It would be good to report a reduced test case upstream. They may not fix it, 
but a lot of things related to static libraries don't work without coaxing and 
they'll never get fixed if people who use CMake with static libraries don't 
make their voices heard.

"Palmer, Bruce J via petsc-users" <petsc-users@mcs.anl.gov> writes:

> Argh, I'm an idiot. I can't write a proper print statement in CMake.
>
> The PETSC_STATIC_LDFLAGS variable is showing all the libraries so probably 
> all I need to do is substitute that for PETSC_LDFLAGS in the GridPACK CMake 
> build (once I find it) when the build is static.
>
> On 2/22/22, 10:39 AM, "Satish Balay" <ba...@mcs.anl.gov> wrote:
>
>     You can run 'pkg-config --static --libs 
> PETSC_DIR/PETSC_ARCH/lib/pkgconfig/petsc.pc' to verify if pkg-config is able 
> to obtain 'Libs.private' values.
>
>     And then you would need help from someone who can debug cmake - on why 
> PETSC_STATIC set by cmake does not reflect this value [as it should - per the 
> FindPkgConfig doc]
>
>     [sorry - I don't understand cmake - or how one would debug cmake issues]
>
>     Satish
>
>     On Tue, 22 Feb 2022, Palmer, Bruce J via petsc-users wrote:
>
>     > The static versions of the variables exist (PETSC_STATIC), but they 
> appear to have the same values as the non-static variables.
>     > 
>     > As I mentioned, I'm a complete novice at pkgconfig, but it looks like 
> if you could add the contents of Libs.private to the link line, you'd be in 
> business. Any idea how to access this information from CMake?
>     > 
>     > Bruce
>     > 
>     > On 2/22/22, 10:22 AM, "Satish Balay" <ba...@mcs.anl.gov> wrote:
>     > 
>     >     
> https://gcc02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fcmake.org%2Fcmake%2Fhelp%2Flatest%2Fmodule%2FFindPkgConfig.html&amp;data=04%7C01%7CBruce.Palmer%40pnnl.gov%7C5533bb688b344af7fb1a08d9f632b7dd%7Cd6faa5f90ae240338c0130048a38deeb%7C0%7C0%7C637811519962671755%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000&amp;sdata=O1tG6y4zgckU2lmyBM9F4hxiOhKNYz8O5WmFLH4k%2Fa0%3D&amp;reserved=0
>     > 
>     >     >>>
>     >     Two sets of values exist: One for the common case (<XXX> = 
> <prefix>) and another for the information pkg-config provides when called 
> with the --static option (<XXX> = <prefix>_STATIC).
>     >     <<<
>     > 
>     >     So perhaps CMAKE is already setting the _STATIC variant of (for 
> PETSC_LIB or equivalent) variable that's currently used)?
>     > 
>     >     Satish
>     > 
>     >     On Tue, 22 Feb 2022, Palmer, Bruce J via petsc-users wrote:
>     > 
>     >     > The contents of the petsc.pc file are listed below. It looks good 
> to me. The Libs.private variable seems to include the -lf2clapack and 
> -lf2cblas libraries. I don't know how this info gets propagated up the build 
> chain.
