Bug rendering smooth point

On Windows, a line (which should not be there) appears above a single point as shown in this picture:
https://www.dropbox.com/s/9y5p78fxsesb75y/screenshot.png?dl=0

Main code:

static void setup_test()
{
	glEnable(GL_POINT_SMOOTH);
	glPointSize(180);
}

static void tick()
{
	float projMat[] = {1209.89f, 0, 0, 0, 0, 2214.84f, 0, 0, 0, 0, -0.3176f, 0, 0, 0, 0, 1};
	float mvMat[] = {0, -0.520227f, -0.854028f, 0, 0, 0.854028f, -0.520227f, 0, -1, 0, 0, 0, 0, 0.172901f, 0.607644f, 1};
	float point[] = {0.608881f, 0.168486f, 0};

	glClearColor(1.0f, 1.0f, 1.0f, 1.0f);
	glClear(GL_COLOR_BUFFER_BIT | GL_DEPTH_BUFFER_BIT);

	glMatrixMode(GL_PROJECTION);
	glLoadMatrixf(projMat);
	glMatrixMode(GL_MODELVIEW);
	glLoadMatrixf(mvMat);

	glColor4f(1, 0, 0, 1);

	glBegin(GL_POINTS);
	{
		glVertex3fv(point);
	}
	glEnd();

	SwapBuffers(hdc);
}

Here is a link https://www.dropbox.com/s/sjl2vs3xkflupds/gltest.zip?dl=0 to the sample program.

Reproduced on a GTX 760, driver version 9.18.13.5306.

Thanks,
Boni

Could you please query the implementation dependent supported point size ranges?

I’m pretty sure the maximum supported antialiased point size in NVIDIA implementations was 63.75 in the past.
If the maximum supported antialised point size is smaller than the 180 pixels you used, the expected result should have been a point size clamped to that maximum, which would be the actual issue. The corruption could then be due to trying to render something inadvertently which isn’t actually supported.
That is, please try if everything works when not exceeding the implementation dependent point size range maximum value.

In any case, your application should not assume arbitrary point sizes are supported by an OpenGL implementation. Even an implementation only supporting point size ranges from 1 to 1 pixels would be spec conformant.

I called

GLint range[2];
glGetIntegerv(GL_POINT_SIZE_RANGE, range);

and get the results {1, 190}. This is with a GTX 760.

-Boni

Thanks. Interesting, I didn’t know that the maximum increased.
I already pointed the OpenGL driver team leads to this for investigation and can reproduce the problem on another board now. I’ll file a bugreport.

Great, thanks!