Traceback rebuilding archive

[Stdivjak]

Witn Mailman 2.1b3, rebuilding a pipermail archive using
$ bin/arch listname
runs fine for 2 lists, but on the third one gives the
following traceback:

[...]
Updating HTML for article 52
Updating HTML for article 53
Schreibe Archivzustand in Datei
/home/mailman/archives/private/humor/pipermail.pck
Traceback (most recent call last):
  File "bin/arch", line 160, in ?
    main()
  File "bin/arch", line 148, in main
    archiver.processUnixMailbox(fp, Article, start, end)
  File "/home/mailman/Mailman/Archiver/pipermail.py",
line 545, in processUnixMailbox
    m = mbox.next()
  File "/usr/local/lib/python2.2/mailbox.py", line 34, in
next
    return self.factory(_Subfile(self.fp, start, stop))
  File "/home/mailman/Mailman/Mailbox.py", line 69, in
scrubber
    return mailbox.scrub(msg)
  File "/home/mailman/Mailman/Mailbox.py", line 89, in
scrub
    return self._scrubber(self._mlist, msg)
  File "/home/mailman/Mailman/Handlers/Scrubber.py",
line 159, in process
    part.set_payload(_("""\
  File "/home/mailman/Mailman/i18n.py", line 76, in _
    return _translation.gettext(s) % dict
UnicodeError: ASCII decoding error: ordinal not in
range(128)

[http://sourceforge.net/tracker/index.php?func=detail&aid=593728&group_id=103&atid=100103]

[Barry Warsaw (barry)]

Can you upload the .mbox file for the list you're
regenerating? If not (perhaps because it's proprietary
information), can you make it available for me to look at
privately?

[Barry Warsaw (barry)]

I'm moving the bug report to pending since I haven't seen an
update. I suspect this is fixed in current cvs.

[Stdivjak (stdivjak)]

Sorry for the long delay. I fiddled around with the 19MB
mbox file, and cut out the part which causes this error
(refer to the attached file).

The traceback also appeared using 2.1b5.

[Stdivjak (stdivjak)]

Just tried to rebuild the other archives: 4 of 21 mbox-files
can't be rebuilt. There are two kinds of tracebacks:

* "ASCII decoding error" (see above)
* "Lookup error: unknown encoding"

I'll send the archive causing the second kind of traceback
to Barry, hence it contains somewhat internal information.

[Barry Warsaw (barry)]

Thanks Stefan,

I got your mbox file, which appeared to contain one message.
 That message has an unknown charset in its RFC 2047 Subject
header. Mailman cvs seems to handle this file just fine, so
I'm closing this as fixed in 2.1b5+

[Stdivjak (stdivjak)]

Barry told me, that it is fixed in the current CVS version in
September 2002, but this bug still appears in mailman
2.1.1 (using bin/arch on the attached mbox-file), so I
switched the status to open again.

[Stdivjak (stdivjak)]

Workaround:

- open the mbox-File in any Editor
- search for "*="
- replace iso-encoded MIME-fields 'name*' and 'filename*'
by plain text, removing the asterisk:
     name*="iso8859-15''Autobahngeb%FChrjpg%2Ejpg"
-> name="Autobahngebuehr.jpg"