Hi,
The more I get to know FlasCC and GLS3D, the more I found that we can actually write GLS3D in C++. When implementing the shaders, I thought I can't create Stage3D objects in C++. However, it turns out that there's AS3++ header file, so that we can create Stage3D objects in C++.

So, the question becomes:

1. Do we actually need the action script code to translate from GL calls to Stage3D calls?
2. What is the initial design of GLS3D? Is it to provide GL layers to both AS code and C++ code? If yes, then it makes sense to have AS code. If not, we can probably write the library entirely in C++ (which can be more efficient). Is there any gotcha's when doing that?

Another advantage of doing everything in C++ is that we can include OpenGL library and thus we don't need to create GL constants for action script code.

Thanks.

-tep

I am sure there is a lot more missing but

glGetAttribLocation and glGetAttribLocation
are missing and it would be great if they can be added.

Hi Alex,
As I'm implementing more and more stubbed functions, I felt there's a need to do a better logging/error reporting. I know in libGL.as you have TraceLog() class to do just this (through trace()).

However, I don't have Flash IDE when implementing this. I use Visual Studio for typing and cygwin to run my test code. Hence, I couldn't see trace() output. Do you know any solution for this?

In addition, we may need to separate code to do tracing and error reporting. I'm thinking to have couple helper functions in TraceLog class:

• trace() - to trace code flow/execution, this will replace send()
• error() - to report any error - will be saved into a string (I can display in a TextField)
• log() - to do misc logging - will be saved into a string (I can display it in a TextField)

What do you think?

Thanks.

-tep

N.B: Which one is better, should I communicate thorugh github or do you prefer this kind of dicussion by emeail?

Can we please add

glGetRenderbufferParameteriv
or
glGetRenderbufferParameterivOES

Hi i want to add VBO support for type GL_ARRAY_BUFFER and after analyzing the code i have a couple of points i want to discuss. I found that is Stage3D the equivalent of a VBO of type GL_ARRAY_BUFFER is VertexBuffer3D.

There is two places where VertexBuffer3D are used:

• GLS3D.as defines glEndVertexData() which is used by the older glEnd() mechanism to build a VertexBuffer3D. I can't reuse it since its assuming a structure of Positions, Color, Normals and Tex coords. We cannot assume anything when creating a VBO since the structure is defined by the user.
• There is also a VertexBufferPool class, also assuming the same vertex data structure.

My problem:
I am in the process of adding support for glGenBuffer(), glBufferData().

• glGenBuffer(GLsizei n, GLuint *buffers)
  This function creates an id (or many if size > 1) and assign it to buffers. The id will be used later on by the user to bind to a specific VBO. I thought it would be ideal to create a list of VertexBuffer3D and the id created by glGenBuffer() would be the index position in the vector, I would also flag it as "Created". When deleting a VBO we will set the it to null in the vector and flag it as "recycled" for the next glGenBuffer() call.

• glBufferData(GLenum target, GLsizeiptr size, const GLvoid *data, GLenum usage)
  Here is my main issue, this function only accepts an simple pointer to data and the data size. We don't know anything on the actual data array. This is a problem with Stage3D since the VertexBuffer3D constructor requires that we specify the number of vertex provided and the stride of each vertex data. I would see to solutions to this :

  -> Create glBufferData() with 2 new params but that will break the OpenGL API and will require the user to add #ifdef FLASCC to add 2 extra params...its kinda bad

  -> Lazy create the VertexBuffer3D at render time, when we specify the glVertexAttribPointer(). This is actually where we specify to opengl how the VBO is constructed...

• Keeping track of the vbos :
  My idea would be to add a new structure for storing VBO allocated by glGenBuffer().

• I would create a Dictionnary keyed on the VBO type , for now i will add only GL_ARRAY_BUFFER but GL_ELEMENT_ARRAY_BUFFER will be another one to add in a near future.

• The value of the dictionnary will be a vector containing VertexBuffer3D and a flag to specify its current state (Ex: Created, Uploaded, Recycled)

So this is it...if someone started to work on this or think there is a better solution let me know. Thx!

No KHR/khrplatform.h in source tree and no information given in readme where to get it.

Hello @alexmac, @tepGithub,
My name is Fil and I’m from Adobe’s Open Source Office. I see this project has been inactive for a long time. Just wondering: is there still value in keeping this project around? Do we foresee community usage/activity happening around this project?

If yes, can I send some pull requests to update the documentation? At a minimum we need contribution guidelines that reference requiring the signing of Adobe’s CLA.

If no, is it OK if we archive this project? Or even remove it altogether?

Thanks for any info.

-Fil

Hi,
We need the implementation of glClearDepthf when porting. I think this might be useful for others too. In libGL.as, I saw this

    public function glClearDepth(depth:Number):void
    {
        // if (log) log.send( "[IMPLEMENTED] glClearDepthf " + depth + "\n")   
        contextClearDepth = depth          
    }

in the function signature, it's glClearDepth, however, in the comment, it's glClearDepthf. Is this a typo?

Thanks.

-tep

Hi Alex,
I'm currently working on loading the shaders with GLS3D. In theory GLS3D should take GLSL and convert it to AGAL under the hood.

I just want to confirm with you the steps I needed to do:

1. Run glsl2agal to get JSON output file (done offline, through glsl2agal)
2. Load JSON file
3. Extract information regarding uniforms
4. Extract agal asm code
5. Run AGALMiniAssembler on agal asm code
6. Create Program3D and upload the AGAL asm code

In your todo list, you wrote about integrating glsl2agal, were you thinking to run the conversion process at run-time? How easy is it to integrate it (how many files I need to integrate)? I can try doing it if it's not that hard.

-tep