src: Use cache file for compose table for launching time

src: Use cache file for compose table for launching time

Save the cache files of compose files under $HOME/.cache/ibus/compose
for the performance.

Load a compose file with the following order:
1. $HOME/.config/ibus/Compose
2. $HOME/.config/gtk-3.0/Compose
3. $HOME/.XCompose
4. /usr/share/X11/locale/$locale/Compose

Check the system compose files of "el_GR", "fi_FI", "pt_BR" only for the
performance because other compose files just load "en_US" compose file.

BUG=https://bugzilla.gnome.org/show_bug.cgi?id=721120
R=Shawn.P.Huang@gmail.com

Committed: 97e28cc04b4601295b22a0aa8e9e21ed4e0758a1

[fujiwara, 2016-02-01 02:50:55]

Porting
https://git.gnome.org/browse/gtk+/commit/?id=a41f02f9b1843e0f0085f801430e55f4...

[Peng, 2016-02-01 15:15:23]

https://codereview.appspot.com/286040043/diff/1/src/ibuscomposetable.c
File src/ibuscomposetable.c (right):

https://codereview.appspot.com/286040043/diff/1/src/ibuscomposetable.c#newcod...
src/ibuscomposetable.c:513: ibus_compose_table_serialize (IBusComposeTable
*compose_table,
How about using GVariant to do the serializing and de-serializing? I think it
will make the code simpler.

[fujiwara, 2016-02-08 05:57:31]

https://codereview.appspot.com/286040043/diff/1/src/ibuscomposetable.c
File src/ibuscomposetable.c (right):

https://codereview.appspot.com/286040043/diff/1/src/ibuscomposetable.c#newcod...
src/ibuscomposetable.c:513: ibus_compose_table_serialize (IBusComposeTable
*compose_table,
On 2016/02/01 15:15:23, Peng wrote:
> How about using GVariant to do the serializing and de-serializing? I think it
> will make the code simpler.

Done.

[Peng, 2016-02-08 15:11:01]

https://codereview.appspot.com/286040043/diff/20001/src/ibuscomposetable.c
File src/ibuscomposetable.c (right):

https://codereview.appspot.com/286040043/diff/20001/src/ibuscomposetable.c#ne...
src/ibuscomposetable.c:535: index_stride * n_seqs,
Arguments do not align.

https://codereview.appspot.com/286040043/diff/20001/src/ibuscomposetable.c#ne...
src/ibuscomposetable.c:537: if (variant_data == NULL)
g_warning()?

https://codereview.appspot.com/286040043/diff/20001/src/ibuscomposetable.c#ne...
src/ibuscomposetable.c:592: g_variant_get (variant_table, "(sq)", &header,
&version);
"(sq)" => "(&sq)", so you do need free the header after using.

https://codereview.appspot.com/286040043/diff/20001/src/ibuscomposetable.c#ne...
src/ibuscomposetable.c:595: g_ascii_strncasecmp (header,
IBUS_COMPOSE_TABLE_MAGIC,
Why not g+strcmp()?

https://codereview.appspot.com/286040043/diff/20001/src/ibuscomposetable.c#ne...
src/ibuscomposetable.c:628: g_variant_get (variant_table, "(sqqqv)",
I think it is not necessary to get the first two value again here and you can
ignore first two values by passing NULL to it. Could you please check it?

[fujiwara, 2016-02-09 08:52:23]

https://codereview.appspot.com/286040043/diff/20001/src/ibuscomposetable.c
File src/ibuscomposetable.c (right):

https://codereview.appspot.com/286040043/diff/20001/src/ibuscomposetable.c#ne...
src/ibuscomposetable.c:535: index_stride * n_seqs,
On 2016/02/08 15:11:00, Peng wrote:
> Arguments do not align.

Done.

https://codereview.appspot.com/286040043/diff/20001/src/ibuscomposetable.c#ne...
src/ibuscomposetable.c:537: if (variant_data == NULL)
On 2016/02/08 15:11:00, Peng wrote:
> g_warning()?

Done.

https://codereview.appspot.com/286040043/diff/20001/src/ibuscomposetable.c#ne...
src/ibuscomposetable.c:592: g_variant_get (variant_table, "(sq)", &header,
&version);
On 2016/02/08 15:11:01, Peng wrote:
> "(sq)" => "(&sq)", so you do need free the header after using.

Done.

https://codereview.appspot.com/286040043/diff/20001/src/ibuscomposetable.c#ne...
src/ibuscomposetable.c:595: g_ascii_strncasecmp (header,
IBUS_COMPOSE_TABLE_MAGIC,
On 2016/02/08 15:11:00, Peng wrote:
> Why not g+strcmp()?

Done.

https://codereview.appspot.com/286040043/diff/20001/src/ibuscomposetable.c#ne...
src/ibuscomposetable.c:628: g_variant_get (variant_table, "(sqqqv)",
On 2016/02/08 15:11:00, Peng wrote:
> I think it is not necessary to get the first two value again here and you can
> ignore first two values by passing NULL to it. Could you please check it?

Done.

[Peng, 2016-02-09 15:48:19]

On 2016/02/09 08:52:23, fujiwara wrote:
> https://codereview.appspot.com/286040043/diff/20001/src/ibuscomposetable.c
> File src/ibuscomposetable.c (right):
> 
>
https://codereview.appspot.com/286040043/diff/20001/src/ibuscomposetable.c#ne...
> src/ibuscomposetable.c:535: index_stride * n_seqs,
> On 2016/02/08 15:11:00, Peng wrote:
> > Arguments do not align.
> 
> Done.
> 
>
https://codereview.appspot.com/286040043/diff/20001/src/ibuscomposetable.c#ne...
> src/ibuscomposetable.c:537: if (variant_data == NULL)
> On 2016/02/08 15:11:00, Peng wrote:
> > g_warning()?
> 
> Done.
> 
>
https://codereview.appspot.com/286040043/diff/20001/src/ibuscomposetable.c#ne...
> src/ibuscomposetable.c:592: g_variant_get (variant_table, "(sq)", &header,
> &version);
> On 2016/02/08 15:11:01, Peng wrote:
> > "(sq)" => "(&sq)", so you do need free the header after using.
> 
> Done.
> 
>
https://codereview.appspot.com/286040043/diff/20001/src/ibuscomposetable.c#ne...
> src/ibuscomposetable.c:595: g_ascii_strncasecmp (header,
> IBUS_COMPOSE_TABLE_MAGIC,
> On 2016/02/08 15:11:00, Peng wrote:
> > Why not g+strcmp()?
> 
> Done.
> 
>
https://codereview.appspot.com/286040043/diff/20001/src/ibuscomposetable.c#ne...
> src/ibuscomposetable.c:628: g_variant_get (variant_table, "(sqqqv)",
> On 2016/02/08 15:11:00, Peng wrote:
> > I think it is not necessary to get the first two value again here and you
can
> > ignore first two values by passing NULL to it. Could you please check it?
> 
> Done.

lgtm

[fujiwara, 2016-02-10 02:41:44]

Committed patchset #3 (id:40001) manually as
97e28cc04b4601295b22a0aa8e9e21ed4e0758a1 (presubmit successful).