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: gcc corrections for better pie support


Peter S. Mazinger wrote:
> On Tue, 9 Nov 2004, Daniel Jacobowitz wrote:
> 
> > On Tue, Nov 09, 2004 at 10:32:49AM -0800, Zack Weinberg wrote:
> > > >> > +%{fno-PIC|fno-pic:-U__PIC__ -U__pic__} %{pthread:-D_REENTRANT}"
> > > >> 
> > > >> This should not be necessary; giving -fno-PIC/-fno-pic should cause
> > > >> flag_pic to be off, and thus the symbols will never get defined in the
> > > >> first place.  If this is not so, then that is itself a bug that needs
> > > >> fixing.
> > > >
> > > > on mips builtin_define __PIC__ and __pic__ are not ifdef'd by flag_pic 
> > > > (probably because it's a PIC arch) so I think the disable logic won't work
> > > > see gcc/config/mips/linux.h line 59/60.
> > > 
> > > Well, like I said, that is itself a bug, and you need to address the
> > > problem there - best is to leave MIPS alone for now and consult with
> > > the MIPS maintainers (see the MAINTAINERS file at top level of the
> > > source tree for their names) how to straighten it all out.
> > 
> > It isn't a bug.  MIPS Linux code is always PIC.
> is than the above line beginning w/%{fno-PIC... necessary?

MIPS Linux Userland code is always PIC, while the Kernel is always non-PIC.
So defining __PIC__ / __pic__ unconditionally seems to be a (until now
harmless) bug.


Thiemo


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