SHT_GROUP Support in binutils

Jakub Jelinek jakub@redhat.com
Tue Jan 8 02:08:00 GMT 2002


On Tue, Jan 08, 2002 at 10:21:15AM +1030, Alan Modra wrote:
> On Mon, Jan 07, 2002 at 01:03:14PM -0800, Jessica Han wrote:
> > Alan,
> >   I've noticed that you added SHT_GROUP support into binutils back in Oct
> > 2001. Could you please tell me what else is on your list in this area? Are
> > you going to provide complete COMDAT support specified by IA64-ABI?
> 
> Yes, although it may be a while before I do so.  I didn't do it at first,
> because I wasn't certain of the requirements.  If you want to contribute
> some code, please go ahead!
> 
> > It seems that gas relies on the assembly file to specify what sections form
> > a group and what's the name of the group, are those information going to be
> > generated by GCC? If so, do you know if anybody is working in this problem?
> 
> I started to make changes to gcc to emit the information, mainly with the
> idea of splitting up dwarf debugging info and grouping it with a function's
> code and data.  That fails because of an unfortunate choice in the sign
> of one of the dwarf offset bytes - see
> http://gcc.gnu.org/ml/gcc-bugs/2001-12/msg00314.html
> 
> I haven't done anything since then, as it appears gas will need a new
> special reloc to handle the subtraction of two syms in dwarf data.
> Current diff attached.  Watch out for the dwarf_cie change!  That's just
> a temporary hack.

Why is it needed when ld now optimizes .eh_frame, which includes throwing
FDEs away if they are against discarded section?
Either you can emit full .eh_frame sections for each group (including CIE;
ld will remove duplicate CIEs), or you can forget about it altogether and
always use just one .eh_frame.

	Jakub



More information about the Binutils mailing list