Drawing to an NSView while inside an event handler - macos

Question

I've tried everything I can think of, but no luck.
I have a project that is essentially an 'interpreter' of a programming language. So events get passed to the interpreter loop, the interpreted code does whatever (in this case, updates a memory bitmap), then the interpreter loop returns and the event handler returns. Eventually the app's drawRect gets called and the memory bitmap gets drawn to the NSView. This is all fine MOST of the time.
But... there's a few situations where the interpreted 'code' wants to cause a short animation, and does that by updating the memory bitmap, usleep()'ing for a few milliseconds, updating the memory bitmap, usleep()'ing, etc. The animation takes a little less than a second, so the thread-blockage shouldn't be an issue.
The problem is that NONE of the animation shows, and the screen doesn't get updated until after the interpreted 'code' ends, and the event returns.
The sleep function that gets invoked when the interpreted code indicates it wants to sleep looks like this:
void KSleep(DWORD tm) {
if( [pView lockFocusIfCanDraw] ) {
inSleep = true;
[pView setNeedsDisplay:YES];
[pView display];
[pView drawRect:NSMakeRect(0, 0, pView.frame.size.width, pView.frame.size.height)];
[pView unlockFocus];
}
usleep(tm*1000);
}
'inSleep' is a global variable I set up for testing purposes, 'pView' is a global NSView* to the window's only view. NOTE: Yes, some of the above code is redundant, I'm just including it to show that I've tried numerous combinations of trying to indicate to the OS that the view is dirty and to update it. None of them have worked.
The drawRect code (removing all of the code that does the run-of-the-mill blitting of the bitmap) looks like this:
-(void)drawRect:(CGRect)rect {
CGContextRef context = [[NSGRaphicsContext currentContext] graphicsPort];
CGContextSaveGState(context);
if( inSleep ) CGContextSetRGBFillColor(context, 0.0, 0.0, 1.0, 1.0);
else CGContextSetRGBFillColor(context, 1.0, 1.0, 0.0, 1.0);
CGContextFillRect(context, CGRectMake(0,0,200,200));
CGContextRestoreGState(context);
}
So:
1) The -mouseDown() event occurs and it calls the interpreter with the event.
2) The interpreted code draws to the bitmap (I'm ignoring that here, as it's not important to the working or not working of the screen update), and invokes 'sleep.'
3) The interpreter, seeing the 'sleep' invocation, calls Ksleep() (above).
4) Ksleep locks the focus, which seems to create a context, as without that, the debugger spit out warnings of a 0x0 context during the drawRect() function, and with the the lockFocus it does not and appears to have a valid context value.
5) Ksleep flags the view as needing an update and calls (varyingly) 'display' and/or 'drawRect', etc.
6) The drawRect routine DOES get control (breakpoints indicate that everything is working okay in that regard). 'inSleep' IS set correctly. It steps through everything in drawRect as expected. But NOTHING shows on the display, until...
7) drawRect returns to Ksleep, the sleep times out, the interpreter keeps interpreting, the interpreted code does more drawing and more sleeps, about 10 times (thus repeating steps 2-7 about 10 times).
From the time the program starts until the mouse action causes the 'animation' attempt, a yellow rectangle is drawn in the view. Once the 'animation'-causing mouse click occurs, NOTHING gets updated in the window until the animation completely finishes (even though drawRect IS executing multiple times throughout that animation attempt), THEN the rectangle turns blue. But breakpoints show that execution IS passing through the drawRect (with 'inSleep' true) each time the KSleep() routine is called.
Is this a thread thing? (The program is not explicitly creating any threads.)
I'm not particularly looking for suggestions for how to avoid the animation/KSleep structure, I realize that's not the preferred Macos method of doing things, but this is an attempt to port an old project from elsewhere, and modifying the 'interpreted' code to avoid this is not a possibility.
Thanks for any ideas or suggestions.

Answer 1

The solution I found was to place ALL of the interpreter code into a second thread, freeing up the main thread to be able to update the display. Even though the drawRect routine was running and drawing, it wasn't getting flushed to the display until after the current "command interpretation" cycle completed and returned. This page has a fair amount of information on the subject:
https://developer.apple.com/library/content/technotes/tn2133/_index.html
Additionally, all of the 'invalidate' calls/statements had to be saved/buffered into a common "smallest rectangle that includes all of the invalidates", and then that invalidate done AFTER the "command interpretation" returned AND/OR before a sleep() was done.
PITA.