Directory.Packages.props

<Project>
  <PropertyGroup>
    <ManagePackageVersionsCentrally>true</ManagePackageVersionsCentrally>
  </PropertyGroup>
  <ItemGroup>
    <PackageVersion Include="Alba" Version="8.2.0" />
    <PackageVersion Include="Aspire.Hosting.AppHost" Version="8.2.0" />
    <PackageVersion Include="BenchmarkDotNet" Version="0.13.10" />
    <!-- Bobcat's out-of-process test supervisor, used only by the Nuke build (#5096).
         It drives each test project's Microsoft.Testing.Platform host as a worker,
         so a failure can be retried in a FRESH process and a pass-on-retry is
         reported as flaky rather than laundered into a clean pass. See
         build/SupervisedTests.cs. Bobcat.Supervisor pulls in Bobcat (the
         Bobcat.Resilience types) transitively. -->
    <PackageVersion Include="Bobcat.Supervisor" Version="0.6.1" />
    <PackageVersion Include="Bogus" Version="35.6.5" />
    <PackageVersion Include="Confluent.Kafka" Version="2.3.0" />
    <PackageVersion Include="coverlet.collector" Version="6.0.0" />
    <PackageVersion Include="DotNet.ReproducibleBuilds" Version="1.2.39" />
    <PackageVersion Include="FSharp.Core" Version="9.0.100" />
    <PackageVersion Include="FSharp.SystemTextJson" Version="1.3.13" />
    <!-- JasperFx 2.5.0: per-tenant event partitioning + tenant-aware async daemon
         surface (jasperfx#407 / CritterWatch#209). Released to nuget.org; this
         consumes the published packages rather than the local-feed pt209 prereleases.
         JasperFx 2.8.0: dynamic-tenancy surface (jasperfx#413) — IDynamicTenantSource<T>
         gains an `AddTenantAsync(tenantId, ct) → string` auto-assign provisioning
         overload. ShardedTenancy implements it so store-agnostic provisioning works
         under the sharded multi-tenancy model.
         JasperFx 2.8.1: fix for marten#4663 — EventPage.CalculateCeiling guards the
         Last() access for the all-events-skipped batch case so projection rebuilds
         no longer crash with "Sequence contains no elements" (jasperfx#416).
         JasperFx 2.8.2: fix for marten#4665 (jasperfx#418) —
         JasperFxAsyncDaemon.CatchUpAsync(CancellationToken) now dispatches on
         _tenantHighWater when the database implements ICrossTenantRebuildSource, so
         ForceAllMartenDaemonActivityToCatchUpAsync actually advances async projections
         under UseTenantPartitionedEvents (previously: the store-global high-water
         path was used, but the global mt_events_sequence is never advanced under
         partitioning, so catch-up stayed pinned at zero). Per-tenant fan-out mirrors
         rebuildProjectionForTenant's ceiling-lookup pattern. Single-tenant / non-
         partitioned stores keep the byte-for-byte global path.
         JasperFx 2.9.0: jasperfx#419 (ShardName.Compose is now the only path used
         for shard name construction), #420 (cap concurrent projection rebuilds per
         database, default 4), #422 (MultiStreamProjection.Options.CacheLimitPerTenant
         now defaults to 1000 instead of 0 — projections that relied on the unbounded
         default still work but per-tenant slice caches now bound by default), and
         #424 (new storage-agnostic IEventStoreInstrumentation hook for telemetry).
         JasperFx 2.9.1: jasperfx#429 — source generator dedupes overridden required
         members in evolver construction (codegen-only fix; no runtime/daemon change).
         JasperFx 2.9.2: jasperfx#431 — ProjectionErrorHandlingDescriptor on EventStoreUsage
         so monitoring tools (CritterWatch) can read the daemon error-handling policy off
         the wire instead of presuming skip-and-DLQ behaviour. See JasperFx/ProductSupport#3.
         JasperFx 2.9.9: jasperfx#443 (marten#4721) — the continuous-startup composite
         "optimized rebuild" now runs on a task separate from the SubscriptionAgent's
         single-consumer command loop. It previously ran inline inside that loop while the
         replay posted RangeCompleted commands back into the same bounded (10k) command
         channel, so the only reader was blocked awaiting its own posts — the shard silently
         wedged at a batch boundary under UseTenantPartitionedEvents once the channel filled.
         JasperFx 2.9.10: jasperfx#444 (marten#4730) — the in-memory aggregate cache is OFF by
         default again (CacheLimitPerTenant 1000 → 0) and, when enabled, only populates after the
         owning batch commits. Previously the cache was written during batch build before commit,
         so a failed/retried batch (e.g. an out-of-order Update-before-Create stream throwing)
         re-applied events on top of the already-mutated cached aggregate, double-applying to
         unrelated, correctly-ordered streams in the same page.
         JasperFx 2.9.13: marten#4729 (two parts) — the optimized composite rebuild now propagates
         the composite's ShardExecutionMode down to its member executions (jasperfx#447). Member
         executions defaulted to ShardExecutionMode.Continuous (CompositeReplayExecutor only set the
         parent's Mode), so during a composite rebuild the stage members fired their RaiseSideEffects
         (PublishMessage/AppendEvent) — unlike the classic single-stream rebuild, which is pinned by
         side_effects_do_not_happen_in_rebuilds. Members now inherit CatchUp/Rebuild and suppress
         side effects during a rebuild. The follow-up (jasperfx#448) decouples slice.Snapshot
         assignment from side-effect publishing in AggregationRunner so multi-stage composites still
         fan the upstream Updated<TDoc> events to downstream stages during a rebuild (previously the
         snapshot was only set on the Continuous path, so post-#447 a rebuild produced no upstream
         snapshot and downstream stages threw NREs). 2.9.12 shipped only #447 and is incomplete. -->
    <!-- JasperFx 2.12.0: dead-letter row drill-in (IEventDatabase.QueryDeadLetterEventsAsync) +
         DeadLetterEvent.TenantId + tenant-aware daemon PauseShardAsync/rebuild fan-out (jasperfx#453).
