mainmenu: Fix minetest.after timing

[SmallJoker]

local start_time = os.clock()

minetest.after(10, function()
    print("This function was called after ".. os.clock()-start_time .."s")
end)

Returns This function was called after 7.0620000000001s, This function was called after 0.45299999999997s or any other value.
This bug was caused by the time calculation of the time last, which is initialized to zero. The time new instead starts with any value.

Tested this pull and got This function was called after 10.469s (some loading delay included).

[sfan5]

👍

[sofar]

os.clock is unreliable, you can't use it for testing the timer. use minetest.get_us_time() instead

[sofar]

I've spent a few nights to make sure that the current code was in order and tested with thousands of timers etc.. I'll review this since I don't want it to regress to the bad timer performance we had before ad884f2

[sofar]

NVM, I see what's going on here. Yes, this is entirely right, thanks for spotting this edge-case.

Please merge.

[sofar]

This seems like a better solution:

diff --git a/builtin/game/misc.lua b/builtin/game/misc.lua
index 4f58710..de41cfc 100644
--- a/builtin/game/misc.lua
+++ b/builtin/game/misc.lua
@@ -6,7 +6,7 @@

 local jobs = {}
 local time = 0.0
-local last = 0.0
+local last = core.get_us_time() / 1000000

 core.register_globalstep(function(dtime)
        local new = core.get_us_time() / 1000000

[SmallJoker]

@sofar Depends what time matters. Either it's the time when Lua loads the scripts or after it loaded all and begins to call the global functions.
Both ways are acceptable IMO

[sofar]

I'd prefer the time initialization method as it removes an entire branch.

[sofar]

My suggested fix (proper initialization) is in #4009

[paramat]

#4009 is merged.