Resolve _minecraft._tcp.<hostname> SRV records before connecting,
matching Java Edition behavior. Players can connect using just a domain
name and the client will look up the actual server address and port
from DNS. Falls back to the original hostname/port if no SRV record
exists or the address is a numeric IP.
triggerEvent() set ignoreUpdate to true at the start but three early
return paths skipped the reset at the end. Once any of these paths was
hit, the TLS flag stayed true permanently, blocking all piston neighbor
updates for the rest of the server session.
Replace the boolean-flag-based async join system with a clean state machine
(eJoinState enum) and move connection progress handling from UIScene_JoinMenu
into UIScene_ConnectingProgress as a dedicated UI class.
Combines the best of two approaches: non-blocking sockets with select()
timeout and SO_RCVTIMEO clearing (prevents random disconnects) with the
upstream's state enum, FinalizeJoin separation, and ConnectingProgress UI.
JoinGame() now returns JOINGAME_PENDING on Win64, and
PlatformNetworkManagerStub::DoWork() polls the join state to finalize
the connection when the background thread succeeds.
triggerEvent() set ignoreUpdate to true at the start but three early
return paths (lines 221, 225, 250) skipped the reset at the end. Once
any of these paths was hit (common with fast redstone clocks where the
signal state changes between event queuing and processing), the TLS
flag stayed true permanently, blocking all piston neighbor updates for
the rest of the server session.
Clear the 5-second SO_RCVTIMEO that was set during the connection
handshake but never removed. The timeout persisted into the game
session, causing the client to disconnect whenever the server paused
for longer than 5 seconds (e.g. autosave, chunk I/O).
Also update README with chunk unloading and connection stability fixes.
Commit a24318ee changed drop() to immediately remove chunks from cache,
bypassing the deferred m_toDrop save/unload pipeline. This caused missing
chunks on dedicated servers, iterator invalidation in dropAll() and
ServerLevel::save(), and entity duplication (item frames) from chunks
being reloaded without their entities first being removed from the level.
- ServerChunkCache::drop(): restore m_toDrop queue instead of immediate
cache removal, so tick() can save/unload/move to unloadedCache safely
- MultiPlayerChunkCache::drop(): restore soft-unload (keep chunk in cache
with loaded=true) instead of nulling cache and hasData
- PlayerChunkMap::setRadius(): remove dropAll() call when reducing radius,
the per-chunk removal loop already handles out-of-range chunks
The dedicated server previously picked a completely random seed with no
biome diversity checks (the client validates but the server skipped it).
On top of that, the client's findSeed() was hardcoded to only check a
54-chunk (Classic) area, so Large worlds had no diversity guarantee
beyond the center.
New server worlds now use findSeed() scaled to the full target world
size. Added override-seed in server.properties to fix existing worlds
without deleting them.
Server list: edits and deletions now update the UI immediately by
calling SearchForGames() in ForceFriendsSessionRefresh() and
UpdateGamesList() on nav-back to LoadOrJoinMenu.
Connection: moved WinsockNetLayer::JoinGame() to a background thread
with non-blocking sockets (5s timeout, 3 retries). Users can cancel
with B/Escape during the attempt. Failed connections always show an
error dialog.
Loki Rautio