Debian Developers - Help needed with FTBFS bug 267413

Author

Help needed with FTBFS bug 267413 - buildd problem?

Frank Küster

2004-08-23, 5:55 pm

Hi,

Very short version:=20

Is it possible that a buildd keeps debconf information of uninstalled
build-dependencies?

Long version:

I am kind of puzzled about http://bugs.debian.org/267413. zope_2.6.4-1.2
has failed to build on ia64, sparc, mips, m86k, powerpc: On nearly every
architecture tried, except s390, and i386 for which the Uploader built
the package. This points to a problem that is not really architecture-
specific, doesn't it?

The actual error message is=20

debiandoc2ps debian/doc/zope-policy.sgml
debiandoc2ps: ERROR: zope-policy.ps could not be generated properly
debiandoc2ps: rerun with the -v option to found out why
make: *** [zope-policy.ps] Error 1

This doesn't tell much without the output that "-v" would give,
therefore, I tried to reproduce it, but essentially I failed. Those
attempts uncovered a different bug, giving the same error message, also
discussed in this bug log.=20

But this cannot, or should not, be the cause of failure on a buildd, and
if this other problem is fixed, I cannot reproduce it; it builds fine
here on i386 as well as on merulo (ia64).

I also had the idea that it might be connected with some special setup
in the buildd's, and set up a sbuild chroot locally (i386): Still not
reproducable, zope builds fine.

So I don't really have an idea what the reason might be, and how to
proceed. The only little thing is that other bug, which I think cannot
(or should not) occur on a buildd.

This other bug is that tetex-bin has switched some debconf defaults from
false to true, without catering for the change in config or
postinst. This has the effect that on machines kept up-to-date in sid or
sarge with noninteractive updates (or hit-RETURN-always), the old
defaults are still in effect. But, and that is what makes it a real bug,
tetex-bin now relies on the new defaults, otherwise it will fail to
operate under some circumstances. Installation will work, however.

Well, of course "debiandoc2ps *.sgml" is one of those circumstances,
giving the error message described above. BUT: On a buildd this bug
should never be triggered, because each time a build-dependency (like
tetex-bin) is installed, it is also purged again: purged, not
uninstalled.=20

- Is there any possibility that debconf answers are conserved between
build attempts?

- Is there any way to have one of the failing buildds execute
"debiandoc2ps -v some.sgml" without uploading a package just for that?=20

Many thanks in advance,
Frank
--=20
Frank K=FCster, Biozentrum der Univ. Basel
Abt. Biophysikalische Chemie

Laszlo 'GCS' Boszormenyi

2004-08-23, 8:48 pm

Warning: I am not a DD yet, I may be wrong!

* Frank Küster <frank@debian.org> [2004-08-23 17:45:15 +0200]:

> Is it possible that a buildd keeps debconf information of uninstalled
> build-dependencies?
I do not think so. See any build log, how the buildd's finish their
job:
/usr/bin/sudo dpkg --root=/home/buildd/build/chroot-unstable --purge m4 bison html2text [...]
^^^^^
As I know, purge removes everything; package, configuration files and
debconf data as well.

> The actual error message is
>
> debiandoc2ps debian/doc/zope-policy.sgml
> debiandoc2ps: ERROR: zope-policy.ps could not be generated properly
> debiandoc2ps: rerun with the -v option to found out why
> make: *** [zope-policy.ps] Error 1
I think the bug is in that package.

> This doesn't tell much without the output that "-v" would give,
> therefore, I tried to reproduce it, but essentially I failed.
Have you tried it with pbuilder?

> I also had the idea that it might be connected with some special setup
> in the buildd's, and set up a sbuild chroot locally (i386): Still not
> reproducable, zope builds fine.
I do not know how sbuild versus buildd versus pbuilder different, but I
bet there are some. Also please note that buildd's are strict (well,
IMHO) about the build tools, I don't think they update it day by day.

> - Is there any possibility that debconf answers are conserved between
> build attempts?
See above, and you also mentioned that packages are purged.

> - Is there any way to have one of the failing buildds execute
> "debiandoc2ps -v some.sgml" without uploading a package just for that?
Ask any buildd admin, lurk on www.buildd.net ; maybe Martin Loschwitz
<madkiss@debian.org> can help you. He is also German AFAIK, and
responsive.

Regards,
Laszlo/GCS

--
To UNSUBSCRIBE, email to debian-devel-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org

Thaddeus H. Black

2004-08-23, 8:48 pm

Laszlo wrote,

> I think the bug is in that package.

I originated the suspect patch. I will look
into it now, and will write again within 72
hours.

