[Crash]: Crash when trying to sync config data with Switchy

[UpcraftLP]

Mod Loader (Required)

Fabric

Minecraft Version(s) (Required)

1.20.1

Mod Version(s) (Required)

8.0.2

Notes (Required)

1. join a singleplayer world (multiplayer kicks you back to the server select screen, does not crash)
2. create a secondary profile with switchy
3. try to switch to the second profile
4. crash!

seems to be a missing null check here: https://github.com/Fuzss/forgeconfigapiport/blob/1.20.1/Common/src/main/java/net/minecraftforge/fml/config/ModConfig.java#L92

Crash Report (Required)

https://mclo.gs/MLVv98Y

latest.log (Optional)

No response

[Fuzss]

Hey, thanks for reporting. Do you know if this is also an issue on just plain Forge?

I checked when the config sync packet is sent, and that happens pretty much at the exact same time as in Forge. A mismatch with the timing there was one thing that came to mind able to cause such issues.

The config data being null would imply the config is not yet loaded on the server, so it is unable to send it. But the server configs are loaded in ServerLifecycleEvents#SERVER_STARTING, which happens before players are even able to log in. Very weird. 😅

[UpcraftLP]

Hmm nope, cannot reproduce on Forge.

However I did manage to reproduce it with a much smaller modpack on Fabric, only had switchy + nether chested + dependencies.

https://mclo.gs/WtstYVh

[sisby-folk]

I'll explain this one

In 1.20, in singleplayer, with default configs, switching profiles with switchy:

• Kicks the singleplayer host
• Prevents the integrated server from shutting down
• Reconnects the singleplayer host

It does this both to save time over a full world reload, and to maintain compatibility with LAN hosting - the cost is, a good few mods assume, implementation-wise, that a flow like this is impossible, and end up either running setup code twice (and crashing) or running clean-up code incorrectly, or failing to run the correct clean-up code - usually by assuming the host leaving the game and the integrated server shutting down always happens at the same time.

setting singleplayerFastReconnect=false instead shuts down the server and restarts it, each switch. this is slower, breaks compatibility with LAN, but won't break mods that assume host kick and shutdown always align.

Basically, there's a workaround, but if you'd like to update the mod for compatibility with this flow that's very welcome too.

[sisby-folk]

generified this info as sisby-folk/switchy/issues/126

[Fuzss]

@sisby-folk Thanks for the insights!
Is it possibly that despite you preventing the integrated server from shutting down Fabric’s ServerLifecycleEvents.SERVER_STOPPED still runs? That’s where the server configs are cleared.

[sisby-folk]

I can confirm that those events specifically are not hit, but that QUERY_START, which you are listening for, is hit (as you'd expect for a normal multiplayer disconnect / reconnect)

[Fuzss]

Alright, I published v8.0.3 which attempts to fix this.
I can confirm the fix is working in singleplayer, however I haven't tested multiplayer.
@UpcraftLP Can you test that please?