Re: PATCH: Don't export libgcc symbol (Re: symbol clarification please?)

[Jack Howarth <howarth at bromo dot msbb dot uc dot edu>]

Franz,
   I'll rebuild my glibc 2.2.5-13 here with glibc-libgcc-compat-ppc-5a
and HJ's proposed glibc-2.2-ppc.patch, install the resulting glibc
and attempt to build debian's gcc 3.2 here. HJ's patch is now suppressing
the symbols as I get...

0002693c g    DF .text	00000020 (GLIBC_2.0)  __divdi3

...now in libc.so.6.
   In any case, now we have the argument that Jakub and Richard have
been demanding...

1) Ulrich demands we not export any symbols for linking that are
in libgcc-compat for resolution at run-time.

2) Satisfying this request breaks the build of gcc 3.2 because
libgcc_s.so.1 then can never provide the symbols since they 
now are no longer available to be linked into libgcc_s.so.1
from glibc.

3) To avoid this paradox we must re-create the symbols in
gcc 3.2 with a new version to keep them from disappearing
from libgcc_s.so.1 entirely.

HJ, is that a coherent read of the situation? The problem
Franz has been having is being able to describe this clearly
enough to Richard and Jakub on the gcc-patches mailing list
to convince them to accept his gcc 3.2 fixup patch. I think
if you are convinced now of the need for the gcc 3.2 fixup
patch, you should speak up on the gcc-patches mailing list.
I suspect you'll be very convincing to them.
                             Jack
ps I think how we eventually implement the specifics of the
patch to libgcc-compat is less important at this point than
convincing the gcc folks to accept Franz's gcc 3.2 fixup
before gcc 3.2 is released. This would likely be the best
point at which to resolve this problem in gcc.