Skip to content

Releases: HeatXD/GekkoNet

v20260316133147-7f1f19e

16 Mar 13:31

Choose a tag to compare

Fix spectators not receiving GekkoPlayerDisconnected events

Decouple disconnect detection from the !locals.empty() guard in
SendPendingOutput. Spectator sessions have no locals, so the host
in remotes was never checked for timeout. Disconnect checks now run
whenever there are connected peers, regardless of local player count.

v20260224163038-7be848c

24 Feb 16:30

Choose a tag to compare

example updates

v20260224144247-7f51e61

24 Feb 14:42

Choose a tag to compare

refactor frame advantage calculations

v20260220015827-e1a5dc3

20 Feb 01:58

Choose a tag to compare

Make the spectator behaviour mimic a latch. the latch opens when fill…

v20260217000837-ad4b516

17 Feb 00:08

Choose a tag to compare

i forgot the nested structs, my bad.

v20260216231753-e7456de

16 Feb 23:17
e7456de

Choose a tag to compare

Fix C name clashes (#44)

* Fix C name clashes by adding prefixes to various enums and structs.

v20260213120409-6dbce37

13 Feb 12:04
6dbce37

Choose a tag to compare

Improve netstats (#43)

* relieved backpressure
* improve netstats

v20260208120351-4810019

08 Feb 12:03

Choose a tag to compare

New GekkoNet update with slight api changes, added stress session for checking state desyncs locally, Improved spectator mode allowing bigger spectator delays.

v20251205001100-92657cf

05 Dec 00:11

Choose a tag to compare

implement input send cache. to cooldown the hot path in addpendinginput

v20251201014944-f61d7fa

01 Dec 01:49
f61d7fa

Choose a tag to compare

Backend rework (#34)

* start redoing examples in sdl3

* bit further. planing to create a 4 player pong example

* good for now.

* ex game

* Sliders ex game

* Consistency with gekko_create

* redo disconnection logic. only fully drop when the remote is genuinely quiet.

* progress. have to make  a new func to fetch received remote inputs to the frontend tho. cuz now its stalling.

* cleanup old-back to front bridge.

* dont forget to jump by input size.

* back to front api

* CULPRIT FOUND

* this should take care of the mem leak.

* allow input bufffers to store multiple incorrectly marked frames and also clear them when solved

* ok the main cause seem to be a missed case in the input prediction logic. added this case now.
the case was that when i had already predicted the input before i was trying to predict beyond my limit rather then using the input that was already generated.

* cleanup

* fixed incorrect received input parsing

* should be spectators covered.

* spectators send inputs in parralel nor series

* ball respawn and bounce based on paddle section

* update game

* remove asserts in critical code otherwise they will get optimized out and cause desyncs

* cleanup

* preety much finish the example. there no win conditions or scores tho.

* Limited saving is not needed when lockstep mode is active

* remove hud from emaple game

* cleanup

* Spectator example added.

* Create init.txt

* Add files via upload

* Add files via upload

* Update project descriptions and logo attribution