Description of problem: When compiling cp2k-5.1 gfortran fails to compile hfx_contraction_methods.F90 with internal compiler error. Version-Release number of selected component (if applicable): gcc-gfortran-7.1.1-2.fc27.ppc64 How reproducible: Always. Steps to Reproduce: 1. download src.rpms from https://rathann.fedorapeople.org/copr/ 2. mockchain -r fedora-rawhide-ppc64 elpa-2017.05.003-1.fc27.src.rpm cp2k-5.1-1.fc27.src.rpm Actual results: gfortran -c -D__FFTW3 -D__LIBINT -D__LIBXC -D__HAS_IEEE_EXCEPTIONS -D__LIBINT_MAX_AM=11 -D__LIBDERIV_MAX_AM1=7 -D__MAX_C ONTR=4 -O2 -g -pipe -Wall -Werror=format-security -Wp,-D_FORTIFY_SOURCE=2 -fexceptions -fstack-protector-strong --param= ssp-buffer-size=4 -grecord-gcc-switches -specs=/usr/lib/rpm/redhat/redhat-hardened-cc1 -specs=/usr/lib/rpm/redhat/redhat -annobin-cc1 -m64 -fPIC -I/usr/lib64/gfortran/modules -fopenmp -ffree-form -ffree-line-length-none -D__COMPILE_ARCH="\"L inux-ppc64-gfortran\"" -D__COMPILE_DATE="\"Mon Dec 18 22:17:13 UTC 2017\"" -D__COMPILE_HOST="\"5e870004bb2a47fabeeefec9f 322d0bf\"" -D__COMPILE_REVISION="\"svn:18091\"" -D__DATA_DIR="\"/usr/share/cp2k\"" -D__SHORT_FILE__="\"hfxbase/hfx_contr action_methods.F\"" -I'/builddir/build/BUILD/cp2k-5.1/src/hfxbase/' hfx_contraction_methods.F90 gfortran: internal compiler error: Killed (program f951) Please submit a full bug report, with preprocessed source if appropriate. See <http://bugzilla.redhat.com/bugzilla> for instructions. make[3]: *** [/builddir/build/BUILD/cp2k-5.1/makefiles/Makefile:428: hfx_contraction_methods.o] Error 4 make[2]: *** [/builddir/build/BUILD/cp2k-5.1/makefiles/Makefile:124: all] Error 2 make[1]: *** [/builddir/build/BUILD/cp2k-5.1/makefiles/Makefile:113: sopt] Error 2 Expected results: cp2k-5.1 should at least compile successfully.
gfortran: internal compiler error: Killed (program f951) ^^ this sounds more like OOM killed process than a real ICE in gcc/gfortran, I will take a look
Dominik, cp2k builds fine in my Rawhide ppc64le VM with 16GB of RAM. What machine did you use for your builds? Mock reports the build machine specs into hw_info.log.
Yes, it's very unlikely this is a gcc bug; simply use a machine with more memory.
Thanks for checking. I did indeed run those builds on VMs with 2GB of RAM. Sorry for the noise.