This is the mail archive of the
binutils@sources.redhat.com
mailing list for the binutils project.
Re: PATCH: Minor ELF32/ia64 cleanup
- From: James E Wilson <wilson at specifixinc dot com>
- To: "H. J. Lu" <hjl at lucon dot org>
- Cc: binutils at sources dot redhat dot com
- Date: Tue, 03 May 2005 14:06:41 -0700
- Subject: Re: PATCH: Minor ELF32/ia64 cleanup
- References: <20050503192121.GA17638@lucon.org>
On Tue, 2005-05-03 at 12:21, H. J. Lu wrote:
> * elfxx-ia64.c (ARCH_SIZE): New.
> (LOG_SECTION_ALIGN): Likewise.
> (elfNN_ia64_create_dynamic_sections): Use LOG_SECTION_ALIGN to
> align relocation sections.
> (get_fptr): Likewise.
> (get_reloc_section): Likewise.
> (elfNN_ia64_tprel_base): Likewise.
> (elfNN_ia64_check_relocs): Support 32bit relocations.
> (allocate_global_fptr_got): Likewise.
> (allocate_dynrel_entries): Likewise.
> (set_got_entry): Likewise.
> (set_pltoff_entry): Likewise.
> (elfNN_ia64_relocate_section): Likewise.
Thanks. This does look like a useful improvement.
There seems to be a few oddities here, e.g. a FPTR64 relocation
presented to a 32-bit toolchain will create a FPTR32 dynamic relocation
for it, and similarly a FPTR32 relocation presented to a 64-bit
toolchain will create a FPTR64 dynamic relocation for it. This is a bit
odd, but I suspect these won't occur in practice. And this is an
improvement over what we had before, since at least we correctly get
FPTR32 with a 32-bit toolchain which we didn't get before.
Do you have access to ia64-hpux to double check any of this? ia64-hpux
should presumably be considered the ILP32 reference, and GNU ld should
be behaving the same as the hp linker. We shouldn't be creating our own
32-bit standard, at least not without an extremely good reason.
--
Jim Wilson, GNU Tools Support, http://www.SpecifixInc.com