Subject: Re: binutils
To: Jason Thorpe <>
From: Jonathan Stone <jonathan@DSG.Stanford.EDU>
List: tech-toolchain
Date: 12/29/1997 18:26:18
on Mon, 29 Dec 1997 17:41:52 -0800,
Jason Thorpe <> writes:

>On Mon, 29 Dec 1997 17:42:47 -0800 (PST) 
> Bill Studenmund <> wrote:

> > What is the advantage of elf over a.out? I'm curious, not trying to say we
> > shouldn't. :-)
>One, it's reasonably standardized.
>Two, it's just more flexible than a.out in so many ways.  This is the main
>reason... it just does more.

I agree with Jason, but it's worth noting that the bloat for `doing
more' can be nontrivial.  I seem to remember a growth of 60%-100% in
statically-linked binaries in /{bin,sbin} when the pmax switched from
a.out to ELF.  Other architectures may be less bloated.  But if it
means growing your root partition, then thats something more the
installation tools needs to support :-).

There's also the question of how much, if any, support we need to add
for the existing a.out binaries until they're finally abanoned (2.0?)
I'd guess that if GDB can read a.out shared binaries, and we can exec
them, that's enough.

Correct me if I'm wrong, but our current kernel ELF exec code doesn't
support the GNU OS tagging standard yet, does it?  We really need to
fix that, or going to ELF will break svr4 and Linux emulation.