[omniORB] Port of omniORB_280pre1 to Compaq C++ 6.2 OpenVMS Alpha 7.1

Bruce Visscher visschb@rjrt.com
Wed, 07 Jul 1999 09:32:46 -0400



Sai-Lai Lo wrote:

> Bruce,
>
> I'll check the _0RL_buildDesc_cstring to make sure other compilers are not
> upset with the change. Sigh!
>

Yes, I started to #ifdef the change, but since I think the change is a
correction I didn't.  I can imagine that it might confuse some compilers
(really, it shouldn't: the change makes data._ptr a const pointer, not a
pointer to const).

>
> > BTW, is the testsuite that used to reside in src/tests/omniORB2_testsuite
> > no longer supported?  If so then we should remove this from the VMS
> > distribution.
>
> It is not the intention to remove the testsuite permanently. At the moment
> it needs some cleanup. I'll try to make it available later.
>

Thanks.

>
> > Unfortunately, it looks like porting 280pre1 to DEC C++ 5.6 is going to
> > take some work.
>
> Sorry about that, I've attempted to clean up the internal template and most
> certainly removed workarounds for gcc 2.7.2. This may have affected DEC C++
> 5.6 as well. The main change that could affect 5.6 is to put back those
> template functions defined in <top>/include/omniORB2/templatedefns.h back
> into inline functions in <top>/include/omniORB2/templatedecls.h.
>
> I appreciate if you spend some time on figuring out what workarounds are
> needed for 5.6.
>

The main problem I'm having at the moment is that for some reason:

template <class _T>
class someclass {
  _T*& out();
};

gives the compiler gas.  Unfortunately, this looks like it will require a
change to omniidl2 as well (I've added a HAS_POINTER_REF_BUG macro to
CORBA_sysdep.h).

I'd like to be able to forget about DEC cxx 5.6, but I can't since that's all
that's available for OpenVMS VAX.

Bruce