act host-mode on Windows does not materialize a JS action's dist/ (actions/checkout
fails with "Cannot find module .../dist/index.js"), so the windows + live-mxaccess
jobs check the repo out with plain git instead (public repo, anonymous https clone
of GITHUB_SHA). The self-hosted windev runner otherwise works (host mode, persistent).
Claude-Session: https://claude.ai/code/session_01DMXXvNuPekkkrTEyPNxEkW
CI now runs end-to-end: portable + java jobs green on fix/ci-selfhosted-tooling.
Flip the tracker status and record the root cause (runner container.network) and
the five real defects the run surfaced and fixed. windows/live-mxaccess jobs
remain queued pending a self-hosted windev runner.
Claude-Session: https://claude.ai/code/session_01DMXXvNuPekkkrTEyPNxEkW
The last portable failure: pytest-asyncio 1.x does not install a current event
loop for synchronous tests and clears it after each async test, so a later sync
test that builds a grpc.aio.Channel (create_channel) hit
"RuntimeError: There is no current event loop" on Python 3.12 depending on test
order (CI hit it; local order did not). Add an autouse conftest fixture that
ensures a usable current loop per test. Local: 127 passed, 1 skipped.
Claude-Session: https://claude.ai/code/session_01DMXXvNuPekkkrTEyPNxEkW
Two real defects the CI run surfaced (driving to fully green):
- OrphanWorkerTerminatorTests hard-coded a Windows path (C:\...); on the Linux
runner Path.GetFullPath does not root it, so the terminator's exact-path match
found nothing (Expected 2/1 kills, Actual 0). Use an OS-appropriate normalized
path so the matching logic is validated on every platform (5/5 pass locally).
- clients/java generated MxaccessWorker.java was stale — missing the canonical
max_frame_bytes field. Refreshed via `gradle generateProto` (protobuf 4.33.1,
the pinned version — the diff is purely the field, no version churn). Broaden
the ci.yml churn-revert to both generated aggregates so verify-clean still
guards real drift elsewhere.
Claude-Session: https://claude.ai/code/session_01DMXXvNuPekkkrTEyPNxEkW
Surfaced by the first real CI run on the self-hosted runner:
- scripts/check-codegen.ps1: `git status --porcelain` returns $null on a clean
tree, so `.Trim()` threw "cannot call a method on a null-valued expression" —
i.e. Check 2 could never report success. Pipe through Out-String (null -> '').
- clients/rust/protos/mxaccess_worker.proto: genuinely stale — missing the
canonical `max_frame_bytes` field. Refreshed from
src/ZB.MOM.WW.MxGateway.Contracts/Protos (Check 3 drift guard, added in P1).
- ci.yml java job: act cannot resolve the gradle/actions monorepo action
(`unsupported object type` on the v4 tag); install Gradle 9.5.1 directly instead
(matches the local build) since the repo has no gradle wrapper.
Claude-Session: https://claude.ai/code/session_01DMXXvNuPekkkrTEyPNxEkW
The portable and java jobs assumed tools GitHub-hosted runners preinstall but the
self-hosted catthehacker act image lacks: `pwsh` (the codegen freshness check uses
`shell: pwsh`) and `gradle` (the java job runs `gradle test`; the repo has no
wrapper). Install PowerShell as a .NET global tool (the SDK is already set up) and
provision gradle via gradle/actions/setup-gradle. No behavior change on
GitHub-hosted runners, which already have both.
Claude-Session: https://claude.ai/code/session_01DMXXvNuPekkkrTEyPNxEkW
Gateway-side follow-up to the shared auth-lib expiry core (delivered via G-2):
- apikey create-key gains optional --expires — absolute ISO-8601 UTC or relative
<N>d/<N>h from now; omitted means non-expiring (opt-in, unchanged default).
Threaded ApiKeyAdminCommand -> parser -> runner into the library's
CreateKeyAsync(..., expiresUtc, ...). list-keys shows an expiry column and an
active/expired/revoked status.
- Dashboard API Keys page surfaces expiry: an Expires column (Never when unset)
and a status badge reading Expired (red) / Expiring (<=7d, amber) / Revoked /
Active. DashboardApiKeySummary.ExpiresUtc projected in DashboardSnapshotService;
StatusBadge maps the new states.
Tests: parser (absolute/relative/invalid/none) + end-to-end past-expiry rejection
through the live verifier; dashboard summary suite green. Docs: Authentication.md
(verification-flow expiry step, CLI table + examples, dashboard badge).
Closes the tracked SEC-10 polish (SEC-10 core was already Done).
The read loop awaited EnqueueWorkerEventAsync inline, which blocks in the bounded
event channel's timed WriteAsync (up to EventChannelFullModeTimeout, default 5s)
when the channel is full with no/slow consumer. A WorkerCommandReply or heartbeat
queued behind an event frame was then stalled, so an in-flight InvokeAsync could
hit CommandTimeout even though the worker replied in time.
Mirror the existing outbound WriteLoopAsync: add an unbounded event staging
channel and a dedicated EventWriteLoopAsync. DispatchEnvelope is now fully
synchronous — the WorkerEvent branch hands the event off with a non-blocking
TryWrite and the read loop never awaits. The event write loop owns the timed
WriteAsync into the bounded channel and the sustained-overflow ProtocolViolation
fault (unchanged contract). Registered in WaitForBackgroundTasks + completed on
close/fault/dispose.
Test: reply arriving after events with a full, consumer-less event channel is
dispatched promptly (no CommandTimeout) — pipe-harness, verified on windev.
Docs: GatewayProcessDesign read/write/event-loop section.
Follow-up to the IPC-02 proto change and IPC-03 headroom default:
- regenerate clients/proto/descriptors/mxaccessgw-client-v1.protoset with the
pinned protoc 34.1 so it carries GatewayHello.max_frame_bytes (IPC-01 descriptor
refresh; ClientProtoInputTests.Descriptor_ContainsEveryContractMessageAndField).
- update GatewayOptionsTests design-default assertion to 16 MiB + 64 KiB reserve.
Worker half of the Wave 3 size/backpressure + write-ordering pass:
- IPC-02: the worker adopts GatewayHello.max_frame_bytes during the handshake
(WorkerFrameProtocolOptions.AdoptNegotiatedMaxMessageBytes) instead of a
hard-coded default; 0 keeps the default, a value above a 256 MiB ceiling is
rejected. Reader and writer share the options instance, applied before the
message loop.
- IPC-04: DrainEvents caps each reply at MaxDrainEventsPerReply (10_000) and
treats max_events = 0 as that cap rather than 'drain the entire queue', so one
diagnostic drain cannot pack a session-killing reply frame.
- WRK-04: WorkerFrameWriter stamps the envelope Sequence at the actual point of
writing (under the write lock) instead of at envelope creation, so the on-wire
order and the stamped sequence always agree under concurrent producers.
- WRK-07: the writer is now a cooperative priority scheduler — callers enqueue at
Control or Event priority and the draining lock-holder writes all control
frames before any event frame, so replies/faults/heartbeats jump ahead of an
event backlog. Per-frame validation/size rejections fail only that frame; a
stream write failure fails all queued frames.
Tests: monotonic gap-free sequence under concurrency, control-before-event
priority (gated stream), negotiated-max adoption, DrainEvents zero-bound.
Worker builds x86 only — verified on windev.
Proto foundation + gateway-side of the size/backpressure pass:
- IPC-02: add GatewayHello.max_frame_bytes (regen Generated/); gateway sends
its negotiated worker-frame max in the handshake so the worker can adopt it
instead of a hard-coded default. Worker read-half lands separately.
- IPC-03: give the pipe frame max envelope-overhead headroom above the public
gRPC cap (WorkerFrameProtocolOptions.EnvelopeOverheadReserveBytes = 64 KiB;
default Worker.MaxMessageBytes bumped to 16 MiB + reserve), cross-validate the
headroom at startup, and pre-check command envelope size in WorkerClient so an
oversized command fails only that correlation (ResourceExhausted) instead of
faulting the whole session.
- IPC-04: reject DrainEvents max_events above a public ceiling in the request
validator (worker per-reply cap lands with the worker half).
