[RFC] unify dynamic_symbol_p implementations
Bob Wilson
bwilson@tensilica.com
Fri Jul 18 21:19:00 GMT 2003
Yup, I was missing something :-)
Maybe it is still OK, though. It doesn't end up in .rodata. GCC's
default_elf_select_rtx_section() function in varasm.c puts it in
.data.rel.ro. I had never even heard of this section before, but it seems to
exist for something like this case.
Help?
On Friday 18 July 2003 13:56, Richard Henderson wrote:
> On Fri, Jul 18, 2003 at 01:09:14PM -0700, Bob Wilson wrote:
> > The R_XTENSA_32 reloc used for the function address turns into an
> > R_XTENSA_GLOB_DAT reloc in the executable. The runtime linker
> > resolves this reloc at startup. Am I missing something?
>
> Really? Even with
>
> extern void foo(void);
> static void (* const bar) (void) = foo;
>
> Doesn't bar wind up in .rodata? And if so, it really shouldn't
> have a dynamic relocation against it.
More information about the Binutils
mailing list