4, N-5)

- N-4: the /sample Unsupported latch is set inside TryDegradeStream's
  generation-checked critical section (latchUnsupported), so a superseded pump can
  no longer latch it on the session that replaced it.
- N-5: a read served from the live pump index clears only the read degradation -
  it is stream evidence, like a chunk, which by the I-3 ruling does not clear an
  unanswered write. The 'cannot stick' wording in the XML doc and
  docs/drivers/MTConnect.md now states the latch-until-next-answered-/current-read-
  or-write behaviour, incl. the subscription-only consequence. Tested (chunk +
  live-index read keep Degraded; an answered write clears it).

Refs #743

Claude-Session: https://claude.ai/code/session_01Ub9hGmYLQvn9XETzw7CMsT
This commit is contained in:
Joseph Doherty
2026-10-22 16:58:37 -04:00
parent c39bcd6e24
commit bc1f341e8a
3 changed files with 57 additions and 14 deletions
@@ -371,8 +371,9 @@ public sealed partial class MTConnectDriver
/// <summary>
/// Whether the last Agent write got no reply (review I-3 / coordinator ruling: an unanswered write degrades).
/// Cleared by the next answered read OR write (<see cref="RecordReadAnswered"/> / <see cref="RecordWriteAnswered"/>),
/// so it cannot latch on a driver that is still talking to the Agent.
/// Latched until the next answered <c>/current</c> read or answered write (<see cref="RecordReadAnswered"/> /
/// <see cref="RecordWriteAnswered"/>). <b>Neither a <c>/sample</c> chunk nor a live-index read clears it</b>, so on
/// a subscription-only driver one timed-out write keeps the driver Degraded until the next write (r1 N-5).
/// </summary>
private volatile bool _writeDegraded;
@@ -855,9 +856,13 @@ public sealed partial class MTConnectDriver
// The read path is answering (from a stream that is delivering), so a degradation a failed /current left
// behind is no longer true of it. The stamp stays the pump's — this read heard nothing from the Agent itself.
if (_readDegraded || _writeDegraded)
// Clears only the READ degradation (r1 N-5): an index read is the stream's evidence, not an Agent answer to a
// request — the same evidence a /sample chunk carries, which by the I-3 ruling does not clear an unanswered
// write either.
if (_readDegraded)
{
RecordReadAnswered(ReadHealth().LastSuccessfulRead ?? DateTime.UtcNow);
_readDegraded = false;
PublishHealthy(ReadHealth().LastSuccessfulRead ?? DateTime.UtcNow);
}
return results;
@@ -2255,12 +2260,9 @@ public sealed partial class MTConnectDriver
// forever, so retrying would be an unbounded hot loop against an endpoint
// that will never stream until an operator changes something. Latch it, say
// so loudly, and stop. A re-initialize clears the latch.
lock (_subscriptionLock)
{
_streamUnsupported = true;
}
if (!TryDegradeStream(generation, outcome.Error!.Message))
// r1 N-4: the latch is set inside the same generation-checked section as the degrade, so a
// superseded pump can never latch Unsupported on the session that replaced it.
if (!TryDegradeStream(generation, outcome.Error!.Message, latchUnsupported: true))
{
return;
}
@@ -2707,7 +2709,8 @@ public sealed partial class MTConnectDriver
/// <param name="generation">The pump's session generation.</param>
/// <param name="message">The failure (redacted text only).</param>
/// <returns><see langword="false"/> when the pump's session has been replaced (nothing written).</returns>
private bool TryDegradeStream(long generation, string message)
/// <param name="latchUnsupported">Also latch the "endpoint cannot stream" verdict (r1 N-4).</param>
private bool TryDegradeStream(long generation, string message, bool latchUnsupported = false)
{
// The mirror of TryMarkStreamDelivered: a superseded pump's failure is not the new session's.
lock (_subscriptionLock)
@@ -2717,6 +2720,11 @@ public sealed partial class MTConnectDriver
return false;
}
if (latchUnsupported)
{
_streamUnsupported = true;
}
_streamDegraded = true;
Degrade(message);