Turkish issue: gnome-language-selector crashes after a fresh install

[Erdem]

Binary package hint: python-aptdaemon.gtk3widgets

After fresh install of Natty Narwhal Alpha 1 using ubuntu-11.04-beta1-desktop-amd64.iso when I try to run gnome-language-selector I get this:

$ gnome-language-selector
Traceback (most recent call last):
  File "/usr/bin/gnome-language-selector", line 8, in <module>
    from LanguageSelector.gtk.GtkLanguageSelector import GtkLanguageSelector
  File "/usr/lib/python2.7/dist-packages/LanguageSelector/gtk/GtkLanguageSelector.py", line 34, in <module>
    from aptdaemon.gtk3widgets import AptProgressDialog
  File "/usr/lib/python2.7/dist-packages/aptdaemon/gtk3widgets.py", line 61, in <module>
    class AptStatusIcon(Gtk.Image):
  File "/usr/lib/python2.7/dist-packages/aptdaemon/gtk3widgets.py", line 66, in AptStatusIcon
    def __init__(self, transaction=None, size=Gtk.IconSize.DIALOG):
AttributeError: type object 'GtkIconSize' has no attribute 'DIALOG'

Also there is another bug in the bug reporting tool which prevents to report bugs in selected locale:

$ ubuntu-bug gnome-language-selector
ERROR: /etc/apport/crashdb.conf is damaged: No default database

$ LC_ALL=C ubuntu-bug /usr/bin/gnome-language-selector

So i created this bug report using C locale.

$ lsb_release -rd
Description: Ubuntu Natty (development branch)
Release: 11.04

$ LC_ALL=C apt-cache policy /usr/bin/gnome-language-selector
N: Unable to locate package /usr/bin/gnome-language-selector

But i learned its version number from synaptic package manager. Which is language-selector (0.28) natty.

Also you may want to have a look at bug 747652 and bug 630927 because they may be related to each other.

ProblemType: Bug
DistroRelease: Ubuntu 11.04
Package: language-selector-gnome 0.28
ProcVersionSignature: Ubuntu 2.6.38-7.39-generic 2.6.38
Uname: Linux 2.6.38-7-generic x86_64
Architecture: amd64
Date: Fri Apr 1 23:54:34 2011
InstallationMedia: Ubuntu 11.04 "Natty Narwhal" - Beta amd64 (20110330)
ProcEnviron:
 LANGUAGE=tr_TR:en
 LANG=tr_TR.UTF-8
 SHELL=/bin/bash
SourcePackage: language-selector
UpgradeStatus: No upgrade log present (probably fresh install)

[Erdem (farukerdemoncel)]

Also when you fall back to the C locale gnome-language-selector works:

$ LC_ALL=C gnome-language-selector

$ env LANG=tr gnome-language-selector

Locale not supported by C library.
 Using the fallback 'C' locale.
locale: Cannot set LC_CTYPE to default locale: No such file or directory
locale: Cannot set LC_MESSAGES to default locale: No such file or directory
locale: Cannot set LC_COLLATE to default locale: No such file or directory
locale: Cannot set LC_CTYPE to default locale: No such file or directory
locale: Cannot set LC_MESSAGES to default locale: No such file or directory
locale: Cannot set LC_COLLATE to default locale: No such file or directory
locale: Cannot set LC_CTYPE to default locale: No such file or directory
locale: Cannot set LC_MESSAGES to default locale: No such file or directory
locale: Cannot set LC_COLLATE to default locale: No such file or directory
locale: Cannot set LC_CTYPE to default locale: No such file or directory
locale: Cannot set LC_MESSAGES to default locale: No such file or directory
locale: Cannot set LC_COLLATE to default locale: No such file or directory
locale: Cannot set LC_CTYPE to default locale: No such file or directory
locale: Cannot set LC_MESSAGES to default locale: No such file or directory
locale: Cannot set LC_COLLATE to default locale: No such file or directory
locale: Cannot set LC_CTYPE to default locale: No such file or directory
locale: Cannot set LC_MESSAGES to default locale: No such file or directory
locale: Cannot set LC_COLLATE to default locale: No such file or directory

[Gunnar Hjalmarsson (gunnarhj)]

Thanks for your help to improve Ubuntu by reporting these issues.

The crash seems to be caused by something in the aptdaemon package, so I make this report an aptdaemon bug, and set the status "Invalid" for the language-selector bug.

The LANG variable expects a locale name, e.g. 'tr_TR.UTF-8'; 'tr' is not a valid locale name. If there still is a problem with apport, please file a separate bug report about it.