Docs: GatewayConfiguration, WorkerFrameProtocol, gateway.md.
Tests: headroom validation, DrainEvents bound, oversized-command per-command
failure (pipe-harness, verified on windev).
Full-host tests passed in isolation but failed 2-4/run under parallel execution on
Windows with 'the process cannot access the file ... because it is being used by
another process' (macOS never sees it — Unix deletes open files). Two shared-state
parallel collisions, discovered while verifying TST-08:
1. Self-signed cert generation. GatewayTlsBootstrapTests sets process-global
Kestrel/TLS env vars that GatewayApplication.Build reads; a parallel host-building
test inherits them mid-run and both generate a cert at the same path, racing on the
fixed-name '<path>.tmp'. Fixes:
- SelfSignedCertificateProvider: stage the PFX in a unique '<path>.<guid>.tmp' instead
of a fixed name, so concurrent/interrupted writers never collide on the temp file
(real robustness: two instances or a restart-during-write no longer clash). The
final atomic Move (last-writer-wins) still yields an equivalent cert.
- TestHostEnvironmentInitializer: default MxGateway__Tls__SelfSignedCertPath to a
per-process temp path so the suite never writes the shared ProgramData default or
fights the deployed service; first host-building test generates, the rest load it.
- GatewayTlsBootstrapTests: [Collection] with DisableParallelization so its global
env-var mutation cannot bleed into parallel tests (new GlobalEnvironmentCollection).
2. AuthStoreHealthCheckTests opened a real SQLite file; Microsoft.Data.Sqlite's
connection pool keeps the .db handle open after 'await using', so the finally
File.Delete threw. Reuse the existing TempDatabaseDirectory helper, which calls
SqliteConnection.ClearAllPools() before deleting.
Server build clean; affected classes 17/17 on macOS. Windows full-suite verified separately.
Empirically verified the full ZB.MOM.WW.MxGateway.Tests suite exits cleanly:
- macOS: 0 surviving testhost after a full run.
- windev (origin/main worktree): 0 new testhost and 0 new worker processes
after a full run, isolated by process StartTime.
Static audit confirms the prime-suspect fixtures already dispose deterministically
(WorkerClient cancels _stopCts + awaits its read/write/heartbeat tasks with a 5s
timeout; GatewaySession disposes the client; harness/E2E fixtures use await using).
Drop the stale 'leaves orphaned testhost processes' rationale from CLAUDE.md's
Source Update Workflow; keep filtered-run guidance as a speed guideline only.
Separately discovered (tracked in 00-tracking.md change log, NOT part of TST-08):
Windows-only temp-file-lock flakiness in ~4 full-host test classes (Sqlite
connection-pool retaining a temp .db handle; SelfSignedCertificateProvider's
fixed-name gw.pfx.tmp racing teardown) that macOS never surfaces.
- SEC-02: DashboardAuthorizationHandler restricts the loopback and
Authentication:Mode=Disabled bypasses to read-only. They now satisfy only a
Viewer-bearing requirement (AnyDashboardRole), never AdminOnly, so anonymous
localhost can view the dashboard but cannot reach API-key CRUD or session
Close/Kill at the policy layer (previously guarded only by service re-checks).
- SEC-12: DashboardSessionAdminService emits canonical AuditEvents through
IAuditWriter (actions dashboard-close-session / dashboard-kill-worker, category
SessionAdmin) on Success/Failure/Denied, mirroring the API-key audit path so
destructive session actions leave durable, queryable rows.
- SEC-20: drop the unbounded session_id tag from the exported
mxgateway.heartbeats.failed counter (per-session detail stays in the snapshot/log).
Docs updated same-change: CLAUDE.md (read-only loopback + 5-min bearer),
GatewayDashboardDesign.md (bypass scoping + session-admin audit), Metrics.md.
Server build clean; 30/30 targeted + 295/295 Dashboard/Security/App/Metrics sweep.
SEC-07: add QueryActiveAlarmsRequest -> events:read scope arm; fix two tests that
constructed StreamAlarmsRequest instead of QueryActiveAlarmsRequest.
