9 Commits

Author	SHA1	Message	Date
romanett	9b034f8291	Enable nullable reference types centrally; migrate all unset and annotations-only projects (#4700) ... # Description Nullable reference types are now enabled once, in `common.props`, and every project in the repository builds warning-clean with them. **Configuration** - `<Nullable>enable</Nullable>` lives in `common.props`. All per-project `<Nullable>` settings are removed: 178 project files (enable, annotations, and the `disable` in the three Aot test projects). - 419 per-file `#nullable enable`/`restore` directives and the 4 `#nullable disable` directives are removed. - Kept on purpose: - The 5 `<auto-generated>` schema files (`UANodeSet.cs`, `UA Type Dictionary.cs`, `OPCBinarySchema.cs`, `SecuredApplication.cs`, `PInvokeHelper.cs`) keep `#nullable enable`. The compiler ignores the project setting in generated code, so the directive is required there. - `tests/Opc.Ua.Interop.LegacyPeer` keeps `disable`. It builds against the 1.5.378 packages and does not import the repository props. - `docs/DeveloperGuide.md` and `.github/copilot-instructions.md` now describe the central setting. **Migration** (the 43 unset and 13 annotations-only projects, roughly 10k diagnostics) - **Production code: annotated at the declarations.** Covers `tools/Opc.Ua.SourceGeneration.Core`, `src/Opc.Ua.Lds.Server` and the LDS sample. - Members, parameters and returns are nullable where null actually flows. - Null-check helpers carry `[NotNullWhen]` (`IsNull`, `IsValidXmlName`, `IsMethodTypeDesign`). - `CoreUtils.Format`, `Utils.Format`, `ITemplateWriter.WriteLine` take `params object?[]`. `Template.AddReplacement` takes `string?` (rendering already skips null replacements). - `IRegisteredServerStore.RegisterAsync` returns `Task<RegistrationEntry?>`: it returns null when a registration is removed. - Dereferences backed by an invariant (for example, the validated model) use `!`, so runtime behaviour is unchanged. - **Behavioural changes in production code** (small, deliberate): - `NodeSetToModelDesign` throws `InvalidOperationException` when the file is not a NodeSet. Before, it threw a NullReferenceException on the next line. - `ModelDesignValidator` starts its lookup tables empty instead of null. A redundant `if (m_nodes == null) return;` guard is gone. - **Test code: minimal annotations.** - `?` on values that can be null, `!` where the test asserts non-null. - `= null!` for fixture members assigned in SetUp. - `TryGetStructure<T>(out T? x)` where the inferred `T?` broke the `IEncodeable` constraint. - A few redundant `== null` checks on structs (`NodeId`, `ArrayOf<T>`) folded into the adjacent `IsNull`/`Count` checks. **Notes for reviewers** - The diff is large (~1,500 files) but almost entirely mechanical: `?`, `!` and directive removal. The non-annotation changes are listed above. - Known follow-up, not done here: `IModelDesign.Namespaces/TargetNamespace/NamespaceUris` stay non-nullable even though the validator returns null before it loads a dictionary. Making them nullable would ripple through every generator call site. ## Related Issues - Fixes #4439 ## Checklist - [ ] I have signed the [CLA](https://opcfoundation.org/license/cla/ContributorLicenseAgreementv1.0.pdf) and read the [CONTRIBUTING](https://github.com/OPCFoundation/UA-.NETStandard/blob/master/CONTRIBUTING.md) doc. - [ ] I have added tests that prove my fix is effective or that my feature works and increased code coverage. - [x] I have added all necessary documentation. - [x] I have verified that my changes do not introduce (new) build or analyzer warnings. - [ ] I ran **all** tests locally using the **UA.slnx** solution against at least .net **framework** and .net **10**, and all passed. - [ ] I fixed **all** failing and flaky tests in the CI pipelines and **all** CodeQL warnings. - [ ] I have addressed **all** PR feedback received. The "no new warnings" box covers net10.0 and net48: full non-incremental builds of `UA.slnx`, `tools/SourceGeneration.slnx`, `tools/MigrationAnalyzer.slnx` and the Aot test projects finished with 0 warnings and 0 errors under the repository's warnings-as-errors settings. net9.0 and net8.0 also had no nullable diagnostics, but they were not rebuilt after the final fixes. Tests have not been run locally yet. 🤖 Generated with [Claude Code](https://claude.com/claude-code) --------- Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>	2026-10-08 08:48:13 +02:00
romanett	fbdd0af13b	Add 1.5.378 <-> 2.0 interop tests; return BadAttributeIdInvalid for unset optional security attributes (#4630) ... # Description Adds an interop test suite between this stack (2.0) and the released **1.5.378** NuGet stack, in both directions, and fixes a 2.0 bug the suite found. ## Interop suite - **`tests/Opc.Ua.Interop.LegacyPeer`** is a console peer built from the `OPCFoundation.NetStandard.Opc.Ua.*` 1.5.378 NuGet packages. - It runs as a child process, because the two stacks share assembly names. - It deliberately does not import the repository build settings (its own `Directory.Build.props/targets/Packages.props`). - Server mode hosts a small interop server with tight limits, or the 1.5 Quickstarts reference server. - Client mode runs named checks and prints one JSON result line per check. - **`tests/Opc.Ua.Interop.Tests`** builds the peer into `legacy-peer/` and runs in every CI test leg as a normal `*.Tests.csproj`. It covers: - sessions over None, the RSA policies and the ECC policies (Sign and SignAndEncrypt), with anonymous and user name logon, plus read/write/browse/call/subscribe; - the existing 2.0 `ClientTest` workers run against the 1.5 reference server, through the new `ClientTestFramework.ExternalServerUrl`; - custom data types (structures, unions, optional fields, enums) loaded, decoded and written back by each side's complex type system; - certificate trust with auto-accept off: untrusted peers, CA-issued certificates, revoked certificates; - multi-chunk messages, encoding limits, response size limits, operation limits, continuation points and security token renewal (RSA and ECC). - Every peer check is reported as its own test result. - `-p:LegacyStackVersion=<version>` tests another 1.5 release. - Docs: `docs/DeveloperGuide.md`. ## 2.0 fix **Unset optional security attributes now return `Bad_AttributeIdInvalid`** (`NodeState.ReadNonValueAttribute`). - Since 3e5f129c6, unset RolePermissions/UserRolePermissions/AccessRestrictions returned Good with a null value. - This made the 1.5 ComplexTypes loader crash against any 2.0 server. - 1.5 and Part 4 (Read) return `Bad_AttributeIdInvalid` here. The suite also found two ECC secure-channel bugs: the Sign-only HKDF salt length, and renewal chaining the previous secret for policies without SecureChannelEnhancements. #4626 fixed both on master in the meantime, so this branch now uses master's version. The ECC Sign and ECC token-renewal interop tests here cover them. ## Reviewer notes - **Behaviour change:** `NodeStateOptionalStorageTests` previously asserted Good with a null value for unset optional attributes. It now asserts `BadAttributeIdInvalid`. - Monitored items on such unset attributes now fail with `BadAttributeIdInvalid`, as in 1.5. - **Known gaps, not changed here:** - An unset Description still reads Good with a null value (same root cause). - A Null write to AccessRestrictions is accepted, as `NodeStateTests` asserts. - **Test runtime:** the suite adds roughly 7 minutes per CI test leg. On net48 the ECC cases skip, because 2.0 has no ECC on .NET Framework. ## Related Issues - None filed. ## Checklist - [ ] I have signed the [CLA](https://opcfoundation.org/license/cla/ContributorLicenseAgreementv1.0.pdf) and read the [CONTRIBUTING](https://github.com/OPCFoundation/UA-.NETStandard/blob/master/CONTRIBUTING.md) doc. - [x] I have added tests that prove my fix is effective or that my feature works and increased code coverage. - [x] I have added all necessary documentation. - [x] I have verified that my changes do not introduce (new) build or analyzer warnings. - [ ] I ran **all** tests locally using the **UA.slnx** solution against at least .net **framework** and .net **10**, and all passed. - Ran: interop suite on net10 (188 passed) and on net48 (passed before the final review fixes), plus Types, Core, InformationModel, Core.Security and Sessions on net10. - [ ] I fixed **all** failing and flaky tests in the CI pipelines and **all** CodeQL warnings. - [ ] I have addressed **all** PR feedback received. 🤖 Generated with [Claude Code](https://claude.com/claude-code) --------- Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>	2026-10-02 19:35:44 +02:00
Marc Schier	f9892020a3	Add configurable server resource isolation (#4541) ... # Description Add configurable server resource isolation on top of the merged #4522 foundation. - Coordinate connection, handshake, retained-message, session-establishment, queue, execution, request-cost and parked-request limits through one server-scoped provider. - Keep configured startup, continuity/reconnect, control and trusted-owner reservations inside existing totals. Shared capacity remains usable up to explicit owner ceilings. - Use authoritative Session membership for reassembly continuity capacity without granting unrelated resource or rate privileges. Preserve explicit owner mappings and bounded tracking. - Cache immutable classification evidence while retaining current Session and classifier checks. - Support weighted request ordering or FIFO independently of resource accounting. Installed providers enforce decoded-request limits in both modes. - Preserve fail-fast reassembly, valid Abort handling, fixed deadlines, active-work protection, reverse-connect handoff and bounded OpenAPI processing. - Support custom-handler parking through `IRequestParkingPolicy`, `WithRequestParking(...)`, and `RequestLifetime.ParkSink`. Parking releases execution capacity only, keeping retained-data and parked charges until completion. Built-in Publish and the global opt-out remain unchanged. - Provide direct, fluent and DI configuration, public incoming-request parking support, effective startup limit logs and user-facing documentation. Balanced remains the default. SharedOnly without an explicit provider retains shared behavior for intentionally incompatible/unlimited configurations. Reservations are finite and are not one full-message allowance per Session. ## Related Issues Builds on merged #4522. Fixes #4554. ## Review and CI corrections All three previously missed findings in [review 5326734718](https://github.com/OPCFoundation/UA-.NETStandard/pull/4541#pullrequestreview-5326734718) are addressed in `9e25e4ecc`: 1. A custom FIFO provider receives RequestQueue, RequestQueueBytes, RequestExecution and ParkedRequest accounting. A blocked head cannot be overtaken. 