This is the mail archive of the
binutils@sourceware.org
mailing list for the binutils project.
Re: C6x gas fix for GOT relocs
- From: "Joseph S. Myers" <joseph at codesourcery dot com>
- To: Bernd Schmidt <bernds at codesourcery dot com>
- Cc: binutils at sources dot redhat dot com
- Date: Thu, 23 Sep 2010 18:56:29 +0000 (UTC)
- Subject: Re: C6x gas fix for GOT relocs
- References: <4C9B8716.6050007@codesourcery.com>
On Thu, 23 Sep 2010, Bernd Schmidt wrote:
> Here's a patch that adds a tic6x_fix_adjustable function to the C6x
> assembler. Like in several other targets, we must prevent the assembler
> from creating GOT relocs with an offset, e.g. for "const.str.1.1 + n",
> rather than just for a symbol. Strictly speaking this isn't necessary
> just yet as the linker has no support for these relocs, but I'm working
> on that.
>
> Is the patch ok? Does it need an extra assembler testcase? This
> failure shows up numerous times in the gcc testsuite.
Adding this for the GOT relocations is OK; there should be an assembler
test added covering each of the affected relocations. I don't believe any
checks for BFD_RELOC_VTABLE_INHERIT and BFD_RELOC_VTABLE_ENTRY should be
added to this assembler port; there are no corresponding C6X ELF
relocations so those BFD relocations (and the .vtable_inherit and
.vtable_entry assembler directives) are useless for this target, and those
directives were to support the GCC -fvtable-gc option which was removed in
GCC 3.4 so it would hardly make any sense to add support for them now.
--
Joseph S. Myers
joseph@codesourcery.com