SEC-05: shorten hub-token lifetime 30m -> 5m; document that the ?access_token= query
carriage must never be request-logged.
SEC-08: gateway-side CachingApiKeyVerifier (short TTL, keyed on a hash of the presented
secret) skips the per-call store read+last_used write; CoalescingMarkApiKeyStore bounds
last_used writes to <=1/key/min; identity constraints are cached. Invalidation is wired
at the gateway admin sites (revoke/rotate/delete); short TTL backstops out-of-process CLI.
SEC-11: fixed-window rate limit on POST /auth/login + a per-peer (key-id) failure limiter
checked before VerifyAsync; new MxGateway:Security options bound + validated.
Also fixes regressions from the SEC-01/04/06 commit (c185f62) that a narrow test filter
missed (all now covered by a full-suite checkpoint):
- Rooting check is cross-platform: accepts Windows C:\/UNC forms on Unix so the shipped
appsettings path validates on the macOS dev box, still rejecting bare filenames.
- AddGatewayConfiguration TryAdds a non-production IHostEnvironment fallback so the validator
resolves in minimal test/tooling containers; the real host + apikey CLI register the actual
environment first (TryAdd no-op there).
- Test assembly defaults ASPNETCORE_ENVIRONMENT=Development (ModuleInitializer) so full-host
tests exercise wiring instead of tripping the SEC-04/06 production guards.
- GatewayOptionsTests asserts the SEC-01 CommonApplicationData-derived default (platform-correct).
archreview: SEC-07/05/08/11 (P1). Verified: NonWindows build clean; full gateway suite
747 passed / 42 failed, where all 42 are the pre-existing macOS named-pipe-harness failures
(Unix-socket path limit) and 0 are validation/regression failures.
- SEC-01: derive AuthenticationOptions.SqlitePath / TlsOptions.SelfSignedCertPath
defaults from SpecialFolder.CommonApplicationData (was Windows-literal strings
that became CWD-relative files off-Windows); validator rejects non-rooted
overrides; delete the stray C:\ProgramData\...gateway-auth.db files that
materialized under src/; add a tree-hygiene test failing on any *.db under src/.
- SEC-04: GatewayOptionsValidator fails startup when IsProduction() && DisableLogin.
- SEC-06: validator rejects LDAP Transport=None in Production; document the
MxGateway__Ldap__ServiceAccountPassword env override + dev-credential rotation.
Galaxy SnapshotCachePath rooting could not be validator-enforced (bound by the
external GalaxyRepository package, not GatewayOptions) — documented instead.
archreview: SEC-01/04/06 (P1). Verified on the Auth-0.1.4 baseline: Server build
clean, GatewayOptionsValidator + GatewayTreeHygiene tests 53/53.
Bump all four ZB.MOM.WW.Auth.* package refs 0.1.2 -> 0.1.4. The shared
ApiKeyVerifier now rejects any key whose ExpiresUtc is in the past;
existing keys have NULL expiry (never expire), so nothing breaks, and the
auth SQLite DB auto-migrates to schema v3 (nullable expires_utc column) on
first boot. Implement the two IApiKeyAdminStore members added in 0.1.3
(SetScopesAsync/SetEnabledAsync) on the test FakeApiKeyAdminStore.
Build green; no new test failures (the macOS worker-pipe IPC failures are
pre-existing/environmental, identical to the pre-bump baseline).
- IPC-01: regenerate the stale client descriptor set and add ClientProtoInputTests
(semantic, protoc-free) so a missing contract symbol fails the build.
- IPC-20: make publish-client-proto-inputs.ps1 -Check source_code_info-normalized
so it is protoc-version tolerant.
- IPC-19: document + guard the "Generated/ must be committed for net48" rule
(docs/Contracts.md, csproj comment, check-codegen.ps1).
- IPC-09: harden the per-client generate-proto scripts (PATH resolution + pinned
grpcio/protobuf/java version asserts) so regeneration is reproducible.
- TST-03: add .gitea/workflows/ci.yml (portable/java/windows/live jobs) running the
build, tests, client checks, and the codegen guards.
