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

Re: stress.pl error with ra_dav+fsfs

From: Jani Averbach <jaa_at_jaa.iki.fi>
Date: 2005-07-05 18:54:05 CEST

On 2005-07-05 17:06+0100, Philip Martin wrote:
> Jani Averbach <jaa@jaa.iki.fi> writes:
>
> > So it looks like that here only SDBM works as activity backend.
>
> As far as I know SDBM is very simple and doesn't do any locking, so
> the apr-util wrappers have to do it. BDB, and I think GDBM, provide
> their own locking.
>
> > With
> > anything else, first activity error pops up during the 10-20 first
> > commit. However, this very same system (chroot) compiles and runs
> > nightly regression tests, so it can't be completely nuts.
>
> The Subversion regression tests don't do any concurrent database
> access, so they won't trigger this problem.
>
> It looks like the locking provided by BDB/GDBM doesn't work on your
> system--so it's a kernel/filesystem/libc/compiler problem, but I don't
> know which :)

Sorry to be unclear, but If I run the stress.pl against BDB(4.3)
backend repository (I double checked that bdb is what I am really
using), it works just fine with __ra_svn__. If I use ra_dav, it fails
immediately with activity related errors. So the general locking
should work, because 10 instances of stress.pl against BDB works just
fine.

So I would say that something is broken on this trail:

stress.pl -> mod_dav_svn -> activity -> apr_dbm_* -> (BDB, GDBM) -> BANG!

BR, Jani

-- 
Jani Averbach
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Tue Jul 5 19:24:07 2005

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