[Pkg-samba-maint] Samba 4.0: Upload to unstable.

[Ivo De Decker]

Hi Steve,

On Wed, Sep 25, 2013 at 12:38:47PM -0700, Steve Langasek wrote:
> On Sat, Sep 21, 2013 at 05:12:46PM +0200, Ivo De Decker wrote:
> > On Sat, Sep 21, 2013 at 02:37:36AM +0200, Jeroen Dekkers wrote:
> > > Lintian still gives a lot of errors and warnings. Before we can upload
> > > to unstable we should fix most of them (or silence them if what we do
> > > is correct). This is the lintian output I get:
> 
> > Some comments on a number of those:
> 
> > > E: libnss-winbind: ldconfig-symlink-missing-for-shlib lib/x86_64-linux-gnu/libnss_wins.so lib/x86_64-linux-gnu/libnss_wins.so.2 libnss_wins.so
> > > E: libnss-winbind: ldconfig-symlink-missing-for-shlib lib/x86_64-linux-gnu/libnss_winbind.so lib/x86_64-linux-gnu/libnss_winbind.so.2 libnss_winbind.so
> 
> > I committed a fix for those.
> 
> This is a wrong fix.  These are not shlibs, they're DSOs, and lintian is
> wrong to report this error for them.

It's obvious my fix is wrong, but lintian is correct that there is an issue:
when installing libnss-winbind (built based on git commit 5b9ec636a911), the
symlink gets created in postinst, and not removed when libnss-winbind is
removed.

I haven't investigated what's wrong, but the 3.6 branch has exactly the same
packaging, and doesn't suffer from this issue.

> > > E: samba-common-bin: python-script-but-no-python-dep usr/bin/samba-tool
> > > E: samba-common-bin: python-script-but-no-python-dep usr/sbin/samba_kcc
> 
> > > E: samba: python-script-but-no-python-dep usr/sbin/samba_dnsupdate
> > > E: samba: python-script-but-no-python-dep usr/sbin/samba_spnupdate
> > > E: samba: python-script-but-no-python-dep usr/sbin/samba_upgradedns
> 
> > Don't spend time on these. Some people think experimenting with :all
> > dependencies should happen in unstable instead of experimental, and don't
> > mind breaking stuff on the way.  This error will likely go away in the
> > next version of lintian (#711988).
> 
> Just because lintian wasn't caught up doesn't mean it's experimental.

It wasn't only lintian. Also wanna-build, britney, ...
But most of this seems to be fixed now (and that isn't really on topic here :)

> > I merged everything into samba_4.0. We should just continue development
> > there.
> 
> Should we delete the other branches now as obsolete?

Sure. I think most of the temporary 4.0 branches can go now.

> What else needs to be done yet before we upload?

Last week, I wrote:

"The main blockers for unstable (from my point of view) are:

- merge the changes that were done in unstable for samba 3.6 after the samba
  4.0 branch was split off (this shouldn't be hard)
- do some basic tests with the new package: it would be nice have at least
  some confirmation (apart from my own tests) that the packages are OK before
  they are uploaded to unstable."

Other than that, the TODO file also mentions the samba4 reverse (build-)deps.
We will break all of those, because they will need to build-depend on
samba-dev in the future. I was planning to file bugs about that once the
latest upload gets out of NEW.

I currently don't see any other blockers.

Cheers,

Ivo