> l773 could use comment here as well (like the other comment noting how CapPnP
> has problems)
> + //std::copy(dnd_handle.begin(), dnd_handle.end(),
> back_inserter(blob->data_));
Fixed.
> +// We miss the "mouseover" occasionally (with valgrind and heavy stress about
> 1/20).
> 1995 +// But it isn't essential for the test and we've probably waited long
> enough
> 1996 +// for the mouse-down needed by the test to reach the window.
> 1997 +// EXPECT_THAT(have_mouseover.wait_for(receive_event_timeout),
> Eq(true));
> does that need a bug?
Possibly. It would appear there is a lack of consistency between the window management "model" that says the surface is ready and has focus (hence the mir_window_attrib_focus notification in paint_window) and the input "model" that does not (yet) route input to the surface. That's what you get by updating observers asynchronously - which, given the code we landed to support it, appears to have been a deliberate decision.
> l773 could use comment here as well (like the other comment noting how CapPnP copy(dnd_ handle. begin() , dnd_handle.end(), blob->data_ ));
> has problems)
> + //std::
> back_inserter(
Fixed.
> +// We miss the "mouseover" occasionally (with valgrind and heavy stress about THAT(have_ mouseover. wait_for( receive_ event_timeout) ,
> 1/20).
> 1995 +// But it isn't essential for the test and we've probably waited long
> enough
> 1996 +// for the mouse-down needed by the test to reach the window.
> 1997 +// EXPECT_
> Eq(true));
> does that need a bug?
Possibly. It would appear there is a lack of consistency between the window management "model" that says the surface is ready and has focus (hence the mir_window_ attrib_ focus notification in paint_window) and the input "model" that does not (yet) route input to the surface. That's what you get by updating observers asynchronously - which, given the code we landed to support it, appears to have been a deliberate decision.
I doubt we'll see an issue in "real life".