- Also: check-codegen.ps1 Check 3 guards the CLI-02 vendored Rust protos against drift.
archreview: IPC-01/09/19/20 Done, TST-03 In review (pipeline authored + validated,
not yet run on a Gitea runner). Verified on macOS: NonWindows build clean,
ClientProtoInputTests 5/5, -Check exit 0.
build.rs resolved the .proto files via a monorepo-relative path and packaging
relied on `cargo publish --no-verify` to hide the resulting failure. Vendor the
three canonical protos into clients/rust/protos/, resolve them via
CARGO_MANIFEST_DIR (in-repo path preferred, vendored fallback for the packaged
crate), ship them via Cargo.toml `include`, and drop `--no-verify` from
pack-clients.ps1.
archreview: CLI-02 (P1). Verified: cargo fmt/check/test (28) + clippy -D warnings,
and `cargo package` (no --no-verify) compiles standalone in the isolated staging
dir — the out-of-repo proof. Vendored-proto drift is guarded by check-codegen.ps1.
Interlocking changes across the gateway server (shared GatewaySession.cs /
SessionManager.cs), committed together:
- GWC-01 (Critical): alarm monitor now attaches as an internal
(non-counted) distributor subscriber instead of a second raw drain of the
single worker event channel; WorkerClient._events -> SingleReader with a
claimed-once guard so a future dual-consumer regression throws loudly.
- GWC-02 (High): faulted sessions are swept in CloseExpiredLeasesAsync
(IsFaultedReapable + FaultedReason); new FaultedGraceSeconds (default 0).
- GWC-03 (High): configurable MaxSparseArrayLength (default 1_000_000)
enforced before allocation.
- TST-02 (High, security): StreamEvents attach now enforces the opening key
id -> PermissionDenied on owner mismatch.
- TST-12 (Medium): CLAUDE.md retention-defaults sentence corrected.
Verified: NonWindows build clean; targeted tests 135/135 on macOS, plus
WorkerClientTests 18/18 on the Windows host.
A long legitimate ReadBulk pumped Windows messages without refreshing
LastStaActivityUtc, so the watchdog false-positived StaHung past
HeartbeatStuckCeiling and then silently dropped every reply. PumpPendingMessages()
now calls MarkActivity() after pumping; a genuinely stuck STA (no pumping)
still accrues staleness and faults correctly. No MXAccess parity change.
archreview: WRK-01 (P0). Verified on the Windows host (x86): worker builds
clean, StaRuntimeTests + WorkerPipeSessionTests 33/33 pass.
Typed write/command wrappers now call ensure_mxaccess_success after
ensure_command_success, failing on hresult < 0 (COM-correct, matching
CLI-08/Python) or any MxStatusProxy.success == 0, via a new boxed
Error::MxAccess variant with credential-safe formatting. The raw invoke
escape hatch stays unvalidated.
archreview: CLI-03 (P0). Verified: cargo fmt/check/test (28) and clippy --all-targets -D warnings clean.
Events()/EventsAfter() previously closed the channel silently when the
16-slot buffer overflowed, indistinguishable from a graceful server end.
Reserve one slot and emit a terminal EventResult wrapping the new exported
ErrSlowConsumer so overflow is always observable. The blocking
SubscribeEvents path (gRPC flow-controlled) is unchanged.
archreview: CLI-01 (P0). Verified: gofmt clean, go build ./..., go test ./... (+ -race) pass.
Adds the 2026-07-08 architecture review (00-overall + six domain reports)
and a remediation/ tree: one design+implementation doc per domain covering
every finding, plus 00-tracking.md as the master progress tracker.
- 153 findings with stable IDs (GWC/WRK/IPC/SEC/CLI/TST), each with
design rationale, implementation steps, tests, docs, and verification.
- Tracker rolls findings up by severity and P0/P1/P2 roadmap tier, records
cross-cutting clusters and per-finding status (all Not started).
- Planning docs only; no source changes.
WorkerLiveMxAccessSmokeTests built EventStreamService with the old
6-arg signature (mapper, dashboard broadcaster, logger) that predates
the SessionEventDistributor refactor; the current ctor takes only
(sessionManager, options, metrics). Pre-existing compile break on this
branch, unrelated to the doc sweep. NonWindows.slnx now builds clean.
