This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: powerpc-linux: C++ looks completly broken
On Wednesday 21 February 2001 22:23, Geoff Keating wrote:
> Franz Sirl <Franz.Sirl-kernel@lauterbach.com> writes:
> > -LIB2FUNCS_EXTRA = eabi.S tramp.S
> > +LIB2FUNCS_EXTRA = tramp.S
>
> Aah, I see. Yes, eabi.S should not be put in the shared library. Or
> if it is, it should have a lot more weak symbols or something (but why
> bother, since eabi.S is _only_ used by the main application?).
Yeah, I suspected that it was only used by the main application, that's why I
suggested in an earlier discussion
http://gcc.gnu.org/ml/gcc/2001-02/msg00132.html on tthis list with Mark
Mitchell:
Well, I'm not really authorative for EABI issues, but as far I can parse
the stuff in rs6000/eabi*.asm, nothing of that needs to end up in
libgcc_s.so. So if that is true and we want to go with minimum changes, I
would suggest to change rs6000/t-ppccomm to something like:
LIB2FUNCS_EXTRA = tramp.S
LIB2FUNCS_NON_SHARED_EXTRA = eabi.S
and change gcc/Makefile.in to link
LIB2FUNCS_EXTRA+LIB2FUNCS_NON_SHARED_EXTRA into libgcc.a and to only link
LIB2FUNCS_EXTRA into libgcc_s.so. That would limit the impact to PPC for
now and all other targets can continue as normal. That should make the
regression tester happy again too.
But if this is a plausible solution probably only the EABI experts like
Geoff or Michael Meissner can tell.
But I guess now I'll just wait on what the other 2 shared libgcc discussions
come up with :-).
Franz.