.weak vs. C++ templates. was: SCO 5.0.2 error

[Robert Lipe <robertlipe at usa dot net>]

Jason, can you advise, please?

What do we give up by turning off weak support in a target?

The problems between .weak and the System V linker have been pretty
well covered in the list.  This is a FAQ (and even more so in
comp.unix.sco.programmer and similar lists) and I'd like to make it
go away.  If we could make it go away for 2.95.2, that would be even
better.

Though I haven't verified what Rob describes, it stands to reason that
one way to end this stalemate is to just turn off WEAK on targets where
GNU ld isn't an option since G++ and System V ld don't see eye to eye on
the subtleties of .weak.

Is this a reasonable thing to do in a point release?

RJL


Rob Kramer wrote:
> >If you have clock cycles to kill, try this
> >	edit config/i386/sco5.h
> >	whack the definitions of SUPPORTS_WEAK and ASM_WEAKEN_LABEL.
> >	do a full build.   (sorry) 
> >	see if the problem is better or worse.
>
> Well, I rebuild gcc without those defs, and now it compiles..! I
> compiled a fairly big C++ application, and it seems to work fine so
> far. That is, until we find something else is broken :)
>
> If it works fine without weak lables, then why use them at all? (Yes,
> I'm clueless here)
>
> Thanks!
>
>     Rob Kramer robk@cyberway.com.sg