I've discovered a problem with my mod that I don't quite understand.  The following code places loot items in chests, and almost the same code works in Minecraft 1.5, 1.6, and 1.7 it work consistently:

 

	public void place(World world, int x, int y, int z, Random random) {
	BlockPos pos = new BlockPos(x, y, z);
	level += random.nextInt(2);
	if(level >= LootCategory.LEVELS) level = LootCategory.LEVELS - 1;
	DBlock.placeChest(world, x, y, z);
	if(world.getChunkFromChunkCoords(x / 16, z / 16).getBlock(pos) != DBlock.chest) return;
	TileEntityChest contents = (TileEntityChest)world.getTileEntity(pos);
	if(ConfigHandler.vanillaLoot && (!ConfigHandler.stingyLoot || random.nextBoolean())) 
		vanillaChest(contents, random);
	int which = random.nextInt(2);
	switch (which) {
	case 0:
		fillChest(contents, LootType.HEAL, random);
		break;
	case 1:
		fillChest(contents, LootType.GEAR, random);
		break;
	}
}

 

	protected void fillChest(TileEntityChest chest, LootType kind, Random random) {		
	int num;
	if(ConfigHandler.stingyLoot) num = random.nextInt(2 + (level / 2)) + 1;
	else num = random.nextInt(3 + (level / 2)) + 2;
	for(int i = 0; i < num; i++) {
		ItemStack treasure = LootCategory.getLoot(kind, level, random).getStack(random);
		if(treasure != null) chest.setInventorySlotContents(random.nextInt(27), treasure);
	}
	if(!ConfigHandler.vanillaLoot) {
		ItemStack treasure = LootCategory.getLoot(LootType.HEAL, level, random).getStack(random);
		if(treasure != null) chest.setInventorySlotContents(random.nextInt(27), treasure);
	}
}

 

Again, almost the same code worked consistently prior to 1.8.  However, in 1.8 it works inconsistently, producing long runs of successfully filling chests mixed with long runs (over several dungeons / many chunks, very many chests) of leaving chests empty.

 

I have found the removing the line "if(world.getChunkFromChunkCoords(x / 16, z / 16).getBlock(pos) != DBlock.chest) return;" leads to crashes with a null pointer exception.  My best guess is that concurrency might be involved, perhaps threads involved in placing blocks lagging behind my code so that the chest isn't there yet when it tries to add loot (or at least comes to the failsafe code) -- but that is purely an educated guess.

 

Does anyone have any other idea of an explanation?  Or a fix that doesn't involve radical changes (I have thought of spitting block and TileEntity placement into separate iterations? -- or, if its allowed, creating a separate thread that can sleep?)  If so I'd be very grateful to hear any relevant info or suggestions.

When in the world gen process is your 'place' method being called? I've generated lots of chests and none of them have had this problem (at least that I've found / heard so far).

 

Aside: why are you using world.getChunkFromChunkCoords to check the block at the position instead of the world method directly? I.e. world.getBlockState(pos).getBlock().

 

Speaking of block states, can you show your DBlock.placeChest method? Perhaps something is going awry in there?

Thanks.  Sorry for the long delay, I've been busy with other things and forgot to check back here.  I guess I was using the get chunk method because of adapting from older code.  I was trying to update and didn't know all the methods that take "pos" as a parameter.

 

The chest placing code is as follows:

 

public static final Block spawner = (Block)Block.getBlockFromName("mob_spawner");
public static final Block chest   = (Block)Block.getBlockFromName("chest");

...

public static void placeBlock(World world, int x, int y, int z, Block block) {
	// This wrapper is a protection against possible changes in block representation,
	// e.g., abandoning the ID system, allowing any needed changes to be made here
	// instead of elsewhere. 
	if(isProtectedBlock(world, x, y, z)) return;
	world.setBlockState(new BlockPos(x, y, z), block.getDefaultState());
}

...	

public static void placeChest(World world, int x, int y, int z) {
	if(!isProtectedBlock(world, x, y, z))
		placeBlock(world, x, y, z, chest);		
}

 

I didn't originally show the code for placing the chest block because the chest itself was always there, it was only the contents that were missing.

 

Then, for all I know this was a Forge bug that might already have been fixed.  I play mostly 1.7 and 1.6 (for MineFantasy), and only used this with 1.8 enough for some general testing, so I wouldn't know if its suddenly works now.

In your place chest method, your isProtected method is redundant since it passes over to the placeBlock method which has the check. That's not why your problem is occurring, but I thought it's still worth pointing out.

Well, placeChest() would have originally called world.setBlock() directly, and may be different in different versions (I maintains versions back to 1.6.4 and have started using git to port bit directly).  I not that worried about minor optimizations until I'm sure it fundamentally works though, so I haven't worried about little things like this for now.