[Bug 19415] new kernel-package breaks official linux-2.6 kernel builds, should not propagate to testing.

Wed Nov 9 16:17:12 UTC 2005

Please do not reply to this email.  You can add comments at
http://bugzilla.ubuntu.com/show_bug.cgi?id=19415
Ubuntu | kernel-package

------- Additional Comments From debzilla at ubuntu.com  2005-11-09 16:17 UTC -------
Message-ID: <20051109145406.GB3246 at localhost.localdomain>
Date: Wed, 9 Nov 2005 15:54:06 +0100
From: Sven Luther <sven.luther at wanadoo.fr>
To: 338283-done at bugs.debian.org, debian-kernel at lists.debian.org
Subject: Re: Point to an actual bug

On Wed, Nov 09, 2005 at 08:06:53AM -0600, Manoj Srivastava wrote:
> Hi,
> 
>         You know, the sheer ignorance displayed in this bug report is
>  staggering. Firstly, there is no upload to stable; all 10.X uploads

Yeah, sorry for thrusting your words, you said yourself that you uploaded it
to unstable, i know now that this was a typo, but you can hardly blame me for
your typos, at worst to not check more in detail, but as i downloaded from
incoming, i thrusted your words.

>  have been to experimental. Not a big deal, but it just  evidence that
>  the reporter has no clue, and is simply reporting out of sheer ill
>  will and does not want to have this arguments get confused with
>  facts.

Yeah, yeah, whatever.

> > As discussed on irc, the new 10.008 kernel-package was prematurely uploaded to
> > unstable, and breaks linux-2.6 builds. The issues at hand are : 
> 
>         This is the first thing that happens not to be true.

Indeed, but then you where the one lying : 

19:21 < Manoj> OK. I think I may have fixed the call to dpkg-shlibs for arches where
there is no dynamically linked elf binary in the header package. Again, everything
worked just fine on i386, so another upload to unstable is on the way, just in time
for dinstall
19:21 < Manoj> this is 10.008
19:36 < jonas> Manoj: great!
19:37 < svenl> i still have doubts, but we will see.
19:38 < Manoj> now all the generated packages are more-or-less lintian clean
(or as clean as I am inclined to make them)
19:38 < Manoj> you have doubts about what?
19:38 < svenl> about the timeliness to upload it to unstable but we will see.
19:39 < svenl> Manoj: on the other hand, upload early and often, and if people
complain loudly about breakage is probably the best way to go forward.
19:39 < Manoj> err, as long as there are no actual bugs in k-p I see no reason
why the time is wrong
19:39 < svenl> Manoj: will it be 100% backward compatible with 9.00x ?
19:40 < Manoj> and yes, uploading to stable gets it a far wider audience
19:40 < svenl> anyway, let's see what comes of it :)
19:40 < Manoj> which means I get more reports, logs, and proposed fixes
19:40 < svenl> Manoj: wrong. the kernel team probably gets more reports to be
triaged.
19:41 < Manoj> svenl: well, the published API has not changed. There are a few
changes, which are mentioned in NEWS.Debian
19:41 < svenl> like said, we will see.
19:41 < Manoj> so It is mostly backwards compatible, but in an internal
reorganization so large, some incompatibilities may have slipped through
19:42 < svenl> Manoj: the true test will be the 2.6.14-3 upload.
19:42 < Manoj> like the disappearance of IMAGE_TOP env var
19:42 < svenl> Manoj: i am not sure we care about that. Is it in incoming or
somewhere already ? I will trigger a full build with it.

> >   1. the missing config caues manuals not to build

Notice that this list was paraphrased verbatim from horms listing on irc.

>         kernel-package has never been meant t function without a
>  .config file; all documentation stresses the need to have one in
>  place before calling make-kpkg.  When presented with a case when
>  there is none, make-kpkg does its best to create a reasonable .config
>  file, as documented, and calls make oldconfig, to let the user answer
>  any options that are nwerer.
> 
>         The older versions of make-kpkg had a bug, meaning that
>  sometimes it failed to look into . and ./debian; this bug is now
>  fixed.
> 
>         The official process was taking advantage of this bug, and now
>   has to be fixed. Not a k-p bug.

Yeah, whatever, you still fail to understand that linux-2.6 depends on
kernel-package to build, and you doing random changes in k-p behavior will
break linux-2.6, and since that is our official kernel which we like to upload
quickly on security issues and such, this is indeed a problem for you doing
random breakages.

> >   2. the headers build binaries uneccessarily (but doesn't break build)
> 
>         It is not at all clear that the build is unnecesary, as the
>  reporter well knows, since he was around when this was discussed on
>  IRC (or perhaps he was unable to follow technical discussions). It is
>  not clear whether the kernel headers are really fully correct, and
>  have all the things in script/ dir compiled, unless a full build is
>  done; hence building anyway is the safe route.
> 
>         So this is either malice, or sheer technical incompetence, I
>  can't tell.

Plain copying of horms list. 

> >   3. missing depends and recommends in the linux-image package.
> 
>         At last report, the build succeeded on i386, and was being
>  retried on ppc, 

Indeed, did i not say this is being worked on ? We have not yet the
confirmation of horms built, and if you didn't scare me away with your insults
earlier, i would have been able to make the build and we would have had a
response to that now maybe.

> > These are probably due to a k-p regression, which means that both
> > the config for the arch-indep build and the control file provided by
> > the linux-2.6 build infrastructure are not used for the final
> > packages, for some yet mysterious reason.
> 
>         Err, I believe it is not k-p regression, it is misusing k-p,
>  or using it in undocumented ways, when not hacking around k-p and
>  making assumptions about internal behaviour would work.  There is a
>  problem, and something needs to be fixed -- but it is not clear that
>  that is kernel-package.

And this is your error, it is a problem in both kernel-package and linux-2.6
if this later is no more able to build, and you should take your
responsabilitied as kernel-package maintainer strongly enough to understand
that and care, and not default responsability for everything, like you have
been doing these past days. The previous version built linux-2.6, this one
didn't this is a regression, and no amount of fast-talking on your part will
change this.

> > This is being worked on, and hopefully a fix will be out soon, but
> > in the meantime, kernel-package 10.008 is useless for official
> > linux-2.6 builds, and since it is in unstable, means we cannot
> > upload anymore linux-2.6 packages until it is fixed, which makes
> > this RC.
> 
>         To this point, the reporter has done nothing apart from troll,
>  insult, and hinder efforts to diagnose the issue, and this BTS abuse
>  is just another step in the process.

Yeah, have i not done builds on powerpc, and was i not the first one to report
you the config failure yesterday evening ? True i lost the log file in a
screen mismanipulation, but still you cannot claim those above without
ignoring this, as for insults, i have never been less than polite to you, and
as i remember it was you calling me an idiot and stuff.

So, i think the but report was perfectly honest, and aimed at stopping the
package from migrating to testing undully, and the only error in it was that
i believed your words when you told us you would upload to unstable, words
which i didn't check, but which you didn't correct also when i told you it was
a bad idea.

Friendly,

Sven Luther

-- 
Configure bugmail: http://bugzilla.ubuntu.com/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the QA contact for the bug, or are watching the QA contact.