Convert on CPU if we are asked to finalize an empty Movit chain.

[mlt] / src / modules / opengl / mlt_movit_input.h

diff --git a/src/modules/opengl/mlt_movit_input.h b/src/modules/opengl/mlt_movit_input.h
index c200cc67f6c2e0536cb33ffb5979c93eee125307..109194f010ed468fbc65edfa34a99b2739f103b2 100644 (file)
--- a/src/modules/opengl/mlt_movit_input.h
+++ b/src/modules/opengl/mlt_movit_input.h
@@ -20,6 +20,8 @@
 #ifndef MLT_MOVIT_INPUT_H
 #define MLT_MOVIT_INPUT_H
 
+#include <framework/mlt_types.h>
+
 #include <movit/flat_input.h>
 #include <movit/ycbcr_input.h>
 #include <movit/effect_chain.h>
@@ -27,7 +29,7 @@
 class MltInput
 {
 public:
-       MltInput();
+       MltInput( mlt_image_format format );
        ~MltInput();
 
        void useFlatInput(movit::MovitPixelFormat pix_fmt, unsigned width, unsigned height);
@@ -36,7 +38,12 @@ public:
        void invalidate_pixel_data();
        movit::Input *get_input() { return input; }
 
+       // The original pixel format that was used to create this MltInput,
+       // in case we change our mind later and want to convert on the CPU instead.
+       mlt_image_format get_format() const { return m_format; }
+
 private:
+       mlt_image_format m_format;
        unsigned m_width, m_height;
        // Note: Owned by the EffectChain, so should not be deleted by us.
        movit::Input *input;