> I believe the problems were only with shlibs, and that _programs_ are not
> affected.

Since only `base relative' relocations are involved - and these occur
only in shared library relocation entries - that is correct.

If indeed these hacks only take care of SET* symbols (but I'm not
sure that's the case) then only shared C++ libraries are involved
(counting only our release sets).

In any case, the impact of breaking compatibility seems much less
severe if only a set of shared libraries need to be replaced along
with the run-time linker.

-pk