[arch-haskell] cabal2arch and clutter

Don Stewart dons at galois.com
Sat Oct 12 07:35:44 UTC 2013

> Hi,
> Would it be possible to add an option to cabal2arch to prevent it from
> creating files outside of the directory which contains the PKGBUILD and
> install file (i.e. "link", "title", "desc", the source tarball and  the
> cabal file)? Tangentially, is all of the necessary metainformation
> contained in the cabal file, or does cabal2arch require further
> information contained in the source tarball? I found it strange that it
> downloaded the latter. Presuming that the cabal file contains all the
> necessary information, I would really prefer it if cabal2arch could
> simply read the file remotely without saving it and then generate the
> directory with the PKGBUILD and install file, and nothing else.

Well, I really shouldn't have had those files be created -- they're a
quick hack to log a bit of metainfo for other tools to process (to
generate the upload logs).
> I ask this because I'm considering adding cabal support to bauerbill,
> which is a full yaourt replacement with additional features such as
> integrated support for building repo packages froms source. I am hoping
> to use cabal2arch to do this but I do now want to deal with the clutter
> that it creates as all I require are the PKGBUILD and install file. I
> do not want to implement my own "cabal2arch" though because it seems
> silly to duplicate the effort and then have to monitor changes in
> cabal2arch for compatibility.

Could you just clean up the files?
> Additionally, please consider changing the cabal2arch help message as
> it does not describe what cabal2arch actually does:

Good idea.
> Finally, the dependencies array does not specify minimum versions
> as specified in the cabal file, e.g. "Cabal-Version >= 1.2" does not
> translate to "haskell-cabal>=1.2". I realize that this example is
> trivial as haskell-cabal is already at a later version, but should this
> not be remedied to prevent possible issues in the future?
> I'm sorry that I haven't supplied any patches, but I'm still only just
> beginning to learn haskell and I doubt that any patches that I submit
> at this point would save more effort than their review would require.

In what sense would you support cabal in your tool? Just transparently
calling cabal2arch?

-- Don

More information about the arch-haskell mailing list