better deinterlace regression fix

Sorry for the noise, but this is a better fix for the regression because
it does not manipulate the image stack the way the regression and its
first fix did.

diff --git a/src/modules/xine/filter_deinterlace.c b/src/modules/xine/filter_deinterlace.c
index d86f938e88d047f6064c4a9498844cd37fb1aa26..2f3cdae28c01187da092d72333f0e33390d17fe1 100644 (file)
--- a/src/modules/xine/filter_deinterlace.c
+++ b/src/modules/xine/filter_deinterlace.c
@@ -244,9 +244,7 @@ static int filter_get_image( mlt_frame frame, uint8_t **image, mlt_image_format
                        if ( !error && !progressive )
                        {
                                // OK, now we know we have work to do and can request the image in our format
-                               while( mlt_deque_pop_back( MLT_FRAME_IMAGE_STACK( frame ) ) );
-                               *format = mlt_image_yuv422;
-                               error = mlt_frame_get_image( frame, image, format, width, height, writable );
+                               error = frame->convert_image( frame, image, format, mlt_image_yuv422 );
 
                                // Check that we aren't already progressive
                                if ( !error && *image && *format == mlt_image_yuv422 )