On Wed, 03 Jan 2007, Garrett Rooney wrote:
> On 1/3/07, Daniel Rall <dlr@collab.net> wrote:
> >When I run 'make svnserveautocheck', I'm getting the following error
> >for pretty much all tests which use ra_svn:
> >
> > svn: SASL(-4): no mechanism available: No worthy mechs found
> >
> >This problem occurs when libsvn_ra_svn/sasl_auth.c's try_auth()
> >detects SASL_NOMECH, either from sasl_client_start() or its
> >handle_interact() function.
> >
> >Malcolm reported the same problem back in last October, and Vlad
> >responded <http://svn.haxx.se/dev/archive-2006-10/0317.shtml> saying
> >that this error can occur if Cyrus auth plug-ins are missing from your
> >/usr/lib/sasl2 directory. For Malcolm's case, Vlad suggested that a
> >missing libanonymous.so was the problem. I have such a plug-in. What
> >else could be causing this problem?
> >
> >
> >Here's the contents of my /usr/lib/sasl2 directory:
> >
> >libanonymous.la* liblogin.so.2@ libsasldb.la*
> >libanonymous.so@ liblogin.so.2.0.21* libsasldb.so@
> >libanonymous.so.2@ libplain.la* libsasldb.so.2@
> >libanonymous.so.2.0.21* libplain.so@ libsasldb.so.2.0.21*
> >liblogin.la* libplain.so.2@ smtpd.conf
> >liblogin.so@ libplain.so.2.0.21*
>
> Well, you're failing on an authz test, which will try to use something
> that has actual authorization, so anonymous won't cut it. You need
> something like libcrammd5.so I imagine.
All the tests fail over ra_svn when attempting the 'svn import' (not
just authz_tests.py), unless I 'configure --without-sasl'.
On IRC, Peter S. mentioned that he has a libcrammd5.so, from the
Debian package 'libsasl2-modules'. I found a similar package for FC5,
cyrus-sasl-md5-2.1.21-10.i386.rpm, which did the trick!
I'm of a similar mind to Malcolm here -- why don't we do something
more user-friendly here? Would falling back to simple_auth.c do the
trick?
- application/pgp-signature attachment: stored
Received on Thu Jan 4 00:16:15 2007