We define the blast intensity as the square of the tnt_radius, divided
by the square of the distance to the explosion center, where distance
is limited to 1 at the lower end.
When destroying nodes, we calculate the intensity for each node, and
only destroy the nodes when the intensity is 1.0 or larger. To avoid
perfectly spherical explosions, we make sure to retain a randomness
factor of 20%. This will make explosion edges jagged and not smooth,
but not too much.
We pass the calculated intensity to on_blast() functions as well,
except we take the jitter here out and make sure it's always 1.0
or larger.
We apply a log scale to the size of the stacks ejected, so that
in larger explosions we are getting larger stacks. For normal r=3
explosions, this gives stack sizes ~6-7 or so, but for r=10 explosions
it could end up giving stacks of 25+.V
The drops list already has quantities, so let's just select the one
with the highest quantity from it, and use that as tile. Fallback
tile will therefore only be used if explosion happens in air. Oh well.
We add a dirt-like particle (drawn from scratch, uses some
colors from default_dirt's palette) to spawn many particles
that have collision enabled around the center of the blast.
This has the effect of obscuring the center of the blast, as
that is a painfully visible empty area when the explosion happens,
as there's only a little spark.
The dirt particles bounce around the walls and floor a bit,
and disappear rapidly, well before the smoke particles disappear.
This is a nice visual distraction that obscures the sudden
appearance of the gaping hole, and makes it a whole lot more
believable.
But not too much.
TNT is a bit underwhelming at the moment. We can make it a bit
more interesting by ejecting not just one or two itemstacks,
but a bunch of them. This code splits up the drops into
separate itemstacks that are 2-5 items together, which
results in generally roughly 10 itemstacks being ejected.
Since now we have multiple ejecta, it makes sense to tune
the ejecta velocities a bit better to get the appearance of
an actual explosion better. The items will not all start
with the same vertical velocity, since that would look
like fireworks. Instead we give them all a different vertical
speed.
If the node is special and has an on_blast() handler, we need
to call it instead of getting node drops manually. However, we
do want to know if drops should be added for the special nodes,
so we modify the on_blast() handler code to allow the nodedef
handlers to pass back itemstacks. This could be used by e.g.
the doors mod to drop door items after a blast.
Since this API is documented in lua_api.txt, a separate PR will
be incoming to update the on_blast() documentation.
Some people borrowed the creative code for their sub-games with an exclusive attribution to celeron55.
This is frustrating since I've largely rewritten, redesigned and carefully maintained this mod for the last months.
I expect to be credited.
Currently any doors viewed from underwater will disappear but removing the line 'use_texture_alpha = true,' seems to fix this. Thanks to Thomas-S for finding this glitch.
The boat model had over 1700 tris (!) before this redo. I've reduced
the amount of tris to ~150, which is very reasonable given that there
are almost 45 faces to this model.
I've also spent a good hour re-UV unwrapping the entire model which
has a huuuge impact on the boats' appearance. As much as possible,
the boat now looks like it's made out of actual blocks of wood,
and I've even attempted to make the grains connect around edges,
appear in the same pattern and spacing, and generally make it look
like it's just a nice crafted thing out of several pieces of wood.
I've had to tweak the rudder part to make the texture actually have
square texture pixels. I did that by varying the vertical position of
the botton of the rudder handle and the top of the rudder bottom, which
worked well. I also had to 'slice' the rear face of the boat to prevent
a strange texture tear, probably due to non-flat surface somewhere,
but I couldn't spot the issue there anymore after adding 2 extra edges.
This looks totally like a new boat now.
- http://i.imgur.com/stiVzsa.jpg
If LVM or some other nonmetadata method is used to place a door,
then metadata is missing that tells us whether the door is left
or right-hinged.
However, we can detect that nodemeta is missing and see if the node
name is _a or _b. In the case of _a, nothing needs to be done and we
can just open the door. In the case of _b we assume the door is right
hinged, and tune the state nodemeta value so that the door opens the
right way. This all of course assumes that the schematic method places
the doors *closed* by default, which is reasonable.
- disallow placing beds in protected areas
- fix rotation of beds(broken after 41c2b2ae)
- allow using others' beds, but don't change spawn location
Fixes#953. #943 isn't something I think was ever implemented, and
this does a fair job of addressing the main concern (spawning in
others' houses)
We were cleverly attempting to use an airlike node as the
top half of the doors, but as airlike nodes are not walkable,
falling nodes would not stop falling and thus remain an entity
stuck on top of a door.
After inspecting the builtin/game/falling.lua code, I considered
the remaining options: (a) revert doors such that the top part is
actually the door, (b) play with nodedef fields and see if other
flags may work, or (c) modify the hidden door part to another
drawtype that properly prevents this issue.
(a) seemed way over the top for now, although it would solve the
issue, it would cause a rewrite of most of the code including the
old-door-conversion.
(b) turned up nothing.
(c) turned out to be relatively simple.
So, here's the implementation where I turn the hidden door top
into a tiny, non-targetable but walkable nodebox that is entirely
inside the door hinge. It's entirely transparent, so you still
can't see it, can't hit it, nor can you place anything in it or
make liquids flow through it. The top part is placed in the right
position on placement and not touched further.
Falling nodes will properly stop on top of these doors. I've
adjusted the door conversion code to properly account for the
issue as well, so the only thing remaining is people who have
been running a git branch - those can upgrade by re-placing the
door.
the ```group:cracky``` group contains all sorts of odds and ends
nodes that we shouldn't connect to. There are a few nodes that
walls now no longer connect to, that probably should get
```group:stone``` added, or something similar, though.
I've created a modified B3Dexport.py version that automatically strips
the embedded texture link to external texture files. These links were
causing the engine to spew "can't find character.png" messages on the
console, but were harmless due to texture loading being done by the
client side and not through irrlicht.
I previously moved character.png to /textures/, which is wrong. I now
understand that character.png was in the same folder as character.blend
simply to make blender load the texture from the embedded linkage
automatically. Nothing more, nothing less.
Subsequently the character.png file should just sit in convenience
in the /models/ folder with the blend file, and not in the textures
file. This patch moves it back. And yes, minetest does load the
character.png from this path.
Tested with nodebreaker, fire.
If called from lua, minetest.remove_node() calls on_destruct() callbacks
before the map is actually updated. This means that we can't look at the
map data to determine if we're done cleaning up adjacent nodes, and we
have to stop recursing some other way.
There's no data we can pass around through functions that would survive
scope to a secondary on_destruct() callback, so we have to maintain
local state somewhere in the mod namespace.
In this case, we keep a bitflag. The bitflag is set to "true" by
default. On the first half removal, the flag is flipped and afterwards
we remove the other half node. When the on_destruct for the other half
is running, it's value is false and we flip it back to true without
removing the other half node.
This thus prevents recursing.
To facilitate easier finding of the bed partner, we tell our on_destruct
whether we're a top or bottom half node through a passed flag.
Now that the top is diggable, we just need to assure that it drops a
bottom bed part.
