[omniORB] G++ 12.2 complains with stringop-overflow warning

[Thomas Braun]

Am Montag, dem 05.06.2023 um 12:04 +0100 schrieb Duncan Grisby:
> On Thu, 2023-06-01 at 19:59 +0200, Thomas Braun via omniORB-list
> wrote:
> 
> 
> > one of the newer gcc versions complains about some omniORB code.
> > I'm attaching the files needed for reproduction.
> 
> [...]
> > In file included from /usr/include/omniORB4/omniInternal.h:44,
> >                  from /usr/include/omniORB4/CORBA.h:63,
> >                  from test.cpp:1:
> > In member function ‘void
> > _CORBA_Sequence<T>::copybuffer(_CORBA_ULong)
> > [with T = unsigned char]’,
> >     inlined from ‘void _CORBA_Sequence<T>::length(_CORBA_ULong)
> > [with
> > T
> > = unsigned char]’ at
> > /usr/include/omniORB4/seqTemplatedecls.h:59:12,
> >     inlined from ‘int main(int, char**)’ at test.cpp:8:10:
> > /usr/include/omniORB4/seqTemplatedecls.h:211:17: error: writing 8
> > bytes
> > into a region of size 1 [-Werror=stringop-overflow=]
> >   211 |       newbuf[i] = pd_buf[i];
> >       |       ~~~~~~~~~~^~~~~~~~~
> 
> I have seen those complaints too. gcc is wrong. I don't know why it
> thinks the region has size 1, because it plainly does not. There is
> no
> error in the code.

I knew it! Thanks for the confirmation.

> 
> [...]
> > In addition I do see some mixing of unsigned long and _CORBA_ULong
> > in
> > seqTemplatedecls.h/copybuffer which I find confusing as IIRC you
> > can't interchangebly use these types in general.
> 
> That is a historical accident, but it does no harm because unsigned
> long is always at least as big as _CORBA_ULong, and the way the
> values
> are used, they cannot overflow. 

Good to know!