chore(deps): Bump the wolverine group with 4 updates#373
chore(deps): Bump the wolverine group with 4 updates#373dependabot[bot] wants to merge 1 commit into
Conversation
LabelsThe following labels could not be found: Please fix the above issues or remove invalid values from |
|
@dependabot rebase |
Bumps WolverineFx from 5.32.1 to 5.39.0 Bumps WolverineFx.EntityFrameworkCore from 5.32.1 to 5.39.0 Bumps WolverineFx.Kafka from 5.32.1 to 5.39.0 Bumps WolverineFx.Postgresql from 5.32.1 to 5.39.0 --- updated-dependencies: - dependency-name: WolverineFx dependency-version: 5.39.0 dependency-type: direct:production update-type: version-update:semver-minor dependency-group: wolverine - dependency-name: WolverineFx.EntityFrameworkCore dependency-version: 5.39.0 dependency-type: direct:production update-type: version-update:semver-minor dependency-group: wolverine - dependency-name: WolverineFx.Kafka dependency-version: 5.39.0 dependency-type: direct:production update-type: version-update:semver-minor dependency-group: wolverine - dependency-name: WolverineFx.Postgresql dependency-version: 5.39.0 dependency-type: direct:production update-type: version-update:semver-minor dependency-group: wolverine ... Signed-off-by: dependabot[bot] <support@github.com>
901c4df to
77aa57a
Compare
|
Looks like this PR has been edited by someone other than Dependabot. That means Dependabot can't rebase it - sorry! If you're happy for Dependabot to recreate it from scratch, overwriting any edits, you can request |
|
Fechado sem merge — combina dois bloqueios estruturais que vão ser resolvidos pelos PRs #378 e #379 (mergeando agora):
A próxima execução semanal do Dependabot (segunda 06h BRT) vai abrir PRs novos com:
Sem perda de informação: bot recria PRs com as mesmas versões na próxima rodada. |
|
This pull request was built based on a group rule. Closing it will not ignore any of these versions in future pull requests. To ignore these dependencies, configure ignore rules in dependabot.yml |
Updated WolverineFx from 5.32.1 to 5.39.0.
Release notes
Sourced from WolverineFx's releases.
5.39.0
What's Changed
Fault<T>on Terminal Handler Failure by @BlackChepo in Fault Events — Auto-PublishFault<T>on Terminal Handler Failure JasperFx/wolverine#2695New Contributors
Full Changelog: JasperFx/wolverine@V5.38.0...V5.39.0
5.38.0
What's Changed
Full Changelog: JasperFx/wolverine@V5.37.2...V5.38.0
5.37.2
What's Changed
Full Changelog: JasperFx/wolverine@V5.37.0...V5.37.2
5.37.0
What's Changed
Full Changelog: JasperFx/wolverine@v5.36.1...V5.37.0
5.36.1
Upgrade to this version please if you use the EF Core transactional middleware and are using a very recent version of Wolverine
What's Changed
Full Changelog: JasperFx/wolverine@V5.36.0...v5.36.1
5.36.0
What's Changed
Full Changelog: JasperFx/wolverine@V5.35.2...V5.36.0
5.35.2
What's Changed
A small bug-fix and dependency-refresh release.
