Eve library

V2 orientation · remaining whole-unit review

Guide

Revision 1

On this page

Source pin: 1ceae52c7896a59ac74a4429d410f3218f0b91c6. This is a source-review continuation map, not another teaching assignment or a claim that the nine-slide promise guide is complete. It was reviewed while the eighteen contracts recordings were being synthesized; no existing slide or speech plan was changed during that run.

Architecture overview#

The five units in assignments/v2-contracts.json cover the conventions and their worked enforcement example. The rest of V2/architecture/v2-product-promise.md remains accountable separately.

Unit Existing explanation Remaining whole-unit work
7c0503646e69930882dbf05d · introduction, lines 1–29 v2-promise-feel establishes one UE5.5 game; contracts evidence-path, reference-language and target-artifacts teach ownership and navigation. Assemble the complete cross-guide evidence map, including 27 modules, the exact 138-section invariant and strict-FP/cook policy. Citations selected only for descendant headings cannot evidence this introduction. Check every explanatory claim before assigning it.
e069b48bd32ae70d56d6032c · What ships, honestly, lines 30–86 The contracts guide qualifies target helpers, editor unity, reserved definitions, module/plugin kinds, static checking and stale descriptors. Promise slides qualify rollback targets and provider availability. The approximate per-module file inventories (~218 Gameplay, ~36 Combat, ~20 Netcode, ~14 Input), 196 KB workflow observation, planned 60-plus GameFeature roster and complete provider/rating-data references need explicit dated teaching and code review. Do not infer all-gate success or a one-to-one checker mapping from this paragraph.
cae0d4c12cfa127a985ead3b · The product promise, lines 87–110 All eleven high-level commitments have related material in the nine-slide promise guide, including the seven traditions, networking envelopes, career length, creator acronyms, platform scope, failure categories and release evidence. Review every commitment's second-column code/status claim, not only the promise wording. The current guide does not explicitly teach every named owner: V2_NetworkPredictionComponent, the Linux server gate, the V2Input ownership/data paths, and the apps/v2/* operations boundary need reconciliation. Do not mark this table covered from eleven related topics alone.
2c37b08bbfb3d90d55d60b92 · Netcode envelope, lines 111–136 v2-promise-netcode-tiers accurately preserves all three mode groups, 60 Hz / ≤8 frames / ≤100 ms targets and the need for deterministic replay. Its prose is less specific than the source diagram: both small and mid-size branches reach V2Netcode, while the dedicated branch names V2OnlineServices, ReplicationGraph and Iris. Reconcile those named paths with actual source before adding them. Preserve the branch conditions and the difference between a documented tier and an executed authoritative path.

The later surfaces, neighbouring-systems and related-reading units (lines 282–343) are not covered by this review. Their diagrams/tables and each distinct source context still need whole-unit comparison. The separate feature overview and documentation README have their own unit identities; matching a general topic is not automatic duplicate coverage.

A concrete replication follow-up#

The source diagram's dedicated branch needs more than copying three names. Pinned V2/ue/Source/V2OnlineServices/V2OnlineServices.Build.cs declares only Core, CoreUObject and Engine as public dependencies. The inspected config sets ReplicationDriverClassName=/Script/V2Netcode.V2ReplicationGraph in V2/ue/Config/DefaultEngine.ini. Those two observations do not establish a complete Iris integration or a dependency originating in V2OnlineServices. Follow the actual V2Netcode graph, engine replication configuration and consumers before drawing a module or deployment boundary. Do not conflate the diagram's Iris name with the platform's Iris memory/conversation domain from its name alone; the relevant engine-versus-platform ownership needs source evidence.

The subsequent header/source review makes the next test boundary concrete: V2Netcode.Build.cs declares ReplicationGraph publicly and NetCore privately. UV2ReplicationGraph inherits UReplicationGraph, but the inspected class declares static band/lookup/routing helpers, not custom engine-node registration or gather overrides. RouteActorForConnection selects OwnPawn first, then AmbientProp, then a fighter at distance ≤2500 cm, then DistantFighter. The 12000 cm distant and 8000 cm ambient distances are declared band data, not distance rejections in this helper. A negative fighter distance also satisfies the nearby comparison; no input-distance validation appears in that function. These are source-reading observations, not executed C++ tests or proof of live replication behavior. A future worked case should distinguish the returned routing description from its actual engine consumer and packet outcome.

The neighbouring channel declarations resolve the naming question further: FV2NetworkQualityAnalyzer::BuildDefaultReplicationChannels in V2NetworkQuality.cpp returns records with UE.Iris.* transport-profile names, while the rollback-input record uses V2.Rollback.CustomSocket. FV2ReplicationChannelSpec::IsValidChannel in V2NetcodeTypes.cpp validates those names and boolean properties, and the Blueprint library returns the same records. This identifies an Unreal-labeled replication contract rather than an Iris-memory-domain call. It still does not prove that declaring bUsesIris creates an engine channel or sends a packet. The inspected automation source checks these record properties and four routing examples; no automation was executed during this reading pass. Future verification must exercise the actual consumer and add boundary/precedence cases before teaching a stronger claim.

Next actions after the contracts delivery gate#

  1. Finish the contracts guide's recordings and final actual-audio, embedded, PDF and preservation checks. Do not edit its frozen narration inputs during synthesis or use intermediate reports as the final receipt.
  2. Review the remaining orientation units above against the complete source and existing guide notes. Add the missing explanation where it materially teaches the unit; avoid packing implementation inventories into an opening comparison merely to obtain a coverage flag.
  3. Follow the orientation's promised next learning task into V2/architecture/combat-system-gas-frame-data-and-determinism.md and V2/architecture/rollback-netcode-and-tag-team.md. Establish their complete source-unit plans and actual code boundaries before choosing guide counts, slides, diagrams or worked tests.
  4. Keep documentation README truth states, role-specific journeys, detailed module topology, build/cook and release gates separately accountable. The current contracts overview is not their exhaustive treatment.

Normal-speed listening, willing-reader comprehension, the broader visual pilot, Eve integration, all Domains and whole-monorepo coverage remain open.