         Marten implements QueryDeadLetterEventsAsync + per-tenant FetchDeadLetterCountsAsync below. -->
    <!-- JasperFx 2.13.2: jasperfx#464 — SubscriptionDescriptor carries ImplementationType +
         AggregateType (TypeDescriptor), populated in the shared
         SubscriptionDescriptor(ISubscriptionSource, IEventStore) ctor that Marten's projections
         funnel through via ProjectionGraph.Describe. See marten#4772. -->
    <!-- JasperFx 2.20.0: jasperfx#487 — IEventStore.DistributesAgentsPerTenant (default false) for
         node-distributed daemon hosts (Wolverine) to fan out one agent per (shard, tenant), plus the
         hardened per-tenant StartAgentAsync: tenant ceilings primed before start (SubscribeFromPresent
         no longer rewinds to 0), exact-identity precedence, loud failure on non-partitioned stores. -->
    <!-- JasperFx 2.21.0: jasperfx#491 (epic jasperfx#486) — distributor-level per-tenant shard
         expansion for the NATIVE daemon: the Solo / SingleTenant / MultiTenanted distributor
         concretes grow a distributesAgentsPerTenant ctor overload that expands store-global shard
         names into per-tenant ShardNames (PerTenantShardExpansion, re-evaluated every
         BuildDistributionAsync so tenant add/remove converges on the leadership polling cycle),
         and the coordinator reconciles running agents against the freshly built distribution
         (stops agents whose identity disappeared). Marten threads
         IEventStore.DistributesAgentsPerTenant into BuildDistributor (ProjectionCoordinator).
         Also DaemonMode.ExternallyManaged (jasperfx#490, wolverine#3290 — external hosts run
         projections; the store hosts no coordination and must not warn).
         JasperFx 2.24.1: jasperfx#499/#500 (marten#4874 case B) — ProjectionCoordinatorBase.executeAsync
         terminates the leadership loop on a disposed data source / wrapped cancellation instead of
         re-polling, so a cold HotCold node's advisory-lock poll no longer spins OpenAsync against a
         data source being disposed during shutdown. Pairs with Weasel 9.16.2 (weasel#349).
         JasperFx 2.25.0: jasperfx#501 (CritterWatch#678) — every published ShardState is stamped with
         its owning database identifier.
         JasperFx 2.26.0: jasperfx#4913 (marten#4913) — under UseTenantPartitionedEvents the store-global
         HighWaterAgent no longer runs its recurring max(seq_id)-over-mt_events scan (which fanned out
         across every tenant partition each tick); tenant high water is driven per-tenant by the
         coordinator + the jasperfx#492 timer, so the global mark is only seeded once at startup.
         JasperFx 2.27.0: jasperfx#506/#507 — Block<T> fault semantics + daemon OnError wiring (block
         failures are loud, dead consumers throw from Post), and teardown catch clauses classify via
         CancellationExceptions.IsCancellationLike instead of discarding every exception type while a
         shard CTS is cancelled; jasperfx#505 — source-gen evolver two-statement fix for the
         GetUninitializedObject cast (rides in the bundled JasperFx.Events.SourceGenerator analyzer).
         JasperFx 2.28.1: jasperfx#528 (marten#4966) — natural-key discovery now builds an event
         mapping for a two-arg static evolve method (static TDoc Apply(TEvent e, TDoc current)), so a
         natural key that CHANGES via an update event is written to mt_natural_key_X on live append and
         rebuild; previously only the create event's key was ever recorded.
         JasperFx 2.29.0: jasperfx#529 (marten#4975) — exact ReadProjectionProgressAsync(ShardName) overload
         on IEventDatabase (no version/shard collapsing), adopted by MartenDatabase here.
         JasperFx 2.29.1: jasperfx#530 (marten#4953) — HighWaterAgent.CheckNowAsync catches up to the
         COMMITTED ceiling via the new IHighWaterDetector.FetchCommittedHighWaterCeilingAsync DIM
         (overridden here with max(seq_id)) instead of looping DetectInSafeZone to the reserved
         last_value, and the loop is time-bounded — rebuild/catch-up during concurrent appends now
         waits for in-flight events instead of pressuring detection to skip them.
         JasperFx 2.30.0: jasperfx#531 (marten#4978) — shared ITenanted marker promoted into
         JasperFx.MultiTenancy; Marten.Metadata.ITenanted now derives from it.
         JasperFx 2.32.0: jasperfx#539 — HighWaterAgent liveness heartbeat, staleness surface, and
         local restart seam (IProjectionDaemon.HighWaterLastPolledAt / IsHighWaterStale /
         RestartHighWaterAgentAsync; DaemonSettings.HighWaterStalenessThreshold; ShardAction.Faulted/
         Restarted). Also carries the jasperfx#540 faulted-start teardown that first shipped in 2.32.0.
         Foundation for the Phase 2 high-water health check (marten#4986).
         JasperFx 2.33.0: jasperfx#543 (GH-3521) — JasperFxOptions.ApplicationAssemblyReuseWarning
         surfaces divergent process-pinned application-assembly reuse across multi-host test processes;
         Marten buffers and logs it at startup (marten#4996). Also ships the step-instrumented
         aggregation fold + MultiAggregateProjectionResult (jasperfx#547) and the serializable
         EventTagQuery DCB source (jasperfx#548).
         JasperFx 2.33.1: jasperfx#550/#551 (marten#5001) — ShardStateTracker.AssignedNodeNumber is
         stamped onto every published ShardState, so a distribution layer (Wolverine-managed
         subscription distribution) can drive the extended-progression running_on_node column. Marten's
         WriteExtendedProgressionAsync already persists it; this bump makes the tracker seam available.
