- Glasgow-haskell-users

RE: ghc --make feature request
by Simon Peyton-Jones 30 Oct '01

30 Oct '01

| Note that in GHC, the "version number" of a function can | often change for hard-to-spot reasons. You just need to | change (for example) the strictness properties of the | function, which can be very easy to do when making changes to | your code. The compiler cares about (much) more than just | the types of imported objects. I tend to expect recompiles | whenever something depends on a function I've changed, even | if I don't think the changes were very significant. But you get many many fewer such "unexpected" changes when you do not use -O or (I believe) when you use ghci. Reason: without -O ghc exports only type info across module boundaries. So I'd still like to understand why Martin is seeing so much recompilation. S

1 0

interpreter features
by S.D.Mechveliani 30 Oct '01

30 Oct '01

Eray Ozkural <erayo(a)cs.bilkent.edu.tr> writes > Still, having come from that realm of old-school-AI-language-interpreters I > have difficulty using GHCi because I can't define functions interactively. > Would it be possible to introduce a mode to do that? > [..] > and perhaps going into "interactive" modules which would be separate > namespaces. > > And if LISP can do it, why can't you? ;) > [..] The following works in ghc-5.02: ghci ... Prelude> let {g:: Integer->Integer; g 0 = 0; g n = n+(g $ pred n)} Prelude> g 0 0 Prelude> g 3 6 But Prelude> let {type T = Int; n = 1 :: T} does not. I wonder, whether there is any good reason for such distinction. I would like too Haskell to behave like a good interpreter, with self-applicability, dynamic features, like, maybe, some Lisp systems do. ----------------- Serge Mechveliani mechvel(a)botik.ru

3 2

GHCi -- thanks ghc team !
by Ronald Legere 29 Oct '01

29 Oct '01

OK, I admit it. I didnt play with GHCi at all when it was first available ... because I was on windows... but I have started to play with it a bit with the latest installer... and let me say: Nice Job GHC team! Cheers! __________________________________________________ Do You Yahoo!? Make a great connection at Yahoo! Personals. http://personals.yahoo.com

3 2

RE: gtk+hs and libgmp
by Gérard Milmeister 29 Oct '01

29 Oct '01

On Mon, 2001-10-29 at 11:11, Simon Marlow wrote: > > > On Sat, 2001-10-27 at 17:27, Manuel M. T. Chakravarty wrote: > > > Gérard Milmeister <gemi(a)bluewin.ch> wrote, > > > > > > > On Thu, 2001-10-25 at 13:53, Simon Marlow wrote: > > > > > It seems very strange that the libgmp in your GHC > > installation doesn't work. How did you install GHC? Do you > > have problems linking other programs, or just gtk+hs programs? > > > > > > > > > As far as I know, it happens only with gtk+hs. Small > > programs I have > > > > written are linked alright. > > > > > > Do these small programs actually use any arbitrary precision > > > arthimetic - ie, use the type Integer? > > > > > > Manuel > > Something like this? > > > > module Main where > > > > main = print (1000000000000000000000000000000::Integer) > > > > Yes, it works. > > I know that gtk+hs adds a -L/usr/lib option to the command line, which comes originally from gtk-config. > > I'm still confused by what you said about GHC not linking with the libgmp.a in /opt/ghc, and yet it still manages to link programs using arbitrary precision integers somehow. Normally GHC won't look in /usr/local/lib for libraries, so it won't even find your GMP 3 library at all. > > Hmm. All very mysterious. I think you'll need to send us output from ghc -v for a link that failed and one that succeeded. > > Cheers, > Simon Here is the output: 1. Main.out from a link that succeeds. 2. GtkHelloWorld.out, a sample program from gtk+hs, compiled with ghc -v -fglasgow-exts -lgtkHS GtkHelloWorld.hs `gtkhs-config --cflags` `gtkhs-config --libs` `gtk-config --cflags` `gtk-config --libs` In fact, you are right, there is a -L/usr/lib before -lgmp. Maybe the libgmp.a included with ghc should be renamed, for example libgmp_ghc.a n.b. I send the message only to you, not to the mailing list because of the attachments. Regards,

1 1

ghc-5.0.2 for RedHat 7.x
by Chris Dornan 29 Oct '01

29 Oct '01

> I am trying to install ghc-5.02 on an NFS partition for use > with RedHad 7.1. The make in-place scheme with the binary > tar balls (as supplied for RedHat 6.2) work very well for this. > Does anyone know of such a tar ball for RedHat 7.1 or should I > be building from source, or can I adapt the RPM packages to put > things where I want them? To answer my own question, the only thing that seems to be stopping from using the RedHat-6.2 distribution on a RedHat 7.1 machine is libreadline.so, it have been changed from .so.3 to .so.4. However, if I create a symbolic link from one to the other (as suggested by the GHC Mini FAQ for SUSE systems) then it appears to work fine. In summary to get the ghc-5.00.2 binary tar ball to work on a RedHat 7.1 system create a symbolic link from libreadline.so.3, in a directory on $LD_LIBRARY_PATH to /usr/lib/libreadline.so.4 and it all seems to work fine. How easy would it be to use libreadline.a in making up the GHC distribution? (GHC implementers: ghc is a very fine system; thanks.) Chris

