This is mainly to get rid of IsContainerEx, because wtf.
The only reason why the categories were restricted was to prevent HasExtraSlot-objets from being found. Some C4D_Objects (or with wrong category in general) wanted to be found however. They are now found just like that.. like it was supposed to be.
We already require support for std::unique_ptr, which itself requires
support for rvalue references. As such, we know we can use rvalue
references, and thus don't have to keep carrying dead code around.
This reverts commit 8ef1f908cc. This seems to be
a performance bottleneck on some graphics cards and/or drivers. Clonkonaut
reports issues with a Radeon HD 7800.
This feature is not vital since textures can always be drawn such that they
tile properly at 512x512. While that particular problem with the Radeon card
could probably be worked around, it's just not worth it.
This will allow better graphics for the clouds movement when wrapping the landscape.
There is some potential for weird behavior; e.g. objects hidden just out of view outside the landscape. But we don't have any goals like Clonk Rage gold mine where you would have to find that nugget lying hidden in that place. Besides, the current behavior of just removing stuff even though it was just halfway out is worse.
On Windows, we need to know whether the library we're linking against is
a static library or a shared library import stub. If it's a shared lib
(and only then), we need to #define GLEW_STATIC.
Other OSes handle library imports differently and don't care.
When switching dialogs in editor mode, the (paint) tools dialog wouldn't
be properly cleared, instead only destroying the dialog window. Thus,
the next time it would be used, it'd try painting onto a device context
which did no longer correspond to any existing window, which made OpenGL
sad.
It would be set by state and then immediately reset in Default(). This had the effect that if you used the color picker on sky before ever selecting a background material, the wrong background would be used.
A texture array is conceptionally what should be used in this case. One
advantage of this is that we don't have to generate mipmaps ourselves but can
let the graphics driver take care of it. Same for selection of the mipmap
level. This would even allow to choose different mipmap levels for different
textures.
This is a somewhat experimental change since it makes OpenGL 3.0 a hard
requirement for OpenClonk. I expect that this is fine, but if this causes
failures during landscape creation on common hardware/drivers we should
revisit.
RejectCollect is not called on CreateContents or other forced entry methods but should still cause the lantern to attach. So move any side-effects to Collection2 instead of direct RejectCollect.
The lantern dummy had Timer=1 and wasn't doing anything. Use a simple overlay instead.
Note that interaction menus still show the lantern always with off-material. But that's a bug in the interaction menu.