[PATCH wayland 1/2 v2] protocol: Add resize from center flag.

[Scott Moreau]

On Mon, Nov 19, 2012 at 11:36 AM, Bill Spitzak <spitzak at gmail.com> wrote:

> Scott Moreau wrote:
>
>  Because resize-from-center depends on knowing what edge was actually
>> clicked on, setting both left and right bits simultaneously means that both
>> left and right cases will be hit, and whichever sets the offset last, wins.
>>
>
> You should be able to tell "which" edge by what side of the original
> center the mouse is on. So if the user grabs the top-right corner and then
> moves the mouse so that it is to the lower-left of the center, they are now
> dragging the bottom-left corner. The few things I have that do
> resize-from-center (drawing programs are all I can find) work exactly this
> way.


My point is that we're not confined to a single byte of data to store edge
flags. We have plenty of room to store additional flags. I agree that it
would be a nice way to structure the data, I don't think it warrants the
additional code required to support it. By dedicating a single bit to
'resize opposing edges', you're avoiding all the code that would have to be
in place to detect the grabbed edge, regardless if you store it in a
different edge mask or compute the pointer position against the original
center, later.


> I think also if the user then lets the shift key go, they expect to now be
> in resize-lower-left mode, though all the programs I tried just go into
> resize-from-center mode but you cannot get out of it without letting the
> mouse go.
>

This the behavior that this patch implements. If you look at the short,
shift is being pressed and released while left-click-grab is held
continuously.


>
> It sounds to me like the "receiver" of the resize is expected to interpret
> shift-up in your scheme. Since the very first thing it has to do is check
> this key (as the user may have pressed or released it between when the
> event was sent and it started working on it), there is now no reason to
> communicate "center" at all.
>

I assume by "receiver" you mean server? Technically, we don't need checks
at all in shell_surface_resize() or reszize_binding() since shift state is
checked on each resize iteration and toggles the state accordingly. That
is, the way this patch works is, only resize-from-center when shift is
held. <mod>+shift+drag doesn't work to initiate resize at all, since
bindings must be matched explicitly. i.e. <mod>+button0 !=
<mod>+shift+button0. Shift+border-edge-click+release-shift will still be in
normal resize mode. While grabbed, pressing shift will activate
resize-from-center, releasing shift will return to normal resize mode.


Scott
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.freedesktop.org/archives/wayland-devel/attachments/20121119/a951031e/attachment.html>