This is the mail archive of the gcc@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]

Re: A completely different approach to EH runtime


On Fri, Feb 23, 2001 at 12:21:05AM -0800, Richard Henderson wrote:
> On Wed, Feb 21, 2001 at 10:05:57PM -0300, Alexandre Oliva wrote:
> > It appears to me that the only problem right now is that the DWARF-2
> > EH table (objects, in frame-dwarf2.c) is currently static.  Couldn't
> > we just make it global, so that ELF symbol-binding semantics resolves
> > it to a single symbol globally?
> 
> No.  Think: the data symbols won't resolve as you wish for
> the exact same reasons that the function symbols didn't.

If we make it common?  Linkonce?  I know there's a way to get exactly
one copy of a data symbol in ELF no matter how many shared libraries
define it.  The new C++ ABI uses it for RTTI.  I just don't know
what it's called.

Are we willing to break the language-independent EH ABI, such that
no EH-using code compiled with 2.x will interoperate with 3.0?
If we are, we have more options available.  I *think* the only
widely distributed EH-using code which is not C++ and therefore
breaking anyway, is glibc.  Unfortunately that's a big "only".

zw


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