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

Re: binNMUs to keep haskell packages in sync



This gets offtopic on debian-release. Please follow to debian-devel.

On Sun, Mar 09, 2008 at 11:55:15PM +0100, Joachim Breitner wrote:
> Am Sonntag, den 09.03.2008, 23:02 +0100 schrieb Bastian Blank:
> > On Sun, Mar 09, 2008 at 07:42:59PM +0100, Joachim Breitner wrote:
> > > haskell libraries have the unfortunate requirement to need exactly the
> > > version of dependencies installed that they were built against.
> > Can you please explain why? And does a rebuild in the same environment
> > also break it?
> AFAIK the compiler, GHC, does heavy cross-module optimization and
> inlining.

gcc does this also for c++. See /usr/include/c++, there is many
to-be-inlined code in it. And with a little bit of thinking it is
possible to have such an ABI fixed.

> > > Currently, we also fix the build-dependencies, to keep the packages in
> > > sync across different arches, and do sourceful uploads of all depending
> > > package when we upgrade a library. I don’t like this situation, and I’m
> > > wondering if we could not do binNMUs to keep the packages in sync.
> > BinNMUs will only work with loose build-deps.
> Exactly, but loose build deps would lead to packages getting out of sync
> unless we do binNMUs to fix that.

This is the case with every library transition. You do one for each
upload.

> I’m just not sure if point 4 wouldn’t be too much of a burden for the buildds and w-b admins.

Well. As buildd admin I already failed several haskell builds because of
not available build-deps. And I refuse to handle a library transition
which matches the size of the last large gnome update for each upload.

> > You are already in touch with security? This is a large problem.
> No, I wasn’t aware that is was such a problem. I’ll get in touch with
> them in case we have to stick to the strict build-dep situation. With
> loose build-deps, there woudn’t be a large problem with security, would
> there?

There would be also a problem. It is more or less the same situation
than static linking.

Bastian

-- 
Spock: The odds of surviving another attack are 13562190123 to 1, Captain.


Reply to: