Hi Sam,

It looks quite good, so as long as CI-build passes I'd gladly incorporate
your PR.
You'll need to update test/tests.json though, otherwise make cross would
fail.
Also as we're currently moving to CMake, you may want to apply the same
changes to CMake build if you happen to have extra cycles for that.

On Thu, Dec 17, 2015 at 11:26 PM Sam Roberts <sam.roberts.1...@gmail.com>
wrote:

> Hi all,
>
> I am working with the latest thrift code from a Haskell program. I have
> made a few changes locally to the Haskell build scripts in order to install
> the Haskell lib and it's dependencies into a sandbox. I would like to run
> these changes past you, and either contribute them back or learn a better
> way of installing the thrift Haskell lib into a sandbox.
>
> My changes can be found here
> <https://github.com/SamRoberts/thrift/tree/enable_cabal_global_flags>, but
> I'll describe what happened below:
>
>    1. I tried to use CABAL_CONFIGURE_FLAGS to use an external sandbox
>    config file, but CABAL_CONFIGURE_FLAGS didn't seem to have any effect. I
>    suspect it has done nothing ever since this commit
>    <
> https://github.com/apache/thrift/commit/c0a7d72c83d93cf266d57b25a3b8850eb81c47d6
> >
>    (released in version 0.9.2).
>    2. As the configure step was deliberately removed from the build, I
>    removed CABAL_CONFIGURE_FLAGS and added CABAL_GLOBAL_FLAGS, a variable
> for
>    cabal global flags that get added to every cabal call in the Haskell
> build.
>    This lets the user specify a sandbox, or to add --no-require-sandbox if
>    necessary.
>    3. This means using cabal exec for all raw ghc calls as well, so that
>    calls to ghc have the sandbox packages available to them.
>
> (I also removed the cabal update command from the build as I didn't really
> approve of a build system running that for me, but that's not related to
> the main changes and I can easily add back the update command if desired.)
>
> Would you be happy for to do a PR with these changes? Or would you prefer
> that I take a different approach?
>
> Many thanks,
> Sam
>

Reply via email to