Pass 2 (2026-10-08 00:21Z, df776170): 17/17. Browse 3162=3162 tags, 20 UDT
templates identical, 200 tags read with 0 unexpected divergences (6 live
values compared structurally + every still element), 1 Forward Open and 1
acked Forward Close per native phase. Latency native p95 105.8 ms vs
libplctag 139.5 ms (limit 153.4) PASS. Pass 1 failures and their fixes are
recorded in the gate-results doc; pass 2 had no independent pcap.
Claude-Session: https://claude.ai/code/session_01Ef2ozxS7QPanGT8uoATbTe
The 5069-L3x supports the Large open, as the press proved. The narrow/refusing case is a pre-v20 1769-L3x or a bridge. AbCipDeviceOptions.ConnectionSize: the native transport honours it on the wire; libplctag ignores it.
Claude-Session: https://claude.ai/code/session_01Ef2ozxS7QPanGT8uoATbTe
M2: after plc_tag_destroy, the release writes 0 back to the wrapper's nativeTagHandle, so a repeat release is a no-op instead of touching a destroyed id. The runtimes keep the destroyed handle for tests and diagnostics.
M1: the rooting fallback is now described accurately, as an unbounded leak (one native tag per failed create), not a bounded one. A test seam (ForceRootingForTest) pins that the fallback roots the tag and destroys nothing.
Both pins are red with their line removed. The orphan tests share a non-parallel collection, because the counters and the seam are process-wide.
Claude-Session: https://claude.ai/code/session_01Ef2ozxS7QPanGT8uoATbTe
I1: a live verdict no longer waives the value check. Only elements that moved between the two libplctag reads are excused.
- An element equal in both libplctag reads must also be equal through native; otherwise **NO: stable element i**.
- A moved element must not read NaN or infinity on native unless libplctag saw it too; otherwise **NO: non-finite element i**.
- Live tags log their full raw bytes and every element, and the summary reports live tags as N of the total.
- The offline pin adds the pass-1 Real[5] shape: a stable element shifted on native fails. It is red with the element check disabled.
M9: after each native phase (differential and latency), wait for the session close. With AB_DIFF_COUNT_OPENS=1, assert exactly one Forward Close was sent, using the new CipConnection counters; acked is logged.
M8: each latency pass logs its GC collection deltas (gen0/1/2) for both transports, plus native per-exchange round-trip min/mean/max.
M4: the latency test also asserts that no libplctag orphan had to be rooted.
Claude-Session: https://claude.ai/code/session_01Ef2ozxS7QPanGT8uoATbTe
CipConnection gains two process-wide gate-diagnostic counters next to ForwardOpenCount: ForwardClosesSent (incremented as the request goes on the wire) and ForwardClosesAcked (incremented when the reply arrives within the 1 s budget).
It also records round-trip statistics for completed exchanges: ResetExchangeStats() clears them and ReadExchangeStats() returns count, min, max and mean. The cost is two Stopwatch reads and a few interlocked operations per request. No behaviour changes.
Claude-Session: https://claude.ai/code/session_01Ef2ozxS7QPanGT8uoATbTe
In T21 press pass 1, two REAL tags diverged only because their live values moved
between the sequential libplctag and native phases (WW_M4ActualTemp 384.887 ->
385.138).
The libplctag phase now reads every selected tag a second time, paced, before
native starts (TransportObservations.ProbeReads). A tag whose two libplctag reads
both succeed and differ is classed live. For a live tag, parity still requires:
- the same status (and the native-TypeMismatch rule);
- no decode error;
- the same raw length;
- the same CIP type code when both sides expose one (native records it now;
libplctag.NET 1.5.2 does not);
- the same decoded shape.
Value equality is not required. Live tags get their own list and a "libplctag
probe" table column, and are counted as neither expected nor unexpected
divergences.
A tag that reads stable on libplctag but differs on native is still an unexpected
divergence, and a missing or failed probe never makes a tag live. The offline pin
A_live_value_is_compared_structurally_and_a_stable_one_by_value goes red with the
live branch disabled.
