Given blue ice introduced in 1.13, it seems that there is good evidence that this behavior is now intended. Perhaps this issue can be closed?
@Matt Can you provide evidence that the frequency is consistent with previous snapshots? I've been exploring in spectator mode on that seed and have only found one nearby outside of a mineshaft.
@Kumasasa In my opinion, the focus of the issue is that dungeons appear to be spawning less frequently. I think the correct approach to confirming this bug would be:
load n random seeds
inspect m random chunks in each seed
count the total number of zombie, skeleton and spider spawners in each inspected chunk
calculate an average number of dungeons per chunk
This all sounds easy, but I don't know of a way to do this efficiently and accurately in game. I'd be surprised if the Minecraft test suite didn't have automated tests to do something like this already.
Here is a seed: 7015474738862924540
And here are the coordinates: /execute in overworld run tp @s 424.78 41.00 -466.80 -176.19 24.85
You will notice that somewhere between 18w09a and 18w10d, the dungeon at that position (in 18w09a) stops generating. In general dungeons seem a bit more rare in the later snapshots than they were previously.
There are other changes in generation as well. In 18w19a you can go to those coordinates, turn around and follow the cave a few blocks to find a lava source that is not there in any previous snapshots. Follow it a bit further until you reach a larger opening and you'll see that a lava source is missing in 18w19a that is there in all previous snapshots. This kind of variation might be expected though.
I disagree. Having read that entire issue, it describes entirely different behavior than what I am describing. That issue concerns multiple clicks being required and the cursor being visible when it should be invisible, and this one concerns the cursor being invisible when it should be visible. Furthermore, that issue describes behavior that has been around for years, whereas this behavior is new as of 1.17.
I have also included a workaround in my description of this issue, which is different than the workaround described in MC-145948.