Re: arm questions

[hjl@lucon.org (H.J. Lu)]

> 
>    Date: Thu, 15 Apr 1999 14:49:16 -0400
>    From: Scott Bambrough <scottb@corelcomputer.com>
> 
>    Actually Phil is in England, Pat Beirne and I are in Canada.
> 
> Oops--sorry.
> 
>    We chose REL because we did not need the addend.  In hindsight I
>    would choose RELA.
> 
> For the benefit of anybody who is considering an ELF port, please
> always use RELA relocs rather than REL relocs.  REL relocs work OK for
> processors with very simple relocation needs, like an i386 or m68k.
> For a RISC processor, they are a real pain.  You always need to
> support an addend which can support your entire address space, in
> order to handle code like this:
>     char a[10000];
>     int foo () { return a[9999]; }
> This ought to turn into a reference to the symbol a + 9999, so you
> need an addend which can hold a value of 9999, or you get inefficient
> code.

Ok. You are saying RELA is better than REL. Does it make any senses
to use RELA for relocatable object and allow REL for excutable and
shared objects?

> 
>    > In the meantime, Cygnus implemented ARM ELF, using RELA relocs.
> 
>    Something of which we were entirely unaware of at the time, otherwise we would
>    have attempted to collaborate in a closer fashion.
> 
> I now recall that I dropped the ball on this one.  I saw the request
> for the ARM ELF work come in to Cygnus, but I didn't manage to figure
> out that it was for a different ABI than the one in H.J.'s snapshots.
> 

If RELA is really desired, can ARM/Linux switch to RELA with a different
soname? Binaries using REL should still run ok, just like what we did
to libc 5 on x86.

-- 
H.J. Lu (hjl@gnu.org)