From help-octave-request at bevo dot che dot wisc dot edu  Tue Jan 12 10:16:13 1999
Subject: Re: compile error on SGI using egcs 1.1.b
From: "John W. Eaton" <jwe at bevo dot che dot wisc dot edu>
To: Brian Tyrrell <tyrrell at ll dot mit dot edu>
Cc: Daniel Heiserer <Daniel dot Heiserer at bmw dot de>, help-octave@bevo.che.wisc.edu
Date: Tue, 12 Jan 1999 10:16:07 -0600 (CST)

On 12-Jan-1999, Brian Tyrrell <tyrrell at ll dot mit dot edu> wrote:

| I had a similar problem.  Take a look at the help-octave and bug-octave
| archives for ~Dec 15.
| 
| What finally worked for me was using Mumit Khan's patch in 1998/270.  Some
| of the patch appeared to have already been implemented, but part of it was
| not.

I assume you mean that the template instantiations are no longer
implemented as Mumit suggested.  That's because the original code was
like this:

  template <class T1, class T2>
  bool
  mx_leftdiv_conform (T1 a, T2 b)
  {
    ...
  }

  template bool mx_leftdiv_conform (const Matrix&, const Matrix&);

which was apparently incorrect.  Mumit suggested changing the
instantiation to be

  template bool mx_leftdiv_conform (Matrix, Matrix);

which worked for egcs, but I really wanted the arguments to
mx_leftdiv_conform to be const reference objects.  So I modified the
template definition to be

  template <class T1, class T2>
  bool
  mx_leftdiv_conform (const T1& a, const T2& b)
  {
    ...
  }

and made the template instantiation look like this:

  template bool mx_leftdiv_conform (const Matrix&, const Matrix&);

which is correct as far as I can tell.  It works for me with egcs
1.1.1, gcc 2.8.1, and gcc 2.7.2.

If it fails for others, can someone please explain what is going on,
and what the correct fix is?

Thanks,

jwe