Your message dated Thu, 20 Dec 2007 16:51:42 +0200
with message-id <[EMAIL PROTECTED]>
and subject line Bug#457151: dpkg-dev -- should not reorder Build-Depends
has caused the attached Bug report to be marked as done.
This means that you claim that the problem has been dealt with.
If this is not the case it is now your responsibility to reopen the
Bug report if necessary, and/or fix the problem forthwith.
(NB: If you are a system administrator and have no idea what I am
talking about this indicates a serious mail system misconfiguration
somewhere. Please contact me immediately.)
Debian bug tracking system administrator
(administrator, Debian Bugs database)
--- Begin Message ---
Package: dpkg-dev
Version: 1.14.12
Severity: serious
Tags: patch
Dear dpkg developers,
Please see this thread for context, though I've pasted relevant stuff:
http://lists.debian.org/debian-mentors/2007/12/msg00361.html
<quote>
Well, here's a further diagnosis. For some reason, the order of
installed packages in the buildds (and my pbuilder) are being
reordered in alphabetical order.
See this old log. The apt-get command follows the same order as
specified in the package's Build-Depends:
http://buildd.debian.org/fetch.cgi?&pkg=octave2.9&ver=1%3A2.9.17-1&arch=sparc&stamp=1195174810&file=log
But here, it is in alphabetical order:
http://buildd.debian.org/fetch.cgi?&pkg=octave2.9&ver=1%3A2.9.19-1&arch=sparc&stamp=1197772112&file=log
And I saw the source package for octave2.9, here's the control:
Build-Depends: g++-4.1 (>= 4.1.1-4), debhelper (>= 5.0.0), autoconf, texinfo,
texlive-latex-base, texlive-generic-recommended, g77,
libreadline5-dev,
[snip]
Notice that refblas3-dev is _before_ lapack3-dev. Now, if apt-get is
called faithfully in this order, atlas does not come in. But if it is
reordered, the lapack3-dev dependencies are honoured first, and that
pulls in atlas3-base, which is the first alternate dependency, and is
satisfiable.
</quote>
As this has caused enough damage already (e.g. octave2.9, numpy etc.)
not unnecessarily depend on atlas. Therefore, I chose to file this as
serious. Please feel free to downgrade severity if you feel so.
I've attached a trivial patch to prevent the ordering.
Thanks.
Kumar
--
Kumar Appaiah,
458, Jamuna Hostel,
Indian Institute of Technology Madras,
Chennai - 600 036
--- a/scripts/dpkg-source.pl 2007-11-23 08:12:51.000000000 +0530
+++ b/scripts/dpkg-source.pl 2007-12-20 12:13:58.000000000 +0530
@@ -324,7 +324,6 @@
error(_g("error occurred while parsing %s"), $_) unless defined $dep;
my $facts = Dpkg::Deps::KnownFacts->new();
$dep->simplify_deps($facts);
- $dep->sort();
$f{$_}= $dep->dump();
}
elsif (s/^X[BC]*S[BC]*-//i) { $f{$_}= $v; }
signature.asc
Description: Digital signature
--- End Message ---
--- Begin Message ---
Hi,
On Thu, 2007-12-20 at 09:25:44 +0100, Raphael Hertzog wrote:
> On Thu, 20 Dec 2007, Kumar Appaiah wrote:
> > Well, here's a further diagnosis. For some reason, the order of
> > installed packages in the buildds (and my pbuilder) are being
> > reordered in alphabetical order.
>
> Relying in the installation order of packages to get the right behaviour
> is asking for troubles.
>
> If you do not want a specific package as alternative, remove it from the
> alternative and/or build-conflict on it. It's as simple as that.
>
> (In fact it's an issue where we want to compile the debian package in one
> given environment but want the user to be able to build it in another one
> but we don't have debian-specific Build-Depends/Conflicts)
Yes, I'd say those are bugs on the packages. There's no guarantee on
what's going to be present on the build environment, except for the
build dependency relationships.
> > As this has caused enough damage already (e.g. octave2.9, numpy etc.)
> > not unnecessarily depend on atlas. Therefore, I chose to file this as
> > serious. Please feel free to downgrade severity if you feel so.
> But I really don't like the justification of the revert, thus I'm inclined
> to not revert it and close the bug. But I'd like to have the opinions of
> other dpkg developers first.
Agreed, and I've just done so.
regards,
guillem
--- End Message ---
