Reverses the conservative call from the previous commit at the user's direction,
and they were right: "surface it on a counter first" left the wire lying. An
MXAccess event can carry Good quality while its payload is a kind MxValueDecoder
has no mapping for, so the old behaviour published null-under-Good — which a
client cannot tell apart from a genuine null reading. A value the gateway DID
send would be consumed as a value it did not, with the only evidence in a log
nobody is tailing. Bad data is the honest wire representation of lost data.
BadTypeMismatch is an existing StatusCodeMap constant, deliberately reused rather
than written inline: StatusCodeParityTests finds status constants by reflecting
over Good/Uncertain/Bad-prefixed uint fields and checking them against the pinned
SDK, so an inline literal would be invisible to that guard (Gitea #497, where 16
wrong constants had accumulated across 6 drivers exactly this way).
Logging is now once per DISTINCT value kind rather than once per pump. The set is
bounded by the MxValue oneof enum so a broken producer still cannot flood, but a
second unrepresentable kind appearing later gets its own line — warn-once-per-pump
would have silently swallowed it, which is the same class of bug as the null this
commit fixes.
The test's fixture matters more than its assertion: the undecodable event carries
Quality = 192 (OPC DA Good), so a decoder that gave up silently would publish Good
and fail. Paired with a positive control — a decodable value at the same quality
must still publish Good — so it cannot be satisfied by a pump that marks
everything Bad. Falsified by reverting the status to Good: the test reds, and only
that test.
Verified: 331 Galaxy tests green, Core.Abstractions 298 (the parity guard),
solution builds 0 errors. Docs corrected in the two places that recorded the old
decision as deliberate.
Closes the two follow-ups left open by the MxGateway 0.2.0 bump. Reading the
gateway's own contract first changed what each of them is, and one of my earlier
claims was simply wrong -- both corrections are recorded below.
Replay gap
----------
A sentinel MxEvent carries family = UNSPECIFIED, so it fell through EventPump's
family switch into the default drop: we silently ignored a documented MUST ("you
missed events -- discard local state and re-snapshot"). It is now intercepted and
answered by re-issuing SubscribeBulk for every tracked subscription, since
MXAccess delivers an initial value per item on subscribe -- re-subscribing IS the
re-snapshot. It does not restart the event pump (the stream carrying the sentinel
is healthy, and the handler runs on that stream's own read loop, so disposing the
pump from there would deadlock on itself), it does not block the read loop, and
re-entry is gated so a burst collapses to one pass instead of a subscribe storm
per sentinel.
Handled on the READ loop, not in the dispatch loop, because everything past that
point is subject to newest-dropped channel pressure -- and "you missed events" is
the one signal whose loss is unrecoverable. Pinned by a test that fills the
channel and still expects the sentinel.
⚠️ This cannot fire in the shipped configuration, and I did not pretend otherwise
or paper over it by enabling resume. The gateway emits the sentinel only for a
stream opened with after_worker_sequence > 0; GatewayGalaxySubscriber passes 0,
which the gateway treats as "fresh stream, not a resume -- plain attach, no
replay, no sentinel". Enabling resume is a larger and genuinely delicate change:
the sequence is only meaningful within ONE gateway session, our reconnect
recreates the session, and replaying a remembered sequence against a fresh
session sets the server's live watermark ABOVE every event it is about to send --
silencing the stream instead of recovering it. That deserves its own change with
a live gate, which I cannot run from here.
Sparse arrays -- correcting the last commit message
---------------------------------------------------
I wrote that the bump "makes handling it possible" on the decode path. That is
wrong: MxSparseArray is documented write-only, the gateway expands it before
forwarding, and "the worker never receives or produces it". So the decode path
is unreachable by contract and the write path has no use for it either -- it is a
wire compaction for a WHOLE-array write, not a partial update (unmentioned
indices are RESET, deliberately, because the gateway cannot read-modify-write
without racing the provider). It is therefore the wrong primitive for an OPC UA
IndexRange write, which expects untouched elements preserved -- and WriteRequest
carries no IndexRange in the first place.
It is decoded anyway, as contract defence: the alternative if a future gateway
ever emits one is a silent null published under Good quality. Fill semantics
mirror the gateway's own SparseArrayExpander exactly (epoch for Time, empty
string, Integer widening to long[] when any element is int64) so a value that
round-trips through a write reads back as the same CLR array.
Also: the decoder is now total by contract, since it runs inside the dispatch
loop where an escaping exception kills dispatch for the life of the pump. A
malformed payload degrades instead of throwing.
