v330

CHANGELOG.md

@@ -4,6 +4,87 @@ All notable changes to this project will be documented in this file.
 This format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+#### [v3.3.0](https://github.com/ergo-services/ergo/releases/tag/v1.999.330) 2026-xx-xx [tag version v1.999.330] ####
+
+* Added **pointer type support** in EDF - `*int`, `*string`, `[]*T`, `map[K]*V`, pointer struct fields. Nil state preserved. Nested pointers (`**T`) not supported. Max encoding depth limit (100) prevents stack overflow on deeply nested structures. See [Network Transparency](https://docs.ergo.services/networking/network-transparency) documentation
+* Fixed logger to preserve Behavior name when process registers name
+* Added **process lifecycle counters** to `gen.NodeInfo` - `ProcessesSpawned`, `ProcessesSpawnFailed`, `ProcessesTerminated` for cumulative statistics
+* Added **mailbox latency measurement** (build with `-tags=latency`). `QueueMPSC.Latency()` returns the age of the oldest message in the queue (nanoseconds), -1 if disabled. `ProcessMailbox.Latency()` returns the max across all four queues. Added `MailboxLatency` field to `ProcessShortInfo` and latency fields to `MailboxQueues` in `ProcessInfo`. See [Debugging](https://docs.ergo.services/advanced/debugging) documentation
+* Added **`Node.ProcessRangeShortInfo`** for efficient callback-based iteration over all processes with their current state. See [Metrics actor](https://docs.ergo.services/extra-library/actors/metrics) for Prometheus integration
+* Added **per-event metrics** - `EventInfo` now includes `MessagesPublished`, `MessagesLocalSent`, `MessagesRemoteSent` counters. Added `Node.EventInfo` and `Node.EventRangeInfo` for querying event statistics. Added `EventsPublished`, `EventsReceived`, `EventsLocalSent`, `EventsRemoteSent` to `NodeInfo`. `EventsPublished` counts only local producer publishes, `EventsReceived` counts events arriving from remote nodes
+* Added **process init time measurement** - `InitTime` field in `ProcessShortInfo` and `ProcessInfo` records the time spent in `ProcessInit` callback (nanoseconds)
+* Fixed **message counters for meta processes** - meta process traffic now propagates to parent process counters, making `ProcessRangeShortInfo` aggregates balanced
+* Fixed **self-send message counter** - `messagesOut` now incremented for self-sends
+* Fixed **simultaneous connect dead loop** - two nodes dialing each other at the same time no longer cause infinite retry loops. Deterministic connection IDs and Erlang-style collision detection (`EnableSimultaneousConnect` flag) ensure exactly one connection per pair. Fixed related connection leaks
+* Fixed **silent data loss on connection pool write failure** - a transient write error could permanently break a pool item's write path without detection, causing all subsequent messages to be silently dropped while the connection appeared healthy
+* Added **software keepalive** for inter-node connections. Application-level heartbeat detects silent failures that TCP keepalive cannot: stuck processes, broken flushers, goroutine starvation. Each side advertises its period during handshake (8 bits in `NetworkFlags`); receiver uses peer's period for timeout. Enabled by default (15s period, 3 misses, 45s timeout). Configure via `NetworkFlags.EnableSoftwareKeepAlive` and `NetworkOptions.SoftwareKeepAliveMisses`. See [Network Stack](https://docs.ergo.services/networking/network-stack#software-keepalive) documentation
+* Added **handshake deadline** (5s) to prevent hung handshakes from blocking connection goroutines indefinitely
+* Added **message fragmentation** for large messages. Messages exceeding the fragment size (default 65000 bytes) are automatically split for transmission and reassembled on the receiving side. Works with compression, important delivery, and all message types. With `KeepNetworkOrder` disabled, fragments are distributed across all TCP connections in the pool for maximum throughput. Both nodes must enable `EnableFragmentation` flag (enabled by default). Configure via `NetworkOptions.FragmentSize`, `FragmentTimeout`, `MaxFragmentAssemblies`. See [Network Stack](https://docs.ergo.services/networking/network-stack#message-fragmentation) documentation
