Re: Why isn't apt 0.5.4 moving to testing?

2001-09-14 Thread Anthony Towns
On Fri, Sep 14, 2001 at 09:36:28AM -0500, Steve Greenland wrote:
> We really (okay, you, Anthony :-)) really need to consider the idea
> of allowing architecture slips in testing, if, there's been a package
> that has been waiting more than (say) 10 days on a rebuild on fewer
> than (say) 30% of the architectures. 

It's not worth it. The package needs to be fixed anyway, it's better in
every way just to fix it and be done with it.

> That way, the affected packages
> wouldn't break on the recalcitrant architectures, they just won't be as
> current. Yes, that might have be tightened up as we approach release.

The point of testing is that it should always be ready for release.
Having random additional breakage appear regularly and automatically
doesn't really do anything good towards achieving that goal.

> But at least the packages would get more testing on the most used
> architectures.

One of the benefits of testing is that it makes release problems obvious
on a package-by-package basis, so they're a little more likely to be fixed
when they appear rather than all as a bunch when it's time to release.

> Now, if the package won't *build* on the problem architectures, that's
> a different problem. But if the autobuilders are just behind, then the
> people who want to support those architectures need to deal with the
> problem.

If any of the autobuilders do get significantly behind, I've been inclining
testing not to worry if they're out of date. This isn't a problem with the
autobuilders, though.

At the moment, woody's consistency across architectures is significantly
better than potato's on everything but powerpc and m68k, fwiw. It's
overall installability is much better too (well, if you ignore hppa
and ia64). Consider that each of these is by its very nature a RC bug
(the former for being distributed without the correct source, the latter
for making packages completely unusable on some architectures), and that
seems a reasonable achievement.

There're much better ways of avoiding these sorts of problems than saying
"oh, just ignore the brokenness". Some of them are even being worked on...

Cheers,
aj

-- 
Anthony Towns <[EMAIL PROTECTED]> 
I don't speak for anyone save myself. GPG signed mail preferred.

 ``Freedom itself was attacked this morning by faceless cowards.
 And freedom will be defended.''   Condolences to all involved.


pgpqNKi67kcf4.pgp
Description: PGP signature


Re: Why isn't apt 0.5.4 moving to testing?

2001-09-14 Thread Steve Greenland
On 13-Sep-01, 17:50 (CDT), Wichert Akkerman <[EMAIL PROTECTED]> wrote: 
> Previously Christian Leutloff wrote:
> > Is it really necessary that the package must be able to be upgraded on
> > every architecture!?
> 
> That's the whole purpose of testing, keep the brokenness to a minimum.
> 

So now we have the situation with testing having a buggy core tool (apt)
on all architectures, to avoid having broken accessory tools on a few
architectures? Yeah, that makes a lot of sense. 

We really (okay, you, Anthony :-)) really need to consider the idea
of allowing architecture slips in testing, if, there's been a package
that has been waiting more than (say) 10 days on a rebuild on fewer
than (say) 30% of the architectures. That way, the affected packages
wouldn't break on the recalcitrant architectures, they just won't be as
current. Yes, that might have be tightened up as we approach release.
But at least the packages would get more testing on the most used
architectures.

Now, if the package won't *build* on the problem architectures, that's
a different problem. But if the autobuilders are just behind, then the
people who want to support those architectures need to deal with the
problem.


Steve




Re: Why isn't apt 0.5.4 moving to testing?

2001-09-13 Thread Wichert Akkerman
Previously Christian Leutloff wrote:
> Is it really necessary that the package must be able to be upgraded on
> every architecture!?

That's the whole purpose of testing, keep the brokenness to a minimum.

Wichert.

-- 
  _
 /   Nothing is fool-proof to a sufficiently talented fool \
| [EMAIL PROTECTED]   http://www.liacs.nl/~wichert/ |
| 1024D/2FA3BC2D 576E 100B 518D 2F16 36B0  2805 3CB8 9250 2FA3 BC2D |




Re: Why isn't apt 0.5.4 moving to testing?

