On So, 2014-07-13 at 09:48 +0200, Sebastian Dröge wrote: > On Fr, 2014-07-11 at 13:47 +0200, Joachim Breitner wrote: > > Dear Sebastian, > > > > Am Freitag, den 11.07.2014, 13:37 +0200 schrieb Sebastian Dröge: > > > On Do, 2014-07-10 at 09:35 +0000, Joachim Breitner wrote: > > > > > > > Changes: > > > > ghc (7.8.20140710-1) experimental; urgency=medium > > > > . > > > > * New upstream release (7.8.3) > > > > The version number is a work-around for a glitch earlier, see below. > > > > > > do you plan to get 7.8.3 together with the (hopefully soon to be > > > released) new Haskell Platform into unstable so it can be included in > > > the next Debian release? > > > > > > Is there anything you could use some help with to make this happen > > > faster? > > > > > > > > > I'm mostly interested in the new IO scheduler in ghc 7.8, and it would > > > be quite sad if that couldn't be included in the next Debian release. > > > > I’m doubtful about putting 7.8 into unstable. > > * The change to dynamic linking might cause problems that we regret > > later. I might be a bit over-cautious here, but I’d rather see that > > well tested by others :-) > > * We’d lose Haskell on arm* and mips*, unless someone steps up to > > investigate these build failures: > > https://buildd.debian.org/status/package.php?p=ghc&suite=experimental > > This is probably where you can help the most. > > * I’m sure many of our packages do not build with 7.8 yet, and not all > > of them have active upstreams. This would imply that once we put 7.8 > > into unstable, we might not migrate _anything_ to testing until _all_ > > these issues are fixed. Given that no new migrations are allowed > > from September 05 on, according to the release team, this might be > > too tight. > > (Ok, I just notice that it says “no new”, not “all have to be done > > by”, so we might be ok) > > > > I guess if one of these issues would be resolve (e.g. #2, hint, hint) I > > could be convinced to move 7.8 to unstable. > > Hi, > > I'll see if I can do anything useful for the ARM/MIPS build failures but > don't expect too much there from me :) > I would've assumed that ARM should be no problem as apparently quite a > few people are using GHC to target Android and iOS... > I could probably be more useful with #3 :) > > For #1, AFAIK Fedora is planning to switch quite soon so maybe that can > help to get some useful numbers there. Actually for the ARM problem a short look in the bugtracker gave this: https://ghc.haskell.org/trac/ghc/ticket/8976 From reading the linked article it doesn't seem that complicated to get it work, most of the work probably being to find a fast ARM machine that can be used for building (I don't want to try this on my RPI!). For MIPS there's a patch in the BTS which would disable dynamic support, so that doesn't seem like the best idea: https://bugs.debian.org/751479 So that needs some further investigation, but maybe some combination of using ld.gold and sprinkling some -fPIC around is already enough.
Attachment:
signature.asc
Description: This is a digitally signed message part