|
Home > Archive > Debian Developers > May 2007 > Can/should debconf notes still be used?
You are viewing an archived Text-only version of the thread.
To view this thread in it's original format and/or if you want to reply to
this thread please [click here]
| Author |
Can/should debconf notes still be used?
|
|
| Frank Küster 2007-05-29, 7:23 am |
| Hi all,
in the past the use of debconf notes has been discussed, and it seemed
to be general consensus that they are mostly useless (or misused), but
that there are still a couple of legitimate uses which prevent their
removal from debconf (or making them a no-op) [1]
Now I am at the point where I would like to introduce a new debconf
note, and I'd like to request comments before we do that.
We are dealing with the cleanup after bug #420390. Briefly, tetex-base's
postrm was written under the assumption that teTeX is the only TeX
system, and the postrm removes a lot of files below /etc which were
installed by older versions of teTeX - but now with the appearance of
texlive, these files are conffiles of texlive-* again. Some of these
files are essential, texlive-base, texlive-base-bin, or
texlive-latex-base won't configure without them. The postrm has been
corrected now, but many people have already done the purge.
To me, the solution is to resurrect these conffiles without prompting,
because prompting doesn't make sense if the only working answer is
"yes". Even if I know that "preserve local changes upon upgrade"
normally means also to preserve conffile removals - I think this cannot
be applied here. I'm also considering to resurrect the non-essential
conffiles while on the way, either unconditionally or only when the
essential ones were missing, too.
Now the question is, how should we notify the user about what we've
done? Since this is a violation of the letter of policy, I don't think
a remark in NEWS.Debian is appropriate, and I'd like to use a debconf
note of priority "high". But notes are considered deprecated. On the
other hand, it's not an error, so the error type doesn't seem
appropriate...=20
Comments welcome,
Frank
[1] http://thread.gmane.org/gmane.linux...general/106537/
--=20
Dr. Frank K=FCster
Single Molecule Spectroscopy, Protein Folding @ Inst. f. Biochemie, Univ. Z=
=FCrich
Debian Developer (teTeX/TeXLive)
| |
| Steve Langasek 2007-05-29, 7:23 am |
| On Tue, May 29, 2007 at 09:01:20AM +0200, Frank Küster wrote:
> Now the question is, how should we notify the user about what we've
> done? Since this is a violation of the letter of policy, I don't think
> a remark in NEWS.Debian is appropriate, and I'd like to use a debconf
> note of priority "high". But notes are considered deprecated. On the
> other hand, it's not an error, so the error type doesn't seem
> appropriate...
IMHO, documenting a policy violation in a package is never an acceptable
solution, whether that's via NEWS.Debian or a debconf note.
I would rather suggest that this is not a policy violation; policy does say
that "local changes" must be preserved, but a) if the config files were
deleted by tetex this is not a "local change", b) I don't think the
requirement to preserve local changes was ever meant to be taken as "admins
have an inalienable right to screw up a package's installation, and packages
must not try to correct for this", which seems to be the case you're
describing.
C.f. the handling of update-rc.d, which considers the removal of all rc.d
symlinks to be an invalid config and recreates them for the user.
--
Steve Langasek Give me a lever long enough and a Free OS
Debian Developer to set it on, and I can move the world.
vorlon@debian.org http://www.debian.org/
--
To UNSUBSCRIBE, email to debian-devel-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org
| |
| Frank Küster 2007-05-29, 7:23 am |
| Steve Langasek <vorlon@debian.org> wrote:
> I would rather suggest that this is not a policy violation; policy does s=
ay
> that "local changes" must be preserved, but a) if the config files were
> deleted by tetex this is not a "local change", b) I don't think the
> requirement to preserve local changes was ever meant to be taken as "admi=
ns
> have an inalienable right to screw up a package's installation, and packa=
ges
> must not try to correct for this", which seems to be the case you're
> describing.
That way to look at it makes me feel much better...=20
Regards, Frank
--=20
Dr. Frank K=FCster
Single Molecule Spectroscopy, Protein Folding @ Inst. f. Biochemie, Univ. Z=
=FCrich
Debian Developer (teTeX/TeXLive)
| |
| Don Armstrong 2007-05-29, 7:23 am |
| On Tue, 29 May 2007, Frank K=FCster wrote:
> To me, the solution is to resurrect these conffiles without
> prompting, because prompting doesn't make sense if the only working
> answer is "yes".=20
Can you test to see if the system is working without the conffiles?=20
Then if it is, then assume the deletion was intentional (and perhaps
give information in NEWS.Debian on how to restore them); if not,
restore them once.
Don Armstrong
--=20
DIE!
-- Maritza Campos http://www.crfh.net/d/20020601.html
http://www.donarmstrong.com http://rzlab.ucr.edu
| |
| Don Armstrong 2007-05-29, 7:23 am |
| On Tue, 29 May 2007, Frank K=C3=BCster wrote:
> Don Armstrong <don@debian.org> wrote:=20
>=20
> There are about 8 RC bug reports which show quite clearly that it
> doesn't. There's no sense in testing that in the maintainer script. If
> a local admin hacked the system so much that it keeps on working without
> them, I frankly don't care - it's no longer a Debian TeX Live system.
