Local vs. non local commands

[Matus]

Martin Geisler <mg <at> lazybytes.net> writes:

> 
> Alpar Juttner <ajuttner.list <at> googlemail.com> writes:
> 
> > On Thu, 2010-08-26 at 06:27 -0400, Steve Losh wrote:
> >> On Aug 26, 2010, at 12:54 AM, Alpar Juttner wrote:
> >
> >>  I think a -S/--subrepos option would be good, because I'm sure there
> >>  are people working with large repos that might not want to pull lots
> >>  of changesets they don't need.
> >
> > Look, my suggestion was to pull only those which has been used by the
> > main repo (and their ancestors, of course). I don't think it results
> > in significantly larger subrepos compared to the current behavior.
> 
> What we can do is to do 'hg pull' inside each subrepository when you do
> 'hg pull -S' in the outer repository. That should give you a consistent
> set of changesets so that you can update to all of them without pulling
> in any extra changesets as long as the set of subrepos does not change.
> 


I agree with Alpar's opinion about current misdesign of hg subrepos, unless I 
misunderstood something. Short example:

What happens if person1 has version1.0 of his software in repo1, and then he 
creates subrepo1 in repo1 as a dependency and creates version2.0 of their 
software. Then person2 clones this repo1 and subrepo1 to his repo2 and subrepo2. 
Then they find a bug in version1.0, so they both update their repositories to 
version1.0 (to fix the bug and create a branch), where subrepo was not present. 
How would they ever get back the subrepo??