On 2008-08-08, Florian Kaufmann <sensorflo@gmail.com> wrote:
> My final goal is that I can do the idea of subversion's 'cherry
> picking' (i.e.applying a single change of a branch to the working copy
> of another branch) alsowith the version control system I am forced to
> use at my workplace - Microsoft'sVSS. I have Cygwin installed, so I
> have all the gnu tools like diff, diff3 etc.
>
> As far as I understand, many svn diff/merge/update operations can also
> be doneusing these gnu tools. So I don't really need svn. E.g. to
> mimic "svn updatemyfile" I get the base and the head revision of
> myfile and then do "diff3 myfilebase head". The only difference is
> that a different diff tool is used.
>
> Now I am asking myself what I have to do to mimic "svn merge -r 10:11
> repos_path/branch/myfile". Anyone?
>
> Anyway, I don't even properly understand if internally, svn merge has
> 3 or 4 input files.
>
> trunk --- 5---...---10---11
> |
> branch \---...---15
> |
> working X
>
> Say I want to merge the changeset 11 into my working copy which's base
> is 15.Now does or doesn't svn merge also look at the common ancestor
> 5? Does it lookat the version 15 (the base)? I'm asking this because
> the manual says svn mergemight doesn't do what I want if I already did
> some local changes to my workingcopy. That suggests that svn merge
> also looks at the base revision (here 15),not only my working copy.My
> final goal is that I can do the idea of 'cherrypicking' also with the
> version control system I am forced to use at my workplace- Microsoft's
> VSS. I have Cygwin installed, so I have all the gnu tools likediff,
> diff3 etc.

I would recommend diff with the 'unified' output format and
the patch program. If the changes to the files are
separated enough (and don't interact functionally in adverse
ways), you could likely apply patches serially to achieve
something analogous to vector addition.

HTH

--
Robert Riches
spamtrap42@verizon.net
(Yes, that is one of my email addresses.)