cvs 4.07 arithmetic errors in Integer arithmetic
Hi there, I am trying to fix some errors in the Integer arithmetic Simon Marlow and I spotted. I just reinstalled cvs from scratch yesterday and build ghc-4.07 from it. I then noticed that one of my programs resulted in completely different output than it did with ghc-4.06 (or even 4.07 from cvs a few weeks ago). I am pretty convinced that the ghc-4.06 output is correct and the 4.07 output is **very** wrong. I am pretty sure the differences are caused by changes in Integer arithmetic (again, I have not noticed these problems with 4.07 and cvs a few weeks ago, so the changes must have been recent.) If I were to guess the errors are caused by changes to quot or gcd. Unfortunately I have very limited time to spend on this. (I already spent more time on this than I had anticipated I would need.) Is there anybody who thinks that a *recent* change made by him could have led to the problems I noticed? I would appreciate suggestions from cvs hackers as to what is a good approach to locate the cause of the problem. Finally I would like to know if there is a way to get my hands on a cvs tree for 4.06 so that I can compare the two and see what happens when I apply a fix to 4.07 against applying the same fix for 4.06? Regards, Marc van Dongen -- Marc van Dongen, CS Dept | phone: +353 21 903578 University College Cork, NUIC | Fax: +353 21 903113 College Road, Cork, Ireland | Email: [EMAIL PROTECTED]
Re: cvs 4.07 arithmetic errors in Integer arithmetic
Finally I would like to know if there is a way to get my hands on a cvs tree for 4.06 so that I can compare the two and see what happens when I apply a fix to 4.07 against applying the same fix for 4.06? cvs co -r ghc-4-06-branch will give you the 4.06 tree. cvs diff -r ghc-4-06-branch will give you the diffs between your current tree and ghc 4.06. `ghc-4-06-branch' includes any patches to the released GHC-4.06 since the initial release (eg, ghc-4-04-branch is GHC 4.04 pl1); `ghc-4-06' is the initially-released GHC 4.06. I believe in this case they're the same. HTH. --KW 8-) -- : Keith Wansbrough, MSc, BSc(Hons) (Auckland) ---: : PhD Student, Computer Laboratory, University of Cambridge, UK. : : Native of Antipodean Auckland, New Zealand: 174d47'E, 36d55'S. : : http://www.cl.cam.ac.uk/users/kw217/ mailto:[EMAIL PROTECTED] : ::
RE: cvs 4.07 arithmetic errors in Integer arithmetic
I am trying to fix some errors in the Integer arithmetic Simon Marlow and I spotted. I just reinstalled cvs from scratch yesterday and build ghc-4.07 from it. I then noticed that one of my programs resulted in completely different output than it did with ghc-4.06 (or even 4.07 from cvs a few weeks ago). I am pretty convinced that the ghc-4.06 output is correct and the 4.07 output is **very** wrong. I am pretty sure the differences are caused by changes in Integer arithmetic (again, I have not noticed these problems with 4.07 and cvs a few weeks ago, so the changes must have been recent.) If I were to guess the errors are caused by changes to quot or gcd. There haven't been any changes to PrelNum.lhs (which is where quot, gcd etc. live) since 4.06. I can't think of any changes off the top of my head that may have caused this. Could you send us an example that shows up the difference? Cheers, Simon
the impossible happend again
Thanks for the bug report. I am investigating. In the meantime you
could change the declaration of ArgrC from a 'newtype' to an ordinary
'data' declaration. It then compiles fine for me.
Hope this helps
regards
Kevin
Volker Wysk writes:
Hi.
ghc -fallow-undecidable-instances -fglasgow-exts -fno-prune-tydecls -O2 \
-c -o Arglib.o Arglib.hs
panic! (the `impossible' happened):
mkWWcpr: not a product
g{-rku-}
- (g{-rku-},
PrelMaybe.Either{-r8t,i-}
((t{-rkv-} - tzq{-rkT-}) - tzq{-rkT-})
[Arglib.GenArgError{-r2p,x-}])
Please report it as a compiler bug to [EMAIL PROTECTED]
