Re: Crazy svn+ssh commit regression in trunk and 1.5

On Feb 8, 2008 7:46 AM, Augie Fackler <durin42_at_gmail.com> wrote:
>
>
> On Feb 7, 2008, at 10:31 PM, Augie Fackler wrote:
>
> >
> > On Feb 7, 2008, at 10:25 PM, David Glasser wrote:
> >
> >> On Feb 7, 2008 8:21 PM, Augie Fackler <durin42_at_gmail.com> wrote:
> >>> Howdy all,
> >>>
> >>> I noticed recently that I can't commit to my FreeBSD server (passwd
> >>> authentication for the users, can log in fine via ssh) with svn
> >>> 1.5 or
> >>> 1.6, but 1.4 would work just fine, even from the same working copies
> >>> (if the WC was downgraded, naturally). After a lot of downgrading
> >>> and
> >>> testing, I've determined the regression was introduced in r21109.
> >>> What
> >>> happens is this:
> >>>
> >>> r21109 client:
> >>> [imladris:~/test-repo/trunk] durin% /opt/experimental/bin/svn ci -m
> >>> 'Checking to see if commit is fixed.'
> >>> svn: Commit failed (details follow):
> >>> svn: SASL(-1): generic failure: Unable to find a callback: 16385
> >>> [imladris:~/test-repo/trunk] durin%
> >>> [02/07/08 10:04PM] 1
> >>>
> >>> trunk HEAD client:
> >>> [imladris:~/test-repo/trunk] durin% svn ci -m 'Checking to see if
> >>> commit is fixed.' svn: Commit failed (details follow):
> >>> svn: Cannot negotiate authentication mechanism
> >>> [imladris:~/test-repo/trunk] durin%
> >>> [02/07/08 10:09PM] 1
> >>>
> >>> But with OS X's stock 1.4 client (10.5.11, svn, version 1.4.4
> >>> (r25188)) the commit works.
> >>>
> >>> I haven't gone diving into the code to figure out what is going on
> >>> in
> >>> the code, that's on the agenda for later. Checkouts and updates work
> >>> no matter what version of the svn client is used.
> >>>
> >>> The server is running FreeBSD, and svnserve is 1.4.4 from ports. I
> >>> control the server, so I can install random versions of svn on it
> >>> for
> >>> testing if needed.
> >>>
> >>> Can anyone try to reproduce this (1.4 server, 1.5/1.6 client, svn
> >>> +ssh)
> >>> to make sure it's not a bug in my setup somewhere? I'm suspicious
> >>> it's
> >>> my setup, since this doesn't seem like the kind of thing that'd be
> >>> missed.
> >>
> >> Is SASL enabled in svnserve.conf?
> >
> > Nope - it's the default svnserve.conf in a freshly created
> > repository. That said, all repositories on that machine fail with
> > the same error.
>
> I just did an install of FreeBSD 6.3 release into a virtual machine so
> there are no configs of mine present. I installed subversion 1.4.4
> from ports and get the same errors with 1.5/trunk, and the same lack
> of errors with 1.4. The repo was freshly created with svnadmin create
> with no changes to the configs from the defaults.
> Also tested against a Dreamhost Linux server, and got the same thing.
> They are running svn 1.4.2, but other than the server version change
> and the OS change everything was the same.

So it seems like the client is trying to use SASL even though it's not
necessary for the server?

--dave

-- 
David Glasser | glasser@davidglasser.net | http://www.davidglasser.net/
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe_at_subversion.tigris.org
For additional commands, e-mail: dev-help_at_subversion.tigris.org