gmane.comp.lang.haskell.general

Perhaps a bug in GHC 6.10

Hi,

One of my students wrote code like this:

prop_blank s (i,j) li =
  if i==li
    then
       do if something then True else False
    else
       prop_blank (...) (i,j) (li+1)

This code is not accepted from GHC 6.8.3 because Bool is not a monad.
In particular the error message is:

Sudoku.hs:269:5:
    Couldn't match expected type `Bool' against inferred type `a b'
    In the expression:
          ......

Surprisingly the same code is accepted from GHC 6.10. This looks like
a bug for me. The easiest way to reproduce this is to try to evaluate:

> do True

in GHCi with GHC 6.8.3 and with GHC 6.10

Is there any reason to accept this code?

Regards,
  Krasimir

(Continue reading)

headers

Eric Walkingshaw | 3 Dec 15:19

Re: Perhaps a bug in GHC 6.10

It looks like the 6.10 behavior is actually the correct one:

http://www.mail-archive.com/glasgow-haskell-bugs <at> haskell.org/msg13632.html

Also, see the do-notation translation section in the Haskell 98 Report:

http://www.haskell.org/onlinereport/exps.html#sect3.14

-Eric

On Wed, Dec 3, 2008 at 5:45 AM, Krasimir Angelov <kr.angelov <at> gmail.com> wrote:
> Hi,
>
> One of my students wrote code like this:
>
>
> prop_blank s (i,j) li =
>  if i==li
>    then
>       do if something then True else False
>    else
>       prop_blank (...) (i,j) (li+1)
>
>
> This code is not accepted from GHC 6.8.3 because Bool is not a monad.
> In particular the error message is:
>
> Sudoku.hs:269:5:
>    Couldn't match expected type `Bool' against inferred type `a b'
>    In the expression:

(Continue reading)

headers

Brandon S. Allbery KF8NH | 3 Dec 16:41

Re: Perhaps a bug in GHC 6.10

On 2008 Dec 3, at 8:45, Krasimir Angelov wrote:
> Sudoku.hs:269:5:
>    Couldn't match expected type `Bool' against inferred type `a b'
>    In the expression:
>          ......
>
> Surprisingly the same code is accepted from GHC 6.10. This looks like
> a bug for me. The easiest way to reproduce this is to try to evaluate:

6.8.3 was changed on the theory that do "ought to" require a monad,  
contra Haskell98.  I suspect Haskell98 compatibility won.

--

-- 
brandon s. allbery [solaris,freebsd,perl,pugs,haskell] allbery <at> kf8nh.com
system administrator [openafs,heimdal,too many hats] allbery <at> ece.cmu.edu
electrical and computer engineering, carnegie mellon university    KF8NH