Persistence fixes
OracleMessageStore.Initializeand the surrounding schema declared timestamp columns (health_check,started,posted, etc.) with a DEFAULT ofSYS_EXTRACT_UTC(SYSTIMESTAMP). That returns a TIMESTAMP without a time zone, and when implicitly cast into theTIMESTAMP WITH TIME ZONEcolumn Oracle stamps it with the session time zone — so for any session in UTC+N the just-persisted value's UTC equivalent was N hours in the past. That tripped theNodeAgentController.DoHealthChecksAsyncstaleness filter and surfaced as aNullReferenceExceptionduring the first leadership-election heartbeat (self!.AssignAgents([LeaderUri])on anullself). All 18 occurrences swapped toSYSTIMESTAMP AT TIME ZONE 'UTC', with quote-doubling for the DDL contexts that go through Weasel'sEXECUTE IMMEDIATE '...'wrapper.Reliability hardening
dbcontrol://andoraclecontrol://endpoints toBufferedInMemory(#2637) — the database-backed control transport carries inter-node leader-election and agent-reassignment commands. Marking either endpoint Durable would route every control envelope through the same store-backed inbox/outbox the durability agent itself owns (deadlock); marking it Inline contradicts the batched-poll semantics. BothDatabaseControlEndpointandOracleControlEndpointnow overridesupportsModeso any policy that tries to flip them offBufferedInMemoryeither silently skips (UseDurableInboxOnAllListeners,UseDurableOutboxOnAllSendingEndpoints, etc., already checkSupportsMode) or fails fast with a clearInvalidOperationException.Source-gen / startup
IWolverineTypeLoaderacross all known assemblies (#2632) by @devployment — since 5.34.0, hosts on the source-generated codegen path silently dropped handlers that lived in referenced assemblies becausetryDiscoverTypeLoaderFromAttributeonly inspectedOptions.ApplicationAssembly. First invocation of one of those handlers threwIndeterminateRoutesException. The runtime now walksApplicationAssemblyand every assembly inDiscovery.Assemblies, collects every[WolverineTypeManifest]loader it finds, and exposes their union via the new internalCompositeWolverineTypeLoader. Existing single-loader semantics are preserved when only one manifest is found.Other
Dependencies
Full Changelog: JasperFx/wolverine@V5.35.1...V5.35.2
5.35.1
What's Changed
A bug-fix and small-feature release covering Oracle-in-Balanced-mode startup, a duplicate-poller defect on RavenDb, an AWS SQS sharded-slot naming fix, and a new SagaDescriptor surface for CritterWatch.
Persistence fixes
Oracle in
DurabilityMode.Balancednow boots (#2622) —OracleMessageStore.Initializewas a no-op, so hosts using Oracle persistence in Balanced mode failed to start withArgumentOutOfRangeException("ControlEndpoint cannot be null for this usage"). The fix mirrorsMessageDatabase.Initializefor Oracle and stands up a parallelOracleControlTransport/OracleControlEndpoint/OracleControlSender/OracleControlListenerset under the neworaclecontrol://protocol — necessary because the sharedDatabaseControlSender/DatabaseControlListenerassume@-prefixed placeholders and Guid values that map directly onto aDbParameter, neither of which Oracle accepts (:-prefixed placeholders,RAW(16)id columns requiringbyte[]). Also fixes a latentOracleMessageStore.EnqueueAsyncno-op that leftLogRecordsAsyncsilently dropping node records, and makesFetchRecentRecordsAsyncDBNull-safe on the description column. Adds a newOracleTests.LeaderElectionproject so the leadership-compliance suite runs against Oracle (marked[Trait("Category","Flaky")]because the suite needs careful TM/DML lock sequencing between back-to-back runs).RavenDb: only one durability agent polls after host start (#2623, #2629) by @Bishbulb (root-cause investigation and original fix) and @jeremydmiller —
RavenDbMessageStore.StartScheduledJobseagerly built and started aRavenDbDurabilityAgentat boot in addition to the agent thatNodeAgentControlleralready builds and starts viaIAgentFamily/MessageStoreCollection. TwoRavenDbDurabilityAgentinstances then polled the same database concurrently, both believed they held the scheduled-job lock, raced to mark the same envelopesIncoming, and surfacedConcurrencyExceptionplus double-fired timeouts. Drops the eagerStartTimers()call; the cluster-managed agent is now the single owner of polling. The agent returned fromStartScheduledJobsis held byWolverineRuntime.DurableScheduledJobspurely for its disposal-timeStopAsync. Comes with a reflection-free regression test (addedRavenDbDurabilityAgent.IsPollingandCompositeAgent.InnerAgentsfor the test to enumerate without poking at private fields). A companion regression-guard test for CosmosDb is included; investigation showed the equivalent CosmosDb path does NOT have the bug today (CosmosDbMessageStore.BuildAgentFamilyreturns null andUriuses thecosmosdb://scheme rather thanwolverinedb://, soMessageStoreCollectionnever registers a competing agent).Other improvements
AWS SQS: correct naming for sharded slot endpoints in #1f294ce5 — fixes a regression where sharded slot endpoint URIs were assembled with the wrong segment ordering, causing the consumer side to listen at the wrong queue.
