bug#28872: [PATCH v2] Fix NS frame resizing issues (bug#40200, bug#28872)

[Alan Third]

On Fri, Mar 27, 2020 at 03:22:08PM +0200, Andrii Kolomoiets wrote:
> On 26 Mar 2020, at 23:21, Alan Third <alan@idiocy.org> wrote:
> > Hopefully this works better.
> 
> Found another changed behavior:
> 1. emacs -Q
> 2. (setq test/frame (make-frame `((parent-frame . ,(selected-frame)))))
> 3. (modify-frame-parameters test/frame '((height . 1) (top . (- 40))))
> 
> Old behavior: child frame is at bottom of the parent frame.

This is wrong.

> New behavior: child frame is at top of the parent frame.

This is wrong too.

> Child frame moves to bottom after code from step 3 applied one more time.

Still wrong.

The bottom of the child frame should be 40 pixels from the bottom of
the parent frame.

This is quite annoying, it means Emacs 27 behaves wrongly. I wonder if
Emacs 26 does too, or has it been broken the whole time?

> Step 3 can be replaced with these two steps to achieve old behavior:
> (modify-frame-parameters test/frame '((height . 1)))
> (modify-frame-parameters test/frame '((top . (- 40))))

I inadvertently made order of operations matter when they shouldn’t.
I’ve got a fix for this, but it’s opening up a whole other can of
worms regarding frame positioning and the reporting of positions.

I think I know what to do but the last time I worked on this we had a
lot of edge cases that weren’t easy to test, with multiple monitors
and spaces. I think I’ll have to look at the old bug reports to work
out how we handle that.
-- 
Alan Third