applestruck

Oops, somehow missed the WAI marking when going to update the report. Unfortunate but understandable. Hopefully it won't destroy too many player creatiions.

I disagree, the amount of machines we have now and that were being developed as this change happened is much bigger than what we might develop within a few updates, not to mention that current behaviour just makes less sense, at least in my opinion, with ptojectiles being redirected or stopped by pistons or water.

Fair enough. If there's a quick point release with a reversion to the old behavior I guess that shouldn't be too disruptive. Regardless it would be nice to hear if this was inadvertently chanaged or if not, hear what the rationale was. If you want to make a clip or a WDL showing off some more of the weird behavior I can attach it to the report.

At this point I assume it's too late for the issue to be addressed before the 1.21 release. In that case it's probably for the best that the change be kept whether accidental or purposeful, since people will soon start building things with the new physics, and reverting the change in the future would just break all those new things.

All I would ask is that in the future significant entity physics changes be given more explicit patch notes instead of being hidden away as side effects of changes to NBT data, in order that players interested in such changes might find out about them in a more timely fashion. I only found out by chance, and apparently too late to make a bug report that could affect the release. The relative consistency of Minecraft's sandbox mechanics is in my opinion one of the game's core strengths, so it's unfortunate whenever parts of that sandbox are changed without note and players are left to guess if the change was accidental or if there was some good reason for it.

Bug still present in 1.21-pre2. 

I highly doubt this is intentional. The behavior was introduced in 1.20, and it breaks a number of automatic tree farms, mostly simpler designs since they tend to use observers closer to the core. Thanks to Lollopollq, we've identified the problematic bit of code. As they put it:

The behavior arises from the fact that TreeFeature#updateLeaves(LevelAccessor, BoundingBox, Set, Set, Set) will add log positions to the BitSetDiscreteVoxelShape used to then send shape updates around it

So any time a tree's canopy grows large enough that its leaves would grow in spots taken by pre-existing logs which are being observed, the observers will get extraneous updates. Generally in tree farms the purpose of these observers is to watch for movement when the core pushes a tree out, powering more pistons to move the logs further from the core. When they fire early due to this new 1.20 behavior, it is likely that they block the log movement instead, bricking the farm.