Hans Breuer writes:
> This is a regression of previous versions. IIRC it was caused
> by the usage of gdk_eindow_new() with parameters 'in range' and
> additional adjusted by SafeAdjustWindowRect and the code
> later calling gdk_window_move with coordinates directly derived
> from the current mouse cursor position.
> Using SafeAdjustWindowRectEx in gdk/win32/gdkwindow-win32.c:996
> appears to fix it.
I just committed changes to gdk/win32 (HEAD and gtk-2-2) that seem to
fix these problems. In fact, I dropped SafeAdjustWindowRectEx()
altogether, I don't think it's GTK's business to prevent an
application from locating its windows outside of the screen if it
wants to :-) And with multiple monitors, coordinates might be negative
The main job of the fix was going through the handling of top-level
window position and size in event generation and functions, and make
sure that it matches GDK's convention (position is that of decoration,
size that of client area). And in the Win32 API, usually both position
and size include decorations.
> I'm still using my somewhat hackish g_spawn_async version attached to
> the bug report and don't have to restart at all under win9x during the
> whole plug-in query phase.
I'll look at g_spawn() improvements tomorrow, hopefully.
Gimp-developer mailing list