Re: Fwd: gcc 4.2.0 compilation failing on x86_64 machine

[Kai Ruottu <karuottu at mbnet dot fi>]

Tim Prince wrote:
> Ryan Propper wrote:
> > See the thread below.  It appears that not having "gnu/stubs-32.h" is
> > a common problem on some AMD and other 64-bit machines, resulting in
> > compilation errors.
> >
> > The solution that worked for both me and Alexander was to use the
> > "--disable-multilib" configuration flag.  However, we did not see this
> > documented in any FAQs or in the output of "configure --help".  We
> > think other users might benefit if this solution was more widely
> > publicized.
> Which situation?  That x86-64 has optional 32-bit support, but 
> attempting to use it without installing all bits is likely to fail?
Agreed, if one wants to install a "complete" x86_64 system,  it can mean 
installing every possible RPM
or other available package... So some parts may be missing in the 
"default" installation.  Maybe there was
a choice for "32-bit development" but the installer didn't want that 
then, so all the 32-bit development
parts were not installed.  But most probably the 32-bit runtimes were 
installed as default so running 32-bit
apps will work still... But not producing them!

GCC expects the builder to want both 64-bit and 32-bit support from the 
produced GCC, that is the
assumption!  If the builder wants only support for 64-bit then the 
support for the 32-bit development
can be disabled via the '--disable-multilib' - no 32-bit libraries will 
be produced with the '-m32' option!
The produced GCC is "brain damaged", some day someone maybe wants to try 
the '-m32' option and
then it gives weird errors again!