[PATCH 00/13] Clean up text protocol and move to wayland

[Jan Arne Petersen]

On 19.02.2013 20:01, Yichao Yu wrote:
> On Tue, Feb 19, 2013 at 1:26 PM, Jan Arne Petersen
> <jpetersen at openismus.com> wrote:
>> From: Jan Arne Petersen <jpetersen at openismus.com>
>>
>> This series finalizes the first version of the text protocol and
>> moves it from Weston to Wayland, so that it can be implemented in
>> toolkits.
> 
> Well, the only problem is that this protocol is totally not usable yet
> and it lacks every single feature a non-layout input method needs.
> Apart from what I have already mentioned ((more important ones first)
> tracking input context, cursor following, getting information about
> im-clients (pid, appname, context), and not really sure about having
> multiple processes drawing the user interface) there is one more
> feature that is missing. Is it possible for the im-client to send a
> key event that is not previously handled by the input method back to
> text_model (or is wl_text_input the new name...)? At least I haven't
> found a request on the im-client to send this out... This is really
> useful when the text model only get input focus after certain key
> event (see pidgin input window as an example).

With this series we try to get the text protocol into Wayland (so it can
be supported by toolkits). The input-method protocol for input methods
is not ready yet and will be included into Wayland later (when it will
be ready).

* tracking input context
  - No changes on text_input side needed. Can be all implemented in
compositor/input-method protocol.
* cursor following
  - text_input::set_cursor_rectangle is there. Missing implementation is
all in compositor/input-method protocol side.
* getting information about im-clients (pid, appname, context)
  - Can be easily added later to text_input (maybe there is even a
better interface to provide application information, which could be
reused by compositor)
* multiple processes drawing the user interface)
  - Already supported (and on compositor/input-method protocol side anyways)
* im-client to send a key event
  - Already supported (input_method::keysym/input_method::key)

So the only thing missing from the text_input interface is a request for
context information, which can be easily added as soon as we know how it
should look exactly.

Regards,
Jan Arne Petersen

-- 
Jan Arne Petersen
Openismus GmbH
http://www.openismus.com