This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: Fix notice_global_symbol


On Sat, Oct 11, 2003 at 07:25:51PM +0200, Jan Hubicka wrote:
> Still there is problem on whether we want to allow global symbol
> overwriting in C++.  WHen allowed, we theoretically can get match on
> GLOBAL_* that is definitly not wanted...

*shrug* I doubt this is really necessary.  Reasonable targets
don't export these symbols at all.  Virtually all object file
formats are capable (notable exceptions are a.out and ecoff),
though not all use it.

Those targets that do not allow GLOBAL_ symbols to be static
in the object file have two options: static linking or dynamic
linking that binds names at link time (a-la windows), or linker
options to hide the symbols (ecoff).

Any that are left over... well, there's not much that can be
done at all.  You Lose.

> Nice would be to relax binds_local_p :) as C++ mandates quite strong
> assumptions about symbol sharing.

No, binds_local_p describes the dynamic linker.  Anything else
would be a bug.


r~


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]