[Svnmerge] revision lists

Konstantin Sobolev konstantin.sobolev at gmail.com
Thu Nov 23 13:48:24 PST 2006


Hi Blair,

On 11/24/06, Blair Zajac <blair at orcaware.com> wrote:
> > We have a branch split from trunk at rev. 55921. At the moment when
> > I've tried to use svnmerge there already were several commits both in
> > the branch and in the trunk.
>
> When you say commits on the branch, do you mean merges from trunk?

No. There was one file added on the branch, then it was merged into
trunk. We were playing with svn branching support. Now there are also
a couple of commits done by svnmerge init attempts.

> > I've switched to the branch:
> > svn switch svn://xxx/branches/release_2006_11/java
> >
> > Then tried to init:
> > kos at kos ~/work/sw2/java $ svnmerge.py init
> > svnmerge: no copyfrom info available. Explicit head argument
> > (-S/--head) required.
>
> You don't normally put a merge tracking at a level below the branch.
> I would recommend using init at the branches/release_2006_11 level
> instead.

I see now. The problem is that we have chosen a slightly non-standard
repository layout. In a couple of words we have 2 almost decoupled
projects, but we want them to share branches and tags. Thus we have
trunk/java
trunk/installer
branches/release_2006_11/java
branches/release_2006_11/installer
etc.
Almost all developers work with 'java' and don't event want to have
'installer' sources in their working copy (they are quite large and
some devs are on a not-so-fast connection). I also have only 'java'
subdirectory checked out.

As I understand from this message:
http://www.mail-archive.com/svnmerge@orcaware.com/msg00291.html
operating on subdirectories is hardly supported and people will have
to checkout full tree to merge something.

> > This was confusing because 'svnmerge.py help init' doesn't mention
> > -S/--head switch. However the following command was successfull:
> >
> > kos at kos ~/work/sw2/java $ svnmerge.py init svn://xxx/trunk/java  -r
> > 0-55921
> > property 'svnmerge-integrated' set on '.'
> >
> > Which I assume means to ignore 0-55921 revisions.
> > The rest you have seen in my original post.
> > I feel that I did something wrong, but I don't understand what was
> > it..
>
> If you're initing on a branch that was copied from the source of the
> branch, trunk in this case, you don't need to supply the -r command
> line option to init, even if there are commits.  If there are merges,
> then you can still init and then merge with the --record-only option.
>
> Maybe it's the 0, and try using a 1 instead?  If you look any
> existing merge tracking information, it starts at 1 and not 0.

I remember that -r 0-55921 was my second attempt. The first one was
svnmerge.py init svn://xxx/trunk/java -r 55921
(initially I thought that I must explicitly specify at which revision
the branch was created), but in this case the full log has also been
requested.

Thanks

-- 
/KoS



More information about the Svnmerge mailing list