Subject: Re: Uhm. Is this a feature?
To: Frederick Bruckman <fb@enteract.com>
From: Thor Lancelot Simon <tls@rek.tjls.com>
List: current-users
Date: 05/30/2000 13:27:37
On Sun, May 28, 2000 at 08:02:26PM -0500, Frederick Bruckman wrote:
> On Sun, 28 May 2000, Brad Spencer wrote:
> 
> >    > #       XXX x86 assy produces lots of RRS text relocation errors on shared
> >    > #       XXX lib, must fix.
> >    > 
> >    > This really should be looked into...  I think that these are fatal, not
> >    > just an annoyance.  They most often happen when a nonposition independent
> >    > code is mixed in with position independent code in the same shared object.
> >    > I just ended up using the C versions of those functions.  If I left the
> >    > assembly ones in there anything built with libdes [like su, etc..] would
> >    > fail trying to find one of the more common des functions.
> > 
> >    Yeah, I noticed that, too. Considering, do you suppose the following
> >    would be appropriate?
> > 
> > 
> > 
> > I am not sure that this isn't trouble on an ELF system.  I really
> > understand all of the details.  It would probably be better to get the
> > assembly fixed.
> 
> Or skip it altogether, then. 'libdes' works fine without the assembly.
> Can anyone produce benchmarks that show a clear benefit?

When I upgraded libdes I did extensive testing which revealed a significant
benefit to using the x86 assembler.  Somehow I've lost some of the context
of this discussion -- are you claiming that as of the last time I touched
the library, the x86 assembly was producing RRS text relocations?