A real defect found in the same code
------------------------------------
A stream the GATEWAY completes -- a closed or evicted session -- ends the RPC
without an error, so nothing threw, the supervisor was never notified, and the
pump exited silently. GetHealth() kept reporting Healthy, which also means the
#527 comms-loss watchdog could never trip, because it requires a non-Healthy
state. The driver went permanently deaf with no signal anywhere. This one is
live, unlike the two above. Now reported as a fault; the falsification control is
the existing CleanShutdown_DoesNotInvokeOnStreamFault, since our own disposal
also ends the stream and must NOT be reported.
Value kinds we cannot represent are now counted and warned once rather than
published as an indistinguishable null. The status is deliberately NOT downgraded
to Bad: nothing can reach that branch today, and changing what an OPC UA client
sees for a live shape is not a change to make without a gate. The counter exists
so the next person has evidence instead of a guess.
Verified: 330 Galaxy tests green (12 new), plus Galaxy.Browser 21, Runtime 558,
Core 351; full solution builds 0 errors with the warning census unchanged. Every
new test was falsified first -- disabling the gap interception reds 3 of 4 (the
throwing-callback test passes vacuously when the callback never runs, by
construction), disabling sparse decode plus the representability signal reds all
7 decoder tests, and both were restored and re-verified green.
Docs: the Galaxy.Performance invariant was stated as received = dispatched +
dropped + in-flight, which this breaks -- a sentinel is received but never enters
the channel -- so it now reads + replay_gaps. Galaxy.md gains the three
operational notes; DriverLifecycle.md records that "reaches a non-Healthy state"
is an obligation a driver can silently fail to meet, which is the general shape
of the clean-end bug.
Gitea #561 ("production writes a 0-byte log") is closed as not-a-defect. The
production log on wonder-app-vd03 listed as 0 bytes with LastWriteTime ==
CreationTime; opening it showed 4,658,893 bytes of correct output, written
continuously since service start.
On Windows the directory entry's size and last-write time are not refreshed while
the writing process holds the file open, so Get-ChildItem, dir and FileInfo.Length
all report size-at-open for the life of the handle. A live, healthy log is
indistinguishable from a dead sink by metadata alone — which is what made the
false reading survive six weeks of investigation.
That same trap was in the 29a202f5 instrumentation: CheckFileSinkGrowth measured
with FileInfo.Length, so it would have reported "self-check FAILED: the sink did
not grow" at every boot on a perfectly healthy Windows service — and a
self-check's false positive reads as confirmation of the very bug it was added to
find. MeasureFamilyBytes now reads Length through an opened FileStream:
FileShare.ReadWrite to open it at all while the sink holds its write handle, and
FileShare.Delete so the probe never blocks a roll or retention delete.
Both instruments are kept. A sink that cannot write its file is still a real and
otherwise-invisible failure — a sibling repo reproduced it via handle contention,
where RollingFileSink logs the lock-out to SelfLog and falls through to _001. Only
the two doc claims that are now false are corrected: the class summary's
"production ran that way for six weeks" and the warning text's "#561 failure mode".
Tests: A_working_file_sink_passes already measured while Serilog held the file
open — precisely the broken case — and passed throughout, because it only ever ran
on macOS, where the length is honest. That platform blindness is now recorded in
the class summary, alongside two tests pinning the sharing contract the fix
depends on.
Four-minor jump 0.2.3 → 0.6.2. The release line exists because a deploy
renaming the app directory took the MxGateway credential store with it
(2026-08-09, two silent days): Secrets:SqlitePath must now be absolute AND
outside the content root, rejected at boot rather than relocated. 0.6.2 over
0.6.1 is message-only (the rejection text now teaches the mounted-volume
case and forecloses the override hunt) — the live rig evidence below was
gathered at 0.6.1 and the validator behavior is identical.
Adoption changes beyond the pins:
- appsettings.json drops the relative "otopcua-secrets.db" (now boot-fatal);
dev gets the library's absolute per-user default, deployments supply their
own (wonder already pins an absolute out-of-root path in the service env).
- Program.cs's pre-host ${secret:} expander now calls the FOUR-argument
AddZbSecrets with builder.Environment.ContentRootPath. This call site is
the actual fix, not the pin: the throwaway container has no
IHostEnvironment, and without the explicit root the under-content-root
rule is skipped at exactly the moment the migrator creates the store file
(the 0.6.0 defect). Do NOT simplify back to the 3-argument overload — it
compiles and silently reopens the gap. Pinned by a Program.cs source-scan
test AND a runtime falsification control that RUNS the migration and
demonstrably creates the store at the rejected path under the 3-arg form.
- docker-dev: Secrets__SqlitePath moves /app/data → /data via a SECOND mount
of the same per-node volume (same file at the volume root, so seeded
stores carry over) because /app is the container's content root and the
new rule correctly rejects it. LocalDb stays on /app/data (secrets-only
rule).
- SecretsStorePathValidationTests: positive-control probes, artifact-first
(relative and in-content-root paths must FAIL leaving no file; outside-
root passes; the no-explicit-root control proves the file assertions can
fail) — a clean boot is equally consistent with the validator never being
wired.
Live-verified on the six-node docker-dev rig (at 0.6.1): all nodes boot
under the new validation with stores at /data; an AdminUI-written secret
converged into BOTH central nodes' independent stores; site-a held only its
own pair's secrets (Phase 6 pair-local isolation intact). Replicator wire
behavior unchanged 0.3.0–0.6.x per CHANGELOG; registration-graph pins and
the full suite pass.
Production ran six weeks with a healthy process and a 0-byte log because
Serilog isolates per-sink failures: a file sink whose every write fails is
invisible from anywhere but SelfLog, which was not enabled. Two instruments:
- LogPipelineDiagnostics.EnableSelfLog (before the first logger builds):
SelfLog to logs/serilog-selflog.txt (rotated at 5 MB) + stderr — the only
channel that carries the exception the sink swallows.
- CheckFileSinkGrowth (after the startup banner): verifies each configured,
unbuffered file sink's rolled-file family physically grew across a marker
event; a silent sink is reported at boot via SelfLog/stderr and the
console sink instead of discovered months later. Growth-across-marker,
not existence — a previous run's bytes must not mask a silent sink (the
false pass an existence check would have given on wonder-app-vd03).
Never throws: both instruments swallow their own failures — a diagnostics
channel that can fail startup is worse than the defect it watches for.
The load-bearing test is the failure case (a sink that writes nothing IS
reported); a check that can only pass is indistinguishable from no check.
The page ran members, the driver-instances heading, and every cluster group
together inside one panel; driver name and id slammed inline; one long driver
exception blew the table out sideways (Controls scrolled out of reach); and the
inline #546 expander row dumped a full-width panel into the table.
- One panel per cluster group, node chips in the panel head.
- Driver id renders on its own muted mono line under the name.
- Last error abbreviated to 60 chars (DisplayText.Abbreviate — #504 rule),
full text on the tooltip and in the Controls dialog.
- DriverStatusPanel moved from an inline expander row to a modal (one at a
time); the dialog re-resolves its row from the current groups each render so
the 5 s rebuild can't strand it, and the Enabled-in-key recycle discipline
carries over unchanged. HostsPanelWiringTests re-pinned to the dialog markup.
Live-verified on docker-dev: separated cards, in-viewport table, dialog opens
with live breaker/status data and grouped Reconnect|Restart, closes cleanly.
Driver.Mqtt.Contracts-004/-005/-006/-007.
-004: retainSeed was the last behavioural key read leniently, so "retainSeed":"false"
— what a CSV import produces — silently became the default TRUE, the INVERSE of the
authored intent. Both mappers now read it through TryReadBoolStrict (the qos
absent/valid/present-but-invalid split) and DescribeRejection names it. The authoring
side needed no hand-mirrored rule: MqttTagConfigModel.Validate already consults this
same parse through TagConfigParseGate, so the save is blocked with the inspector's
wording — pinned by a new AdminUI parity theory pair.
-005: the seven [Range] attributes on MqttDriverOptions/MqttPlainOptions never ran —
the options are bound by JsonSerializer, and nothing in the four MQTT projects calls
Validator.TryValidateObject. Two of them CONTRADICTED the real authority (KeepAlive's
[Range(1,…)] forbade the zero Normalize admits as legal MQTT; MaxPayloadBytes' forbade
the non-positive value its own remarks document as falling back to 1 MiB), so wiring
them up would have broken deliberate behaviour. Deleted, with each property naming
where its bound actually lives. The one escape they papered over is closed: an
out-of-range defaultQos rode into the SUBSCRIBE verbatim, and MqttSubscriptionManager's
constructor now clamps it exactly as the AdminUI driver form already persists.
-006: the DateTime write arm cast a negative long to ulong, so 1969-12-31 published a
timestamp ~584 million years in the future; it is now refused as ValueNotRepresentable.
An Unspecified kind was handed to ToUniversalTime(), which reads it as LOCAL — the same
authored instant encoded differently on the two nodes of a redundant pair; it is now
read as UTC.
-007: Format/FormatState enforced only null/empty, and nothing at all on deviceId, so
the type was not round-trip-closed. They now apply the parser's IsValidSegment (plus
'/', which only the format side can see). MqttWriteEncoder's own guards stay — TryEncode
returns an error rather than throwing, and that contract does not change.
IWritable shipped behind write.enabled and docs/drivers/MTConnect.md told the operator to
set it, but MTConnectDriverForm — the only editor DriverFormMaps maps for this type, with
no raw-JSON escape hatch and no driver-instance API — exposed no write section, so the
documented enablement path was a hand edit of DriverInstance.DriverConfig in the config DB.
- Write-back section: switch + form text stating it is observation injection, that the
Agent's AllowPut is an independent second gate, and that it does not by itself make a
node writable over OPC UA (Tag.AccessLevel still composes CurrentWrite).
- Write added to the mirrored MTConnectConfigDto and to the form model, written in BOTH
states — an omitted block would leave a previously-authored true standing, so the switch
could never turn writing off again.
- The false 'MTConnect is read-only' header comment is gone; the doc names the new section.
No bUnit here, so the model/serialization seam is unit-tested both sides: the AdminUI test
pins the emitted bytes, and MTConnectFactoryTests parses that same document and asserts
MTConnectDriverOptions.Write.Enabled. Live page verification remains a rig step.
The first form compared the publish's managed thread id with the caller's, which flaked
(1 in 6): the evaluation pump may reuse the pooled thread the test released while awaiting
the drain, so an id match proves nothing. It now sets a ThreadLocal flag for exactly the
OnDependencyValue window, so 'published inline' is answered on the stack rather than by
thread identity. Re-falsified by making SchedulePump run the pump synchronously (red
immediately, reverted); 5 consecutive clean suite runs after.
The two value-bearing fields were read asymmetrically and nothing said why.
whereValue went through ReadScalar (string/number/boolean — the design doc's own
wide-row example authors it as the number 7); keyValue went through the string-only
ReadString, so a numerically-authored key refused the tag outright. Nothing downstream
justified the split: both are BOUND as DbParameters and both are matched back as text —
SqlPollReader.IndexByKey stringifies the key cell precisely so an authored key can face
an INTEGER key column.
The refusal was also mis-explained. Inspect branched on ReadString(root, "keyValue") is
null, so a numerically-authored key drew "Sql KeyValue tag has no 'keyValue'" — telling
the operator the field is missing when it is right there in the blob, the #549 shape of
an authoring surface disagreeing with the runtime about WHY.
keyValue now uses ReadScalar; Inspect mirrors it with HasScalar and says what a usable
key is; SqlTagConfigModel loads it with the shared TagConfigJson.GetScalarText (which
also replaces SqlRowSelectorModel's private copy), so the editor is not left stricter
than the parser. An object/array/null key is still refused. docs/drivers/Sql.md updated.
RED first: 4 failing in Driver.Sql.Tests.
Driver.Mqtt.Browser-002/-003/-004/-005.
-002: ObservePlainTopic's pre-pass judged segment LENGTH only, so a plain topic
segment carrying a newline or an ANSI escape became the picker's displayed label —
a rule ObserveSparkplugBirth already applied via IsUsableLabel. The control-character
half is now single-sourced (HasControlCharacter) and the plain marker text converges
with the Sparkplug one. An EMPTY topic level stays accepted on purpose: it is legal
MQTT and its terminal leaf is a real committable address.
-003: GetOrCreate read _nodeCount and incremented it later, so N observers could each
see cap-1 and all insert. The slot is now reserved with one Interlocked.Increment and
given back on both non-consuming paths.
-004: Project ordered a level by Segment alone, but the level is keyed by Path
precisely because two Sparkplug siblings CAN share a Segment — leaving the tie to
ConcurrentDictionary enumeration order, which .NET's randomised string hashing makes
differ per process. .ThenBy(Path) makes the sort total.
-005: no behaviour change — DefaultNodeCap's doc now states the real per-session
ceiling and records that the aggregate across sessions is bounded by the reaper's
idle TTL, not by any session cap.
Inspect parsed the TagConfig blob once itself (root kind + the present-but-discarded
changeTriggered/timerIntervalMs checks) and then called TryReadTagConfig(reference, out _),
which parsed the same string from scratch again for the scriptId rule. The field-by-field
read is now TryReadRoot(JsonElement), which both entry points share, so Inspect consults
the one authority over the root it already has. The public string-based TryReadTagConfig
signature other modules consume is unchanged.
Pure refactor: added CalculationTagDefinitionFactoryInspectTests (17 cases pinning the
exact wording, the rejection-vs-advisory split and their order, and the
never-stricter-than-the-runtime-read invariant). They pass unchanged against the
pre-refactor code — verified by stashing the production change.
ISqlDialect gains MaxBindParameters — the per-command ceiling the seam was
missing while modelling everything else that differs per backend: 2098 usable
of SQL Server's 2100 (sp_executesql spends two), 1000 for Oracle's IN-list
limit (ORA-01795, tighter than its 64k bind count), 65535 for the Postgres
Bind message.
SqlGroupPlanner.AddKeyValuePlans now emits several plans for one group when
its distinct keys exceed that ceiling, instead of one statement the provider
refuses after the connection opened — which left every tag in an EAV group
publishing BadCommunicationError on every poll, permanently, and reported the
database endpoint as down. Chunks share the group key; members follow their
own key so a shared keyValue never splits. SqlPollReader is unchanged: it runs
one command per plan and slices each plan's own Members back, proven by a
fixture read at a ceiling of 1.
OnDependencyValue ran every change-triggered dependent's script INLINE, under _lock,
via TimedScriptEvaluator.RunAsync(...).GetAwaiter().GetResult() — on the Akka dispatcher
thread of Runtime's DependencyConsumerMuxAdapter, for up to RunTimeout (2 s default) per
arrival, with the state lock held for the whole run.
The callback now records the value, PREPARES the affected tags under the lock (definition
+ warm-up gate + dependency snapshot taken at trigger time) and queues them for a single
serialized evaluation pump. _lock is a short-held state lock again; a new _evalGate
serializes the pump against the timer ticks and is taken only by evaluation workers, so
no host thread can wait on a script. Lock order is always _evalGate -> _lock.
Semantics preserved: per-tag ordering (one FIFO pump), the warm-up gate BEFORE the
evaluator runs, dedup/Bad-transition/recovery publishes, and — because the dependency
snapshot is taken when the trigger fires rather than when the pump catches up — a burst
of arrivals still evaluates against the values that triggered it.
TagConfigJson.ClassifyEnumText classified via Enum.TryParse(text, ignoreCase: true),
which accepts a numeric STRING for any enum: "1" returns the member at that ordinal
and an out-of-range "99" parses just as happily. Every driver contract's strict enum
read inherits that, so a Sql tag authored "type":"99" was ACCEPTED, carried
(DriverDataType)99, and SqlPollReader.TryCoerce fell through its switch to the String
arm — a stringified cell published Good under a type nothing can map.
Refuse a digit- or sign-leading spelling before the parse (judged on the TRIMMED text,
because Enum.TryParse trims), and require Enum.IsDefined — which covers what the
leading-character check cannot, since TryParse also ORs a comma-separated list for a
non-[Flags] enum. Mirrors RawBrowseCommitMapper.ParseNativeType and
MTConnectTagDefinitionFactory.TryParseDataType, which already guard their own parses.
Three AdminUI restatements of the same rule now ask ClassifyEnumText instead of
Enum.TryParse (TagConfigJson.GetEnum, SqlTagConfigModel.ReadEnumStrict,
MqttTagConfigModel.DescribeInvalidEnum/GetEnumNullable) — leaving them would have opened
the #549 direction the moment the runtime tightened: a blob the modal accepts and the
driver skips.
RED first: 10 failing in Core.Abstractions.Tests, 4 in Driver.Sql.Tests, and 10 of the
11 new fleet-wide parity cases (re-confirmed by reverting the Core fix).
Driver.MTConnect-005/-006/-007, each TDD'd red-first.
- -005: _options was the one field at the StartCoreAsync commit point published by a
plain write and read lock-free from ReadAsync/WriteAsync/DiscoverAsync/HostName/health.
Now Volatile.Write/Volatile.Read like _session/_binding/_announcedInstanceId, pinned by
a source scan (MTConnectOptionsPublicationTests) that fails on any unfenced touch.
- -006: TryFormatValue guarded null and Array, but Convert.ToString renders ANY
non-IConvertible value as its type name under a Good status. Guard is now the property
the remark relies on; the accepted set (string/bool/numeric/DateTime) is unchanged.
- -007: MTConnectAgentClient.Dispose was a check-then-set on a non-volatile bool, so two
racing disposals both passed and the loser cancelled an already-disposed CTS. Now an
Interlocked.Exchange claim over an int, with the request seams reading it volatile.
(1) DisposeAsync disposed _gate unconditionally. SemaphoreSlim.Dispose does not
complete pending async waiters, and the in-flight call's Release() then threw
ObjectDisposedException from inside the semaphore (swallowed by the finally), so
a caller parked in WaitAsync was never released and its call never returned. The
gate is now left alive - the pattern SqlPollReader documents - so the waiter is
admitted and meets the session's own guarded ThrowIf(_disposed, this).
