On 26-01-2017 22:16:51 +0100, Clemens Lang wrote:
> Hi,
> 
> On Thu, Jan 26, 2017 at 11:03:19AM -0800, Will Yardley wrote:
> > On Thu, Jan 26, 2017 at 10:40:37AM -0800, Kevin J. McCarthy wrote:
> > > On Thu, Jan 26, 2017 at 06:42:10PM +0100, Fabian Groffen wrote:
> > > > Please find attached a patch against current Mutt tip to add LMDB
> > > > backend support. I must stress that this patch wasn't written by
> > > > me, I've made sure it applies to current Mutt and works correctly.
> 
> Thanks for finishing this up. I originally did some optimizations that
> can be seen in https://dev.mutt.org/trac/ticket/3691 and got this work
> into neomutt so it would see some initial testing, but haven't had the
> time and energy to get it submitted here over what was already in the
> ticket.

I took the neomutt commit and applied it to the patch:
https://github.com/neomutt/neomutt/commit/7e5380cd4c40d119ff83b2cf5f51f2cdb8a95ab3
It references you.


> Two minor issues:
>  - You may want to set LMDB_DB_SIZE to a larger value, see
>    https://github.com/neomutt/neomutt/issues/267

Ok, I'll add that.

>  - There's a missing free(&ctx) in hcache_lmdb_open() when
>    mdb_env_create() fails.

I'll add it, thanks.

> > > > The performance aspect of LMDB over BerkelyDB is significant, and
> > > > in the light of licence aspects LMDB gets preferred by some
> > > > distros these days. FWIW, the patch is part of Mutt on Gentoo.
> 
> Here are some measurements in case it helps:
> https://github.com/neomutt/neomutt/issues/244#issuecomment-274349073
> 
> > As discussed a while back, maybe we should also see if someone's
> > willing to propose a patch to add Kyoto Cabinet support and possibly
> > to remove / deprecate Tokyo Cabinet support?
> 
> I did this work a while ago. It's in neomutt at
>   https://github.com/neomutt/neomutt/blob/neomutt/hcache-kc.c
> now, but this won't apply as-is because neomutt now has separate files
> per header cache backend.
> 
> My original patch before the backends were split into separate files is
>   
> https://github.com/neomutt/neomutt/commit/944435025d5939aa7c568e54a9a1a9b2e8e09c1a
> 
> If somebody has some time to make that apply in mutt, that would be
> great.

I'll leave this up to Kevin and others.  I guess this is also very
related to whether the backends would be split out like NeoMutt did or
not.  It makes it cleaner for sure.

Thanks,
Fabian

-- 
Fabian Groffen
Gentoo on a different level

Attachment: signature.asc
Description: Digital signature

Reply via email to