SagaDescriptorexported viaServiceCapabilitiesin #f0998c7b — adds a saga-shape descriptor (saga type, id type, current state) to the capabilities exporter so CritterWatch can surface saga inventory and current state in its dashboard.Full Changelog: JasperFx/wolverine@V5.34.0...V5.35.1
5.34.0
What's Changed
This release introduces three new features (Claim Check pattern, declarative Polecat data requirements, and a new opt-in
WolverineFx.RuntimeCompilationpackage as the first step toward AOT compatibility), a fix for a long-standing service-location footgun, several real RavenDB and EF Core durability fixes, plus the cold-start prep work done as part of issue #1577.Highlights
[Blob]attribute (#2412) by @jeremydmiller in #2617 — off-load large message-property payloads to external storage on send and re-hydrate them on receive, so the on-the-wire envelope stays small. Core abstractions (IClaimCheckStore,ClaimCheckToken,[Blob],FileSystemClaimCheckStore,opts.UseClaimCheck(...)) live inWolverine.Persistence. Two new backend NuGet packages ship alongside:WolverineFx.ClaimCheck.AzureBlobStorageandWolverineFx.ClaimCheck.AmazonS3. New documentation page under/guide/durability/claim-checks.[DocumentExists<T>],[DocumentDoesNotExist<T>], andPolecatOps.Document<T>().MustExist(...)/.MustNotExist(...)mirror the Marten ergonomics, with the same batch-query optimization that folds multiple existence checks into a single PolecatIBatchedQueryround-trip.IMessageBus/IMessageContextnow see the active context (#2583) by @jeremydmiller in #2616 — when user code service-locatesIMessageBus(e.g., constructor injection on a service the handler resolves at runtime), it now receives the sameMessageContextthe handler itself uses, preserving outbox semantics. Implemented as a per-chain opt-in (gated onChain.UsesServiceLocation) so chains that don't service-locate pay zeroAsyncLocaloverhead per message.WolverineFx.RuntimeCompilationopt-in package (#1577) by @jeremydmiller in #2613 and #cc00ca0a — first pass at the cold-start / AOT roadmap. Wires theWolverine.SourceGenerationanalyzer into theWolverineFxNuGet output (so source-generatedIWolverineTypeLoaderdiscovery flows transitively to consumers), pre-populates theWolverineMessageNamingcache during startup to eliminate first-message reflection cost, applies[DynamicallyAccessedMembers]annotations on theActivator.CreateInstancecall sites for trimmer-friendliness, and addsConfigureAwait(false)across 117 awaits on the per-message hot path. The newWolverineFx.RuntimeCompilationpackage becomes the future-facing opt-in API for runtime Roslyn compilation; default behavior is preserved.RavenDB durability fixes
EF Core / ancillary stores
DurableLocalQueueto route incoming envelopes to ancillary stores (#2611) by @jeremydmiller — handler chains targeting an ancillary message store now have their incoming envelopes persisted in that store rather than the main store, fixing a class of "envelope stuck as Incoming forever" bugs.WolverineEnabledannotation collision in ancillary-store EF Core scenarios as part of #2618 —MapWolverineEnvelopeStorageis now idempotent when called against the same model graph more than once.Other improvements
MessageBusclobbering per-messageEnvelope.Sourcein #26e38002 —MessageBus.Sendwas overwriting the per-messageSourcewith the host'sServiceName, masking the originating service in causation tracking.ServiceLocationPolicy.NotAllowedv6 default (#2584) by @jeremydmiller in #2609 — chains that resolve dependencies via service location now log a warning at codegen time, with a clear migration path to the new opt-inWolverine.RuntimeCompilationstory for v6.CI stabilization
IDisposable→IAsyncLifetimeconversions, lifecycle cleanup; chronically broken-on-CI classes (SqliteTests.Transport.multi_tenancy_with_multiple_files,CosmosDbTests.LeaderElection.leader_election,Wolverine.RabbitMQ.Tests.send_by_topics,Wolverine.RabbitMQ.Tests.sending_raw_messages) tagged[Trait("Category", "Flaky")]with header comments documenting the symptom and the real follow-up fix.JasperFx.CodeGenerationandJasperFx.RuntimeCompilernow compile cleanly after theInitializeSynchronouslyextension method was moved intoJasperFx.CodeGenerationupstream.Issue-tracker housekeeping
#2507(Quartz.Net / TickerQ first-class integration) added to the 6.0 milestone with a phased plan documented on the issue.#1577(cold-start optimization roadmap) updated with a six-phase plan and current-state checklist; phases 1–4 are non-breaking on v5.x and largely landed in this release.Full Changelog: JasperFx/wolverine@V5.33.0...V5.34.0
5.33.0
What's Changed
This release includes a major reliability fix for distributed leader election, a port of the Polecat 2.x event store integration, and several other quality-of-life improvements.
