Bug fixes in `bukkit-support` branch

[FallenCrystal]

Changed:

• Inject ChannelInitializer in JavaPlugin#onLoad if late-bind not enabled.
• Remove FallbackInboundHandler if player not need to verify.
• Lazy for ServerConnection to prevent get it before late-bind binds.

Tested on 1.8.8 PandaSpigot and 1.21 server.

Compatibility tested:

• 1.21:
  • [x] Geyser-Spigot
  • [x] ViaVersion
  • [x] PacketEvents
  • [x] ProtocolLib
• 1.8.8:
  • [x] late-bind (The logs for enable late-bind were not deleted.)
  • [x] ViaVersion
  • [x] PacketEvents
  • [x] ProtocolLib

ByteBuf#release located at FallbackBukkitInboundHandler, has also been deleted. Because there seems to be no sign of a memory overflow.

This was removed because there were of duplicate releases after removing FallbackInboundHandler.

These changes work though. But it looks a little messy. Accept any suggestions on code styles.

Known issues:

• Throw exceptions when sonar not loaded but injected.
• Players can join servers that don't have late-bind enabled in the same version as the legacy server before enabling the plugin. (Add bukkit listener to implement a temporary plan.)

[jonesdevelopment]

@FallenCrystal Are you able to test this on 1.12.2 and 1.16.5? 1.8.8 and 1.20.4 seem to work, but @F3F5 I using 1.12.2 (I think), and there've been some errors. I'm not sure if they're fixed yet, but it would be good to check if they are.

[FallenCrystal]

Are you able to test this on 1.12.2 and 1.16.5

I don't have these versions of the server. I'll test that later.

[FallenCrystal]

Unknown disconnect error found on 1.21 bukkit server with sonar (on login state). Set the PR to a draft until the investigation is clear.

[jonesdevelopment]

This resolves https://github.com/jonesdevelopment/sonar/issues/349?

[FallenCrystal]

This resolves #349?

This only fixes the issue where disconnect messages are not sent correctly and the number of online messages is not reduced because of the change of that PR.

[FallenCrystal]

btw can't use viafabric anymore

That issue may not be our main objective. Because Sonar only sends the registry that the vanilla client required for. If install Sonar on a server then cannot join the server (not fallback). It may be that this issue will be taken seriously.

[com.github.retrooper.packetevents.PacketEventsAPI] Could not find encoder/decoder handler in channel pipeline!

Check if the warning message still exists in latest commit.

[jonesdevelopment]

Ugh... why is this all so complicated? I'd really like to just avoid all these lazy-loading mechanisms and make it as simple and straightforward as possible, but I don't think we can pull this off without some "hacks". What do you think @FallenCrystal?

I've tested it and it seems to work. I'd refactor some codestyle stuff after I eventually merge this PR, but - all in all - this looks to be functional.

I love Minecraft. I love Bukkit. I love Spigot. ... I also love all people who are responsible for all those "async optimized mega fast multi threaded server forks"

[FallenCrystal]

What do you think?

Almost everything in this PR is to guarantee that it works. I didn't think much about it. tbh

I removed class Lazy and got ServerConnection directly in the FallbackBukkitInjector#inject method. After reading your comment, I thought it might be better.

About initializeListener located in SonarBukkitPlugin. I don't know what else could replace it. Since I wanted to avoid flooding bots before Sonar was enabled. Maybe we can let method FallbackBukkitInjector#inject to accept a CompletableFuture<?>. It may be neater this way.

[jonesdevelopment]

After reading your comment, I thought it might be better.

My problem is just that it's just too complicated for what we're trying to achieve. I don't have a problem with merging this as it is. I'd just like to clean it up and simplify it as much as possible.

Almost everything in this PR is to guarantee that it works

That is good. I'll also contact some other people for helping me test this.

[FallenCrystal]

My problem is just that it's just too complicated for what we're trying to achieve. I don't have a problem with merging this as it is. I'd just like to clean it up and simplify it as much as possible.

I know. The reply to that paragraph is aimed at:

I'd really like to just avoid all these lazy-loading

[jonesdevelopment]

lgtm