         JasperFx 2.35.0: jasperfx#561 — new backwards-compatible virtual overload on the aggregation
         projection base, RaiseSideEffects(TOperations operations, TId id, IEventSlice<TDoc> slice).
         AggregationRunner now always invokes the 3-arg overload with slice.Id, so a projection can
         recover the slice identity even when slice.Snapshot is null (e.g. a deleted MultiStreamProjection
         slice) to emit a follow-on event or publish a message. The old 2-arg override still works; the
         default 3-arg implementation delegates to it.
         JasperFx 2.36.0: three daemon/projection changes Marten consumes directly.
         jasperfx#564 (marten#5047) — DaemonSettings.StopAndDrainTimeout bounds how long a single
         shard's graceful stop-and-drain may take before it is cancelled; the default of 5 seconds is
         what was hardcoded before, so nothing changes unless configured.
         jasperfx#565/#567 (marten#5048) — ShardFailure/ShardFailureCategory/EventFailureDetails and the
         IEventFailureContext seam an exception implements to declare its own failure category and name
         the event it broke on. ShardState.Failure rides along on the paused/stopped states, and
         IEventDatabase.WriteExtendedProgressionAsync documents persisting (and clearing) it.
         jasperfx#569/#571 (marten#5052) — NaturalKeyEventMapping.Extractor widens from
         Func<object, object?> to Func<IEvent, object?> so an IEvent<T> [NaturalKeySource] handler is
         bindable, and an unbindable one now fails loudly from AssembleAndAssertValidity().
         JasperFx 2.36.1: jasperfx#572/#573 (found here) — ShardStateTracker now takes one lock across
         recording a published state and capturing the listener list, and ShardStatusWatcher subscribes
         and reads the current-state snapshot through it as a single atomic step. 2.36.0's #568 narrowed
         the lost-wakeup window between those two paths but left it open, so a watcher could fall between
         them and see the state on NEITHER — and once a high water agent reaches the head it has nothing
         left to publish, so the wait could only end in a timeout no matter how generous. That surfaced
         as HighWaterAgentTests.skips_multiple_gaps_and_keeps_advancing failing ~8 of 10 runs.
         JasperFx 2.36.2: jasperfx#574/#575 (marten#5055/#5056) — the daemon stop path is now safe to
         run after disposal: JasperFxAsyncDaemon.Dispose() is idempotent and StopAllAsync() no-ops on a
         disposed daemon, ProjectionCoordinatorBase.PauseAsync logs ObjectDisposedException at Debug
         instead of Error, and ProjectionCoordinatorBase.StopAsync calls the new abstract
         ClearResolvedDaemons() seam after disposing daemons so subclass caches drop the disposed
         instances (implemented here on ProjectionCoordinator + ExplicitProjectionCoordinator).
         JasperFx 2.36.3: jasperfx#592/#593 (marten#5090) — ProjectionCoordinatorBase.StartAsync now
         cancels AND drains any existing leadership loop instead of disposing the old
         CancellationTokenSource without cancelling it. Disposing a CTS does not cancel it and _runner
         was overwritten, so a ResumeAsync not preceded by PauseAsync orphaned a live executeAsync for
         the rest of the process — one that re-attained the leadership lock right after StopAsync
         released it, and that won an advisory-lock handle after the lock had been disposed. Pairs
         with Weasel 9.20.1 (weasel#396), which is what makes that stranded handle impossible.
         JasperFx 2.37.0: jasperfx#590 — OptionsDescription no longer loses an entire diagnostic
         description to one throwing property getter; set-only properties and indexers are skipped, and
         a getter that throws is recorded as OptionsValue.Unreadable (exception TYPE only, never the
         message, since descriptions ship to monitoring consoles). Also carries the jasperfx#594 patch
         half (a timed-out blue/green side-effect gate re-reads progression and succeeds when the replay
         had already reached the mark, plus a configurable DaemonSettings.SideEffectGateTimeout),
         jasperfx#595 (BatchingChannel could deliver its trailing batch twice on shutdown) and #597.
         JasperFx 2.37.2: compliance wave 2 (marten#5118) — four more shared suites plus the seam
         members they need. Nothing outside JasperFx.Events.ComplianceTests changed; the other
         packages move only because the line versions together.
         JasperFx 2.37.3: jasperfx#611 — JasperFx.Events.SourceGenerator now discovers an
         EventProjection's published document types SEMANTICALLY. Discovery used to match only
         GenericNameSyntax, so an explicit ApplyAsync override writing `ops.Store<Doc>(x)` registered
         Doc while the equally valid `ops.Store(x)` registered nothing at all, silently (storage is
         still provisioned on demand, so only schema-ahead-of-time, AllKnownDocumentTypes and rebuild
         teardown came up short). Binding the invocation gives the same answer for both spellings.
         Also stops `Store<object>(...)` registering `object` itself as a document type: registrability
         is now a SpecialType/TypeKind question, because object/string render through ToDisplayString()
         as C# keywords and slipped past the old name-prefix check. New JFXEVT005 (Info) flags a call
         that binds to the projection's session but whose document type cannot be named.
         JasperFx 2.38.0: two changes Marten consumes directly.
         jasperfx#616 (marten#5127) — the ProjectionScenario test harness is lifted into a new
         JasperFx.Events.TestSupport namespace so Marten and Polecat share one implementation
         instead of maintaining parallel ports.
         jasperfx#617 (marten#5095) — a projection registered through AddProjectionWithServices with
         a Scoped/Transient lifetime is now reachable from AggregatorFor, so live aggregation and
         single stream rebuilds run the actual projection instead of silently falling through to
         conventional aggregation off the aggregate type; and the projection a scoped wrapper wraps
         is validated rather than only the wrapper, so an invalid configuration fails at startup for
         every lifetime instead of only Singleton.
