rebase with --keep

[Peter Arrenbrecht]

On Mon, Feb 1, 2010 at 4:35 PM, Waldemar <waldemar at beechwoods.com> wrote:
> In our development model, our client keeps sending us frequent updates to
> the code base we're modifying, too. Typically, there would be several
> updates, over 100, before we get a chance to merge our code with upstream.
> We are using rebase to cope with the situation.  In essence, we keep
> rebasing our csets on top of incoming updates.  It works very well except it
> has a side effect that we're not quite sure how to handle.  We use --keep
> with rebase because we share our code with several developers and have to
> preserve history. This produces a dangling head on every iteration.  We have
> to kill it somehow because in our system we can push only one head per
> branch. We found one solution to this which was to keep merging the extra
> head with the rebased csets.  This works but it produces an ever expanding
> sequence of trivial merge csets plus an ever expanding graph that is
> increasingly difficult to analyze.
>
> Is there another way to remove the head while preserving history?

You might want to look at pbranch, too. I use it extensively to
maintain a set of patches on top of regular Hg. But it also messes up
your graph, to be honest.
-parren