Re: [pkg-haskell-tools] 01/01: Fixed bug that caused builds to fail under certain conditions outlined below. Let's consider three haskell source packages: <A>, <B> and <C>. <A> build-depends on a library package built by <B>, called <B.a>. <B.a> (runtime-)depends on an executable package built by C, called <C.b>. Neither A nor B build-depend on packages built by C. When <A> is built, it is possible that <C> is not built yet, thus <B.a> fails to install, failing the build of <A>.

[Sven Bartscher <sven.bartscher@weltraumschlangen.de>]

On Fri, 21 Aug 2015 09:03:14 +0200
Joachim Breitner <nomeata@debian.org> wrote:

> Hi,
> 
> Am Donnerstag, den 20.08.2015, 20:59 +0000 schrieb Sven Bartscher:
> > commit f2bcce12244b66e93410574545cdbad3ebe810af
> > Author: Sven Bartscher <sven.bartscher@weltraumschlangen.de>
> > Date:   Thu Aug 20 22:47:13 2015 +0200
> > 
> >     Fixed bug that caused builds to fail under certain conditions
> > outlined below.
> >     Let's consider three haskell source packages: ,  and .
> >      build-depends on a library package built by , called .
> >      (runtime-)depends on an executable package built by C, called .
> >     Neither A nor B build-depend on packages built by C.
> >     When  is built, it is possible that  is not built yet, thus
> >      fails to install, failing the build of .
> 
> quite right that this is a shortcoming of the system. I tried to make
> it take binary dependencies into account, but then I noticed that I
> would have to handle Provided/Depends, and reverted. I did not think
> of this particular problem with non-library dependencies, so the
> variant that only works with non-libraries is still useful, thanks!

I made the change on branch, in case you want to review the changes
first (although it has become a one-commit branch).
Do you want to review the changes before they're merged or am I free to
go?

> I take it that this was the cause for Alexandre’s hspec problem?

Yes it was.

Regards
Sven