The truncated-snapshot caveat moved from poll-only to push-driven. The page
already held an in-process alarm-feed subscription for the provider badge; it
now also handles the feed's snapshot_status frame, so a capped provider fetch is
caveated when the monitor decides it rather than up to three seconds later.
The poll's assignment stays as the reconcile baseline — both sources read the
same monitor verdict, and the frame is consumed, never synthesized page-side.
StreamAsync primes every subscriber with a snapshot_status frame at open, so a
page attaching mid-truncation needs no priming logic of its own; the loop is
renamed StatusFeedLoopAsync because it now feeds two indicators, not one.
Both lists predate the scope rename and would mislead anyone creating a key:
CLAUDE.md's Authentication section still named the pre-rename scopes, and
docs/Authentication.md's ops.alice example passed 'read,write', which
GatewayScopes.ValidateScopes rejects outright. Same defect family as the
Build/Test/Run sample fixed in a5f843c.
Recorded as a follow-up: code review finding Server-012 claims it fixed the two
CLAUDE.md lists on 2026-05-18, but neither correction was present — a Resolved
finding is not re-examined, so the sibling Server-0xx doc resolutions want a
spot-check for the same pattern.
The prior plan's "Follow-ups recorded, not started" block described pre-branch
behavior; every item is now closed, narrowed, or restated with its evidence, so
the block no longer misleads a reader who lands on it first. The stale Rust-guard
bullet is corrected in place rather than deleted: Check 3 always existed, and
saying so is the only way the reader learns what the real (one-directional) gap was.
Also fixes the CLAUDE.md apikey sample, which named a verb the parser has never
accepted ('create'; only 'create-key' exists, no alias), omitted the required
--key-id, and listed non-canonical scope strings that GatewayScopes now rejects
at create time — the sample could not have run.
The final integration review's non-blocker reservations, all documentation
or comment truth except one test arm.
The alarm feed opens provider_status -> snapshot_status -> cached
active_alarm -> snapshot_complete, which is what GatewayAlarmMonitor has
done since the snapshot_status frame landed. Two places still described
the old order: docs/Grpc.md said provider_status arrived *after* the
initial snapshot, contradicting its own snapshot_status section two
paragraphs down, and AlarmFeedMessage's leading proto comment named
neither status frame at all. Both now state the sequence the monitor
emits, so a client author reading either one gets the frame order right.
The proto comment change flows through the generated trees (Contracts,
Go, Java) and the client descriptor set; the Rust vendored copy stays
byte-identical to canonical. Python's generator does not carry proto
comments into its output, so it has no delta.
AlarmsHubPublisherTests' valueless-payload case covered snapshot_complete
and provider_status but not snapshot_status, leaving the newest arm
unpinned against the redaction switch that must ignore it. Added.
WnWrapAlarmConsumer's ack comment led with the 2026-05-01 reading that
-55 tracks the 8-arg overload, then refuted itself six lines later with
the 2026-08-18 probe. It now leads with the observation labelled as
narrower than it reads -- mirroring the correction already in
docs/AlarmClientDiscovery.md -- so the block argues one thing: the 6-arg
call site stays for parity, and rc semantics are per the probe. A
paragraph orphaned by an earlier splice is rewrapped. Comment interior
only; the file compiles on Windows.
TST-16 gets a dated closure note rather than a rewrite: the flag it
called dead was implemented 2026-08-18. GatewayDashboardDesign's /browse
paragraph gains the failed-read carve-out GatewayConfiguration already
documented, so the two agree that a failed read keeps its - placeholder.
Follow-up to 90331b6. Three comment/prose corrections, no behaviour change.
AcknowledgeByName's comment still said the 6-arg overload "works and reaches the
alarm-history path correctly", which the same commit's own findings contradict in
three other places. It now says what was observed: rc=0 means accepted, not
applied — the 2026-08-18 probe acked a live alarm six ways and the snapshot, the
OPERATOR_NAME field, and the extension's .Acked attribute all stayed put. The -55
tracks the consumer, not the overload. Subscribe's comment gets the same
treatment: "lets AlarmAckByName succeed" becomes "return rc=0".
AlarmProbeFindings.md said the re-raise arrives as "a separate record" alongside
the returned one, which reads as coexistence and is wrong. The snapshot carries
one record per tag: the cap=1024 replies bracketing the re-raise are both
elementCount=3 (one per TestMachine_00{1,2,3}) at an identical 1613 bytes, and the
old GUID is absent from the later one. The re-raise replaces the record, so a
single poll spanning it sees the Clear and the Raise together.
Worker diff verified strictly comment-only; builds x86 on windev, 0W/0E.
AuthenticateUser("Administrator", "") + WriteSecured raises the alarm UDAs that
plain Write could not touch (SecurityError detail=1008), so the 2026-08-17 blocker
was the verb, exactly as that run's own Unblocking list predicted.
Two of the three open questions are now observed rather than assumed:
- ALARM_RECORDS/@COUNT reports the records in the reply, not the total active
count. With three alarms active it read 1 at cap 1 and 2 at cap 2. There is no
exact truncation signal to switch to, so IsTruncatedFetch's conservative rule is
the design rather than a placeholder — behaviour unchanged, only the comments.
- Clear-then-re-raise mints a new GUID; the ALM->RTN leg keeps its GUID
(reconfirming the 2026-05-01 capture). ComputeTransitions already reads the
re-raise correctly as one instance ending and another beginning.
The acknowledge leg stays unobserved for a narrower reason: every wnwrap ack
surface is inert on this rig. AlarmAckByName returns 0 from the ack-only consumer
and -55 from the SetXmlAlarmQuery-applied one, for both the 6-arg and 8-arg forms,
and neither the snapshot STATE, OPERATOR_NAME, nor the extension's own .Acked
attribute moves. That corrects AlarmClientDiscovery.md, which read the zero return
as a working ack.
Comment- and prose-only; no behaviour change. The three throwaway probes ran from
the windev CI clone and were deleted; that clone is a clean tree at ab3ff16.
Task 1 added AlarmSnapshotStatus and AlarmFeedMessage.snapshot_status = 5.
Carry it downstream from the canonical Contracts protos:
- Rust vendored protos under clients/rust/protos, refreshed byte-identical
(build.rs falls back to them for out-of-repo tarball builds)
- client descriptor set (protoc 34.1 pin)
- Go (protoc-gen-go v1.36.11 / protoc-gen-go-grpc 1.6.2)
- Python (grpcio-tools 1.80.0 pin)
- Java (gradle generateProto)
.NET needs no regeneration: the client compiles against the Contracts
Generated/ output committed with the proto change.
The hand-written CLI feed renderers switch on the payload oneof, so codegen
alone does not carry the arm. Add snapshot-status to the .NET, Go, Rust, and
Java renderers; the .NET and Go renderers were also missing provider-status,
which has been on the wire since the provider-mode work, so add it there too.
Java's renderer is an exhaustive switch expression and did not compile until
the new case landed. The Python CLI renders generic protobuf-JSON and needs
no change.
Each client README gains a paragraph on the feed-level frame next to its
existing from_truncated_snapshot paragraph: it arrives at stream open after
provider_status and before the cached active_alarm frames, then on every
verdict change including the clearing frame a monitor restart emits, so a
live consumer can track set completeness without polling QueryActiveAlarms.
GatewayDashboardDesign: list the two new payload cases the AlarmsHub forwards,
and — separately — record the GroupToRole / GroupToTag / UntaggedSessionVisibility
rows the settings page already renders but the bullet list omitted.
The form split tags with the shared ParseList and attached the result verbatim,
so "team-a, TEAM-A" persisted as two entries and the constraints column read
dashboard_tags=[team-a, TEAM-A] — one grant reported as two on the page whose
job is to show what a key was granted. Enforcement never saw it (a session holds
its tags in a case-insensitive set), which is exactly why the display was the
only place it could surface.
De-duplicates ordinal-ignore-case at the attach point only, first spelling
winning, matching ApiKeyAdminCommandLineParser.ParseDashboardTags. ParseList is
untouched: the five glob lists are matched literally, so near-duplicates there
are not necessarily the same rule and must survive verbatim — pinned by a test.
The help text claimed to mirror the CLI flag; it now claims only the shared
separators and the dedupe, since the form still drops an empty segment silently
where the CLI hard-fails. A browser form has no exit code to fail with, so that
difference stays, and Authorization.md now records it.
The constraints column enumerated only the eight positional ApiKeyConstraints
members, so a key whose sole recorded policy was a dashboard tag summarised to
an empty string and rendered as "-" — the same cell a key with no policy at
all gets. ApiKeyConstraints.IsEmpty counts DashboardTags, so that key is not
unconstrained, and the column was quietly telling operators otherwise about a
grant that decides who can watch a session's events.
The create form had no dashboard-tags input either, so tagged keys could only
be minted from the apikey create-key CLI. Adds the field beside the other
constraint lists (same ParseList separators) and attaches it through the
record's init-only member, since it postdates the eight-member constructor.
CreateModel, OpenCreateDialog and TryBuildCreateRequest widen to internal for
the new render tests: the create form is behind a click and static rendering
cannot dispatch one. That is the assembly's existing InternalsVisibleTo seam.
The per-session dashboard event ACL shipped in 693a78d + 7ec0b35 with unit
coverage over a fabricated principal. What a fabricated principal cannot show is
that the group names the shared directory actually returns -- short RDN values,
not DNs -- are the ones Dashboard:GroupToTag keys match. Two [LiveLdapFact]s
close that: gw-viewer binds for real, its GwReader membership grants team-a, and
IDashboardSessionAcl then admits a team-a-tagged session and refuses a
team-b-tagged one; multi-role takes the Administrator bypass. The mapping is
config-side only -- no GLAuth entry, group, or membership was added, and
glauth.md records that explicitly so a future reader does not go looking for a
directory change that never happened.
multi-role is a member of GwReader as well as GwAdmin, so it holds team-a too.
Its bypass is therefore asserted on team-b and on the untagged session -- the two
it would lose if the Administrator branch were ever dropped -- rather than on
team-a, which would pass either way.
One cheap hardening from a prior review: a GatewayOptionsTests case binds
Dashboard:GroupToTag through a real ConfigurationBuilder and looks the group up
mis-cased. The property initializer seeds an OrdinalIgnoreCase dictionary, but
only the binder decides whether that instance survives; if it did not, a
mis-cased group name from the directory would grant no tags and the ACL would
deny with no diagnostic.
Docs follow the shipped shape: docs/Sessions.md gains the session-tag model
(owner-key sourced, immutable, visibility-not-access), gateway.md and CLAUDE.md
gain the ACL in their dashboard-auth paragraphs, and three
GatewayDashboardDesign.md passages that still described the ACL as outstanding
now describe both gated seams and the decision order. GatewayConfiguration.md's
ShowTagValues row no longer claims the redaction is the only thing between a
Viewer and another session's values -- it is now the second of two independent
layers. gateway.md's hub-token lifetime corrected 30 minutes -> 5, matching
HubTokenService. Authentication.md disambiguates --dashboard-tags as the only
constraint flag that splits on commas. The plan doc header is Implemented; its
as-built section 12 already existed and is not duplicated.
Verified: NonWindows.slnx builds clean; GatewayOptions/DashboardSessionAcl/
EventsHub filters 37/37; the live-LDAP suite skips cleanly without the env var
and runs 7/7 green against the shared GLAuth with it.
Follow-up to the per-session event ACL. Part of that change rode into 693a78d
via a concurrent agent's pathspec-less commit; this commit carries the review
fixes and uses pathspecs on the commit itself so it cannot recur in either
direction.
Gating the page's subscribe seam made AttachEvents asynchronous — it awaits the
authentication state — and that await is a suspension point the synchronous
version did not have. On a rapid A -> B navigation the suspended A continuation
resumes after B's parameter set has run to completion, re-reads the live
SessionId (now B's), and attaches B a SECOND time. The ACL is not bypassed —
the newer attach already cleared that same session — but the fields holding B's
first subscription are overwritten in place, so nothing ever disposes it: its
EventsHubViewerRegistry entry is never released, which keeps the mirror cloning
events for a session the page is no longer watching through that handle, and
its pump is never cancelled. A resource leak the ACL work introduced.
OnParametersSetAsync now claims a monotonic _attachGeneration synchronously,
before its first await, and AttachEventsAsync re-checks it after the await and
before any field write or Subscribe call. A stale attach returns rather than
detaching: it owns nothing, and tearing down there would destroy the newer
attach's subscription. DetachEventsAsync needs no such guard — it captures and
nulls the live fields synchronously before it awaits, so a resumed detach only
unwinds what it already took ownership of. Same dispatcher-owned identity idea
as the existing ReferenceEquals guards in PumpEventsAsync and
MarkDisconnectedAsync, one level up.
The interleaving is not expressible with the static HtmlRenderer idiom the other
page tests use: it renders a root component once and exposes no parameter-update
seam. The new test therefore adds a minimal Renderer subclass whose only job is
to mount a component and drive a second SetParametersAsync into it while the
first is parked on a gated AuthenticationStateProvider. That subclass is the
lone reason for a narrowly scoped BL0006 suppression, justified in place: it is
test-only scaffolding that never ships, and the cost of the warning coming true
is a compile break in one test file on an SDK bump. Confirmed non-vacuous by
mutation — with the generation check disabled the test goes red on the doubled
subscription and the two passing ACL tests stay green.
Two decision-table corners are now pinned rather than implied. Admin x
nonexistent session id resolves to ALLOW, because the admin bypass is evaluated
before the registry lookup; a plausible "look the session up first, it reads
better" refactor would flip it, so a test documents the ordering. EventsHub's
remarks said "an unknown session id is denied" without qualification, which read
as universal; they now state that the bypass is checked first and every rule
below it is a non-Admin rule.
HubTokenServiceTests gains the truly-absent-field case: a hand-built payload
JSON with no Tags key at all, protected through the same purpose, which is the
shape every in-flight token has across the deploy that introduces the field. The
existing test covered present-but-empty, which does not exercise the null
coalesce that stands between a legacy token and a crash on the hub auth path.
ProtectorPurpose became internal so the test cannot drift from the real purpose
string.
Tag-count cardinality cap considered and recorded as a deliberate non-goal.
Build 0 warnings / 0 errors; 48 filtered (ACL/hub/token/page) and 257 dashboard
tests pass.
The truncation-cliff fix made alarm transitions truncation-safe but silent:
when GetXmlCurrentAlarms2 returns exactly maxAlmCnt records the worker
suppresses absence-implies-Clear inference and says so only in a rate-limited
stderr warning. No client and no operator could tell a complete active set
from a capped one.
Two additive proto3 booleans carry the verdict out:
- QueryActiveAlarmsReplyPayload.snapshot_truncated = 2 (worker IPC reply)
- ActiveAlarmSnapshot.from_truncated_snapshot = 16 (per record)
The per-record field is not an aesthetic choice. QueryActiveAlarms returns a
bare `stream ActiveAlarmSnapshot` with no envelope, header, or trailer, so a
per-record boolean is the only carrier that stays wire-compatible; an envelope
message would change every existing client's stream element type. The reply
payload states it too because a prefix filter can leave zero records and a
truncated fetch with nothing to report still has to say so. The flag means
"this set may be incomplete", never "this record is unreliable" — it is
independent of the subtag-fallback `degraded` field.
Detection is deliberately UNCHANGED: IsTruncatedFetch remains
`fetchedRecordCount >= maxAlarmsPerFetch`. The live probe (docs/AlarmProbeFindings.md,
ce5d8ae) could not verify whether ALARM_RECORDS/@COUNT reports the total active
count or only the records in the reply, so @COUNT is not parsed for detection;
switching to it stays blocked on probe evidence. The probe's comment
annotations in WnWrapAlarmConsumer.cs are preserved.
Reset semantics: not latched. WnWrapAlarmConsumer.FoldFetch replaces the
verdict on every poll under the same lock as the snapshot merge, so the first
sub-cap fetch clears it; GatewayAlarmMonitor.ClearCache drops it with the cache
generation it describes. A caveat that never turns off is one operators learn
to ignore.
Flow: WnWrapAlarmConsumer.LastSnapshotTruncated -> AlarmDispatcher (stamps every
record) / IAlarmCommandHandler (payload) -> MxAccessCommandExecutor reply ->
GatewayAlarmMonitor._snapshotTruncated -> IGatewayAlarmService.SnapshotTruncated
-> DashboardAlarmQueryResult -> AlarmsPage warning banner (render-side only; the
poll loop and DisposeAsync drain are untouched). The public QueryActiveAlarms
RPC forwards worker snapshots unmodified, so the per-record flag needed no
mapper change — a test pins that.
Parity: this describes OUR fetch mechanics — additive gateway metadata — not
MXAccess provider behavior. No event is synthesized and no MXAccess-observable
semantics change, so it is not a parity deviation.
Tests: worker LastSnapshotTruncated set/reset/consecutive-burst (windev-run);
gateway end-to-end truncated reply -> monitor -> public stream, with the
complete-reply control as the load-bearing assertion; AlarmsPage banner
present/absent. Docs: gateway.md alarm surface, docs/DesignDecisions.md entry.
Both questions stay open, and the reason is the finding: the dev rig's alarm
UDAs reject a plain MXAccess Write with SecurityError/detail=1008 from the
responding automation object, so no alarm instance can be created to follow
through an acknowledge and no population can be built to overflow a capped
fetch. The rig is otherwise live — objects deployed and on scan, wnwrap
subscribed, GetXmlCurrentAlarms2 returning well-formed XML — which is what
makes the blocker specific and the unblock (engine-side script, or
AuthenticateUser + WriteSecured, or reclassifying the UDAs) actionable.
Comment-only changes in WnWrapAlarmConsumer: scope the GUID-identity claim to
the leg live capture actually covers, and record that ALARM_RECORDS/@COUNT
exists as a candidate exact truncation signal but is deliberately not trusted
because its semantics under a capped reply are unverified. No behavior change.
Adds a dashboard event-visibility tag to ApiKeyConstraints, riding in the existing
constraints JSON blob so no auth-store schema migration is needed (design
docs/plans/2026-07-10-dashboard-session-acl-tst15.md sections 3/3.1, open call
settled per its own recommendation). The tag is visibility-only: no read, write,
browse, or subscribe path consults it, and HasRead/HasWriteConstraints ignore it.
GatewaySession gains an immutable, ordinal-ignore-case Tags set stamped at
construction from the owning API key, forwarded by MxAccessGatewayService.OpenSession
from the resolved ApiKeyIdentity — never from the wire request, so a client cannot
label its own session with another tenant's tag. ISessionManager gains a tag-carrying
OpenSessionAsync overload whose default implementation forwards to the tagless one, so
an implementation that does not model tags opens an untagged (least visible) session.
apikey create-key gains --dashboard-tags team-a,team-b (repeatable, trimmed,
de-duplicated; an empty segment is rejected rather than dropped) and list-keys prints
the tags column. No enforcement yet — the EventsHub ACL that consumes the tag is a
later change.
WriteAsync enqueued its frame and then contended unconditionally for the write
lock, so a caller that lost the race stayed in WaitAsync until the winning
drainer released — even though that winner writes, flushes, and completes the
loser's control frame at the control-to-event class boundary, part-way through
its pass. The boundary flush made the delivery point honest; the awaited task
was still charged for the whole event backlog it had just been flushed ahead of.
WriteAsync now awaits its own frame's completion racing the lock acquisition.
