Re: [AD] Problem with remove_mouse() and video bitmaps.

[Andrei Ellman <ae-a-alleg@xxxxxxxxxx>]

Peter Wang wrote:
> On 2005-12-29, Andrei Ellman <ae-a-alleg@xxxxxxxxxx> wrote:
>
> > Peter Wang wrote:
> >
> > > Just to be sure, calling show_mouse(NULL); before destroying the video
> > > bitmap had no negative side effects, correct?
> >
> > That is correct. Adding show_mouse(NULL) also solved the problem that 
> > was reported on the other machine failing to change the resolution back, 
> > so it looks like the documentation needs to be changed.
>
>
> How's this?  The text goes in the entry for show_mouse:
>
> --- docs/src/allegro._tx	(revision 5645)
> +++ docs/src/allegro._tx	(local)
> @@ -2860,6 +2860,10 @@
>     pointer whenever you draw onto the screen. This is not needed if you are
>     using a hardware cursor.
>  
> +   Note: you must not be showing a mouse pointer on a bitmap at the time that
> +   the bitmap is destroyed with destroy_bitmap(). This does not apply to
> +   `screen' since you never destroy `screen' with destroy_bitmap().
> +
>  @@void @scare_mouse();
>  @xref unscare_mouse, scare_mouse_area, show_mouse
>  @shortdesc Helper for hiding the mouse pointer before drawing.


Seems OK to me, but I would add a bit about calling show_mouse(NULL).


--- docs/src/allegro._tx	(revision 5645)
+++ docs/src/allegro._tx	(local)
@@ -2860,6 +2860,11 @@
    pointer whenever you draw onto the screen. This is not needed if 
you are
    using a hardware cursor.

+   Note: you must not be showing a mouse pointer on a bitmap at the 
time that
+   the bitmap is destroyed with destroy_bitmap() (make sure that you call
+   show_mouse(NULL) before the bitmap is destroyed). This does not apply to
+   `screen' since you never destroy `screen' with destroy_bitmap().
+
 @@void @scare_mouse();
 @xref unscare_mouse, scare_mouse_area, show_mouse
 @shortdesc Helper for hiding the mouse pointer before drawing.


Also, I would add something to the entry for derstroy_bitmap() about 
making sure that show_mouse(NULL) is called if it's the bitmap that 
contains the mouse-cursor.


Stef wrote:
> I have a hunch that the crash is caused by the draw_mouse() function  accessing the _mouse_screen bitmap (which in this case is invalid, since  it is destroyed).

It's show_mouse() (called from remove_mouse() ) attempting to remove the 
mouse cursor from _mouse_screen (the crash happens in 
bmp->vtable->acquire(bmp) which is in acquire_bitmap() ). By the time 
remove_mouse() is called, _mouse_screen is an invalid bitmap.

Even if destroy_bitmap() took in it's argument by reference and set it 
to NULL (or the programmer set the pointer to NULL immediately after 
destroy_bitmap()), it will not help. This is because _mouse_screen is 
stored internally.

So what we would need is a way of setting _mouse_screen to NULL, and 
guess what, that's exactly what show_mouse(NULL) does. Therefore, I 
think that only the documentstion needs to be changed.


AE.