Re: fix eh on tru64 [Re: libjava status on Tru64 UNIX V5.1]

[Roger Sayle <roger at eyesopen dot com>]

Rainer Orth wrote:
> > >> Excess errors:
> > >> /usr/bin/ld:
> > >> conpr-3b.o: foo_t::foo_t[in-charge](): weak symbol multiply defined
> > >
> > >Stupid ld.  Consider patching testsuite/lib/prune.exp to remove these.
> >
> > I've been annoyed with these warnings from the native Tru64 linker
> > myself.  I've been using "-Wl,-S" to pass the "-S" flag to the linker
> > to silence these problems.  I'm not sure how easy it would be to do,
> > GCC's SPECs are cryptic to me, but it might help to always pass -S
> > to the native linker from g++ on alpha*-dec-osf*.
>
> that's what I tried when implementing #pragma weak support for Tru64 UNIX.
> Unfortunately I found no way to silence the ld warning, not even -S.  I
> tried both /usr/ccs/lib/cmplrs/cc/ld from Tru64 UNIX V5.1 Patch 358.00
> which is currently installed on my V5.1 system and the latest V5.1 ld
> from Patch 605.00.

I've found that -S does indeed silence the "weak symbol mutiply defined"
warnings emitted from ld V5.1 (I've no idea which two patchs levels I've
tried).  The problem is that on Tru64 you still get "weak symbol mutiply
defined" warnings from the native ar, with no way of turning them off.
It might have been these messages that led you to believe that the -S
flag wasn't working with ld.  To get around these warnings, we use GNU
ar from binutils on Tru64 which works fine with g++ generated objects.
Between -Wl,-S and GNU ar we get warning free compilations on the alpha.

Roger
--
Roger Sayle,                         E-mail: roger@eyesopen.com
OpenEye Scientific Software,         WWW: http://www.eyesopen.com/
Suite 1107, 3600 Cerrillos Road,     Tel: (+1) 505-473-7385
Santa Fe, New Mexico, 87507.         Fax: (+1) 505-473-0833