[Pkg-haskell-maintainers] Bug#676874: Bug#580038, #676874: ghc fails to configure after install (update-alternatives priority out of range)

[Joachim Breitner]

Hi,

Am Sonntag, den 10.06.2012, 21:44 +0200 schrieb Guillem Jover:
> On Sun, 2012-06-10 at 21:11:04 +0200, Joachim Breitner wrote:
> > Am Sonntag, den 10.06.2012, 03:46 -0400 schrieb Aleksandar Micovic:
> > > update-alternatives: priority is out of range
> > 
> > this is likely due to a call setting the priority to 8600000600, which
> > has been there since ghc6_6.4-1 (13 Mar 2005!). Was there a recent
> > change in update-alternatives that disallows large priorities? Sounds
> > like this change in dpkg_1.16.4:
> >   * Check all parsed integers for out of range errors; i.e. that no negative
> >     values are allowed if not appropriate, and that no overflows occur.
> >     Closes: #580038
> > And indeed, the diff at
> > http://anonscm.debian.org/gitweb/?p=dpkg/dpkg.git;a=commitdiff;h=2bf4b48#patch12
> > confirms that. CC’ing 580038 at b.d.o for that; looks like the fix introduced
> > a regression.
> 
> Well that number is bigger than INT_MAX which means it cannot be
> stored in the priority field used by u-a, as such that number will
> wrap-around and will not be what it's meant to be. This silent
> truncation has happened up to now, but it's catched correctly from
> now on.

could you, just to not break existing packages, change the code to only
print a warning and set the value to INT_MAX? After all the dpkg from
wheezy should be able to install packages in squeeze, shoudn’t it?

We’ll change the number in ghc, or rather replace the alternative by a
regular file, but I’d like to not do such changes so short before the
release.

> > (Not saying it is sane to use that number, but its there and its been
> > there for long, so I don’t think dpkg should break that so shortly
> > before the freeze.)
> 
> The above has happened probably since the rewrite of u-a from perl to
> C in 1.15.0, which is already included in squeeze.

I guess by „the above“ you mean the wrap-around, which theoretically
lead to wrong results, but in this is harmless as since squeeze (or even
longer) only ghc was providing an alternative for runhaskell. The
explicit check in 1.16.4, though, broke the existing packages in
practice.

I’m sorry for having to ask for the work-around outlined above, but I
don do think its the right course of action.

Greetings,
Joachim

-- 
Joachim "nomeata" Breitner
Debian Developer
  nomeata at debian.org | ICQ# 74513189 | GPG-Keyid: 4743206C
  JID: nomeata at joachim-breitner.de | http://people.debian.org/~nomeata
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 198 bytes
Desc: This is a digitally signed message part
URL: <http://lists.alioth.debian.org/pipermail/pkg-haskell-maintainers/attachments/20120610/bc0d82da/attachment.pgp>