[petsc-dev] [petsc-maint #45539] petsc installation failure

[Satish Balay]

Barry,

The code that crashed for me is the fortran configure test [not the c
configure test].

login1.surveyor:~/junk > cat foo.F 
                       program main
                       write (*,*) 'hello'
                       end
login1.surveyor:~/junk > mpif90 foo.F
./a.out
login1.surveyor:~/junk > ./a.out
Illegal instruction
login1.surveyor:~/junk >


Note: one difference between the machines BG/L and
BG/P are the version of gnu compilers used. [gcc 3 vs gcc 4]. And the
results could be different for IBM compilers. [Didn't try yet].

Satish

On Fri, 30 Apr 2010, Barry Smith wrote:

> 
>    If you don't change ./configure then which code does not run on our  
> BG and triggers the "crash" of .configure. Is it the same one that did  
> for that guy? If you then compile and run that code standalone outside  
> of .configure does it not run?
> 
> 
>      Barry
> 
> On Apr 29, 2010, at 9:20 PM, Satish Balay wrote:
> 
> > On Thu, 29 Apr 2010, Barry Smith wrote:
> >
> >>
> >>    THIS IS INSANE! What a stupid useless error message. Why isn't it
> >> suggesting --with-batch???
> >>
> >>     Well, look what happens -----
> >>
> >> In checkCCompiler() it builds an executable and seems to run without
> >> error. This is where configure checks for failed running and prints
> >> the helpful --with-batch option. But since the "execute" succeeds no
> >> useful error message.
> >>
> >> Then several other executables seem to run fine.
> >>
> >> Then it builds the sizeof(char) test code and that is not executable
> >> and hence the terrible "Unable to determine ..." error message.
> >
> > Yes - some code compiled for compute nodes also run on the frontend -
> > but not all..
> >
> >>
> >>    Satish,
> >>
> >> Does this same problem occur on our BG?
> >
> > Nope.. I get:
> >>>>>>>>>>>>>>>>>>>>>>>>>
> > login1.surveyor:~/petsc-release-3.1 > ./configure --with-fc=mpif77
> > = 
> > = 
> > = 
> > = 
> > = 
> > = 
> > = 
> > = 
> > = 
> > ======================================================================
> >             Configuring PETSc to compile on your system
> > = 
> > = 
> > = 
> > = 
> > = 
> > = 
> > = 
> > = 
> > = 
> > ======================================================================
> > TESTING: checkFortranCompiler from config.setCompilers(config/ 
> > BuildSystem/config/setCompilers.py:742)                     
> > *******************************************************************************
> >                    UNABLE to EXECUTE BINARIES for config/configure.py
> > -------------------------------------------------------------------------------
> > Cannot run executables created with FC. If this machine uses a batch  
> > system
> > to submit jobs you will need to configure using/configure.py with  
> > the additional option  --with-batch.
> > Otherwise there is problem with the compilers. Can you compile and  
> > run code with your C/C++ (and maybe Fortran) compilers?
> > *******************************************************************************
> > <<<<<<<<<<<<<<<<<
> >
> >>
> >> Can you have the original test of the executable in checkCCompiler()
> >> make a code with a print statement in it and see if that is not
> >> executable and hence stops with the correct error message?
> >
> > I have tested with sizeof.c [in src/benchmarks] and this code compiles
> > fine with mpicc - and runs on the front-end. I guess I need to find a
> > trigger that stops this from working..
> >
> > login1.surveyor:~/junk > mpicc sizeof.c
> > login1.surveyor:~/junk > ./a.out
> > long double : 16
> > double      : 8
> > int         : 4
> > char        : 1
> > short       : 2
> > long        : 4
> > long long   : 8
> > int *       : 4
> > size_t      : 4
> > login1.surveyor:~/junk >
> >
> > Satish
> >
> >>
> >>   Barry
> >>
> >>
> >> On Apr 29, 2010, at 8:16 PM, Satish Balay wrote:
> >>
> >>> Its not AIX - looks like its BlueGene/L - and you need --with- 
> >>> batch=1
> >>> option.
> >>>
> >>> For a preconfigured BG/P options list with IBM compilers - check:
> >>> config/examples/arch-bgp-ibm-opt.py
> >>>
> >>> To use PETSc on BG/L with the default GNU compilers - you can try
> >>>
> >>> ./configure --with-batch=1 --download-f-blas-lapack=1
> >>>
> >>> [and follow instructions as printed by configure]
> >>>
> >>> Satish
> >>>
> >>>
> >>> On Thu, 29 Apr 2010, Shree Krishna wrote:
> >>>
> >>>> Hi,
> >>>>
> >>>> I was trying to install Petsc on CCNI ( IBM AIX cluster at  
> >>>> Rensselaer
> >>>> Polytechnic institute) and it gave me following error.
> >>>> -------------------------------------------------------------------------------
> >>>> Unable to determine size of char
> >>>>
> >>>> I am attaching the configure.log file with this email. Will it be
> >>>> able for
> >>>> you to help me configure Petsc.
> >>>>
> >>>> Thanks.
> >>>> Shree.
> >>>>
> >>>>
> >>>
> >>>
> >>
> >>
> >>
> >
> 
> 
>