2. `IResourceIsolationReassemblyProvider` replaces the sessionless threshold independently of `UseFairScheduling`. The aggregate byte ceiling remains enforced. 3. Runtime `ReservedReassemblyBytes` includes all configured protected stage floors, including ControlReserved. The review's two already-resolved changes remain intact. The CI collection race was fixed at its source in `e1a994a4b`: `ArrayOf.ToArrayOf` treats a separately read collection count as an initial capacity hint and grows/trims for the actual enumeration. Deterministic tests cover growth, shrinkage and stale-zero counts. The subsequent recovery-test race was fixed in `05753ce77`: fake time advances only after a gated keepalive response has been processed and reset its timer, not merely after request arrival. Recovery-order and timeout assertions remain intact, with additional response-state assertions. Instrumentation reproduced the original failure twice; both fixed variants passed 400 executions per framework. CodeQL alert 388 was verified and dismissed as a false positive. Startup logs contain aggregate numeric capacities, not credentials or caller identities. ## Validation | Scope | Result | | --- | --- | | Latest automated checks at `05753ce7775a58801b4ba9d8f6d7781b709d0b89` | 167 successful, one package-publication check skipped, no failures | | Remaining external prerequisite | `license/cla` pending: Contributor License Agreement is not signed yet | | FIFO/weighted queues, custom parking and server-channel selection | 175 passed on each net10.0/net48 | | Isolation, configuration and parking-hosting selection | 259 passed net10.0; 257 passed net48 | | ArrayOf conversion tests | 73 passed on each net10.0/net48 | | Recovery-audit and keepalive fixtures | 40 passed on each net10.0/net48 | | NativeAOT smoke from preceding updates | Published and executed successfully | No tests, assertions, coverage thresholds, or required checks were weakened. The remaining CLA prerequisite must be completed through [CLA Assistant](https://cla-assistant.io/OPCFoundation/UA-.NETStandard?pullRequest=4541). This is not CTT certification or an unconditional network-DDoS, whole-process-memory or fleet-wide guarantee. ## Checklist - [ ] I have signed the [CLA](https://opcfoundation.org/license/cla/ContributorLicenseAgreementv1.0.pdf) and read the [CONTRIBUTING](https://github.com/OPCFoundation/UA-.NETStandard/blob/master/CONTRIBUTING.md) doc. - [ ] I have added tests that prove my fix is effective or that my feature works and increased code coverage. - [x] I have added all necessary documentation. - [x] I have verified that my changes do not introduce (new) build or analyzer warnings. - [ ] I ran **all** tests locally using the **UA.slnx** solution against at least .net **framework** and .net **10**, and all passed. - [x] I fixed **all** failing and flaky tests in the CI pipelines and **all** CodeQL warnings. - [x] I have addressed **all** PR feedback received. --------- Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com> Co-authored-by: marcschier <11168470+marcschier@users.noreply.github.com>	2026-09-27 21:11:52 +02:00
Marc Schier	46fa7d364a	Harden recovery, concurrency, and resource lifetimes across the OPC UA stack (#4516) ... # Description Bound managed-session recovery, preserve Publish workers and recovery ownership, and address the client, server, transport, and certificate audit findings and subsequent concurrency review. Also reduce the memory retained by exhaustive session-test diagnostics. ## Recovery and Publish ownership - Add `ManagedSessionOptions.ChannelReconnectTimeout` through direct construction, the fluent builder, and DI. The automatic duration is `max(3 * KeepAliveInterval, revised SessionTimeout, OperationTimeout)`, sampled at cycle start. Positive overrides are supported; `Timeout.InfiniteTimeSpan` explicitly opts out. Raw sessions retain their unlimited default. - Add `CreateReconnectBudget(TimeProvider)` as a required `IReconnectParticipant` member, as agreed in review. Custom participants implement it and may return null for no participant-specific bound. - Enforce the earliest participant/caller deadline through transport reconnect, activation/recreation, and subscription restoration. Late callers can shorten an active cycle but cannot extend it. - Cancel and unwind recovery, fence late responses, and release channel recovery ownership before the existing `Faulted` event hands control to the outer reconnect/failover policy. Preserve session recreate admission through final Publish cleanup and keepalive suppression until shared recovery is complete. - Treat participant budget creation/evaluation failures, including exceptions at the exact deadline, as terminal shared-recovery failures. Notify participants before exposing `Faulted`, preserve the original diagnostic, and never silently replace a finite budget with unlimited recovery. - Keep caller-supplied replacement channels, managed bindings, wrapper event subscriptions, and recovery-owner identities consistent across success, failure, and cancellation. Reject replacement during active recovery and fence retired callbacks so they cannot clear newer ownership or suppress later keepalive recovery. - Retain eligible Publish workers through temporary subscription-identifier clearing without retaining workers for never-created or removed subscriptions. Preserve cancellation, acknowledgement rollback, and drain-before-replacement ordering. - Coordinate automatic subscription updates with session recovery: cancel admitted updates, keep them paused across a failed-restoration handoff, and retry only after explicit restoration completes. This prevents competing CreateSubscription requests and updates against a retiring session, while preserving manual recreation, transfer fallback, and pending option changes. - Exercise the real ManagedSession, V2 subscription engine, and channel ready gate with scripted transports and an injected clock. Cover transfer fallback, deadline handoff, late completions, and resumed notification delivery. ## Audit and review corrections - Correct certificate/trust-list transactions, change-pump lifetimes, validation, peer revalidation, and reset/resource ownership. - Harden node admission, attribute/range/dimension validation, monitored-item sampling and lifecycle transitions, event fanout, and server shutdown. Run lifecycle readiness and compensation outside shared monitored-item admission, with stale-completion ownership checks. - Close synchronous node-manager admission before teardown and retain resources until admitted service, lifecycle, and asynchronous method work returns. Dispose the monitored-item manager outside both locks, reject late calls before they can reach it, and forward asynchronous disposal through the adapter so the master awaits the same completion. Preserve reentrant shutdown initiation, cancellation/failure unwinding, and shared cleanup-failure reporting. - Bound WSS pending-request retention using connection-local wire handles while preserving caller-visible request handles and translating Cancel targets. Safely ignore retired replies and drain admitted send/close users before resource disposal. - Clarify that the shared `ClientBase` caller-handle sequence is distinct from the non-wrapping WSS wire allocator. Existing exhaustion/reopen, late-reply, and Cancel-translation regressions verify the connection-local contract; no allocator behavior was changed by this clarification. - Preserve virtual-node validation and resolver service errors instead of converting them to cached misses. A mismatched NodeId reports `BadNodeIdInvalid`; callback type mismatches and resolver service errors keep their exact status. Per-item boundaries preserve unaffected reads, writes, history operations, method calls, browse paths, and monitored-item creation/restoration, including permission-validation paths. Caller cancellation still propagates; stored-item validation failures are logged without aborting later restores. - Correct client file/role/state-machine/streaming APIs, node-cache behavior, redundancy/failover ownership, and session configuration/identity handling. - Cancel each owned peer-discovery scope before disposing it, including when the bounded wait wins a cancellation-delivery race. Keep late results out of the endpoint cache and preserve single-flight lookup ownership. - Keep materialized alias Browse results synchronized with the live store by default, preserving explicit startup-only opt-out. Apply HistoryUpdate service results per value while retaining explicit all-or-nothing provider APIs. - Update the directly related feature documentation and add focused positive, negative, concurrency, and ownership regressions. ## Latest review-feedback corrections - Fence stale failed lease observations with a successful-confirmation generation distinct from the revocation epoch. Snapshot it before each store read, including failed-CAS revalidation, so an old failure cannot revoke or reschedule a newer lease while fresh ownership loss still revokes leadership. Preserve independent overlapping acquisitions, expiry, disposal, and late-success protections. - Isolate attributable WSS response-decoding errors to their request, including retired handles. Keep admitted socket sends connection-owned so caller timeout/cancellation cannot abort sibling requests; shutdown still cancels and drains those sends. - Restart keepalive and classic Publish replenishment after successful session reuse. Retire V2 worker demand after intentional deletion/disable without confusing that state with temporary recovery resets. - Browse virtual directories lazily across continuation pages without the 1,024-entry snapshot cap. Preserve active-page cancellation, bounded metadata retention, pushback, and exactly-once asynchronous cursor cleanup. - Apply certificate/CRL staging in last-operation