templ decl for non-templ ?

[Kurt Garloff <garloff at hft dot e-technik dot uni-dortmund dot de>]

Hi,

here comes a message that I don't understand ...

garloff@marvin:/home/plasma/numerix/lib/test > egcc -c bug1.cc 
 bug1.cc: In method `class tensor<T,(rank + r1)> tensor<T,rank>::operator*
  (const class tensor<T,r1> &) const': 
 bug1.cc:18: template definition of non-template `class tensor<T,(rank+r1)>
  tensor<T,rank>::operator *(const class tensor<T,r1> &) const'
garloff@marvin:/home/plasma/numerix/lib/test > egcc -v
 Reading specs from /usr/local/lib/gcc-lib/i586-pc-linux-gnulibc1/egcs-2.90.21/specs 
 gcc version egcs-2.90.21 971202 (egcs-1.00 release) 

------------------------------------------------------------------
/* bug1.cc, 18.12.97 by garloff@hft.e-technik.uni-dortmund.de */

template <typename T, unsigned rank>
class tensor
{
 protected:
   T* data;
 public:
   tensor () : data (0) {};
   tensor (const tensor<T, rank>& t) : data (t.data) {};
   
   template <unsigned r1>
     tensor <T, rank + r1> operator * (const tensor <T, r1>& ) const;
};

template <typename T, unsigned rank, unsigned r1>
tensor<T, rank+r1> tensor<T, rank>::operator * (const tensor <T, r1>& t1) const
{
   tensor <T, rank+r1> tr;
   return tr;
};
     
int main ()
{
   tensor <double, 2> t1; tensor <double, 3> t2;
   tensor <double, 5> (t1 * t2);
};
------------------------------------------------------------------------------
AFAIK, the declaration in ll.12/13 declares a templated operator, which
should be possible. The definition in ll.18 ff matches the declaration.
So why is there an error message? Isn't it allowed to do computations
(rank + r1) in template parameters? Hope somebody can fix the bug in egcs
or in my code.


Kurt Garloff
garloff@hft.e-technik.uni-dortmund.de

------------------------------------------------------------------------------
P.S: I found another bug ...
When compiling some code with deeply nested templated classes, I got error
messages from the assembler. -fname-mangling-version-0 fixed it.
I'm gonna submit a full bug-report, when I managed to isolate it.
Here some details: (Maybe someone can figure out if it's the compiler or
the assembler that is wrong here.)
(The file tensortest.s was created by egcc -O3 -S tensortest.cc)

[dortmund@etpvis test]$ as tensortest.s
tensortest.s: Assembler messages:
tensortest.s:7911: Error: Rest of line ignored. First ignored character is `*'.
tensortest.s:7912: Error: ignoring unrecognized symbol type ""
tensortest.s:7912: Error: Rest of line ignored. First ignored character is `*'.
tensortest.s:7913: Error: invalid character '_' in opcode
tensortest.s:10955: Error: expected comma after name
`contract__H2ZdUi5_RCt6Tensor2ZX01UiY11UiUi_t6Tensor2ZX01Ui' in .size directive
tensortest.s:10955: Error: Rest of line ignored. First ignored character is `*'.
--------
tensortest.s:7909
.section .gnu.linkonce.t.contract__H2ZdUi5_RCt6Tensor2ZX01UiY11UiUi_t6Tensor2ZX01Ui *6,"ax",@progbits
        .align 4
        .weak   contract__H2ZdUi5_RCt6Tensor2ZX01UiY11UiUi_t6Tensor2ZX01Ui *6
        .type   contract__H2ZdUi5_RCt6Tensor2ZX01UiY11UiUi_t6Tensor2ZX01Ui *6,@function
contract__H2ZdUi5_RCt6Tensor2ZX01UiY11UiUi_t6Tensor2ZX01Ui *6:
        pushl %ebp
--------
tensortest.s:10954
.Lfe16:
        .size    contract__H2ZdUi5_RCt6Tensor2ZX01UiY11UiUi_t6Tensor2ZX01Ui *6,.Lfe16-contract__H2ZdUi5_RCt6Tensor2ZX01UiY11UiUi_t6Tensor2ZX01Ui *6
.section        .rodata
.LC38:
        .string "Swap identical indices"
--------
garloff@marvin:/home/garloff > as -v
GNU assembler version 971211 (i586-linux), using BFD version 971211