[darcs-users] darcs patch: darcs prefs: remove type witness stuff, breaks compiles

Jason Dagit dagit at codersbase.com
Wed Apr 30 03:50:40 UTC 2008


On Tue, Apr 29, 2008 at 8:34 PM, <gwern0 at gmail.com> wrote:

> Tue Apr 29 23:19:21 EDT 2008  gwern0 at gmail.com
>  * darcs prefs: remove type witness stuff, breaks compiles
>  It's worth noting that the test as it stood ran on every patch record,
> and allows patches to be recorded only if one can compile with type
> witnesses.
>  It's also worth noting that 6.8.x can't compile everything with type
> witnesses, and lispy says 6.6.1 can't either. So, in general, turning on
> type witnesses in this particular place strikes me as a very very bad thing
> to do.
>  If Dr. Roundy wants to automatically test compilation with type
> witnesses, the thing to do is to have the configure script turn it on if an
> example compiles, say. Forcing it on just leads to breakage and considerable
> annoyance for people who have to rm _darcs/prefs/prefs just to record some
> patches.


Gwern, I think it would be better if we checked (in configure) for the GHC
version and disable type witnesses on 6.8.  This has the draw back that
people changing type witnesses code on 6.8 have to be really careful (or
wait for the buildbots to try to build their code).  All things considered,
I think it's probably smart idea to do your 'darcs test' on ghc-6.6 at the
moment.

Ganesh has mentioned that fixing the type witness code on 6.8 should really
just boil down to adding type signatures in more places but I can't tell
where from glancing at the error messages.  Fixing the type witnesses for
6.8 and onward is something I plan to do, but it's lower priority than
writing my thesis and getting type witnesses to work fully on 6.6.

Jason
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.osuosl.org/pipermail/darcs-users/attachments/20080429/08c48c98/attachment.htm 


More information about the darcs-users mailing list