[Pkg-samba-maint] Help needed with Samba 4.0.5 package

[Ivo De Decker]

Hi,

On Sat, May 11, 2013 at 09:34:34AM +1200, Andrew Bartlett wrote:
> > I imported the tarballs samba_4.0.0+dfsg1 and samba_4.0.3+dfsg1 tarballs from
> > previous samba4 uploads into the repo and created a new samba_4.0.5+dfsg1
> > tarball. I don't have a good git-import-orig workflow for this yet. Maybe
> > someone can point me at a good workflow to manage repackaging upstream
> > tarballs. It would be nice if the samba upstream release process would strip
> > out the non-free files before creating the tarball. That would save us the
> > work of repackaging them. Do you think this is something that could be
> > considered? 
> 
> If it were that easy, we would.  The 'non-free' files are of course
> RFCs, which contain tables which are extracted by our build process.  As
> the Debian package uses Heimdal from the system, the problem is punted
> to the Heimdal maintainer.  That is the only reason the build works with
> these files removed. 

The samba heimdal package has a script to extract the tables from the RFC and
only ships the extracted table. Could something like that be considered for
samba?

> Thanks, I've now build with that configuration.  The only thing I
> couldn't do is inject CC="ccache gcc" into it, as far as I can tell.  I
> use that to speed up builds a little.

my .git/gbp.conf has this:

[DEFAULT]
# the default build command:
#builder = debuild -i -I
builder = debuild -i -I -us -uc -j12

You could replace the builder with a wrapper which calls ccache.


> From here, I think we need to address the three TODO commits, sort out
> the debconf mess and the upgrade/prerm/postint scripts.  I would very
> much appreciate assistance with that.

I fixed the debconf stuff for now, by moving the templates to the right names
for the current packages. This version builds and installs fine. Just running
the 'samba' package (with smbd and nmbd) is fine. The samba-ad-dc package
installs fine now too, but I didn't get it to work (probably because I never
tried running the samba4 ad dc before).

We should think about the naming of the packages and the init scripts. The
'samba' package doesn't ship the 'samba' daemon, and the 'samba' init script
doesn't control this daemon. Looking at the history, this is expected, but it
gets confusing very quickly.

Cheers,

Ivo