Paul Hughes

Oho. That's a weird one. Since reloading fixes it, it's obviously not too serious, but it is pesky. Haven't had any repeats, yet.

Yeah, I'd suggest having a visibility box as well as the boundary box, but doubling the number of boundaries each entity has to track might make them laggy. Perhaps it could be added only to the ones with different visible/mechanical boundaries, or defining each entity's visibility box by a single floating-point number as a scaling factor on the boundary box? I'm not conversant with the code and how it decides whether an object is visible or not, but I imagine there must be some way to resolve the issue without making the engine do a lot of extra work for each entity.

Update: Quitting the game and reloading seems to have brought the map wall back, and filled in the new map. Interesting.

@Kumasasa: In Creative, I was able to find them with exceptional rarity, and also documented their anomalous despawning. As I'd suggested earlier in this thread, I suspect what most people reasonably interpreted as total failure to spawn was actually something to do with the spawning/despawning behavior. In an issue thread on the subject of spawning/despawning (MC-13716), it was resolved "works as intended"; but now people are reporting (here and on the forums) that horse spawning appears to happen at a more normal rate. I haven't had a chance to sit down and look for them in my standard testing world to see if they're willing to appear, so I can only go on those reports... but "fixed" and "as intended" seem to be conflicting resolutions. Maybe it was done intentionally and then someone decided to change it? Idunno.

Interesting... I'm not sure how the issue can be fixed if it worked as intended, but I guess one shouldn't look a gift horse in the mouth, so to speak.

I strongly suspect this is actually just another manifestation of MC-14269.... which claims that it's deliberate. Probably as a way of making horses extraordinarily rare, and available only to players willing to either go through really tedious searches or cheat. I don't particularly like this method of content-exclusion, but poor judgement isn't a code bug, so it's unlikely to be fixed.

Confirmed. In a game on Creative, I spawned like 20 horses with a spawn egg on an island with a tall marker, flew off until the marker disappeared, then flew back again, and there was only one untamed horse left. A tamed donkey and a tamed horse left on an adjacent island were still there, however, suggesting this only (or at least mainly) affects untamed horses.

However! I spawned a new batch. Saving, closing the world, opening another one and mucking about a bit, quitting out of Minecraft, and then restarting it and returning to the world with the horses, they were still there. Flying away until the chunks unloaded again, the horses once more disappeared, but the nearby 4 cows and 2 pigs remained. So it seems to be a chunk-loading thing, not a Minecraft-restarting thing.

I can confirm they disappear if their chunk is removed, unlike all other passive mobs. I wonder if this implies they also randomly despawn when more than 32 blocks away? This might partly explain the disputed reports of failure to spawn naturally.

I started a new world explicitly to look for horses, and after flying through three pretty expansive plains biomes (with and without hills) I've seen not a single naturally-spawned horse, although I saw tons of pigs and cows. I wonder if the spawning rates might be screwed up? Either way, I can't think of a good reason for them to not spawn in old worlds... it seems like we need more study to be sure what's going on with the spawning.

A theory: This could be partly a manifestation of MC-14269, an error where horses seem to despawn in the manner of hostile mobs rather than animals. Since animal spawning is so rarely, if something does inhibit horses spawning in old chunks (maybe even just passive mob like spawn-checking frequency) and they frequently disappear from new chunks before you can even see them, it would produce this observed "very very rare spawning" behavior.

Confirmed to affect minecarts and signs, as well, as of 13w04a.

I can confirm this happens at Normal and at least one intermediate FoV setting (91), so I suspect it's more about the client getting confused as to when you can actually see the object.

I've seen it happen with signs and minecarts, as well. I'm running up-to-date 64-bit Java on a Win7 machine, usually with over a GB of RAM free while Minecraft runs. I've seen it on the most recent snapshot as well as a few previous ones, but I might have just failed to notice it before since it only happens at the very edge of the FoV.

Currently uploading example video to http://youtu.be/3GxpZHG9DTA

Aha, interesting. At a glance, I think you're right.

I think I might have found the trouble--not why it was fixed by using another circuit, but the original cause. I had two back-pistons that pushed two fore-pistons, and a block to the side of each fore-piston's forward position was powered by a redstone trace prior to the back-pistons getting power, so they'd activate instantly, giving a quick two-block push. However, one of the two fore-pistons was apparently getting power through a redstone-powered block diagonally ahead and to its right. The redstone trace also pointed at the block diagonally ahead and above. When I changed this arrangement so that the redstone trace was one block lower and pointed directly at the foreward position of that fore-piston, it worked fine. The problem seems to have stemmed from this early activation, and the inability of a piston to activate when it would mean pushing an already-active piston.

I'm not sure how that fore-piston got power, however--it shouldn't have been able to, as far as I understand redstone power transmission, since there was no redstone trace on or pointing toward an adjacent block... or how using another circuit fixed it.

I'm honestly not sure--it's only this one circuit that I've seen produce this peculiar behavior, and as stated, it's intermittent. I'll continue playing with it, and see if I can identify the source of the problem. It's difficult to trigger the mechanism while standing in a position from which I can see the circuit's insides, so I may have to reproduce it in another map with a more transparent layout, and hope it misbehaves in the same mysterious way.

Huh. I didn't realize this desync issue could happen even in SSP with the standalone client. Sorry about the duplicate!