Claude-Session: https://claude.ai/code/session_01Ef2ozxS7QPanGT8uoATbTe
The T21 press pass 1 test host died at exit with "A callback was made on a garbage
collected delegate of type ...plctag+callback_func_ex::Invoke". The cause is in
libplctag.NET 1.5.2. When a tag's create fails, for example with NotFound,
InitializeAsync throws with the callback still registered. Dispose() destroys a tag
only if it is initialised, so the native tag outlives its wrapper. The GC then frees
the delegate, and libplctag's atexit plc_tag_shutdown raises DESTROYED into it.
Reproduced against ab_server: the production LibplctagTagRuntime with a NotFound tag
exits 134 with that exact message, and exits 0 with this fix.
A timed-out create unregisters its callback but never destroys the handle either,
so that native tag keeps reconnecting for the life of the process.
The fix adds LibplctagOrphanRelease (AbCip and AbLegacy). After a failed
initialise/read/write that leaves the tag uninitialised, it reads the handle through
the pinned wrapper's private fields, unregisters the callback and destroys the tag.
If those fields cannot be read, it roots the tag for the life of the process
instead. It is applied in both tag runtimes, the symbol-listing reader and the
template reader. GetStatus keeps the status the tag held before it was destroyed.
Pins: offline refused-connection tests in AbCip.Tests and AbLegacy.Tests, a live
NotFound test on ab_server, and a reflection-shape pin. All go red with the release
disabled. The differential harness also asserts that no orphan had to be rooted.
Claude-Session: https://claude.ai/code/session_01Ef2ozxS7QPanGT8uoATbTe
- AbCip/AbLegacy driver forms gain a Transport InputSelect; FromOptions/ToOptions now carry
Transport, so saving any other field no longer silently reverts Native to libplctag.
- Both drivers log the constructed transport once per init.
- AbCip factory now takes the host ILoggerFactory (it ran on NullLogger in production, hiding
every AbCip log line); DriverFactoryBootstrap passes it.
- Gate results: AdminUI /run (T19) native, inverted (dead port) and libplctag legs.
The libplctag adapters assigned Tag.Path straight from the device's parsed
CipPath. A device address with no route (ab://host:44818/ - the normal
Micro800 form, and a MicroLogix with no route) parses to "", and libplctag
rejects an empty path attribute with ErrorBadData; omitting the attribute
(null) is what it accepts. So Micro800 could not connect on the default
libplctag transport.
This predates the native transport work and exists on master.
Add an internal LibplctagPath.Normalize helper per driver project (blank ->
null) and route all four Tag constructions through it (AbCip tag runtime,
symbol-list enumerator, template reader; AbLegacy tag runtime). The runtime
tag construction is extracted to an internal BuildTag so tests can inspect
Tag.Path without connecting. Tests fail with Normalize reverted to identity.
Update the Micro800 smoke-test comment: both transports are expected to pass.
ab_server answers every Execute PCCC with a fixed requestor id (pccc.c
PCCC_RESP_PREFIX: 07 3D F3 45 43 50 21) instead of echoing the request's, so
the R14-3 echo check rejected every native reply sent under the 0x1337 CIP
originator identity as BadReply. The pooled runtime now sends libplctag's
identity (PcccServices.RequestorVendor/RequestorSerial), which passes the echo
check against ab_server and against a real controller that echoes. Pinned by a
FakeEipServer round trip that answers with ab_server's fixed prefix.
ab_server's RegisterSession sanity check rejects any non-zero sender context
with encapsulation status 1 and a header-only reply, so every native session
against the simulator failed with BadCommunicationError. libplctag sends zero
(logix-register-session.hex); the codec test now matches that capture byte for
byte instead of excluding the context, and a client test pins the zero on the wire.
Scheduler tests (exclusive mid-queue and at head) and a template-reader test that
genuinely queues a Read Template behind a small read; bounded QueuedCount wait
replaces the fixed delay. Falsified: removing exclusive from the reader turns the
reader test red; removing the Exclusive handling in RequestScheduler turns all three red.
Joseph Doherty