[Ecls-list] Some new issues with building on Solaris 8
Juan Jose Garcia-Ripoll
juanjose.garciaripoll at googlemail.com
Fri Sep 11 08:49:52 UTC 2009
This is pretty weird. The "configure" program I sent you before was a
version in which changes had been reverted... I have not really
changed much about the way GMP builds, really.
One last check would be configuring GMP out of place and see whether
it detects your platform correctly. Like you leave the ecl directory.
Make some other empty directory and do path/to/ecl/src/gmp/configure
--prefix=$HOME --disable-shared and see whether it still uses the v8
files or not.
If problems persist I might just merge in the MPIR library, which is
said to be a more sensible version.
On Fri, Sep 11, 2009 at 2:50 AM, Raymond Toy <toy.raymond at gmail.com> wrote:
> Juan Jose Garcia-Ripoll wrote:
>> May I ask you a favor? I have cooked up a new file "configure", that
>> does not tell GMP, neither the host nor the build type. I presume this
>> might be the actual problem. I would like to see whether it still
>> misidentifies your system as V8.
> I tried out this configure. Didn't really change anything, but I did
> see configure print out:
> checking for egrep... egrep
> using ABI="32"
> CC="gcc -fPIC"
> CFLAGS="-g -O2 -fPIC "
> MPN_PATH=" sparc32/v9 sparc32/v8 sparc32 generic"
> checking for function prototypes... yes
> And a closer inspection shows that all of the files in build/gmp/mpn are
> linked to gmp/mpn/sparc32/v9 files. I am running on an Ultrasparc II
> (and III) machine, so v9 with v8plus is ok.
> This explains why I get the error about using v8plus instructions in a
> non-v8plus target binary. I think either configure should link to a
> different set of files or configure should pass the correct flags for
> the v8plus build.
> Haven't figured out how this works, but a build from a few days ago did
> this correctly. I think it used the v8plus compiler flags.
Instituto de Física Fundamental, CSIC
c/ Serrano, 113b, Madrid 28006 (Spain)
More information about the ecl-devel