Canonical Movement #82085
Canonical Movement #82085
Conversation
san7890
added
Refactor
|
people were complaining about a bug that would "lock people up" transiently for a few seconds on Sybil Round 225755- I pulled the test-merge and demanded they make an issue report regarding it (some maintainers were on, hopefully they can collect good data i hope) |
|
It's been testmerged on sybil for 2 rounds and both times apparently movement just seemed to stop. Even observing I find myself unable to move, trying to move instead just types in my chat bar and tabbing doesn't swap to move mode. arrow keys also don't work and instead just moves where I'm typing. 8mb.video-8oL-9THp18BN.mp4Runtimes that might be related (ID 225755): Kyler also noticed errors about divisions by 0 in FOV, screenshot stolen from him on discord |
|
I have changed the code such that The |
|
No complaints about this over the last two days of dual-server test merges, you think it's ready to ship? |
|
About The Pull Request
Currently, TG's movement chain is not canonical, meaning movement can resolve in an order that doesn't actually represent what it going on. For example, if something inside of an Entered() call would move the currently moving object, it resolves in this order:
This makes Moved() unreliable at tracking things like spatial grid locations. This is a massive problem.
This PR introduces
active_movement, a list containing arguments forMoved(). At the start ofMove()anddoMove(), if the list is present, it will call Moved(), concluding the original movement, before proceeding. The originalMove()call will not end inMoved(), due toactive_movementbeing null.This is touching some of the most important code in the game and needs extensive testing.