paramat / watershed

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

water, water everywhere

cyisfor opened this issue · comments

http://imgur.com/aB7rV8G

I knew was dealing with the devil when I tried seed 666, and at somewhere around (1800, 0, 1000) everything went below water and never came up. Tried at x = 2000, 5000, 10000, 20000, no matter where I looked it was nothing but water (or ice with a flat layer of snow on top for the snow biome). Some weird glitch with that seed specifically, right? No, I tried changing the seed. Some seeds delayed the effect, but by x=20000, it was always forever under water, no matter what seed. The log was spitting out things like
[watershed] chunk minp (20048 -32 -20112)
[watershed] 675 ms
[watershed] chunk minp (20048 -32 -20032)
[watershed] 732 ms
[watershed] chunk minp (19968 -32 -20112)
[watershed] 755 ms
[watershed] chunk minp (19968 -112 -20112)
[watershed] 633 ms
[watershed] chunk minp (19968 -112 -20032)
[watershed] 627 ms

Y always being negative, and always either -32 or -112. (0x20 and 0x70 respectively.) Ended up making the ocean floor slightly hilly, but pretty consistently around -60.

I wasn't using any real funny mg_flags, and no extra mods, just v7 mapgen, and watershed. It looked real cool near 0,0,0, but out at x=20000, z = 1000 it was just blah ocean.

Now here's where it gets weird. I figured it was a consistent radius by flying around at (20000,0,20000), (20000,0,-20000), (-20000,0,-20000) but then I went to (-20000,0,20000) and I was buried in the soil of a huge jungle. There was no endless ocean. Everything was normal. Whatever was causing the land to drop low forever doesn't seem to work when x is negative and z is positive?

I've noticed that perlin noise often has long ridges or valleys aligned with the world grid, i suspect these can appear in the same place even with different seeds.

It would be nice if not all seeds produced a single central continent on a world largely filled with water. I wouldn't mind if seeds existed that weren't always plunging the land height to abyssal depths out far away from (0,0)

This isn't an issue of a particular seed making world sized features, but of every seed making every map underwater in all quadrants except the southwest one.

One thing that will help with this is to add a random offset to the xlscale perlinmap at line 241:
local nvals_xlscale = minetest.get_perlin_map(np_xlscale, chulens):get2dMap_flat({x=x0+2994, y=z0+7133})
Those 2 random offsets should be somewhere between 0 and 8192.
Also, increasing persistence of xlscale noise parameters to 0.4 should help a little.

Also raising octaves by 1 and/or increasing persistence to 0.4-0.5 should help.

It sounds fascinating. I sort of haven't been able to play minetest much due to migranes but I'd like to know if it produces good results!

I recommend buying a flicker-free LED monitor then, most LED monitors dim the LEDs by PWM (pulse width modulation) strobing, some people are sensitive to this. Look for a LED monitor with DC (direct current) LED dimming, they aren't more expensive either, just a limited choice.