[Pkg-haskell-maintainers] Bug#701059: Bug#701059: haskell-conduit: FTBFS: test failure form doctest adopted

[Colin Watson]

On Thu, Feb 21, 2013 at 12:36:28PM +0100, Joachim Breitner wrote:
> Am Donnerstag, den 21.02.2013, 13:10 +0900 schrieb Debian ppc64 build
> daemon:
> > Build test of haskell-conduit fails form ver. 0.5.5 which adopted doctest.
> > http://buildd.debian-ports.org/status/logs.php?pkg=haskell-conduit&arch=ppc64
> > 
> > Unfortunately, because transitions of ghc is not applied to experimental repository,
> > this error is not checked at present except on ppc64 port.
> > https://buildd.debian.org/status/package.php?p=haskell-conduit&suite=experimental
> > http://buildd.debian-ports.org/status/package.php?p=haskell-conduit&suite=experimental
> > 
> > So, since it may be a bug peculiar to doctest on ppc64 port, this is reported by "Severity:  important". 
> 
> doctest requires GHCi, which is not supported on all architectures. The
> cleanest fix would be if doctest would detect this situation and fail
> gracefully: https://github.com/sol/doctest-haskell/issues/46

I don't think this fix will cover armhf, where GHCi is present:

  $ ghc --info | grep 'Have interpreter'
   ,("Have interpreter","YES")

... but broken:

  http://hackage.haskell.org/trac/ghc/ticket/7794

I don't know what would be best here, as I haven't managed to nail down
a proper fix yet; would it be worth explicitly excluding armhf in
doctest for now, or would it be better to disable GHCi on armhf until
it's fixed properly (which will probably require some set of packages to
be removed on armhf), or something else?

-- 
Colin Watson                                       [cjwatson at ubuntu.com]