compare OpenGL contexts with == instead of isEqual

[glowcoil]

I don't entirely understand why, but starting with a recent update to macOS, is_current has started returning false even when the OpenGL context is in fact current, tripping this assert in Glium (see this issue; I experienced it in a project I was working on as well). However, this seemed to fix it.

[goddessfreya]

Hey, does this solve the same problem as #1099? If so, which PR do you and @yvt think is better?

[yvt]

Both PRs fix the issue in the same way except for the difference in the ways of checking equality — pointer comparison (this PR) or - isEqual: (#1099 and master).

I think they make no difference in 99.99% of use cases, but might break in the other rare cases. Pointer comparison is in general more fragile against changes in the internal implementation. For example, since NSOpenGLContext is a wrapper of a low-level object, it's possible that currentContext could return a different object for the same underlying object.

Imagine the case where you have a NSOpenGLContext and for some reason you activate it by directly calling CGLSetCurrentContext. Now, what would currentContext be supposed to do? (a) It could maintain a global hash table that stores mappings to the original NSOpenGLContext. But then it would require a process-global critical section. Furthermore, it would still have to return nil for contexts not created via NSOpenGLContext, thus deviating from the documented behavior. (b) Or, it could simply construct a new NSOpenGLContext pointing the same CGLContextObject, which behaves exactly the same as the original object, except for its pointer identity.

[glowcoil]

Yeah, I'll concur that @yvt's is the more correct of the two.

Unfortunate that the original mistake wasn't able to be caught by the type system.

[goddessfreya]

#1099 has been merged into #1058 in favor of this PR, as you both seam to agree it's the more correct of the two. Thanks folks!