<<< Date Index >>>     <<< Thread Index >>>

Re: header_cache broken in cvs



Brendan Cully wrote:  [Sun Mar 12 2006, 04:47:38PM EST]
> Have you removed and repopulated the cvs version of the cache? I think
> if there's a version mismatch between mutt and the cache, mutt will
> just silently ignore the cache.

Yes, as mentioned in my last email, the first timing was empty cache,
second timing was with cache populated (by the first run).  I clean it
out each time.

My testing methodology was a script that does the following.  I used
it to bisect the dates from Sept. 15, 2005 (1.5.11 release date) to
the present to find where the behavior changed.

    0. delete existing working copy
    1. cvs checkout fresh from local rsync'd cvs repo
    2. cvs update to date in question using -D
    3. build
    4. remove existing cache files, which exist in the directory
       ~/.mutt-cvs/header_cache.  Additionally remove any files that
       might be in ~/.mutt-cvs:
            rm -rf ~/.mutt-cvs/*/* ||:
            rm -f ~/.mutt-cvs/* 2>/dev/null ||:
    5. Get timing for opening an IMAP inbox twice:
        time ./mutt -f imaps://mail.server/INBOX -e 'exec quit'
        time ./mutt -f imaps://mail.server/INBOX -e 'exec quit'

It's also observable that the cache is having no effect within
a running instance of mutt.  For example, changing to an alternate
folder, then back to INBOX, shows that it has to refetch all the
headers from the server.

> Another thing to try is to use a directory for the cache instead of a
> file. What's your architecture? Maybe a 32-bitism crept in when I
> added a couple of extra fields to the cache...

I'm running 32-bit kernel and userland on amd64.

Now, the question is: Does this actually work properly for you?

Regards,
Aron

Attachment: pgpGsx9JYFULs.pgp
Description: PGP signature