         JasperFx 2.39.2: jasperfx#630 (marten#5167) — the IEventDatabase batched extended-progression
         contract now REQUIRES one row per transaction, and ExtendedProgressionWriter flushes each
         batch ordered by shard name. A batch amortizes the CONNECTION, never the transaction: a
         multi-row statement holds a row lock on every shard in the batch until it commits, so one
         slow projection batch on one row stalled every other shard's telemetry on that database
         (Marten's side of this is #5186, which does not depend on this bump). The ordering closes
         the cross-writer deadlock hazard — the tracker is per-database and shared, and
         BuildProjectionDaemonAsync is not cached, so two writers can race over the same rows.
         JasperFx 2.39.3: two extended-progression fixes Marten consumes directly.
         jasperfx#631 — a status transition published while the shard has NO progression row to
         decorate landed nowhere, because every store's extended-progression write is update-only.
         That is every fresh shard's Started: the agent starts before its first batch commits. With
         the periodic beat off by default (jasperfx#622) there was no later write to correct it, so
         agent_status/heartbeat stayed NULL for the entire life of a healthy agent. Such a shard is
         now remembered and written again on the first publication carrying a committed sequence.
         jasperfx#630 follow-up — a duplicate ExtendedProgressionWriter on one database now logs a
         warning. It used to announce itself as lock contention; one-row-per-transaction writes made
         it harmless to correctness and therefore silent, while still meaning two daemons are
         started for one database. Reported, never refused: the duplicate is the symptom.
         JasperFx 2.43.0: jasperfx#639 — compliance wave 6 (#5153, #5154). The library gains
         StreamCompactingCompliance and EventDataMaskingCompliance, taking it to 24 suites / 199
         tests, plus the seam those needed (EventStoreComplianceFixture.ApplyEventDataMaskingAsync
         and the two ComplianceStoreConfig.AddMaskingRule overloads). Compliance sources only —
         nothing outside JasperFx.Events.ComplianceTests changed, so this bump is behaviourally
         inert for the shipping libraries and only affects the test projects that enrol the
         suites.
         JasperFx 2.44.0: jasperfx#641 — compliance wave 7 (#5149, #5150), the last two event
         sourcing suites in the backlog. RebuildAndCatchUpCompliance and DeadLetterCompliance take
         the library to 26 suites / 216 tests. Neither needed a seam addition: IProjectionDaemon
         already declared the whole rebuild surface, and the error policy
         (IEventStore<,>.ContinuousErrors) plus dead-letter reads (IEventStore.AllDatabases() ->
         IEventDatabase.QueryDeadLetterEventsAsync) were already shared and already overridden here.
         Compliance sources only, so again behaviourally inert for the shipping libraries.
         JasperFx 2.45.0: jasperfx#642 — compliance wave 8 (#5148, #5151), the last two suites in
         the event sourcing backlog. ConjoinedEventTenancyCompliance and SubscriptionCompliance take
         the library to 28 suites / 230 tests, and with them the ES compliance backlog is empty.
         Two additive seam members: ComplianceStoreConfig.ConjoinedEventTenancy and
         IComplianceStoreRegistrar.Subscribe(ComplianceSubscription), the latter paired with a
         per-consumer partial completing ComplianceSubscription against Marten's own ISubscription
         (IChangeListener is per-product, so that signature cannot be written once upstream).
         #5148 needed no tenant-scoped session seam after all — OpenSession(IEventDatabase,
         tenantId) was already on the shared generic IEventStore<,> and TenancyStyle was already a
         shared JasperFx.MultiTenancy enum. Compliance sources only; behaviourally inert for the
         shipping libraries.
         JasperFx 2.48.0: jasperfx#664 (#663) stamps stream identity on string-keyed appends, plus a
         batch of source generator projection-dispatch fixes (jasperfx#652-#656): ShouldDelete now
         composes with Create/Apply for the same event type, a genuine override of the dispatch
         virtuals is told apart from a `new` hiding member, the evolver is handed the registered
         projection instead of a shadow instance, `partial` is only required where the dispatcher
         actually lives, and unregistered published types are reported rather than skipped in
         silence. The dead opt-out for the removed lambda registration APIs is gone.
         JasperFx 2.49.0: jasperfx#665/#666 (marten#5241) — IDocumentReadOperations gains
         LoadAsync<T>(object id), so store-agnostic code can load a document keyed by a strong-typed
         identifier. Marten needs NO product change for it: IQuerySession already declares the
         overload (IQuerySession.cs) and QuerySession implements it publicly, dispatching on
         id.GetType() — Marten is in fact the store the member was modeled on. It ships with a
         default implementation (unbox a Guid or string and forward, else NotSupportedException) so
         Polecat and Fisher can take the bump before adopting it, which means the enforcement lives
         in DocumentLoadAndStoreCompliance rather than the compiler. What Marten does owe is the
         fixture line below: DocumentComplianceConfig gained ValueTypes/RegisterValueType<T>() and a
         fixture that does not replay it fails the two strong-typed facts at runtime.
         JasperFx 2.50.0: jasperfx#669/#671 — two additive changes. (1) The document session
         contracts gain an Events accessor: IDocumentReadOperations.Events (IQueryEventStore) and
         IDocumentSessionOperations.Events (IEventStoreOperations), both with throwing defaults.
         ⚠️ Marten did NOT satisfy either implicitly, on BOTH tiers: its sessions declare Events as
         Marten.Events.IQueryEventStore / Marten.Events.IEventStoreOperations — Marten's own
         subtypes — and C# interface implementation is not return-type covariant, so both members
         bound to the throwing defaults with no compile error anywhere. QuerySession and
         DocumentSessionBase now carry one explicit interface implementation each, pinned by
         DocumentSessionEventsCompliance. (2) IEventBinarySerializer and BinaryEventAttribute are
         promoted out of Marten.Events into JasperFx.Events; Marten's interface now derives from the
         core one, its attribute is checked alongside the core one (which is sealed, so no
         inheritance), and the registration surface is widened to the core interface.
