rid backend seems to be initialized twice?

simo idra at samba.org
Sat Oct 13 16:18:44 GMT 2007


On Sat, 2007-10-13 at 11:06 -0500, Gerald (Jerry) Carter wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
> 
> Simo,
> 
> >> This is one major reason why I would like to see us move away from
> >> all the file reorganization that people like to do so often.
> > 
> > This seem a deficiency in git.  I agree unnecessary file 
> > re-organization is not healthy, but the SCM tool should not
> > constrain us.
> 
> This is entirely a social thing.  Git puts no restraints on you
> in this regards.  It has more to do with introducing unnecessary
> change.  And when working with large groups of people,  we should
> be more cognoscente of how unnecessary changes affects other
> peoples work.

Yeah in the same way as tar does not :-)
I am not really blaming git here, most other tools are basically
incapable of doing decent merge even with the same files :-)

> But if you are patching file A in one branch and file B in another,
> it really is a new patch and should not be considered the same.
> Every change that requires human interaction requires testing.
> Only if you are making the exact same change to a file in two
> identical branches is the change really the same and should have
> the same hash because in that case the entire tree is the same.
> 
> Even if you make the same change to file A which is identical
> in two branches but the branches themselves differ, it may still
> have additional side affects due to tightly coupled code and
> is *not* the same change.  Therefore I believe that git's response
> of assigning a new hash is the right thing to do, both mathematically
> and socially.

Maybe, I need to think about this, I don't buy it completely right now.

Not because what you say does not make sense, but because the same thing
maybe true even if the file is in the same position with exactly the
same patch applying perfectly. Once you have multiple trees you pull
from there is always the chance of side interaction between your code
and the code that differs, so if you bring this conclusion to the
extreme it means you should never merge. But we know that merge is
necessary, powerful, and also is ok most of the time.

But as I said, this is a minor issue, let's move on.

Simo.

-- 
Simo Sorce
Samba Team GPL Compliance Officer <simo at samba.org>
Senior Software Engineer at Red Hat Inc. <ssorce at redhat.com>



More information about the samba-technical mailing list