- Glasgow-haskell-users

Deforestation of literal lists
by Carsten Schultz 17 Sep '04

17 Sep '04

Hi! Just a thought... Would it be a sensible thing to not desugar [x, y, z] into (:) x ((:) y ((:) z [])), but into build (\ c n -> c x (c y (c z n)))? Alternatively, how about the following rules? {-# RULES "singleton" [~1] forall x . (:) x [] = build (\c n -> c x n) #-} {-# RULES "cons/build" [~1] forall (x::a) (f:: forall b . (a->b->b) -> b -> b) . (:) x (build f) = build (\c n -> c x (f c n)) #-} This should be useful for example in conjuction with sequence_ or // applied to short lists. Has this been discussed before? Greetings, Carsten -- Carsten Schultz (2:38, 33:47), FB Mathematik, FU Berlin http://carsten.codimi.de/ PGP/GPG key on the pgp.net key servers, fingerprint on my home page.

1 0

RE: problem building for cygwin
by Simon Marlow 17 Sep '04

17 Sep '04

On 17 September 2004 14:31, Simon Marlow wrote: > On 16 September 2004 13:04, Simon Marlow wrote: > >> It looks like MingW comes with a libreadline.a these days. Readline >> appears to build fine in GHC 6.2.2. I'll see if we can include it >> with the 6.2.2 release... > > So I compiled up 6.2.2 on Windows with readline support (various > patches to build system & configure script were required). It > doesn't work though - typing control key combinations just results in > odd characters. It doesn't work in an ordinary cmd.exe window either. > Anyone have any ideas? Aha - I forgot to fix something in InteractiveUI. It works now, albeit only in an ordinary command shell - if you try to use it from a cygwin terminal then it crashes when you press enter. I think we'll stick with plain non-readline for Windows for the time being. Cheers, Simon

1 0

RE: problem building for cygwin
by Simon Marlow 17 Sep '04

17 Sep '04

On 16 September 2004 13:04, Simon Marlow wrote: > It looks like MingW comes with a libreadline.a these days. Readline > appears to build fine in GHC 6.2.2. I'll see if we can include it > with the 6.2.2 release... So I compiled up 6.2.2 on Windows with readline support (various patches to build system & configure script were required). It doesn't work though - typing control key combinations just results in odd characters. It doesn't work in an ordinary cmd.exe window either. Anyone have any ideas? Cheers, Simon

1 0

RE: problem building for cygwin
by Simon Marlow 16 Sep '04

16 Sep '04

On 16 September 2004 12:11, Beni Kavanagh wrote: >>> Has anyone else ran into this problem. Am I making a mistake? >> >> In the terminology of the build system, you're building with >> host=i386-unknown-mingw32 and target=i386-unknown-cygwin32. Try >> configuring with those options. >> >> Sounds like the compiler's Makefile should be checking the host >> rather than the target when adding -package unix too. >> >> Cheers, >> Simon > > > (Again, I am building from 6.2.1). > > If I try to configure cygwin with host = "i386-unknown-mingw32", > target = "i386-unknown-cygwin32" it complains that cross compilation > is not supported. Did you mean I should go through the cross > compilation process? Well it *is* a cross-compilation, of a kind. It is true that the build system isn't really set up to handle cross-compilation. I'm afraid this is going to be a bit of a non-answer. Building for cygwin is something that isn't officially supported, and it almost certainly requires tweaking the build system and Makefiles to get it to go through. I can't tell you how much, because I've never done it. > Is there an unofficial binary release for ghc-6.2.1 on cygwin? Not that I know of. > All I really need is Readline which is not built into the mingw binary > release. Initially I tried to get the mingw build to work. This > failed. The script 'configure' hangs in the step "checking if your > perl works in shell scripts". > > Has anyone had this problem? Nope, never seen this one. Perhaps try a different Perl? The one shipped with GHC seems to work well. It looks like MingW comes with a libreadline.a these days. Readline appears to build fine in GHC 6.2.2. I'll see if we can include it with the 6.2.2 release... Cheers, Simon

1 0

RE: problem building for cygwin
by Simon Marlow 16 Sep '04

16 Sep '04

On 16 September 2004 01:57, Beni Kavanagh wrote: > Hi I'm trying to build for cygwin. > >> From a clean windows box I > > Installed cygwin to c:/cygwin > Added c:/cygwin/bin to path > Installed MSys and MinGW in c:/msys and c:/mingw (did not add path) > Installed ghc to c:/ghc/ghc-6.2.1 > Untarred source to c:/ghc/ghc-6.2.1 (same dir) > Copied perl from ghc dist to /cygwin/bin/perl > Downloaded and build happy and alex (not that this is necessary > but built just fine) > Ran ./configure --with-ghc=c:/ghc/ghc-6.2.1/bin/ghc --with- > gcc=c:/cygwin/bin/gcc > > make > > failed while trying to build ghc/compiler. The build rules give the > argument -package unix but the binary ghc release from > www.haskell.org is a mingw version and so doesn't have package unix. > Of course inorder to build the unix library I need > ghc/compiler/ghc-inplace and I'm now skuppered. > > Has anyone else ran into this problem. Am I making a mistake? In the terminology of the build system, you're building with host=i386-unknown-mingw32 and target=i386-unknown-cygwin32. Try configuring with those options. Sounds like the compiler's Makefile should be checking the host rather than the target when adding -package unix too. Cheers, Simon

2 1

RE: Release candidate for 6.2.2 availabe
by Simon Marlow 16 Sep '04

16 Sep '04

On 16 September 2004 06:12, Adrian Hey wrote: > I thought we were going to get a new FFI function soon.. > http://www.haskell.org/pipermail/ffi/2004-March/001740.html > > AFAICS it isn't in the libs for this release :-( > Has this got lost or forgotten about? We don't make any changes or additions to library APIs in a patchlevel release. Bug fixes only. 6.4 will be out "soon", with loads of new stuff (and new bugs too, no doubt!). Cheers, Simon

1 0

problem building for cygwin
by Beni Kavanagh 16 Sep '04

16 Sep '04

Hi I'm trying to build for cygwin. >From a clean windows box I Installed cygwin to c:/cygwin Added c:/cygwin/bin to path Installed MSys and MinGW in c:/msys and c:/mingw (did not add path) Installed ghc to c:/ghc/ghc-6.2.1 Untarred source to c:/ghc/ghc-6.2.1 (same dir) Copied perl from ghc dist to /cygwin/bin/perl Downloaded and build happy and alex (not that this is necessary but built just fine) Ran ./configure --with-ghc=c:/ghc/ghc-6.2.1/bin/ghc --with- gcc=c:/cygwin/bin/gcc make failed while trying to build ghc/compiler. The build rules give the argument -package unix but the binary ghc release from www.haskell.org is a mingw version and so doesn't have package unix. Of course inorder to build the unix library I need ghc/compiler/ghc-inplace and I'm now skuppered. Has anyone else ran into this problem. Am I making a mistake? Cheers Ben

1 0

RE: URI Typeable & Data
by Niklas Broberg 15 Sep '04

15 Sep '04

>Could we possibly have derived instances of Typeable and Data for >Network.URI.URI in the 6.3 CVS please? I second this request, and also ask for an instance of Typeable for Control.Concurrent.MVar (and the other Control.Concurrent types as well). /Niklas _________________________________________________________________ Add photos to your e-mail with MSN 8. Get 2 months FREE*. http://join.msn.com/?page=features/featuredemail

1 0

Re: [Haskell] arrows
by ross＠soi.city.ac.uk 15 Sep '04

15 Sep '04

On Wed, Sep 15, 2004 at 03:38:43PM +0200, Sebastian Boldt wrote: > I am a bit confused about command stacks. I tried to run the following > code in ghc: > > mshowA :: (Arrow a, Show b) => a (b, String) String > mshowA = proc (x,s) -> returnA -< s ++ show x ++ s > > f :: Arrow a => a Int String > f = proc x -> do a <- (mshowA -< x) "***" > returnA -< a > > To my understanding (this doesn't mean anything :-) this is correct > because (mshowA -< x) "***" should be equivalent to (mshowA -< (x,"***")), > but somehow ghc is rejecting the code with the error message "Non-empty > command stack at command: mshowA -< x". Yes, the error message isn't terribly friendly. In the GHC implementation, the typechecker forbids application commands (-< and -<<) from being applied to anything. As you point out, this seems an unnecessary restriction, so I've removed it from the version that will become GHC 6.4. (A similar restriction on do commands is still there.)

1 0

RE: linker errors
by Frank-Andre Riess 15 Sep '04

15 Sep '04

> The compiler has generated intermediate code which uses a 77-tuple, but > unfortunately the libraries don't contain code for 77-tuples (we only go > up to 62-tuples). > > This is kind of a known bug, occasionally tickled by Happy-generated > parsers. > > If you're not generating your parser with 'happy -agc', then that should > help. > > Cheers, > Simon Thank you a lot. As simple as it is, it really helped. Thanks to Volker and Johannes, too. Greetings, Frank-Andre

1 0