So... I've had a look at my code, in an attempt to understand what it actually does (it's old, and was only written as a hack to try and view some old OBJ models in a stereoscopic way...)

Frankly, I have no idea why it works, but it appears to, on Win32 and macOS, and I remember it working on X11 at the time (have not checked today.)

So, the first thing is that the only Mac I have now is old, running 10.13 or so IIRC, so not all that current, and can not run any macOS more recent.

Next, all platforms (including OSX) seems to be using GLEW to access the "higher" API levels. (I might be wrong about this though!)

The OBJ models are rendered with some old code that uses GL1-style methods, these are written into buffers, which are then read into various other textures to apply the (anaglyphic) colour filtering and anti-aliasing, before getting merged together in yet more textures and rendered to the display. These later stages are using later GL API to do their magic, but I'm looking at the code and I do not understand what I did, nor have any idea why... (I also can't see why I needed so many intermediate buffers to get to the  end result.)

Anyway: The end result is that a picture does appear, and when viewed through the anaglyph glasses the stereoscopic effect works (more or less, as well as anaglyphic rendering, anyway) but...

I suspect the crux is that the GL-1.x and GL-3 stuff never actually happens in the same context (my recollection was that it did, that appears not to be true) - the GL-1.x stuff is effectively written out to a pixelmap and then read in by the GL-3 code.

Anyway, what I have is a mess and may only work by random chance!

It certainly does not look like something we should strive to reproduce.