Boost logo

Boost Users :

From: Frank (opcode_at_[hidden])
Date: 2005-11-04 14:47:52


John Maddock <john <at> johnmaddock.co.uk> writes:

>
> > I 'm trying to put the implementation into a dll and mark the explicit
> > instantiation of Normal as exported with the following results:
> > 1) With the small_dim_ctor's enabled I get the static assertion when
> > marked for export
> > 2) With the small_dim_ctor's enabled everything is ok when not marked
> > for export
> > 3) With the small_dim_ctor's disabled everything is ok regardless of
> > export marking
> > 4) If I comment out the BOOST_STATIC_ASSERT and build my app with
> > export enabled I can use the resulting dll without any problems
> >
> > I'm thinking that marking for export is somewhow generating an
> > implicit conversion somewhere but I don't know how.
>
> Unfortunately, when you explicitly instantiate a template every non-template
> member function gets instantiated, and that triggers the static asserts.
>
> The only workarounds I can see are:
>
> 1) Use regular asserts whenever the class is being exported rather than
> static asserts.
> 2) Use partial template specialisations for the N = 1, 2 or 3 cases and give
> them different constructors (no need to static asserts then).
> 3) Use template constructors instead:
>
> template <class U>
> explicit Vector( U x )
> {
> BOOST_STATIC_ASSERT(DIM==1);
> data_[0]=x;
> }
>
> Hope this helps, John.
>

I thought I was only marking the explicit specialization of a single method (
eg Normal) for exporting or maybe it's because its not possible to specialize
just one method.?

After reading your reply I had a look at enable_if but I couldn't figure a way
to test the DIM
ie

template <typename T>
explicit Vector( T x, typename boost::enable_if< DIM==1, void*>::type = 0 )
 {
 data_[0]=x;
 }

but I'm not sure I understand how to use it and am still surfing for examples.

Thanks for the help.

Frank


Boost-users list run by williamkempf at hotmail.com, kalb at libertysoft.com, bjorn.karlsson at readsoft.com, gregod at cs.rpi.edu, wekempf at cox.net