[prev in list] [next in list] [prev in thread] [next in thread] 

List:       kde-core-devel
Subject:    Re: RFC: The Road Ahead
From:       Martijn Klingens <mklingens () yahoo ! com>
Date:       2001-07-27 13:17:17
[Download RAW message or body]

On Friday 27 July 2001 14:26, Dirk Mueller wrote:
> Moin Martijn!
>
> > But aren't all these methods ways of working around the true problem? I
> > know all of this, but I still have to _try_ the exact branch name. And
> > calling it "UNSTABLE_BRANCH" instead of "UNSTABLE" is maybe even better.
>
> I understand, however I didn't see it as a problem up to now ;-)
>
> The real issue is that you _want_ to have different names for the branches
> because otherwise you loose the development in the "old" UNSTABLE_BRANCH as
> soon as you start the new "UNSTABLE_BRANCH" because tags have to be unique.
>
> so I rather want to do a "cvs di -rmake_it_cool_branchpoint
> -rmake_it_cool_branch" instead of finding out when the development in a
> UNSTABLE_BRANCH for the previosu feature stopped and then update to handle
> around with timestamps.
>
> The disadvantage with CVS is that its amazingly difficult to diff between
> dates in branches. So why your idea sounds nice at first sight, it makes it
> actually a lot more difficult for development, as you need separate
> checkouts you can diff (as you have to update backwards first to diff
> against the branch, which screws all your local modifications because of
> the CVS conflicts up). Either that, or you end up with not working with CVS
> at all because its so horrible to use in such a case.

Hmm... I really forgot about the backporting issue that you would get in the 
end...

Maybe there is a way around, but maybe we should indeed drop my idea instead. 
Anyway, for those interested:

Assumption:
I don't think there are multiple "unstable" branches active at once. If there 
are then at least the secondary and later branches need their own naming 
scheme, or you would end up with the diffing mess you described.

Ok, and now the "solution":
If a branch is not unique by name, but at least isn't used by multiple 
separate unstable branches at once (see assumption) then one could tag the 
unstable branch as soon as the branch is used again. So you get a _tag_ 
KSIRC_RICHTEXT instead that marks a timestamp on the UNSTABLE branch. 
Backporting can be done against the tag, but people who just want to checkout 
the code don't have to worry about the naming and can just checkout the 
UNSTABLE branch.

If it's worth it? Don't know. Depends on how it is actually used in CVS I'd 
say. Life would have been so much easier if the backporting issue hadn't been 
there :-)

So feel free to speak up...

Martijn

[prev in list] [next in list] [prev in thread] [next in thread]