(2) LastUsedUtc advanced only on completion, so a slow catalog call (the AdminUI
allows 20 s) was invisible to the TTL reaper for its whole duration and a session
at the edge of its TTL could be disposed mid-call. It is now stamped on entry
too.
PreferredTimestampColumn nominated the first column mapping to DriverDataType
.DateTime, and that set includes date-only types (SQL Server / PostgreSQL
'date'). A table whose first date/time column is a 'date' committed
rowSelector.topByTimestamp = <date>, making the wide-row read ORDER BY <date>
DESC LIMIT 1 - which ties for every row written that day, so 'the newest row'
became storage order among the ties.
Time-bearing candidates now outrank date-only ones; a date-only column is still
stated when it is all the table has, so such a table stays bindable. Oracle's
DATE carries a time of day and is exempt - pinned by a control dialect that
differs from the SQLite one only in the Provider it reports.
Driver.Mqtt-006 (Low, concurrency). DisposeAsync disposed _gate while StopAsync
cancels the reconcile loop WITHOUT awaiting it, so a ReconcileAsync already in
flight could be inside _gate.WaitAsync, or on its way to the finally-release,
when the semaphore went away. The release then threw ObjectDisposedException
naming SemaphoreSlim, logged as a reconcile failure — telling an operator
nothing about what happened and hiding whatever the reconcile was really
failing on.
Left undisposed, matching MqttConnection's documented rule for its own lifecycle
primitives and its stated rationale: a late callback racing teardown should be a
benign no-op rather than an ObjectDisposedException naming the wrong type. Two
lifecycle types in one module disagreeing about this is what gets copied into
the third. Nothing is leaked — the gate holds no timer and no registration — and
ClaimAsync still re-checks _disposed under it, so no claim can be created after
dispose either way.
Test is deterministic rather than timing-dependent: the gate is taken before the
connect starts, so a MiniBroker in blackhole mode (TCP accepted, CONNECT never
answered) accepting the socket proves the claim is inside the gate when dispose
runs. RED: log.Captured contained "ObjectDisposedException".
Driver.Mqtt-007 (Low, error handling). OnReconnectedAsync throws when a
re-subscribe is granted zero filters — MqttConnection's contract is that a
throwing Reconnected subscriber tears the fresh session down and retries under
backoff, because a connected-but-deaf driver is the one outcome this path must
never serve. But when the live subscribed topics resolved to zero filters in the
first place it logged a warning and returned, reaching the same "nothing was
re-established" state by a different route and giving it the opposite answer.
Throw there too, and say so in the remarks so the type states ONE rule. It costs
no other tag — every live topic is in the set that resolved to nothing, so there
is nothing left flowing to protect, which is the whole reason the partial-
rejection case is exempt — and it is not permanent: the retry runs against
whatever Register has published by then.
Test reaches the state the only way production can: Register prunes
_subscribedTopics of the topics its new table no longer carries, but a SUBSCRIBE
already in flight records its own topic AFTER that prune ran, so a deploy landing
inside the round-trip leaves a live topic no authored tag resolves to. Paired
with the control that a reconnect with no live subscriptions at all stays a
no-op — throwing there would tear down the session of every driver whose tags
have not been subscribed yet.
Driver.Mqtt.Transport-002 (Low, correctness). Credentials were attached only
when Username was non-empty, so a config carrying a Password and no Username
connected anonymously with the secret silently discarded — the deployment reads
as authenticated and is not, which is the direction an operator cannot see from
the config.
Warned rather than sent. MQTT 3.1.1 may not carry a password without a
username, so the password-only CONNECT is not available on every protocol
version this factory serves, and synthesizing one under V500 only would change
the wire behaviour of deployments that connect happily today. The configured
values stay authoritative; the drop just stops being silent. The secret is
never formatted into the message.
Tests: the warning fires and does NOT contain the password, plus the falsifying
control that an ordinary anonymous config (neither credential set) stays silent
— without it the fix could warn on every connect and still pass.
NAMEDATALEN-1 (63) and Oracle 12.2's 128 are byte limits in the database
encoding, not character counts, so a 32-character Cyrillic name passed
ident.Length and was then silently truncated (Postgres) or rejected by the
server (Oracle). Both dialects now measure Encoding.UTF8.GetByteCount and say
'bytes' in the rejection. SqlServerDialect deliberately keeps ident.Length —
sysname is nvarchar(128), a real character limit — pinned by a control test.
SqlDriverProbe surfaced 'Config JSON is invalid: ' + ex.Message, where
SqlDriverBrowser.ParseDocument deliberately withholds the identical message
because the malformed text may itself carry a connection string. The two read
the same blob shape - the AdminUI Sql picker composes a document with a
top-level connectionString literal - so a malformed paste reaches both. The
probe now reports the failure class only, matching the browser.