+* Fixed **important delivery use-after-release** - reference ID for acknowledgment was read from buffer after it was returned to the pool, causing corrupted ACK responses under load. Affected `SendImportant` for PID, ProcessID, and Alias targets
+
+#### [v3.2.0](https://github.com/ergo-services/ergo/releases/tag/v1.999.320) 2026-02-04 [tag version v1.999.320] ####
+
+* Introduced **mTLS support** - new `gen.CertAuthManager` interface for mutual TLS with CA pool management (`ClientCAs`, `RootCAs`, `ClientAuth`, `ServerName`). See [Mutual TLS](https://docs.ergo.services/networking/mutual-tls) documentation
+* Introduced **NAT support** - new `RouteHost` and `RoutePort` options in `gen.AcceptorOptions` for nodes behind NAT or load balancers. See [Behind the NAT](https://docs.ergo.services/networking/behind-the-nat) documentation
+* Introduced **spawn time control** - `InitTimeout` option in `gen.ProcessOptions` limits `ProcessInit` duration for both local and remote spawn. Remote spawn and application processes limited to max 15 seconds. See [Process](https://docs.ergo.services/basics/process) documentation
+* Introduced **zip-bomb protection** - decompression size limits to prevent memory exhaustion attacks
+* Added `gen.Ref` methods for request timeout tracking. See [Generic Types](https://docs.ergo.services/basics/generic-types#gen.ref):
+  - `Deadline` - returns deadline timestamp stored in reference
+  - `IsAlive` - checks if reference is still valid (deadline not exceeded)
+* Added `gen.Node` methods. See [Node](https://docs.ergo.services/basics/node) documentation:
+  - `ProcessPID` / `ProcessName` - resolve process PID by name and vice versa
+  - `Call`, `CallWithTimeout`, `CallWithPriority`, `CallImportant`, `CallPID`, `CallProcessID`, `CallAlias` - synchronous requests from Node interface
+  - `Inspect` / `InspectMeta` - inspect processes and meta processes
+  - `MakeRefWithDeadline` - create reference with embedded deadline
+* Added `gen.RemoteNode.ApplicationInfo` - query application information from remote nodes. See [Remote Start Application](https://docs.ergo.services/networking/remote-start-application) documentation
+* Added `gen.Process` methods. See [Process](https://docs.ergo.services/basics/process) documentation:
+  - `SendWithPriorityAfter` - delayed send with priority
+  - `SendExitAfter` / `SendExitMetaAfter` - delayed exit signals
+  - `SendResponseImportant` / `SendResponseErrorImportant` - important delivery for responses
+* Added `gen.Meta` methods. See [Meta Process](https://docs.ergo.services/basics/meta-process) documentation:
+  - `SendResponse` / `SendResponseError` - respond to requests from meta process
+  - `SendPriority` / `SetSendPriority` - message priority control
+  - `Compression` / `SetCompression` - compression settings
+  - `EnvDefault` - get environment variable with default value
+* Added `gen.ApplicationSpec` / `gen.ApplicationInfo` fields:
+  - `Tags` - labels for instance selection (blue/green, canary, maintenance). See [Tags for Instance Selection](https://docs.ergo.services/basics/application#tags-for-instance-selection)
+  - `Map` - logical role to process name mapping. See [Process Role Mapping](https://docs.ergo.services/basics/application#process-role-mapping)
+* Added **HandleInspect** implementations for all supervisor types (OFO, ARFO, SOFO)
+* Fixed **LinkChild** in `RemoteNode.Spawn` / `RemoteNode.SpawnRegister`
+* Fixed **args persistence** for Simple One For One supervisor - child processes now restart with their original spawn arguments
+* Fixed **critical bug**: terminate signals (Link/Monitor exits) were incorrectly rejected due to wrong incarnation validation in network layer. Thanks to [@qjpcpu](https://github.com/qjpcpu) for reporting [#248](https://github.com/ergo-services/ergo/issues/248)
