On Tue, 12 Jul 2016 16:10:42 +0200 Joachim Breitner <nomeata@debian.org> wrote: > Hi, > > Am Dienstag, den 12.07.2016, 13:46 +0200 schrieb Sven Bartscher: > > On Tue, 12 Jul 2016 11:31:22 +0200 > > Joachim Breitner <nomeata@debian.org> wrote: > > > > > Hi, > > > > > > I lost a bit of oversight. Where are we right now? > > > > > > Has all of LTS-6 migrated to testing? > > > > I still we are still waiting for a testing migration, which is blocked > > by reverse dependencies of openglraw that aren't removed yet. I started > > filing bugs about them, but problems in my mailer prevented me from > > filing all of them. > > according to #debian-haskell, testing migration was just done, after > three removal hints by pichu: > > <pochu> # haskell transition > <pochu> remove haskell-gloss-rendering/1.10.1.1-1 haskell-gloss/1.10.1.1-1 gitit/0.12.1-2 > <pochu> and things migrated to testing Yay! > > > Should we update to LTS-6.7, to get some newer packages in? > > > > > > Is it worth starting to stage GHC-6 and LTS-7 (based on nightly) on a > > > branch in the repository? > > > > [snip] > > > I still don't have comprehensive test about my changes to > > haskell-devscripts and dh-haskell still needs to be updated to handle > > the new ABIs correctly. > > Ok > > > Another problem is that updating GHC-8 with the new haskell-devscripts > > causes a one-time bootstrapping problem, as GHC-8 needs the > > haskell-devscripts-nocolour >= 0.10.3 to handle the ABIs correctly, but > > the new version of haskell-devscripts needs GHC-8 as earlier versions > > of ghc-pkg aren't able to retrieve the abi field, like > > haskell-devscripts requires it. > > Hmm. Any chance to make haskell-devscripts fall back to parsing the > package id, to avoid such problems. Sounds reasonable. > > We can prevent this by shipping the next version of GHC with it's own > > version of dh_haskell_provides again and making it depend on > > haskell-devscripts-nocolour once we have GHC-8 in unstable. That > > shouldn't be hard, but still needs to be done. > > Might work, but then we’d have to do that as well for backports? And so > on? Probably, so your proposed solution above is probably better. > Oh, that reminds me of another reason why haskell-devscripts depends on > ghc (and not the other way around): > > When upload, say, ghc-8 to experimental, we want to upload all Haskell > packages to experimental. To ensure that they are built with ghc, what > we did previously, was to bump the dependency of haskell-devscripts on > ghc to ghc (>= 8.0), and then the individual packages only get their > haskell-devscripts build-dependency bumped. > > Nevermind, we can still do the same in haskell-devscripts, as before. I don't see the problem here. Even if we didn't split haskell-devscripts, it would still have its dependency on GHC, which we would bump. > But can we find a better name than -nocolour? There might be other > packages that we might want haskell-devscripts to depend on (not fully > realistic, but an example: hoogle). Maybe haskell-devscripts-files? > haskell-devscripts-stage1? haskell-devscripts-common? haskell- > devscripts-bootstrap? - haskell-devscripts-files: Sounds reasonable - haskell-devscripts-stage1: Feels a bit strange to me, even though I can't tell why - haskell-devscripts-common: Seems out of place, as the -common suffix is AFAIK usually used for other situtations. - haskell-devscripts-bootstrap: Seems alright. I think haskell-devscripts-minimal might also fit. Regards Sven
Attachment:
pgpswOrKUxj4K.pgp
Description: Digitale Signatur von OpenPGP