[CMake] rdynamic and fpp flags failing

Brad King brad.king at kitware.com
Tue Oct 6 15:31:15 EDT 2009


Jason Slemons wrote:
> Hello Brad, 
> 
>> Is this a cross-compiling environment? 
> 
> Yes it is, both are linux systems though. I'm mainly curious to see if it will build for now though. Ill deal with testing the build later. Oh and I noticed in CMakeSystem.cmake that it says:
> SET(CMAKE_CROSSCOMPILING "FALSE")
> But I notice this with cmake 2.6.4 and cmake 2.8, both when it works and when it fails. 

See here if you want real cross compiling:

  http://www.cmake.org/Wiki/CMake_Cross_Compiling

> I compiled CMakeFortranCompilerId.F(no problem) with pgf90 and pgf95 and I always get an executable that spits out just :
> INFO:compiler[PGI]
> INFO:platform[Linux]
> 
> You also asked about the output of strings a.out? it's a long file that I'll attach(pgi.strings). When I switch to a compiler that works(like gnu) strings a.out is still a long file that looks similar(gnu.strings) to me. 
> 
> It's starting to sink in that " The Fortran compiler identification is unknown " and that this is really the root of the problem. What I can't figure out is why it's not found, since the output from the executable of CMakeFortranCompilerId.F seems correct.

In order to handle cross compiling we do not actually run the
compiler id executable (the C and C++ one's do not actually
print anything even if they are executed).  Instead we parse
the binary file itself to extract encoded INFO strings.

Based on the string files you sent me I think that the format
used by your compiler to store strings does not separate the
INFO:compiler[] and INFO:platform[] text.  I'll have to fix
the parsing code to handle this case.

Please send me the 'a.out' from each compiler
(off-list since they are not of general use).

Thanks,
-Brad


More information about the CMake mailing list