0

I'm getting some intermittent crashing when calling glReadPixels on an FBO. Currently causing more than 1000 crashes per day in my live app.

I'll include below the code for reading the pixels and generating the FBO. Happy to include other code used if needed.

typedef struct
    {
    GLuint frameBufferID;
    GLuint textureID;
    int textureNum;
    } FBO;


+ (void) readPixelsFromFBO: (FBO) fbo  withWidth: (int) width  andHeight: (int) height  intoData: (NSMutableData*) data
    {
    glActiveTexture( GL_TEXTURE0 + fbo.textureNum );
    glBindTexture( GL_TEXTURE_2D, fbo.textureID );

    glBindFramebuffer( GL_FRAMEBUFFER, fbo.frameBufferID );
    glReadPixels( 0, 0, width, height, GL_RGBA, GL_UNSIGNED_BYTE, data.mutableBytes );

    GLenum error = glGetError();
    if ( error != 0 )
        NSLog( @"readPixelsFromFBO ERROR: %d", error );
    }


+ (FBO) createFBOWithWidth: (int) width  andHeight: (int) height  andTextureNum: (int) textureNum
    {
    FBO fbo;

    fbo.textureNum = textureNum;

    // Generate the IDs
    glGenFramebuffers( 1, &fbo.frameBufferID );
    glGenTextures( 1, &fbo.textureID );

    // Create the texture
    glActiveTexture( GL_TEXTURE0 + textureNum );
    glBindTexture( GL_TEXTURE_2D, fbo.textureID );

    glTexParameteri( GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_LINEAR );
    glTexParameteri( GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_NEAREST );
    glTexParameteri( GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE );
    glTexParameteri( GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE );

    GLubyte* bytes = (GLubyte*) calloc( width * height * 4, sizeof( GLubyte ) );    
    glTexImage2D( GL_TEXTURE_2D, 0, GL_RGBA, width, height, 0, GL_RGBA, GL_UNSIGNED_BYTE, bytes );

    free( bytes );

    // Create the frame buffer
    glBindFramebuffer( GL_FRAMEBUFFER, fbo.frameBufferID );
    glFramebufferTexture2D( GL_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, fbo.frameBufferID, 0 );

    // FBO status check
    GLenum status = glCheckFramebufferStatus( GL_FRAMEBUFFER );

    switch(status)
        {
        case GL_FRAMEBUFFER_COMPLETE:
            NSLog(@"FBO created");
        break;

        case GL_FRAMEBUFFER_UNSUPPORTED:
            NSLog(@"FBO unsupported");
        break;

        default:
            /* programming error; will fail on all hardware */
            NSLog( @"Framebuffer Error: make sure that the context was created!" );
        break;
        }

    return fbo;
    }

The crash logs look like this (with some variation):

Exception Type:  SIGSEGV
Exception Codes: SEGV_ACCERR at 0x1
Crashed Thread:  0

Thread 0 Crashed:
0   libGPUSupportMercury.dylib          0x36d0ae22 <redacted> + 10
1   IMGSGX543GLDriver                   0x324bff31 <redacted> + 245
2   IMGSGX543GLDriver                   0x324c0011 <redacted> + 37
3   IMGSGX543GLDriver                   0x324c3a81 <redacted> + 405
4   GLEngine                            0x348712f9 <redacted> + 1257
5   OpenGLES                            0x34915da3 _glReadPixels + 55
OnesAndZeroes
  • 315
  • 1
  • 9
  • For your NSMutableData, have you used `+dataWithLength:` or `-setLength:` to make sure you have sufficient capacity for the framebuffer size you're reading? You might be writing to memory past the end of where your NSMutableData has allocated its internal byte buffer, leading to intermittent crashes. – Brad Larson Jul 09 '13 at 17:27
  • Yes, I'm using `+dataWithLength:` and have ensured that enough bytes have been allocated. Just as an aside I once used `+dataWithCapacity:` which caused me all sorts of trouble (mentioning this just in case someone reads this and is using `dataWithCapacity:`). – OnesAndZeroes Jul 09 '13 at 21:13
  • In that case, are the crashes segmentation faults or out of memory exceptions? What is logged at the console? Is it possible that `glReadPixels()` is being called on a thread where your OpenGL ES context has not been made the current context? Could you be attempting this read at the same time as another rendering operation is accessing the OpenGL ES context on a different thread? These are all common triggers for crashes with this function. – Brad Larson Jul 09 '13 at 21:17
  • The crashes are all SEGV_ACCERR segmentation faults. Unfortunately, I've never been able to reproduce the crash on local devices (try as I might) so I'm unsure of console output. All operations are happening in the main thread. For debugging I've added `if !NSThread.isMainThread` checks everywhere. One extra thing to note. I have another app that uses the same basic FBO/glReadPixels code, but creates only 3 FBOs and never crashes. The problematic app is creating 6 FBOs. I'm assuming 6 FBOs (at 768x1024px) is not too many to create? – OnesAndZeroes Jul 10 '13 at 01:32
  • I believe I found what the problem was. I was allocating OpenGL resources each time a needed frame was initialized, but I was neglecting to deallocate the resources when the frame was closed (ie., the render buffers, frame buffers, fbos, and program). – OnesAndZeroes Jul 10 '13 at 03:06

0 Answers0