I think it'd be worthwhile to provide packages with haddock-generated documentation for all haskell packages. However, I'm not convinced that the docs should be packaged with the binary components of the package. I've been experimenting with "split packages" (in pacman as of version 3.3) as a means of building both packages from the same PKGBUILD. I've used dataenc for my experiments and have arrived at something that makes some sense, but I'd like comments and feedback before I start coding it into cabal2arch. I've attached the modified PKGBUILD and the added .install file for the -doc package. I believe the same approach could be taken to providing packages with profiling support as well. /M -- Magnus Therning (OpenPGP: 0xAB4DFBA4) magnus＠therning．org Jabber: magnus＠therning．org http://therning.org/magnus identi.ca|twitter: magthe
Resending this because I forgot to cc the list when I responded to Magnus. On Thu, Aug 06, 2009 at 12:07:35PM +0100, Magnus Therning wrote:
I think it'd be worthwhile to provide packages with haddock-generated documentation for all haskell packages. However, I'm not convinced that the docs should be packaged with the binary components of the package.
Generally, Arch users expect docs to come with the package as provided by upstream, and not as some separate package. Historically Arch stripped docs out of the packages, but that policy was reversed over a year ago. See the hue and cry over here: http://www.archlinux.org/pipermail/arch-dev-public/2008-April/005574.html Cheers, Matt
matt:
Resending this because I forgot to cc the list when I responded to Magnus.
On Thu, Aug 06, 2009 at 12:07:35PM +0100, Magnus Therning wrote:
I think it'd be worthwhile to provide packages with haddock-generated documentation for all haskell packages. However, I'm not convinced that the docs should be packaged with the binary components of the package.
Generally, Arch users expect docs to come with the package as provided by upstream, and not as some separate package.
Historically Arch stripped docs out of the packages, but that policy was reversed over a year ago. See the hue and cry over here: http://www.archlinux.org/pipermail/arch-dev-public/2008-April/005574.html
We also need to specify the haddock version, have that on every person's machine (easier now it is part of the Haskell Platform) and be robust in the tolerance of doc generation failure. -- Don
On Thu, Aug 06, 2009 at 01:20:31PM -0700, Don Stewart wrote:
We also need to specify the haddock version, have that on every person's machine (easier now it is part of the Haskell Platform) and be robust in the tolerance of doc generation failure.
We could define a haskell-base package group (this may not necessarily be the entire platform) that all haskell packages have an assumed build dependency on. This is how other PKGBUILDs work: anything in base-devel is assumed to be installed and up to date on the machine you're building packages on. Candidates for haskell-base would be GHC, Haddock (if we stop using the internal GHC one), Alex, Happy, and C2Hs. That being said, I don't really have too much personal stake in the AUR packages, since I primarily use cabal-install these days. Take my thoughts with as much care as they deserve. Matt
participants (3)
-
Don Stewart -
Magnus Therning -
Matthew William Cox