Highlights
LeadershipLostnode record type and anIWolverineObserver.LostLeadership()hook for monitoring.JSONtype.Other Improvements
launchSettings.jsonto sample projects by @jeremydmiller in #2600Full Changelog: JasperFx/wolverine@V5.32.1...V5.33.0
What's Changed
New Contributors
Full Changelog: JasperFx/wolverine@V5.32.1...V5.33.0
Commits viewable in compare view.
Updated WolverineFx.EntityFrameworkCore from 5.32.1 to 5.39.0.
Release notes
Sourced from WolverineFx.EntityFrameworkCore's releases.
5.39.0
What's Changed
Fault<T>on Terminal Handler Failure by @BlackChepo in Fault Events — Auto-PublishFault<T>on Terminal Handler Failure JasperFx/wolverine#2695New Contributors
Full Changelog: JasperFx/wolverine@V5.38.0...V5.39.0
5.38.0
What's Changed
Full Changelog: JasperFx/wolverine@V5.37.2...V5.38.0
5.37.2
What's Changed
Full Changelog: JasperFx/wolverine@V5.37.0...V5.37.2
5.37.0
What's Changed
Full Changelog: JasperFx/wolverine@v5.36.1...V5.37.0
5.36.1
Upgrade to this version please if you use the EF Core transactional middleware and are using a very recent version of Wolverine
What's Changed
Full Changelog: JasperFx/wolverine@V5.36.0...v5.36.1
5.36.0
What's Changed
Full Changelog: JasperFx/wolverine@V5.35.2...V5.36.0
5.35.2
What's Changed
A small bug-fix and dependency-refresh release.
Persistence fixes
OracleMessageStore.Initializeand the surrounding schema declared timestamp columns (health_check,started,posted, etc.) with a DEFAULT ofSYS_EXTRACT_UTC(SYSTIMESTAMP). That returns a TIMESTAMP without a time zone, and when implicitly cast into theTIMESTAMP WITH TIME ZONEcolumn Oracle stamps it with the session time zone — so for any session in UTC+N the just-persisted value's UTC equivalent was N hours in the past. That tripped theNodeAgentController.DoHealthChecksAsyncstaleness filter and surfaced as aNullReferenceExceptionduring the first leadership-election heartbeat (self!.AssignAgents([LeaderUri])on anullself). All 18 occurrences swapped toSYSTIMESTAMP AT TIME ZONE 'UTC', with quote-doubling for the DDL contexts that go through Weasel'sEXECUTE IMMEDIATE '...'wrapper.Reliability hardening
dbcontrol://andoraclecontrol://endpoints toBufferedInMemory(#2637) — the database-backed control transport carries inter-node leader-election and agent-reassignment commands. Marking either endpoint Durable would route every control envelope through the same store-backed inbox/outbox the durability agent itself owns (deadlock); marking it Inline contradicts the batched-poll semantics. BothDatabaseControlEndpointandOracleControlEndpointnow overridesupportsModeso any policy that tries to flip them offBufferedInMemoryeither silently skips (UseDurableInboxOnAllListeners,UseDurableOutboxOnAllSendingEndpoints, etc., already checkSupportsMode) or fails fast with a clearInvalidOperationException.Source-gen / startup
IWolverineTypeLoaderacross all known assemblies (#2632) by @devployment — since 5.34.0, hosts on the source-generated codegen path silently dropped handlers that lived in referenced assemblies becausetryDiscoverTypeLoaderFromAttributeonly inspectedOptions.ApplicationAssembly. First invocation of one of those handlers threwIndeterminateRoutesException. The runtime now walksApplicationAssemblyand every assembly inDiscovery.Assemblies, collects every[WolverineTypeManifest]loader it finds, and exposes their union via the new internalCompositeWolverineTypeLoader. Existing single-loader semantics are preserved when only one manifest is found.Other
Dependencies
Full Changelog: JasperFx/wolverine@V5.35.1...V5.35.2
5.35.1
What's Changed
A bug-fix and small-feature release covering Oracle-in-Balanced-mode startup, a duplicate-poller defect on RavenDb, an AWS SQS sharded-slot naming fix, and a new SagaDescriptor surface for CritterWatch.
