[Mailman-Users] problem with mm.arch

[David Relson]

On Sun, 17 Apr 2005 14:08:28 -0700
Mark Sapiro wrote:

> David Relson wrote:
> >
> >Unfortunately, I don't know the full history of mylist.mbox.  I don't
> >recall mucking with the mbox file and May 2004 was quite a while ago
> >and there's no telling.  I don't _think_ I did anything with the mbox
> >file -- especially since I use the .mbx extension and don't recall
> >seeing a .mbox file before the recent problems.
> >
> >I _do_ notice that the messages added via bin/arch haven't been added
> >to mylist.mbox (which has a Feb date as I write this), which doesn't
> >sound like the expected behavior.
> 
> Unless you have set ARCHIVE_TO_MBOX = 0 in mm_cfg.py (presumably it's
> not -1 or 1 since if it were you'd have no pipermail archives at all),
> the archiving process will append each new archived message to the
> archives/private/mylist.mbox/mylist.mbox file. bin/arch doesn't do
> this because in normal operation it is reading this file to rebuild
> the archive.
> 
> If I understand correctly what has been done in your case, you
> overwrote this file with the equivalent of
> 
> cd archives/private
> cat mylist/*.txt > mylist.mbox/mylist.mbox
> 
> and then ran
> 
> path/to/bin/arch --wipe mylist mylist.mbox/mylist.mbox
> 
> which is equivalent to just
> 
> path/to/bin/arch --wipe mylist
> 
> which resulted in a problem because the Date: headers in the *.txt
> files for messages that were originally archived with a previous
> bin/arch were all the date of the bin/arch run instead of their
> original dates.

Correct.

> It also caused a second problem which you may not be aware of or care
> about, but the 'cat' caused a reording of the messages in the
> mylist.mbox/mylist.mbox file which caused the archive messages to be
> numbered differently in the rebuilt archive which invalidates any
> saved URLs to archived messages.

Yes and that's been corrected (for the most part)  After realizing the
problem, I was able to load a fairly recent backup (presumably with the
correct message numbers).  Unfortunately, there's nothing to do about
very recent messages, but I can live with that.

> You then fixed the first problem by processing the
> mylist.mbox/mylist.mbox file with a perl script to put the date from
> the "From " separator into the Date: header and reran bin/arch.
> 
> This is probably the best you can do at this point, but now, if my
> above understanding is correct, you have a
> archives/private/mylist.mbox/mylist.mbox file which is complete and
> has correct Date: headers. Again, assuming ARCHIVE_TO_MBOX has its
> default value of 2, new messages will be appended to this file, so if
> in the future you ever need to rebuild the archive, you just use this
> file directly and not the individual *.txt files.

Understood!