This one crashes. Turns out the 169031fb7b
contained a random temporary test state instead of the real solution
(and so the comment didn't even match the code, it should have been
resetting that to 0). That also made some tests fail with DSA disabled,
but none of the tests were actual Mesh tests, just accidentally hitting
the problematic code path.
I took the opportunity to look at this more closely and investigate
*why* this failed -- turns out, in setIndexBuffer(), I was resetting
the state tracker to a VAO state that was about to be set in the very
next step, and then, when doing that next step the state tracker
"optimized away" the state change because it thought it was already done
(even though it wasn't). The new test in MeshGLTest covers this
particular case.
What's left is *a lot* of places taking monstrous
std::vector<std::reference_wrapper> and that can't be changed to
std::vector<Containers::Reference> in a source-compatible way. Even that
would be only a temporary change, since the goal is to fully avoid
dependency on STL in those cases.
The final version of these APIs should take
Containers::ArrayView<Containers::Reference> and be implicitly
convertible froom e.g. std::vector<Containers::Reference>. That's
definitely possible, but not in time for 2019.01, so instead of forcing
users to temporary pass a `{vec.begin(), vec.size()}` everywhere instead
of just `vec`, I'm rather keeping these APIs intact.
Since Range1D is now used all over Animation, the vector made it very
annoying to use. That's fixed now. This is a backwards-incompatible
change, but I don't expect the 1D range to be used much, mainly because
it was so shitty to use. Generic code that needs a vector can always
cast to it, like this:
Math::Vector<dimensions, T>{range.min()}
Test for the constructor from pair is no longer accepting pairs of 1D
vectors. I have no idea what I meant by that test case (it's testing the
same thing twice), so I removed one of these.
The change done in 680144f1c5 was not
properly handling these cases:
* Mesh(NoCreateT) and wrap() were not constructing the internal vector,
which blew up when move-assigning another instance.
* ~Mesh() was not destructing the internal vector if the VAO ID was
zero or non-owning wrap() was used.
Strangely enough none of these were causing *any* problems for me on
Linux (even ASan was totally quiet and due to the unfortunate
combination of bugs even when I assigned totally random data to the
storage vector). This however blew up on MSVC, assuming there the
implementation is more checked.
Because it's possible to construct Mesh with no GL context available,
the move construction and destruction needs to avoid accessing Context
unless really necessary (it would be also unclear which type of vector
should be constructed if we have no context).
Extended the tests to handle hopefully all the cases.