The templated version had the unfortunate "feature" of not being able to
figure out the type when an array view or a C array got passed to it.
That led to worse-than-ideal UX and even though it's now a bit more
verbose on the implementation side, it's the preferred solution.
No matter how broken iOS is in CMake 3.6, $<CONFIG> seems to work there,
so reducing the amount of code and putting the configure into a single
place independently of what generator or what system/build is used.
Compared to current state it always adds Debug/configure.h instead of
putting it directly to the ${CMAKE_CURRENT_BINARY_DIR}, but the
alternative would be some CMake branching again and I just removed that,
so no.
This also prepares everything for plugin libraries being put into a
central place -- the config files don't depend on their location
anymore.
Also move it to a new GenerateNormals.h header so we can easily add
generateSmoothNormals() to it. The old API and header is deprecated and
will be removed in the future. I can't be bothered rewriting the old
code using the new thing, so it's preserved there as a mausoleum until
it gets finally nuked from the orbit.
Provide a version taking StridedArrayView instead, taking an arbitrary
view and also arbitrary index type. Also introduce duplicateInto() which
doesn't even allocate, but rather writes the output to pre-existing
location.
All functionality is now available through free functions. The classes
are now just deprecated wrappers and/or typedefs and will be removed in
some future release.
As with Corrade, this is not exactly backwards compatible, but for
common use case without OBJECT libraries this should not be a problem.
In any case, recreate the build dir and update your copy of all
Find*.cmake modules to avoid weird things happening.
User-facing changes:
* Documentation of all Find*.cmake modules converted to
reStructuredText to follow official CMake guidelines.
* The newfangled way to use the libraries is to link to Magnum::Shaders
instead of adding ${MAGNUM_SHADERS_INCLUDE_DIRS} to include path and
linking to ${MAGNUM_SHADERS_LIBRARIES}.
* The old ${MAGNUM_*_LIBRARIES} are deprecated and now just expand to
Magnum::* target. Use the target directly. These are also enabled
only when building with MAGNUM_BUILD_DEPRECATED.
* The old ${MAGNUM_*_INCLUDE_DIRS} are removed as the Magnum::* targets
cover these too.
Internal changes:
* Global state such as include_directories() was replaced with
target-specific settings.
The final release doesn't have the issue with non-explicit
default std::vector constructor. Most of the conflicts resulted from
Mesh::Primitve -> MeshPrimitive refactoring.
This reverts commit c2ad09706e.
Conflicts:
src/Magnum/Primitives/Capsule.cpp
src/Magnum/Primitives/Circle.cpp
src/Magnum/Primitives/Crosshair.cpp
src/Magnum/Primitives/Cylinder.cpp
src/Magnum/Primitives/Icosphere.cpp
src/Magnum/Primitives/Implementation/WireframeSpheroid.cpp
src/Magnum/Primitives/Line.cpp
src/Magnum/Primitives/Plane.cpp
src/Magnum/Primitives/Square.cpp
src/Magnum/Primitives/UVSphere.cpp
src/Magnum/SceneGraph/Object.hpp
src/Magnum/Text/GlyphCache.cpp
src/Magnum/TextureTools/Atlas.cpp
src/Magnum/TextureTools/Test/AtlasTest.cpp
The point of this change is to allow greater flexibility and reduce
confusion.
When instanced meshes are implemented, MeshTools::interleave() can be
used for creating interleaved buffers with per-instance data and then
the call to Mesh::setCount() will be harmful and/or confusing, becuase
the user would in fact want to call Mesh::setInstanceCount() instead.
Similarly, MeshTools::compressIndices() can be used to create index
buffer for more than one mesh.
GL 4.4 has ARB_buffer_storage, which (in relatively distant future) will
mean that the current way of Buffer::setData() will be deprecated in
favor of Buffer::setStorage(), similarly as Texture::setStorage()
replaced Texture::setImage(). Thus any function which calls
Buffer::setData() internally is not future-proof.
The old MeshTools::compressIndices() and MeshTools::interleave()
overloads are marked as deprecated and will be removed in future
release.
Basically a version of interleave() which takes existing buffer instead
of creating new zero-initialized one. The last bit needed for real
runtime usage.
Everything are now pure functions, no shared state, no useless functor
class wrapping. Still needed to wrap attribute count and stride
computation functions in struct to make proper overload resolution
working without the need to deinline the functions (they have cyclic
dependencies).
It now returns new index array instead of operating on already existing
one and also custom vector size is removed. The internal implementation
is now much simpler and cleaner. The old way (but still without custom
vector size) is now alias to the new implementation, is marked as
deprecated and will be removed in future release.
Renamed the parameters and reworded the documentation so it doesn't talk
about vertices, but rather about generic floating-point vector data.
Vladimír Vondruš