This is the mail archive of the binutils@sources.redhat.com mailing list for the binutils project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [PATCH/RFA] Mark arm-*-netbsdelf* binaries as ELFOSABI_NETBSD


Hi Guys,

> > > EI_OSABI without introducing any new values.  You can take ELFOSABI_ARM
> > > to mean your existing ABI, and assume that at the point binutils adopts
> > > an "official" ARM EABI it will start writing EI_OSABI as zero. 
> > 
> > Is this was binutils is going to do?  
> 
> I believe so, yes.

I second this.


> From: Richard Earnshaw <rearnsha@arm.com>
> 
> Hmm, I think it would be far better to set EI_OSABI to ELFOSABI_NETBSD 
> now, thus indicating that these files don't really conform to the EABI.  
> Then, when the transition to the conforming ABI happens, we can set the 
> field to 0, or whatever the ARM EABI mandates.

> From: Philip Blundell <pb@nexus.co.uk>
> 
> What does that gain you compared to leaving them as ELFOSABI_ARM?

I think I agree with Phil on this one.  We are currently using
ELFOSABI_ARM to distinguish between an old implementation of the ARM
relocs and the current implementation.  Neither of these
implementations fully conforms to the yet-to-be-finalised EABI for
ARM.  Once the EABI is published and binutils has been fixed to follow
it, we ought to reset EI_OSABI to 0, dropping support for old-abi
altogether, and then use this to distinguish between pre-EABI
conforming binaries and post-EABI conforming binaries.

NetBSD can use it for the same purpose.  If EI_OSABI is set to
ELFOSABI_ARM then it means that the binary is conforms to today's
NetBSD implementation.  If it is 0, then it conforms to tomorrow's
implementation.



> From: Richard Earnshaw <rearnsha@arm.com>
> 
> > > BTW, who allocated ELFOSABI_ARM?  I'm fairly certain it wouldn't
> > > have been ARM, 
> > 
> > It certainly wasn't ARM, and I don't think ELFOSABI_ARM was ever
> > codified by SCO either.  I don't remember who exactly made the
> > choice of 97 as a value -- it might have been Nick, Uli, Pat,
> > Scott or possibly even me -- but at any rate it was just a GNU
> > thing.
> 
> Grr... ;-)

Umm, sorry ?  I think that it was me who made this choice, although I
do not recall the exact circumstances.  I certainly checked it in to
the sources.  Given the value (97 = ASCII 'a') I assume that it was an 
attempt by a handsome young naive programmer to find a way to
distinguish between the old implementation of the ARM ABI in binutils
and the "new" implementation.  (Of course using the terms 'old' and
'new' was also an unfortunate decision.  Ah how innocent I was back
then).

Cheers
        Nick


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]