<class 'UnicodeEncodeError'>: 'ascii' codec can't encode characters in position 18-25: ordinal not in range(128)

[Антон Дерлюк]

Traceback (most recent call last):
  File "/usr/lib/python3/dist-packages/aptdaemon/worker.py", line 1181, in _simulate
    trans.unauthenticated, trans.high_trust_packages = self.__simulate(trans)
  File "/usr/lib/python3/dist-packages/aptdaemon/worker.py", line 1239, in __simulate
    deb = self.install_file(trans, simulate=True, **trans.kwargs)
  File "/usr/lib/python3/dist-packages/aptdaemon/worker.py", line 740, in install_file
    deb = self._check_deb_file(trans, path, force)
  File "/usr/lib/python3/dist-packages/aptdaemon/worker.py", line 1326, in _check_deb_file
    if not os.path.isfile(path):
  File "/usr/lib/python3.2/genericpath.py", line 29, in isfile
    st = os.stat(path)
UnicodeEncodeError: 'ascii' codec can't encode characters in position 18-25: ordinal not in range(128)

ProblemType: Crash
DistroRelease: Ubuntu 12.10
Package: aptdaemon 0.45+bzr861-0ubuntu5
ProcVersionSignature: Ubuntu 3.5.0-15.23-generic 3.5.4
Uname: Linux 3.5.0-15-generic i686
Annotation: Installation of the package file failed
ApportVersion: 2.5.2-0ubuntu4
Architecture: i386
Date: Fri Sep 28 16:53:16 2012
ExecutablePath: /usr/sbin/aptd
InstallationMedia: Ubuntu 12.10 "Quantal Quetzal" - Beta i386 (20120926)
PackageArchitecture: all
SourcePackage: aptdaemon
Title: <class 'UnicodeEncodeError'>: 'ascii' codec can't encode characters in position 18-25: ordinal not in range(128)
TransactionDepends: [[], [], [], [], [], [], []]
TransactionErrorCode: error-unknown
TransactionKwargs: {'path': dbus.String('/home/silentlight/Загрузки/google-chrome-stable_current_i386.deb'), 'force': dbus.Boolean(False)}
TransactionLocale: ru_RU.UTF-8
TransactionOutput:

TransactionPackages: [[], [], [], [], [], []]
TransactionRole: role-install-file
UpgradeStatus: No upgrade log present (probably fresh install)

[Launchpad Janitor (janitor)]

Status changed to 'Confirmed' because the bug affects multiple users.

[Josef Andersson (northar)]

Trying to install something from /home/myuser/Hämt/fs-uae_1.3.29-0_amd64.deb
gives this bug with lates quantal updates (x64)

When skipping folder /Hämt/ in the path it works.Uniode prob?

[Nicolas Delvaux (malizor)]

This is indeed a unicode problem.
The strange thing is that it works when killing and launching aptd from the command line.

According to http://docs.python.org/py3k/howto/unicode.html#unicode-filenames , this is because the LANG environment variable is not set when aptd is spawned by the sotware-center. In this case, the default encoding for pathes is ASCII.
I hacked arround and was able to confirm this (by writing os.environ in a file).

So here is the bug.
Though, I'm not sure about the best way to fix it.

[walterclozet (walterclozet-u)]

trying to install chrome amd64 for ubuntu, download from official site. crushes every time when double click to install.

[Michael Vogt (mvo)]

I can confirm this, when LANG/LANUAGE is unset and LC_CTYPE is set to posix/C the sys.getfilesystemencoding() call
will return "ascii".

[Michael Vogt (mvo)]

Looking at this a bit more it appears to be a bit of a can of worms. It appears that the dbus activation sytem
is launching the daemon with a almost empty env:
environ(
{'DBUS_STARTER_BUS_TYPE': 'system',
 'DBUS_STARTER_ADDRESS': 'unix:path=/var/run/dbus/system_bus_socket'})

which means that even the systemwide default language is not set.

[Launchpad Janitor (janitor)]

Status changed to 'Confirmed' because the bug affects multiple users.

[Michael Vogt (mvo)]

AFAICT the default fsencoding is read only at python startup and can not be changed later. The dbus activation does run the aptd without a env (well, just a minimal one).

<cjwatson> Or else there needs to be a way to break the caching when we change Python's locale at run-time
<cjwatson> That might be a good idea anyway
<mvo> cjwatson: yeah, that would be the cleanest I think, ensuring this is reset on setlocale()
<pitti> yeah, a locale.setlocale() call ought to change the fs encoding, toovalid env

Allowing to change the encoding later is probably the best option.

[Michael Vogt (mvo)]

<pitti> just saying that if all else fails, hardcoding utf-8 shoudl avoid 99% of the problems with a really trivial/safe patch?

[Sebastian Heinlein (glatzor)]

We should unset those variables in the test suite or even use dbus activation for the daemon testing. But makes cleaning up in the tests a little bit tricky. Someting to target for R.
--
Diese Nachricht wurde von meinem Android-Mobiltelefon mit K-9 Mail gesendet.

[Sebastian Heinlein (glatzor)]

We could also set lang to C.UTF8 in aptdaemon only.

Michael Vogt <email address hidden> schrieb:

><pitti> just saying that if all else fails, hardcoding utf-8 shoudl
>avoid 99% of the problems with a really trivial/safe patch?
>
>--
>You received this bug notification because you are a member of
>Aptdaemon
>Developers, which is subscribed to aptdaemon in Ubuntu.
>https://bugs.launchpad.net/bugs/1058038
>
>Title:
> <class 'UnicodeEncodeError'>: 'ascii' codec can't encode characters in
> position 18-25: ordinal not in range(128)
>
>To manage notifications about this bug go to:
>https://bugs.launchpad.net/ubuntu/+source/aptdaemon/+bug/1058038/+subscriptions

--
Diese Nachricht wurde von meinem Android-Mobiltelefon mit K-9 Mail gesendet.

[Michael Vogt (mvo)]

@Sebastian:
we would have to use a shell wrapper or fork/exit as the env is read on interpreter statup.

[Launchpad Janitor (janitor)]

Status changed to 'Confirmed' because the bug affects multiple users.

[Michael Vogt (mvo)]

The attached branch works around the problem by setting up a basic LANG env and doing a a os.execv()

[Launchpad Janitor (janitor)]

This bug was fixed in the package aptdaemon - 0.45+bzr861-0ubuntu9

---------------
aptdaemon (0.45+bzr861-0ubuntu9) quantal; urgency=low

  * Workaround that ensures that for sys.getfilesystemencoding()
    "utf8" is used when the daemon is dbus auto-activatied (LP: #1058038)
 -- Michael Vogt <email address hidden> Tue, 09 Oct 2012 14:27:36 +0200