This is the mail archive of the
mailing list for the binutils project.
ARC port broken reloc processing
- From: Alan Modra <amodra at gmail dot com>
- To: Andrew Burgess <andrew dot burgess at embecosm dot com>
- Cc: binutils at sourceware dot org
- Date: Fri, 27 Nov 2015 14:09:38 +1030
- Subject: ARC port broken reloc processing
- Authentication-results: sourceware.org; auth=none
- References: <1448580380-2724-1-git-send-email-andrew dot burgess at embecosm dot com> <20151127024411 dot GA8901 at bubble dot grove dot modra dot org>
On Fri, Nov 27, 2015 at 01:14:12PM +1030, Alan Modra wrote:
> Why allow zero for the base address, and the variant ranges?
I had a look. If it was for arc-elf, please don't weaken the test.
arc-elf reloc processing via howto is broken.
#define ARC_RELOC_HOWTO(TYPE, VALUE, RSIZE, BITSIZE, RELOC_FUNCTION, OVERFLOW, FORMULA) \
[TYPE] = HOWTO (R_##TYPE, 0, RSIZE, BITSIZE, FALSE, 0, complain_overflow_##OVERFLOW, arc_elf_reloc, #TYPE, FALSE, 0, 0, FALSE),
static struct reloc_howto_struct elf_arc_howto_table =
..results in all reloc howtos having 0 for dst_mask, ie. don't
actually apply the relocation. Clearly this needs to be fixed by
modifying arc-reloc.def to add the appropriate bits to be used in
Australia Development Lab, IBM