--
Thaddeus H. Black
508 Nellie's Cave Road
Blacksburg, Virginia 24060, USA
+1 540 961 0920, t@b-tk.org

Frank Küster

2004-08-24, 3:19 am

"Thaddeus H. Black" <t@b-tk.org> schrieb:

> Laszlo wrote,
>
>
> I originated the suspect patch. I will look
> into it now, and will write again within 72
> hours.

Which patch?

Regards, Frank
--=20
Frank K=FCster, Biozentrum der Univ. Basel
Abt. Biophysikalische Chemie

Frank Küster

2004-08-24, 3:19 am

Laszlo 'GCS' Boszormenyi <gcs@lsc.hu> schrieb:

> Warning: I am not a DD yet, I may be wrong!
>
> * Frank K=FCster <frank@debian.org> [2004-08-23 17:45:15 +0200]:
>
> I do not think so. See any build log, how the buildd's finish their
> job:
> /usr/bin/sudo dpkg --root=3D/home/buildd/build/chroot-unstable --purge m4=
bison html2text [...]
> ^^^^^
> As I know, purge removes everything; package, configuration files and
> debconf data as well.

This is how it should be, I was rather suspecting a bug. But I was
stupid, everything is okay: I had a closer look at the install messages
of tetex-bin in the failed buildd logs, and I can see that the debconf
answers are as they should be.

=20[vbcol=seagreen]
> Ask any buildd admin, lurk on www.buildd.net ; maybe Martin Loschwitz
> <madkiss@debian.org> can help you. He is also German AFAIK, and
> responsive.

Thanks very much, Frank
--=20
Frank K=FCster, Biozentrum der Univ. Basel
Abt. Biophysikalische Chemie

Thaddeus H. Black

2004-08-24, 8:19 am

Frank asked,

> Which patch?

The Zope password patch (bug #251038).

I wish that it were not, but as the last patch
applied before the FTBFS, my password patch is
necessarily suspect. This is unavoidable.
Three points:

* The password patch does at least seem to
relate to bug #267528.

* It is not obvious to me how a password patch
could affect debiandoc2ps activity. Perhaps
it does not affect it; but to establish this
theory, I need to diagnose the debiandoc2ps
FTBFS bug #267752.

* The zope package seems to build fine under
pbuilder. Why it apparently does not want to
autobuild (sbuild?), I do not yet know; but
I am working on it.

(This message relates potentially to bugs
#220370, #229368, #232648, #251038, #267384,
#267394, #267528 and #267752, variously against
the zopectl and zope sources. Some of these
bugs probably could be merged, but rather than
to spend time analyzing the web of bug reports,
the effort now is to find and to attack the root
of the RC problem.)

I am neither a Zope user nor the Zope
maintainer, incidentally. A Zope malfunction
could easily escape my notice. For this reason,
the new bug reports are appreciated.

I wish that the maintainer were somehow
available.

--
Thaddeus H. Black
508 Nellie's Cave Road
Blacksburg, Virginia 24060, USA
+1 540 961 0920, t@b-tk.org

Blars Blarson

2004-08-25, 4:24 am

In article <87n00loo38.fsf@alhambra.bioz.unibas.ch> frank@debian.org writes:
>Is it possible that a buildd keeps debconf information of uninstalled
>build-dependencies?

Pbuilder always starts from the same point, by default base +
build-esential. Sbuild does not always purge packages installed for a
build, (I don't think it does after a successful build) and some
packages may have broken purge that will cause problems for later
builds.

--
Blars Blarson blarson@blars.org
http://www.blars.org/blars.html
With Microsoft, failure is not an option. It is a standard feature.

--
To UNSUBSCRIBE, email to debian-devel-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org

Wouter Verhelst

2004-08-25, 4:24 am

Op wo 25-08-2004, om 06:19 schreef Blars Blarson:
> In article <87n00loo38.fsf@alhambra.bioz.unibas.ch> frank@debian.org writes:
>
> Pbuilder always starts from the same point, by default base +
> build-esential. Sbuild does not always purge packages installed for a
> build, (I don't think it does after a successful build)

It does. Well, it tries at least; but if there's an error in one of the
postinsts, it just doesn't care and goes on with the next build. This is
based on the premise that usually, a problematic dpkg run doesn't mean
that the next dpkg run will fail as well, but that trying to recover is
very likely to cause another failure.

--
EARTH
smog | bricks
AIR -- mud -- FIRE
soda water | tequila
WATER
-- with thanks to fortune

--
To UNSUBSCRIBE, email to debian-devel-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org