On Wednesday 26 November 2003 13.42, Jostein Chr. Andersen wrote:
> On Tuesday 25 November 2003 16.05, firstname.lastname@example.org wrote:
> > establish that it works fine with Subversion, then maybe we won't
> > have to ship SVN 1.0 with a version of Berkeley DB so ancient it has
> > dinosaur tracks in the source files
> FWIW, I have a "make check" failure with r7852 and db 4.2.50 (the
> libtool I used is 1.4.3 ;-)
Well, this may make me look like a fool, but I can't hear you and I'm
sure that I'm not the first one to do this. Anyway, I think this may be
As suggested by Branko, IRC and the log file, the test suite was using
another version of svn under "make check", here is my system:
* My standard Subversion are configured with --prefix=/usr
* The Subversion to test with all related packages (httpd,
berkeley-4.2.50, neon etc) are prefixed with
* Before I compile and test a new subversion and/or related
packages I add this first on my /etc/ld.so.conf:
/usr/local/svnrel/lib and "export
PATH=/usr/local/svnrel/bin:$PATH". The I run ldconfig after
each make install.
* Running "echo $PATH" tells me that /usr/local/svnrel/bin
are first in the path.
The thing that happened was that when running "make check", the test
suite was looking for svn in the path (at this point, it's no svn or
svnadmin in /usr/local/svnrel/bin which is the way notes/releases.txt
says) and not in $PWD/subversion/clients/cmdline/svn.
So all I had to do was running "make install" _before_ "make check" and
then everything was just fine.
Someone may wonder (and have asked or commented) why I want to install
the untested stuff in another place (I prefare in a chrooted environment
or in a clean box). The reason is that I have a running Subversion
server on my workstation. It should not be a problem in a a well
constructed package to install in another place anyway.
Jostein Chr. Andersen <email@example.com>
To unsubscribe, e-mail: firstname.lastname@example.org
For additional commands, e-mail: email@example.com
Received on Wed Nov 26 21:12:55 2003