1 0

Re: ghc --make feature request
by cwitty＠newtonlabs.com 29 Oct '01

29 Oct '01

"Simon Marlow" <simonmar(a)microsoft.com> writes: > > I've seen unexpected compiles using ghc --make. I've got a system > > with modules A, B, and C; A depends on B and B depends on C. I've > > seen the following sequence of events: > > > > I change C > > ghc --make A compiles C, B, and A > > I change A > > ghc --make A skips C and compiles B and A > > > > I don't know of any reason why it would have compiled B in the second > > case. > > > > If this is not a known bug, I can try to reproduce it and submit a > > formal bug report. By the way, a quick try at reproducing this bug -- with what I remember of the sequence of events that caused it -- didn't work; things proceeded as they should have. > From your bug report on SourceForge, it looks like GHC has inferred > the wrong dependency graph for your program. It appears that > Point.lhs doesn't depend on GtkExtra, but nevertheless GHC has > decided to compile GtkExtra before Point, but after the other > dependencies of Point, which would seem to indicate that it thinks > Point depends on GtkExtra. > > If you do a 'ghc --make Point.lhs' (with whatever other options you > need), does it try to compile GtkExtra? No. Carl Witty

1 0

RE: ghc --make feature request
by Simon Marlow 29 Oct '01

29 Oct '01

> "Simon Marlow" <simonmar(a)microsoft.com> writes: > > > GHC actually has rather sophisticated recompilation checking which > > goes beyond just checking whether the interface changed - it keeps > > version information for each entity exported by a module and only > > recompiles if any of the entities actually used by the module have > > changed (this is described in the user's guide under the section on > > recompilation checking). > > I've seen unexpected compiles using ghc --make. I've got a system > with modules A, B, and C; A depends on B and B depends on C. I've > seen the following sequence of events: > > I change C > ghc --make A compiles C, B, and A > I change A > ghc --make A skips C and compiles B and A > > I don't know of any reason why it would have compiled B in the second > case. > > If this is not a known bug, I can try to reproduce it and submit a > formal bug report. >From your bug report on SourceForge, it looks like GHC has inferred the wrong dependency graph for your program. It appears that Point.lhs doesn't depend on GtkExtra, but nevertheless GHC has decided to compile GtkExtra before Point, but after the other dependencies of Point, which would seem to indicate that it thinks Point depends on GtkExtra. If you do a 'ghc --make Point.lhs' (with whatever other options you need), does it try to compile GtkExtra? I tried a small example with similar module structure to your setup, but couldn't reproduce the bug. Cheers, Simon

1 0

RE: ghc compiled program crashes
by Simon Marlow 29 Oct '01

29 Oct '01

> from time to time ghc compiled programs simply crash on > my windows2000 without any meaningfull error message. I guess > this has something to do with storage management (I vaguely > remember having read something about errors durcing gc on this > list, but did not find the corresponding posts in the archive). We need more info, in particular an example of a program that crashes. Does it crash in the same way every time? Does it depend on what GC options you use? Cheers, Simon

1 0

test
by Eray Ozkural 29 Oct '01

29 Oct '01

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 test! - -- Eray Ozkural (exa) <erayo(a)cs.bilkent.edu.tr> Comp. Sci. Dept., Bilkent University, Ankara www: http://www.cs.bilkent.edu.tr/~erayo GPG public key fingerprint: 360C 852F 88B0 A745 F31B EA0F 7C07 AE16 874D 539C -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.0.6 (GNU/Linux) Comment: For info see http://www.gnupg.org iD8DBQE73XOzfAeuFodNU5wRAt2mAJ9fmzCmrO5hLN9x2UJtVDLiH/6+9gCeNMks I2r1Ts1IPwh37u01s5p64tg= =iHjs -----END PGP SIGNATURE-----

1 0

RE: Existential Typing (was Multi-parameter OOP)
by Simon Peyton-Jones 29 Oct '01

29 Oct '01

| If GHC had true existential typing, as opposed to just | existential datatypes, | you could reasonably code what I think you want like this: | | class A a where | basicA :: Bool | nextA :: a -> (EX a'. A a' => a') | basicA = True | nextA = id Curiously enough, this thread intersects with the "higher-ranked types" thread on the Haskell list. If I do implement the Odersky/Laufer higher-ranked types inference mechanism, then I'll also allow existentials in arbitrary positions. The elimination of "stupid" existential wrapper constructors is discussed in a bit more detail in the paper Mark and I wrote recently: First class modules for Haskell http://research.microsoft.com/~simonpj/papers/first-class-modules/index. htm Simon

2 1