Building on Solaris 10 with gcc 3.4.x

[Matt Ball]

I've got a Solaris 10u6 sparc and a Solaris 10u8 x86 system, with gcc 3.4.3 and 3.4.5 installed.  After extracting the source and typing './configure' and 'gmake', I get the following error while compiling TestAssert.cpp:

    In file included from TestAssert.cpp:2:
    ../../include/cppunit/portability/FloatingPoint.h: In function `int CppUnit::floatingPointIsFinite(double)':
    ../../include/cppunit/portability/FloatingPoint.h:43: error: `finite' was not declared in this scope

This is a little strange, because the ./configure script determined that 'finite' worked on this system (and accordingly defined CPPUNIT_HAVE_FINITE).

I got around this problem by manually changing FloatingPoint.h to not use finite, but the bigger question is why 'configure' thought 'finite' was available.  I couldn't see any obvious way to fix configure so that it correctly identifies 'finite'.

Anyone else have similar issues compiling on Solaris 10 with gcc 3.4.x?

[Stephan Bergmann]

I just ran into the same problem, trying to build CppUnit 1.12.1 on Solaris 10 using the Sun CC C++ compiler.

configure checks for finite with AC\_CHECK\_FUNCS, which only checks that finite is in libc (which it is), but not which header needs to be included (ieeefp.h on Solaris).

So, I think a fix would be:

    -- cppunit-1.12.1/include/cppunit/portability/FloatingPoint.h
    +++ cppunit-1.12.1/include/cppunit/portability/FloatingPoint.h
    @@ -3,6 +3,11 @@
    
     #include <cppunit/Portability.h>
     #include <math.h>
    +
    +#if defined(__sun) && !defined(CPPUNIT_HAVE_ISFINITE) && defined(CPPUNIT_HAVE_FINITE)
    +#include <ieeefp.h>
    +  // <math.h> is still needed for usage of fabs in TestAssert.cpp
    +#endif
    
     CPPUNIT_NS_BEGIN
    

[Stephan Bergmann]

… see (https://sourceforge.net/tracker/?func=detail&aid=2912590&group_id=11795&atid=311795)