This seems to be a better place to fix the OS X driver bug via glFinish; glFinish() will get called much less frequently and add less delay.
It seems to me (subjectively) that this further improves accuracy on OS X, though I have no objective measurements right now.
| Lint Skipped |
| Unit Tests Skipped |
I don't know much about OpenGL, but I wonder if the glFinish call is even needed at all? How about replacing it with glFlush?
I'll also need to check any changes on Windows with both desktop OpenGL and ANGLE.
I'd be tempted to put the entire change, in KisCanvas2 too, within #ifdef Q_OS_OSX for clarity's sake. It shouldn't make any difference on any other OS, because it only moves the Q_OS_OSX block of code.
Yes, it's an OS X only bug fix. It actually needs glFinish (full synchronous stop, argh), I tried glFlush and then you get artefacts (at least with mip mapping texture modes).
I'm currently investigating getting rid of the glFinish altogether, by (1) using fences or (2) using double buffered pixel buffers. Might actually abandon this if one of the other approaches works.
May I ask which Qt version are you testing with? The current OS X release is using Qt 5.7.0 but Boud will be checking if it's worth updating to Qt 5.9.2 (specifically for https://bugreports.qt.io/browse/QTBUG-61384).
(Related: https://bugs.kde.org/show_bug.cgi?id=373676)
@alvinhochun Hi Alvin, sorry for the late reply, I didn't see this one. I'm locally still testing using QT 5.9.1.