[darcs-users] Upcoming 2.8

[Ben Franksen]

Archi wrote:
> On Sat, 25 Feb 2012 07:21:00 -0700, Stephen J. Turnbull
> <stephen at xemacs.org> wrote:
> 
>> Archi writes:
>>
>> > What scenario were you wanting rebase for that patch commutation does
>> not
>> > solve? There are specific corner cases, but I don't see how cornercases
>> > make for a good marketing feature.
>>
>> Rebase is mostly about presentation, rather than solving problems.
>>
>> Also, in some cases you want to "force commute" across a dependency
>> (again usually for presentation purposes, but sometimes to make a
>> branch work better).
> 
> The new hunk splitting allows those commutes.

But only if you re-record everything back to the conflict point. This can be 
a lot of work and most of it is completely unnecessary because it could and 
should be automated.

The typical situation where I would like to have rebase (or a 'deep' amend-
record) is when I have accumulated a number of patches on a working branch 
and then someone else makes a change that conflicts with one of mine. I pull 
this patch and I get conflicts. Now, none of my patches have been published 
yet, so I would like to amend the one or two that conflict with the one I 
just pulled. If I am lucky, my conflicting patch(es) are at the top, or can 
be commuted there, because in that case I can just do an amend-record and 
avoid the conflict. The problem is that often other patches depend on teh 
one I want to amend and darcs does not allow me to do that.

What I want is a way to say to darcs: yes, I know there are other patches 
depending on the one I am going to amend, but do it anyway, and change those 
other (depending) patches where necessary, give them new identities, I don't 
care. What I care about is that I don't have to unrecord everything (back to 
where the conflict happens) and then re-record because that means selecting 
and splitting hunks etc etc.

If you can explain how to use hunk splitting so that I can avoid repeating 
all the work, then I'd be happy.

Cheers
Ben