Suppress Gtk warnings like...

Suppress Gtk warnings like

(test_shell:2476): Gtk-WARNING **: /usr/lib/gtk-2.0/2.10.0/immodules/im-uim.so: wrong ELF class: ELFCLASS64

(test_shell:2476): Gtk-WARNING **: Loading IM context type 'uim' failed

This happens when Gtk tries to open 64 bit shared object from 32 bit binaries.

As this change suppress warnings only when the above three lines come the same order, we can still see other important warnings.


Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=15973

[tony, 2009-05-12 21:28:39]

Does this cause layout tests to fail?  Can we disable IMEs from within test
shell (only if --layout-tests is passed in)?

[hamaji, 2009-05-12 21:52:51]

On 2009/05/12 21:28:39, tony wrote:
> Does this cause layout tests to fail?  Can we disable IMEs from within test
> shell (only if --layout-tests is passed in)?

Thanks for the review. Yes, it makes layout tests fail. I think there are no way
to disable IME without changing configuration files.

There is a similar chromium bug

http://code.google.com/p/chromium/issues/detail?id=9643

and it seems that this is a bug of Ubuntu.

https://bugs.launchpad.net/ubuntu/+source/ia32-libs/+bug/190227

[tony, 2009-05-12 22:15:11]

Can we disable IME by setting some environment variables before we launch
test_shell in run_webkit_tests.py (or test_shell_thread.py)?  Maybe unset
XMODIFIERS and GTK_IM_MODULE?

https://help.ubuntu.com/community/Japanese_Input_and_Fonts_in_Ubuntu_7.04

[hamaji, 2009-05-12 22:45:07]

On 2009/05/12 22:15:11, tony wrote:
> Can we disable IME by setting some environment variables before we launch
> test_shell in run_webkit_tests.py (or test_shell_thread.py)?  Maybe unset
> XMODIFIERS and GTK_IM_MODULE?
> 
> https://help.ubuntu.com/community/Japanese_Input_and_Fonts_in_Ubuntu_7.04

Yeah, I tried them and GTK_IM_MODULE_FILE but it seems that the GTK always reads
the wrong configuration (from
/usr/lib/gtk-2.0/2.10.0/immodule-files.d/uim-gtk2.0.immodules) and then loads
the 64bit shared objects. The name specified in GTK_IM_MODULE seems to be used
after the load process...

I agree that this hack looks bad, please feel free to reject this patch. It
seems that there are some workarounds which require changing configuration of
system (moving the configuration directory to elsewhere, set GTK_IM_MODULE_FILE
explicitly for 64bit applications, and unset GTK_IM_MODULE_FILE before we use
test_shell).

[tony, 2009-05-12 23:02:23]

Maybe another way to handle this is to have a custom glib log handler using
g_log_set_handler?  If it's an IME warning, just ignore it in layout test mode? 
See the code in chrome/app/chrome_dll_main.cc for an example of this.

[hamaji, 2009-05-13 16:40:12]

On 2009/05/12 23:02:23, tony wrote:
> Maybe another way to handle this is to have a custom glib log handler using
> g_log_set_handler?  If it's an IME warning, just ignore it in layout test
mode? 
> See the code in chrome/app/chrome_dll_main.cc for an example of this.

Thanks for the suggestion. I re-wrote this patch with g_log_set_handler
approach. Could you take a look again?

[tony, 2009-05-13 17:12:34]

LGTM!

http://codereview.chromium.org/115263/diff/10/1007
File webkit/tools/test_shell/test_shell_platform_delegate_gtk.cc (right):

http://codereview.chromium.org/115263/diff/10/1007#newcode35
Line 35: g_log_set_handler("Gtk",
Nit: Can this all fit on one line?

[hamaji, 2009-05-13 17:58:36]

Fixed. Thanks for the review and suggestions! I'll submit after tree opens.

http://codereview.chromium.org/115263/diff/10/1007
File webkit/tools/test_shell/test_shell_platform_delegate_gtk.cc (right):

http://codereview.chromium.org/115263/diff/10/1007#newcode35
Line 35: g_log_set_handler("Gtk",
On 2009/05/13 17:12:34, tony wrote:
> Nit: Can this all fit on one line?

Done.