[Agda] Agda with GHC-6.10.4, continued...
kahl at cas.mcmaster.ca
kahl at cas.mcmaster.ca
Fri Apr 9 22:58:03 CEST 2010
On 2010-04-03, Nils Anders Danielsson wrote:
> On 2010-04-02 12:09, James Chapman wrote:
> >> Darin introduced a dependency on recent versions of Cabal in his work on
> >> Cabal support for Agda.
> >
> > Is it possible to turn this off temporarily?
>
> You could try the following command:
>
> darcs obliterate --match='author dwm at cs.nott.ac.uk'
I succeeded to get Agda installed with GHC-6.10.4 after that,
but now pattern matching on pairs does not work anymore
(it is used also in the standard library):
/export/home/kahl/doc/Lang/Agda/lib/src/Data/List.agda:150,7-19
Not implemented: Pattern matching for records
when checking that the pattern just (x , s') has type
The missing patches are:
$ darcs pull --dry-run
Would pull from "http://code.haskell.org/Agda"...
Would pull the following changes:
Wed Feb 3 11:30:05 EST 2010 dwm at cs.nott.ac.uk
* initial commit for cabal stuff
Wed Feb 3 12:20:50 EST 2010 Nils Anders Danielsson
<nils.anders.danielsson at gmail.com>
* Removed a hard-coded version number.
Wed Feb 3 12:27:38 EST 2010 Nils Anders Danielsson
<nils.anders.danielsson at gmail.com>
* Made the build dependency constraints stricter.
Wed Feb 3 12:28:34 EST 2010 Nils Anders Danielsson
<nils.anders.danielsson at gmail.com>
* Switched to warning flags which are accepted by cabal check.
Wed Feb 3 12:34:02 EST 2010 Nils Anders Danielsson
<nils.anders.danielsson at gmail.com>
* Various changes and fixes.
Fri Feb 5 08:27:23 EST 2010 dwm at cs.nott.ac.uk
* cleanup
Mon Feb 8 08:57:26 EST 2010 dwm at cs.nott.ac.uk
* Fix package DB reading to work with new DB format (Cabal >= 1.8.0.2 &&
GHC >= 6.12)
Mon Feb 8 09:10:41 EST 2010 dwm at cs.nott.ac.uk
* Basic error handling
Mon Feb 8 12:43:31 EST 2010 dwm at cs.nott.ac.uk
* Ported some parts other parts of the package handling code to the new
format. Incomplete and untested.
Making no changes: this is a dry run.
It is not obvious where the problem might be hiding,
and quite surprising that it depends on the ``initial commit for cabal
stuff'' --- is there an easy way to get pattern matching for pairs back?
Wolfram
More information about the Agda
mailing list