Working with Subversion

[Mark A. Flacy]

On 2008.02.23 10:25, Patrick Mézard wrote:
> Antoine Pitrou a écrit :
> > Le mercredi 20 février 2008 à 22:30 +0100, Patrick Mézard a écrit :
> >>> IMO, using both named branches and clones is much more practical
> >> than using mq.
> >>
> >> I don't really see how the merge trick can be applied to more than  
> one
> >> revision.
> >
> > I'm not sure what "merge trick" you are talking about. I don't push  
> to
> > SVN, perhaps that's the reason :)
> 
> Say you have (U for upstream, P for patches)
> 
> U1--U2--U3--U4
>      \
>       \--P1--P2
> 
> How do you rebase P1 and P2 on top of U4 ?
> 
> The merge trick would be to end with:
> 
> U1--U2--U3--U4--P1*--P2*
>      \         /     /
>       \------P1----P2
> 

Why wouldn't it be...

U1--U2---U3---U4
      \          \
       \--P1--P2--P2*


...if the P1 is not interesting in and of itself?

> MQ does that better. But I may be missing something here.

IMOP, MQ is horrendous pain in the ass to work with, in that your work  
flows are much different with MQ than without.  You *always* have to  
have out-of-band knowledge on whether the patches have been applied or  
not before you can do anything constructive (such as "clone" or  
"update").

YMMV, of course, but that's how I see it.

-- 
Mark A. Flacy