[Mailman-Developers] Mix of 8-bit ascii and unicode in Archivemarshals

[Mark Sapiro]

Mads Kiilerich wrote:
>
>Now I see crashes in Archiver like shown below. My problem is similar to 
>but different from 
>http://archives.free.net.ph/message/20060116.022354.167a1fae.en.html
>
>It turns out to be because both 8-bit ascii and unicode strings appears 
>as first element in key tuples in the 2006-February-author marshal. When 
>sort tries to compare these unrelated encodings python don't know what 
>to do. Neither do I...
>
>How should the problem be solved? Should keys be stored as 8-bit ascii 
>or as unicode?


This appears to be the same problem as in the thread at
<http://mail.python.org/pipermail/mailman-users/2006-February/049333.html>
which was apparently worked around by rebuilding the archive with
'bin/arch --wipe'.

I haven't looked at this in any detail, but it looks like maybe we
changed the encoding of author and/or subject so if you upgrade in the
middle of a month, you get this problem in the database.

-- 
Mark Sapiro <msapiro at value.net>       The highway is for gamblers,
San Francisco Bay Area, California    better use your sense - B. Dylan