My point was that in such a system you'd likely end up breaking it by
reinstalling the conffiles; if such a test is relatively easy, then
it'd be worth it. If not, then a NEWS.Debian entry on how to return to
the previous configuration is probably good enough for such an admin.
Don Armstrong
--=20
I shall require that [a scientific system's] logical form shall be
such that it can be singled out, by means of emperical tests, in a
negative sense: it must be possible for an emperical scientific system
to be refuted by experience.
-- Sir Karl Popper _Logic of Scientific Discovery_ =A76
http://www.donarmstrong.com http://rzlab.ucr.edu
| |
| Frank Küster 2007-05-29, 7:23 am |
| Don Armstrong <don@debian.org> wrote:
> On Tue, 29 May 2007, Frank K=FCster wrote:
>
> Can you test to see if the system is working without the conffiles?=20
There are about 8 RC bug reports which show quite clearly that it
doesn't. There's no sense in testing that in the maintainer script. If
a local admin hacked the system so much that it keeps on working without
them, I frankly don't care - it's no longer a Debian TeX Live system.
> if not,
> restore them once.
Yes, we'll do it only once anyway.
Regards, Frank
--=20
Dr. Frank K=FCster
Single Molecule Spectroscopy, Protein Folding @ Inst. f. Biochemie, Univ. Z=
=FCrich
Debian Developer (teTeX/TeXLive)
| |
| Frank Küster 2007-05-29, 1:23 pm |
| Don Armstrong <don@debian.org> wrote:
> On Tue, 29 May 2007, Frank K=FCster wrote:
>
> My point was that in such a system you'd likely end up breaking it by
> reinstalling the conffiles;=20
No, the trick would be to hack formats or executables to simply ignore
those files - reinstalling them won't have any effect.
> if such a test is relatively easy, then
> it'd be worth it. If not, then a NEWS.Debian entry on how to return to
> the previous configuration is probably good enough for such an admin.
We'll do that; the test would be quite hard, because there are tons of
possibilities to misconfigure your system such that the error looks
quite similar to a missing file.
Regards, Frank
--=20
Dr. Frank K=FCster
Single Molecule Spectroscopy, Protein Folding @ Inst. f. Biochemie, Univ. Z=
=FCrich
Debian Developer (teTeX/TeXLive)
| |
| Steve Langasek 2007-05-30, 7:17 am |
| On Tue, May 29, 2007 at 01:22:10PM +0200, Frank Küster wrote:
> Don Armstrong <don@debian.org> wrote:
>
[vbcol=seagreen]
[vbcol=seagreen]
> There are about 8 RC bug reports which show quite clearly that it
> doesn't. There's no sense in testing that in the maintainer script. If
> a local admin hacked the system so much that it keeps on working without
> them, I frankly don't care - it's no longer a Debian TeX Live system.
[vbcol=seagreen]
> Yes, we'll do it only once anyway.
How do you plan to ensure that? Since old versions of the tetex packages
with the broken postrm are still in the wild, and a user may have already
removed the tetex package but *not* purged it yet by the time fixed tetex
packages became available, AFAICS any strategy that will make sure that the
conffiles are only restored once (e.g., by only restoring the conffiles when
upgrading from an older version of the texlive packages) will not be able to
reliably fix this problem for all users.
--
Steve Langasek Give me a lever long enough and a Free OS
Debian Developer to set it on, and I can move the world.
vorlon@debian.org http://www.debian.org/
--
To UNSUBSCRIBE, email to debian-devel-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org
| |
| Frank Küster 2007-05-30, 7:17 am |
| Steve Langasek <vorlon@debian.org> wrote:
>
> How do you plan to ensure that?=20=20
Sorry, no plan, I was just dreaming. Or deliring.=20
> Since old versions of the tetex packages
> with the broken postrm are still in the wild, and a user may have already
> removed the tetex package but *not* purged it yet by the time fixed tetex
> packages became available, AFAICS any strategy that will make sure that t=
he
> conffiles are only restored once (e.g., by only restoring the conffiles w=
hen
> upgrading from an older version of the texlive packages) will not be able=
to
> reliably fix this problem for all users.
Exactly. And I think it's not worth the effort to do some weird "touch
/var/whatever" stuff which is error-prone again. As long as Debian
packages of texlive-base require the presence of modes.mf to be
configured, and Debian packages of tl-base-bin require pdftexconfig.tex,
we simply restore it with the original content if we find it removed.
In case someone really wants to hack their system to not need the file
(and loose any support by us), they can just have a file with only
"%%\bye" or "\endinput" in it, respectively, which will of course
respected.
Regards, Frank
--=20
Dr. Frank K=FCster
Single Molecule Spectroscopy, Protein Folding @ Inst. f. Biochemie, Univ. Z=
=FCrich
Debian Developer (teTeX/TeXLive)
|
|
|
|
|