[PATCH] [mq] addchangegroupe finalize mq patches if changeset children are added

[Matt Mackall]

On Sat, 2011-03-12 at 19:29 +0100, Martin Geisler wrote:
> Dirkjan Ochtman <dirkjan at ochtman.nl> writes:
> 
> > On Sat, Mar 12, 2011 at 18:57, Pierre-Yves David
> > <pierre-yves.david at ens-lyon.org> wrote:
> >> # HG changeset patch
> >> # User Pierre-Yves David <pierre-yves.david at ens-lyon.org>
> >> # Date 1299952567 -3600
> >> # Branch stable
> >> # Node ID bbfa6f943b255fec0d6a4183597939f022bfbf00
> >> # Parent  38c9837b1f75f53cd3e5414e4283d21da1eb125f
> >> [mq] addchangegroupe finalize mq patches if changeset children are added.
> >>
> >> Before this change pulling descendant to applied mq patches result in standard
> >> changesets descendants to mq managed patches. This state is invalid mq wise.
> >> With this change, any mq patches getting standard changeset descendant get
> >> finalize uncondfitionnaly. Status messages are displayed for the user.
> >>
> >> Commiting a new changeset over a mq patches is still forbiden.
> >
> > Seems to me like we should instead just refuse to add changesets from
> > the pull to the repository if they descend from changes that are in
> > MQ. i.e. abort with a hint to finish the patches first.
> 
> Why is that? If the pull brings in children of the mq patches, then that
> is because what is mq patches in our repository also exists as
> changesets in another repository. Then I think it makes good sense to
> finalize the mq patches since they've spread to somewhere else.

I'm leaning slightly to this side.


-- 
Mathematics is the supreme nostalgia of our time.