LWJGL 0.9 released!

[Matzon]

LWJGL 0.9 has been released.

http://www.lwjgl.org/

stuff changed from 0.89 of the top of my head:
* bug fixes
* setGrabbed support for grabbing cursor, or showing OS cursor
* Software OpenGL support via flag (mental note, document these)
* win32 cursor animation (still only 1 bit cursor transparency on win32, will fix soon)
* input now accumulates over multiple Window.updates
* more extensions
* automatic read of input, now you only need to call Window.update

One gotcha, due to the accumelated delta's on devices, you can only call getD* once. A second call will return 0, untill the device has actually moved.

more comments to come, perhaps

ps. due to a small f*ckup, the Version string is still 0.9pre. This isn't a biggie, and would require a full recompile on all platforms, which I am not in the mood for, so I'll let it slip.

[elias]

GL/AL changes include splitting each extension out into separate classes, like ARBVertexProgram instead of the usual GL. This works the same way as the core GL which is split into GL11/GL12/GL13/GL14/GL15. This was probably in 0.89 too, but it's still important to note in order to hinder confusion...

- elias

[spasi]

Have a look at opengl.org's headlines

[Matzon]

w00t, I posted it last night

[cfmdobbie]

Request: More helpful messages in BufferOverflowExceptions.  Example, from GL11.java:

public static void glTexImage2D(int target, int level, int internalformat, int width, int height, int border, int format, int type, ByteBuffer pixels) {
		if (pixels.remaining() < BufferChecks.calculateImageStorage(format, type, width, height, 1)) {
			throw new BufferOverflowException();
		}
		nglTexImage2D(target, level, internalformat, width, height, border, format, type, pixels, pixels.position());
	}

There are a huge number of blocks almost identical to this one.  Maybe they should be refactored together and given a nicely formatted exception message?

[princec]

Most annoyingly BufferOverflowException doesn't have a String constructor for an error message Sorry.

Cas

[cfmdobbie]

Wha...? *blink*  What a bunch of muppets! :x

That's the second time this week I've cursed the ground the Sun engineers walk on.  (The first time was for the entire java.awt.imaging API.)

Sigh...

[AndersD]

Well, as you're already limited to >= 1.4 you could use a chained runtime exception to get a more user friendly error message:

throw new RuntimeException("BLABLABLA", new BufferOverflowException());

Which will first print the error message and then the cause (bufferoverflow in this case)

[elias]

How about an IllegalArgumentException as it is done in BufferChecks.java already?

- elias

[princec]

Yeah, seems a bit cleaner than a chained exception.

Cas

[AndersD]

yep, even better  :oops:

[tomb]

I think I've found a bug in gluBuild2DMipmaps. I get artifacts when I use non power of two textures.

I'll investigate further and post a fix if I find the bug.

[tomb]

Found it. The scaling algo was only designed to sale images down not up. Here is the fix. It's to class org.lwjgl.opengl.glu.MipMap, line 219:

for (int iy = 0; iy < heightOut; iy++) {
			for (int ix = 0; ix < widthOut; ix++) {
				int x0 = (int) (ix * sx);
				int x1 = (int) ((ix + 1) * sx);
				int y0 = (int) (iy * sy);
				int y1 = (int) ((iy + 1) * sy);

				int readPix = 0;

				// reset weighted pixel
				for (int ic = 0; ic < components; ic++) {
					c[ic] = 0;
				}

				// create weighted pixel
				for (int ix0 = x0; ix0 < x1; ix0++) {
					for (int iy0 = y0; iy0 < y1; iy0++) {

						src = (iy0 * widthIn + ix0) * components;

						for (int ic = 0; ic < components; ic++) {
							c[ic] += tempin[src + ic];
						}

						readPix++;
					}
				}

				// store weighted pixel
				dst = (iy * widthOut + ix) * components;

				if (readPix == 0) {
					// BUGFIX: Image is sized up, caused by non power of two texture as input
					src = (y0 * widthIn + x0) * components;
					for (int ic = 0; ic < components; ic++) {
						tempout[dst++] = tempin[src + ic];
					}
				} else {
					for (k = 0; k < components; k++) {
						tempout[dst++] = c[k] / readPix;
					}
				}
			}
		}

[tomb]

Another possible bug in gluBuild2DMipmaps. The following code causes a
java.lang.IndexOutOfBoundsException in org.lwjgl.opengl.glu.MipMap.gluScaleImage.

public class GLUMipMap {
	public static void main(String args[]) {
		try {
			org.lwjgl.opengl.Window.create("", 600, 0, 40, 40, 1, 0, 16, 0);
			int width  = 383;
			int height = 226;
			ByteBuffer buf = ByteBuffer.allocateDirect(width*height*3);
			IntBuffer intBuf = ByteBuffer.allocateDirect(4).order(ByteOrder.nativeOrder()).asIntBuffer();
			GL11.glGenTextures(intBuf);
			GL11.glBindTexture(GL11.GL_TEXTURE_2D, intBuf.get(0));
			int error = GLU.gluBuild2DMipmaps(GL11.GL_TEXTURE_2D, GL11.GL_RGB, width, height, GL11.GL_RGB, GL11.GL_UNSIGNED_BYTE, buf);
		} catch (Throwable t) {
			t.printStackTrace();
		} finally {
			org.lwjgl.opengl.Window.destroy();
		}
	}
}

The problem has got something to do with PixelStoreState as the rowstride do not match imagewidth*3. It works in the old GLU so I think it's a bug.

[Matzon]

regarding GLU, I am sure erikd will pop in - he did the basic GLU port from native to java.
We changed from requiring native GLU to a java GLU implementation - certain to be some bugs to iron out