These changes are necessary for programs whose set of active attributes is not fixed at compile time by a template parameter pack, but rather varies based on the generated shader text at runtime. In such cases, the attribute location of a given named attribute may vary between instances of the same Program.
Previously, attribute bindings were implicitly associated with a location based on template parameter order, and -1 was used to indicate an inactive attribute. This left us unable to disable the appropriate attribute when it went from active to inactive. Now, the state tracker for bindings explicitly associates locations and state, and an empty optional is used to indicate an inactive attribute.
In addition, a gl::VertexArray class is now exposed, allowing more flexibility in the relationship between Programs, Segments, and attribute bindings. In this commit, that relationship does not change, but the subsequent commit adjusts it to match gl-js, reduce rebinds, and work around buggy VAO implementations.
VertexArray uses a pimpl idiom in order to support implementations that lack the VAO extension. In that case, all VertexArrays share global binding state, reflecting the platform reality in the absence of VAOs, while still providing a uniform API.
Introducing two new attributes to enable property functions for line-width (#9250) pushed the attribute count over GL_MAX_VERTEX_ATTRIBS on some devices. Now we selectively bind only attributes that are used, making it unlikely to surpass GL_MAX_VERTEX_ATTRIBS.
The current value is not passed to `SymbolSizeBinder::uniformValues`, so we shouldn't check `currentValue.isConstant()` in `SymbolSizeBinder::attributeBindings`. If it were true, then we might end up using attribute bindings that are appropriate only for a constant property, but uniform bindings that are appropriate only for a source or composite function.
Instead, just wait for a new bucket to be generated. This will happen automatically, since icon/text-size are layout properties.
Make sure all the implementations, currently Qt and Default, throw
an exception with error code CANTOPEN when opening a database that
doesn't exist without the CREATE flag.
ThreadLocal should not own the pointer it is managing because
the use case in Mapbox GL is to keep a pointer to a stack allocated
object, like:
```
MyObject foo;
threadLocal.set(&foo);
```
To keep consistency, it is required that we clear the managed
object before ThreadLocal gets destroyed by setting it to `nullptr`.