RawBrowseCommitMapper.MapLeaf typed a browse-committed raw tag's OPC UA node
from the leaf's probe-inferred DriverDataType, but its BuildTagConfig MTConnect
branch emitted `new MTConnectTagConfigModel { FullName = address }` — and that
model's DataType defaults to String with ToJson() unconditionally writing
`dataType`. A non-blank authored type wins the driver's coercion precedence, so
the Agent-probe inference step was never consulted and every browsed SAMPLE
published text into a Double/Int64 node. Silent: the blob is perfectly valid, so
neither the authoring gate nor the deploy gate had anything to say.
Same defect class as #536, which is why the AbCip / TwinCAT / FOCAS branches
forward the browsed native type via NativeType(addressFields, …). MapLeaf now
passes driverDataType into BuildTagConfig, whose MTConnect branch uses it as the
NativeType fallback — a stated native `dataType` address field still wins, so
the branch reads like its siblings; the difference is only that MTConnect's
fallback is the leaf's own inference rather than a fixed model default. Its
coercion type IS a DriverDataType, the very enum the leaf carries, so nothing is
re-derived. NativeType's numeric-string and unparseable-name guards are shared
via the extracted ParseNativeType, so an unusable name still keeps String.
Tests assert the PAIR (row OPC UA type and blob coercion type agree with the
inference) — the blob alone would pass against a mapper that stopped typing the
row — plus the parity control that a leaf inferred (or reported) as String still
commits `"dataType":"String"`, so the forward cannot be implemented as a
deletion. Observed red first: expected "Float64" but was "String".
CalculationDriverProbe.Opts (Test-Connect) omitted AllowTrailingCommas and
ReadCommentHandling that CalculationDriverFactoryExtensions.JsonOptions
(spawn) and CalculationDriver.ReinitJsonOptions (in-place ApplyDelta) both
set, so a config blob the real deploy path tolerates could fail Test-Connect
as "invalid JSON" even though deploying it would succeed. Aligned the three
option sets. Also adds CalculationDriverProbeTests.cs, which previously did
not exist at all for this driver type's Test-Connect surface.
Part of code-reviews/Driver.Calculation/findings.md (-003, -004).
Code review of Driver.Calculation found two real defects, both TDD-fixed:
- CalculationDriver's ctor tried to reuse its ILogger<CalculationDriver> as the
evaluator's ILogger<CalculationEvaluator> via an `is` check that can never
succeed (unrelated closed generics), so CalculationEvaluator always ran with
a NullLogger regardless of what the driver was constructed with. Fixed by
having CalculationEvaluator accept the non-generic ILogger instead.
- EvaluateTag called the evaluator BEFORE checking whether every declared
dependency had arrived, so a multi-dependency change-triggered calc tag
evaluated its script against a partial snapshot on every warm-up tick.
VirtualTagContext.GetTag returns a Bad snapshot (.Value == null) for a
missing dep, so the idiomatic `(double)ctx.GetTag("x").Value` cast throws
NullReferenceException — caught and logged as a spurious "script threw"
script-log Warning on ordinary deploy/restart warm-up, for the standard
multi-input calc-tag shape. Fixed by gating evaluation itself on
"all deps arrived", not just the publish.
See code-reviews/Driver.Calculation/findings.md for full detail (5 findings:
001-004 Resolved, 005 Open/cross-module).
Findings from the per-module review of Driver.Mqtt at b990adae
(code-reviews/Driver.Mqtt/findings.md records all ten; these are the fixed
ones).
Driver.Mqtt-001 (High) — MqttSubscriptionManager._subscribedTopics is a claim
about the broker session it established filters on, and AttachTo never cleared
it, but the manager OUTLIVES a session rebuild (AdoptOptions keeps it whenever
the ingest identity is unchanged). After a driver-host reconnect
(ShutdownAsync + InitializeAsync, the #527 watchdog's retry) or a
ReinitializeAsync that only changed the endpoint/credentials, the host's
re-subscribe resolved an empty pending set, sent no SUBSCRIBE at all, and left
the driver Connected/Healthy on a session carrying ZERO subscriptions — no
error, no Bad status, values frozen until the new session happened to drop.
AttachTo now forgets them, mirroring SparkplugIngestor.AttachTo's reset.
Driver.Mqtt-002 (High) — InitializeCoreAsync published _connection and
_primaryHost over whatever the driver already held, so a second initialize
stranded a connected, self-supervising broker session (plus a second Sparkplug
primary-host claim) that no later dispose could reach. Reachable without any
teardown of ours via DriverInstanceActor.AdoptConfigDuringInit. Now tears down
first, under the gate both callers already hold.
