On Mon, Jun 15, 2009 at 3:38 PM, Dr. David
Kirkby<david.kir...@onetel.net> wrote:
> William Stein wrote:
>>
>> On Sun, Jun 14, 2009 at 3:16 PM, Dr. David
>
>>> Could one or both of you verify you pick up gcc 4.4.0 when you log in,
>>> and
>>> that the compiler will compile a simple c program. It should be ok, but
>>> no
>>> so easy for me to check how it works for others, as the paths are in my
>>> own
>>> directory.
>>
>> yep it works:
>
> Good
>
>> wst...@t2:~/tmp$ gcc sage.c
>> wst...@t2:~/tmp$ ./a.out
>> sage:
>> sage:
>> wst...@t2:~/tmp$ gcc -v
>> Using built-in specs.
>> Target: sparc-sun-solaris2.10
>> Configured with: ../gcc-4.4.0/configure --with-gnu-as
>> --with-as=/home/kirkby/bin/as --with-gnu-ld
>> --with-ld=/home/kirkby/bin/ld --with-gmp=/home/kirkby/dependencies/
>> --with-mpfr=/home/kirkby/dependencies/
>> --enable-languages=c,c++,fortran --prefix=/home/kirkby/dependencies/
>> Thread model: posix
>> gcc version 4.4.0 (GCC)
>>
>>
>>> At a later date, I'll build a complete new gcc which should be easily
>>> downloadable and work for anyone without messing around with enviroment
>>> variables and editing hard-coded paths.
>>
>> And even better also make a GCC spkg that includes binutils, which
>> anybody can (probably) build.
>
> I've been thinking about the use of the GNU binutils, and in particular the
> GNU linker. I think it might actually be better to use the Sun assembler and
> linker.
I am happy with whichever option works, and encourage you to try
whatever you want to see. Thanks for investigating this.
>
> I would not be surprised if two of the three issues I've come across (sqlite
> and linbox) might actually disappear if the Sun linker was used, rather than
> that GNU one. The sqlite issue was clearly related to the linking, and so I
> think is the linbox, since it fails to find libraries which clearly exist.
> The sqlite issue a bug in libtool, but I don't think that bug would have
> mattered with the Sun linker.
>
> There are a few reasons I am thinking this:
>
> 1) The gcc compiler Sun ship in /usr/sfw/bin uses the Sun tools. I would
> have thought if the GNU ones worked better with the compiler, Sun would have
> used them. (They do in fact ship the GNU compler and linker, but don't use
> them)
>
> 2) The guy from Sun on the sqlite mailing list suggested the Sun linker
> would be preferable.
>
> 3) The developer of libtool advised against the use of the GNU linker.
>
> 4) Past experience of mine.
>
> 5) A 64-bit port is likely to be easier with the Sun tools.
>
> 6) There is a front end for the Sun compiler and linker that takes gcc
> options, which might be a useful way of getting the performance of the Sun
> compiler, but without changing all the options.
>
> I just created a small shell script, which should (in the process of
> compiling it) build gcc with the only requirement being that the three
> source files gcc-$version.tar.gz, mpfr-$version.tar.gz and
> gmp-$version.tar.gz all exist. I think there are sufficient other files in
> /usr/sfw/bin to enable this to be built.
>
> That shell script could be converted to a .spkg file - I'd just rather get
> it right first in a script I fully understand and can edit quickly.
If you can make a straightforward automated shell script way to build GCC,
then I can very easily make an spkg from that.
Any chance you make it depend on MPIR instead of GMP?
http://mpir.org/
mpir is a drop-in replacement for gmp, which I like better for various reasons.
---
William Stein
Associate Professor of Mathematics
University of Washington
http://wstein.org
--~--~---------~--~----~------------~-------~--~----~
To post to this group, send email to sage-devel@googlegroups.com
To unsubscribe from this group, send email to
sage-devel-unsubscr...@googlegroups.com
For more options, visit this group at http://groups.google.com/group/sage-devel
URLs: http://www.sagemath.org
-~----------~----~----~----~------~----~------~--~---
