[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Changed base ABI hash, Was: Accepted ghc6 6.12.1-9 (source all i386)


Am Dienstag, den 16.02.2010, 17:21 +0200 schrieb Kari Pahula:
> On Mon, Feb 15, 2010 at 09:44:29AM +0100, Joachim Breitner wrote:
> > Am Samstag, den 13.02.2010, 16:51 +0000 schrieb Kari Pahula:
> > >  ghc6 (6.12.1-9) unstable; urgency=low
> > >  .
> > >    * Remove /usr/lib/ghc-$(ProjectVersion)/lib/haddock (ie. the internal
> > >      haddock binary that ghc6 used at build time) from the ghc6 package.
> > >    * Install haddock's files from /usr/share/haddock-$VERSION in ghc6
> > >      package, not ghc6-doc.
> > 
> > caused, at least on AMD64, a change in the base ABI hash:
> Ouch.  I didn't think of testing for those since all I did was to
> change haddock related things.  One thing that implicitly changed from
> -8 to -9 was that the former was built with 6.10.4, the latter with
> 6.12.1.  But with a stage 2 compiler, it shouldn't matter.  Perhaps it
> did.  Anyone still have -8 and -9 ghc6 debs for amd64 on their hard
> drive?  I'd like to hear what's the output of
> http://people.debian.org/~kaol/iface-diff run on those two.

On ravel, in ~nomeata/.

> > This means we need to do binNMUs, and we need to do them ???blindly???
> > because we can not tell from the outside what ABI versions ghc provides.
> I could remember to check if the interfaces' hashes change from
> version to version even when there's no specific cause to suspect it.
> No harm in playing it safe.  But I would have needed to do it on the
> amd64 package to catch this one.  I built the i386 packages myself and
> they were indeed fine, now that I checked them.

Ok, so it does not even affect all arches, making binNMUs even harder...
did anyone hear if other arches besides amd64 are affected?

> > We could consider using dh_haskell_provides also for ghc6 and ghc6-prof,
> > so that they list the ABI is they provide. Any other ideas?
> I still wish we could get away with just relying on not changing
> ghc6's ABI in Debian revisions.

This is of course desirable in any case, but I wonder if we really
should rely on it. What speaks against using Provides to enforce this?


Joachim "nomeata" Breitner
Debian Developer
  nomeata@debian.org | ICQ# 74513189 | GPG-Keyid: 4743206C
  JID: nomeata@joachim-breitner.de | http://people.debian.org/~nomeata

Attachment: signature.asc
Description: Dies ist ein digital signierter Nachrichtenteil

Reply to: