Tags from closed branches?

[John Mulligan]

On Tue, Jun 09, 2009 at 11:24:06AM -0500, Matt Mackall wrote:
> On Tue, 2009-06-09 at 17:40 +0200, Dirkjan Ochtman wrote:
> > On Tue, Jun 9, 2009 at 17:26, Matt Mackall<mpm at selenic.com> wrote:
> > > I think that it's probably wrong for heads() to know anything at all
> > > about the concept of branches, be they open or closed.
> > 
> > Our concept of closed is actually about heads, not about branches, AFAICS.
> 
> Then it is broken. Only ==branch heads== should be closable and branch
> heads are distinct from heads. 
> 

In that case I'm pretty sure that my recent changes f'ed things up.

> If you start 'closing' heads, you start breaking a whole ton of things
> in Mercurial that have never heard of the branches concept. Tags is one.
> Push and pull are another. I suspect with the latest changes that you
> will be unable to push/pull/bundle/in/out the branch-closing markers
> because the changeset discovery protocol will not be able to find the
> heads in question.

I was concerned that my changes would have affected push & pull myself so
I tested it and it worked. Either I screwed up my test or it is going
through a different path to get the "real" list of heads.