On Fri, Aug 29, 2014 at 10:46 AM, Joachim Breitner <[email protected]>
wrote:

* Allow building of documentation when when package has no Haskell modules
> ✓
> * Fix a bash bug in the dev library install recipe
> ✓
> * Remove cases for binary debs in the ghc package - haskell-devscripts is
> not a build dependency of ghc
> While the package is not a build-dependency (which would induce too many
> build dependency cycles), we _do_ use ./dh_haskell_provides in GHC,
> after copying it there. So this code needs to stay.
>

Ok


> * Remove support for obsolete doc package prefix "haskell-"
> ✓
> * Pass --package-db to the cabal configure command
> ✓, but why?
>

This has so far been necessary for building ghcjs libraries.


> * Pass --with-haddock and --with-ghc to cabal haddock
> again ✓, but why? What does this change?
>

This is also required by ghcjs - it has its own ghcjs-pkg and haddock-ghcjs
executables.


> * Add functions to Dh_Haskell.sh to parse library package names, compute
>   compiler names and compiler dependent paths
> You use “ghc -e” which requires GHCi which is not available on all
> architectures, so this is not good.
> The proper way to do it is to parse the output of "ghc --info".
> I’m pulling it for now.
>

I will try to make this change.


> * Move the make recipes from hlibrary.mk to Dh_Haskell.sh
> Nice cleanup
> * Add a postinst script to the ghcjs dev library to run recache
> Shouldn’t this be handled by a dpkg trigger in your ghcjs package?
> Not pulling.
>

Yes, I would appreciate some guidance on how to do this.


> It looks like not pulling
>  * Remove cases for binary debs in the ghc package - haskell-devscripts
>    is not a build dependency of ghc
> prevents me from pulling these patches:

 * Large patch to parameterize the name of the haskell compiler in order
>    to support ghcjs
>  * Add duplicates of the libghc rules modified to build libghcjs
>    packages
>  * Remove set -x directives in hlibrary.mk
>  * Add improved debugging code (disabled)
>  * Supply default compiler to packages_hc call in dh_haskell_blurbs.
> I’m afraid that this means that the patches I did pull left me with
> something broken.
>
> Not sure how we should proceed from here. Do you want to integrate my
> review until I’m satisfied with the overall result, which I then can
> pull in one go?
>

Yes, I will create revised patches.  Is it too late for you to unpull?


>
>
> I see that os() and cpu() are only used when building ghcjs packages.
> But still, reading that data from ghc --info is saner. Or maybe even
> from dpkg-architecture.
>
>
> Greetings,
> Joachim
>
>

Reply via email to