Re: [cp-patches] [patch] fix BE image issues

[Thomas Fitzsimmons]

On Tue, 2005-08-16 at 19:11 +0200, Andreas Tobler wrote:

> 
> Index: classpath/native/jni/gtk-peer/gnu_java_awt_peer_gtk_GtkImage.c
> ===================================================================
> RCS file: 
> /cvs/gcc/gcc/libjava/classpath/native/jni/gtk-peer/gnu_java_awt_peer_gtk_GtkImage.c,v
> retrieving revision 1.1.1.1
> diff -u -r1.1.1.1 gnu_java_awt_peer_gtk_GtkImage.c
> --- classpath/native/jni/gtk-peer/gnu_java_awt_peer_gtk_GtkImage.c 
> 16 Jul 2005 00:30:51 -0000      1.1.1.1
> +++ classpath/native/jni/gtk-peer/gnu_java_awt_peer_gtk_GtkImage.c 
> 16 Aug 2005 17:06:45 -0000
> @@ -176,7 +176,7 @@
>     int width, height, rowstride;
>     guchar *pixeldata;
>     jint *src_array_iter, *src;
> -  int i;
> +  int i,j;
> 
>     gdk_threads_enter ();
> 
> @@ -184,15 +184,26 @@
>     height = gdk_pixbuf_get_height (pixbuf);
>     rowstride = gdk_pixbuf_get_rowstride (pixbuf);
> 
> -  src = src_array_iter =
> +  src = src_array_iter =
>       (*env)->GetIntArrayElements (env, pixels, NULL);
> 
>     pixeldata = gdk_pixbuf_get_pixels (pixbuf);
>     for(i = 0 ; i < height; i++)
>       {
> +#ifdef WORDS_BIGENDIAN
> +      /* Convert and copy the native BigEndian pixels to the 32-bit
> +        AABBGGRR format the Java uses.  */
> +      for (j = 0; j < width; j++)
> +       {
> +         unsigned int *pix_conv = (unsigned int *) pixeldata;
> +         *pix_conv = SWAPU32((unsigned)src[j]);
> +         pixeldata += 4;
> +       }
> +#else
>         memcpy((void *)pixeldata, (void *)src, width * 4);
> -      src += width;
>         pixeldata += rowstride;
> +#endif
> +      src += width;
>       }

This, and the existing code, assumes rowstride == width * 4.  This
depends on whether or not the image has an alpha channel though.  I
think we should replace the 4 constants in this code with n_channels,
obtained from gdk_pixbuf_get_n_channels.

Shouldn't you also swap bytes in getPixels?

Tom