Skip to content

Release v4.1 (2026-04-02)

Choose a tag to compare

View all tags
@github-actions github-actions released this 02 Apr 00:42
· 159 commits to master since this release
v4.1
aba98cd
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

What's Changed

Fixes

  • While in Release build upon login the empty while loop is optimized out and the app hangs. by @Xian55 in #1
  • Replace Thread.Sleep spell delay with queue-based tracking by @Xian55 in #5
  • Fix npc movement zig-zag at tile boundaries by @Xian55 in #6
  • Fix: ItemEffect hotfix handling for duplicate keys and record removal by @Xian55 in #9
  • Feature: .NET 10 migration with performance optimizations and CI modernization by @Xian55 in #16
  • Fix CMSG_ENABLE_NAGLE causing decrypt failure when Optimize Network for Speed is disabled 2960e77

New Features

  • Added metrics collection to measure opcode processing latency by @Xian55 in #7
  • Feature/perf/span pooled packet handler by @Xian55 in #10
  • Fix all 1,902 nullable compiler warnings by @Xian55 in #14

Performance

  • Refactor WowGuid to value-type record structs by @Xian55 in #2
  • BnetTcpSession: optimize packet parsing with Span by @Xian55 in #3
  • BnetTcpSession: zero-allocation buffer management with Span by @Xian55 in #4
  • Modernize threading, allocations, and runtime patterns for .NET 10 by @Xian55 in #15

New Contributors

  • @Xian55 made their first contribution in #1

Performance comparison

Measured with --metrics flag, 8-minute sessions, same game world and activity pattern. Baseline from PR #7 (master branch).

Client → Server (top improvements by P50)

Opcode Count (old/new) P50 before P50 after Speedup P99 before P99 after
CMSG_CAST_SPELL 121 / 45 16.145ms 0.092ms 175x 17.164ms 2.301ms
CMSG_DB_QUERY_BULK 2 / 2 3.478ms 1.544ms 2.3x 5.720ms 1.924ms
CMSG_MOVE_START_FORWARD 112 / 93 0.070ms 0.043ms 1.6x 0.461ms 0.351ms
CMSG_MOVE_HEARTBEAT 93 / 32 0.072ms 0.045ms 1.6x 2.361ms 1.607ms
CMSG_QUEST_GIVER_STATUS_MULTIPLE 24 / 27 0.035ms 0.030ms 1.2x 1.676ms 0.946ms

Server → Client (top improvements by P50)

Opcode Count (old/new) P50 before P50 after Speedup P99 before P99 after
SMSG_CAST_FAILED 149 / 55 15.959ms 0.002ms 7,980x 17.043ms 0.885ms
SMSG_SPELL_GO 160 / 62 16.041ms 0.065ms 247x 17.173ms 2.616ms
SMSG_ENUM_CHARACTERS_RESULT 1 / 1 29.215ms 17.247ms 1.7x — —
SMSG_SEND_KNOWN_SPELLS 1 / 1 4.159ms 1.367ms 3x — —
SMSG_ACCOUNT_DATA_TIMES 1 / 1 8.113ms 4.250ms 1.9x — —
SMSG_ITEM_QUERY_SINGLE_RESPONSE 42 / 24 0.256ms 0.570ms ~same 25.301ms 10.203ms
SMSG_LOOT_RESPONSE 44 / 31 0.081ms 0.072ms 1.1x 3.508ms 2.507ms

Analysis

The spell-related opcodes (CAST_SPELL, SPELL_GO, CAST_FAILED) showed a ~16ms median latency floor on master — the classic signature of Mutex kernel-mode transitions under contention. Replacing with System.Threading.Lock eliminated this entirely, bringing medians to sub-100μs.

One-shot startup opcodes (ENUM_CHARACTERS_RESULT, ACCOUNT_DATA_TIMES, SEND_KNOWN_SPELLS) show 1.7–3x improvement from BinaryPrimitives / allocation reduction in serialization paths.

┌──────────────────────────────────────────────────────────────┐
│              HOW THE PROXY SETUP WORKS                       │
│                                                             │
│   Your PC                                    Internet        │
│  ┌──────────┐     same computer     ┌───────┐      ┌──────┐   │
│  │   WoW    │ ◄──── instant ──────► │ Proxy │ ◄──► │Server│ │
│  │  Client  │     (0ms delay)       │       │ slow │      │  │
│  └──────────┘                       └───────┘      └──────┘  │
│                                                              │
│  The WoW client talks to the Proxy on your own computer.     │
│  The Proxy talks to the game server over the internet.       │
└──────────────────────────────────────────────────────────────┘

Shall i set the Optimize Network for Speed enabled or not ?

┌──────────────────────────────────────────────────────────────┐
│         "OPTIMIZE NETWORK FOR SPEED" = ON (default)          │
│                                                              │
│  WoW sends every tiny message immediately, one at a time:    │
│                                                              │
│  WoW ──► Proxy: "move"         Proxy ──► Server: "move"      │
│  WoW ──► Proxy: "jump"         Proxy ──► Server: "jump"      │
│  WoW ──► Proxy: "cast spell"   Proxy ──► Server: "cast"      │
│                                                              │
│  3 separate deliveries across the internet                   │
│  Like sending 3 separate letters in 3 envelopes              │
└──────────────────────────────────────────────────────────────┘

┌──────────────────────────────────────────────────────────────┐
│          "OPTIMIZE NETWORK FOR SPEED" = OFF                  │
│                                                              │
│  WoW batches nearby messages together before sending:        │
│                                                              │
│  WoW ──► Proxy: "move+jump+cast"                             │
│                      │                                       │
│                      ▼                                       │
│               Proxy ──► Server: "move+jump+cast"             │
│                                                              │
│  1 delivery across the internet instead of 3                 │
│  Like putting 3 letters in 1 envelope                        │
│                                                              │
│  The batching between WoW and Proxy is FREE                  │
│  because they're on the same computer (instant).             │
│  But it saves real time on the internet hop!                 │
└──────────────────────────────────────────────────────────────

Full Changelog: https://github.com/Xian55/HermesProxy/commits/v4.1

Contributors

Xian55
Assets 6
Loading