Re: [GOLD][PATCH] Set ARM ABI information in ELF file header.

[Viktor Kutuzov <vkutuzov at accesssoftek dot com>]

Good work, Doug!
Are you going to parse input objects ".ARM.attributes" section as well?

-Viktor

----- Original Message ----- 
From: "Doug Kwan (Ãö®¶¼w)" <dougkwan@google.com>
To: "Joseph S. Myers" <joseph@codesourcery.com>
Cc: "Ian Lance Taylor" <iant@google.com>; "binutils" <binutils@sourceware.org>; "Viktor Kutuzov" <vkutuzov@accesssoftek.com>
Sent: Tuesday, October 27, 2009 9:12 AM
Subject: Re: [GOLD][PATCH] Set ARM ABI information in ELF file header.


> Wow.  This makes gold bug-compatible with ld :).  I can remove that.
> Thanks for catching it.
>
> -Doug
>
> 2009/10/27 Joseph S. Myers <joseph@codesourcery.com>:
> > On Mon, 26 Oct 2009, Doug Kwan (Ãö®¶¼w) wrote:
> >
> > > Hi,
> > >
> > >     This patch adds code to set proper flags in the processor-specific
> > > flags in the ELF file header.  Previously we always set e_flags to 0
> > > on all target.  This patch add methods processor_specific_flags and
> > > set_processor_specific_flags in Target so that Output_file_header can
> > > get the flags.   For targets that merge processor-specific flags from
> > > input objects,  we also track whether the flags have been set at least
> > > once.  Currently only the ARM target sets the flags correctly.
> >
> > It looks like this patch is replicating the GNU ld bug whereby
> > EF_ARM_HASENTRY is set on EABI objects.  There is no such flag in AAELF
> > and so it is a bug to set it on such objects.
> >
> > --
> > Joseph S. Myers
> > joseph@codesourcery.com