You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@subversion.apache.org by ja...@jrv.org on 2003/09/29 15:39:28 UTC
commit fails on a dual processor FreeBSD 5.1 system.
commit fails on a dual processor FreeBSD 5.1 system.
$ svn --version
svn, version 0.29.0 (r6976)
compiled Sep 9 2003, 01:47:49
$ db4_verify -V
Sleepycat Software: Berkeley DB 4.0.14: (November 18, 2001)
Four instances of the test script (below) are started. So far one has
failed thusly:
...
Finished loop 218
svn: Item already exists in filesystem
svn: Commit failed (details follow):
svn: successor id `33.0.nx' (for `33.0.nt') already exists in filesystem '/home/james/svn-test/repos/db'
$
The other three are still running at around 700 iterations.
The poor man's test script:
$ cat commit.sh
FILES=${2:-100}
REPOS=file://$PWD/repos
LOOP=0
if [ .$1 = . ]
then
echo usage: $0 dir [size]
exit 1
fi
mkdir $1
I=1
while [ $I -le $FILES ]
do
echo $LOOP > $1/x$I
I=$(($I + 1))
done
svn import -m - $1 $REPOS/$1
rm -rf $1
svn checkout $REPOS/$1 $1
cd $1
while svn commit -m - -q
do
I=1
while [ $I -le $FILES ]
do
rm -f x$I
echo $LOOP > x$I
I=$(($I + 1))
done
echo Finished loop $LOOP
LOOP=$(($LOOP + 1))
done
$
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: commit fails on a dual processor FreeBSD 5.1 system.
Posted by kf...@collab.net.
Philip Martin <ph...@codematters.co.uk> writes:
> Most of the reports do involve dual CPU machines, although I don't
> know if Karl's report
>
> http://subversion.tigris.org/servlets/ReadMsg?list=dev&msgId=233543
>
> was a dual.
It was a single-processor machine.
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: commit fails on a dual processor FreeBSD 5.1 system.
Posted by Philip Martin <ph...@codematters.co.uk>.
"C. Michael Pilato" <cm...@collab.net> writes:
> james-tigris@jrv.org writes:
About a "successor id ... already exists in filesystem" error.
>> commit fails on a dual processor FreeBSD 5.1 system.
>
> I believe this is a known bug, and I believe that Sleepycat has
> addressed the problem in their forthcoming 4.2.x release.
Interesting. Yes, this error has been seen before, for example
http://subversion.tigris.org/servlets/ReadMsg?list=dev&msgId=245179
Most of the reports do involve dual CPU machines, although I don't
know if Karl's report
http://subversion.tigris.org/servlets/ReadMsg?list=dev&msgId=233543
was a dual. I haven't been able to reproduce the error recently,
possibly not since I switched to BDB 4.1.
--
Philip Martin
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: commit fails on a dual processor FreeBSD 5.1 system.
Posted by Justin Erenkrantz <ju...@erenkrantz.com>.
--On Monday, September 29, 2003 10:46:49 -0500 "C. Michael Pilato"
<cm...@collab.net> wrote:
> I believe this is a known bug, and I believe that Sleepycat has
> addressed the problem in their forthcoming 4.2.x release.
>
> Justin, can you confirm (and potentially add something to our
> Inconveniences website page)?
Hmm.
> svn: successor id `33.0.nx' (for `33.0.nt') already exists in filesystem
'/home/james/svn-test/repos/db'
Nope, sorry, this doesn't look like the problem I've seen with BDB 4.1.
(Note, this report is using BDB 4.0 *not 4.1.)
The problem I've seen with BDB 4.1 (and is fixed in 4.2.x) is that the
httpd processes go into an infinite loop in the shared memory allocation
code. There are lots of races in that code, and all of the locking code
has been rewritten for 4.2.x and is confirmed to work. If you get an
infinite loop and you are seeing this problem, it should be an infinite
loop around line 239 in db-4.1.25/env/db_salloc.c. FWIW, my reproduction
case is an import of linux kernel source.
Sleepycat has promised that 4.2.x should be out the first of October, so
we'll want to recommend that all 4.1.x users on MP boxes upgrade. -- justin
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: commit fails on a dual processor FreeBSD 5.1 system.
Posted by "C. Michael Pilato" <cm...@collab.net>.
james-tigris@jrv.org writes:
> commit fails on a dual processor FreeBSD 5.1 system.
I believe this is a known bug, and I believe that Sleepycat has
addressed the problem in their forthcoming 4.2.x release.
Justin, can you confirm (and potentially add something to our
Inconveniences website page)?
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org