Persistence fixes
Oracle in
DurabilityMode.Balancednow boots (#2622) —OracleMessageStore.Initializewas a no-op, so hosts using Oracle persistence in Balanced mode failed to start withArgumentOutOfRangeException("ControlEndpoint cannot be null for this usage"). The fix mirrorsMessageDatabase.Initializefor Oracle and stands up a parallelOracleControlTransport/OracleControlEndpoint/OracleControlSender/OracleControlListenerset under the neworaclecontrol://protocol — necessary because the sharedDatabaseControlSender/DatabaseControlListenerassume@-prefixed placeholders and Guid values that map directly onto aDbParameter, neither of which Oracle accepts (:-prefixed placeholders,RAW(16)id columns requiringbyte[]). Also fixes a latentOracleMessageStore.EnqueueAsyncno-op that leftLogRecordsAsyncsilently dropping node records, and makesFetchRecentRecordsAsyncDBNull-safe on the description column. Adds a newOracleTests.LeaderElectionproject so the leadership-compliance suite runs against Oracle (marked[Trait("Category","Flaky")]because the suite needs careful TM/DML lock sequencing between back-to-back runs).RavenDb: only one durability agent polls after host start (#2623, #2629) by @Bishbulb (root-cause investigation and original fix) and @jeremydmiller —
RavenDbMessageStore.StartScheduledJobseagerly built and started aRavenDbDurabilityAgentat boot in addition to the agent thatNodeAgentControlleralready builds and starts viaIAgentFamily/MessageStoreCollection. TwoRavenDbDurabilityAgentinstances then polled the same database concurrently, both believed they held the scheduled-job lock, raced to mark the same envelopesIncoming, and surfacedConcurrencyExceptionplus double-fired timeouts. Drops the eagerStartTimers()call; the cluster-managed agent is now the single owner of polling. The agent returned fromStartScheduledJobsis held byWolverineRuntime.DurableScheduledJobspurely for its disposal-timeStopAsync. Comes with a reflection-free regression test (addedRavenDbDurabilityAgent.IsPollingandCompositeAgent.InnerAgentsfor the test to enumerate without poking at private fields). A companion regression-guard test for CosmosDb is included; investigation showed the equivalent CosmosDb path does NOT have the bug today (CosmosDbMessageStore.BuildAgentFamilyreturns null andUriuses thecosmosdb://scheme rather thanwolverinedb://, soMessageStoreCollectionnever registers a competing agent).Other improvements
AWS SQS: correct naming for sharded slot endpoints in #1f294ce5 — fixes a regression where sharded slot endpoint URIs were assembled with the wrong segment ordering, causing the consumer side to listen at the wrong queue.
