This bug was introduced when lighting computations was changed from eye-space to object-space.
The light position need to be transformed back to object-space each time the modelview matrix changes which requires us to compute the inverse of the modelview matrix. This computation was done with the assumption that normals where transformed (which was the case when the computation was made in eye-space), however, normals only require the inverse of the upper 3x3 matrix while transforming positions requires the inverse of the whole matrix.
This caused the interesting behavior that lights were more-or-less transformed properly, but not translated at all, which caused improper lighting with directional lights in particular.
There was also another smaller bug affecting directional lights: when vertices are read, only the active component are read, the other ones are ignored, later, the transformation operations are set up to ignore the unset values, howver, in the case of lighting, we use the vertex in object space (that is, before it is transformed), and therefore were using uninitalized values; in particular w.
JSR239 and android.opengl gl Pointer functions (glColorPointer, etc.)
now respect the current setting of the Buffer position.
This fixes a regression introduced when we started requiring the
Buffers passed to the Pointer functions to be direct Buffers.
This was always a documented restriction, but was not enforced by the runtime until now.
Until now, if you passed in some other kind of buffer, it would sometimes work, and
sometimes fail. The failures happened when the Java VM moved the buffer data while
OpenGL was still holding a pointer to it.
Now we throw an exception rather than leaving the system in a potentially bad state.
Remove include of an internal agl header file. We should not depend on any implementation details
of our software renderer, since they may not be correct if another renderer is used.
Fix glGet number-of-elements logic for GL_FOG_COLOR GL_LIGHT_MODEL_AMBIENT,
and GL_COMPRESSED_TEXTURE_FORMATS.
This is just plumbing. The Java APIs existed already, but there were no C APIs to hook the Java APIs
up to. Now there are C APIs, so we can call them.
Of course, whether or not the C APIs actually work when you call them depend upon the
capabilities of the active OpenGL driver, which must be checked at run time.
Also, while we're here, make the glGetString method static. It was always supposed to be static,
but was accidentally implemented as non-static, because the code was copied from the non-static
OpenGL ES classes.
This change adds four new public classes that expose a static OpenGL ES 1.1 API:
android.opengl.GLES10
android.opengl.GLES10Ext
android.opengl.GLES11
android.opengl.GLES11Ext
Benefits:
+ The static API is slightly faster (1% to 4%) than the existing Interface based JSR239 API.
+ The static API is similar to the C API, which should make it easier to import C-based
example code.
+ The static API provides a clear path for adding new OpenGL ES 1.1 extensions
and OpenGL ES 2.0 APIs, neither of which currently have a JSR standard.
Example:
import static android.opengl.GLES10.*;
...
glClear(GL_COLOR_BUFFER_BIT | GL_DEPTH_BUFFER_BIT);
Note that it is possible to mix-and-match calls to both the static and JSR239 APIs.
This works because neither API maintains state. They both call through to the same underlying
C OpenGL ES APIs.
Implementation details:
This change enhances the "glgen" "gen" script to generate both the original JSR239 and
new static OpenGL ES APIs. The contents of the generated JSR239 classes remained the same as before,
so there is no need to check in new versions of the generated JSR239 classes.
As part of this work the gen script was updated to be somewhat more robust, and to
work with git instead of perforce. The script prints out commands to git add the generated files,
but leaves it up to the script runner to actually execute those commands.
+ gen script is really a bash script rather than a sh script,
so declare that to be true. (For example, it uses pushd,
which is a part of bash, but not a part of sh. Not sure
how this worked until now. Possibly gen was only run in
environments where /bin/sh was really bash.
+ Check the results of the java compile of the code generator,
and abort the script if the compile fails.
+ Turn on the bash shell option that guards against using
uninitialized variables in the script.
+ Remove the generated class files.
Refactor JniCodeEmitter into two classes: a general-purpose
JniCodeEmitter and a specific Jsr239CodeEmitter. The hope is
to use JniCodeEmitter as a base for emitting static OpenGL ES
bindings.
Added @Override to overridden methods.
Removed unused imports.
Converted tabs to spaces.
Removed \r characters from end-of-lines.
Add .gitignore file to ignore the .class files that are
generated when the "gen" script is run.
