Repository navigation
Serialization.V2: integration proof suite (remoting, persistence, Delivery, DData) - #8407
Closed
Aaronontheweb wants to merge 3 commits into
Closed
Aaronontheweb wants to merge 3 commits into
Aaronontheweb wants to merge 3 commits into
Conversation
…e, delivery, DData (tasks 6.4-6.10, 8.5-8.6) Implements the remaining openspec integration-validation tasks from messagepack-sourcegen-validation/tasks.md: source-generated V2 MessagePack serializers now have end-to-end proof through real Akka.NET machinery, not just direct Serialization.cs round-trips. - ClassicRemotingSpec (Akka.Serialization.V2.Tests): two independent ActorSystems exchange generated V2 payloads over classic akka.remote.dot-netty.tcp, HOCON-registered the same way ArteryControlMessageSerializer is registered in Remote.conf. Covers a scalar+IActorRef payload and a nested-value-object payload. (task 6.4) - PersistenceIntegrationSpec (Akka.Serialization.V2.Tests): events persist and recover through a small test-only WireFormatJournal that genuinely serializes bytes via Serialization.FindSerializerFor/ToBinary/FromBinary (the built-in in-memory journal holds live object references and would prove nothing about the wire format); snapshots save and load through the real, shipped akka.persistence.snapshot-store.local plugin. (tasks 6.5, 6.6) - ReliableDeliveryIntegrationSpec (Akka.Cluster.Tests): a generated V2 payload survives the real (internal) ReliableDeliverySerializer wrapper via ConsumerController.SequencedMessage<T>, plus a live ProducerController/ConsumerController exchange under serialize-messages=on. (task 6.9) - GeneratedV2WrapperSpec (Akka.DistributedData.Tests): a generated V2 payload survives ReplicatedDataSerializer's OtherMessage wrapper as both an LWWRegister<T> value and ORSet<T> elements. (task 6.10) Tasks 8.5/8.6 (focused Remote/Persistence generated-serializer test runs) are satisfied by the above specs. Placement: ReliableDeliverySerializer and ReplicatedDataSerializer are production wrappers that live in Akka.Cluster / Akka.DistributedData, so those two specs live in their existing test projects (with new test-only ProjectReferences to Akka.Serialization.V2) rather than pulling Akka.Cluster/Akka.DistributedData into Akka.Serialization.V2.Tests. Classic remoting and persistence specs live in Akka.Serialization.V2.Tests per the task brief, with new test-only ProjectReferences to Akka.Remote and Akka.Persistence. No product code (generator or V2 runtime) was touched. Verification: full suites pass with no regressions - Akka.Serialization.V2.Tests 110/110, Akka.Cluster.Tests 381/381, Akka.DistributedData.Tests 192/193 (1 pre-existing unrelated skip). All new remoting/persistence/delivery/DData specs were run multiple times with no flakiness observed. dotnet format --verify-no-changes and dotnet build -warnaserror both clean on all touched projects.