Correct three stale facts against the current tree: the repo IS a git
repo at top level, the Java client targets JDK 17 (not 21), and note the
NonWindows.slnx build entry point for the macOS working tree. Flag the
gradle JDK 17 toolchain requirement in ToolchainLinks.
The hardcoded clientVersion() string lagged the Gradle version bump in the
JDK-17 retarget, so version/--json self-reported 0.1.2. Align it (and the two
CLI version-command test assertions) to 0.2.0. 97/97 tests pass.
Ignition 8.3 runs on JDK 17; the 0.1.2 client was compiled --release 21
(class-file v65) and would throw UnsupportedClassVersionError when loaded
in an Ignition module. Drop the toolchain and bytecode target to 17 so the
artifact loads on the gateway (a 17 build still runs on 21+). No source uses
Java 18-21 APIs, so this is a pure target change. Bump 0.1.2 -> 0.2.0.
Update the toolchain-pinning smoke test to assert feature()==17.
All 97 tests pass; all jar classes verified class-file v61. Roadmap G0.
Dashboard Galaxy summary now copies volatile fields fresh and memoizes only the
O(N) breakdown; added the DI registration-order wiring test and the memoization
regression/guard tests. All modules back to 0 pending.
The 2026-06-25 re-review observed that the gateway server no longer uses the
Contracts-generated Galaxy types (it consumes the wire-identical types from the
ZB.MOM.WW.GalaxyRepository package). That made galaxy_repository.proto look like
dead code. It is not: Go/Rust/Java/Python clients compile it by path, the .NET
client consumes Contracts.Proto.Galaxy.* via project reference, and
clients/proto/proto-inputs.json publishes it. Document this at the csproj entry
and in CLAUDE.md so it is not deleted in a future cleanup.
Resolves Server-059, Tests-041, Server-060 (2026-06-25 re-review).
DashboardSnapshotService memoized the whole Galaxy summary keyed on the cache
Sequence, but the shared library bumps Sequence only on a heavy refresh: the
steady-state tick, the refresh-failure path, and the age-based status getter all
replace the entry via 'previous with { ... }' at the SAME Sequence. The dashboard
therefore froze LastQueriedAt and, during a Galaxy SQL outage, kept showing
Healthy with no error for the whole outage.
Split DashboardGalaxySummaryProjector into ComputeBreakdown (the O(N) template/
category work, the only sequence-bound part) and BuildSummary (cheap volatile
fields copied fresh). ResolveGalaxySummary now memoizes only the breakdown by
Sequence and rebuilds the summary from the current entry each tick. Removed the
redundant DashboardGalaxyProjector wrapper.
Tests: same-sequence status/error/timestamp now reflected (the regression);
memoization-hit and sequence-invalidation guards; GatewayApplicationTests asserts
the DI container resolves IGalaxyBrowseScopeProvider to GatewayBrowseScopeProvider
(pins the registration-order invariant over the library's no-op default).
Mark A2 handoff and stillpending §2 adopted; note the host-side design
(GatewayBrowseScopeProvider, dashboard summary projector), the lib 0.2.0
upstream changes, and caveats (NSSM deploy config, pre-existing NU1903 +
IntegrationTests EventStreamService breaks). Point CLAUDE.md at the package.
Add GalaxyRepositoryHostWiringTests.BrowseChildren_BrowseSubtreesConstraintThroughHostWiring_FiltersChildren:
constructs the real GatewayRequestIdentityAccessor + GatewayBrowseScopeProvider, passes the
provider as IGalaxyBrowseScopeProvider to the lib GalaxyRepositoryGrpcService, and asserts
two children (unconstrained) then empty (BrowseSubtrees=["NonExistent"]) — proving the full
production authz-filtering chain is correctly wired.
Strengthen DashboardSnapshotServiceTests.GetSnapshot_ProjectsGalaxySummaryFromHierarchyCache:
add Assert.Equal(2, TopTemplates.Count) and Assert.Contains($Area, InstanceCount==1) so the
test guards the complete summary output, not just the $Pump entry.
Joseph Doherty
dohertj2