Driver.Mqtt-003 (Medium) — connectTimeoutSeconds is [Range]-annotated but
unvalidated on the runtime path; 0 makes every deadline fire immediately and a
negative value throws ArgumentOutOfRangeException out of ShutdownAsync and
DisposeAsync. Normalized once, at the seam every options instance passes.
Also: ShutdownAsync keeps the live last-message instant (-004), three usings
left behind by #512 removed (-005), stale remarks corrected (-009, -010), and
docs/drivers/Mqtt.md now distinguishes the two kinds of reconnect.
Tests 708 -> 713. All three regression tests were confirmed to fail against
the un-fixed code, with two controls that pass in either state.
Driver.Sql-001 (Medium): SqlCatalogLoader.ReadDefaultSchemaAsync documented a
fallback to the sole visible schema when the server reports no default schema and
did not implement it, so a login with no DEFAULT_SCHEMA (or a Postgres search_path
that resolves to nothing) rejected EVERY unqualified authored table at the catalog
gate — a whole-driver BadNodeIdUnknown outage against a healthy database. Fixed;
SqlCatalogDefaultSchemaTests written red first, with a control proving the
ambiguous multi-schema case still refuses to guess.
Driver.Sql-003 (Medium): docs/drivers/Sql.md spelled the WideRow row selector at
the top level, but the parser reads those keys only inside a nested rowSelector —
a tag authored from the doc is refused and reads BadNodeIdUnknown (the #552 class,
via the documentation). Also removed the documented-but-nonexistent pollIntervalMs
per-tag field.
Driver.Sql-007 (Low): dropped the unused Microsoft.Data.SqlClient PackageReference
left behind by the .Dialect extraction (#526).
Open: 002 (no IN-list chunking vs the 2100/1000 provider bind caps), 004, 005, 006,
008, 009, 010.
Driver.Historian.Gateway-003 (Medium): both were exercised only by the
env-gated Category=LiveIntegration suite (skips cleanly offline) or only
indirectly through FasterLogHistorizationOutboxTests with single-char
ASCII tags. Adds direct offline unit coverage for the value writer's
ack-to-bool mapping, timestamp presence/absence, and non-throwing
contract (via the existing FakeHistorianGatewayClient double), and for
the serializer's Serialize/Deserialize round trip incl. a multi-byte
UTF-8 tag, an empty tag, and ushort/double boundaries.
Driver.Historian.Gateway-001 (High): AlarmEventMapper wrote the alarm's
Severity wire property as the .NET enum name (e.g. "High"), but
EventMapper's read side int.TryParse()s that same property and silently
falls back to the minimum severity on any unparseable value - so every
alarm event's severity round-tripped through this module's own mappers
collapsed to 1 (Low), contradicting AlarmHistorianEvent.Severity's own
documented contract ("mapped to Historian's numeric priority"). Now
mapped through a numeric priority mirroring the existing
NativeAlarmProjector.MapSeverity convention (200/500/700/900).
Driver.Historian.Gateway-002 (Low): GatewayTagProvisioner never disposed
its IHistorianGatewayClient, unlike its GatewayHistorianDataSource /
GatewayAlarmHistorianWriter siblings. Now implements IAsyncDisposable so
the DI-owned singleton's gRPC channel closes gracefully at shutdown.
Both TDD: new/corrected tests confirmed RED against the pre-fix code.
SparkplugRebirthScope (the AdminUI's typed input) rejects '+'/'#' and its own
remarks say it is 'advisory, not the gate' because MqttBrowseSession
.ResolveRebirthTargets 'refuses an unusable one BEFORE any message leaves'. That
claim was false: the bare {group}/{edgeNode} branch — honoured on purpose, since a
node that has not birthed is the prime rebirth target — gated on IsUsableLabel,
which checks length and control characters only. A scope of 'Plant1/#' therefore
published spBv1.0/Plant1/NCMD/# , a PUBLISH the MQTT spec forbids and a conformant
broker answers by closing the browse connection; '+' would have fanned one
operator click across every matching edge node.
The check is a separate IsUsableRebirthSegment rather than a tightening of
IsUsableLabel: that helper also judges Sparkplug METRIC names, which are arbitrary
publisher-chosen UTF-8 and never part of a topic, so tightening it there would drop
legitimate metrics from the tree. Cases 1 and 2 cannot reach this shape — their ids
come off SparkplugTopic.TryParse, which already rejects both characters.
Driver.Mqtt.Tests 724 (was 708).
Joseph Doherty