https://gcc.gnu.org/bugzilla/show_bug.cgi?id=123976

--- Comment #132 from Sergey Fedorov <vital.had at gmail dot com> ---
(In reply to Peter Dyballa from comment #130)
> I just checked which "bootstrap" ports I have installed – and to my surprise
> I see:
> 
>   cmake-bootstrap @3.9.6_0
>   gcc10-bootstrap @10.5.0_0
>   glib2-bootstrap @2.88.1_0+x11
>   gobject-introspection-bootstrap @1.86.0_1
>   py-bootstrap-modules @2025-10-11_0
>   xz-bootstrap @5.8.3_1
> 
> gcc10-bootstrap surprises me, I was so sure it failed to build (could be it
> did so in Snow Leopard).

AFAIK, gcc10-bootstrap does not build as +universal on SL, because it tries to
build ppc64 slices, and SL does not support ppc64. But the port disables
universal builds on 10.6 ppc/rosetta. So no manual intervention should be
needed, it is enough not to force +universal.
On 10.5 is builds fine at least for ppc+ppc64 (on a G5). I might have also
tried i386 + ppc ages ago.

To save time, just build as `gcc10-bootstrap -universal`.

> This would mean that I can deactivate much more but
> activate gcc10-bootstrap and xz-bootstrap plus possibly this or that library
> the "castrated" GCC-16.1 would need to build (finding it out "a bit" later).
> 
> Understanding you correctly, it would need to add possibly two more
> directories, those containing the executables from gcc10-bootstrap and
> xz-bootstrap?

If you mean what you need to build gcc outside of ports, then gcc10-bootstrap
is the only requirement (because you need a compiler with C++11 support, so
Xcode gcc cannot work).
If you build inside of *ports, then do not do anything manually without a good
reason. Deactivate everything first, then build the first port in the chain (in
gcc case it is libgcc15 or libgcc16). If that fails, only then changing
something becomes justified.

Reply via email to