Bug 887685

Created attachment 664616 [details]
File: backtrace

Created attachment 664617 [details]
File: core_backtrace

Created attachment 664618 [details]
File: dso_list

Created attachment 664619 [details]
File: environ

I was looking through sealert entries when exception got thrown.

What lanquage were you using?

Every time on boot.

Package: setroubleshoot-server-3.1.19-3.fc18
OS Release: Fedora release 18 (Spherical Cow)

(In reply to comment #6)
> What lanquage were you using?

Sorry, missed your reply.

I'm using Ukrainian language.

$ echo $LANG
uk_UA.utf8

I'm using russian and suomi.

Created attachment 674340 [details]
I think this will fix the problem

Could some one apply this patch to catchall_boolean.py and see if it fixes the problem

# cd /usr/share/setroubleshoot/plugins
# patch < /tmp/unicode.patch
# python catchall_boolean.py

Then run sealert.

Dan: there are two issues here:
File "/usr/share/setroubleshoot/plugins/catchall_boolean.py", line 60, in get_if_text
    return _("you want to %s") % txt[0].lower() + txt[1:]

Issue 1:
I believe you're missing some parentheses: I believe you meant:

    return _("you want to %s") % (txt[0].lower() + txt[1:])

whereas precedence order means that python is interpreting this as:

    return (_("you want to %s") % txt[0].lower()) + txt[1:]

Issue 2:

FWIW you're relying on the side-effect of importing GTK on the default encoding:

>>> import sys
>>> sys.getdefaultencoding()
'ascii'
>>> import gtk
>>> sys.getdefaultencoding()
'utf-8'

>>> txt = '\xd0\x9a\xd0\xb5\xd1\x80\xd1\x83\xd0\xb2\xd0\xb0\xd1\x82\xd0\xb8 \xd0\xbc\xd0\xbe\xd0\xb6\xd0\xbb\xd0\xb8\xd0\xb2\xd1\x96\xd1\x81\xd1\x82\xd1\x8e \xd0\xb2\xd0\xb8\xd0\xba\xd0\xbe\xd1\x80\xd0\xb8\xd1\x81\xd1\x82\xd0\xb0\xd0\xbd\xd0\xbd\xd1\x8f mmap \xd1\x83 \xd0\xbd\xd0\xb8\xd0\xb6\xd0\xbd\xd1\x96\xd1\x85 \xd0\xbe\xd0\xb1\xd0\xbb\xd0\xb0\xd1\x81\xd1\x82\xd1\x8f\xd1\x85 \xd0\xbf\xd1\x80\xd0\xbe\xd1\x81\xd1\x82\xd0\xbe\xd1\x80\xd1\x83 \xd0\xb0\xd0\xb4\xd1\x80\xd0\xb5\xd1\x81 \xd1\x83 \xd1\x81\xd0\xbf\xd0\xbe\xd1\x81\xd1\x96\xd0\xb1, \xd0\xb2\xd0\xb8\xd0\xb7\xd0\xbd\xd0\xb0\xd1\x87\xd0\xb5\xd0\xbd\xd0\xb8\xd0\xb9 /proc/sys/kernel/mmap_min_addr.'
>>> print(txt)
Керувати можливістю використання mmap у нижніх областях простору адрес у спосіб, визначений /proc/sys/kernel/mmap_min_addr.

>>> print(u'foo: %s' % txt[0].lower())
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
UnicodeDecodeError: 'utf8' codec can't decode byte 0xd0 in position 0: unexpected end of data

Given that the format string is unicode, the above code implicitly attempts to convert the %s argument to unicode, using the system default encoding.

The issue is that txt is a string of bytes in UTF-8 encoding, and txt[0] is accessing the first *byte*, but 0xd0 is the first byte of a two-byte encoding of the unicode code-point 041A.  Hence the string that's a single byte 0xd0 is malformed UTF8, and python complains.

To do it correctly, you need to work at the level of unicode code points, not bytes:
>>> unitxt = unicode(txt, encoding='utf8')
>>> print(unitxt)
Керувати можливістю використання mmap у нижніх областях простору адрес у спосіб, визначений /proc/sys/kernel/mmap_min_addr.
>>> unitxt[0]
u'\u041a'

So you want something like:

unitxt = unicode(txt, encoding='utf8')
return _("you want to %s") % (unitxt[0].lower() + unitxt[1:])

so that it's working on unicode code points rather on individual bytes, and so that the precedence is definitely correct.

Fixed in setroubleshoot-plugins-3.0.47-1.fc18