
Hi all, pandoc in the [community] repository is by now quite outdated. While there is a haskell-pandoc package in AUR, it cannot be used by people who do not want to install the whole gamut of GHC and the haskell toolchain. I've prepared a PKGBUILD for pandoc 1.1 which depends only on gmp, but makedepends on ghc, haskell-utf8-string and haskell-zip-archive. Is it OK if I move these to [community]? Cheers, Abhishek

abhidg:
Hi all,
pandoc in the [community] repository is by now quite outdated. While there is a haskell-pandoc package in AUR, it cannot be used by people who do not want to install the whole gamut of GHC and the haskell toolchain.
That also means the package will be installed broken -- only the binary part of it will work. The library components will be useless. http://aur.archlinux.org/packages.php?ID=19804 it is used by a number of other things too, as you see.
I've prepared a PKGBUILD for pandoc 1.1 which depends only on gmp, but makedepends on ghc, haskell-utf8-string and haskell-zip-archive. Is it OK if I move these to [community]?
Could you attach the PKGBUILD? I'm fairly uncomfortable repackaging pandoc to intentionally remove dependencies that other tools will need. What happens when we move gitit into [community] and it also needs the (full) pandoc? Maybe we can have the pandoc author separate the library and the executable? -- Don

The PKGBUILD I've now worked with ghc 6.8 but is now broken with ghc 6.10... I'll try fixing it. I think the best solution to this problem is to wait for makepkg splitting support to be integrated into pacman. Then we could have two packages built from the same PKGBUILD: * pandoc: containing the binaries and man pages. * pandoc-dev: containing the libraries. Then gitit and other stuff which need the libraries could depend on pandoc-dev. -- Abhishek

abhidg:
The PKGBUILD I've now worked with ghc 6.8 but is now broken with ghc 6.10... I'll try fixing it.
I think the best solution to this problem is to wait for makepkg splitting support to be integrated into pacman. Then we could have two packages built from the same PKGBUILD:
* pandoc: containing the binaries and man pages. * pandoc-dev: containing the libraries.
Then gitit and other stuff which need the libraries could depend on pandoc-dev.
Knowing John, you could probably just ask the developer to make this split upstream. -- Don

2009/2/15 Don Stewart
Knowing John, you could probably just ask the developer to make this split upstream.
OK, I've opened an issue in the upstream bugtracker: http://code.google.com/p/pandoc/issues/detail?id=126 -- Abhishek

abhidg:
2009/2/15 Don Stewart
: Knowing John, you could probably just ask the developer to make this split upstream.
OK, I've opened an issue in the upstream bugtracker: http://code.google.com/p/pandoc/issues/detail?id=126
Excellent!

2009/2/15 Don Stewart
abhidg:
2009/2/15 Don Stewart
: Knowing John, you could probably just ask the developer to make this split upstream.
OK, I've opened an issue in the upstream bugtracker: http://code.google.com/p/pandoc/issues/detail?id=126
Excellent!
The split can be easily done using cabal or using runhaskell, from John's comment on issue 126 in pandoc's bugtracker. I've prepared two packages: haskell-pandoc and pandoc. haskell-pandoc depends on the ghc toolchain and the other haskell libraries; it also depends on pandoc. pandoc contains only the binary 'pandoc', some helper scripts and the manual pages. Thus the pandoc package only needs to depend on gmp and zlib. Doing `pactree -u haskell-pandoc` (pactree is in pacman-contrib) gives the dependency chain of haskell-pandoc as: ghc gmp gcc-libs glibc kernel-headers tzdata texinfo ncurses findutils gzip bash readline libedit haskell-utf8-string haskell-zip-archive * haskell-binary * haskell-digest * zlib haskell-zlib pandoc Of these the * ones are not yet in [community], but they need to be if pandoc is to be updated to 1.1. -- Abhishek

abhidg:
2009/2/15 Don Stewart
: abhidg:
2009/2/15 Don Stewart
: Knowing John, you could probably just ask the developer to make this split upstream.
OK, I've opened an issue in the upstream bugtracker: http://code.google.com/p/pandoc/issues/detail?id=126
Excellent!
The split can be easily done using cabal or using runhaskell, from John's comment on issue 126 in pandoc's bugtracker.
I've prepared two packages: haskell-pandoc and pandoc. haskell-pandoc depends on the ghc toolchain and the other haskell libraries; it also depends on pandoc. pandoc contains only the binary 'pandoc', some helper scripts and the manual pages. Thus the pandoc package only needs to depend on gmp and zlib.
Doing `pactree -u haskell-pandoc` (pactree is in pacman-contrib) gives the dependency chain of haskell-pandoc as:
ghc gmp gcc-libs glibc kernel-headers tzdata texinfo ncurses findutils gzip bash readline libedit haskell-utf8-string haskell-zip-archive * haskell-binary * haskell-digest * zlib haskell-zlib pandoc
Of these the * ones are not yet in [community], but they need to be if pandoc is to be updated to 1.1.
And my general view is that if a Haskell app needs some Haskell libs, we should move them in.
participants (2)
-
Abhishek Dasgupta
-
Don Stewart