This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: A clue for the libstdc++ problem.
On Sun, Apr 01, 2001 at 04:40:55PM -0300, Alexandre Oliva wrote:
> On Apr 1, 2001, "H . J . Lu" <hjl@lucon.org> wrote:
>
> > * ltcf-cxx.sh: Make sure we never use g++ for the shared
> > libstdc++-v3.
>
> This is outrageously misleading. BTW, you've duplicated Zack's
> investigative work, and the patch for the top-level Makefile.in I
> wrote on Friday to address his findings hasn't worked, it seems, but
> it's the Makefile.in that has to be fixed. Maybe with some
> cooperation from the top-level configure.in, for example, setting a
> different CXX_FOR_TARGET for use for libstdc++-v3. But this still
> wouldn't fix problem of taking the libstdc++-v3 flags created only
> after libstdc++-v3 is configured. I'm still aiming at a
> single-CXX_FOR_TARGET solution. The tricky thing is how to defer the
> evaluation of CXX_FOR_TARGET.
>
Please keep in mind that building the shared libstdc++ is a special
case, just like building libc.so. You cannot use the normal C++ rules,
just like we don't use the normal C rule for building libc.so. I think
the best way to fix it may be to have something like ltcf-stdcxx.sh,
which is only used for building libstdc++.so. Anything else probably
will be just another kludge which may introduce other problems.
H.J.