[Sebastian Heinlein (glatzor)]

Please update your system. Seems to be an issue in python gtk

[Erdem (farukerdemoncel)]

Hi I meant tr_TR.UTF-8 as locale name. Also my system is already set to tr_TR.UTF-8 locale.

$ locale
LANG=tr_TR.UTF-8
LANGUAGE=tr_TR:en
LC_CTYPE="tr_TR.UTF-8"
LC_NUMERIC="tr_TR.UTF-8"
LC_TIME="tr_TR.UTF-8"
LC_COLLATE="tr_TR.UTF-8"
LC_MONETARY="tr_TR.UTF-8"
LC_MESSAGES="tr_TR.UTF-8"
LC_PAPER="tr_TR.UTF-8"
LC_NAME="tr_TR.UTF-8"
LC_ADDRESS="tr_TR.UTF-8"
LC_TELEPHONE="tr_TR.UTF-8"
LC_MEASUREMENT="tr_TR.UTF-8"
LC_IDENTIFICATION="tr_TR.UTF-8"
LC_ALL=

I updated my system as you mentioned. But the problem still exists.

$ LANG=tr_TR.UTF-8 /usr/bin/gnome-language-selector
Traceback (most recent call last):
  File "/usr/bin/gnome-language-selector", line 8, in <module>
    from LanguageSelector.gtk.GtkLanguageSelector import GtkLanguageSelector
  File "/usr/lib/python2.7/dist-packages/LanguageSelector/gtk/GtkLanguageSelector.py", line 34, in <module>
    from aptdaemon.gtk3widgets import AptProgressDialog
  File "/usr/lib/python2.7/dist-packages/aptdaemon/gtk3widgets.py", line 61, in <module>
    class AptStatusIcon(Gtk.Image):
  File "/usr/lib/python2.7/dist-packages/aptdaemon/gtk3widgets.py", line 66, in AptStatusIcon
    def __init__(self, transaction=None, size=Gtk.IconSize.DIALOG):
AttributeError: type object 'GtkIconSize' has no attribute 'DIALOG'

And I think I don't need to set locale with a command line parameter. Because my default locale is already set to tr_TR.UTF-8.

[Martin Pitt (pitti)]

I cannot reproduce this on current natty. Does it still happen to you?

If so, can you please copy&paste the output of these commands:

  python -c 'from gi.repository import Gtk; print Gtk.IconSize.DIALOG'
  dpkg -l gir1.2-gtk*|grep ^i

Thanks!

[Erdem (farukerdemoncel)]

Sure :)

$ python -c 'from gi.repository import Gtk; print Gtk.IconSize.DIALOG'
Traceback (most recent call last):
  File "<string>", line 1, in <module>
AttributeError: type object 'GtkIconSize' has no attribute 'DIALOG'

$ dpkg -l gir1.2-gtk*|grep ^i
ii gir1.2-gtk-2.0 2.24.4-0ubuntu2 The GTK+ graphical user interface library -- gir bindings

[Erdem (farukerdemoncel)]

Yes it still happens to me. With "tr_TR.UTF-8" locale.

 I installed this version of Natty from a daily repository. Does it matter. And also 32 bit version is installed.

[Martin Pitt (pitti)]

Erdem,

please also give me the installed version of python-gobject, with

  dpkg -l python-gobject|grep ^i

and the output of

  python -c 'from gi.repository import Gtk; print Gtk.IconSize, dir(Gtk.IconSize)'

Thanks!

[Ayan Öztürk (ayanozturk)]

it is still happening to me too. I'm at work right now, i will send these
outputs tonight.

On Wed, Apr 27, 2011 at 00:18, Martin Pitt <email address hidden> wrote:

> Erdem,
>
> please also give me the installed version of python-gobject, with
>
> dpkg -l python-gobject|grep ^i
>
> and the output of
>
> python -c 'from gi.repository import Gtk; print Gtk.IconSize,
> dir(Gtk.IconSize)'
>
> Thanks!
>
> --
> You received this bug notification because you are a direct subscriber
> of the bug.
> https://bugs.launchpad.net/bugs/747796
>
> Title:
> gnome-language-selector crashes after a fresh install
>
> To unsubscribe from this bug, go to:
>
> https://bugs.launchpad.net/ubuntu/+source/language-selector/+bug/747796/+subscribe
>

--
*Ayan Öztürk**
*www.ayanozturk.com

[Erdem (farukerdemoncel)]

Here's the output requested:

$ dpkg -l python-gobject|grep ^i

ii python-gobject 2.28.3-1ubuntu1 Python bindings for the GObject library

$ python -c 'from gi.repository import Gtk; print Gtk.IconSize, dir(Gtk.IconSize)'

