AMassBattleNetworkPlayerControllerBase

内部 RPC(非 BP)——所有上行锁步命令的统一入口。落地:取 GetPlayerIndex() 盖章 ValidatedCmd.PlayerIndex 后转发 GM->ReceivePlayerCommand 做 K-buffer 调度。BP 侧请用 SendLockstepCommandToServer 便捷版。

Internal RPC (not BP) — the unified entry for all uplink lockstep commands. On landing: stamps ValidatedCmd.PlayerIndex with GetPlayerIndex() and forwards to GM->ReceivePlayerCommand for K-buffer scheduling. BP should use the SendLockstepCommandToServer convenience version.

客户端→服务器锁步指令便捷版。自动分流:序列化负载 ≥ MassBattle.AutoFloodSwitchBytes(默认 16KB)时,负载打包压缩为 TCP 上传块(SendLockstepDataBlockToServer,BlockId = AllocLocalBlockId 的 LocalBlockId),命令只带 DataBlockId/DataBlockCrc/ExtraDeferTicks 经 Server_SendNetCommand 上行——服务器汇聚时把 LocalBlockId 重映射为 ServerBlockId 并回广播(发送方像任意客户端一样收到);小负载直接经现有可靠 Server RPC 上行。

Client→server lockstep command convenience. Auto-flood: when the serialized payload ≥ MassBattle.AutoFloodSwitchBytes (default 16KB), the payload is packed/compressed into a TCP upload block (SendLockstepDataBlockToServer, BlockId = AllocLocalBlockId's LocalBlockId) and the command carries only DataBlockId/DataBlockCrc/ExtraDeferTicks via Server_SendNetCommand — the server remaps LocalBlockId → ServerBlockId on convergence and rebroadcasts (the sender receives it like any client); small payloads go through the existing reliable Server RPC.

客户端主动报告失步。落地:盖章玩家索引后转发 GM->ReportClientDesync——服务器比对 hash,不一致则记录失步并隔离该客户端。

Client-reported desync. On landing: stamps the player index and forwards to GM->ReportClientDesync — the server compares hashes and records/quarantines on mismatch.

DedicatedServer:上报本地模拟 hash 用于投票。落地:盖章后转发 GM->ReportClientHash → relay 聚合。

DedicatedServer: reports the local simulation hash for voting. On landing: stamped and forwarded to GM->ReportClientHash → relay aggregation.

DedicatedServer:上传压缩快照用于灾备。落地:盖章后转发 GM->ReceiveSnapshotUpload(入口验证大小与 tick 窗口)。

DedicatedServer: uploads a compressed snapshot for disaster recovery. On landing: stamped and forwarded to GM->ReceiveSnapshotUpload (entry validation of size and tick window).

客户端心跳确认:上报已模拟到的最新 tick 及本地哈希;服务器据此做步调控制(pacing)、延迟追踪与 bPacedMode 超时踢人。落地:盖章后转发 GM->AckClientTick——推进 ack 追踪并在 hash 不一致时隔离。失步校验链路中每个 tick 都经此上报(见 EvaluateDeferredDesyncChecks)。

Client heartbeat ack: reports the latest tick simulated along with the local hash; the server uses acks for pacing, lag tracking and the bPacedMode timeout kick. On landing: stamped and forwarded to GM->AckClientTick — advances ack tracking and quarantines on hash mismatch. Every tick in the desync-validation chain reports through this (see EvaluateDeferredDesyncChecks).

客户端请求重同步(客户端自恢复失败后发出)。落地:盖章后转发 GM->HandleClientResyncRequest 进入服务端 resync 调度。

Client resync request (sent after client self-recovery fails). On landing: stamped and forwarded to GM->HandleClientResyncRequest for server-side resync scheduling.

客户端通知 resync 完成。落地:忽略传入 PlayerIndex,以 PlayerState 的 PlayerId 为准转发 GM->OnClientResyncComplete。

Client notifies resync completion. On landing: the passed PlayerIndex is ignored; the PlayerState PlayerId is authoritative and forwarded to GM->OnClientResyncComplete.

客户端报告 resync 结果(非 TCP 路径)。落地:以 PlayerState 的 PlayerId 盖章转发 GM->OnClientResyncResult。

Client reports a resync result (non-TCP path). On landing: stamped with the PlayerState PlayerId and forwarded to GM->OnClientResyncResult.

服务器向本客户端单播一条锁步指令(SendLockstepCommandToClient 的下行落地)。非权威端(HasAuthority() 为 false)才处理:入队 EnqueueNetCommand。

Server unicast of one lockstep command to this client (the downlink landing of SendLockstepCommandToClient). Handled only on non-authority (HasAuthority() false): enqueues via EnqueueNetCommand.

服务器命令客户端追帧到目标 tick。非权威端调用 Net->BeginFastForward(TargetTick, MB) 进入快进状态机。

Server directs the client to fast-forward to a target tick. Non-authority calls Net->BeginFastForward(TargetTick, MB) to enter the fast-forward state machine.

