[FLASH-USERS] Compiling LaserSlab for FLASH 4.4 error in Diffuse_computeFluxLimiter.F90

[Klaus Weide]

On Tue, 3 Jan 2017, Marissa Adams wrote:

> Thank you very much Klaus. Your patch actually did not apply successfully,
> it says that Hunk #2 Failed at 194.

Hmm.

This probably was just a case inconsistency (1.0d-100 vs 1.0D-100 in the 
original file), I hope the fix was obvious.

And indeed, I seems you were able to progress to the next problem, not 
related to the previous one:

> > /opt/ibmcmp/xlf/bg/14.1/bglib64/libxlfmath.a(math.o): In function `_pow':
> > math.c:(.opd+0x408): multiple definition of `_pow'
> >
> 
> Surely as a result of my change. Would you suggest I talk to the system
> admins at this point?

Always a good idea to get local support - especially with a problem that 
may be specific to versions of installed libraries.

We have seen similar issues using XLF compilers some time ago.
Apparently, in your system there are conflicting versions of the math 
function '_pow' (which presumably is used to implement the Fortran '**' 
operator).  The version contained in libxlfmath.a somehow disagree with
another one linked in - it is not clear from the error message quoted
how that one comes in.

A possible workaround may consist of modifying your Makefile.h so that
you change the libraries to link with. You may want to change LIB_OPT / 
LIB_TEST / LIB_DEBUG. In our case, a workaround consisted in changing
MASS from '-lmass' to empty, but this may have been very specific to our 
Makefile.h and our problem.

Klaus