         JasperFx 2.51.0: three additive changes, each with its own Marten adoption issue. All are
         purely additive — nothing here is a compile break, so this bump lands on its own and the
         adoptions follow separately.
         (1) jasperfx#672/#677 (marten#5248, marten#5249) — DocumentComplianceConfig gains a nullable
         StreamIdentity so a document suite can declare the stream identity style it needs instead of
         leaving each fixture to guess it (DocumentSessionEventsCompliance appends by stream key and
         failed three facts on every store's Guid default), and JasperFx.Events.BinaryEventAttribute
         is unsealed so Marten can derive its own back-compat attribute from it and collapse
         EventGraph.ResolveBinarySerializerFor back to one lookup.
         (2) jasperfx#673/#675 (marten#5250) — IDocumentSessionOperations.PendingStreams
         (IReadOnlyList&lt;StreamAction&gt;), so store-agnostic code can read the stream actions a session
         has queued but not committed. ⚠️ Same non-covariance trap as 2.50.0's Events accessor:
         Marten's PendingChanges.Streams() returns IList&lt;StreamAction&gt;, which does NOT satisfy an
         IReadOnlyList&lt;StreamAction&gt; member, so this binds to the throwing default until Marten
         writes the forward. Pinned by PendingStreamActionsCompliance.
         (3) jasperfx#674/#676 (marten#5251) — IAggregateWriteCache, AggregateCacheKey,
         NulloAggregateWriteCache, RecentlyUsedAggregateWriteCache and AggregateWriteCacheOptions move
         into JasperFx.Events.Fetching, with AggregateWriteCaching/CacheAggregatesForWriting&lt;T&gt;() on
         EventRegistry so all three stores inherit the registration surface. Grade 1 semantics only:
         the cached snapshot is a baseline, version and delta are always re-read, OCC is untouched.
         The default cache is backed by JasperFx.Core's RecentlyUsedCache, deliberately NOT
         Microsoft.Extensions.Caching.Memory — the prototype's IMemoryCache dependency would have been
         pushed onto every consumer of JasperFx.Events rather than one store.
         JasperFx 2.52.0: jasperfx#679/#680 (marten#5258) — IDocumentCommitListener, the
         store-agnostic post-commit SESSION hook, plus IDocumentChangeSet and IDocumentDeletion in
         JasperFx.Events.Documents, and DocumentComplianceConfig.CommitListeners/AddCommitListener so
         a compliance suite can register one before the store is built. Purely additive.
         ⚠️ NOT the same trap as 2.50.0's Events accessor or 2.51.0's PendingStreams, and the
         difference is worth stating because it changes where the risk lives: no member of this
         contract has a default implementation, so a near-miss is CS0535 at build time rather than a
         silent bind to a throwing default. What the compiler still cannot see is the WIRING — a
         store that declares the interfaces perfectly and never invokes the listener compiles clean
         and passes every other suite in the library.
         Marten adopts it as an ADAPTER rather than by widening its own types, and that is forced
         rather than stylistic. IChangeSet.Inserted/Updated are IEnumerable&lt;object&gt; against the
         contract's IReadOnlyList&lt;object&gt;; IChangeSet.Deleted is
         IEnumerable&lt;Weasel.Storage.IDeletion&gt; against IReadOnlyList&lt;IDocumentDeletion&gt;; and
         DocumentSessionListenerBase.AfterCommitAsync differs from the contract's signature in three
         of its four positions. Putting IDocumentChangeSet on IChangeSet would therefore be a
         breaking change to every existing Marten listener, and putting it on ISessionWorkTracker
         would additionally rope in ProjectionUpdateBatch, whose Inserted/Updated/Deleted all throw
         NotSupportedException. See DocumentCommitListenerAdapter/MartenDocumentChangeSet and
         StoreOptions.AddCommitListener, pinned by DocumentCommitListenerCompliance. -->
    <PackageVersion Include="JasperFx" Version="2.53.0" />
    <PackageVersion Include="JasperFx.Events" Version="2.53.0" />
    <!--
      The shared cross-store event sourcing compliance suites (#5116). Source-only package: the
      suites compile inside EventSourcingTests so JasperFx's aggregate source generator can bind
      Marten's own session types. Marten.Testing needs it too because MartenComplianceFixture,
      which closes the seam, lives beside the rest of the harness.
    -->
    <PackageVersion Include="JasperFx.Events.ComplianceTests" Version="2.53.0" />
    <!--
      Deliberately AHEAD of the rest of the JasperFx line, which stays at 2.41.0 until Marten adopts
      the strong-typed identity compliance suite that landed in 2.42.0. This package is analyzer-only,
      declares no dependencies, and Marten bundles its dll into Marten.nupkg (see
      _BundleEventsSourceGeneratorAnalyzer in Marten.csproj, marten#4557), so its version is
      independent of the runtime family's.

      2.42.1: jasperfx#637 (marten#5192) — the generator used to register an EventProjection's
      discovered published document types by emitting a parameterless constructor into the user's
      partial class. That is illegal on a type declaring a primary constructor, so
      `partial class MyProjection(ILogger logger) : EventProjection` failed to build with CS8862; and
      for a projection registered through AddProjectionWithServices the container calls the
      dependency-taking constructor, so the generated one never ran and the published types went
      silently unregistered. Registration now rides an override of ProjectionBase.PublishedTypes(),
      which does not care how the instance was constructed. The generator is otherwise byte-identical
      from 2.38.0 through 2.42.0, so this is an isolated swap of that one fix.
