Re: [patch] Set calling convention of methods

[Jonas Maebe <jonas dot maebe at elis dot ugent dot be>]

On 30 Sep 2009, at 02:02, Joel Brobecker wrote:

> > I've attached my first patch. Other patches will modify code in this
> > one, so I'd prefer to get this one out of the way first.
>
> But it looks like this patch actually introduces code that will be  
> dead
> until you actually set the calling convention, right? It seems strange
> that you'd prefer to do it this way.

It will already be active for regular functions/procedures, because  
the DWARF reader does propagate the calling convention in that case  
(as it was already done for the DW_CC_GNU_renesas_sh calling  
convention). It simply won't be taken into account yet for methods  
(because that requires a patch to the DWARF reader), that will indeed  
require a further patch.

> One of the concerns that never got resolved from what I've read in the
> archives, was the use of a DWARF constant outside of DWARF code.
> I am not sure I understand the problem, though. Was it the use of
> constant zero when populating this field when reading stabs debug
> info, or anything else?

Yes, it was the use of an unnamed constant: http://sourceware.org/ml/gdb-patches/2009-04/msg00196.html

> As far as I am concerned, I can't see a problem with using DWARF
> declarations even from stabs.

We could include the dwarf2.h header in the stabs reader and set the  
calling convention to DW_CC_normal in all cases.

> Just a couple of formatting nits:
>
> > +      if (struct_return
> > +	  && reg_paras[para_regnum-1] != nargs)
>
> and
>
> > +  while (type
> > +	 && can_dereference (type))
>
> You probably want to join the two lines in one. gdb_indent.sh, our
> automatic indentation program would (though no one uses it, it makes
> pretty bad choices sometimes). I think it'd make the code a little
> easier to read too.

Ok, will do.

> diff --git a/include/elf/dwarf2.h b/include/elf/dwarf2.h
> > index a7448dc..efa786e 100644
> > --- a/include/elf/dwarf2.h
> > +++ b/include/elf/dwarf2.h
> > @@ -662,7 +662,8 @@ enum dwarf_calling_convention
> >     DW_CC_normal = 0x1,
> >     DW_CC_program = 0x2,
> >     DW_CC_nocall = 0x3,
> > -    DW_CC_GNU_renesas_sh = 0x40
> > +    DW_CC_GNU_renesas_sh = 0x40,
> > +    DW_CC_GNU_borland_fastcall_i386 = 0x41
>
> This part is maintained by binutils, I believe.  You'll need
> to ask them for approval of this change.

Tom said it came from gcc (http://sourceware.org/ml/gdb-patches/2009-04/msg00063.html 
) and I did submit a patch there: http://gcc.gnu.org/ml/gcc-patches/2009-04/msg00301.html 
 . I did not get any reaction to that patch (and I guess it's not  
been applied).

Later on, Tom clarified that he thought that the gdb and gcc versions  
of dwarf2.h should actually be merged into a single copy, but that I  
shouldn't worry about this since the divergence started before my  
patch: http://sourceware.org/ml/gdb-patches/2009-04/msg00099.html


Jonas