Page 1 of 1

Undefined CPP variables across the source code

Posted: Mon Feb 22, 2010 6:42 pm
by pouillon
Hi,

My 6.1.0 branch contains an improved version of my CPP option checker script. Running it on a 6.1.0 source tree reveals the following:

  • the HP and NEC Fortran compilers are checked for in the source, but not supported by the build system;
  • the ASL, Compaq, MLib, and SGIMath implementations of FFT are checked for in the source, but not supported by the build system;
  • the MLib mathematical functions are checked for in the source, but not supported by the build system;
  • an unused TEST_AIM CPP option appears in the source;
  • an unused TEST_WRAPPER CPP option appears in the source;
  • a T3E CPP option appears 54 times in the code, but is not supported by the build system;
  • a VMS CPP option appears 76 times in the code, but is not supported by the build system.

This situation has been lasting for 5 years for some of these options, which means that the associated code could be safely removed.

What can we get rid of, and what should we keep?

Re: Undefined CPP variables across the source code

Posted: Tue Feb 23, 2010 6:03 pm
by gonze
Dear Yann,

My feeling is that these can be safely removed.

In any case, we cannot maintain them, because for some, we do not have
the material to test, and for others (e.g. libraries) we are already supporting
a sufficient number of other possibilities.

But ... wait a bit for people to react eventually ...

Best regards,
Xavier

Re: Undefined CPP variables across the source code

Posted: Wed Feb 24, 2010 10:54 am
by torrent
Yann,

We are not using these features (at the CEA and on all the architecture of the "GENCI" consortium).
So, from our point of view (strictly), you can remove them

Re: Undefined CPP variables across the source code

Posted: Wed Feb 24, 2010 12:15 pm
by pouillon
OK, I'll wait for the next cold and rainy week-end. ;)