Re: [Monotone-devel] Commit a child of 2 parents

[Justin Patrin]

On 1/16/06, rghetta <address@hidden> wrote:
> Nathaniel Smith wrote:
> > As I understand, tailor does work with monotone as both a source and a
> > target; but tailor can only handle linear histories.  Even if
> > transferring from one DAG VCS to another, it has to throw away
> > everything except a single linear subgraph.
> >
> > This could presumably be fixed.  I don't know if that would be more or
> > less work than starting from scratch.  (Presumably even if starting
> > from scratch, one could steal their VCS backend codes where useful.)
>
> Yes, tailor can use monotone both as source and target, but supports
> only a single ancestor.
> When used as source, the monotone backend forces a linear history by
> faking the graph edges; i.e. the target will end up with all the
> revisions stored on monotone, but with bogus ancestors.
> Changeset data as generated by monotone contains also the *real*
> ancestors, however, so a capable backend *could* use those additional
> informations to recreate the real history.
> None of the current backed does that, not even monotone, because it
> lacks merge-via-working-directory.
> With that feature, I think adding full history support to tailor would
> be reasonably easy to implement.
>

Again, I suggest asking around the OpenEmbedded community. I was under
the impression that the OpenEmbedded person playing with it had nearly
full history being transferred between multiple systems such as
mercurial, monotone, and svk.

--
Justin Patrin