    -->
    <PackageVersion Include="JasperFx.Events.SourceGenerator" Version="2.53.0">
      <PrivateAssets>all</PrivateAssets>
      <IncludeAssets>runtime; build; native; contentfiles; analyzers; buildtransitive</IncludeAssets>
    </PackageVersion>
    <PackageVersion Include="JasperFx.SourceGenerator" Version="2.53.0" />
    <PackageVersion Include="Jil" Version="3.0.0-alpha2" />
    <PackageVersion Include="Lamar" Version="7.1.1" />
    <PackageVersion Include="Lamar.Microsoft.DependencyInjection" Version="15.0.0" />
    <PackageVersion Include="Marten" Version="8.2.1" />
    <PackageVersion Include="MemoryPack" Version="1.21.4" />
    <PackageVersion Include="Microsoft.CodeAnalysis.Analyzers" Version="3.3.4" />
    <PackageVersion Include="Microsoft.CodeAnalysis.CSharp" Version="4.8.0" />
    <PackageVersion Include="Meziantou.Analyzer" Version="2.0.257" />
    <PackageVersion Include="Microsoft.AspNetCore.OData" Version="8.2.5" />
    <PackageVersion Include="Microsoft.AspNetCore.SignalR" Version="1.1.0" />
    <PackageVersion Include="Microsoft.Extensions.DependencyInjection.Abstractions" Version="6.0.0" />
    <PackageVersion Include="Microsoft.Extensions.Logging" Version="8.0.0" />
    <PackageVersion Include="Microsoft.Extensions.Logging.Abstractions" Version="6.0.0" />
    <PackageVersion Include="Microsoft.Extensions.Logging.Console" Version="10.0.0" />
    <PackageVersion Include="Microsoft.Extensions.Hosting" Version="10.0.0" />
    <PackageVersion Include="Microsoft.Extensions.TimeProvider.Testing" Version="8.3.0" />
    <PackageVersion Include="Microsoft.FeatureManagement" Version="3.2.0" />
    <PackageVersion Include="Microsoft.IO.RecyclableMemoryStream" Version="3.0.1" />
    <PackageVersion Include="Microsoft.NET.Test.Sdk" Version="18.0.1" />
    <PackageVersion Include="Microsoft.VisualStudio.Threading.Analyzers" Version="17.14.15" />
    <!-- Neovolve ships xunit v3 support under a distinct package id, not a major bump. -->
    <PackageVersion Include="Neovolve.Logging.Xunit.v3" Version="7.2.0" />
    <PackageVersion Include="NetTopologySuite" Version="2.5.0" />
    <PackageVersion Include="NetTopologySuite.IO.GeoJSON" Version="4.0.0" />
    <PackageVersion Include="Newtonsoft.Json" Version="13.0.3" />
    <PackageVersion Include="NodaTime.Serialization.JsonNet" Version="3.1.0" />
    <PackageVersion Include="NodaTime.Serialization.SystemTextJson" Version="1.1.2" />
    <PackageVersion Include="Npgsql" Version="9.0.4" />
    <PackageVersion Include="Npgsql.DependencyInjection" Version="9.0.2" />
    <PackageVersion Include="Npgsql.Json.NET" Version="9.0.4" />
    <PackageVersion Include="Npgsql.NetTopologySuite" Version="9.0.4" />
    <PackageVersion Include="Npgsql.NodaTime" Version="9.0.2" />
    <PackageVersion Include="NSubstitute" Version="5.3.0" />
    <PackageVersion Include="Nuke.Common" Version="10.1.0" />
    <PackageVersion Include="OpenTelemetry" Version="1.8.1" />
    <PackageVersion Include="OpenTelemetry.Exporter.OpenTelemetryProtocol" Version="1.8.1" />
    <PackageVersion Include="OpenTelemetry.Exporter.Prometheus.AspNetCore" Version="1.8.0-rc.1" />
    <PackageVersion Include="OpenTelemetry.Extensions.Hosting" Version="1.8.1" />
    <PackageVersion Include="OpenTelemetry.Instrumentation.AspNetCore" Version="1.8.1" />
    <PackageVersion Include="OpenTelemetry.Instrumentation.Http" Version="1.8.1" />
    <PackageVersion Include="OpenTelemetry.Instrumentation.Runtime" Version="1.8.0" />
    <PackageVersion Include="Pgvector" Version="0.3.2" />
    <PackageVersion Include="Shouldly" Version="4.3.0" />
    <PackageVersion Include="Spectre.Console" Version="0.53.0" />
    <PackageVersion Include="StronglyTypedId" Version="1.0.0-beta08" />
    <PackageVersion Include="Swashbuckle.AspNetCore" Version="6.5.0" />
    <PackageVersion Include="System.IO.Hashing" Version="10.0.3" />
    <PackageVersion Include="Vogen" Version="7.0.0" />
    <!-- 9.16.2: the hold at 9.2.1 (were newer builds actually being published?) is resolved —
         Weasel.EntityFrameworkCore is on nuget.org, so it tracks the rest of the Weasel line. -->
    <PackageVersion Include="Weasel.EntityFrameworkCore" Version="9.25.1" />
    <!-- Weasel.Postgresql 9.1.x: managed LIST partitions under multi-database (sharded) apply.
         9.1.2 (marten#4706): managed-partition tables are no longer destructively rebuilt — the
         out-of-band AddPartitionToAllTables path creates partitions while the generic schema diff
         leaves managed tables alone (Marten sets IgnorePartitionsInMigration on them; see DocumentTable).
