Please check my thinking on bug 646979

[John Arbash Meinel]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On 10/5/2010 9:27 AM, Barry Warsaw wrote:
> On Oct 05, 2010, at 09:16 AM, John Arbash Meinel wrote:
> 
>> I've been thinking about it, and I'm pretty confident that what you are
>> trying to do is inherently "criss-cross". Specifically consider a
>> semi-ideal case:
> 
> This is all fascinating, and while I have nothing constructive to add, I
> wonder: does it make life any easier if you were only considering the debian/
> directory?  I'm not sure that's even possible or desirable, but it seems like
> that would at least take upstream out of the picture.
> 
> -Barry
> 

That was part of my "if you have a distinct set of changes on both sides
then per-file merging should handle it cleanly".

To try to describe, go back to my graph:


 A
 |\
 | d
 | |\
 B | u
 |\: |
 | \ |
 | :\|
 C | v
  \|
   e


imagine if debian and ubuntu only change debian/* and upstream is the
only time you change the content. In that case, the per-file graph for
tree content looks like:

 A
 |
 B
 |
 C

and the per file graphs for debian content looks like:

 d
 |\
 e u
   |
   v

Notice that neither of them is particularly complicated, or has any
crossing lines. :)

Now, I would imagine that the *interesting* merges are not clean like
this. Why would you really care about merging if debian isn't adding
patches to the upstream code? (Other than procedurally being the right
thing to do, it doesn't seem *interesting*.)

John
=:->

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (Cygwin)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/

iEYEARECAAYFAkyrOEEACgkQJdeBCYSNAAM3mwCfTaR959x9z0HMEWZL+wHLkbet
i38AoJ3/FlNk18q6/mJW5pw1d6X3xKxX
=7i13
-----END PGP SIGNATURE-----