4 Commits

Author SHA1 Message Date
Joseph Doherty 31af7acde0 test(mtconnect): make the STAB-20 retention test deterministic again (review I-2)
29c829b5 let the test accept 'already finished' for the stale-session pump, which
since T-1 is nearly always true, so deleting the retention reintroduced STAB-20
silently. A new CannedAgentClient.HoldSampleExit knob keeps that pump alive past
the retire point; the test now asserts strictly that it is still running
(positive control) AND retained, then releases it and asserts shutdown joined
it. Red with the _supersededPumps.Add removed. The pump's own self-exit on a
gone generation is covered by MTConnectStaleCursorTests (pump A completes, the
current pump keeps streaming).

Refs #743

Claude-Session: https://claude.ai/code/session_01Ub9hGmYLQvn9XETzw7CMsT
2026-10-22 16:29:04 -04:00
Joseph Doherty 29c829b540 test(mtconnect): a stale-session pump may finish before the restart retires it
Since the T-1 generation guard a pump stops on its own once its session generation
is gone, so the STAB-20 test's 'the retired pump is retained' assertion raced it
(intermittent failure). The test now accepts either outcome - retained for shutdown
to join, or already finished - which is what the retention exists to guarantee.

Refs #743

Claude-Session: https://claude.ai/code/session_01Ub9hGmYLQvn9XETzw7CMsT
2026-10-22 15:45:54 -04:00
Joseph Doherty 7628402893 fix(drivers): R5 minors - MTConnect session publish+generation atomic and superseded pump joined; MQTT birth-typed writes use the birth's own datatype; AbLegacy dead init join removed + Dispose join test; docs (#602)
Claude-Session: https://claude.ai/code/session_01Ef2ozxS7QPanGT8uoATbTe
2026-10-08 07:52:59 -04:00
Joseph Doherty e76df0741a fix(mtconnect): stamp the /sample pump with its session generation and supersede a stale one (#602 STAB-20)
Claude-Session: https://claude.ai/code/session_01Ef2ozxS7QPanGT8uoATbTe
2026-10-08 07:24:43 -04:00