         9.1.5 (marten#4713): the additive reconcile now SAVES/RESTORES IgnorePartitionsInMigration
         instead of permanently clearing it on the shared DocumentTable singleton, so a re-apply after
         tenant provisioning stays idempotent (no 42P07 / no rebuild). -->
    <!-- 9.3.0 (marten#4779 / weasel#318): RANGE partition bound round-trip fix so date-keyed (and
         quoted/integer) partitions no longer drift into a destructive rebuild on every migration. -->
    <!-- 9.4.0 (marten#4811 / weasel#321 + #322): lifts the closed-shape IIdentification identity
         family into Weasel.Core.Identity (with the ISequenceSource sequence seam) so Marten and
         Polecat share it; Weasel.Core 9.4.0 flows in transitively. -->
    <!-- 9.5.0 (weasel#323): adds a shared Weasel.Core.ISerializer base for the Critter Stack
         dedupe. Weasel.Core 9.5.0 flows in transitively. -->
    <!-- 9.6.0 (weasel#324): adds DbParameter[]-returning AppendWithDbParameters to the non-generic
         ICommandBuilder, unblocking the closed-shape operation param path (#4821/#4828). Weasel.Core
         9.6.0 flows in transitively. -->
    <!-- 9.7.0 (weasel#327): shared Weasel.Core SQL-gen contract (neutral ICommandBuilder +
         ISqlFragment) that the Postgres/SqlServer ones derive from. Marten disambiguates the new
         simple-name collision via a GlobalUsings alias to the Weasel.Postgresql types (API-preserving). -->
    <!-- 9.8.0 (marten#4830 / weasel#329): new Weasel.Storage package holds the 5 neutral closed-shape
         storage contracts (IStorageDialect, StorageColumnType, BulkColumnValue, IStorageSerializer,
         IVersionTracker) lifted out of Marten so Marten and Polecat share them. -->
    <!-- 9.14.0 (weasel#339): dialect-neutral grouped-parameter seam (Weasel.Core.IGroupedParameterBuilder
         + ICommandBuilder.CreateGroupedParameterBuilder/AppendParameter(object)) + StorageColumnType.Binary,
         so the closed-shape EVENT binders + per-event ops drop their direct Npgsql reference (#4821 event E1). -->
    <!-- 9.15.0 (weasel#341): closed-shape EVENT descriptor core moved into Weasel.Storage — the 3
         descriptors, IEventMetadataBinder + 5 metadata binders, IEventStoreSqlDialect — so Marten and
         Polecat share them (#4821 event E2). -->
    <!-- 9.16.0 (weasel#343): closed-shape EVENT storage hierarchy + operations + new neutral op
         bases moved into Weasel.Storage — EventStorage<TId>, the 3 storages, EventStorageBuilder,
         and the per-event/insert/update/assert ops — so Marten and Polecat share them (#4821 event E3). -->
    <!-- 9.16.1: latest Weasel patch consumed for the 9.13.0-alpha.3 Marten alpha. Also carries
         weasel#345/#346: PostgresqlDatabase.DisposeAsync() honors an OwnsDataSource flag so a
         caller-owned NpgsqlDataSource is not disposed on the async teardown path — the Weasel
         sibling of marten#4874 (the sync path was guarded Marten-side by #4903).
         9.16.2 (weasel#349/#350, marten#4874 case B): AdvisoryLock guards against a disposed
         NpgsqlDataSource during shutdown — a disposing flag short-circuits TryAttainLockAsync and
         disposed-pool ObjectDisposedException is swallowed to a non-acquire, so the HotCold cold-node
         leadership poll no longer aborts OpenAsync against a data source being disposed. Pairs with
         JasperFx 2.24.1 (jasperfx#499).
         9.16.3 (weasel#354, marten#4915): the ObjectDisposedException catch in AdvisoryLock.TryAttainLockAsync
         now latches the disposed flag AND rethrows instead of swallowing to a non-acquire. 9.16.2's swallow
         made jasperfx#500's terminate-on-ODE catch in ProjectionCoordinatorBase unreachable, so the HotCold
         cold-node leadership loop re-polled a dead pool forever; rethrowing lets the coordinator terminate.
         9.16.4 (weasel#356/#359): db-apply bounds its connection usage over many databases and extends the
         pool-release/retry to SqlServer, MySql and Oracle.
         9.17.0 (weasel#360): archive/tombstone/progression seams added to Weasel.Storage.Events. Additive for
         Marten — bumped to stay current on the Weasel line (compile-verified).
         9.20.1 (weasel#396/#397, marten#5090): AdvisoryLock can no longer strand a handle acquired after
         DisposeAsync drained. TryAttainLockAsync checked _disposed only at method entry and then stored the
         winning handle unconditionally, so an acquire in flight during disposal put its handle where nothing
         would ever dispose it — with the default transaction-scoped leadership lock that is a backend left
         'idle in transaction' on pg_try_advisory_xact_lock until the process exits, which the #4953 gap
         detection then reads as a live pre-gap reserver and never advances past. The store now happens under
         the same lock the drain latches under, a displaced handle (lock lost in monitored mode, re-attained)
         is disposed rather than overwritten, and HasLock/ReleaseLockAsync move under that lock too.
         9.20.2 (weasel#399/#400, found here): REQUIRED to move off 9.17.0. Weasel 9.18.0's computed-column
         delta detection (weasel#373) routed every column through TableColumn.MatchesForDelta, where a bare
         Equals(actual) binds to the protected NON-virtual Equals(TableColumn) overload and silently bypasses
         subclass overrides of Equals(object). That override is exactly how RevisionColumn declares an
         existing bigint mt_version acceptable for an integer-desired column instead of emitting a lossy
         narrowing cast (#4614/#4742), so on 9.18.0–9.20.1 the column landed in Columns.Different, the table
         was classified Update, and AssertDatabaseMatchesConfigurationAsync threw with an EMPTY change set —
         Bug_4614_revision_column_int_for_IRevisioned's assert-check test fails on every version in that
         range. 9.18/9.19/9.20 are otherwise additive for Marten and ride along (compile-verified).
         9.21.0 (weasel#401, marten#5093): ManagedRangePartitions — the RANGE analogue of
         ManagedListPartitions. A RollingWindowPolicy (period size, periods ahead, periods retained)
         attached through RangePartitioning.UsePartitionManager makes the partition set a pure function
         of the clock, so RangePartitioning.CreateDelta turns purely ADDITIVE instead of resolving a
         rolled-forward window to PartitionDelta.Rebuild. That is what lets Marten own a time-series
         document table's partitions end to end (PartitioningExpression.ByRollingRange) rather than
         pushing users onto ByExternallyManagedRangePartitions, which also opts them out of Weasel's
         ordering and dependency management (see JasperFx/CritterWatch#886).
