When using git versions of PETSc you cannot use this option --with-sowing=0
unless you turn off Fortran with --with-fc=0 Sowing builds much of the infrastructure for Fortran. Barry > On Dec 11, 2016, at 4:14 PM, Pierre Jolivet <[email protected]> > wrote: > > On Sun, 11 Dec 2016 16:02:02 -0600, Barry Smith wrote: >> I don't think this has anything to do with MUMPS. > > For sure, otherwise I would have bugged my teammates (who are developing > MUMPS) instead of you. > >> Seems the Intel ifort >> ifort (IFORT) 17.0.0 20160721 >> >> doesn't like our Fortran modules. I'll try to reproduce this and >> determine the problem. > > Actually, I just realized that I tried to fix the issue myself first. I have > re-uploaded make.log. The initial error was (correctly shown here > http://jolivet.perso.enseeiht.fr/make.log): > > petscsysmod.F(49): #error: can't find include file: > ../src/sys/f90-mod/ftn-auto-interfaces/petscsys.h90 > > But I noticed that the file src/sys/f90-mod/petscsys.h90 exists (no > "ftn-auto-interfaces"), so I did a quick `find . -name *mod.F -type f -print0 > | xargs -0 sed -i 's/ftn-auto-interfaces\///g'` and I ended up with the old > error instead. > > Sorry for the confusion. > > Thanks. > >> Barry >> >>> On Dec 11, 2016, at 3:48 PM, Pierre Jolivet <[email protected]> >>> wrote: >>> >>> On Sun, 11 Dec 2016 15:17:14 -0600, Matthew Knepley wrote: >>>> On Sun, Dec 11, 2016 at 3:12 PM, Pierre Jolivet wrote: >>>> >>>>> Hello, >>>>> It seems --with-fortran-interfaces is no more. >>>>> Any clue on how to compile packages which require a Fortran >>>>> compiler (e.g., MUMPS), without having PETSc compile the Fortran >>>>> interfaces? >>>> >>>> That flag just compiled the F90 interface. Now they are built by >>>> default. >>> >>> That's too bad for systems like mine which fail to build the PETSc >>> interfaces but have no problem whatsoever compiling MUMPS. >>> Following Barry's message, here are my logs (large files, sorry): >>> http://jolivet.perso.enseeiht.fr/make.log >>> http://jolivet.perso.enseeiht.fr/configure.log >>> It's been at least a good couple of months since I've been able to compile >>> the Fortran interfaces, and --with-fortran-interfaces=0 was the easiest fix >>> I could find. Now that it's gone, it looks like I'll have to bother you >>> with this. >>> >>> Thanks for looking. >>> >>> PS: this is from a clean dir (git clean -x -f -d) >>> >>>> Matt >>>> >>>> >>>>> Thanks in advance. >>> >
