I'm puzzled; if try this (on 10.4.11 _ but with my atlas (3.9.9)),
everything goes fine.
And there shouldn't be any difference as to the treatment of those 2
symbols...
(lib is linked in basically the same way).
[Only I ran the test under gfortran of gcc44, because the is
currently a buildlock
on it for a (long!) build of atlas-3.9.10].
And "otool -L test" shows /sw/lib/gcc4.3/lib/libgcc_s.1.dylib,
so gfortran did add it to the link line...
Do you get indeed :
# nm -m /sw/lib/liblapack.dylib |fgrep powi
(undefined [lazy bound]) external ___powidf2 (from libgcc)
(undefined [lazy bound]) external ___powisf2 (from libgcc)
and does /sw/lib/gcc4.3/lib/libgcc_s.1.dylib appear in the output
of `otool -L /sw/lib/liblapack.dylib` ?
If you have gcc44, could you try the same command after installing
its gfortran?
Jean-Francois
On 29 Mar 2009, at 17:42, Sébastien Maret wrote:
> The programs bellow fails to link against lapack library from the
> atlas
> package:
>
> % gfortran -L/sw/lib -llapack -lcblas -lf77blas -latlas test_dgesv.f
> /usr/bin/ld: Undefined symbols:
> ___powisf2 referenced from liblapack expected to be defined in libgcc
> ___powidf2 referenced from liblapack expected to be defined in libgcc
> collect2: ld returned 1 exit status
>
> This is with gcc43 4.3.3-1000 and atlas 3.8.2-2 on MacOS 10.4.11 PPC.
> The powisf2 and powidf2 symbol are defined in
> /sw/lib/gcc4.3/lib/libgcc_s.1.dylib. Indeed the program links fine if
> one add -lgcc_s.1 to the link command:
>
> % gfortran -L/sw/lib -llapack -lcblas -lf77blas -latlas -lgcc_s.1
> test_dgesv.f
>
> Why does this library need to be added by hand? Is it a bug in
> atlas on
> 10.4/PPC? The same program compiles well on 10.5 without -lgcc_s.1.
>
>
> PROGRAM TEST_DGESV
> C...Resolution d'un systeme lineaire
> IMPLICIT NONE
> DOUBLE PRECISION MAT
> DOUBLE PRECISION VEC
>
> DIMENSION MAT(4,4)
> DIMENSION VEC(4)
>
> INTEGER I
>
> INTEGER INFO
> INTEGER IPIV
> DIMENSION IPIV(4)
>
> C...Elements de la matrice MAT
> C...Premiere ligne
> MAT(1,1)=1
> MAT(1,2)=2
> MAT(1,3)=3
> MAT(1,4)=4
> C...Deuxieme ligne
> MAT(2,1)=4
> MAT(2,2)=1
> MAT(2,3)=2
> MAT(2,4)=3
> C...Toisieme ligne
> MAT(3,1)=5
> MAT(3,2)=6
> MAT(3,3)=1
> MAT(3,4)=2
> C...Quatrieme ligne
> MAT(4,1)=-7
> MAT(4,2)=-8
> MAT(4,3)=-9
> MAT(4,4)=1
>
> C...Elements du vecteur VEC
> VEC(1)=-10
> VEC(2)=-4
> VEC(3)=-12
> VEC(4)=-22
>
> C...Resolution de A*X=B
> C...DGESV(N,NRHS,A,LDA,IPIV,B,LDB,INFO)
> CALL DGESV(4,1,MAT,4,IPIV,VEC,4,INFO)
> C...En sortie, MAT contient sa decomposition LU
> C...En sortie, la solution est dans VEC
> C...En sortie, INFO=0 si tout va bien
> C...En sortie, IPIV contient la liste des permutations
> WRITE(6,*)'INFO=',INFO
> WRITE(6,*)'LISTE DES PERMUTATIONS:'
> DO I=1,4
> WRITE(6,*)'IPIV(',I,')=',IPIV(I)
> ENDDO
>
> C...Affichage du resultat (1,-2,3,-4)
> WRITE(6,*)'SOLUTION DE MAT*X=VEC:'
> DO I=1,4
> WRITE(6,*)'X(',I,')=',VEC(I)
> ENDDO
>
> END PROGRAM TEST_DGESV
>
>
>
>
>
> ----------------------------------------------------------------------
> --------
> _______________________________________________
> Fink-devel mailing list
> [email protected]
> http://news.gmane.org/gmane.os.apple.fink.devel
------------------------------------------------------------------------------
_______________________________________________
Fink-devel mailing list
[email protected]
http://news.gmane.org/gmane.os.apple.fink.devel