<class 'gi.repository.Gtk.GtkIconSize'> ['BUTTON', 'DND', 'DiALOG', 'LARGE_TOOLBAR', 'MENU', 'SMALL_TOOLBAR', '__abs__', '__add__', '__and__', '__class__', '__cmp__', '__coerce__', '__delattr__', '__dict__', '__div__', '__divmod__', '__doc__', '__enum_values__', '__eq__', '__float__', '__floordiv__', '__format__', '__ge__', '__getattribute__', '__getnewargs__', '__gt__', '__gtype__', '__hash__', '__hex__', '__index__', '__info__', '__init__', '__int__', '__invert__', '__le__', '__long__', '__lshift__', '__lt__', '__mod__', '__module__', '__mul__', '__ne__', '__neg__', '__new__', '__nonzero__', '__oct__', '__or__', '__pos__', '__pow__', '__radd__', '__rand__', '__rdiv__', '__rdivmod__', '__reduce__', '__reduce_ex__', '__repr__', '__rfloordiv__', '__rlshift__', '__rmod__', '__rmul__', '__ror__', '__rpow__', '__rrshift__', '__rshift__', '__rsub__', '__rtruediv__', '__rxor__', '__setattr__', '__sizeof__', '__str__', '__sub__', '__subclasshook__', '__truediv__', '__trunc__', '__weakref__', '__xor__', 'bit_length', 'conjugate', 'denominator', 'iNVALiD', 'imag', 'numerator', 'real', 'value_name', 'value_nick']

Thanks ..

[Martin Pitt (pitti)]

Erdem [2011-04-27 16:35 -0000]:
> <class 'gi.repository.Gtk.GtkIconSize'> ['BUTTON', 'DND', 'DiALOG',

What the heck? This almost looks like an April's fool..

Can you please give me the output of

  apt-cache policy gir1.2-gtk-2.0

and try

  sudo apt-get install --reinstall gir1.2-gtk-2.0

?
--
Martin Pitt | http://www.piware.de
Ubuntu Developer (www.ubuntu.com) | Debian Developer (www.debian.org)

[Cengiz Can (cengizcan)]

I have exact same module/package versions with Erdem, and reinstalling the package didn't help with anything.

The problem appears to be the letter uppercase dotted "I", which is a Turkish only character.
http://en.wikipedia.org/wiki/Help:Turkish_characters

So instead of DIALOG, the icon identifier has a 1 letter shifted name with DiALOG.

The alphabet has two letter I in Turkish, first one is universal I and the following is I with dot above.

Anyway, we need a manual way to fix this.

At least please tell how to revert to en_US locate alltogether.

[Cengiz Can (cengizcan)]

EXTRA:

Removing language-pack-*-tr / language-pack-tr-* series packs from synapic resolves many problems with applications using python libraries.

[Erdem (farukerdemoncel)]

$ apt-cache policy gir1.2-gtk-2.0
gir1.2-gtk-2.0:
  Kurulu: 2.24.4-0ubuntu2
  Aday: 2.24.4-0ubuntu2
  Sürüm tablosu:
 *** 2.24.4-0ubuntu2 0
        500 http://tr.archive.ubuntu.com/ubuntu/ natty/main i386 Packages
        100 /var/lib/dpkg/status

The problem continues after reinstalling gir1.2-gtk-2.0.

And I suspect the real problem may be that Cengiz points out. For example in English uppercase of i is I. But in Turkish uppercase of ı is I and uppercase of i is İ.

[Gunnar Hjalmarsson (gunnarhj)]

I propose a language-selector fix, after all - see linked merge proposal. Have a feeling it's a workaround, though - your call, Martin. If it's acceptable for now, I assume it should go to Nattu as an SRU.

It would be great if those who are affected by this problem can test the fix. You can simply replace /usr/share/pyshared/LanguageSelector/gtk/GtkLanguageSelector.py with the modified file in the linked branch.

[Erdem (farukerdemoncel)]

I downloaded GtkLanguageSelector.py from here:

http://bazaar.launchpad.net/~gunnarhj/language-selector/lp-747796/files/head:/LanguageSelector/gtk/

And replaced with the old one. This time it doesn't crashes. But fonts are not displayed correctly this time. There are ?? symbols etc..

[Gunnar Hjalmarsson (gunnarhj)]

Aha, I hadn't tested with Turkish as the selected language; thanks for catching that.

Improved the fix. To prevent regression risk, it now applies only when a Turkish regional formats setting is in effect.