SagaDescriptorexported viaServiceCapabilitiesin #f0998c7b — adds a saga-shape descriptor (saga type, id type, current state) to the capabilities exporter so CritterWatch can surface saga inventory and current state in its dashboard.Full Changelog: JasperFx/wolverine@V5.34.0...V5.35.1
5.34.0
What's Changed
This release introduces three new features (Claim Check pattern, declarative Polecat data requirements, and a new opt-in
WolverineFx.RuntimeCompilationpackage as the first step toward AOT compatibility), a fix for a long-standing service-location footgun, several real RavenDB and EF Core durability fixes, plus the cold-start prep work done as part of issue #1577.Highlights
[Blob]attribute (#2412) by @jeremydmiller in #2617 — off-load large message-property payloads to external storage on send and re-hydrate them on receive, so the on-the-wire envelope stays small. Core abstractions (IClaimCheckStore,ClaimCheckToken,[Blob],FileSystemClaimCheckStore,opts.UseClaimCheck(...)) live inWolverine.Persistence. Two new backend NuGet packages ship alongside:WolverineFx.ClaimCheck.AzureBlobStorageandWolverineFx.ClaimCheck.AmazonS3. New documentation page under/guide/durability/claim-checks.[DocumentExists<T>],[DocumentDoesNotExist<T>], andPolecatOps.Document<T>().MustExist(...)/.MustNotExist(...)mirror the Marten ergonomics, with the same batch-query optimization that folds multiple existence checks into a single PolecatIBatchedQueryround-trip.IMessageBus/IMessageContextnow see the active context (#2583) by @jeremydmiller in #2616 — when user code service-locatesIMessageBus(e.g., constructor injection on a service the handler resolves at runtime), it now receives the sameMessageContextthe handler itself uses, preserving outbox semantics. Implemented as a per-chain opt-in (gated onChain.UsesServiceLocation) so chains that don't service-locate pay zeroAsyncLocaloverhead per message.WolverineFx.RuntimeCompilationopt-in package (#1577) by @jeremydmiller in #2613 and #cc00ca0a — first pass at the cold-start / AOT roadmap. Wires theWolverine.SourceGenerationanalyzer into theWolverineFxNuGet output (so source-generatedIWolverineTypeLoaderdiscovery flows transitively to consumers), pre-populates theWolverineMessageNamingcache during startup to eliminate first-message reflection cost, applies[DynamicallyAccessedMembers]annotations on theActivator.CreateInstancecall sites for trimmer-friendliness, and addsConfigureAwait(false)across 117 awaits on the per-message hot path. The newWolverineFx.RuntimeCompilationpackage becomes the future-facing opt-in API for runtime Roslyn compilation; default behavior is preserved.RavenDB durability fixes
EF Core / ancillary stores
DurableLocalQueueto route incoming envelopes to ancillary stores (#2611) by @jeremydmiller — handler chains targeting an ancillary message store now have their incoming envelopes persisted in that store rather than the main store, fixing a class of "envelope stuck as Incoming forever" bugs.WolverineEnabledannotation collision in ancillary-store EF Core scenarios as part of #2618 —MapWolverineEnvelopeStorageis now idempotent when called against the same model graph more than once.Other improvements
MessageBusclobbering per-messageEnvelope.Sourcein #26e38002 —MessageBus.Sendwas overwriting the per-messageSourcewith the host'sServiceName, masking the originating service in causation tracking.ServiceLocationPolicy.NotAllowedv6 default (#2584) by @jeremydmiller in #2609 — chains that resolve dependencies via service location now log a warning at codegen time, with a clear migration path to the new opt-inWolverine.RuntimeCompilationstory for v6.CI stabilization
IDisposable→IAsyncLifetimeconversions, lifecycle cleanup; chronically broken-on-CI classes (SqliteTests.Transport.multi_tenancy_with_multiple_files,CosmosDbTests.LeaderElection.leader_election,Wolverine.RabbitMQ.Tests.send_by_topics,Wolverine.RabbitMQ.Tests.sending_raw_messages) tagged[Trait("Category", "Flaky")]with header comments documenting the symptom and the real follow-up fix.JasperFx.CodeGenerationandJasperFx.RuntimeCompilernow compile cleanly after theInitializeSynchronouslyextension method was moved intoJasperFx.CodeGenerationupstream.Issue-tracker housekeeping
#2507(Quartz.Net / TickerQ first-class integration) added to the 6.0 milestone with a phased plan documented on the issue.#1577(cold-start optimization roadmap) updated with a six-phase plan and current-state checklist; phases 1–4 are non-breaking on v5.x and largely landed in this release.Full Changelog: JasperFx/wolverine@V5.33.0...V5.34.0
5.33.0
What's Changed
This release includes a major reliability fix for distributed leader election, a port of the Polecat 2.x event store integration, and several other quality-of-life improvements.
Highlights
LeadershipLostnode record type and anIWolverineObserver.LostLeadership()hook for monitoring.JSONtype.Other Improvements
launchSettings.jsonto sample projects by @jeremydmiller in #2600Full Changelog: JasperFx/wolverine@V5.32.1...V5.33.0
What's Changed
New Contributors
Full Changelog: JasperFx/wolverine@V5.32.1...V5.33.0
Commits viewable in compare view.
