Borland model, Gnu-ld 2.8

[Yotam Medini]

>  Cc: egcs@cygnus.com
>  From: Jason Merrill <jason@cygnus.com>
>  Date: 16 Sep 1997 09:33:15 -0700
>
>  >>>>> Joe Buck <jbuck@synopsys.com> writes:
>
>  > From: Yotam Medini <yotamm@tmai.com>
>
>  >> > I also prefer the Borland method, provided the linker can support it.
>  >> > That is to ignore multplied defined methods.
>  >> > Is such linker is in egcs's scope?
>
>  >> Just for you, voila, presto, it has been added, please get and use GNU
>  >> ld, with ELF.  :-)
>
>  > For anyone confused by Mike's witty reply: any linker that can handle
>  > weak symbols will do, and older versions of GNU ld, as well as vendor
>  > linkers on SVR4-style systems, will work fine.  This is not "in egcs's
>  > scope" since the work was performed well before egcs began.
>
>  But wait, there's more!  With GNU ld 2.8, the linker will discard the
>  duplicates, like Borland's.  Older versions, and other ELF linkers, keep all
>  copies even though only one is actually used.
>
>  Jason

Very nice! So could you please clarify:

 * Should we build GNU ld 2.8 on all platforms ?

 * When building egsc's g++, how do we tell it to use that great 'ld' ?

-- yotam