with help from Arne Bergene Fossaa i get to this solution:
int size = m_TexBitmap.getRowBytes() * m_TexBitmap.getHeight();
ByteBuffer bb = ByteBuffer.allocateDirect(size); // size is good
bb.order(ByteOrder.nativeOrder());
m_TexBitmap.copyPixelsToBuffer(bb);
bb.position(0);
ETC1Texture etc1tex;
// RGB_565 is 2 bytes per pixel
//ETC1Texture etc1tex = ETC1Util.compressTexture(bb, m_TexWidth, m_TexHeight, 2, 2*m_TexWidth);
final int encodedImageSize = ETC1.getEncodedDataSize(m_TexWidth, m_TexHeight);
ByteBuffer compressedImage = ByteBuffer.allocateDirect(encodedImageSize).order(ByteOrder.nativeOrder());
// RGB_565 is 2 bytes per pixel
ETC1.encodeImage(bb, m_TexWidth, m_TexHeight, 2, 2*m_TexWidth, compressedImage);
etc1tex = new ETC1Texture(m_TexWidth, m_TexHeight, compressedImage);
//ETC1Util.loadTexture(GL10.GL_TEXTURE_2D, 0, 0, GL10.GL_RGB, GL10.GL_UNSIGNED_SHORT_5_6_5, etc1tex);
gl.glCompressedTexImage2D(GL10.GL_TEXTURE_2D, 0, ETC1.ETC1_RGB8_OES, m_TexWidth, m_TexHeight, 0, etc1tex.getData().capacity(), etc1tex.getData());
bb = null;
compressedImage = null;
etc1tex = null;
i know about the ETC1Util.compressTexture and ETC1Util.loadTexture, but they were giving corrupted textures. good thing is that i went from 100MB down to 26MB with native memory consumption. but this solution is slow as hell. and even though it is done on a separate thread with min priority, the rendering thread is totally blocked. is there a more efficent way? or do i have to create these ETC1 textures on the first run on a new device and save them to SD card for later reuse?