Incorrect distance field of view calculation for negative elevation

[HarelM]

This is in continue to the following PR which removed the elevation offset from terrain 3D: #1578

The situation can be see in the following diagram:

And can be reproduced in the local environment here when running npm run start-debug
http://localhost:9966/test/debug-pages/terrain-satellite.html#13.45/31.54164/35.33225/49.6/54

More info in the following comment:
#1578 (comment)

[prozessor13]

I did not analyzed in detail, but for documentation here is an example where the far-clipping plane is not correct:

http://localhost:9966/test/debug-pages/terrain-satellite.html#10.23/43.0154/9.1969/-97

[HarelM]

There's also this strange artifact around there, not sure it is related to this issue but might be a problem with the DEM:

http://localhost:9966/test/debug-pages/terrain-satellite.html#11.53/42.7216/9.247/173.4/73

[prozessor13]

I think the spikes are a data-problem.

And i think the far-clipping-plane error is because the dem-tiles have the bathymetry included (e.g. contains negative-coordinates, like the dea-sea).

[HarelM]

Assigning bounty M.
Link to parent Bounty: maplibre/maplibre#189

[typebrook]

I did not analyzed in detail, but for documentation here is an example where the far-clipping plane is not correct:

http://localhost:9966/test/debug-pages/terrain-satellite.html#10.23/43.0154/9.1969/-97

Attach image for this case:

[HarelM]

I think I manages to think about a solution while looking at the code.
There's a method of calculating the minimum DEM for a tile.
If we store that along side the elevation in the terrain object we can use that when calculating the FOV.
I'm not sure the code is correct, but it seems to behaving nicely in general.
I'll open a draft PR later today so that other can see the horrible code I wrote :-)
cc: @maxammann.

[HarelM]

I've opened a PR for feedback: #2858
cc: @acalcutt, can you check if this fixes the issue you've seen with the bathymetry stuff?

[birkskyum]

Where is the reproduction found now? The link suggested http://localhost:9966/test/debug-pages/terrain-satellite.html#13.45/31.54164/35.33225/49.6/54 doesn't appear to be working anymore

[HarelM]

I've took the page from an old version of the repo, added it, fixed the bug and removed the page as I'm not sure we are allowed to use that data in the examples.
You can take a look at the commits in my PR.

[maxammann]

If this is not urgent, then I can review the changes. Though right now there is not much time I can invest.

[HarelM]

I was looking for your input since you said you were looking in this code.
I generally would like to push this bug fix through...

[acalcutt]

I will also try to take a look at this soon. I just got back from a week of vacation.

[HarelM]

Removed bounty as this is mostly solved by the linked PR.
Any chance for a review on it? Would be great to have it merged and released...