[darcs-users] two-way sync between cvs and darcs

[Nathan Gray]

On Sun, Apr 10, 2005 at 12:27:38AM +0200, Lele Gaifax wrote:
> >>>>> "Nathan" == Nathan Gray <kolibrie at graystudios.org> writes:
> 
>     Nathan> On Sat, Apr 09, 2005 at 11:47:06AM +0200, Thomas Zander
>     Nathan> wrote:
>     >> On Saturday 09 April 2005 11:11, Lele Gaifax wrote: > Tailor
>     >> could > learn about yet another command line option to disable
>     >> that, or I may > completely drop the code and assume nobody
>     >> will manually update the > repo...
>     >> 
>     >> Or; you could an a cvs up -A on the file prior to patching it.
>     >> You would immediately get feedback on failure if the file
>     >> actually changed (the P line being printed by cvs).  Since if
>     >> you are not looking at the last version the merge would be
>     >> wrong anyway since you can't put darcs changes before the
>     >> changes of somebody elses cvs commit.
> 
>     Nathan> That sounds reasonable.  So if tailor was asked to update
>     Nathan> a cvs repository from a darcs repository, it would call
>     Nathan> 'cvs up -A' before pulling in the darcs patches, and then
>     Nathan> pull and commit as normal.
> 
> Uhm, maybe I should rethink the stuff, because I'm missing the
> point... or does Thomas mean that tailor should reset with -A before
> update and set again afterward?

That's the idea I was getting.

> IMHO, a person who is going to maintain a two-way sync should be
> expert enough to *know* that that repos must manipulated *only* by
> tailor... and from this point of view, I'd be inclined to offer an
> option to not force sticky tag. 

That's probably true.

> Nathan, is this blocking for you? I'm short of time, and (luckily) CVS
> pressure is (slowly) fading away, so I won't be working on this very
> soon... 

Well, I am blocked right now, but I can get by with a non-code solution, if
'cvs update -A' works, and resets/removes the sticky tag.

I might peek at the code myself, and see what I can discern.

-kolibrie