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

--- Comment #126 from Iain Sandoe <iains at gcc dot gnu.org> ---
(In reply to Sergey Fedorov from comment #124)
> (In reply to Peter Dyballa from comment #120)
> > (In reply to Iain Sandoe from comment #119)
> > > additional:
> > > 
> > > To bootstrap, i used a GCC-10.5 built using Apple gcc-4.2.1
> > 
> > It does not play a role which compiler is used to start building since in
> > the end it's GCC that builds its final stage. So I won't start building
> > GCC-10. And I am not sure which compiler was used to start building
> > GCC-15.2, could be GCC-7 I then had and used since MacPorts' default
> > compiler for this job or port had failed a few times.
> 
> While I believe the statement in the narrow context is correct (provided gcc
> build has succeeded), what is problematic is that once environment diverges
> from what was tested, there is no way to know what could have affected the
> failing build.
> We know, for example, that a few specific ports do break the build of gcc.
> What we don't know is whether all other do not have that effect.
> 
> P. S. BTW, if you build outside of *Ports, make sure to set the PATH so that
> /opt/local/bin is NOT in it. (Alternatively, all ports are deactivated.)
> Otherwise we are back to the same problem: arbitrary binaries can be picked.

agreed if adding the PATH to the bootstrap compiler enables a whole bunch of
other tool changes - once again we are in the "too many variables" land...

Reply via email to