On Sun, Apr 25, 2010 at 4:22 AM, Ganesh Sittampalam wrote: > We've just encountered a GHC 6.8 specific problem: > http://bugs.darcs.net/issue1830
This problem may be specific to 6.8.2 and below only, since it is related to GHC bug #2093. That bug was fixed in 6.8.3. It would be a *lot* easier to backport 6.8.3 to lenny than to backport 6.10 and above. > In view of this, and given there is a general desire to drop it anyway, I > think we should explicitly drop it from the 2.4 series. To put things in perspective: issue 1830 was reported by people having trouble on the community.haskell.org server, where many of the most important haskell projects are hosted. That server is currently running Debian etch, where the standard GHC is 6.6. We are now in the process of upgrading that server to lenny. There is still a possibility that we'll be able to run squeeze, but my personal opinion is that it's unlikely. In any case, the standard for hosted servers across the Internet is lenny (or the equivalent in other distros). I hope that even if we are on lenny we'll be able to backport more recent versions of GHC and darcs to run on the server. But that will take work. So it will mean a further delay in being able to get to darcs 2.4. Overall, I would say that dropping support for GHC 6.8 would mean in practice that a significant proportion of darcs users would be sticking with 2.3 and below for a long time to come. That's not necessarily so bad. It would just mean that in order to provide good ongoing support for the majority of the darcs user base, darcs developers would need to support not only older versions of GHC, but also older versions of darcs. If 6.8.3 solves the mmap problem, would it be difficult for the 2.4 series to support GHC 6.8.3? Thanks, Yitz _______________________________________________ darcs-users mailing list [email protected] http://lists.osuosl.org/mailman/listinfo/darcs-users