Completion first: the caller returns at its frame's delivery point and the
outstanding acquisition is detached, not dropped — a continuation drains
whatever is queued and releases, so the lock is never acquired and silently
held and a frame enqueued between the previous drainer's last dequeue and its
release is still written. Lock first: drain as before. Cancellation keeps the
WRK-22 tombstone semantics exactly, and a wait cancelled after the caller has
already detached releases nothing (SemaphoreSlim hands no count to a wait it
cancels), so no count leaks and no queued frame is stranded. A token that fires
after the frame's completion won the race changes nothing — the frame was
delivered. WriteBatchAsync deliberately keeps the plain wait-then-drain shape:
its last completion resolves at the end-of-pass flush anyway.
Three tests: the latency win (a control caller returning while the winning
WriteBatchAsync event burst is demonstrably still blocked mid-pass), a
mixed-priority concurrency soak pinning exactly-once writes and a single
drainer, and the cancel-after-detach corner (a wrongly released count would
surface as the drainer's own Release throwing SemaphoreFullException).
edited on macOS, windev verification pending (plan Task 11). Verified here by
compiling and running WorkerFrameWriter plus the writer suite against net10.0
in a scratch harness: 31/31 pass, and the two behaviour-pinning tests fail
against the pre-change parked implementation.
Groundwork for the per-session dashboard event ACL (docs/plans/2026-07-10-dashboard-session-acl-tst15.md 3.2): a dashboard group can now grant visibility tags, and untagged sessions default to AdminOnly. Enforcement lands with the EventsHub ACL; nothing consumes the grant yet.
GroupToTag is deliberately uncoupled from GroupToRole - a group may appear in either map, both, or neither - and is validated for shape only. Tags gate dashboard event visibility, never data access.
ISessionManager.ReadEventsAsync had zero production call sites: the worker
event channel is drained once by GatewaySession.MapWorkerEventsAsync (the
distributor pump), and every consumer — gRPC subscribers, the dashboard
mirror, the alarm monitor — attaches to the distributor. The interface
member, SessionManager's forwarder, and GatewaySession.ReadEventsAsync are
gone; IWorkerClient/WorkerClient.ReadEventsAsync is untouched, it is the
live worker-channel claim.
No test was removed or rewired: nothing invoked the member through the
interface. Nine ISessionManager test fakes carried a required-member stub
(seven threw NotSupportedException or yielded nothing; EventStreamServiceTests
and GatewaySessionDashboardMirrorTests forwarded to the session; the two
MxAccessGatewayService fakes yielded their Events list) — all nine stubs were
deleted. The MxAccessGatewayService suites' streaming tests already run
through FakeEventStreamService, which reads the same Events list, so their
coverage is unchanged; only the now-inaccurate doc comments on Events /
LastReadEventsSessionId were reworded.
The MapWorkerEventsAsync comment no longer describes a twin to keep in step;
it now states the single-reader claim directly. docs/Sessions.md drops
ReadEventsAsync from the SessionManager member list and from the Run-state
prose. The 2026-08-15 deferred-remediation as-built note records the removal.
The two-class writer already got control bytes out ahead of a queued event
backlog, but a frame counts as delivered only once flushed, and the drain
deferred its single FlushAsync — and every TrySetResult — to the end of the
pass. A heartbeat, command reply, fault, or shutdown ack was therefore written
first and completed last, behind up to a full 128-frame event batch.
The drain now records each frame's priority class on PendingFrame and flushes
at every control-to-event boundary, completing and clearing the written set
there. Cost stays bounded: a pure-event pass still pays exactly one flush, a
run of control frames still pays one for the run, and only a pass that mixes
both classes pays a second — never one flush per control frame, the
syscall-per-heartbeat cost WRK-12 removed.
A boundary flush that itself fails is a new failure window and is handled like
the end-of-pass flush failure, additionally failing the event frame the drain
had already claimed off its queue and every frame still queued. Frames a
boundary flush completed leave the written set, so a later failure in the same
pass can no longer reach back and fail an already-delivered control frame.
The awaited task of a caller that lost the write-lock race is still bounded by
the winning drainer's pass — that enqueue-then-contend parking is unchanged and
now documented on WriteAsync and in docs/WorkerFrameProtocol.md.
Joseph Doherty