…net#8385+) The worktree's original base (84524b3) predated the Akka.Serialization.V2 surface reshape that landed via PR akkadotnet#8385 and successors. After the coordinator merged current dev into this branch, the four new integration specs from the prior commit no longer compiled against the new surface: - Attribute/base-class rename: `[AkkaSerializer(Name = ..., SerializerId = ...)]` + `MessagePackSerializer<TProtocol>` became the generic `[AkkaSerializer<TProtocol>("name", id)]` (positional, required args) + non-generic `AkkaSerializer` base class. `CreateRegistration()` partial method declarations were unaffected. Fixed in all four new files (ClassicRemotingSpec, PersistenceIntegrationSpec, ReliableDeliveryIntegrationSpec, GeneratedV2WrapperSpec). - Type collision: dev's GeneratedUnionSpec.cs (new since the base commit) declares a top-level `Akka.Serialization.V2.Tests.OrderPlaced` record with manifest "order-placed-v1". PersistenceIntegrationSpec.cs declared its own `OrderPlaced`/`OrderState`, causing CS0101/CS0579/CS8863 and a knock-on AKKASG012 + CS0534 on the unrelated UnionTestSerializer (the duplicate symbol confused the generator's per-serializer manifest scan). Renamed to `PersistedOrderPlaced`/`PersistedOrderState` with distinct manifests ("persisted-order-placed-v1"/"persisted-order-state-v1"); confirmed no other identifier collisions exist against the merged tree (checked every type declared across all four new spec files). Verification against the merged tree: - dotnet test src/core/Akka.Serialization.V2.Tests -c Release: 260/260 passing (was 106 pre-merge baseline + our 4; merged dev tip's own baseline is 256, confirming the orchestrator's original "256 tests" figure referred to current dev, not the stale base this worktree started from). - dotnet test src/core/Akka.Cluster.Tests --filter ReliableDeliveryIntegrationSpec: 2/2 passing. - dotnet test src/contrib/cluster/Akka.DistributedData.Tests --filter GeneratedV2WrapperSpec: 2/2 passing. - dotnet format --verify-no-changes: clean on all three touched projects. - dotnet build -warnaserror: clean on Akka.Cluster.Tests and Akka.DistributedData.Tests. Akka.Serialization.V2.Tests only fails -warnaserror when built as a standalone .csproj outside solution context, because the Verify package's build props can't auto-discover SolutionDir/ SolutionName from a project 3 directories below the repo root (its search only walks 2 parents) and downgrades that to a real warning under -warnaserror; this is pre-existing since dev's wire-format-snapshot Verify tests (86976dd), reproducible on an unmodified file in that project, and resolves cleanly once SolutionDir/SolutionName are supplied (confirmed with explicit /p:SolutionDir/SolutionName, matching how CI builds via the solution) -- not something introduced by or fixable in these test files. All new specs re-run twice post-fix with no flakiness. No product code (generator or V2 runtime) touched.
Aaronontheweb
force-pushed
the
feature/serialization-v2-integration-proofs
branch
from
July 18, 2026 12:37
655d1ed to
86a13cf
Compare
This was referenced Oct 2, 2026
Member
Author
Member
Author
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Serialization.V2 integration proof suite
Implements openspec tasks 6.4–6.6, 6.9–6.10, 8.5–8.6 of
messagepack-sourcegen-validation: proves source-generated V2 serializers work end-to-end through real Akka machinery, not just unit round-trips. Tests only — zero product code changed. Headline result: no product gaps found — every integration point worked against the generated serializers without a single workaround.Coverage
Akka.Serialization.V2.Tests/ClassicRemotingSpec.csAkka.Serialization.V2.Tests/PersistenceIntegrationSpec.csLocalSnapshotStorefile pluginReliableDeliverySerializerwrapper + liveProducerController/ConsumerControllerexchange underserialize-messages = onAkka.Cluster.Tests/ReliableDeliveryIntegrationSpec.csReplicatedDataSerializerwrapper interactionAkka.DistributedData.Tests/Serialization/GeneratedV2WrapperSpec.csNotable test-integrity decision
The built-in
akka.persistence.journal.inmem(MemoryJournal) stores live object references and never serializes — an event round-trip "proof" through it would be a no-op. The event test instead uses a small test-onlyWireFormatJournalthat does what real journal plugins do (FindSerializerFor→ToBinary/FromBinary, storing only bytes + manifest + serializer ID), documented inline. Snapshots use the real shippedLocalSnapshotStore, which already round-trips through actual serialization.Placement
The Delivery/DData wrapper specs live in
Akka.Cluster.Tests/Akka.DistributedData.Tests(those serializers are internal to those assemblies); the projects gained lightweight test-onlyProjectReferences toAkka.Serialization.V2— no shipping dependency changes anywhere.Verification
Akka.Serialization.V2.Tests: 260/260 (256 pre-existing + 4 new), stable across repeated runsAkka.Cluster.Tests: full suite 381/381;Akka.DistributedData.Tests: 192 passed / 1 pre-existing unrelated skipdotnet format+-warnaserrorclean on touched projectsmessagepack-sourcegen-validation/tasks.mdNote for reviewers: the branch history contains a mid-flight port commit — the specs were originally authored against a stale base that predated the #8385 attribute-surface reshape; they are fully ported to the current
[AkkaSerializer<TProtocol>(name, id)]/AkkaSerializersurface and validated against currentdev.