+* Completely reworked internal **Target Manager** (`node/tm/`) - improved architecture for process, event, and node target management with comprehensive test coverage
+* Completely reworked internal **Pub/Sub** mechanism - improved reliability and performance
+* Improved **ProcessInit state** - more `gen.Process` methods now available during initialization:
+  - `Link*`, `Unlink*`, `Monitor*`, `Demonitor*`
+  - `Call*`, `Inspect`, `InspectMeta`
+  - `RegisterName`, `UnregisterName`, `RegisterEvent`, `UnregisterEvent`
+  - `SendResponse*`, `SendResponseError*`
+  - `CreateAlias`, `DeleteAlias`
+* Introduced **shutdown timeout** - `ShutdownTimeout` option in `gen.NodeOptions` (default 3 minutes). During graceful shutdown, pending processes are logged every 5 seconds with state and queue info. After timeout, node force exits with error code 1. See [Node](https://docs.ergo.services/basics/node) documentation
+* Added **pprof labels** for actor and meta process goroutines (with `--tags pprof`) - each process goroutine is labeled with its PID, each meta process with its Alias, making it easy to identify stuck processes in pprof output
+* Improved API documentation - comprehensive godoc comments for all public interfaces
+* **Documentation rewritten** - complete documentation now included in the repository (`docs/`) and available at [docs.ergo.services](https://docs.ergo.services)
+* New documentation articles:
+  - [Project Structure](https://docs.ergo.services/basics/project-structure) - organizing projects with message isolation levels, deployment patterns, and evolution strategies
+  - [Building a Cluster](https://docs.ergo.services/advanced/building-a-cluster) - step-by-step guide to distributed systems with service discovery, load balancing, and failover
+  - [Message Versioning](https://docs.ergo.services/advanced/message-versioning) - evolving message contracts in distributed clusters with explicit versioning strategies
+  - [Handle Sync](https://docs.ergo.services/advanced/handle-sync) - synchronous message handling patterns
+  - [Important Delivery](https://docs.ergo.services/advanced/important-delivery) - guaranteed delivery mechanism
+  - [Pub/Sub Internals](https://docs.ergo.services/advanced/pub-sub-internals) - event system architecture
+  - [Debugging](https://docs.ergo.services/advanced/debugging) - build tags, pprof integration, troubleshooting stuck processes
+
+* **Extra Library - Actors** (https://github.com/ergo-services/actor):
+  - Introduced **Leader** actor - distributed leader election with Raft-inspired consensus algorithm. Features: term-based disambiguation, automatic failover, split-brain prevention through majority quorum, dynamic peer discovery. See [documentation](https://docs.ergo.services/extra-library/actors/leader)
+  - Introduced **Metrics** actor - Prometheus metrics exporter that collects node/network telemetry via HTTP endpoint. Features: automatic collection of node metrics (uptime, processes, memory), network metrics per remote node, extensible for custom metrics. See [documentation](https://docs.ergo.services/extra-library/actors/metrics)
+
+* **Extra Library - Meta Processes** (https://github.com/ergo-services/meta):
+  - Introduced **SSE** (Server-Sent Events) meta-process - unidirectional server-to-client streaming over HTTP. Features: server handler for accepting connections, client connection for external SSE endpoints, full SSE spec support (event types, IDs, retry hints, multi-line data), process pool with round-robin load balancing, Last-Event-ID for reconnection. See [documentation](https://docs.ergo.services/extra-library/meta-processes/sse)
+
+* **Benchmarks** (https://github.com/ergo-services/benchmarks):
+  - Introduced **Distributed Pub/Sub** benchmark - demonstrates event delivery to 1,000,000 subscribers across 10 nodes. Achieves 2.9M msg/sec delivery rate with only 10 network messages (one per consumer node) instead of 1M
+
+
 #### [v3.1.0](https://github.com/ergo-services/ergo/releases/tag/v1.999.310) 2025-09-04 [tag version v1.999.310] ####
 
 **New Features**