October 2001 - Glasgow-haskell-users

RE: Readline version
by Simon Marlow 29 Oct '01

29 Oct '01

> I've been having a bit of trouble compiling ghc5.02 from source on > Solaris 2.6. The problem is that it can't find > readline/readline.h which > is not in a standard place since I had to compile readline > myself. I've > been messing around with various environment variables and > configuration > options, but have failed so far to communicate where this > file is. I'm > hoping someone can give me a suggestion. (I've reviewed the > mailing list > log and FAQ and have seen that readline is a problem on Solaris, but > haven't seen anything that obviously is what I've been missing.) > > I hacked mk/config.mk to say GhcLibsWithReadline=YES, since I have > libreadline.a and libtermcap.a in my LD_LIBRARY_PATH. BTW: the "right" way to modify these variables is by overriding them in your build.mk file, since config.mk is auto-generated every time you run configure. > That produced the > aforementioned missing readline/readline.h. I tried CFLAGS=<blah> so > that readline/readline.h should be visible but that didn't > work either. You can set ReadlineIncludePath in build.mk to the path containing readline/readline.h. > One more thing, is there a particular version of readline I should be > using? (I'm using 4.2) 4.2 should be fine. Cheers, Simon

1 0

RE: gtk+hs and libgmp
by Simon Marlow 29 Oct '01

29 Oct '01

> 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

1 0

ghc compiled program crashes
by Sven Eric Panitz 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). sven eric

1 0

RE: gtk+hs and libgmp
by Gérard Milmeister 27 Oct '01

27 Oct '01

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. -- Gérard Milmeister Tannenrauchstr. 35 8038 Zürich gemi(a)bluewin.ch

2 2

Re: Reading files via http or ftp
by Marcin 'Qrczak' Kowalczyk 26 Oct '01

26 Oct '01

25 Oct 2001 21:01:27 +0100, Colin Paul Adams <colin(a)colina.demon.co.uk> pisze: > Has anyone used the Socket library to read files specified via an > http and/or ftp URL? Yes (http), although I implemented only a small subset of the http client protocol. http://qrczak.ids.net.pl/Haber-0.3.tar.gz -- __("< Marcin Kowalczyk * qrczak(a)knm.org.pl http://qrczak.ids.net.pl/ \__/ ^^ SYGNATURA ZASTÊPCZA QRCZAK

1 0

Readline version
by Frank Seaton Taylor 26 Oct '01

26 Oct '01

Good day folks. I've been having a bit of trouble compiling ghc5.02 from source on Solaris 2.6. The problem is that it can't find readline/readline.h which is not in a standard place since I had to compile readline myself. I've been messing around with various environment variables and configuration options, but have failed so far to communicate where this file is. I'm hoping someone can give me a suggestion. (I've reviewed the mailing list log and FAQ and have seen that readline is a problem on Solaris, but haven't seen anything that obviously is what I've been missing.) I hacked mk/config.mk to say GhcLibsWithReadline=YES, since I have libreadline.a and libtermcap.a in my LD_LIBRARY_PATH. That produced the aforementioned missing readline/readline.h. I tried CFLAGS=<blah> so that readline/readline.h should be visible but that didn't work either. One more thing, is there a particular version of readline I should be using? (I'm using 4.2) Thanks, ---Frank Seaton Taylor P.S. I'm a bit new at some of this stuff, so I may have misrepresented the situation. I'd be glad to clarify anything.

1 0

Re: ghc --make feature request
by cwitty＠newtonlabs.com 26 Oct '01

26 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. Carl Witty

1 0

RE: Strictness Information (Was: ghc --make feature request)
by Simon Marlow 26 Oct '01

26 Oct '01

> On Friday 26 October 2001 14:41, Jan-Willem Maessen wrote: > > > 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. > > Is there a way to automatically discover the strictness of a > function? I > think it would be really usefull to have some kind of > extended :t command in > ghci which would do something like: > > main> :T foo > foo :: A (Strict) -> A (Strict) -> Bool > main> :T bar > bar :: a (Lazy) -> [a] (Lazy) -> [a] > > Perhaps this is possible already, and I just don't know about it. This may be possible for entities from compiled modules, but GHCi can't do strictness analysis for interpreted modules (due to quite boring reasons - the byte-code backend can't quite handle the output from some of the optimisation passes). That's an interesting suggestion anyway, we'll bear it in mind. Cheers, Simon

1 0

RE: ghc --make feature request
by Simon Marlow 26 Oct '01

26 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). > > > > So the upshot is that what you're describing shouldn't > happen, and it > > may be a bug. Could you send us more info? > > 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. Good point. But I'd also like to add that these hard-to-spot reasons only apply when optimisation is turned on - otherwise the only way a function's version number may change is if its type changes. In GHCi, optimisation is never on so you shouldn't see any of these effects. Cheers, Simon

1 0

RE: ghc --make feature request
by Jan-Willem Maessen 26 Oct '01

26 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). > > So the upshot is that what you're describing shouldn't happen, and it > may be a bug. Could you send us more info? 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. -Jan-Willem Maessen

2 1