[Fwd: Re: [GAP Forum] Help Configuring GAP Packages on iBook 900 MHz]

[Justin C. Walker]

Dear Dimitrii,

On Feb 9, 2008, at 22:17 , Dmitrii Pasechnik wrote:

> Dear Justin,
>
> On 2/10/08 6:35 AM, "Justin C. Walker" <justin at mac.com> wrote:
> [...]
>>> Now, I have a new problem. Make complains that it can't find a
>>> certain file in the carat directory:
>>>
>>> (cd src; make GNU_MP_LIB=. GNU_MP_INC=.)
>>> gcc -O3 -DLONGLONG -I../../carat/carat/include -L../../carat/carat/
>>> lib -DVERSION='"2.0 January 2003"' nq.o consistency.o pc.o
>>> relations.o word.o system.o time.o addgen.o tails.o eliminate.o
>>> engel.o glimt.o instances.o presentation.o mem.o pcarith.o
>>> collect.o combicol.o trmetab.o gap.o -L. -o nq -lgmp -static
>>> /usr/bin/ld: can't locate file for: -lcrt0.o
>>> collect2: ld returned 1 exit status
>>> make[1]: *** [nq] Error 1
>>> make: *** [compile] Error 2
>>>
>>> The file isn't in /usr/local/lib/gap4r4/pkg/carat/carat-2.0/lib.
>>
>> I've reproduced this on my system.  I really don't think this is a
>> 'gmp' issue,
> well, the problem evidently lies in the way carat builds gpm  
> library. Hence
> it IS a gmp issue :)
>
> I don't see why we really should discuss buidling gmp in this forum,
> especially as there are well-known ways to avoid messing around  
> with this on
> the particular platform we talk about, e.g. to use fink.
> Otherwise you are of course right that the tool chain gets confused  
> here.
> But it's gmp's toolchain...
>
>> and I would not (yet) start fooling around with fink or
>> MacPorts.
> without one of them, developing "generic" unix soft, such as gap,  
> on a Mac
> becomes a major pain, that's why it's not "fooling around", but  
> rather a
> very reasonable thing to do.

I wasn't trying to provoke a discussion on the issue of fink/macports  
(and I'm not disagreeing that fink and macports are good ways to deal  
with Unix software on Mac OS X).  I wanted to isolate what was going  
wrong for Jeffrey.

For someone not familiar with software development, tackling the job  
of installing either of these schemes can be daunting.  Throwing that  
into the mix at the start didn't seem like the right approach.

It may be that it works out to be a GMP issue, but for now, my  
concern is getting to the bottom of a specific failure.

In my case, the problem seems to be related to differences between 64- 
bit and 32-bit toolchains (and the fact that I had a 64-bit GMP and a  
32-bit nq).  I'm not sure yet what is wrong in Jeffrey's case.

Justin

--
Justin C. Walker, Curmudgeon-At-Large
Institute for the Enhancement of the Director's Income
--------
When LuteFisk is outlawed,
Only outlaws will have LuteFisk
--------