order, preflight required issuer storage before writes, select actual policy-compatible RSA keys for None endpoints, frame certificate chains independently of signature OIDs, and preserve admitted validation results during rejected-writer shutdown. - Preserve base asynchronous disposal when a legacy override omits its base call; avoid double derived disposal in the adapter. Let genuinely admitted nested work finish while rejecting stale captured contexts and returning null from closed lookup APIs. - Re-park recovered monitored items if AddNodes rolls back. Give event compensation a fresh finite deadline, and preserve healthy Browse/Translate siblings at every reference/path boundary while retaining starting-node errors and caller cancellation. - Give final reconnect notifications a live, bounded shutdown-linked lifetime. Synchronize child state before parent lifecycle handlers outside the parent transition lock. - Document the accepted existing policies: direct reverse connections are single-use without a ReverseConnectManager, and multi-node stream capacity is shared rather than reserved per node. The expanded readiness controls also exposed a real generation-handoff race: an old source could finish draining after activation was deferred but before readiness tasks were collected, causing the wanted replacement to be omitted from the wait set. A deterministic test gates that exact interval; the fix keeps the replacement in the readiness wait. The original race passed 200 repetitions after the fix. A separate control now awaits both producer readiness and iterator entry under the same original deadline, rather than treating those independent signals as equivalent. Temporary logging and repetition are removed. ## CI integration contracts The History and Server CI failures exposed integration tests that still asserted the behavior intentionally changed by the audit fixes. A Windows WoT negative test raced its caller deadline against upstream retries. A subsequent Ubuntu Client failure exposed a real automatic-update versus explicit-restoration race: - History insertion collisions now verify exact per-value statuses and all persisted timestamps/values, including preservation of the existing conflicting entry. Explicit atomic-provider rollback controls remain. - Alias materialization tests wait for actual publication and verify the added node plus forward/inverse `AliasFor` references, instead of racing a startup-only snapshot expectation. Explicit refresh opt-out controls remain. - The unavailable-upstream WoT test uses an explicit finite retry policy through existing DI configuration, preserving its original four-minute guard and requiring a mapped-value service error before caller cancellation. Production retry defaults remain unchanged. - The Client fix retains the exact one-create and late-response assertions. Three deterministic real-subscription/real-manager cases cover paused creation, disposal while paused, and cancellation of an admitted create followed by explicit restoration and a pending option update. Both composed restoration-deadline cases also prove automatic updates stay blocked at the outer handoff and resume after restoration. - The merged-head net48 correction prevents lost peer-resolver cancellation and distinguishes the harness's delayed initial `Connected` event from actual outer recovery completion. A deterministic timer-order test covers both cooperative and cancellation-ignoring resolvers, cancellation before return, rejection of late cached results, and a fresh subsequent lookup. - The remaining CodeQL alert was a false content flow through `ArrayOf.Count` into an integer-only trust-list summary. Projecting the same count through `Span.Length` keeps cardinality distinct from serialized contents without allocating, dropping diagnostics, suppressing the rule, or dismissing the alert. Five log-contract cases cover null/empty/nonempty results and mixed chain-capable/fallback listeners, asserting the exact count, message, and structured properties. - Classic transfer acknowledges messages selected for asynchronous Republish only after their responses arrive. The existing locked replay-selection decision returns the immediate-acknowledgement subset to Session, preserving the public transfer API, advertised availability, and immediate acknowledgement of older gapped messages not selected for replay. A callback changing `RepublishAfterTransfer` cannot change the already-selected acknowledgement ownership. - Transfer integration tests explicitly control acknowledged versus retained origin messages. Exact assertions distinguish retained replay from requested fresh initial values by sequence number, client handles, values, status, and source timestamp. Atomic counters, observed delivery/keepalive progress, and failure-safe session/file cleanup replace timing assumptions and racy resets. - Server lifecycle tests verify the documented independent five-second compensation deadline rather than requiring a non-cancellable token. They cancel the actual request, prove compensation and unrelated monitored-item creation still proceed one tick before the deadline, preserve the original failure and ownership assertions, and verify the completed deadline is retired. - Synchronous and asynchronous directory-browser tests require the exact provider exception instead of an empty-directory success, then verify the failed cursor is not enumerated again. These correct expectations left behind by the intentional, documented provider-error propagation change; production behavior is unchanged. No tests were disabled, assertions loosened, coverage thresholds lowered, or CI checks made optional. Existing platform exclusions and Explicit debug variants remain. ## Related Issues - Fixes #4514 - Fixes #4515 - Builds on merged #4503 and retains the #4508 publishing-drain behavior. ## Validation **Lease-election review correction (`a13afa641182c70983e16bbcfebf507e666a45a5`):** The correction is based on the existing upstream merge `43079e2b851fe338a4082066f8d14abfdafb6e32`. Three deterministic stale-verdict cases fail on the original source and pass with separate confirmation fencing. Six new cases cover the initial foreign-owner verdict, failed-CAS revalidation in both invocation orders, and fresh ownership loss through both read paths, including revalidation begun after another successful confirmation. | Target and configuration | Passed | Failed | Skipped | | --- | ---: | ---: | ---: | | net10.0 Debug | 120 | 0 | 0 | | net48 Debug | 120 | 0 | 0 | The selection is the full `Opc.Ua.Core.Tests.Redundancy` namespace, including all 38 `SharedStoreLeaseElectionTests` cases and the existing expiry, clock rollback, disposal, and overlapping-success regressions. New witnesses are `StaleForeignLeaseReadDoesNotRevokeNewerAcquisitionAsync`, `StaleFailedRevalidationDoesNotRevokeNewerAcquisitionAsync`, and `FreshForeignLeaseObservationRevokesLeadershipAsync`. They assert resulting leadership, exact transitions, independent progress, and the unchanged expiry boundary. Retaining the initial generation for the fallback read independently makes the fresh-loss-after-confirmation control fail; that mutation was removed before the final rebuilt runs. The successful scoped command is `dotnet test tests\Opc.Ua.Core.Tests\Opc.Ua.Core.Tests.csproj -c Debug --no-restore --filter "FullyQualifiedName~Opc.Ua.Core.Tests.Redundancy"`, with `CustomTestTarget` and `-f` matching each row and that framework's assets restored. Both changed C# files pass formatting and analyzers through informational severity on both frameworks. These are scoped results, not a full-solution test or measured coverage-delta claim. CI monitoring is paused for this feedback-only operation; no current-head CI success is claimed. **Latest CI contract correction (`1cd1f2047bb5b50c9bf300b1ff710b0b2e935a29`):** All four failing Server shard 23 jobs in [run 36120946740](https://github.com/OPCFoundation/UA-.NETStandard/actions/runs/36120946740) reported the same six cases. Two local replays reproduced all six failures. The source changes and documentation from `cd43b831a` establish both intentional contracts: independent bounded event compensation and surfaced provider errors. Only the two test files are corrected; production code and CI policy are unchanged. | Target and configuration | Passed | Failed | Skipped | | --- | ---: | ---: | ---: | | net48 Debug | 113 | 0 | 0 | | net10.0 Release, CI Coverlet settings | 113 | 0 | 0 | The selection contains `MonitoredItemManagerLifecycleTests`, `DirectoryObjectStateTests`, `DirectoryBrowserBoundedEnumerationTests`, and `EventCompensationReadinessHasFiniteDeadlineWithInfiniteSourceTimeoutAsync`. Both changed files pass formatting and analyzers through informational severity on both frameworks. Independent mutations restoring a non-cancellable compensation token and swallowing provider errors fail all four lifecycle cases and both browser cases, respectively. Both mutations were removed before the final passing runs and commit. The scoped command is `dotnet test tests\Opc.Ua.Server.Tests\Opc.Ua.Server.Tests.csproj` with `CustomTestTarget` and `-f` set to the selected framework, `--no-restore`, and `--filter "FullyQualifiedName~MonitoredItemManagerLifecycleTests|FullyQualifiedName~DirectoryObjectStateTests|FullyQualifiedName~DirectoryBrowserBoundedEnumerationTests|FullyQualifiedName~EventCompensationReadinessHasFiniteDeadlineWithInfiniteSourceTimeoutAsync"`. The modern run additionally uses `-c Release /p:CollectCoverage=true --collect "XPlat Code Coverage" --settings tests\coverlet.runsettings.xml`. These are scoped results, not a full-solution test or measured coverage-delta claim. Hosted checks for that commit were pending at publication. The fail-fast CI watch was stopped when the requested operation switched to review feedback; no current-head CI success is claimed. **Earlier master merge (`8b15ea6ef9fc2214ac8ec886f373bb3846dd6d6f`, base `b4136d8e9bab0df680f41d9ef985f519160534b1`):** The `TcpListenerChannel` conflict preserves both lifetime responsibilities: disposal clears the retained peer-certificate chain and unregisters the quota-scoped message-assembly scheduler. Both fields and all upstream deadline handling remain intact. The only additional manual source adjustment adds the parentheses required by the