VC Language - Template Compiler Error (array/pointers)

Author

Template Compiler Error (array/pointers)

Mycroft Holmes

2005-05-30, 9:00 am

Hi. The following sample is fine for Comeau Compiler.
Howver VC7.1 does not compile line #2, but accepts line #1.
It looks like it-s trying to unnecessarily let "double [3][4]" decay into
"double (*)[4]" (or whatever)

For the moment we'll stick to the pointer workaround, but can anyone try
VC8.0 as well?
Thanks
MH

//////////////////////////////////////////////////////////////////////////
#include <cstddef>

using std::size_t;

template <typename matrix_t>
struct matrix_traits
{
typedef typename matrix_t::value_type value_type;
typedef matrix_t copy_type;
};

template <typename scalar_t, size_t R, size_t C>
struct matrix_traits< scalar_t [R][C] >
{
typedef scalar_t value_type;
typedef value_type copy_type[R][C];
};

template <typename matrix_t>
size_t FAKE_rank(const matrix_t& m, const typename
matrix_traits<matrix_t>::value_type epsilon = 0)
{
typedef matrix_traits<matrix_t> traits_t;

typedef typename traits_t::copy_type copy_t;
copy_t A;
copy_t B;
copy_t* C = (epsilon>0) ? (&A) : (&B); // LINE #1

copy_t& D = (epsilon>0) ? (A) : (B); // LINE #2

return 0;
}

int main(int argc, char* argv[])
{
double b[3][4];
FAKE_rank(b);
return 0;
}

Igor Tandetnik

2005-05-30, 4:01 pm

"Mycroft Holmes" <m.holmes@nospam.it> wrote in message
news:eZD9COQZFHA.1088@TK2MSFTNGP14.phx.gbl
> Hi. The following sample is fine for Comeau Compiler.
> Howver VC7.1 does not compile line #2, but accepts line #1.
> It looks like it-s trying to unnecessarily let "double [3][4]" decay
> into "double (*)[4]" (or whatever)
>
> For the moment we'll stick to the pointer workaround, but can anyone
> try VC8.0 as well?

Looks like a bug, and it still produces an error with VC8 Beta 2:

..\Test.cpp(36) : error C2440: 'initializing' : cannot convert from
'double (*)[4]' to 'copy_t (&)'
.\Test.cpp(46) : see reference to function template
instantiation 'size_t FAKE_rank<double[3][4]>(const matrix_t (&),const
double)' being compiled
with
[
matrix_t=double [3][4]
]

Consider reporting it at http://lab.msdn.microsoft.com/productfeedback/
.. If you do, post a link here, I'll validate it.
--
With best wishes,
Igor Tandetnik

With sufficient thrust, pigs fly just fine. However, this is not
necessarily a good idea. It is hard to be sure where they are going to
land, and it could be dangerous sitting under them as they fly
overhead. -- RFC 1925

Mycroft Holmes

2005-05-30, 4:01 pm

----- Original Message -----
>
> Looks like a bug, and it still produces an error with VC8 Beta 2:
> . If you do, post a link here, I'll validate it.

thanks for the reply.
this should do:

http://lab.msdn.microsoft.com/Produ...ackId=FDBK28317

Tom Widmer

2005-06-01, 9:09 am

Mycroft Holmes wrote:
> Hi. The following sample is fine for Comeau Compiler.
> Howver VC7.1 does not compile line #2, but accepts line #1.
> It looks like it-s trying to unnecessarily let "double [3][4]" decay into
> "double (*)[4]" (or whatever)
>
> For the moment we'll stick to the pointer workaround, but can anyone try
> VC8.0 as well?

This simpler program exhibits the same bug:

int main()
{
double a[1];
double b[1];
double (&c)[1] = true ? a : b;
}

For some reason, it is doing an array-to-pointer conversion on the
result of the conditional expression. 5.16/4 makes it clear that this
shouldn't happen.

Tom

Sponsored Links