>     >     > 
>     >     > Bruce
>     >     > 
>     >     > 
> prefix=/pic/projects/gridpack/software/petsc-3.16.3/linux-openmpi-gnu-cxx-complex-opt
>     >     > exec_prefix=${prefix}
>     >     > includedir=${prefix}/include
>     >     > libdir=${prefix}/lib
>     >     > ccompiler=mpicc
>     >     > cflags_extra=-fPIC -Wall -Wwrite-strings -Wno-strict-aliasing 
> -Wno-unknown-pragmas -fstack-protector -fvisibility=hidden -g -O
>     >     > cflags_dep=-MMD -MP
>     >     > ldflag_rpath=-Wl,-rpath,
>     >     > cxxcompiler=mpicxx
>     >     > cxxflags_extra=-Wall -Wwrite-strings -Wno-strict-aliasing 
> -Wno-unknown-pragmas -fstack-protector -fvisibility=hidden -g -O  -std=gnu++11
>     >     > fcompiler=mpif90
>     >     > fflags_extra=-Wall -ffree-line-length-0 
> -Wno-unused-dummy-argument -g -O
>     >     > 
>     >     > Name: PETSc
>     >     > Description: Library to solve ODEs and algebraic equations
>     >     > Version: 3.16.3
>     >     > Cflags:   -I${includedir} 
> -I/pic/projects/gridpack/software/petsc-3.16.3/include
>     >     > Libs: -L${libdir} -lpetsc
>     >     > Libs.private: 
> -L/pic/projects/gridpack/software/petsc-3.16.3/linux-openmpi-gnu-cxx-complex-opt/lib
>  -L/share/apps/openmpi/3.0.1/gcc/6.1.0/lib 
> -L/qfs/projects/ops/rh6/gcc/6.1.0/lib/gcc/x86_64-pc-linux-gnu/6.1.0 
> -L/qfs/projects/ops/rh6/gcc/6.1.0/lib/gcc 
> -L/qfs/projects/ops/rh6/gcc/6.1.0/lib64 -L/qfs/projects/ops/rh6/gcc/6.1.0/lib 
> -lspqr -lumfpack -lklu -lcholmod -lbtf -lccolamd -lcolamd -lcamd -lamd 
> -lsuitesparseconfig -lrt -lsuperlu -lsuperlu_dist -lf2clapack -lf2cblas 
> -lparmetis -lmetis -lm -lstdc++ -ldl -lmpi_usempif08 -lmpi_usempi_ignore_tkr 
> -lmpi_mpifh -lmpi -lgfortran -lm -lgfortran -lm -lgcc_s -lquadmath -lpthread 
> -lquadmath -lstdc++ -ldl
>     >     > 
>     >     > On 2/22/22, 8:39 AM, "Satish Balay" <ba...@mcs.anl.gov> wrote:
>     >     > 
>     >     >     The relevant pkg-config commands are:
>     >     > 
>     >     >     balay@sb /home/balay/petsc (release=)
>     >     >     $ pkg-config --libs arch-linux-c-debug/lib/pkgconfig/petsc.pc 
>     >     >     -L/home/balay/petsc/arch-linux-c-debug/lib -lpetsc 
>     >     >     balay@sb /home/balay/petsc (release=)
>     >     >     $ pkg-config --shared --libs 
> arch-linux-c-debug/lib/pkgconfig/petsc.pc 
>     >     >     -L/home/balay/petsc/arch-linux-c-debug/lib -lpetsc 
>     >     >     balay@sb /home/balay/petsc (release=)
>     >     >     $ pkg-config --static --libs 
> arch-linux-c-debug/lib/pkgconfig/petsc.pc 
>     >     >     -L/home/balay/petsc/arch-linux-c-debug/lib -lpetsc 
> -L/home/balay/soft/mpich-3.4.2/lib -L/usr/lib/gcc/x86_64-redhat-linux/11 
> -llapack -lblas -lm -lX11 -lstdc++ -ldl -lmpifort -lmpi -lgfortran -lm 
> -lgfortran -lm -lgcc_s -lquadmath -lstdc++ -ldl 
>     >     > 
>     >     > 
>     >     >     And more example usages in share/petsc/Makefile.user
>     >     > 
>     >     >     Satish
>     >     > 
>     >     > 
>     >     >     On Tue, 22 Feb 2022, Barry Smith wrote:
>     >     > 
>     >     >     >  Bruce,
>     >     >     > 
>     >     >     >   Can you please send the PkgConfig calls that you make to 
> get the PETSc values? And then exactly what PETSc PkgConfig returns.
