[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Removing Mssgs., Inconsistency

On May 5, 2008 at 18:19, Douglas Kline wrote:

> I wanted to expunge an accumulation of spam,  When other attempts left the
> links to messages in the index file and in the message files pointing to the
> wrong files, I decided to reconstitute the archive from scratch.  I started
> with an empty directory and ran mhonarc on all of the spooling-type files of
> messages.  Then I recompiled the list of spam messages because I couldn't use
> the previous list because the message numbers might be different.  Then I
> converted the spam message file names to message numbers and ran 

FYI, message number consistency is a known limitation wrt to
rebuilding archives.  I.e.  If you rebuild an archive, but the
set of messages have changed from the original archive set, then
message numbers will not match.

mharc works around this problem by utilizing namazu's message-id
index to allow own to have "permanent" location for a message.

> with the list of message numbers as arguments.
> That removed those files.  So far as I can tell, the links in the message fil
> es
> to other message files are now correct.  It also re-wrote the .mhonarc.db fil
> e.
> So that part worked.

IIRC, you may want to run some tests on the latest version of mhonarc.
I do not know when, but some of the logic for tagging things to update
on message removal were improved (check NEWS file).

> The index files (date1.html, date2.html, auth1.html, thrd1.html, etc.) still
> had references to the deleted spam messages.  So next I ran the command we ru
> n
> routinely to incorporate new messages with a dummy message to re-write the
> indices.  The dummy message was necessary because if it doesn't find any new
> messages it won't act.  That worked too.

Have you tried -editidx?  It rewrites ALL archive pages.

> Then I ran the scripts which compile the master indices (datedir.html,
> authdir.html, thrddir.html) and that worked.
> So what's the problem?  Some of the links in the message files to indices are
> wrong.  They refer to non-existent date[0-9]*.html, auth[0-9]*.html, and
> thrd[0-9]*.html files.  How can I fix that?

See comments above.  Later versions of mhonarc may fix this.

Also, -editidx should provide a brute-force way to correct the files.


[Index of Archives]     [Bugtraq]     [Yosemite News]     [Mhonarc Home]