[omniORB] Announce: omniORB 4.1.0 and omniORBpy 3.0 release candidates

[Duncan Grisby]

On Thursday 19 October, Thomas Lockhart wrote:

> I'm including patches to fix some compiler warnings about non-virtual
> destructors, or rather about classes which have virtual methods but do
> not have a virtual destructor defined.

Thanks. However, it's by design that those classes have non-virtual
destructors. They are always created on the stack or as global statics,
so they are always deleted by most derived class. To give them virtual
destructors just adds code along the call chain for no reason other than
to suppress a compiler warning.

> There are additional patches in the attached patch file which fix a
> few compiler warnings about possible uninitialized variables. You may
> or may not want to actually apply those, but if so I can give you more
> patches for more of those cases if so.

I've applied your patches, but I'm not particularly inclined to go on a
big crusade to remove all such warnings. There are no cases where the
variables really are used uninitialised -- gcc just isn't clever enough
to work it out.

> Also, I see warnings from omniidl which were not present in earlier
> versions:
> 
> omniidl -bpython  -I/usr/share/idl/omniORB test.idl
> corbaidl.idl:15: Warning: Forward declared interface 'CORBA::IDLType'
> was never fully defined
> omniidl: 1 warning.
> 
> for cases where #include <orb.idl> is present in my IDL (used to get
> access to typecode definitions in TAO; it was formerly a no-op for
> omniORB afaicr).

I've fixed that by suppressing the omniidl warning about unresolved
forward declared interfaces for interfaces in the CORBA module.

Cheers,

Duncan.

-- 
 -- Duncan Grisby         --
  -- duncan at grisby.org     --
   -- http://www.grisby.org --