On Sat, Sep 22, 2007 at 02:39:08AM -0400, Blair Sadewitz wrote:
> For gcc4, I think it might pay to start enabling this on a per-package
> basis (after the freeze, of course).  I don't know enough about it to
> know if there are any common cases where this would be
> contraindicated, though.
> 
> For smaller libraries, AFAIK this is not that important, but for huge
> ones with lots of long symbol names, it significantly reduces the
> amount of work the run-time linker has to do.
> 
> Opinions?

Please don't. It makes assumption about the libraries, changes the ABI
and I generally believe is something only upstream should do.

Joerg