TCP 快照控制平面:服务器通知客户端将通过 TCP 接收快照。流程:确认玩家角色快照模板(EnsurePlayerCharacterSnapshotTemplate,失败仅警告不阻挡——逐 agent 模板可能已经 Agent.Spawn 预注册存在);EnterDesyncRecovery 进入恢复暂停;EnsureSnapshotTcpTransport 惰性创建客户端传输;传输缺失或已连接失败时上报失败;未连接则从 NetConnection->URL.Host 取服务器地址(空则 127.0.0.1)调 ConnectClient(非阻塞,含连接/认证截止);连接建立后等待 TCP 快照到达(由 OnSnapshotBytesReceived 反序列化并排队应用)。

TCP snapshot control plane: the server tells the client to expect a snapshot over TCP. Flow: confirms the player-character snapshot template (EnsurePlayerCharacterSnapshotTemplate; failure only warns — per-agent templates may already exist via Agent.Spawn pre-registration); EnterDesyncRecovery enters the recovery pause; EnsureSnapshotTcpTransport lazily creates the client transport; a missing transport or failed connect reports failure; otherwise the server host is taken from NetConnection->URL.Host (fallback 127.0.0.1) and ConnectClient runs (non-blocking, with connect/auth deadlines); once connected the client waits for the TCP snapshot (deserialized and queued for apply via OnSnapshotBytesReceived).

TCP 预连接握手:服务器在 PostLogin 分配槽位后立即下发,在任何大载荷锁步命令(自动分流)或 resync 需要之前先建立客户端快照/块 socket——缺了它 socket 要等首次 resync 才出现,期间自动分流上行块被静默丢弃(无 socket)、下行广播也跳过该客户端。实现:权威端直接返回(监听服务器主机共享服务端模拟);EnsureSnapshotTcpTransport 惰性创建传输后,若未连接则从 NetConnection->URL.Host 取服务器地址(空则 127.0.0.1)调 ConnectClient。连接失败非致命——传输 Tick 的节流自动重连会用保存的端点持续重试。

TCP pre-connect handshake: sent from PostLogin right after slot allocation, establishing the client's snapshot/block socket BEFORE any large-payload lockstep command (auto-flood) or resync needs it — without this the socket only appears on the first resync, until then auto-flood uplink blocks are silently dropped (no socket) and downlink broadcasts skip the client. Implementation: returns immediately on authority (the listen-server host shares the server sim); after EnsureSnapshotTcpTransport lazily creates the transport, if not yet connected it takes the server host from NetConnection->URL.Host (fallback 127.0.0.1) and calls ConnectClient. Connect failure is non-fatal — the throttled auto-reconnect in the transport Tick keeps retrying with the saved endpoint.

客户端向服务器报告 TCP 快照结果(成功 ack 或失败)。落地:以 PlayerState 的 PlayerId 盖章转发 GM->OnClientResyncResult——成功时其清理 ClientsWithPendingResync 并冲刷延迟生成。

Client reports the TCP snapshot result back to the server (success ack or failure). On landing: stamped with the PlayerState PlayerId and forwarded to GM->OnClientResyncResult — on success it clears ClientsWithPendingResync and flushes deferred spawns.

流场基底修复请求(非锁步命令):客户端已应用的流场基底版本与差分基准脱节(跳过差分的晚加入端、分叉链)时,请求该场的显式全量重发。基底修复带版本且幂等,无需确定性定序。落地:取 PlayerState->GetPlayerId() 盖章(缺失 -1)后转发 GM->HandleFlowFieldFullResyncRequest——服务器侧节流在该 GameMode 内。

Flow-field substrate repair request (NOT a lockstep command): when the client's applied flow-field substrate version missed a delta's base (skipped-delta late joiner, forked chain), it asks for an explicit FULL re-send of that field. Substrate repair is versioned + idempotent and needs no deterministic ordering. On landing: stamps PlayerState->GetPlayerId() (-1 if missing) and forwards to GM->HandleFlowFieldFullResyncRequest — the server-side throttle lives in that GameMode.

ServerRequestFlowFieldFullResync 的客户端侧节流包装:把连续被拒的差分链收敛为每 5 秒窗口、每场一次修复请求(服务器侧 GameMode 门限按场跨所有端收敛,此处保持单端安静)。仅客户端入口(ROLE_AutonomousProxy,FieldKey 为 None 时直接返回);任意客户端导入失败路径均可调用,重复调用安全。窗口内重复请求直接吞掉,过期则记录墙钟时间并发起 Server RPC。

Client-side throttled wrapper for ServerRequestFlowFieldFullResync: collapses a stuck chain of rejected deltas into one repair request per 5-second window per field (the server-side GameMode gate throttles per field across ALL peers; this one keeps a single peer quiet). Client-side entry only (ROLE_AutonomousProxy; returns immediately when FieldKey is None); safe to call repeatedly from any client-side import-failure path. Repeat requests inside the window are swallowed; an expired window records the wall-clock time and fires the Server RPC.

返回 PlayerState->GetPlayerId();PlayerState 缺失时返回 -1。

Returns PlayerState->GetPlayerId(); -1 when the PlayerState is missing.

把 AgentUniqueID 解析为当前有效的实体句柄:AgentUniqueID < 0 或 UMassBattleNetworkSubsystem::LookupEntityByUniqueID 查不到有效实体时返回 false;成功时写入并返回 AgentHandle。

Resolves AgentUniqueID into the currently valid entity handle: returns false when AgentUniqueID < 0 or UMassBattleNetworkSubsystem::LookupEntityByUniqueID finds no valid entity; on success writes and returns AgentHandle.