Some servers running ubuntu in particular were reporting issues with teleport tubes not working.
On investigation, all tube entries were colliding as string.format("%d", ...)
was returning either -2^31 or 2^31-1 depending on system bit width,
causing hash entries to be overwritten.
This is possibly related to the use of C sprintf within lua.
Fix this by using %g instead to interpret as double without int conversion,
with a large enough number of digits such that all possible 2^48 values from
minetest.hash_node_position() can be correctly serialised.
(for situations where one might want to run
several closely-spaced parallel pipelines)
Also fixed a bug where a pipe "entry panel" would
report full of water if a pipe next to it (but not
connecting) was also full of water
Only tested under "classic" flowing logic mode.
to higher quality. Way more polys than before, but
without a measurable drop in FPS.
Also tweaked the textures for the pump and flow
sensor, and tweaked the low-poly models for them,
so that low-poly models use a different part of
the texture for the "light" on the side (allows
the higher-poly models to just put a glow there
and use a raised dome for the actual light).
Also fixed a Z-fighting issue with the grating on
the bottom of the pump (both versions)
This allows the raw offset to be visible inside that for-loop, which will be needed for calling the directionfn for directional neighbours to determine if they can flow in the given direction.
The "new flow logic" name was supposed to indicate that it was a continuation of the old branch by the same name, but it is beginning to become clear that it's not "new" any more and it might lead to confusion with "classic mode" flow logic while that still co-exists.
Explicitly name the subdirectory "pressure logic" to give a better idea of what goes in it, init.lua edited accordingly.
