[parisc-linux] 2.5 kernel and CVS -- opinions?

[parisc-linux] 2.5 kernel and CVS -- opinions?

[Paul Bame]

It's time to start tracking the 2.5 kernel at cvs.parisc-linux.org
and there are a few different ways to do it -- I'm looking for opinions.
My favorite option at the moment is 1A.  SPEAK UP BY FRIDAY because I'll
probably make CVS changes pretty soon after that.  I'll send a
HOWTO to the mailing list after the CVS changes are made.

The first decision:

    1. Start a separate 'linux-2.5' in CVS just like 'linux' with both
       an upstream branch and a development branch.  This is more manual
       perhaps, but using 'diff' and 'patch' and having 2.4 and 2.5 trees
       checked out all the time, it can accomplish the same things as #2.

    2. Add the 2.5 upstream and development branches to CVS 'linux'.
       IMO this is the most functional and flexible -- allowing us to
       use CVS to do the myriad of interesting 'diffs'.  CVS could also
       be used to keep single versions of 2.4/2.5 pa-specific files where
       no fork is required, but that will require some developers to acquire
       additional CVS expertise so I'm not sure whether to recommend that.
       2.4-only developers would not need to change their use model.  Not
       that it's an issue, but #2 requires less CVS disk space.

The second decision:

    Date-based CVS update on the trunk is broken (bugs filed -- no
    action for a long time).  It works fine on non-trunk branches.

    A. We've lived with this defect so far, don't change anything.
       If we choose this with #2 above, the only thing on the trunk
       is 2.4 development -- 2.5 devel date-based checkout will
       work fine.  However 2.4 devel date-based checkout will work
       less often than it does now.

    B. Move everything in CVS to non-trunk branches so we can use
       date-based CVS update.  This would cause another little change
       in how we use CVS.

    Option A is even more acceptable if we do either aggressive tagging
    (for example, each -paXX release gets tagged) or keep a fairly
    large set of the daily source tarballs.  IMO tagging is the "right"
    answer, but it's pretty slow due to CVS limitation or perhaps
    we could do a daily tagging from cron so nobody'd have to wait.

Re: [parisc-linux] 2.5 kernel and CVS -- opinions?

[Grant Grundler]

Paul Bame wrote:
> It's time to start tracking the 2.5 kernel at cvs.parisc-linux.org
> and there are a few different ways to do it -- I'm looking for opinions.
> My favorite option at the moment is 1A.

I'm ok with 1A as well:
o *someday*, I hope the CVS date-based check out will get fixed.
o TAGs work and I'm ok with tag being used.
o We could TAG weekly (or monthly) and keep enough nightly
  checkout's to cover two or three tags. Or whatever we can
  afford for diskspace.


thanks!
grant

Re: [parisc-linux] 2.5 kernel and CVS -- opinions?

[Thomas Bogendoerfer]

On Tue, Dec 11, 2001 at 09:29:25PM -0700, Grant Grundler wrote:
> Paul Bame wrote:
> > It's time to start tracking the 2.5 kernel at cvs.parisc-linux.org
> > and there are a few different ways to do it -- I'm looking for opinions.
> > My favorite option at the moment is 1A.
> 
> I'm ok with 1A as well:

I don't have a hard opinion whether having two tree makes more sense than
trunk/branch hoping. People will probably work for some time on 2.4
and switch over to 2.5, so I guess two trees are ok.

> o *someday*, I hope the CVS date-based check out will get fixed.
> o TAGs work and I'm ok with tag being used.

tags are sometimes nice, but too much tags aren't. Try a cvs log with hundreds
of tags. That's not funny. And btw. tags are (re-)movable and they don't have a
history, which makes them pretty useless for any serious build system. At
work we use seperate cvs controlled sourcelists, which lists every file with
its revision of a release. That way it's always possible to rebuild an old
release even if a dork moved/removed a tag.

> o We could TAG weekly (or monthly) and keep enough nightly
>   checkout's to cover two or three tags. Or whatever we can
>   afford for diskspace.

I don't see a point for needing that. But maybe I'm just overlooking something.

I personaly wouldn't touch 2.5 at the moment. Right now 2.5 looks to
me like an adventure park of out of control kids. That's ok but IMHO it
doesn't help Linux/PARISC development. Hunting generic bugs mixed with
architecture bugs isn't fun. Let's wait until 2.5 settles and Linus will
accept PARISC inclusion (which he won't at the moment).

Thomas.

-- 
Crap can work. Given enough thrust pigs will fly, but it's not necessary a
good idea.                                 [ Alexander Viro on linux-kernel ]