2001-09-13 Thread Christian Leutloff
Anthony Towns  writes:

> On Thu, Sep 13, 2001 at 11:16:53AM -0400, Paul D. Smith wrote:
> > I'm seeing lots of problems with apt 0.5.3 in my testing release
> > installation which have apparently been fixed in 0.5.4.  So, I wanted to
> > see why it wasn't in testing yet.
> > The excuses say:
> >   apt 0.5.4 (currently 0.5.3) (important) (low) 
> > Maintainer: APT Development Team <[EMAIL PROTECTED]> 
> > apt uploaded 22 days ago, out of date by 12 days! 
> > valid candidate (will be installed unless it's dependent
> >upon other buggy pkgs)
> 
> The List says:
> 
>apt: alpha: apt-move aptitude gnome-apt stormpkg
> 
> which means that the new apt breaks apt-move, aptitude, gnome-apt
> and stormpkg get broken unless they're upgraded at the same time as
> apt. Unfortunately aptitude isn't able to be upgraded (there's an update,
> but it's not built on arm yet) and additionally the updated gnome-apt and
> stormpkg were built with the old apt on powerpc, so can't be successfully
> upgraded either.

Is it really necessary that the package must be able to be upgraded on
every architecture!? If we have 10 (or more) architecture to support,
it is very likely that packages break on a single architecture and
hold on the whole improvement process aided through more people
testing new software (packages).

Bye
  Christian

-- 
Dipl.-Ing. Christian Leutloff, Aachen, Germany  [EMAIL PROTECTED]
   http://www.oche.de/~leutloff/[EMAIL PROTECTED]

Debian GNU/Linux - http://www.de.debian.org/
   taz muss sein. [EMAIL PROTECTED]




Re: Why isn't apt 0.5.4 moving to testing?

2001-09-13 Thread Anthony Towns
On Thu, Sep 13, 2001 at 11:16:53AM -0400, Paul D. Smith wrote:
> I'm seeing lots of problems with apt 0.5.3 in my testing release
> installation which have apparently been fixed in 0.5.4.  So, I wanted to
> see why it wasn't in testing yet.
> The excuses say:
>   apt 0.5.4 (currently 0.5.3) (important) (low) 
> Maintainer: APT Development Team <[EMAIL PROTECTED]> 
> apt uploaded 22 days ago, out of date by 12 days! 
> valid candidate (will be installed unless it's dependent
>upon other buggy pkgs)

The List says:

   apt: alpha: apt-move aptitude gnome-apt stormpkg

which means that the new apt breaks apt-move, aptitude, gnome-apt
and stormpkg get broken unless they're upgraded at the same time as
apt. Unfortunately aptitude isn't able to be upgraded (there's an update,
but it's not built on arm yet) and additionally the updated gnome-apt and
stormpkg were built with the old apt on powerpc, so can't be successfully
upgraded either.

Cheers,
aj

-- 
Anthony Towns <[EMAIL PROTECTED]> 
I don't speak for anyone save myself. GPG signed mail preferred.

 ``Freedom itself was attacked this morning by faceless cowards.
 And freedom will be defended.''   Condolences to all involved.


pgpCIckcMQrdZ.pgp
Description: PGP signature


Re: Why isn't apt 0.5.4 moving to testing?

2001-09-13 Thread Daniel Burrows
On Thu, Sep 13, 2001 at 11:16:53AM -0400, "Paul D. Smith" <[EMAIL PROTECTED]> 
was heard to say:
> Packages which satisfy both of those dependencies _are_ currently in
> testing, so... why isn't APT?

  I'm not sure, but I think deity would break if apt moved to testing.

  Daniel

-- 
/ Daniel Burrows <[EMAIL PROTECTED]> ---\
|   DROP THE SCYTHE AND TURN AROUND SLOWLY.   |
| -- Terry Pratchett, "Reaper Man"|
\ Real Programmers don't have braces. -- http://www.python.org ---/