Problems with an unusual NSOpenGLView setup

Question

Hi,

This is a bit long, so feel free to skip to the numbered parts if you want to avoid the lengthy intro.

I'm trying to set a subclassed NSOpenGLView in an unusual way and I am running into some problems. Basically, I am writing a program to perform a bioengineering simulation for my PhD and I need to be able to compile it under both MacOSX and Unix (my machine is a Mac, but the sim will eventually run on a more powerful Unix machine). Since the code will get longer and longer over the next year and a half I'd rather not have to keep track of two completely different versions of the program. So, I'm hoping to be able to compile the ObjectiveC code under Unix by avoiding ObjectiveC-2.0 and keeping the interface optional (it will mostly be there to perform setup before the long simulations and monitor things for the short ones during development).

The current version works well without the interface - the simulation is performed correctly and the program is capable of rendering OpenGL frames and exporting them into image and video files without any problems. Since I am now adding the interface (right now just a simple window with an NSOpenGLView subclass and a "start" button") on top of that (so that I can run the code with an alternate version of main() without it) I have to "wire" OpenGL together in a weird way, since the drawing code is not in the drawRect function, or even anywhere in the subclassed view, but instead in the "basic" program.

What I've done so far is this:

1. The main program (using an object called "Lattice") performs all the simulations and rendering, correctly outputing images and video to files. This also contains the NSOpenGLContext and calls [renderContext flushBuffer];

2. A subclass of NSOpenGLView called PottsView contains an instance of a lattice, which is initialized together with the view like this:

   • (id)initWithFrame:(NSRect)frame { if(![super initWithFrame:frame]) return nil;

     // code frameSize.width = WIN_WIDTH; frameSize.height = WIN_HEIGHT; [self setFrameSize:frameSize];

     init_genrand64(time(0));

     latt = [Lattice alloc];

     if (SEED_TYPE) { [latt initWithRandomSites]; } else { [latt initWithEllipse]; }

     [[latt context] makeCurrentContext];

     return self; }

drawRect() is empty.

1. PottsController is the object instanced in the InterfaceBuilder which connects the start button to the view. The start button simply tells the lattice to run for a number of steps.

Now, pressing start results in the simulation running correctly (i.e. output to files and terminal), but the PottsView is not working correctly. It remains white, but if I cmd+tab parts if it change to sections of a rendered frame. Same if I press Expose (F3).

I've tried several combinations of flushing, setNeedsDisplay, etc, but frankly speaking I'm lost. I haven't done any programming before this April and with this being (as far as I can tell) a completely backwards way of using NSOpenGLView I'm out of ideas. I'm hoping someone can suggest how I can make the current setup work or how to completely rewire the program (while still keeping the interface optional).

Thanks, Bartek

Answer 1

It's not clear how you think that you have 'wired' the context and the view together. You can have as many openglContexts as you like - just by drawing into one won't make it's contents show up in a random NSOpenGLView. Apologies if i have missed something.

NSOpenGLView is a fairly simple subclass of NSView that creates the context and pixel format. As you already have those you can do away with NSOpenGLView and use a custom NSView subclass.

You should look at this instruction.. http://developer.apple.com/library/mac/#documentation/GraphicsImaging/Conceptual/OpenGL-MacProgGuide/opengl_drawing/opengl_drawing.html

To draw to the screen you must flush the graphics context from -drawRect:

This will block the main thread while the gpu processes your instructions, this could be a problem if you have many instructions. It also can not happen more than 50fps.

If you are already rendering your frames to files woudn't you be better observing the output directory and drawing the image each time a new one is added, no opengl required?