Oops: user displayname has non-ascii value when setting the contact address field for a team

salgado, can you take a look at this one?

Is this really a high priority one?

Code to handle Unicode email addresses can be found at http://stuartbishop.net/Software/EmailAddress/index.html

However, the last time I went over the RFC's the example email address is invalid, as the account name portion should be US-ASCII. This is because the character set used to encode the account name is undefined. Some systems assume LATIN-1 so things appear to be ok, at least until they attempt to communicate with a system that assumes something else (UTF-8, ASCII, system locale). If this is no longer correct and the RFCs have been updated then it should be trivial to update the linked code to cope.

Bouncing to Mr. Mailman for an opinion.

Unsetting milestone as this won't be a quick fix given the number of tests that will be needed.

I get the OOPS by just entering a normal email-address. The error seems to crop up when trying to display the error message. It seems NOT to be about the email address!

Anybody ever noticed, that the character in question u'\xe1' is _not_ ó but á? Same for me. I enter a valid email-addres and it complains about ü, which is not part of the address.

I just tried it out with <email address hidden> which seems to work. I suspect that some module produces a localised error message that cannot be encoded.

This is strange. I cannot reproduce the error locally on launchpad.dev!

Entering fó@bar.com will correctly complain about "u'\xf3' in position 1". Entering "henning (AT) box25 (DOT) de" works just fine.

I hoped to be able to triage this but I have to give up here for now ...

On Wed, Nov 12, 2008 at 10:16:10AM -0000, Henning Eggers wrote:
> This is strange. I cannot reproduce the error locally on launchpad.dev!
>
> Entering fó@bar.com will correctly complain about "u'\xf3' in position
> 1". Entering "henning (AT) box25 (DOT) de" works just fine.
>
> I hoped to be able to triage this but I have to give up here for now ...

Looking at the OOPS (OOPS-775EB177) the user supplied an e-mail address
that was ASCII only. The OOPS happens when the code does a str(error),
and my guess would be that the error is something like:

    "The e-mail address is already used by Nicolas Valcárcel"

Thus str(error) fails, since the error message contains non-ascii
characters.

    subscribe bjornt

Looks like Bjorn has the right diagnosis. There are several places where we don't properly handle non-ASCII in user's display names. This one should be fixed, but I'm not sure it's a high priority. OTOH, reviewers should really be watching for str'ing error messages, especially when users are involved.

We may also want to give persons a __str__() or __repr__() that DTRT.