On Sat, 1 Jul 2006, Mathieu Taillefumier wrote:
I still have to check the example code which does compile with. I have some suggestions for this modifications.
I've pushed some changes [I -> Ii] to a bunch of examples to make them work. There could be some more examples that need this change. Some complex examples were giving errors - I'm not sure if this is some PETSc bug or a bug with the C99 complex change. This needs debugging.
- Since I don't really know python, I made a very bad modification on the language.py file. I put in comment two lines. However I think that it is better to still have the c++ version as default language at this time since I don't have enough feedback for C version. So the best is I think that c++ is the default language for complex but have an option to activate C version for complex numbers. I should also indicate that the we need to have a C99 compliant compiler gcc 3.4 and after work well but they are still not very optimized. Intel compiler works also, but I don't know about the other. So use the option with care. The modifications should not break anything.
I've made further fixes to configure to get complex working in C mode. Now the option --with-scalar-type is decoupled from --with-clanguage. This mode is a bit cleaner than trying to preserve the old default of complex => c++. BTW - gcc295 worked on my laptop [which also had gcc-3.4]. Not sure what that is.. [I have glibc-headers-2.3.6-0.fc3.1 - which is providing /usr/include/complex.h - which gcc295 is able to use] You might want to get the latest petsc-dev & BuildSystem [or petsc-dev.tar.gz snapshot] and see if it works. Satish
- As I said before. I still have to look at the example code with do not compile when Complex are activated. It is just question of time.
Regards
Mathieu