existing analyzer around the session-count fallback. Master advanced during validation. The initial transport merge `80303c434ce813cf75aa7fe54a6a32fe2aeafc30` was retained, and the newer base was merged normally in `8b15ea6ef9fc2214ac8ec886f373bb3846dd6d6f`, without rewriting history. The additional recovery, fluent-pipeline, and test-framework overlaps were inspected; the branch's cancellation, validation-error, and acknowledgement ownership fixes remain alongside the upstream changes. Full `UA.slnx` builds passed on net10.0 and net48 with **0 warnings and 0 errors**. The final scoped tests rebuilt against the committed tree `2074f8efcb61d8d9c6106e7d6f76ee941940e358`: | Target | Passed | Failed | Skipped | | --- | ---: | ---: | ---: | | net10.0 | 2,653 | 0 | 0 | | net48 | 2,529 | 0 | 7 | The net48 skips are existing platform exclusions: six RSA_DH_AesGcm nonce cases and one ClientWebSocket certificate-validation case requiring .NET 7 or later. The selection covers transport cleanup, reassembly budgets/deadlines/admission, peer-chain ownership, server session binding, fluent lifecycles and source generation, reconnect and keepalive, transfer acknowledgements, Publish queues, and JSON encoding. Modern-.NET WSS transport and real-server integration cases also pass. The resolved file passed solution-relative formatting and analyzer verification through informational severity on both frameworks. These are scoped test results, not unfiltered full-suite or measured-coverage claims. Hosted CI was not monitored or repaired as part of this merge-only operation. Selected local results below are overlapping validation runs, not distinct-test totals. **Earlier review-feedback batch (`cd43b831a8f11de3f7d65c3681712a61e3c669a5`):** The final net10.0 scoped selection passed **471 tests, with no failures or skips**. The final net48 solution-root selection passed **439 tests, with no failures and one unchanged platform skip** (`ValidateServerCertificateFallsBackToTlsPolicyErrors`, which requires newer ClientWebSocket certificate-validation support). Both full nonincremental solution builds completed with **0 warnings and 0 errors**, and all **40 changed C# files** passed informational-level formatting/analyzer verification. Both frameworks validated the identical final tracked diff. The focused matrix covers every one of the 20 current review threads, including real mid-frame WSS cancellation, stale admission contexts, directory continuation ownership, multi-element Browse/Translate failures, and the readiness drain race. The net10.0 solution-root test command has a separate pre-existing runner conflict: three unrelated AOT projects reject the VSTest target under Microsoft.Testing.Platform 2.4.0 and SDK 10. Per-project net10.0 results are recorded separately; no runner migration, test exclusion, assertion relaxation, or CI-policy change is included. These are bounded selections, not unfiltered full-suite or measured coverage claims. At publication of `cd43b831a`, hosted CI was not green: the `OPCFoundation.UA-.NETStandard` check reported `ACTION_REQUIRED`, AppVeyor reported `FAILURE`, and the two signed-build jobs were in progress. No CI jobs were rerun, cancelled, approved, or repaired as part of that feedback-only operation. **Classic transfer correction (`f425376fca045794122e19aaf9cafebc4a717d91`):** The preceding head's [macOS net10.0 shard 25/30](https://github.com/OPCFoundation/UA-.NETStandard/actions/runs/35980326299/job/107570512458) failed `TransferSubscriptionTest(DisconnectedRepublish).TransferSubscriptionAsync(True,False)` at the static-node count assertion: expected 10, received 20. The trace contains a successful, explicitly requested republish of retained sequence 1 and a fresh initial-value batch at sequence 3, not duplicate dispatch of one response. Controlled lost origin acknowledgements reproduced that incorrect test expectation on both the preceding PR head and exact master base `a2539cb53dd1666cef856f33812d7a8830d68909`. The maintainer explicitly authorized correcting this existing issue in the PR. Retained replay and requested initial values are independent, as specified by [OPC UA Part 4, TransferSubscriptions](https://reference.opcfoundation.org/specs/OPC-10000-4/5.14.7). The reason that the original hosted run retained its particular origin message remains unproved. The expanded full-suite run then exposed a **real production race**, not just a count assertion: Session queued acknowledgements for all transferred available messages before the asynchronous Republish responses arrived. Publish could delete those messages first, causing `BadMessageNotAvailable`. A gated real-Session/classic-engine regression failed both replay-enabled cases on the original implementation while both no-replay controls passed. The ownership fix passes all 24 combinations of empty/consecutive/unsorted-gapped queues, replay preference, sequential delivery, and a reentrant preference change. All 40 normal real-server transfer cases pass on both frameworks. Independent mutations suppressing actual replay and duplicating observed sequence records each fail the intended assertion; both mutations were removed. The full classic suite passes on both frameworks. Hosted CI was pending when that correction was published; the earlier failed hosted job was not rerun for a lucky pass. **Local net48 caveat:** the full Client run timed out in the pre-existing fake-clock test `ReconnectAsyncSwapsFaultedLeaseEntryAsync`, which does not exercise subscription transfer. Its unchanged isolated replay passed, but that does not resolve the full-suite timeout. No unrelated code or assertions were changed, and a clean full Client net48 run is not claimed. | Scope | net10.0 | net48 | | --- | --- | --- | | Full Client.Tests after transfer acknowledgement correction | 2,549 passed | 2,552 passed, 1 timeout failure noted above, 1 existing platform skip | | Full classic-subscription project after transfer correction | 57 passed, including all 40 normal transfer cases | 57 passed, including all 40 normal transfer cases | | Gated transferred-message acknowledgement ownership | 24 passed, included above | 24 passed, included above | | Saved-session consumer of the shared subscription helper | 40 passed | 16 passed, 24 existing unsupported-security-policy skips | | Audit regression matrix | 2,339 passed | 1,934 passed, 34 existing skips | | Full Client.Tests after review follow-ups | 2,520 passed | Focused selections only | | Full Client.Tests after CI recovery correction | 2,523 passed, CI coverage collector | 2,527 passed, 1 existing platform skip | | Latest merged-head recovery/redundancy selection | 139 passed | 139 passed | | Full Client.Tests after merged-head net48 correction | Focused selection above | 2,529 passed, 1 existing platform skip | | Lifecycle and actual-dimension regressions | 146 passed | 146 passed | | Focused recovery follow-ups | 7 passed | 7 passed | | WSS-related fixtures | 93 passed | Not applicable | | CI History integration corrections | 28 passed | 28 passed | | CI alias, dispatcher, and explicit-atomic controls | 63 passed | 63 passed | | Full WoT sample project with CI coverage collector | 29 passed | Not applicable | | Trust-list effects and count-only summary contract | 22 passed | 22 passed | | Virtual-node feedback, node-manager, dispatcher, and lifecycle selections | 661 passed, 63 existing not-executed cases; all 48 virtual-node cases passed | 661 passed, 63 existing not-executed cases; all 48 virtual-node cases passed | | WSS wire-handle exhaustion, late replies, and Cancel translation | 9 passed | Not applicable | | Synchronous node-manager disposal and related service/lifecycle suites | 1,037 passed, 64 not-executed cases from existing fixture assumptions; all 93 sampling/disposal cases passed | 1,037 passed, 64 not-executed cases from existing fixture assumptions; all 93 sampling/disposal cases passed | - Full solution builds passed on net10.0 and net48 during the audit work. The subsequent full non-incremental net10.0 build also completed with zero warnings/errors. - Modified source/test formatting and analyzer checks passed through informational severity. - Original composed recovery regressions passed repeated runs on both frameworks. Independent mutations removing Publish cancellation, bypassing the recreate drain call, or ignoring ready-gate cancellation each failed the intended assertion; all mutations were restored before passing runs. - The latest recovery correction passed 100 executions of each original restoration-deadline scenario with unchanged assertions. Restoring premature update admission independently made both strengthened handoff cases fail; the correct retention behavior was restored afterward. - Startup-event and cancellation-delivery timing probes reproduced the subsequent net48 failures, then passed 100 executions of each of the three scenarios with the fixes. All timing probes and repetition were removed. The permanent timeout-order regression also went RED on the original cancellation cleanup without sleeps or scheduling amplification. - Final post-format recovery/subscription/engine checks passed all 151 selected cases on net10.0. The first local full net48 attempt aborted in `clr.dll` with native access violation `0xc0000005`; an unchanged-assembly full diagnostic rerun passed as recorded above. That native host crash has not been attributed or claimed fixed. - A minimal replay with the same CodeQL CLI and query-pack versions reproduced the custom-count alert and confirmed that the standard span-length projection is clean while direct sensitive-value and raw-identifier logging controls remain flagged. The source was already logging only a count; no identifier or certificate disclosure was established at the reported location. - The virtual-node feedback regressions first reproduced five error-preservation/cache failures; expanded service-dispatch cases exposed two additional whole-batch failures before the operation-boundary fixes. The final selections pass on both frameworks. Genuine misses remain cached, successful results remain reusable, failed validation allows same-cache retry, and request cancellation still stops the batch. - Three deterministic synchronous-manager disposal regressions reproduced early resource release and lifecycle calls reaching a manager