- Previous message: Bron Gondwana: "Re: cyrus.expunge/cyrus.cache mistmatch on replica"
- Maybe in reply to: Bron Gondwana: "Large Mailbox Append Fix"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ] [ attachment ]
|
|
|
From: Bron Gondwana (no email)
Date: Tue Sep 25 2007 - 09:35:28 EDT
On Tue, 25 Sep 2007 07:11:36 -0400, "Ken Murchison" <> said:
> Bron Gondwana wrote:
> > I'm afraid I still don't see how your code addresses the case where there
> > turn out to be no more messages with higher UIDs which need uploading even
> > though we've hit exactly max_count messages. It doesn't pre-scan to the
> > end and check if there's going to be another one so it can predict if it
> > needs to set last_uid to the mailbox value.
>
> Ahh, my rusty old brain finally figured out what you're talking about.
> I added a check for that case (see below). In my infinite wisdom, I
> also didn't have a way for the while() loop to exit cleanly.
Yeah, I didn't see that clearly in all the diff and I was concentrating
on the other issue.
> > Can you please email the sync_client.c with the patch applied so I can see
> > the function in its final form and run some thought experiments against it.
> > At the moment I feel I might be attacking a strawman, but given how bad the
> > performance hit of un-necessary cache rewrites on big folders was, I want
> > to make sure we get this logic right!
>
> Here's what I currently have:
>
> [snip code]
As they would say in the Linux kernel world:
Signed-off-by: Bron Gondwana <>
:)
Looks good to me.
Bron.
-- Bron Gondwana
|
|
|