Interpolation + event updates causes visual glitch when mining #214
Labels
Comments
Merged
sourencho
added a commit
that referenced
this issue
May 23, 2020
- Replace rewind and replay with the client playing in the past and interpolating to game states recieved from the server stored in a fixed size buffer. The size of the buffer is configurable through the var CLIENT_GAME_STATE_BUFFER_SIZE in game_settings.json. - The logic for updating the local game state is as follows: - If the clients tick is too far behind (more than MAX_BEHIND_TICK_DELTA ticks) the smallest server state in the buffer then the local state is overwritten with the smallest server state. - If the client tick is not too far behind the smallest server state then then a game state is interpolated for the next tick using the smallest server state. - If the client is ahead of the smallest server state then the next game state in the buffer is used to recalculate the current tick's state but the tick is not incremented. This is in order to stall and fall back behind the smallest server state. If there aren't any states in the buffer ahead of the current tick, then the game state doesn't change. - Add debug metrics around interpolation, ahead, behind, etc and update the layout to be more readable - Implement hermite interpolation to be able interpolate curves rather than straight lines given start and end velocity of the rigid bodies. - Send collision events in game state updates. This was necessary because we no longer extrapolate, we only interpolate, which means we no longer run the physics simulation locally and can't trigger visuals caused by collisions. The events recieved from the server are now applied and trigger collision animations. This introduced a visual bug (issue #214). - Reduce server input window sleep to 16ms to make the client and server ticks/sec in sync. This is a messy dependency, it just so happens that the server sleeps for that window causing it to control it's steps per second. Not really a good way to control tick speed and the client and server tick speeds goes out of sync as more clients connect. We might need to implement some clock sync in the future. - Seems like this change also broke the winner being displayed on the client (issue #215). - Minor code cleanup.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
PR #213 introduces a bug where a collision event causes mine to appear to be mined twice in a quick visual glitch.
This happens because now that the client is just interpolating, it no longer computes collisions locally. In order for the appropriate visual side effects of a collision event occur (such as the collision animaton), the server now sends events along with the game state. The callbacks for this event are then triggered on the client. This was done to trigger the collision animation but the "mine" callback also triggers. This causes the mine to lose a resource. Since the server already sent the mine with the updated number of resources we are now showing the state where the mine was mined twice. In the next frame this is likely to get corrected when we interpolate again but this causes a quick visual glitch.
A solution could be to have distinct callback subscriptions on the server and the client. That way the client can just subscribe the animation callback and not the "mine" one.
The text was updated successfully, but these errors were encountered: