Add IBUS_ERROR domain and reply IBUS_ERROR_ON_ENGINE in org.freedesktop.IBus.InputContext.GetEngine

Add IBUS_ERROR domain and reply IBUS_ERROR_NO_ENGINE in org.freedesktop.IBus.InputContext.GetEngine


BUG=None
TEST=Manually

[Yusuke Sato, 2011-06-14 15:34:52]

LGTM w/ one question.

The code itself LG, but the motivation of the change is a bit unclear to me. Is
the change just for making it possible to call g_debug instead of g_warning in a
certain situation?

[fujiwara, 2011-06-15 08:09:25]

LGTM.

Thanks. I also thought g_debug might be enough.
BTW, I think this also might be a workaround because if the engine already emits
REQUEST_ENGINE, the warning might be an error while we don't reproduce that case
at the moment.

[Peng, 2011-06-15 14:39:02]

On 2011/06/15 08:09:25, fujiwara wrote:
> LGTM.
> 
> Thanks. I also thought g_debug might be enough.

In some cause, the remote function call could be failed by some other reasons.
Time out, lost connection, input context is not exist. We need print them out as
warnings.

> BTW, I think this also might be a workaround because if the engine already
emits
> REQUEST_ENGINE, the warning might be an error while we don't reproduce that
case
> at the moment.

It is very possible the engine is null, even if rquest-engine has been emitted.
For example, there is no engine installed in system or enabled.

[fujiwara, 2011-06-16 02:08:21]

On 2011/06/15 14:39:02, Peng wrote:
> On 2011/06/15 08:09:25, fujiwara wrote:
> > LGTM.
> > 
> > Thanks. I also thought g_debug might be enough.
> 
> In some cause, the remote function call could be failed by some other reasons.
> Time out, lost connection, input context is not exist. We need print them out
as
> warnings.

Yes, right. However the original problem is that the message is output as
warning even though the null engine is the expected situation. 

> > BTW, I think this also might be a workaround because if the engine already
> emits
> > REQUEST_ENGINE, the warning might be an error while we don't reproduce that
> case
> > at the moment.
> 
> It is very possible the engine is null, even if rquest-engine has been
emitted.
> For example, there is no engine installed in system or enabled.

OK, I see. I meant this kind of situations are not handled but it seems you
prefer this patch to g_debug only.