3.1 branch libstdc++ problem with ctime on hpux

[law@redhat.com]

 In message <20020326214821.A8298@disaster.basement.lan>, Phil Edwards writes:
 > On Tue, Mar 26, 2002 at 04:35:01PM -0700, law@redhat.com wrote:
 > > 
 > > The underlying problem is "ctime" provided by libstdc++-v3 defines "_CTIME
 > ",
 > > which is a special symbol for HPUX -- if _CTIME is defined, then HPUX will
 > > not provide a prototype for the "ctime" function.
 > > 
 > > The only reason _CTIME is defined apparently is to deal with including
 > > "ctime" multiple times. 
 > 
 > Usually there's a smarter multiple-inclusion guard than just this; _CPP_foo
 > or __SGI_STL_INTERNAL_foo_H are the two most common.  If we can think up
 > a more consistent scheme I'll volunteer to do the gruntwork changes for
 > 3.1 and 3.2.
I've got no particular preference.  So whatever scheme you and Benjamin
pick I'm happy with.

In the mean time I'll hack something up locally to avoid the problem so that
I can proceed with HPUX testing.

jeff