Updated WolverineFx.Kafka from 5.32.1 to 5.39.0.
Release notes
Sourced from WolverineFx.Kafka's releases.
5.39.0
What's Changed
Fault<T>on Terminal Handler Failure by @BlackChepo in Fault Events — Auto-PublishFault<T>on Terminal Handler Failure JasperFx/wolverine#2695New Contributors
Full Changelog: JasperFx/wolverine@V5.38.0...V5.39.0
5.38.0
What's Changed
Full Changelog: JasperFx/wolverine@V5.37.2...V5.38.0
5.37.2
What's Changed
Full Changelog: JasperFx/wolverine@V5.37.0...V5.37.2
5.37.0
What's Changed
Full Changelog: JasperFx/wolverine@v5.36.1...V5.37.0
5.36.1
Upgrade to this version please if you use the EF Core transactional middleware and are using a very recent version of Wolverine
What's Changed
Full Changelog: JasperFx/wolverine@V5.36.0...v5.36.1
5.36.0
What's Changed
Full Changelog: JasperFx/wolverine@V5.35.2...V5.36.0
5.35.2
What's Changed
A small bug-fix and dependency-refresh release.
Persistence fixes
OracleMessageStore.Initializeand the surrounding schema declared timestamp columns (health_check,started,posted, etc.) with a DEFAULT ofSYS_EXTRACT_UTC(SYSTIMESTAMP). That returns a TIMESTAMP without a time zone, and when implicitly cast into theTIMESTAMP WITH TIME ZONEcolumn Oracle stamps it with the session time zone — so for any session in UTC+N the just-persisted value's UTC equivalent was N hours in the past. That tripped theNodeAgentController.DoHealthChecksAsyncstaleness filter and surfaced as aNullReferenceExceptionduring the first leadership-election heartbeat (self!.AssignAgents([LeaderUri])on anullself). All 18 occurrences swapped toSYSTIMESTAMP AT TIME ZONE 'UTC', with quote-doubling for the DDL contexts that go through Weasel'sEXECUTE IMMEDIATE '...'wrapper.Reliability hardening
dbcontrol://andoraclecontrol://endpoints toBufferedInMemory(#2637) — the database-backed control transport carries inter-node leader-election and agent-reassignment commands. Marking either endpoint Durable would route every control envelope through the same store-backed inbox/outbox the durability agent itself owns (deadlock); marking it Inline contradicts the batched-poll semantics. BothDatabaseControlEndpointandOracleControlEndpointnow overridesupportsModeso any policy that tries to flip them offBufferedInMemoryeither silently skips (UseDurableInboxOnAllListeners,UseDurableOutboxOnAllSendingEndpoints, etc., already checkSupportsMode) or fails fast with a clearInvalidOperationException.Source-gen / startup
IWolverineTypeLoaderacross all known assemblies (#2632) by @devployment — since 5.34.0, hosts on the source-generated codegen path silently dropped handlers that lived in referenced assemblies becausetryDiscoverTypeLoaderFromAttributeonly inspectedOptions.ApplicationAssembly. First invocation of one of those handlers threwIndeterminateRoutesException. The runtime now walksApplicationAssemblyand every assembly inDiscovery.Assemblies, collects every[WolverineTypeManifest]loader it finds, and exposes their union via the new internalCompositeWolverineTypeLoader. Existing single-loader semantics are preserved when only one manifest is found.Other
Dependencies
Full Changelog: JasperFx/wolverine@V5.35.1...V5.35.2
5.35.1
What's Changed
A bug-fix and small-feature release covering Oracle-in-Balanced-mode startup, a duplicate-poller defect on RavenDb, an AWS SQS sharded-slot naming fix, and a new SagaDescriptor surface for CritterWatch.
