paint-nodes: Clarify color handling for TextureNode
The TextureNode premultiplies the blend color passed to the node constructor, so we need to document the fact properly to avoid causing premultiplication twice. We can also allow passing NULL for a color, and use a fully opaque white, to make the code slightly more friendly.
This commit is contained in:
parent
0c39138200
commit
bceca34ef9
@ -652,7 +652,7 @@ clutter_scaling_filter_to_cogl_pipeline_filter (ClutterScalingFilter filter)
|
||||
/**
|
||||
* clutter_texture_node_new:
|
||||
* @texture: a #CoglTexture
|
||||
* @color: a #ClutterColor
|
||||
* @color: (allow-none): a #ClutterColor used for blending, or %NULL
|
||||
* @min_filter: the minification filter for the texture
|
||||
* @mag_filter: the magnification filter for the texture
|
||||
*
|
||||
@ -661,6 +661,10 @@ clutter_scaling_filter_to_cogl_pipeline_filter (ClutterScalingFilter filter)
|
||||
* This function will take a reference on @texture, so it is safe to
|
||||
* call cogl_object_unref() on @texture when it returns.
|
||||
*
|
||||
* The @color must not be pre-multiplied with its #ClutterColor.alpha
|
||||
* channel value; if @color is %NULL, a fully opaque white color will
|
||||
* be used for blending.
|
||||
*
|
||||
* Return value: (transfer full): the newly created #ClutterPaintNode.
|
||||
* Use clutter_paint_node_unref() when done
|
||||
*
|
||||
@ -686,12 +690,18 @@ clutter_texture_node_new (CoglTexture *texture,
|
||||
mag_f = clutter_scaling_filter_to_cogl_pipeline_filter (mag_filter);
|
||||
cogl_pipeline_set_layer_filters (tnode->pipeline, 0, min_f, mag_f);
|
||||
|
||||
if (color != NULL)
|
||||
{
|
||||
cogl_color_init_from_4ub (&cogl_color,
|
||||
color->red,
|
||||
color->green,
|
||||
color->blue,
|
||||
color->alpha);
|
||||
cogl_color_premultiply (&cogl_color);
|
||||
}
|
||||
else
|
||||
cogl_color_init_from_4ub (&cogl_color, 255, 255, 255, 255);
|
||||
|
||||
cogl_pipeline_set_color (tnode->pipeline, &cogl_color);
|
||||
|
||||
return (ClutterPaintNode *) tnode;
|
||||
|
Loading…
Reference in New Issue
Block a user