2021-11-13 23:29:08 +01:00
|
|
|
|
local test_minetest_find_nodes_in_area_can_count = function(dtime)
|
|
|
|
|
-- This function tests that minetest.find_nodes_in_area() can
|
|
|
|
|
-- count nodes correctly. If it fails, the engine can not be
|
|
|
|
|
-- trusted to actually count how many nodes of a given type
|
|
|
|
|
-- are in a given volume. Yes, *it* is bad if that happens.
|
|
|
|
|
--
|
|
|
|
|
-- If you are looking at this function because it executes at
|
|
|
|
|
-- startup and crashes the game, by far the most stupid thing
|
|
|
|
|
-- you could do is disabling it. Only an absolute moron would
|
|
|
|
|
-- disable tests that ensure basic functionality still works.
|
|
|
|
|
--
|
|
|
|
|
-- Experience has taught me that such warnings are mostly not
|
|
|
|
|
-- taken seriously by both Minetest mod & core developers. As
|
|
|
|
|
-- there are very few situations in which someone would read
|
|
|
|
|
-- the code of the function without a crash, you are probably
|
|
|
|
|
-- asking yourself how bad it can be. Surely you will want an
|
|
|
|
|
-- example of what will break if this test breaks. The answer
|
|
|
|
|
-- to this simple question is equally simple and consists of a
|
|
|
|
|
-- heartfelt “What the fuck did you say, you stupid fuckwad?”.
|
|
|
|
|
--
|
|
|
|
|
-- Alrighty then, let's get it on …
|
|
|
|
|
|
|
|
|
|
local pos = { x=30999, y=30999, z=30999 }
|
|
|
|
|
-- You think there is nothing there? Well, here is the thing:
|
|
|
|
|
-- With standard settings you can only see map until x=30927,
|
|
|
|
|
-- although the renderer can actually render up to x=31007 if
|
|
|
|
|
-- you configure it to. Any statements given by minetest core
|
|
|
|
|
-- devs that contradict the above assertion are probably lies.
|
|
|
|
|
--
|
|
|
|
|
-- In any way, this area should be so far out that no one has
|
|
|
|
|
-- built here … yet. Now that you know it is possible, I know
|
|
|
|
|
-- you want to. How though? I suggest to figure the technique
|
|
|
|
|
-- out yourself, then go on and build invisible lag machines.
|
|
|
|
|
|
|
|
|
|
local radius = 3
|
|
|
|
|
local minp = vector.subtract(pos, radius)
|
|
|
|
|
local maxp = vector.add(pos, radius)
|
|
|
|
|
local nodename = "air"
|
|
|
|
|
local c_nodename = minetest.get_content_id(nodename)
|
|
|
|
|
|
|
|
|
|
-- Why not use minetest.set_node() here? Well, some mods do
|
|
|
|
|
-- trigger on every placement of a node in the entire map …
|
|
|
|
|
-- and we do not want to crash those mods in this test case.
|
|
|
|
|
-- (Voxelmanip does not trigger callbacks – so all is well.)
|
|
|
|
|
--
|
|
|
|
|
-- And now for a funny story: I initially copied the following
|
|
|
|
|
-- code from the Minetest developer wiki. Can you spot a typo?
|
|
|
|
|
-- <https://dev.minetest.net/index.php?title=minetest.get_content_id&action=edit>
|
|
|
|
|
local vm = minetest.get_voxel_manip()
|
|
|
|
|
local emin, emax = vm:read_from_map(
|
|
|
|
|
minp,
|
|
|
|
|
maxp
|
|
|
|
|
)
|
|
|
|
|
local area = VoxelArea:new({
|
|
|
|
|
MinEdge=emin,
|
|
|
|
|
MaxEdge=emax
|
|
|
|
|
})
|
|
|
|
|
local data = vm:get_data()
|
|
|
|
|
for z = minp.z, maxp.z do
|
|
|
|
|
for y = minp.y, maxp.y do
|
|
|
|
|
local vi = area:index(minp.x, y, z)
|
|
|
|
|
for x = minp.x, maxp.y do
|
|
|
|
|
data[vi] = c_nodename
|
|
|
|
|
vi = vi + 1
|
|
|
|
|
end
|
|
|
|
|
end
|
|
|
|
|
end
|
|
|
|
|
vm:set_data(data)
|
|
|
|
|
vm:write_to_map()
|
|
|
|
|
local npos, nnum = minetest.find_nodes_in_area(
|
|
|
|
|
minp,
|
|
|
|
|
maxp,
|
|
|
|
|
{ nodename }
|
|
|
|
|
)
|
|
|
|
|
local nodes_expected = math.pow( 1 + (2 * radius), 3 )
|
|
|
|
|
local nodes_counted = nnum[nodename]
|
|
|
|
|
local nodes_difference = nodes_expected - nodes_counted
|
2021-12-15 12:55:33 +01:00
|
|
|
|
-- Originally, there was an assertion here that made the game
|
|
|
|
|
-- crash at startup if Minetest forgot how to count. This was
|
|
|
|
|
-- originally intended to avoid buggy engine releases, but it
|
|
|
|
|
-- mostly made people upset and hindered debugging. Also, the
|
|
|
|
|
-- assertion contained no error message hinting at the reason
|
|
|
|
|
-- for the crash, making it exceptionally user-unfriendly. It
|
|
|
|
|
-- follows that a game or mod should only assert on behaviour
|
|
|
|
|
-- of the Lua code, not the underlying implementation, unless
|
|
|
|
|
-- engine bugs are bad enough to permanently corrupt a world.
|
|
|
|
|
if ( 0 ~= nodes_difference ) then
|
|
|
|
|
minetest.debug(
|
|
|
|
|
"minetest.find_nodes_in_area() failed to find " ..
|
|
|
|
|
nodes_difference .. " nodes that were placed. " ..
|
|
|
|
|
"Downgrading to Minetest 5.4.1 might fix this."
|
|
|
|
|
)
|
|
|
|
|
end
|
2021-11-13 23:29:08 +01:00
|
|
|
|
end
|
|
|
|
|
|
|
|
|
|
minetest.after( 0, test_minetest_find_nodes_in_area_can_count )
|
|
|
|
|
|
|
|
|
|
local test_minetest_find_nodes_in_area_crash = function(dtime)
|
|
|
|
|
-- And now for our feature presentation, where we call the
|
|
|
|
|
-- function “minetest.find_nodes_in_area()” with a position
|
|
|
|
|
-- out of bounds! Will it crash? Who knows‽ If it does, the
|
|
|
|
|
-- workaround is not working though, so it should be patched.
|
|
|
|
|
|
|
|
|
|
local pos = { x=32767, y=32767, z=32767 }
|
|
|
|
|
-- Note that not all out of bounds values actually crash the
|
|
|
|
|
-- function minetest.find_nodes_in_area()“. In fact, the vast
|
|
|
|
|
-- majority of out of bounds values do not crash the function.
|
|
|
|
|
|
|
|
|
|
local radius = 3
|
|
|
|
|
local minp = vector.subtract(pos, radius)
|
|
|
|
|
local maxp = vector.add(pos, radius)
|
|
|
|
|
local nodename = "air"
|
|
|
|
|
local npos, nnum = minetest.find_nodes_in_area(
|
|
|
|
|
minp,
|
|
|
|
|
maxp,
|
|
|
|
|
{ nodename }
|
|
|
|
|
)
|
|
|
|
|
-- That's all, folks!
|
|
|
|
|
end
|
|
|
|
|
|
|
|
|
|
minetest.after( 0, test_minetest_find_nodes_in_area_crash )
|