Re: [sigc] Re: [Bug 147391] - test_lambda does not compile with SUN Forte 5.5

[Martin Schulze <martin-ml hippogriff de>]

Hi Damien!

Sorry for the break - Now I have time to look at this error. Obviously  
it occurs in the compatibility module, the reason being that the FORTE  
does not handle the "const" keyword correctly when building a slot from  
a const member method.

However, the implementation causing the error does not differ from  
libsigc++-1.2's so most likely slots referring to const member methods  
didn't work with the FORTE before. Therefore, the compatibility module  
doesn't need to solve this non-trivial problem.

Instead, I will remove the const member method test from  
test_compatibility.cc and upload to:

 http://www.hippogriff.de/libsigc++-2.0.4-test5.tar.gz

Please try this one!

Regards,

 Martin


Am 11.08.2004 11:03:08 schrieb(en) Damien Carbery:
> /sgnome/tools/sparc-solaris/forte/SOS8/SUNWspro/bin/CC  -I. -I. -I..
> -I.. -I..    -g -c -o test_compatibility.o test_compatibility.cc
> "../sigc++/object_slot.h", line 175: Error: Could not find a match  
> for
> sigc::bound_mem_functor1<void, foo, const std::string
> &>::bound_mem_functor1(const foo, void(foo::*)(const std::string
> &)const).
> "test_compatibility.cc", line 57:     Where: While instantiating
> "SigC::slot<void, const std::string &, foo, foo>(const foo&,
> void(foo::*)(const std::string &)const)".
> "test_compatibility.cc", line 57:     Where: Instantiated from
> non-template code.
> 1 Error(s) detected.
> make[2]: *** [test_compatibility.o] Error 1
> make[2]: Leaving directory `/jds/libsigc/libsigc++-2.0.4/tests'
>
>
> FYI: I used tarball test4.
> I work with a clean directory each time, removing the old stuff  
> first.
>
> Damien
>
> ----- Original Message -----
> From: Martin Schulze <martin-ml hippogriff de>
> Date: Tuesday, August 10, 2004 6:24 pm
> Subject: Re: [sigc] Re: [Bug 147391] - test_lambda does not compile
> with SUN Forte 5.5
>
> > Am 10.08.2004 18:09:36 schrieb(en) Damien Carbery:
> > > make failed.
> > >
> > > To be sure I have the correct tarball:
> > > $ cksum libsigc++-2.0.4-test3.tar.gz
> > > 761559274 1767826 libsigc++-2.0.4-test3.tar.gz
> >
> > The tarball is correct. I don't understand why the FORTE doesn't
> > choose
> > the template specialization:
> >
> >  template <class T_type>
> >  struct convert_array
> >  { typedef T_type& type; };
> >
> >  template <class T_type, int N>
> >  struct convert_array<T_type[N]>
> >  { typedef T_type* type; };
> >
> >  [...]
> >
> > 'convert_array<const T_type>::type' (with T_type='char(&)[2]')
> > gives
> > 'const char(&)[2]' instead of 'const char*'. gcc works as expected.
> >
> > This is unnerving! Especially since glibmm uses the same trick to
> > distinguish between array types and other types.
> >
> > Probably there is a compiler bug. I will upload a tarball that has
> > the
> > offending lines in test_lambda.cc disabled so that we can move on.
> > Please try:
> >
> >  libsigc++-2.0.4-test4.tar.gz
> >
> > in two hours or so.
> >
> > Regards,
> >
> >  Martin
> >