[svn.haxx.se] · SVN Dev · SVN Users · SVN Org · TSVN Dev · TSVN Users · Subclipse Dev · Subclipse Users · this month's index

Re: [PATCH] Disable fsync at transaction commit

From: Greg Stein <gstein_at_lyra.org>
Date: 2003-02-22 05:59:01 CET

On Sat, Feb 22, 2003 at 05:41:28AM +0100, Branko Cibej wrote:
> Greg Stein wrote:
>...
> >Seriously... just use a hash. Forget the svn_config_create stuff.
>
> The point of using an svn_config_t is that you can put
> FS-backend-specific options in fs-backend-specific sections, without
> having to know in advance which backend you're using. Yes, I know we
> have only the BDB backend at the moment. I'm sure gat appreciates every
> BDB-specific top-level interface we add. :-)

Quick. What was the name of the cmdline option to svnadmin?

  --bdb-txn-nosync

Quick, what would the hash key be?

  bdb-txn-nosync

You've got ample namespace protection for the backends. Each backend is
going to announce the specific things that it might look for. Hopefully,
there won't be many.

> >Hmm. It seems like there are two pieces here. One piece that passes config
> >to affect the creation, and one piece to pass params during normal
> >operation. But I didn't see any of the latter, so maybe the patch can be
> >trimmed back to just the creation params [for now] ?
> >
> >We can always come back at some future point to add runtime params.
>
> Well, svn_fs_new has to take the config parameter because that's where
> you create the svn_fs_t, not in svn_fs_create_berkeley.

Well, I might suggest that you pass the creation options to the
svn_fs_create_berkeley() function.

Not sure if I like that, but it doesn't mean you have to change all callers
of the svn_fs_new() function, especially given that we have no runtime
parameters that we know of.

That said, when we move to SQL, we *would* have runtime params for the
connection string (or user, pass, etc separate params). Would those be
passed to fs_new() or to fs_open()?

> I guess the
> changes to the repos interface could be limited to just svn_repos_create
> for now, with svn_repos_open always sending a NULL config to svn_fs_new.
> That should work, yes.

Ah. That would be nice. We could expand svn_repos_open() later, yes.

Cheers,
-g

-- 
Greg Stein, http://www.lyra.org/
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Sat Feb 22 05:54:44 2003

This is an archived mail posted to the Subversion Dev mailing list.