Fix Itanium vtable construction

Our use of bridges::cpp_uno::shared::VtableFactory::Slot to model all the
elements of a vtable is an abstraction that doesn't quite match the reality of
<https://itanium-cxx-abi.github.io/cxx-abi/abi.html>, as vtables are not
homogenous sequences of function pointers, but are rather a mix of offsets, data
pointers, and function pointers.

The data preceding the virtual table address point is the offset to top (an
offset) followed by the typeinfo pointer (a data pointer).  On other platforms
where offsets, data pointers, and function pointers are all of the same size, we
model those as two additional Slots at index -2 and -1, resp.  On Itanium, where
function pointers (and thus Slots) are twice the offset and data pointer size,
we should model those as one additional Slot at index -1.

The code has been this way ever since its introduction in
0c25631972809c752624b4883c71671c8e83e797 "INTEGRATION: CWS ia64port01_DEV300
(1.1.2); FILE ADDED".  It should never have caused any issues as the existence
of the excess Slot at index -2 should never have gotten in the way.  But it is
probably better to clean this up anyway.  (This is a blind fix, as I don't have
an Itanium build environment available.  And this Itanium bridge may well be
dead code by now, anyway.)

Change-Id: I98d58785c054e1cdc621e5968c41b06d8788998f
Reviewed-on: https://gerrit.libreoffice.org/c/core/+/145035
Tested-by: Jenkins
Reviewed-by: Stephan Bergmann <sbergman@redhat.com>

0 files changed, 0 insertions, 0 deletions