You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@subversion.apache.org by Ben Collins-Sussman <su...@collab.net> on 2004/12/16 15:08:26 UTC
locking branch progress report.
Oops, I meant to send this to all of dev@.
Begin forwarded message:
> From: Ben Collins-Sussman <su...@collab.net>
> Date: December 16, 2004 9:07:39 AM CST
> To: Jani Averbach <ja...@jaa.iki.fi>
> Subject: Re: svn locking r12327: FAIL (x86_64-unknown-linux-gnu shared
> ra_dav bdb)
>
>
> On Dec 16, 2004, at 7:38 AM, Jani Averbach wrote:
>
>> On 2004-12-16 01:40-0700, svntest@jaa.iki.fi wrote:
>>>
>>
>> SVN Locking, over ra_dav:
>>
>>
>
> ...doesn't work at all, as expected. I've not implemented RA->lock()
> in ra_dav yet!
>
> Here's the state of locking right now:
>
> * BDB: done by sussman.
> * FSFS: fitz is working on it.
> * svnserve: done by lundblad.
> * ra_svn: done by lundblad.
> * mod_dav_svn: done by sussman.
> * ra_dav: NOT done. sussman is working on it.
> * client changes: lundblad is working on it.
> 'svn lock/unlock': done.
> changes to 'svn up/commit': NOT done.
> changes to 'svn info/status': NOT done.
> svn:needs-lock property and ro/rw behaviors: NOT done.
>
> I don't think it makes much sense to run nightly regression tests on
> the locking branch yet... not unless you run ra_svn/bdb only.
>
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: locking branch progress report.
Posted by "Peter N. Lundblad" <pe...@famlundblad.se>.
On Thu, 16 Dec 2004, Ben Collins-Sussman wrote:
> Oops, I meant to send this to all of dev@.
>
> Begin forwarded message:
>
> > * svnserve: done by lundblad.
> > * ra_svn: done by lundblad.
> > * mod_dav_svn: done by sussman.
> > * ra_dav: NOT done. sussman is working on it.
> > * client changes: lundblad is working on it.
> > 'svn lock/unlock': done.
> > changes to 'svn up/commit': NOT done.
> > changes to 'svn info/status': NOT done.
svn info WCPATH done.
//Peter
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: locking branch progress report.
Posted by "Brian W. Fitzpatrick" <fi...@red-bean.com>.
On Dec 16, 2004, at 9:08 AM, Ben Collins-Sussman wrote:
>> Here's the state of locking right now:
>>
>> * BDB: done by sussman.
>> * FSFS: fitz is working on it.
I'm a bit behind, however, as I got pulled off to work on some other
stuff for about 3 days.
-Fitz, catching up
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: locking branch progress report.
Posted by Ben Collins-Sussman <su...@collab.net>.
On Dec 16, 2004, at 11:52 AM, Jani Averbach wrote:
> On 2004-12-16 10:09-0600, Ben Collins-Sussman wrote:
>>>
>>> The intresting thing is that the Apache was segfaulting
>>
>> I have no idea why apache would segfault, though. Can you look more
>> into that?
>
> Not very much, sorry.
>
Aha, I can reproduce!
During a commit, ra_dav sends a PROPPATCH on a 'working baseline'
object as a means of attaching the log message to the commit
transaction.
Apparently mod_dav tries to 'validate' the request by checking for
locks on the resource first. In the old days, this was a no-op,
because mod_dav_svn didn't provide a locking vtable to mod_dav. But
now mod_dav is asking mod_dav_svn to call svn_fs_get_locks() on a
resource which has a NULL fs path... boom, segfault.
TMI? :-)
Fix coming up. I guess the moral of the story is that I should run
'make davcheck' once in a while, not just the unit tests for fs locking
funcs.
Thanks for finding!
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: locking branch progress report.
Posted by Jani Averbach <ja...@jaa.iki.fi>.
On 2004-12-16 10:09-0600, Ben Collins-Sussman wrote:
> >
> >The intresting thing is that the Apache was segfaulting
>
> I have no idea why apache would segfault, though. Can you look more
> into that?
Not very much, sorry.
> I mean, we have no python tests that call 'svn lock', so I'd like to
> know what you're seeing, or how to reproduce.
HTTPD:
URL: http://svn.apache.org/repos/asf/httpd/httpd/branches/2.0.x
Revision: 122552
APR-0.9:
URL: http://svn.apache.org/repos/asf/apr/apr/branches/0.9.x
Revision: 122552
APR-UTIL-0.9:
URL: http://svn.apache.org/repos/asf/apr/apr/branches/0.9.x
Revision: 122552
SVN:
URL: http://svn.collab.net/repos/svn/branches/locking
Revision: 12331
The server OS is Linux/amd64.
In general I am running plain test framework against locking branch,
but it seems that you could trigger the same thing with just import or commit.
(with client from locking branch or from trunk, doesn't matter)
It seems that you could read (ls, co) but not write (commit/import)
anything to the repository, if that is any help.
mod_dav_svn:
<Location /test>
DAV svn
SVNPath /tmp/svntest/test
</Location>
[10:27:40] $ ./inst/svn_locking/bin/svn import -m ' ' \
<some dir> http://127.0.0.1:42024/test
svn: applying log message to
/test/!svn/wbl/f87034f9-5eeb-0310-b9dd-f55ca1238f8f/0: Could not read
status line: connection was closed by server. (http://127.0.0.1:42024)
$ tail -n 3 inst/httpd-2.0.x/logs/error_log
[Thu Dec 16 10:26:21 2004] [notice] Apache/2.0.53-dev (Unix) DAV/2 SVN/1.2.0-dev configured -- resuming normal operations
[Thu Dec 16 10:27:51 2004] [notice] child pid 29767 exit signal Segmentation fault (11)
[Thu Dec 16 10:27:52 2004] [notice] child pid 29807 exit signal Segmentation fault (11)
I have rebuild everything and this still happens, so it is something
in the locking branch or in my system. However, svn/trunk, and
svn/1.1.x are both working with the very same setup. Could you
reproduce this?
BR, Jani
--
Jani Averbach
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: locking branch progress report.
Posted by Ben Collins-Sussman <su...@collab.net>.
>
> The intresting thing is that the Apache was segfaulting
I have no idea why apache would segfault, though. Can you look more
into that?
I mean, we have no python tests that call 'svn lock', so I'd like to
know what you're seeing, or how to reproduce.
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: locking branch progress report.
Posted by Jani Averbach <ja...@jaa.iki.fi>.
On 2004-12-16 09:08-0600, Ben Collins-Sussman wrote:
> Oops, I meant to send this to all of dev@.
>
> Begin forwarded message:
>
> >From: Ben Collins-Sussman <su...@collab.net>
> >
> >On Dec 16, 2004, at 7:38 AM, Jani Averbach wrote:
> >
> >...doesn't work at all, as expected. I've not implemented RA->lock()
> >in ra_dav yet!
> >
> >Here's the state of locking right now:
> >
> > * BDB: done by sussman.
> > * FSFS: fitz is working on it.
> > * svnserve: done by lundblad.
> > * ra_svn: done by lundblad.
> > * mod_dav_svn: done by sussman.
The intresting thing is that the Apache was segfaulting, so even when
ra_dav is undone, we have a client side server exploit which could
bring down the server/mod_dav_svn.
This applies only of course in the case that mod_dav_svn is really
ready and done.
> > * ra_dav: NOT done. sussman is working on it.
> > * client changes: lundblad is working on it.
> > 'svn lock/unlock': done.
> > changes to 'svn up/commit': NOT done.
> > changes to 'svn info/status': NOT done.
> > svn:needs-lock property and ro/rw behaviors: NOT done.
> >
> >I don't think it makes much sense to run nightly regression tests on
> >the locking branch yet... not unless you run ra_svn/bdb only.
> >
Ok, I will probably disable ra_dav test for the present. Please let me
know when I should re-activate them.
BR, Jani
--
Jani Averbach
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org