UnicodeDecodeError in Testament.as_short_text

[Carlos Perelló Marín]

Hi,

I was trying to execute bzr commit on one of my branches but got an UnicodeDecodeError that prevents me to do it.

I'm using:

carlos@aragorn:~/Work/Canonical/bug-1887$ dpkg -l bzr bzrtools
Desired=Unknown/Install/Remove/Purge/Hold
| Status=Not/Installed/Config-files/Unpacked/Failed-config/Half-installed
|/ Err?=(none)/Hold/Reinst-required/X=both-problems (Status,Err: uppercase=bad)
||/ Name Version Description
+++-=========================-=========================-==================================================================
ii bzr 0.8~200602240300 bazaar-ng, the next-generation distributed version control system
ii bzrtools 0.8~200602240659 Collection of tools for bzr

I got them from jbailey's .deb snapshots.

[Carlos Perelló Marín (carlos)]

Here you have the whole entry at .bzr.log when the UnicodeDecodeError appears.

[David Allouche (ddaa)]

I hit this bug when working on the bzr transition of importd. It was fixed by upgrading bzr to a recent bzr.dev (or so lifeless tells me).

The relevant thread on the mailing list is:
http://thread.gmane.org/gmane.comp.version-control.bazaar-ng.general/8839

[David Allouche (ddaa)]

Actually, it was merged in bzr.dev on 2006-02-18.

bzr.dev is now <http://bazaar-vcs.org/bzr/bzr.dev>.

[Carlos Perelló Marín (carlos)]

Then, Why Am I getting this error with a snapshot created on the 24th ?

[David Allouche (ddaa)]

Likely, it's a packaging problem (not packaging from the right branch). If you can reproduce the problem with the current bzr.dev, I'll reopen the bug.

[David Allouche (ddaa)]

Or maybe I'm just confused with what the date means... I'm not sure how the move to a shared branch was managed.

[Carlos Perelló Marín (carlos)]

The problem is still there.

jbailey fixed the snapshot creation and after getting the new package: 0.8~200602281718 I still have the problem.

[Carlos Perelló Marín (carlos)]

Confirmed, using the bzr tree got with: "rsync -av bazaar-vcs.org::bazaar-ng/bzr/bzr.dev ." gives me exactly the same error and I'm sure it's not using the old installation because the log shows all paths pointing to the rsync tree.

[Carlos Perelló Marín (carlos)]

Here you have the log I got using a fresh checkout of bzr.dev

[Carlos Perelló Marín (carlos)]

After some checks with David, the problem was with my ~/.bazaar/bazaar.conf file, I had there my name with non ascii chars using latin1 encoding. It was working on the past but I don't know why it start failing now... Removing the non ascii chars allows me to commit again.

Should it be using UTF-8 now or bazaar does not supports non ascii names at all?

[Wouter van Heyst (larstiq)]

On Thu, Mar 02, 2006 at 11:27:33AM -0000, Carlos Perelló Marín wrote:
> Public bug report changed:
> https://launchpad.net/malone/bugs/33029
>
> Comment:
> Confirmed, using the bzr tree got with: "rsync -av bazaar-vcs.org
> ::bazaar-ng/bzr/bzr.dev ." gives me exactly the same error and I'm sure
> it's not using the old installation because the log shows all paths
> pointing to the rsync tree.

If you copy over with rsync, be _sure_ to bzr revert afterwards. The
working tree may or may not be in sync (and I think it is not for
bzr.dev)

Wouter van Heyst

[Carlos Perelló Marín (carlos)]

The rsync download was done over an empty directory so there is no problems with the checkout.

Cheers.

[Wouter van Heyst (larstiq)]

On Thu, Mar 02, 2006 at 01:41:22PM -0000, Carlos Perelló Marín wrote:
> Public bug report changed:
> https://launchpad.net/malone/bugs/33029
>
> Comment:
> The rsync download was done over an empty directory so there is no
> problems with the checkout.

I wasn't clear enough. The problem is that the workingtree on the remote
side that gets rsynced over is out of date. To get a workingtree that
matches what is actually in .bzr, you should do a revert.

When you do a bzr pull/branch, bzr does that for you, but rsync doesn't
know how to do that.

Wouter van Heyst

[Martin Pool (mbp)]

On 2 Mar 2006, Carlos Perelló Marín <email address hidden> wrote:
> Public bug report changed:
> https://launchpad.net/malone/bugs/33029
>
> Comment:
> After some checks with David, the problem was with my
> ~/.bazaar/bazaar.conf file, I had there my name with non ascii chars
> using latin1 encoding. It was working on the past but I don't know why
> it start failing now... Removing the non ascii chars allows me to commit
> again.
>
> Should it be using UTF-8 now or bazaar does not supports non ascii names
> at all?

That configuration should be UTF-8; we should probably give a
warning/error if it's not.

--
Martin

[Carlos Perelló Marín (carlos)]

Using UTF-8 it still fails.

[John A Meinel (jameinel)]

I have manually set a utf8 username, and committed a unicode filename with a unicode message. And there were no problems.

There is an issue that you cannot have unicode characters in the domain portion of an email address.