bug#25408: Remove Decorations Around Emacs Frame (NS port)

[martin rudalics]

>>> It appears that making the child frame invisible ‘disconnects’ it from
>>> the parent frame, so the next time it’s made visible it no longer
>>> moves with the parent. I guess I’ll have to make sure that when a
>>> frame is made visible it’s reconnected with it’s parent. Or find a way
>>> to prevent it disconnecting. It seems a really odd thing for it to do.
>
> I’ve pushed a fix for this now.

Thanks.  One of the largely untested behaviors with frames is what a
frame looks like when it was invisible/iconified for a while, in that
period operations (resizing, moving, reparenting) have been applied to
it and the frame is made normally visible again.  I've been struggling a
while with the similar problem how to set a frame's size when it was
fullscreen for a while and in that period a resize request was posted
for it.

> It turned out that there was no NS version of
> mouse-absolute-pixel-position, so I’ve created one and suddenly all of
> the above things work. :)

Fine.  Please add a

(declare-function ns-mouse-absolute-pixel-position "nsfns.c")

to frame.el.

> I’m slightly worried that there may be issues if the frame is resized
> across a screen edge, as it treats each screen as it’s own ‘space’,
> starting at (0, 0) at the top left. This is how the existing
> set-mouse-absolute-pixel-position works so it makes sense to me to
> keep them the same.
>
> It might make more sense to treat multiple screens as one ‘space’,
> though. I’m not sure.

Can you give me an example?  I never use multiple screens so my ability
to imagine how a mouse behaves there is very limited.  There's also the
problem that some systems (Windows, IIRC) consider (0, 0) the top-left
corner of some sort of "primary" screen and allow negative coordinates
while others consider (0, 0) an abstract location at the top-left corner
of all screens.

Thanks, martin