Re: constraints_experiments is ready

[Elijah Newren <newren gmail com>]

Didn't realize a couple thinkos I had made until after I hit send,
sorry.  A couple corrections/clarifications...

On 11/17/05, Elijah Newren <newren gmail com> wrote:

> > - in window_new_with_attrs you are now using
> >   window->size_hints.win_gravity for the initial resize,
> >   not convinced that's right but reading further ;-)
> >
> > - similarly in apply_session_info
> >
> > - OK, reading further you have a big comment about not understanding
> >   this, so I'm not sure I feel more confident ;-)
> >   (I'm not saying I fully get it either...)
>
> I think I can explain them.  Let me start by looking at cases other
> than new window and session restore.  If the window's gravity is not
> passed to meta_window_move_resize_internal() then bugs like 109953 and
> 312007 result.  So, it is necessary for configure requests and user
> resize operations to do that.

That's poor wording.  It is necessary for user operations to specify
the gravity associated with those operations, and for application
configure requests it is necessary to pass the window's size hints
gravity.

>  The ICCCM says that initial window map
> should be treated like a configure request (at least, according to one
> of your comments in the code).  So it makes sense to treat it the
> same.
>
> As for session restore, yeah, I'm not so sure.  I can hard code
> NorthWestGravity if you like (that's what used to be done before), but
> since hardcoding NorthWestGravity was done in multiple other places
> (in some cases in a hidden way) and was wrong and buggy and had to be
> fixed by using the window's gravity, I figured it probably made sense
> to change the session_restore one too.

> > - The point of the original code was that the gravity passed to
> >   move_resize_internal was simply the gravity we programmatically
> >   inside metacity wanted to use. i.e. which corner to move in
> >   order to implement a resize.
>
> The fact that you didn't use the window's gravity caused multiple
> bugs, 109953 and 312007 in particular.  The gravity you should have
> wanted to use _was_ the size hints gravity in all cases other than
> maybe session restore.

Again, bad wording.  User resize operations shouldn't specify the
window's gravity but rather the gravity associated with the given
operation.  Nuke that paragraph and replace with "True, but in many
cases the gravity being specified was not actually the one that we
programmatically wanted to use and caused bugs like 109953 and
312007."