On Aug 10, 2005, at 8:54 AM, Jack Howarth wrote:
FX,
I forgot to mention that since the gcc cvs I built last night now
contains your ISATTY and TTYNAM intrinsics patch, I regressed the
portion of the xplor.patch which worked around the prior absence
of the ISATTY intrinsic in gfortran. The resulting xplor-nih build
works fine.
One other question. Do you know much c++? I had to make one change
to some buggy c++ code to allow gcc 4.0 and main branch to compile
xplor-nih...
I am trying to figure out if this is catching some new buggy c++ which
is no longer accepted by the g++ compiler in gcc 4.1 or if this is a
regression in that branch.
The code is invalid C++ as "friend class" does not define a forward
declaration. Also for 4.1, the function norm is not considered as it
is declared after the use as 4.1
Both are invalid C++.
The example for the second issue is:
template<class T> class FixedVector;
template<class T> void norm(const FixedVector<T>& v);
template <class T> T g(void);
template<class T> void f(void) { norm(g<T>()); }
double norm(const double& v) {}
static void _wrap_sqrt__SWIGPY_1(){f<int >();}
there is only one norm before f, so the overloaded set of norm in the
template f is
just that one norm. If you move the other norm before the f, it works
as expected in
both GCC 4.1.0 and ICC in strict mode.
Thanks,
Andrew Pinski
