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

Re: Packaging my first haskell package



Hi Joachim,

On Wed, Dec 09, 2015 at 07:32:01PM +0100, Joachim Breitner wrote:
> 
> It looks as if you created this with an old(ish) version of cabal-
> debian...

Thinking about it - yes it was done on a Jessie system.
 
> Anyways, in general we heavily advocate against Haskell library
> packages being maintained outside the Haskell group.

The target package I was talking about is actually not a library (at
least what *I* would call library - no idea about Haskell terms here).

> The
> interdependencies between packages are intricate and it is just so much
> less hassle if you can work on them all at once. In fact, we even have
> all of them in one git repository.

I'm perfectly fine to adopt to your workflow and put things in whatever
repository.
 
> Now in your packaging, you build both the library and the binary
> packages.

This was actually cabal-debian's idea - not mine. ;-)

> Are you sure the library packages are of importance?

As far as I know there is no specific interest and I would like to
keep things as simple as possible to get the final user application.

> Often in
> such cases, only the binary package (phybin) is useful for Debian, and
> not having the library packages greatly simplifies things.

Please teach me how to simplify!! ;-)
 
> Also, the upstream package has not seen an upload in over two years.

That's not unusual for scientific packages.

> For hackage packages, this is often a sign of insufficent upstream
> activity, which can cause the package to cause problems for us, and
> eventually a swift removal. Are you sure this is an actively maintained
> project?

I'm sure that you as developers are talking about a different scale of
"active maintenance".  The package is used as is by biologists and works
successful in the target environment.  It might happen that it does not
work with the latest and greatest Haskell compilers.  I would assume
that upstream is interested in updating but I can not answer your
question reliably.  In other programming languages I was always able to
keep the code up to date with the help of other Debian people in case
upstream was not as responsive as hoped but there are also very active
and interested upstreams.  I would assume that if an upstream is
choosing a *very* unusual language for this field at will be an engaged
developer. :-)

> If so, we’d have a lot less to worry if it were part of stackage
> (which, in a way, is a Distribution-independent distribution of Haskell
> package). I suggest you contact the author and talk him into adding his
> package to stackage, which is easy and provides free QA to him:
> http://www.stackage.org/authors

I can do so but top fully understand this strategy:  Is it true that I
would then point the d/watch file to stackage.org?

> If you really need the library (and not just the binary), I suggest
> that the Debian Haskell Team takes it over.

I admit I'm perfectly fine if you take over at all and I'd be happy if I
can provide the package to our users (on a Jessie system) - no matter
who did the work. ;-)
 
> > I also need to add that I intend to backport the package to Jessie which
> > might add some extra effort since I noticed that one of the dependencies
> > (libghc-prettyclass-*) is not available in Jessie and besides this it is
> > quite tricky to install a mix of Haskell packages from Jessie and
> > Jessie-backports.
> 
> Right. That’s why we generally shun away from backporting of anything
> more than simply GHC and cabal-install.
> 
> > I'm also aware that libghc-hierarchical-clustering-* is not even
> > available in unstable but cabal-debian has added this to the
> > Build-Depends.  I guess I need to package this as well.
> 
> Right; or even better, the DHG has to. That package looks more actively
> maintained, but should also enter stackage before it gets into Debian.

OK, I'll talk to upstream. 

Thanks for your additional explanation

   Andreas.

-- 
http://fam-tille.de


Reply to: