-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
On Dec 8, 2006, at 11:10 AM, Paul Tomblin wrote:
In that case, shouldn't any message that reaches mailman with an unescaped From_ line in the body already be handled properly? It seems like something else must be borked. That or all of the
messages in a list mbox that contain unescaped From_ lines got there from really old versions of Mailman/python.
That is distinctly possible. The archives in question go back to
1998. I didn't keep track of when the *last* unescaped From_ line was put
in the archives.
Sorry, I should have been clearer that the /default/ behavior of the
generator is to mangle From_ lines. So it's true that nothing in
Mailman should need to be changed.
However, it's also true that in the distant past, there were some
bugs in the mbox implementation which would cause broken mbox files
to be written. A quick scan through the svn logs jogs my memory:
r6341 on 2003-04-17 was added to fix a message separation bug. I
don't know how long that bug was lurking, but the fix puts it just
before the 2.1.2 release according to the NEWS file. I'll bet that
it existed from 2.1 final (Dec 2002) until 2.1.2 (Apr 2003), the
latter which was probably released specifically to fix this problem!
Note that this bug had no effect on the archiving of new messages on
the fly. Those always got archived correctly. But the message was
appended to the mbox file incorrectly which meant that if you
regenerated your archives, you'd be screwed. This was what bin/ cleanarch was intended to fix.
BTW, one less ambitious way to participate here to help fix things
would be to improve bin/cleanarch. At the very least, you should be
able to run that script and get an mbox file that bin/arch can use to
DTRT. It would also be nice if bin/arch was able to compensate for
running out of memory, possibly by changing it to fork a sub-process
to do the actual archiving with the parent process pre-chunking the
workload for the child.
Anyway, I'm cc'ing mailman-developers. Further discussion of how to
improve matters should be conducted on that list (and mailman-users
should be removed).