Fix issue 19081: Linux: Not able to switch input method correctly in web page...

Fix issue 19081: Linux: Not able to switch input method correctly in web page using SCIM

This CL fixes issue 19081 by setting im context's client window when host view's native widget is realized rather than focused in.
When setting im context's client window, GtkIMMulticontext will destroy and recreate its slave im context, which will cause this issue.

BUG=19081
: Linux: Not able to switch input method correctly in web page using SCIM
TEST=Export GTK_IM_MODULE=scim then open chrome, open www.google.com and place cursor in search box, hit ctrl-space to open SCIM then click scim bar or tray icon and select another input method, then check if input method was switched correct.


Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=24442

[James Su, 2009-08-21 11:03:00]

Hi,
  Please help review this CL.

Regards
James Su

[James Su, 2009-08-25 07:09:12]

Ping.

On 2009/08/21 11:03:00, James Su wrote:
> Hi,
>   Please help review this CL.
> 
> Regards
> James Su

[Hironori Bono, 2009-08-25 09:21:05]

Sorry for my slow review.
Unfortunately, I have been working from home without any access to Linux PCs.
So, I will try this change tomorrow.

By the way, is it possible to explain why this issue happens only with
"im-scim.so"? To read the description of this change, this issue seems to happen
with any input modules. On the other hand, to read <http://crbug.com/19081>,
this issue happens only with "im-scim.so" So, I'm a little confused.

Regards,

Hironori Bono

On 2009/08/25 07:09:12, James Su wrote:
> Ping.
> 
> On 2009/08/21 11:03:00, James Su wrote:
> > Hi,
> >   Please help review this CL.
> > 
> > Regards
> > James Su

[James Su, 2009-08-25 09:29:02]

GtkIMMulticontext will destroy and recreate its slave context instance if client
window is changed. And unfortunately im-scim.so record current input method's
information in context instance, thus it's affected.
If the im module records the information globally, then it won't be affected.

Regards
James Su

On 2009/08/25 09:21:05, hbono wrote:
> Sorry for my slow review.
> Unfortunately, I have been working from home without any access to Linux PCs.
> So, I will try this change tomorrow.
> 
> By the way, is it possible to explain why this issue happens only with
> "im-scim.so"? To read the description of this change, this issue seems to
happen
> with any input modules. On the other hand, to read <http://crbug.com/19081>,
> this issue happens only with "im-scim.so" So, I'm a little confused.
> 
> Regards,
> 
> Hironori Bono
> 
> On 2009/08/25 07:09:12, James Su wrote:
> > Ping.
> > 
> > On 2009/08/21 11:03:00, James Su wrote:
> > > Hi,
> > >   Please help review this CL.
> > > 
> > > Regards
> > > James Su

[Hironori Bono, 2009-08-26 08:15:19]

LGTM with a nit.
Thank you for your change.

Regards,

Hironori Bono

http://codereview.chromium.org/174217/diff/1/3
File chrome/browser/renderer_host/gtk_im_context_wrapper.cc (right):

http://codereview.chromium.org/174217/diff/1/3#newcode64
Line 64: if (widget->window) {
nit: it may be better to move this 'if (widget->window)' check to
HandleHostViewRealize() and call the function here?

[James Su, 2009-08-26 15:16:46]

Thanks. CL submitted.

http://codereview.chromium.org/174217/diff/1/3
File chrome/browser/renderer_host/gtk_im_context_wrapper.cc (right):

http://codereview.chromium.org/174217/diff/1/3#newcode64
Line 64: if (widget->window) {
On 2009/08/26 08:15:19, hbono wrote:
> nit: it may be better to move this 'if (widget->window)' check to
> HandleHostViewRealize() and call the function here?

Done.