[PATCH wayland-protocols v7] text: Create second version of text input protocol

[Silvan Jegen]

On Wed, Oct 12, 2016 at 03:50:33PM -0700, Bryce Harrington wrote:
> On Wed, Jul 06, 2016 at 12:26:28AM +0200, Jan Arne Petersen wrote:
> > On Thu, Jun 9, 2016 at 1:01 PM Carlos Garnacho <carlosg at gnome.org> wrote:
> > 
> > Hi Carlos,
> > 
> > Thanks for the feedback.
> > 
> > >
> > > Chiming in, and kinda late at that... hopefully we'll get this moving :).
> > 
> > I do not think we want to change text_input_unstable_v2 version 1 anymore
> > (since it already got shipped in Qt 5.7.0 and Plasma 5.7.0). Sorry for that.
> > But I already started to  work on text_input_unstable_v2 version 2 and also
> > text_input_unstable_v3 where I like to include the feedback.
> >
> > > First of all, I'm aware that some of my comments are directed towards
> > > stuff that's unchanged between v1 and v2, so please bear with me, I
> > > hope the feedback is useful.
> > 
> > Sure that is perfectly fine, I think that is the idea of the unstable protocols
> > anyways that we can still change everything and adapt them with real
> > world experience.
> 
> > > On Mon, May 30, 2016 at 11:41 AM, Jan Arne Petersen <janarne at gmail.com> wrote:
> > > > There were some shortcomings in the first version of the protocol which
> > > > makes it not really useful in real world applications. It is not really
> > > > possible to fix them in a compatible way so introduce a new v2 of the
> > > > protocol.
> > > >
> > > > Fixes some shortcomings of the first version:
> > > >
> > > > * Use only one wp_text_input per wl_seat (client side should be
> > > >   handled by client toolkit)
> > > > * Allow focus tracking without wl_keyboard present
> > > > * Improve update state handling and better define state handling
> 
> I'd love to see a re-rev of this patch.  Looking at a diff between v1
> and v2, those three changes seem quite suitable, although I'd like to
> see an exact minimal diff of the changes so holding off on detailed
> review.  But worth an acked by at least:

I share this sentiment!


>
> Acked-by: Bryce Harrington <bryce at osg.samsung.com>
>
> The input-method and text-input protocols both deal with similar
> functionality (text input).  I'm not sure if we already have a high
> level description somewhere that describes their relationship, but at a
> minimum I think the protocols ought to cross-reference each other.  (For
> instance, something like, "See also the input-method protocol, which
> provides ...")  When I first looked at the two protocols, just reading
> their descriptions it wasn't obvious at all how they related; it only
> clicked after studying the weston code.

I spent quite a long time looking at those protocols and their
client/compositor implementations (in Weston as well as in QTWayland). By
now I think I have at least a basic understanding of how they should
work together.

I was wondering where a high-level documentation describing how text
input is handled (with the help of an input method) should live. Since
those are unstable protocols, maybe adding an .md file somewhere in
wayland-protocols/unstable/ would be appropriate?

I have one question regarding the following request (I hope this is
the right place to ask): 

>    <request name="set_cursor_rectangle">
>      <description summary="set cursor position">
>	Sets the cursor outline as a x, y, width, height rectangle in surface
>	local coordinates.
>
>	Allows the compositor to put a window with word suggestions near the
>	cursor.
>      </description>
>      <arg name="x" type="int"/>
>      <arg name="y" type="int"/>
>      <arg name="width" type="int"/>
>      <arg name="height" type="int"/>
>    </request>

I assume that the "window with word suggestions" that this request
enables the compositor to draw will be supplied by the input method
(though the current version of the input-method protocol does not seem
to provide an event/request for this yet). Is this assumption correct?


Cheers,

Silvan