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

Re: Subversion's use of Berkeley DB [#11511]

From: Branko ─îibej <brane_at_xbc.nu>
Date: 2004-12-10 02:47:04 CET

Justin Erenkrantz wrote:

> --On Thursday, December 9, 2004 4:28 PM -0600 Ben Collins-Sussman
> <sussman@collab.net> wrote:
>
>> This ignores the fact that there are a number of libsvn_fs users that
>> are
>> *not* RA layers. Like svnlook, viewcvs, and so on.
>
>
> *sigh*
>
> I meant svn_repos and svn_fs not RA (sorry, mistyped). It'd be up to
> the server side to figure out when to call
> svn_repos_init()/svn_repos_check()/svn_repos_foo(). The point is that
> the shmem check (as you have to attach to the segment, create the
> segment if it doesn't exist, then map it in, check the exclusivity,
> and then close it) is going to be extremely expensive. This doesn't
> need to be done for mod_dav_svn if it's done in the right location:
> i.e. before httpd comes up. -- justin

Like I said, we can't use shmem segments because tbere's no automatic
cleanup if a process crashes. What Pthilip suggests is close to a
workable solution, because file locks _do_ get cleaned up when the
owning process exits.

-- Brane

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Fri Dec 10 03:13:52 2004

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