[cxx-abi-dev] comparing type_infos (was Re: Patch for throw/catch problems)

[Jason Merrill]

>>>>> "Jakub" == Jakub Jelinek <jakub at redhat.com> writes:

[ message delayed going out ]

> This should be no problem if NTBS are private to each library and
> type_infos unique accross the whole program, if NTBS' are only referenced
> from type_infos with the same name (Jason, does your "only pointers to
> incomplete types are represented differently" mean this condition is
> always satisfied?).

No; the symbol for the type_info for a pointer to incomplete class type is
mangled differently from that for a pointer to complete class type, but the
NTBS is shared.

> I don't know how could we make _ZTI symbols GLOBAL/WEAK while making
> relocations against them be resolved at static link time to that symbol,
> unless we special cased this in the linker, which is not a good idea IMHO.

Looking at the gas docs, it seems that .protected would do that.

I notice that the old HP ia64 EH document specifies that entries in the
type table should be GP-relative, which doesn't work if the type_info might
be coming from some other object; rth dealt with this for gcc by adding a
level of indirection.  I wonder what Intel and HP are doing; last we heard
from Intel, they were trying to use gprel.

Jason