branch command crashes with UnicodeEncodeError exception

[davidedb]

bzr 1.9 crashes when executing the command:

bzr branch http://gigo-ice.org/repos/extdiff

with the error message:

bzr: ERROR: exceptions.UnicodeEncodeError: 'ascii' codec can't encode characters in position 75-77: ordinal not in range(128)

See the attached log for more details.

For your convenience, here is the output of command 'bzr --version':

Bazaar (bzr) 1.9
  Python interpreter: /design/scripts/activestate/python/2.5.1.1/Linux.i686/bin/python 2.5.1
  Python standard library: /design/scripts/activestate/python/2.5.1.1/Linux.i686/lib/python2.5
  bzrlib: /design/scripts/canonical/bazaar/1.9/Linux.i686/lib/python2.5/site-packages/bzrlib
  Bazaar configuration: /home/MCDD/users/cad/davide/.bazaar
  Bazaar log file: /home/MCDD/users/cad/davide/.bzr.log

[John A Meinel (jameinel)]

My best guess is that you have a non-ascii filename in your branch, but your local configuration is indicating that your filesystem only supports ASCII filenames.
Which is also supported by the line:
0.169 encoding stdout as sys.stdout encoding 'ANSI_X3.4-1968'

You probably could do "export LANG=en_US.UTF-8" or some equivalent. I'm not sure why your system feels that it is ascii-only. It is far more common to have systems default to UTF-8 or even latin-1/iso-8859-1.

[davidedb (davidedb-gmail)]

Your guess is right: in my environment LANG is set to "C" for compatibility with old programs giving wrong results with other locales.
After I changed LANG's value to "en_US.UTF-8", I was able to create the branch successfully.

However it would be nice if bzr could catch such an exception and quit gracefully, instead of emitting a stack trace.

[Jelmer Vernooij (jelmer)]

I can still reproduce this. Indeed, we can't really do much at the moment but report a sensible error and exit.

[Martin Packman (gz)]

From bzr 2.5 file handling defaults to using utf-8 even in the C locale.