Improves the Bukkit detection as a core mod.

[brunoxkk0]

Instead of using the World class to detect if it's a Bukkit core mod, it's safer to use the DyeColor class.
Using this class shouldn't break Bukkit itself.

That will fix #33

[mitchej123]

I'm not entirely following, what's the problem with using World?

Also is DyeColor there on the various forks, like thermos, etc?

[brunoxkk0]

It fixes the mentioned issue #33. We don't know why it happens, but loading Bukkit classes too early does cause issues with plugins and breaks reflection. Alternatively trying to detect bukkit without loading classes should be safer.

And about the second question, yes it will work on detecting all bukkit-like implementations.

[mts2200]

We don't know why it happens, but loading Bukkit classes too early does cause issues with plugins and breaks reflection.

This happens because Bukkit refers to Minecraft classes by notch/nms names, mods do this using SRG names. Initially Minecraft is still covered with notch names, Forge changes them at runtime to SRG using DeobfuscationTransformer and FMLRemappingAdapter. Judging by the stacktrace from the issue, the person still has Notch names,
although it should have changed to SRG. If we decompile org.bukkit.World (I took it from Crucible), then there will be just the class that was not found in runtime:

I dont know how Thermos/Crucible transform Bukkit NMS names to SRG, but its not working there.

[brunoxkk0]

It does not transform anything, All bukkit related code is written under the patch dev workflow forge has and reobfuscated like a normal forge release. To go more into details we don't know why loading that class too early skips that deobfuscation step, nor why the server still works if that method was actually missed which should utterly break bukkit.