Re: ambiguous clean (was Re: Re: [cdv-devel] more merging stuff (bit long...))

On Thu, 11 Aug 2005, Bram Cohen wrote:

>   a
>  / \
> |   b
> |   |
> b   a
>
> b wins (implicit undo)
>
>   a
>  / \
> |   b
> |   |
> b   c
>
> c wins (convergence)

I'm not convinced here. I agree with the c wins (convergence) but the
other case looks bizarre. Consider:

       a
      /|\
     b b b    <- patch to mailing list (three developers pick it up)
     | | |
     a | a'   <- dev1 & 3 realize it is a bad patch and back it out
     | | |       dev1 backs out with patch -R dev 3 backs out manually[1]
     |/|\|
     b | a'   <- dev1 & 3 now merge with 2 who's been on holiday


And what happens when we merge the heads?


And would we expect a different result if everybody had merged after
applying the patch - i.e. the graph looks:

        a
       /|\
      b b b
       \|/
        b     <- everybody merges here to get a single head
       /|\
      a | a'
      |/|\|
      a | a'


[1] I think this is most likely to occur if dev3 first attempted to
"fix" the patch before backing it out.

Tim.

--
God said, "div D = rho, div B = 0, curl E = - @B/@t, curl H = J + @D/@t,"
and there was light.

     http://tjw.hn.org/      http://www.locofungus.btinternet.co.uk/