Raspberry pi Alpha transparency

SDL Development
1

Hi All,
I’ve been working on this for a few days now and not getting any closer to a result.

i’m trying to use sdl to make a overlay type display on top of the raspberry pi camera.
however no matter what i try i can’t seem to get it to work i always end up with a black background that is over writing the display under it.

i know it can be done as I’ve got a proto type mocked up using the dispman hello font example.

the code i’m trying is

Code:

#include <stdio.h>
#include “SDL.h”
#include “SDL_image.h”

int main(int argc, char **argv)
{
if (SDL_Init(SDL_INIT_VIDEO) !=0) {
printf(“Error SDL_Init %s \n”,SDL_GetError());
}
SDL_Window *window;
window = SDL_CreateWindow(
“Test”,
SDL_WINDOWPOS_UNDEFINED,
SDL_WINDOWPOS_UNDEFINED,
640,
480,
SDL_WINDOW_OPENGL
);
if (window == NULL) {
printf(“Error with Window %s \n”,SDL_GetError());
}
SDL_Renderer *ren;
ren = SDL_CreateRenderer(window,-1,SDL_RENDERER_ACCELERATED |
SDL_RENDERER_PRESENTVSYNC | SDL_RENDERER_TARGETTEXTURE);

SDL_Texture *tex;
tex = SDL_CreateTexture(ren,SDL_PIXELFORMAT_ARGB8888,
SDL_TEXTUREACCESS_TARGET,500,500);
if (tex == NULL) {
printf(“textur error %s \n”,SDL_GetError());
}
SDL_SetTextureBlendMode(tex,SDL_BLENDMODE_BLEND);

SDL_SetRenderTarget(ren,tex);
SDL_SetRenderDrawColor(ren,0,0,0,0);
SDL_RenderClear(ren);
SDL_SetRenderTarget(ren,NULL);
SDL_Rect rect;
rect.x=0;
rect.y=0;
rect.h=300;
rect.w=300;

SDL_SetRenderDrawColor(ren,255,255,255,255);
SDL_RenderFillRect(ren,&rect);

SDL_RenderPresent(ren);
SDL_Delay(3000);
return 0;
}

the only way i’ve been able to get close is to modify the following in the source code.

Code:
/* Disable alpha, otherwise the app looks composed with whatever dispman is showing (X11, console,etc) */
dispman_alpha.flags = DISPMANX_FLAGS_ALPHA_FIXED_ALL_PIXELS;
dispman_alpha.opacity = 100;
dispman_alpha.mask = 0;

this works but makes the entire window transparent including anything i draw onto it.

i’m using arch Linux with a PKGBUILD version of sdl that disables opengl and makes SDL2 use the dispmanx backend.

can anyone point me in the right direction with this???

Thanks
Matthew

2

I’ve made the change and it’s not made a difference although i’m sure it needs to be there for it to work.

i’ve done some tests and can get a renderer to display under the camera by setting the camera preview layer to 30 and the sdl2 layer to 10.

i also tried removing the following lines

Code:

dispman_alpha.opacity = 0xFF;
dispman_alpha.mask = 0;

and again no difference,

this is the relevant SDL section for the test I’m running,
i should get a semi transparent colour across the screen.

Code:

if (SDL_Init(SDL_INIT_VIDEO) !=0) {
	printf("Error SDL_Init %s \n",SDL_GetError());
}
printf("Before Window %s \n",SDL_GetError());
SDL_Window *window;
window = SDL_CreateWindow(
	"Test",
	SDL_WINDOWPOS_UNDEFINED,
	SDL_WINDOWPOS_UNDEFINED,
	640,
	480,
	SDL_WINDOW_OPENGL
	);
	
if (window == NULL) {
	printf("Error with Window %s \n",SDL_GetError());
}
SDL_GLContext context;
context = SDL_GL_CreateContext(window);
glClearColor(0.0f,1.0f,0.0f,0.50f);
glFlush();
glClear( GL_COLOR_BUFFER_BIT | GL_DEPTH_BUFFER_BIT );
printf("After Window %s \n",SDL_GetError());

SDL_Renderer *ren;
ren = SDL_CreateRenderer(window,-1,SDL_RENDERER_ACCELERATED | 
	SDL_RENDERER_PRESENTVSYNC);
if (ren == NULL) {
	printf("Render error %s \n",SDL_GetError());
}

SDL_Rect rect;
	rect.x=0;
	rect.y=0;
	rect.h=300;
	rect.w=300;
SDL_SetRenderDrawBlendMode(ren,SDL_BLENDMODE_BLEND);
SDL_SetRenderDrawColor(ren,255,255,255,255);
SDL_RenderFillRect(ren,&rect);	
printf("ERR Check %s \n",SDL_GetError());
char *me;
SDL_GetWindowData(window,"test");
printf("SDL Window Data %s \n",&me);	

SDL_GL_SwapWindow(window);
//SDL_RenderPresent(ren);

it’s messy but I’m still in the testing stage.

is it possible that I’ve messed up the linking some how and it’s using the mesa gl instead of raspberry pi’s built in.

Many Thanks
Matthew

3

Cool, i’m gonna see if i can try this now and will report back.

out of interest as i understand it the Raspberry pi has 3 layers do you know why it’s been set to 1000 or 10 for example instead of just 0-2

Thanks
Matthew

4

it does yes,
I’m using code from the Raspberry pi’s video record program seemed silly to rewrite the hole thing again.
you can change the layer of the preview window in source but not from the cmdline.

regarding the layer numbering I had it the wrong way I though the 64 elements limit was to-do with the number of objects on the screen not the number of layers on the screen.
is that right?

Thanks
Matthew

5

No worries I just wanted to much sure I understood it correctly.