Persistence fixes
Oracle in
DurabilityMode.Balancednow boots (#2622) —OracleMessageStore.Initializewas a no-op, so hosts using Oracle persistence in Balanced mode failed to start withArgumentOutOfRangeException("ControlEndpoint cannot be null for this usage"). The fix mirrorsMessageDatabase.Initializefor Oracle and stands up a parallelOracleControlTransport/OracleControlEndpoint/OracleControlSender/OracleControlListenerset under the neworaclecontrol://protocol — necessary because the sharedDatabaseControlSender/DatabaseControlListenerassume@-prefixed placeholders and Guid values that map directly onto aDbParameter, neither of which Oracle accepts (:-prefixed placeholders,RAW(16)id columns requiringbyte[]). Also fixes a latentOracleMessageStore.EnqueueAsyncno-op that leftLogRecordsAsyncsilently dropping node records, and makesFetchRecentRecordsAsyncDBNull-safe on the description column. Adds a newOracleTests.LeaderElectionproject so the leadership-compliance suite runs against Oracle (marked[Trait("Category","Flaky")]because the suite needs careful TM/DML lock sequencing between back-to-back runs).RavenDb: only one durability agent polls after host start (#2623, #2629) by @Bishbulb (root-cause investigation and original fix) and @jeremydmiller —
RavenDbMessageStore.StartScheduledJobseagerly built and started aRavenDbDurabilityAgentat boot in addition to the agent thatNodeAgentControlleralready builds and starts viaIAgentFamily/MessageStoreCollection. TwoRavenDbDurabilityAgentinstances then polled the same database concurrently, both believed they held the scheduled-job lock, raced to mark the same envelopesIncoming, and surfacedConcurrencyExceptionplus double-fired timeouts. Drops the eagerStartTimers()call; the cluster-managed agent is now the single owner of polling. The agent returned fromStartScheduledJobsis held byWolverineRuntime.DurableScheduledJobspurely for its disposal-timeStopAsync. Comes with a reflection-free regression test (addedRavenDbDurabilityAgent.IsPollingandCompositeAgent.InnerAgentsfor the test to enumerate without poking at private fields). A companion regression-guard test for CosmosDb is included; investigation showed the equivalent CosmosDb path does NOT have the bug today (CosmosDbMessageStore.BuildAgentFamilyreturns null andUriuses thecosmosdb://scheme rather thanwolverinedb://, soMessageStoreCollectionnever registers a competing agent).Other improvements
AWS SQS: correct naming for sharded slot endpoints in #1f294ce5 — fixes a regression where sharded slot endpoint URIs were assembled with the wrong segment ordering, causing the consumer side to listen at the wrong queue.
SagaDescriptorexported viaServiceCapabilitiesin #f0998c7b — adds a saga-shape descriptor (saga type, id type, current state) to the capabilities exporter so CritterWatch can surface saga inventory and current state in its dashboard.Full Changelog: JasperFx/wolverine@V5.34.0...V5.35.1
5.34.0
What's Changed
This release introduces three new features (Claim Check pattern, declarative Polecat data requirements, and a new opt-in
WolverineFx.RuntimeCompilationpackage as the first step toward AOT compatibility), a fix for a long-standing service-location footgun, several real RavenDB and EF Core durability fixes, plus the cold-start prep work done as part of issue #1577.Highlights
[Blob]attribute (#2412) by @jeremydmiller in #2617 — off-load large message-property payloads to external storage on send and re-hydrate them on receive, so the on-the-wire envelope stays small. Core abstractions (IClaimCheckStore,ClaimCheckToken,[Blob],FileSystemClaimCheckStore,opts.UseClaimCheck(...)) live inWolverine.Persistence. Two new backend NuGet packages ship alongside:WolverineFx.ClaimCheck.AzureBlobStorageandWolverineFx.ClaimCheck.AmazonS3. New documentation page under/guide/durability/claim-checks.[DocumentExists<T>],[DocumentDoesNotExist<T>], andPolecatOps.Document<T>().MustExist(...)/.MustNotExist(...)mirror the Marten ergonomics, with the same batch-query optimization that folds multiple existence checks into a single PolecatIBatchedQueryround-trip.IMessageBus/IMessageContextnow see the active context (#2583) by @jeremydmiller in #2616 — when user code service-locatesIMessageBus(e.g., constructor injection on a service the handler resolves at runtime), it now receives the sameMessageContextthe handler itself uses, preserving outbox semantics. Implemented as a per-chain opt-in (gated onChain.UsesServiceLocation) so chains that don't service-locate pay zeroAsyncLocaloverhead per message.Description has been truncated