Simple investigation shows:
when build against cairo-gl, display_create_egl_window_surface() is using type(struct egl_window_surface_data) to set_user_data, while display_get_buffer_for_surface() is using type (struct surface_data) to get_user_data. Two structures are incompatible

(In reply to comment #1)
> Simple investigation shows:
> when build against cairo-gl, display_create_egl_window_surface() is using
> type(struct egl_window_surface_data) to set_user_data, while
> display_get_buffer_for_surface() is using type (struct surface_data) to
> get_user_data. Two structures are incompatible

That's not the cause of the black boxes bug, but that does cause the dnd client to terminate when a drag is started, since it sends garbage as the wl_buffer for the surface attach in dnd_button_handler().

The black boxes problem is that since weston commit bad42973 [1], the cairo surfaces for the dnd items is created from an wl_egl_window. This surface will have an embeded operand of type CAIRO_GL_OPERAND_TEXTURE, but no texture is actually created. The end result is that texture 0 is bound on 
_cairo_gl_context_setup_operand().


 [1] 
    commit bad429738087a851ab293e5a9329efdf30393333
    Author: Kristian Høgsberg <krh@bitplanet.net>
    Date:   Tue Apr 10 11:49:45 2012 -0400

        window: Drop EGLImage surface type

>
>
> Reproduce steps:
> ----------------------------
> wayland test
> 1. start weston
> 2. ./dnd
>
>
It's the same here. dnd exhibits black squares for dnd objects and clicking
on any of them cause a crash with "read error: Broken pipe"


Scott

(In reply to comment #2)
> (In reply to comment #1)
> > Simple investigation shows:
> > when build against cairo-gl, display_create_egl_window_surface() is using
> > type(struct egl_window_surface_data) to set_user_data, while
> > display_get_buffer_for_surface() is using type (struct surface_data) to
> > get_user_data. Two structures are incompatible
> 
> That's not the cause of the black boxes bug, but that does cause the dnd client
> to terminate when a drag is started, since it sends garbage as the wl_buffer
> for the surface attach in dnd_button_handler().

You are right :)

(In reply to comment #2)
> The black boxes problem is that since weston commit bad42973 [1], the cairo
> surfaces for the dnd items is created from an wl_egl_window. This surface will
> have an embeded operand of type CAIRO_GL_OPERAND_TEXTURE, but no texture is
> actually created. The end result is that texture 0 is bound on 
> _cairo_gl_context_setup_operand().

Moving bug to cairo. The thing that does not work here is doing something like this:

item_surface = cairo_gl_surface_create_for_egl();
cr = cairo_create(item_surface);
...
cairo_paint(cr);
cairo_destroy(cr);

and later

surface = cairo_gl_surface_create_for_egl();
cr = cairo_create(surface);

...

cairo_set_operator(cr, CAIRO_OPERATOR_OVER);
cairo_set_source_surface(cr, item_surface, x, y);
cairo_paint(cr);

Created attachment 60397 [details] [review]
Workaround for the issue

-- GitLab Migration Automatic Message --

This bug has been migrated to freedesktop.org's GitLab instance and has been closed from further activity.

You can subscribe and participate further through the new bug through this link to our GitLab instance: https://gitlab.freedesktop.org/cairo/cairo/issues/211.