[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Emdebian and wheezy-grip - HELP!



On Wed, 30 May 2012 13:33:27 -0400
Adam Magaluk <adammagaluk@gmail.com> wrote:

Adam, none of this is to you specifically, this is a general appeal to
everyone on this list.

> Currently i am using Emdebian on a project and am happy using the
> mutlistrap with the stable suite to build root file systems.
> 
> However I would like to experiment with the testing branch (wheezy).

Wheezy-grip is not getting regular updates currently, even though I have
been able to do a single set of ~4,000 updates recently. The
integration into Debian meant that the old scripts had to be rewritten
and there probably isn't time now to get that in place for the freeze.
Wheezy-grip (Emdebian Grip 3.0) will be a "best effort" release,
depending on how much time I can allocate between now and the end of
the Debian Wheezy release. I need a stable release based on wheezy-grip
as much as anyone else, if not more, so it's not as if this has been
overlooked.

Unstable-grip has been improving steadily over the last 4 months.
Automated dependency checking should fix most issues but, being
unstable, breakage is quite likely, depending on how your package
selection fits in with coordinated and uncoordinated library
transitions in Debian.

The problem is that testing migration rules are *HARD* to duplicate and
it isn't possible to just create wheezy-grip in isolation. Packages must
migrate, updates need to be identified, removed packages need to be
identified and handled. There are already 800 RC bugs in Debian and
those bugs don't stop coming just because the freeze starts. We take
10% of those packages but that amounts to several hundred updates.

> Is it possible to build Emdebian rootfs with mutlistrap for the testing
> branch?

No, not yet. Several hundred dependencies are missing.

apt-grip can be used to create a secondary repository containing just
the packages you need from Wheezy - include the converted packages into
a local repo using reprepro but that is only for a single architecture
and only gives you a snapshot of wheezy at the time you downloaded
stuff.

Until the freeze kicks in, the number of panicky maintainers cramming
their "urgent" fixes into Wheezy means that it will be unlikely that
wheezy-grip will be able to catch up and then once the freeze starts,
the work of bringing wheezy-grip into line with wheezy via
wheezy-proposed-updates just gets more complex.

In some ways, it may be easier to not push too much into wheezy-grip
until Wheezy is released as then it is a sitting target, not a moving
one.

So I'll do what I can for wheezy-grip 3.0 (based on Wheezy 7.0) but if
you want a truly stable Emdebian Grip package set using Wheezy, it's
likely that 3.0.1 (based on Wheezy 7.0.1) is more realistic. This is
how I expect to be using wheezy-grip as stable, 3.0.1 or even 3.0.2.

My objective with any stable release is to minimise broken dependencies
in the release itself. If that means that some packages in 3.0 are
behind the equivalent in Wheezy 7.0, then those are precisely the
updates which will have to wait for 3.0.1. We've had to do this before
for Lenny and Squeeze, it's not a new problem. It's not a nice problem
but we'll do what can be done.

Dependency resolution will be the primary tool to picking which packages
need updates prior to the release. Updates in leaf packages or updates
which do not have strict reverse dependencies will just have to wait.

This also means that the chances of getting *new* packages into the set
for wheezy-grip are slight. I will still add new packages to the set in
unstable-grip - people just need to ask - but I'm not going to make the
dependency resolution of wheezy-grip any harder by adding new chains
or new links to existing chains.

I'm expecting to do a few more mass-migrations / mass-removals before
the freeze but then it will be a case of carefully picking the packages
which fix the highest number of dependency problems. (Removals can be
just as useful in fixing dependency problems, don't forget.) There's no
key functionality likely to be removed for wheezy-grip which was in
squeeze - unless the packages themselves get removed from Debian for
whatever reason. I expect wheezy-grip to have a few more packages than
Squeeze but not much more than 3,000 per architecture.

If you're miffed by this (and you probably should be), then it's
high time to help and get involved at the Debian end. If you want
better Emdebian stable releases in the future, if you want to have an
Emdebian testing suite once Wheezy is released, you need to get involved
in Debian *now*, speak up for Emdebian, show that you're willing to put
in the time and fix stuff in Debian and get more Debian people thinking
that Emdebian is a helpful, positive, resourceful bunch of people who
deserve a bit more help with getting their stuff into Debian in the
future. We desperately need more people doing helpful stuff in Debian
who are identified with Emdebian. If we do more to help Debian, we gain
more support for Debian helping us.

Get involved with the Debian release process, keep an eye on packages
which you want to see in wheezy-grip (rc-alert), fix their RC bugs to
stop them being removed from wheezy and you'll help ensure the packages
survive into wheezy-grip. Oh, and if one of your favourite packages has
a new upstream release - *tell the maintainers you do NOT want it in
wheezy!* and make everyone's life easier. Tell them it's because you
want a stable Emdebian release, tell them not to be stupid by rushing
new upstream code into a stable release, tell them it's already too late
(which it is), tell them whatever you like as long as they don't annoy
the release team. The less work it is to release wheezy, the better
will be the quality of wheezy-grip. It is that simple.

If you want to help Emdebian then help Debian get Wheezy out quickly,
smoothly and without complicating things by having a new upstream
version in unstable only to find that the version in testing needs an
RC bug fix. Encourage people to put new stuff into experimental or just
wait until after the release - and explain why. Don't wait for the
freeze, get involved *now* and help unstable stay clean and manageable
to make it easier to get critical updates into wheezy.

Help me get a good stable release of wheezy-grip, even if that is only
really possible at the first point release of wheezy - by helping
inside Debian. FIX the existing RC bugs and dissuade maintainers from
causing new ones!!

-- 


Neil Williams
=============
http://www.linux.codehelp.co.uk/

Attachment: pgpkTMntqxXgh.pgp
Description: PGP signature


Reply to: