Re: [Ekiga-devel-list] Win32 UTF-8 vs. Codepage or what so ever

[Eugen Dedu <Eugen Dedu pu-pm univ-fcomte fr>]

Michael Rickmann wrote:
> Ekiga's handling of Windows file names and device names is done mostly 
> in UTF-8 which works as long as there are no national characters in the 
> names. I have a German Windows and a USB headset which in Ekiga shows up 
> as "USB-Ger" instead of "USB-Gerät" (USB device). Ok, I can click 
> Ekiga's update devices and then there is a "USB-Gerät" which I can 
> select thanks to the g_locale_to_utf8 support in 
> ekiga/lib/gui/gmpreferences.c. However, the next time Ekiga is started 
> it complains because it is unable to open the "USB-Ger".

Hi Michael,

Until one month ago, linux usb subsystem have gotten the device name, 
translated it from utf-16 to latin-1 and give device names as latin-1 to 
applications.

The 
http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commit;h=a853a3d4eb2edb066248a39f0634f6f5858816a0 
commit changes that, so starting with 2.6.31 my ekiga patch should be 
removed (or make #ifdef linuxversion <= 2.6.30...)

In the original ubuntu bug report, someone with such a camera and under 
windows had no problem.  In fact, I was told that windows gives device 
names already in utf-8.

If I understand you correctly, this is not the case for you?!

> I tried a similar approach as Eugen did when fixing bug #575907 (git 
> c9aedde44825, 2009-04-21). But similar changes at the level of 
> lib/components/ptlib device managers using g_locale_to_utf8 and back 
> would not work for Windows. Presumably some library calls have been in 
> between. For Win32 Ekiga the separation of (UTF-8) vs. (Codepage) with 
> respect to Ptlib's device names is
> (GUI + gconf) vs. (C-library + device strings).

In linux, it's the same to my knowledge.

I think that in ekiga storing in locale in gconf entries was a very old 
code, which should now be removed, is that right (your third patch)?

> A good place to convert the device strings to UTF-8 is 
> gm_prefs_window_convert_string_list in ekiga/src/gui/preferences.cpp. 
> Conversion from UTF-8 can only be done where we know that it is a device 
> string, i. e. in the device gmconf-bridges. I have prepared three 
> patches and tested them under Windows and for adverse effects also Linux.
>
> ekiga01_filename.diff is not so important but easy. It allows special 
> characters in the sound files. It shows the principle of converting from 
> UTF-8, i.e having a specialized routine in lib/gmconf/gmconf-glib.c 
> which is called for WIN32 from audiooutput-gmconf-bridge.cpp.
>
> ekiga02_devstring.diff involves quite a few files. It does the device 
> string conversions summarized above.
>
> ekiga03_isutf8.diff cleans up ekiga/lib/gui/gmpreferences.c. Linux is 
> completely UTF-8 and for Win32 we have done the conversion to UTF-8 with 
> the second patch earlier. Moreover double conversion to UTF-8 does not 
> work when there are any special characters.
>
> In some parts the patches seem a bit bulky. Do you have any idea how to 
> solve the device string issue for Win32 in a more elegant way?