Don't crash when forest mapgen requests groundcover from a biome with no groundcover

[akrieger]

Summary

None

Purpose of change

Fixes #72877, #72872. #72812 exposed that some forest mapgen code assumes adjacent biomes have groundcover. This is not true.

Describe the solution

Patch the code to fallback to regional groundcover like the other fallback cases use, in case the chosen adjacent biome in the feathering algorithm does not have groundcover of its own.

Also, noticed the adjacent biomes array wasn't being fully populated, with the last entry being skipped due to an off by one error in the initialization loop. Fix that too.

Describe alternatives you've considered

Testing

Use one of the repro saves from the issue, it doesn't crash anymore when mapgenning an innawoods cave tile.

Additional context

[akrieger]

So to test the crash, what I did is I moved cave_innawood out of the first slot for forest terrains, and the crash stopped happening even on the save that already had a cave generated. There's something funky that arrangement is exposing, but I can't take the time to look harder so I'm just monkeypatching the crash away.

[Procyonae]

There's something funky that arrangement is exposing, but I can't take the time to look harder so I'm just monkeypatching the crash away.

It probably expects the first one in the list to be the oter the others are mimicking somewhere along the line, I didn't really think twice when alphabetising them.
I'm going to bed now but I'll have a look in the morning.

[mqrause]

I got a theory: cave_innawood is a terrain that does not exist in vanilla dda. So because that is loaded first, it can't use the actual region settings from vanilla and creates a new empty biome, which causes mapgen to trip over it.

Cataclysm-DDA/src/regional_settings.cpp

Lines 175 to 177 in a1e69f1

	std::string first_ter = forest_biome_ter_keys.get_string( 0 );
	load_forest_biome( forest_biome_jo, forest_mapgen_settings.unfinalized_biomes[first_ter],
	overlay );

[akrieger]

squint

I'll change this to a debugmsg then which explains what needs to happen in the json, and fix the json for innawood.

[akrieger]

The other option is changing it so the forest biome regional setting loading code always prefers to inherit from the forest biome of the name of the region, if its available and in the terrains list (which seems to be always?), otherwise to use the first biome.

[mqrause]

The other option is changing it so the forest biome regional setting loading code always prefers to inherit from the forest biome of the name of the region, if its available and in the terrains list (which seems to be always?), otherwise to use the first biome.

From what I can see, the loading code is kinda sketchy because it treats regional settings and overlays the same. But overlays can skip all the required checks, because they expect a regional setting to already exists to inherit all the required data from. Except they don't in cases like this. So it should be forbidden to just create a new forest_biome object when loading an overlay.

[akrieger]

oic, because forest_mapgen_settings.unfinalized_biomes[first_ter] will create an empty biome if the string is novel. And this function is called from both loadpaths. I can throw an error instead if overlay = true and the first terrain is not seen yet.

[Procyonae]

I had a look yesterday but couldn't determine what the best way to actually make it work sensibly is, sorry for radio silence tho I were having an off day