during disposal. The fix passes direct and adapted calls, a sampling-worker join after the last operation exits, reentrant shutdown, cancellation and failure paths, repeated disposal, and exact rejection across 39 service/lifecycle entry points both during drain and after cleanup. The original outside-node-lock disposal regression remains and now requires the worker to observe closed admission. - Prior-head CI on `c985f8bded77a60449d9cd381c8a004d8590ccf6` completed with **168 passing checks, one expected release-package-publication skip, and no failures or pending checks**. All test shards, signed builds, container builds, the coverage check, and the CodeQL PR check passed. Hosted CodeQL analysis `1829262912` completed successfully and marks count-flow alert 385 fixed. This is evidence for that head, not a claim that subsequent review follow-ups have completed hosted CI. ### Session-test diagnostic memory Complete browse listings are streamed to unique NUnit attachments rather than retained in inline NUnit/VSTest output. Detached regression-test results are cleaned up without deleting runner-collected attachments. - Two covered session-suite runs each completed with 782 passed and 379 existing skips. Peak working set fell from 4.86 GiB to 2.44-2.78 GiB. - TRX output shrank from 303.1 MiB to approximately 40.7 MiB, retaining all 46 complete browse attachments. - Seven output regressions cover exact content, empty/single/large listings, disabled/enabled output, bounded inline summaries, and unique filenames. The separate Azure build-18452 VM deprovisioning cause remains unproved without provider logs. The memory measurements do not establish its cause. Unfiltered local `UA.slnx` tests, including the long stress/soak suites, have not been run on both frameworks. A local coverage delta was not collected; builds and focused tests are not substitutes for those claims. ## Checklist - [ ] I have signed the [CLA](https://opcfoundation.org/license/cla/ContributorLicenseAgreementv1.0.pdf) and read the [CONTRIBUTING](https://github.com/OPCFoundation/UA-.NETStandard/blob/master/CONTRIBUTING.md) doc. - [ ] I have added tests that prove my fix is effective or that my feature works and increased code coverage. - [x] I have added all necessary documentation. - [x] I have verified that my changes do not introduce (new) build or analyzer warnings. - [ ] I ran **all** tests locally using the **UA.slnx** solution against at least .net **framework** and .net **10**, and all passed. - [ ] All checks on the latest review-follow-up head have completed successfully; prior-head CI passed as recorded above. - [x] I have addressed **all** PR feedback received. Tests have been added and pass as recorded above; the hosted coverage check passed on the prior head identified above. The combined tests/coverage-increase checkbox remains open because no local coverage delta was measured. --------- Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> Copilot-Session: 68f26810-22b8-476f-abcd-0a8c82d6e208	2026-09-25 14:59:34 +00:00
romanett	e0c56a7c00	Replay the fuzz crash corpus from the fuzz-corpus orphan branch (#4539) ... # Description Since the move to GitHub Actions, CI replays only the checked-in fuzz corpus. The full crash corpus used to be the Azure DevOps secure file `FuzzingArtifacts.zip`, which `.azurepipelines/test.yml` overlaid onto `Opc.Ua.Encoders.Fuzz.Tests`. The nightly `private-corpus` job that was meant to replace it depended on an Azure storage account, federated credentials and a protected environment. None of these were ever provisioned, so the job never ran. **The corpus now lives on the orphan branch [`fuzz-corpus`](https://github.com/OPCFoundation/UA-.NETStandard/tree/fuzz-corpus)** (commit `843ba944a7b144a7ae6429151846d9e5ab28d29f`, already pushed): - It contains 22,267 `crash-*`, 25 `slow-unit-*` and 1 `timeout-*` input under `Opc.Ua.Encoders.Fuzz.Tests/Assets/`. - The files are byte-identical to the secure file (SHA-256 `85198e46…fe1d61b`), recovered from the `drop` artifact of Azure build 18397. - The branch shares no history with `master`, so clones of `master` don't get 22k extra files in their checkout. **Why the corpus is no longer treated as private:** - The inputs date from April–June 2024. - The `drop` artifacts could be downloaded anonymously. - Every master build printed up to about 7k base64 reproducers into anonymously readable Azure logs. - Current `master` replays every input without a test failure. New, unfixed crashes still go through [SECURITY.md](https://github.com/OPCFoundation/UA-.NETStandard/blob/master/SECURITY.md) and are added to the branch only after the fix ships. ## Changes - **`nightly.yml`:** `private-corpus` is replaced by `crash-corpus`. The new job: - checks out the commit pinned in `FUZZ_CORPUS_COMMIT` (no secrets, environment or OIDC); - verifies the SHA and copies each top-level folder over `fuzzing/<folder>`; - runs `Opc.Ua.Encoders.Fuzz.Tests` on the 12 non-macOS profiles through the shared `run-dotnet-tests` action and uploads `crash-corpus-results-*`. The `run_private_corpus` input and the INCOMPLETE label are removed. The job runs on every scheduled and manual run, and the nightly summary fails if it was skipped. - **`fuzzing/CrashCorpus.md` (new):** the Azure history, branch layout, how CI replays the corpus, local replay, and how to add inputs and move the pin. `Fuzzing.md` links to it. - **`docs/DeveloperGuide.md`:** the private-corpus section and its secrets table are replaced, and the migration to-do list is updated. - **`nightly.yml` AOT job (fix for a failure found by the first nightly run):** the Linux and macOS legs looked up the published binary by `BaseName`/`Extension`. A NativeAOT executable has no extension on those platforms, so PowerShell splits `Opc.Ua.Aot.Tests` into `Opc.Ua.Aot` + `.Tests`, and every leg reported *No NativeAOT executable was produced* even though publish succeeded ([run 36017168717](https://github.com/OPCFoundation/UA-.NETStandard/actions/runs/36017168717), `aot-ubuntu-latest`). The lookup now matches the exact file name. A failure now reports the publish errors, a missing publish directory, or which files were produced instead. - **`Opc.Ua.Redundancy.Samples.Tests` / `Opc.Ua.WotCon.Samples.Tests` (fix for failures found by the first nightly run):** the test runner's framework guard rejected both projects on the nightly net8.0, net9.0, net472 and netstandard2.1 profiles. Redundancy built a net48 placeholder for every target except net10.0, and WotCon was pinned to net10.0. Both now build the requested target on net8.0/9.0/10.0 and use `RestrictForLegacyTfm` to become explicit not-applicable shells on net48, net472, netstandard2.0 and netstandard2.1. The redundancy tests no longer hard-code `bin/<cfg>/net10.0` when locating the sample binaries, and `SampleAppProcess` falls back to an `object` lock below net9.0. The net48 placeholder remains for Azure's plain `dotnet test -f net48` leg. Checked locally by building Redundancy for `CustomTestTarget=net9.0` and WotCon for `netstandard2.1`. - **`DefaultServerRedundancyHandler` (fix for a failure found by the first nightly run):** `ResolveServerAsync` gave the resolver a token linked to the caller's, but waited on the caller's token directly. On cancellation the wait could finish, and its `using` could dispose the linked source, before the linked source's own callback had run. That unhooks it, so the in-flight peer lookup never saw the cancellation (`PeerDiscoveryPropagatesCallerCancellationAsync`, netstandard2.0 build on net48). The wait now uses the linked token. A timeout is still logged as a discovery failure, as before. A companion test fix makes `PeerTimeoutPreservesOtherEndpointsAndCoalescesPendingDiscoveryAsync(False)` deterministic. The timed-out lookup drains on the thread pool, so the test's retry could join it and then have its own time limit expire before it started a fresh lookup. The test now waits for the second resolver call before advancing the fake clock. - **`Session` keep-alive logging (found while investigating the net48 nightly failures):** `OnKeepAliveError` and the keep-alive read failure log used `Endpoint?.EndpointUrl`. `ClientBase.Endpoint` dereferences the transport channel, so once the channel was detached it threw a `NullReferenceException` on every keep-alive tick, and `BadNoCommunication` never reached the `KeepAlive` subscribers. Both now log the configured endpoint's URL. Three new `SessionKeepAliveTests` cover a detached channel. - **`ClientTestFramework` shared session (fix for 67 `Opc.Ua.Sessions.Tests` failures on the net48 nightly):** the fixture-wide shared session used the 10 s default session timeout. A test-host stall of about 11–18 s while other fixtures restarted their servers let the server expire it, and every later test in the `ClientTest("https")` and `ClientTest("opc.https")` fixtures failed with `BadSessionIdInvalid`. The shared session now requests `SharedSessionTimeout` = 120 s; per-test sessions keep 10 s. - **`run-dotnet-tests.ps1`:** `-QuietOutput` help text only; it no longer names the removed job. The Azure secure file stays in place, because `master378` and the 1.x release pipelines keep their own `test.yml` and still use it. ## Verification - `nightly.yml` parses as valid YAML, and no job, output or summary field still refers to the removed job. - I ran the job's overlay script against a clone of `fuzz-corpus` at the pinned commit. It overlaid all 22,293 files, left the tracked `Assets/Repo` untouched, removed the checkout folder and wrote the summary line. - **Not yet run:** the workflow itself. It is `schedule`/`workflow_dispatch` only, so it needs a manual `nightly.yml` run on this branch. On master the corpus has so far only been replayed on Windows (net48/net10.0 in Azure), so the Linux legs are new ground. ## Follow-ups (admin) - Add a ruleset for `fuzz-corpus`: require pull requests and block force pushes and deletion, so the pinned commit stays reachable. - The `fuzz-private-corpus` environment, if it was ever created, can be deleted. ## Related Issues - No issue exists yet; this is part of the Azure DevOps → GitHub Actions migration (#4513, #4524). ## Checklist - [ ] I have signed the [CLA](https://opcfoundation.org/license/cla/ContributorLicenseAgreementv1.0.pdf) and read the [CONTRIBUTING](https://github.com/OPCFoundation/UA-.NETStandard/blob/master/CONTRIBUTING.md) doc. - [ ] I have added tests that prove my fix is effective or that my feature works and increased code coverage. - [x] I have added all necessary documentation. - [ ] I have verified that my changes do not introduce (new) build or analyzer warnings. - [ ] I ran **all** tests locally using the **UA.slnx** solution against at least .net **framework** and .net **10**, and all passed. - [ ] I fixed **all** failing and flaky tests in the CI pipelines and **all** CodeQL warnings. - [ ] I have addressed **all** PR feedback received. 🤖 Generated with [Claude Code](https://claude.com/claude-code) --------- Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>	2026-09-25 07:49:00 +02:00