[Ayan Öztürk (ayanozturk)]

Problem is solved with latest fix for me. Special characters looks fine too

[Gunnar Hjalmarsson (gunnarhj)]

Great! Thanks for letting us know.

[Martin Pitt (pitti)]

Thanks Gunnar for providing a workaround. Disabling locale settings for pygi calls might be an acceptable workaround for a natty stable update, but I'll give the real pygobject bug some examination first. It needs to be fixed either way, and if the fix is nonintrusive, we can SRU it as well.

[Gunnar Hjalmarsson (gunnarhj)]

Wonder if there is a really good solution to this problem. The root cause is that the Turkish locales map two ASCII characters to non-ASCII characters with respect to case conversion:
    "I" (capital I) is mapped to "ı" (lowercase i-dotless)
    "i" (lowercase i) is mapped to "İ" (capital I-dotted)

I played with an alternative Turkish locale (diff below), and it proved to work as well as the l-s workaround I propose. Guess the issue has been discussed many times, and it would be interesting to hear some Turkish voices on this. Have a feeling that programmers will keep overlooking this special case and with that cause troubles for Turkish computer users.

Still looking forward to know about your conclusion regards pygobject.

--- /usr/share/i18n/locales/tr_TR 2011-03-15 17:38:26 +0100
+++ /usr/share/i18n/locales/tr_TR@uclc 2011-05-02 09:09:04 +0200
@@ -797,7 +797,7 @@
 toupper /
    (<U0061>,<U0041>);(<U0062>,<U0042>);(<U0063>,<U0043>);(<U0064>,<U0044>);/
    (<U0065>,<U0045>);(<U0066>,<U0046>);(<U0067>,<U0047>);(<U0068>,<U0048>);/
- (<U0069>,<U0130>);(<U006A>,<U004A>);(<U006B>,<U004B>);(<U006C>,<U004C>);/
+ (<U0069>,<U0049>);(<U006A>,<U004A>);(<U006B>,<U004B>);(<U006C>,<U004C>);/
    (<U006D>,<U004D>);(<U006E>,<U004E>);(<U006F>,<U004F>);(<U0070>,<U0050>);/
    (<U0071>,<U0051>);(<U0072>,<U0052>);(<U0073>,<U0053>);(<U0074>,<U0054>);/
    (<U0075>,<U0055>);(<U0076>,<U0056>);(<U0077>,<U0057>);(<U0078>,<U0058>);/
@@ -815,7 +815,7 @@
    (<U0115>,<U0114>);(<U0117>,<U0116>);(<U0119>,<U0118>);(<U011B>,<U011A>);/
    (<U011D>,<U011C>);(<U011F>,<U011E>);(<U0121>,<U0120>);(<U0123>,<U0122>);/
    (<U0125>,<U0124>);(<U0127>,<U0126>);(<U0129>,<U0128>);(<U012B>,<U012A>);/
- (<U012D>,<U012C>);(<U012F>,<U012E>);(<U0131>,<U0049>);(<U0133>,<U0132>);/
+ (<U012D>,<U012C>);(<U012F>,<U012E>);(<U0069>,<U0049>);(<U0133>,<U0132>);/
    (<U0135>,<U0134>);(<U0137>,<U0136>);(<U013A>,<U0139>);(<U013C>,<U013B>);/
    (<U013E>,<U013D>);(<U0140>,<U013F>);(<U0142>,<U0141>);(<U0144>,<U0143>);/
    (<U0146>,<U0145>);(<U0148>,<U0147>);(<U014B>,<U014A>);(<U014D>,<U014C>);/
@@ -1001,7 +1001,7 @@
 tolower /
    (<U0041>,<U0061>);(<U0042>,<U0062>);(<U0043>,<U0063>);(<U0044>,<U0064>);/
    (<U0045>,<U0065>);(<U0046>,<U0066>);(<U0047>,<U0067>);(<U0048>,<U0068>);/
- (<U0049>,<U0131>);(<U004A>,<U006A>);(<U004B>,<U006B>);(<U004C>,<U006C>);/
+ (<U0049>,<U0069>);(<U004A>,<U006A>);(<U004B>,<U006B>);(<U004C>,<U006C>);/
    (<U004D>,<U006D>);(<U004E>,<U006E>);(<U004F>,<U006F>);(<U0050>,<U0070>);/
    (<U0051>,<U0071>);(<U0052>,<U0072>);(<U0053>,<U0073>);(<U0054>,<U0074>);/
    (<U0055>,<U0075>);(<U0056>,<U0076>);(<U0057>,<U0077>);(<U0058>,<U0078>);/
@@ -1019,7 +1019,7 @@
    (<U0118>,<U0119>);(<U011A>,<U011B>);(<U011C>,<U011D>);(<U011E>,<U011F>);/
    (<U0120>,<U0121>);(<U0122>,<U0123>);(<U0124>,<U0125>);(<U0126>,<U0127>);/
    (<U0128>,<U0129>);(<U012A>,<U012B>);(<U012C>,<U012D>);(<U012E>,<U012F>);/
- (<U0130>,<U0069>);(<U0132>,<U0133>);(<U0134>,<U0135>);(<U0136>,<U0137>);/
+ (<U0049>,<U0069>);(<U0132>,<U0133>);(<U0134>,<U0135>);(<U0136>,<U0137>);/
    (<U0139>,<U013A>);(<U013B>,<U013C>);(<U013D>,<U013E>);(<U013F>,<U0140>);/
    (<U0141>,<U0142>);(<U0143>,<U0144>);(<U014...

