[Trilinos-Users] Build regression on BGP

[John R Cary]

Thanks for your response.

On 4/29/10 2:25 PM, Brad King wrote:
>
> It looks like your toolchain file may not set everything it needs to set.
> Did you get this error with Trilinos 10.0?  Please also review this page:
>
>    http://www.cmake.org/Wiki/CMake_Cross_Compiling

Believe me, I have read and reread that page, googled, and so
forth.

I started with something like that page:

login3.intrepid$ grep -v '^ *#' 
/gpfs/home/cary/projects/vpall/numpkgs/Toolchain-BlueGeneP-mpixl.cmake | 
grep -v '^$'
SET(CMAKE_SYSTEM_NAME BlueGeneP-MPIXL)
set(CMAKE_FIND_ROOT_PATH
   /bgsys/drivers/ppcfloor/comm/fast
   /bgsys/drivers/ppcfloor/comm/default
   /home/cary/projects/vpall/numpkgs
   /home/cary/projects/vpall/CMake
)
set(CMAKE_FIND_ROOT_PATH_MODE_PROGRAM NEVER)
set(CMAKE_FIND_ROOT_PATH_MODE_LIBRARY ONLY)
set(CMAKE_FIND_ROOT_PATH_MODE_INCLUDE ONLY)
set(CMAKE_SHARED_LINKER_FLAGS " ")
set(HAVE_GCC_ABI_DEMANGLE_EXITCODE 0)
set(HAVE_TEUCHOS_BLASFLOAT_EXITCODE 1)

adding the last three lines to get past compilation, which *did work*
with 10.0.5.

Any suggestions much appreciated.


>
>> -- Error, the directory fmangle could not be found so we can not
>>     determine Fortran name mangling!
>
> This is the real problem that leads to the error message you get from
> the TrilinosFortranMangling.cmake module.  I'm not familiar with this
> version of Fortran mangling detection.  Hopefully one of the Trilinos
> developers can jump in here.
>

Okay.  Will wait.

Thx...John