[PATCH/RFC] Use submodule for dnspython

[Andrew Bartlett]

On Mon, 2015-05-11 at 06:36 +0100, Andrew Bartlett wrote:
> On Sun, 2015-05-10 at 20:31 +0200, Stefan (metze) Metzmacher wrote:
> > Hi Jelmer,
> > 
> > > What changes for users
> > > ======================
> > > 
> > > A consequence of this is that if "dnspython" is not installed on the
> > > system and if samba wasn't cloned using 'git clone --recursive',
> > > the user will get the following message during configure:
> > 
> > The problem is still that this only works for the current checkout
> > and not for all branches/tags (all commits) in a repository.
> > Unless this is fixed I'm against using submodules.
> 
> I've been chatting with Jelmer about this, and the way I see it, if 'git
> checkout' doesn't also make the submodules update to match, then this is
> a non-starter.  They need to be a bit more transparent than that, while
> gaining us the advantage that we can point to a different revision by
> SHA1, not having to import the actual tree and then (in a correctly
> paranoid fashion) verify that a delta hasn't been introduced. 
> 
> It is this later feature that I'm after from submodules - being able,
> for our third party code, to say 'it is JUST what you find at official
> repo $HERE, see the SHA1 matches!', rather than re-running the update
> script and hoping for no non-while space differences. 

If we could work around the non-updating issue (say by having waf
confirm that the submodule checkout is correct), there would be one very
big advantage, which is I think where this started from:  We would never
again have silly discussions about patching/forking upstream code, third
party in Samba.  That might just be enough to make it all worthwhile...

Andrew Bartlett

-- 
Andrew Bartlett                       http://samba.org/~abartlet/
Authentication Developer, Samba Team  http://samba.org
Samba Developer, Catalyst IT          http://catalyst.net.nz/services/samba