After searching the web, the problem appears to be associated with compilation on the Athon64 chip. I can't check it though, so Chris if you could test this it would be appreciated. Try compiling the module on the command line by typing:
$ gcc -I/usr/include/python2.4 -c -o maths_fns/c_chi2.os maths_fns/c_chi2.c $ gcc -I/usr/include/python2.4 -c -o maths_fns/exponential.os maths_fns/exponential.c $ gcc -I/usr/include/python2.4 -c -o maths_fns/relax_fit.os maths_fns/relax_fit.c $ gcc -fPIC -shared -o maths_fns/relax_fit.so maths_fns/c_chi2.os maths_fns/exponential.os maths_fns/relax_fit.os Does the test data attached to the bug report cause a segfault after using these options? If this works, (which it does on my system) I should be able to modify the 'sconstruct' script to work on x86_64 chips. On 11/03/06, Edward d'Auvergne <[EMAIL PROTECTED]> wrote: > I've submitted this as a bug report in the bug tracker system on the > Gna! relax site. Attached to the bug report is an archive containing > all the files necessary for testing the relaxation curve-fitting C > module. Between the 1.2.2 and 1.2.3 releases, I specifically removed > the -fPIC option in the Sconstruct script (which is only added as an > option when the gcc compiler is used). The reason was that when using > the -fPIC compilation option, the shared object > relax/maths_functions/relax_fit.so would compile but would segfault > during execution. My system is Mandriva 2006 and, hence, the problem > may only arise in that distro. Chris, if you have time, would you be > able to run relax on the files I've attached to the bug tracker? It > would be much appreciated. If you use the 1.2.2 release or the 1.2.3 > release with the option 'SHCCFLAGS' removed from the 'sconstruct' > script, the information of whether the module segfaults on SUSE 10 > would be useful in fixing the problem. Another useful test would be > to compile the module by hand by typing: > > $ gcc -I/usr/include/python2.4 -c -o maths_fns/c_chi2.os maths_fns/c_chi2.c > $ gcc -I/usr/include/python2.4 -c -o maths_fns/exponential.os > maths_fns/exponential.c > $ gcc -I/usr/include/python2.4 -c -o maths_fns/relax_fit.os > maths_fns/relax_fit.c > $ gcc -shared -o maths_fns/relax_fit.so maths_fns/c_chi2.os > maths_fns/exponential.os maths_fns/relax_fit.os > > then running the test files, re-typing these commands with the -fPIC > option, and re-running the test files. > > > > > > On 10/03/06, Chris MacRaild <[EMAIL PROTECTED]> wrote: > > Build of relax1.2.3 failed with the following error: > > > > gcc -shared -o maths_fns/relax_fit.so maths_fns/c_chi2.os > > maths_fns/exponential.os maths_fns/relax_fit.os > > /usr/lib64/gcc/x86_64-suse-linux/4.0.2/../../../../x86_64-suse-linux/bin/ld: > > maths_fns/c_chi2.os: relocation R_X86_64_32S against `back_calc' can not > > be used when making a shared object; recompile with -fPIC > > maths_fns/c_chi2.os: could not read symbols: Bad value > > collect2: ld returned 1 exit status > > scons: *** [maths_fns/relax_fit.so] Error 1 > > scons: building terminated because of errors. > > > > The box is a 2 x Athlon64 machine running suse 10 > > > > relax1.2.2 builds ok on the same system, though I'm not currently using > > the curve-fitting routines so have done no testing... > > > > > > _______________________________________________ > > Relax-users mailing list > > [email protected] > > https://mail.gna.org/listinfo/relax-users > > > _______________________________________________ Relax-users mailing list [email protected] https://mail.gna.org/listinfo/relax-users
