darcs patch: Typo correction in String Constant (and 2 more)
Sun Aug 24 01:06:28 EDT 2008 Matthew William Cox
Hello, I hacked on cabal2arch a bit tonight. I removed the hand-coded list of core packages in favour of generating the list dynamically. I use the GHC API to parse the global package database. This entails dependencies on ghc and ghc-paths (oh well.) I've assumed that any package with all its library-dirs inside /usr/lib/ghc-6.8.3 (actually, the libdir from GHC.Paths) is a core package. This seems to exclude all the external packages, which get installed to /usr/lib/$PkgName-$PkgVer/. Anybody know if this is a bad assumption? I also sent these patches to dons. Thanks, Matt
On Sun, Aug 24, 2008 at 04:24:16AM -0400, Matthew William Cox wrote:
I've assumed that any package with all its library-dirs inside /usr/lib/ghc-6.8.3 (actually, the libdir from GHC.Paths) is a core package. This seems to exclude all the external packages, which get installed to /usr/lib/$PkgName-$PkgVer/. Anybody know if this is a bad assumption?
I went to #ghc to confirm this. It sounds safe. One possibility is for vegai to modify the GHC PKGBUILD to copy the package.conf to package.conf.shipped at packaging. Then cabal2arch could query this for an *authoritative* list of packages in bootlibs. The other possibility is to load the bootlibs into the provides array of the PKGBUILD. I don't think this would be very easily done with makepkg... hmakepkg time? :) Cheers, Matt.
matt:
Hello,
I hacked on cabal2arch a bit tonight. I removed the hand-coded list of core packages in favour of generating the list dynamically.
I use the GHC API to parse the global package database. This entails dependencies on ghc and ghc-paths (oh well.)
I've assumed that any package with all its library-dirs inside /usr/lib/ghc-6.8.3 (actually, the libdir from GHC.Paths) is a core package. This seems to exclude all the external packages, which get installed to /usr/lib/$PkgName-$PkgVer/. Anybody know if this is a bad assumption?
This looks great. It saves us the problem of redoing that list on a compile upgrade. Cheers! -- Don
matt:
Hello,
I hacked on cabal2arch a bit tonight. I removed the hand-coded list of core packages in favour of generating the list dynamically.
I use the GHC API to parse the global package database. This entails dependencies on ghc and ghc-paths (oh well.)
I've assumed that any package with all its library-dirs inside /usr/lib/ghc-6.8.3 (actually, the libdir from GHC.Paths) is a core package. This seems to exclude all the external packages, which get installed to /usr/lib/$PkgName-$PkgVer/. Anybody know if this is a bad assumption?
I also sent these patches to dons.
Resolving dependencies... cabal: dependencies conflict: ghc-6.8.2 requires Cabal ==1.2.3.0 however Cabal-1.2.3.0 was excluded because cabal2arch-0.3.8.3 requires Cabal >=1.3.11 && <1. Hmm. I can't remember what we need new Cabal for... Ah, yes, ./Main.hs:32:7: Could not find module `Distribution.Simple.PackageIndex': it is a member of package Cabal-1.4.0.1, How did you manage to build this?
matt:
On Mon, Aug 25, 2008 at 10:31:26AM -0700, Don Stewart wrote:
How did you manage to build this?
I built it on ghc-6.8.3. I've attached the setup-config from the build.
hmm. I'm not sure how to make progress on this. Do you have 6.8.2 installed via vanilla Arch?
On Mon, Aug 25, 2008 at 02:31:54PM -0700, Don Stewart wrote:
hmm. I'm not sure how to make progress on this.
My suggestion would be to sit on these patches until vegai bumps ghc to 6.8.3. These will ease that transition at least. Cheers, Matt
matt:
On Mon, Aug 25, 2008 at 02:31:54PM -0700, Don Stewart wrote:
hmm. I'm not sure how to make progress on this.
My suggestion would be to sit on these patches until vegai bumps ghc to 6.8.3. These will ease that transition at least.
Cheers, Matt
Ok. Will do. I've released the latest stable version of cabal2arch, and updated our most popular Haskell packages using it (mainly it is a bit smarter about haskell-cabal dependencies, and knows the pacman names of a few more C libraries). In other news, Arch Haskell has picked up haskell-http, but we still need to take over ScriptDevil's key packages, haskell-binary haskell-fingertree I've pinged him a few times, so might seek to just take over these packages. The next step is to prepare the Haskell platform, for ghc 6.10's release, with reference to how we'll integrate that into Arch. -- Don
matt:
Hello,
I hacked on cabal2arch a bit tonight. I removed the hand-coded list of core packages in favour of generating the list dynamically.
I use the GHC API to parse the global package database. This entails dependencies on ghc and ghc-paths (oh well.)
I've assumed that any package with all its library-dirs inside /usr/lib/ghc-6.8.3 (actually, the libdir from GHC.Paths) is a core package. This seems to exclude all the external packages, which get installed to /usr/lib/$PkgName-$PkgVer/. Anybody know if this is a bad assumption?
I also sent these patches to dons.
Resolving dependencies... cabal: dependencies conflict: ghc-6.8.2 requires Cabal ==1.2.3.0 however Cabal-1.2.3.0 was excluded because cabal2arch-0.3.8.3 requires Cabal >=1.3.11 && <1. Hmm. I can't remember what we need new Cabal for... Ah, yes, ./Main.hs:32:7: Could not find module `Distribution.Simple.PackageIndex': it is a member of package Cabal-1.4.0.1, How did you manage to build this?
participants (2)
-
Don Stewart -
Matthew William Cox