Re: [PATCH] Implement x86 SIZE32/SIZE64 relocations

["H.J. Lu" <hjl dot tools at gmail dot com>]

On Thu, Jan 17, 2013 at 9:52 AM, Jakub Jelinek <jakub@redhat.com> wrote:
> On Thu, Jan 17, 2013 at 09:46:37AM -0800, H.J. Lu wrote:
>> On Thu, Jan 17, 2013 at 9:25 AM, Jakub Jelinek <jakub@redhat.com> wrote:
>> > On Thu, Jan 17, 2013 at 09:10:28AM -0800, H.J. Lu wrote:
>> >> On Thu, Jan 17, 2013 at 8:53 AM, Jakub Jelinek <jakub@redhat.com> wrote:
>> >> > On Wed, Jan 16, 2013 at 08:32:54PM -0800, H.J. Lu wrote:
>> >> >> I checked x86 size relocation support into binutils and
>> >> >> pushed it to master branch in glibc.
>> >> >
>> >> > Please revert it, this should have never been a dynamic relocation.
>> >> >
>> >>
>> >> That is not true.  Please see binutils size relocation linker tests.
>> >> 2 tests scan for run-time size relocation.
>> >
>> > You've added a feature that is not really useful for asan, without
>> > discussing it first.  So we just end up with dead code in both glibc and
>> > gcc.
>>
>> First, they aren't dead code. Secondly,  they are specified in the
>> psABI:
>>
>> http://web.archiveorange.com/archive/v/KGyKyS8JaGMawGhL882p
>
> They were added for a different OS.  If they are in psABI, they can be kept
> in elf.h, but there is no point to slow down the dynamic linker to handle
> something that won't be ever used.

I disagree.  Address sanitizer doesn't use it doesn't mean it isn't useful.

> I'm sorry I haven't responded immediately to your changes (but it wasn't
> obvious that you are actually intending to submit this to binutils/glibc
> without actually first checking whether what you've implement can be useful
> for something), but it was a new feature, something not appropriate for
> GCC 4.8 at that point and thus work on GCC 4.8 stabilization took
> precedence.
>

It is never intended for GCC 4.8 due to dependency on glibc and
binutils.


-- 
H.J.