         9.21.1 (weasel#416): partition bound VALUES are now escaped. FormatSqlValue doubles an embedded
         single quote, and no longer returns an already-quote-wrapped string verbatim — that short-circuit
         existed for idempotency but meant a quote-wrapped value skipped escaping entirely. Marten reaches
         this with a raw tenant id as the partition value on every provisioning path
         (ShardedTenancy.createPartitionsForTenant, AddMartenManagedTenantsAsync), and Marten's own
         DatabaseScopedTenantPartitions.Partitions() override only covered the schema-apply delta, not the
         runtime resolveBuckets path. REQUIRED alongside the tenant-id identifier escaping in this commit.
         9.22.0: carries the JasperFx 2.37.0 floor (weasel#417) so this matrix stays coherent rather
         than resolving transitively.
         9.23.0 (weasel#416): AssertValidIdentifier now rejects a quote or semicolon outright. This is
         the belt to 9.21.1's braces — 9.21.1 made partition bound VALUES escape correctly, and this
         closes the identifier side of the same surface.
         9.23.1 (weasel#420/#422): the same weasel#416 identifier validation extended to the other
         provider migrators (SQL Server previously validated nothing at all; Oracle, MySQL and Sqlite
         each missed their own quoting characters), with the PostgreSQL check folded onto the shared
         Weasel.Core helper so the providers cannot drift apart again.
         9.23.2 (weasel#424/#425): Sqlite and MySql implement the non-generic
         Weasel.Core.ICommandBuilder (weasel#423), and the JasperFx line moves to 2.38.0 so this
         matrix stays coherent with the JasperFx pins above rather than resolving transitively.
         9.25.0: "the identifier release" (see weasel docs/release-9-25.md). Most of it lands on the
         other providers — SQL Server bracketing, MySQL identifier normalization, the SQLite rebuild
         fix — but two things reach Marten. weasel#462 makes PostgreSQL quote for SHAPE as well as for
         case and keywords, and weasel#458 stops TableColumn silently rewriting a space in a column
         name to an underscore. Both are no-ops for conventional lowercase schemas, which is every
         name Marten generates itself; a user-supplied duplicated-field column containing a space
         would now be emitted quoted rather than folded. weasel#468 also widens identifier validation
         to every identifier a table writes rather than only the ones naming objects, which is the
         same surface as the 9.21.1/9.23.x hardening and matters most to the per-tenant partition
         names. Verified against the schema-heavy suites rather than assumed — see the bump commit.
         9.25.1 (weasel#485/#487, found here): REQUIRED to move off 9.24.0 — do not pin 9.25.0. That
         weasel#468 widening ran local identifiers through AssertValidIdentifier, which also enforces
         the 63-character limit, so a long primary key CONSTRAINT name was refused where PostgreSQL
         had always just truncated it. Weasel's own TableDelta compares those names through
         TruncatedNameIdentifier precisely so a truncated one matches, so the two halves disagreed.
         9.25.1 adds Migrator.AssertValidLocalIdentifier: same safety rules, no length limit.
         It cost two Marten failures on 9.25.0, and the second is the one to remember — the same
         rejection aborted a projection's schema application and surfaced as a 23505 on
         pk_mt_event_progression during per-tenant catch-up, an error naming a table with nothing to
         do with identifiers. A validation failure on this path does not only throw where you can
         see it.
         Requires JasperFx.Events >= 2.46.0; this matrix is already at 2.47.0. -->
    <PackageVersion Include="Weasel.Postgresql" Version="9.25.1" />
    <PackageVersion Include="Weasel.Storage" Version="9.25.1" />
    <PackageVersion Include="WolverineFx.Marten" Version="4.2.0" />
    <!-- The whole test suite is on xunit v3. VSTest stays the execution mode, so the
         Nuke targets and the GitHub workflows are unaffected. -->
    <PackageVersion Include="xunit.v3" Version="3.2.2" />
    <PackageVersion Include="xunit.runner.visualstudio" Version="3.1.5" />
  </ItemGroup>
  <!-- EF Core packages - requires net9.0+ due to Npgsql 9.x compatibility -->
  <ItemGroup Condition="'$(TargetFramework)' == 'net9.0'">
    <PackageVersion Include="Microsoft.EntityFrameworkCore" Version="9.0.2" />
    <PackageVersion Include="Microsoft.EntityFrameworkCore.Relational" Version="9.0.2" />
    <PackageVersion Include="Npgsql.EntityFrameworkCore.PostgreSQL" Version="9.0.3" />
  </ItemGroup>
  <ItemGroup Condition="'$(TargetFramework)' == 'net10.0'">
    <PackageVersion Include="Microsoft.EntityFrameworkCore" Version="10.0.0" />
    <PackageVersion Include="Microsoft.EntityFrameworkCore.Relational" Version="10.0.0" />
    <PackageVersion Include="Npgsql.EntityFrameworkCore.PostgreSQL" Version="10.0.0" />
  </ItemGroup>
  <!-- Framework-specific package versions -->
  <ItemGroup Condition="'$(TargetFramework)' == 'net9.0'">
    <PackageVersion Include="Microsoft.Extensions.Diagnostics.HealthChecks" Version="9.0.2" />
    <PackageVersion Include="Microsoft.Extensions.Hosting.Abstractions" Version="9.0.0" />
  </ItemGroup>
  <ItemGroup Condition="'$(TargetFramework)' == 'net10.0'">
    <PackageVersion Include="Microsoft.Extensions.Diagnostics.HealthChecks" Version="10.0.0" />
    <PackageVersion Include="Microsoft.Extensions.Hosting.Abstractions" Version="10.0.0" />
  </ItemGroup>
</Project>