How to design for the following requirement

[Adrian Buehlmann]

On 27.08.2009 20:16, Jatin wrote:
> I am trying to achieve following mercurial design on my machine.
> 
> I have a main repository on which I am working. I have cloned the main
> repository at some point realeasing a workable version and after that I have
> added some more new features in the project(main repository) . At the same
> time, I am fixing bugs on the released version (clone). I fix the bugs and
> provide patches to the user. I don't wan't to include the new features from
> the main repository on the "workable version"(clone). In short I am trying
> to achieve this.
> 
> Main Repository - Add new Features to the project.
> Clone Repository - Clone a "workable version" from Main repository. Change
> this only for bug fixes i.e. no new features from Main Repository must
> appear here.

That's fine. It's done on mercurial development itself like that as well.

Your "clone repo" is usually called the the "stable" repo -- it contains
only bugfixes (no new features) and is a strict subset of the main repo.

Mercurial repos (as an example):

stable: http://selenic.com/repo/hg-stable
main:   http://selenic.com/repo/hg

> In this set up, when some bugs are fixed in the workable version, I will be
> pushing the same to the main repository. This seems to work for only the
> first push of workable version(clone) to main repository. Subsequent pushes
> result in error "push will not be performed as it will create remote
> branches. You must first resync by doing a pull from target repository and
> Merge".

Pull stable into main. This will add a new head to main, which is fine.

Then merge and commit. This will merge the two heads into one again, creating
a merge changeset (merge changesets have two parent revisions).

As an example in the mercurial repo itself, look for changesets with change
messages like "merge with...".

See also http://mercurial.selenic.com/wiki/UnderstandingMercurial

> Now I dont wan't to resync the clone with the main repository, as I don't 
> want the new features added in main to appear in clone. I am relatively new
> to mercurial, so this seems to be some design issue. What I am doing wrong
> here. Why can't I keep pushing from clone to main repository as and when
> needed.

See above.

> Note that there is only one pc holding main and clone.

That's fine. A repo can always be cloned to everywhere you want it.
You can pull from local repos as well as from remote repos (if there is
a server running on the remote side).

See hg help pull.

Just make sure you have backups in case your PC dies. Use hg verify
to check the backups.