>     >     >     > 
>     >     >     > Thanks
>     >     >     > 
>     >     >     >    Barry
>     >     >     > 
>     >     >     > 
>     >     >     > > On Feb 22, 2022, at 11:03 AM, Palmer, Bruce J via 
> petsc-users <petsc-users@mcs.anl.gov> wrote:
>     >     >     > > 
>     >     >     > > Hi,
>     >     >     > >  
>     >     >     > > We recently switched the CMake configuration on our 
> GridPACK application to use the PkgConfig utility instead of Jeb Brown’s 
> FindPETSc.cmake module. This seems to work on a number of platforms but it is 
> failing to link on others. It appears that the build cannot find the LAPACK 
> and BLAS libraries. The PETSc library I’m linking to (v3.16.3) was configured 
> with -download-f2cblaslapack so it should have these libraries, but when I 
> try and link one of the test applications in GridPACK I get the errors
>     >     >     > >  
>     >     >     > > /share/apps/gcc/6.1.0/bin/g++ -pthread -g -rdynamic 
> CMakeFiles/greetings.dir/test/greetings.cpp.o -o greetings  
> -Wl,-rpath,/qfs/projects/ops/rh6/openmpi/3.0.1/gcc/6.1.0/lib 
> ../math/libgridpack_math.a libgridpack_parallel.a 
> ../timer/libgridpack_timer.a ../environment/libgridpack_environment.a 
> ../math/libgridpack_math.a 
> /pic/projects/gridpack/software/boost_1_65_0/lib/libboost_mpi.a 
> /pic/projects/gridpack/software/boost_1_65_0/lib/libboost_serialization.a 
> /pic/projects/gridpack/software/boost_1_65_0/lib/libboost_random.a 
> /pic/projects/gridpack/software/boost_1_65_0/lib/libboost_filesystem.a 
> /pic/projects/gridpack/software/boost_1_65_0/lib/libboost_system.a 
> /pic/projects/gridpack/software/ga-5.7/build_pr/lib/libga++.a 
> /pic/projects/gridpack/software/ga-5.7/build_pr/lib/libga.a 
> /pic/projects/gridpack/software/ga-5.7/build_pr/lib/libarmci.a -lrt 
> /usr/lib64/librt.so /usr/lib64/libdl.so 
> /qfs/projects/ops/rh6/openmpi/3.0.1/gcc/6.1.0/lib/libmpi.so 
> ../timer/libgridpack_timer.a 
>      libgri
>     >      dpack_
>     >     >      parallel.a ../configuration/libgridpack_configuration.a 
> /pic/projects/gridpack/software/petsc-3.16.3/linux-openmpi-gnu-cxx-complex-opt/lib/libpetsc.a
>  
>     >     >     > > 
> /pic/projects/gridpack/software/petsc-3.16.3/linux-openmpi-gnu-cxx-complex-opt/lib/libpetsc.a(zstart.o):
>  In function `petscinitializef_':
>     >     >     > > 
> /pic/projects/gridpack/software/petsc-3.16.3/src/sys/objects/ftn-custom/zstart.c:280:
>  undefined reference to `mpi_init_'
>     >     >     > > 
> /pic/projects/gridpack/software/petsc-3.16.3/linux-openmpi-gnu-cxx-complex-opt/lib/libpetsc.a(baijfact.o):
>  In function `MatSolve_SeqBAIJ_N':
>     >     >     > > 
> /pic/projects/gridpack/software/petsc-3.16.3/src/mat/impls/baij/seq/baijfact.c:1462:
>  undefined reference to `zgemv_'
>     >     >     > > 
> /pic/projects/gridpack/software/petsc-3.16.3/src/mat/impls/baij/seq/baijfact.c:1475:
>  undefined reference to `zgemv_'
>     >     >     > > 
> /pic/projects/gridpack/software/petsc-3.16.3/src/mat/impls/baij/seq/baijfact.c:1478:
>  undefined reference to `zgemv_'
>     >     >     > > 
> /pic/projects/gridpack/software/petsc-3.16.3/linux-openmpi-gnu-cxx-complex-opt/lib/libpetsc.a(baijfact.o):
>  In function `MatSolve_SeqBAIJ_N_NaturalOrdering':
>     >     >     > > 
> /pic/projects/gridpack/software/petsc-3.16.3/src/mat/impls/baij/seq/baijfact.c:1407:
>  undefined reference to `zgemv_'
>     >     >     > > 
> /pic/projects/gridpack/software/petsc-3.16.3/src/mat/impls/baij/seq/baijfact.c:1420:
>  undefined reference to `zgemv_'
>     >     >     > > 
> /pic/projects/gridpack/software/petsc-3.16.3/linux-openmpi-gnu-cxx-complex-opt/lib/libpetsc.a(baijfact.o):/pic/projects/gridpack/software/petsc-3.16.3/sr
>     >     >     > >  
>     >     >     > > I suspect that the reason it worked for others and not 
> for me is that they had viable blas and lapack libraries in their path and I 
> don’t. Is there anything special you need to do to make sure that the build 
> is pointed at the libraries that get created with the -download-f2cblaslapack 
> option?
>     >     >     > >  
>     >     >     > > Bruce Palmer
>     >     >     > 
>     >     >     > 
>     >     > 
>     >     > 
>     > 
>     > 

Reply via email to