https://gcc.gnu.org/bugzilla/show_bug.cgi?id=116887
--- Comment #16 from Xi Ruoyao <xry111 at gcc dot gnu.org> ---
(In reply to chenglulu from comment #15)
> (In reply to chenglulu from comment #14)
> > (In reply to Xi Ruoyao from comment #13)
> > > Hmm I do think 2 is better. It seems we are just "reinventing" the GOT
> > > in 1.
> > >
> > > So OK with your approach if it passes regtest.
> >
> > I agree, I'll test the patch
>
> extern struct rtld_global _rtld_global;
> static struct dl_find_object_internal _dlfo_main __attribute__ ((section
> (".data.rel.ro")));
> extern void _dlfo_process_initial_noncontiguous_map (void);
> void
> _dlfo_process_initial (void)
> {
> if ( _dlfo_main.map == &_rtld_global._dl_rtld_map)
> _dlfo_process_initial_noncontiguous_map ();
> }
>
>
> I've been looking at this code for the past two days, and I think the second
> modification is incomplete. The reason for the problem is:
>
> If decls are put in ".data.rel.ro" or ".data.rel.ro.local" using the
> __attribute__ construct （_dlfo_main）, then the obtained section flags
> contain only SECTION_WRITE and not SECTION_RELRO.
> But labels are placed in the constant pool ("data.rel.ro" or
> "data.rel.ro.local") (_rtld_global._dl_rtld_map), the section flags contain
> both SECTION_WRITE and SECTION_RELRO.
> Since two sections are the same, but the flags are set differently, then can
> cause a section conflicts.
So can we avoid putting labels into constant pool? If not we'll have to use
patch 1 like hppa.