[Martin Pitt (pitti)]

I now found the root cause of this, keeping notes in the upstream bug. We'll fix this properly in pygobject, closing the language-selector task.

[Erdem (farukerdemoncel)]

Yes finally it works. Thanks for the fix :)

Sorry I don't know Python. I study D programming language. We also had a issue with Phobos when sorting Turkish strings. So a few friends of mine wrote a small library to change Phobos's default behavior. As I know it also supports other problematic languages and alphabets like Irish and other Turkish alphabets. Here is the project page of the library:

http://code.google.com/p/trileri/

If you want may be you may contact developers for further exchange of ideas.

[Gunnar Hjalmarsson (gunnarhj)]

I have prepared a pygobject branch for natty-proposed, that will update Natty with a patch that Martin Pitt has proposed upstream (https://bugzilla.gnome.org/show_bug.cgi?id=649165). It will be released as soon as upstream has approved the patch.

In the meantime, the best manual fix of this bug is to replace /usr/share/pyshared/gi/module.py with the module.py attachment.

[Martin Pitt (pitti)]

I got the go ahead, I pushed the fix to upstream master and 2-28 branch:

http://git.gnome.org/browse/pygobject/commit/?h=pygobject-2-28&id=ad96a3f1d187a640ec0a463b766fd5e027f3c16a

[Launchpad Janitor (janitor)]

This bug was fixed in the package pygobject - 2.28.3-1ubuntu2

---------------
pygobject (2.28.3-1ubuntu2) oneiric; urgency=low

  * debian/patches/21_fix_symbol_names_to_be_locale_independent.patch:
    Case conversion issue, that proved to break language-selector for
    Turkish users in Natty, fixed.
    (LP: #747796; https://bugzilla.gnome.org/show_bug.cgi?id=649165)
 -- Gunnar Hjalmarsson <email address hidden> Thu, 05 May 2011 07:16:38 +0200

[Martin Pitt (pitti)]

Accepted pygobject into natty-proposed, the package will build now and be available in a few hours. Please test and give feedback here. See https://wiki.ubuntu.com/Testing/EnableProposed for documentation how to enable and use -proposed. Thank you in advance!

[Reşat SABIQ (tilde-birlik)]

FYI: some other Turk languages like Crimean Tatar (Crimean Turkish) and Azeri are also affected by this. Hopefully this will be in the main repo soon, and most users won't upgrade until then...

P.S.
I installed python-gobject 2.28.3-1ubuntu1.1 which is now in proposed, and was then able to user language-selector using Crimean Tatar locale.
P.P.S.
Side note: there was a related issue in Java's locale collation rules. It was fixed in Java 6:
http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=4688797
Though this python issue wasn't really a i18n implementation issue, but rather a logical bug, it appears programming languages bump into this nuance in various ways. It's good to see Java and now Python handling this correctly...

Thanks.

[Martin Pitt (pitti)]

Thanks for testing!

[Erdem (farukerdemoncel)]

Also I may confirm that it works! :) Thanks ..

[Launchpad Janitor (janitor)]

This bug was fixed in the package pygobject - 2.28.3-1ubuntu1.1

---------------
pygobject (2.28.3-1ubuntu1.1) natty-proposed; urgency=low

  * debian/patches/21_fix_symbol_names_to_be_locale_independent.patch:
    Case conversion issue, that proved to break language-selector for
    Turkish users in Natty, fixed.
    (LP: #747796; https://bugzilla.gnome.org/show_bug.cgi?id=649165)
 -- Gunnar Hjalmarsson <email address hidden> Thu, 05 May 2011 07:16:38 +0200