Marc Schier	06ae192084	Fix CI failures tracked by #4299 (#4308) ... ## Summary - stabilize the History, authentication lockout, Robotics, redundancy, and DI tests reported in #4301-#4305 - fix the closed-transport buffered-write completion deadlock that hung `ConnectMultipleSessionsAsync` - add macOS crash diagnostics for #4306 and retain evidence for root-cause analysis - run macOS tests and NativeAOT in GitHub Actions pull requests, using supported Intel and ARM64 runners, while keeping every Azure macOS definition disabled by default - move all remaining source-level `InternalsVisibleTo` declarations into project files so signed OpenUSD builds receive the repository public key ## Local validation - MCP PubSub: 29 passed - History focused tests: 3 passed - Sessions lockout: 3 passed - Sessions multiple-session regression: 3 passed - deterministic binary client channel tests: 22 passed - Robotics runtime: 34 passed - Redundancy focused tests: 4 passed - OpenUSD net10.0: 914 passed - affected OpenUSD projects build on net10.0 and net48 Refs #4299, #4300, #4306 Fixes #4301 Fixes #4302 Fixes #4303 Fixes #4304 Fixes #4305 --------- Copilot-Session: 7cb486de-8b4c-43c5-a900-e7d2e0ad04a8	2026-08-22 19:30:19 +00:00
Marc Schier	aad8350bc1	Pluggable symmetric cryptography, an awaitable asymmetric path, registrable security policies, and PubSub crypto offboarding (#4206, #4207, #4208, #4210) (#4232) ... # Description Closes the three crypto-offboarding follow-ups split out of #4192. Each was recorded there as knowingly *not* done, with the reason and what would justify revisiting it. Those conditions are met, and where one of them turned out to be unachievable that is written down rather than worked around. The safety property throughout: **every seam is opt-in and null-checked, and in the default software configuration the added code is observably inert.** No interface dispatch on the per-message path, and the asynchronous paths complete synchronously — so channel locking order, buffer ownership and error sequencing are unchanged for a deployment that configures nothing. ## #4207 — symmetric, key-derivation and RNG provider seams `ISymmetricCryptoProvider`, `IKeyDerivationProvider` and `ISecureRandomSource`, added on the terms #4207 set: only once a concrete requirement arrived (a validated module that must perform *every* operation, not just the asymmetric ones), behind a fast path, with `SymmetricChannelCryptoBenchmarks` as the gate. They are **facets discovered by type test**, not members of `ICryptoProvider`, so providers written against the shipped interface still compile. `PlatformCryptoProvider` implements all three, which answers the "an interface nothing implements" objection recorded in the issue. `CryptoProviderFacets` returns `null` both when no registry is configured *and* when resolution lands on the platform, because the platform facets are the inline code. Resolution happens once, in `CalculateSymmetricKeySizes`, and is held for the life of the channel. `CryptoCompliance.GetUnservedOperationPurposes` closes the failure the seam would otherwise introduce: a provider bound to `ChannelSymmetric` without implementing the facet is silently replaced by the platform, so a deployment could believe its validated module performed the per-message cryptography while the platform did. Under `FipsOnly` that now refuses to start. ## #4208 — the asymmetric path is awaitable `RSA` and `ECDsa` are .NET's contracts and cannot be replaced without making every ready-made hardware and cloud implementation unusable. Instead an implementation may **also** declare `IAsyncRsaKey` / `IAsyncEcdsaKey`, which the stack finds by type test. Now asynchronous: the secure channel open and renew path, user identity token signing and decryption, and session activation. A software key declares neither facet, so those paths complete synchronously and nothing about their ordering changes. `ReadAsymmetricMessage` is refactored into shared helpers rather than duplicated — adding an `await` was not worth two hundred cloned lines of signature, padding and endpoint-selection logic. `AsymmetricMessage` / `AsymmetricWriteResult` exist because an asynchronous method cannot have `out` parameters. The channel no longer serialises its state on a monitor. A monitor cannot be held across an `await`, and a plain `SemaphoreSlim` was not an option at the time: the code relied on re-entrancy in eight places (`HandleIncomingMessage` holds the lock and calls `ForceChannelFault`, which takes it again; `ForceReconnect` and `Shutdown` do the same). `ChannelGate` replaces it at 32 sites and `DataLock` is `[Obsolete]`. That reliance has since been removed as well — see [Removing the gate's re-entrancy](#removing-the-gates-re-entrancy) below. ### Defects this surfaced, all fixed and regression-tested **Detached work started inline.** Writes were started with `_ = WriteBuffersAsync(...)`. An async method runs its prologue on the caller's stack, so `LeaveInheritedContext` — which exists to stop *detached* work inheriting the right to re-enter the gate — was stripping the **caller's own** entitlement; when the send then completed synchronously, the completion blocked on a gate that very thread was holding. The client sent its OpenSecureChannel request and never processed the response. Writes are now queued. **A synchronous gate handle held across an `await`.** `Enter()` records the acquiring thread so an inline completion callback can re-enter — `ChannelAsyncOperation` invokes its callback both inline and detached. That record is only sound while the holder is synchronously on that thread; the token renewal path held it across an await, so a reused pool thread was recognised as the holder. **`SignAsync` rejected the policy that signs nothing.** It validated the certificate before the algorithm, so `SecurityPolicies.None` — which carries no certificate — threw where the synchronous `Sign` returns `null`. **`EnterAsync` lost the holder when contended** *(found in review)*. The `AsyncLocal` recording ownership was written inside the awaiting method, and .NET restores the caller's execution context when an async method completes, so the write was discarded. The caller returned owning the semaphore while `IsHeldByCurrentContext` reported `false` — and any nested `Enter()` then blocked forever on a gate its own context held. Reachable remotely: `HandleIncomingMessageAsync` → error path → `ForceChannelFault`. The holder is now published from the caller's frame before the wait, and only its depth is raised in the continuation. **Queued writes lost their order** *(found in review)*. Sequence numbers are assigned under the gate, but independently queued work items reach the transport's FIFO send lock in arbitrary order, so a peer could reject a chunk with `BadSequenceNumberInvalid`. Writes are now appended to a chain: still off the caller's stack, but a write cannot start before its predecessor finishes. ### Four more this surfaced in CI, on paths only the integration suites reach Queuing the writes to fix the deadlock above turned out to have moved three implicit guarantees, and the gate had one more liveness bug. Each was reproduced locally against a clean baseline before it was changed. **A faulting channel's error message never reached the peer.** A listener channel sends a UASC `Error` describing why it faulted and closes the transport as its next statement. That worked only because the write started on the caller's stack, so the bytes reached the socket first; queued, the close discarded them and the client saw `BadConnectionClosed` instead of the certificate status it was told to expect. Nine `SecurityCertValidationTests` cases assert exactly that status. The terminal error is written inline again, on a path that touches the gate only to disclaim what it inherited and reports nothing — a channel that is already faulted is owed nothing beyond its buffer. **The certificate of a rejected sender leaked.** An asynchronous method cannot have `out` parameters, so the asymmetric read returns a record instead. That silently dropped an ownership guarantee: an `out` parameter reaches the caller even when the method goes on to throw, so the caller's `catch` disposed the sender certificate and reported it to the audit. A return value does not — nine leaked certificates, one per rejection case, which the leak detector fails the run on. The parsed certificate is now handed to the caller before anything that can reject the message, restoring both the disposal and the audit. **The gate could strand its own permit.** It released only when the holder's depth returned to zero, but that depth lives in an `AsyncLocal` and is therefore shared with every context forked from the owner's — a timer callback whose timer was created inside the region, for instance. Such a context entering the gate raised the shared depth, so the owner's own exit saw a non-zero depth and returned without releasing, and nothing afterwards could: the nested exit is not the owner. Every later entry on that channel blocked forever. `Opc.Ua.Features.Tests` hung after its tests had finished, with a receive loop waiting on the gate and the server unable to stop. The owner now releases when it leaves; depth still gates re-entrancy, but liveness no longer depends on a count another context can move. **Write completion throttled the writer.