Pryaxis
Rewrite: Partially fix bed spawning for SSC
This PR rewrites the player spawnpoint saving system for SSC.
When SSC is enabled, player spawnpoints are only saved server side but not client side. However, due to the game's netcode logic, it is most likely impossible for the server (TShock) to change the client's spawnpoint value Terraria.Player.Spawn{X,Y} since the server provided value gets overwritten by its own client sided spawnpoint value in sp{X,Y} in Terraria.Player.FindSpawn() after packet 12 has been received.
In order to workaround this, TShock would keep track of the client's initial spawnpoint value on the first player spawn, it uses this value to compare if the player has changed his spawnpoint. This way, we can teleport the player with the correct server saved spawnpoint after the player tries to spawn with its incorrect client sided spawnpoint, at least before the player changes his spawnpoint mid-game.
What this means is that we purposely desync the client and server Terraria.Player.Spawn{X,Y} value before we know that the player had changed his spawnpoint. Once we know its changed, both client and server spawnpoint values would be synced, and we won't need to force teleport the player anymore.
Note that there is one limitation with this approach - this only works if the player spawns at his changed spawnpoint before logging off, otherwise the server would have no idea that the player's spawnpoint has changed.
I'm not sure if I covered all test cases, but please let me know if there is any bugs :)
What problem are you trying to solve here?
I'm mainly trying to solve #1966 (and #2844, #3011) which were known issues for quite some time. This PR doesn't fully solve the bed spawning problem with the limitations above. There are also other related issues like #1849.
Need testers for this PR! You can help test out with this Docker image with:
docker run -p 7777:7777 -v $(pwd)/tshock:/tshock -v $(pwd)/worlds:/worlds --rm -it ghcr.io/potatocider/tshock:general-devel
(This will create two folders tshock/ and worlds/ in your current working directory. You can then enable SSC via tshock/sscconfig.json)
Please do let me know if there's any bugs or potential issues with this workaround.
Upon dying and respawning, any action I take seems to not register. Can anyone reproduce this?
Steps to Reproduce
- Run & initialize the server
docker run -p 7777:7777 --rm -it ghcr.io/potatocider/tshock:general-devel
- Enter the world
- Die
- Attempt to place a block or hurt enemies
- The enemies will not register your hits server-side
- Leave and re-enter the server
- The blocks placed before will be gone as if never placed
Also, there seems to be some debugging output in the console. Is this intended?
@xNarnia
Also, there seems to be some debugging output in the console. Is this intended?
Oops. That's not intended.
I've fixed the death issue you mentioned, could you help test it again?
Tested scenarios in https://github.com/Pryaxis/TShock/issues/1966, https://github.com/Pryaxis/TShock/issues/2844, https://github.com/Pryaxis/TShock/issues/3011, and https://github.com/Pryaxis/TShock/issues/1849. All passing with no errors.
Also tested some scenarios with bombs destroying beds during death. It works similarly to previous versions of TShock, where the client sees destroyed blocks but it doesn't register on the server. In this scenario with the new change, the user is teleported back to the spawnpoint after the blocks are destroyed during death. However when relogging, the destroyed blocks and destroyed bed return. The player's spawnpoint is set back to the bed, and that is where they spawn in.
Tested with both SSC on & off.
Looks good to me! ✅
Okay, the case where beds are broken/invalid is now handled. This PR should be ready for merging ^^
Tested bed breaking with both the magic mirror and death.
Also tested Player A setting spawn point, Player A exiting server, Player B destroying Player A's bed spawnpoint, Player A logging back in. Player A returns to the world spawnpoint as intended.