Re: [leaf-devel] bering-uclibc source tree

[Mike N. <mh...@us...>]

On Wed, 2002-11-27 at 14:31, Erich Titl wrote:
> At 22:42 27.11.2002, Mike Noyes wrote:
> >On Wed, 2002-11-27 at 12:50, Erich Titl wrote:
> > > With the
> > > consent of the lead developers of each branch it should be possible to
> > > build a tree which does not necessarily have to be maintained by the lead
> > > developer.
> >
> >I don't think so. By definition a lead developer is in charge of the
> >release/branch purpose and direction. Aren't both lost when abdicating
> >source tree control? I believe a release/branch source tree in our
> >repository not endorsed by it's lead developer would be a new
> >release/branch?
> 
> The important word IMHO is _endorsed_. What is the canonical way for this 
> endorsement? I doubt Linus Thorvalds still manages his own kernel CVS tree.

Erich,
In my opinion endorsed means the developer uses the tree to generate
releases. Or, in the case of bering-uclibc K.P. approached Jacques and
asked permission to use the bering name for a uclibc based tree.

> > > Maybe we could invite the lead developers of the various branches to
> > > mirror their respective cvs tree(s) to a public place where it is
> > > possible for the other developers to make branches/modifications which
> > > eventually would be either rejected or make it to the base. Of course
> > > this might change the development cycle a little.
> >
> >I'd like to avoid numerous cvs branch creations in our repository.
> >Merging multiple cvs branches is a significant challenge.

To ensure we're talking about the same things, I have these two
definitions:

    cvs branch == a cvs function
        5. Branching and merging
        http://www.cvshome.org/docs/manual/cvs_5.html#SEC54
    
    release/branch == LEAF project release/branch (e.g. Bering,
    Dachstein, Oxygen, WISP-Dist, PacketFilter, Lince).

> This is true and that is why I personally favour an open tree. CVS is a 
> wonderful tool for distributed development. Unfortunately this tool is IMHO 
> not used to its full potential in the LEAF community (which you pointed out 
> in the mail which triggered this thread).

I agree wholeheartedly.

> For example in my little CVS tree 
> I am the only one doing anything. If someone spots an error in something I 
> did, he can easily get the code, modify it, and make a _redundant_ copy in 
> his own CVS tree. Reporting this back to me is a compulsory thing. I might 
> not spot the same error  for a long time, continuing on my erroneous way 
> and someone else might later on find the same error again at nauseam. If I 
> understand CVS correctly that is not the way it was meant to be. CVS allows 
> concurrency and conflicts.

Correct. (see note at bottom of post)

> What is done with the various branches of a tree 
> is something which can be dealt with at release time (by of course the lead 
> developers or someone charged with the release task)

I'm not quite grasping your meaning here. Please elaborate.

> It is even less if someone just adds a little (hopefully not harmful) 
> extension to a part of the software. Unless this extension makes its way to 
> the base distribution it remains in its niche at the developers private 
> tree.

Correct. It is up to the release/branch lead developer and team members
to incorporate these extensions/patches. I expect people that repeatedly
provide useful additions would be asked to join the team.

> For example if I look at the ulibc Bering distribution they 
> certainlly keep parts of Bering as it is, as Jaqcues and Erik kept parts of 
> Dachstein. Some improvement made in a new branch may not find its way back 
> to its predecessor because they are living in different trees in the same 
> forest. Alike development in the main branch requires merging with a 
> foreign CVS tree to make its way to  the new branches. This IMHO leads to 
> redundancy.

Are you talking about LEAF releases/branches or cvs branches? Also note:
not all of our releases/branches derive from the same root.

> But all this is at of course the discretion of the lead developers.
> 
> OK, I hope I have not insulted anybody.

Not in the least. I find this type of discussion stimulating. :-)

> Folks, please use my CVS, and fix my bugs on the way ;-)

I second that.

Note: it is possible with our cvs_acls script to give write access to
certain sections of your personal tree to other project members.

-- 
Mike Noyes <mhnoyes at users.sourceforge.net>
http://sourceforge.net/users/mhnoyes/
http://leaf-project.org/  http://sitedocs.sf.net/  http://ffl.sf.net/