** Reporting a completed write sat inside the ordering chain, and the client channel implements that callback by entering the gate — so every write queued behind a gate the receive loop contends for on each incoming chunk. A session with fifty subscriptions publishing continuously had its publish requests throttled enough that the client kept topping them up past its own limit, which `Subscriptions.Classic` asserts against. Completion is now reported off the chain, which covers only what ordering actually requires. ### Still synchronous, and why Certificate, certificate request and revocation list signing cannot be made asynchronous at all: `X509SignatureGenerator.SignData` is invoked by .NET's own builders. Service faults and the `Reconnect` override are reached from synchronous call sites (`SendServiceFault` alone has eight) and are left as they are behind scoped, commented suppressions. All three are documented rather than left to be rediscovered. ## #4210 — PubSub This one produced a finding rather than the API the issue asked for, and the finding is the point: > **With a standard Security Key Service the key necessarily exists in process memory.** `GetSecurityKeys` (Part 14 §8.3.2) returns raw key bytes over the wire, so the "key never leaves the device" property #4192 achieved for client and server **cannot** be achieved for PubSub through the SKS pull profile. That is a property of the specification, not of this stack. A wrapped-key envelope would change what is on the wire and break interoperability with third-party key services and publishers, so it was rejected rather than built. What is achievable is delivered: the per-message AES-CTR and HMAC route through `ISymmetricCryptoProvider`, so a validated module performs them. `IPubSubSecurityPolicy` is **not** modified — new members would break every implementation of it; the provider is supplied through the policy constructor. `PlatformSymmetricCryptoProvider` gained AES counter mode, which the seam should have covered anyway since the algorithm enum names it. Key lifetime is also tightened: unpacking an SKS response copied key material into intermediates that `ByteString.Create` then copied again, leaving plain keys in the heap for the lifetime of the process. Those are now zeroed. ### Two wiring gaps this depended on, both found in review The registrations above were **inert**, so this section originally claimed more than the code did: - `CryptoProviderConfiguration` was applied nowhere outside tests, so `AddCryptoProvider(configure)` never reached the registry in a real host. The registry now applies every registered configuration when it is first resolved — which also affects #4207's DI story. - `PubSubSecurityWrapperResolver` selected `PubSubAes256CtrPolicy.Instance`, the provider-less static singleton, ignoring the provider-backed bundles registered for it. A `FipsOnly` deployment therefore certified while the platform did the work. The resolver now selects from the registered policies; the container and `PubSubApplicationBuilder` both supply theirs. A third, smaller one: `CryptoProviderBuilder.Use` published a **policy-scoped** random source as the process-wide nonce source, so `.For(purpose, policyUri).Use(hsm)` redirected nonce generation for every other policy too. Only an unscoped binding does that now. ## #4206 — a provider can contribute a security policy Added in review: the policy set was fixed at compile time, so a provider could contribute key custody but not a *profile*. It is now table-driven and open, and the blockers #4206 listed are gone. ```csharp using IDisposable registration = SecurityPolicies.Register( new SecurityPolicyInfo("urn:vendor:SecurityPolicy#CustomProfile", "CustomProfile") { PlatformSupport = () => true, SupportedCertificateTypes = [ObjectTypeIds.RsaSha256ApplicationCertificateType], IsDefault = true }); services.AddOpcUa() .AddCryptoProvider(crypto => crypto .For(CryptoPurpose.ChannelSymmetric, "urn:vendor:SecurityPolicy#CustomProfile").Use(module)) .AddSecurityPolicy(customPolicy); ``` - **The reflection is gone.** Four lookup tables were built by reflecting over the fields of `SecurityPolicies` and `SecurityPolicyInfo`. They are one explicit table now, which removes the **last reflection in `Security/Constants`** — so this is a trimming and Native AOT improvement as much as a feature. - **`SecurityPolicyInfo` is constructible from another assembly** and carries the metadata the helpers need, so `IsPlatformSupportedName`, `GetDefaultUris`, `MapSecurityPolicyToCertificateTypes` and `GetCurveFromCertificateTypeId` read that table instead of each being hand-written. Adding a policy is one entry rather than five edits. - **The dead `#if CURVE25519` conditional is deleted** from `Security/Constants`. - Registration is copy-on-write, so the read path stays lock-free. Shadowing a built-in policy requires `replaceExisting: true` and is reversible by disposing the registration. `RegisterLightsUpCurvePoliciesFromOutsideCore` is the acceptance test the issue asked for: it registers `ECC_curve25519` and `ECC_curve448` from the test project and asserts they become discoverable through the public API. **Registering a policy makes it advertised and resolvable; it does not supply the cryptography behind it.** For those two curve profiles the in-tree key agreement is still behind a compile symbol no project defines plus a BouncyCastle dependency, so a deployment lighting them up provides the operations through a provider. That division is the point: the policy set says *what* is offered, the provider says *who* performs it. ## Removing the gate's re-entrancy Folded in from #4242, which was merged into this branch. `ChannelGate` was re-entrant, and that forced the design behind most of the defects above. A holder that awaits releases its thread, so ownership could not be tracked by thread identity — it had to live in an `AsyncLocal<Holder?>`. A logical context is *inherited*, so anything started while the gate was held inherited the right to re-enter and ran inside the guarded region alongside whatever started it, unless it opted out by hand via `LeaveInheritedContext()` at each of eight sites, enforced by review alone. Five of the defects listed above trace to that mechanism, including both CI hangs. So the *requirement* was removed rather than the mechanism, and the `AsyncLocal` then deleted itself. Every path that re-entered now calls a lock-free `Core` method (`ForceChannelFaultCore`, `ShutdownCore`, `ForceReconnectCore`, `CompleteConnectCore`), and three regions stopped running foreign code under the lock: the `RequestReceived` dispatch is deferred until the gate is released, `OnTransportError` no longer wraps the `HandleSocketError` virtual, and the callback setters and `EndpointDescription` publish a single reference with a volatile write instead of locking — the last is load-bearing, because `AsyncResultBase.OperationCompleted` invokes the reverse-hello callback inline while the gate is held. `ChannelGate` is now a `SemaphoreSlim(1,1)` behind a `Releaser`; `Holder`, the thread-identity fallback, `LeaveInheritedContext`, `IsHeldByCurrentContext` and all eight disclaimer calls are gone, with the call shape at all 32 sites unchanged. `SaveIntermediateChunk`, `GetSavedChunks` and `DoMessageLimitsExceeded` gained a `gateHeld` argument, because the client reaches them both with and without the gate and must pick the matching teardown — without it an oversized `OpenSecureChannel` response deadlocked the receive loop against itself. The remaining paths were not found by reading the code, since that is what missed them originally. A temporary strict mode made every re-entrant acquisition throw with its stack, the suite was run under it until clean, and it was deleted with the re-entrancy it measured. It caught two sites review had missed: the `EndpointDescription` **property getter** taking the gate, and the `RequestReceived` dispatch to application code happening inside the guarded region. Two deliberate residuals are documented for reviewers: `Dispose` no longer takes the gate (unavoidable — it is reached from `ChannelFaulted` with the gate held, where the old re-entrant acquisition excluded nothing anyway), and the deferred request dispatch builds `SecureChannelContext` just after gate release rather than under it. ## Testing New: the three seams and their platform implementations, facet discovery and resolution precedence, `FipsOnly` refusing a provider that cannot serve what it was bound to, AES counter mode (including a non-block-aligned round trip and a check that the counter carries, since one that did not would silently reuse key stream), `ChannelGate` (exclusion, re-entrancy across an await, re-entrancy on a reused thread, and the deadlocks above), the asynchronous channel paths, and the PubSub policy provider. Each review finding got a test that fails without its fix: - `ContendedEntryStillRecordsTheHolderAsync` — established contention deterministically, then re-entered; failed on the pre-fix gate. Removed with the re-entrancy it covered; `ChannelGateTests` now asserts the non-re-entrant contract instead. - `SecureClientAndTcpListenerKeepChunkOrderUnderConcurrentRequestsAsync` — 60 concurrent multi-chunk `SignAndEncrypt` requests over a real loopback listener; failed 3 runs out of 3 with the previous dispatch, passes 3 out of 3 with the chain. - `DependencyInjectionRoutesPubSubCryptoThroughTheRegisteredProviderAsync` — builds a container, registers a counting provider, wraps a message and asserts the provider actually performed the encryption and the signature. This is the assertion the PubSub claim above now rests on. - `PolicyScopedRandomBindingDoesNotEscapeItsPolicy`. Two further assertions are the ones that would catch a silent defect: a provider and the platform must produce **byte-identical** output — a validated module that disagreed could not interoperate — and with a software key the returned `ValueTask` must report `IsCompletedSuccessfully`, which is what makes "ordering is unchanged" checkable rather than merely asserted. `SymmetricChannelCryptoBenchmarks` gains a provider-registered variant beside the existing baseline, so the cost of the indirection is measured rather than assumed. Verified locally, every suite at or above its recorded baseline: `dotnet build UA.slnx -c Release` **0 warnings, 0 errors on all six target frameworks**; Core **4413** (net10.0); Core.Security **617**, with its certificate-leak detector clean; PubSub **1350** plus PubSub.Diagnostics **54** and PubSub.Udp **233**; Features **156**; Subscriptions.Classic **32**; Server integration **4053**; Client integration **2123**; MigrationAnalyzer **151** and MigrationAnalyzer.Core **13**. Zero failures in each. Core moved from 4388 to 4381 when the gate change landed — eight re-entrancy tests were removed along with the behaviour they covered, and one deadlock regression test added — and has since grown to 4413 with the registry isolation, provider-output and security-review tests. `ChannelGateTests` was rewritten around the new contract — exclusion across both entry modes, an uncontended asynchronous entry not suspending, re-entry blocking by contract, and the property that used to be the hazard: work started while the gate is held is excluded without having to opt out. Every CI failure this PR produced was reproduced locally, diagnosed against a clean `origin/master` baseline in a second worktree, and fixed at the source. Two hangs were found from `--blame-hang-timeout` dumps read with `dotnet-dump`. One failure that looked like a candidate was confirmed **not** to be: an intermittent `Opc.Ua.Aot.Tests` connect failure, pre-existing on `origin/master` (5 of 8 baseline runs failed, versus 3 of 14 here). ### The publish request overshoot `Subscriptions.Classic.PublishRequestCountAsync` failed on the Windows net10.0 leg with **207** outstanding publish requests against a cap of `max(maxServerPublishRequest, subscriptions)` = 50. It failed in 2 of 6 pipeline builds on this branch, never on `master` in 13 sampled builds, and never locally in 12 attempts including under deliberate CPU contention. The mechanism is a genuine race. `ClassicSubscriptionEngine.QueueBeginPublish` was a **check-then-act**: it read `GoodPublishRequestCount`, compared it against `GetDesiredPublishRequestCount(false)`, and issued a publish if it was below. It runs on every publish completion, so *n* completions that overlap each read the same value, each conclude they are below the cap, and each issue — overshooting by up to *n* per round. The read also *lags*: a request only becomes visible to `GoodPublishRequestCount` once `AsyncRequestStarted` has recorded it, which happens after the request has already been sent. That code is pre-existing and was untouched by this PR. What this PR changed is how hard the race is hit: the server used to hold the channel gate across request processing, which serialised every request on a channel, so publish completions arrived one at a time. Dispatching `RequestReceived` outside the gate — an intended improvement, since a channel should not serialise on application request-processing time — lets them complete concurrently. Rather than change the publish pipeline blind, the race was first made reproducible. `ConcurrentPublishReEvaluationDoesNotExceedDesiredRequestCount` drives 64 barrier-released completions against a mocked context whose count lags the send exactly the way the session's does. It issued **64 requests where 5 were desired** — the same defect the CI leg hit, now deterministic and running in milliseconds. The engine now tracks the requests it has in flight and reserves a slot with a compare-exchange before sending, so the decision and the send are one atomic step. The reservation is released when the request completes, when the send is declined, or when it throws. The repro test then issues at most the desired count. `StartPublishing` is deliberately left **uncapped**. It is the recovery valve that refills a pipeline whose requests are outstanding but no longer expected to return, so gating it on the in-flight count would remove the only path out of a stalled pipeline. An earlier revision of this fix did gate it, and that is the one change here that could not be justified on the evidence — it was reverted. The assertion was **not** loosened, retried away, or marked flaky: it caught a real throttling regression earlier in this PR and is worth keeping sharp. Validated: the repro test red before the fix and green after; `Subscriptions.Classic` 32/32 **five** consecutive runs on net10.0 and once on net48; `Opc.Ua.Client.Tests` 2124 on net10.0 across four consecutive runs and 2127 on net48; `Opc.Ua.Subscriptions.Tests` 604 passed with this fix in place. ### The SSH.NET advisory that turned CI red repo-wide Partway through this work every build job began failing with `NU1903: Package 'SSH.NET' 2025.1.0 has a known high severity vulnerability` ([GHSA-q939-rpr3-3284](https://github.com/advisories/GHSA-q939-rpr3-3284), `<= 2025.1.0` affected, first patched `2026.0.0`). Nothing in the dependency graph changed — the advisory was published. Build 17026 failed on it where build 17000, two hours earlier on an identical graph, had passed every build job. SSH.NET is not referenced anywhere in the repository; it arrives transitively through `Testcontainers.Kafka`. `master` landed the remediation independently while this was being investigated (a central transitive pin to `2026.0.0`, which `CentralPackageTransitivePinningEnabled` exists for). This branch briefly carried its own identical pin, which made the merge commit hold **two** `PackageVersion` entries for SSH.NET and fail restore with `NU1506`. The duplicate is removed and `master` is merged, so the pin now comes from one place only. The audit was **not** suppressed: `NuGetAudit` stays enabled and `NU1903` stays an error. Verified after the merge — `dotnet restore UA.slnx` reports neither NU1903 nor NU1506, and `dotnet build UA.slnx -c Release` is **0 warnings, 0 errors** with the audit live. The remaining red check in build 17026 was not from this change either. `aot-ubuntu-latest` is the intermittent `Opc.Ua.Aot.Tests` failure already noted above — `ReconnectSessionAsync` timed out at 30s with 135 of 136 passing. It remains below the `origin/master` baseline rate (4 of 15 sampled here against 5 of 8 on the baseline), and the managed reconnect paths are green throughout: `Opc.Ua.Client.Tests` 2124 across four consecutive runs and `Opc.Ua.Subscriptions.Tests` 604. ## Documentation `docs/CryptoProvider.md` gains sections on substituting the symmetric primitives, using a key served over a network, and PubSub — including what cannot be claimed and why, the scope rule for a random-source binding, and how to supply a provider-backed policy outside the container. `docs/MigrationGuide.md` covers the `[Obsolete]` channel virtuals and `DataLock`, and the non-re-entrant gate's contract for channel subclasses. `docs/WhatsNewIn2.0.md` is updated. `plans/cryptooffboard.md` is deleted now that the work it planned has landed. ## Note for reviewers This touches the secure channel handshake, which is the most security-critical code in the stack. Nine distinct concurrency and lifetime defects were found and fixed in it during this work — three during development, two in review, four from CI — so `ChannelGate`, the write path and the certificate ownership across the new suspension points deserve the closest reading. Five of those nine traced to the gate's `AsyncLocal`, which is why the re-entrancy was subsequently removed and the `AsyncLocal` deleted. The wrapper/`Core` split and the two documented residuals in that section are the parts most worth a second reading. The #4206 work removes the last reflection from `Security/Constants` and rewrites the policy lookup tables, so `SecurityConfiguration.SupportedSecurityPolicies` and the endpoint set it feeds are worth a second look: an earlier revision of it changed that set and broke six server tests, which is now covered by `SecurityConfigurationBuildsExpectedSupportedPolicySet`. ## Related Issues - Fixes #4206 - Fixes #4207 - Fixes #4208 - Fixes #4210 - Follow-up #4250 — this PR makes the security policy set injectable, but nothing in the stack resolves `ISecurityPolicyRegistry` from DI yet, so every production path still lands on the `Default` fallback. #4250 tracks threading it through `Session` and `UaSCBinaryChannel`, which is what makes the injection points already added here load-bearing. ## Checklist - [x] I have signed the [CLA](https://opcfoundation.org/license/cla/ContributorLicenseAgreementv1.0.pdf) and read the [CONTRIBUTING](https://github.com/OPCFoundation/UA-.NETStandard/blob/master/CONTRIBUTING.md) doc. - [x] I have added tests that prove my fix is effective or that my feature works and increased code coverage. - [x] I have added all necessary documentation. - [x] I have verified that my changes do not introduce (new) build or analyzer warnings. - [x] I ran **all** tests locally using the **UA.slnx** solution against at least .net **framework** and .net **10**, and all passed. - [ ] I fixed **all** failing and flaky tests in the CI pipelines and **all** CodeQL warnings. - [x] I have addressed **all** PR feedback received. --------- Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com> Co-authored-by: marcschier <11168470+marcschier@users.noreply.github.com> Copilot-Session: 7a68ca72-83e8-496c-b2f5-3df4c718320b Copilot-Session: 3f5ab76a-c0ba-4f47-87ac-54d1f4182c6b	2026-08-15 05:23:20 -07:00
Marc Schier	2a4bc0fe24	Make reverse connect lifecycle fully asynchronous (#4059) ... ## Summary - add cancellation-aware asynchronous start, lazy initialization, stop, reload, and disposal for `ReverseConnectManager` - make listener replacement transactional with explicit ownership, rollback/restoration, stale watcher protection, and deterministic concurrent start/stop/dispose behavior - add injectable async configuration providers, remove the former protected configuration hooks, and retain obsolete synchronous lifecycle compatibility wrappers - integrate eager Generic Host startup plus non-host lazy startup, including custom application-configuration providers - serialize `ReverseConnectHost` lifecycle and extend raw TCP, Kestrel TCP, WSS, hosting, concurrency, rollback, and NativeAOT coverage - document the async lifecycle, hosting behavior, cancellation guarantees, provider model, and migration path ## Validation - `Opc.Ua.Client` Release build across all supported target frameworks - reverse-connect client and core tests on net10.0 and net48 - raw TCP, Kestrel TCP, and WSS integration tests on net10.0; raw TCP on net48 - NativeAOT publish and 115-test executable run Closes #4013. Related: #3985, #4009.	2026-07-23 10:40:32 +02:00
Marc Schier	7cec2d614d	Move repository files to the consolidated layout	2026-07-16 06:25:54 +02:00